G.P. Gavrilov, A.A Sapozhenko: Selected Problems in

G.P. Gavrilov, A.
A Sapozhenko
Selected Problems in
DISCRETE
MATHEMATICS
MIR PUBLISHERS MOSCOW

selected Problems In
r n raapUJJOB, A A. Canod\eHKO
C6opHIIJt aa~a'l no AJlChpeTIIOii
MaTe)t3TitRe
G.P. Gavrilov,
A.A. sapozhenko
Mir Publishers Moscow

Translated from Rus.siaJl by
Ram S Wadhwa and
NatalJa V Wadhv.a
FJrse pub1l5bed J 989

Rev1.sed froru the t 977 Ru.ssian e
ISISN 5 ·03·000522-6 ~ r Jtallaast peAa11U.V.R tJluan«o- va te¥&Tn.qe~KOi.

nu Tepa IY p~ ua.u;a ren ~cTJJa • Hay ;Kat , i 977
@ EngiJsh trBllslat1on~ M1r Puhl~sbers, 1989
Contents
Preface 7
Chapter 1. Boolean Functions: Methods of Defining and

Basic Properties 10
Ll. Boolean Vectors and a Unit n-Dimensi onal Cube 10
i 2. Methods of Defining Boolean Functions . Elementar y
Functions. Formulas. Superposi tion Operation 22
i 3 Special Forms of Formulas. Disjunctiv e and Con-
junctive Normal Forms. Polynomia ls 33
i.4. Minimizat ion of Boolean Functions 42
1.5 Essential and Apparent Variables 49
Chapter 2. Closed Classes and Completeness 55

2. i. Closure Operation . Closed Classes 55
2.2. Duality and the Class of Self-Dual Functions 59
2.3. Linearity and the Class of Linear Functions 63
2.4. Classes of Functions Preserving the Constants 67
2.5. 1\lonotonicity and the Class of Monotonic Functions 70
2.6. Completeness and Closed Classes 76
Chapter 3. k· Valued Logics 82

3.1. Represent ation of Functions of k-Valued Logics
Through Formulas 82
3.2. Closed Classes and Completeness in k-Valued
Logics 88
Chapter 4. Graphs and Networks 101

4.1. Basic Concepts in the Graph Theory 101
4.2. Planarity, Conn~>ctivity, and Numerical Charac-
teristics Qt Graphs H{}
4.3. Directed Graphs 117
4.4. Trees and Bipolar Networks 123
4.5. Estimates in the Theory of Graphs and Networks 1.37
4.6. Represent ations of Boolean Functions by Contact
Schemes and Formulas 143
CONTE'lTS
Chapter 5 Fundamentals of Codlng Theoey t5S

5 \ Codes w1tb Co-rretllon~ 1.55
5 2 Linear Codes 160
5 a A1phabe\tc Cod1ng 1'03
Chapter 6 Fi.o I te .Automatons 174

6 t Determinate and Boundedly D:eterm1nate Functions 174
6 2 Representation ol Determinate Funct1ons b_y Afoore
D1agrams Canontcal Equat1ons Tatdes.and Schemes
Operations In volvtng Detetm1na te Funetlons i87
6 3 Closed Cla~ses and Completeness In the Sets of
Determanatt! and BlJundedly Determtnale Fune
llODS 206
Chapter 7 Fundamr;ntRls of the Algorithm Theory 212

1 I Tur1Dg s },faehln~s and Opera. t1ons "all th Them
Funt1lons Cnmputahl~ an Turlng ! l\.1 ath1n~s 212
1 2 Classe~ ol Computable and Recurs1'e Functions 233
7 3 ComputabJltty and Comple1:~ty of ComputatJons 2~1
Chapler 8 Elements ol Combioat<-riaJ Analy5m 248

8 J Permutations and CombtnatJons Propert1es of Bu1o
ml al Coeflic len ts 2.48
8 2 I ncl ws ion and E .ze) us Ion Formulas 259
8 8 Recurrent Sequencee Generat1og Funetlans and
Recurrence ReJal1ons 265
8 4 Pol)A s Theory Z75
8 5 Asymptot 1c E1press1ons .~&nd Inequallttes .280
Sol u.tions, Answers t and Hints 289
B1bJ 10graphy 403
Notations 405
S11bjec l Index 40!1
Preface
This collection of problems is intended as an accompani-

ment to a course on discrete mathematics at the universi-
ties. Senior students and graduates specializing in mathe-
matical cybernetics may also find the book useful. Lectur-
ers can use the material for exercises during seminars.
The material in this book is based on a course of lec-
tures on discrete mathematics delivered by the authors over
-
a number of vears at the Facultv. of Mechanics and Mathe-
matics, and later at the Faculty of Computational Mathe-
matics and Cybernetics at i\loscow State University.
The reader can use Introduction to Discrete ~1fathema
tics by S. Yablonsky as the main text. when solving the
problems in this collection.
The book consists of eight. chaptet·s. The ftrst two chap-
ters are devoted to Boolean algebra which forms the ba-
sis of discrete mathematics. About a quarter of the total
teaching time during lectures and practica1s at the Com-
putationall\fathelllatics and Cybernetics Facult.y at Mos-
cow University is devoted to Boolean algebra. The ma-
terial in this part introduces the student to the concepts
of discrete. functions, superposition, and functionally
complete sets. It also acquaints the student with various
methods for specifying a discrete function (tables, poly-
nomial representation, normal forms, geometrical repre-
sentation using an n-dimensional unit cube, etc.). Me-
thods for testing the completeness and closure of sets of
functions are also considered.
The third chapter is devoted to k-valued logics. The
problems presented are intended to acquaint the reader
with the canonical expansions ot k-valued functions, equiv-
alent transformations of formulas, closed classes of the
k-valued functions, and methods for testing the complete-
ness and closure of functions. Several problems in the
8 PREFACE
ehapter 1llustrate the d1ilerence betv.een k valued log1cs

(k > 2) and Boolean algebra
The fourth chapter conta1ns problems on the theory of
dlrected and undtrected graphs and the network and ctrcutt
theory The chapter descr1bes the baste concepts methods
and terms of graph theory v..h1ch are w1d~ly 1.1sed to de
scr1he and 111Vest1gate the structural propert1es of obJects
In various branches of sctence and technology TJ1e prob
le-ms a.re intended t.o eonsolldate the bas1c concepts of
graph theory to Illustrate the application of network and
graph theory to the construction of cJrcuJts 1 epresent1ng
Boolean functtons to count the number of obJects w1th a
gtven geometrJcal structure etc The authors hope that
the lecturer wtll also find problems 1n thts chapter to
help h1m demonstrate the mathemattcal r1gor dur1ng the
proof of geometrtcally ob-v1ous statements
'rhe fi(tb chapter des~tlbeS the baste concepts o[ c.od1ng
tlteory The problems concern the properties of error cor
rect1ng codes, alphabetJcal codes and m1n1mum redun
dancy codes
The chapter conta1ns problems demonstrating
SI>..tii
d1fierent ways of descr1btng d1screte transformers (auto-
matons) Problems a1med at revealtng deternltntst•c and
boundedly determinl.SlJc automatons are also gtven Other
problems conc.~l"n the d1fierent ways of representing auto-
matons (diagrams canonical equations and schemes (ctr
CUlts)) the 1nvestJgat1on ol the funct1onal completeness
and closure of sets of autontaton mapptngs and also the
properties of operations Involving such mappJngs
The seventh chapter deals Witll the elements oi algo-
rithm theory and 1s Intended to provide an 1dea about
effect1ve computabtl1ty and complexity of computations
lt 18 also about eertaln ways fot spec1fy1ng algortthms
suclt as Tur1ng s maeh1oes and recursive functions
!he e1ghtb chapter descr1bes the elements of co.mbtna
tor1al ~nalys1s \Vhtle studytng dtscrete mathemattcs
one frequently comes across questtons concerning the
ex1stence eount1ng and est•mat1on of v arJous combtna
tor1al obJects Hence comblnator,al problems are 10
eluded In the book
For the sake a[ c.onven1ence th~ authors have started
each 5ecl1on uttb a theoretical background
H1nts and answers are prov1ded far most (but not c;J.ll)
PREFACE 9
problems. Solutions are given in a concise form in the

form of notes, and trivial conclusions are omitted. In
some cases, only the outlines of solutions are presented.
The exercises in the book have various origins. Most of
the material is traditional and specialists on discrete
mathematics are all too familiar with such problems.
However, it is practically impossible to trace the origin
of the problems of this kind. l\Iost of the problems were
conceived by the authors during seminars and practical
classes, during examinations, and also while preparing
this hook. Some of the problems resulted from studying
publications in journals, and a few have been borrowed
from other sources. Sevel·al problems were passed on to
us by staff at the Faculty and by other colleagues. The
authors express their sincere gratitude to them all.
The authors are deeply indebted to S. V. Yablonsky
for his persistent interest during the preparation of this
book. His comments and suggestions played a significant
role in determining the structure and scope of this book.
We are also grateful to our reviewers V .V. Glagolev
and A.A. Markov for their critica.l comments and sugges-
tions for improving the collection.
G.P. Gavrilov
A.A. Sapozhenko
Chapier One
Boo\ean Functions: Meihods

of Defining and Baslc Properties
t.t. Boolean Vectors

1
and a Unit 1l-Dimensional Cube
A vector (a1 a 2 an) whose coordtnates assume
"t ,
values from tJJe set ( 041 1} IS caJ led a blnary or Boo

lean, veetor (tuple) We shaJI denote such a vector by
,.,; ~
a• or t:L The number n 15 ~ailed the length aj the

vettor The set of all Boolean vectors of length n 1s
called a unlt n d1rnenslonal cube and IS denoted by nn
/"'ftJ
The vectors a.n are called the verttces of the cube nn

~ """'
The wnght or norm. ntzn H of the vector a,n. lS the nnm
her of -coord1nates of thts vector that are equal to \'ln1ty.
n
1 t U~,.,. U~ ~ at The set of all verttces of the cube B"
... ~,
having a we1ght k ts eall~d k th stratum of tile cube
~
11" and IS denoted by B~ To each Boolean vector a" .

,..., .,..,
there corresponds a number v (a.n) ~ ~ ct. 12n- 1 1 cal ted
-~-J
ltW ~
the number of the vector etn. The tuple et,. IS obviously

......
a blnary expans1on of tne number v (~n} The (Ham-
;i"'V ,._
ming) dt.stance between the verlltes a and P of the tube

,..) ,..,; t\
s• is the nurnhet p (a .. Pl ~ ~
[:;:;;.; I
1a., ...... P1 (, equal to tne
number of coord1.r1ates Jn wh1c.h they dJHer The Ham..
minR d1stance IS a....,metnc,,._ an 6 the cube B~ 1s a rnetr1c
space The tuples o: and P from B" are called adJacent
~ II'V I"'V 1'-¥
ll p (cr., ~) = 1, and oppostte 1f p (at p) = n An unor
T b.ls set.t~r.an 1s auxlliuy \V-e s.ha\1 h~ Ui>\.UI t\-n.\y pt~'h\~m!.

I
I t i ·1 1 6 1 1 1f t f ~ 14 • 1 1 f 5 t t 31 1 t 34 I t 35
and 1 i -" i t •
1.1. BOOLEAN VECTORS 11
dered pair of adjace nt vertic es is called an edge of the cube.

~ ,..., I"<J f"oJ
The set B~ (a) ..-= {~: ,....,p (a, ~) = k} is called a sphere, ~~
while the set S~ (a) = ,..., {~: p (a, ~) ~ ,...,

k} is a ball of ra-
dius k with a centre at a. The tuple an is said to precede
"" ""71 ""
the tuple ~n (notat ion: a ~ ~n) if a 1 ~ ~~ for all
""11 "" "'
i = 1, n. If in this case a =1= ~n, the tuple all is said
"' "'
to precede ~ strictly (notat ion: a" < W'). If at least one
""11
,..., /"'J f"oJ ,..., f"'oJ
of the relatio ns a 11 ~ ~n or ~ll ~ an is satisfied, a" and

"' ""
W' are called comparable. Otherwise, all and are said ""
w·
"" ""
to be incomparable. The tuple a'' directly precedes W'
tV N f"tJ ,....;
if an < ~n and p (an, W') = 1. The precedence relatio n

between the tuples is the relatio n of partia l order in 8 11
•
1111
111
1110
110 1100
01
100 001 0001
000
83
- Fig. 1
Figure 1 shows the diagra ms of partia lly ordere d sets

2 3
B 1 8 and B.t. The sequence of vertic es of the cube
..... "" ,...,
{ao, al, ... , an} is called a chain connecting ~ and
• ,..., ,...., ,..., ti"'oJ
roJ
a,, (notatwn: [a 0 , a 11 ]) if p (a 1_ 11 a 1) = 1 (i = 1, k).

The numbe r k is called the length of the chain [;0 , ~11 j.
,..., "" ,...,
The chain {a 0 , a 1 , • • • , a 11 } is called an ascending chain
,..., ""
!! at-1 <at (i = 1, k). A chain z of the type{~0 , ~1 , ••• ,
an} is called a cycle of length k if ~0 = ~ 11 • Let ~=(a
"" l'
a~l .. . ,an)and~ = (~1• ~21 • • • , ~n) be vector s from 8 11 ,
We denote by ; EB ~ the vector (a 1 EB ~ 1 , a 2 EB ~ , • • • , 2
12 CH l DOOL:EAN PUNC.TtONS
..... ~
a:n $ ~n) obtatned by extl us1-v e sum of ve-ctors a. and

,..., ,._
f.
By cc. U ~ \\e denote a vector whose t th coord1nate ts
equal to zero 11 and only lf C(l ~ ~~ = 0 and by a. ~
.l"ft.J
n
"'
a ~ector wl1ose t th c.oord1nate 1s equal to 1 1f and only 1(

-,...., ~
a: 1 = ~~ == 1 By a. we denote a vector (opposzte to a)

v..hnse t-th eccrdlnate assum-es the value 0 1.f a. 1 ~ 1 and
,liiiV
th.e ~alue 1 s,_f a.t ~ 0 If a E. {0~ 1 },,., we ,....

put ott ~
(oa.1 , oa 1 , aa.n) The s-ymbols 0 and 1 are used to-
denote vectors (0, 0~ ~ 0) and (t ~ 1, ~ 1) res pee·
tively
The set n;. it 0"~
CtM) from B"
'k of all tuples {<:ll'
for wh1ch a~_ = o 1 (J = t, k) IS called a face oj the
1
cube Bn The set I ;;;;: {l 1 ~ '" J ts tall ed the d trectwn oj
the fact the number A the rank of the /a(.e, and n - k
the dlmtns~on (!/ the face
.......
f t t. (1) Ftnd the number f B~ I of tuples an havJng
a we1ght h •
{2) 'Vhat IS the total number of \ ertJces 1n the
cube nn?
1 t.2., (1) F1nd the numbers of tupJes (tOOt)~ (01101)
and (110010)
(2) F1nd a vector of length 6 wh1ch ts a btnary
expansiOn of the number 19
.l"w
1.1.3. F1nd the number of tuples a. E B~ sat1sfy1ng the

~
tond1t1on 2" '~ v (a) < 2"

1.. 1.. 4.. Show that the !ollow1ng relatlons hold fur any
I"'W ,.,. ~
al ~~ V tn Bn
fly ,., ..... ,.,., f'Y f"V
{1.) $) {a..1}~ p {o. ~} + p ~, 1},

_, .-...t ,.., t'IW ,... l'W
(2) P (a. l} ~ P (t:£ $ ~' l ID ~},

,_ il"'J .,..;
n u u n- n
I"W ~ ~
(3) p (a., ~) =- a + tt 2 Ct n ~ n'

u
,...,. ~ ~ f/4!
(4) p (~~ ~> = tt e ~ u

t t 5 (1) Ftnd the number of unordered PBilrs of ad]a--
cent vertices of B"'
(2) Ftnd the number of unordered pa1rs of tup
J/1¥ ~ ~ ~
les (etn, 11"), such that p (c:tn, ~n) ~ k

U. BOOLEAN VECTORS i3
~ ~
1.1.6. Let a and ~ be the vertices of a cube Bn,

"'
p (~. ~) = m. Find the number of vertices y satisfying
the condition
tv ~ ¥"./
+
l"J f'V ,...,
(1) p (a, y) p (y, ~) = p (a,~);

..... "" ,.., ""
(2) p (a,'\') = k, p (~, y) = r;
"' "" "" ~
(3) p (a,'\')~ k, p (~, '\') = r;
"' "' ~ ~
(4) p (a, y)~ k, p (~, y)~ r.

1.1.7. Prove that the following systems of relations are
~ "' "'
incompatible for a, ~ and y in En, n ~ 2:
"' ,..., """" ,.....; ,...,
>
,-.v
(1) p (a, ~) > 2n/3, p (~, y) > 2n/3, p (y, a)

2n/3;
,...., - ,...., "' ,..., rJ f"ooJ
(2) v (a) < v (~ E9 y), v (~) < v (y E9 a), v (y) <

"' .....
v (a. El) ~);
,..., l"'tJ "' "" "' ,..., ,...,
(3) II a II > II ~ EB y II , II ~ II > II y E9 a II, II y II >

II~ EB ~ II , II ~ n <If ny) II = 0;
,...,
- - - ,..., l"'oJ ,...., "' "'
(4) II a. tB ~ E9 y II = o, II a EB ~ EB y II = n - 1.
..... "' "'
1.1.8. Let a, ~ and y be the vertices of a cube Bn. Show
that:
,..., ,..., roJ - t"J
(1) a:::;;~ is equivalent to a n~ = 0;

-,..., ,..., ,..., -,..., -,....,
(2) aU ~ = 1 is equivalent to ~:::;;a;
t'Ool ,..., ¥"J ,...,
(3) a n (au ~) = a;
a. u (a n ~) = a;
1"'<1 ,..., """' ,...,
(4)
(5) <; u (3) n <~ u Y> = <~ n Y'> u ~;
((a. n '\') u ~> n -v = n u n
,.., - I"W "' ,..., ,..., ,...., -
(6) (a y) <P y); ~
ny;
,...., """' f'V
(7) a:::;; 'Y leads to the relation au(~ ny) = (au p)

,..., ,..., ,...., ,....,
1"-1 ,-v ,...., ,..., ,..., ,..., ,..., ,...,
(8) a~ y is equivalent to aU (~ n y):::;; (aU ~) n'Vi

n n n n ( u y)
,..., f"o,J ,..., ,..., ,...., ,..., - ,..., ,...., -
(9) (a ~) u (~ y) u (y a) = (au ~) ~
n <:Y u ~).
14. CH t: liOOLLAN FUNCTIONS
2
1 I 9 How many vecto1s (a 1 az as 2) 1n B. sattsfy
the relation ~ a.1 ~ m/2 Ior all m 1 121
~~ ~m ..,._
f 1 10 F1nd the number of' ectors a. ln B~ 1 ~ J.. ~ n/2
1 =:;; r~ (n __../..)/(A 1) whtch 11ave at least r zero
coordinates between tv.o un1t coord1nates
i \ t 1 (t) S1~ow that R contat.ns a set -ccns1st1ng of
( ) pauwtso wcomparabl~ vectors
111121
(2) Sl1ow that any subset conta1n1ng not less
th.an n +
2 vectors Includes a patr of tncornparable
\ec.tQrs
;ow
J t i2 Let 0 ~ l < k -s;;;, n and let A (a) be a set of all

#"'IJ
vectors 1n B C()mparable w1th a. F,Dd tl\e pow-er of the

set C
......
n
~
(1) C ~ A (rz) B~ a E B( ~
n Bi
~ ~
(2) C - A (a.) a E B~
1"¥ f".rr
(3) C ~ A (n) a. E B~t

1 l 13 Let A s Br and B be a .set of all tuples •n
B,., tl1at are comparable wtth at least one set tn A
Prove that 1 A J ~ B 1
(~} ~(:)
t 1 14 (1) SllOW that sn containS 11, palrWI5e different
ascendtng cha1ns of lengtl1 n
(2) Sl~ow tl1at the number of pa1rw1se d1fferertt
ascendt ng cl1atns of length 1z con tatntng a fixed vertex
~
a 10 B~ 1s equal to k1 (n ~ k)J
I t 15• (1) Show that the power of any subset of
pa1rw1se 1ncomparable tuples o1 cuhe _Brt does not exceed
( r r/2J )
(2) Show that If tlle subset A ~ nn conststs
.-.I ,.,.,..
of pairWISe Incomparable sets and {J a II~ k for any aEA

then I A I~ {;) for k~n/2
f 1 t 6 Let p 1 p, Pn. be pa1rwtse dtfferent pr1me
numbers and N be the set of all numbers that can be
presented Jn the form p~•p~• p~n where a~ E (0, 1}
~ = 11 n Let A c N be a subset such that none the or
1.1. BOOLEAN VECTORS 15
numbers a E A is a divisor of any of the numbers b EA

I A I~ ( 1 ,t~ 21 ).
1
other than a. Prove that
"' ,..,
~
11
1.1.17. Let a and be the vertices of a cube 8 , sueh
~ ,...., ,...., ~ .
that a < ~. p (a, ~) = k. Prove that the numbet· of
"' ,..., ,...., ro.l
tuples 'V satisfying the relation a~ 'V ~ ~ is equal to 2h.

1.1.18*. Prove that the cube 8 11 can be represented as
a union of pairwise nonintersecting ascending chains
having the following properties:
(1) the number of chains having a length (n- 2k) is
equal to { ~) -{ k~!), k=O, [n/2],theminima l tuple
of each chain having a weight k and the maximal tuple
a weight n-k;
,.., "' "'
(2) if ah a; +1 , and a 1+2 are three consecutive vertices
of an ascending chain having a length n - 2k,,...., the ,....,ver-
,...., ,...... rv ,....,
tex ~ satisfying the relation a;< ~ < ai+ 2 , ~ =I= Ctt+ 1

belongs to a certain chain of length n - 2k - 2.
"' ,..,
1.1.19. Leta and~ be the vertices of the cube B 11
such
"' I"V I"V ,....,
that p (a 1 ~) = k, and let C (a, ~) be a set of all verti-

,.., "' "'
ces 'V for each of which there exists a chain [a, ~] of
length which is not more thank + 2r containing this ver-
"' "' "'
tex 'V. Find the power of the set C (a, ~).
1.1.20. The set A s 8'1 is called a complete set in B"
"'
if any vector ~ E B" can be uniquely reconstructed under
"' ,..,
the condition that the distance p (a, ~) is known for each
"'
a E A. A
11
complete set in B is called a base set if for each
,.., "'
vector a in A, the set A"- {a} is not a complete set.
(1) Prove that any ascending chain a"'0 , a"'1 , • • • ,
"'
a;ll-1 in nn forms a base set.
(2) Show that the sets B~ and B~!-t are complete in
for n > 2. Indicate a number n > 2 for which 8"1
11
B
•
ts not a base set.
(3) For what n and k is the set B~ not complete in B"?
(4) Prove that any base set A in B" satisfies the con-
11
dition 1og~ (11 + 1> ~lA l~n.
CH t BOO L.EAN F Ut-. Ctro llrl"S
(5) Prove that none of the laces havtng a d1men:non

n -2 ts a complete set 1n B"
(6) Prove tl.J at the 11 umber 'l'n of base sets 1n Bn
sabsf~es the lnequahttes 2 {nl) ~lf'n ~ ( ~ )
1. t ..21. Let r,p be a one to one mapplng of Rn onto If

self The mapp1.ng ,}
tween verltees tf p (a, p) ~ p (rp (a), rp (~)) for all a ~

1n Bn Prove that the mapp1ng q; pteser\cs the distance
1f and on1y Jl 1t can be obtained wJtll the he)p of
(f) a certa 1n com muta t1on of coord 1na tes ~~ mu l tan eon~
ly tn all vee tors 111 Bn,
(2) a substltut1on of 0 hy t anrl J by 0 1n certa1n coordt
nates of all vee tors
t 1 22 The mapptng r.p of the set IJ' onto ttself ts calle<l
""""' ~
TnJ)nolonlc: 11 the condttton v (a)~ v (~) leads to the
~ ~
lnequaltty v (~ (a.})~ v (<p (~ )~ F1nd the number of

m-nnotontc mapp1ngs of the C\lhe B"
,.., ,.w """" ...,
t ~1.23.
~ ,....,
Let a E B; The set AI% (a)=== {y v (1')~
v (a.) 1 E B~} lS c.alled the 1»lttal segment of the stra.
tum B~ ,_, Let A £ n•, we denote by Z7 (A) the set oi all ~
tuples P E R? for each of wh1ch tl1ere ex1sts an a: E A

~
s ut h t}l at f. .s;,; a.
~
(1) Let a En:. and l,, ,,, , lr.. (1 ~ I, < i2 <

< tk~ n) be the numbers of those coordtnates of
the vector a""" wh1cb are equal to unttY Show that
J flf~(~) l=l+(ll~lt)+{'~:': )+ +{n~ht)

(2) Show that l ~ k and A 1s the 1n1 tta1 segment of
1f
the stratum B; ~ the set ZF (A) 1s the 1DJ t1al c;egmen t ol
the stratum
~
nr
(3) Let ttand ~ 1 , ~ 2 ,
E B~ , l" (1 ~ 11 < 12 <
< ~It~ n) be tl1e numbers of those coordinates
~
of the vector a whtclt are equal to un1 ty Let A =

"""
Alt(a.) and l ~ k Show tl1at
1Z~(A) 1=1+ ("~ l )+C',:~· )+

1
+ (n-;t1)
1.1. BOOLEA N VECTOR S 17
(4)* Let1 ~m~ ( ;~). Prove that. the minim um of

the quanti ty I Z~-t (A) 1 for all As= B~ satisfy ing the
condit ion 1 A 1= m is attain ed at the initial segme nt
of the stratu m BJ:.
(5) Let 1~m~ ( ~) and l~k. Prove that the mini-
mum of the quant ity I Zl' (A) I for all As; m:
satisfy -
ing the condit ion 1 A I =-~ m is attain ed at the initial
segme nt of the stratu m m:.
(6) Let a 0 , a 1, a2 , ••• , a., be given numbe rs for which
there exists a set A of pairwi se incom parabl e tuples
such that I A n m: I = ak (k = 0, n). Then the mini-
mum of the quant ity I Zl' (A) I for all such sets A is
attaine d when the set A n
B~' is the initial segme nt
of the stratu m B~' for all i > l.
1.1.24 *. Let A s; B" be a set of all tuples such that
I'V
there,..., are .....,no tuples

.....,
l"o..
a,
,....,
~' y in A for which a

,...,
~ = 0 n
,..., ""
and aU ~ = y. Let ak = I A n
B~ I . Show that
for all natura l numbe rs k and m that do not exceed n .

....., "' .....,
1.1.25. The set r (A)= {a: a EB"'-. ..A,p( a, A)= 1}
is called the boundary of the subset A c B 11
• Let 1 ~ i
1
<
l
.2 < • • •
. - n. UT
< lk"'=:::: ne d eno e y t b A i,, i,, . . . , ik tl
cr,cr •.•. crk 1e
set of all tuples in A for which the coordi nate with numbe r
ii is equal to u1 (j = 1, k).
The centre of set A is define d as a tuple whose i-th coor-
dinate is equal to zero for 1A~ I~ 1/2 I A I , and equal
I"J • • •
1 1 1
to unity otherw ise. Wede noteb ya ' ' " • • •• k the tuple ob-
.....,
tained from a by revers ing the values of coordi nates with
numbe rs i 1, i 2, ... , ik.
(1) Show that for any A s B" there exists an A' s; B"
such that lA' I= lA I, lr(A' .....,
)I =I r(A) I and
the centre of A' is the vertex 0 = (0, 0, ... , 0).
(2) We shall say that the set A s B" has the prope rty
I if for all i = 1, n it follow s from;_ E A! that ~;EA.
Show that for all A s; B 11 there exists a set A' with cen-
2-osas
18 CH t BOOLEAN I"UNCT10'S
treat o. sueh that 1A"' 1 == I A I , A

Jllll-.l
possesses the
propc1ty l, and J r (A) l~ l f (A) l
(3} \Ve s1tall say that tlte set A ~ possesses the nn
property II if for all I, j (1 ~ l < J~ n) It follows from
#IW .....,.
a EA!4i that a~,J EA Show that for any A havtng 1ts

~
eentre at 0 and possesstng tho property I, there extsts an

l'ltl
A' with tentrs at 0 pos.scssing tl10 properlles I, 1I, and

such that ~ A J ~ f A J , and ' r (A') I ~ I r (A) J
(4)• Show that the mlnimum of I r (A) I for all A ~ Bn
~-1
saUsfylDg the inequality iJ ( 7) <
t~t
rA I~ }J
(~~
-
l:)
~ rf!fW
ts attained on the set A = Sl-t (OJ Uill~ (a), \\here

f'lw
.t~Ir (~) 1s a certa1n fin1te segment of the stratum B~

(see Problem 1 1 23)
1
i.1.2S. Let cp (n) (rp (n)) be,., tl1e,.., max1mum pov.cr,., of
~
the set A sa Bn such that Ua. f> U~ 1 {re-sp Ua. n n
P U~ 1) for any two different vee tors in A Shaw that
(i) rp (n) ~ n; (2) q>' (n) = 2"-l
t .. t .27 ~ Let F (n, k) be a fam1ly of subsets A of the
,.,
se;t nn such that Ucc. n~ U k fer any a.. 1\ 1n A Let
~
> ~ ~
f (n,. k) == ma1 r A ' Prove the lolloW!Dg state

A£ F(n ~)
ments
n
(i) q~{n, k};;t ~ ( ~).
(.._j~·tj+ I [
(2) Let. A c F (n. k);then) An Btl +1 An B~-1+~-~~ ~

( ~).
(3) For I;> --!-::'-7 ~- t · tha equaltty I A n Bf I+ IA n
n:-I+A-tl = ( ~) llo1ds only when A n B:~r+A-1 ""'* 0
111
(4) cp (n, k) == ~
, . . .
n+k
:aTT 57
2
1.1. UOOI.P..AN YP..CTORS 19
•
m
'~"'
(n, k) = n+k-1
n-1 + h (1i1) for odd n+k.
? . n+k+1
- 1= ?
1.1.28. Let F r (n) be a family of subsets A s; B" such

....., ~ ,..., ,...,
that p (a, ~):::;;2r for all a and ~ in A. Let CJlr (n) =
max I A I·
•.J.EF/")
(1)* Show that the maximum of I A. I is attained for
~
all A EF r (n) on sets of the type S~ (a) and is equal to

r
~ ( ~).
h=O
(2) For odd n and r = (n - 1)/2, give an example of a
set A E Fr(n) which is not a ball of radius r, and for
which I A I = IPr (n).
1.1.29*. The subset A s B" is said to possess the
property 1 if any two vectors in A have a common unit coor-
~
dinate, property 2 if for any a E A the opposite vector

-
~
a does not lie in A, and property 3 if it follows from

roJ _,....,
ti"'Y ....,
a E A and a:::;;; ~ that ~ E A. Let 1jJ (n) be the number of

subsets A s B" possessing properties I and II simulta-
neously, and cp (n) he the number of subsets B s B 11
possessing property 3. Show that:
(1) 1jJ (n)~2(l< ~~>~2~);

71 (2) 1p (n):::;;<p2 (n-1).
1.1.30. Let A c:B", and R (.;1) be the number of such
~,..., ,..., ~
edges (a, p) that a E A, ~ E B"-......A. Show that I R (A) I~

min { I A I , 2" - I A I }.
1.1.31. Prove the following statements:
(1) The number of different faces in a fixed direction
{i1 , i2 , ••• , i,} is equal to 2".
(2) Two different faces in the same direction do not
intersect.
(3) The union of all faces of the cube B" having a
given direction is the entire cube.
(4) The number of all faces of rank k in the cube B"
is equal to ( ~ ) 2k.
(5) The total number of faces of the cube B" is equal
to 311 •
CH J BOOLEAN FUNCTIONS
(6) The number of faces of dimension "- conta1n1ng a

g1ven vertex ;;: IS equ.cil to { ~ )
(7) The number of faces of d1mens1on k con ta1n 1ng .a
give:P faee of d1menswn l Is equal to {= =!)
(8) The number or k dlmenstonal faces Intercectlng
a g1ven l d1mens1onal face of the cube B" 1q equal to
m-..n<,l I)
~ ( ~) 21-J {~:;)
1-0
t 1.32. \Ve define th6 Interval of the cube nn as a set
~ f'ltJ ~ -.1 ,..,. ,..,
of the form {Y C£ ~ y ~ fl }, where a. and ~ are certatn

,..,.. .,.., l"w ;l'lw
vertices such that a~ P The number p (a, ~) t.s called

the d1.menswn of the Interval Show that a face of d1men
ston k 1s an tnterval of dlmenston k
j • t ~33~ Let lit g It g a be the faces of the cube on SJto w
that the relations fl n g, ==.:: 0, gl n13 f21t l3 n +
gl * ~ lead to the lnequaltty (gl n g2) n g,
1~ 1. 3" Let n,~ ~~ . , n~ be non negatt ve tntegers
#="
such that
'~ ~i~2 11
ln th1s case~ B" contains pa1r
a I
:&::!!!:'
Wise non 1ntersect1ng faces g 1, g1 ~g. whose d1men l
stons are respect1vely equal to n 1, 'lz n~

t.t~35~ The faces g1 and g2 of the cube Bn a~e cal(ed
tncomparable tf netther of the 1nclustons g 1 s g 1 and
t1 s; g 1 1s sat1sfied
(1) Show that there e-x1sts a set of faces of the cube Bn
c.Dnslstmg of ( r
31
[n~~~ ) paU''WISe •ncomparable faces
:;3)} (II
(2) Show that the power of any set of pairWise 1ncompa
rable faces of the cube _Bn does not e:x:ceed ( rn~SJ) X
2$-1"111
1.1 . 36. Let L (n, k) he the m1n 1mum number oi "er
t1ces ID Bn such that each J... dimenstonal face conta1ng at
least one of these verttces Prove that
(l) L (n~ 1) ~ zn-1 t
(2) L (n, n - 1) ~ 2,
(3) L (n, 2) ~ (2"/31.
(4*) m~ L (n, n- 2)~ m + 2t where m IS the smal
lest mt@ger for wb1ch ( {rn.~l } ;;;.: n;
U. BOOLEAN VECTORS 21
{n/lt]
k}~ ~ ( ~ i
1
(5) L (n, ) ;
i=O
(6) L (n, k)~2PI-T>L (r, k), k~r~n. .
11
1.1.37. Show that tuples in B can be expanded mto
"" "" ""
a sequence a 1 , a 2 , • • • , azn which is a cycle.
1.1.38. Show that Bn does not contain cycles of odd
length. .
1.1.39. The binary vector {a 0 , a 1 , • • • , aN_1) ts
called n-universal if for each vector (~ 0 , ~ 1 , • • • , ~n-1)
.in sn there exists a number k such that ~i = aMH (i =
0, n - 1), where k EB i = k + i (mod N). Will the
vector a"" be n-universal if
-
(1) a= (0011), n -- ?·... ,
(2) -
a = (01011), n - - ...?·,
""
(3) a = (00110), n = 2;
""
(4) a = (00011101), n = 3?
1.1.40. Prove that for each natural number n there
exists an n-universal vector of length 2n, and no n-univer-
sal vector having a length smaller than 211 •
11
1.1.41. The cycle Z of the cube B is called a 2d-cycle
""
if I Z nS:l (a) I = 2d + 1 for all a EZ, i.e. if for
""
any vertex ""a in the cycle Z the set of vertices of the cycle
situated at a Hamming distance not exceeding d coin-
cides with the set of vertices at a distance not exceeding
d "along the cycle" .
. (1) Let l (n) be the maximum length of a 2-cycle in
11
B • Find the value of l (n} for n = 2, 5.
11
(2) Let Z he a 2-c.ycle in 8 • Show that any face of di-
mension 4 does not contain more than 8 vertices of the
2-cycle Z.
(3)* Show that the maximum length of a 2-cycle in
sn does not exceed 211 -1, n > 3.
"" ""
1.1.42. The pair (ah aH 1) formed by two consecutive
vertices of the cycle Z in cube B 11 is called an edge of the
cycle. Show that B 5 contains a cycle of length 32, any
four consecutive edges of which have pairwise different
directions.
]
I
'22 Cil t BOOLEAN .FUNCTIO.._.S I
J 1.43. Let a 1 a..a be ~uch numbers that aj > 2 (t ~

.1. . . . . .n)
. . . Show that the number of sums
n
iJ (- 1)~ 1 a, e11 E{0 1), J = fn
1:~1
')1
sattsfJ. Jng the con dJ lJOD , L:

i-l
(- i r:J I a, I~ l does
not exceed {(li;Zl)

1 I 44 Let A s;; nn I A J > 2n-1 Show that at least
n edges of the cube are completely contained 1n A
1.2. Methods of Defining Boolean Func11ons.

Elementary Funct1ons~ Formulas.
Superpos1tao11 Operation
f"'IIJ f't-1
We shall aSSign tbe symbol :r:n (or x) to lhe tuple of

variables (.r, .:r2 Xn} and denote the set of the
~
var1ables by X 11 The funct1on j (ztt) whJch 1s defined on

the set Ir and wh1ch assumes values from the set {0 1 J
ts termed a Boolean (unt:t.wn We .shall denote the set of
all Boolean functtons of vartables z- 1 Zn x.,
by Pt. {X") ,..,..
The Boolean functton I (.t") can be presented 1n tabular
~
form T (/) (seo Table 1) Here the tuples a ar~ arranged

Table t
~
s
"'• rn 1 2:,.,. I(:.:, z,. "=n 1 zJII}
0 0 0 0 f (0 0 0 OJ
0 (} fl \. 1 \0 \)
0 0 t 0 I (O
Q
{l
"J 0)
i 1 l 1 J {1 1 1 l)
1n a.scendtng order oJ the1r numbers Assum1ng such an

arrangement to be the standard procedure .\Q the lol.,.
1.2. l\tETHODS OF DEFINING BOOLEAN FUNCTIONS 23
'""11
lowing, we shall defmo tho function I (x ) through the vee~
~211 • , •
tor (J../ = (rl 0 , (J.. 1 , ••• , (J.. 2n_ 1 ) m wh1ch the coordmat e
is the value of the function f
Ct.t
number i (i = 0, 1, ... , 211 -1).
on tho tuple with <:;h a
,..., ,..,
The symbol N 1 will be used to denote the set {0': (<1 E
,...,
B 11 ) & (f (<1) = 1) }.
~
Tho Boolean function I (x 11 ) can be also presented

through the rectangu lar table Th,n-h (f) (see Table 2) in
which the value f (<11 , 0' 2 , ••• , 0' 11 ) of the function flies
at the intersecti on of the "row" (a1, 0' 3 , ••• , all) and
the "column" (0'/!Ht O'h+2• , • ,, C1n) 1 1 ~ k < lt,
Table 2
0 0 • • • a,,+l • • • i
0 0 • • • C1h+3 • • • 1
• • • • • • • • • • • • • • •
0 1 • • • CTn • • • 1
xl X~ • • • XII
•••
I
0 0 • • • 0 I
••
0 0 • • • 1 •I•
••
• • •
•
• • • •
a1 Oz • • • Oh .,., ...•.•....... "'
I (o)
• • • • • • •
1 1 • • • 1
Boolean functions described in Tables 3 and 4 are

assumed to be elementary functions .
Let us now consider the notation and names of these
functions .
Table 3
X 0 1 ,, t.
0 0 1 0 1
1 0 1 1 0
CH t BOOLEAN FUNCTIONS
fable 4
-X':~ ~. I; t. f, t. 11
'•
,,
0 0 0 0 0 t i t t
0 t 0 1 t 0 t 1 0
{\ (\ 1 t \} (\ 1. {\
1
1 1 I I 0 i t 0 0
(1) The functtons 0 and 1 are called :era (ot 0 funclton)

and untty (or 1functzon.) respectively.
(2) The funct1on / 1 1s called an tdent1ty junctton and
1s denoted by x
(3) The functton / 2 IS called tha negatton o/ x t denoted
by X or lxl and o[ten read as ''nat x"~'
(4) The function / 3 Is called the con}unctlon of x 1 and
x 2 ,. denot~d by :tr&:z 2 , or :r1 ~ 2 or :t1 r 2 or m1n (:t 1 x 2),
and often read as "x1 and x 1"
(5) The function/, IS called tl1e dtsJunct,on o} x 1 and x 2 ,
denoted by Xt V x 1 , or .x1 + x 2 or max (x 1 ~ x1 ), and often
reads as ' x1 or x 2'"
~6) The funct1on /~ IS called tlte exclu.stve sum of x 1 and
X 2 ~ denoted by Z 1 ffi Z.z, and o(ten read as '¥1 plus X2n
(7} The iunctlon / 6 ts called tile equlualence of x 1 and x 2 ~
denoted by .x1 ,..., z 2 ,. or x 1 ~ • x 2 , or x 1 x 2 t and often
read as "zl ts equtvaltnt to z,n
(8) The unction / 1 tS called the lnlpltcatzon of Xt and
z 2 , denoted by .x1 • x 2 , or x1 ~ x 1 t and often read as
4
" x1 zm.pltes z 1"
(9) The functlon / 1 IS called Sheffer's &lroJ...e for x 1 and
:.z 2 , denoted by x 1 1.x 2 , and often xead ns unot both x 1 and
:r.I'
(10) The function f 9 IS called the Petree's arrow of
.x1 and z, denoted by xl ' x 2 ~ and often read as ,.neztlter
z 1 nor x ~,,
Sllmettmes} the functlons 0 and 1 are tteated as
functtons depending on the empty Bet of vartables
The symbols l' &, v'
m ' ;.., , etc are used ln the
notations of elementary functions and are called senten
tzal connectJves
Let us fix a eerta1n (fi.n1te or countably tnfin1te) alpha
bet oj varJables X Let tlJ ~ {fin~}, fin,),.. J he a set
1.2. :.IETHODS OF DEFI;\I NG BOOLEA N FU;\CTI ONS 25
of junctio nal symbo ls, where the supers cripts indica te the
numbe r of sites where the symbo ls can be placed . Some-
times the supers cripts are omitte d if the arity of the
functi onal symbo ls is assum ed to be known .
Defini tion 1.1. A formul a generated by the set cD is
such (and only such) expres sion as
(1) h.. and fi (xi 1 , xi,• ... , Xi 11 ), where In and fi are
functi onal symbo ls of zero and n argum ents respec tively,
and xi 1 , xi~• .•• , xin are variab les in the set X;
(2) /m (~I 1 , W2 , • • • , WJ, where f mis a functi onal sym-
bol of s argum ents and \">I; is either a formu la genera ted
by (J> or a variab le in X, i = 1, s.
In order to accent uate the fact that formu la ~I con-
tains only variables in X (or only junctio nal symbo ls in cD),
we shall write \){ (X) (resp. l2f [CI>]).
Somet imes, formu las of the type f (x, y are writte n
iii the form (xjy) or xfy, and formu la f (x in the form
(fx) or jx. The symbo l f is called a connective.
Usuall y, conne ctives are denote d by symbo ls from the
set Ei = {l, &, V, ffi, ""'• ~. I, t }.
Defini tion 1.2. A formul a generated by the E is such
(and only such) expres sion as
(1) x, i.e. any variab le from the set X;
(2) <l\10, (12f & m), (12f vm>. (\">I EB m), (\"lf""' m),
(\"lf ) Q3), (\">I I !8), (\){ t !8), where \)( and !8 are for-
mulas genera ted by @).
The follow ing conve ntion is usuall y adopte d for ab-
brevia ting. the notati on of formu las genera ted by the set
@) of connec tives:
(a) the outer bracke ts in the formu las are omitte d;
(b) formu la (l \)f) is writte n in the form §f;
(c) formu la (\">I & m) is writte n in the form (\)£.!H) or
(\)Im);
(d) it is assum ed that the conne ctive lis stronger than
any conne ctive of two variab les in @);
(e) the conne ctive & is assum ed to be stronger than any
of tl~e connec tives V, ffi, ""'• ) , j, t·
W1th the help of this conve ntion, we can write, for
~nmple, the formu la ((lx) J>o ((x&y) Vz)) in the form
x ) (xy V z).
"il!ixe if' form of notati on is also used , for examp1e,
X ffi f (y, z) or X1f (x 2 , 0, X3) V xtf (1, ~~ X3).
26 Cl:1 f BOOLE.AN FUNCTJO'\S
(ll.l)
Suppose that each funettonal symbol f, tn the set
(nt)
¢1 has a corresponding funct1on F 1 B -.. ,. B !he con-
cept of the junctlon fPw~ represented by j&rmula '}I gep.era-
ted by the set ¢Jt IS defined by Induction
(1) rf ~ ==- h (X"Jti x 1••
{n i) h
, zJ.,.. ), t en /or each tup e
1
1
((tl ~ ., a.n ) of values of the v.artables x, 1,
1
x; 1 x; the value of the funet1cn (fm lS equal
"' u
to F1 (aI a2 , ctn 1)'
(2) Jf W::.;: I (l!! 1 \!z \! mJ, where fEd>, lllA =

~A (YA 1 y, 2 t YA•tc.) IS a formula generated by .P or
a variable Jn X"' and on each tuple (etJ4lt a~~.. 2 , , ct.\,k)
of the \Olues of var1abJes Y~tt-t Ykl-~~ , Ytt 1 ~ the function
1pw~ 1s equal tc ~4 (k ~ 1~ m}t then
cr 0' ((X tIt a 12 t t a 1I 1' ' Ct At (%A 2' ' Clk 'l , elm I '
am,,_)= F (~t;
am21 , , ~tt~ ~ O.:m)
(het"e F ts the functiOn correspond1ng to tho iunct1onal
symbol f)
If 'i'l = f~n~ 1 (x11 , XJ1 , . , ZJn ), '+'0' 1s usua.Uy denoted by
1
F, (x; 1 , Z 32 , , ZJn. ) If, llO\\ever,
I
'21 =1(\1l ~2f ' ~m),

then rpg- JS denoted by
F (~W 1 {y llt Y12.~ t Y 11 1 ) ' ~it,... (Yml" Ym2' , Yrll• 11,))
The co nee p t of the junctlon fJlw represented by formula 21
generated by the set of connect r1.-es 6 ts 1n trod uced a.s foJ lo\\ s
(l) lbe formula 9l = x v.here :c EX 1.s JU'tl.aposed
to tl1e tdenlllY funet1on f9.t (x) ~ z,
(2) If W :;=; llfB) (or 21 = (lliOU) \\here 0 E {& V $
-
~, . . I t J), then cp 91 =: ~ (resp (j) J == 'Pa Offer~ \\ltere
the .symbol 0 should now be consldered as the notatton for
the correspond1ng elementary Boolean Cunctton see
Tables 3 4)
Let (.r1 z 2 ~
t , z n} be a set of varJab]es ul11cli aro
ellcountered Ill at least 008 Of the fonnu}as f! Or f8
F Ol(llu}as 1[ and ~ are called eguz"'alen.t fnota tton 21 :;::;
•
1.2. METHODS OF DEFINING BOOLEAN FUNCTION S 27
\:8 or 121 = ll:!) if on each tuple (a 1 , a:!l ... , an)

of the values of variables x 1 , x 2 , • • • , Xn the values of
functions !p~ and IP!8 represent ed by formulas ~ and 18
respectiv ely coincide.
Let be a set of functiona l symbols (or sententia l con-
nectives) , and P be the set of functions correspon ding to
them. The superposition generated by the set P can be
defmed as any function F that can be obtained through a
formula of the set .
1.2.1. What is the number of functions in P 2 (X")
that assume the same values on opposite tuples?
11
1.2.2. Find the number of functions in P 2 (X ) which
assume opposite values on any pail· of adjacent tuples.
11
1.2.3. Find the number of functions in P 2 (X ) that
11
assume the value 1 on less than k tuples in 8 •
1.2.4. Using the functions f (x1 , x~) and g (x 3 , x4 ) spe-
cified in the vector form, write the function h in the
vector form:
~ ~
1) a 1 =(1011), ag=(100 1),

h (x 2 , x 3 , x~) = f (g (x 3 , x 4), x 2 );
,...,
ag= (1001),
,....,
h (x 4) = f (x 1, x 2) V g (x 3 , x4);
~
• ,...,
3) a 1 = (1000), ag = (0111),
,...,
h (x ) = f (x 1 , x2) & g (x3, x~).
4
1.2.5. Let v 1 be a number having its binary expansio n

in the form of the tuple (x1 , x 2 ), and let v 2 be a number ,...,
with its binary expansio n in the form (x 3 , x 4 ). Let fi (x4 )
be the i-th order of the binary represent ation of the num-
bet· I v 1 - v2 I , i = 1, 2.,..., Construc t ,...,
a rectangu lar table
nl,2 of the functions /1 (x4) and ,....,
/2 (x4).
1.2.6. (1) The function f (x 3 } is defined as follows:
it is equal to unity either for x 1 = 1, or when the varia-
bles X2 and x 3 assume different values while the value of
the variable x 1 is less than that of the variable x 3 • Other-
wise, the function is equal to zero. Compile the tables
T (f) and nl,2 (f) of the function I (xa) and write down
the tuples of the set N 1.
•
CH I BOOLEAN FtJNCT J0"\l S
""-
(2) The funt\lon j (xt} 1s d~fiJ'~d ,.., as follo-V¥s 1t 1s

equal to t.ero only on sucll tuples o:. == (a1 rx 1 ~ a.,""' a..,)
that sattsfy the 1nequal1ty aJ + a 2 > aJ + 2cx, Com
p1le the table T U) and wr1 te d ov, n the tuples of the set
N 1 of thts function ~
1412.7. The funet1on f (zll) JS c.alled symmetrlc tf

f (x1 , x!, ~ x~} == f (x. 11 x.h:' , .t1n) upon any sl)bstl~
tUtJon (
l 2
IJ i•
n}
ln
r-.1
( t) Show that 1f f (xn) IS a syrometr1c func tton, the

rw ,..,.. ~
equaiJt:t
,..,
tr a" U= U~,. J1 leads to the equal1 ty f (ex") ===
I(~"}
(2) Ftnd the number of symmelr1t funct1ons tn P'J (Xn)
~
1~2~8 Let I (xn) be an atbttrary Boolean funGlton

(n ~ f) 7
'\ e associate w1th 1t a symmetr1c functton
........,:
Ym)~ m = 2 1 S(a.'") ~
11
S(yi, !12
,.., ~
where
.!lw
-
/ (~ )
11
1f Uam Jl = v (~n) Prove that
S lxh ' :J:,, Zt,
.. _ I
, x2, , zl F
'II' 2
Z times
1. 2.9. 'Vh1ch of the followJng expressions are form u
)as genera ted by the set of sen tent 1al connect 1\ es
<l~ &~ v~ l •f -,
(1) :r • y, (5) (x - • (y & (l x))),
(2) (x &) ly, to) {x & y) l:,
(3) (z "' - y) (7) ( l z - . _ z)?
(4) (y .- (x))t
t .2 10 In l~ow many V\.a')~ can brac~e's be arranged
1n the e>..press•on A :so that a formula generated h) the set
{l, & V ~ } of sen tent tal connect1 ves IS obt.a 1ned e8ch
lime tf
(t) A = lz -• y & z;
(2) A ::::: X & y & l l % x; v
(3) .A = x .. • lv __.. z & lx?
1412 1 t .. The complexllJJ oj a formula geoerated by the
set ofsententtalconnect\\ rs 6 ls defined as the number nf
1.2. METHODS OF DEFINING BOOLEAN FUNCTIONS 2!)
connectives in it. By induction with respect to the com-

plexity of a formula, show that in a formula
(1) of nonzero complexity, there exists at least one pair
of brackets;
(2) the number of left brackets is equal to the number of
right brackets;
(3) two connectives cannot be put next to each other;
and
(4) two variable symbols cannot be put next to each
other.
1.2.12. The connectivity index in a formula is defined
as the difference between the number of left brackets
preceding the connective under consideration in the for-
mula and the number of right brackets preceding this
conne-ctive. Prove that any formula of nonzero com-
plexity generated by the set { l, &, V. ~ }
(1) contains a single connective of index 1;
(2) can be uniquely presented in one of the following
forms: (l~), (~( & \8), (~ V \8), (~ ~ \8), where ~ and
m are formulas generated by { l. &, v. . }.
1.2.13. Let us consider the formulas generated by the
set of connectives {l, &, V, ) } in a form without bra-
ckets. Instead of ( l x), (x & y), (x V y) and (x .. y), we
shall write lx, &xy, Vxy, ~ xy.
For example, the formula ((l x) ) (y V (l z))) will
assume the form ) lx V y lz. Prove that if each of the
connectives &, V, ... is evaluated as the number + 1,
each val'iable symbol as the number -1, and the con-
nective l as zero, the expression for A in a form without
brackets will be a formula if and only if the sum of esti-
mates of all occurrences of the symbols in A is equal to
·-1 and if this sum is non-negative in each proper initial
segment of the expression A.
1.2.14. Find out if the expression A is a formula gene-
rated by the set <l> if
(1) A = j<2J(g<2J(x, y), j<2J (x)), <l> = {1<2J 1 g<2J 1 cp};
(2) A = )x (cp< 1>, <1> = {g, cp};
(3) A = cp< 1>(f< 2>(1, x)), Q> = {f<2J, <p<ll};
(4) A = g<2l( cp(l), j<3J (x, y, cp(ll)), <1> = {1<3), g(2J, cp! I)};
(5) A = (cp<ll(j<2J(x, cp<ll(x)))), = {1<2J, cp<lJ}.
~
1.2.15. Find the vector o:ljl of the function cp represented

by the fo~mula mgenerated by th~ set <l> ={til) I ~~2 ) I g(2)} if.
the functwnal symbols /~u, /~21 , g<-> are put in correspond-
30
enca w•th Boolean functtons defined by vectors (10),

(1011), and (1.000) respect1vely
(1) i2I == fzt)(/~ u (g(~l(x, /~ u (y)) ), y)'
(2) 21 === ct''Ui'~U~'}(x t y)), ct") (.r /~ . . ~ (y)))
{3) 0C = f1°(J~t., (x~ tr )([~ J( x~ Y)~ /~ (y))))
2 1 0
1 2.16 Comp1Je the tables of funct1ons represented hy

ll1e Jollowttlg lorntulas (generated by the set ol conn~c
tl\ es 9'}
(1) (I ' )J. y) EB ((y - ,. z) ffi (z .. .. z))'
(2) (; vy)
- - v (z z) ~ (x ,._, y)
(3) X · , (z ~ (y ffi- xz)),
(4) (((.x t y} i
z) ( y) ~ z
f 2 t 7 A formula genera ted by a set 6 IS lied ldentl.- c,
talfy truth (tdenrzcalfy jal3e) 1f the fun~t1on rep1esented
by 1t 1s equal to 1 (resp zero) on any tuple of values of
variables Ftnd out whtc}l of the follow1ng formulas are
1d~nl1cally truth a.nd wlllCh a1e 1dent1eally false
( 1) (X ~ Y) · ~ ((X V Z) - * (y V %) ) t
(2) ((r e
y) ~ z) (x .. yz)t
((X v y) ~ {x e y)) e (x , ,. y ~{ivy)),
------~
(3)
(4)
-
{(z V y) z - ,.. {(x ~ z.) EB y}) (x(yz})
1.2 18 F1nd out Jf the follow1ng formulas 21 and Q3 are
equ1' alent
(t) '! = ((z V y) V z) - ... ((x V y) {x V t)), \8 :== ~t,
(2) W = (x .... y) • :z ~ = x ..... (y -~ z),
(3) 21 = ((x ED y) J-o (x V y)) ((z - • y) -.- (x ffi y))
m.~xty
(4) 21 -:;;: (.r ;.- y) V ((x ,. z) y), tll = (xy)
...- - -
(y - • xz)
1. 2 19 F1nd all funct1ons whJch depend only on
var1ablss (}{ t.he set {xt y} and are superpos1t1ons ge-nerated
by the set P
(t) P ~ {u1 $ U5u 1}-t
(2) P ~ {u 1u 1 Gl (u1 V u 1J },
(3) P ~ {/ (u1 , u 2) = (1101), g (ul, u 2 u.,) t
= (100101 tO) )1
t .2 20. Depth of a formula '! generated by the set W
(notat1on dep4l (~) 1s determtned by Indl.lctton
(I) 1f ~ 1s the s-ymbol of a var1able or a functional s3-m
bol of 0 arguments~ then dep¢t w ::.-= at
• Here and below~ while presenting a ru.nctlOD lD the vector
f"W #"'t.J ,..,
f arm \\f:' thall use t he notat•on f ~ 6 1nstead ci ~ ::::::= 6

t.:!. ::IIETHODS OF DEFimXG DOOLRAX FUNCTIOXS 31
(II) if ~( = j< >('2! 1 ,

11
11
••• , '21 11 ), where /< > E cll, then
dep (12I) = max dep<ll (121: t) + 1.
I ""i..;n
Find the formula '21 generated by the set {t}. having

a minimal depth and representing the function
(1) f = x V y; (2) I = x EB y; (3) I = xy.
1.2.21. Find out if a function I can be represented by a
formula ~( of depth k generated by a set of connectives S
if
(1) I = xy, k = 2, S = {I};
(2) I = X ~ y, k = 3, s = {V. ""'};
(3) I = .l: EB y ffi z, k = 2, S = { ~ , & }.
1.2.22. Prove that a function f cannot be represented
by a formula generated by a set of connectives S if
(1) I = X EB y, s = {&};
(2) I = xy, S = { >- };
(3) I = X v y, s
= {,...., }.
1.2.23. Is it possible to represent a function I by a
formula of depth k + 1 generated by a set S if it can be
represented by a formula of depth k generated by the
same set?
1.2.24. A function I in P 2 can be represented by a
formula of depth k generated by a set S. Show that the
function f can be represented by a certain formula of
depth more thank generated by the same set S.
ln operations involving Boolean functions, it is of-

ton expedient to use the following equivalences (we shall
call them basic equh·alences in the following):
x 0 y = y 0 x (commutativity of a connective O,
where the symbol 0 is a common notation for the con-
nectives &, V, EB, ~ , I, .j,;
(x 0 y) 0 z = xO(yOz) (associativity of the connective
0, where 0 is the common notation for &, v.
EB, ......,);
y
x & Y = ~ V and x V y = x
& y(De l\Iorgan's laws);
x V (x & y) = x and x & (x V y) = x (absorption
laws);
-
X V (x & Y) =XV Y and X & (XV y) = X & y;
x & (y V .::) = (x & y) V (x & z) (distributiv ity of
conjunction relative to disjunction);
. x V (y & .::) = (x V y) & (x V z) (distributivity of dis-
JUilction relative to conjunction);
X &. (y ffi z) - (z &. y) ffi (.z & z) ( dt~\r1bultv 1t y ol
COfilUnCliOU reJat1ve to CxcJU~IVe ~UJU)
0-x&'i--x&O--JtBx
1=xvz-xV1==-z,....,:z-
-
x=~~xvx x&x-x&t-xvo
x --- x G1 1 x ~ y ---- (x ffi y) $ 1
x ED y - (x & y} V (i & y) x - *' y --- ((x &_ y) ti.1 x)
$1
1 2 25 Ver1fy 1! tl1e Iollowtug re1attons are val1d
(1) z V (y ~ z} ~ (.r V y) ~ (x V z)
(2) X -:.. (y .,....., Z) - (X y) ~ (X ~ Z)
= ,_
(3) :c & (y;..., z) ~ (.t &. y) ..._.. (x &. z)

(4) x - ..- (y V z) -- (x fF y) V (x ..... z)
(5) x - .- (y & z) - (x ,. y) & (x - tr- z)
(B) X ID (y • .z) -- (.r ea
y) ,. (x $ z)
(7} x l"" (y ,.. .z) --- (z --... y) •· (x _, z)
t 2 26 Represent the functton I by a formt1la gener
ated by t}Le set of connectives S tf
(t) f - X -·t-oy s - l V}
(2) I .._. X y s-v .. }
(3) I - X~ !I s ~ {&. .. )
(4) I - X l y S' ~ {4}
t 2 27 Us1ng the bas1c equtvalences pro\e the equav
alence OJ forro u}as l'( and m If
(t) ~ ~(x& ) V (x&y) V (x&.t)
v7 & z
£8 ==:t&.y& z
(2) ~ = (x- y) .. ((z & y) EB (x ~ y)}
=)o
~ (x V y) & (z v y)
(3) 2{ == z J.r (.r & y .._ ((.r -~ y) - :. y) & z)
my -· =:;:: (z = ,_ z)
I 2 28 Let ~ and 23 be formulas generated by the
~et {l V m ,...., -Jt= 1 ~ J and~.= (2! __ ,. ~) _,
&
58 ~., -- '21 _.., (~ -a-- l (tt -• ~ )) Prove that 11 ....... m==
1 t[ and only tf l (ill. 1 ~ ~~) --- t
I 2 29 Prove that formula 121 contaan1ng only the con
nectl\ e,..., IS tdent1cally truth 1f and only tl any varra
ble from formula £{ appears 111 tt an even number o(
tlmes
i 2 30 By S~2I ~ \\e denote the formula obtained
[rom formula ~ as a result o[ a s~multaneous rep~ace
l.:l. SPECIAL FORMS OF FORMULAS 33
ment of each occmrence of variable x by formula $.

If formula ~ does not contain .x, then by definition we
put s~~I I = ~.
(1) Prove that ~{ is identically truth if and only if
the formula s~.-+Us ~ I is identically truth, Yt and Y2
being variables not appearing in formula ~.
(2) Can the condition that y 1 and y 2 do not appear in
formula ~ be neglected in (1)?
1.2.31. (1) Suppose that the function I (x, y) from P 2
satisfies the relation
I (f (x, I (y, z)), I (f, (x, y), I (x, z))) - 1.
Prove that the foUowing equivalences are truth in this
case:
(a) I (x, x) = 1;
t t
(b) (x, (y, x))- 1;
(c) I (f (f (x, y), I (x, z)}, I (x, f (y, z))) 1;
(d) I(/ (x, y), f (f (x, f (y, z)), / (x, z))) 1;
(e) f (/ (x, f (y, z)), f (y, I (x, z))) 1.
(2) Do the equivalences (b), (c) and (d) follow from (e)?
1.3. Special Forms of Formulas.

Disjunctive and Conjunctive Normal Forms.
Polynomials
The formula xf.' & x?: & ... & x~r (formula x?,' V
'r
x~: V ... V x~~), where all E {0, 1 }, xYh = x;h,
xlh =xi"'. ih E {1, 2, . . . , n} for all k = 1, r, is
called a conjunction (resp. disjunction) generated by the set
of. ~ariables X 11 = {xl, x2, •.• , X n }. A conjunction
(disJunction) is called elementary (abbreviated as e.c. and
e.d. respectively) if xi =1= xi for j =1= k. For the sake
of brevity, the symbol {& in a~ e.c. will be omitted. Ex-
pressions of the type x~h will be called characters. The
h
number of characters in an e.c. (e.d.) is called the rank
of the e.c. (e.d.). The constant 1 will be assumed to be
the e.c. of the rank zero, and 0 will be called the e.d.
of the rank zero.
Formulas of the type
:JJ - Kt V K, V VK, (1)
,.
(con~1se nn\at1on V K 1), wh~re
K 1 (1- =- 1, s) are con]u.nt
·~·
\1onst are ealled tlu~ dtsJUnct' Lte n-orma t fortn ft ( abhre\ 1
a ted as d n I)
Formulas of the type
JC-=D 1 &D 3 & &D~ (2)
(conctse notat1on '& D 1) where D 1 (i == 1:=S) are diSJune

-· ~
i~t
t1ons, are called conjunclttt normal forms (abbreviated
as c n f ) Tile number s 18 caUed tlte length of the d n I
(c n f) The sum of tlke ranks of conJunctions (dlSJUUctton.s)
ts called the complez~ty oJ d n f (c n f) Tl1e d!sJuncll\e
(eonJUDCtlve) normal form generated by tJ1e set X'~ ::::
{x-1 1 X 1 , ;x n } of vartables JS cal led perfect tl 1t
1s formed by paltw1se different elementary eonJUntt1ons
(dlSJunct1ons) of the rank n
,f¥
Let I (xn) be a Boolean funct1on and 16t f ~ 1 1 < lz <

i:l t. 1l ~
< '" ~ n By /o a
(or somet1mes by
1 1 o• (z~)
i.l i.l tik t'¥
S d.a. G_. I (zn)) v. e shall denote tl1e funct1on obtained
,..,
from f (xn)
by subst1tut1ng the constants a2 • a,. a,.,
res}W!etlvely fer the var1ables ~ 11 , 3:i,. ~ lriA The
[unctinn f i..0 ,o-.i: a 0 Jt.~ ~ {x,.....n ) 1s eal I ed the xi:a x a~..• aA
x ,11 w
,.,., ~
component oj the junct!on f (xn) or a subfunc~ton (J.f f (xn}

Th e su bf unc t1on i:. fa
fa"' :a, a-.
•• ......on
(x ) 1s called pro per 1
f k +n
k +0 The subfunct1ons of I (.x•) are d~Oerent 1f they dtffer
r'leF
as funct1ons of the var1ables .r1t z 2• , r,.

Let 1 ~ l 1 < :11 < < 11 ~ n In t hls ease, tlle f ol-
lowing representation 1s valtd
,.,.,
f (xn) = V xf; ~•: (3)
(a,. o. dA)
where the dtSJUnt\lon lnt1udes all \ettors (a., o 2 o"')
10 B" For k = n th1s represen ta ttan has the form
~
f (.xn) == V ' a!\) (4)

(01 cr.
{.3. SPECIAL FORMS OF FOR,.,IUI,AS 35
Representation (4) can be written in the form

"'
f (xn) = V x1'x~· ··· (5)
"' ""
a: /(a)=l
The right-hand side of this formula is the perfect d.n.f.

.....
of f (x")· Similarly. the following representations are
valid:
{6)
-
I ("'n)
X = & (Xta, v a. v v
X2 , •,
0
X 11 n) • (7)
7?1: j{;J 11 )"'0
An elementary conjunction is called monotonic if it
does not contain negations of variables. The formula
,..,
P (x") = K 1 $ K 2 $ ... $ K8 , {8)
where K 1 {i = 1, s} are pairwise different elementary mo-
notonic conjunctions generated by the set X", is called
Zhegalkin's polynomial or mod 2 polynomial. The highest
rank of elementary conjunctions appearing in a polyno-
mial is called the degree of this polynomial. The numbet·
s is called the length
,.., of polynomial (8). For s = 0, we
assume that P (x") = 0.
1.3.1. Using equivalent transformations, reduce the
following formulas to d.n.f.:
(1) F = (x1 V x 2"X3 ) (x1 V x 3);
_(2) F = ((x1 V x;i3x 4) ((x 2 V x4) ~ x1"Iax4) y XzXs) V
(xl V xl); '
(3) F = ((xl )o XJXs) {x2x4 $ Xa) ~ xlx4) v xl.
1.3.2. Present the following functions as a perfect
d.n.f.:
-
(1) I (x ) = (x1 ED Xz) )o X2Xs;
3
"' = (01101100);
(2) 1 (r)
"'3) = (10001110).
{3) f (x
:!•
CR I BOOLEAN FUNCTlO'lS
1 . 3.3~ Ustng transformations of the type A == AzV ~
A~ A VA =A~ reduce a g1ven d n f D (r) to a

perfect d n f 1f
,.,1
(1) D (x ) = :t1 V :t2X3t
-
"'¥ - ~
(2) D (r) :=: :r1.r1 V xlx3,
,., -
(3) D (r) -= x1 V .z1xt V J: tXs
-
f 3 .. 4. Ustng relattons of the: type :r V yz = (z V y) X
(z V z), transform the d n f 1n the prev lOUS problem to
a t. n f
1 ..3.5 Construe t a perfect e n f for each of the func-
tions •n problezn 1 3 3
l'w
l 3 641 Count the number of funct1ons I {x~) for whach

a perfect c n f iS Simultaneously a d n f
I 3 .. 7 F1nd the length of the perfect d n f of the func
r..s
tlOD I (.t")
,..,
(1) I (xn) ==: XI e x2 ED ED z~, n~ t,
{2)
~
I (xn) = (x1 V x. V V z,) -(.:tl V XI- V

......
vx,.) n~ 2,
(3) I rf'¥ 11
(x ) .=:: (r1. v x, v .x1;) (x1 v .rf. v x,) e
- - -
X( e
x1 Ell $ .t~t n~ 4
I ~3 8. Let us suppose that the sets X" ~ {x1 xt, ,
.:r-.} and Y"' == {y1 .. Yt , Ym.} do not Intersect
,..,
Assumtng that the perfect d n f s of functtons I (xn) and
I'W
g (y"') have k and l terms respectively, find the length

of the perfect d n I of the follow1ng lunct1ons
{{) l 1'¥
(~n) & ~{ym}t

~
{2} J (:t\'l} V g (yrn},

""""t' ,.,.
(3} t (:z:"\ $ g (ym\

1 3419 .. Present the Zt ,.z1 and xlxa components of the
~ -
Jlllt.,l
function I (x') by us1ng the mlntrual complextty d n f for

1f'llfl#
(t) t (r) = {Oitottot),

r-.# -
(2) I (zl) = (z1 · ... .z 2) E£l z 1x 3 ,

.,..,
(3) I (r) =(z1 -;. :r.zs) $ x.
-
t ..3 ~tO., Ftnd out whleh of the funet1ons d~pend1ng on
var1ables .r1 and x1 have largest number of patrw1se d1 ffer-
ent .suhlunc.tloo.s
1.3. SPECIAL FORMS OF FORMULAS 37
~
1.3.11. Find the number of Boolean functions I (xn)

which are transformed into themselves upon a commuta-
tion of x1 and x2. ~ ~
1.3.12. Two functions I (x11 ) and g (x11) are commuta-

tiL'ely eqllivalent if there exists a permutation n of numbers
1, . . . n such that I (x 1 , • • • , X 11 ) = g (xnc1> , ••• ,
X;~(n>). Find the number of classes of commutatively
equivalent functions in P 2 (X 2 ).
~
1.3.13*. The function In (xn) is defined by the fol-

lowing recurrence relations:
~
/4 (x 4) =
-X1X2 (x3 V x4) V X2- (x1x3
-- V x1x4) V X1X2x3x4,
--
ln+l (x +1) = fn (xl,
f'OJ - - - -
71
X2, • • ., Xn) Xn+l V X1X2,
(n;;:;: 4). ••• X 11 Xn +1
For each function in the sequence {/ n} find the number of
. ~
different subfunctions of the type f ~ (x"), i = 1, n,

cr E {0, 1 }, such that no two functions are commuta-
tively equivalent.
1.3.14. Prove that the number of different functions
"'11 "'
I (x ) for which a given function g (xk) is a subfunction is
not less than
22n- (2zk -1)zn-k (k~n).
0 ~ • ("tJ
1..3.15. Let I~ (x11) =I: (x11) for any i (1~i~n). Prove

"'11
that I (x ) is a constant.
"'11
•
1.3.16. Find the number of functions I (x ) such that
0 ...... • • l"oJ
lo'o (x")=/i't 1 (x 11 ) ,....,for all 1~i<j~n

1
,...,
(n~3).
,....., ""
1.3. 17. Let h (x 11 ) =I (gdx 11 ), . . . , g 1_t(x11 ), g!(x11 ),
~ ~
gi+t (x ) , • • • , g 11 (x")), n;;;: 2, and the following relation

11
is satisfied for any i=i, 2, ... , nand uE{O, 1}:

sjah <"'n)
X
I -I
-
(Si (~n)
ogl X I ' • • •'
Si(jgi-1 <"'n)
X
I' -U,
i - 11
!, ... ,
• ,-J
S~gi+i(x") S 0g 11 (x ) 1),
in other words, the x?-component of the superposition
of the functions I, g 1, ••• , g 1_ 1, g 1, gi+l• ... , g 11 is
-
equal to the superposition of the x?-components of the
CH f BOOLEAN ll1NCTIO'JS
functions 1. g1, , g,_ .. g 1+1, • ' " l\loreover, let

a G1 ~
the .x;lz-~ --component of the functton g, (z'Q) c01nC1de
w1th the x1
q -
a
J,x_,.JI.-eomponent of the same function g 1 (z"),
~
I= i:--
n, lor all It k {1 :s=;/ < k~n) and f1J~
jllllfll
al from
{0, 1} Prove: that h { z") 1s a constan l
1 3~18 . . Find the number of monotonte elementary eon-
JUn<:tton.s of ttle rank r generated by the set X"
l 3.f9. Ftnd the number of polynom1als of power r
generated by the set of variables X"'
1 3 20 F1nd the number of d1fferent polynomials of
length k generated by the set Xn, vanlShtng at the tuples
,.., ill.,;
0 and 1 (Palynomtals are eonstdered to be dtfferent tf

they d1ffer tn the composition of e e )
\Ve adopt the follow1ng numeratiOn of monotontc ele
menta.ry CODJUnCtlOnS generated by the set Xn :::::::
{.r1, Z 1, Zn} of var1ables \Ve correspond each
rtW'
monoton1o e c K wtlh a vector a (K} === (o 1, a1, ..

a..,J 1n B", tn whtch Ut == 1 1f and only tf ,... Zt ap ..
~
pears 1n K Tb.e num her K of an e e 1s defined as" (a {K)) :=:;

n
~ a 12"-" The constant I w1ll have the number zero
i.-t ,..
In tbts numerat10n Thus, each polynomial P (x"') can
be presented In the form
where K e IS an e c w1th number 1 (1 :::= 0~ 2" -- 1)

~
The veetor ~P = (~ 0 ~ 1 , ~ n
2 1
) 1s called the
,-w
t-'tctor oI coeIPt: l ents of poly ntJm~ a l P (zn}

The method of tndeterm1natt coe{/itlenta for construcllng
..,
Zhegalk1n s polynomial reptesent.lng the func.t1on f (x11 )
can be descr1bed as follows We constder a polyt:tcm1al
tn the form (9) and for eaeh a E we compose an equa-
~ """
~
n•
tion I (a) === P (a) The solut1on of these equat1ons g1ves
1-¥
the coeffictcnts of the polynom1al P (.t")

1.3. SPECIAL FOR4\IS OF FOR:\1ULAS 39
"' ......2
Example. I (x 2) = x 1 l- X2, P (x ) = ~o ED ~txt ED
~2x1 ED ~aXtXz.
I (0, 0) = 1 = ~o ED ~~ ·0 ED ~2 ·0 ED ~3 ·0;
I (0, 1) = 1 = ~o ED ~~ ·1 ED ~ 2 ·0 ED ~3 ·0;
I (1, 0) = 0 = ~o ED ~ 1 ·0 ED ~ 2 ·1 ED ~ 3 ·0;
I (1, 1) = 1 = ~o ED ~ 1 ·1 ED ~ 2 ·1 ED ~ 3 ·1.
We obtain ~ 0 = ~~ = ~3 = 1, B1 = 0. Hence x 1 ) x2 =
1 ED x 1 ED x1x 2 •
1.3.21. Using the method of indeterm inate coefficien ts,
find Zhegalkin 's polynom ials for the following functions :
,..,
(1) f (x 2
,..,
) = (1001);
(2) f (x,..,3 ) = (01101000);

(3) 1 (x3 } = (11111000).
· 1.3.22. We introduce the operation T generated by
n #'OJ -
vectors in B • If n = 1 and a. = (a. 0 , a 1 ), then T (a.) =

2
- n
(a 0 , a 0 ED cc 1 ). Suppose that for each a E B 2 the
211 1
vector,.., T (u) is defmed and the vector ;: in has the B +
form CG = (B 0, BP ... , ~ 2 n_ 1 , y 0,y 1, •.• , y 2n_ 1). Let
T(Bo, ~I• . . . , Pzn-I)=(O o, Ot, ... , 02n-l),
T(yo, '\'t• ... , '\'za-l)==(eo, el, ... , e2n-t)•
In this case,
,..,
T(cc.)=(Oo, 01, ••• , 02n_l' 00 ED eo. 01 ED el,
... , 02 n_ 1 ED s2 n_ 1 ).
,.., ""
Fot· example, if cc. = (1011), we obtain T (cc.) = (1101).
"' of the values of the function
Show
,.., that the vector cc. 1
I(x ") is related as follows to the vector B""P of the coeffi-
. ""11
ctents of the polynom ial P (x ) represent ing the f11nclion
,...., "' f'-.1 ,...., -
CGJ = T (~p}. ~P = T (cc.J)·

11
I (x }:
"' ,..,
1.3.23. Fo1· a function I (x~} such that a.1 =
(1011001000101101) find the vector ~P of Zhegalki n's
polynomi al coefficien ts.
C1l 1 DOOLEAN FUNCTIONS
One of the methods oJ con1trucl1ng Zhegalk1n s poly

nom1al "from formula F ~nvolves tile construet1on of an
equtvalent formula- generated by the set of connectt' es
{&~ - }~ followed by the replacement of X by X ffi f
-
throughout, the open1ng of the brackets the appltcatton
of the d1str1hu t1 ve law (z EB y) z ==- .t: $ y: and the col·
lect1on of tertns
Example.. z 7
.x 1 .=;; :r1x 2 ~ x 1 (x 2 $ 1) ffi 1 =

•
;zl:t2: m.:tr $ 1
t 3 24 Construct the polynomtals lor the followtng
funCtlODS
t
~
(1) f(r)
.......
= (x1 l z"l) x3-.
(2} f (x ) == {x 1
3 7
XJ) {x 2 ~ x3),
...
IV -
(3) 1 (.tt) .= ((xt +- x2) V Z3) 1X1
t 3 25~ Any Boolean funct1on I cs.n he wr1tten ln the
form of a polynomial by us1ng conventional arit}lmettc
operations of mult1pltcat1on, add1t1on and subtr.actron
For th1s purpose 1t JS sufficient first to express /tn terms
of a cOnJunctton and a negat1on and then replace ~ub-
formulas o.f the type A by 1 - A and open the brackets
Us\ng the arlthme\lc opetat1ons find the expression:! for
the following functtons
~
(1) f (r) -= ~l e x2
(2) 1 (r)
,.,
~
= (xl ,.. x2) . ~ x,.

(3) I (~) = (10000001)
~
1.3 26. F1nd the function j (x11) whose polynomtal has

a length 2n \tmes the length of 1ts perfect d n f
1 3 27 ~ Prove the val1dlty of the followlng formula for
ex pans1on 1n k variables
1~3.28~ Prove that

Jlillol ~ I"W ~
(1) f (zn) ~ tr1 (f! (xn) tB /~ (.rn)) EB /! (.xn)t

f'W ,.., .......
1; (.rn))}...... ~ /~ {%11 )
~
(2) f (xn) = (zt V (f~ (.t 11 ) ;-v
1..3.29 Show that the funetton f (a:n) represented by a

polynomial of degree k > 0 becomes equ.al to 1 Q11 at
)east 2n-A VeCtorS lll 8 11
1 3. SPECIAL FORJIIS OF FORMULAS 41
1.3.30. Determine the number of tuples in Bn on

"" 11
which the polynomial P (x ) becomes equal to unity:
"" = x ... xll ffi xh+l ••• Xn (1 ~k < n);
(1) P (x")
,..., 1
) = 1 $ x 1 $ x 1x 2 $ ... $ x 1x 2 ••• Xn =
11
(2) P (x
n
1$ ~ X1 , , • X1•
i=1
1.3.31. Show that for any l (l~ 211) there exists a
,...,
polynomial P (x11) of length not exceeding n, so that
I N P(x"'n ) I = l.
,...,
1.3.32. Prove that any function I (x 11
) other than 0 at
3
n = 1 can be presented in the form I {;; 11

) = ~ K io
i=1
where K 1 (1 = 1, s) are elementary conjunctions con-
taining not more than one negation of the variable,
and s ~ 211 - 1•
1.3.33. Show that if the symbol V in a perfect d.n.f.
is replaced by $ everywhere, a formula equivalent to
the initial one is obtained. Is this statement true for an
arbitrary d.n.f.?
""11
The derivative of a Boolean function f (x ) with respect
to the variables x;,, x;,, ..• , x;h (or the Boolean differ-
ence) is defmed as the function
-
,..,
(For k = 1, the notation a;x1,(:z:n) is used.)
1.3.34. Prove the following properties of a derivative:
""11 ,...,
(1) d l' df (x ) ) = d ( dj (xn) ) •
dxJ dxi dxi dxJ 1 '
(2)
""
iJf(xn) = - ""
cJf (xn) ;
iJ {xi,• ., , , X;h) 0 (:z:; 11 ... 1 X;h)
42 CU t SOOLE.A.N Ft.TNCTIONS
III'W' ,.,
m d~~~) ~~~~.,
.,., .,., fl'v .,.,
(5) ! (( !~'lt (.t_nu =I (i•) .d~J~nJ_ EB 8 (~) df (~1l)

dzt dz~ dz 1
,.,.
{b) d~(~) =0 1-l and o-nly 1f ~~ does not appear expllc.
Zl
....
1tly in the Zl1egalk1n polynomtal of the function f (~"').
,.,;
(7) 1f / (z.,.,.) == %1g (:r2, Za,
'~4- Minimization of Boolean Functions

The admissible conJunction or tmplican.t of the tunc
N
t1on I (x,) 1s an elementary con)unctton K of the set of
N
VS.t1a.bles {:c 1 ~ X 1 ,
.;..,
l a:n J,. sueh lhat K VI (xt•) ~
1 (x") The 1mpl1cant K of a funetton f 1s called prlme
1f th~ reJettlon of any ehatac.\er from K leads to an ele
mentary conJunetJ..on that lS n0:t an lmpllcant o{ the rune
tton f The dtsJunctlon of all prlme 1mpl1ea.nts nf the
funct1on I It; called a contracted d n f of the funet1on f
U. MINir.UZATION OF BOOLEAN FUNCTIONS 43
A disjunctive normal form is called

minimal if it has the smallest number of characters
among all equivalent d.n.f.s;
shortest if it has the smallest length among all equiva-
lent d.n.f.s;
terminal (irredundant) if the omission of any term or
character leads to a non-equivalent d.n.f.; and
a d.n.f. of the function f if it represents the function f.
If an elementary conjunction K is an implicant of the
~ ~
function I (xn), the set N K of such vectors a in Bn for

......
which K (a) = 1 forms a face belonging to the set N 1•
......
This face is called the interval of the function f (xn) cor-
responding to the implicant K. The interval of the func-
tion f which is not included in any other interval of the
function f is called the maximal interval. Maximal inter-
vals correspond to the prime implicants of the function f .
....
1.4.1. Isolate prime implicants of the function f (x'')
from a given set of elementary conjunctions Gr. if
- - "'
(1) ~· = {x1 , x 3 , x 1x 2 , x 2x 3 }, I (r) = (00101111);
- ....
(2) tit.' = {x1x 2 , x~3 , x 10 x 1x 2x 3 }, I (r) = (01111110);
- - --
(3) 1!/C = {x1 , x~, x 2x 3 , x1 x 2x 3 },
-
I (x") = (1010111001011110).
Blake's method of obtaining a contracted d.n.f. from
an arbitrary d.n.f. involves the application of the fol-
lowing rules: xK1 V xK 2 = xK 1 V xK 2 V K 1K 2 (gen-
eralized pasting) and K 1 V K 1 K 2 = K 1 (absorption).
It is assumed that these rules are applied from left to
right. At the first stage, the operations of generalized
pasting are continued as long as possible. At the next
stage, the absorption operation is carried out.
Example. Obtain the contracted d.n.f. for ::tJ (~) =
X1X2 V
-
X1Xa
-
V X2Xs·
After the first stage, we obtain
:JJ1 = X1X2 V - X1Xs V X1.Xs V - X1Xs V Xs V X1Xs•

After the second stage, we get
44 CH l :BOOLEAN FUNCTION~
1.4 .2.. U s1ng Blake s method~ eon5lru~t the oonlract~

d n f for a g1ven d n f ~
-
~.... ....
{1) 9J = x1xr V .xix,x.. V x2.r,z4,
--
(2) :D ~ .t1x2x., V ~rXti-1 V X3.t4
-
(3) !Z
- - -
x 1.t 1 V x 1x 3 V z1 x~x:r.x 4 V x 1x 2x 3 x 4
:;:::=
.....t
The contracted d n f -of the funet1on I (r ) spee1fied 11
1n the form o,{ a c n f can be obtalned as fo1lows F1rst

the parentheses are "Opened by us1ng the dlstr1hUl1V1ty
law After this, characters and terms are cancelled from
the obta1ned d n f by us1ng tl1e relatt ons xx = 0 -
xx -= x, x V .z ~ x~ K 1 V K1 K 1 -;:::::. K1
Example~ F1nd the contracted d n f for
I (x )
,..,..,
3
-
~ (.rl V x 2) (xl V x, V x,)
After opening the parentbesest we get
!:IJ1 = - .X1X1
-
V XtZJ V x1x, V .t2x1 V x.~rz V X2X3
Applying the ab-ove me-nt1oned rules,. we obta1n
9J = XtX.3 V .Xt
t .4.3. Construct the contracted d n I by us1ng the
c n I._.s
followtng g1 ven ..... _,.. -
(1) (xl V x, V Xa) (xt V X2 V X3) (.tz V x3),
(2) (xl V .x,) (x~ vZ:~
- .....
V X, V z3),
- vx4) (x, -
(3) (.x1 V z.,. V xJ) (z, V x,} (x, V .ra V ..x,)
i &4 4 Let N I be a set of tuples a E an such that
I"W
I (a) = 1 and Jc(/) 1~ tJte length of the contracted d n f

of the fun~hon f Sb.owthat lc(J}~ { I N 1 I ( I N 1 I + t)
1.4 54P Ftnd the 1ength of the contracted d n f of the
following functions
{1) l"1 ffi X:z $ $ Xn•
_. -
(2) (xl V ZJ V J'a) (XI V Z2 V Xa) ffi Xf $ Xs (B
-
@ Zn,
(3) (rl v x3 v .z.,) ~ - -
(.rl v .X:z v .r3) (.r4 $ z6 e
ffi Xn)f
(4) (xl EB e ZJi) (xl+J ED mXn). 1 ~ k~ n.
(5) (zl v v J;l\)(.r1 v v XI\ v X).+l v vXn)
t~k~n
1.4. 1\IINIMIZATION OF BOOLEAN FUNCTIONS 45
~ ~ ~
1.4.6. Letl(x")besuchthat Nf = {a.: k~ II a. II~ k+

"'
m }, and k::;;;, l::;;;, k + m, a. EBY.
(1} Find the number of terms in the contracted d.n.f.
of the function I, which become equal to unity on the
~
tuple a.
(2) Show that the length of the contracted d.n.f. of
the function I "'
(x11 ) is equal to (nk) (n-mk).
1.4.7. Suppose that the functions f (x") "' "'
and g (ym)
do not have common variables, K is a prime implicant
of the function/, and La prime implicant of the function g.
Show that K & Lis the prime implicant 111
of the function I & g.
Each elementary conjunction gener-
ated by the set of variables {x1 , x 2 , ••• , ~ 1 011
x 11 } has a one-to-one correspondence
100 001
with the face of the cube B" formed
"'
by the vertices a. at which the e.c. he-
comes equal to unity. This allows us
to construct a contracted d.n.f. from Fig. 2
the geometrical representation of a
Boolean function. Vertices of the set N 1 of the function
fare marked on the cube B 11 • The faces contained in N 1
and not contained in any other face formed by vertices
from the set N 1 are written down. Each such face is as-
signed a prime implicant.
Example. Let the function I "'
{x3 ) be defined by the vee-
~
tor ctt = (11111000). Find its contracted d.n.f..

Solution. The set N 1 is formed by {(000), (001},
(010}, (011), (100) }. The faces have the form g 1 =
{(000), (001), (010), (011) }, g2 = {(000), (100) }.
Thefacesg1 correspond to the e.c. ~.and g 2 to 2 3 • The xx
x xx
contracted d.n.f. is given by 1 V 2 3 (see Fig. 2).
1.4.8. Construct the contracted d.n.f. of the function
,...
f (x11):
(1) I(;;"') = (1111100001001100);
(2) I (;4 ) = (0000001111111101);
(3) I{;;~) = (0001101111011011).
46 CH I SOOLEA :r-. FU'l'C TIO' ~
"jjllljjj
1 4 9. Calculate the number of functions I (z"J for

whtch a g1ven elementary conJUnctton of rank r is
(1) an 1mp11tant,
(2) a pr1me lmpllcant
t 4 tO Let t,U) be the number tmpl1cant.s of ran.k
1'¥
or
r for the functton f (z"} and a, (j) be the number of
pr1me Jmplican ts of rank r Let P n be the set of Boolean
functions of the var1ables .:r1 , z 3 • Zn. ar1rl
I" (n) =
2
!n ~
JEPn
l, (/)~
- (h):::
S:r t n .t:J
" s, (f)
22 t~Pn.
Show t1lat
(f) t, (n) = ( : ) 2r-z\\-',
(2) i, (n)""'" { : } 2r-l!'~-t (1-2 :tn ")r

For small values ol n, the contracted d n f of the June
~
t1on I (x'ft) ean be found w1th the help of the reetangular

table (m1nimi:ing chart ot a Karnaugh
,.., map) For exam
pie. suppose that the lunct1on (.z1) IS def•ned With the
1\elp of Table 5 Combinlng t e tells torresponding to
T.ahle 5
.rl 0 0 1 ,
.r4
x, x2 0 1 1 0
•r . ., ,.,. ..... ... --~
0 0 . ~\_,) ~
0 I\. ....' __ ~
~~-
. --..
0 1 0 \ 1\ 0
I
~ I
..
~ ...... ~
1 1 {,..... ~~1 ~ .
~ ~ ._.1_, J 0
~
1 0 0 ..._1.-~ 0 0
-
U. MINI!IlJZAT ION OF BOOLEAN FUNCTION S 4i
unit values of the function I into the maximal intervals

as shown in Table 5 and comparin g them with the e.c.,
we obtain the contt·ucted d.n.f.:
.rlx2x3 v
- v XaX4 XzX4
---
v XtX:!Xa v---
XtXzX4.
1.4.11. Find the contracte d d.n.f. for functions de-

fined by the following tables:
(1) (2) (3)
Xa x,
x, 00I I x, 00I I
0I I 0 0 I I 0
x,x, x,x.
0 0 11 10 0 0 0 0 0 110
0 1 01 1 1 () 1 1 l) 1 1101
1 1 1 0 11 1 1 10 11
l 0 I 11 01 1 0 1 0 0 11 I)
A prime implican t is called a core implican t if its re-

moval from a contracte d d.n.f. leads to a d.n.f. that is
not equivalen t to the initial one. For each core implican t
K there exists a tuple of variables which turns K to uni-
ty and the remainin g terms of the contt·acted d.n.f. to
zero. Such a tuple is called a proper tuple of the core im-
' plicant.
1.4.12. Isolate the core implican ts from thP- contracte d
d.n.f.:
(1) XlXa v x2x3 v XlXz v xlx2x4 v ~XaX~ v xlxaxol;
') - - - - v --
(... ) xlx2x3 vxlx2x3 Vxa:c4 Xzx,.v XlX2Xa v x2x4 v xlx4;
-
(3) X2X3 V X2X3 V X1X2 V X1X3 V XaX4 V X2X4 V X1X4.
1.4.13. Show that the number of core implican ts of an
arbit1·ary function I (;'n} does not exceed 211 - 1 ,
1.4.14. Find the numbe1• of core implican ts of the
functions in Problem 1.4.5.
~H t BOOLEAN fUNCTtQtrrtS
1.4.15 F1nd tl1e number of Booleatl funel1ons I (xn)

for whtch the e c .l\ xt IS a core Implicant
f 4.t{i. Let K K 1 and K 1 be conJunctions from the
contracted d n f , and r, r 1 and r 2 be the ranks of these
conJunctions Let K V K1 V k 2 = K 1 V K 2 Show that
r 1 + r,# r + 2
t ~4.~ 17 Cons true t all term 1nal d n f s of the follow1 ng
4
functJons
~
(1) I (r) = (Oitt11tO)t

,.,
(2) f (x 4) = (t t 1001 f00001010t)l
f'W
(3) J {x') ~ (01 1010111.10111. iO}

t4 .. 18 F1nd the number of termanal and mtntmal
d n f s for the funct1ons appear1ng 1n Problem 1 4 5
1.4919. Show that the number of terminal d n f s
/'V
for an arbitrary B oo]ean functJon f (;z:-11 ) does no' exceed

(~:)
t 4..20*.. I-Iow many tetm1nal d n f s exlSt for a [unc·
lton hav1ng 2n-1 core tmpllcants?
1 ~4 21. F1nd out 1f the follow1ng d n f s are terminal
or shortest~ or minimal
(f) 9J ~ XtX2
...... -.
v ~,
_. ........
(2) :21 ~ x,x. V x1r2x4
- V Z"t.t'zZs,
......
(3) 2 == X1Z1 V xl.r3 V .rtxtxt V Za.l'a
1.4.22. Let L (f) be the complexity of the mtntmal, and
l {j) the length of the shortest d n f of the: funct1on f
,.., 1"'1.1
,.,., that L U (xn)) ~ nl V (x"'')) for an arbitrary function

Show
I (z'l)
N ,...,
I ~4 ~23. Show that l (/ (.rn)) ~ 2n-l,. L (j (xn)) ~ n211- 1
~
for any funct1on f (xtt)

f"t..
t ~4 .24 ~ For haw many func tton.s f (x11 ) are the fol--
lowing rela t1ons v a ltd
lV ~
(i) L(f(xn))==n2n-•, (2) L(f(xn))~n2" 1 -n?

1~4 .. 25. G1ve an example of a number k (0 < k~ n2n)
,......
:such that there IS no f (zn) havtng a m1n1mal d n f of
comple.xtty k
t ~4.26. For the funcltons ol Problem 1 4 5 ftnd the
1.5. ESSENTIAL AND APPAR ENT VARIAB LES 49
complexity of the minim al and length of the shorte st

d.n.f.
1.4.27. Let us,..,consid er a family of belt functi ons, i.e.
the functions f (xn) for which there exist numbe rs k and m
such that ,.., ...,
N 1 = {a: k~ II a II~ k + m }.
(1) Find the numbe r of core implic ants of the belt func-
,..,
tion f (xn) for differe nt values of,..,k and m.
(2) How many belt functi ons f (x11) have the maxim um
number of core implic ants?,..,
1.4.28. The functi on f (x ) is called a chain (cyclic}
11
function if the set N 1 can be arrang ed in a sequen ce that

is a 2-chain (2-cycle).
(1) Find the numbe r,..,of termin al and minim al d.n.f.s
of a chain functi on f (x11
) if I N f I = l. ,..,
(2) Find the same for a cyclic functi on f (x11
) such that
I N t I = 2m (m > 2).
1.5. Essential and Apparent Variables
_The variab le x 1 of the functi on I (x1 , x 2 , ••• , X 11 ) is
'""' ,..,
called essential if there exist tuples a and f) such that ~
~
a= (a 1 , • • • , a,_1, 1, ai+i• ... , a 11), ~ = (a 1, ... , ai-l•

,.., ,..,
0, ai+l• .•• , an) and I (a) =I= f (f)). Otherw ise, the
variab le Xt is called an appare nt variab le of the functi on
,..,11 ,.., ,..,
~ (x ). Two functi ons f (x 11
) and g (x 11
) are called equal
1f the sets of their essent ial variab les coinci de and on any
two tuples, differi ng perhap s only in the values of appar-
ent variab les, the values of the functi ons are identi cal.
Let 1.~ i 1 (x1 , • . • , X;,_1 , x, Xi,+l! • •• , Xi,- ..., 1 , Xts+ll ••• ,
Xin-1• Xia+l• . • • , X 11 ) is obtain ed from I (x 11

) by identif i-
cation of variables xi,, xi,, ... , Xi if q> is obtain ed from

f by substi tution of x in place of ~ariablesxi,, Xi,, ••• ,
Xik· For x we can take any variab le not belong ing
to the Set X 11 '-""' {x·1 ' ' X·111 • • • , X ik } .
1.5.1. Show that the statem ent "xt is an essent ial variab le
of the function f (;' )" is equiva lent to each of the fol-
11
4-0636
CH J BOOLEA' FU\CTIO\lS
loWtng statements ,.., ;..,;
(f) /! (x ) =F f: (J:n),
11
.. -
(2) there are va.rtables .:r,;Lt , x •._ ('J ::fo ,, I == t, k)
and constants o 1 , o" suth that the funet.1on
h 1Jt
f 01 «-. r-
(zn) depends. cssenttalty on z 1
1 .. 5 2. Enumerate the essenttal variables of the fol·
low tn g tunc.
,..,
t 1.nus
(1) I (:x3)
,., === (x 1 • (x 1 V Zz)) - ,.. X :a
(2) I (z 2) ~ (x1 V X2) -)- X2,
,., - ..... -
~
(3) f (z 4) == (x1 V Xz V x 2 r 3 V r 1 x 1z 3 ) z,t

(4) I (r}
,..,
= :t2 V za)
-(xt V - -
(xl V xst V
-
::3} (zl V %s V
.... .,.... ...... -
z.,) (zl V Xz V x,)
1.5 3~ Show that x1 1s an apparent va.rtable of the func
t•on f (express I through a formula that does not con
tatn x1 expltcltly)
,..;
(1) t (:t1) (x, e :t1) (xl ~ %1),

~
-
(2) f (r)
~
~
:== (((:l' x"1) Xt) (r, = )rn v

Zt) ;t ,xl} ED x, .
~ - == ..
-
(3) I (r) = ((x1 V ;r2) (x1 V x 3 ) :jjl- (.x1 ,.. z 2xs)) Z:~
I ~5.4 lnd1cate the apparent varJables of the funct1on
,.,
(t) I (zS)
....,
= (11110000)
(2) 1 (r)
,..,
= (001 toot1),
(3) t (r) ~ (001111.00)
,y
1 5.. 5. Let the func.tron f (xn) be defined by the vee ..

""""
tor a,= (a•.u ct1, , ct 2n._ ) Show that 1f Z~t IS an appar...
1
ent variable.. ct1 == a2n-A+t for all 1 on the segment
[s.2"-Jt+t, {2s+1)2n k-1)t s~O 2" 1 -1
1.5 6~ Show that tf there are apparent variables among
~
the varrables of the functtan f (xn)t n ~ 1 the funetton

assumes the value 1 on an even number of tuples Is tho
converse true?
~
l 5~7. Let the funct1on I (x") be such that J Nt I=

2m (2l - 1) \Vhat IS the maxtmum poss1ble number of
apparent var1ables of the func:tron J?
t ~5w8 ~ Ustng the results of Problems 1 5 5.. and 1.. 5 6 ,
1.5. ESSE NTIA L AND APPA REN T VAR IABL ES 51
find the vari able s on which the func tion I depends essen-
tiall y:
N
(1) f (x~) = (1011100111001010);
,...,
(2) f (x~)
,...,
= (0011110011000011 );
(3) f (x~) = (0111011101110111);
,...,
(4) f (x~) = (0101111100001010). •
1.5. 9. For each func tion in Pro blem 1.5.8. con stru ct
an equ al func tion that depends esse ntia lly on all its
vari able s.
1.5.10. Find the valu es of n (n "> 2) for whi ch the fol-
lowing func tion s dep end esse ntia lly on all thei r vari ab-
les:
~
(1) f (x 11) = (:z: 1 Vx 2 ) EB (x2 V x3) EB ••. EB (Xn- t V Xn)

EB (xn Vx,);
,...,
(2) f (x11 ) = (x 1 ; x2 ) (x 2 ; Xt) EB (x 2 ) x 3 ) (x3 ) X2 )
• •• EB (Xn ) X1) (xt >- Xn);

,...,
(3) f (x 11
) = ( ... ((xd . x 2)t x3)t ... } Xn)
:>-( xtl( x21 (x3 1··· 1xu ) ... ));
,...,
(4) f (x 11) = (x 1 ) x2) (x 2 :>- x3) ••• (X 11 _ 1 ~ X 11 ) (xn . . x 1)
--r)' (xt E9 x
2 E9 .•. EB X 11 Ef) 1);
& ( . . V -- -
X;,x ;, •.• x;[n / 2]) ) (x 1 EB x 2 EB
1~t 1 <1 1 < . , ·<i[n/2]~11
••• EB Xn)·
,..., ,....,
1.5. 11. Let the func tion s I (x 11) and g (ym) dep end es-
sent ially on all thei r vari able s and let the vari able s
xl, · .. , Xn, Y1 , • • • , Ym be pair wise diff eren t. Sho w
that the func tion f (x1 , . . • , X 11 _ 1 , g (y 1 , • • • , Ym)) de-
pends esse ntia lly on all its vari able s.
1.5. 12. Let Pc(X") be a set of all Boo lean func tion s
depending, and that too esse ntia lly, on the vari able s
xl, X2, ••• , Xn.
(1) Enu mer ate all func tion s in pc (X2). •
(2) Find the num ber I pc(X 3 ) I " , •

{_( tJ h 1
52 en 1 BOOLEAN FUNCtfiONS
(3) Show that I~ (Xn) I = ~ (- f)lt ( : } 2:t"-lt

.~~:~o
{4.) Snow tbal ltm 2- 2n I pe (X 11

) ~= 1
n ...... r::a
{ltW tl'v ~
1 .5.13. Let a., ~ and 1 b~ such tuples 1n B" that

........ ~ ~ ,._r I"'W
et ~ p~ ;~ Let tl1e function/ (zll) he such that I (a) =

,.., ~ ,.,._..
I (l)-:/=- I (fi.) Sl:tow th~at f (x'•) depends essenl1ally on at
least t"'o var1abl~s
1~5.14*. Sh()W that. x 1 1s an essent1al var1.able of the
funet1on f 1f and only 11 th1s var1ahle appears exp\1c~tly
ln the t.ontratted d n f of the funct1on 1
I .5 "I 5, Show that ~ 1 IS an essen t1a 1 v ar1able of the
function I 1f and only II x, appears e.xpllCttly tn the
Zhegalk1n polynom1al ol the func.t1on j
~
1 a.t6. Let .. - .... ~I <.~~~t - - =0 for any non empty

a (~i:a• z,..\ j)
.il'
set {x,1., , .x.,"-} ol vartables deHerent from x 1 Do~s

ltv
/ {x~}
have an apparent dependent on z 1?
,..,
1.5 . 17. Show that any symmetric funct1on f (xn) otl1er
than a constant depends essenttally on all tts vartable.s
f"'IJ
1. 5.. 18 .. Suppose that the functton f (J::n) changes tts

..... rN
value m llm~ at the verttces of the cba1n a, ilt ,... ,
.,...; ...., f!atl #ktl
~A-lt y connecting these verttces a., y of the cube

,..., ,..,
=
~
Bn t for whlch p (ct, 1) k Show that f (x,.) depends es-

sentially on a.t least,., m. var1ables
1.5~19. Let j (.:t'fl) depend essentially on at least two
I'¥ II¥ ,..
variables Sho\v that there are three verttces a,. ~J 1

of the cube B" satlsfytng the condttton
"'¥ ~ l"fee ;..,; ltv ./"V """' ,.,... ~
p (a. P) = p (p, y) ~ t, a :#: Yt I (a) ~t <v> ~I 

,...
I .. 5,.20. Let I (x ) depend essentially on all tts varla-
11
hles Prove that for any ~ (1 ~ l ~ n) there ex1sts a 1

such tha\ a certa1n subst1lUt1on of constants 1n place of
vatlahles other than x 1 and x 1 leads to a functton de--
pend\ng essent1.ally on x 1 and x 1
1.5. ESSENTIAL AND APPARENT VARIABLES 53
1 .5.21. Show that for any function I (xn) "" that depends
.
essentially on n variables there exists .a ,...,variable ,..,x 1 and
a constant a such that the function I~ (xn) = I (x 11 • • • ,
x 1_ 1 , a, x 1+1 , • • • , Xn) depends essentially on n- 1
variables. ,..,
1 .5.22. Let 1 (xn) depend essentially on all its varia-
bles. Check the validity of the following statements:
(1) there exists an i such that for any j there exist con-
stants whose substitution into I (xn) for variables other "'
than x 1 and x1 leads to a function depending essential-
ly on x 1 and xi;
(2} for any two variables x.....,1 and XJ there exist constants
whose substitution into I (xn) for variables other than
x 1 and x 1 leads to a function depending essentially
on x 1 and x 1.
1.5.23. Enumerate the functions in P 2 (X 2 ) that can be
obtained by identifying the variables of the following
functions: ....., .....,
(1) I (x f ) = (11111101);
3 3
.....,
) = (10010110); (2) (x
(3) I (x ) = x1x 2 V x 2x 3 V XaX1;
3
"'
(4) I (x3) = x 1x 2x 3 EB x 2x 3 E9 X 3 X1 E9 X 2 EB 1.
1.5.24. Show that ~he identity operation on the func-
,.., "' ""
tion I (xn) can lead to a constant if and only iff (0)= I (1).
1.5.25. Find the number of functions I (x"'2 ) for which
the identification of variables cannot lead to a function
'
depending essentially on one variable.
1.5.26. Can ,.., the identity operation on the symmetric
function I (xn) lead to a function that depends essentially
on all its variables and that is not symmetric?
....., ....., "'
1.5.27. Let a, ~. y be three vertices in Bn such that
,..., """' ,..., ,..; ,..., f"'tJ
a~ ....., ~ ~ y, and let I (xn) be such that I (a) = I (y) =/=

I (~). Show that it is possible to identify certain varia-
bles of the function f in such a way that the function
depends essentially on at least two and at the most three
variables. •
1.5.28*. Show that for the function f (;n), n ~ 4, re-
resented by Zhegalkin's polynomial of power not less
QI t BOOLEAN IUNCTlOI\S
th.an 2f there extst two vartables whose 1denttfieatton de

creases the number o[ essent1al var1:ables hy one
~
t lt5~29fro Enumerate all the functtons I (r) that depend

e~sent1ally on tl1ree varlables and 1n wh1ch the 1dent1fi.ca
tiOn of any two variables leads to a function depending
essentia1ly on exactly one var1able,..,
1.5"30 Let the f unct1 on f (z'1) be such that ) N 1 1 >
2n-1 Show that the Idetlttficatton of any two vari8h1es
of the (unctton le.ads to a function that 1s not tdentu~all)
equal to tero
1 5 31 • Show that the number o[ functions f (x1, x 2 f ..
Zn+ 1 ) whtch ]ead to a certa1n funct.ton g (;r1 • z 2 t •t:

x") as a result of Identrficatton ts asymptotically
equal to ( i
11 1
) 2"" as n -;.. co
~ ,..,
! . . 5.32. Show that 1f f (xn) depends apparentJy on z.h
the tdeuttficatlon of tb.1s va.rtable to any other var1able
leads to a fut1t\1Dn that depends essent1ally on the same
.fla.lt.
vartables as the tunet1on f (.:r )11
l tl .. 33~ Let n
........
> 1 and let the funct1ons I (x"'n) and
At/
g (xn) b_! ~u-~h that ~ N 1$g 1 = 1 Show that for any

l = 1 ~ n, at least one of the functtons f or g depends e.s
senttally on .r 1
1 .5.34. Show that 1f t N ,..; i ,..; ~ JS odd, the fun c..
f J.i(xntBI t!{~11.}
,...,
t1on q' obtained from f (x ) by tdenttfytng the var1ables
11
z, and ZJ depends essentially on n- 1 variables {n~ 3) .f"'t,
1.51135 • Let the funct1on f (.tn) depend essen ttally on

"""-J ~
n var1ables Let v1 (a:) be the number of vert1ces ~ .for wb1ch

"""-6 ~ "' t""'W ......
f {a.)+ I (~} and p (a, Jl) = 1 Let v U} = max Vf (a) ......

a(Btl
F1nd v (n for the fo]J oy, 1ng funct1ons f
~
(1) j (xll) =:; z 1 $ X 2 (13 ffi X 11 ,

'lv
(2) I (.xJI Cnt-"1) :::= (x1 V V xJt) & (zJl+l V V X2.k)

& & (r.~~; C[n/~1 -I )+i V V Zl (n/Al), 1 ~ k~n,
_, A N lc.
(3) j (z'l+Z ) = Ct4+ i+v (a:, Q.._) If .:z:~+'2i = (aJt t%2"
~ a~" a fl.~~~ ~ a~+ 1 ~t )\ where v («~ 1 ~a.") 1s the num
her of the tuple (et1, • ,. a ~c)
Chapter Two
Closed Classes and Completeness
2.1. Closure Operation. Closed Classes

Let J}f be a certain set of Boolean functions. The
closure [MI of the set M is defined as the set of all func-
tions from P 2 that are superpositions of functions in the
set JY!. The operation of obtaining the set [MI from M
is called the closure operation. The set Miscalled a func-
tionally closed class (in short, closed class) if [MI = M.
Let M be a closed class in P 2 • The subset A in lvf is
called a functionally complete system (in short, complete
system) in M if [A I = ill. The set A of Boolean func-
tions is called an irreducible system if the closure of any
proper subset A 1 in A is different from the closure of
the entire set A, i.e. [A c [A] and [A I =I= [A]. An
1
]
1
irreducible complete system in the closed class !vi is called

the basis ,of the class M. The set 1W contained in the
closed class M (among other things, in the entire set P 2 )
is called aprecomplete class in M if it is not a complete set
in M, but the equality [ill' U {!}1 = M is satisfied for
any function I E M"-M 1
•
Functions f 1 and I 2 will be called congruent if one of

them can he obtained from the other by a change of
variables (without identification). For example, the func-
tions .r·y and y :; are congruent while the functions
.r ·y and z ·z are not. While considering questions concern-
ing closed classes, it is convenient to indicate one repre-
sentative each from the set of pairwise congruent func-
tions. For example, the class {x, y, z, ... , x 1 , x 2 , • • • }
formed by all identity functions will be denoted by
lx).
If
11
M is a certain set of functions, then M (X 11
) (or
M ) will denote a subset of all functions in JI;J depending
only on the variables .r1 , :~: 2 , • • • , X 11 •
tH 2 CLOSED CLASSES A.Nn<.OMPLE1'BNESS
.2 t 1 J ust1fy the follow1ng closure properties

(1) [{Mil = (Mil
(2) 1f Af1 ~ M 2 {Mtl r= [Af 21t
(3) {A/1 UJlf 2 1 i2 {Aftl Ul~f2],
(4) 1.01 ~ 0
2. t 2.- Does relatton (4) Jn 2 1 1 follo'v from rela t1on.s
(i) (3)?
2~ I ~3. Is the set A a closed class? Assume that together
Wtth each functton I tn A 1 the set A also c.ontatns all
funet1ons 1n P 1 that are congruent to f
(1) A = (0, 1 }~ (2) A .=; {x} (3) A = {it z},
-
(-4) A ==: {.:t1 t z1 ~z 2 f .XJ z, z,, , x 1 .:t2 X
xl\.'t },
(5) A :=:: 0, .:rl V .r~ V V Zn-. n~

i l~
(6)A~ x1 f.Bx 1 t:B EDzn, n>t).,
(7) A ~ {0, xl aX e
I e Xj n n~ 1 }
2 t 4. \Vr1te down all funttlons from the closure of the
set A whtch depend only on varJables x 1 x 2 x 3 and .are
paJrw1se non congruent
(f) A === {z - - • 1 },
-
(2) A = {0, xJ,
(3) A ===- :t V y },
(4)A~ zll}
(5) A ~ :r: $ Jl ffi z),
(6) A ~ 00000001}
2 t 5.. Prove that If a closed class 1n P 1 con ta1ns a
functton dependtng essenttally on n ~ 2 vartables tt c.ou
talns an lnnn1tely larg-e number of pa1:rwlse Don congruent
funettons
2~ 1 .. 6~ Ennmerate all closed classes 1n P 2 whtth con
ta1n only a fintte number of patrwtse non congruent
funct1ons
2 J.. 7 In Pal
(t) IS the 1ntersect1on of closed classe! al"avs a closed
class~
(2) 1s the d1fference between closed ~lasses alwa}s a
closed c.lassJ
(3) can the com plet•on of a closed class never be a
closed class?
2 1.8. Through reductton to a prtorl complete set
1n P 2.1 show that the set A IS complete 1n P, If
(f) A =: {.t ~ y},
(2) A = {x y E9 z {:z:...., !I) e z},
)
2.L CLOSURE OPERATION. CLOSED CLASSES 5'7
(3) A = {x ~ y, x $ y $ z};
(4) A = {x ) y, (1100001100111100) };
(5) A = {0, m (:t, y, z), xll EB z}l;
(6) A = {(1011), (1111110011000000) }.
2.1.9. Which of the relations::::>, c, 2, s=, = , ~
is satisfied for the sets 2 K 1 and K 2 (the relation ~means
that none of the relations:::l, c, 2, s;, = is satisfied)?
(1) Kt = [Mt n lvl2l, K2 = [Mtl n lM2l;
(2) K 1 = [Mt '-.,1vl2], K2 = (Mtl'-JM21;
(3) K1 = [Mt U (M2 n
M:~}l, K2 = [Mt U M2l n
[Mt U M3l;
(4} K1 = [Mt n(M2 U M3)], K2 = lMt M2l U n
[Mt M3]; n
(5) K1 = [Mt '-.,(Jvft n
M2)), K2 = [.Mt)'-.,[Mt Mzl. n
2.1.10. Let M 1 and 1vf 2 be closed classes in P 2 , such
that M 1'-.,M 2 =I= 0. Give examples of concrete classes lv/1
and M 2 that also satisfy the following conditions:
(1) .MtnM2=0, M2'-.....Mt=F0, U\tlt UM2l=
Mt U M2;
[Jv!l uM2J =
M1 U M2;
(3) Mt";:) M 2 , UV!t '-.....M2] =/= lv/ 1 '-.....M 2 ;
(4} M1 n M2 =1= 0. M2 "MI =1= 0. [M1 '-.,M 2 1=
Mt '-.,M2;
(5) M1 n M2 =I= 0. M2 '- Mt =/= 0, [Mt $ M2l=
M 1 $ M 2•
2.1.11. Isolate the basis from the set A that is com-
plete for the closed class Jvl = [A I.
(1) A = {0, 1, x, x};
(2) A = {1, X EB y EB z $ 1 };
(3) A= {xVy, x·y·z, xVy·z, (xVy)·z};
(4*)A = {xED 1, x $ y EB z, m (x, y, z)};
(5)A={xVyV z, x·y·z, (x )y) )z,
(xVy) )oz};
(6) A = {(x ) y) )- (y ) z}, x V y V (y $ z} };
(7} A = {x ·y, x V y, x , y, x $ y ED z $ u }.
2.1.12. Show that any precomplete class in P 2 is
a closed class.
1
By m (x, y, z) (or h2 (x, y, z)) we denote the function xy V
X% V yz called the median (or majority junction).
2
The sets 11re taken in P ~·
.58 CU .2 CLOSED CLASS£5 ,Al"r.~ COl\JPLETE"'lESS
2. t. 13~ Let AI1 ~nd AI2 be d1 fferent preeromplete classes

1n the same closed class J.lP Show that 1f At: -=# Jfl,
then A:f! +
Jf! (1 e classes ft.f 1 and A! 2 tJ d 1fi Pr" even on
a set of funct1ons that depend on only one var1ahle)
2.f41f4:. Enumerate all pretomplete classe.s 10 the clO!ed
class Af
(t) ~f = (Ot il, (4) ~~ :: (0, l: y]f v
(2) Af = [0, 11~ (5) Af = CO, x y ·zJ
(3) A-/ == (z y),
2,. 1.1 S ~ Ve11fy 1( the r()l\ ow1 ng sets form. elose d el asses
In P,
(t) the set of all symmetric functions,
f"'aa
(2) the set of all funct1ons f (x n ~ Ot sa t1sfy1ng the

71
),
A.J ,..,.
cond1 t1on f (0"') = f (1-n) = 0-',

,..,.
(3) the set of all functions f (x11), n ~ t for wh1ch
IN,'~ zn-t
2. t. t 6. Show that If AI 1s a elosed class In P '" then
IAI U {x}l= Af U {z)
2 .. 1. (7 Pt0Ve that the set P'" of all Boolean funttlons
I
cannot be presented as a un1on U J.~f, (s ~ 2) on patr

t,.-1
w1se non 1ntersecttng closed classes 1n P 1
2 .. 1.. tB~ Prove that any closed class tn P 2 contatntng
a funct1on other than a constant also contains the
funetton :z
2.t.t9~ Prove that If a closed class 1n P 2 has a fi.nlle
bas1s~ then each basis of this cla.ss 1S finJte
2,. t ~20. 1-rlaJOri:te the power of the .set of all closed
classes 1n P 2 eon taln1ng fi.n1te complete sets
2.1 .. 21. Prove that 1f a non~mpty closed class 1n P 1
d1ffers from the sets {OJt {1} and {0 1), 1t cannot be
extended to a bas.ls 1n P 2
2~ ( 22 Let Jf be a closed class ln P 1 con tatn 1ng a fi
n1te number of precomplete classes (1n ~tf) It 1s assumed
t1la.t any elosr:d tlass 1n Jlf ean be extended to a precom
plete class 1n Jll Prove that the number of functtons
tn any bas1s of class 1/ does not exceed the number of
pretomplete classes (1n '!)
a \Ve assume that AI~ Pt

• 11 n = {} \ben f lS a f"onstant equal to zero consJdered as a
fun ct1 on of ::zero arguments
2.2. DUALITY 59
2.1.23. Prove that a closed class [x ~ y] contains only

such functions in P n which can be presented (barring the
notation of the variables) in the form x 1 VI (xx, x2, · · · •
,...,
x 11 ), where f (x ) E P 2 •
11
,...,
2.1.24. Let the function I (x11) belong to the closed class
[x ~ y], and depend essentially on at least two variables.
Prove that l N 1 l > zn-t.
2.1.25. Prove (without the help of Problem 2.1.18)
that each precomplete class in P 2 contains an identity
function.
2.2. Duality and the Class

of Self-Dual Functions
The function g (x 1 • x 2 , x 11 ) is called dual to the

function I (x1 , x 2 , ••• , X
••• ,
11 )
--
if g (x 1, x 2 , ••• , X 11 ) = f (x 1 ,
~ •... , X,,). By definition, the dual function for con-
stant 0 is constant 1 and, conversely, constant 0 is a func-
tion dual to constant 1. The function dual to I (x1 ,
x 2 , • • • , X 11 ) is denoted by/* (xi, x 2 , • • • , X 11 ).
The following statement, called the duality principle,
is true: if <D (xi, x 2 , • • • , X 11 ) = f (f 1 (x1 , x 2 , • • • , X 11 ),
••. , fm (x 1 , x 2 , • • • , X 11 )), then <D* (x 1 , x 2 , • • • , X 11 ) =
f* (f~ (xi, Xz, • · ., Xn), • .. , /in (xll Xz, ••• , Xn)).
Let M be a set of Boolean functions. By Jv!* we shall
denote the set of all functions that are dual to the func-
tions in the set M. The set 111* will be called dual to the set
M. If ill* = M, the set Jl{ is called self-dual.
The function I (x1 , :r 2 , • • • , x 11 ) is called self-dual if
f* (xl, X2, ••• , Xn) = f (x 1 , X 2 , • • • , X 11 ). The set of all
self-dual functions is denoted by S.
It follows from the definition of self-dual functions
that a function is self-dual if and only if it assumes oppo-
site values on any two opposite tuples of values of vari-
ables.
The following statement, called the lemma of non-self-
dual function, is valid: if the function f (;h
is non-
self-dual, the substitution of the functions x and x
for
its variables will lead to a constant,
60 CR 2 CLOSED CLASSES AND CO"!PLETENESS
2,2. t" Is the iu1lct1on g dual to the funcl1on I 1f

(i) f ~ X ffi Y g ~ E,..., y..,
(2} f = x ,.. y g = y ., x~
(3) f == xy V x: V y:, g ~ xy ffJ xz (B yz,
(4) t == x my m z c =:; x E& u mz,
(5) f = :cyz V x (y ~ :)t g (.t, y, z) ~ (01101101)?
"'-ttt
11
2 2.2. Let the function j (x ) be defined by the vector
~
«J =:;;. ( <X 1 , a. 1 .. f • ~
,. <X~; 11 _ 1 ) Prove that the unctton I (.x )
1s dtfined by the vect-or (a. 2n _1,
- ~ -
, <It • «,)
2 2.3~ U.smg the dualtty pr1nc1ple~ der1ve a formula
for representrng a function dual to f Slmpltfy the ob
ta•ned expresston (b) presenting 1t tn a dlSJUDCtJve normal
form or 1n the form of a Zhegalktn polynomial)
(1) f == xy V yz V xt V zt,
(2) I =z x f V Y (zt V 0) V "iiJ:,
t
(3) I = (x -)i- .Y) ED ((x y) I (X~ yz)).
(4) t = (x v Y v (yz fB
t))- ... 1
;tv
2.2414 Suppose that the run~tlon I (:t11 ) lS represented

by the formula \l genera~d by the set {0 1. 1, &} V}
.,.,
Prove that the tunct1on /* (x") 1s represented by the
for1nula ~ *. call~d tile formula dual to ~ and ohtalned
from 1t by replac1ng each symbol & by V V by & .. 0
by 1 and 1 by 0
2~2.5 Prove that If formulas 21 and generated by sa
the .set {0~ 1, l, & ~ V} are equ1 valent, formulas 21 •
and m• are also equivalent
;to./
2 2 6. Prove that tf the function f (.xn) depends essen..

1-ttJ'
tta]Jy on the varlab]e .x, (1 ~ 1 ~ n), the function /* (x )

11
also depends essentially on x~

2 2.7. Pro' e that
(1) the set dual to ~~· 1s Identical to Jl,
(2) the set Af lfs a ~losed class 1f and oll1y 11 the set ~!•
ts a closed clas.s,
(3) tf the set A/1 rs a complete system (or has1s) 1n the
closed class M 1 the dual set At: forms a complete system
{resp bas1s) Ill the closed class ill• I
(4) 1f M 1 ~ M 2 , then M~ ;2M~

2~2.8" Count the number of ftlnCtlons dependtng on
11 ar1ables x 1 ., z2:, , .x n 1n the ~t (A!•l ,,JA!1
2.2. DUALITY 61.
(i) M = {0, x}; (2) ill = {x E9 y}; (3) M =

{xy, x V y, i}.
2.2.9. Is the function I self-dual in the following case:
(1) f = rn (x, y, z); •
(2) f = (x ) y) ) xz ;. (y ;. z);
- --
(3) f = (x V y V z) t V xyz;
-
(4) 1 = (0001001001100111?);
(5) I = x1 $ .r 2 ffi ... EB Xzm+l ffi a,,...,where a E {0, i }?
2.2. 10. Prove that the function f (x") is self-dual if
~ -
its xc
and only if its x 1-component /~ (x") is dual to
~
component f~ (x"). ,...,

2.2.11. Prove that if f (x") is a self-dual function,
then I Nt I = 2'1-1.
2.2.12. Show that there are no self-dual functions
depending essentially on two variables.
2.2.13. Determine the number of self-dual functions
depending essentially on variables x 1 , .r 2 , • • • , X 11 •
2.2.14. Enumerate all self-dual functions that depend
essentially on variables x, y, z. Show that each of these
functions can be presented in the form m (xa, yf'>, zi') or
x ffi y EB z ffi a, where a, ~. y, rJ belong to the set {0, 1 }.
2.2.15. Determine the values n~ 2 which make the
,...,
function f (x") self-dual:
,...,
(1) f (x") = Xt (xz V X3 V · · · V Xn) V Xz (x3 V · · · V Xn) V
,...,
•· • V Xn-:! (Xn-1 V Xn) V Xn-tXn;
(2) f (x") = x 1 (x2 $ x 3 E9 ... Ea x 11 ) Ea Xz (x3 E9
• •• E9 X 11 ) ffi ... $ E9 X 11 ) $ Xn-tXn;
Xn-z (x 11 _ 1
,...,
(3) f (x") = (x 1 ,. x:J Ea (x2 ,. x 3) $ ... E9 (x11 _ 1 ) X 11 )
E9 (xn ,. x 1);
~
(4) f (x 11
) = . . V . x;,x;, ... x; 11112 r
· · · <•Jn/2[~11
1
~••<•s< .
~ere the disjunction is taken over all monotonic conjunc-
ttons of length ln/2[ formed by variables x 1 , x 2 , • • • , x 11 •
2.2.16:.., Let f (;.') = x 1 $ x 2 E9 . . . E9 X 111 gf (xm) =
~
gn-i+l• (xm), i = 1, 2, . . . , n. Is the function f (g1 (xm),

"'
· · ., gn (xm)) self-dual?
62 CH !! CLOSCD CLASS:CS A~D C0'1PL:CTC~~SS
2.24117. Pro\ e that 1f ......

(t) f {t1 , :tru , Xn) E Sl then f (x. x, x} E {xl z lt <t
(2) IE s, then m (xl1 7,x2 mx3 e /) ffi Xs ED X4 Es~

(3) I E S then i,.t EB x 2x 3 EB i;x3 $ xJ ffi x4 EB xs E S
~ l"tV
2~2~ 18 Let the func t1ons f (xn) and t• (xn), n ~ 1,

'N
A
sa l1siy the cond1t1on ~ N 1 ~ ~ t• ~ Prc•ve that

t•
~
(1) tf IV const~ then f E S,

(2) If I e
I ,. ~ const, then I E s -
2~2.19. Subst1tut1ng the funct1ons x and x for the
vatla bl~s ohtatn a constant us1ng the non self-dual
funct1on f
(t} 1
(0011100tJ.
-
=;;
-
(2) I = (x V y V z) t V .zyz,
(3) I ~ (x ~ y) · ~ (x ID z)~
(4) I = xy V xz V yt V zt
2412.20. Prove that tl a non self dual function depends
essentially on at least three variables the Ident1fictlt1on
of some of Its variables can lead to a funct1on that de-
pends essentially on two var1ables
2.2~21, Prove that 1.t tdentlfi.cat1on ot ~ ar1ables lll the
~
function f (zn)~ .n ~ 3, cannot lead to a funct1on d~

pend1ng essentially on two v ar1ables, •tile functton f IS self-
dual
2 2.22,. Let f {z. y, z) = rn (li', ytJ, z')t where a, ~'
V belong to the set {0, 1 J Prove that for any n ~ 4, v. e
can apply the superpos1t1on operation to the functlon /
to obtatn a functton dependang essent1ally (}n n var1ables
2.2.23~ Prove the following equivalences
~
(1) x EB y 6) z = m (m (x, y, z), m (:r, y., z), m (.:t't y, - -
z)) = m (x, m (xt !/t z), m (X 1 1/t z)) ~ m (m (z, Y, :z)1:
m (t, y, z)t i},
(2) (.t V y V z) t V :ty: = m (m (y, z t), :r, t) ===
~ (nt (r y, t), ~ (xt z, t), n1 (Yt z, t)),
{3) m (z, y, z) ~ ,..,m (m (i. y, z), y, z)
2 2.24. Let I(~) E P 2 and n~ 3 Prove tl1e fol·
11
lowing relat1on
f (Zt• X t.t Z 1.. X4.t ~ X n}
=I (z1, m (x1, x 1 , .x,)'t m (.ttt Xt, z 1 ] .. x 1 ,. , .rnJ
2.3. LINEARITY 63
$ 1 (11~ (x1 , x 2 , x 3 ), x 2 , m (x1 , x 2 , x 3 ), x 4 ,, • •• , Xn)
$I (m (x1 , x 2 , x3 ), m (x1 , X 2 , x 3 ), x 3 , x 4 , • • • , Xn)·
2.2.25. (1) Using the problems 2.2.12., 2.2.14., 2.2.23.

(item 1) and 2.2.24., prove that [m (x, y, z)l = [m (x,
-
y, Z)J = s.
(2) Prove that any basis of the class S of all self-dual
functions contains not more than two functions .
2.2.26. Can the superpos ition operation be used to
obtain the function g using the function f if:
(1) 1 = (10110010), g = (1000);
(2) 1 = (1111011100010000), g = (00010111);
(3) 1 = (11001100), g = (00110011)? ~
2.2.27. The function f (x ), n ~ 2, has the following

11
properties : 1 (x11 ) EJ: S, and the identifica tion of any

2[ Vnlvariables in it leads to a function from class S.
What is the largest number of essential variables in the
~
function I (x11 )?
2.2.28. Enumera te all functions dependin g essential ly
on the variables x 1 , x 2 , x 3 , x 4 , any x?t-comp onent of which
is a self-dual function.
2.2.29. Is the set j}J self-dual in the following case:
(1) l'ri = {x E9 y ED z, m (x E9 y, X......, z, y......, z)};
(2) J}f = {x·y, x V y, xED y ED m (x, y, z)};
(3) JU = {(x ) y) ) y, (x V y) ED x E9 y, (x V y V z)
-
tV xyz}; ·
(4) iVI = S'-., {x ED y ED z, m (x, y, z) }?
2.3. Lineariiy and ihe Class of Linear Functions

~
The function I (x11) is called linear if it can be pres-

ented in the form
~
I (xn) = a 0 E9 et 1x 1 E9 et 2X 2 ED ••• ED et~Xn,

where ai E {0, 1 }, 0 ~ i ~ n. The set of all linear func-
tions is denoted by L, while the set of all linear func-
tions dependin g on the variables x 11 x~, ... , Xn is
denoted by L". The set L is a closed and~ precomp lete
c~ass _in P 2 • The following statemen t (lemma on non-
lmeartty of a function) is valid:
If I~ L, the substitut ion of the functions 0, 1, x, y,
CII 2 CLOSED CLASSES AND CO\JPLETENESS
i, g for --
the variables can lead etther lo xy or to ry
If ( (t L~~. f JS called nonltnear
2 3 .. t.. Expand 1ng the functton f tnto a Zhegalktn poly-
nom1al, lind 1f 1t lS l1near
~ ~-
(1) I (.t 3} ~ (.rt.rt V .z1:r2) @ z,~

~
(2) f (z2) = x 1x 2 (x1 ED .r,.)7

(3) I
_,4
(x )
-
~ X1X2 V - - -
.TzXa V X3X4 V .z.,:tt,
(4)
.#fiP
1 (r) ~ (zl- -• x~} (.x,

-Jo.x1} ~ ,..., x,
2.3.2. Prove that tf tlte funetton j (x4f).) assumes oppo.. t'9fJ ~
s1tB value-s on any two adJace-nt vert lees a. and ~ 1n B",.

1t ts a linear funct1on Is the converse true'
#fiJ'
2.'3 . 3. Pt~~~ that ,.t 1 \xn} 1~ a \l.n~af tuntrt\~ ~tMr

than a constant, ) N 1 t = 2n-l Is the converse tru:e1
2.3 .. 4~ Ftnd out tf the function f IS ltnear
...w
(1) 1 {x~) = (1010 1010 0110 1000),

......
{2) I (x~) .,.,.,
= (i001 0110 tOOi 0110),
(3) t (x 4) = (1001 Otto 0110 1001)1
IV
(4) j (z1J :.==: (0110 1001 1010 0101) ,...,

2: . 3~5~ Show that tl\e number of l1near funet1o-ns I {xn)
dependtng essentially on exactly k var1ables of th~ se:\
{x1 , x:l!, ~ :c~} 1s equal to 2 (~)
2.3.6~ F1nd the number of self dual functtons be·
long1ng to the set Ln
2 3 ..7 Sh-ow that the run~llon x !I cannnl he o~
If )1-
ta.1ned from the funttlOfiS X ffi y {!) z, X $ 1~ X 9 g Wt\h

the help of the superposttaon operat1on
2.3.8~ Can the funct1on f lead to :rlj 1f the functions
- .....
0, 1, z., y, x, y are subst,tuted fnr ltS var1ables"'.l
.tv
(1) I (x') ~ (fitO 1000),

~
(2) J (.x') ~ (0111 1111) f
(3) I (.?) =: (1001 1001),

(V
(4) I (xn) =::: (.t1 -l- x,) (z 2 -,.. J:l)f!l (z2 =+- X~) (X' 3 ... xj) e
ij) (Zn--1 __ ., r 1)) (Xn - ~ Z,_ -1)
2.3.9. F1nd out 1f the functton :z _., y_ can be represent-
ed by a formula generated by the set r}l, where
2.3. LINEARITY 65
{1) tD = L U {xy V xz V zy };
(2) cD = L"S;
(3) ell = (L U {xy V yz V zx }),S;
(4) <D = (L U {xy})" S. ~
2.3.10*. Prove that if the function I (xn), depending

essentially on all its variables, is linear if and only if
tile substitution of any subset of constants for any subset of
variables leads to a function that depends essentially
on all the remaining variables.
2.3.11 *. Prove that a polynomial of degree k ~ 3 can
lead to a polynomial of degree k - 1 as a result of iden-
tiftcation of variables.
2.3.12*. Show that by identifying the variables in the
"'
function 1 (x"), n~ 4, we can obtain a nonlinear function
depending on not more than three variables. Enumerate
"'3
all nonlinear functions I (x ) from which a nonlinear
function cannot be obtained as a result of identification
of variables.
2.3.13. Show that the identification of variables in
a non-linear function can lead to a function that is con-
gruent either to xy ffi l (x, y), or to xy $ yz $ zx $
l (x, y, z), where l (x, y) and l (x, y, z) are linear
functions.
....,
2.3.14. Show that if I (x' ) (f L, there exists a two-
1
dimensional face in B 11 on exactly three of whose vertices

""11
the function I (x ) assumes the same value.
"'1
2.3.15. Let the function I (x' ), n~ 3, be such that for
...., "'11
all i, j (1 ~ i < j ~ n) I~ (x 11 ) = l:f (x ). Show that
""11
f (x ) E£.
· 2.3.16. Which of the following sets form a basis in £:
(1) {1, X ffi y };
(2) {x ....... y, x ffi y ffi z};
(3) {X'"'-' y, X$ y, 0};
(4) {0, X ffi y ffi Z ffi 1 };
(5) {1 ffi x EEl y EB z, x EEl y EB z}?
2.3.17. Single out all bases in the following sets
complete in £:
(1) {0, 1, x E!l1, x- y, x EB y EB z};
(2) {x ....... y, (x,.... y) ,.... z, x $ y $ z, x $ y $ z EB t };
(3) {0, (x,....y)~z, x$1, x$y}.
a-0636
CH 2 CI..OSE.l> CUSSES \ 'D COliPLSTE"JESS
2 3 t8 Prove that any set complete 1n L conta1ns

at JeGSL L\\ ~ J unt-1 l u1:,
2 J 1.9 l 11ove tJlat any has1.s 1n L conta1ns not more
than three funcltons
2 3 20 Pro\e tl1at only a i•n1te numbC!r ol closed cld.S
ses can be fornted by l1near functxons alone L~:st 1ll
such classes
2 3 21 Show that any closed elass fotJned hy d finltL
numb.Br nt pa1rw1se non congl uent ( unt.llotls ts con
tained 111 L
.......
2 3 22 Tlae lunet1on j (xu), n ~ 1 sat1.sfLes ll1e condl
t1ons
{1) 1 V1 i ===: 2n t (2) J e) J* IS a constant
i'a,1
bhow tllat I (~) E L U S for 11 ~ 3 G1ve an ~" 1mple

wlu~n tl1e functlon I sal1sly1ng tlte condlllOilS (l) :tnd (2)
doe.s not belong to tl1e se l L U S
2 3 23 Pro' e t1la t 1 n S --- lz ED y ffi z ffi t l
2 3 24 Show that L* ....... L
2 3 25 Enumerate ali patr" l~e llOll-congruent {'1nc
tlons I sat1sfy1ng the follov.. 1ng condllloJtS
(i) t '!. L
(2) any proper ~ubfunct1on of j1s )1ne1r
2 3 26 Delermlne the numbet of ~elf du11 llr1Par
~
r tnc
l1ons I (.rn) tha L depend es~en t 1.a lJ y on lll 1het r v J. t 1 tbles
2 3 27 In huw man) \\a)s ean b1ae~ets b8 ~rtanged
tn the expresston .r1 • .x1 .. x 3 =),.- x, -. ~ x 1 so tl11t to g1ve
a ltnear functtan?
~
2 3 28 Let I (xn) E L n S I (0 0 0) =
f (0 0 0 1) and J (o. 1 a.i a.n) ~ 0 F1nd "\he
value of the functton I (al a-2 I Cln) an
2 3 29 For wl\at '"al\tes of n r-an tile ltr..cat fl1ncttnn
r-.1
I (r ) sat1sfy the condttton f (0 (} 0) =/=; f (1 1 1}

and be symmetr1c'
"-
2 3 30 Let f (x } be ~uch that

:J 4
/ 11 t
,...,
Show tllat I (.x»)
L US ~
2 3 31• bnnmera te all pa 'rv.- 1se non congruent uon
Ia near runct 1ons I sucl1 that 1 n y tden t1 fica t 1on of \a rta bles
leads to a [Ltnct ton 1n L
2.4. FUNCTIONS PRESERVING THE CONSTANTS 67
2.3.32*. Enumerate all pairwise non-congruent func~

tions f !£ L U S for which any identification of variables
leads to a function in L S. n
2.3.33. Let Q = {0, x,
11 , /2, /a}, where 111 / 2 , / 3 are
pairwise different functions depending essentially on
variables x 11 x 2 , ••• , X 11 (n > 1). Prove that the system Q
is complete in P 2 •
2.4. Classes of Functions Preserving

the Constants
"'11
The function I (x ) is said to preserve the constant 0
(orconstant1) if/(0, 0, . . . , O)=O(resp. if/ (1,1, ... , 1)
= 1). The set of all Boolean functions preserv-
ing the constant 0 is denoted by T 0 , while the set of
functions preserving the constant 1 is denoted by T 1 •
The set of all Boolean functions that depend on variables
x1 , x 2 , • • • , x 11 and preserve the constant 0 (constant 1)
is denoted by T~ (resp. by T~). Each of the sets T 0 and
T1 is a closed and precomplete class in P 2 •
2.4.1. To which of the sets T 0 U T 1 , T 1 "'-.. T 0 do the
following functions belong:
(1) ((x V y) )o (x I yz)) t ((y ~ z) ..,. x);
(2) (xy )o z) I ((x >- y) t (z ED xy));
(3) (x >- y) & (y t z) V (z ..,. y)?
""11
2.4.2. For what values of n does the function f (x )
belong to the set T 1 "'-.. T 0 :

,....
(1) J(x 11
) = x 1 ED x 2 $ . . . ED X 11 $ 1;
,....
(2) I (x = ( ... ((x1
11
) , xJ , x 3) >- .. • >- X 11 );
,....
(3) I (x11) = ( ... ((:c 1 >- x 2) ~ x 3) }' • •• >- X 11 )
$ (( · · . ((x 2 >- x3) ) x,) )o • • • >- X11 ) >- x 1) $ ...

ED( ••• ((Xn ) X1) ) xz) ~.... r Xn-1);
i
(4} J(; = 1 ED 11
) E Xi,Xi,Xi,?
i~i,<ia<h~n
2.4.3. For what values of n does the recurrently defined
function
•
f (;11) belong to the set (T 0 "'-..T1 ) U (T1 "'-..To):
~·
at '2 CLOSED CLASSES. AND COMPLETENESS
(1) /"1 (.t 2) ;;;; Zt $ Z1

~ ,..., ./"tU
f" (:~:») -= (Zn ... ._ In t (zn-l )) (z" V fn-l (zt'l-i )) l n > 2t

(2) f, (x1) = .x1.. {., (z", x 1 ) = z t $ zi,
,.., .,..,
f D (~ 11
)-= ;tn-tXn ef J1 J (zn 2
) n, 3t
(3) f 1 (X J) :=o Zt t f 2 ( :r i • %2) -== X l V .Z:
.,.... ,., "'-~
t 11 (z"') =In I (z"ll-l) e Zn/n .2 (xn- 2), n~3'

2.4 . 4 In how many ways can the brackets be arranged
1n the express1on .t1 ,. z 2 · to x 1 -+ .x2 -· ...- .z1 In order to ob
ta1n a formula representing a functton tn T V'l?
2 4 .. 5. Count the number of functions dependtng on ;tit
x1 t t Zn. •n each of the followtng sets
(5) L'(T0 t 1 ), n (9) S n{To U T1),
(6) L'(T 0 U T 1), (10) n(To,Tt),
S
(7) T 1 n S~ (11} S ' (T, U T1}1
(B) T~, s, (12) (S, TDJ n rlJ
(ta) L ns n rl,
(14) L ' (T9 U (T1 nS)).
(15) (L U S),(T o U T1)
2.4 6 Ftnd the function f (x xt t $) 1l
(1) I (.rt, Xst r Xr~) E T1' TG,
(2) I (x •• l:s• f Za) E L ' (Tl n S),
(3) f (.xll Xat ~ :rn) E S ' Ta
2.~ 7. Is 1t posstble to obta1n the funet1tJn I as a result
of the superpos1t1on opera t1on genera ted by the set fll1f
(l) I == :z t11 y, «b = {z ~ y ).
(Z) f = :x-,., y, <b:: (:ty, X v y},
(3} f = 3: Vy, '1> = T, U(S' (L U Tl)}.
(4) t ~ xi en ~ (T1 ' L) U {z W y }?
2.4~8~ Prove that
(t) T, = [xg, X EB u1 = lx v g X e ylf
{2} T 1 ~ {x yy, x~ yl ~ {:cy, z;..; ylt
(3) T 0 nT1 =(xy. z$y$zl
2 .. 4.9•. Prove that any bas1s 1n T 0 conta.tns not more
than three funct1ons Gtve examples of bases of the clas.9
T 0 cons1st.rng of one, two or three lunct1ons
2,,, FUNCTIONS PRESERVIN G THE CONSTANTS 69
2.4.10*. Prove that any basis in T 0 n T 1 contains not

more than two functions . Give examples of a basis con-
sisting of one function.
n
2.4.11. Does the class T 0 T 1 contain a function de-
pending on three variables and forming a basis in it?
2.4.12. Prove that
n
(1) L T 0 = [x E9 y];
(2) L n
T1 = [x,...., y];
n
(3*) s T 0 = [xy vyzV zx, X EB y $ z] =
= [xy V yz V zX'J. ~
2.4.13. The function I (r), which is not defined every-

where, is equal to zero on the tuples (000), (001) and equal
to unity on the tuples (011), (100), (110). Extend the
,....,
definition of the function I (r) on the remainin g tuples
so that the obtained function forms a basis in T 0 •
2.4.14*. Prove that if the function f depends essential -
ly on at least two variables and f ~ T 0 U T1 , then
n
L s s ttl. ··
2.4.15. Are the following sets bases in T 0 :
(1) {xy v
yzV Zx, X EB y $ z};
(2) {xy, x EB y EB z, x V y };
(3) {xy EB z };
(4) {xED y EB z, xy, xyy Vz}?
~
2.4.16. Give an example of a symmetr ic function
f (x~) forming a complete system in T 0 •
2.4.17. Prove that
L nTo nT1 = L nS n T 0 = L nS nT1
= L n S n To n T1.
2.4.18*. Prove that the classes T 0 n L, T n S, To n
0
Tt are precompl ete in T 0 •
2.4.19. Prove that the system of functions {1} U (T 0 "'-
(Tl U L U S)) is complete in P 2 •
2.4.20. Prove that
(1) r:
= T 1;
(2) (To U T1)* = T 0 U T 1 ;
((~) (To U T1 U S U L)* = T 0 U T 1 U S U L;
'*)) ((Ton Tt) US)*= (T 0
(5 (T 0,S)* = T1 " S;
n T1) US;
(6) (S" T0)* = S"'-T 1•

10 CH 2 CLOSED CLASSES J..Nb CO'IPtETENESS
,..,
2~4~2111 Let I (x 4) E snLn T 0 , I (i, 1, 0, 1) = 1~ and
1'¥
let f (t•) depend essentially on at least two var1ables
;.,/
F1nd I (.~:')
2~4..22• ~ Prove that the classes L nTa• L n T1 , L nSt
[0, x]
are the only preeomplete classes In L
2 4.23~ Prove that I E (TfJ n .rl, If and only If us
not a Slngle constant can be obta1ned from I by carry1ng
out the superpos1t1on operat1on
2~5. Monofonlcity and the Class

of Monotonic Functions
,.,..
The Boolean functton
.,.., ......
f ~xn)
ts called mon.atantc 1f for ,.., ,y
any two tup1~s « and ~ 1n B"' such that a~~' the 1ne
~ ~
"q,U?J~\tJ 11n}~ 1 ~f,) 1.~ "521.\l~r~red 0\htiW\~, \h-e tun~\\\">r.

ifllt.l
! (xn) IS called non·monotonlc The set of all monotonic:

Boolean funct1ons \S d~noted by M, whlle th-e. set of all
monoton1c functtons depend1ng on variables .t'1 't x 1 , ., Zn
1s denoted by Ar
The set AI 1s a closed and precom
p]ete class 111 P 2 The following statement (lemma on
a non monotonic junct1on) IS valid 1f f f1 M, substl\U\t-
ons of the ft.tnc.trons 0 t, x for 1ts var1ables can lead to
the funetton X
~
The vertex a. of the cube B"' 1s ealled the lower urnty

rAil ;..,
(upper uro) of the monotontc funet1on 1 (% 11 } 1f I (a} ~ t

""" ,. ~ ~
(resp 1 (a) = 0) For any vertex ~' 1t follows from J} < a.

.1'¥ """
that I (~) ~ 0 (resp f {~) = 1 follows from the eond1

f!'IW ~
tton a<~)
2.5,1 Whleh of the iollow1ng funt\lons are mono\on1e

t/'tll
(1) x ,.. (z ., y), {5) f {r)

,.., == (00110111).
(2) x ._ .. (g , .x), (6) J (z~) = (01100111),
.#ltJ
(3) :rv (z ED y), (7) I (x1) == (0001010101010it1)t

f'a.J
(4) xg e !JZ e u es. (8) I (.x~) == ( 101f1111)?

2 5 lltoNOTONICITY 71
"'
2.5.2. For what values of n is the function f (xn)
•
monotomc:
(1) I (;'n) = E xl:cl;
(2) f (:en) =""

X1X2 •• • Xn-tXn
- )o {xl $ Xz $ · · · E9 Xn);
n
(3) f (;") = x 1x 2 ••• Xn $ Ex 1 ••• Xf-!Xl+l ••• Xn?

i=1
,...,
2,5.3. Prove that for the monotonic functions f (xn)
the following expansion formulas are valid:
rv
t (xn) = x,J: (xn)
,_
v n(xn); f (xn) = (x, v n(xn)) & t: (xn).
,.., "' I'V
2.5.4. Prove that for any monotonic function f other

than a constant, there exist disjunctive normal forms and
conjunctive normal forms that do not contain negations
of variables and that represent f.
"'3
2.5.5. For how many monotonic functions f {x ) are the
relations I (0, 1, 1) = f (1, 0, 1) = 1, I (0, 0, 1) = 0
valid? How many of these functions belong to the set
M '- S? Do they include functions with apparent vari-
ables?
2.5.6. Let t (x~) E S n M, f (0, 0, 1, 1} = t (0, 1,
"'
"'4
1, 0) = f (1, 0, 1, 0) and let I (x ) depend essentially on
"'
all its variables. Compile the table for the I (x").
2.5.7. Prove that if I is not a constant and f V /* a
constant, then f (i M U S.
,...,
2.5.8. (1) Is it true that if 1 (xn) is monotonic, the
condition
I'V
II > II a II, I
"-J ,...., ,...., "' "' '"""'
a, ~ E nn, v (~) > v (a), II~ (a) = 1,

leads to the equality f (~) = 1?
~)* Suppose that for all k (0 ~ k < n} the conditions
f (ctn) = 1, v(~n) ~ 2n-1 _ 2h, V (~n) = v (~n} 2h +
lead to the relation 1 (jfn) = 1. Prove that I (~n) E M.
2.5.9. Enumerate all functions 1 (;4) EM satisf.ying
the following conditions;
72 CH 2 CLOSED CLASSES .AND COltfPLETENESS
•
(1) i {1 r 0, 0,
0) ~ 1, I {Ot ,....
t. 1, 1) ~ 0,
(2) I (11 0, 0, 0) :::=: 1, I (x4) E L~
~
{3) J (Ot 1, 0, 0) ::P j (1 t 0 t, 1)t I (x4) iS syrometr1c,

(4) J (t, 0, 0, l) ~ 0, I E S
~
2.5 .. tO. Show that lf f {~) ts nQn monotontc~ there ex 1st

,.., lfll¥
tWO VeCtors C% 1 P 10 9n that d1ffer exactly IR ODe COOT...

#9tJ ~ ,..., """'
d1nate, and for wh1eh a.< ~ but I (a) > f (~}

.2.5.1 f. Show that the lunct1on /, whJcb depends essen-
tially on a't least two variables" Js monotoniC tl and
only If any (proper) subfunct1on of the funct1on I ts
monotonic
2.. 5.t2~ G1ve an example of a non monotonic functton
i"W i. = ,_ ~
J (zn) whose each subfunclton of the type J,(xn)t r = 1, n.

a € {0, 1 }., is monotontc How many of such functions
depend (not always) essentially on the vartables of
the set {x1, z 1 ') ... l zb }?
~
2415 .. 13. Shew that the funct1on J {.t'\;) 1s nl.on<JtonJ.c 11

and only 1f for any k (k = 1: ii-_:_ f),. for any non-empty
subset {t1 , , ~~}c{1, t n} and for any tuples
~ f"'tJ ,..., 1'-P
a= {o1 , 1 al) and "t ~ (t1 , , TA:) With o~ T,

the foiiow1ng relatton 1s sa t1sfted
211i5~ 14. Prove that no Slmple lmpltcant of a mono ton 1c

funetlun C.Gnta1ns a negat1on o[ var1ables
2.5.15_.. The monotontc elem~ntary conJunctton K gen•
era ted by the set of "ar1 abl es x1 xf. x n IS .saJd to
correspond to the vector (a1 , a. ., an) tn Bn when
a~ = 1 for any & ;::=:: 1·, ii lf and o~ly lf z 1 ts 1nclttded tn K
,IAtl
Prove that 11 the vector a lS tl1e lower unJty of the mono

II¥
ton1e function f (xn)t the Zhegalkln p~lynom1.al cn:nlalns

as one of the ,..,terms the conJunetlon K eonespond1ng to
the vector a
."-J
2~5.16•.Let J (x'1) he a monoton1c symmetric func-

I/IWJ ~ ;..,
tton, such that N 1 .=;= {a, Ua. U;;>- k, a, E 8 )4r For \vhat
11
values of n and k does \ts Zhegalklll polynom1al contatn

2.5. l\!ONOTONICITY 73
at least one elementa ry conjunct ion of rank k +

2 as one
of its terms?
2.5.17. Show that there are no monoton ic self-dual
functions with exactly two lower unities.
2.5.18. Can monoton ic self-dual functions have three
lower unities and depend essential ly on:
(1) three variables , (2) more than three variables ?
2.5.19*. Show that the maximum number of lower
unities of the monoton ic function I (;n) is equal to{ [,;; 21 ).
2.5.20. Show that if the values of a monoton ic function
are arbitraril y replaced by zeros at some of its lower
unities, the resulting function will also be monoton ic.
2.5.21. Show that I j}J" I ~ 2 { [,;; 21 ) •
2.5.22. Find the number of functions in each of the
following sets:
n n n
(1) 1VI""<T1 ra); (4) ~Jfl L" S ;
11
(2) lll''"(T'l UTa); (5) Ln'-....(.Mn US").

(3) .M"nLn;
2.5.23*. Show that
( 1) I S" nM" I < 1111"-1 I for n~1;
(2) I illn 1< 111r-t 12 for n~1;
(3) I jvfn I~ I Mn-z 12 22"-2 for n~2.
2.5.24. Let m (n) be the number of monoton ic functions
dependin g on the variables x 1 , x 2 , • • • , x 11 • Show that
m (1) = 3, m (2) = 6, m (3) = 20, m (4) = 168.
2.5.25. Count the number me(n) of monoton ic fun<?-
~
tions I (x 11
) dependin g essential ly on n variables for
n = 1, 4.
2.5.26 *. Prove 1 < I S" I for n ~ 4.
11
that I Jv/
2.5.27. What is the number of monoton ic self-dual
""
functions I (xt) that depend essential ly on all their
variables ?
2.5.28 *. Using the fact (see problem 1.1.18) that a cube
11
B" can be divided into { [n/2) ) non-inter secting ascendin g
.
chams, prove that
(1) 1M" I ~(n+1)(£n~ 21 );

74 CH 2 CLOSED CLASSES .AND tOMPLET~E!S
2.5~29"' .. \Vlth the help of the problem 1 i t8~ show

tba t I IfF I::;;;;;; 3 ((~;n)
2,5,3041 Prove that the number of monoton1e functJons
"'¥
f (z"') eaeh o! whose lower un1t1es has a we1ght not ex-

eeedmg k (0 ~ k~ n/2) ts not more thll.n t k( : ) +
2415.31. Let (A. ~) be a part1ally ordered 5et The
functJon j, detined on the set A and a.ssum1ng llS values
from the set (O~ t )J 1s called monotonlc tf for any ex and
~ in A, such that a~ ~' the relallon f {a.)~ J (~) 1s
s.attsfled Let m. (At~) be the number of d1flerent mono
ton1e lunct1ons defined on A Ftnd mJn m (At ~
IAr~
max m (A, ~) and find the parttally ordered
JAI ..... n
sets wttb n elemeilts on wh1cb these IDIDimum and maxi-
mum values are atta1ned
2.5 32•. G1ve an example of a sequence of monoton1c
#t.ti
funel1ons f • (z'n), n = 1, 2, t such that the number of

~
lower un1t1es of the functton In. (.r1 "} exceeds the n um

ber of upper zeros by s. factor of 211 /n
2.5,33. Prove that 1f the ...,number of lower un1ttes of
the monotonte funct1on 1 (:c11) IS not less than 2, the
funct lon depends essen ttally on at )east two variables
2~5634. •. Let t (f) be the number of lower untttes of the
monoton1e funct1on I and let p (f) he the number o( its
essenl1al vanahles Prove that p (j) ~ log 2t (/J -
log 1 Jog, t (/)
2. 5 . 35.. Let I E Jr.. and let m" (/) be t h.e number .of
vectors ~ ~
a tn B:, such that f (a) -= 1 qt (/) = ml (j) k {R)
Show that (/)~I[.. (j)ff k == fit
qA.-t
Jill¥
n
2.5.36 .. The functtan rp (z~) f defined on B"" and assum
9
ing arbitrary real values~ 1s called a generalized mon.o-

#Y ,.., ~ ,..,
ton1e functiOn 1f 1t follows from as; ~ that tp (a.)~ rp (~)

Prove tba t a generaJtzed monoton1o funclaon can be
2.5. :MONOTONICITY 75
presented as a linear combination of monotonic Boolean

functions of the type
""' ""'
rp (xn) =c + ~ a/' f (xn),
t ~nJEliinT•
•
where c is real and a 1 ~0.
""' be a generalized monotonic func-
2.5.37*. Let rp (xn)
tion and qh (rp) = ( ~ ) -I ~ rp (~). Show that gh-l (rp) <
~
et.EB]:
qh (rp), k= 1, n.
""'
2.5.38*. Show that if /(xn) ~ Jv!(n~4), an identi-
fication of variables in it can lead to a non-monotonic
function depending on not more than three variables.
2.5.39. Is it possible to obtain from x xyz
V t (xy ~ z):
(1) by identification of variables; .
(2) by identification of variables and substitution of
the constant 0 for certain variables;
(3) by identification of variables and substitution of
constants?
2.5.40*. Show that if I "'
(xn) ~ M U S (n ~ 3), Lhl'
identification of variables in it can lead to a non-mono-
tonic non-self-dual function that depends essentially on
two variables.
2.5.41. Show that if I EM, I* EM.
2.5.42. Show that any monotonic function is contained
in not more than two classes in T 0 , T 1 and L.
2.5.43. Show that none of the classes T 0 , T 1 , S or L
contains M.
2.5.44. Can the function xy V yz V zx lead to the
function xy with the help of the superposition operation?
2.5.45. Is it possible to obtain 0 from the functions
xy, x V y and 1?
2.5.46. Show that the set {0, 1, xy, x V y} forms
a basis in M.
2.5.47. Isolate all subsets from the set {0, 1, xy,
x V y, xy V z, xy V yz V zx} which are bases in Jltf.
2.5.48*. Show that any basis in M contains not more
than four and not less than three h•nctions,
76 CU 2 CLOSED CLASSES AND COMPLETENESS
295.49. Show that any has1s 1n Jf cons~sttog of three

functions conta1ns a [unction that depends essent1ally on
three or more variable'
2.5 50 Gtve e.xantple.s of bases tn the fo11owtng classes
(1) To 0 j\-f, (2)
,..,..
T1 n A! t (3) L A! n
2~5 51. Let f (:en.) E AI (n ;;a: 3) Prove the valtdlty
of the repre~enta.tlon
~
J (.xn) ;:= m (/ (.xi, X1,. Xs, .t4, ' Z n)

/ (.rl, Z:, x.~, z.,. , Zn}, f (.r3, Z2, x,, Z.t, .x.)),
where m (x~ y, z) = :zy V yz. V zx
2 5.52. Show that
(1) :LY V !JZ V z:t forms a basts tn J/ n S,
(2) any function from }l[ n S, Whlth depends essentla}-
ly on more than one \ artahle fotms a baSl.S. ln Af t ns
2.5 53•. Let !lJ be a set constst1ng of all monoton1c
elementary dtsJunctions, and let ;tC be a .set of all mono
tonu~ elementary conJUnCtions Show that only the :1ets
3J U {0, i.}, ;;;( U {0, i}, Af T 0 , 1\f T 1 are Jlre n n
complete 1n Al
~ fl - I '2!\.-.& :2;k- I
2.5.54*. Let f (y2") = & & (g 1 - • Uu:~") Sbow
_.~o ~~a
that rN
1J = ~Jf•l ,.,
-""-'~
2.5 55. Let f (xn} and g (.t11) be monoton1c functiOns

and h =I & g Prove that
J N A J~ J N t I I N t. I 2-n
2.6. Completeness and Closed Classes
The follow1ng completeness crlter1on 1s valtd In P 1
Theorem (Post) A system A tS complete ln P 2 lf and
only l / l t lS not conta1ntd enttrely 1n one of the classes T fJ•
T 1 , L S, and Af
"""-~
The function f (z'') IS called Sheffer's Junctfon 11 1t

fc.rms a bas1s 1n P 1
f"'<J
Let the funct1on j ,_(.x") depend essentially on all 1ls

var1ables By ~ U (xt\)) we denote the set of all such ,.,.
funct1ons whtcb are obtained from the function I (x")
through ldentlnt:a\Ion of var1ables, the fun~t1on f does not
2.6. COMPLETENESS AND CLOSED CLASSES 77
belong to the set ln (/). If n < 2, then by defmi tion

'"" "'
~n (j (xn)) =f::. 0. The set \)~ (! (xn)) is called the hereditary
"'
system of the/unction f (xn). The functi on f is called irreduc-
ible if lln (/)] =I= [/). The basis 58 of the closed class K
is called prime if the substi tution of its heredi tary system
for an arbitra ry functi on f from 58 leads to an incom plete
system in K. The functi on f not belong ing to the closed
class K is called a prime function with respect to K if
its heredi tary system ~n (/) is contai ned in K.
2.6.1. Show that P 2 does not contai n precom plete clas-
ses other than the classes T 0 , T 1 , L, S, M.
2.6.2. Using the compl etenes s criteri on, fmd out if
the system A is compl ete in the follow ing cases:
-
(1) A = {x ~ y, x y·z};
- -
x.....- yz};
(2) A= {x·y,
-
(3) A = 0, 1, x (Y"" z) V x (y EB z)};
(4) A = (01101001), (10001101), (00011100) };
(5) A = {(0010), (1010110111110011) };
(6) A = (S'-..M ) U (L'-..(T o U T 1 ));
(7) A = (S nM) U (L'-..M) U (To '-..S);
(8) A= (M'-..(T 0 nT 1)) U (L '-.. S).
2.6.3. Prove that if the functi on f depen ds essent ially
on at least two variab les and belong s to the class s n JY!,
the system {0, /} is compl ete in P 2 •
'"" "'
2.6.4. Is the system A = {h (xn), f 2 (xn)} compl ete if:
(1) It E S'-.M , /2 (£ L US, ft ~/2 1;
(2) /1 a To U L, /2 aS, /1 ~ fz 1;
(3) ft a n
To Tt, /2 E M""-T 1, ft ~ !2 -1?
"'
"' "'
2.~.5. Is the system A= {It (xn), / (xn),/ (xn)} com-
a
plete tf It L U (T 0 n
2 3
T1 ), /z E M'-..L, / 1 ~f 2 = 1 and
It V/3 === 1?
2.6.6. From the compl ete system A in P 2 , isolate all
possib le bases in the follow ing cases:
(1) A = {(x V y) (x Vy), xy EB z, (x EB y),.., z,
_ m (x, y, z) };
(2) A= {1, x, xy (y,..., z), x EB y ED m (x, y, z)};
(3) A = {0, x ED y, (x ~ y) .j. (y"' .t}, (x 1 (xy)) >-z};
(4) A = {x V (x EB y) V z, (x ,...., y) '7" z, xy EB zu,
--
m,(x, y, z)}.
78 Cll .2 CJ.OSED CLASSES A ""i D C0~1 P LETEliE SS
.
2 .. 6~7~ G1ve three examples of bases 1n Pa each con-
talnJng onet t\vo, three and four functions.
2.6AS..- Enumerate all d1tleren\ bases 1n P a conta1n1ng
only such functions that depend essentially on two vart-
ables z and y (bases are considered to h~ dJflerent tf
one o! these cannot be reduced to the other through a rede
stgna t1on of variables)
2A6~9A Ftnd out 1f the set 2l can be extended to a. bas1s
10 P 1 1n the f ollow1ng eases
(i) '"! = {.r G) Y~ m (x y, z)J, (3} ~-= Jf,(T,UT1))
(2) ~ = {x ~ y, x V gz}, (4) 2I ~ L 0 AI

2 .6 .. 1.0•. Doe-s P 1 eonta1n a baS!! eons1al1ng of four fltnt-
tlons / 1, ( 1 , 13. and /,1 such that / 1 t1 T 0 , / 1 4 T 1, / 1 tt L
and / 4 Ef S?
2~6.11. Us1ng the operat1ons 10 the theory ol .sets
express the closure of the set ~ through known closed
classes T {H T1 t L S. AI and P 1 1f
(t) '21 ;;::::: PI '(T, u Tl uL us u AI),
(2) ~ = ~J,(T. u L), to) \! ~ s,tr o' T1>•
(3) m ~ Af,(T 0 0 T1 ). (7) ~-== L'(Ta U Tt)~
(4) ~ ~ Jf,L. (8) 2! = Ta 'T1,
(5) W := T 0 (L,S),n (9) ~ = (T0 Tl),hl n
2 .. 6.1-l .. \Vh1ch of the relat1ons :=J, c:,. ;2, ~' =, :
•s valid for c]asses K 1 and K a (the rela.t1on ~ means that
none of the other f1ve relat1ons ts satt:sfied)
(f) Kl ~ (:.r v (x m g) v :). KJ = (x v
g, X ED yJ.
(2) K1 :;;;= l.z ~ g, x V yz), K, = lx ffi yz1~
{3) Kl = {zy' z e Yl~ Ks =. [xI )!-- y' X~ yll
(4) K1 = It~ x vul, K, = [z m y!. x v uz.
(5) Kl ;::=: l:t m y, ;t,...., y.z], x. = lm (y, X, .z), z e
y $ :Z e 1. xy m z],
{6) K 1 == l:t ~ yl, K 1 =: {x ffi y, m. (x. y, t)l?
2~6~t3~ (1) Prov-e lh&t. P 1 (X1 ) eonta1ns e.xaetly two
Sheffer s functions.
(2) FJnd the number of Sheffer's functroos 1n P 1 (X1 )
2~6.14u (1.) Prove that 1f I E: T 0 U T 1 U S., then f 1s
a Sheffer function
(2)• F1nd the number of Shefierts funct1ot1S 1n P 1 (Xn)
2.6.15~ Prove that by 1den\l.nta\1on bf vartablest we
c.an obta1n from Sheffer's functJons depeodtD'i essent1ally
:! t3 CO:\IPl CTC:-<E:>S AND CLOSED CLASSES i9
on at least three variables a Sheffer function depending

essentially on two val'iables.
2.6.16. For what values of n (n~ 2) is the function f
a Shef£er function if:
(1) 1=1 (E)

l~l<j~n
E :l'jXj; •
(2) I-= 1 $ .1' 1..r:! $ .r::x3 E9 •.•

(3) f = (.r 1 r .1'2) E9 (:t2 ~ X a)$
.... $ (Xn-l ~ X11 ) ® (Xn r ,1:1);

(4) I= (x 1 !.1'2) ® (x2 lx3 ) ® . . $ (xn-tlxn) E9 (xnl·rt);
(5) f = 1 $ (x 1 ~ x 2) (x 2 ). x 3) ••• (xn-t r l 11 ) (Xn ~ x,);
(6) f = V -X; -X -Xj]n/~[?
1 11 • • •
l~i 1 <i,< ... <i]n/2[~n -
2.6.17. The function I does not belong to the set T 1 U J1l

and has exactly one value equal to zet·o. Prove that it
either is a Shl•ffer function, or depends essentially on
one variable.
2.6.18. Give an example of a Sheffer fm1c.tion that de-
pends essentially on the smallest possible number of var-
iables and assumes the value unity on exactly half the
tuples of values of variables. ,...,
2.6.19. Give an example of a function f (.c") such
that fm· any k (1 ~ k~ n - 2) and for any subset i 1 ,
i 2, •.. , ill in {1, 2, ... , n} the function
a
dx.
( / x.
( • . . ( a.
dx,
I (;·") ) . . . ) )
lt lJ 1{
is a Sheffer function.
2.6.20. Let the function/ be monotonic with exactly two
lower unities. Prove that f is a Sheffer function.
2.6.21. Prove that any self-dual function not belonging
to the set 1' 0 U 1' 1 U L U i1I forms a basis in the class S.
2.6.22. (1) Prove that any function in class L belongs
to at least one of the classes T 0 , T 1 , S and iV[.
(2) Give examples of linem· functions contained exactly
in one of the classes 1' 0 , T 1 and S. .
2.6.23*. Give example of a function in class T 0 that
does not form a basis in T 0 and does not belong to the
set T 1 U L U S U Jvf.
80 tl-1 2 CLOSFD C.LAS~~S AND. C0\1PI FifE.hES~
.....
2 ~6~24.. G1 ve example of a noulltlear f tUlC ~!ott I (.c,•)
that depends essentially on the smallest posstble number
of var1able.9 and sat1sfies the follow1ng cond1tton for any
l"V
number ~ (1 ~ 1 ~ n), each of the functJolls 1: (xn) and

,.,
j~ (2") assumes the "Value un1ty at exaetly 2"-~ tup1es
of values of var~ebles
2 ..6-~~ 25• ~ Prove that ldent1iu~at1on ol var1ahles 1n
a nonlinear func.lton depend1ng essen\.1ally on n ~ 4
va.tlables leads to a nonlinear func.t1on dependtng essen
t1ally on n - t vartables
2.6 ~26 •.. Prove. that tdenttfic.atton o[ v artables Ln
a lunctlon I nol beJong1ng to lhe set L US and d.ependlng
essentially on n ~ 4 vartables leads to a non self dual
nonlinear funct1on dependtng essenttaJiy on n - 1 vart
ables Is tb1s statement valld for n :=:: 3?
2.6.27 •.. Prove that Identtticatton of varta.bles 1n
,;..,
a Sheffer functton f {x11 ) depend1ng essenttally on n # 3

variables leads lo a Sheffer functton dependtng essen
ttally on n - 1 variables
2.6.28~ F1nd out 1f the following tmpltcattons are true
(l} 1 ~(To U T 1} 'S~f EL UAI-.
(2) I E£ T 4 U T, U AI z>r I ts a Sheffer func.tton •
(3) /fl T, US Ufll..->1 (£ (L '-. Tt) n(S '-.ill)~
(4) I~ L U S U l'rf ~I ts a Sheffer funclton
2.6 . 29 ~ Let the subset \l from P 1 (Xn} contaJn more
.2ft I
than 2 - fun~tlons Prove that ~ 1s a complete system
1n P, for n> 2
2.61130.. Prove that each funct1on from a pr1me bas\s ln
Pt; ts 4 prtme functton Wtth respect to precomplete class
tn P,.
2.6.31, F1nd ali patrwtse non-congtuent functions that
are pr1mo functtons wtth respect to the class K
(1) K = T 0 1 (3) K = lO, 1 xJ,
(2)* K = L 0 S, (4) K ~ CO, 1\ x, xl
2.6.32. Prove that prtme funct1ons With respect to the
class K that are palrwtse non congruent, are exhausted
by funct1ons of the set 3'
(1) A = T 0 0 Tt•
~
W ;:=: {0, 1, s}t

(2) K =ToOL ns~ 21 = {0, 1\ Z~ :y., zVy .
m {z, y, z), nt (z, y, z)]. -
2.1i. CO~IPLETENESS AND CLOSED CLASSES 81
2.6.33. Prove that P., contains:

-
(1) only two prime bases {x I y } and {x t y} con-
sisting of one function;
(2)* only three prime bases {0, 1, x EB y EB z, xy },
{0, 1, x El1 y EB z, x VY} and {0, 1, x EB y El1 z, m (x, y,
z)} consisting of four functions .
'
6-0636
Chapter Three
k·Valued Logjcs
3" t. Representation ol Functions
of k-Valued Logics Through Formulas
J.f.t. Elementary Functions of k-Valued Log1cs

and Relations Between them
J, (x)
the smaller of x or y m1n (x y) (alternatJ \ell xy or x&. y),

the larger of x or y max (.x, y) (alteina.ttvely, xV y),
3.1. REPRESENTATION 83
•
mod k sum: x +
y (mod k) (read as "mod k x plus y") ;
1
mod k product: x ·y (mod k) (read as "mod k product of x

and y" 1 );
truncated difference:
0 if O~x<y~k--t,
x....!,_y=
x-y if 0 ~ y ~ X~ k- t;
implication:
k-1 if O~x<y~k-1,
x::;:Jy= (k-1)-x+y if O~y~x~k-1;
joint denial: max (x, y) + 1 (mod k), denoted by vh (x, y);
mod k difference:
x--y if 0 ~ y ~X~ k-1,
x-y=
k-(y-x) if O~x<y~k-L
The functions (operations) min, max, + and • are com- •
mutative and associative. Moreover, the following rela-

tions are valid:
•
(x + y) ·z = (x ·z) + (y ·z)
called distributivity of multiplication with respect to
addition; ·
max (min (x, y), z) = min (max (x, z), max (y, z)}
called the distrihutivity of the operation max with respect
to the operation min;
min (max (x, y), z) = max (min (x, z), min (y, z))
called the distrihutivity of the operation min with respect _
to the operation max;
max (x, x) = x, min (x, x) = x
called idempotency of the operations min and max; and
min (-x, ......... . y) =,..,max (x, y), max (--x,,..,.. y)
= ,..,min (x, y)
•
called the analogs of De Morgan's rules (laws) in P2 •
.
1
Unless otherwise stated, the symbols + and . in this chapter
will denote mod k sum and product.
6•
CU ' Jr-VALUED LOGICS
The follow1ng equal1t1es ate 1ntrodueed by defin1l1on

max {z1 • .r2 , , .x" _1 , .:t n) == max (max (r 1 ~ .x 2
Zn -l)t Xn)t A~ 3,
tntn (x1 ~ z 1., ~ *'n-l~ Zn) == min (0110 (.zt, xt
Zn-t), Zn)• n> 3,
0 tf z~ 0,
-x- -- k - Z If :t :;b 0
In v1ew of the assnela\tve. nature of mod k produet, the
product x ·x x • ·X (l colactors, l ~ 1) IS often wr1tten
1n the power form z 1
3.1~l~ Prove the validaty of the follow1ng equaltt1es
-
(I) - (x) -==: #¥Jx.
(2) z ~ y ~ ,..,.,; (z - Y)t
(3) :t - (.:c -- y) -= mtn (x, y),
(4) (x:;:l; y)~ y ~max (x, y).
(5) {z~ y) + i ~ m1n (x. y).
(6) x. - y ::=: x - m1n (x, y);
(7) z - y ==- max (z,. y) - y,
(8) (,-._.r.z} - (y z} == ...-...fmax (xl: y),
(9} (.........t)- (~yJ = y ~ x,
-
(10) ~(.r + J/) = (~x) + (~y)J
(it) ~~:Y) =
{~x) Y~
(12} max {(x + 2) -1, l~t 1 (z)) ~ Z~
(1.3) m1n (AIJJ A-t (z), (k - 2)~ x) = z.
(14) :X- g = (x- y) +X }If. 1 (y) + Y Jtt.-1 (z),
.II-- - - -
z g
(15) v" (rt y) + •J~t-t (y) + Jil-l- (.:r) ~ max {i y),
(16) max (z, y) + /o (g - z) + J~c -1 (.t) y = max c.t, g)t
(1 7) min (x, y) + J 0 (y -- x)- ll-t (x) -~y = min (.i:' y),
(18) 1 o (max (J o (.z), J 1 (.r)t , J k 2. (x))) = J 11 1 (.x).
(19) J 1{max {z, 1. / 1 (x), J 1 (z), ., J 11. -a (x))) =
JfJ(z),
(20) z •Jo Vt (x)) + Jo (z) •Jt (z) = x + Jo (z) -- It (x),
(21) 1 0 (z-1)---J 0 (z-(t-1)) = J,(z), t ~
t, 2t ., k - 1.
(22) (,-w(("-'.X)- t))- ( (((k-t)-
(~ x))- (~ x)) ----- - (.-wx)) ~X;
~-\ t.~mu
(23) ( (((k- 1) lo (x}) / 1 (x - 1)) -
It (:t- (k- 3})) ........ ((k- 1)/9 (~ z)) ~ ~.
... tl-
3.1. REPRESENTATIO!Il 85
3.1.2. Prove that the function f in Pk is generated by

the set of functions A (A c Pk) as a result of superposition
operation.
(1} f = J 1 (x), A = {J 0 (x), 1 2 (x), max (x, y)},
k = 3;
(2.) f = ,_.x, A= {J 0 (x), J 1 (x), min (x, y), max (x,y)},
k = 3;
-
(3) f = X, A = {1, r, J 1 (x), max (x, y) }.
k = 3;
(4) f = j 0 (x), A = {x- 1, x 2 }, k = 3, 5;
(5) f = ii (x), A = {x ·Y +_ x - y3 + 1 }, k =3, 5;
(6) f ='""X, A = {1, x·y}, k = 3, 5;
(7) f =x, A = {3, j 0 (x), x...:... y}, k = 4;
(8) f ='""X, A= {x + 2,1 0 (x), 1 1 (x),
max (x, y), x·y}, k = 4; •
(9) f = j 4 (x), A= {x .:....1, 1 2 (x)}, k = 6;

(10) f = h (x), A = {x + 2, x 2 , 1 3 (x)}, k = 6;
(11) 1 = h (x), A = {x, - x, J k-1 (x) };
(12) f = J k-1 (x), A = {'""x, x ...:... y };
(13) f = Jk-'J (x),A = {k- 1, X+ 2, X...:... y};
(14) f = j 0 {x}, A = {1, '""X, x...:... 2y };
(15) I= X, A = {1, ,...,X, X...:... y}.
3.1.3. Prove that if a belongs to Ek and is coprime to k,
each function J; (x), 0~ i~ k - 2 can be presented as
a superposition generated by the set {x + a, J k- 1 (x) }.
3.1.4. Show that the function t:p from Pk can be repre-
sented by a formula generated by the set {0, 1, ... , k-1,
X ...:... 2y} if
(1) t:p = ik- 1 (x), k =2m (m~ 2);

(2) t:p = j 0 (x), k = 2m+ 1 (m~ 1).
3.1.5. Let h 1 (x) = -x, h;+ 1 (x) = x::> h 1 (x),
i ~ 1. Prove that '""hk- 1 (x) = J k-1 (x).
3.1.6. For what values of k (k~ 3) are the functions x 2 ,
r and x 4 pairwise different?
3.1.7. Let k = 3, 4, .•. , 9, 10. How many different
functions in Pk, depending only on the variable x, can
be presented in the form x 1 (l ~ 1 and the power is taken
over mod k) for each k?
3.1.8. Prove that each function f (x) in PJ> can be pre-
CH 3 Jt VALUED LOGlC:S
sented as a superposition generated by the set {l, J h-J (x),

:l + y}
3~ 1.9 ~ The [unct tons I 1 (x} and f, (.r} tJl P 3 sat 1sfy the
follow1ng condttlons / 1 (x) ~ const, / 1 (E3) =;'=- E:~ and
/, (Ea) = Ea
Prove that the .funt tton g (.x) = / 1 (x) + I 2 (z). where
mod 3 sum 1S taken, does n()t assume at least one value
tn E 3 1 e g (E 3 ) ~ E3
3.1 .1 0. Let the funct1 on f (.t) tn P 3 be presented 1n the
form aoX 1 + a1x + a2 (mod 3 sum and product are taken.
and a 0 , a 1 , a 2 belong to £ 3 ) \Vhat values can be assumed
by the -cneific1e:nts 4~h a 1 aud a 1 1f the funct1on I (x) lS
known not to assume at least one value 1n E,,. 1 e
I (EaJ =PEa
l.i 2. Decompos1tzon of Functions ol k·VaJued Logics

into F1rst and Second Forms
Any functton I (x1 , x 2 ,.rn) rn Pk n~ 1., can be
presented tn the first form wl1teh JS an 8nalog of the per-
fect dls)unct1ve normal form for Boolean funet1ons
, xn) === max {m1n

~
(f (a1 , a 1.,
fJ
.....,
Here the max1mum lS taken over all tuples (u1 cr ~
O:u , an) of values of the vartables r 1 .zt, , zn t
Another representation of the lunet1ons of k va1ued

log1e, called the second jonn 1s also poss1ble
/ (;~) ~ ~ J (a) Ta, (x 1.J

~
fa D
(.rn)~
a
~
where the sum 1s taken over all tuples c1 =(a,~ , a nl of

values of var1ables x1.~ x 1 " ~ .:rn. (mod k sam and product
are taken)
Example. Present the function I (z, y) ->- max (1 0 (.r) X
Ju (y), :c• {11 (.y) + 2J 2 (y))) 1n P, tn the fi.rst and second
f0:rms
Solulton. F1rst of all~ we comptle the table of values
of the function f .
3.1. REPRESENTATION 87
y
0 1 -
?
0 1 0 l)
1 0 1 -1
')
-
•) 0
-
•)
With the help of this table, we obtain the first and

second forms of the function f.
f (x, y) =max {min (1, J 0 (x), J 0 (y)), min (0, 1 0 (x),
J 1 (y)), min (0, J 0 (x), J 2 (y)), min (0, J 1 (x), J o (y)),
min (1, J 1 (x), J 1 (y)), min (2, 1 1 (x), J 2 (y)), min (0,
J 2 (.r), J 0 (y)), min (2, J 2 (x), 1 1 (y)), min (1, J 2 (x),
J 2 (y)) }.
Using simple manipulations, we get
f (x, y) =max {min (1, 1 0 (x), J 0 (y)), min (1, 1 1 (x),
J 1 (y)), min (J 1 (x), J 2 (y)), min (J 2 (x), Jl(y)),
min (1, l2 (x), J2 (y))}.
f (x, y) = 1·io (x) •j 0 (y) + 0 ·io (x) ·j 1 (y) + 0 ·io (x) X
i2 (y) + O·jl (:x·)·io (y) + 1·it (x)·jl (y)+2·jl (x)·i2(Y)+
O·i2 (x)·io (y)+2·j 2 (x)·i 1 (y) + 1·j 2 (x)·j 2 (y) = io (x)X
io (y) +it (x)·jl (y) + 2·jl (x)·j2 (y) + 2·j 2 (x)·it (y) +
i2 (x) ·i2 (y).
3.1.11. For a given k, present the function I in the
first and second forms (simplify the obtained expressions).
-
(1) f = x, k = 3;
(2) I =""" 1·, k = 4;
(~) I = - io (x), k = 5;
'
(4) I= 2·1 1 (x),I k = G;
(5) I = J 2 (.r 2 + x), k = 5;
(6) f = (......... x) 2 + x, k = 4;
(7) I = 3 ·j 1 (x) - j3 (x), k = 4;
(8) I = :x· + 2y, k = 3;
(9) I = max (.r, y), k = 3;
(10) I =X...!.... y 2 , k = 3;
(11) I== x 2 ·y, k = 3;
(12) I = .r ·Y. k = 4.
3.1.12. Prove the validity of the following relation,
which i~ the analog of the perfect conjunctive normal form
•
88
of the functions 10 PA..

;..,;
f (xfl) ~ m1n
,.;
(max UCat, 0'2,
(J
~ J ,,.(xJ), t ,.., J f1n (.t n))}

>
(here, n 1, and the m1n1mum 1s taken over all tu ple.s
o= (a1 , a 2 } , an) of values of the varwbles x 1 z 1
' Zn)
3.2. Closed Classes

and Completeness in k-Valued Logics
3.2.1 .. Some Closed Classes of k-Valued Log1c:s.
Representation ol Funeflcns
in Pk through Mod k Polynomu,ls
.......
Lel S be a subset of the set E 11 The function f (.rn) tn
P,., (n~ t) LS satd to preserve these' ,g tf tt assumes a va
.,., fl<J'
lue f (an)~ also belongtng to 'l, on any tuple aJ1 =

(a1 , asu ~ an) such that al E t (t = 1, 2, n)
For n = 0, 1t. 1s assumed by defin1t1on that the Iunctlon
f =-=a (a E E 1 ) preserves the set ~ only 1f a E f The
set of allju.ncttons 1n P 11 preservlng the set o lS a closed
tlass, deno-ted by T (~) and c3lled the class preser1J1ng
the set ~ If ~ 1s a proper snbset of the set E~c (1 e
0 ~ 3 ~ E,.J, T (if} =P P~c (see Problem 3 2 3)
Let D = {~ 1 , I :t' " ~ ,} denote the part1t1on of
I
a s~t. E1, 1 e £,. = U ~~~ ~~ 0 lor t = 1 2. t *
~·
.sand 1 1 n ~J ==== 0 for t =I= 1 Elements a and b are sa1d
to be equ1valtnl w1th respect to the partltlotJ D (notat1on
a..._, b (mod D)) 1f a and b belong to a certatn (the samer)
,..,11 ~
subset IJ of the part1t1on D Two tuples a and p-n are

called equtLalent Wtth respect to the part1t1on D (nota
,wt'V
tton et" 13" (mod D)) tf ,..., ....w p, (mod D) for z = t,

J"'IW' a,
2, , n The function f (.xn) 1n P~ (n ~ 1) 1s satd to
,., I'Y
preserve the partltEon D tf for any tuples ttn and ~" the
~ ~
equ1valenee a." lin (mod D) leads to the equtvalence

;r..tJ
3.2. CLOSED CLASSES AND CO:IIPLETENESS 89
f (~ 11 ) """ ~ 11 ) (mod D). It is assumed by definition that

any constant function (i.e., a function depending on zero
variables of the type f = a, a E Eh) preserves any parti-
tion D. The set of all functions in Ph. preserving the parti-
tion D = {c 1 , Cf, 2 , • • • , l;s} is a closed class, denoted by
U (D) or U (c 1 , ;f- 2 , • • • , ~ .• ) and called the class pre-
serving the partition D. If s =I= 1 and s =I= k, the class
U (D) is different from Ph (see Problem 3.2.4).
The function f (x "'11 ) in PIt (n ~ 0) is called linear if it
can be presented in the form a 0 + a 1 ·x1 + ... +
a 11 ·x 11 , where ai E E 1., j = 0, 1, ... , n, and the sum
and products are taken over mod k. The set of all linear
functions in Ph forms a closed class of linear functions,
denoted by £ 11 (or L). The class Lh. differs from Ph for
all k ~ 3. A mod k polynomial of variables x 1 , x 2 , • • • , X 11
is an expression of the type a 0 + a 1 ·X 1 + ... +
am ·Xm, where the coefficients a; belong to the set Eh,
and Xi is either a variable in {xl, X2, •.•• ., X11 } or a pro-
duct of variables in this set (j = 1, ... , m). A certain
function from P 1, can be presented by a mod k polynomial
if there exists such a polynomial equal to this formula.
The set of all functions in Ph that can be presented by
mod k polynomials (or, in short, the set of all mod k
polynomials) forms a closed class in P 11 •
Theorem. Each function in P 11 can be represented by
a mod k polynomial if and only if k is a prime number
(in other words. the set of mod k polynomials in Ph. is
complete in Ph if and only if k is a prime number).
If k is a composite number, then Ph contains functions
that can be t·epresented by polynomials, as well as func-
tions that cannot (for example, the constant functions 0,
1 •... , k - ·1 an d " po l ynomta ~ x ·y,
. l" f nne t.wns x, x-,
x + y can be represented by polynomials, while the
functions j 0 (x), max (:r, y). min (x, y), x...!.. y cannot).
Example 1. Represent the function f (x) = x 2 ...!.. x in
P5 by a mod 5 polynomial.
Solution. Let us first represent the function f (x) in the
second form:
I (x) = ~ f (a) ·ia (x) = f (0) ·io (x) +I (1) ·j 1 (x) + f (2) ·j., (x)
a -
+f (3)·j3 (x) + f (4)·j 4 (x)=O•j 0 (x)+O·j 1 (x}+2·j 2 (x)
+ 1·j3 (x) + O·j 4 (x) = 2·h_ (x} +
j 3 (x~
C.11 a. r.. \ ALL EO LOGIC'S
Then v.e sl1all use t}le iact that Jo (.r) = 1 - xi-1 1f k

1s a pr1me number (.I.> 3) and that ]J {3:) ~ Jo (z - i)
' -= 1 A. - 1 (Here, as usualt \'\ e take mod k
dJfle1enc-r and mod k po\\er) \Ve l1avc
1'1. (z) == 1 - (x --- 2)"' =- -- .x" - + x 2xt
2
2r +
13 (z} = { - (.r - 3)"' ;;=- 1 - (x + 2)"' = --- x~ +
2J:S + :tt - 2.t
Consequently .r1 - .r = 2.t4 - 2r - 2..r 2 + 2x
(fhts
e "t pres.ston can be v. rt t ten Jn a more compact form
2.r· (x- 1}' (x + 1} )
The polynomial represent1ng tlte functton f (x) can
also be determ1ned by ns1ng the method of t.ndetermEnate
toef{tctents Let f (x) = au +
a 1 t.X +
a!. x 2 + Da jlr +
a>4. z' (higher po\\ers need not be cons1dered Stnce x + =4 1
r 1 (m()d 5)} \Ve contprle tlte folloWLng set of equattons

a 0 =f(O) ==0,
«o-+ «1 + <t"l+ «3_, a._~ I (t} ~ 0~
a0 + 2a 1 + 4a 2 _l.... 3a.a +a, .=:;;; / {2) = 2,.
+ +
a 0 + 3a, 4a 1 2a 1 +a~:=:;; f (3)-== 1~
o0 + 4a 1 --t... a2 +4a 3 + a•.:;: f (4) =0
Sol vJng tl11B set of cqua t 1ons., for exam p lc, b~ tl1e n1eth
od of ellminatlon~ l\e obta1n an = 0 a 1 = 2 a"l -;;;;;::. 3
a 1 = 3't at ~ 2
Example 2. Pro\c tlJat tl1e funct1on f (r, y) ~ x 2 - y
In P, cannot he rep1esented by a mod 4 pol)nomtal
Solut1on .. \\re assume tl1at the con\ersc IS true~ t e
that I (x y) cau be repreSPJlfed by a mod 4 pol}nOrulal
and wr1te th1s h)potheticaJ po1)nomJal 1n tJ1e general
form (with 1ndeterm1nate coefficients) j (.r y) ~ (a 410 +
a1oX + a2.,.xJ+aaoX') + (antY + a11xy + attX Y + a,try) +
2
(a(ltY' + a,t.ryl + attzty2 + a3tX'Y2) + (a-u3Y3 + tJ,~y'+
attx"g3 + a a:tx'lt)
(Powrrs l1Igher tl1an 3 need not he considered s1nce lor
l ~ 1 the re Ia t1ons x2'• 2 ~ x 2 (mod 4) and .x21 +3 ::;;;
.ts (mod 4) c1re valtd ) \Ve ha\e
f(t, O)~a 00 +a 1 g+a20 +a30 ~ 1.,
I (1 ~2)=a 011 +aJ 0 + a 20 +a30 + 2 {aot+a 11 +.a21 +a 31) = 0 (•)
3.2. CLOSED CLASSES AND CO;\IPLETENESS !ll
Since the equation 1 + 2a = 0 (mod 4) has no solu-

tions (this can be proved directly by successively putting
a=O, 1, 2 and 3), the set (*) has no solution either.
Consequently, we cannot choose appropriate coefficients
au which would ensure the representation of the func-
tion j in the form of a mod 4 polynomial. Therefore, the
function j cannot be represented by a mod 4 polynomial.
Remark. In the problem considered above, it would
suffice to write two relations between the coefllcients.
However, a more complete set of equations should be
often considered. In the following subsection, we shall
demonstrate other methods to prove that the functions
in Ph cannot be represented by mod k polynomials.
3.2.1. (1) Which of the classes, T ( {0, 2 }) or U ( {0, 1 }.
{2 }), do the following functions in P 3 belong to:
(a)_,.._, x; (b) j 1 (x); (c) ! 2 (x); (d) x...!... y; (e) x + y;
(f) min (x, y).
(2) Which of the classes, T ({1, 3}), U ({0, 1 }. {2},
{3}) and/ or U ({0, 3}, {1, 2}) do the following functions
in P 4 belong to:
-
(a) x; (b) ,..., x; (c) j 0 (x); (d) x + 2y; (e) max (.r, y);
(/) x2 ·y.
3.2.2. For a function f in Ph and for a given k, choose
the classes of the types T (~) or/and U (D) to which this
function belongs. Here G must be a propm· subset of the
set Eh (i.e. ~=I= 0 and o T- Eh), and D must be
a partition {c 1 , • • • • cs} of the set Eh, such that the
following inequalities are satisfied: 1 < s < k.
(1) k = 3, I = x 2 + 1;
(2) k = 3, I = J 0 (x 2 + 2x);
(3) k = 3, f = (x 2 ..!.. y 2 ) + 1;
(4) k = 3, j = X ."y - y + 1;
(5) k = 4, I = j 2 (x- x 2 );
(6) k = 4, I = J 3 (2x + x 2);
(7) ,, = 4, I = X •y - y + 3;
(8) k = 5, f = min (x2 , y);
(9) k = 5, I = (2x2 ..!.. y) - 1;
(10)!.: = 6, 1 = .1:3 ·y + 2;
(11) k is an arbitrary integer not less than 3, 1 =
j 1 (2x ..!.. x 2 ); •
(12) k is an arbitrary integer not less than 3, f =
Jh-1 (x·y- 1}.
92 GH: 3 k VALUED LOGtCS
3.2~3~ (i) Let. t; s;; EJt Prove that T (~)* p._ tf and
only 1f t 1S a proper S\tbset of the set E ~ (1 e 'Q. ::::fo 0
and r-=/:
E~)
(2) Determ~ne tlte number of dlffetent closed classes 1n
Pk whlen preserve tl1e sets'
(3) Let ~ ~ E k Calculate the number o( functtnns
tn P~t contatned tn tlte tlass T (~) and dependtrlg on
tiLe var1ables .r1 , x~"t , x n (n ~ 0)
3~2 4~ (1) Let D = {W 1 ~ ~ 3 } be it part1tton of
the set Ek Prove that U (D) ;:/:: P~a;. 1f and only 1f 1 <
s<A
(2) Fori.- 3, 4 and 5. calcuJate the number of differ-
-===
ent c]osed clas.s~s 1n Ph whtch preserve the part1t1on.s
(3) Let D :;: : : : {t1 , , 'I, } be a part1 t1on of the
set E~t Calculate the number of functtons 1n P1 con
tatned 1t1 the class U (D) and depending on the variab1es
z1, x 2 , Xn (n~ 0)
3~2 5. Let ~ be non empty subset tn E~t other than
the enltre E" and Jet D = {~. E" 'WJ Calculate the
number of luncttons tn Pt~ \VhLch depend on the vartables
x1 z 2 , Zn (n;a: 0) and are conta1ned 1n the set
(1) T (6)" U (D), (2) U (D)' T (~)t (3) T (~) U
U (D)
3 2.6 By S • we denote the set of all dlfJerent valued
junctzons zn P,. whEth depend on a Slngle tartable (t e g (x)
belongs to 5 11 1f and only 1f g (E 11) == E 1 ) Let p~t~ be the
set of all junctlons of the k talued logtc Pit~ whrch depend
on a .s~ngle va.rtable and CS 11 ~ P~11 'S It
(1) Prove that the sets S,. and CS ~ are closed cla.5ses
(2) Caleulate the number oi funet1ons de:pend1ng on
the Varlable X and belongtng to the tlass S l n U ( {0
k--- 2) {1, k-- 3], {k-- i}) (Fnr k ~ 3, we
assume that {1, " k - 3) ~ 0 )
3~2~7 ~ Expand a funct1on f 1n P-. 1nto a mOO k poly-
nomtal
(t) f = 2 r - x1 ! === 5.
(2} f = m1n (.x 2 .t3 } k = 5~
(3) I = max (2 x - 1~ x 2 )~ k 5~ =::;;
(4) / ~ 3 r - (x- 2 z}~ k ~ 1~

(5) / = max ((x- 1) 2 ~ .z-3) k ~ 7,
(6) I == tn1n (x 2, y)t k ~ 3,
(7) I = max (2 ~.r - y, x y) k = 3 1
(8) I == z .. vt k ;;; a.
3.2. CLOSED CLASSES AND COMPLETENESS 93
(9) f = J k- 2 (x), k is an arbitrary prime number;

(10) f = j 2 (x- x 2 ), k is an arbitrary prime number.
3.2.8. ('1) Prove that the function 2j i (x) in P 4 can be
represented by a mod 4 polynomial for any i = 0, 1, 2
and 3.
(2} Prove that if a function in P 4 depending on a single
variable assumes the values either from the set {0, 2} or
in the set {1, 3}, it can be represented by a mod 4 poly-
nomial.
(3) Using the function/ (x, y) = 2·j 0 (x)·j 0 (y) (in P 4 },
prove that statement (2) cannot be generalized to func-
tions depending on more than one variable.
3.2.9. Prove that if a function f (x) in P 4 cannot be
represented by a mod 4 polynomial, then for any integer
m ~ 2 the function (/ (x))m cannot be represented by
a mod 4 polynomial either.
3.2.10. Calculate the number of functions in P 4 which
depend only on a variable x and can be represented by
mod 4 polynomials.
3.2.11. (1) Let a function f (x) in P 6 be represented by
a mod 6 polynomial. Prove that it can be represented
by a mod 6 polynomial of the form a 0 +a1 x +a 2x 2 •
(2) Prove that the number of functions in P 6 , which
depend on a variable x and can be presented by mod 6
polynomials, is equal to 108.
(3) Enumerate all the functions f (x) in P 6 which can
be represented in the form a + b ·j 0 (x) (a and b belong
to E 6), cannot be represented by mod 6 polynomials,
and satisfy the condition that the function (/ (x))2 can
be represented by a mod 6 polynomial.
3.2.12. Find out whether a function f in Pit can be re-
presented by a mod k polynomial:
· (1) 1=3·x....:... 2·x2, k=4;
(2) f=3·j 0 (x), k=6;
(3) 1=2·(ldx)+J4 (x}), k=6;
(4) I= (x....:... y)....:... y, k=4;
(5) I= (max (x, y)- min (x, y))Z, k = 4.
CH 3 k VALU..CD LOGICS
1.1~1~ Completeness Test far Sets ol Functions

cf k-varued Lag1c
In k valued log1cs, the completeness te~t1ng of an ar

h1trary set of funct1ons 1s fraught w1th .set1ous technical
dtlfi.cult1e.s because tl1e application of the completeness
cr1tet1un based on an analys1s of the se.t cf all precom
plete classes 10 P,.. Involves the verificat1on of a large
number of condtt1ons even for A = 3 and 4 (since there are
exactly 18 precomp1ete c1asses 1n P 3 and 82 In P t) The
completeness of concrete sets In P.~c Js normally proved
by redutang \hem to sets known to be oomplt~e before
hand (l1ke the Rosser-Turquette set {0~ i ~ k - 1"
J 0: (.t), J 1 (x)t , I A-t (.z), mtn {z, y) max (x y)J or
-
Post's set {x 1 max (x, y)J) Bestdes, there ex1sts a number
of completeness criteria tn wbtch we cons1der sets of
functions conlalnlng some funct.1ons of a s1ngle v anabl~
and only one (unctlon \\h1cl1 essent1ally depends on a.t
least two variables Let us formulate the most tmportant
criteria of th1s k1nd It should be recalled that
Sit IS a set of all d1fferent valued functions tn P 11 • wh1ch
depend on a stngle: var1able .
l}h.u 1s a set o{ all func.tlons of one var1able 1n PJJ. and
/'IW
CS Jt ~ P~ 1 ' S, The funct 1 on 1 (:c) E PA 1s referred

to as essenttal 1f lt essentlally depends on at least two
variables and assumes all k values 1n the set EA
~
Theorem 1 (Slu peck1 s cr1terton) A set p~u U {/ (.r}}

JIWJ
t.s complete in P)t (fork~ 3) t/ and only tf j (.x) ts an tsun--

ttal function
Theorem 2 (S V Yablonsky s cr1ter10n) A set CS~t. U
....... ,..,
{f (x) }1s complete tn p.,_ (jor k ~ 3) 1./ and only 1( f (.:) 11
an es~ent1al junclton
,.,
Theoren1 3 (A Salomaa s crtterton) A set S..,.. U {/ (.x)}
;l'ltJ
IS complete 1n Pl\ (jor k ~ 5) z/ and only t/ f (.z) ts an essen·

t ~al juncllon
The statements conta1n1ng var1ous completeness crtte.rta
for sets of lunct1ons 1n the sets P~"\ SJt. and CS~e C.tln be
used along Wlth ll\e~ theorems Let us co-ns,der an
example of appl1cat1on of such crtterJa Let a function
3.2. CLOSED CLASSES AND CO:IIPLETENESS 95
hij(x) where 0 ~ i < j~ k - 1. be defme d as follow s:

i if x=j,
h ii (x) =-= j if x = i,
:z. otherw ise.
Theore m 4 (S. Piccar d). Each of the sets (x, hot (x}.
x + j 0 (x)} and {hot (x), h 02 (x), . . . , ho0<-1) (x}, x +
j 0 (x}} is complete in Pj1 °.
3.2.13 . By choosi ng an approp riate class of the type
T (~} or U (D), provo that the set A is incom plete in
pk if:
(1) A= {......, x, min (x, y), X· y2};
(2) A={'i , 2, x...:....j 2 (x), max(x , y)};
(3) A={2 , j 0 (x}, x+Mx )+Jdx )+Jh- t(x),m in(x,y )};
(4) A={J 2 (x), x+j 0 (x},x+ j 0 (x)+J t(x), max(x ,y)};
(5) A={k -1, J 0 (:r), .r...:....y, X·Y·z} ;
(6) x
A={2 ·x 3 • 2·x+ y, x 2 ·y, X·lo(Y ). 2 +( '"""'y)};
(7} A={x ·y, max(x , y)-z+ 1};
(8) A={ -x 2 • max(x . y)+z] ;
(9) ...1={1. -x·y. x2...:....y};
(10) A= {2. max (x, y), x...:.... y};
(11) A={k -2. -h- 2 (x), max(:r , y). x+y};
(12) A={j 2 (x), x+j0 (x)+J d:r), X·!J, x...:....y};
(13) A=[1 , J 0 (x), .1"+h-dx), min(x , y). max(x , y)};
(14) A={1 , 2, ... ,k-l, x+j0 (x), j 2 (x)71 , max(x ,y)};
(15) A={1 , ""x, x...:....y, min(x , y)};
(16) A={,.. ..,x, l 0 (x),Jd :r), ... ,Jh-d x), min(x , y)+
Uo (x) + io (y)) (x + y)};
(17) A={1 -x, iu(x), jdx}, ... , h.-dx) , x·y, x...:....y,
-
min (a., y)}.
3.2.14 . It i!' known (s£>e Sec. :3.1) that the set
A= {0. 1, ... , k-1, j 0 (x), j 1 (x), ... , ik-dx} , :r+y, :e·Y}

is compl ete in Ph.
(1) Prove that it is possib le to isolate from A a subset
which is compl ete in Ph and consis ts of two functi ons.
CII 1 l \ .ALLED J OGIC~
(2) Prove that an~ subset of tl1e ~et A whlch conststs

of a szngle funcl1on 1s 1ncompJete In P/l
3 2 t 5 The Rosser Turq uet te set
A1 - {0 I ~ - 1 J n (x) 11 (;r) J h 1 (.r)
min (x y) ma1: (l y)}
lS .known to be complete 111 11 , (see Sec 3 1)
(l) Verify that h:,. omill1ng 1n A 1 an) constltnt other
than 0 and h. ---- t v.e obtain a subset contained 111 a cer
ta1n class ot th~ l)pe T (l') \\heie 0 # t + 1!.11 (4.\nd
hence 1ncomplele In }J"}
(2) Isolate f1om the .set A1 a subset \\hich 1s c ,mplele
tn l)~c and constsls of 21.. --- 2 functJons
3 2 16 For g1ven k test tl1e comp1eteness ol tl1e Iol
l()w~ng subsets of 1lu~ Rosser TllrQ\lette set
(1} h ~ 3 {t / 1 (x) J1. (z) mln (.r y) max (x y}}
(2) 1.. ~ 3 {1 2 J 2 (.t) m1n (x y) rna 1: (x y)}
(3) A - 4 {1 2 J 0 (x) J 1 (.r) m•u (.x y) ma"< (z y)}
(4) 1..-- 4 {1 2 J 0 (z) J, (x) ntln tx Jl) max (x y)}
3 2. \1 Pro\-e \\\a\ ~atl1 ui \ht iD\\Dw"lng st\~ 11~ t.tm.
plete 1n S ~
{1) {h 01 (x) h 62 (x) ho(ll tl (x)}
(2) {hfll (x) lt 11 (x) h, (l +J) (.t)
(3) {~ hrH (x) J
3 2 tS Prove that tl1e ~et {h01 (.t) h 0 'l. (x)
x + Jn (x)) IS complete 1n Ptt'
3 2 19 Us1ng the metl1od o{ reductlon. to sets wl11th
are known beforeh_.and to be complete prove tl1e com
p]eteness In P 11 of the following .set.s
(J) {Jo (x) Jt (z) JA 1 (x) x 2 x - Y}
(2) (k-1 x---y x+yJ
(3) { ~x x+2 x- y)
(4} ( - x 1 - xz ;1 - y J
(5) {x y ( ~ x) -- 2y}
-
(G) {.r min (.r y)J
(7) {m1n(x y)-1}
{8) {Ill .(r) x+y X yJ
(9) {i zZ+ y x2- Y}
(10) {/6 (~) x+y ry'}
(11) {k-2 x v+t {,._,z)-yJ
3 2. CLOSED CT.ASSF.S AND CO:IIPLETENESS 97
(12) u~-1. x:! - y, x 2 ...!.. y};

(13) {1, 2·x + y, x--=- y};
(14) {1, .'t2 -y , min (x, y)};
(15) {1, x+ y+ 2, :r2--=-y};
-
(Hi) {.r- j 0 (y), min (.1-, !f)J.
3.2 .20 . Using the Sln pec ki crit erio n, pro ve the com -
pleteness in P" of the foll ow ing sets :
(1) {k -1 , .c -y +2 , x 2 --=-y};
(2) {j ~ (.r). .c + y\ .L' ·y + 1 };

(3) {.c--=-y, (-. r) -y} ;
-
(4) {j I (.c), X - !J, ,t:2 - y };
-
(5) {.1·, j 0 (x), .C· y}; .
(o) {x -1, (x+ h(x ))·( 1-- =-y )+( 1-- =-x )·(y -jt( y)) };
-
(7) {(1--=- x) · y+ x· (1--=- y)};
(8) {x· j 0 (x - Y) + (x - i 1 (xj) · jo (y) + Y· io (x)};
(9) {j0 (x - y) +X· j 0 (y) + (x - j 1 (x)) · j 1 (y)];•
(10) {x· j0 (Y )+j0 (x) ·(Y +jo (Y )-j, (y) )
+ -
j, (x) · (y - iu (y))};
-
(11) {y · io (x) + j dx) · (y + io (y)) + jt{y ) · (j 2 (x) -j 1 (x)) }.
3.2 .21 . Tes t the completenes~ in Ph of the foll ow ing
sets :
(1) {k -1, x+ 2, max (x, y)J;
(2) {1, 2, x·--=- y);
(3) {k- 2, x+ y, min (x, Y)};
-
(4) {0, 1, x--=-("'y)};
(5) {2, 2·x +y , x2--=-y};
-
(6) {1, 2, max (x, y)};
(7) {2 --=- x, x · y, max (x, y) };
(8) {k- 2, 2·x +y , x--=-y};
(9) { """x, -x ·y, min (x, y)};
(10) {2, x+ y, x--=- y};
(11) {"' .c, 2·j0 (.c), Jt( x), x--=-y};
(12) {1, ....., x, J 0 (x) + ldx ), max (x, y));
(13) {0, 1, ""'x, 2- j0 (x) -2· j 1 (x), min (x, y)};
(14) {1, k- 1, x-- =-[ ;J, min (x, y)};
(15) {k- 2, ......,. X, X...:... y}.
7-06 36
98 CH 3 A \ALL ED LOGIC:t
3.2.22, I'rov~ that the flJllo\Vtng sets arfJ complete 1n

fJ,. t( ·'-~ul (~nl \ t ( I t ~ 1 prlmP number
(i) {1, .I: T Y T J. •zJ,
(2) .r-y+t x'-y :ry~)
(3) ;r .__ 1t ;t + y, .t1! y Jt
(4) {~ - f, X •y T X - 1/ + .Z },
(5} {J. - 2~ $ - 2 y, X y2 } ~
(6) {~x, x...-y, x 2 ·yJ,
(7} {X- 2, X+ 2 y 1, X y +
X-- y},
(8) {f. 2. X+ y, X•!l 2 - X+ y),.
(9) {x +
11 + 1. x y - xt},
(f0) {z - 2gt X y X ~ 1 }, +
(ff) {f + x 1 - Xs + :r1 ·X~ • ~ X~c}
3t2.23. Pro\e tl1at the folloYttng fun<:t1ons 1n Ptt cannot
be reprs.sen ted by mod h pol vnom lals 'n the ca ~p or
a eompos,te k
(1) It (x), 0~ l~ 1.. ~ 1~
(2) max (.t, y) *
(3) mtn (x y}.
(4) X - Y•
(5) z::> y.
(6) (x - y) - zf
(7) any Slteffer s £unctton (l ~ Ct ft~nct ton £ormtng
a compJ~te system 1n. Plf.)
3.Z.24 IIav~n~ thos~n {or a funcl \On J {.r y) t lle pol:..
nom1als Q0 (.r), Q1 (2) a11d Q2 (.r) "uch t Ita t at least one
of the t u ncttons Q() (/ (QJ (x) Qr. { y))) or Qo (/ { Q1 {.r),
Q 2 (x))) 1s known not to he e\pandabJe 1nto ,a mod J.. pol)
nomtal pro\e that for h == 4 and 6 the folJowtng funr
ttons cannot be represented b} mod A pol} nomta I'h)
-
(1) /=(2-x3) y,
C2l t=<( ~ x>+
y)- (;--- 1)
(3} f=mln(~.:r~ y}-(1-x)
(4) I= max (x, y) .._, (x- 2)
3.2.,25. Isolale the be.s1s of ll1e ~t A v.h1cl~ \S tom
plete 1n 11 1
(1) A= {h- 1, ] 0 (.r) It {x), ]II; t (.r) X y,
:z:- Y}t
(2) A-=; x _. 2, J 0 (x) mtl't (x g) x- y2, r2 yJ,
(3) A=:; ,...., z~ m1n (xl y)l x y, x + y],
(4) A==:{k-tt x+2 ma~(xt y}'t x~yJ•
(5) A=;; l2, }0 {x). x + yz, x" ~- y, x~ y z}
3.2.26. Let B be an arbitrary basis isolated from a Ros-

ser-Turquette set. Prove that in this basis
(a) there exists at least one of the functions J 1 (x},
0~ i~ k- 2,
(b) there does not exist the constant 0,
(c) if J 0 (x) E B, then k - 1 E£ B.
3.2.27. Prove that if a closed class in Ph. has a finite
complete set, the set of all different bases of this class is
not more than countable. (Two bases are assumed to be
different if one of them cannot be obtained from the
other by redesignating the variables without identi-
fying them.)
3.2.28. (1) Let A be a non-empty set of functions of one
variable from Ph., which differs from the entire set Ph.u
and satisfies the following conditions: there exists a pre-
complete class B in Ph., such that B n Ph.0 =A. Prove
that such a class B is unique.
(2) Prove that the number of precomplete classes
in P", each of which does not contain the complete set
PJ.ll, is smaller than 21!".
3.2.29. (1) Prove that the closed class K 1 = [x 2 ·y 2 ) in
P 3 contains neither constants nor identical function.
(2) Prove that the closed class K 2 = [h (x) ·j 2 (y)] in
P 3 does not contain functions which essentially depend
on a single variable.
3.2.30. Two functions in P" are referred to as congruent
if they can be obtained from each other by redesignating
their variables without identifying them. Prove that the
closed class K 3 = [j 2 (x) ·j 2 (y)l in P 3 consists of a finite
number of pairwise non-congruent functions which essen-
tially depend on all their arguments.
3.2.31. Let us consider the closed class K 4 = [11 (x1 ),
lz (xl, X2), 13 (xl, X2, X3), · · ., In (xl, X2, ••. , Xn), ••• ],
~
where In (x") = j 2 (x1)·j 2 (x 2)· ... ·j 2 (xn), n = 1, 2,

3, . . . . It obviously contains an infinite number of
pairwise-non-congruent functions. Prove that K 4 does not
contain precomplete classes.
3.2.32. Calculate the number of essential functions
in Ph., which depend on the variables x 1, x 2 , • • • ,
X 11 (n~ 2).
3.2.33. It is known that for k ~ 3, there exist in Ph.
closed classes having no bases and closed classes with
countable infinite bases. Let us consider the class Ah. =
7tt
100 CH 3 A VALUED LOGICS
fm, ] where
Xm)
1 for z 1 = =.%1-t =Xt+t =
= x 1 ~1(i=1, ,. m}
0 otherw1se
m~ 2 The set {/2 , fm, } Is a basts 1n A,._ Us1ng
>
t.be elass A 1.., prove that there ex1st.s 1n P k (k 3} a conttn
ual famtly {Bv} of closed classes whtch form a chain
upon lneluSlOD, l e only one ol the lDCluslODS Byl c: n,..
or B-v t c. B~t 1 1s valtd for any two classes B~"I and B~, t
1 1
1n the fam1Iy {B v}
Chapter Four
Graphs and Networks
1
4.1. Basic Concepis in ihe Graph Theory
Let l' be a finite non-empty set and X be a tuple of
pairs of elements in V. The tuple X may contain both
pairs with identical elements and identical pairs. The
set V and the tuple X define a graph with loops and multi-
ple edges (to make it short, a pseudograph) G = (V, X).
The elements of the set V are called vertices, while the
elements of the tuple X the edges of the pseudograph.
The edges of the type (v, v) (v EV) are called loops. A pse- •
udograph without loops is referred to as a graph with

multiple edges (or a multigraph). If none of the pairs is .
contained in the tuple X more than once, the multigraph
G = (V, X) is called a graph. 2 If the pairs in the tuple X
are ordered, the graph is referred to as oriented. The
edges of an oriented graph are often called arcs. If the
pairs in the tuple X are disordered, the graph is called
non-oriented and is referred to simply as a graph. If x =
(u, v) is an edge of a graph, the vertices u and v are known
as the ends of the edge x. If v is an end of the edge x, v
and x are known to be incident. The vertices u and v of
the graph G are called adjacent if there exists an edge of
the graph G connecting them. Two edges are called
adjacent if they have a vertex in common. The degree pf
a z:ertex v is defined as the number d (v) of the edges of
the graph which are incident with the vertex v. The
degree of a vertex v in a pseudograph is equal to the total
number of the edges incident with this vertex plus the
. ~ The definitions given in this section coincide with, or are

similar to, those presented in ref. [23] and in the chapter "Graphs
and Networks" in ref. [6].
2
Hence!~rth, all definitions are given for graphs. As a rule,
these defimt1ons can he extended naturally to multigraphs and
pseudographs. When the differences in definitions .are significant,
definitions for pseudographs are formulated separately.
{02 CH I G:RAPBS AND NETWORKS
number of the loops tnctdent wtth tt A vertex of a graph

of degree 0 ts referred to as an lsolated vertex and a ver-
te~ of degree 1 Ls known as a pendant vertex
The sequence
(1)
1n whtch vert1ces and edges alternate and such~ that for

ea(rh t ~ 1.,· n··_:-1. the -edge x 1 has the form (vi 11 1 +J)•
is called the walk cnnneettng the vert1ees v1 and u,.
The number of edges 1n a walk 1s termed as the length
c/ walk A sequence cons1stln.g of a Slagle vertex lS called
a walk of length. 0 A walk 1n wh1eh all edges are palrW~se
dtffe.rent IS known as a chatn A walk 1n wh1eh all verltces
a[e palrwlse d1tlerent is called a stmple chatn The walk (t)
IS ea.lled cla.red tf v1 = v n A closed walk all whose edges
are pa.lrwlse dtfferent ts called a cycle A cycle 10 whtcb
att vert1ces except the first and the last~ are pa1rv.1se
dtflerent ts called a stmple cycle A graph 1s termed
connected tf for any of tts two vetltces there. extsts a chatn
connecttng these verttces The dlstance between verttees of
a connected graph LS defined as the length of the shortest
chatn conneettng these vertttes The dtameter of a. connect-
ed graph ts the diStance between two verttces having
the largest separatton The dtan1eter of a graph G J.S
denoted by D (G) A subgraph of a graph G IS a graph all
whose vertices and edges are eonta1ned among the ver-
tices and edges of the gcaph G A sub graph ts proper tf 1t
d1fters from the graph 1lself A connected component of
a graph G 1s 1ts connected subgraph whtch IS not a proper
subgraph of any other connected subgraph of the graph G
A subgraph contatntng all the vertices of a graph 1s called
a spann1ng subgraph A subgraph of the graph G == (V X)
generated by the subset U ~ V ts a graph H = (U Y)
the set of whose edges consists of those and only those
edges of the graph G whc>se both ends l1e tn rJ The graphs
(pseudograpb.s) G = (V, X) and H = (U Y) are lSomor
phzc 1f there ex1st two one to one conespondenees rp V .. ~
U and'(.~ X . .- \' Y such that for any edge x = (ut v) 1n X
we have ,., (z) -= (rp (u), q; (v)) For graphs, we can for-
mulate the following definition GriJphs G = (Y, X) and
H == (U. Y) are zsomorphfc 1f there exists a one-to one
mapp1ng q; V .... U such that (u v) E X 1f and odly 1f
•
U. BASIC CONCEPTS f03
(cp (u), cp (v)) E Y. Such a mapping cp is termed an isomor-

phic mapping (or isomorphism). An automorphism is an
isomorphic mapping of a graph onto itself. The operation
of omitting a vertex from a graph G consists in the removal
of a certain vertex together with the edges incident with
it. The operation of omitting an edge from a graph G =
(V, X) consists in the removal of the corresponding pair
from X. Unless stipulated otherwise, all the vertices are
-
preserved in this case. The complement G of a graph G
is a graph in which two vertices are adjacent if and only
if they are non-adjacent in G. The operation of subpartition
of the edge (u, v) in the graph G = (V, X) consists in the
removal of the edge (u, v) from X, addition of a new
vertex w to V, and addition of two edges (u, w) and (w, v)
to X '- {(u, v) }. A graph G is called a subpartition of
a graph H if G can be obtained from H by a consecutive
application of the edge subpartition operation. Graphs G
and H are homeomorphic if they have subpartitions which
are isomorphic. Let G = (V, X) and H = (U, Y) be two
graphs. We shall denote by G ED H a graph known as
the symmetric difference of the graphs G and H with a set
of vertices W = V U U and the set of edges Z = X ED Y
consisting of those and only those edges which are in-
cluded exactly in one of the sets X or Y. We shall denote
by G X H the Cartesian product of the graphs G = (V, X)
and H = (U, Y) i.e. a graph whose vertices are the
pairs of the form (v, u) (v E V, u E U) and such that its
vertices (v1 , u 1 ) and (v 2 , u 2 ) are adjacent if and only if
at least one of the pairs v1 , v2 (in the graph G) or u1 , u 2
(in the graph H) is adjacent. The union of the graphs G =
(V, X) and H = (U, Y) is the graph E = (V U U,
X U Y).
A tree is a connected graph containing no cycles. A graph
without cycles is known as a forest. A graph is com-
plete if each two of its different vertices are connected
by an edge. A complete graph with n vertices is denoted
by K n· An empty (null, completely disconnected) graph
is a graph without edges. A single-vertex graph without
edges is referred to as trivial. A graph is bipartite if the
set of its vertices can be divided into two subsets (two
fractions) V 1 and V 2 so that each edge of the graph con-
nects the vertices of different fractions. A bipartite
graph with fractions YJ. and V 2 and the set of edges X
CH f. GRAPHS AND NETWORKS
w1ll he denoted by (V1 , V ,, X) If each vertex of V 1 IS

connected by an edge w1th each vertex of V 2 , the graph 1s
referred to tt.s a complete b1parttte graph A complete
bJparttte graph (V1 , V 2 , X) such tl1at l V1 l = n1 , l V 2 l=
n~ 1s denoted by K:n J n 2 A graph 1s assumed to be k ...
t
connected 1f the om1sston of any of 1ts k - - 1 vertices

tesults Jn a connected graph differing from triVIal The
vertex of a graph wl1ose removal tncreaseq the number
of connected components 1s known as a separating z.-ertez.
o.r a cu.t vertex A graph 1s called a regular grapk {jf degree cl
1f all 1ts vert tees are ol degree d A regular graph of
degree t IS known as a matchtng A regular graph of
degree 2 1s called a cub1c graph The k-factor of a graph
1s Its spanntng regular subgrapb of degree k 'Fhe 1 factor
1s known as a perfect matchtng The martmum mat.ch.Eng
of a graph G IS the match1ng conta1n1llg the maxtmum
number of edges. The Hamlltdnlan cycle of a graph 1s
a Simple cycle conta1n1ng all tllP vertices of the graph
The H amt llonfan cliain 1s a s1mple el1a1n conta1n1ng all
the vertices of a graph A u.nlt n--dlmen&Lonal cube rs
a graph onwhllse vert1ces are Boolean vectors of length n,
and the edges are one d1mens1onal faces (see Sec 1 t)
4.1.1. Prove tl1at for an arb1ttary graph G === (V, X),

the equal1ty 2 J X l = 2.1 d (v) IS val1d
~ev
4 ~ 1.2. Let 111: (G) be the number of verttce.s of degree k
JD a graph G Determ1ne the number of pairWise non-
lsomorphle graphs for wh1eh
(t) l 2 (G) = l:~ (G) = ,_. (G) = 2.. i~ (G} ~ 0 !or k =F
2., 3,. 4.
(2) 12 (G) ~ r., (G) ~ ,,. (G) === 3, 1• (G) = 0 for k +
2.. 3, 4
4. .. t .3. Prove that 1n any graph Wltho\lt mllltlple edges
and loops wh1c:h has not less than two vert1res, there
eXIst two vert1ees w1tl1 identical degrees
4~1 .. 4. Prove thnt for any tuple of non·negat1ve Inte-
gers (ko, k1 ,. ).,. such tbat 2J lk 1 = 2m., there e.x1sts
t
a pseudograph With m edgest wh1ch has exactly k, ver
ltces of degree i for each l == 0, t,
4. i .5. Let d 0 (G) be the m1n1mal of degrees of the
verttces of an n·vertex graph G
4.1. BASIC CONCEPTS 105
IL-f h
(1) Prove that if d 0 (G) ~ 2
, t e graph is con-
nected.
(2) Can [
11
;-
1
J he substituted for (n-1)/2 in the
previou~ statement?
4.1.6. Prove that any closed walk of an odd length
contains a simple cycle. Is a similar statement valid for
walks of an even length?
4.1.7. Prove that a connected graph with n vertices
contains at least n - 1 edges.
4.1.8. Prove that a connected graph with n vertices
and c connected components has a number of edges not
exceeding !- (n- c) (n - c + 1).
4.1.9. Prove that any non-trivial connected graph
c.mtams a vertex which is not a separating vertex.
4.1.10. Prove that any two simple chains of maximum
length in a connected graph have at least one vertex in
common. Do they always have a common edge?
4.1.11. Prove that if an arbitrary edge contained in
a simple cycle is removed from a connected graph, the
latter will remain connected.
4.1.12. Show that if a graph with n vertices contains
no cycles of an odd length and if the number of edges
exceeds (n- 1)2/4, the graph is connected.
4.1.13. Determine the number Ph. (n) for which any
graph with n vertices and Ph. (n) cycles of length k is
connected.
4.1.14. Let graphs G and H be isomorphic. Prove that
(1) for every d~ 0, the number of vertices of degree
d in the g1·aphs G and H is the same;
(2) for every l, the number of simple cycles of length
l in graphs G and H is the same.
4.1.15. Prove that conditions (1) and (2) in Problem
4.1.14 are insllfficient for the graph G and H to be iso-
morphic.
4.1.16. Indicate the pairs of isomorphic and non-iso-
morphic graphs among those shown in Figs 3-6. Justify
your answer.
4.1.17. Let graphs G and H be 2-connected, have six
vertices and eight edges each. The graph G has exactly
two vertices of degree 2, while the graph H has four ver-
100 CH ' G.R.APUS .AND NETWORKS
(a) (b)
(a)
(a) (b)
4.1. BASIC CONCEPTS 107
•
tices of degree 3. Can we state that the graphs G and H

are (1) isomorphic, (2) non-isomorphic?
4.1.18. Graphs G and H are 2-connected, have six
vertices and ten edges each. One vertex in each graph is
of degree d (1 ~d~5), while the remaining vertices are
of degree d 1 (d1 < d). Prove that graphs G and H are iso-
morphic.
4.1.19. Prove that in a graph having no non-trivial
automorphisms
(1) the distance between any two vertices of degree 1 is
larger than two;
(2) there exists a vertex of degree 3 or higher.
4.1.20. What is the number of aulomorphisms of a
graph which is a cycle of length p?
4.1.21. Construct a graph without cycles which does
not contain non-trivial automorphisms and has the
minimum possible number of edges.
4.1.22. What is the smallest number n (n > 1) of
vertices in a graph having no non-trivial automorphisms?
4.1.23. Let the degrees of all six vertices in a 2-con-
nected g1·aph having nine edges be the same, and the
number of simple cycles of length 3 be two. Reconstruct
the graph and determine the number of its automorphisms.
4.1.24. Let 0 (v) be the set of all vertices adjacent to
v and 0' (v) = 0 (v) U {v }. Let R n be the set of all graphs
G with n vertices, which possess the following properties:
for any two non-adjacent vertices u and v, either 0 (v) £
0 (u) or 0 (u) £ 0 (v), and for any adjacent vertices
u and v, either 0' (v) £ 0' (u) or 0' (u) £ 0' (v). Let
G E R n· Prove the following statements:
(1) all the vertices of the same degree in a graph G
are either pairwise adjacent or pairwise non-adjacent;
(2) there is at least one vertex of degree n - 1;
(3) if for a certain d the vertices of degree d are pair-
wise adjacent, the vertices of a degree higher than d are
also pairwise adjacent;
(4) a graph G is uniquely determined (within isomor-
phism) by specifying the degrees of its vertices;
(5) the removal of a vertex in G leads to a graph in
Rn-1·
4.1.25. Let n~ 2, and let the family F (G) =
{HI, H2, ... , H n} of graphs be specified so that a
certain graph H; is obtained from an n-vertex graph G
108 CB '- GRA.PliS .AND NE'tWOI\t(.S
by the removal of a vertex w1th a number l (r = t n)

It should he noted that the vert1ces lfl graphs H 1 are not
labelled Prove tl1at UsJng the fam1ly F (G) 1t 1s pos
stbJe
(t) to determine the number of edges of the graph G
(2) to determane the degree of the vertex 1n each H 1
whose removal from G leads to the graph H 1
(3) to find out whether an arbtlrary graph L havtng
not more than n - t 'ert1ces JS a subgraph of the graph
G
(4} to find out whether the graph G 1s connected and
(5) to reconstruct the graph G 1f tl IS dtsconnected
-
4 f 26 Let D (G) be the d1ameter of a graph G and G
be a complement to the graph G Prove that D (G)~
-
3 af the graph G LS dtsconnected or lf D (G)# 3
4 I 27 A graph G IS termed self complementary If the
-
graphs G and G are tsomorph1c
(1) F1nd a self-complementary non tnv1al graph w1th
the .smallest number of \Crttce.s
(2) Prove that a ~e]f complementary graph IS connected
(3) Prove that 1f G IS a .self-eolll.plementary graph then
2~ D (G)~ 3
4 1 .28 Ftnd the number of pa1rw1se non Jsomorphlc
graphs WJth 20 vert•ces and 188 edges
4 t 29 Pro'\? e that 1f graphs G and H are home om orph1c.
then
(1) for elery d +2 the number of vert1ces of degree d
1n both graphs ts the same
(2} there e:ttsts a one to one mapptng of the set (]f
stmple cycles of the graph G onto the set of s1mple cycles
of the graph 11 for wbtth the number of vert1ces of de
gree d lh the correspond1ng cye)es \S the sa.me (or all
d*2
4 I 30 F1nd out whether the graphs shown 1n F1g 1
contiu.n subgr:ap:hs homeo-morph1e to the graph G 1f
(l) G = K 4 {see F1g Sa)
(2) G ==:: K & (see Ftg 8b)
(3) G = K 1 1 (see F1g 8c)
4 1 3 J The superdtVlSlon operat1on tonsJsts 1n the
replacement of two adJacent edges (u v) and (t w)
whose common vertex v 1s ol degree 2 by a stngle edge
(u w} Applytng the superdtvtsion operatton step by
4 1. BASIC CONCEPTS 109
step, we can obtain from an arbitrary graph G containing

vertices of degree 2 a pseudograph which does not con-
tain vertices of degree 2. This pseudograph will be refer-
red to as a complete superdivision of the graph G.
(a) (c)
Fig. 7
(a) (c)
Fig. 8
(1) Prove that a complete superdivision of graph G

does not depend on the order in which the superdivision
operation has been applied to pairs of adjacent edges of the
graph G.
(2) Prove that the graphs G and H are homeomorphic
if and only if their complete superdivisions are isomor-
phic (as pseudographs).
4.1.32. Prove that the Petersen gmph (Fig. 7a) has
no Hamiltonian cycle, but the graph obtained from it
by the removal of a vertex has a Hamiltonian cycle.
4.1.33. Prove that each of the graphs K 11 , Kn,n• and
Bn has a Hamiltonian cycle.
4.1.34. Let n he odd and B~ be a subset of vertices of
11
a cube B , consisting of vertices of weight k. Let G be
a subgraph of the cube B 11 , generated by the set B~- 1 U
2
B~+l· Does a perfect matching exist in the graph G?
2
IIU CU :r. ORAPllS A~D NtT\VORl\.S
4.1.35 .. A graph G conta1ns a HamlltonJan eyelet \\hlle

a graph H c.ontatns a Ham1lton1an chaln Is 1t true that
a Ham1lton1an C) cle ex1sts tn the graph G X H?
4.1 ~36 .. Prnve that 1f the condttton d (u) + d (~};;;, n
1s sa ttsfied for any t \\ o verttces u ant( v of a con nee ted
n vertex graph, the latter has a Ham1lton1an cycJe
~.1 .. 37.. Pro\e that any graph Wlth n vertices~ con
n 1.
ta101n g at least
2
+ 2 edges has a HamJ lton18.t1
cycle
4.1.38. Prove that a graph conta1n1ng two non adJa
cent vertltes ot degree 3 and other verttces of a degree
not htgher than 2 does not ha \ e a Ho.mtltontan cycle
4.t 39. (t) Pro'e that D (G x 11)~ D (G)+ D (H)
(2) Is Jt true that D (G X H)~ max {D (G), D (H)}~
4.1 4()•.. Prove that each regular connected graph of
degree 2d can be represented an the form of a un1on of
disJoint 2 factors
4 ,.1.41•. Prove that a graph K 1 n can be represented
1.n the fo-rm of a \1n1on of a 1 factor and n - 1 Ham1lto-
tltan cycles
4 1 .. 42. Prove that a K 2 n+t graph can he represented
1n the form of a un1on of n Ham1ltonJan cycles
4.. t 43 Prove that a graph K n wtth labelled vert tees
conta~ns (n-;i)l d1Her~nt Ham1llonlan c)cl~s
4.1 ~44 . Prove ll1at the numb~r of d1Herent perfect
match1ngs of the graph Ktll wtth the labelled vert1ces ts
(2n)!
2nn1
4-l. Planarity, Connectivity,

and Numerical Characterisfics of Graphs
A graph ts called planar 1f 1t can be drawn on a plane
so that the arcs ot the curves representing the edges an
tersect only at po1nts oorrespondlng to the vetl1ees of
the graph ~loreover,. 1rrespect1ve of tbe potnt of Inter ...
sec.t1on, only the arcs corresponding to the edges tnCI-
dent exactly w1th the vertex correspond1ng to a gtven
potnt converge at thts patnt Such a repres.entatton
wb•ch maps a planar grapl1 1s caJied a plane (or topolo-
4.2 PI. \::\ARITY AND COr\NI:CTIVlTY 111
gical planar) graph. The internal face of a plane connected

graph is defmed as a finite region of a plane, bounded by
a closed walk and containing neither vertices nor edges of
the crraph. The walk bounding a face is called a boundary
of tl~e face. The walk of a plane consisting of the points
which do not belong either to the graph or to any of
its internal faces is called the e:ctemal face. For 2-connec-
• •• ...
•• •
(a)
Fig. H
ted plane graphs (multigraphs) having n vertices, m

edges and r faces, the Euler formula 12 - m + r = 2
is valid. The following planarity criterion is also ful-
filled.
Theorem (Pontryagin-Kuratowski). A graph is planar
if and only if it does not contain subgraphs homeomorphic
to the graphs I(, and K 3 , 3 (Fig. 8b, c).
The thickness of a graph G is the smallest number t (G)
of its planar subgraphs whose union is equal to G. Each
non-trivial connected plane pseudograph G can be put
in correspondence with a dual pseudograph G* as follows.
Within each face of the pseudograph G, we choose a ver-
tex belonging to the graph G*. If x is an edge of the
graph G, lying on the boundary bet ween faces g 1 and g 2 •
and V 1 and v 2 are vertices of the pseudograph G* taken
on these faces, the vertices v1 and v 2 are connected by
an .edge in G*. A plane pseudogmph (graph) isomorphic
to Its dual pseudograph is refened to as self-dual. The
cycles Z1 , Z 2 , • • • , Zk of a graph G are linearly depen-
_....- t·1 < t·2 < • • •
dent if f or some t·1 , 1·2 , • • • , t·5 ( 1 ::::::::::
< is~ k) the relation Z;, EI1 Z;, EI1 ... EI1 Z;, = 0,
!s satislied, where 0 is a graph without edges. and Z EI1 Y
Isthesymmetl'ic difference between the graphs Z and}'.
~therwise, the cycles zl. z2 . ... , zk are refened to as
llltearl~ independent. The maximum number ; (G) of the
cycles tn the aggregate of linearly independent cycles of
the graph G is termed the cyclomatic number of the graph G.
112 CJI & GJIAPHS A 'lD NrT\VORKS
fhe colouring ol the vcrttces (edges) uf a grap11 called IS

eortect 1f ad)atent vcrt1c.es (edges) l1~,. e d 1fieren t co tours
The smallest number x (G) of colours for l\hich there
extsts a correct colour1ng of the vert1ces of the graph G
IS known as the chromatiC number of the graph G Tlle
smallest number x (G) of colours for wl1tcl1 there exJ.sts
a correct colour1ng of the edges of \he grapll G 1s called
ttle edge chrontat!C number of the graplt G 11te subset U
of the verttces (edges) xs called the tot-er1ng of the set of
vertlces (or edges) o.f a gt'dph G tf each vertex (edge) of the
graph ettber coincides W1tl1 a certain element of the .set
U or IS adJacent to a certain element ot U (accordingly
lS lne1d~nt wlth a terta1n element of U) A cover1ng 1s
termtnal tl tt ceases to he a CO\ ertng as a result o( the
ornisSion of any element The mJntmum power of a
subset U of the vert1ces of a grapl1 G, such that each edge
of the grapl1 1s 1nc1dent W1tl1 at least one vertex 1n U 1
lS t\enoted by et 0 (G) and IS called the l}trte.r covering
number Tlu~ m1n1mum po~er -o[ tl1e. suhset. Y -of \l~e cd
ges of a graph G., sucl1 tb.at eacJ1 'ertex of the graph IS tn
c1dent With at least one edge of Y 1s denoted by a: 1 (G)
and IS called the edge tovertng nutnber of the graph G
The set of 'erttces (edges) of a grap1t G ts called Endtpen
dent lf none of 1\s t\\o elements are adJacent By ~ 0 (G)
(resp P1 (G)) we denote t(\e max.tmttm po\\er o( an tn
dependent .set of verttces (resp edges) of a graph G The
number Po (G) (number ~~ (G)) lS called tile ~ertex (resp
edge) 1.ndependence number of the graph G By a 00 (G)
\\e denote t\u;': mtnlmltm po\\er of a set of vertices U S\lth
that each. vertex o{ tl\e grapb G wh1ch \S n-ot contained 1n
U IS .adJacent to at least one "ertex In U
4 2 1 Are the graphs sho,vn tn F1gs 6a and b and 7a b

and c planar'
4~2 2. For wh1ch n ~ 2 are tlte graphs .shown 1n F1gs Ya
and b planar?
4 2~3 Let G" ~ (V1 V, X) he a bapatltle graph
v1 = {aJ a,, ' an),

v_ ~ {b1 blto t bn}.,
X === {x 1 ~ x 1 ' .Xn, !/1, Yr. 1 Yn, z,. Za Zn},
'..!!. PLANARITY AND CONNECTIVITY 113
where
xi= (a 1, bi), i = 1, n;
Yi=(ah bi+t),
::, = (ai, bi+:J, i= 1, n-2,
=n-1 = (an-1• b1), Z 11 =(an, b:J.

Determine n > 2 for which the graph Gn is planar.
4.2.4. (1) What is the minimal number of edges that
have to be removed from the cube B 4 for the obtained
graph to be planar?
(2) What is the minimal number of vertices that have
to be removed from B 4 for the obtained graph to be pla-
nar?
4.2.5*. Let G be a 2-connected plane graph having at
least two internal faces. Prove that there exists a simple
chain belonging to the boundary of the external face and
such that its removal leads to a 2-connected plane graph
with fewer faces.
Remark. The removal of a chain involves the removal
of all its edges and intemal vertices, while end vertices
of the chain remain in the graph.
4.2.6. Using the result obtained by solving Problem
4.2.5, prove by induction that the number of internal
faces in a plane 2-connected graph with n vertices and m
edges is rn - n + 1.
4.2.7. Prove by induction over the number of edges
that the cyclomatic number ; (G) of a pseudograph with
n vertices, rn edges and c connected components is equal
to rn- n +c.
4.2.8. Determine the cyclomatic number ; (G) for
the graphs sho,vn in Figs Sa, Ga and 7a and isolate a sys-
tem consisting of ~(G) linearly independent cycles.
4.2.9. Prove that any planar graph contains a vertex
of degree not higher than 5.
4.2.10. Prove that any planar graph having at least
four vertices contains at least four vertices of degree not
higher than 5.
4.2.11. Prove that if each simple cycle in a connected
planar graph with n vertices and m edges contains at
_...- k (n-2) •
least k edges, th en m;:;:,...
~ k-2
S-0636
114 CH I G.RAPfJS A 'lD NET\1/0it.KS
4.2.f2. A plane connected graph whose every face

{1nclud1ng the e~ternal face) 1s bounded by a C}tle o(
length three IS ealled a trtangulatJon
Prove that any tr1angulat1on \\Ilh n>
3 vertices has
3n - 6 edges and 2n 4 faces #-
4 2 .. 13. Let t (G) be the thtckness of a graph G Prove

that
(f) t (K11 ) >( t 11 7
]•
(2) t(Knm)~ l~tlJf_:::. 2r[•

(3} t(Bn) ;;z, Jntt [
4.2 t4•. Ustng Euler! formula, prove that graphs
K, 1 and K 1 are not planar
4.2.15* ~ Prove 'lbat. a 2 connected graph w1th a cye-
lomatlc number~ - 1 can be ohtatned from a 2 connected
grapb. Wllh a c.yclomallC number ~ (i > i) by removtng
a ehatn
4 .. 2.16. Construct a graph dual to the graph shown In
Ftg Sa
4.2.11 . Prove that a pseudograph dual to a connected
plane graph lS connected and plane
4.2.18. Prove that the cyclomatlc number of a dual
graph cotncldes Wlth tl1e cyclomat1.c numbe-r of the
or1gtnal graph
4.2.19. Prove that 1f G has a separat1ng vertex, G•
has a separat1ng vertex as v..ell
4.2 20 .. Prove tllat a pseudograph G• dual to a 3 con
nected plane graph G has no loops or n1ulttple edges
4.2 2!. Determ1ne the number of pa1rwtse non tsomor-
phlc self dual plane 2-conneGted graphs w1th s1x vertlces
4.2.22 . Prove that there ex1st no 6 connected planar
graphs
4.2.23 .. Prove that a graph dual to an n vertex (n > 3)
trtangulat1on 1s a 2 connected plane cub1c graph
4~2.24, (t) Prove that a plane cubic graph each of
whose faces has at least five \ ert1ces conta1ns not less than
twelve vert1ces
(2) Prove that 1f r 1 1s the number of faces of a plane
c.ub'c graph~ whteh are bounded by ' edges, then
~ (6 - l) ,., = 12
4.2. PLANARIT Y AND CONNECTIV ITY 115
4.2.25. Determin e the chromati c numbers of the

graphs shown in Figs 4-8.
4.2.26. Determin e the edge chromati c numbers of the
graphs shown in Figs 3-7.
4.2.27. Determin e the chromati c and edge chromati c
number of
(1) the graph Kn;
(2) the graph Kn,n;
11
(3) the graph B •
4.2.28. What is the number of the .correct vertex col-
11
ourings of the cube B for the minimum number of
colours?
4.2.29. Prove that the edge chromati c number of the
Petersen graph shown in Fig. 7a is equal to four, but for
any of its subgraph s G with eight vertices x' (G)~ 3.
Is the inequalit y x' (G)~ 3 valid for an arbitrary proper
subgraph of the Petersen graph obtained as a result of
the removal of a vertex?
4.2.30. Prove that four colours are sufficien t for the
correct edge colouring of any cubic multigra ph.
4.2.31. Prove that all edges of a plane cubic graph
can be coloured in two colours a and b so that each vertex
is incident with an edge of colour a and to two edges of
colour b.
4.2.32. Prove that it is enough to have six colours
for a correct colouring of the vertices of any plane graph.
4.2.33. Prove by induction over the number of vertices
that the inequalit y x (G)~ 5 is valid for a plane graph G.
4.2.34. Construc t a plane graph G with the minimum
possible number of vertices and such that x (G) = 4.
4.2.35. The vertices of a graph G are numbere d in
the order of increasin g degrees. Prove that if k is the high-
est number for which k ~ d (v 11 ) + 1, then x (G)~ k.
4.2.36. The operation of contracti on consists in the
removal of two adjacent vertices from a graph and ad-
dition of a new vertex adjacent to those of the remainin g
vertices which were adjacent to at least one of the remo-
ved vertices. Prove that the graph obtained as a result of
contracti on to a planar graph is planar.
. 4.2.37. Let l be the length of the longest simple chain
Ill a graph G. Prove that X (G)~ l + 1.
4.2.38. Let d be the highest of the degrees of vertices
of a graph G. Prove that
li6
(1) X (G)~ d 1~ +
(2) X (G) ~ d + 1
4 2 39 Let p be the maximum number for '\\h1cb there
ex1st.s a subgraph of a graph G tsomorphlc to the graph
Kp Prove that X (G)> p
4. 2 4.0 Prove that for a graph G with n \ erttces the
follow1ng Inequalities are val1d
(1) ~ 0 (G) X (G)~ nt
(2} x (G) 1 (G)# n
4.2 41 \Vhat 1s the smallest number n for whJch there
ex1sts an n vertex non planar graph w1th a non planar
complement?
4 2 4.2. Determine the thickness ol the graph K,
4 2 43 Prove that
a. 0 (G) 1 Po (G) -= a 1 (G) + ~1 (G) = n
for an arbitrary connected graph G w1th n (n > 1) \Cftl
c.es
4 2.44. (1} Gtve an example c.ontradtttlng the follow
1ng .statement a.ny vertex co\ er1Ilg conta1ns the minimum
vertex covertng
(2) Prove that any vertex cover1ng containS a tenntnal
vertex cover1ng
4..2.45 Determine the number of tern11nal and mint
mum edge cover1ngs Cor
(1) a cha1n of length m.
(2) a cycle of length n,
(3) Petersen s graph (Ftg 7a)
4~2 46 Prove that for any graph G the tnequallt,es
a, (G)> P1 (G} and et 1 {G)> ~ 1 (G) are valid
4.2.47 Prove that. for any graph G the 1nequal1t}T
a 00 (G)~ a. 1 (G) ts valid
4~2 . 48 Prove or refute the follov.-1ng tnequallty
p., (G)~ aoo (G)
4 2 .. 49. Let U ~ V be a subset ol vertiCes of a graph
G= (V, X) and let v (U) be a number of \ettlces v E v,u
wh1ch are not adJactnt to any vertex 1n U Let
- 1
k v=:v lU f--.\
(1) Prove that a.,G (G)~ k + -;A (G).

4.3. DIRECTED GRAPHS if 7
(2) Let d 0 be the lowest of the degrees of vertices of

a graph G. Prove that
k-1
~dG) ~lVI II (1- IV~o-i) ·
i=O
(3) Prove that the following inequality is valid for

a regular graph G of degree d with n vertices:
~ ~ ct 00 (G) ~ 1+ -J- (1 + l n d).

4.2.50. Prove that if d 0 is the lowest of the degrees of
the vertices of a graph G, then a 0 (G);::;:.: d 0 •
4.2.51 *. If G is a bipartite graph and m is the number
of its edges, then m~ a 0 (G) ·~ 0 (G). Prove that the equal-
ity is attained only for complete bipartite graphs.
4.3. Directed Graphs

A dzrected pseudograph D = D (V, X) is defmed by
specifying a non-empty (finite) set V and tuple X of ordered
pairs of elements in V. The elements of the set V are called
vertices, and the elements of the tuple X are called arcs
(or dzrected edges) of the directed pseudograph D (V, X).
The tuple X may also contain pairs of the form (v, v)
referred to as loops, and identical pairs lmown as mul-
tiple (or parallel) arcs. Pairs (u, v) and (v, u) are assumed
to be identical if and only if u = v. A directed multigraph
is the term applied to a directed pseudograph containing
no loops. If a directed pseudograph has neither loops nor
multiple arcs, it is called a directed graph (or simply
digraph). An oriented graph is a digraph having no sym-
metric pairs of directed edges, i.e. the tuple X cannot
contain simultaneously an arc (u, v) and an oppositely
directed arc (v, u).
Let x = (u, v) be an arc of a directed pseudograph.
Then the vertex u is called the initial vertex (origin),
and the vertex v is the terminal vertex (end) of the arc x.
In this case, the arc x is also said to emerge from the ver-
tex u and terminate at the vertex v. If a vertex v is the
origin or the end of an arc x, v and x are said to be incident.
The out-degree of a vertex v (of a pseudograph D) is the
number of arcs of the pseudograph D emerging from the
f18 CH ' GB.APHS .AND ~£TWOJtKS
vertex v The out-degree of the vertex v 15 denoted by od(v)

or d+ (v) Similarly, the in-degree of a verte:c v (denoted
by td (v) or d- (v)) IS the number of arers of the pseudo
graph term1nBt1ng at the vertex v
By replac1ng each ordered patr (u, v) 1n the tuple X of
a dtrected pseudograph D (V, X) by a dis ordered pa1r
{u, v} cons1St1ng of tlte same elements u and v, we obtatn
a pseudograph G -=; ( V XII) a..sooct.ated wtth D (V X)
D1teeted pseudograpl\S D 1 (V1 , X 1) and D 1 (V ~· X J)
are assumed to be tsomotphic tf there extst two one to~one
correspondence ~ V1 .., l'" V2 and 11' X1 .. ,. X~ such that
for any arc z == (k~ v) E X 1 the relation "'(r) ~ {r.p (u)t
<p (v}) IS val1d The tsomorphu~~ mapping of a directed pseu-
dograph onto tlself IS referred to as the autom.orph1sm oi
the pseudagraph All automorph1sms of a d1rected pseudo
graph form a group relat1vo to the operat1oi1 of multlplJ-
catioJl (successive application) of automorph1sms Th1s
group 1s referred to as a group (oj automorphlsms) of the
dlrected pseudograph
The operatton of omlSSlDn of a vertez and an arc. as well
a.s the concepts P/ subgraph, ~pannlng subgraph and gene·
rated subgraph. fa-t d1reeted pseudograpns are defined as
1n the ease of non~d1reeted pseudographs
In contrast to the definttlons of the correspondtng
"non..d1rected'' concepts~ the defintttons of directed walk~
closed walk . . chaln._ cycle~ simple chain and s1mple cycle eon ..
ta1n the requirement that the sequence (of vert1ce:s and
a res) v1 t x 1 • v 1 t z 2 , • r" _2 , vn- 1 ., :c 11 _ 1 , vn. ( n ~ 2)
sat1sfy the cond1tton that each aro z 1 (1 ~ t ~ n -- 1) has
the form (vi, v, .... 1)t 1 e the vertex v1 1s the or1g1n of the
arc x, and the vertex u1+1 IS zts end It 1s assumed that
a dtrected (u-v)-walk 1s directed front 1ts Jnlt1al ver·
tex u to the termtnal 'ertex v The length oj a walk IS equal
to the number ol arcs conta1ned 1n 1t The dlstance p (llt v)
from the vertex u to the verte;:z v 1s defined as the shortest
length of the {u·v)~walk A d1reeted walk 1s oft.en re-
ferred to as JUSt a pathf wh1le a dlreeted s1mple cycle is
known as a clrcu't
A d1rected simple spann1ng chaxn 1s called a Ham,l·
tontan path (Ham•ltonlan chain) A Hamtltonlan &lrtult
JS the spanntng cxrcuit of a d1rected pseudographjl If a
directed pseudograph conta1ns a Ham1lton1an c1rcu1t, 1t
ts called a Hamllton1an pseudograph
• 4.3. DIRECTED GRAPHS 119
A veL·tex v of a directed pseudograph is said to be ac-

cessible from a vertex u if there exists a (u-v)-path in the
pseudograph, i.e. the path emerging from the vertex u
and terminating at the vertex v.
A directed pseudograph is strongly connected (or strong)
if any vertex in it is accessible from its any other vertex.
A directed pseudograph is called unilaterally connected
(or unilateral) if for any two of its vm·tices at least one
is accessible from the other. A directed pseudograph
D (V, X) is assumed to be weakly connected (or weak) if
a pseudograph (V, X 0 ) associated with it is connected.
If a directed pseudograph is not connected even weakly,
it is referred to as disconnected. A trivial digraph consist-
ing of only one vertex is assumed (by defmition) to be
strongly connected.
A strong component of a digraph D is the term applied
to any of its directed subgraphs which is a strong di-
rected graph contained in no other strongly connected
directed subgraph of the digraph D. Similarly, a unilat-
eral component is the maximum unilateral subgraph of
the digmph D, while a weak component is the maxi-
mum weak subgraph. The concepts of strong, unilateral
and weak components are naturally generalized to the
case of a directed pseudograph.
Let y = {S 1 , S 2 , •. • ., S n} be a set of all strong com-
ponents of a digraph D. The condensation D* of the di-
graph D is a digraph in which the set of ve1tices is '\' and
the arc (S 1, S 1) is contained in the digraph D* if and only
if the digraph D contains at least one arc emerging from
a vertex of the component S 1 and terminating at a vertex
of the component j·s
If D = D (V, X) is a digraph, the digraph D' inverse
to it is specified by the same set of vertices V and a set
of arcs X' such that an arc (u, v) belongs to X' if and only
if the arc (L', u) belongs to X.
A vertex v of a digraph D is called a source if any other
vertex of the digraph D is accessible from it. The sink
of a digraph D is any of its vertices v which is a source
in the digraph D' inverse to digmph D.
Let D he a digraph for which a digraph associated with
it is a tree. Then the digraph G is called au arborescence
(or a growing tree) if it has a source. A directed pseudo-
CB 4 GRAPHS .AND NETWORKS
graph 1s assumed to be complete 1f .any two dtfferent vertL

ces tn 1t are JOined by at least one arc
A complete orten ted graph IS kno" n a.s a tournament
Let D be ann vertex d1graph and let V = {v1 , 1..,, .,
v n} he tl\e set of 1ts vert tees The adJacency matrtx
of a d1graph D IS an (n X n) matriX A (D) ~ fl a,, U1n
whJch a, 1 -= 1 1f the arc (v,., vJ) belongs to D and a~J = 0
1n tl1e oppos1 te case b-1 ore over, "e assu111e that tl1e set
of all arcs of the dlgra.ph D ts also ordered X :=:
(r1 ~ x 2 , , z 111 ) The tnctdence matrtx (or matrt.r
oj1ncufence) of a dJgraph D 1s an (n X m) matrix B (D) =
= n b11 UIn WhiCh
1 1f the vertex vl JS the end of t1le arc ZJ,

b11 = -1 1f the vertex v1 ts tl1e or1g1n of the ar~ x 1,
0 1f the vertex vi IS not 111C1dent w1th the are ZJ
4.3.1. D1sprove the follow1ng statement tf out and

In-degrees of Bny vertex of a dJgrapJt are postll\ e and even,
there ex1sts lor each verte1: ol the d1gr~ph a CirCuit
eon\.a1n1ng ,t
4.,3A2~ Let a d1graph D (Vt X) be at least v.eakly con
nected, V = {v 1, v2 ., , vn.}t n> 2 and d• (v1 ) -
d- (v1) = t, d+- (v1 ) d~ (v 2) ~ . - t, d~ (v1) ~
d- {v1) for J -= 3, • 1 n Pruve that the d1graph D
also conta1ns tn lhts case a dJrected (v1 - vt) cl1a1n
1ncludtng .all the arcs of the d1graph
4.3 3. Prove that a d1grapb JS strongly connected 1f
and only 1f 1t containS a d1reeted spanning closed \\alk
4~3 .. 4.. Prove that a \\eak digraph IS strongly connected
1f and only If It has a dtrected closed walk containing
eath are of the d1graph at least once
4~~5 Let us suppose that a d1graph D Gan be presented
•n the form of a un1on of 1ts certain directed closed walks
Dt, D 1 t , D,. (k ~ 1) wh1ch sat1sfy the follo\\tng
cond1t1on every two walk" D1 and D1 +1 (l ~ 1 ::s; k - 1)
have at least one vertex 1n common Prove that. the
d1graph D 1s strongly connected 10 thts case
4.3~6 .. Prove that any tournament contains a Ha..
mtltonlan path
4~3~ 7. Prove that a tournament T 15 a strong d1graph
tf and only tf T has a spann1ng ctrcutt (t e Js a Ham1lto·
n1an tournament)
U. DIRECTED GRAPHS 121
4.3.8. Let {v 1, v2 • . • . , vn} be a set of vertices of a

n
"
~ (d+ (vi)) 2 = ~ (n-1-
tournament. Prove that
-
i=l i=t
d+ (vi))2.
4.3.9. Let the out-degree of the vertex v of a tourna-
ment T be not lower than the out-degree of each other
vertex of the tournament. Prove that the distance from
t11e vertex v to any other vertex of the tournament does
not exceed 2.
4.3.10*. We denote by Sa set of arcs of a toumamentT.
The m·cs of the set S are called concordant if the vertices
of the tournament T can be renumbered so that the fact
that the arc (vi, vi) belongs to the setS implies that i < j.
Let f (n) be the largest integer such that each n-vertex
tournament (n ~ 3) contains the set S consisting of I (n)
concordant arcs. Prove that I (n) ~ [ • -i J·[ nt 1
4.3.11 *. Prove that the number of directed cycles

of length 3 in an n-vertex tournament does not exceed
11(11~-i)
24
if n is. odd,
t (n) =
It (rt2 -4)
24
if n is even.
4.3.12. Prove that the group of automorphisms of
any tournament has an odd order (i.e. consists of an
odd number of elements).
4.3.13. Prove that the condensation D* of an arbitrary
digraph has no cit·cuits.
4.3.14. Prove that a digraph D is unilateral if and
only if its condensation D* has a unique oriented spanning
chain.
4.3.15. A digraph D (V, X) is called transitiL·e if the
fact that arcs (u, v) and (v, w) belong to the set X implies
that the set X contains the arc (u, w). Prove that the con-
densation of any tournament is a transitive tourna-
ment.
4.3.16. A circuitless digraph is a digraph containing
no circuits. Prove that a circuitless digraph contains a
vertex with zero out-degree.
4.3.17. Prove that a digraph is isomorphic to its
condensation if and only if it is circuitless.
4.3.18. Let a digraph D b~ we{l.kly connected but not
122 CB ' GRAPHS A'l> 'E1WORKS
untlaterat Prove that there ts no vertex tn D such that

•ts om1ss1on leads to a .strong dtgraph
4 .3. f 9 Prove that a weak d1grapl1 1s an ar borescence
1I and only 1f only one of 1t.s verl1ees has zero tn degree,
the 1n degree of any of the rema1n1ng vert1ces he1ng equal
to un1ty
4~3.20. Let S" he a 5ymmetr1e g1oup of permutations
operatJng on the set (1. 2,. t >
n ), n 2 Let us con
s1der an arbitrary aggregate T of transpositions tn the
group S n We pu\ In correspondence to the set T a dtgraph
D (V, XT} 1n wh1eh V ~ {1, 2, , n) and the are
(L, 1} belongs to X r 1f and on I y 1f t < J and tJte tran~
posll1on (1/) 1s conta1ned Jn the set T Prove that the set
T forms a bas1s 1n S n (or, 1n other words., the set T IS an
trredutlble system of generators of the group S .,) Jf and
only 1f the d1graph D (V, X 7 ) 1s an arbores~enee
4 .. 3 .. 21. Prove that a complete strongly connected dt~
graph ts a Ham1ltontan dtgraph
4.3.22. Prove that a complete d1graph contains a
source
4~3.23. Ver1fy that any trans1t1ve tournament l1.as a
stngle Hamiltontan path.
4.3.24 . Prove that tbe number of aJI dtfferent 1-IamJI
ton1an paths 1n each tournament 1s odd
4.3.25. Let a complete strongly connected d1graph
have n vert1c.es (n ~ 3) PrO"Ve that ior any I. (3 ~ k ~ n)
there extsts for any vertex of the digraph a CltCUll of
length k containing thts vertex
4.3.26. Let D (l'. X) he a complete strongly connected
dJgraph for WhiCh 1 v r~ 4 Prove that the dJgraph D
conta1n.s two d1Herent vertices v1 and v1 sat 1sf ytng the
f-ollowing c-ond1lton dlgraphs D 1 and D 1- ohta1ned !rum
the dtgraph D as a result of removal of the vertJces v1 and
v 2 (respectively) are strongly connected
4.3 .. 27. \Ve shall denote by A' the q-th degree of the
adJacency matriX A (D) := uDtJ nof a dtgraph D Pro\ e
\hat the (f, 1)·\h element tJ\~ of the matr1x A q IS equal
1
lo the nu.mher of a\1 (Vt•V J)-.walks of length q (1n the

dtgraph D)
4~3.28. Let B be the tnc1dence matrtx of a dJgraph
D (V, X) Pro\e that the subset X 1 of arcs of the digraph
D(X 1 s X) generates a stmple C}Cle (not necessarily
or1ented) t[ and only 1f the set of columns correspond1ng
U. TREES AND BIPOLAR NETWORI).F 123
to these arcs (in matrix B) is linearly dependen t, and

each of its proper subsets does not possess this property.
4.3.29. Prove that the determin ant of any square sub-
matrix of the incidence matrix B(D) of a digraph D is
equal to 0, or +1, or -1.
4.3.30. Let B be the incidence matrix of a weakly con-
nected n-vertex digraph D, and the matrix lJ be obtained
from B by deletion of any (one) row. Prove that there
1
exists a one-to-on e correspon dence between different
directed trees of the digraph D (which are treated as di-
rected subgraphs of the digmph D) and non-degenerate
(n - i)-order submatt·ices of the matrix B.
-
4.3.31. Let a matrix Bbe a submatri x of the incidence
matrix B of a weakly connected digraph D, described in
-
the previous problem. By B' we shall denote a matrix
transposed relative to the matrix B. Prove that the
number of different directed trees in the digraph D is
equal to the value of the determin ant of the matrix B .Jj'.
4.4. Trees and Bipolar Networks

A pseudograph G = (V, X) in which k vertices known
as poles are isolated is termed a k-pole network. The pseu-
dograph G will be called a graph of the correspon ding
k-pole network. A network r with a set of poles P and a
graph G = (V, X) will be denoted by (P; V, X). Two
k-pole networks are isomorphic if their graphs are isomor-
phic and there is one-to-one correspondence between
their poles. A one-pole network whose graph is a tree
is called a rooted tree. The only pole of such a network
is called a root. A plane rooted tree is the map of a graph
on a plane. This concept can be defined by induction
as follows. The network shown in Fig. 10a is a plane
rooted tree. If A and B (see Fig. 10b) are plane rooted
trees, the figures C, D and E (Figs 10c and d) are also
plane rooted trees. We shall assume that an arbitrary plane
~·ooted tree is mapped onto a plane with a cut which
1s a ray emerging from the root (see Fig. 10e). Here we can
assume that the edges incident with the root are numbered
Here w~ assume that two trees are different if they are non-
1
.
Isomorphic d1rected trees with labelled (numbered) vertices.
124 CH ' GRAPHS Al\D l\ETWOR:h..S
clockwtse b} 1 m where m ts the degree of the

root If "e remo\ e the edge w1th number 1 from such
a pi a ne rooted t 1 ee v. e sha II obta1n a graph \\ 11 h two
connected components The component \\h1eh does not
con ta1 n a 1oot w1ll be cal 1ed the l th branch of the or1g1 nal
tooted tree Tl1e vertex tnctd~nt w1th the l th edge (tn
the orJg1nal tree) w1JI be treated as the root of the ' th
C1 0 b
c
(a) (b) (c)
branch For an arbttrary rooted tree A \\e denote by

d 0 (.~t) the degree of 1 ts root Plane rooted trees A and B
ate called tdenttcal 1E e\ther da (A} -=- d 0 (B) ~ 0 or
d n (A) - d 0 (B) -= m > 0 and Cor any t ~ 1 m the
l th branches of the trees A and B are 1dentlcal Tv..o
trees wh1ch are not 1dent1cal are referred to as different
TJ1u.s trees D and E are dttlerent (see Ftg 10d) 1f the
trees A and B (see F1g 10b) are different Each plane
rooted tree T With m edges can be put 10 one to one
correspondence wttll a h1nary vector of length 2m known
as a code of the lree A tree w1th. a stngle edge corresponds
....... 1'-J
to the vector 01 If vectors rz and ~ are put In correspon

dence w1th trees A and B (Ftg iOb) respeCtively then
~
tJ1e 'ector Oat put 1n correspon de nee WIth the tree

lS
~~ .At!.._.
C lFlg 10c) and vectors a~ and r>a are put. tn corre

spondence With the trees D and E (F1g tOd)
Hencefort\1 unless st1p11lated otherwlsr the term
netYtork shaiJ tnean a two pole network The network
r {{a b l l X) Will be briefly denoted by r (a b} The
subgraph of such. a net\\ork 1s the subgraph of the grapJ1
(V X) A vertex of a non lrtvtal ~uhgraph 6 of the net
WOrh f IS Cal)ed boundary tertex If tt IS either a pole
or Jne1dent wtt1t a certa1n edge of the network wh1ch does
not belong to the subgrapll G 'Ve shall call a non tr1V1al
subgraph of a network a shoot 1f 1t has a untaue boundary
4.4. TREES AND BIPOLAR NETWORKS 125
vertex. A subnetwork of a two-pole network is its sub-

graph having exactly two boundary vertices. These ver-
tices are the poles of the subnetwork. A network is connected
if its graph is connected. A connected network (or sub-
network) having a single edge is called trivial. A con-
nected network is called strongly connected if each edge
lies in a simple chain connecting the poles of the net-
work. A strongly connected network is known to be
decomposable if it has at least one non-trivial subnetwork.
Otherwise, it is assumed to be non-decomposable. Let
r (a, b) be a decomposable network, G (c, d) its non-tri-
vial subnetwork, and f 1 (a, b) be a network obtained
from r (a, b) by substituting an edge (c, d) for the sub-
network G (c, d). Then the network r (a, b), in turn, can
be obtained by substituting the network G (c, d) for the
edge (c, d) of the network r 1 (a, b). Thus, the decompos-
able network r (a, b) can be defined by specifying the
network rl (a, b), edge (c, d) of the network rl (a, b),
and the network G (c, d). Such a definition is referred to
as the decomposition of the network r (a, b). A network
f 1 (a, b) is called external and the network G (c, d) in-
ternal network of the decomposition. A network r (a, b)
is called the superposition of the networks r 1 (a, b) and
G (c, d). A network consisting of m parallel edges joining
the poles a and b is denoted by r~:t (a, b) or simply r~:t.
A network whose graph is a simple chain of length m and
which joins the poles a and b is denoted by r:n (a, b) or
simply r:n. A network that can be obtained from networks
r~ and r~ by applying a fmite number of substitutions
of a network for an edge is referred to as a series-parallel
network, or a rt-network. A non-trivial non-decomposable
network r (a, b) differing from r~ (a, b) and n (a, b) is
called an H-network. -
A decomposable network is called p-decomposable
(s-decomposable) if a certain external network of the
decomposition has the form rr,l (resp. r~l), m ~ 2. If
a certain external network of the decomposition of a net-
work r is an H-network, then r is referred to as H-decom-
posable. Any decomposable network is either p-, or s-,
or H-decomposable. A canonical p-decomposition of a
network is a p-decomposition for which the internal net-
works of the decomposition are other than networks of
the type r~ and other than p-decomposable networks.
126 CIJ ' GRA.PHS .A.!'rr.D NETWOJlKS
'fhe tanonfcal s decomposition 19 defined 1n a s1m1lar way

The canon!ca.l H*"ckcomposit,on 1s a decampositt~n whose
external netv,ork JS an H net\\ork Each n network r
"1 th m ~ 1 edges can be put 1n correspondence Wl th a
plane rooted tree T (f) w1tlt m pendant vert1ces, such
tbat (a) each vertex of the tree T (P) otl1er than a pendant
vertex Js labelled by a symbol p or s, (b) the labels p
and s alternate on eath chatn emerglng !rom the r-oot to
a pendant vertex, and (c) 'e rtrces other than the root ha ~e
a degree other than two Pendant ' erttces of the tree
T (r) are not labelled The tree T (f) 1s defined by tnduc-
tlon J r r has the form r~ (or r~) then T (r) lS a tree
r(r)
r
a
(a)
whose root IS labelled by the symbol p (resp s) wlti1e tlte

rema1n1ng tn vert1ces are pendant adJacent to the root
a.nd are nQt labelled lf a ltetwork r lS decomposable and
other tltan the networks of the t)pe descrthed abov(!, and
lhe exterJial network of the decompos1t1on has the form
ff (or fl) f wh1Ie tlte Internal net works are Just G1 G, ,
G" the tree T (r) 1s constructed as follows Let
T {G1), T (G 2) T (GJt) b~ the trees correspondtng
to \be 1nternal networks of tl1e decomposition !hen lor
tl1.e roct of T {f) we take a vertex of degree k labelled
by the symbol p (rasp s) The verttces adJacent to the
root are labelled by the .symbol s (resp p) The vertices
v 1 • v1 vJt adJacent to the root are 1dent1fied w1th
the roots of the trees T (G1). T (G,), T (G 1,.) For
example the n network ~hown Jn F1g 11a corresponds
to the tree sh()wn ln F1g 11 b A tre-e T (f) 1s called a
dlagram of the canontcal de.com.pas4tUJn of a n network r
It should be noted that 1f the externaJ network of the
decomposition of a network r has the form n (a, b),

and the internal networks G1 {a, u 1), G2 {u1 , U2), • • .,
G11. (u~~._ 1 , b) are substituted for the edges (a, u1 ),
(u 1 , u 2), • • • , (u~~._ 1 , b), respectively, the . ver~ices
v1 , v2 , • • • , Vn in the tree T (f), which are 1dent1fied
with the roots of plane rooted trees T (G1), T (G2), ... ,
T (G~~.), follow one another from left to right in
r(r')
r'
a
(a)
Fig. 12
the increasing order of numbers. Thus, the tree T (f')

shown in Fig. 12b is a diagram of the canonical decom-
position of the network f' (Fig. 12a) but is not a dia-
gram of the network r (see Fig. 11a).
A vertex of a network other than a pole is referred to
as internal. A vertex v depends on a vertex u if any simple
chain joining the poles and passing through v passes
through u as well. Vertices v and u are equivalent if
v depends on u and u depends on v. A vertex v is weaker
than a vertex u, while the vertex u is stronger than the
vertex v if v depends on u hut is not equivalent to it.
A vertex v is called minimal if it is not weaker than any
other internal vertex of a network. Henceforth, we shall
call a simple chain joining the poles of a network a
chain of the network. The cases when the term "chain"
is used in different sense will always be stipulated. A
chain of a network will be assumed to be the shortest if
it has the minimum possible length. The length of a
network is equal to the length of its shortest chain. A
cut is a set of the edges of a network whose deletion de-
stroys all the chains. A cut is called terminal if none of
its subsets is a cut. A cut is referred to as minimal if it
C:::H l GRAPH"~ AND NCT\VORJ....S
has tl1e mtn1mum poss1ble number of edges fl1e number

or edg-es ln the m1n1mal cut 1s called the wtdth of a net
Y.ork
4 4 1 Let G be a graph Wtth n ~ 2 verttcec; Prove

the equivalence o( the followlng statements
(1} G JS a connected grapfl w1tft n - 1 edges
(2) G JS tl connected grapl1 wh1ch however becomes
dt~connected as a result ol delet1on of any of 1ts edge.s
(3) any paLr o( dtHerent vert1ces of the graplt G ts Jou:ted
by a un1que chain
(4) G JS a graph Without cycJes but the addition of an
edge JDlnlng any t\Vo verttces leads to tlte emergence of a
c~cle
4 -4 2 Prove that any tree wt th n ~ }, \ ert 1ces has
at least two pendant vert1ces
4 4 3 Prove that •f the number of pendant vert1c~s
~n a non trLvtal graph G lS equal to the ntlmber of edges
G lS e1ther disconnected or IS a tree
4. 4 4 Let F (G)= {H 1 H 2 Hnl be a famtly
of graphs tn whtth the graph H 1s ohlttln~d from an
n vertex graph G by delettng a vertex wtth the number
' (i == 1 n) The vertices In the graphs Rt are not label
led Prove that
(1) 1t can be found out Irom tl•c fam1ly F (G) wl1etl1er
the graph G 1S a tree,
(2) tf G 1s a tree 1t can he un1q_uely reconstructed from
F (G) (accurate to an lsomorphlsru)
4 4 5 The lntersectlCn o} two graphs G and 1/ts a grapl1
n
G H .all whose vert1ees and edges oolong tG bnth G and
H Prole that a non empty lnlerseetton of two subtrees
of a tree ts a tree
Let Po (v u) be the dastance between vert1ces v and
u tn a grapll G = (l' Y) Th.e vertex u. 0 for v.hlt.l\
max 1lc {u 0 v) ~ mln max PG {u v)
,E V uEU ~EV
IScalJed the centre of the graph and the number R (G) ~

ma-x Pa (u 0 v) IS called the rad1us oj the gro.ph
't'~ v
4 4 6 (1) Delertn1ne the Jlumber of the centres tn
the graph
(a) G (see Ftg 6a)
(b) G (see F1g bb)
•
(c) G = Kn 1, n~·
(2) Prove that any tree contains not more than two
centres.
4.4. 7. Prove that a tree has one centre if its diameter
is an even number and two centres when the diameter is
odd.
4.4.8. (1) Let D (G) be the diameter and R (G) the
radius of a gt·aph G. Prove that R (G)~ D (G)~ 2R (G).
(2) Prove that if G is a tree, then R (G) = J D ~G) [ •
(3) Give example of a graph G for which R (G) = D (G).
4.4.9*. Prove that a tree can be uniquely reconstructed
(correct to an isomorphism) if pairwise distances between
its pendant vertices are specified.
4.4.10. Prove that any two simple chains of maximum
length in a tree with an odd diameter have at least one
edge in common.
4.4.11. An infinite tree is the term applied to a graph
with a countable set of vertices satisfying the following
condition: for any two vertices u and v of the graph, there
exists a unique simple (u-v)-chain with a finite length.
Prove that if the degree of each vertex of an infinite tree
is fmite, there exists for any vertex a simple chain of
infmite length, which contains this vertex.
4.4.12. Let P = {v1 , v 2 , • • • } be an infinite simple
chain. Let G = K 2 X P. What is the power of the set of
all spanning trees of the graph G?
Let G = (V, X) be a multigraph with a set of vertices
V = {v11 V 2 ••• , v 11 }, and M (G) = II a 11 II be a quadratic
matrix of order n, where
for i = j;
for i =I= j and (vh v1) EX;
for i =I= j and (vh v1) ~X.
It is well known [23] that the number of pairwise different
spanning trees of the graph is equal to the minor of any
element of the principal diagonal of the matrix M (G).
4.4.13. Determine the number of the spanning trees
of the graphs shown in Fig. 4a, b and Sa, b, if the vertices
of the graphs are labelled.
4.4.14. What is the chromatic number of a tree with
n ~ 2 vertices?
9-0636
130 CU .l GRAPHS A~D NETWORkS
the d1ameter ot a graph G IS

4.4~15-t Is 1t true that 1f
h (/. > 2), there C"<lsts a spann1ng tree whose d1ameter
1s equal to k'
4.4.t6jl Prove that for n~ 3, the number of pa1rw1se
non·-Isomorph1c rooted trees w1th n vertiCes ts at least
double the number of pairWISe non •somorphtc trees
wtth n vertices hav1ng no roots
4~4. 17. Let a rooted tree wtth n (n >
2) pendant ver-
t tees have no vert1ces ol power 2 other tl1an a root Prove
(a) (b)
Ftg 13
that tl)e total number oi vert 1tes of \he tree does not
exceed 2n- 1
4 .4,. f8. Construe t the codes of plane rooted trees shown
tn F1g 13a and b
~
4.4.19. Us1ng a gtven code a~ construct a plane rooted

tree 1f
~
(1) a,.,., == (00101001 1011 )1:

(2) a ~ (Ot0001100t101011),
~
{3) tt ~ (0001010110011011)
4~4 20. (1) Determtne the number of pa1rwtse non
tsomor-phtc rooted trees wtth four edges
(2) Determlne the number of pa1rw1se dtflerent plane
rooted trees w1th four edges
,..., ,.., ~ ~
4.4.21. D1v1de the set of \ ectors {a 1 , a 2 , a at a,.

,...,
rx,} 1nto -e.qu1.valenc.e elasses so that the \-ectors of one
~
class are the codes of patrw1se tsomorph1e trees a 1 ==

U. TREES AND BIPOLAR NETWORKS 131
"' ,..,
(0010101101),
,...
a 2 = (0100101101),
,...
ct 3 = (0101001011),
a 4 = (0100101011), a 5 = (0010110101).
,...
4.4.22. Prove that the code a = (ct 1 , ct 2 , • • • , ct 2 n)
of a plane rooted tree with n edges possesses the follow-
ing properties:
2n
(1) ~ ct; = n;
i=i
k
(2) for any k (1::;;:; k::;;:; 2n), the inequality ~
i=1
k/2 is valid. ,..,
4.4.23. Prove that any binary vector a of length 2n,
which satisfies conditions (1) and (2) of the previous
problem, is a code of a plane rooted tree with n
edges.
4.4.24. (1) Prove that the following recurrence
,.., rela-
tion is valid for the number tjJ (n) of vectors a EB 211 that
satisfy conditions (1) and (2) of Problem 4.4.22:
11
'ljl(n)=~ tjJ(i-1)tjJ(n-i), where tjJ(O)=tjJ(1)=1.

i=1
(2*) Derive an analytic expression for tjJ (n).

4.4.25. Let G be a connected plane graph. Starting
from a certain vertex v belonging to an external face, we
shall go around (circumvent) this boundary so that it
always remains on the right. At a certain moment of
time, we shall return to the initial vertex v. The circum-
vention will be continued if there remain uncircumvent-
ed edges incident with the vertex v and belonging to
the boundary of the external face. Otherwise, the circum-
vention is terminated.
(1) Prove that G is a tree. if and only if each edge is
circumvented twice in the above process.
(2) Let G be a tree with a root v. In accordance with
the circumvention procedures described above, the tree
G can be put in correspondence with a binary vector. By
circumventing edges consecutively, we shall write 0 if
an edge is circumvented for the first time and 1 if it is cir-
cumvented for the second tim~. Prove that the vector
obtained in such a way is a ·code ._of the tree G.
9*
132 CB ' GRA.Pl!S AND NETWORKS
4.~4126. Let G be a graph w1th n vertices and m edges

The graph 1s called balanced Jf ne1ther of tts subgraphs
has verttces of power htgher than 2mln
(1.) Prove that for n > 2, a tree lS a non balanced
graph
{2) Glve. an example of a balanced graph for whtch
m=-n+3
4.4.27~ Let G he a graph With a real non negat1ve num-
ber ascr1bed to each edge as a \\eight The wetght of a
subgraph of the graph G 1s the sum of the we1ghts of the
edges of thts subgraph The mintmal spannlng tree oj the
graph G 1s 1ts spanntng tree hav1ng the mlntmum v,e1ght.
Prove that the minimal covertng tree can be ohta1ned by
u.stng the follow1ng algor1thm At the first stage, we
choose the edge w1th the m1n1mum v.e1ght 1hen In the
next stage we choose the edge that has the mtnlmum
we1ght among the edges which do not form a C} cle W!th
the edges chosen earl1er The algor1thm IS terminated when
there 1c; no sueh edge
4~4.28. Is It true that any subnetwork of a strongly
connected network ts strongly connected'
4.44129. Let a connected network have a untque shoot
w1th k verttces~ whtch 1s not contained 10 another shoot
wtth a larger number of vert1ce.s Prove that It IS suf,..
fieient to draw k - 1 add1tional edges to make the net~
work strongly connected Can we alwa)'s manage wtth the
m1n1mnm number of additional edges'
4~4~30 . Is 1t true that lor arty n and rn (m;a= n~ 3)
there exist decomposable networks wtth n vertices and
m edges?-
~-4~31. Is It true that the graph of a strongly connect--
ed network 1s 2-eonnect~d1
4.4.32. Is 1t true that 1f we ehoo.se two non adJacent
vert1c~s ln an atbltrary 2-eonneeted t.uh1e grapn as poles,
we shall obta1n a non decomposable network~
4.4.33 ~ Determtne the number of patrwtse non-tSo
morph1c non decomposable networks tltat c.a.n be obtained
by ehoostng two verttces 1n an n dimensional cube as
poles
4.4.34. (1) Prove that 1f a non decomposable network
has n > 2 verttces and m edges, then
3n~ 2m+ 2~ n (n 1)
U. TREES AND DIPOLAR NETWORKS i33
3
(2) Prove that for any nand m such that m ~ 2 n - 1,
n~ 4, there exists a non-decomposable network with n
vertices and m edges.
1
a b
(a) 5 (b) 6
a b
3
(c)
Fig. 14
4.4.35. For the networks shown in Fig. Ha, b andc

determine
(1) the type of decomposition;
(2) external networks of the canonical decomposition.
4.4.36. Let f be the net-
1 2 3
work shown in Fig. 15. ,..-----
(1) Indicate all minimal
vertices of the network. Q b
(2) Divide all the inter-
nal vertices of the network
r into the classes consist- 7 8 9
ing of pairwise equivalent Fig. i5
vertices.
(3) Verify whether there exists a vertex in this network
which is weaker than any other vertex.
4.4.37. Prove that for any vertex v of a strongly con-
nected network there exists a chain containing all the
vertices that are stronger than or equivalent to it.
4.4.38. Let S (f) be a set of all vertices v of the net-
work r which are not minimal.
(1) Is it true that if r is an H-decompoable network
without multiple edges, we obtain a non-decomposable
network by joining each vertex v E S (f) through edges
with each pole to which v is not adjacent?
t34 CH 4i GRAPHS .AND NEtWORKS
(2) For obta1n1ng a non decomposable network IS a\

suffic1ent to )Otn eath vertex v from S (f) Wlth exa.etly
nne pole from th()se whtc.h are not adJacent to u)
4.~4..39 .. (f) Prove that any separat1ng vertex IS mt
ntmal
(2) Prov6 that any vertex adJacent to both poles lS
m1n1mal
4.4~40 . Let all vertrces of a strongly connected net~
work r he mlnlmal
(1) F1nd out whether the network r ea.n he
(a) p-decomposable._
(b) s decomposable,
(c) H-decomposable
(2) Let the network r be s decomposable Prove that
each 1nternal verte1 IS a ~eparat1ng vertex
(3) Let r be H decomposable Ver1fy whether or not
any of Internal networks can be
(a) an H network,
(b) an H--decotnposable network,
(c) a network other than networks r~ (m~ 1)
4~~~41. Let a strongly connected network r w1th etght
edges be ne1ther p-. nor s-deeomposable and have no sub·
networks of the t:ype r~. r~ (m > 1) Pro.'le that r lS
a non decomposable network
4.4.42. Let G = (V1 ,. V 1 ,. X) be a connected btpart1te
graph the degree of each of whose verttces IS b1gher than
or equal to two Prove that 1f we construct .a. network
r (at b) by JOining the pole a by an edge with each vertex
of the set V1 and the pole b w1th eaeb vertex of the s~t V 1 ,
then r (a, b) lS an H-netwotk
4 4 .. 43. Does there exist a p decomposable network for
wh1ch any subnetwork contaJntng at least three edges JS
p decomposable?
4.4.44. Construct canontcal decompos1t1on diagrams
for the- networks shown In FJg 16a and b
4.4.45. Construct n. n~etwotks hav1ng \he eanGnl~al
desomposttton dragrams shown tn Ftg i.1a and b
4-4~46. Prove that lf a-networks r1 and r I are not
1Somorph1c, they have d1fferent canonical decomposition
dtagrams
4 . 4.47. Let A and B be two cha1ns of a network r {a, b)
and let a ve-rtex u belong to cha1n A and not belong to
chatn B l whtle a vertex u bel<lngs tG cha 1n B ~nd does nQt
U. TREES AND BIPOLAR NETWORKS 135
belong to A. Further, let [u, v] be a simple chain which

joins u and v and does not intersect the chains A and B
at any point other than its ends.
(1) Prove that there exist at least two chains of the
network r whose vertices belong to the union of the
chains A, B and [u, v] and which contain the chain [u, v].
Do at least three such chains always exist?
(2) Prove that if the chains A and B have internal
vertices in common but have no edges in common, there
b b
a
(a) (b)
Fig. 16
s (a)
Fig. 17
exist at least four chains of the network, which are con-

tained in the union of the chains A, B and [u, v] and
satisfy condition s (1).
4.4.48. Prove the equivalen ce of the following two
definitions of a n-networ k:
(1) A network r (a, b) is a n-networ k if its edges can
~e. ~riented so that all the edges in each simple chain
JOimng the poles a and b are directed from a to b.
(~) The n-networ ks are those and only those networks
WhiCh are obtained as a result of the following inductive
process:
(a) The networks r~ and ri (Fig. 18a) are · t.wc
- • '>
136 Cil ' GRAPBS .A'n NETWORKS
(b) If the networks A and B (Ftg 18b) are n networks

the net-works shown 1n F1g 1.Sc are also n n~tworks
4.4.. 49 Prove that among all :t networks wtth
m (m > 4) edges.. the s decomposable networks have
the largest number of shortest chains
ao • ob a
a b c d a-.c
(a) (b) (c)

F1g iS
4.4.50. (1) Indicate the type of n networks whtch

have the max1mum number of s1mple eha1ns JOlDing
the poles
(2} lnd1cate the type of n netv.orks having the maxim
um number of terminal cuts
4,4 51 Let fP (m) be the ma.x1mum number of cha1ns
of a n~network With m edgeS: Prove tl1a\
(1) '(f) ~ 1~
(2) rp (3n) ~ 3" (n~ 1),
(3) rp (3n + 1) = 4 X 3n (n ~ 1)
(4) tp (3n + 2) =
1
2 X 3n (n ~ 1)
4 .. 4.52~ Is 1t true that among all the networks With
m edges, the n networks have the maxtmnm number of
s1mple chains.,
4.4.53. F1lr e:aeh n >-
51 1nd1c.ate an H network With n
verttces,. whtth has the maxtmum number of chaJns
connecttng the poles For eath k (1 ~ k < n), calculate
the number ol cha1ns of length k between the poles
4 4 54~ Pro-ve that for a network w1th .m edges hav1ng
a length l and w1dth t the 1nequal1ty m ~ l X l IS
val1d
4.4~55. Prove that 1n a n network the 1ntersect1on
of any Simple eha1n between poles w1tb any tetmlnal
cut conta1n.s exactly one edge
4.4~56·. The set of chatns lD a network r
J5 called a
defin~ng .set lor a vertex v 1f {v) 1s the Intersection ol
the sets of the 1nterna.l vertlces of these eha1ns Pro-ve
or d1sprove the followJng statement the necessary snd
~.5. ESTIMATES 137
sufficient condition for a non-p-decomposable network

without multiple edges to be non-decomposable is that
there must exist a defining set of chains for any internal
vertex.
4.4.57. Prove the following statement: the necessary
and sufficient condition for making a decomposable
network r non-decomposable by adding edges is that
r must not have multiple edges and must have at least
four vertices.
4.4.58. Construct a decomposable network with the
smallest number of edges and n vertices (n ~ 4), which
cannot be made non-decomposable by successively re-
placing subnetworks of the type r~ and r~ by edges.
4.5. Estimates in the Theory of Graphs

and Networks
•
A graph (dig1·aph, pseudograph, etc.) is called labelled

(or numbered) if its vertices are assigned labels (or !).Urn-
hers). We shall denote by '§ 11 the set of all n-vertex graphs
(in short n-graphs) whose vertices are labelled by numbers
1, 2, ... , n. The subset of all graphs from '!J 11 , each
of which has exactly m edges, will be denoted by rJ n, m·
A graph with n vertices and m edges will be briefly called
an (n, m)-graph. Graphs G and H in '!J 11 are assumed to be
different if there exist two vertices j and k that are ad-
jacent in one graph but not in the other.
Let q> 11 (P) denote the number of all graphs in '!J 11
having the property P. It is said that almost all n-graphs
have the property P if lim f~ (P~ =1. Let m = m (n) be
n-+oo n
an integral non-negative function, and let cp 11 , m (P) be
' the. number of all graphs in 9' 11 , m having the property P.
It 1s said that almost all the (n, m (n))-graphs have the
property P if lim cpn, m (P) =1.
n-oo I :§u,m I
4.5.1. Prove that
{1) I'!Jni=2G); (2) l'!ln,ml= (;) .
m
4.5.2. (1) Find the number of different tournaments
with n vertices, labelled by 1, 2, .. , , n.
138 r::s .1. GRAPHS ~....."D NE11'WCillXS
(2} F1nd the number of ortcnted pseudographs w1th

n labelled verttee$ and m arcs
4.5 . 3.., {1) Show that the number of graphs tn Y7n
(11-JI)
whose k glven vertices are l!Olatedl ts equal to 2 2
(2) Sb.a:w that the nutn ber: of graphs w1thout 1sola ted
\ ert.l :e.es l n Y tt 1s eq-ual to
n
k 2
t;") •
(3) Sltow that almost a]l n graphs have no Isolated

vertices
4 5.4 Let the subset Y c Y li cons 1st of N pairwise
d1ffercnt graphs Show that the number of pairwise non·
ts.omorphu; graphs tn § lS not less than Nfnt
4.3 5.. Let ~ {m) he the number of palrwtse non 1so·
mcrph1e co.nnected graplts wtlh m. edges Show that
{t) *(m}~ ~ {; } •
-r' (t + n +im)~n<m+ l
- -
m '
(2) ~ (m) ~ (2m)m for m • )'- oo
4~5 6. Show that the num her of pa1rw1se non·lsomor-
phlc pseudographa that do not have Isolated vert1ces but
have m edges doe.s not exceed {em)"', where c 1s a constant
1ndependent of m
4.5.7 ~ SJtow that the number of patnVISe non tso~
morphte k pole n~t"' orks w1 th m edges without loops and
wtthout tsolated verttces does not e:(cced (2m)A (crn) 1na,
wllete c 1s a constant independent of rn and k
4.5~8. Prove that the number of palr\\lse non-lS(Y
morph1c trees Wl\h m edg~s does not exceed the number
of pa1rwtse d1fferent plane rooted trees 'W1th m
edges
4.5 9. (1) Prove that tlJe number of pa1rw1se d1fferent
plane rooted trees w1th rn. edges does n~t exeeed { : )
2
(2) Determ1ne the aS)Dlptottc behavtor of the number

q (m) of plane rooted trees with m edges for m • co
4.5~10~ Us1ng Cayley theorem\ according to \\-hiCh
the number of PaLrwtse dtfferent trees Wlth. n labelled
v~rttces lS equal to n~-•. prove that the number of pa.1r-
4.5. ESTI:\fATES 139
wise non-isomorphic trees with n vertices is not less

than Cnn-1.5en, where lim Cn =
n-..oo
V2rt.
4.5.11. Show that the number of pairwise different
trees with n labelled vertices, where the vertex with num-
ber 1 has a degree k, is equal to (~=;) (n - 1)n-h-1 •
4.5.12. Find the number of graphs in '§ n constituting
forests.
4.5.13*. Show that the number of forests in '9 11 , whose
given vertices j and k belong to different components,
is equal to 2nn -S.
4.5.14. Let <p (n) be the number of pairwise different
rooted trees with n pendant vertices, such that the degree
of a root is equal to 2, and the degree of each vertex other
than a root or a pendant vertex is equal to 3.
(1) Show that the number <p (n) is equal to the number
of ways in which brackets can be arranged in the expres-
sion b1 : b2 : • • • : bn so that the new expression obtained
in this way is meaningful.
(2) Show that <p (n) = ! (;_=-
2 2
1 ).
4.5.15. (1} Show that the number of pairwise non-
isomorphic bipolar rt-networks with m edges does not
exceed twice the number of pairwise different plane rooted
trees with m pendant vertices.
(2) Show that the number of pairwise non-isomorphic
Jt-networl<s with m edges does not exceed 2 ( ~:=~).
4.5.16. Find the number of pairwise non-isomorphic
networks r (a, b) with n vertices and m edges, having the
following properties:
(1} the network r (a, b) is s-decomposable,
(2) all the vertices in the network r (a, b) are minimal.
Remark. The poles a and b of the network r (a, b)
are not equal: the former is the entrance to the network
and the latter is the exit from the network. Upon an
isomorphic mapping of the network r to the network G,
the entrance (exit) of the network r must correspond to
the entrance (exit) of the network G.
4.5.17. Let CIJ (n, m) be the number of different formu-
las generated by the set of connectives {&, V} and the
set of variables {x1 , x 2 , • • • , xn} with m entries of the
connectivity symbols, .
(1) Show that 4> (nt m) 1s equal to the number of
pa1rv. tse dtfterent rooted trees t?'ach of whose pendant
vertices IS labe}led by a certain sywbol 1n the set
{x1 , :t 2 , z n. }, wh1le each non-pendant vcrle.x 1s
labelled by one of the symbols & or V
(2) Show that $ (n~ m) = m~ t ( ~ ) 2mn"'•t
Ren1ark Formulas are assumed to be dtflerent 11
they forn• different words tn the alphabet (&t V, (t),
z1 ~ z 1 , xn}
4.5 18. Let (J) (n, m k) be the num her of d1fierent tor
mulas generated by the set of c.onnect1ves {&, V, -} and
the set of v.artables {x t x 2 ~ , .r"} Wtth m entries
of the symbols of V<mab\es and k entrtes of the symbol~
Prove that
<lJ (n m. k) ~ 1
' ' -.. . .;: : rn
(2m--2}
m. - J
sna-tnm
4415.19 Show that the number of d1seonnected graph~

In Y n m does not e"'tceed
[n/~1 ( n)
~ ( : ) 2 -k (n - k)
A-=t m
4~5.20. Show that the number of graphs 1n Yin m.
havtng exactly two connectiVIty components does not
sx.eeed
(n/~l ( ~)
~ ( k- I)
~ ( ~ ) kk-~ (n-k)ft--'-z L} 2
l-t J~l-l J-k+t~ m-J-n+k+t
~ .5.21. Show that the number of k con nee ted graphs
+
tn '<J n m wh1cb are not (h. 1) connected (J.. ~ n - 2)
does not exceed
1
["2 1
( ~} +k (n-k) }l (n;k)
1 r-.=1
Y n-k}
( 2 -r (tt- k--r)
m.-1
U. ESTIMATES 141
Hint. For k~ n - 2, a graph that is not (k + i)-

connected contains k vertices whose deletion results in
a disconnected graph.
4.5.22. Show that the number of pairwise different
11
connected subgraphs of a cube B , which are generated
by subsets with k vertices does not exceed 2n (4n)k- 1 •
(Assume that the vertices of the cube B" are labelled
by the numbers from 1 to 2 11 ).
4.5.23. By majorizing the number of graphs in <§ 11
having a vertex of degree n - 1, show that almost all
n-graphs have a radius larger than unity.
Let p (G) be a certain numerical parameter of graph
G. Let p(n)=2-G) ~ p (G) be the mean value of
GE~n
parameter p, and Dp (n) = 2-G) ~ (p(G)- p(n))2

GE~n
be the variance of p. In the same way, we can determine
the mean value and variance of parameters of .• the graphs
in'§ n, m· Let e > 0, c5 11 (8) be the fraction of those graphs
G in '9 n for which p (G)~ eI and !1 n (8) be the fraction
of such graphs Gin'§ n• that I p (G) - p
(n} I~ e. Various
estimates and proofs of the properties of nearly all graphs
are frequently obtained with the help of the following
(Chebyshev's) inequalities:
-
p (n) •
On (8} ~ (1)
e '
Dp (n)
t1n {8) ~ 6:1 • (2)
For example, let p (G) be the number of isolated ver-

tices of the graph G. We must show that p(G)=O for almost
all n-graphs. Let g 11 (i) be the number of graphs in ~ 11 ,
in which the i-th vertex is isolated. In this case,
n-1)
Obviously, g 11 (i) = 2 ( 2 for all i = 1, n. Hence
p (n)=
n· 2-n. Putting e= 1/2 in (1) I we find that the fraction
of graphs G in rl 11 , for which p (G)~ 112. does not
142 CU t. 0 R A PllS Alrrr. D N Et\VOR K S
exceed n~~+.l But l1m n 2-"•1 := 0 Hence p (G)< 1/2

:n, ... CL11
lor almost all n graphs 1 e p (G)~ 0.,.
Let u~ now suppose that p (G) 1s the nun)ber of
edges ln the graph G \\e shalJ show that for nedrly all
n graphs p (G)= 1/2 {;) (1 + ta.) where hm siL =0 \\e
ft ... OI:II
ba\e P(n) = 2-(;) ~ p (G)= 2 G) ~ g,. (l, J)t where

GE~n t J)
g 11 (1 J) = 2G)-t 1s the number of graphs m whtch the

pa1r (a, ]) O:( ~er\1ces 1s JO~ned through an edge Thus
P(n) = 1/2 {;} Let u~ calt.ula1.e the. vananee
- (p (n))1
Let us label all pa1rs o! the type (t, l} 1 ~ l <
r ~ n by numhe:rs from t t() { ~} , and let B (\I, i-ll 00 g
the number of graphs G tn Yn~ 1n whJch pa1rs w1th num
hers v and 1-4 are edges Then
But ~ (vt fl) = 2(~)- z If v '=F J.t Hence
Dp{n)={ {;)+{- (;) ({~} - i } - ( } (;} } 2

~T( T}
Put tiDg 9 = V ~p (n) m (2)~ we find that the fracllon
of those graphs G E §"' for w hu:h p (Gl ~ {;) ~
] / ; (;) does not exceed lin Hence for a1moc;t all
graphs p (G)= f/2 {;) (1 + l!:n), where hm t!:q =0
:fl...,..CIO
~G. RJ:PRESJ::::\TATIO::\S Ol' DOOLE..\::\ J:T::\CTIO::\S 143
4.5.24. Let p (G) be the number of pairs of different

vertices of graph G in f) n• for which there is no chain
-
of length smaller than 3 joining these vertices. Let p (n) =
n
2-(2) ~ p (G).
G<:~n
(1) Show that p(n) = ; (~) ( ~ r- 2

•
(2) Show that almost all n-graphs have no vertices
separated by a distance larger than 2.
(3) Using the results of problems 4.5.23 and 4.:>.24 (2),
show that the radii and diameter s of almost all n-graphs
are equal to two.
4.5.25. Show that the average number of Hamilton ian
l · I G · . equa l t o (It-t)!
P n IS
eye es grap 1s m :; ·> .,.
4.5.26.
111
- •
Find the average number of cycles of length 3

11 1
in graphs G of '!J n. m·
4.5.27*. Using the Chebyshe v inequalit y (2), show
that nearly all in (n, m (n))-grap hs, where m (n) =
[n/ln (n Inn)], the number of isolated vertices is equal
to n (1 - e (n)), where lim e (n) = 0.
n-oo
4.5.28*. Let k be an integer (k~ 2). Show that if
~
2- -
m = m (n) = 'P (n) ·n h.-1(where cp (n) oo for
n ).. oo), nearly all (n, rn)-graph s contain a complete sub-
graph with . /;: vertices.
4.5.29. Find the average number of /;:-vertex indepen-
dent sets in g1·aphs G in ,'rJ,..
4.5.30. Let k be a natural number. Calculate the aver-
age numbe1· of vertices of degree k in g1·aphs G of .'9,.. m·
4.5.31. Let p (G) be an integral non-nega tive parame-
ter and let p (n) be its average value for graphs G in f) n.
Show that if lim p
(n) = 0, then p (G) = 0 for almost
all graphs. n-oo
4.6. Representations of Boolean Functions

by Contact Schemes and Formulas
A network r with k poles in which each edge is la-
~elled by a letter from the alphabet {xl, X2• • • • , Xn, X1
- 1
X2, • •• , Xn} is called a k-pole contact scheme (circuit) repre-

14:-1 Cll 4 C'"t\AP~IS AND NrTVtORKS
$en.ttng Boolean functlons of vartables .x1 .r1 Xq

or 1n snort a (k n}-scheme Th.e (2 rl.) sGhemes wt(l be
called xn schemes The: netv...ork r lS called the contaet
scheme netuork A contact scheme ts called connected
(strongly connected sertes parallel etc ) 1f 1ts network
l1as the same property A sertes parallel contact
scheme 1s brtefiy denote-d as ~ scheme Tlte edges
of a .sc1Ierne labelled by the symbols ot varrables or
the1r negat.Jons are .calJ~d conlacts A contact marked
by tne symbol of a var1able (or lts negat1on) ts called
a makt (resp break) cc-:nta.ct Let. l: 1 and S ~. he tWD k pole
contact schemes whose poles are labelled by the letters
a1 a 2 a,._ T.he schemes l: 1 and X 1 are called isomor
phtc 1f thetr networks are 1.somorpntc and 1l (a) the corres
pond1ng edges are labelled 1n the same way and (b) the corr
espond1ng poles are labelled 1n the same \Yay Let a
and b be two poles of tl1e contact scl1eme S and let {a b)
M. a cha.\u. 3Q.-.,n\n.g « arul tt Let Ki a t.l be the canlt.Lnc
t.1on of letters ass1gned tC) the edges of the eha1n (a b1
,....
Tb.e lunctton I atr (x } defined hy tlte formt1la
N
fnb (:c~) = V Ara 61 (3)

la b}
1n wh1ch tha dtsJunctlon IS taken over all stmple cbatns

tn the scheme )Oln1ng th.e poles a and b lS called the
:conducttvtty funttron between poles a and b of the scheme .,.,.,
~ Ttte scheme t ts sard. to represent tlte functton g (.rn)
f"'tJ .....,
tl It con tatns the poles a and b sucJ1 tlta t g (xn) == f ab (xn)

A contact scheme Wllh k + t pol~s tS Galled a (1 A)
pole network 1f one of 1ts poles 13 Isolated (tl11S pole w1ll
be denoted by a) and tlu.~ rema1nang poies are Intercl1ange
able (tbese poles "tll be des tgnated by 6 (l ~ ·t-kj}
;.,,
The functton g (x") 1s :sa•d to be represented b) a (1 k) pole

network If tJtere extsts a pole b (t ~ l ~ k) sucl1 tltat
~ .......
I a~-c (x~) = g (xn) lf the number of po1es 1n a scl1eme
1s not lnd1cated we. shall always mean two pole contaGt
schemes Two contact schemes are called equEvalent
1f they represent the same Boo1ean funct1on The ctJm
plex1ty of a contact scheme IS the number of ats tontacts
A contact sclteme ha.v1ng tl1e lo~est coroplex1ty among
4,0. REPRESENTATIONS OF BOOLEAN FUNCTIONS 145
all equivalent schemes is called minimal. The complexity

of a Boolean function f in a class of contact schemes (nota-
tionLk (f)) is the complexity ofthe minimal contact scheme
representing f. The complexity of a Boolean function
fin the class of n-schemes is the number of contacts in the
minimal n-scheme realizing f (notation Ln (/)). The com-
plexity of a Boolean function f in a class of formulas gener-
ated by a set of connectives { V, &, -}is the number of
entries of the symbols of variables. In this class of for-
mulas, the complexity of a function f is denoted by
L (f).
A directed contourless network whose poles are divided
into input and output poles is called a scheme of junctional
elements. The input poles are labelled by variable sym-
bols. The output poles are called outputs of the scheme.
Each vertex other than an input is labelled by
a functional (logical connective) symbol. The following
conditions must be satisfied in this case: (1) the
in-degree of each entrance pole is equal to zero; (2) the
in-degree of each vertex other than the input pole
is equal to the number of arguments of a functional
symbol (or connective) which labels a given vertex.
The concept of a function f t represented at the vertex
i of a scheme Z is defined as follows. If the vertex i coin-
cides with the input pole labelled by x, then ft = x.
Let the vertex i be labelled by a functional symbol cp
of r arguments, and let cp 1 , . • • , CJlr be functions repre-
sented at the vertices from which the arcs terminating
at the vertex i emerge. In this case, It = cp (cp 1 , • . • , CJlr)·
It is said that the function f is represented by a scheme
~ if the latter has an output where this function is repre-
sented. Functional element schemes with one exit whose
entrance poles are labelled by the symbols x 1 , • • • , Xn,
and the vertices other than the input poles are labelled
by the symbols v. &, - will be called xn-functional
schemes. The complexity of a functional element scheme is
the number of its vertices other than entrance poles. An
xn-functional scheme :z representing a function f is
called minimal if every other xn-functional scheme re-
presenting f has a complexity not lower than that of the
scheme Z. The complexity of a Boolean function f in
a class of functional element schemes is the complexity
of a minimal xn-functional scheme representing the func-
10-0636
CH ' GRAPHS A~D NEtwORh.S
tJon f The com plex1ty of a funct1on 1 ln thts clas~ of

schemes WJ ll be denoted by L lj)
#'It# ,.,
4A6.1 ~ Let f (x 2) ~ x1 61 ..r1 Pro\ e that L (/ (:r 2 )) ~ 4

4 6~2~ Show tl1at for a DooJea11 funct1on nther than a
constant, the m1n1mal contact 5Cherne representing tJIJS
func t1on IS strongly r onnected
4.f6 3. F1nd the number of Boolea11 functJons / (z 1 ~ x2)
represented by contact schemes of complextty 3
4.6 4. (1) Show that for each natural m there ex1sts
a mtn1mal contact schente of comple1tty "'
(2) Show that there are no m1n1mal contact schemes of
complextty 4 conta101ng only clos1ng contacts With labels
1n the set {x1 • x2 .r3 ]
4.6.5. Show that 1f m > n 2»-t~ none of the xn
schemes of complexity m IS mintmal
4 6.6. Show that the functton 1 depends essenttally
on v.artable z If and only 1f the minimal scheme rrpre
a a
(a) (b)
b
(c)
F1g 19
sent1ng I eonta1ns a contact labelled by the varlahl~ .r

.flr by 1.ts negat1on
4 6-7. For the contact schemes sho-wn ln F1gs 19a-.
b, c .and 20a, b, c find the conduGllvttY function fab
411.6 ..8. Construct the contact schemes reprecent,ng ll1e
func.ttons f 1n the follow1ng cases
(i) I (;a) = (z. V .z._ V .:r3) (~ V ;I! V X3},
;t.;
(2) j ~X )
3
-:=. X1Xz $ 2a%J $ X3X1i
4 6. REPRESE~TATIONS OF BOOLEA~ FUNCTIONS 147
~
(3) t (r) = cooo11111};

(4) f (;a) = (11010001);
~
(5) f (x3) = X1 El1 x! El1 x3.

4.6.9. For each function in Problem 4.6.8, construct
X 3 -functional schemes.
z y- u
-
y
z z
a b a b
- z y -
!J u :r
(a) (b)
z
z
(c)
Fig. 20
4.6.10. Construct a contact scheme with a complexity
not exceeding L for a function j, after simplifying its
formula if:
-- (1} (x1 V x 2x 3) ((x1·V x 2 ) (x2 V x 4) V (x 3 V 4 ) X x

(xlx2x3 V x 4 x 5 )), L = 5;
-- - --
(2) (x1 V X 2 ) ((x1x 2 V X 2 x 3 ) x 6 V (x 2 V x 3x 5) X
- - --
(x2 V x3) V (x1X2 V X1X2 V Xs) Xa V X1X2 V XsXs),
L = 5;
-- - -
(3) x 1x 2x 3 ((x 4 V X1X5) (x 6 V x 1x 4x 7) V (x 6x 7 V
- - - - - -
X2x 4x 5) & (x 4x 5 V x 3 x 6 x 7 ) V (x1 V x 2 ) (x 4 V x 6 ) X
(x 5 V x 7 )), L = 6.
4.6.11. For each function in Problem 4.6.10, con-
struct a functional element scheme of complexity not
exceeding 6 and representing this function.
~ n
4.6.12. Let v (ex) = ~ 2iexi be the numhPr of the
~
i=t
tuple ex = (ex 1 , ex 2 , ••• , exn)· Construct a contact scheme
10*
CH '- GRAPHS AND NEtWORKS
w1th not more than tO contacts representing the func

t1on
"" 1 1f v (a 1 a 2 a 3 ) ~ v (a.., a: 5 cte)
flr::')-- 0 otherwise
t3 Construct a contact scheme representing the
4 6
"""'
function f (x') whtch ts equal to unt ty 1f and only 1I
(.rl z 2 x 3 ) ~ (x4: 2 5 x,)
4 6 14 Construct a con.taet scheme representtng the
add1t1on of two d1g1t htnary numbers Speaktng more
precisely construct a contact scheme WJth poles a b., b1
and ba 1n wh1~h fur each ' - 0 1 2 the eonduc.tiVlty
~
funet1on /ob 1 (xc) 1s equal to : 1 where z1 E {0 1} and

15 determ1ned by the relation 4z1 + 2z, + z, ~ 2 (rl +
.x.~) + .Xt + x,
4 6 15* Construct a functiOnal element scheme sat1s
fy1ng the follow1ng eond1 tlons
(1) 1t represents three functions 1 :r-;. X, x
(2) the vert1ces other than entrance poles are labelled
by V & and-,.
(3) the scheme has not more ttian two vettlces lah~lled
by nega t1on symbols
4 G 16 Show that L¢t U) ~Lit U> for any Boolean
funet1on 1 wh1ch l.S not a constant.
""'3
4 6 17* G1ve an e-xample of a funtttlon 1 (x ) fo.r wh1ch
L,. U} > L}t U)
Hmt Consider the functton represented by a scheme
shown 1n F1g 19a
Let a 2 connected two pole contact scheme I be a pia

nar sclleme (t e let tts network r (a b) be plane) 8nd
let 1ts poles a and b l1e on one face \Ve draw an edge
(a b) on this face 1n such a way that the network r
obtatned from r by addtng the edge (a b) rematns planar
We choose one vertex each on the filces of the network r
On the vert1ces ehusen 1n th~s way we construct a graph
G• dual to the graph G of the network f Each edge of the
graph G• other than (a b) Intersects a contact of the
s~heme ~ We label th1s e-dge by the same letter that 1s
us~d to label the contact intersected by 1t Let us denote
4.6. REPR ESEN TATI ONS OF BOO LEAN FUNC TION S i49
the vertices of the grap h G* loca ted on the faces of the

network r' divi ded by the edge (a, b) by a*, b* and call
them poles. We dele te the edge (a*, b*) from G*. As
a resu lt, we obta in a 2-connected scheme 1:* with poles
a* and b*. The scheme 1: * is call ed a sche me dua l to 1:.
4.6. 18* . Show that the scheme 1:* dua l to a plan ar
2-connected scheme 1: repr esen ts a Boo lean func tion dua l
to the func tion repr esen ted by the scheme 1:.
Hin t. Esta blis h a one-to-one corr espo nde nce betw een
the chai ns of scheme 1: and the cuts of scheme 1:*.
4.6. 19. Con stru ct schemes dua l to thos e shown in
Figs 19a, b, and 20a .
4.6. 20. Prov e that for every Boolean func tion I the
equ ality Ln (/) = Ln (/*) is satis fied . -
4.6. 21. Show that if £11. (/) :::;;;; 7, then Lk (f) = Lk (/).
A con tact scheme is call ed simp le if the labe llin g of an
x
edge by a lett er x or mea ns that ever y othe r edge has a
labe l othe r than x or x.
4.6. 22. Show that a stro ngly connected sim ple con tact
scheme represents a func tion that depends esse ntia lly on
all vari able s enco unte red in z
the scheme.
4.6. 23. Show that a
stro ngly connected sim ple a "'-
scheme is min ima l.
4.6. 24. Show that if a
min ima l sche me is sup plie d v
with a new con tact labe lled Fig. 21
by a new vari able in such
a way that a stro ngly con nect ed scheme is repr esen ted,
the newly formed scheme will also he min ima l.
4.6.2.5* .. Let I be a func tion repr esen ted by the scheme
shown m F1g. 21. Prov e that a func tion dua l to 1can not be
represented by a sim ple scheme.
4.6. 26* . Is the rela tion £~~. (/) = Lk [f) true for all
Boolean functions f?
4.6. 27 · Let I ab and I ad be the con duc tivi ty func tion s
of a three-pole con tact scheme 1: 1, while 1 and 1 h are
the con duc tivit y func tion s of a thre e-po le c;~tact s~heme
1:2. Let 1: be a scheme with poles a and e, obta ined from
the schemes 1: 1 and ~ 2 by iden tify ing pole b with pole g ·
and pole d with pole e. Is it true that the scheme ~ re-
f. 50 C.U l GRAPHS Alrr.D NETWOltKS
presents the functton

fat!= (fab & /eg) V (fab & feg)?
4.61128. (1) Shnw that the funct1on I \S monoton1~ tf
and only 1f tbete e:x1sts a -c.nn tact seheme representtng f
and not ~onta1n1ng open1ng contacts
(2} Is Jt tr-ue that the mtn1mal conta~t stheme repre
sent1ng a monotonte func.t1on does not conta1n opentng
,..,
contacts? The Boolean functiOn f (xn) IS called mnnotonk
ln variable z 1 1f 1t can he presented ID the form.
.......
f (zn) = g (Ztt Z 3 ,. f Xn) V x:ih {x2 t Zat , Xn)
,.....
The funttlon f (:z:n} 1s guas1. monotonu: ~on varlable x1 tf 1t
lS monot.on1e 1n x1 or become-s m()notontc 1n x1 after sub
z
stJtnttng 1 for :r1 The monotontcity and quast monoto
Diclty 1n x 1 (l :f.;. 1) are defined In an 1dent1eal manner
~
4~6 .. 29 Prove that the function f (zn) 1s quasi mllno

tontc In x~ 1f and only 1f there extsts a contact scheme
""""
represent1ng the funct1on I (z"} and c-onta1n1ng ntn the-r
mak~ nor break contac.ts
4jt6.30. {1) Prove that the m1ntmal GOntac.t stbeme re
presenting the functton X1 EfJ Xg contalDS 4 contacts
(2*) Prove that the sth:eme ~hown 1n F1g 20c 1s m1n1
mal
4 ..6 3f • . Construct a m1ntmal contact scheme for the
funet1on J 1f
f'N
(1) I (z") = (.rt V :t1 V x.s) x4 V z 1;r2 r.J;

~
{2) I (x } = z 1xr: 3 EB ~xi;:z: 4 $ r 1 X:tX 4 EB x 2.xa-t 4,

4
#'¥
(3) I (x') = (0001011101111111)

4116.32 .. Show that the number S (n, m) of connected
pa1rw1se Don J.somorph1c X" contact schemes of complex
1ty not more than m does not e1:ceed (cnm)mt where c
IS a constant Jndependent of n .and m
4.6 . 3341 Show that the number P (nt m) of connected
pa1rvt1se non 1somorph1c n schemes of complexity not
more than m and real1ztng Boolean functions of variables
x1, .r2, ~ , Zn does not exceed (en)''\. where c 1s a con
stant Independent of n and m
4.6.,34. Show that the number Cll (n, m) of pa1rw1se
d1fierent formulas of complexity m generated by a set of
4.6. REPRESENTATIONS OF BOOLEAN FUNCTIONS 151
connectives (V, &, -) and by a set of variables .r1 ,

x.,, ... , x 11 does not exceed (cn)m, where c is a con~l an l
-
independent of n and m.
The lower estimates of complexity in representing
different classes of functions by schemes and formulas
are frequently obtained by using "power considerations".
The following statement is an example.
Let S (n, m) be the number of schemes from a class K,
each of which represents a Boolean function depending on
the variables x 1 , x 2 , • • • , X 11 , and has a complexity
higher than m. Let q> (n) be the number of Boolean func-
-
tions I (x") in a set ~m. Then, if S (n, m) < q> (n), there
"'
exists in m a function I (x") that cannot be represented
in class K by a scheme of complexity lower than or equal
to m.
4.6.35. Show that for any e > 0 and a sufficiently
large n, there exists a self-dual function I (x") for which -
the following inequalities are satisfied simultaneously:
211-1
(a) Ld!) ~ (1-e);
II
?Il-l
(b) L 11 (f) ~ l~gzn (1- e).
4.6.36. Show that for any e > 0 and a sufficiently
-
large n, there exists a function I (x11 ) which is a superposi-
tion of functions q> (x, y, z) = xy V z and is such that
n-1
n-1
Let> (/) ~ .....;..:='Io-g-
2
n-='-
4.6.37. Let L (n) = max L (f). Show that for any
/EP~
self-dual function I "'
(x11 ) the inequality -
L (! (x'')) =s:;;
L (n - 1) + 4n is valid.
4.6.38. A Boolean function F (ym) has the property -
. "'
U" If any Boolean function I (x") can be obtained from
~
F (y"') by substitution of constants and redesignation of

variables (identifications are admissible).
y
(1) Sl_ww that the function y 1 2 has the property U 1 •
(2) Fmd the function with the smallest possible number
of variables having the property U•.
-
!52 CU ' GRAPHS Al\0 NETWORKS
(3) Let m (n) be the smallest possible number of \art

ables of a lunctlon hav1ng the pyoperty U n Show that
r+
I og n:
11 ... )
9 ~ m (n).::;;;; 3
2"
1
4 6 . . 39 .. Show that there eXI.Sts anxn runcllOnal ~cheme

having a com.pfex1 ty 2 22~ ~ n and represen t1ng a II func
tions depending on variables x1 ~ , $n
For constructing twa pole contact scl1emes representing

Boolean functions, the method of cascades ts frequently
used We shall descrtbe th1s method here Let I {x11
.r2 , , :rn)t n ~ 2 b~ a Boolean runctlbn wh1cll has
to be repiesent~d by a c.cntact scheme By ~, (t ~
1, n - 1. we deno.te the. set of all Bool-ean funettons, each
of whtch depends only on varw.bles x 1+l' ..z 1+1 ., ~ rn,
~
and can be ohta1ne-d from the funct1on f (xn) by an apptop

rtate subslltUllon of t.eros and UnltleS for the varlable~
.t1 .t,,. ~ x 1 Each set 2[ 1 IS put 1n a one to one cortes
pondence Wl th the set Vi v. hose elements are potn ts 1n
the plane~ called ' th rank verlrtes \Ve add t\\o more
poles, vtz the 1nput pole a and the output pole b Pole a
IS a zero rank vertex, wh1le pole b JS a vertex of rank n
Tbe set of -vert1tes 1n the scheme ~ representing the func
,._ n-l
t1on I (xn) Will cotnc1de WIth {a} U {b} U U Vi The
t--1
set of contacts of the scheme may be descrthed as follows
Jet v 1 be an arbltrary ' th ra.uk vertex (n - 2 ~ l ~ 0)
and let the funct1on q; (x1 + 1 ., xi+:, x Q) In the set
~, correspond to rt (for 1. = 0, the funct1oi1 tp 1s rdenttcal
,..,.
to the functton I (.x11)) The funct1ons 4F (0, xf+t, , xn.)
and rr (f t Zl+lt , Xn) belong to the set 2l.t-t1 and
have certain Vertrees v,+ 1 and v;..,.. t re~pectt vell corre
spondtng to them (tf the functions rp (0, i; +1 , , .xh)
and £P (1 ~ x 1+s, • Zn) are tden t1cal, the vert tees v,+-t
and v;+, colnclde) The vertex L 1 ts JOlned 1n scheme S
to the vertex vi+ 1 through tha contact i, -t- 1 and to the
vertex vi+1 through x 1 +1 F1nally, the (n- 1) th rank
vertJces are Joined w1th the n th rank vertex (pole 6)
accord1ng to the following rule
(1) 1f a vertex u 1n V n-1 corresponds to a functton :En,
1t IS connected to tlie pole b t)Jrougb the contact Xn •
(2) 1f a vertex v corresponds to a functtop .r;,
1t 1s
JOined to b through the contact Zn;
4.6. REPRESENTATIONS OF BOOLEAN FUNCTIONS 153
(3} if a vertex v is put in correspondence with a func-

tion identically equal to unity, it is joined to b through
-
two parallel contacts x 11 and X 11 ; and
(4) if a vertex v is put in correspondence with zero, it
is not joined to pole b.
4.6.40. Using the method of cascades, construct a
scheme representing the function 1:
(1) I (x3) = x 1xz EB x3;
~
(2) I (x~) = x 1x 2 V x 2 x 3 V x 1x,;

~
(3) f (x = Xt EB Xz EB ••• EB Xn;

11
)
"' -- -
(4) I (x") = x 1x 2 •.• Xn V x 1x 2 ••• Xn;
- 11
(5) f (X 11
) = V X1 ••. Xj-tXi+l ••• Xn•
t=l
~
4.6.41. (1) Show that if I (x") =,;S 0, we can delete

from all sets 9{ i the functions that are identically equal
to zero when constructing by the cascade method a con-
tact scheme representing the function I.
(2) Show that the contact scheme obtained through
such a construction is strongly connected.
~
4.6.42. (1) Suppose that the function I (x"}, n ;;;:::: 2,

depends essentially on all its variables. Prove that the
scheme representing the function f and constructed by the
cascade method is strongly connected and does not contain
parallel contacts of the type x 1, x 1 if and only if I is a
linear function.
~
(2) Give an example of a nonlinear function I (x"),

n ;;;:::: 2, depending essentially on all its variables and
such that the scheme representing it and constructed by
the cascade method does not contain parallel contacts
of the type x i• xi·
-
4.6.43. Find out if the following statement is true or
~
false: if a function I (xn), n;;;:::: 2, depends essentially on

all its variables and the contact scheme representing I
and constructed by the modified cascade method described
in Problem 4.6.41 (1) contains n contacts, then I =
XtGr Xz a~ ••• Xnan , wl1ere a i E {0 , 1} •
4.6.44. Disprove the following statement: if a func-
tion I is such that for a certain i each of the sets ~ i
and ~ i +1 used in the cascade method does not contain
GH 4. GRAFHS A 'D tr.tTW'ORh.S
funtt1ons tdentically equal to zero and unttyt hut does

conta1n at least three functtons, then the stheme repre
senttng f and cons true ted by the method of cascades 15
1
not plane
4 6 45 G1ve an example of a function f for wl11C11 the
cascade stltema teallz.tng 1l 1s not plane
l \Ve mean the ~heme! w1thout an add1ttonal edge JOJnJng the

poles
Chapter Five
Fundamenials of Coding Theory
5.1. Codes with Corrections

Let A and B be two finite alphabet s and let R be a
certain set of finite words in the alphabet A. A single-
valued mapping <p of the set R into a set of words in the
alphabet B is called a coding of the set R. The counterp art
C of the set R for the mapping rp is referred to as a code
of the set R. Words in C are known as code words. If a
word w in R maps a word v in C, then v is a code of the
word w. Words in R are called messages, and the alphabet s
A and Bare called a message alphabet and a coding alphabet
respectively. If a coding alphabet B consists of two letters
(in this case we shall assume that B = {0, 1 }), the coding
<p and the correspon ding code C are binary. A code is
referred to as a uniform, or bwck, code if all code words
are of the same length. A block binary code in which
every code word has length n is a subset of vertices of
~
a unit n-dimensional cube. A Boolean function fc (x")

which is equal to unity in the set C and to zero outside
C is called the characteristic function of the binary block
~ "'
code C. Let p (a, p) he the ordinary Hammin g distance
,.... "'
between vertices a and ~ in B", which is equal to the
"' "'
number of coordina tes in which a and~ differ. The quant-
"' ,....
ity d (C) = min p (a, ~), where the minimum is taken
over all pairs of different vertices belongin g to the code
C s sn, is called the code distance of the code C. The
code C s B" with a code distance d is abbrevia ted as
the (n, d)-code. The maximum possible power of an
(n, d)-code is denoted by m (n, d), while an (n, d)-code
whose power is m (n, d) is known as the maximum code.
A compact code is an (n, 2d + 1 )-code which satisfies
"'
the following condition : for any vertex a E B", there
• 1'1 "' I"J
exrsts a code word ~ for which p (a, ~) ~d. A _binary

1.56 CH 5 FU"DAI\.IENTALS OF C0D1NG THEtrnY
block code 1s referred to a.s equtdtstant Jl the dt.stance be

t"een any two rode words JS constant A code C £ Bn
15 kno\\ n to he an equal welght code tf any code word bas
the same v. e1ght 1 e 1f there extsts an Integer k (0 ~
k ~ n) such that C £ Bh Thts number k 1s calle:d the
wetght of an equal we1ght code The quantity max i C ~
where the maximum lS taken uver all <n. d) codes -of
we1gb.t k 1s denot-ed by m (n, k, d)
Let lhe words of a htnaey block code C be tr.ansrn1tted
th.rougll a cotnmuntcatton channel where a dlslortton of a
transmitted word roay occur The transmiSSIOn vta such
a channel can be regarded a.s transformations of the words
be1ng transmitted Here we shall consider only such
transformattons of binary words wh1ch do not change the
length of a word and consist 10 the replacement of certarn
letters by the opposite, 1 e ln the replacement of 0 by f
;.v
and 1 by 0 If a word a ts transformed tnto a word ~
.f"W
d1ffer1ng from ct as a result of transm1Ss1oo th1s mean!i

that errors have been lntrvduced 1nto the channel If the
/flu
t th letter of a word a be1ng transmitted dtffers from the

.tltU
r. th Jetter of the obta1ned word ~'! the error IS satd to

occur 1n the ' th dlglt lf the ohtatned word dtffers from
the word betng transmitted 10 t d1g1ts, t errors are sa1d
to have been tntroduced Obv1ously tbe number of errors
made durtng a tran.smtsston 1s equal to the Hammtng
d 1stan.ce between the transm1 tted and recei \ ed
"\\Ords
Let cs
nn be a binary code An arbitrary Single
valued mapp1ng ..p of the set 1P onto the set C IS called a
~ ~
decodtng Let a E C, and let \b- 1 (ex) be a set of all ver

........ ,_ ....... r"w
ttc:es ~ tn Bn, such that "¢ {~) == ct Let S~ {a,) be a set

,..,.,
of all words wh1cb are obtained lrom a code l\Ord a
(4J
as a result of not more than t errors (obviously S7 (a)

......t
ts a sphere o( radtus t w1tb. center at r.t) A code C 1s sa1d
to correct t errors tf there extsts a decodtng ~ such that
,.., ...... """"'"
S7 (a.) s; ..p
{ct.) for any a E C A code C reveals t errors
1
1f any word whtth -can be obtained {rom an arh1trary code

"'tJ
"ord a as a result of not,._ more than t distortions d1ffers

from any word tn {~} c,
5.1. CODES WITH CORRECTIONS 157
5.1.1. Prove that a code C s; B" corrects terrors if and

only if p (v, w);;;::: 2t +
1 for any two different code words
v and win C.
5.1.2. Is it true that a code C s; B 11 correcting t errors
reveals
(1) at least 2t + 1 errors;
(2) at least 2t errors;
(3) not more than 2t errors?
5.1.3. Prove that it is possible to obtain from any sub-
set C s; B 11 a code detecting an error by deleting from C
not more than half the vertices.
5.1.4. Determine the number of errors corrected and
revealed by a code with a characteristic function f if:
~
(1) f (x ) = x 1 El1 X 2 $ ... $ Xn;

11
(2) f (x") = x 1x 2 ••• X 11 V x 1x 2 ••• X 11 ;

f"tJ - - -
,..., - - -
(3) f (x ") = X1X2 • • • X3n V X1X2 • • • XznXzn+tXz~+2
3
.. • X3n
- - - -- -
~
(4) f (x") = x 1x 2 ••• Xn-t $ x 1x2 ••• Xn-.!•'l:n $ ...

EJ1 XzX3 • •• Xn •
5.1.5. Let the words of a binary code C be transmitted
via a transmission channel so that not more than one error
can be introduced during the transmission of a code word.
~
For each code word ex, construct the set of words ,...,
which
can be obtained as a result of the transmission of ex through
the channel:
(1) c= {01100, 00111, 1'1010, 10001 };
(2) c= {11110, 10100, 01011, 11001 }.
Let p be the probability of the fact that an error is

introduced in the i-th digit of an arbitrary word w in B"
during its transmission via a channel for all i = 1, n.
Let C s; B be a code of power m, and tjl: B" ~ C be a
11
decoding. The quantity

Q..p(p, C)= !I~ ~ pP(v,w>(l-p)"-p(v,w)
t-EC wE¢-•(v)
is ~alled the authenticity of the decoding tjJ for the code C.

<>.1.6. Prove that for 0 < p < 1/2, the maximum of
the quantity Q.p (p, C) over all possible decodings for a
158 Cit 5 FUNO.A\It::,~ALS OF COOI'\G THEORY
fi.x-ed C 1s attatned prov1ded that for each wEB" the

equality p (w, 1p (w)) = m1n p (w, v) IS satisfied
~'EC
5 !.7. (1) For the code C 1n Problem 51 5 (1),. con
struct a decod1ng w
"1th the maxlm.llm authentlct\)
Q.; (p, C), 0 < p < 1/2 and lndtcate the set 1p 1 (w)
for each w E C Determine max 0¢ (1/4, C)
tb
(2) Determ1ne the number of different de.codtngs 'P
w1th the maxtmum authent1c1ty for the code C ID Prob
lem 5 1 5 (1) and 0 < p < 1/2
5 l 8 . Detetmtne the max1mum posstble po1&er of ,.,a
code ,..C ~ Bn possesstng
_,
the follow1ng property for any a
and ~ 10 C, p (a, ~) are even
5.1.9 Detenn1ne the number of maxrmum {n, 2)
codes
5,.1 fO. Let n === 3k Prove that m (n, 2n/3) ~ 4
5 .. 1. 1 t.. Pro'\e \lrat the pov,.er ol a compact (n,
d
2d + 1} code 1~ r;.n ~ j (; )
1-0
.5.1 .. 12,. Does there extst a compact {nt 3) code for
n = 147?
5,.1 l3. Prove that for n > 7, there eXIsts no compact
(n, 7 )-codes
5.1.14 .. Pro\e that a code C ~ Bn 1s not maximum 1f
J C I= 3
5. t 15 Prove tha l 1f there exIsts a compact (n... 3) code
In nn 't there ex•sts a partJtlOD of the cube Bfl+l Into dlS·
)o1nl spheres of tadlus 1
5 .. 1. 16. Let C be a compact (n, 2d + 1) code Prove
n
th~t tn thts case ts exactly dlVtstble by
d+1
2d+f
d
5.1.17 Prove that there ex1st no equtdtstant (n, 2d +
1} codes of power exceedlng 2
5. t .. 18.. Prove that for an even d there e.xtsts an eq u 1
dtstant code of power l2n/d)
5.1 .. 19. Prove that m (n t d) 1s a non decreasing func
tton of parameter n.
5.1. CODES WITH CORRECTIOXS 159
5.1.20. Prove that

(1) m (n + d, d)~ 2rn (n, d);
(2) m (2n, d)~ (m (n, d}) 2 ;
(3) m (n, d) ~2m (n- 1, d).
5.1.21. Prove that
21
m(n, 2t-!-1)~2"/h
i=O
c:).
5.1.22. Prove that
m (n, k, -:--:--....:.-~--.
~ {~) (n-;k)
i=O
5.1.23. Prove that for n < 2d, the following ine-

quality holds
--- 2d
d) =::::::::
m (11 ' 2d -11 •
5.1.24. Prove that
m(n, k, nd
d)~ [ 2k 2 -n(2k-d ) J
'
if 2k2 -n (2k-d) > 0.

5.1.25. Prove that
(1) m (n, k, d):::;;[~ m (n-1, k-1, d)J;
(2) rn(n, k, c/)~[1-[~=:[ .. ·[~_::J .. ·JJJ:
(3) m(n, k, cl)~[ ~~~k m(n-1, k, d)J.

5.1.26. Let q (n, d) be the maximum number of ver-
tices in B", the pairwise distances between which do
not exceed d. Prove that
m(n, d-f-1)q( n, d):::;;2".
5.1.27. Prove that m (n, d)~ 2"-d+ 1 ,
5.1.28. Prove that from any set C s; B" such that
I C I ~ 2c1, it is possible to isolate a subset D of power
not less than 2-d+l 1C I which is an (n, d)-code.
160 CH 5c FUfrr.OA \IE'ITALS OF CODING TIIEORY
5.2. Linear Codes

The express1on of the form
{1)
,jA,I
where a., E Bn A1 E (0 1} ....,;

1.
"""'-J
~ 1 s \s called the ltnear
l"'fV
comblnatlon of the vectors a 1 , a~ a.s Tlle It near .com

b1na t1on (1) called trl.t-lal tf A1 ~ Ai ==
J.s = As 1 0 !==
and non trtvtal otl1erwtse Any f1near comblnat1nn ......of

vectors In nn lS obVIOUs}y a vector In Bn Vectors at
,..,.. r-.1
a1 , cx:-8 tn B to~ are ltnearly lndepen.dent tf any of the1r

,_
non trtvtal ltnear comh1nattans dtHers [rom 0 = (0

,..., f"Y ~
0 0) Otherw1se the vet tors a.i, a..2 r.1.~ are

said to be ltnearly dtptndent A subset G s; s~ lS called
a. group tf G ts closed wtth respect to the exclustve so.m,
"""""' ........ ....... """'
1 e for any ~ and ~ 1n G the -vector a ~ ~ belongs toG 1
It follo\\s {rom the closure of G Wtth respect tD tll.e oper

at\on e tllat any llnear oombtnatlGU of vect-ors lUG
.f'llt.,
also belongs to G (1n part1cular, 0 E G) Thus any group

G 1n Bn lS a llnear vector space generated by the field F 2 =
( {0 1} ffi ) The max1mum numb-er k ~ k (G) fnr
\\htch there extst k lLnearly Independent vectors tn the
group (ltnear space) G IS called the dlmenslonal,ty of G
The set of k linearly Independent vectors of a l1near k
dtmensional space IS known as the basls of th1s space
If a code G s; Bn forms a group, tt Js called a llnear or
group, code If a ltnear code rn Bn has a dtmensionality k,
lt Js referred to as an (n k) code A h1nary l1near code
correcting one error lS known as a Hammlng code
L1near codes can be conve-nlently defined 1n tern\s of
matrtces A matrtx H (C) "hQse ro~ s are code \\ords of
a code C s n» IS called a code C matrz..x A matrix AI (C)
formed by k arbitrary ltnearly Independent vectors whtch
are code words of an (n k) code C IS called a generat1ng
matrix of the code C If H IS .an arbitrary matr1x formed
1
The definJtJons of mod 2 sum of vectors and mod 2 product
of a .sea lar by a vee tor are g1 ven 1n the see t lon Boo Iean \ ee tor!
tr.nd a Un1t n drrnens~Dna.l Cube
5 2. LINEAR CODES 161
by zeros and unities with n columns, the set C (H) of

all vertices of the cube 8'', which are linear combinations
of rows in the matrix H, is called a code generated by the
~ ~ ~
matrix H. Vectors a = (a1 , a 2 , • • • , an) and ~ =

(~ 1 , ~ 2 , • • • , ~n) are referred to as orthogonal if a 1 ~1 EB
a 2 ~ 2 EB ... EB an~n = 0. The set V (H) of all vectors
in Ell which are orthogonal to each of the rows of the
matrix H is known as the null space of the matrix H. Let
C be a binary code whose each word is orthogonal to each
row of a certain matrix H. If C is an (n, k)-code, and the
matrix H consists of n - k linearly independent rows,
then H is called a checking matri:c of the code C. The set C*
of all vectors that can be represented as a linear com-
bination of the rows of the checking matrix of the (n, k)-
code C is called a code dual to the code C. By g (n, d)
we shall denote max I C I, where the the maximum is
taken over all linear codes C s 8n with a code dis-
tance d.
5.2.1. Let a set C s B" consist of k linearly indep-

endent vectors. Prove that any two linear combinations
of vectors in the set C, having different coefficients, are
different vertices of the cube E".
5.2.2. Prove that there exists in E" a system of n linearly
independent vectors, but there exists no system of n 1+
linearly independent vectors.
5.2.3. Prove that the number of vectors in B 11 that
can be represented by linear combinations of type (1),
where ±J..i ~ t,
•=1
does not exceed ±( ~ ).
t=O
5.2.4. Prove that any (n, k)-code has a power 2h.
5.2.5. Prove that in a binary linear code either each
code vector or half the code vectors have even weights.
5.2.6. Prove that the number of different bases in Ell is
(2." -1) (2n-2.) .•. (2n-2,1l-l)
nl• •
5.2.7. Prove that the number of different (n, k)-codes

in 8 11 is
(2n-i) (211 -2) ... (2n-2h-1)
(2"-1) (2"-2) .•. (2h-zh-1) •
11-0636
f62 Cli .5 FUNDAMENTALS O.F CODlhG 1JIEORY
5~2.8~ Dete.rmme the number of vectors 1n B" wh1ch ate

.,.,., 1l
orthogonal to a. g1ven vector a. in B,.
5.2.9. Pro\e that the set of all vectors 1n Bn which are
orthogonal to each row of a (k X n) bJnary matri.s: H
form a linear space Is th1s space always {n - k) d1men
sional1
5,2 .. 10. Is 1t true that for any lmear code there exists
a matrt:s: H which J!
(i) a generatmg matru.,
(2) a check1ng matrts?
5.2.11. From a gt"Yen matnx H, deterrotne \he power
m (C (H)} of a code C (H) generated by 1t, as well as the
cOde dtstance d (C (H))
Here H 1! an {n X n) matr1x,
(4) H = (J ~tP). where 1A ts a untt (k X k) matriX a.nd
P :LS an arbitrary b1nary (k X (n - k)) matru:~ formed by
dtfferent rows~ each of which cc.nta1ns at least two uni·
\1es, and k ~ 11 .......
log1 (n + 1),
(5) H =- (I 1Q)t where 1 1 l! a (5 X 5) un1t matrix and
1110110010
0111100101
Q== 1 1 0 1 t 0 1 1 1 0 41
0110011111
1101011101
5.2. LINEAR CODES 163
5.2.12. Let V s Bn be a space consisting of linear

combinations of rows in a matrix H = (I ~~.P), where I 11.
is a unit (k X k) matrix and Pis a (k X (n - k}) matrix
consisting of zeros and unities. Prove that V is a zero
space of the matrix G = (PT I n- 11 ), where In-h. is a unit
matrix of dimensionality (n - k) X (n - k) and PT
is the transpose of the matrix P.
5.2.13. Construct a checking matrix for a code gener-
ated by the matrix H from Problem 5.2.11 (1).
5.2.14. Prove that if C s B 11 is an (n, k)-code, the
code dual to C is an (n, n - k)-code.
5.2.15. Let H (C) be a matrix of an (n, k)-code C s
8 11 , which does not contain null columns. Prove that
(1) each column of the matrix H (C) has 211.-I unities
and the same number of zeros;
(2) the sum of the weights of rows of the matrix H (C)
is n·2h-t.
5.2.16. Prove that the code distance of a linear code
C s 8 is equal to the minimum weight of its nonzero.
11
vectors.
5.2.17. Prove that the code distance of an (n, k}-code
does not exceed [n2"- 1 /(2h - 1)1.
5.2.18. Prove that for n = 2d - 1, the power of a
linear (n, d)-code does not exceed 2d.
5.2.19. (1) Prove that the maximum possible power
g (n, d) of a linear (n, d)-code satisfies the inequality
g(n, d)~ 2g (n- 1, d).
(2) Using the result obtained by solving Problem 5.2.18,
prove that g (n, d) ~ d. 2n-2d+2.
5.2.20. Let a code C be a zero space of a matrix H.
Prove that the code distance of the code C is not smaller
than d if and only if any combination of d - 1 or a
smaller number of columns of the matrix H is linearly
independent.
d-2
5.2.21. Prove that if ~ { n-; 1
) < 211., there exists
i=O
a (k X n) matrix consisting of zeros and unities, in which
any d - 1 columns are linearly independent, and hence
there exists an (n, n - k)-code with a code distance
equal to or longer than d.
5.2.22. Let H h, n be a (k X n) matrix, where k =
llogll (n + 1)[ 1 in which the i-th column is a binary de-
11•
164.: CH 5 FUNDAMENTALS OP CODING TBEORY
composition of the number l (i = 1~ n) For example,

for n = et the matr1x H '·' has the form
0 0 0 1 1 1
H= 0 1 1 0 01
1.01010
(1) Prove that the zero space ol the matrix 1s a Ham ..
tnlng cadet 1 :e. a llnear cod-e eortect1ng one. error
(2) Construct a code wh1ch IS a zero space of the matrix
H3 e
(3) Determine the number of errors that are corrected
by the. code generated by the matrtx H,. 11 ......
5.2.23. The specttum of a set C £; B" ts a. vector s =
{s.ot s1 t , sn), where s, Is the number of pa1rs of ver-
tlces 1n C that are r apart Prove that lor any (n. k) code
C, there ex1sts an (n, k) code C' having the same spectrum
and a generattng matrtx of the form (!llP), where I •
1s a un1t (k X k) matriX and PIS a (k X (n - k)) .matr1x
con.s1St1Jlg of zeros and untttes
5 . 2~24. (1) Prove that g (9, 5) = li
(2) Prove that m (9 .. 5) ~ 5
5.2.25. Let an (n, k) code have a code distance d
ls 1t true that there ex1sts an (n, k)-cod e wt th a code drs-
tance d - 1')
5.3~ Alphabet Coding

Let A he an atptlabet, then A • 1s a set o( all tinLta
words In the alphabet A, tnclud•ng an empty word The
length (number of letters) of a word w IS denoted by A (w)
An empty word 1s denoted by A A concatenatlon ojworda
w 1 and w, obtained by wrtttng th.e word w2 sus' to the
right of the word w 1 1s denoted by w 1w2 The word W1
1s called the prefiz, and the l\ord w2 the suffix of the word
W1Wt The prefix w 1 (suffix W2 ) of the word w1w2 1s called
proper If w1 ::# A and w, -=#-A stmultaneously A v. ord v
1s known as a subword oj a word w 1f there extst \\Ords u1
and u 2 such that w = u1vuJ
Let A =- {a.1 , a~, , atJI} be a message alphaoot
and B a cod1ng alphabet Let ~ he a Single valued rnap-.
ptng of letters of the alphabet A onto B• The cod1ng of
-words 1n the alphabet A when each \\ ord (mes.sage)
5.3. ALPHABETIC CODING 165
a1,a1, ••• a1k is juxtaposed to a word cp (a;,) cp (a 1,} •••

cp {a;k) is called alphabetic (or letter) coding. An alpha-
betic coding is completely determined by the mapping cp
generating it and is denoted by K,.,. The set {cp (a}: a E A}
is called an alphabet code and is denoted by cp (A}. An
alphabetic coding K,., and the corresponding code cp (A}
are assumed to be uniquely decoded or divisible if each
relation of the type
cp (a;,) cp (a;,) ... cp (a;k} = cp (ai 1) cp (ai) ... cp (ai)
for the words in the coding alphabet B implies that l = k

and it = it (t = 1, k). A code cp (A} is called a prefix
code if none of the words in cp (A} is the beginning of
any other word in cp (A}. A divisible alphabet code cp (A}
is assumed to be complete if for every word w in a coding
alphabet B the following statement is valid: either w
is a proper prefix of a certain word in cp (A}, or a certain
word in cp (A} is (not necessarily proper) a prefix of the
word w.
One of the algorithms of detecting an alphabet code
divisibility consists in the following. Let cp (A) =
{wl, w2, ... , Wm} be an alphabet code. Let sl be the
set of all proper suffixes of code words and S 2 the set of all
words each of which is a prefix of a code word. Let us
consider an oriented multigraph G,., whose vertices are
the elements of the set S = (S 1 n S 2 ) U {A}. Let o
and 1: be two different vertices in S. The arc joining a
and ,; in the graph G,., exists if and only if there exist
such a code word w and such a sequence P = w1,,
w;,, ... , w;k of code words that wand aw 1,w1, •.• w 1k-r
are equivalent as the words in a coding alphabet. More-
over, if a =1= A, the sequence P may be empty. The arc
joining a with -r is assigned the word w;, w;, . . . w 1 •
The multigraph G,., has no loops and is referred to as t~e
graph of the alphabet code cp. The following theorem is
valid.
Theorem 1 (Al. A. Markov). A code cp is divisible if and
only if the graph G,., contains no contours passing through
the vertex A.
Example 1. Let cp (A) = {cc, cca, bcca, aa, ab }. Then
S = {A, c, cca, a, b }· The graph Gcp is shown in
100 CH 5o fUNDAMENTALS OP CODtNG TltEORY
F1g 22a There ex1sts a contour 1n G" pass1ng through

the vertices A~ a and b By wrt t1ng the words ass1gned to
the vert1ces and arcs of thts contour, we obta1n a word
whlc.h ~an be decoded 1n two ways
ccabcca = (cca) (bcl"a) = (cc) (ab) (cca)
Example 2. Let cp (A) -= {at abt acbb, bb, bbacc}
Then S = (A, b, bb) The graph G;p shown 1n Ftg 22b
has no contours passtng through A The code ts dlvtstble
Let us suppose that we llave a message source wh1ch
CODSCCU\lV ely generateS at Tandom t.hl) letters of the
a b
ceo
(a)
Flg 22
alphabet A = {a1 , a 2 t , a01 } It IS assumed that the
emergence of letters of the alphabet A 11 stat1st1cally
Independent and obeys the probability d1.str1button
m
P ={Pit Pt~ , Pm}t PI> 0, ~ p~ ~ 1 Any binary
i-l
alphabet code C = (w 1 , w2 t , Wm} can be assoe1ated
Wllh a number
tn
Zc (P) === )J P1"- (wl),

1~1
whtch ts called the cost of the code C for the dlstrlbUtlon

P The number :tc (P) 1s eqnal to the average n11mber of
letters of the cod•ng alphabet per letter of the alphabet
A The prefix code C0 1s assumed to he opt~mal for ths
dutrzbut,Dn P tf Z c., (P) -= 1n£ Z c (P)'I) where the lower
c
bound IS taken over the set of prefix b1nary codes con·
SlStlng of m words
~.3. ALPHABETIC CODING i67
The Huffman method for constructing an optimal code

is based on the following theorem.
Theorem 2. If C = (w11 w2, ' •• , Wm) is an optimal
binary code for the distribution P = {p 1 , p 2 , • • • , Pm} and
Pi = q1 + q2, where P1 ~ P2 ~ • · ; ~ PJ-1 ~Pi ~
PJ+t ~ .•• ~ Pm ~ qi ~ qJ, the code C = {w1, w2 , ••• ,
WJ_ 1, WJTl• ••• , Wm, w;O, WJ1} is optimal for the distri-
bution
The code C' is known as an extension of the optimal cixle C.

The Huffman method consists in the following. Let
Pm- 1 and Pm be the last two probabilities in the original
list of probabilities compiled in a non-increasing order.
These probabilities are excluded from the list, and their
sum is inserted into the list in such a way that the prob-
abilities in the new list are arranged in a non-increasing
order. This procedure is repeated until we obtain a list
of two probabilities, in which symbol 0 is assigned to one
probability and symbol1 to the other (this is the optimal
code for a two-letter alphabet of messages for any prob-
ability distribution). Then, in accordance with the
theorem, an optimal code is constructed for three letters
for the corresponding list of probabilities, and so on until
we obtain an optimal code for the original list of prob-
abilities. The Huffman method is illustrated by the
following example:
•
l
1 0.4 0.4 0.4 0.6 0 1 1 1
2 0.2 0.2 0.4 0.4 1 01 000
0.2 0.2 0. 001

•
4 0.2
•
168 Cll 5 FUNDAJ.lENTALS OF CODING THt:ORY
For a g1ven probabtl1ty dt5tribution, there also ex1st

other opttmal codes such as
Table 6
i .P, 1Dt tclt i ,, IDi ~
tD"'
t 04 t 00 4 0 t 0010 tto
2 02 01 Ot 5 0 t 0011 tit
3 02 000 to
Fano's method of constructing codes close to opt1mal

.consists 10 the followtng A l1st of proba b1l1 ties comp1led
tn a non 1ncreasiog order ts divtded tnto two (cousecuttve)
parts so that the sums of the probabtlitJes consttluttng
these parts are as close .as po.s5I ble Each letter of the
message alphabet tDnespondlng to a probab1l1ty 1n the
first part Is ass1gned the symbol 0 (or f)~ anlt the re
mainlng letters are asstgned the symbol 1 (resp 0) Thts
procedure IS conttnued on these parts as loog .as .a part
-conta.lns at least tv.o prahab1l1t1es \Ve ecntln\le 1n tbe
same manner unt1l the enttre 1tst IS d1v1ded 111to parts
containing one probabtl1ty each The examples of codes
construeted by Fano's method are gtven tn Table 6
5 3 I Us1ng a g1ven alphabet code cp (A), construct the

graph G(f and find out v;betber or not the code IS dtvlSlhle
(1) cp (A) = {ab, dct a, bcadd ca) A = {z J ~ t 5)
_ !2) ~ (A) = {dda.c~ dd~ cddab., a~ Cddd, b }~ A ~ {t, t .e::
1' 6}'
(3) f9 (A) ~ {a abt acbb abb, bbacc J, A == {l ' :=
--
f 5),
(4) v- (A) ~ {abc, bbc, bcb, ella, acbbt cbtb bctabb
abcacbbb J, A === {i~ I = -1, J~ rs
(5) q> (A) =- {abc abb, bcc1 ccaa, bcabbb(c, bbccaaabca,
abcabbabbbcca )~ A ~ {i~ i = 1, 7} t
(6) (J) (A) = {ab, bb, ca cba~ abb, bac" aabc, cabba},
A =- {i, 1 -=- 1., "8 }
5 3 2. Let the numbers f '* 21: 4,. 17 all£\ 98 be encoded
by the1r binary dtvts1ons of the m1n1mum poss1ble length
5.3. ALPHABETIC CODING
For example, the code of unity is 1, the code of two~

is 10, and the code of four is 100. Is this coding div-
isible?
5.3.3. Using a given indivisible code <p (A) = {aa, ab,
cc, cca, bcca} and the word IV in the coding alphabet B =
{a, b, c}, fmd out whether the word w is a code of a
message. If this is the case, find out whether the word w
is a code of exactly one message:
(1) IV = ccabccabccabcc;
(2) IV = bccaccabccabccacabcca;
(3) IV = abbccaccabccaabab.
5.3.4. Let <p (A) be an alphabet code and G.., be a graph
of this code. Let a graph G~ be obtained from G.., by delet-
ing all the vertices which are code words. Will Theorem 1
remain valid if G~ is substituted for Grp? .
5.3.5. Let <p (A) be an alphabet code and G.., the graph
of this code. Let a graph G~ be obtained from G<j) by delet-
ing all the arcs which are labelled by A and which ter-
minate at the vertex A. Will Theorem 1 remain in force
if we substitute G~ for Gcr?
5.3.6. For a given divisible code <p (A), construct a
prefix code with the same set of lengths of the code words:
(1) <p (A) = {01, 10, 100, 111, 011 };
(2) <p (A) = {1, 10, 100, 0100};
(3) <p (A)= {10, 101, 111, 1011}.
5.3.7. Let <p (A) be an alphabet code of power m, in
which the sum of the lengths of the code words is equal
to N, and the maximum length of the code words is l.
Using Theorem ·1, prove that the code <p (A) is divisible
if and only if each word in the coding alphabet having
a length not larger than (l - 1) (N - m + 1) + 1 is
either a code of exactly one word composed of the letters
of the alphabet of messages A, or is not a code of any
word in A*.
5.3.8. Let k be the minimum and l the maximum length
of the code words of an alphabet code <p (A) and let N
~e the sum of the lengths of the code words. Prove that
~n .order to establish the divisibility of the code <p (A),
It IS sufficient to verify the uniqueness of decoding of all
codes of words whose length in the alphabet of messages
does not exceed Nllk.
5.3.9: We shall assume that two words IV and v in the
alphabet {0, 1} are equivalent if there exists a word u
t 70 CH 5 FUNDA.JtE'fTA.LS. OP CODING TliEOJlY
wb1ch can be obta1ned bath from w and from v by us1ng

the follow1ng fi.n1 te number of operat1on.s
(a) delet1ng snhVtords -of the form 10 and tOOt,
(b) 1nserttng words of the form 10 and 100f 1nto the
spaces bet ween letters 1
Are the follow1ng pa1rs of words w and v equ1valent
(1) w ~ 1010101 t v == 0101010.
(2) w = 10010010 tt :=: 010010010 ~
(3) w = 11011010, v :== 01101101?
5.3 ~I 0. Let AI be a set constst1ng of m non empty words
tn au alphttbet A conststtng of k letters Prove that
(1) there e:x1sts a word tn Af, whose length ts not le:s!
than log11. {1 + rn (k - 1))
(2) for any e > 0, the fraetton of the words in AI,
whose length IS smaller than (1 - e) logA (1 + m (k - i)),
does not exceed { 4- }J -Em e for m ~ 2t k ~ 2
5 3 .. t f • Let each word 1n an alphabet b1na ry eo de C
of power 2"' + 1 ltave a length not exceeding n
(1l Prove that the code C ts not a prefix code
(2) Can the code C be dtVl!nble:?
5 3~ t2. For glven prohabtl1ty dtstrtbuttoll of occurrence
of letters, construct optimal codes by Huffman's method
(1) P t=; (0 34 0 i B, 0 17 , 0 16 0 t 5) ,
(2) p == (0 6, 0 11 0 09 0 08 0 07 t 0 06),
(3) p 1:=+ (0 4 t 0 4' 0 1 ' 0 03 t 0 03 0 02 f 0 02)'
(4) p := (0 3, 0 2t 0 2, 0 t, 0 t1 0 05, 0 05)
5 3 13~ (t) Construct the codes for the probabtlaty
dlstributlons of the prevJous problem by ustng Fano's
method
{2) Give an example of a probahlltty dlslr1but1on for
'vh1ch a code constructed by Fano's method tS not optam.al
5.3.14.. Let C = {w1 w2 wmJ be an opltmal
btnary code corresponding to the d 1str1 bu t1on P =
(pli P2' , Pm)• Pt ~ P2 ~ ~ Pm Prove that
(t) ). {wt) ~ 1 (wJ) 1f P1 > PJ'
(2) the code C IS complete,
(3) there ex 1st t'' o code words of length A, (lum) wlttch
have Identical prefixes of length 1 (w,11.) - -l
5.3.15~ Prove that tf m 1s not a power of two there
ex\st two words of dl'tlerent lengths 1n the opt1mal btna.ty
1t JJ also allowed. to wt1 te tbe wo.rd s 10 and 100 I Just to tb e
1
right and. to the left o [ t h.e word b~tng traD.! forDled
5.3. ALPHABETIC CODING 171
code for any probability distribution P = (pl, p 2 , • • • ,

Pm)•
5.3.16. Using the solutions of Problems 5.3.14 and
5.3.15 and the definition of an optimal code, explain
why the following codes are not optimal for given prob-
ability distributions:
(1J (2) (3)
to. WI l'j 10.

Pj I PI I
0.6 0 0.6 1 0.2 000

0.2 10 0.2 01 0.2 001
0.1 11 0.15 001 \ 0.2 010
0.1 01 0.05 0001 0.2 011
0.2 100
5.3.17. Determine the smallest m and a probability

distribution P = (p 1 , p 2 , • • • , Pm) for which there exist
optimal codes differing in tuples of the lengths of the
code words.
5.3.18. (1) Prove that the maximum length of a code
word in an optimal code of power rn does not exceed
m-1.
(2) Prove that for any integer ni (m ~ 2) there exists
a probability distribution P = (p1 , p 2 , • • • , Pm) satis-
fying the following condition: there exists an optimal
code corresponding to the distribution P such that the
maximum length of the code word in it is equal tom - 1.
5.3.19. Prove that there exist not more than ( m"~ )
2 1
• complete prefix binary codes of power m .

5.3.20. A code is referred to as almost uniform if the
lengths of its code words differ by not more than unity.
Prove that for any natural m a nearly uniform code is
optimal for the distribution P = (1/m, 1/m, ... , 1/m).
Is the converse true?
5.3.21. (1) Prove (by induction in m) that the sum of
the lengths of the code words in an optimal code with m
messages does not exceed .} (m. + 2) (m - 1), ~ ~ 2.
w
•
172 CH s .FUl\DAMENT.ALS OP COlll"lG THEORY
..
(2) Prove that for any IDtegsr m ~ 2 f the est1m ate Ill
the prev1 ous ptobl em IS accessible
5 .. 3 22. Let g 1 , g 1 , , gm be arh1tr.ary pairWise dts--
JOint faces of the cube Bn. with dimen.slonalttles r 11
r,t , rm. respectively
m
( 1) U sJ ng t}Je obVIous 1nequa)I ty ~ 2,.i ~ 2-n prove t
l~l
that the Jengths A:(w 1), h.(w 2), , h. (rom) of the -vwords
of an ar h1 trary htoary pre:fix code C ~ {w1 Wz, , Wm}
,.,.
satlsfy the condttton ~ 2-Mwi)~t
a-1
(2) Prove that a complete prefix code satJsfies the
1D
equal1ty ~ 2-u~,~== 1
1~1
m
(3) Prove that J f ~ 2-l., ~ 1.. where 1 1 are nat ural
__.........,...__ "(::;;;: t
numbers. ':==: 1., mt there exJsts a prefi:t code wtth
words of Iengths,.A1 , l 2 , , Am
5 3~23 .. (1) 18 1t true that the nuntber of code words
of max1mum length 111 an optimal b1nary code ts even?
(2) Is 1t true that tbts number IS a pol\ er of two?
5.3.24. . Let C -.= {wt, w 2 , t wm) he a complet-e
prefix h1nary code and let h (w 1) ;;=: h (w 1 +1 ) for all
i ==: 1, m - 1 Prove that for m ~ 2 and ail t :=:
t, m - 1 the lnCqua1tty 1 (wl+.rl - l (w 1) ~ 1og2 m - 1
IS valid
5 ~3 .25 • Let Lm be the smallest 1nteger for which
there e1: tsts a btnary prefix code of poVt er m, WJ th the
sum of the lengths of the code "ords equal to Lm Prove
that Lrn ~ m (Jog1 mJ lor m ~ 2
5.3 .26.. Pro' e that the lengths of the code words A(w1} 't
A. (w 2 )~ , l (Wn1 } of an optttnaJ btnary code C =

nt
{wl., W2, , Wrn) sattsfy the cond1t1on 2} 2 ltw.t) =1
I~ I
5 ..3.27 .. Let P == (pl, Pt• , Pm) be a probabtllty
distribution (P 1 > 0, ' = 1, m), and :e
(P) ~ IDf Zc (P).
c
where the lower hound ls taken over all b1nary prefix
cades of powe~ m (m. ;;.=:. 3)
(t) Prove. that Z (P) > t~
5.3. ALPHABETIC CODING :173
(2) Prove that for any e > 0 and any integer m ~ 1,

there exists a probability distribution P = (p 1 , p 2 , • • • ,
Pm) such that X (P) < 1 + e.
Shannon's method for constructing almost optimal
codes consists in the following.
m
Let P=(p 1, P2• . · ., Pm). Pi>O, LJ
i=l
P1 = 1, Pi~Pi+t•
i= t, m-1, be the probability distribution for the
occurrence of the letter!> in the alphabet of messages
A= {a 1, a 2 , ••• , am}· Then a letter· ai is assigned a code
word of length l 1 = log J ;i [ , composed of the first
(after the decimal point) zi digits of the decomposition
i -I
of a number qi-1 = ~ Pi into an infinite binary fraction
i=l
(rounded down).
For example, if P = (0.4, 0.3, 0.3), the code of the
letter a1 is 00, the code of a 2 is 01, and the code of a 3
is 10.
5.3.28. Using Shannon's method, construct the codes

for the probability distributions in Problems 5.3.12 (1)
and (2).
5.3.29. Indicate the smallest m for which there exists
a probability distribution P = (p 1 , p 2 , • • • , Pm) such
that the code constructed by Shannon's method for the
given probability distribution is not optimal.
5.3.30. Prove that the code constructed by Shannon's
method is a prefix code.
5.3.31. Let X* (m) = sup inf Xc (P), where the up-
P C
per bound is taken over all distributions P = (p 1 , p 2 , ••• ,
m
Pm) such that Pi> 0, i = 1, m, LJ
i=l
Pi = 1.
(1) Prove that X* (m) ~ Uog 2 ml.
(2) Prove that X* (m) ~ [log 2 ml + '1. ',
Chapter S1x
F1n1fe Automatons
6. f,. Determtnate
and Boundedly Determ1nate Fu nct1ons
Let A he a non empty f1n1le alpl)abet Th.e elements o!
the alphabet are called letters (or symbols) A word tn an
alphabet A 1s a sequence formed by the letters of th1s
alphabet The length of a uord lo (the number of letters 1n
.1'¥
the word) ts denoted by '-. (w) The set of all words x• -

z (l) z (2) z (s) of length s (s ~ 1) to the alphabet
A \Vill be denoted by A • A word of length 0 (empty u~ord)
1s denoted by the symbol A By A• we denote the set
{A} U UA" whlle A" denotes the aet of aU words
,fl~ t
;(,J -
x {1) x (2) where x (t) € A t ~ 1 2 The
words ln the set A w are called ~nfin~te lVords tn the alpha
betA
The word w obtatned by \vr1t1ng a word w1 to the right
Clf a f1n1te (or empty) word w 1 1s called a concatenat~on oj
words w1 au.d u. 2 and IS denoted hy w1 w 2 The word w1
IS called the beg~nntng (prefix) and wa tl1e end (sujfiz)
of the word w
Let A and B be finJte non empty alphabets The map
p1ng ~ A(I) - ) B(J) 1s called a determlnate junction or a
determtnate operator (tn abbrev1ated Corm a d futtetwn
or ad operator) 1f 1t satJs.fies the following condttlon
1'-W
for
,_
any s ~ 1 the s th symbol y (s) 1n a word y(iJ =
f9 (~) JS a stngle valued functton of ,..,the first s syrnboJs
x (1} z (2} z (s) of the word zlll
..,.
[( the "ords zOl and
~
r:
have identical prefixes of length
("¥ #¥ ~ ,.,;
s (s ~ 1) tfle words YT =:. \ll {rt) and y~ t = q> (J::) l1ave

also 1deDt1cal prefixes of length s
The set of all d funct1ons of the type rp A w - ~ BW w1II
he denoted by Ill A s
•
•
li.l. DETERMINATE FUNCTIONS ti5
If A = A 1 X A 2 X ..• X An and B = B1 X B2 X
••• X Bm, then the mapping cp: All) ,. Bill induces
m functions each of which depends on n variables, the
variable X1 running through the set Aj (i = 1, 2, ... ,
n). These functions are defined as follows. Let ,...,xlll =
-
x (1) x (2) . . . .c: (t) . . . be a word in A Ill and yll) =
cp (xlll) = y (1) y (2) ... y (t) . . . . In this case, x (t) =
"'Ill
(x1 (t), x 2 (t), ... , Xn (t)), XJ (t) E A}t xi =
1
,...., ,..., ""'J ,...,
x1 (1) x1 (2) ... x 1 (t) . . . , xlll=(x~, x~, ... , x~), y(t)=

....,
yi=
(yl (t), Y2 (t), ... , Ym (t)), Yt (t) E Bt.
,..., ,...,
Yt (1) Yt (2) · · • Yt (t) · .. , ylll = (y~. y~, · .. , YWt),
- ,....;
- - ,..., ,...,
!pt (x~, x~, ... , x~) = yf.
By proceeding in the reverse order, we can. construct
the mapping - 8"'.
The variable X 1 of a function <p (X 1 , X 2 , ••• , Xn):
(A 1 X A 2 X . . • X An )Ill - Bill is called essential if
,...., "' ,...., ,...., ,...,
(x~11 x~11
... , x~,) and (x~2 , x~11
-
there are two tuples
.
. . . , x~,) of values of the variables X 11 X 2 , • • • , Xn
which differ only in their first components and such
f"tJ ,...., I'V ,..., ,..., ,...,
that !jl (X~II X:\1, • • • , X~ 1 ) =fo !p (X~2 , X~p • • • 1 X~ 1) • Jf

the variable X 1 is not essential, it is called fictitious.
All other essential and fictitious variables X 1 on which
a given function depends are defmed in an identical man-
ner.
The function cp (X 1 , X 2 , • • • , Xn) is said to depend
essentially (fictitiously) on a variable X 1 (1 ~ i ~ n) if
this variable is an essential (resp. fictitious) variable of
the function IP·
If A is a set of all vectors of length n with elements
in En, and B is a set of all vectors of length m with ele-
ments in E 11 we shall use the notation £D~;;n. instead of
ll>A.B· For n = m = 1, the superscripts in <l>k'i will
be omitted. If k = l, we shall write only one sub'script:
mn 1 m
""'h •
!76 eft 6 F lNtTF.: .AUTOMA TO I\ ct
Sometimes It IS convenient to assume that a d funct1on

rp In l!J A 8 1s represented by a dtstrlte devl£e (automaton)
!., operat1ng at discrete lDStants of t1me t ~ 1 2
At each 1nstant t of ttme a stgnal z (t) 1s suppl1ed lo the
lnput of th1s de\ 1ce and a slgnal y {t} appears at the nut ,-w
p11 t (F1g 23) !he "ords ~

"""'
are called tnput li)ord:, and y~
r .. tl1e output words The alpha
bets A an.d B are called lnput
Fllf 23 o.rvl Qu.tpu.t o.lpkabel$ of the
automaton 21~ If A = A 1 X
A1 X X An and B = B 1 X B 2 X X Bm It can
be assumed that the automaton 21~ represents m determ1
nate functtons each of whtch depends on n vartables
Any twD funet1ons 'jJ1 and tp 2 tn <1> A B are called dis f'IW
t~nguuhable 1f there ex1sts an Input word x~ "h1ch IS

processed 1nto d1fierent output words b} them 1 e
~1 (.x:)
1'¥
,..., *' 1
f"'V
(~) If, however. the eq

,...,.
ual1 ty fl' 1
J"'ftJ
(xw) ;;;:::
-r.p 1 (J:M) 1s satisfied for any Input word zw tl1en fP 1 and rp 2
are called equfvalent or tndlstlngulshable d juncttona
Let tp E$A s and t E $A n If there extsts a word
,.,.,.. ,...., ,..., ,.. ,...,
X: E A • sucl:t that t rp (x~) = cp (:c:) 1p: (ZW) for any word
i"W'
xw E At.~ the opera tor 1p Is called a resulual operator of the

operator r.p generated by the word z: and lS denoted by
~
J'v
(J);. The set Q (q> x~) of all res~duat operf.ltors of the

0
opera tor --F' eq 111 valent to the opera tor f./>"""
-x.J
forms an
I
eqv.tvalence. cla.u called the state of operator {}) conta.tntng

a resldual operator c.p.....
:rl
The state conta1n1ng the operator
' .,..,
qJ IS called the lnttial state € Q (rp X:), the operator lp
If ,p
~
JS satd to be represented by the state Q ('ll ~:> of the oper

a tor Q' The opera tor rp ts called a boundedly determ1.nate.
operator (abbreviated as a b d operator or a b d functton)
1f 1t has a fintte number of pa1rw1se different states The
, .... ----
,..,
1 H ert " (.r:) den ot:es a prefa of I engt h 1 to the output word
r¥./fi/V
'(~~.w)
G.l. DETERM INATE FUNCTIONS 177
nmube r of differe nt states of a b.d.-fu nction is known as

its u;eight. If the pairwi se differe nt states of an operat or cp
form an infinit e set. 1ve shall assum e that the weigh t of
"
the operat or cp is equal to oo. B~~ cD A. n we shall denote
.t.
the set of all functi ons in (I> n that are b. d.-fun ctions .
It is conve nient to analyz e rl-func tions by consid ering
their graphi c. repres entatio n in the form of infinite in/or-
• •
• •
• •
•
••
Fig. 24 Fig. 25
mation trees. Let A be an alphab et of n letters . By D A

we shall denote an infinit e orient ed rooted tree satisfy ing
the follow ing condit ions:
(a} the out-de gree of each vertex , includ ing a root, is
equal to n;
(b) the in-d-egree of a root is equal to zero, while every
other vertex has an in-deg ree 1;
(c) each arc of the tree D A is assign ed a letter of the
alphab et A, and differe nt arcs emana ting from the same
vertex of the tree (in partic ular from a root) are assign ed
differe nt letters . A root of the tree is assum ed to be a zero
rank vertex. If a vertex v is the end of an arc emana ting
from an i-th rank vertex (i ~ 0}, it is called a vertex of
(i + 1)-th rank. An arc of j-th level (i ~ 1) is an arc
emana ting from the (j - 1)-th rank vertex. Each infinit e
oriente d chain in the tree D A corres ponds to a word in
A(~. Figure 24 shows a fragme nt of the treeD A, A = {0, 1 },
consis ting of the first three levels of this tree (here and
below, we assum e that the left and right arcs emana ting
from a vertex are assign ed the symbo ls 0 and 1 respec tive-
~y). Thick lines are used to isolate the chain corres pond-
lug to the word 101 ·
12-0636
t78
A loaded tret D A 8 1s obtained from the tree D A by

assJgn1ng a letter from the alphabet B to each arc Each
or1ented Infintte chain 10 the tree D A B bas a correspond
1Dg v. ord l!l Bw formed by the letters as.s1gned to the
arcs of thts cha1n l-Ienee 1t can he assumed that the
loaded tree D A B defines (represents) the mapping
tp A fJ) . . ,.~ BoJ "b.1eh 1s a d funt.tlon Tl~e fragment uf a
loaded treeD A B' \\here A = {0 1} and B = (0 1, 2},
IS shown tn F1g 25 The d funet1on corresponding to th1s
tree processes for example" the word 1010 tnto the word
2012
Let D A be a loaded tree representing the d function cp
8
The testdual operator ~~. (.r ~ 0) of the operator r.p has
,.., ;;!:"
subtree D , JJ (x;) growing from an s th rank vertex
a ,......
v (X:) at wh1eh the t.ha\n st~rt1ng from the root and c.on
ta1n1ng e'(actly s ares termJnates In the 1 th level thts
chaln has a corresponding arc labelled by the Jetter
z, (t} E A
If the restdual operators cp~'X., and fJ'..v.
'l::lt'
are equ1valent,
f'W' I ,.., I I '
the vert1ces v (zt l and v (z2!) eorrespondtng to them and
1
the .subtrees gro\\ tng from these vert tees are also called
equivalent The we1ght of a tree representing a d funct1on
IS equal to the we1gbt of thts functJon and hence to the
max1mum number of pa1rw1se non equ1valent vert1ce.s
(or subtrees) of the g1 ven tree
6 t f, Ftnd out 1f the mapp1ng rp {0, 1 }~ -... {01 1}~

1s a d funet1on
(i) «p (x (f) .t (2) ) ~ z (2) x (3) 1 e y (t) =
% (t + 1). t
,..,;
~ 1,
(2) f (xw) === 10100100010 010 010 for any
L , •
I:-.,
,., n t 01fS
1n p:u lr word zw,

(3) 'P (z (f) Z (2) X (3) ) = X (1) X (2) X (J) X (2)~ (J)
, 1 e y (I) -= z (t), y (2) = x (2) and y (t) -=
~ (t · -- 2) for t ~ 3,
(4) q> (z (t) z (2) ) = Ox (1 z (2)) x (1 +
x (3)) +
• J u y (1) = o~ Jl (t) = z (t ::t (t)) for t ~ 2+
6.1. DETERMINATE FU:-;CTIO:-;s 179
6.1.2. Isthefunctioncp: {0, 1}w ~ {0, 1}wdescribed

by the following description a d-function?
-
(1) The word xw = x (1) x (2) ... is processed into
-
a word ow = 000 . . . 0 ... , if there exists a t such that
-
x (t) = 0; otherwise, cp (xw) = 1w = 1'11 ... 1 . . . . -
~
(2) The s-th letter y (s) in a word y"' = cp (xw) is equal
-
to zero if the inequality 3x (t) ::;;;; x (t 1) x (t + 2) + +
is satisfied for a certain t =:;;;; s; otherwise, y (s) = 1.
~
(3) The s-th letter y (s) in a word yw = cp (xw) is equal
-
to 0 if there exists a t > s such that x (t) ::;;;; x (s); other-
wise, y (s) = 1.
(4) y (1) = 0, and for s;::: 2 the prefix y (1) y (2) ...
-
y (s) in the word yw = cp (xw) contains one more zero -
than in the word x (2) x (3) .•. x (s).
Each word xw = x (1) x (2) - . . . x (t) . . . fr9m
,...,
{0, 1 }w corresponds to a number v (xw) in the segment
[0, 11. The binary expansion of this number is 0, x (1)
x (2) ... x (t) . . . . If a E [0, 1), its binary expan-
sion2 0, a1a2 • • • a1 ••• generates the word (a) =
x (1) x (2) ... x (t) ... , where x (t) = at(t ;::: 1).
6.1.3. Find out if the function cp: {0, 1 }w ) {0, 1 }w
is a determinate function:
"'
(1) cp(xw)=(1/3};
-
(2) cp (xw) = (7/15};
-
(3) cp (xu')= (2/7);
(4) cp (;w) ={f2/2);
(5) cp (;w) = (1/Y3}.
6.1.4*. Find out if the function cp (X 1 , X 2 , • • • , Xn):
{0, ·l}W X {0, l}w X . . . X {0, ·l}w r {0, ·l}w is de-
n t lmes
terminate:
,..., ,..., "'
,..., ,.., x~ if v(x~)~v(x~),
(1) cp (x~, x~) = ,..,·
x~ otherwise;
2
• If c; = p/211
;> 1), we consider a binary expansion contain-
(n
Ing an mfimtely large number of zeros (p < 211 ). If a= 1, then,
by definition, (a) = iw.
{8Q CH 6 FINITE AUTOMATONS
other\\ 1se,
tf v <r:l + 1J (.z:) ~"(XC:}~
,_ ,..,.,; ,..., i/flllW
~ ~ ~ lw
(4) fP (z~t 3.~, x~) =:=: ~
otberw1se z:
6. t .,5. The partial fune ttons cp {0, 1 }~ - )' {0, 1) b)
f"'J
g1ven below are not defined only on th~ l\ ord ~=

00 0 Wh1ch of these funet1ons can he red~
fined to determ1nate funet1ons and whttJh ()nes cannot1
,..,
zw 1f the :number of zeros tO each prefix
lfll.,i
{1) 4Jl (~) = f of the word zi'l lS not less than the
num ~~ a,{ un\\,\t:"S,
~
1n~ otherwtse•
..,.
(2} <p (~J == y (1) y (2) y (t) • where
0, tf 3s ((s~t) & (x (s) ;=;f)),
y (t) = 1 otherw•se,
.f/'¥
(3) rp (x'd)= y (1) y (2) y (t), where

f 1f for a cert.a.1n s ~t 1n the prefix
t) :=::;: z (f) z (2) x (s), the number of zeros ts
Y( larger than the number ol un1 t1esf
z (t) otherwtSe,
,.....
(4) rp (xw) = 11 (1) y (2} y (t) , where
1 1f ~(z(1)x(2) :z: (t) 00 0 )~t/2,
y (t) = 0 otherw)se
6~f .6* ~(I) Dlsprove the Collow1ng statement 1f the
funct1on r.p (Xi, X 2) {0, 1}w x {0 1 1 }(I) ~ {0, t} r.l de-
pends essent1al yon thevartable X 1 and Jf, for each (fixed)
,.., (V
word ~ E {0, 1 J0 .,
'P (X1 .r:) 1s a d-funct1on, the func·
J:
t1on q; (X1 • X 1) 1s also a d funct1on

(2) Let the funct1on ~ (X 1 ~ X,_} {0~ 1 )~ X {0, 1 }td ~
6.1. DETERM INATE FUNCTIONS 181
{0, 1 }(I) satisfy the follow ing condit ion: for any words
~ ~ ~
x~ and x~ in (0, 1}(1), the functi ons rp (X 11 x~) and
~
rp (x~, X 2 ) are determ inate. Will the functi on rp (Xu X 2 )

be determ inate as well?
6.1.7. Const ruct a fragm ent of a loaded tree contai ning
~ ~
s first levels for the functi on y(l) = rp (x(l)) belong ing to the
set <1> 2 :
(1} y (1} = 1 and y (t} = x (t - 1) for t ~ 2, s = 3;
(2} y (1} = 0 and y (t} = x (t) EB y (t - 1) for t ~ 2,
s = 4; .
(3) cp (;;(I)) = <~) , s = 3;
(4) cp (';(I))=--< ~ ) , s = 4.
~
6.1.8. For a given functi on cp (x(l)) E <1> 2 , repre-

sent in a loaded tree a chain corres pondin g to the
~ ~ ...., 3
prefix xs of the input word x(l) and write the prefix Y
....,
of the outpu t word y(l):
....,
(1) cp (x(l)) = 10100100010 .. . i.e. y (t) = 1 only for
•
L ....,
t= , i = 2, 3, . . . , x7 = 0101001;
2
(2) IJl {x(l)) = (
~
! ) , (a) ;'1 = 1111101,
(b) x7 = 1010110;
if X(1) +X (2) + ... +X (t) > t/2,
1
(3) y (t) = 0 otherw ise,
"' 0 ...., 0
(a) xt = 0101010110, (b) xi = 1100101110.
6.1.9.....,The loaded tree corres pondin g to a certai n func-
tion cp (x(l)) E <1> 2 has the following form: the left and
right arcs emana ting from a root are labelle d by 0 and 1
respec tively; if v is an i-th rank vertex (i ~ 1) and the
i-th level arc termin ating at the vertex v is labelle d by
cr E {0, 1 }, the left arc emana ting from v is labelle d by
the same symbo l (J, while the right arc is labelle d by
(negation of a).
a
(1) Is it true that the s-th letter of the outpu t =word
182 CH 6 FI,ITE AU1'0MAT0"'S
i"W ,.,.
y• := ft' (~} for th1s funetton ~an be determ1ned fron\ the

relatton
(a) v (1) = x (1)~ y (s) == ::& f) ED .z (s) for s ~ 2.
(1 -
(b) !J {i) =:X (1)t y (2) ~ :t (1) G) X (2)1t Y (s) ::
z (B- 2) $ z (s) for s ~ 3,
(c g (1) =
z (1) m :x (2) e ED ,; (s) 1
(2 Ftnd the wet.ght af the funct1on ~
6~ t .. fO. Are the res1dual operators rp.,... and rp"'V of
%1 :r;;r
0 I
the d funetton rp E$ 2 equtvalent?
,.,
(1) <p (~) = 10100100010 1 e y (t) = 1 only for
t ....
t= 2 • 1 = 2. 3, z: = 101,
.r!::::: fQ z: == 00101.
...... 8 ""w' ~
(2) ~ (xw) ~ iS ~
{3) fJI (~) = ; . ;~ = l .......-~._

$1 _.
OOt '
'f"'oo:
(4) «p (z~) == y (1) y (2) and

0 11 z(t)+z(2)+ +x (t) < tl2~
!I {I)--=
(1) ~
6.1. DETERMINATE FUNCTIONS 183
(4) q> (xw) = (!: ) • q>d;w) =

7
• 15 ) •
6.1.12. Find out if the function q> E <ll;l' m is a b.d.-
function, and determine its weight
"' y(1)=y(2)=1,
(1) q>(xw):
y(t)=x(t-2), t;;;::3;
y(2t)=x(t+1), t;;;::i,
-
y (2t-1) =X (t), t;;;::i;
y (1) = 1,
,..,
(3) q> (xw): y (2t-1) =X (2t-1) ED y (2t-3), t;;;::2,
y (2t) =X (2t-1), t;;;::1;
Yt (1) = X1 (1) ED X2 (1},
y 1 (t) = x 1 (t) EB x 2 (t} ® Y2 (t-1), t;;;:.2,
y2 (1) = x1 (1) x 2 (1),
y 2 (t) = xdt) x 2 (t) $ x.(t) y 2 (t -1)
EBx2 {t)y2 (t-1), t;;;::2;
yt(1)=1,
-
Yt(t)=y 2 (t-1), t;;;::2,
Y2 (1)=0,
Yz(t)=Ydt-1), t;;;:.2.
6.1.13*. Let DA,B be a loaded tree representing a
b. d.-function q>: A w )- Bw. Each vertex of the treeD ~1.. 8
is assigned a number equal to the weight of its subtree. We
-
obtain a new tree D A. B·
(1) For any r ~ 1, give an example of a b.d.-function
-
q>, such that each vertex of the tree D A, 8 corresponding
to the function q> is labelled by the number r.
(2) For any r ~ 2, give an example of a b.d.-function
such that for j = 0, 1, ... , r - 1, each j-th rank vertex
-
of the tree D ,t, B corresponding to the function q> is la-
belled by the number r - j.
-
6.1.14. (1) Prove that the tree D 8 construyted.t.
according to the conditions of Problem 6.1.13 has the
following property: the sequence of numbers v (v1 ),
v (v 2 ), • • • assigned to the vertices of the oriented chain
vl, (vl, V2), V2, (v 2, v3}, v3, ... (finite or infinite) is mo-
CH 6 FINITE AUTOl\.l.ATONS
noton1cally non tncreasing, and that th1s sequence IS

stab1l1zed lor an tnfinlte cbatn
(2) Show that any -res1dual operator of the function
~ E ¢A u has a wetght not exceeding the we1ght of the
function -tp ........
6.1 ~t5 . The tree representing the funct1on cp, (x(J}) E <b,
1s of the form the symbol 1 1s assigned onl) to the arcs
belonglng w the orlented chaln. emanat1ng from tl\e root
and correspondtng to the 1nput word 10100100010 .
(hero :r {t} = 1 only for t = ( ~ ) , ' = 2~ 3, }
the rema1n1ng arcs are labelled by the symbol 0 Prove
that the funCtiOn £Po bas an Infinite we1ght and lS therefore
not boundedly determtnate
6.t~t6., For each r ~ 2, obtatn an el.atnlllc o,( a b d-
functJon of we1gl1t r sati~fy1ng the follo\vtng condal1on
1n the loaded tree represent1ng th1s function, the symbol 1
ts a.ss1gned only to some (not necessarily all) arcs of a
slrtgle tnftntte ortented chain Z em.anattng from the root
The remaining arcs (not bel ongtng to the chain Z) "~~are
ass1gned th8,...symbol 0
The word :xtr:. E A 1s called qua.JL-ptTlOdtc 11 there ex 1st
tl)
lntegers n 0 and T, suclt that n 0 ~ 1 t T ~ 1, and x (n +

T) ~ z (n) for n ~ n, In this case~ the preflX x (t)
,..,.
z (2} 1) of the word .x~ ts called a pr«-perlod.,
x (n, -
the number n 0 - 1 the length of thls pre perlodto the word
:c (no) :c (n 0
~
+1) z (n 0 +
T -- 1) tl1e pertod of the
word r't and the number T the length of thls perlod
Such a quas1 per1odtc word can be conveniently pre.sented
1n the form
3, (1} x (2) {n 0 -
:.t 1) [.t (n 0) % (n 0 + 1)
z (n 0 + T - ~- 1)}f.O
row '*'
6 .1. 17. (1) Prove that 1£ the (unct1on cp E ClJ A n't then
any quas1·per1od1e word 1n A est 19 tran5formed by the
functton q. Into a quasi periodic word tn BIJJ
(2) Confin1ng yourself to the set $ 2 " show that the
statement converse to the one formulated 10 part (1) IS
not true
Hint. See Problem 6.1 15
6. t . 18 *. D1sprove the follow1ng sta tementt: t.f a d func-
6.1. DETERMIN ATE FUNCTION S 185
tion is also boundedl y determin ate.
6.1.19. Suppose that all the vertices of a loaded tree
are divided in the normal way into equivalen ce classes.
Prove that any equivalen ce class has a vertex v satisfyin g
the condition : all the vertices in an oriented chain ema-
nating from the root of a tree and terminat ing at the ver-
tex v are not pairwise equivale nt.
6.1.20. (1) Prove that for each vertex of a loaded tree
having a weight r, there exists an equivale nt vertex
having a rank not higher than r - 1.
(2) Can we replace r - t in the previous problem by
r - 2 if r ~ 2?
Let 11 (x 1 , x 2 , • • • , X 11 ), 12 (x 1 , x 2 , • • • , X 11 ), • • • ,
fm (x1 , x 2 , • • • , x 11 ) be a function of the type
Ek XEh X . . . X Ek ) £ 1, whereka nd l are not lower than
n times
2. The operator fPt •• 12. .... 1m in cD;::r is called an operator
generated by the Junctions I1> I 2 , ••• , 1m if the following
equalitie s hold for all t ~ 1:
Yi (t) = li (x 1 (t), X:3 (t), ... , Xn (t)),
i = '1, 2, ... , m.
6.1.21. Show that an operator in cD~· 1
1 is generated
(by a function of the_ type Eh X Eh X . '•• X Eh ) £ 1)
-----~--------~
• n times
If and only if its weight is equal to 1.
6.1.22. Find out if the operator rp E cD~· 2
defined by the
following relations is a generated operator:
Yd1) =0,
ydt) = x (t-1) EB y2 (t-1), t~2,
!/2 (t) =X (t).
6.1.23. A partial operator cp: {0, 1 }(J) ) {0, 1}w is
defined only at the input words ow
and i(J). Redefine it in
such a way that the resulting operator has the minimum
. possible weight.
~ ,....
(1) fj) (Ow) = 0101 [100}w, cp Wll) = 11 [10}w;
CU '6 11N ITE AUTQ:MA.'rO', S
of'tJ ,;tV
(2) ~ (Ob)) ~ Of {011 l(&)~ rp (1 (a)) =

10100100010000100
l
1 e y {t) = 1 only ior t ~ , 1 -=== 2" 3,.
2
6 1~24. (1} Prove that .tf (A l > 1 and t B I> t\ the
power of the set llJA. B ts equal to the po\\er c (of the
continuum)
(2) Ftnd the power of a set <b A a for the case when
e1 ther ( A r = 1 or I B ' = 1
6~t.2.i.~ The words
~
and
"'''
,....,
xr x:
ln AWare called s--equlva-
f'W
lents (notatton .r~ ,..; x:) If thetr prefixes of length s

I
(s ~ 1) are eqt1al fhe relation ~ 1s the equJvalence re-
I
lat1on and d1v1des the setA(&) 1ntoclasses KJ (s)of s~qUlV-
alent words
(i) Flnd the pov.er of eacb. class KJ (s) (.s lS a fixed
number)
(2) How many d1fferent classes KJ (s) extst for a fixed s?
(3) How many d1.fferent classes K 1, (s + I) ex1st tn
each class K 1 (s) (here l 1 )' >
(4) Prove that tlte mapp1ng '1J A w -• B~ lS a d fun.c-
f"V
t1on 1f and only 1f for all s ~ 1 and for any words

f't<tl (lal
xf
~
and X: belong1ng to the set A c.l' the relation ~ ~ xf

'
,.,; ,.,.,
leads to the relat1on 'P (~) ~ ~ (.t~)
• .....
6.1~26. (f) Show that the set tVA~ a IS countable 1nfin1te
for I B r 2 > ...
{2} Fl.nd i()-r ~ B ~ -= 1
th~ p~"h~t ~t \h~ ~e,t ~A 8
6.1 ~27 ~ Let D A., B be a lnoaded tree represent1ng a
b d functtnn 'P A~ -).- B{Jl of ~etght r Let us change the
output symbol of snme are 1n the J th l~~el of the tree
D A B (we assume that ~ B t >
2} TillS leads to a tree
DA.. s.. representlng a. certain (new) b d Iunct1on tp'
of wetght r
(1) Show tltat t ~ r ~ r j +
(2) G1ve an example of a b d funct1on 11 at wh1ch the
upper esttmate ,~ = r 1 ts attatned +
6.2. REPRESE:-;TATION OF DETER:\liN ATE FU:\CTIO:-;s !8i
6.2. Representation of Determinate Functions

by Moore Diagrams, Canonical Equations,
Tables and Schemes. Operations Involving
Determinate Functions
Let Q = {Q 0 , Q1 , ••• , Qw-d be a set of all states of
r"'
A
the function cp in <ll ,\, n· We associate a digraph

with the function cp:
(1) the set of vertices of the digraph r"'
is the set
E w = {0, 1, ... , w - 1}; the vertex j is assumed to
correspon d to the state Qi;
(2) if cp and cp<iJ are residual operators of the func-
tion cp represent ed by the states Qi and Qi respectiv ely
(q;ti> is also a residual operator of cpti) and correspon ds to
,...., • 1'1 • ,_
the prefix x1 = a and if cp(l) (axt~) = bcpUl (xw)), then

the digt·aph r"' contains the arc (i, j) and is assigned the
expressio n a (b);
(3) the arc (i, j) exists in r <r only when the condition s
in (2) are satisfied.
The vertex in r q:• correspon ding to the initial state of
the function cp, is often marked by an asterisk. Suppose
that m arcs, which. are assigned the expressio ns a1 (b 1),
a 2 (b 2 ), • • • , am (bm}, pass from vertex i to vertex j
(here, it is essential that ap =I= aq fot· p =I= q, although
some or all of the symbols bs may be identical ); we shall
then join i with j through a single arc (i, j) which is
assigned all the expressio ns a 1, (b 1,), p = 1, m. The
digraph r '~' is ~alled the Moore diagram of function cp.
Two functions can be associate d with this diagram:
(a) F: A X Q ) B is output function,
(b) G: A X Q ) Q is transition function.
The functions F and G in r"'
are defined as follows:
from the pair (a, j), we determin e the vertex j and an arc
emanatin g from j which is assigned the input symbol a
(suppose that this arc has the form (j, r)); the value of
the. fun.ction .F on the pair (a, j) is an output symbol
whiCh Is asstgned to the arc (j, r) and which appears
after the symbol a in the parenthes es. The value of the
function G on the pair (a, j) coincides with r , i.e. it is
I " ,
equa to t I1e number' of the state which "is the end" of
the arc U, r).
t88 CH f Pl~l'I'E AlJTOM.ATONS
The system o[ equat1onq,

y (t) ~ F (.z: (t), q(t-1))
q <tJ ==; G c~ (t), q (t .-1)). (1)
q (OJ~ qo
where x (I) EA y (t) EB q (t) EQ (t = 1 2 )
and q0 EQ IS called the canon teal eqUIJ.tton-s for tht oper
ator q1 u ttk tke lntttal condttlon q0
\V1th the help of 1\loore s diagram and csnon1eal equa
t1ons, we can also spec1iy determtnate opetators that are
D<lt boundedly det-erm1nate The set of verttc.es of the
dtgraph r,. corresponding to such a d operator ccln
c1des WJ th the natural scale N = {0, 1 2 f J
If rp 1s a boundedly de term 1na te operator, the funct1ons
F (z (t) q (t - 1)) and G (r (t), q (t - 1)) (see Eqs (l))
and the arguments on wh1eh these lunct1ons depend a.s
sume a finl\e number of va1ue-s Hence 1t 1s posslb1e to
eomptle a canonical tabu of the b d operator r.p
Tahle 7
%(U f (t- f) II (t} q (f)
lJ 1 F(a J) G (1 J)
Instead of the canonlcal equattons (1) 1t 1s ccnven1en~

to constder canonical equa t 1ons In wh1ch the tnput and
trans1t1on functtons are functions of the k valued logtc P 4
(k ~ 2) In order to obta1n a corresponding represen ta tton
of the operator rp the alphabets A B and Q are coded by
vettors whose coordln&tes belong to the set Elt (k 2) >
If <p 1s a b d operator and n-===- 1log 11 ' A I { m =-
llog~ l B ll-. r = llog~ ( Q ll~ then tt lS suffictent to take
vectors (whose coordinates belong to EJt) of length n m,
G.2. REPRESENTATION OF DETERMINATE FUNCTIONS 189
and r to code letters from the alphabets A, B and Q re-

spectively. 3 In this case, the system (1) is transformed as
follows:
Yt (t)=Fdxt(t), ... , Xn(t), qt(t-1), ... , q,(t-1)),
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
Ym (t) = F m (xl (t), ... , Xn (t), ql (t- 1), ... , qr {t-1)),
ql(t)=Gdxt(t), ... , Xn(t), qt(t-1), ... , qr(t-1)),
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
qr(t)=Gr(xdt), ... , Xn(t), qt(t-1), ... , q.(t-1)),
ql (0) = qol• · · ·, qr (0) = qor·
(2)
Henceforth, we shall also use the "vector form" of systems
similar to (2). In such a notation, the system (2) has the
form
y<"'l (t) = F<m) (z(ll) {t), q<•> {t-1)),
q<•> (t) = Gt•l (a:tnl (t), qtr> (t -1)), (3)
q<•) (0) q~r).
;=
The functions F i and GL in (2) are, in general, parHal

functions, i.e. they are not defined everywhere. Usually,
they are redefined in such a way that the right-hand sides
of Eqs. (2) acquire the simplest possible form.
The canonical table describing the b.d.-operator cp
corresponding to (2) has m n + +2r columns and kn+r
rows. .
The equations in system (2) are called canonical equations
in scalar form, while the equations in (3) are canonical
equations in vector form.
It can be assumed that system (2) defines an operator in
the set ~·m.
Let a d-operator cp be defined by system (2), and let
each of the functions F i and Gt be a function of the k-val-
ued logic Ph. (k ~ 2), which are defined everywhere. We
shall consider the operator cp as an element of the set ~·m.
The scheme ~"' of the operator cp is defined as follows:
~q; is a network whose poles are assigned the symbols of
input and output variables, while some of the vertice&
other than the poles are assigned symbols of some
3
If the d-operator :
w1ll be presented 1n the form of a rectangle
m
{Flg 2G) wtth n ~ nput and m output channels The 1npu.t
chann~l:; are represented hy arro\\s ema.na.llng {rom lhe
1n put poles \\ ht te the out put channels are represented
by arrov.s termtnat1ng at the output poles The poles are
.x,
x1 Y1 x2
~2 . • Y2
•
:Cn %n
F1g 26 F1g 27
represented lD the form of CitCles 1f m ~ 1 the .scheme

:E~ or the operato.r 1:p can somellmes be represented ln
the form of a tt1angle (F1g 27) wtth n lnput poles and one
output pole
\\ e assume that an 111 put S) m bol x t (t) EE k IS sup
pltetf to the 1 th 1nput z 1 at earh Instant of t1rue t =
1 2 \\htle the value Vl (t) = F J (.t1 (t)
4.,. (t) q1 (t - 1) g, (t 1)) IS delttertd (represented)
at the j th output YJ The output y1 l.S sa1d to have a delayed
de pendente on the Jnput x 1 tf the {unc ltcn F1 (~ 1 (t} 71
qcr) (t - t)) dQes not depend essenllally on the \ a.r1ahle

x, (t}
The concept of delayed dependence can be zntroduced 1n
a dtffertnt way Let us cons1der for example a d funct1on
of the type «p (XJ X 2 X 11 ) A 1 X A 2 X X
A 11 , .. B and define 1ts delayed dependence on the vartable
X 1 The iunctJon ~ has a delayed dependence on X 1 1f lor
any 1nput words ';.r ~ ,._
;~
,.,...,
{;: E A~
,..,
1 =" f 'R)
~
the s th letter of the yUJ = rp (~ x~ x:) JS defined

...... ~
by the first s S) mbols of the words zt1 x-;: and s - t 11
At!'
nrst symbols of the word :z~

Suppose that a d funct1on lfj 1s defined by Eqs (2) and
that l:'P 1s the scheme of thas functton \Ve define three
operat1ons ln'Volving ' and 1:~
•
6.2. REPRES EN"rAl' lON OF DETERJ:IIINATE FUNCTIONS 191
(1) Operation 0 1 ~nvolves th~ identificatio~t of. two . or

more input variables m the functiOn cp and the zdenttficatzon
of the input poles corres pondin g to these variab les i~z the
scheme ~<r· The identif ied poles are consid ered as a smgle
•••
•••
• ••
•• •
!Jm
!J,
Fig. 28
x, x,
••• •••
••• • ••
!Jm
Fig. 29
pole in 'the new schem e. Figure 28 shows a schem e ~.P

obtain ed from ~'P by identi fying poles x 1 and x 2 •
(2} Operation 0 2 involv es the deletion of an output vari-
able Yi in the functi on rp (which is equiva lent t'o the dele-
tion of the equati on Yi (t) = Fi (x< 11> (t}, q<r> (t - 1)) from
(2)) and the deletion of the output channel and the pole
corres pondin g to the outpu t variab le Yi from the scheme
~'P (see Fig. 29, which shows the schem e ~.P obtain ed from
~'P after the deletio n of the output chann el and the pole
Yt)·
Remark. If m = 1, the deletio n .of the variab le y 1
(the only outpu t variab le) leads to an automaton without
an output.
192
(3) Operat1on 0 3 Involves the tnlroductton of a jeedbach

of one 1nput and one output vartable~ Let u~ take the
var1ables zi and y 1 as the lnpt.lt and outpttl vartables The
operat1on o,can be appl1ed (to the function rp and scheme
l: 4) only 1f the output YJ has dela)ed dependencP on the
1np\1t x 1 The canontca) eq\1at1ons for 1\v~ new fnnctlon ~
are ohta1ned by deleting the equation y 1 (t) ~ FJ (zP )(t),
q''l (t- 1)) from system (2) and by replacJn.g' lhc vartabl~
Xn
.... .. .....
F1g 30
Zt tn each of the funet1ons Fq (q =F J} attd G1 by the

(t)
funet1on F1 (x1 (t), , .t 1_1 (t), :t1 +l (t), x,. (t)
1}1 (t - 1), , q, (t - 1)) obtarned from the fnnct1on
f 1 (z(n> (t)~ q'') (t - 1.)) by omllt1ng tl\e Hctttt<l"-'S va.rt..able
:r, (t) The tntttal cond1ttons rema1n the same as before
The scheme ~ • IS obta1ned from tlte schema I" by tden
tt(y,ng the o,utput YJ w1th the 1nput :rl l11 th1s t.(:\se, the
1dentlfied poles are referred to as the 111 ternal 'ertex of
the scheme l':, F 1gure 30 sho\\ s tJ1e sc}Jeme .1: w obta1 ned
from :E, by tntrod1te1ng the feedback ln var1abl~s .:r1 and
Yt
Remark 1.. If n = 1 the tn trod ue t ton of a feed back tn
vanable x 1 (and any output var1able) 1eads to an auto
maton ltlthout lnput
Remark 2.. \Vh1le applytng the operattvns de.scrtbed
above't 1t 1s convenient to 1nd1cate 1n parentheses (lollow-
tng the notat~on of these operatlons) tl~e cl1annels (poles
and variables) to wJ11ch these operations are appJ1ed
6.2. REPRESENTATION OF DETERMINATE FUNCTIONS 193
For example, we can use the notation 0 1 (x1 , x 3),

0., (y 5) and 03 (xlo• Y2).
·Let us define two more operations involving d-func-
tions.
(4) Operation 0 4 is the union of two (or more) functions.
Let rp 1 E cP~ 1 ,m 1 and rp 2 E <l>~ 2 'm 2 • We shall assume that
these functions have input variables x;, x;, ... , x~ 1
and x;, x;, ... ,
x; 2 respectively. For the output vari-
ables, we take y~, y;, ... , y:.n 1 and y;, y;, ... , Y ~~ 2
respectively. All these variables are assumed to be pair-
r-
I
--
•••
--- ---
• •• -.I
I I
:E<j> : I Yl1 Yl2 I
I I
I I
••• •••
L.- -- --- --- _.J
Fig. 31
wise different. Let ~m and ~'I' be the schemes of the

"'1 2
functions rp 1 and q> 2 respectively. In this case, the scheme
~.P of the function 'lj>, which is equal to the union of the
functions rp 1 and rp 2 , will appear as shown in Fig. 31.
The output (input) poles of the scheme ~"' are all output
(input) poles of the initial schemes ~'1' 1 and ~'1' 2 • The
system of canonical equations (and initial conditions)
of the function 'lj> is obtained simply by a union of the
corresponding systems of functions rp 1 and q> 2 (naturally,
it is assumed that the sets Q< 1> and Q< 2> of all the states of
the functions q> 1 and q> 2 are disjoint).
(5) Operation S is the superposition operation. Let
fP1 E cD A, B and q> 2 E ci> n,C· The superposition q> 2 (rp 1)
of operators q> 1 and rp 2 is an operator 'lj> E ci> A,c, such that
~ r>J ,..., ,....,
'lj> (xw) = rp 2 (rp 1 (x 00 )) for any input word x<o> in A 00 • Let

the operators rp; E cD~i·m; (i = 1, 2), and let the schemes
~'1' 1 and ~'1' 2 correspond to them (we assume that the input .
and output variables and the states of the functions rp 1
13-063 6
194 CH 6 FINITE .AUTOMATONS
and rp 2 ato tJ1a samo as 1n the precedtng paragraph) In

tlus c1se "'e can con.sJder a 'd1ffe1 cnt, superposttlon for
th£'sc operators lor e,..ample v..e Jdent1iy the 1nput pole
%1 of the scheme l:~ w1th the output pole Yi+l of the
I M
stl1emo !. "' tl1e pole X a "1tl1 the pole y 1+2 and sn on t

l ~
f1u 1.lly, \\ o 1dent1fy tl1e pole Xm 1 1 \Vlth the pole yffl,
As a result, \\e get a scheme l!;p (F1g 32) for wh1ch (a) all
x, X'n .. Xm;-L+, Xn:z

,-- -----..--.-------
... !Ill
0
.....
0
--~--...,
I'
I
\
'' r
i
I }
I ~ • • ~r
I I
~t.+ri1 Ym~-';;,_ 1
a
I
'I
~
411 •
'
r
~
~
I '
I
'
I
_____ ___ __
r
I
L_ """"'~~- ,__ .. ..
...., to
...... _.-~,_ ..,.J
0
~lg 32
tl10 lnpul poles Wlll he the \nput poles of tl\e scheme I_,1
and the 1nput poles of the scheme I, 1 not parttc1pat1ng
1n tl1e above 1dentlflcat10:n proce-d\1[0 (b) the output poles
arc a]l output. poles of the scheme 1:., and those output
:a
poles of tlte scheme ~., whtclt were not tdenttfted wtlh
1
llUY urtltc Input poles of the sclteme l:q-1 The ldenllfied
poles are call-ed t.l~e lnternal vert 1ces o.[ the scl1eme Z. -t
fhe- scheme 1: • 1s called tl1e superpost llon of schemes l: fP;~
and k~ 1 (1n variables .x; -
11 1• 1.. YI+J , , .r~1 t - x; --
6.2. REPR ESEN TATI ON OF• DETE RMIN ATE FUNC TION S 195
y~ 1 ). If the ope rato r rp1 is defined by the syst em of equ a-

•
twn s
u; (t) = F~ (x~ (t), ... , x~.(t), q~ (t -1) ,

... ,q;l (t-' 1)) ,
• • • • • • • • • • • • • • • • • • • • • •
y~,(t)=F~,(x;(t), ... , x;,( t), q;( t-1 ) 1
... , q;l (t-1 )),
q~ (t) = G~ (x~ (t), ... , x;, (t), q; (t- J), (4')
... ' q;l (t-1 )),
• • • • • • • • • • • • • • • • • • • • • •
q~l (t) = c;l (x; (t),... ' x~. (t), q~ (t- "1)1
... ' q;l (t-1 ))1
q: (0) = q(ll, ... , q; 1 (0) = qiJr 1 1
and the ope rato r rp 2 is defined by the syst em of equ atio ns
y; (t) = F; (x; (t), .... x~. (t), q; (t-1 ),

.•. ' q;q (t- "1)),
-
.... . ... ..... . .. .. ....
y;, (t) = F;;,. (x; (t), ... 1 x~. (t), q~ (t- '1) 1
.. 'I q;2 (t-t )),
q;(t)=G~(x7(t), ... 1 x~.(t), q;( t-'1 ), (4")
... ' q;2 (t-1 }), •
• • • • • • • • • • • • • • • • • • • • • •
q;2 (t) = c;2 (x; (t), .. 'I x~. {t), q; (t -1) ,
• • • 1 q;2 (t- '1)),
q~(O)=qo11 ···1 q;2 (0)=qor 2 •
13*
196 CH 8 FINITE AUTOMATONS
then the operator 'P represent ed by the scheme I• has

the follow1ng system correspon d1ng to 1t
y1 (t)=F1 (x1.(t} Zn (t} q 1 (t-1)

' q,.l (t -1))
Yl (t) = Fl (Z1 {t) .tn. {t)f q 1 (t-1)

q!"l (t -1))
y 1 (t) = F~ (Ft+l Fm z:n 1 t+t (t)
Z~ 1 (t), qt(t--f) q, 1 (t-1))
Ym.{t):= Fm.(Fl+lt ; Fm.t X~ 1+1 (t)

tn. (t), qJ (t- f)w q,.., (t- 1))
q1 (t) = G1 (x1 (t) Zn 1 (t)J g 1 (t- J)
q,.l(t-1) ) (5)
g1 (t)=G1 (Fl+t• "F,.

., Zn (t), q, (t l)
1
q,.. (t) ~a;, (FI+t ' Fm ' Xma l+l (t).

, x~J (t) q1 (t -1) q,.a (t 1))
91 (0) =: qot q,.1 (0) =::::; 9or1 qt (0) = q(j 1
l
where
F 1 ~F,(x 1 {t), , J:n.{t)~ q1 (t-1)
J =:;.l + 1 1 i' m1
The element of unit delay (in the set <l>k) is a b.d.-ope-

rator cpd defined by the system of equations •
y (t)= q(t-1),
q (t) =X (t), (6)
q (0) = o.
A
Example. The operator cp in Cll 2 is defined with the help

of Moore's diagram presented in Fig. 33. The canonical
q (t-1)
::x:(t)
d
y(t)
o( 1}, 1(0) o(o)
1( 0)
Fig. 33 Fig. 34
table (see Table 8), canonical equations, and the initial

condition for this operator have the following form:
Table 8
X (l) q ( l - 1) y(l) q(t)
0 0 1 1
(\ 1 0 1
1 0 0 1
1 1 0 0
y(t)=x(t)·q(t-1),
q(t)=x(t)Vq(t-1),
g (0) =0.
198 CS 6 FINITE AUTOMATONS
.,.,.
F1gure 34 shows a scheme representing this operator
Thls d1agram has been constructed by us1ng an element of
...
un1t delay and an operator 10 ! ', generated by a S)mbol
of Jotnl negattonz. y Here., the operator generated by the
JOint negation symbol 1s represented through the scheme
whlle an element of un1l delay ts represented by the scheme
....
Every b d operator In ill~ 1"ft. may be represented by a
scheme generated by a set conta1ntng (1) schemes repre·
sent1ng an element of un1t delay; (2) schemes represent1ng
Opera\ors generated Dy lunC\lDnS from a certalD se\ tha\
ts complete tn Ptt In other words,. lhe set of b d ~ope
rators formed by the element fPd and the operators gen·
erated by functions from a certain set complete 1n P"
form a set that ts complete 1n U ¢.r"' wttb respect. to
n.m
the operations 0 1 , 0 1 , 0 3 -r o., and S
Remark. Henceforth~ wb1le constructing a scheme of
any b d opera tor q:> genera ted by a certa1n gt ven set M
ot b d -operators, we :shall adopt the follow1ng procedure
the scheme Itt of the operator 41' 1S constructed by nstng
only single output schemes representJng operators In the
set Af (ln th1s case unless stipulated othenv1se, we can
only carry out the operattons 0 1 ~ O~., 0 3 0 4 ~ and S)
'4
6 .. 2A 1. Construct the l\1 oore d1.a gram~ the canonical

..
table, and the canontcal equattons for the funct1on (~) y(2t· -1)=J:(2t-J), t~·~
y (2t) = z (2t) EB y (2t--1}t t~ t:
y(1)=1,
v<t)=: ;t{t f) t!) z(t), t~2,
y(3t-2) =x(3t-2 ), t>1,

y(3t-1) =x(3t-2 ), t~1,
y(3t)=x (3t)y(3 t-1), t~1;
,.,
(4) q> (x(J))
,., = (2/3);
( 5) cp (x(l)) = (5/8).
6.2.2. Redefine the partial operator cp: {0, 1 }w >-
{0, 1 }Cll in such a way as to obtain a b.d.-oper ator. Con-
struct the Moore diagram, the canonica l table and the
canonica l equation s for the new operator if:
(1) cp (W11010l(J)) = f011H'\ cp (0 [1J(J)) = o [11(1);
(2) cp (O(J)) = 1(J), cp (1 [Ol(J)) = [10J(J);
(3) q> (1 [10](J)) = [01](J), q> ([001](J)) = 1 [10l(J);
y (3t - 1) = X (3t - 1), t ~ 1,
4
< > cp: y (3t) = x(3t - 1), t ~ 1.
A
6.2.3. Find the weight of the b.d.-oper ator cp in cl> 2 ,

defined by the canonica l equation s:
( y(t)=x( t)qt(t-1 ) V x(t)q 2 (t-1),
(1) cp:
I qdt)=x (t)qt(t- 1}Vx(t} q2 (t-1},
q2 (t} = x(t} q (t-1) Vx (t)q (t-1) q
2 1 2 (t-1),
q1 (0) = q2 (0) = 0;
y(t)=x(t ) ED q1 (t-1) ED q2 (t-1),
qt (t) =X (t) ~ qt (t-1),
(2) cp:
q x q
q2 (t) = (x (t) ) 1 (t -1)) >- (t) 1 (t-1),
qt(O)=q2 (0)=1;
(3) cp is defined by the cannonic al equation s of the pre-
ceding problem, but with different initial condition s:
ql (0) = 0, q2 (0) = 1;
• y(t)=(x( t) ~ g2 (t-1)) > qdt-1),
(4) cp: qt (t) = qt(t -1) >- (x (t) >- q.
(t-1)),
qz (t) = q2 (t -1) >- x (t),
qt (0) = q2 (0) = 0.
A
6.2.4. Let (C!>~·rn; X1, X 2 , Xn; Y 1 , Y 2 , • • • , Y m)

••• ,
A
denote the set of all functions in cl>ff.mwith input varia-

bles X1, X 2 , ••• , Xn and output variables Y 1, Y 2 , • • • ,
Y m· Show that the number of functions with a wei~ht w
200 Cll 6 PtNitE .AUTOllATO'iS
111
lD the set ((Jll' , x1 .. XI. ,. X.,., Yl. Yl, t y m)
does not exceed (wkm)wl,.,.
6 2 5 Let the operator 'P tn a>; "' be defined by (2)
.and let. the funet1ons G1 , G1 , , Gr be canneeted through
0(1)
1(1) 0(1 1(0)
___0(1} 1(0)
---o(o)
10
F1g 35
the relat1on G1 e G'~ m •

EB G, ==----= 0 Show that the
we1ght ol the ttperator cp does not exceed 2" 1
.
62S Repxesent the operator rp E ill~ m through a
scheme g~neratl!d by a set formed by untt delay elements
and the operator generated by Sheffer s strole
y(t)~q(t-1)
(1) q (t) = x(t),
q (O)~ 1,
y(t)~(x(t) V q~(t
1)} .,-q1 (t-f)
ql (t) ==X (t) _,_ qJ (t --f)*
(2)
9: (t) =it (t --1) v q2 (t- f)
91 (0) == 0, 9:! (0) == 1,
~ ........
(3) ~ (~) = 10~(L) 1f ~ = OOz (3) z (4)
1c..- otherw1se
(4) the operator f ts defined by ltfoore s dtagram sho1VD
1n F1g 35,
Yt (t)~ X (t} I -1),
)v 91 (t
Y: (tl ==;; q1 (t-1) q2 {t- ·= t),
-
(5) ~ q 1 (t)~z{t),
9: (t)~(g 1 {t- --1) V q2 (t-1)) %{t),

ql (0)-== q1 (0) ~ 0
6 2. REPRESENTATION OF DETERMINATE FUNCTIONS 201
6.2. 7. Using the scheme of the operator rp E <IJg.m,

construct the canonical equations, the canonical table
and Moore's diagram.
(1) see Fig. 36a; (2) see Fig. 36b; (3) see Fig. 36c.
6.2.8. For the superposition 1jJ = rp 1 (rp 2 ) of operators
A
rp 2 and rp 1 in <1> 2 , construct the canonical equations,
x(t)
x(t)
d d
d d
z(t) z(t)
(a) (b) (c)
Fig. 36
Moore's diagram, and the scheme ~¢· The scheme ~ ¢

must be generated by the set consisting of a unit delay
element and operators generated by the implication
-
x ~ y and the negation x. Consider
Yt (t) = xt(t) qt(t-1),

(1) ql (t) = ql (t -1) ~ XI (t),
ql (0) = 0,
Y2 (t)=x2(t) 61 q2 (t-1),
rp 2: q2 (t) = X2 (t) V q2 (t--1},
q2 (0)=1;
(2)
the operator w1 J.S defined by l\.foore' s d1agram shown 1n
Ftg 37~
,.,., y(f)~Ot
(3) lJ'i (~)
y(t)~y<t '"=t) e z(t~t), t>2,
y (f):;;:: 1,
y(t}~x{t) _,_ :J:(f .. ·-1}, t'#2..
y{1)==1t
y(t)= y{t-1) ~ z(t}~ t~2~
~
fP2 (z¢) ~ (f/7)

6.2,9. Construct the canonical equations and ~(oore's
d1agram of the b d operator ohta1n~d frum the uperator tp
'{o}. o{ 1 ~
0{0)
1{1)
Ftg 37
by mtroduclng the feedback 1n vartables x 1 and y1 lf

-
Y1 (I) = g (t- 1) (t) Z3 (t)
5
). Zt 1
Ya (t) = z2 (t) V (:rl (l)- ~ g (t-- 1)),

(1) -
q(t)~.xz(t) e
q (t--1),
g (0) = 0;
-
Yt (t) :::::: q (t .-1) - .._ % 1 (t) Z~ (t) 1
Yz (t) ~ (xt {t) ~ q (t \)) \J z 2 (t),
(2)
g (t) ~ .z1 (t) EB q (t --1)~
q(O)~ 0
6.2.16. Ftnd the v.e1ght of the b d -operator obtarned

frorn the b d operator 'P by 1ntroducmg the feedback 1n
Variables z, and y 1 11
Y1 (t) = Z 1 (t) , ,. (x1 (t) _,.. q (t-1)),
Yz (t) ~ x, (t) ,. x 1 (t),
(f) rp q(t)~x {t) ,.{x.(t)'i2 (t) ,.{za(t) ... q(l 1))),
1
q(O) 0; =
y 1 (t) = x 3 (t) > q (t-1),

X 1 (t)
Y2 (t) = X 2 (t) q (t -1),
(2) cp: -
q(t)=x2 (t) V x 3 (t) V q(t-1),
q (0) = 0.
6.2.11. Suppose that the operator 1jl is obtained from
the operator cp through the operation 0 1 (identification
of input variables).
(1) Give an example of a pair of operators cp and 1jl
for which the following inequality holds: the weight of
the operator cp is more than the weight of the operator 1jl.
(2) Can the weight of the operator cp be less than the
weight of the operator 1jl?
6.2.12. Let cp be a b.d.-operator of weight r and let 1jl
be an operator of weight r', obtained from cp by intro-
ducing the feedback in a certain pair of variables. Is
it always true that
(1) r>r'; (2} r = r'; (3) r~r'?
6.2.13. Suppose that the b.d.-operators <p 1 and <p 2
have weights r 1 and r 2 respectively. What will be the
1{o)
1{--.;.0)........
0 (1 ) ______ 1( 1 } o( 1) 1( 1 )
o( 1} o( 1}
(a) (b)
•
Fig. 38
weight of the operator 1jl obtained from cp1 and cp 2 with

the help of the operator 0 4 (union)?
6.2.14. The b.d.-operators cp 1 and cp 2 have weights
equal to r 1 and r 2 respectively. Will the weight of the
superposition cp1 (cp 2 ) be
(a) larger than
(1) r 1 , (2) r 2 , (3) r 1 +
r 2 , (4) r 1 • r 2 ?
(b) smaller than
(5) r 1 , (6) r 2 ?
6.2.15. Find the weight of the superposition cp 1 (cp 2 ) if
. (1) the operators cp 1 and cp 2 are defined by Moore':'!
diagrams shown in Fig. 38a and b;
CH 6 FINITE AUTOMATOl\S
(2) the operators r.p1 and cpz are defined by 1\loore s

d lag rams shown tn Ftg 39.a and b
Yi (t)-= x, (t) .. fJI (t -1)
(3) ft'~ Qi (t) == Qt (t -1) ~ Z, (t) I =::; 1 2
1
q1 (O)~O
The opera tor fP 10 rs called autonomous (constant
til A B
;w
or Wlthout tnput operator) 1f for any 1nput word 3fJl E A~

1(0) 1{1)
1(o) 1( 1) 00 0(1} 01
----
0 (0) o{., ) 1( 1)
0(1} ...._..... 10 o{o} tt ......

1(0)
(a)
(b)
1"¥
the value of the operator f.P on z:'li ts equal to the same

(output) word In Bw
6 2 16 Is the opera1.oY tp E-<1>1 autonomous tn the fol
Iow1ng cases?
~ y(l)~ y (2) = 1
(1) cp (Z:W}
y(t)~y(t- !) ffJ y(t-2)
y(t)~x(t) .. q(t-1)
(2) q(l)= :r:(t) v q(t-1)
q(O)~ t,
y (t)--- {t) -* q (t- 1)
X
(3) 9 {t) ~X (t) V q (t -1)
q-(0} ~ 0,
y(1)-=0
(4) y(2l- 1)= y(2t-2), tp2,

y {2t} = Ieos ~ t 1 t~1
6.2. REPRESENTATION OF DETERI\IINATE FUNCTIONS 205
'
6.2.17. Let cp be an autonomous operator of weight

r (r < oo).
(1) Show that the output word of the operator cp is
quasiperiodic.
(2) Prove that the sum of lengths of the period and
pre-period of the output word of the operator cp does not
exceed r.
6.2.18. (1) Construct a scheme ~'I' using the set
M= o10l l•o, '
such that it contains exactly four elements in iVJ and repre-

A
sents a certain autonomous operator cp in <1> 2 •

(2) Is it possible to construct a scheme ~ 2 and containing less than

four elements?
Operation 0 5 , called the branching operation (for a cer-
tain output of a b.d.-function) can be defined as follows.
A
Let cp be an operator in <l>i:·m and let ~'I' be the scheme

representing it. Applying the operation 0 5 to the out-
put Yi of the function cp, we obtain an operator "ljl,
represented by a scheme I: .p which acquires, instead of one
channel (and pole) y 1, several "equally operative" channels
y<j>, ••• , y<j> (s;;;;;:: 2), each of which represents the same func-
tion as y1 does in the scheme I:'l'.
6.2.19. (1) What changes must be made in the system
of canonical equations and initial conditions describing
a function cp in order to obtain a corresponding system
for the function 1jJ resulting from the application of the
operation 0 5 to the output Yi of the function cp?
(2) Is it possible to construct by using operations
01-05, Sand the set M (see Problem 6.2.18) a scheme ~'I'
which contains not more than three elements and which
A
represents the autonomous operator cp in <1> 2 ?

6.2.20. Suppose that the operator 1jJ is obtained from
A
cp E cp~,m by applying the operation 0 5 to a certain out-

put of the operator cp. Show that the operator 1jJ can be
constructed with the help of the set {cp} u.e. by using
206 CO 6 FINITE .AUTOlrA.toNS
the operator cp a few times) through the operat1ons 0 1

0 2 , .and 0 4 A
6 2 21 The operators tp 1 and <p 2 1n CfJ 2 have a \\etght

2 each, and are defined by the followtng canontcal equa
ttons
y 1 (t) = F1 (x, (t), q1 (t- 1)},
ft1, q1 {t) = G1 (x~ (t) q 1 (t t)), ~ =::; 1 2,
q1 (0):;;:;;; 0
where G1 (z, q) G2 (x, q) It 1s known that 1f the zeros
assigned to the arcs are replaced by unJ ttes and the un1t1es
are replaced by zeros, 10 hloore s dtagram of the operator
q> 1 , we obtain ~loore s diagram of the operatol ({Jz Prove
tha.t 1f the output values af the Qpe.ratots ~ 1 and qt:t cotn
c1de on a certa1D prefix of length 2 (1 e , tl «p1 (a1o 2) =
<Fa (crt<12 ) for some values of cr1 and o2 )* these operators
are 1den ttcally equal
....
6 2 22 Count the number of b d functions 1n W1
wh1ch sat1sfy the follO:Wlng ~ondlt1ons {a) a funet1on
depends on 1ts Input variable X, (b) the wetgh t of a func
tJon Js equal to 3, (c) lll the hJoore .d1agram descrJbJ.ng
a funct1ont the Ill degree of each vertex IS the same and
equal to the out degree
6 2 23 For each
A
r> 2, gtve an example of an opera
tor 1 such that the we1ght of the ~uperpos~tton
rp (rp) IS equal to r Is 1t poSSible to do the same for non
autonomous operators?
...
6 2 24 The weight of a function rp E (Jl 2 IS equal to r
and the wetght of the superposition rp (tp) IS equal to 2r
\\'1ll the superpos1t1on qJ (~ (<tl)) have a we1ght equal
to 3r?
6.3.. Closed Classes and Completeness

in the Sets of Determ1nate
and Boundedly Determ1nafe Functions
Let AI be a set of d funct1ons (or b d functions), and
let G be a set of operat1ons the results of whose appltcat1on
also rematn 10 the set of all d functtons (or b d func
t1ons) In other words, 1f Cl E G the application of a
to a-rb1trary (o-r adm1ssl.ble) d- (or b d l futlctlODS wtll
6.3. CLOSED CLASSES AND CO:'IlPLETENESS 2lli
again !'esult in d- (or b.d.) functions. The closure [Mle

of the set M relative to the set of operations 0 is the set
of all d- (or b.d.) functions which may be obtained from
the functions of set M with the help of operations in ~,
and these operations can be applied any fmite number
of times. Usually, it is assumed that [M]l') ~ M. Hence-
forth, we shall always assume that this inclusion is
satisfied (if this is not the case, a special mention will be
made). The operation of obtaining the set [.M]l') from M
is called the closure operation. The set JVI is called (func-
tionally) closed class relative to the set of operations (9
if [Mle= M. Let M be a class of d- (or b.d.-) functions
closed relative to the set of operations (9. The subset ff in
ill is called a (functionally} complete set in jli relative to the
set of operations (9 if [.fle = Jv!. The set & of d- (or b.d.-)
functions is called an irreducible set relative to the op-
erations (9 if for every proper subset &' in P the (strict)
inclusion [:f']l'>c [&le is observed. Any complete and
irreducible set in A/ is called a basis of tho closed class JV/
relative to the set of operations (9. The set M', con-
tained in the closed class M is called a precomplete class in
ill if it is not a complete set in Jl/, but if the equality Ull' U
{cp }le = M is satisfied for every function cp E jvf "-.. M'.
Remark 1. Henceforth, we shall omit the expression
"relative to the set of operations <:)" if the set <:J relative
to which the closure, complete sets, bases, etc. are con-
sidered is known.
Remark 2. Usually, we shall use tho set {01 , 0 2 , 0 3 ,
0 4 , S} formed by the operations described in Sec. 6.2,
for the set of operations (9.
It was mentioned in Sec. 6.2 that the set of b.d.-func-
tions containing a unit delay element «Fct and the operators
generated by the functions of a certain set that is com-
plete in P 4 (k~ 2) forms a complete set in the set
C<) OQ A
U U ct>J:.m relative to the operations 0 1 , 0 2 , 0 3, 0 4,

11=0 m=O
and S (i.e., in a set of all 1~-L'alued b.d.-functions).
C0 (\) A
The set U U C!J~· (k~2)

111
is !mown to contain
n=O m=O
h.ases relat.ive to the set {01 , 0 2 ,0 3 , 0 4 , S}, which con-
Slst of a smgle function. An example of such a b-asis is
(.lt & PIN11E AU'I'OM.ATONS
the .set {cro (X 1 t x2t X 3' 'Pd (X t)) }t where Cfd IS a Ulllt
A A
delay element from <l>.~t, and (j)o lS an operator ID (J):'

generated by the funCllo!l max (x1 .,x 4 + :r 2 (l-~,} :r 3) +
t (IILod k sum. d1fference and product) ...
Everywhere tn th1s section., we shall use the sym hoi <Doc)
(«Dc 11 )) to denote the set of all k valued b d (resp d )
functtons tacludtng funclLQn5 wtlhout tnputs~ or Wlthout
outputs,. or Without both, 1 e
lJ ~~
d>cu = U m-0p-Q
m and ll>cli) = U m-0
n--0
U tbA ,.)~
Is the set AI a closed class rela t1 ve to the set
6~3.1.
of operat1o.ns G?
(1) 1\f cons1sts oi a.ll k-valued b d funct1ons Wl\)J.O\lt
(llj). QC ....
outputs, and of all functions 'P belonging to U U $~ .,.

n~D m~l
..-../ ~ (I) 1'\,1
and "preservtng o(t) ', J e <p (0 '

I F
t 0"') =
7 rl
"" 4
n tJmes
t FT 4 2a r
m IJm~,a.
(2) AI conststs of alJ k valued b d funeltons belong1ng

(IQ A
to the set U lll~ rn and havtng a wetght that lS mul-

m--o
ttple of k., t) ~ {S}
(3) l'tf cons1sts of all k valued d functions he]onglng
to the set
~
U U
:n==l m-1
Q;)
$: m, and l1a VI ng e1 ther an e\ en or

1nfin1te ue1ght, <9 = {01 , 0 2 , 0 4 , 0 5}
1 Ql) ...
(4) At~ U U {$~ m~<Di! m)t G ~{0 3 t S}

n---o m--0
....
6.3~2. (1) Let fJld be a un1t delay element tn the .set W1
and G ~ {0 .:J• 0;,, S} Prove that the class (cpd}t) contatns
only functions that are tdentu:ally equal to zero (t e op
f""9'
erators whose each output yields the word 06)), functions

Without Inputs and outputs whtch are always In the same
lnttial state, and tunct1ons havtng an even w~1gbt
(2) G1.ve an example of a funet1on 1n dl2 hav1ng an
even weight and not belonging to the class [cpd]t) de-

scribed in Problem 6.3.2 ('1).
A
6.3.3. Let 'Po {X 1 , X 2 ) be an operator in (D< 2 > generated

by the Sheffer stroke x1 I x 2 , and let 'Pct be a unit delay
A <V
element in (D 2 • Prove that the operator 2 • Find the maximum peri-

od of the output word (the majorization is carried out
over all input periodic words of period 3):
y(t)=(x(t) EBqdt-1)) q2 (t-1),
qt(t)=x(t) qdt-1) (j2(t 1)Vx(t) q2(t 1),
( 1) <p: - -
q2 (t)=x(t) qdt-1) q2 (t 1)Vx(t) q2 (t-1),
qi (0) = q2 (0) = 0;
y(t)=x(t) qt(t-1),
qdt>=x(t) ~<t-1)Vq2<t-1),
(2)
q2 (t) = qi (t-1),
ql (0) = q2 (0) = o.
6.3.5. Let A. B.

"'
(1) Prove that if each input word xw is quasi-periodic
"'
with a period T, the output word <p (xw) is also quasi-
periodic and its period does not exceed the number r· T.
(2) Find an upper estimate for the length of the pre-
period of the output word <p (xw) "' if the pre-period of
"'
the word xw is known to have a length p.
6.3.6. Let the functions ct (q>ct ( ••• q>ct (X) ... )) is the
superposition of s functions q>ct.)
~
If I (x") E P11., then q> f (x!l)

"' (X 1 , • • • , Xn) will denote
a b.d.-operator (in <f>~· 1 ) generated by the function 1 (:i').

llt-0636
210 CU 6 FINITE .AUTObiATO~S
6.3 .1 Are the follow1ng sets of b d funct1ons complete

1n ll>(ll rela t1 ve to the set of opera t1ons {01 , 0 2, 0 s.
0 4 ., SJ?
(1) { '-Pmax(:.: 1 x~1 (X 11 X 1), C)ld (X)), k ~ 2t
(2) { ctlxt .T2 (X l' X 2)' 4l'=c1!.x:~ (X1, X 2)' rp ~L- x 1 ( fPd (X),
X 1 ) cp::: 1 (X), 'P::::A 2 (X) J, k ~ 3 (here, cp;::J (X) 1s
a b d operator generated by a function 1n Ph, Jdent1cal
ly equal to J),
(3*) {fPx +1 (X)1t qld ( tpd (X)) t ti'x-u (cpd (X), (X 2))
6.3.8. Isolate at least one basis Jn the .set ~ wh1ch 1s
.....
complete 1n Cll<kl relattve to the set of operatiOns {0 1 ~
02t 0 3-. 0• S}
( 1) ~ == {qJd (X) ~:::.o (X), <v~1 {X), \f:t 1 !t 1 {X1 Xz)
ffx 1$~2ffi..r 3 (X 1 , X2 t X3 )} t k == 2
(2) !f = {rpd (X) tps o (X), ltl== 1 (X)t fllJ~(:~e) (X)
\Px1+x2 (X,, X2)} k~ 3~
(3) [f -=== { 'P=o (X) £Px 1 ~x~ (qJd (X)t X 2), cp x+t (X),
fPmtn(x 1 1~} (Xlt X 2 )}~ k~ 3
...
6. . 3 a_•. Does lll(2l contarn a bas1s relat1ve to the set of
operations {01 0 2 ,0 :Jt 0 4 , S} wtth f1ve functtons?
f
....
6~3~10* Pro\e that any ~losed class ln $(Al e.xeept the
...
ent1re set <ll(~t} can be extended to a precomplete class 1
6~3.11 • Ustng the Problems 6 3 8 (1) and 6 3 10, show
...
tl1at f!J(t.l contarns at least four precomplete classes
....
6 3 12 Let ft1d E <b: Enumerate all the classes that
a.re pre-complete 1n {q:rd],o 3 oa s) Telat1ve to the set. of
operations {0 5 , S)
6 3 ~ t3, Let ~=-: -= rp ~ (X) be a b d opet at or lU $(Jr.}
generated by the functiOn x Prove that whatever the type
of the closed class' Af1n ~{k), the equal1ty lAf U {'P %}1 -=
AI U {'il :x J 1s always satisfied (here, as usual, we take to
gether w1tb. the tunct1on all funct1ons that are equal to
1t or congruent to It)
6.3 . 14 . Can th~re ex1st a funct1.on helcng1ng to each
..
precomplete class 1n <be.~~,?
' The cIesure and precom p le tton ate ta k.eu rela tl ve to the set
of operatJons {01 0~, o,, 0", S}
t
6.3. CLOSED CLASSES AND COl\IPLETENESS 211
6.3.15. Can the set <D<kl be presented in the form of

s
a union U .ilii (s~ 2) of pairwise disjoint closed5
i=l
~
classes in <D<kl?
6.3.16*. Prove that the power of the set of all closed
~
classes in <D<kl is continual.

6.3.17. Determine the power of the set of all closed5
A
classes in <D<kh having finite complete sets.

6.3.18*. Prove that the power of the set of all closed5
classes in <D(kl is equal to 2c {the power of a hyper-
continuum).
6.3.19*. Find the power of the set of all closed classess
in <D<k> having finite bases.
6.3.20. Disprove the following statements:
A
(1) the set <D(kl forms a precomplete class in <D<kl;

(2) the set <D(kl contains a function forming together
~
with the set <D<kl a complete set in ({><kl·

6.3.21. Let M 0 be a set consisting of all k-valued d-
functions that do not have outputs, and of all functions cp
00 00
belonging to U U <Dr·m and assuming the value 0

n=O m=l
r-J ....., "' /"'J
for t = 1, i.e. cp (x~, ... , x~) = (y~ 1 ••• , y~) and

Yi (1) = 0 for j = 1, m..
(1) Is 1v!0 a precomplete class in $< 11 >?
Is the intersection M 0 n <D<kl a
A
(2) precomplete
A
class in <D(kl? _
I 00 A
6.3.22. {1) Is the set U U <D~· m a precomplete

A
n=O m=O
class in <D<11 >?
1 00
(2) Is the set U U <D~· m a precomplete class in
n=O m=O
<D(k)?
6
See footnote on p. 210.
Chapfe r Seven
Fundamentals of the Algorithm Theory
7.t. Tur1ngf:s Machines

and Operations with Them.
Functions Computable
on Turing'"s Machines
A Turtngts m.athlne 1s an (abstract) device cons1st•ng of
a tape a controll1ng deLlCe and a scannlng head
The tape IS divided 1nto squares Any square at each
(d1serete) tnstant of t1me conta1ns exaetly one symbot 1n
an exttrnal alphabet A ~ {a 0 , a1 , ~ an._ 1 }t n> 2
A eertatn symbol of the alphabet. A 1s ~alled a dummy
syTTlbol, and a square contalntng a dummY symbol at
a g1ven moment 1s known as a blank square (at th1s mo
ment) A dummy S)- mbol 1s usually represented by 0
(zero) The tape 1s assumed to be 1ntln1te vn both d1ree
ttons 1
At any 111stant of lime the conlrolltng dev1ce 1s 10

a certa1n state q1 whtcl1 belongs to a set Q = {qo, 91t
4J t
qr_ 1 ), r > 1 The set Q 1s ref(;rNd to as an 1nterna.L alpha

bet {or a set of internal states) Sometimes, dtsJotnt subsets
Q1 and Q0 of tntttal and final states resp-ectlvely are tso
Iated from the set Q
Rema~:k Henceforth unless stated oth~rv.lse,. we shall
assume that f Q 1~ 2. and take only one state q1 for the
tn1.ttal state As a rule, the state q0 w1ll be the linal atate
The head mo\ es along the tape .so that at each anstan t
tt -c;cans exactly one square of the lape The head can
scan a symbol In the square and record (prtnt) on at a new
symbol of the external alpl\a.het tnstead Tb.e symbol
uprrntedu 10 the ~quare can, lD partlCUlelr, COlllClde With
that betng scanned (a.t that 1nstant)
Dur1ng tts operat1on, the controll1ng devtce changes or
does not Its (1nternal) state dependtng on tts state an.d on
the symbol .sean ned by the head orders the bead to prtn t
1
Th1s should be understood a.s follows at each 1nstant the tape
J s lin1 te (con tatns a fin.t te number o ( sq tlates l, but the ~ tength~ of
the tape (the number of squares Jn 1t) can be Jncre-ased (•f requtred)
7.1. TURING'S MACHINES 2i3
in the square scanned a symbol of the external alphabet ,

and either to remain at the same square, or to move by
one square to the left or right.
The operation of the controlli ng device can be charac-
terized by three functions:
G: QxA )Q,
F: QxA >-A,
D: QxA ){S,L,R }.
The function G is called a transition function, F an out-
put function, and D is a function of displacem ent (of the
head). The symbols S, L and R denote respectiv ely the
absence of head displacem ent, a left movemen t and
a right movement by a square.
The functions G, F and D can be specified by a list of
quintuples
q,ap (qi, a1) F (q17 a,) D (qi, a1) (1)
which is abbrevia ted as q,aiqiiaii dil. These quintuple s are
referred to as commands. The functions G, F and D are,
in general, partial functions (specified not everywhere).
This means that a quintuple of the type (1) is defined not
for each pair {q1 , a1). The list of all quintuple s which
determine the operation of a Turing's machine is known
as a program of this machine. The program is often
given in tabular form (see Table 9).
Table 9
q. • • • q. • • • qr-1
'
•
•
ao
•
• • •
aj • • • qiJa•ldii • • •
• • •
•
an-1
•
•
214 CII 7 FUNDAl'-lENtAtS OF THE ALGORlTHJ\.1 THEORY
If there 1s n-n qutntuple of the type {1) tn the program

for a pair (q~, a 1) a dash appears 1n the table at the Inter
sect1on of the row a1 and the column g,
The operatton of Turtng's roach1ne ca..n be con-vent
ently descr1bed ID the "language of confi.gurattonst-t
Let at Instant t the extreme left non blank square cl of
the tape con tatn a symbol at1 • wh1le the e.x treme r1gh t non
blank square C, (s;;a- 2) contatns a symbol a,, (there are
s -- 2 squares between C1 and C~) In th1s case, the word
P = a 1,a12 aJp aJ,~t where a1P t.s the symbol wn
ta1ned at 1nstan t t tn the square CP (1 ~ p ~ s), IS sa1d
to be recorded on the tape at lnstant t For s = t. 1 e when
the tape ~onta\ns Gnly on-e no1l dummy symbul, P ~ a11
Let us suppose that at this 1nstant of t1me the controll1ng
device 1s 1n a state q1 and the head scans a symbol a 1 ~ of
the word P ( l ~ 2) Then the word
ail a J,_~q~aJ, aJ. ( 2)
ts called the configuratton of the machEne (at a glven ~~~

stant t) For l == 1, the configut&.tlon has the form 9JtiJ1
a 1,. If at Instant t the head scans a blank square located
to the left (or r1ght) of the word P, and there are v~ 0
blank squares between thiS SQUare and the first (last)
square of the word P, the tonfiguratwn of tht' mathlne at
the moment t ts presented by the word
q1A Aa11 a1, (3}
t~+t Urnes
(resp the word a1 a1,A Aq1A)t where A denotes

t "'- ,
a ....... a
v tiznes
the dummy symbol of the alphabet A If at the moment
t the tape Is blank'" 1 e a blank word conststtng of only
dummy symbols of the ex tern a I alphabet 1s recorded on
1t, the word q;A wtll be the confi.gurat~on of the machlne
at the instant t
Let the configuration of tl1e mach1ne at a moment t have
the fo:rm (2)t and suppose that the program of the ma
chtne contains the command
q~a J~q,ila l/~dtJ 1
7.1. TURING'S MACHINES 215
Then for diit = L, the configuration of the machine at the

next instant will be expressed by the word
(a) qiit Aau 1aJ 2 • • • ais if l = 1;
(b) q;j2aj 1aiipi3 • • • ais if l = 2;
(c) ail· .. aiJ-2qiitaii-taiizai l+l" .. ais if l > 2.
The cases when d;J 1 = R or dii 1 = S, or when the ma-
chine configuration corresponds to the head outside the
word P (as in the word (3)), or when the word Pis dummy
are described in a similar way.
If a program of the machine contains no quintuple of
"the form (1) for a pair (q;, a i 1), or if a "new" state qii1 is
the terminal state, the machine stops, and the "resulting"
configuration is referred to as the terminal configuration.
The configuration corresponding to the beginning of
operation of the machine is referred to as the initial
configuration.
Let K be the configuration of the machine at a certain
instant and let K' be its configuration at the next instant.
Then the configuration K' is called a configuration directly
derived from K (and denoted by K ~ K'). If K 1 is the
initial configuration, the sequence K 1 , K 2 , • • • , K n"
where K; ~ Kt+ 1 for 1 ~ i ~ m - 1, is called the
Turing computation. In this case, Km is said to be deriva-
ble from K 1 , and the following notation is used: K 1 ~
Km. If, in addition, Km is a terminal configuration, it
is said to be terminally derivable from K 1 and the nota-
tion K 1 1 . Km is used.
Suppose that a Turing machine T starts operating at
a certain (initial) instant of time. The word recorded at
this moment on the tape is called the initial word. To
put the machine T in operation, it is necessary to place
the scanning head against a square of the tape and indi-
cate the state of the machine T at the initial moment.
If P 1 is an initial word, the machine T which starts to
operate at the "word" P 1 will either stop after a certain
number of steps, or will not stop at all. In the former
case, the machine T is said to be applicable to the word P 1 ,
?nd the result of application of the machine T to the word P 1
IS a word P corresponding to the terminal configuration
(notation P = T (P 1 )). In the latter case the machine T
is said to be inapplicable to tlze word P 1 •
Henceforth, unless otherwise stipulated, we shall as-
2.16 CH 1 YVNDAMENTALS. OJ THE ALGOR11H\'I T11EOl\'Y
sume that (t) the tnJtial word IS non dummy (2) at the
Jnittal :moment the head 1s against the extreme left non
blank square on the tape and {3) the ma~h,ne starts to
operate v,hen 1t 1s tn the state qL
The zone of operatlon of the mach1ne T (on the uord P 1l
1s the set ol squares wh1cb are scanned by the head at
least once durtng the operation
\Ve shall often use the notatton (pJm for words of the
form P P P (m t1mes) If P -= a 1s a word of length t
v.e shall Wille am Instead of aa a {m tltnefi) and (a.lm
\Ve shall denote by JV art arbttrary fin1te word 1n the
external aJphahet of a Turing machtne (1n particular
a dummy word l e that cons1St1ng of dummy symbols
of the external alphabet)
\Vhtle descr1b1ng the operation of a Tur1ng mach1ne 1n
terms of contiguratJon language we shall use an ex
presston stmllar to
q\1 ~otrOJJ f- f 110q6lV
x~ l and y ~ 1 The express1on should be understood as
follov.. s the machine erases the word 1iX and stops at the
f1rst symbol oi the word lV If ho\vever lV IS a dummy
word the mach1ne stops at the seeond 0 (ze:ro) follJ)Wlng
the word tv
Tur1ng s machrnes 1 1 and T 1 are called equzvalent (1n
alphabet A) 1f for any tnput word P (tn alphabet A) the
rei a tton T1 (P) ~ T, (P) hoi ds Th1s express ton has the
foil o w1ng rncan1ng T 1 (P) and T z (P) are etther both
spec1fied or both not spec1f1ed2 and II they are spettfi.ed
then T1 (P) = T~ (P) The symbol~ 1s called the sym.
bol of condttwnal equaltty
Let maChlnes T I and T J have programs Ill and n.
respect1vely \Ve shall assume that the 1nternal alphabets
of these machtnes do not Intersect and that qa ts a final
state of the machine T 1 wh1le g1 IS an IDlt1al state of the
r
m aehlne 2 We shall replace the state q1 by the state q1
throughout the program fit and combrne the oblatned
program Wl th nJ The new program ll determines the
mach1ne T called the compos1twn of the machmes T1
and T 1 (over the palr of states (q1 q,)) and denoted by
' T (P l IS defined (not d e.fined) 1f the mactvue T IS a ppJu;a.ble

(mapphca.hle} to the wcrd P
7.1. TURING'S l\IACHINES 217
T1 0 T 2 or T 1 T 2 (in detail, T = T (T 1 , T 2 , (q~, q;))).

The external alphabet of the composition T 1 T 2 is the
union of the external alphabets of T 1 and T 2 •
Let q' be a llnal state of a machine T, and q" be a state
of this machine, which is other than fmal. We shall replace
the symbol q' throughout the program ll of the machine T
by q ". We shall obtain a program ll' determining the
machine T' (q', q"). The machine T' is called the iteration
of the machine T (over the pair of states (q', q")).
Let Turing's machines T1 , T 2 and T 3 be defined by pro-
grams Ilp TI 2 and TI 3 respectively. We assume that the
internal alphabets of these machines are pairwise non-
intersecting. Let q; and q~ be two different final states of
the machine T1 . We shall replace the state q~ throughout
the program TI 1 by an initial state q~ of the machine T 2 ,
q; q;
and the state by an initial state of T 3 • Then we com-
bine the new program with fl 2 and TI 3 • We obtain a pro-
gram II defining the Turing machine T = T (T1 , (g~, q;),
T 2 , (q~, q;), T 3). This machine is known as a branching of
the machines T 2 and T 3 , controlled by the machine T 1 •
In defining complicated Turing machines, the operator
notation of algorithms is often used in the form of a line
consisting of symbols of machines, transition symbols
(of the form Jq~ and ~~), and also symbols a and w used
to denote the· beginning and end of algorithm operation.
In operator notation (of an algorithm), the expression
Ti jq~o Ti ... Tm q~ 1 1 T 11 denotes the branching of ma-
chines T i and T 11 , which is con trolled by the machine T i.
Here the final state q io of the machine T i is replaced by
the initial state q111 of the machine T n, \vhile any other
final state of the machine T i is replaced by the (same)
initial state of T i· If the machine T 1 has a single final
l
state, the symbols jg 10 and ga 1 denote unconditional
transition. The symbols q10 and q111 are omitted whenevet·
a confusion is unlikely.
Example. The operator scheme
I
----.:2 1
j T 1aT2 / q.,w T 3 / q30 T 4w
I 2
describes the following "computational process". At ftrst

the machine T 2 starts to operate, If it sto.vs in the 'state
218 Ctl 7 FUNDAf-.lENT.ALS OF THE ALGORITJIM Tli.CORY
q20t the machlne ri beginS to operate When rt stops. r,

starts to operate again If t however, T 2 stops ID a final
state other than q1 tH the • work 1s conttnued,, by the rna
ch1ne Ta If thiS mach1ne comes to the final state qao~ then
T 1 starts to operate If, however, T a tetm1nates at a nnal
4
state other than q30 , the 'work 1s continued" by T 4 If the
mach1ne T" stops 1n a certa1n state, the computatxonal
process •s term1nated
....;
Let ct ~ (a 1 , a 2 , >
, an), n 1, he an arbitrary
tuple of non neg a t1 ve tn tegers The word 1 et, 01 ~;t+ ~ 0 +'
Ot ~n +I lS called the baszc machlne code (or Just a code) of
,..., ,/IIIJt,l
the tuple a (1n the alphabet {Ot 1}) and IS denoted by k (a)
ln part1cular, the word fa.+l IS the baste machtne coda
af the number a
Henceforth, we shall ma1nly cons1der partt.al numerltal
junctl.ons !he fnnctton I (x1 ~ x 2 , , :rn ), n> 1, 1s
a parttal numericaljunctlon 1f the vartables z, assume the
values In the natural scale N == (0, t, 2, , m, },
.,...,
and 1f I (a) E N when the funct1on J ts denned on the
I"W
tuple <Xr ~ {~, a.1 ., , d.n)

A numertcal part1al functton 1s satd to be computable
{~n the TUrlng sense) Ii there ex1sts a Turning ma
chtne T 1 hav1ng the folloWing properties
.,., """" ,..,
(a) 1f I (a) ts spec1fied t then T1 (k (a)) ;;;::;; k (I {a)),
(b) 1f I (a) 1s not specified, then e1thet Tt (k (ct)) 1s
""4 "'
not a. c.ode. of any number 1n N t or the maeh1ne T 1 cannot

N
be appl1ed to the word k {a)
Remark. Henceforth, we shall assume that at tl1e 1n1
t1al moment ,.,.,;
the head scans the extreme left uo1t:y of the
word k (a.) It 1s known that thiS. restrtc llon does not make
the class of computable functions narrow
If a funtt1on J 1s ~omputable 1n the Tur1ng senae by
us1ng a machtne T1f the machine T 1 w1ll be sa1d to compute
the funct~on I
7 l 1 ~ Find out whether a Tur1ng maeh1ne T spec1
tJ
tied by a program n lS appltc.able to a wotd p If tt 1Sl

determ.1ne the result of appl1catton of the machtne T to
the word P It 1s assumed \hat q1 ts the lnttial and fJa the
final statest .and at the tnltial moment the head scans
7.1. TU RI NG 'S l\IACHINES 219
th e ex tr em e le ft un it y on th e ta pe .
q10q 20 R
q11q 11R
q20q 30R
( 1) IT: qz1qt1 L (a) P = 1302 12,
q30q 00S (b) P = P O P ,
2 1.
q31q21R (c) P = 10 [01]
r q 0q 1R
1 2
I q11q30R
IT: qz0qa1L (a) P=1~01,
(2)
q2 1q21S (b) p = !3012,
6
q3 1q 11R (c) P = 1 •
( q10q 11R
q 1q20 R
I q 0q 1R
1
(a) P = 1012,
(3) IT: { 2 1 2 02 1,
q2 1q3 1L (b ) p = 1
q30q 11L (c) P = { 1 0 ] 21.
ab et {0 , 1 }, co ns tr uc t a T ur in g m a-
7 .1 .2 . In th e al ph sy m bo l
g pr op er ty (f or du m m y
chine ha vi ng th e fo ll ow in
we ta ke 0): w or d
is ap pl ic ab le to an y no n- du m m y
(1) th e m ac hi ne
in th e al ph ab et {0, 1 }; in th e
ap pl ic ab le to an y w or d
(2) th e m ac hi ne is in or d is
th e zo ne of op er at io n in ea ch w
al ph ab et {0, 1 }, an d
infinite;
in ap pl ic ab le to an y w or d in th e
(3) the m ac hi ne is d by th e
an d th e zo ne of op er at io n is li )l li te
al ph ab et {0, 1} nd on th e
re s, w hi ch do es no t de pe
same nu m be r of sq ua
chosen w or d;
ne is ap pl ic ab le to w or ds of th e fo rm
(4) th e m ac hi e fo rm
an y w or d of th
P " (n ~ 1) an d in ap pl ic ab le to
1 3
n+ cx, w he re a = 1, 2 an d n ~ 1;
hi ne is ap pl ic ab le to w or ds of th e fo rm
(5) th e m ac if a =F ~
1cx0 11 3
1cx01cx, where a.~ 1, an d in ap pl ic ab le to
(a.~ 1 an d ~~ 1) . it ia l
n T ur in g m ac hi ne T an d an in
7 .1 .3 . Using a gi ve ig ur at io n
in e th e te rm in al co nf
configuration K 1 , de te rm
(qo is th e final st at e) :
220 Ch 7 tuNDAME~T.ALS OF THE .ALGORI'tU\1 ,.HEQR...,
q, \1'1
(1) T 0 q.,1S q10R

r
~
t q;2DR q1tL
t1• q. Ql
(2) T 0 tJtOS qo1L q1tL
1 q 1 1R qsOR g10R
~
(a) K1 =;;;;; 1q1 t~, (b) K1 = q1 tS01, (c) K 1 =10glt~

7 I 441 Construct a Tur1ng macbtne 1n the alphabet
{0 1) whtch transforms the configuratron K 1 1nto KG
(1) K1 = q11n K 0 = q0 11tQ1" (n> 1)
(2) K 1 = q10n1", K o -= q0 {01.1n (n i ), >
(3) K 1 -== !nqtO~ K 0 = q 01.211 (n~ i)
(4} K1 = ft1q1 0tm4J K"' =1m q 001n (m~ 1 n~ 1)
7.1415. (1) Prove that for any Turtng maehtne there
exJsts an e.qutvalent machine whose program does not
contAin the symbol S
(2) Prove that for any Turing ma~htne tt IS possible to
construc-t an equtva1ent machtne whose program does
n0t conta1n final states
7 t~6- The program of a Turlng macb1ne T has the
form (the dummy symbol 1s 0 and tbe 1n1 ttal state IS ql)
q, <h 'II q,
0
I q1 1R q1 fR
I q1 2L j q1 3R
t q.,1L
I q,1R
I q~i.R
I q,1.R
~
2 qt2L
J
q10R
I q1 2R -
3 - I q,3JI
t q30R q:t3L
7.1. TURING'S ;'.IACHINES 221
'
(1) Prove that starting the operation with a blank tape,
the machine T will construct a word of the form P 11 =
110102 103 • • • 1011 1 (n is an arbitrary positive integer) in
t (n) steps, and at any instant t;;;: t (n) (for n;;,: 3) the
head of the machine will be to the right of the word P 11 _ 2 •
(2) Construct a machine possessing the same property
and having the set {0, 1} as an external alphabet.
(3) Provo that if the program of a Tul"ing machine does
not contain the symbol L (naturally, it may contain
symbols S and R), the machine, starting the opl'ration
with a blank tape, cannot construct a pre1'1x of the word
110102103 1 ... 0 11 1011 +1 1 .... however long.
7 .1. 7. Prove that for any Turing machine T (in an
alphabet A) there exists a countable set of machines
T1 , T2 , • • • , T 111 , • • • equivalent to it (in the alphabet A)
and diffedng from one another in their programs.
7.1.8. Let A = {a 0 , a 1 , • • • , a 11 _tJ he a certain alpha-
bet containing at least two letters. We shall encode its
characters as follows: the code of a; is the word 1Qi+11 in
the alphabet {0, 1 }. Accordingly to !his coding, the code
of an arbitrary word P = a;,a;, . .. a; s (s;;,: 1) in the
alphabet A will have the form 1Qi,+111Qi,+l1 . . . 10i~+11.
Prove that irrespective of the type of a Turing machine
T with the external alphabet A there exists a Turing
machine T 0 with the external alphabet {0, 1 }, which
satisfies the following condition: for any word P (in the
alphabet A) the machine T 0 is applicable to the code of
this word if and only if the machine T is applicable to
the word P. Moreover, if T (P) is specified, the code of
the word T (P) coincides with the word which is a result of
application of the machine T 0 to the code of the word P.
? .1.9. Construct the composition T 1 T 2 of Turing rna-
dunes T 1 and T 2 (for the pair of states (q10 , q:n)) and de-
termine the result of application of the composition T 1 T 2
to the word P (q 20 is the l'mal state of the machine T 2 ).
II q .. q" 'lu q.,
0 q120R q1 0 iL T,: 0 q~z1R qu1R

-
1 q 12 1R q 110R 1 qnOL q2 0iS
(a) p = 130212
'
2:22 CH 1 FUN OA~fE'lTALS OF THE ALGORJTII\f THEORY
q l ll • Qu V:t t.u
{) 'l1ofJL q 1,oR q1 10R 0 9:taiL 'ltoOR
1 gt1fR qlatR 'It lOR 1 qJtlL q110L
'~· qit qu Q'lt .,.. Qu
--- - ,.. I
- .... ., £&I
0 9t10R 9't30R q'to1L 0 q~,oL q 130L qz_z 0 0R
1 q 1 t..1R q11fR - 1 q~, t.f.l q~,1L q1,1L
P == 12 01 301' (b) P ~ 12 0f2Q212

(a)
7 t 10 Determtne the zesult of appl1eat1on of 1terat1on
of a machtnc r (for a palr of states (q.o q,)) to the word
P (q 0 .and q0 are the final states)
q:~ 4• q. q., q.
{1)_ t~1, T 0 <IoCS 9,os 9,os 1,},i.R q;,f.L
1 q;gO.R fa OR q10R - -
ql 'll t/1 'l.a q,. q;
(2) I=: J, T 0 g 00R q0 0R qtOR g1fL q.oL g00R
f fJsOR qsOR q1 tR q~fR g1 1L q,tL

7.1. TURING'S .l\IACHINES
--
')')3
q, q. 'J• ,,, •Is q,
0 q~OL - q~OR qslL q60L q~OR
(3) i = 3, T:
1 q1ZR q~lR - q~lR q;,iL qslL
2 - q3 1R - - - q0 1R
P=F0lY(x~1, y~1).
7.1.11. Detemine the result of application of the ma-

chine T = T(T 1 , (q; 0 , q21 ), T 2 , (q; 0 , q31 ), T3 ) to the word
P (q20 and q3() are the fmal states of the machines T 2 and
T3 respectively):
I q" I 'lu I 'h• I q.. I q,.

(1) T 1: Olq 120Riq~ 00R
-
T?: I 0 q20 1S TJ: ojq 3 ~1L,q 30 1L
11q12 1R,q;0 1L 1 IqziOR tjq3ltLj -
(a)P= lOP, (h)P =POL
'lll qtz 'l\3 q ••

"
0 ql20R q~ 00L q~ 00R T?:

-
1 q11 1R qia1R q13iR
I
q.,
I q,.
T3: 0 q3zOR q3o1S
1 q31lR q 31lR
(a) P = 1"'02 1 (x;;;;;:: 1),

(b) P= 1"'0101YQF (x~i, y~ 11 z~ 1).
224 en: 1 FU~OA~fF"'lTAtq OF TIIF ALGORITU\I THrORY
7.1 12 Us1ng mach1nes Tit T 1 T3 T" and Ti con

st rue t the opera tor scheme of the algor 1 t h m 21 (here
q10 t q$ 0 q26 qJrJ, q~ 6 and q50 are the final states of tlte
't
corr~s pond 1 ng mach lnes)
1} l Q;u q:t, <1~1 Chl
i) 'lliOR qtDfiR {} 9 ~~nR '/2s0R q~D:cs
t fJ•GOR q 1 ttS i 9ttiR rJt,_IR f/2,tR.
t.ltt q.... q., q"'t q"'~ q~ ..
0 q 3:a1R q301.S () q -,OL

4 q,. 3nL VtoOR 0 ifilot~
l q, 11R - 1 "iti1L q4 .tL q 43 tL 1 q.llt .R
(1) ~ q11;r h- q 0 1~ (:t#1)t

(2) ']l qt1x-.lJ- - qotl~ (x~O}t
(3} ~ lVOq 1.1.t:+J ( ~ IVOqo.12.;l'+t (z~O)
Remark. If z -= 0, then 1¥ IS assumed to be a dummy

word
79 I" 13. Us1ng the operator scheme of the algot1thm ~
and the descr1plton of the maeh1nes appear1ng 1n the
scheme of the algor! thm .. construct the program of s. ma-
chlne and determtne the result of appl1catton of the
ma.chtne spectfied by th1s scheme to the word P
7.1. TURIN(;'!> l\IACHINES 225
(1) lli: = a.Tl IT?.T31 q30 w,

l 1
'111 q,. q., q., q,. q.,
0 q100L 3 3
() q~ 0R q~ 0R q2 0 iL 0 qazOL q; 00R
1 qu:!.R 1 qHlR - q~ 3 lR 1 qat1L q3 2iL
?
~
- •)
~ q!~lR - - '}
~ - qao1R
(q11 , q21 and q31 are the initial and q10 ,

the final states of the machines}.
(2) ~l=a. 2jT,T2,t!oT31;3o.......,...,lw,
'lll 'lu q,, fJu q,. q,. IJu
0 - q130R q14 0L - 0 q220L

T2:
1 q120R qt21R q13 1R qtoOL 1 q~ 0 1S
,, :11 q,. q,. 1),.
T3: () qa2llL qa30R q~ 0 1S q3o1R
1 q31 1L q3 J1L - q3 J1L
P = 1xo1 Y (x;:;: ·1, y;:;: 1)

(qu, q21 and q31 are the initial and q10 , q 20 , q; 0 , q30 , and q~ 0
the final states of the machines).
t 5-0636
226 Clf 1 FU.SDA \IE\ T.ALS OF Tfi.C -\LLORITU\1 TJII:OR1
7 .t . 14. Construct a Turtng mach1ne computing the

luncl1ort I
0 tf z=-0
(1) f (x) ~ sgn z ~
1 if z~1,
1 If X= 1.
•
(2) f (z) = ( ! ]- · 0
not defined
If
If
x~2,
x=O,
JS
{3) f (.x) = ( ~ ]= m 1f ~ = 2m or x = 2m+ 1,

m~O;
0 lf :t~y.
(4) I (x y) ===; x - y =-
X - y 1f X> y.,
(5) I (:r, yJ == : t - g.
4.----.r
(6) f (.t, y) = !I~
Remark. I!crc (and below}~ some "ell kno\\n ~e]emen

t.a.ry' f unctJo.ns .are often used I or "anal yt 1C' represent a
t.1nn of num~tleal functtons An f,'analJt 1enlly'' presented
funct1on ts assumed to he spect{ied on1y on Integral tuples
of values of ~ar1ables (belonging to the natural scale 1V)
for wbtch all the uelementary functtOnS appearing In tt
g1ven analytiC representation of the function under con
s1derat1on are tptctfitd and assume 1nttgral non ntgatH.e
values For e.xample tho funct1on z2/(3- y/2) 1s specJfie.d
only when y/2Js a non negat1ve 1nteger ~ 3- yJ2 JS a po'i:1t1' .e
1n teger and .x1 /(3 • y/2) ts a non neg a 11' e 1 n teger
A Tur1ng mach1ne T ts sa1d to correctly compute a func
1'¥
t ion J (x"")
.,...,
1! ...... ~
(t) f(a.n) IS spec1f1ed, T (k (a.n)) ~ k (/ (a.fl)), and th€

head of the .machtne tn the terminal configurat1on scanct
.....
the left un1ty of the code k (/(a")),
~
(2) J (an) JS not spec1f1ed, and the mach1ne T start1ng lo

if/1¥
operate from the lefl unlty of the code k (an) does net
.stop
1. t .15. Pro-ve that for any com pu 1able f unct 1on, there
extsts a Tur1ng machine that correctly computes th1ot;
functton
7.1. TURING'S l\IACHINES 22i
7 .1.16. Construct a Turing machine which correctly

computes the function f:
(1) f (x) = x ..:.- 1; (4) f(x, y)=x+y;
X
(2) I tx) = sgn x = 1- sgn x; (5) f(x, Y)= 2 _Y.
. 1
(3) I (x) = x-~
'> ;
7.1.17. Using the program of the Turing machine T,

write the analytic expression for the functions /(x) and
1 (x, y) computable by the machine T. (For the initial
and final states everywhere, we take q1 and q0 respec-
tively.)
q, 'h q, 'h 'h q, 'I•
(1) T: 0 q2iL q00R (2) T: 0 q2 0R q 10L q 10L q 40L q0 0R
1 q1iR qziL 1 q1lR q30R q30R q,lL q6 1L
q, q, q, 'h 'I• q, q, q, 'It
(3) T: 0 q20R q30R q0 1S q5 0R q3 0L q70L - q 9 0L q10R
1 q20R q4 1R q3 1L q4 1R q6 1R q6 1R q80L q8 1L q91L
7.1.18. Which functions of one variable can be com-

puted by Turing's machines (in the alphabet {0, 1 }) whose
programs contain only one command?
7 .1.19. Is it true that
~ ,...,
two different computable func-
tions I 1 (xm) and f 2 (xn) can be computed on the same
Turing machine if and only if m =1= n?
7 .1.20. Let .M be a countable set of computable func-
tions and let T (M) be the minimum possible set of
Turing's machines, such that for any function f in M
there exists a machine in the set T (M) that computes
the function f.
(1) Prove that if for a certain n ~ 1 there e_xists in the
set l'vi an infinite subset consisting of functions of n
15•
228 CH 'I FUNDAMENT.ALS OF THE ALGORITHM THEORY
variables then there ex1sts ID T (M) a machrne wtth an

Indefinitely large number of states (1 e for any l 0 ~ t
there ex1sts 1n T (M) a maeh1ne w1tb a number of states
exceeding l,)
(2) Wha.t are the necessary and suffi.ctent condtt10ns
for the set T (M) to be fin1te?
7 .. 1. 21 Construet an operator sehellle of a T\lrtng ma.
ch1ne computing a functton f For elementary operators
use the machines T 1 (' = 4, 5t 6 7, and 8) In Problems
(1) and (2), first construct an operator scheme only through
the mach1nes T1 , T•, T1 and then represent eaeh of
them by an operator scheme us1ng T tot T ,, , T 1 The
Initial states 1n the macb1nes under con~uderat1on are
qtt• 9st' , q81 and the final st.ates are q 10 t q1.,, qllt
q:ao' q,o, q•ru ' q,o, !l,a
'Iii 9u o••
0 q~ 0 0R T~ 0 q~ 00L 0 "•~os
~
t 'ltuJB. 1. g•0 tL t .q11tR
q~ l flu.
0 fJ.'HIOS 0 q 10fS
f q1 t1L 1 g,~os
(1) f(x, y)~~+Y (mod 2},
... ltr Qu faa ,.. qu
ri) q,.~L -zu~O.R 10 qptL flt30L g10 ts ro 'lioOL
1 9'11tR i111L I q110L fl1lOL

- f g11 1R
(2) f (x) = 2x,

qui! • qto1 2 ,
Tt: q 11 ix+tl 7 101x0q~ 0 1 2 for x>O;
1x010 q ;1 11 l. qzo1 11 ,
1
2
Tz: fX(f+tf0 1q:u1111. fX+t01z0 1 q~ 0 1" for z>O.
Here x > 0, y > 0, t > 0;
T 3 : W0q 31111+1 I. W0q 30 P 11+1, y~ 0;
(3) f (x) = 3x;
(4) f (x, y) = x ·y;
(5) f (x, y) = x - y.
The distance between two squares C and C' on a tape is
equal to the number of squares between C and C' plus one.
In particular, adjacent squares are separated from each
other by unity. Let l be a positive integer. The subset of
all squares of the tape, such that every two of them are
separated by a distance multiple of l, is called a lattice
with spacing l. Thus, the tape can be regarded as a union
of Zlattices with spacing l. Let Rm be a lattice with spac-
ing l. Two squares of this lattice will be called adjacent if
the distance between them (relative to the entire tape)
is l. The word P = a1a 2 • • • am is said to be recorded on
the lattice R m if
(1) the symbol a 1 is recorded in a square C1 of this lat-
tice;
(2) ~he symbol a 2 is recorded in the square C2 which
is adjacent to C1 in the lattice R(l) and is located to the
right of c1, and so on;
(m) the symbolamisrecorded in the square Cm separat-
ed from the square cl by a distance (m- 1)l and located
to the right3 of cl.
A Turing machine T 1 is said to simulate a Turing
machine T on a lattice R(l) (with spacing l) if irrespective
of the word P (in an alphabet A), the following condition
is satisfied: let the word P be recorded on the lattice Rm
and let at the initial instant the head of the machine T1
scan the extreme left character of the word P; the machine
T 1 stops if and only if the machine T is applicable to
the word P. If T (P) is defined, the word T (P) will be
3
We .assume that only dummy symbols of the external alphabet
are outs1de the squares C1, C2, ••• , Cm on the lattice R(l).
.230 CH 7 FUNDAMENTALS OP THE .ALGORI111M THEORY
recorded on the lattice Rcn after the mach1ne 1 1 term•

nate-s 1ts operatton
7. t ~22. S1mula.te the operat1on of a maehtne T com
put1ng the lunctton f on a la t tJce WJ tl1 spactng l
(1) f(;t) = [ ..~ ] l =4, t
(2) f (z._ y} == sgn .t" ) l = 3,

!I
(3) f(:e, y):;;;:z+y, l=-3
7 1.23. Prove that lor any Turing machine T and any
1nteg~r l ~ 2, there ext~ts a math1ne T' s1mulat1ng th-e
mathtne T on a latllce Wlth spac1ng l
A word 1n the alphabet (0, 11 hav Jng the form
flCa:~+1)() 1 1l(a:,+ t)Q' 0 11 l(a ~+ J) (l ~ 2) w1ll be called
' ~
an l multl ple code of the tuple a/' -== (cc1 , a -a., .. a 11-l
7. t .. 24. (1) Prove that a machtne transforming the bas-
~
ac code of a tuple a.n to an l mulltple code of thts tuple

(l ~ 2) can he defined by the follow1ng operator scheme.
a.Tl IT, I qiO (t},
1 1
where
(a) T1 has the 1D1t1al state q11 and the final state q 1 ~~
and
quiQ;t+iotQ:•+io Otan. +l
~ 1a 1 + 1 03t~•+i01~ 1 + 1 0 01 m.n. +1 QZ•t J I,
r;- 910
(b) Tz has the JnltJal state qz 1 and two final states
g20 and q,~ and
q2t t~HQ%1 tzi+1+1(}1~Xh·l+l0 01 an+iOt.tti t(~t+t )Qlfl(o:fl+t )()l
oz1lla,+1 ~) l 1ZQ2j ct ••l+ t 01 Clf+l +I 0

~ q!O.
01 Cln + t Ol*lil(a l +I >ot"'"i t'<a, +I >ot ot 1l{a.i + 2 ~ ~)

for $>0,
q2.1102t~~.u+ 101ata+1o Ola"' +JQltttt<aJ +Oo•
Ol1l(Q.i +Ill - t""l•t+ t (}2J«i+:s!+ t Of~l+.tO
t • q,_o
0 tan+ l O'+ •li(a,l +I lQ~ .. .. O't I<a:t+ s,ot tl'
q2L1QI+L11(o:1+1>0'11(o:2+1)0I ••• 0'11<an+1)
11(a1+1)Q111(a2+1)Q! • •• Ql1!(o:n+l) .
I• q20
(2) Construct the programs of the machines T 3 , T 4 , • •• ,
T11 using the following description:
the machine T 3 , which starts to operate from the last
unity of the array of unities, "shifts" the array by one square
to the left (without changing the "remaining content"
of the tape 4 ); the head stops against the first unity of the
"shifted" array;
for a given l ~ 1, the head of the machine T 4 , which
starts to operate from an arbitrary square containing
unity, moves to the right until it passes through an array
of l + 1 zeros; the head stops in the first square behind
this array and prints 1 on it;
for a given l ;;;;::: 1, the head of the machine T 5 , which
starts to operate from any square and moves to the right,
prints l unities in succession and stops at the last unity;
the machine T 6 starts to operate at the extreme
left non-blank square; for a given l;;;;::: 1, the first left
array of l + 1 zeros is "detected", and the head stops at
the last of these zeros ("the content of the initial piece
of the tape" remains unchanged);
the machine T7 starts to operate from the extreme left
non-blank square, and detects the unity to the left of the
first array consisting of three zeros "bordered" by unities,
the head stops at the unity detected ("the content of the
initial piece of the tape" remains unchanged);
the machine T8 prints 0 in the initial square, and the
head stops after shifting to the left by a square;
the head of the machine T 9 is shifted to the right by
two squares of the "initial" square, and the machine stops
at the state q90 if the new square contains 0 and at the
q;
state 0 if the "new" square contains 1 (the content of
the tape remains unchanged);
the head of the machine T10 is shifted to the left by
one square (after that the machine stops, and no changes
occur in the tape);
~ In other words, we assume that not a single "new" unity has

appeared and the changes in the initial piece of the tape occurred
only in the indicated array.
232 cu 1 FUND.A:&lE~TA.Ls o.F Trne ALGORJTitii THEORY
the head of T11 starts to move to the r1gltt of an IIUttal"

square. io"detects" tl1e tirst (1n th1s displacement) un1ty
and., hav1ng made another step, stops al tha square lotat
ed to the rtght of the ''detected ' un1 ty-' (the con tent of
tho tape rema1ns unchanged)
(3) Taktng T 'it T • T11 ror tnt t1 a1 mach1 nes, con-
struct the opera tor scheme for the machtnes T 1 .and T1
and for the mac.h1ne transfurm1ng the bas1c ~ode of the
tuple 1nto an l mult1ple code
7 f 25 Construct the operator scheme of a Turing rna
#IW
ch 1ne transforming the l mul tJ pie code of the tuple a,"

1nto the bas1c code of thls tuple For tnittal machines cor
responding to elementary operators 1 use tl1e maeh1nes T,,
Ti~ , T 1 of Problem 7 1 21 and the follo,vtng three
maehtnes
the mach1ne T1 , such that for l~ t 1ts head. mov1ng
to the r1ght of a blank square, detects the first (In th1s d1s
placement} array c.ontatntng at least l urut1es then the
bead erases the first l unt t1es 1n th1s array and stops at the
square eontalnlng the last unt.t} wl11ch has been era.sed
( the rema1n1ng" content of the tape rema1ns uncbsnged),
the mact.ttne T 2 whose head 1s sh1fted from the ' 1nr\.1 a1
pos1t1on~ to the left by l squares (lis spec1fied)9 the ma
eh1ne stops at the l th square, and the content of the
tape rema1ns unchanged 1n th1s ease.
the maeh1ne Ta operates 1n the same way as m.acb1ne
T 1 , but the head 1s shtfted to ,.,.,the r1ght
The lattu:e code of the tuple an ~ (ct1 a. 1 , a") xs
the word 1n the alphabet {0., 1) recorded on n Jatt1ces
With spae1ng n so that the first 1att1ce conta1ns the word
1a:t+1 the secrund eonta1ns the word ta.r+t" ete , and the
l
n th latttce contatns the word 1o;n+J The beginnings of the

"ords on the Iatt1ces must be agreed u,ith each other
1 e the extreme left untty on the first latt-Ice Immediately
precedes (on the tape) the extreme left un1ty tn the second
latticef and th\s unlty rmmedlately pteeedes the ex
treme left un1ty of the third latttce~ etc
7 .t .26~ (1) Construct the operator scheme of a Turtng
'i"'J
machtne transforming the basiC tode of the tuple a; 11 Into
• Jf un1 t Y 1s recorded lD the .., D1tia Itt .sqll a.re. the he.ad stops a t
the ne1t. square: ta the t1ght
7.2. COliiPUTABLE AND RECURSIVE FUNCTIONS 233
a lattice code of this tuple. For the initial machines· cor-

responding to elementary operators, use the machines
1'4 , T 5 , • • • , T 8 in Problem 7.1.21. and the following
two machines: .
the machine TI, whose head is shifted from the "initial
position" to the right by n squares, stops at the n-th
square; the content of the tape remains unchanged;
the machine T 2 operates in a similar way, but its head
is shifted to the left.
(2) Using the same machines as in the previous prob-
lem, construct the operator scheme of a Turing machine
~
that transforms the lattice code of a tuple a" into the

basic code of this tuple.
7 .2. Classes of Computable

and Recursive functions
The functions considered in this section are partial nu-
merical functions.
The function F (xi, ... , Xn) = f (gi (xi, ... , Xn), •.. ,
gm (xi, ... , Xn)) is called the superposition of functions
I and gi, ••• , gm and is denoted by S (f<m>; g<~>, ..• ,
gl::l). Here the function F is_ defined on the tuple
f'oJ ,..., ~ ,...,
a" and F (an) = f (gi (a"), ... , gm. (a")) if and only
if each function gi (1 ~ i~ m) is defined on the tuple
~
an and, besides, the function f is defined on the tuple

~ ~
(gi (a"), ... , gm (a")).

Let g (xi, ... , Xn-I) and h (xi, ... , Xn-I• Xn, Xn+l)
be two functions and n ~ 2. We shall define the third
function I (x 1 , • • • , Xn-I• x 11 ) with the help of the fol-
lowing scheme:
I (xi, ... , X 11 _ 1, 0) = g (Xp ••• , Xn_ 1),
f (xil · · ·, Xn-1• Y + 1) (1)
h (xh ... , Xn-h y, I (x 1, ... , Xn-1• y)), y~O.
=
This scheme is known as a primitive recursive scheme
for the function f (;11) in variable Xn (and x 11 +I) and gives
~
a primitive recursive description of the function f (x11) in

terms of functions g and h. The function f is also said to
CH 7 FtTh.D.A.MEl\T.ALS OP 111£ .ALGORJTH\1 TB.t.ORl'
be abt4tned from g and h by u.&t.ng the prunittLt recurswn.

ope:ratlon tn. t.arcabk :r.,. (and r,~ 1 ) In th1s ease~ the fol~
low1ng notat1on lS used f ~ R {g~~ h) (the vartahles 1n
wh1ch the reeurs1on ts carried out. are 1ndtcated sep.arat~
ly}
In the prt.Dlltlve recurs1ve deS(rlpl1Gn of the funet1on
1 (x} depend1ng un a stngle v.ar1ahle the pr1m1tlve recur
SlOn scheme has the form
I (O)~ a,
(2)
J (y + 1) ~ h (y, I (y}}, y~O,
where a JS a constant (a number from the natur~l scale
N -= {0, 11 2 })
Let I (xtt t Xn. 1 , Zn) n~f be a certatn func-
tion \Ve shall define the function g (x1 , , X-rt-lt .:rn)
as follol\ s let a. """"" ==: ( a ., i: a. 11 _ 1 , Un) he an arbttrary

1
tuple of non neg a lr\ e 1.ntegers \V e shall con51det the
equat1on
f (a:J' , rJ.n-1 t y) :::==: Cln (3)
{a.) If ~ua.\J:a.n {3} ha~ a s.a.\\).\\\\n y0 E ~.v nnd {'3t ~ll
!I E N such that 0 ~ y < y 0 the funct1on I (a.1 •
an_1 , y) IS defi.nedt and 1ts values d1fier from a,. . , "e as-
,.,;
.sume that g (a) = YD

(b) If Eq (3) does. not have a solutton tn non negalt\8
,...,
Integers, we assume that g (a) 1s non defi.ned
(c) If Yo IS the smallest non negat1ve Integral solutton
of Eq (3) and for a certatn y 1 E Nand y1 < y 0 , the value
of I (ct.t t t an-t f y 1 ) 1s non defined "e assume that
./'¥
g (a.} 1s non defined

~
11
The funGtlon .,.., g (x ) eonstTue:.ted 1n th1s wal from the
function I (z11) ts sald to be obtalned from the functEon
I (x1 , ., Zn- 1 Zn) by uszng the operation tJ/ mlnimtJallon
w1th respect lrJ the vartable ..t'n (or Just the m~ntmizat1on 1n
Xn.) ln th1s case the following nola tton 1s used g === lfj,
rtW ~ ,..;
or g (z ) = ~I~-..,.,(j {.t71 )} ~ or g (xn) = 1:1. 11 (I {x1 ,

11
~
, Xn-tl
y) ~ Xn), or g (zt\) ~ Jl%a U (x-~'))

Remark. The opera t1on.s of prt:alt t1' e recurston and m1n~
tmlzation can be applted w•th respect. to any vartables
7.2. COMPUTABLE AND RECURSIVE FUNCTIONS 235
in the functions f, g and h (however, the variables should

always be indicated).
The following functions will be henceforth referred to
as simplest:
(a) s (x) = x + 1 is the successor function,
(b) o (x} 0 is the nullary function, and
(c) I~ (x1 , • • • , Xn) = Xm(1 ~ m~ n, n = 1, 2, ... )
is the selection function, or the function of selecting argu-
ments.
The class Kpr.r of all primitive recursive junctions is the
set of all functions which can be obtained from the simpl-
est ones by operations of superposition and primitive
•
recurs10n.
The class Kpar.r of all partial recursive functions is the set
of all functions that can be obtained from the simplest
ones by operations of superposition, primitive recursion
and minimization.
Remark. In the definition of the Kpr.r and Kpar.r classes
it is assumed that while constructing each concrete func-
tion, the corresponding operations are applied a finite
number of times including zero (some or all of the opera-
tions may not be applied at all).
The class Kg.r of all general recursive functions is a set
of all partial recursive functions defined everywhere.
It can easily he shown that Kpar.r :::> Kg.r (the inclusion
is strict!). The following strict inclusion is also valid:
Kg.r :::> K pr.r·
We shall denote by Kc the class of all partial numerical
functions that can be computed an Turing's machines.
The following statement is valid: the classes Kpar.r and
Kc coincide.
Theorem (R. Robinson). All primitive recursive functions
of one variable, and only these functions, can be obtained
from the 'junctions x + 1 and sgn (x • [ Vxl 2 ) by applying
a finite number of times the following three operations:
(a) absolute difference f (x) = I f 1 (x) - r!. (x} I:
(b) composition f (x} = f 1 (f 2 (x});
f (0) = 0,
(c) iteration
f (x + 1) = f1 (/ (x)).
7 .2.1. Apply the primitive recursion to the functions
g (xt) and h (x1 , x 2 , x 3 ) with respect to variables x 2
(and Xa). Write the function f (x1 , x~) = R (g, h) in the
2.36 CH 7 FUl\DAlrEl'r.TA.LS OP l11E ALGORITSM THEORY
analyt1c" form
(1} g (zl} == .rl h (Zt Z1 Z,t) = Zt + Za
(2) g (x1) = .r1 h (XI Zt x 3 ) -= Z 1 + %1
(3) g (z1) - 2% h (x1 z2 .r3 ) = z: (we put oo === t)
(4) g {x1) = 1 h (x1 Z 1 Z.t) =
:r 3 (t + sgn I x 1 + 2 2x 3 I )
h (.l\ :r'l. z 3} = (.r3 + 1} sgn (1 + =f)
g (z 2 ) =0h (x1 .r1 Z:~)

~ x,~ + sgn l rt + 1 - (.:r 1 + 1) z 1 J + sgn
--- a
the recurs1on ts carr1ed out 1n var1ables z 1 (and z 1 )

.x,
(3) 1(xl) ~ :rt- lV%;f'
g-- 0 h (1\ z~ ==(~2 +1) sgn(4x 1 - (z:+4xJ)
(4) I (x1)-- \V.X.1
g ~ 0 h (.TJ .t2) ~ Xz T Sgn
((x2 + f) 2 - (r:t +f))
7 2 4 Prove that If the functtons g (y} t11 (.r} <p, (z)
and ff 1 (x) are prJ m1 t1 ve recursive the fuuc tion
rt't (z) 1f g (y) ~a
f(x y)~ <p2 (z) tf a<g(y)~b
q 3 (r) tf g (y) > b
where 0~ 4~ b ts also pr1mtt1ve rec.urstve •
• The c.cndlt.t.an~ tmposed on the funct1nn. g (y) mus\ be undtr-
~tood as follo"'s we c:onsJder all such values of .. for wh1c:h 1he
f unc t1 on 1 (g) 6a t sfies the 10 d ca ted rela ~~ oD
7.2. COMPUTABLE AND RECURSIVE FUNCTIONS 23i
7.2.5. Let g1 (y), g 2 (x) and g'J (x, y) be primitive recur-

sive functions. Prove that in this case the function
1 (x, y) deftned by the scheme
I (0, y) = gt (y),
I (x + 1, 0) = g2 (x),
f(x+1, y+1)=g3 (x, y)
is also primitive recursive (here x ~ 0 and y ~ 0).
7.2.6. Let the functions g (x 1 , • • • , Xn_ 1 , Xn),
hl (Xtr ••• , Xn-1• Xn) and h!! (XI, • • • , Xn-1> Xn}, n~1,
be primitive recursive. Prove that the following
functions are also primitive recursive in this case:
xn
Xn) = 1} g (xl, ... 'Xn-1• i);
i=O
xn
(2) I (xl, · · · • Xn-t• Xn) = 11
i=O
g (xl, • · · r Xn-tr i);.
z .
•
~ g (x1, ..• 'Xn-t• i)
t=y
(3) f (xt, · · · r Xn-1• y, z) = for y:::;;;;, z,
0 for Y> z;
z
•
11
t=V
g (xl, ... , Xn-t> i)
(4) f (xl, ... 'Xn-1• y, z) = for y~z,
1 for y>z;
h2(xr. ... , xn_ 1, xn>

- ~ g (xl! •.. , X 11 _ 1, i)
i=ht(Xto ... , x 11 _ 1 , xn)
(here, as usual, the sum is considered to be zero if the
upper limit of summation is smaller than the lower one);
(6) f (xl! ... , Xn-l, Xn)
h2(xl, ... , xn-1' xn>
- IJ
i= h1(x1, ... , "'n- 1 • xn>
g (xl, ... 'Xn-t·, i)
(in the case when the upper limit of the product is smal-
ler than the lower limit, the product is assumed to be
unity).
238 C1I 7 FUND AI'c-fC 'l TALS OF TIJ.I: .ALGOR l TIJ 'l TO .tORY
7 .2..7 ~ Apply the mintmtzatto no peratton to the functton

1 1n x1 Represent the resultant functton 1n an "'analy-
t1c" farm
(t ) f (z t) === 3 ' -- 1J
{2) I (x1) - [ ~ ) 1
i= 1,
(1} I (x1 .ri)- 1: (x1 r 2)1 l!::::;;: 2,:
{4) ( (X l ~ 3: ~) :== E1 - x 2 "t l = f ,. 2.,
(5) /(X1 r,) = X, - _1 - 1
Z:t
I = 1\ 2,
( 6) I (x1 ~ x2) -- T' L (2 z: + 1) l i ==- 1, 2
7 2.8 AppJy1ng the mtntmtzat1on operation to an
appropriate pr1m1tavo recursive function, prove that
the function f 1s partially recurSIVe
(1) /(r 1 )~2-z 1 ~ {3) /(z1 , xz)~z 1 -2x2 ,
7.2~9. Js the follo'\\Jng statement correct 1f at Jeast

one of parttal recur~J ve functtons g and h IS not defined
everywhere~ then I = R (g~ h) tf K, r'
742 .. JO. (1) Can a funct1on that ts not defined anywhere
be ohta1ned by a s1ngle appltcallon of mtUimitat1on of a
runctton 'vh1cl1 1s defined every,vhere?
(2} Gtve an example of a prtmtttve tecursl\9 functton
whtch leads to a funct1on that 1s nut deflned anywhere by
apply1ng the m1ntm1tat1on operat1nn tw1ce
7 ~2~1 I. Prove the oomputabtllly of the follow1ng
funet1ons
(1) f (z:t Yt -')·zi f] (:t:- S"gn' (2" y}) -- (x+ t}r,

= [
(2) f (x fl, t} = { z~ 1 + 2 J} hf- xz), (:t/
2
(3) f (x y, z) ~ 4/t --"~ ~(zl 3

+ i}t_._l t
(4) f (3.:, y, t) = z;~:' 2(:.: 11 +viSgri(~ ... ll.tl

7 ~2~12 \Vhat ate the poVters of the ctasses Kp-r ,,. K 1 r.,
Kp..T 1 and Ke?
7.2. COMPUTABLE AND RBCUHSIVE FUNCTIONS 239
The function <p (x, y), defined by the scheme

cp (0, y) = y + 1,
rp(x+1, O)=rp(x, 1),
cp (x + 1, y + 1) = cp (x, cp (x + 1, y)),
where x;;;::: 0 and y;;;::: 0 is usually called Ackermann's
function.
7 .2.13. Prove that the Ackermann function satisfies
the following conditions:
(a) rp (x, y) > y for any x and y,
(b) cp (x, y) is strictly monoionic in both variables,
(c) cp (x + 1, y) ~ rp (x, y + 1) for any x and y.
7.2.14. Using the solution of Problem 7.2.13. and Ro-
binson's theorem (stating that the set {x + 1, sgn (x...:....
[V~P)} is complete relative to the composition, itera-
tion and absolute difference operations in the class of
all primitive recursive functions of one variable) prove
that irrespective of the form of the primitive recursive
function I (y) of one variable, there exists an x fo:r which
I (y) < cp (.r, y) for any value of the variable y.
7 .2.15. Prove that the Ackermann function is general
recursive but is not primitive recursive.
7.2.16*. Let us denote by KW.r and K~~~ the sets of
all unary primitive recursive and all general recursive
functions of one variable respectively. Prove that the
sets K~'i.r U {x + y} and K~~~ U {x + y} are complete rel-
ative to superposition in the classes Kpr-r and K~ r re-
spectively.
7.2.17. Let a Turing machine T compute the function
ft (x) E Kg.r'""-Kpr.r· Is it always true that the function
l2 (x, y) computable by this machine does not belong to
Kpr.r?
7.2.18. (1) Let Turing machines T 1 and T 2 compute
primitive recursive functions 11 (x) and 12 (x) respective-
ly. Is it always true that the composition T 1 T 2 also com-
putes a primitive recursive function I (x)? What will
happen if the machines T 1 and T correctly compute the
q
functions 11 and 12 ? •
(2) Let a machine T compute a primitive recursive
function f (x). Is it always true that if an iteration of the
rn~chine T computes a function g (x) defined everywhere,
tlus function is necessarily primitive recursive?
240 CH 1 JUNDJ.."M.ENtAt ~ O"F THF AlMORlnf~t 't'HrGRY
7 . 2.19 ~ Are the following relat1ons valtd1

(1) ~:c: (x -1)~(}£ 2 (x- 2)) .- 1
+
(2) l-':~:e. (x1 (x2 - x 1)) ~ 11-xt ((xt - Xa) + ~~)
(3) ~t$ ( Z - [ 1-]) EKFr r
(4) f-1~ (x- [YiJZ) EKFr r

(5) J.l~ ((}'ii)) EKpr r
7.2 20• Let the functions / 1 (x) and !J (a:) belong to the
set K, r '-._K p.t , Can the followtng s tatetnen ts be cor
rect?
(1) 1t {/2 (:r)) EK Pt rt b11t lz (f 1{x)) 4K Dr r
(2.) Is (x2 ) E KJlr ft but ll' f t {x 2)l EKpr r
+
(3) / 1(x) + /2 (x) E Kpr r• but /1 (z) 2/2 (:t) EK pr r
7 ~24121*. Let/ (x) EK 1 r ,K,r r Are the followtng rela-
ttons always saltsfied?
(1) f (2.r) t1 Kpr: r, (4} 1 - f (x) tt Kpr r'
(2) J (x + y) E Kvr:r't (5}1 {x- y} E Kprr
(3) I (x y) GKp r'
7 112 .. 22. (1) Botl1 vartables ol a functaon I (.r, y) 1n
Kr rare essenttal \Ve assume that l'xf (r y) and J.ly/ (x, y)
are the functtons defined everywhere Can at least one of
these functtons essenttally depend only on one
variable?
(2) The !unet~on j (:1:, y) E Kr t has one {let1tlous var1-
able Can both var1ables of the function J.lxi (x~ y) be es..
sent1 altl v. e assume 1n add1 tton that 1t 1s a general recur-
SIVe functron?
7 ~2 •.23~ Formulate the necessary and sufficient condr·
t1on that a funct1on Vaxf (r} ts not defined any\\here
7.2 .24.. ( t} \Vha t 1s the necessary and .suff1Cl en t co ndt-
tlan for the relatlun r=rf (x) E K 1 r to b~ lulnlled?
(2) Can at least one o the funet1ons ~xf (.x, y) or J.l" I (.t Y)
be general recursive tf I {x, y) E K J)l: r r""' Kg r,
7.2~25•41 Let f (x) E Kg r 'Kpr r Can the relat1on
p.J (z) E Kpr r he true?
7.2 ~2b It IS known that f (x) E K, r at1d that for any
z~O, I (2z + 1) =I (x) and f (2x) ::::= j (z + 1) It 1s true
that / (z) E Kvr r?
'1.2.27. Ftnd out "hether the set Kg r'
plete rela.b.ve to supel'poslhon m the class K~ r?
K r r lS com.
7 .:I. \.Ol\IPUTADILITY AND COJIIPLEXITY 241
7 .2.28. Does a set iii form a complete set relative to

the set of operations (9 in the class Kpar.r?
(1) ill=Kpar·r"Kpr.r• (9={R, f.Ll;
(2) M""' Kpar.r" K g.r, (9 = { S};
(3) M = Kg.r" Krr-r• (9 = {S, ~t};
(4) M = K pr.r• (9 ={f.!}·
2. 7.29. Let a general recursive function f (x) be such
that f (N) = {I (x): x EN} is an infinite proper subset of
the set N. Will the following equality hold
[(K~ ~r.r "K& .'r)
1
K~ ~r.r,
1 1
U {f (x)}J =
where KWr.r and K~~-~ are the sets of all unary partially
recursive and all general recursive functions of one variable
respectively and the closure is taken relative to super-
position and minimization?
7.3. Computability and Complexity

of Computations
A partial function F (x 0 , x 1 , . • . , Xn) is called uni-
versal for a family Y of functions of n variables if the
following two conditions are satisfted:
(a) for any i (i =0, 1, ... ) the function F (i, x 17 ••• ,
x") of n variables, belongs to :9;
(b) for any function f (x 1 , • • • , Xn) in '9, there exists
a number i such that for all values of variables x 17 ••• , Xn,
F (i, x1 , ••• , X 11 ) =f (x1 , ••• , Xn).
The number i is called the number of tlze function f (x1 , ••• ,

xn), and the numbering of functions of the family '§,
obtained in this way, is called a numeration corresponding
to the universal function F (x 0 , x 1 , . • • , Xn). Conversely, if
a. numeration of the family rJ is specified, i.e. if a map-
pmg cp: i ) f t of the natural scale on '§ is defined, the
function F (x 0 , x 1 , • • • , xn) defined by the formula
F (xo, X1, • •• , Xn) = fx. (xl, · · ·• x,.)
is u~iversal for §. Each primitive (partial) recursive
func~10n /(x 1 , •• • ,xn) can be associated with a term re-
flechng the way of representation of the function f (x 1 , ••• ,
16-0636
24.2 Cll 1 FU\D.AhiENT.ALS OF tHE ALGORlTH!'il THEORY
xJl) in terms of the functions /~ (x1, , z,..) ;;;; Xm

s (z) .= x + 1 and o (z) a 0 us1ng superposttaon. pr1m1 ..
t1ve recurs1on (and mtnlmtzat1on) By numbering all the
terms \\O can obta1n the numeratton for nil pr1m1t1\8
(partial) recutstve fu11CltOtts A numer1llon of th1s. l} pe
IS called a Godel numberlng1 IIenccforth \Ve shall as-
sume that a certain Godel numberang IS flxed A part1al
recurSt\ e funet1on of n vartahJes, havJng a number x
ln tb15 number1ng., w1ll ho denoted by f.P~u 1 he superser1pt
will be omitted tf there are no other stipulations con
cernlng the number of vartables of the function q:J~"' 1
The following statements are val1d
Theorent f (o11 a lunct1on wlllch is unlversal for the set
of all prtmlt1ve recur.slve functaons of n vartables)
The clasa of all prtmtttve recurslte juncttOIII of n tartablel
has a general recursive untter.ral fu.uctlon
Thts un1ver.sal funct1on (correspondtng to a cltosen
Godel number1ng) w1ll be denoted by D (z 0 t z 1 Zn)
Theoren1 2 (on a unl\ ersal !unct1on) There e:rtsts a
partfal recurlit.:e junction U (Xt• z 1, , Zn) tthich U
un t~ersal for the set of all partial recursive /unction.& of 11
~tartable.s
The concept of un1versal function 1s frequently used in
proofs Involving "diagonalizatton' As an example, \\e
can cons1der the following proof of the ~x1stence of a
general recurs1 ve func t ton w)lleh is not pr1m1 tt ve recur
slvQ Let D (r 0 ~ .x1} be a universal functton tn the class
of pr1mlt1ve recurst\e Cunct1ons or one var1able lf fol-
lo"s from part (a) of the defin1tio11 of the univcrsallunc-
tion thai D (x 0 .r1) is deftncd e\ erywhere Let us cons1d
et g (x) ~ D (;t :t} + 1 Tl1e funttton g (:t) is not prlm-
ltlve recurs1ve Indeed 1f th1s v..ere sot the equality
D (j, z) = g (x) \Vould be sat1sfied for a certa1n 1 and
for all x However, for :t ~ 1-. this equality ts transformed
lnto an inconsistent relatton
D U, /) = g U) == D (/. J) + 1
The follow1ng (s m n} tl1eorem~ put forth by Kleene
plays an important role
' Fo.r a formal defi.natlOD or the Godel numbering Bee, for

f\lmf.le, "Theory ol Reeunlve FunctJona and ERective Computa·
b1hty' by Rogelll II • !,lcGraw-HiJl New York. t967.
•
7 .3. COMPUTABILITY AND CO:\IPLEXI'rY 243
Theorem 3. For any m, n ~ 1, there exists a primitive

recursive function s<m+n such that for all x, Y1, ... , Ym,
the following relation holds:
cp~m+n) (yp •..• Ym· z" ...• Zn) = !pWk. !II• .... !lml (zt, ...• Zn)·
Let T be a Turing machine and K be a certain configu-
ration. The time complexity tr (K) of a computation process
is defmed as the number of steps taken by the machine T
during a transition from K to the final conftguration if
T can be applied to K. The function tr (K) is not defined
if T is inapplicable to K. Tho computation process zone
is the minimum part of the tape containing all squares in
at least one configuration encountered in the computation
process. The storage complexity sT (K) is defined as the
length of the computation process zone with the initial
configuration K if T is applicable to this configuration.
The function sr (K) is not defined if T is inapplicable to
K. By wr (K) we denote the number of changes in the di-
rection of the head displacement, and by rr (K) the maxi-
mum number of passages of the head across the boundary
between two adjacent squares of the zone during compu-
tation (the maximum is taken over all pairs of adjacent
squares). The functions wr (K) and rT{K) are not defined
if T is inapplicable to K. If for K we take the initial
configuration for the word P, the complexity functions
are denoted by tr (P), sr (P}, Wr (P} and rr (P) respec-
tively. If qr is a complexity function, then qT (n) =
maxq7' (P).
P: )..(P)~n
7.3. t. Prove that a function that is universal for
primitive recursive functions of one variable
(a) assumes all the values in N = {0, 1, 2, ... };
(b) assumes each value in N an infinite number of
times.
7 .3.2. Prove that each primitive recursive function of
one variable in a Godel numbering has a countable set
of numbers.
7.3.3. Prove that there is no partial recursive univer-
sal function for a family of all general recursive functions
of n variables .
. 7 .3.4, Prove that there exist partial numerical func-
tiOns that are not partial recursive. Give the proof based
on a comparison of powers of two sets: that of partial
16•
CH 7 J'VND.AMENT.ALS OP THE .ALGORITHM TIIEORY
recurslve functtons and that of all parttal numer1eal

funcliOJlS Gtve also the other proof based on ud•agonaiJ-.
zatton1 '
7.3 5_. G1ve an example of a parttal numertcal lunc..
tlon whtch assumes. exactly one value and whtch ts not
parttal recurstve
7 .3.6 Prove that the funct1on f 1s not a parl1al recur-
S\ ve functtan
tf the value of ll'::c. (y) 1s defined ..
1
(1) f(z,y)=
0 otherwtSe.,.
tf tp,x (z) 1s defined,
(2) I (x) ~
0 otherwise,
fl'.-.: (y) 1f 'P:c (y) 1s defined,
0 otherwise,
1 lf tF.c (y) = z,
0 otherw1se.
1 If there e:x1sts y for wh1ch
~% (y) ~ z.
0
otherw1se.
7 .3.7. Is the funtt1on I part1al recurs1ve tn the fol-
lowtng ca..sas
f 1f CfJ.r(x):= 1~
(l) I (x) ~ 0 otherwise,
JS not defined 1f 41'~ (x) IS defin~d,
{2} t (:t} =:=;
1 otherw1se,
f If the Untty belongs. to the set af
(3) f(x) ~ values of funetton "Px'
0 otherwise,
1 1f ~:t IS pr1.m 1\t ve recurs1 ve,
(4) f(z)~
0 otherwtse.
(5) I (z)
1 If the decimal decomposition of the number
n conta1ns an 1nfi.ntte number of zeros.
0 otberw1se.
7 .,3 .84' Let u (x,, z 1 , .. • , z n.) he a part1al rec urstve
func.ttan that. 1s un\Yersal for a non-emp\y subset J,f of
7.3. COJIIPUTABILITY AND Go:\IPLEXITY 245
•
general recursive functions, such that Kg.r "M is infi-

nite. Using the "diagonalization" process, indicate a count-
able set of general recursive functions not belonging
to M.
7.3.9. Show that if a function f (x) is partial recur-
sive, any function differing from f (x) on a finite set of
values of the argument is partial recursive.
7 .3.10. Let U (x, y) be a function that is universal
for the set of all partial recursive functions of a single
variable. Prove that the function f (x) = U (x, x) + 1
does not have recursive extensions (in other words, a
function defined everywhere, coinciding with f (x) where-
ver I (x) is defined and arbitrary otherwise, is not a par-
tial recursive function).
7.3.11. Let a partial recursive function f (x) be such,
that the function h (x), defined by the condition
0 at points where I (x) is defined,
h (x) = 1 at points where f (x) is not defined,
is recursive. Prove that the function f (x) has a recursive
extension.
7.3 .12. Give an example of a binary sequence a 0 ,
ct 1 , • • • , an, . . . generated by an appropriate autono-
mous deterministic operator and satisfying the following
condition: the function f (x), that is defined by the equali-
ty I (n) = an for all n ~ 0, is not general recursive.
7.3.13. (1) Prove that if the sequence a 0 , a 1 , ••• , an,
... is the output for a certain autonomous b.d.-operator,
the function f (x), defined by the relation f (n) = an for
all n ~ 0, is general recursive. .
(2) Is such a function always primitive recursive?
~
7 .3.14. Give an example of an infinite binary sequence
a.~ cto, a 1 , • • • , an, ... satisfying the following con-
dltwns:
. (1) there is no autonomous b.d.-operator for which a "'
1s an output sequence;
(2) there exists a Turing machine which starts operat-
.mg on a blank tape and constructs the sequence "'a;
moreover, for each n there exists an instant of time t 0 =
to (n), such that fort~ t 0 , the head of the machine does
not scan the squares of the tape to the left of the square,
where the symbol an is recorded.
CH 7 1U'VDA \IEu.'lT.ALS OF THE ALOORITll 'f Tltt:ORY
7.3~15.Give an example of transformatlon of nnlte

words wh1ch may be carrted out by a suitable Turing
mach1ne, but cannot be carr1ed out by any determtnlsti.c
operator
1-3"16~ For a giva.n function (, cons.truct a Turlng rna.
clllne T which can correcely compute I with .an uppet es
tim ate for 1ts t1me complexity, and maJortze the rema1n-
1ng eomplexit1 funetion3 The 1nput data. are giVen 111.
a unary form The external a1phabet 1$ g1ven by A=
{0. 1}
(1) I (Zt g) == z + vi tT (n) ~ en,
(2} I (t) ~ 2:r. tT (n) ~ tn 1 ,
(3) I (z) ~ f z ...... y
tr (n) ~en', f,
(4) I {z. g) = ( f]
t.,. (n} ~ cn 2 ,
(5) I (x) := Jlog 1 zf ~ tT (n) ~ cn
2
7 3. f 7. Construe t a Turtng machine T traasform1ng the

unary notatton of a number into a btnarY not.atJon w1th
gtven testttetions on the eomplexlty function t~ (n) <
c1n1 ,. sr (n) ~ n~ ror (n) ~ c1n rT (n) ~ c,n "here c1 f
c1 and cs are con!lants
7 3 ts. Constlntt a Turlng mach1ne T '\\hlth transform!
a b1nary notat1on into a unary natation, and such that
lr (n) ~ .:'tn2111 Sr (n) c1 n + 2'1 :r (n) ~ t 1 n2n,
fi'W t '-'
rr (n) ~ c,.n2 \ where t 1 , c1 c3 and &4 are constants

1
t
7 3 .. 19. Cnnsttutt a Tur1ng n1atl~ine T with an 1nput

alphabet A of m characters. transfor:n1ng a "ord P
1nto P•Pf where the sy:rnbol • does not appear •n Pt
<
and such that. lr (n) c1 n' sr (n) ~ c1n (j)r (n) .=:;; e~n
rT (n}~ c,n
7.3.20. (1) Construct a mach1ne Tt recogniZing the li
f'¥
near1ty 1 of an arbitrary Boolean funct1on I (zn) The

i'w .......
functton I (z") is defined by a h1nary vector a 1 of length

N ~ 2n The tnput alphabet of the machane 1s A =
{Ot 1, A} The funct1ons tT (N) and 'f.. (N) must sati&fy
the inequalt ttes t 1 (N) ~ cl N2, sr (N} ~ c 2N
(.2) Construct a mach1ne T, recogniz1ng the self dual1ty
Jll¥
of an 8cbltra.ry Boolean luactron 1 (;tn), and such that

trr (N) ~ cJ.N 1 , sr (N) ~ c1N
~--· ad 5 L
1
A machine recognlting a property of an input word has two
fmal stalest q; and. q; The mach1ce stOpl at the state q; 11 \he
property i.s observed, and at the Btate q0 if It is not
7.3. CO:\IPUTADILITY A1'\D CO:\IPLEXlTY 24i
(3) Construct a machine T, recognizing the monotoni-

city of an arbitrary Boolean function I (x "' 11 ), and such
that tr (N}~ c1 N 2, sT (N)~ c2N.
7 .3.21. Show that for any Turing machine T and for
any word P
(1) ST (P) ~ tT (P) + I p I ;
(2) there exists a constant c, such that tr (P)~ c&riP>.
Chapter Etghf
Elements of Combinatorial Analysis
8.,. Permutations and Combinations.

Properties of Binomial Coefficients
The tuple of elements a 1 • , , a 11 an tl1e set U =
{a1, , an} ts cal led an r mult,subset of n set U, or
an (n, r) multtset If the sequent1al order of e1ements 1n ~
multtset 1s spec1f1ed,. we call 1t an ordered multtset TVto
ordered multtsets~ dtfferlng only 1n the ~equent1al order
of their elements. are assumed to be dtfferent If the se
quent1al order of elements 1s not s•gnlr1cant, the mult1
5et 1s c.allt!!d a dlsordtred mu11Jset Elements may, or may
not~ repeat themselves ln a m ul t1set An ordered (n r)
multtset 1n wh1ch the elements may be repeated 1s calJed
a permutation of n elements wtth repellltons taken r al a
ttme, OT an (n, r) pf'rrnuta.t1on 1c1th repetltlons If an ordered
(n~ r) mult1set has pa1rw1se d1Iferent elements, we Lalltt an
(n, r)~permutatlon Wlthout repetlllons., or simply an (n,
r) permu.tat1on The number of (n, r) permutat1ons v.,Ill
he denc-ted by the symbol P (nt r), whlle the number of
(n"
... r) permntattons w1th repetitions w1ll be denoted by
P (n" r) A disordered (n:" r) mult1set~ 10 wh1ch the ele-
ments can be repeated, IS c:aJied a comblnatlon of n ele·
ments taken r at a ttme, or simply an (n, r)-eombinatton
wtth repe11taons If tl1e elements of a d\sordered mult1set
are pairwise dtfferent, It 1s called a t:omblnallon (wlthout
repetltlon) of n elements tal..en r at a t~rne't or an (n, r) . .
comblndllon Each such combanatton ts a subset of power
r ln the set U The number of combltl&tlons of n elements
taken r at a ttme w1ll he denoted by C (n~ r), 'vh1le the
number of combtnatJons (w1th repetttJons) of 11 elements
A
taken r at a ttme wtll be denoted by C (fl, r)

Example 1. Let U -= {a, b, c] r = 2 Ju thiS case
-we obta1n
nine permutations witb repetitions, "1z.. (ld, Q.b~ ac\
ba, bbr be, ca, cb, c&l
8.1. PERliiUTATIONS AND CO:'IIDINATIONS 249
six permutations without repetitions, viz., ab, ac, ba,

be, ca, cb,
six combinations with repetitions, viz., aa, ab, ac,
bb, be, cc, and
three combinations without repetitions, viz., ab,
ac, be.
The product n (n - 1) ... (n - r +
1}, where n is
real and r is a positive intege1·, will be denoted by (n)r.
By defmition, we put (n} 0 = 1. If n is a natural number,
(n) 11 is denoted by the symbol n! and is called n-factorial.
For n = 0, we assume that 0! = 1. For any real n and
a non-negative integer r, the quantity (n}r/r! is called a
n
binomial coefjicient 1 and is denoted by the symbol •
r
Let r 1 , r 2 , • • • , rlt be non-negative integers and let
r 1 + r2 -r1 ... -r1 r 1, = n. Tl1e quan t't1 y n! ·
1s
r1.1 r 2.•..
1 rh.1
called a polynomial coefficient, and is denoted by
n
•
r 1, rl, •.. , rlt
The following two rules are taken into consideration
while counting the number of different combinations.
Multiplication rule. If an object A can be chosen in m
different ways, and if after each such choice an object
B can, in turn, be chosen in n different ways, the choice
"A and B" in this order .can be carried out in m X n ways.
Summation rule. If an object A can be chosen in m
different ways and an object B in n other ways, and if
the choice "A and 8" is not possible, the choice "A or B"
can be made in m + n ways.
Example 2. Two dice (with six faces each) are cast.
In hmv many different ways can they be cast so that
each of them shows either an even number, or each shows
an odd number.
Solution. Let A be the number of ways in which an even
numbe1· is shown by each die, and let B be the number of
ways for an odd cast. Then, according to the sum rule,
the required number is A + B. Let C be the number of
ways of an even cast by the Jirst die and D the number of
ways for an even cast by the second die. Obviously,
1
. The notations Cf., nCro and (n, r) are also encountered in the
111erature.
=
C ::::: D 3 ,\ccording to the mtlll1pltcat1on rule
A === C X D == 9 S1milarly B -= 9 Hence lhe total
pQSSlble number o[ ways ~s equal to 18
Example 3 Show that the number of (n r) permutat1on~
w1thout repet1t1on.s 1s equal to n'
Solullan Indue tlon on r Fat r == 1 tlLe number of
ways 1n v..hrch one obJect can be chosen from n 1s equal
to n' Suppose that the equal1ty P (n r) == (n),. 1s sati.s
fted for a c.ertatn r ~ 1 \Ve shall prove that an analogous
equalaty also holds for r + t Each set of r + 1 obJeCts
can he obtained hy first choostng r obJects com prHung
the (11 r) perm uta t1on fol lo\\ ed by the addttJ on of
the (r + 1) th obJect to It If r objects are cho~en there
are n, · T ways 1n wh1.Ch the (r + 1) th ObJeCt. can be
cltosen Accord1ng to the mult1pltcat1on rule we ohta1n
P (n r + 1) = P (n r) X (n ,..._ r)
Ustng the 1nduct1on hypothesi! we can wrtte
P (n r + t) =:M (n), X (n. r} ::t (n)r+J
A largo number of combluatotlal problems can be re
duced to a computatton of the nurrtber or binary vectors
£xatnple 4 In how man)' Wa)! can the number n he pre
sented as a sum of k non negative components (repre~en
tat1ons d1ffering only 1n the order t r the components are
R.ssumed to he d1Herent)?
Solution Each decornpoSllJon or the number n inlo
k non negative 1ntegral components i.!l asstgned a 'ector
of lt.nglh n + k 1 w1th n \lnlttes and k -1 teros
such that the number of un1 t 1es just before the first zero
1s equal to the first component the number of un1t1es be
tween the nrs.l and Ee~>ond tero.s is equal to ll\e s-e-cond eom
ponen t and so on Th1s ts a one to one correspondence
It should be noted that each b1nary vector \\Jth n + 'h -
t coord1no.tes and n un1.t1es can ln turn he assoel&t.ed
With an n clement subset A of the .set U == {a1 a 1
an+-' 1 } as follo'\\S tbe 1 th coord1nate of the \ector 1s
equal to 1 tf and only tf a 1 E A t = t 2 n +
k - 1 But the number of such suh:sets 1s C (n r)
8 t i In l1ow many \\als can tbrce t1ckets hP d1str1b
uted among 20 students 1f
(1) The ttekets are for dtfferent theatre$ and each stu
dent can get not more tltan one t1cket'
8.1. PER:<\IUTATlONS AND COMDINATlO:-JS 251
p (n, r) = nr,
C (n, r) = ( ~ )
1
,
(3) C(n, r)= ("~.:_~ ). 1
,...,
8.1.4. Find the number of vectors a = (ct. 1 , • • • , ct.n)
whose coordinates satisfy the following conditions:
(1) a 1 E{O, 1, ... , k-1}, i=1, ... , n;
(2) a. 1 E {0, 1, ... , kl-1}• i = 1, ... , n;
(3) a 1 E {0, 1 },i = 1, ... , n, and a. 1 + ... +au =r.
8.1.5. (1) Find the number of matrices having n rows
and m columns with their elements in the set {0, 1 }.
(2) Find the same, provided that the rows of the matrix
are pairwise different.
8.1.6. Given n objects of one type and m objects of
another. Find the number of multisets containing r objects
of one type and s objects of the other type.
8.1. 7. Words are formed by using n letters, of which a
and b appear a and ~ times respectively, while the re-
maining letters are pairwise different. How many differ-
ent r-lettered words can be formed, so that the letters a
and b appear h and k times respectively?
8.1.8. A deck containing 4n (n ~ 5) cards has n cards
each of four different suits, numbered 1, 2, ... , n. In
how many different ways can five cards be chosen, so
that they contain
(1) five consecutive cards of the same suit?
(2) four of the five cards with the same value?
(3) three cards with one value and two cards with
some other value?
(4) five cards of the same suit?
(5) live successively numbered cat•ds?
.2S2 Clf 8 EL.E)[EJrr. TS OF COlm I~ A TOR IAL A "'l A. LYSIS
(6) three of tbe h"\ e cards w1th t l1e same value'

(7) not more tl1an two cards of the same sutt?
8.1.9. Sol\e the lollov.tng problems by usJng the .sum·
matJon and mult1plJcatton rules
(l) In ltow many \\ays can two ch1ps be choseu irom
among: 2S 1n a. g~t.me of dom1no, suc.h that they can M: put
next to e.ach other (1 e ~ the s.ame number IS encountered
on both ch1ps)1
(2) Th1ee d1ce are cast In l1ow many wa}s can they
snow \he same numbe-r or patr\\lse d1fterent numbers?
(3} It JS customary to g1ve more than one chr1sttan name
to the newly horn baby 1n England In how many d1ffer
en t wa} s can a baby be named 1f he ts to be g1ven three
names from a total of 30(Y~
8~1-10 (t) In how many ways can the number n be
presented as a sum of 1., natural components (representa
t1on.s dJffer1ng only tn the order of components arc as
sumed to be dtHerent)'
{2) In l\OW many ways can the num~~ 7n be 9te.rented
1n the form of a product of three cofactors (repre.senta
ttons dttler1ng only 1n the order of cofactors are assumed
ro be dtflerent)?
(3) The same. provided that representations dlBer-
lng only tn the order of cofactors are assumed to be 1den·
t1cal, and n ¢. 3s?
8.1. t 1. In how many wa}s can n zeros and k un111es be
arranged so that @-ath two un1t1es are separated by not
less than m zeros'
{2) F1nd the number of non negattve Integers not ex-
ceeding 10n whose d1g1ts are arranged tn non dtmtnishtng
order
(3) A rectangular c1ty IS d1v1ded by streets lnlo
squares There .are n such squares from north to south, and
/.. squares from east to \\est Ftnd the number of shortest
walks from the north eastern end of the ct t y to the south
western end
8 .. t -12~ Let n be the product p~ 1 X X p~' of
powers ol palrwrse dJfll!rent prfme numbers F1nd
(i) the number of ali natural diVIsors of Lhe number
nJ.,
(2.) the number of a1l d1"Y1sors that cannot be dtv1ded by
a square of any tnteger other tb.an t, and
(3) the sum of dtvtsors of the number n
8.1. PERMUTATIONS AND COMBINATIONS 253
8.1.13. Prove the following properties of binomial

coefficients:
(1) (; ) = ( n~k ) ;
(2) ( ~) ( ~) = (~=~) (;) ;

(3) (~)=(n-;i)+G=!);
(4) (k~r )/( ~ )= (n~!)+r)r ;
n
8
<> ~
r=h
(~J=GtD·
8.1.14. Prove that
(1) ( ~ ) increases with n for a· fixed k;
(2) (~=~) decreas'es with r for fixed n and k;
(3) if n is fixed, then ( ~ ) increases with k for
k < ]n/2[ -1 and decreases for k~ ]n/2[;
(4) max (
o,;;;;k,;;;;n
~ ) = ( [nj21 ) ;
(5) the minimum value of the sum ( nk 1
) + ( ~2 + )
•
. . . + ( n; ) provided that ~ n; = n is equal to (s-r) X
i=1
(~)+r(qti), where q=[n/sJ, r=n-s[n/sJ;
(6) the maximum value of the sum ( ~ ) + (; + 2
)
· · · + ( k~ ) provided that O:::;,;k 1 < ... < k 5 ~n, 1 ~s~

n + 1 is equal to ~ ( ; );
n-s,;:: .<n+s
1
2 "" 2
(1) for a pr1me p and for any p > k~ i, the number
( : ) Ul a mult1ple of p.
(8) II p,-.r::;;, ( z;) t where the product is taken
n<p,~2n
o\ er all prime numbers p 1 (n < p 1 ~ 2n)~
841t ~15. (f) Let m be a non . .negat1ve 1ntegert and let
n == n (m) be the m1n1mum 1nteger, such that m <:: n1
Prove that there IS one and only one way 1 n wh1ch the
.....
number m can he assoctated w1th a vector a (m) ==
(a, a 2 , , O:n _ 1). such that m ::.:: a 1 x 1 r + a 1 X 2J
+ + a~- 1 X (n.-f)J. 0~ a.1 < 1 t == ~..
~
n-t
f"'V ~ f'IV
Let Jl (a.) be a number, suc.h that a -=== a. (~ (a))

#w
(2) F1nd the vector a (m) for m ::=: 4~ 15, 37

~ fw
(3) Ftnd t.t (a) !rom the fo I to\\ tng values of «
~ ,..; /fly
(a) a == (Ot 2. 0, 4), (b) a ~ (Ot 2t t)t (c) a === (!. 2,

3, 2)
(4) The permutat1on ol a set Zn = {1 2, t n}
1s defined as an arb1trary mapp1ng of Z n onto 1tselt Any

permutation n ean he put 1n a one..to one correspDndence
... =
"'1 t l1 the "ector n (n (t ), n (n)) tn wh1cb tl1e
11 (') coord1nate tndtcates the pos1t1on of tbe element l
Eaeb permutat1on :r IS put 1n correspondence w1th a
number v (n:), 0 ~ v (n) < nl, called the number of the
permutatton
,., For th1s purpose, we first construct a
vector a, === o:" _1 ) \Ve put a"_' :=: n (1) -
(a 1 • a 2 , f
1 lf Gln-1 , ~ CXn-J are already defined and s (I) =

1{i <I J n (~) < n {!)))~ we put an J~t = tt (J + 1)- ,.
s (J + 1) -- 1 The number v (n) JS defined as 1.1 (a.)
._.. ~
For example, 1f n. -= (3., 4t 2,. 1), then C£n {0, 2t 3), ::%.
v (.n) :;;:; 0 X 1 l + 2 X 21 + 3 x 3! = 22 Frcm the

....
pennutat1on .n defined by the vector n. find the nurnber
...
v (n) 1n \ht) follow1ng cases
... .....
(a) :~ :::; (2, 3, 1, 4)t (b) n ::; (3f 5. 2, 1 t 4), (e) :t =
(1. 4, a. st 2)
(5) F1nd the algorithm for constructing the permuta-
tion n from its number~ (n)
S.l. PER:.\tuTATlO:-\S AXD CO:IIBIX.-\TIOXS 255.
(6) From the number m, find the permutation n on the

set Z 11 , n! > m~ (n- 1)!, such that v (n) = m:
(a) m = 7; (b) m = 18; and (c) m = 28.
8.1.16. (1) Let k and 11 be natural numbers. Prove
J
that an~· integer m [ O~m < ( ~ } can be uniquely
,..,
associated with an integral vector ~ (m) = (~ 1 • ~ 2 , .... ~ 11 ),
sulisfyiug the condition n > ~~ > ~ 2 > ... > ~lt~O.
m= { ~1 ) +(
k~ 1 ) + ... +
(~h) In this case, the num-
.
""'
her m is culled the number of the tuple~ (denoted by
Tn= ~ (~)).
(2) For given ""'

m, n, and k, construct the vector ~ (m):
(a)m=19, n=7, k=4;
(b) m = 25, n = 7, k = 3;
(c) m = 32. n = 8, k = ft.
,..,
(3) From a given vee lot· ~ = (~Jt ... , ~It), fmd a num-
ber m,... satisfying the conditions
,..,
of Problem 8.1.16
,..,
('1):
(a) ~ = (6, 3, 0); (b) ~ = (5, 4, 3, 1); (c) ~ = (6, 4.
3, 2, 1).
(4) Let B~ be the set of nil vectors of length 11
having k unities and n- k zeros. U~ing Problem
8.1.16 (J), enumerate all tuple~ in B~ from 1 to (%),
i.e. construct a one-to-one mapping v of the set B~ onto
the set { J, 2, ... , ( ~ )} .
8.1.17. Using the !elation
prove the following identity by induction on n:

II
(1 + w= ~ ( ~) t". (1)
h=O
. 8.1.18. Let nand m be positive integers. Using the iden-

tity (1) or some other method, prove the following
2.56 CH 8 ELC!\IENTS OF CO\tSI"iATORIAL A~ALYSJS
equaiJtles
n n-k
(13) J: (- f)ll r { ; } :--~ ~ ( -1Y' ' r { ; ) ,
,.~Jrt Jl"::zD
S.t. PERMUTAT IONS AND COl\!BJNATIONS 257
n-k
(14) ~ (k~r ) {~t) m+n).
( 11-k '
r=O
for m =!= 11,

for m=n.
k=n
8.1.19. Prove the following identities :
(!) ~ U~)=~ (2k~1 )= 2"-';

k k
(2) 4 ~ u~) =2"+ 2y+i cos n~l ;

k
(3) if o:::;;;r < m, then
m-1 2nirv 2niv
m~ (mk ~r)=~ e-
1
m (l+e m )", whez·e i2 = -1;
k V=O
( ")
.q
~
LJ
(
4k+ r
It ) 1 (2
= 4
n+ 22f+t cos n (n-2r)) ,
4
k
(5) if o:::;;;r<m, m~1. then
~ ( 1-(m-1 )cos
11
~ )~~ (mt+r)
h
:::;;; ~~ (t + (m-1) cos" ~ ). .

8.1.20. Prove that
k
mk+r v=O
Using the identity (*), calculate

(2) ~ 311. (2~);
k
(3) ~ ( --1)k 211. (' n ) •

k 4k+ 1 '
(4) ~ ( _ t )k 31t ( n ) ( 2k + 1 )
k
2k+ 1 r •
17-0636
258 CU & EI.E:'iiEl\ TS OF CO 'lDI'VATOR I A L .A' AL\'SIS
8 1 21 F1nd the number of rat1onal terms 1n tl1e ex

pans Ion
<t> CJ12,- V3) 2
f (2) (Va + V5) 50
<~) <Vt;+V2) ~
10
(4) (Vi2'+ Ya>3a
8 1 22 F1nd the coeffictent of l 4 Jn the elpanstons
(t) (t + 2t - 3t 1) ' h = fJ
(2) (1 - t +
2t1 )l0 k ~ 7
(3) (2 + t - 2t 3 ) 20 h = f0
(4) (2 t1 + +
t 7) 1' h. == 17
8 l 23 Show that the follow1ng Identlties are valJd for
1ntegral m ~ 0 and n 0 >
n m
(2) ~ (m+:-t }_. ~ ( n+: t }
k~G ~-~
B f 24. Let a and b be real and let k m n and r be

non n.egattve 1ntegers Pro\e that
(1) (: )+{k~t} :-{atf)~
-ao
{2) {1 + t)a = ~ { : ) tA I tl < 1

A-D
(3) (-:)---<-tl'{ +:-·) 4

«>0
(4}
"'~ (4J--;k)-- {:t!)- (~+~)
..\;eO
.,..
(5) :E (~ ) ( ~ k) ·-- c•t b)
n (add1t lOll tlleorem)
A~D
Jl
(6} ~ (-1) _.(:}~(~~~)

~~.o
•
• 8.2. INCLUSION AND EXCLUSIO:-< FORMULAS 259
•
00
(10) ~ ( 2k+t 112 ) ( 1 )h._

2 -
v- -.
5- V3,
k=O
(11) m "' ( a ) bmk+r

LJ mk+r
h.
~
2
~:rv (1 +be ~"' r,
1 2
= e- 1 b I < 1.
v=O
8.1.25. (1) Find the number of all words of length Jnn
in an n-lettered alphabet, in which each letter is encoun-
tered m times.
(2) In how many ways can a set of n elements be decom-
posed into s subsets the first of which contains k 1 ele-
ments, the second k 2 elements, and so on?
(3) With the help of combinatorial analysis, prove that
the following equality holds for any non-negative inte-
gers k 11 k 2 , • • • , k 8 , n,. such that k 1 k2 ks = + + ... +
n:
( ~) (n~kl) ... c~-kt-kzk-:- ... -ks-1) kt! k2!n: .. ks!.

(4) Prove the following identity by induction on s:
ks
• . • ts •
8.2 Inclusion and Exclusion Formulas

Let us consider N objects and n properties A 1 , • • • , A 11 •
Each object may, or may not, possess any of these prop-
erties. We denote by N ;, , ... , i the number of objects
having the properties A i, , . • . , h. A ;h. (and perhaps some
A
other properties as well). In this case, the number N 0

17•
260 CH 8 .ELEMENTS OF COMBINATORIAL ANAL YSlS
of obJects not possessing any of the propertieS A 1, -. A.

ts defined by the equalt ty
A
No:=S0 -S1 +S:.- +(-t)nSn, (2)

where S0 = N, and
s,. =::;; )J Nj L ,n
t<-t51< <1~~;~"
Formula (2) 1s called the lnclusion and exclusion formula

Example t. Suppose that a deck conta1ns n cards, num
bered from 1 to n In how many ways can the cards be
arranged in the deck so that card number l does not
occupy the 1-th poo1t1on for any value of t (1 ~ 1 ~ n)?
Solution. w~ have n properlleS a., t){ the form ~'the
t-th card occuptes tl1e t-th positlon 1n the deck'' The
number of poss1ble arranltements of the cards In the deck
1s equal to n! The number of arrangements N t, ih
for whl~h a card W\th number f~ occupies the
positlon tv (v ~ t, k)t equal to {n - k)t Th.ts g1ves
Us1ng formula (2) we find that the nurnber N 1 of the

arrangements for wh1eh none of the properties a 1 1s sat1s·
fi.ed, IS equal to
ft n
~ (- -1)~8~ =nl ~ (---·1),. :c

A~o 1~o
If the number of properttes IS not very large, such prob·

lems can be conveniently solved w1th the help of Venn's
Circles
Example 2., ln a group of 25 students, 20 have successful
ly passed theJr exam1natrons Out ol the 12 students ell-
gaged 1n sports) 10 have passed thetr exams How ntany of
th-e unsuc.eessful students do not take part )n spo1t.s
acttVlttes?
Solution. We represent the group of students who have
.successfully passed thetr examtnat1ons by a Clrcle
8.2. INCLUSION AND EXCLUSION FORMULAS 261
•
marked A (see Fig. 40), and the group of students participat-
ing in sports activities by a circle marked B. The intersec-
tion of the circles corresponds to the group of successful
students who participate in sports, while the union of
the circles corresponds to the group of students who have
either passed their examinations or take part in sports
Fig. 40
activities. The number of such students is equal to 20 +

12 - 10 = 22. The number of unsuccessful students
who do not take part in sports, is equal to 25-22 = 3.
Venn's diagrams are used for solving problems involv-
ing the counting of number of elements in the set U =
{a1, ... , aN} having given properties. Venn's diagram
for n properties has the form of a rectangle divided into 2n
squares. Each square corresponds to one type of elements.
The type of an element is determined by whether or not
the i-th element possesses the i-th property for each i,
1 ~ i ~ n. Hence it is convenient to encode the type
of an element and the cell corresponding to it by a binary
vector (a 1 , • . . , an), in which ai = 1 if the i-th property
takes place for a given type of elements, and ai = 0
otherwise (i = 1, ... , n).
Example 3. Let X, Y and Z be the subsets of the set
u = (al, ... , aN), satisfying the conditions X s (Y n
~ U U,X, Z s (X n Y) UX, Y s (X n Z) UZ,
A = U......_A, A E {X, Y, Z}. Find the number of such
triads.
Solution. For each a E U, three properties are defined:
a. EX, a E Y, a E Z. Each element belongs to one of the
eight types depending on its affiliation to the sets X, Y,
and Z. Apparently, the inclusion A s; B is equivalent to
A nB = 0 Hence the conditiOn X s;; ( y nZ) u y lS
equ1valent to X n (Y Z) nY == X n ({Y U Z) n Y)~ n
X n(Z n }} -= C6 The rema1ntng two condJttons are
equtvalent to the equallttes z n (Y n X) = 0 and Y n
(X f\ Z} =::: :Q1 The~ candtllons are -preseat~d \.n tM
d1agram below as empty squares Hence X Y and Z
sat1sfy the cond1t1ons of the problem 1f and only 1f the
elements of the set U ==- {a 1 av} do not eootaln
the elements of tllree l} pe' VIZ XYZ xYz and XYZ
An1 elerne.ot 1n U can belong to any of the rema1n1ng
five types Hence the number of the requtred trJads as
equal to ~n
y -y y y
X kl 0
----
X {6
z z -
z z
..........
8 2 1* (f) Pro\e formula (2) by tnductton

,.
(2) Let JV m be the number of obJects possess1ng exactl-y
m. propertu~s from among n Prove that
n m
Nm.= ~ {-1)1!. (m~k) Sm+A (3)
1--ft
(3) Let N nt be the number of obJects pos:;ess1ng not less

than m properttes from among n Prove that
"' m
Nnl= ~ (~J)~(m~~tk)sm+ls. (4)
h-f}
(4) Pro"e that
n
s.. =~(~)N,., (5)
t.'J'l~l
'q
S,. = ~ ( ~=:
11\~J\.
}N m (6)
8.2. INCLUSION AND EXCLUSION FORMULAS 263
•
(5) Prove that ~
Sm-(m+ 1) Sm+t ~Nm~Snl' (7)

v
Sm-mSm+I~Nm~Sm. (8)
8.2.2. Four persons deposit their hats in the cloak-
room. Assuming that the hats are returned at random,
find the probability that exactly k persons will get their
hats baclL Consider all the values of k (0 ~ k ~ 4).
8.2.3. Let E (r, n, m) be the number of ways in which r
different objects can be arranged in n boxes of which m
are empty. Let F (r, n,-m) be the number of arrangements
for which at least m boxes are empty. Show that
II
(1) E(r, n, 0)= ~ (-1)h{ ~) (n-ky;

h=O
n-m
(2) E(r,n,m)=(;,:) ~ (-l)h(n-;;m)(n -m-ky;·
h=O
(3) F (r, n, m)
n-m
= {~~~) ~ (-1)h (n-;;m) (n-m-kY m~k •

h=O
8.2.4. A survey into the reading habits of students
revealed that 60% read magazine A, 50% read magazine B,
50% read magazine C, 30% read magazines A and B,
20% read magazines B and C, 40% read magazines A
and C, while 10% read all three magazines. What percen-
tage of students
(1) do not read any magazine,
(2) read exactly two magazines,
(3) read at least two magazines?
8.2.5. A university department has thirteen staff mem-
bers, each of whom knows at least one foreign language.
Ten know English, seven German, and six French.
Five know both English and German, four know English
and French, and three know French and German. •
(1) How many know all three languages?
(2) How many know exactly two languages?
(3) How many staff members know only English?
8.2.6. (1) Show that the number of positive integers
that are divisible by n and that are not larger than x is
equal to [x/n],
264. CH 8 ELElrENTS OF COMElN.A'fORlAL .ANALYSIS
(2) F1nd the number of pnsttlve 1ntegers not exeeed1ng

1000 that are not dtvtstble by 3~ 5, and 7
(3} F1nd tl1e number of pos1l1ve Integers not exceed1ng
1000 that are not diVISible by 6 1 10, and 15
(4) Show that 1f n == 30 m, the number of posttlve 1n
tegers not exceed1ng n. and 1nd1Vtsible by 6, 10~ and 15
IS equal to 22m
(5) Let p 11 , p,. be all pr1me numbers not exceed1ng
)fli Show that the number of pt1me numbers p such

r
that ]Iii< p~ n, 1s equal ton- 1 + 2]
~~~~
(-- t)~SA,
where the sum
P:i,.
p,.
r
IS taken over all poss1bJe k tuples of pow.ers a 1 ~ , o:,
of Vthlch exactly k powers ate equal to i, whlle the re ..
matntng po\\ers are equal to rero
(6) Find then umber of prtm a numbers not e.xceedtng 100
8.2~7~ Let V be a set of n (n~ 3) elements
(t) F1nd the number of pa1rs (X, Y) of subsets oi the
set U sat1sfyrng the condttion X n
Y === 0
(2) F1nd the number of pa1rs (X, Y)t such that X~ U,
Y ~ U, \ (X,Y) U (Y".X} i ~ i
(3) Ftnd the number of tr1ads (X, Y, Z) for wh1ch
X s;; U ~ Y £; U, Z £ U, X U (Y f1 Z) X UY =
(4) Find the number of patrs (X, Y} of subsets of the
set U for v.l11cb X n Y ~ 0 f I X r ~ 2, t Y I~ 3
(5) F1nd the nJ.Imher of pa1rs (X, Y) for wh1ch X ~ U,
Y G V, I (X,Y) U (Y,X) J == ft J X (# 2
1y 1> 2
{6) Flnd the J)lllD her of tr1ads (X 1 Y, Z). snch that
X ~ U Y s U, Z s;; U, X U (Y 0 Z) = X UYf
( X J ~ 1 , ~ lF ' ~ t, I Z I ~ 1
8~2.841 Butler s problem n pa1rs of warrtng kn1ghts
(n~ 2) are 1nvtted to dtnner ShQw that there are
n
2 (- f}rt ( : ) 2" (2n - k-1}! ways of .seatwg them by
1~o
a round ta.hle, so that no two enemtes s1L s1de by s1de

8.3. RECURRBNT SEQUENCES 265
8.2.9. Married couples problem. In how many different

ways can n married couples be seated around a round
table so that men and women occupy alternate positions,
and none of the couples is seated side by side?
8.3. Recurrent Sequences, Generating Functions,

and Recurrence Relations
The sequencea 0 , a 1 • • • • , a 11 , • • • is called recurrent if,
for a certain k and all n, a relation of the following type
is satisfied:
an+h +p 1an+h-I + . · . + Phan = 0, (9)
where the coefficients p;, i = 1, k are independent of n.
The polynomial
p (x) = xk +
Pixk-1 + ...
+ p" (10)
is a characteristic polynomial for recurrent sequences {a 11 }.
8.3.1. (1) Prove that a recurrent sequence can be com-
pletely defined by specifying its first k terms and the
relation (9).
(2) Let A. be the root of a characteristic polynomial.
Prove that the sequence p.n} satisfies the relation (9).
(3) Prove that if 1..1 , • • • , A.h are prime (not multiple)
roots of the characteristic polynomial (10), the general
solution of the recurrence relation (9) has the form
an = Ct'A.Y + ... + C~~A.~.
(4) Let 1.; be a root of multiplicity r, (i = 1, s) of
I the characteristic polynomial (10). Prove that in this
l
case, the general solution of the recurrence relation (9)
has the form
s
an=~ (C;,+C;.n+···+C; nri-
1
)/..i',
t=l r;
I where C; 1 (i = 1, s, j = 1, r;) are arbitrary constants.

l 8.3.2. Find the general solutions of the following
recurrence relations:
' (1) an+2- 4an+l + 3a 11 = 0;
' (2) an+2 + 3an = O· •
\ (3)} an+•- - an +1 - a' 11 = O·'

(4 an+z + 2a +l + a = 0;
11 11
\
I
266 CH 8 ELE)1ENTS OF COUBIN'.ArD.RIAL AVALYSIS
(5) an+s + 10an+t + 32an+t + 32a:n. ====- 0

(6) an+3 + 3an+2 + 3an+l +a"' = 0
8 3 3 Ftrtd an fram the followtng recurrence relattons
and 1l11t1al condtttons
(1} an+2 ~ 4an+l + 3an -= 0
a 1 - 10 a 2 = 16
(2) «n+a-- 3an+2 + dn+l- 3an = 0
a1 === 3 a 2 === 7 a 3 = 27
(3) On+- 3 -- 3an+t + 2an = 0
a 1 :=:.= a a 2 -=- b a3 === c
(4) fln+ 2 - - 2 ros-a. 4n+t +an ~ 0
a 1 = cos ct =
a~, cos 2a
(5) a 11 +a -- a 11 :=: 0
a1 - 2 a2 ~ 0
(6} an+t - 6an+l + 9a., = 0
a, -= 6 a :1 = 27
8 3 4 Show that
(1) 1f x = 1 1s not a root of the polynomtal z 1 +
pz +
q the part1al sol-ution of the recurrence relatton
an+:: + pan+I + qan = an + ~ (11)
where a ~ p q are gtven numbers IS the sequence a~ ==
an + b ltnd a and b
(2) 1f .1: ~ i 15 a Slrople root o{ the polynOml.al x2 +
px + q the partta) solut1on can be obtained 1n the form
a: -= n (an + b) find the values of a and b
(3) tf .z = 1 1s a multtple root of the polynomtal .x1 +
px + q the part1al solut1on may be ohta1ned 1n the
form a:= n 1 (an + b) find a and h
(4) tn eaeh of the above cases find the general solutton
of tho relatJ on (11)
8 3 5 Solve the following recurrence relat1ons
(1} an + 1 - - iln -- n a 1 ==:: t
(2) an +2. + 2o11 +I - Ban 21 X sn
a1 -- --- 9 a 1 -- 45
(3) 4~ +2 - - 2an+ 1 + 2a 11 = 2n
11: 0 = i 41 = 2
(4) an.+ 2 + an+ 1 - 2a 11 - - n
a0 - 1 a 1 ..._ - 2
(5) an+2 - 4an+J + 4an = 2n
4 0 :==: t a1 = 2
(6) On+ 2 + an+t--.. 6an = 5 X 2~• 1
a.o = 2 al = - 1
8.3. RECURRENT SEQUENCES 267
9.3.6. (1) Let {an} and {bn} be two sequences whose

terms are connected through the relations
an+l = plan + qlbn,
bn+l = Pzan + q2bn,
!:l = P1q2 - Pzq1 =F 0,
where p1 , q11 p 2 , q2 are given numbers. Find the expres-
sions for an and bn, assuming a 1 and b 1 to be given.
(2) Find the solution of the following system of recur-
rence relations:
an+l = 3an + bn, bn+l = -an + bn,
a 1 = 14, bl = - 6.
(3) Find the general solution for the following system

of recmrence relations:
an+l =b+5, bn+l =-an +3.
8.3.7. Fibonacci's sequence {Fn} is defined by the
recurrence relation F n+2 = F n+l + F n and by the
initial conditions F 1 = F 2 = 1. Prove that
(1) F n+m = F n- 1F m + F nF m+l for any natural num-
bers n and m;
(2) the number F n is divisible by F m for any m and
n = km;
(3) two adjacent numbers F n and F n+t are mutually
•
pnme;
(4) any natural number N (N > 1) may be uniquely
presented as a sum of Fibonacci's numbers, such that
each number enters the sum not more than once and no
two adjacent numbers enter together;
(5) Fn= Js [( 1+Y5r-( 1-y5 rJ;
(6) F1 + F3 + · .. + Fzn+l = F2n+2;
(7) 1 + F 2 + F4 + ... + F2n = Fzn+t;
(8) F~+ t + F;: - FL t = F 3 n.
Each sequence a 0 , a 1 , ••• , an, . . . can be associated
with a series A (t) = a 0 + a1 t + ... + antn + ...
called a generating junction for the sequence {an}. If the
series A (t) converges to a function f (t), the latter is
also a generating function for {a 11 }. The exponential
generating function for {an} is the series E (t) = a 0 +
26S CH .f! EL~tE~TS (llr COJ.fBI"iATORlA~ ANALYStS
nl
+ D'ntn +
The operations of addl
t1on mult1pl1cat1on, and multlpltcatton by a constant
ran be defined ro~ generatlng Cunetlons by treat.tng them
as formal serJes Let A (t) and B (t} be the generattng
functions lor tl1e sequences {an} and {bn.} respectively
and let a and ~ he constants In th1s case,
aA {t) + PB (t) + ~b 0 + (<W + l!b1)t
~ a.a& 1
+ + (ctan + ~b.n)tn +
A (t) B (t) == a btl + (a b + a b )t
0 0 1 1 0
+ + (a bn + a bn- +
0 + artb )tn +
1 1 0
If E 6 (t) and Eb (t) are exponent1al geDerat1ng functton.s
for the sequences {an} and {bn} respect1velyt the oprra
ttons of addttton and mult1p11cat•on by a constant are
defined 1n the same way as for normal generating func-
tions, wh1le thetr product IS defined as
+
Ett (t} Eb (t) = c0 + c,t
where Cra = a 0b~~. + ( : } a btH + . 1 +

artba
8.3 B,. F1nd tl1e generat1ng function I (t) for the se·
quence {an} 1f
(1) an := t for all 0, n>
(2) an= -1 ior 0~ n~N ana Dn ~ 0 ior n>N,
(3} a 11 ~ a."',
(4) an == ~ntn•,
(5} a" ~ (- t)tt ..
(6) aa === nl
(7) an ~ n (n - 1),
(8) al'i = {~ ) , m 1S a natural n-umber,
(9) an = ( :) ~ a IS a real numbt!r,
(tO}an = n 1 ,
(11)an. = stn an~
(12) an. = cos an
8.3.9. F1nd th~ exponenttal generating funct1ons E (t)
for the sequence {an} 1f
(1) Un = i,
(2) an ~ (l.n t
8.3. RECURREN T SEQUENCE S 269
(3) an = n;
(4) an = n (n - 1);
(5)an = (m)n;
(6)an = n2 •
8.3.10. Using the identities connectin g generatin g
functions , derive the following identities for binomial
coefficients:
+ +11 111
(1) (1 t) (1 t)rn = {1 t) + ;
11
+
k
h (~) (k:s)-Cztm);
s=O
(2) (f-Wl-n {1-W1- m=(1-tt 2-n-m;

k
2} ci~S) (m+;-s)=(m+ntk+t)j
s 0
(3) (1 + tr (1 + ttm = (1 + t)n-m;

k
2} ( _ -l)k-s ( ~) ( m+;=:-1) _ (n-;;m) j
•=0
• •
2k
h (- 1)•(n;s) (n+!k-s)=(ntk);
s=O
(5) (1 + t)n (1- t2ttl = ('1- tt11 ;
[k/2]
2} (k~2s) (n+;-t)-(n+z-1);
s=O •
(6)
•
(1 + t)n (1- t) 11
= (1- t 2t; .
k is even,
s=O 0, k is odd.
~.3.11. Find the general term a 11 in the sequence for
whtch the function A (t) is a generatin g function.
{1) ,A (t) = (q pt)m; +
(2) A (t) = (1 - t)-l;
(3) A (t) = V 1-t;
{4) A (t} = tm (1 - t)m;
2.70 CJI 8 ELEhiZ::frr.TS OF CO\fiiJNATORIAL ANALYSIS
(5) A (t) := (I + t1 + + tr)m,

(6) A (t)= (1+·~~)-m,
(7) A (t) = (1 + 21rm ( t -f) -m;
(8) A (t) = t 1 (t - l) (t + 2t)-m1-
(9) A (t) := ln (t + t),
(10) A (t) =
arc tan t,
(1 i) A (t} ~ are s1n £..
(12) A (t) = e- 2tJ,
(13)
'
A (t) = ) e-z:l dz,
fJ
(14) A(t)=(t+'trJS
8.3~12. Derrve the follow1ng tdent1t1es2
{l) ~(- i)-•(;)(:t;) (n:.t)'
'
(2) ~(-1)'(~ }(2n~s)=(-Hn(;:),
...
(3) h (~ 1)' ( m~+•) { n~') ={m-;-t) •
'
(4) 2 ~ ( ~) ( 1m~21) ·-= (
.
t }+ (-i)"' ( ; }•
(5) 2~ (n:2.) r·+~+t2a+t)=(2nt~t2),

'
(6) ~ ( -1)"-• 4• ( n+~+i) = n+ 1
'
8,.3. 13. Let A (t) and E (t) he a generat.1ng and an ex 411
ponent1al generating funct1ons of the sequence {an} re-

spectively.
Cia
(1) Us1ng the eq_uahty nl = ~ e......-xn dz~ show that

0
'Qg
A(~}=~ e~E(3:t)dz. {•)

0
• Sum rna t1 on IS carr1 ed out over all 1 for wluch these ex pres . .
slon.s are mta.Jullgful
8.3. RECURRE:-;T SEQUENCES 2il
(2) Vedfy that formula (*) is valid for generating func-

tions A (t) = (1 - t)-1 and E (t) = e 1 of the sequence
a0 = a 1 = a2 = . . . = L
(3) The same, for a sequence with a general term
0, n < j,
(n)i, n~j.
8.3.14. (1) Let (a)n = a (a - 1) ... (a- n + 1).

With the help of exponential generating functions, prove
that
II
(a+ b)n = :6 ( ~ ) (a)n-db)lt"

h=O
Hint. Use the identity (1 t)a+b = (1 +

W (1 t)b. + +
(2) Let (a)n. h =a (a +h) ... (a- h (n- 1)).
Prove that
"
(a+ b)n, h = :6 ( ~ ) (a)n-k, h (b)k, 11·
h=O
8.3.15. Let {an} and {bn} be sequences, and let A (t)

and B (t) be their respective generating functions. Pt·ove
that if .
(1) an=bn-bn-t• then A(t)=B(t)(l-t) ;
1
(2) an= bn+t- bn, then A (t) = B (t) -; t - b0 t-t;
B (1)-B (t)
(3) an= bn+t + bn+ 2 + ... , then A (t) =
1- t
;
(4) an= nbn, then A (t) = t ~ B (t);

2
(5) an= n b11 , then A (t) = t :t (t :e B (t));
(6) we defme the operation S 11 (k ~ 0) on the sequence
{bn} with the help of the relation ·
+
Sh (bn) = bn ( k1 ) bn-t + ... + (k+ ij bn-j -1)
.
+ ... + (k+;:-1) bo
and we put «n.-=--=- S"- (bn) tl1e~1 A (t) ~ (t - t) h B (t)
(7) a11 = b2" then A {t) =-}
(B (t 1n.) + B {- til 2 ))
(8) an==b&+b 1 +bJl~ + a 0 =0 then A(t)--
8 (t) t (1 ~ t) J
8 3 t6 Let A {l) and B (t) be the generaltng functlons
of tlte sequences {an J and {br.} respectlvely and let
A (t) B (t) == 1 Ftnd {b n } and B (t) from the g1 ven se
que~~ce {aJL} lf
m
(t} an ~
n
~ (2) al) = a 1' (J) an = n + 1-.
(4) a0 = a 2-= 1 an ;:::=: 0 for n ~ 0 2
(5) lln=(-1)'~, (6) a~=(-i)" e:}4 tl
8 3 f7 Suppose that the seqoence {an) sattsfles the

recurrence relalton an+':! + pan+i + qan ~ 0
(l) Pro\e that A (t) = .~-±..<al+_P~~.t.
1+ pt-t-9f~
(2) Let 1 + pt + qt 2 ~ (t- A1t) (1- L~t) j.1 =F ~
Pro\e that
.1n-.ln A.n+•-A~+~
a"= {a,+ pa,) A'
1--
At
I
+ ao A1-- 'A.:a2
(3) \Vr1te an expresston for a 71 for tb.e case when
1 + pt + qf 1
= (1 - lt)'
8 3 f8 Let
n
an = ~ (n;j 1 } , n-=0 1 2,
J 0
and let A (t) and B (t) he th.e corre.spo!ldtng generatlng

functLons
(1) Prove that an and b n are counected tl1ruugh telat1ons
ol the type
a J1 +1 + bR +t
.::::::=: ll n.
bn+I ~ 4n + b,. flo ~ 1 b, =- 0
(2) Pro\ e that A (t) and B (t) .sa t1.sfy the system of equa
t1ons
A (t) - 1 == tA (t) B (t} +
B (t) ~ tA (t) + aB (t)
8.3. RECURRENT SEQUENCES 27::1
(3) Find A (t) and B (t).

(4) Show that
• (
hm 2 _ )n a 11 = 1 +liS
_ , . (
hm 2 - )n bn =
n-oo 3+115 21/5 n-oo 3+115
8.3.19. Suppose that the terms of the sequence
satisfy the relation
an=aoan-t+atan-2+ ···-i-an-tao, Go= 1.
00
(1) Provt> that the generating function A (t) = LJ

11=0
ant"
satisfies the equality tA 2 (t) =A (t)- a 0 or, if the initial
1-l11-4t
conditions are talwn into account, A (t) = 2
t •
(2) Expanding A (t) into a power series in t, show

that a 11 = 1 {2n)
n+ 1 n .
(3) Find the sequence {an} whose terms satisfy the re-
lations
a 0an _1 + a1 a 11 _ 2 + . . . + an _1a 0 = 211 a 11 ,
ao = al = 1.
8.3.20. Derive a recurrence relation for the sequence
{a 11 } and solve this relation if
(1) an is the number of ways in which a convex: (n 2)- +
gon can be divided into triangles by diagonals that do
not intersect within this polygon.
(2) a 11 is the number of ways in which parentheses can
he arranged in the expression b1 : b 2 : • • • : b 11 +l, so that
the resulting expressions are meaningful.
8.3.21. Using the method of mathematical induction,
find the sequence {an} from the following recurrence re-
lations and initial conditions:
(1) an+t = (n + 1) an, a 0 = 1;
(2) na 11 +l + a 11 = 0, a 1 = 1;
(3) (n + 2) (n +
1) au+ 2 - n 2a 11 , a 0 = 0, a 1 = 1;
(4) (1~ + 2) 2 an+z +an = 0, a 0 = 1, a 1 = 0;
(5) wan+z + (n + 2fl an = 0, a 1 = 1, a 2 = 0;
(6) a~+l - a nan +z = (-1) 11 -t, a 0 = 1, a 1 = 1.
00
8.3.22. Let An (t) = 2J a (n, k) tk be an arbitrary

k=O
function for a sequence satisfying the relation
a (n, k) =a (n, k- 1) +a (n - 1, k)
18-0636
214 Cll 8 ELEMENTS OP CO~tBtNA.1fORIAL AN.ALYStS
wttb tnittal cond1t1ons a (n, 0) === {, a (0, k) = 0 for

k > 0 Show the.t
(1) (1- t) An (t) =An-I (t),
{2) An {t) = (1- -tJn,
{3} n(n, h)= ( n+:-f }
8~3 23 (f) In bow many ways can a 10-kopeck co1n be
changed tnto 1., 2, 3 and 5-kopeek eo1ns?
(2) The same, hut eath eo1n must have a dupllta.te
(3) The same problem under the condltton that the
change rs to be made from four 1 kopeck co1ns. three
2 kopeck cotns,. two 3-kopeck GOins, and one 5 kopeck
COlD
8.3 24~ F1nd the generat1ng func\ton A (t) for the se-
quence {an}, where
(1) Da IS the number of solutions Jn non negative Inte-
gers for the equatton
2z + 3y + 5z = n
(2) an 1s the number of solut1ons of the ;same equat1on
g1ven that x, y, .z assume values In the set {01 1}
(3) a,. 1s the nltmber of tntegra..l solutl.Gns of the same
equation g1ven that 0 ~ z ~ p, 0 s:;; y ~ ,.t 0 ~ z ~ s
S 3 ~25~. F1nd the value of a" us1ng a given generattng
functton A (t) for the sequence {an}
Ol:t
(1) A (t) ~ II (1- qlt)t

A~G
Iq '< 1
H1nt. Show that A (t) = (1 - qt) A (qJ), and tompare
the coeffictents of tn on the left- and rtgbt hand stdes of
thJs equation
00
(2) A (t) =: II
l-1
(f + qt2hJ
Hant. Prove that a,. = gbn, where b11 1S the number of
un1t1es 1n the b1nary expan:s1on ol tho number n
8 3~26~ Let
S(n, k, l)= ~ (-1)n-v (:) (v+lt'

~0
Prove that
(1.) S {n + +
1. k 1) = S {n, k, 1 1) ... S (n, k, l),
(2) S (n. k. l + 1) = S (n, 1, l) - S (n + 1, kt l),
•
8.(1. POLYA'S THEORY 275
(3) S (n, k + 1, l) = (n + l) S (n, k, l) + nS (n- 1,

k, l); .
(4) S (n, k, l) = 0 for n > k;
(5) S (n, n, l) = n!;
(6) S (n, k, l) > 0 for n ~ k;
(7) S (n, k, l) is an increasing function of parameters
k and l for n ~ k;
n
'
(8) S (n, n+ 1, l) = Lj (k+ l) k!;
h=O
(9) S(1, k, 0)=1 for 1~k.
8.3.27. Let S (n, k) = S (n, k, 0) = ~ ( -1t-v ( ~) -vh
•
v
00
and an (t) = Lj S (n, k) th. Prove that for I t I< 1

h=O
n! t 11
(1) an (t) = {1-t) (1-2t) ... (1-nt);
11
(2) an (t) = t ~ ( -1)"-h k ( ~) (1-ktt 1•

h=i
8.4. Polya's Theory

The permutation on a set Z 11 = {1, 2, . . . , n} is the
mapping of Z 11 onto itself 3 • The permutation :rc =
1, 2, ... , n
• • • will be frequently specified by the
t1 • ~2 • • • • • ln
line (i 1 , i 2 , • • • , i 11 ). A permutation is called cyclic (or
a cycle) if a certain number h is substituted for j 2 , j 2
for j 3 , and so on, h-1 is substituted for h, and h for h,
while all the other numbers remain unchanged. Such a
cycle is denoted by (it, j 2 , • • • , h). The number k is
called the length of the cycle. Any permutation can be pre-
sented as a product of cycles. For example,
1, 2, 3, 4, 5 •
= (1, 2) (3, 4, 5). A cycle of length 2

2, 1, 4, 5, 3
is called a transposition. The permutations on the set Z 11
form a group relative to the multiplication operation.
The multiplication operation of the permutations :rt 1
and :rt 2 involves their successive application. For example,
3
Any set of n elt!ments can he taken as Z 11 •
18*
276 en B r:LEhfENTS OF COMBINATORIAL ANALYSIS
1 2~ 3~ 4 1t 2' 3,
rf n 1 = , n1 = then
2t 4,
3, 1 1' 4, 3, 2
t t 2 3, 4
"-1tlt =-
4 2 3 1
, It can be eas,l:y venf1t:od \hat
the m ul tlpltca tton opera lt on 1s assoc1a t1 ve :n (at) ;: : : :
(na) t The unLt element of a groLlp ts tlte tdenttty pef""
1 2., ,n
mutation The permutation Inverse to
l,. 2t n
2, ,n ltt 1 2,
ts grven by
ltt l' ln 1' 2' n
The pet1nUta\}on group on the set Zn lS called a symn1etry
group of n th degree and IS denoted by S n. The order of
a symmetry group of n th degree (the number of 1ts
elements) 1s equal to n'
If the permutation on a .set N 11 JS represented as a prod
uct of b1 cycles of length 1, b2 cycles of Jength 2, .and
so on, and b n C) cles of length n the permutatton IS satd
to be of the type (b 1 ~ b1 .. bn} For example the per
1, 2, 3, 4
mu.tat1on 1s o1 the t;ype (1 o. 1, 0}
3' 2t 4 1
)f G ts a subgroup ol the group Snt the polynomial
PG=PG(tl t,.)~IGJ- 1 ~1~ 1 t:n,
1t~G
where (b 1 ~ , bn) 1s the t:ype of permulatton nt lS

called the cgcllc ~ndex of the group G Let G be a group of
permutations on Z., Th.e elements a and b 1n Zn are called
G-equtLalent (notat\un a ,..., b), 1{ there ex1sts a permutation
n E G, such that .na -= b (or, whrch ts the same, nb =- a)
The classes of G-equ1 valence are called tran.sttlve sets or
orb1t.s
Bernside . s Lemma. The number of orbtt.s v (G) fn the set
zn t defined by the group G' lS glven by the equality
v (G)==--- f G r-• nEG
:2j bt (n)
Let Af and N be nntte sets, and let G and H he the pet-
mutat1on grcups oi M and N The pnv..er group HG
consists of all posstble pa1rs (:n: 1 a), where n EG, a E H,
and acts on the set NAt of all funct1ons j AI -• N ~lore
8.4.. POLYA'S THEORY 271
over, by definition, (n; cr) f (x) = crf (n (x)) for all .c EM

and f E NM. Suppose that a weight function w: N )o
{0, 1, ... } is defined on the set N, and that q rt is the
number of elements having a weight n inN. The generat-
00
ing function Q (t) = 2} q 11t 11 is called a figure counting

11=0
series. The weight of the function f in NM is determined
from the equality w (t) = 2} w (f (x)). The functions f 1
xEl\I
and f 2 in Nlll are called equivalent (notation f 1 ,..., f 2 )
if there exists an element n E G, such that f 1 (nx) = f 2 (x)
for all x E j\-J. If f 1 and f1. are equivalent, they have the
same weight. Hence we can determine the weight w (F)
of the equivalence class F as the weight of any element
f in F. Let <ph be the number of equivalent classes of
00
weight k in NM. The generating function <D (t) = 2} <phth

h=O
is called a function counting series.
Polya 's Theorem.
<D (t) = Pa (Q (t), Q (tZ), ... , Q (t11)),
where n = I G 1. Pa (t 1 , t 2 , • • • , tn) is the cyclic index
of group ·c, and Q (th) is substituted into Pa in the place
of variable th, k = 1, 2, ... , n.
8.4.1. Find the type of permutation n:
(1} n = (2, 3, 4, 1); (3) n = (3, 4, 5, 6, 1, 2);
(2) r£ = (4, 2, 3, 1); (4) r£ = (8, 2, 1, 7, 4, 6, 3, 5).
8.4.2. Represent the permutations in Problem 8.4.1
in the form of a product of transpositions.
8.4.3. Find the cyclic index of the group G, where
(1) G is a group of permutations of the vertices of a
square, which are obtained by rotating the square in
a plane; '
square, which are obtained by rotating the square in
space;
tetrahedron, which are obtained by rotating it;
(4) G is a group of permutations of the edges of a tetra-
hedron, which are obtained by rotating it;
(5) G is a group of permutations of the faces of a tetra-
hedron, which are obtained by rotating it;
27& GK JJ ELEllE.NTS OP COl-IBIN.ATORIAL ANALYSIS
(6) G = Sa,
(7) G ~ A._~ where A, ts an alternating group of po~er
4., I e a Subgroup of group 84. 1 formed by permutatlons
r~presentab]e as products ol an even number of transpos1
ltons.,
(8) G 1s a group of permutatlons of the faces of a cube,
wh1ch are obta1ned as a result of rotatton, and
(9) G IS a group of permutations of the verttces an or
octahedron
8.4 4~ (l) Pro'\Te that the group G defined 1n Prob
lem 8 4 3(1) specifies one orb1t on the set N 4
(2) Ftnd the number of orbtts 1n the set N 4 , wh1ch are
defined by the group G and formed by the perm.utat1ons
tr 1 = (f 2, 3, 4), .n2 ~ (t t 2t 4t 3), n, = (2, 1 3, 4) t
and 1t 4 === (2 t t , 4 , 3)
8 .. 4. 5. Prove
_,..
Berns1de,s lemma
8 . 4 . . 6 . . Let b = (h1't b 2 ~ , b~} be a vector corre-
spon d1ng to the de-com pos1 t1on of the number n~ 1 e
b1 + 2b 1 + +
nb" = n, where b-' are
......
non negatt\e
1nteger:s (t ~ k ~ n} \Ve denote by H (b) the se:t of a.1l
permutations of the symmetry group s n whose type
...... ..... ....;.
COIDCides Wtth b, and let h (b) = r H (b) f Prl1Ve that
..... 11
(t) k {o)~nt ( l' 6
k AbAt}-J,
~~·
n
(2) Psn (th 12• , ln) = ~l 2} h {b) IT t!•.

,
--fl' A:~l
(3) the cy.cl1c Jndex P8 ,. rs equal to the coefficient of

z,. 1n the expans1on of the funtt1on
Into a power sertes tn x

S.li.7 ~ Let An be an alternatlng groop of power nt • e
a .subgroup of the group S n, consiSting of all 1 ts perm u
tattons, wh1ch can be represented as a product of an even
numher of transpos1t1nns Pro-ve that
' ttl) =;:;:=:; P s,. ( t1, • tn)
1
+ Psll (f 1t __, tz, .... , ( -1)n tn)

~
8.&. POLYA'S THEORY 279
8.4.8. Let G be a group of permutations of the set X,

and H a group of permutations of the set Y, X n Y = 0.
Any pair of permutations n E G, a E H is put in corre-
spondence with the permutation n X a of the set X U Y,
defined as z ~ nz for z EX, z ~ crz for z E Y. Prove that
(1) the permutations n X cr form a group of the order
l G I I H 1. This group is called· the direct product of the
groups G and H, and is denoted by G X H;
(2) if the permutations :rt E G and cr E H, are of the
type (b 1 , • • • , bn) and (c 1 , • • • , cn), then :rt X cr is of
the type (b 1 + c1, ••. , bn + Cn)i
(3) PaxH = Pa X PH.
8.4.9. (1) Find the number of necklaces that can be
made from beads of two colours, if each necklace contains
seven beads. If one necklace can be obtained from the
other by turning it, the two are assumed to be identical
(mirror reflections are not allowed).
(2) Using Polya's theorem, find the number of necklaces
from beads of k colours, if each necklace consists of n
beads. The necklaces are considered to be identical if
they can be obtained from one another by rotation (with-
out reflection).
8.4.10. (1) Find the number of different colourings of
a tetrahedron vertices in two colours. Two colourings are
assumed to be different if the colours of the vertices cannot
be made to coincide through a rotation of the tetrahed-
ron.
(2) Find the number of different colourings of the ver-
tices of a tetrahedron using three colours.
(3) Find the number of different colourings of the
faces of a cube, such that three faces are painted in red,
two in blue, and one in white.
8.4.11. Let G be a group of permutations of the set
Zn, and E be a unit group acting on the set N and trans-
forming each element x EN into itsPlf. Find the number of
orbits defined by the power group EG on the set Nzn.
8.4.12. (1) Let Th be the number of pairwise non-iso-
oo
morphic rooted trees, and let T (x) = 2J Thxh be the
h=l
generating function for the sequence {T h }. Prove that
(X)
T (x) = x + x 2J P8 n (T (x), ... , T (x11)).

11=1
280 CH ~ ELE1.1EN1S Olf CO)r1BlNATOR.L\L ANALYSIS
(2) Prove the valldtty of the followtng teeurrence rela.

tton
n
T,.~ 1 = ~ ~ ~ rT,Tn-Jt.+h T1 = 1.
h~l ,. I A
B 4~13. Let gn be the nl1mber of palrwtse non-tsomor..

phtc graphs With n vert1ces, and let l,.. be the number of
pa1rwrse non-Isomorphic connected graphs w1lh n ver
ttces Let
(Q
g (t) ~ )J gnt 71 ,
:n-J
be the correspondtng generat1ng funct1ons Prove that
~
g (t) := 2J
n~t
G!J
n
(l (t),
8 4. f 4 Let p (n, h) he the n unther of perm uta t1ons of

the group S n, whtch cons1st of h eyc]es, and let p" (t) ~
~
~ p (nl k) t 1
be the correspond lng gei1erat1ng functton
-n~o
Prove that
~-1.
<tl Pb<tl~ II <t+ ')~

1~0
(2) p(nt k)==p(n-1, k-1)+(n-l)p(n~ J k)
8.5. Asymptotic Expressions and Inequalities

The follow1ng notation w1ll be used while esttmatJng
the growth of f11ncttons The notation lp (z) ~ 0 ('P (x))
for .x E X means that there e.xJ.Sts a constant C such that
w
' fP (.t) 1~ C l (x) ~ for x E X If cp (x) == 0 (tl> (x)) and
""' (x) ~ 0 {~ (x)) for .t E X, we can wr1te q> (.x) ::< 1}> (x)
for x E X Tl1e notation rp (.:r) == 0 (~ (.z)) for x ,...'"a
means that llm _(P (z)}_ ==. 0 The fun~l1ons (.r
'Ill (z) are satd to be asymptotlcally equal (notatton cp (z) ;...~
~ (2)) for x - a! tf lfJ (:t) ~ 1:- (x) + 0 (~ (z)} for x • t1
)r
For different types of esttmates 1t 1s con'Ven1e.nt to use

StJrl1ng s formula
nt ,._., V 2nn.n. e-"11
(12)
8.5. ASYliiPTOTIC E..'(PRESSIONS AND INEQUALITIES 281
•
For more accurate estimates, the following inequalities

are used:
V2nnn 11 exp ( -n+ 1 ~ 11 - 36~ 113 ) <nl
<V2nnn exp( -n+ 1 ~ 11 ).
11
(13)
8.5.1. Prove the following inequalities: ~
(1) n 12<n!<(
11 11 1
t r
for n>2;
(2) (2n)!<[n(n+1)]";
(3) ( 1 + ~ r < 3;
(4) ( i r <n!;
(5) (n!)!l < ( n(n2+1) )n' n> 1;
n(n+l)
xn ~( : )
11 211 1 2
(6) 1x22 x3 3 x ... ;
( ~/} (211-1)!!
? )II < 1 ., (10) (1 +ct)n ~ 1 +ctn, ct~- 1·,
( -n ·· 11 3n+ 1 11
(8) (2n-1)!!<n", n>1; (n+21n+1) +t>(n+iln)".
(9) n! > e-11 n 11 •
8.5.2. Prove the following inequalities:
~ +~ r < ( ~ ) < ( r ' n > k > 1;
3
( 1) ( 2 ;
(2) {-i-)k.<(~)<kk(n:lk)n-R.• n>k>O;

(3) 4 11
(?n)
,,- < - ~ ,/ 4
, n > 1.
11
2vn n v 3n+1
1
8.5.3. Using Stirling's formula, show that for n ~ oo,

the following asymptotic equalitie!l are valid:
(1) (2n-1)!1"""' V2 (2nt e-";
(2) ( 21t) """' 1 411;
n 11 :rm
3 n! 3~ ~.
11 '
(4} (m+1) (m+2) ••• (m+n)"""" kl n"'-k

(k+1) (k+2) ... (k+m) m!
for
integral
---~
" The symbol (2n - 1)!! is used to denote the number 1 X
3 X 5 X • . . X (2n - 1), while (2n) I! = 2 X 4 X 6 X . •• X
(2n).
282 CU 1 ELEMENTS OP COMB1NA10RIAL .ANALYSIS
non-negatrve k and rn,

(2rt)\ t ... / :rt
(SJ {Zn-t)H- V 2 n
8 5 4 Pr(}ve that for n , )r oo. the followlng asymp-
tattc eq,ualttleS are valid
n
( 1) h k! i ( ~} - ~+~. t
k~l
(2) ~ {r+\v) --}2n~ O~r<k,

v
(3) ~
" ( r+kv
n ) QT+h ,_..!k {1+a.}n ' O~r<k, a.>Ot
n
(4) ,2l (-!)k-l ( : } -Inn,
k~l
n
(5) ~ k {: )
2
-- i y' -:~ 4"
h-1
8 5 5 Let b0 , b1 , bn be such numbers that 0 <
el ~ b"~~- ~ ~ < 1 Are the foll ow1ng 1nt!quall t1es val1d 1
n
(1) (1 +a)"&; ~ ( ~ ) bt~(1 + c)n~
k~o
n
{2) (f-ct~ h (-1),.{~)h.. ~(1-aY~

It-O
8.5.6 (1) Prove the Chebyshev 1nequaltty Jn the
follow1ng form Let A= {a 1• a2 • ,. an} be a set of
n n.
numbers, and ii"=~ ~ ah Da=f ~ (4~-€t)2 Then the
~l \-l
-
fraction l5, of those a( for which f a1 -- a 1~t, does not
Da
excee d . t,_.
(2) Ustng CbebJ.shev s tnequallty. prove that
~ (; )+ ~ (; )~ (:e;*
o~~i -tVii i . . ,~.;~·~n
8.5. ASYMPTOTIC EXPRESSIONS AND INEQUALITIES 283
(3) Prove that
k=l
8.5.7. Using Stirling's formula, show that
(1) if k ) oo and k- = 0 (n 31~) for n -. oo, then i
II ) """' 211+1 e-(2k-n)'/2n;
(
k V2itii
(2*) if a>O, k -. oo and k- a~t = O(n213) for n -. oo,
n+ 1 (lt(a+i)-an)'
then ( ~) 1
k
ah = (i+a)
V2nna
e- 2an (1 +0 f..!..+
\ n
(k (a+i)-an)3)).
n2 '
(3*) if a>O, k<m, k -..oo and k, m=a~ +0(n21 3 )

1
as n -. oo, then
m -
2} ( n ) av /"Oo' (a+ 1)~ na
v=k v V2n
((a+1)m-an)' ((a+-i)k-an)'
e
- e
- na
X ...;__--:--:-:~--
m (a+i)-na
na
- --:---:-...,.-,.,..---
k (a+1)-na
,
•
lilt '
(4*) if a>O, k ... oo and k- a+i =O(nZ/3) as n ... oo,
t'
then ~ (
11
) av"""' .:_ e- z (1 +a)".
v' V2n t
v> na +t V""iiii
a+1 a+1
8.5.8. Let 0 < '}, < 1, A11 be an integer, f.1 = 1- 'J., and
A.- A.n - J.Lll
let G (n, 'J.) = V 1.1. • Using formulas (12) ·and (13),
2n:A.!ln
show that
(1) ( ;,t),..,. G(n, 'J.) as n -. oo;
(2) G (n, 'J.) exp ( - 1; 11 ( i+i)) < ( ;,t )< G (n, 'J.);
(3) ~i G (n, 'J.)::;;;:; ( ; 11 ) ;
for
CH 8 ELE\fENTS OF COMBJN.ATO.R.I.AL .ANALYSIS
n
(5) 2} (~)<A -l.nl,-pn for ;\>{, n~3
-'-~"
8.5~9. Let k and n be natural numbers (/,. < n) Show
that
00 h--1
{1) (n), ~ n-. exp - ~ iv ~ i'~ •
-vn.
v~t .,-1
(2) 1f for n- ~ oo k ~ 0 ( J/n), tben (n)~c ~ nJe,
{3) 1f k~ o(n) as n ~ oo~ then for any m> f
(4) for n _.., oo and k 1:::; .o (n3/4)

(n)1 = nJJ. exp ( "t k'
--211- - bn:a + o (I) ) '
(5) 1f m. t- t,n.; n; m ..,.. Pt n.
; m
_,. 1/ and h X (k-
rp): x Cot m+ n-
)II t-r 00 1 then
{:} (T:k }/ (m~n) =h~(x)t

where h = ((n + m) pqt (1 t))-11 2 , fiJ (x) === (2n)-J/2 e-~•;2
8 . 5 .. 10~ (f) Let k = k (n) and s == s (n} be such that
for ll"t-00 s=o(Vk) Showthat {~::)/(~}-{!r
(2) Shaw that tf &= o (kt~t ) then t
n--- ')
( k--1
( n }
k ~
}~
( }c
n ~p -
~ ,~+ ( f
1
kJ v (v+ 1) ·kv -
1 )
n.V
'¥~1
8~5-li. Let s==s(n) and k=-k(n) be non negat1ve

Integral funct1ons of a natural argument Show th.at
{t) 1l s+J.==o(n) for n-).oo, then
8.5. ASYllfPTOTIC EXPRESSIONS AND INEQUALITIES 285
(2) if s2+k2=o(n) for n oo, then (n7/)/ ( ~),..., 1;

~
sk. ' 1 k sn ) )
(3) exp ( - n { + n-k + 2 (n- k} (n-k-s) <
(n~s) sk
-=--..:.:.-..:...<ex - n > k +s.
d
8.5.12. Let f (.z·) be a. continuous, monotonically in-
creasing function on the segment [n, m). Show that
111 m
f(n)~ ~ j(k)- ~ f(x)dx~f(m).

h=n n
8.5.13. Using the previous problem, show that for

m ~ oo, the following relations are valid:
m
(1) ~ lnk......., m ln m-m+O (ln m);
1<=1
m
(2) Y,
~
11
k = 1
n+1
m'~+t + 0 (m 11) ' n > 1;
11=1
m
(3) ~kink x\nln k = ln lnln m+c+O {m lnm\nln m)'

h=2
c is a constant;
m
'V lo" k 1 ( log m) .
(4 ) L.J k 2
=:rlog m+c+0 m ,ctsaconstant;
h=t
m
~ kl!2k=c+ 1 a m+O(nd~em),
1
(5) cis a constant;
11=1
m
~ k~. =v!_i ( n \!-1- m v-1

1
(B) )+0( n\.- m\ ) ·
h -II
8.~.14. The sequence {Pn} is defined by the recurrence

relahon Pn = p 11 _ 1 -apL 1' p0 = 1, 0 <a< 1, ~> 1.
Show that
(1) 0< Pn < 1, n;;;;d;
(2) Pn decreases monotonically with increasing n;
1
(3*) Pn....,(CG(~-1)n)1-~, n ; oo.
286 Clf 8 .LLE)l.ENTS OF C0'1V1NATORIA.L .ANALYSJS
Hmt .. Use the 1nequal1t1es

n Pn n+l
n= ~ Pk-1'11-'J < f cU < ~ 1,Jt7,"P11-~. =n+i.
h~ I
--.ap~I, - 4
..
'
t
-a~ .It~ I
~ap6h
8 5~15. (f) Show that the solutJon of the equation
xe$ ~ t has the form
--Jo. t--+ o((lnlnt)2)

x = ] tl t- In In t + lnln' In I - •
t ...... 00
(2) Prove that tlte solutton of the equatton ex + In z =
t for t _., oo has the form
z= In t- o.;nl + 0 ( (!n !~) ).
1 2
8,5.16.. Let 1(t) > 0 and et 1t1l =: j (t) t +0 {t) for

lnt
+
0 < t < co show that t (t) = t - + 0 (t-2.) for t 00 a .,...
8 ~5. t 7 Suppose that the generattng funct1on A (t)

ol the sequence {a Ill has the: form. A (t) = ~ where g: .,
Q (t) and P (t) are polynomlals w1tb real coeifit:.1ents
Let ~ be the smallest (tn magn1tud-e) root 0:f the poly-
nomlal P (t} Prove that
(1) tf 1 1 IS a s1mple (not mult1ple) real root, then for
n -) oo
a 11 - -
1
--t)(~ }f 11 11
-
1
, w lu~re P' (A.J) --;;: .:, · P (t) It-At'
(2) If A1 1s a real root of multtplu~tty r') th.en for n -,. oo
-~ ( -t)r ,r 1.J (J. 1} { r+ n.-- t) ( 1,1~ )n+r
a" Ptr) (.11.) r- { t
where pr (A1) 1s a derrvat1ve of P {t) of the order r at

the po1nt t -= h1
8~5 .. 18.Let A (t) be the generat1ng functtan of the
sequence {an.} Find the asymptot1e behavtour of aR-
for n · ~ oo •
(t) A{O=at•:_~:+t;
t
(2 } .A (t)= 6a'+SJ~ 6'
1Zt1 +1<k1
(J) A (t) = 6t'+5t- 6;
8.5. ASYMPTOTIC EXPRES SIONS .AN}) INEQUA LITIES 28i
t
(4) A (t) = 4t2+ 1;
2t3
(5) A (t) = 6t~-11ta+ 35t2 - 22t 4; +
1-3t
(B) A (t} = +
(8t3 -1) (t 2 1);
1
(?) A (t) = (t2+ 2t- 2)2;
t3-1
(S} A(t)= (2t 2 +1) (t 2 -J-i.4t+ 0.49) ·
8.5.19 . Find the asymp totic behav iour of an for n ) oo
from the follow ing recurr ence relatio ns and the corre-
spondi ng initial condit ions:
(1) an+ 2+ 3an+t +
2an = 0, a 0 = 1, a1 = 2;
(2) an = qan-t + P (1 - an-t)r a 0 = 1, P q = 1, +
p, q > 0;
(3) an+ 2 + 2an+t + 4an = 0, a 0 = 0, a 1 = 2;
(4) an+J - 9an+ 2 +
26an+t - 24an = 0, a 0 = a1 =
1, a2 = -3;
(5) an+ 4 - 4an+ 2 + 4an = 2", a 0 = 1, a~= 0,
a 2 = 2, a 3 = 0.
8.5.20 *. Find the limit of the sequen ce {an}, given by
the recurre nce relatio ns
(1) an+t = (an +
bla 11 )12, b > 0, a 0 > 0.
(2) an+l = (2a 11 + blaW3 , b > 0, a 0 > 0.
(3) an+t = (b - a;i), 0 < b < 1, a 0 = b/2.
8.5.21 . Suppo se that an satisfi es the relatio ns
an~2-n- 1 + (n -1) 4-n- 1
, (14)
an+2 ~ 2-n-3 + (n + 1) 4-n-3
+an ( ( { r+ 2
+ (n + 2) 2 +
11 2
+4an )' (15)
a1 = 0, a2 = 1/16.
Show that
(1) an~ 1/8;
(2) an ~ 9 (3/4)n;
(3} an = 2-n-1 (1 + 0 ((3/4)n}).
8.5.22 . Suppo se that the sequen ce {an} satisfi es the
Condition an+m ~ an + am, a 1 > 0. Prove that an <
a1n for n ~ 1.
8.5.23 . Let k and n be intege rs. Calcul ate k = k (n)
to the neares t intege r, for which the functi on f (n, k)
fll 8 F'Lr.r--IENT~ OF COMBIN'ATORJAt ANALYSIS
assumes the maximum value

(1) f(n, k) = ( ~) 2-z",
{2) I {n~ k) = l :) 2'1lo-R-'l"' (1- 2-'l.!t)n-11
8 .. 5 .. 24. F1nd the mtnJmum and maxtmnm values of the
expression
f (n, r, k}= ( : } {~:;) 2-r+2~
as a function of r (0 <
r ~ k ~ n, r._ k., n are tntegers)
8 .. 5~25~ F1nd the asymptotic behaviour of the quanttty
g (n) = m1n f (n, k) (k IS an Jnteger) for n -.., oo, tf
0~1~n
(1) /{n, k) = 2n h + 2~'\

(2) I (n, k) = h2k +-} z:~n--'
Solutions, Answers, and Hints
CHAPTER ONE
1.1.
1.1.1. {1) (: ; (2) 2n. 1.1.2. (1) n, 13, 50; (2) (010011).
n-1 It 2n-1.
1.1.3. • 1.1.5. (1) n2n-1; (2)
k-1 k
m n-m k m
{2)1 k+m-r k-m+r •
I (3) 1 ~ i+m-r X
2 2 i=O 2
n-m
J+r-m •
2
~
1.1.9. Hint. Prove that the number 1jl (11) of vectors a E B~11
m
such that ~ ai ~m/2 for all m=i, 2n satisfies the recurrence
1=1
n
relation 1jl (n) = ~ 1jl (i-1) 1jl (n- t), 1jl (0} = 1jl (1) = 1. Using this
i=i
relation, prove that 1jl (6) = 132.
n-r (k-1)
1.1.10. •
k
1.1.11. (1} Hint. Consider the stratum B[n 121 . (2} Hint. Among
n + 2 tuples in Bn, there always exist two tuples of the same
weight .
•
n-l k
1.1.12. (1) k-l ; (2) ; (3) 2k+211 -k-1.
l
1.1.13. Assuming that 0 ~ l < k ~ n, we calculate the number
f"'J ...... ....., i""J
P (n, k, l) of pairs (a, ~) such that a E A, ~ E B. On the one hand,

n-l
P (k, l, n) = I A I , while on the other hand, p (n, k, l) ~
k-l
1 n
For a non-integral a, is assumed here to be equal tn 0.
a
'
19-0636
SOLUTIO~S ANSWERS AND Rll\lTS
\B\ k Usmg the ,denUt}

n
l
n-l)
J.. ~ = "
1
.l k)
1
'I.e (lb\a. n
1
the r\! qu l r~d 'neq ua! Ll} Fat' l ;;a J: the ~~ult -v. Lll he the ~mt
j I 14 (I) An ascend 1ng cha 1n of Iength n .con ta tn s one "erte:r
Jn each slratum lD a stratum B~ :. vertex can be chosen Jn n dlffer
eut W3}S After a \ettex or
the chain has been chos~n 10 Bn
the v erttl. of the c ha 1n 1o the 13econ d stratum caD he c hoseD JD
n - t ways and won
I I 15 (I) Let A be a set of pa1rw1se 1ncompara ble tuples ln
n ~
Bn A~ A n B 1 and z ( a) be the set of ascend1ng cha1ns oJ
fv
Jengtb n that con ta 1ns the vert e.J a. \ Ve have
>[m;;-]'X];[ X:E \A 1
- lA\ Y(-~ )t X }~-(t
1~0
11:
Hence lA J :r; (2) H1Dl for '~ k ~ nJ2 tbe 1nequal1ty
(n/21
,, (n-- J) r ;a: k' (11- v.al1d
k) I IS
1 t 15 lltnt \Ve assoc1ate ea.c.h number ol the lorm p« p~ 1

p~ 12 \\lth a \ector (a1 a-1 a.,J 1n. on Then the set A
corresponds to the set A s JJn .cons1St1ng of paJr-wJse JDcomps.rahle
tuples Us ng- the sotut1on bf Probl-em 1 1 1.5 {t) '\\ e obta1n the
requJred statement
1 1 17 Let 1J i:t 'A: he the numbfrs of those coordJna tes
"""'-' ,._ ,_ ~ f'W f"V
at "htch a and ft are dtfierent Any vector y such that a~ y ~ ~

~
c& n be nhta 1ned fN m a. by t e-plat l ng 0 by i 1n -s{)roe oi the aho'V t-

mentloned coordinates The number of ~ays 1n v.bJch Ci:iUCh A sub
.btl tution can be made JS 211
i I 18 For the cube B 1 the part1t1on conststs ol a s1n~le cha1n
Z == {(0) (f)J Z 1 {(10}} and z == {(00) (01) (ff)j_ c.an he
ta k~n as a pa.ttt. t 10n t ba'--ns tor BI 'r h-e pto p~rt~e-s {1) and f2} are
satudied for these pattJtJons \\e assume that the statement JS
pt"O\ ed I or cubes v. 1t b a d 1mens1 ana }It y not exceed t ng n and then
prove the stateme-nt !\Jr nn+t B} h) pothes.~s there exlst.g a part1
tiOD of faces B~+ 1 "+1 and nn+l + 1 lnto a cha D \\ htch .satJ.s6e.s
the condlttoD.! (1} and (2) Pattlt ons, tn BS+1 11 -+• and 8~-+l '~+l can
he chosen Ul such a \\ay that they .are 1somorphtc 1 e such that
~ ~ f'IIIJ
the chaJn Z 0 = (a.1 a 1 a,) beJongs to a part1 t1on of the cube

III'P ~
n.,..l 1f and only 1f the cha1n Z1 = {a a 1

r¥
B~tl a .. } obtained
from z. by substitUting
,., un1ty for zero at the (n t} th coordUJate +
lP each tuple ctt In Zo belongs to a partltlOD of the face Br-~" 1 n+l
AtJ ~ ~
z. ~ {at
./"ttl .,...; ,..,
Let a* a.} and Z 1 ~ {a. . a, a .. ) he '''{O 1.so

CHAPTER ONE 291
morphic chains in partitions of the faces B~1 + 1 , 11 + 1 and 8~1 + 1 , 11 + 1 •

" l"'oJ "" I'OJ "'
We construct two new chains Z 0 = {a. 1 , a.2, . . . , ct. 5 , a.~} and
A l"'oJ N 1'1 A
Z1 = {a.;, a.;, ... , a.;_d. The chain Z 0 is obtained from Z 0 by
~
adding the vertex a.; E Z1 , while the chain Z 1 is obtained from Z1

.
~
by omitting the v~tex a.;. We proceed in th~s way wi~h: each pair
of isomorphic chams. As a result, we obtam a partitiOn o_f. the
cube Bn+I into non-intersecting chains. The fulfilment of conditiOns
(1) and (2) can be easily verified.
r
"'V 11- k
1.1.19. 2k ~ . •
!
t=O
1.1.20. (2) We prove that the stratum Br is a complete set for
n :::::. 2. Let a."'1 be a vector in Br•. in which the i-th coordinate is
~ ~
equal to unity. Let the distances p (a. 1, ~) = r1, i = 1, n be spe-

~ "'
cified. Let us demonstrate how ~ can be reconstructed. If II ~ II =
k > 0, we obviously have r 1 E {k - 1, k + 1 }. Let max ri ;;;;;= 2
1... i ...n
~
•
and A (~)
= {i: r 1 = rum ri }. Then the i-th coordi-
.<'
1...)~11
nate of the vector "'~ is equal to unity if i E A (~)
"'
and equal to zero
~ "'
otherwise. If max ri = 1, then ~ = 0. For n = 5, the
1:s;;j,.; II
stratum B~1 is not a basis set since the set {(00001), (00010), (00100),
(01000)} is complete. (3) For an even nand k = n/2, for any n > 1
~ "'
and k = 0, n. (4) Let A = {a.1 , • • • , a.k} be a complete set. Let
"' , "'
us consider the set of equalities p (a. 1 ~) = r 1, i = 1, k.
Obviously, r 1 E {0, 1, ... , n}, i = 1, k. Each tuple (r1, r 2, •.• , rk)
for which the set has a solution uniquely determines a certain
~
vector ~ E B 11 • Hence it follows that (n + i)k ;;;;;= 211 • Hint. The

upper bound for the number of vectors in the base system can be
"' ~
obtained from the fact that the set of equations p (a.it ~) = r 1,
1 = 1, k, can he reduced to an ordinary set of linear equations.
For k > n, some of the equations of this system can be expressed
as linear combinations of remaining equations. Such equations, and
hence the vectors corresponding to them, can he omitted.
1.1.21. Sufficiency. The commutat ion of coordinates of all
vectors in B 11 , as well as the inversion of the coordinates i 1 , i 2 , • • • ,
lk of a~l vectors in B 11 do not change the separation between
the vertices. Necessity. Let <p he a mapping of the cube B 11 into itself
,...; - ~ l"oo ,...., ""
s~ch that for any a. and ~. p (<p (a.), <p (~)) = p (a., ~). Let us con-
stder the set D = B~1 U Br which is complete in B 11 (see the pre-
Vtous problem). Let <p (D) = {<p (0), <p (a. 1), • • • , <p (a. 11 )} be an
• "' "' rJ
.Image of the set D and a."' be a vector in Br whose i-th coordinate

•
1
IS equal to unity. We put ~~ = cp (0) e cp (a.j), i = 1, n. We
"" ,...., f'J
292 SOLUTIONS ANSWERS Alr-.D HlVT.S
./"tV ~ ~ ,_. f""'J ,.., ,.,
have II P.t U= P (0. ~~) = P (~ (0) $ ~ (OJ~ cp (0) ffi rp (a,)} :.:;
~ ~
..... oN
p (CJ' (0}, cp ((Xi}) = p (01 a 1) = 1 Let 1 (1) be the number of the
~
eoord1nate oi the vector fi 1 \\bleb .ts equal to untty ~ = 1, n t
Then the mapp1ng n t-+ J (l) IS a permutJ1t1o.n on the set {t 2

"""""'
n} For an a.rbttrary ct (aJ =
;.,
, GCn) JD Bn, \\e put .n {a) =
('tot
(l.Xn(d , B:c(n)) We show that for &D) y E Bn, the equah'y

#'IV ,-w .l'fltJ
'P (y) = n (y) e cp (~ Js valuJ Thus the necess1 ty 1s proved Jn ...

"'¥ {'W f'V i"V ~ ~ ~ II¥
deed, p(~ {y) ~ (0)) ~ p, (l 0), p (~ (yl~ ~ («,}) == P {y, (tl} On

...... ,...,. ~ #"tJ ...... ~ ..,.;
the other hand p (n: (Y) fP f9 (0}" r.p (0))

r.v ~ ~
P (Y~ O)t P (n (V} $ '9 (0}, =
cp (a~)} = p (Y a.1) Ohvulusly, the .stts ,ql (D) and D are complete ~ ,., -!!'ow
lD Bn Hence tt folJo-ws that rp (y) and n (y) $ cp (0) COlDCldt

2n•L-\
I I 22.
21'-1
1.1 23 (1} For n= 1 and k= Ot t. the statement J.s true.
i)
\Ve assume that == 0 for all k:::::;;; n 1 Let us suppo~e
k
tba t the statement 1s proved for a II d 1men~u onal1t1e!t wb1ch ,.....are
less than or equal to n -1 and for all k = o. i, Let a be
f"'W
(a.1 ('L1l) and a/ be {tti a., l} =: ( «:t t Q. n) l f fl.1-==. a!'1
~ ,. n----f--{lt--1}
we ohvtously have ~At~ (a.}~~ ~At~-~{~)\:;;:: 1 + k +
+ n-1-(ht-'l) __
l ~ 1 -r
n-f 1
k + + n-iJt} If a 1 ~ I
1
~ n--t t .... n-f
then JM~ (tL){ ::: k + }M~:i {a.')\ =- k +1+
n-i-(1 1 -1) n-i-{lA-1) n---11 +
k-f + + I =f+ k +
n -- 'lt """' l"'ftJ
t {2) For arbLtraey lc and a€ B~, the number '-'~ (tX}=

t"t.. ,.,
I Mf (a.)t w1ll be called the number of the tuple a 1n the st ra
.,..,
tum ~ Le\ a, 'E B~ be a VRtor whose un1t cootdlnates are
i1• ~ l4~ t ~ £1 < < lk ~ n Then In VlfW' of \\bat has
n:- 11
been stated above, fLl (~) = t
1
~
+ k
+ + n-Ul
Fur·
....... .,..,
ther .. let the tuple I} EB! be obtained from a by substttutlng zeros
...,
lor the last k - l un1t coord1nates \Ve show that Zj (M~ (a)) =::::'
Mi
,..,
(p) or, whleh the same, (MA (~)) ~ (y 1'1 (l) ~ f'I (~))
JS
l'a.l
zr n ~ .....,
.......
~
.......
ri'Y
We .first pro"Ve that for .any y EBlsuch that l'l (y) < 1-'4 (~}, there
CHAPTE R ONE 293
exists a tuple '& E B~ such that < 6' and Ilk (o) ~ Ilk (;:;:), We
~
y ~
choose for o a tuple obtaine d from y by substit uting unities for ~
the last k - l zero coordin ates. We must prove that Ilk {o) ~
~ ~ "'
[lit {et). If Ill! {0) > Ill! {et}. there exists a t such that Ctt = oi for
i < 1, at = 0, Ot = 1. Consid ering the tuples "'
a' and "'
o' obtaine d
from(; and by ~ubstituting zeros for the last k - 1
Brespect ively
~ "'-J ,...., ,...., ,...., '
unit coordin ates, we have Ill (a') ~ Ill (o'), a' = ~. !l! (o') ~ ~
~
Ill (y}. We arrive at a contrad iction to the inequa lity Ill (~}>
~ "'
Ill (y). Similar ly, we can prove that for any y E Bl1 such that
rt. rJ
v.
""""' ""'J ,_
!lt (y} > 1 {~) there does not exist a Luple o E Mk (a} such that
~ "' n
y o.
< {3) It follows from (1) and {2). {4) Let A s Bk and A be
other than the initial segmen t of the stratum B~. We
1} v (~ < ,..,~ v (~.
constru ct the set F .s B~ such that IF 1=1 A I, ,..,
aEF aEA
I ZL1 (F) I ~ 1 zL1 (A) 1. Thus, the statem ent is proved . We
~
denote by K (a) the set of the number s of unit coordin ates of the
11 """"
tuple a. Let a E B k "-A and ~ E A be the tuples on which
,...., f'V
~ ~
min I K (o}"-K (-r) 1 is attaine d, where the,...., minimu m,...., is taken

,....,
"""" ,...., n ro.~
over all pairs (o, -r) such that a E Bk"-A, -c E A, v (a}< v (-r}. ,...., ,...,
n
,..., f'V
Let M = K (a}"-.,K (~), N = K (~)"'-K (o:). Obviou sly, M N =

"' "'
0, I M I = IN I· For arbitra ry a and -c in Bn, we denote by
0".
,....,
f'V
.....,
the tuple of the

,...., rv
form a
,.._,
ffi (a,....,
n -r),
,....,
and assume that o:'
,..., ""J ,...., ,....,
=
a"-~ and W= ~"-a· Clearly , v (ex'} < v {W\. We put
,...., ,...., ,...., ,..., ,...., ,...., ,...., ,..., ,...., "'
D = {y : y E A, ex' n y = 0, ~· < y; <Y"-Wl U a' ~ A},
,_ ,...., ,...., 1"0.1 ,...., ,....,
E = { o : o = <v"-Wl U ex', y E D },
F = E U (A"'-D) .
,.., "'
Obviou sly, F .s B~, I F I = I A I and 1} v (ex) < ~ v (ex),
~ ,..,
. ,...., ,...., ,...., cxEF"..
aEA
smce v ((y"'-Wl U a') < v (y} for all y ED. It remain s for us to
prove that I ZL ZL 1 (F) I ~ I 1 (A) I· We shall just give a hint
prove that if a E Z~_ 1 (F)"'-
~
towards the solutio n. We,....,must first

,, ,...., ,...., "" ,....,
Zk- 1(A), then a n W = 0 and ex' < a, and then show that the
,...., f'V f'V ,....,
tuple e =.<a"'-a') U Wis contain ed in Z~- 1 (A )"-Z~- 1 (F).(5) Hint.

Pr~ve by mductl on on k - l. (6} Let k be the highest integer for
which ~k > 0. Then l = k - 1, the statem ent is reduced to (4}.
By takmg an inducti ve step (induct ion is carried out on k - l},
k
we assume that Ai =A n B~,
1
A~=A
t
n( U • .
B }1} ) and note that
1=1
Zi' (A)= Zi' <Z/'+1 (Az+1)) U At.
294 SOL lJTlO"'JS ANSWERS AI\ D DINTS
It 1s clear that
cs;B{
min
l C t =-m
!Z~ (C)~ s:;
CSI:l/
)UID
l Cl >m
I zr (C) I
By tnduct1vehypotbes•s mLll J Zf+t (At+ 1) 1 14 atlaaned when each
z~ (A) , > l 1 1! an 1n1t1al segment ol the stratum B~ But
lh m &ceord l ng to {2) the 5e-t Zf+1 (A l +1) IS an ln1 t1 a.1 y.gm-en t By
an approprlate eh.otce of A 1 the set Zf (A)= Zf (Zf+ 1 (A z+ t }) UAt
can he made an 1Dit1al segment of the st-ratum Bi Th.Js
1eads to the stat em en t
1 1 24 For each I= 0 n \\e suppose that A,::;::
J"'ffl ,_
An ~i ~
\\e
denote by t (n m k) the number of p~urs (a ~) such that cxEAm.
~ ,..,.. ~ ~ l"'tJ ~ ,[[''ljJ l'fii.J .......
~~A ..Sn a= 0 or II Uct=j3 Jl EBk UB!-+~r.. By hypothesJs JlfA

n--..m
On the one- hand t {n m h)= om On the other hand
k
t(n. m k) ~ m+k -6-rn-.!4. n)m + ~ --~ . . . .. )(m+ k) Thu ':_ . . . _) (n .m

. . . k)
\:e:a6s \.() \ n:e- -s\a\~~\
~ ..... """"
t. t 2.5 (l} U!\ a. be the ct:ntet of the &-e\ A Then A = {a $ ~
~
~ E A } l s. the teil u 1red s-e-t {2) \\ e ean assume that 1n v u~w o! (i)
.....
0 1s the cent~~ ot A In ord-er to eons truet A 1\ 1s &"Uf fte1-en. \ 1-rtr
each i
-
=::: i. t
-n. tO: subsll u le th'S! t up te ,.... a fc.r -e-ae b tuple
Q. a. 1n A~ !ot
~
JIAi .......
wh1c.h a.i il A (3} H1nt Let A ~ B 11 ha.ve \b~ tent-er 0 and poss-ess
the property (1) Let S t J he,., a transform a t1 on cons1
,...,
st 1n g J n that
:G'
~
,.... each a E A
for such that ai J E A the tuple a 1s replaced by
a 1 J Let us arrange tbe pa1rs (r 1} !uch thst 1 -::::; , < 1 ~ n as
follo\\""S
(t n) (f n - i} (1 2} (2 n) (2 n -- f)
(2 3} (11 - f n)
App ll tng to A eonset u t~: ve Iy all tra.n~f or mat1ous Sl J s ta ttlng

from (c 1) ::=; (f n) and Jin1shJng Wl th (z 1) === (n - 1 n} y, e obtatn
the required A
I t 26 {2) Fur A such that ~ A 1 :=:::: zn t v..e can take any
~
(n - t) dLmenstonal f.aee wntat.nlng t F<lt &n. odd n we can take

,... .f'V
+
the: .set {tX ~~ d !~ ~ {n. i)/2) On the o th~r hand ~f \ A \ rtJ >
2• 1 there ex.15t t n A two oppcSl te tup1es w be s.e 1nle ts.ec.t1on l& 0
(f) Htnt Cons1der the set {,

a. . II ,.,;
a rJ ~
] n+ k+i [
2
CHAPTER ONE 295
~ ~
1.1.28. (1) The proof is similar to 1.1.25. (2) A= (a:: a: E b n,

~
11 a: 11 is even}.
' .
11 11
.
1.1.31. (1) Each face B 0 ' ' 0···• 'k of direction I= (i;, ... , ih.}
1 •• • II
can be put in one-to-one correspondence with its rode, viz. a vector
(y 1 , • • • , Yn) with coordinates Yi E (0, 1, - }, i = 1, n, such that
. = o 1, • • • , y. = o, , the remainin 0rr coordinates being blanks.
Yt( lll "
The fixed k places can be filled with zeros and unities in 2" ways.
(4) The required number is equal to the number of vectors (y 1, • • • ,
Yn) with k coordinates in the set (0, 1} and n - k blanks. The k
places from n for significant symbols (i.e. for 0 and 1) can be chosen
in ( ~) ways, and the fixed k places can be filled by zeros and unities
in 2h ways. (6) The code of the k-dimensional face containing a given
vertex (a. 1 , • • • , a.n) can be specified by arranging k blanks instead
of certain k coordinates of the vector (a: 1 , • • • , a.nl· (8) The inter-
section of two faces is a face. A j-dimensional face can be chosen
in the face of dimensionality l in ( ~) 21-j ways. Thus, l coordinates
have already been fixed in the code of the k-dimensional face.
Among the remaining n - l coordinates, we must arrange k - j
blanks. This can be done in (
"' ~ ~
1
~= /) ways.
1.1.33. Let y1 , Yl• y3 be the codes of the faces g 1 , gl, g3 respec-

tively. Two faces do not intersect if and only if there exists a coor-
dinate at which the codes of these faces have significant symbols,
and these symbols are opposite. If, however, the faces g 1 and g 2
intersect, the code of their intersection contains significant sym-
"' "'
bols of coordinates at which at least one of the codes y 1 and y~
has a significant coordinate, and significant coordinates of the
code of the intersection coincide with the coorresponding coordi-
"' ~ ~
nates of at least one of the codes y1 and y.,. If /) is the code of the
intersection of the faces g 1 and gz and the face g3 has no vertices
in common with this intersection, there exists an i such that the
"' "'
i-th coordinates of the vectors /) and y 3 are significant and do not
~
coincide. Then the i-th coordinate of the vector y 3 does not coincide
. ,.., ~
With the i-th significant coordinate of one of the vectors y 1 or y 2 •

This means that the corresponding faces do not intersect. We
arrive at a contradiction.
1.1.34. The problem can be easily reduced to the case when
s
2j 2n; = zn. We shall carry out the proof by induction on 11.
i=l
~or = 0, 1, the statement is obviously true. We make a transi-
11
tiOn n-. n + f. Let min n; ~ 1. We put n:1 = n; - 1. Then n:
1.;;; i.;;;s t
SOLUTIONS .ANSWERS A.l\1~ HlNTS
il
ue non negattve oumhen~ and ~ 2"i '=:;$ 2,. By mduct1ve hy

~~
pothesJs there ensts a ~fttllOD of each ol the fates B~+t n-+l
a.nd B~+ 1 n+ 1 of the cube an+l mto the laces. Q( dtme.nstontllttu
n, n.. We choose these p~rt!_!.ton s .so that they are 1denb
eal, 1 e .such that lor each t = 1 11 ' 'he UDIOD of 11 -d1 mens• onal
111 1
faces Jorms an ~~~~1mensJonal face g~, We ohta1n the requuet\

parttt 1on Far Dll n n1
i~h,.
=
0'!. let m he the Jl umber of tb ose i
J:
for """hu~h n• ij 711
== 0 The condJtion 2 = zn..t.a 1mpl1es thal
1-1
m 1s even Lel 11! t:onsl d ~:r a new set n ~... nz, , n.-m1 1 obtalntd
1111'
from the prevtous one hy suhstJtutll1g m/2 UnltJes lor m zeros

As ear11 er, v.. e part 1t1oD t.he cube B "'+1 1n to facest and tben spl1 t
eertatn m/2. faees of d1menstonal1ty t mto m fates of dtmens1on
alttY 0 ,..., ~
~
I I 36 (I) MaJOr I z.a t IOD Hint Consl d er N ~ Cl E t
;::$. { on
rl ~ I~J ls even lf1uonzat1oD Let N G B"'* ~ N ~ < 2" 1 There
t:us\.3 (see 1 1 31 and l 1 34} a par\1t1on of the cube Bulnlo tLDDln
tersect1ng faces of dJmens1onal1ty t The number of facet~ ID such
a part J t1on 1s .211 1 Hence J t fallows that at Jeast one of the ,..,
faces
;v
does Dot con l8.1D \: erllces f.rom N ( 3) If fn t Consld er v = {a: Ct

raJ /!"'V ,.., ~
e
B"~ u~ u~ 0 (mod 3}} OY NJ. ~[a. a eBr\ ~{a;)~ 3 (mod~)}
......
(4) )!ioorJzat1on lt should be noted that for a vertex a = (a 1 11
Ctn) to he eonta tned ln. an (n --- 2)-dun.en.s.tona l fate w Lth a code

~
'V = (t1 ~ Yn.) (see sol uuon of Prohle m. l t 3t) l\t th Sl gn 1fican t
coord tna t e-s 1n the places , and 1 11 1 t 1.S n e-eessary 1hat a 1 ~ 1~ and
aJ ~ ?1 Lel N s B'tl.ll ) N l ~ m. be a set ,.., .such that Jls eacb ( n- 2r
dlmens1onal face- c-onta\n! the -..;ertr::x a frmn N Lt\ us r:oll!ldH
a b1nary matrJx .AI ,.,hose l1nes are vectors JD N The abo\eremar.k:
1mpl1es that for each patr of numbers (l,. JJ t ~ 1 < 1 ~ n. and
.......
each pa1r (0' 1") a T € {0 f) tbere e.JJsts a ltne a ~ (a1 an)
such. that «-t = «, a,J == 't Hence any t'Ao columna of the mat
nx M are pau-v.Jse Incomparable The number of pa1N1.se 1ncozn
parable bmary ,~;ets of length m does Mt exceed ( , ; ) Hence
1 21
;<r- " Ma)on~abf)n L-e\ rn be the mmlmum mtegeuu~b
({m7:ZJ}
that ( (m7 ) ~ 11 \\e coru:truet a btllary matru: ftl \1o Ith ndmes
21 ,.., ~
and a pa1J'1\.1Se Incomparable columnB \Ve add the lmes 0 and t to

the ma tn.x M Then the ohtarned set ol vectors bnes 1s 'p1erc1ng"'
lor lhe stt of all (tt - 2) dlllltu~onal faces ol \he tube B"-
AI" ,..
t. t l1. F{)r n ~ 1 2 the sta\t-mfYn.t ~~ oh-v1ous Let. a 1 a 2 ;
~ ~ ~
a. 1n ht a C) cle 111 Bn L"¥."\ f. 01 -,.. ber:e- 1\- E. Bu\ o E \0, 1 }~ denot-e

CHAPTER ONE 297
a vector of length n + i, whose first n coordinates coincide with

the corresponding coordinates of the tuple "'
l"'J
~. and the (n
I'V ~
1)-th
,.....,
+
coordinate is cr. Then the sequence a: 1 0, 0: 20, ..• , a: n0, a: n1, . · .,
2 2
~ ~
a 21, a 1 i is a cycle in B 11+1 , containing all its vertices.

1.1.39. (1) Yes. (2) No. (3) Yes. (4} Yes.
1.1.43. Hint. It should he noted that if the tuples "'
a = (cr1 , • • •
~
cr 11 ) and a' = (a;, ... , cr~) are comparable, then one of the sums
n n ,
~ (- l)ai ai and 2j (- 1) aiai is greater than unity in
1=1 i=i
absolute value.
1.2.
.,n-1
1.2.1. 2w (n;;;:,: 1). 1,2,2, 2 (n;;;:,: 1). 1.2,3, qn + C}n +
... + c~-i' where rn = 2n (n;;;;:,: i, k;;;;:,: i).
1.2, 7. (2) 2 11 +1 (n;;;;:,: 1). 1.2.10. (1) In five ways. (2) In
nine ways.
1.2.12. (1) Basis of induction. If the complexity of a for-
mula is unity (generated by the set of connectives M = {-,, &; V,
-+}), thi5 formula has one of the following forms: ( lx), (x & y),
(x V y) or (x-->- y) (within the notation of variables). Therefore,
the statement is obvious for formulas of complexity 1. Induction
stC'p, Let the statemPnt be valid for all formulas (generated by
the set of connectives M) of complexity not exceeding a number
l (l;;;;. 1). We take an arbitrary formula '2! (generated by M) of
complexity l + i. We assume for the sake of definiteness, that
ca' = (!B V C£) (remaining cases are considered similarly). The
connective V between formulas !B and G: has an index equal to
unity. Further, if the connective in formula !B (or <l:) treated as
a connective of formula \8 (resp. G:) has an index greater than 1,
the index of this connective in formula '2! cannot be lower. If,
however, the connective in \0 (or f,n has an index equal to unity,
its index in formula Ill will he equal to 2 (the index increases since
formula '2! contains a left parenthesis before subformula !8).
1.2.15, (3) (0 1 0 0). 1.2.18. (3) Ill......, \B. 1.2.19. (2) / 1 (x, y) ==
0, !2 (x, y) = x, / 3 (x, y) = y, ! 4 (.r, y) = xED y. 1.2.20. (1) ((x t
y) t (x t y)). (2} ((x t y) t ((x t x) t (y t y))). 1.2.21. (1)
((x I y) I (x I y)). (2) ((x V (x V y))- y}. (3) No, it cannot.
1.2.22. (2) The function represented by a formula generated
by the set f-+} assumes the value 1 on at least half the tupt~s of
values of its arguments. The function xy does not satisfy this
condition.
1.2.23. In general, it is impossible. For example, for S =
x
{I}. the function cannot be represented by a formula of even
depth generated by S.
1.2,24. If only constant functions can be represented by a for-
mula generated by the set S, the statement is obvious. Let a cer-
tain function f ~n) ¢ const be represented by a formula generated
by S. Identifying all the variables in the function f (;n) and in
298 SOL UTIO~S ANSWERS A.l\D JIJt..jTS
the Iorm ula re pres en tt tJ g It we ob ta 111 a f unc t1on iF (:1) (and the
form u Ia W (z) r orres pond 1n g to it) "\\ e must eon.!U der the f ollowmg
three eases (a) ~ (:r) Js a constant {h) 'fP (z) ~ s and {c) 'F (x) ===
Ha v Lng o b ta lned (Ln cases (.a) and (c \} a !orm u1a representtng the
z
function x (and h.avlng a depth not less than 1) we tan construct
a formula tlf a.n 1ndefi nt te Iy large depth for each speetfic fur.tct\Qn.
represented by a formula generated by S Let <p (2') ~ 0 Then
"""
there ex1sts a tuple a - (a 1 ctl an) on •htch the functton [
1s equal to f Subs.t1 tut1ng ~ or all ~l. corresponding to at
and rp (.r) s 0 for the remaiDtng z 1 we obtain a Iunct1on $ (z) = ~
=
-
If howev-er <p {.t) ~ t the fun-ell on :r. ~~ u be a h tat n-ed 1n a Sl. m\W
¥a-a.y
t 2 26 (2) :J: V 11 ~ {z ....... y) ..- !I (3) ~ ft# y~(.t......,.. y) &. ~~}
I 2 30 (2} In general 1t cannot For example W y ~ z 19 =
a fcrmula wb1eh 15 nnt ldtn\1cally thlt However the Iormu[a.
8: 1
=
... Wt,W ( I/1-..... (YI -F- Y1) HI Jdentlca.ffy true
t 2 31 (I) II 11.e assume that 1 (0 0) .._ 0 then putllog :r = lJ=
z - 0 "e oh laJn I (! (0 t(0 0)l f U (0 0) t (0 0}}) = t (J (0 Ol
t (0 0)) = 0 wh1 ch contra dJc ts the cond 1t1 oru: of the prob le rn
Thus I (0 0) :=::; 1 Putt1ng f (0 i \ = Q we aht.aln ([ot ~ :::::: J ~
.: ~ 0) U (0 I (0 0}) I (j (0 0) I (0 on) = I U (0 1) I (I t)) ----
1 (0 t (1 1)) and hy hypothf!~l! th'2; t~la t" oo. J (l 1} - () must bt
satJ:sfied However considering that z = 0 y == 1 = f we .see
t h.a lf U ~o f f\ 1}} I (J (0 i.} 1 (0 t)}} ~ 1 (I (0 0} I (0 G)) -- 0
whLch lS Jn eon tradJet1on to the condttJons of'fhe problem Thus
I {0 1) = 1 F1nttlly assuming that t {t t) = 0 and putt1ng ~ ==
!1= 0 and • = t "e get I (I (0 I (0 t)) 1 (f (D 0} t (0 t) J) ==
1 U (U t) / (t t)) .==:: 1 (1 0) Tak1ng tnto account the condt tron
of the problem v. e ohta.Jn f (t 0) 1 But ID thts ca:,e I (/ (I
I 11 ! )) I (/ (l l) I (t t))) = I f/ (1 0) t (0 O)) = t (1 t) = 0
wh lch con trad 1cts the cond 1t1 on of the probIeill Consequently
I (t t) = t ~s well Thus f (0 0) = f (0 t) == 1 (t t) == I Hence
I (x g) = ~-+ y or I (z v) ~ f The equivalences (a) (e) are verr
.tied d Irec tl y (for e.xa m p le by usJ n g the hast c eq ni valences) (2)
No they do not It 1s su fficJent to consider the runctton f (z y} ==
~,.., g
t 3
f 3 6 2n I 3 1 ( t) 2n 1 (2) 2n-- 2 (3) 2n 1 J 3 8 (I) It X l
(2) k X 2m+ l X 2n..._ k X l (3) k (21D-lJ+ I (2,.~k) I 3 9 (2)
1~ (;I)=~~~., 11 (~) ;1 1~ 1 3 {;s) = :z, I 3 1~ 2( l/4. l 2." t 3 12
It
8
2 ;.~2SJ; S~= h
1
12 f 3 16 4 I 3 IS f 3 19 n
t i
~--o
f 3 20 0 1( k 1s odd and 1s even t 323

(f10lfllOiif001fO} f 3 26 .:t'n
CHAPTER ONE 299
i.3.29. Hint. Using the decomposition formula, carry out the

proof by induction on n.
1.3.30. (1) 211-k + 2h- 2; (2) ~ (211+1 +(
-1)11).
1,3,31. For n = 1, the statement is obvious. We proceed by

induction from n - 1 to 11. If l ~ 211 - 1 , by inductive hypothesis
there exists a polynomial P (xn-"' 1 ) of length ~ n- 1, for which
~
1N p 1= l. Then the polynomial x 11 P (x11 - 1) becomes equal to unity

at l vertices of the cube B 11 • If 211 -l < l ~ 2 11 , we consider the
"' 1 ) which becomes equal to unity at 211 - l
polynomial P (xn-
vertices of the cube B11 - 1 • Then the polynomial x 11 P "' (x11 - 1) E9 1
is the sought one.
1.4.
1.4.4. Hint. Let K=:z:~ 1 ... x~n, L=x~ 1 ... x~n and let KaL
be an e.c. obtained from K by deleting the letters x~i for which
"'11
a; =1= ~i· Note that each implicant of the function f (x ) can be
represented in the form KoL, where K and L are suitable elementary
conjunctions from a perfect d.n.f. of the function f (.x "'11 ), which
may coincide.
1.4.5. (1) 211 -1. (2) 2n-1. (3) 3X2n-a. (4) 211-2. (5) k+(n-k) X
(n-k-1).
I 11-l
1.4.6. (1) ( .
,k k+m-l
1.4.8, (2) ;-1X2X3 Vx;z2 VX1~ VX1X~ VX2X3X~ •
r
211 2n-r 11 n-r 'V r • 11-r
1.4.9. (1) 2 - . (2) 22 -2 LJ (-1)i . 2- 2
1 •
L
--
1.4.11. (1) x2x3V-x1x~ V x 2x 3 Vx1x 4• - i=O
1.4.12. (!) x1x3, x1x2, x;,;;2x4, x2x3x4 •

. 1.4.13. Hint, Prove that there are no two proper tuples of two
dtfferent kernel implicants which are adjacent, making use of the
fact that the set N c. B 11 such that 1 N 1 > 211 ·1 always contains
the edge (see Problem i.1.44).
1.4.14. (1)211-1. (2) 211-3_ (3) 0. (4) 211-2. (5) k.
211 211-r
2 -
1.·1.15, (1- (1-2-2r)n-r).
1.4.18. (1) 1. (2) 52n-r. (3) 52n-l. (4) 1.
1.4.20. One. Prove that all simple implicants are kernel impli-
cants.
1.4.23. Hint. Carry out induction on n, using expansion (3) of
t he function in the variables.
1.4.24. (i) For two functions. (2) For 211 +1 functions.
1.4.25, For example, k = n211 - i.
300 SOLUTIOzrr.. S A.l"-4 SWERS A'\ D HINTS
n
f427 (t)01fkif:;.O k+m<n If k=O
rrt
(2) On t" o functtons for even ~~o and on four
odd n
j 5
1 5 2 (t} X;~ (2) :J!1 Zt <J} ~f. (4.} .t'1 1 !j 4 (f) .t51: .2:~ (2} Zl
Jl (3) ~a 1 5 7 m 1 5 8 (1) x 1 ~~ ~~ .:r, (2) ~~ z 1 z 1 (3} z 1
('"t) ~ z,
t 5 tu
.z,
(t} n ~ a
(2) No values. o( .n (i} n # 3 (4) £,eo. r.r.
2' ..
(5} n ::::$. 4k -- 2 k === f 2

t 5 12 (2) ) Pt {X') ~ = 218
I 5 13 H1nt Ucung the condJlton of the problem prove that
~ ,...,. ~ .f'OtJ ,..,.. .-.,. ,.,.,
there e.lJst tupJes a ~ l such that r (a ~ l = p (~ y} ::= 1

~ ,l'9til .l"'ttJ ,...,.
Let a: and p d1fler tn the i tb. and ~ and "l 1n the l tb -place Then
the vartahles z 1 and x 1 are essent1al _
I 5 i:t H1nt Jt JS sufftctent to prave that 1I r 1 or ZJ appear
Ln a PrLme tmpllc..ant K of the funet1otl I then l' 1 lS an ~~Ullal
'ar1ahle Delf'tJng .z~ from K we o.bla1tt an elementary cOnJUDC
t1on A that 1snot an 1mphcant Hence Jl follows that there e.:rJSt.!l
~
1"'fJ-
a tupJe a -= (a. 1 1 a ai+t a, ..-¥

a 11 ) !iuch that A (a) === t
N
l"'W
but I (a) 0 At the same tJme I c~') = f SlDCe K (ct 1) = 1

Hence z 1 Js essen t1a1
~
I v 15 H1nt If z 1 explltlt1y enters the po}ynomJal P (~)

l he Ia t t er can be presented 1n the form :r1Q -EB R "'here Q and R
~
are poJynom1aJs 1ndependent of :r, and Q ¢ 0 Let the tuple a =

~
(al al 1 '-'t 1 fl. ) be such that Q (a}== t Then P (at

rxl 1 :Ct a +J """' The statement now folio\\ s
an l = .:rt E9 R (ct)
from Problem t 5 I (1)
~
I 5 16 No It does not Hint Conslder I (.tn) = x, e3 .rl e

e Xn
""""
t 5 17 H1 o.t Let f (xn) sat1s£y the e \)U d ttl nn o{ tb-t ptahl'E! m.,... ~
Then there e~l!.tS an lnt-egtr [ {1 ~ I ~ n} Sut:h that. ror any (t and ~

f"'pJ l"w ,...; ;l'lw
{sueh that t:J. E an ~ -E 8 11 l) f (a.)+ I (~} Thf: stat-ement M~

fnlJows from the fact tba t for any, 1n B, t U B~ th~re e.J:J.st ~dges
of .;iDy d 1reettoo of ,
l a 18 H1nt See Problem I 5 i3
~
I 3 t~ If I (z-'n) ts ex pressed through a po}J: nom al of dcgrre

greater th.an un t) \\lthout loss of .gen~ra.hty tt can be pre~nted
10 the form :r1 z 2 P 1 @ ~ 1 P~. ffi z 2 P 3 $ P, where P, ~ 0 .......and
P1 P• P'3 and P1 dPpend on the \ar1ahles x~ z~ Let a
.......
(a, an.) be a tuple .such that P 1 ((l) = ~ Then th.e tom
3. 'R. ,..,;
ponen t I a. 3 e1 n. (z1l) ha:s the form z1 z 1 EB i.. 1.r1 ED A,.rt e i\.J and
CHAPTER ONE 301
hence this component essentially depends on .r1 and "x 2 • Hence it

follows that the ,....,required three tuples will be found in the face
Bn; 3, ···• n. Iff (:en) is expressed by a first-degree polynomial and
a a ••• O:n
essentially depends on :r1 and :r 2 , then for any tuple (o: 3 , ••• , O:n)
any three vertices in the face Ban;3 3u•• ·a··'n n are the required vertices.
1,5,21. We assume the contrary. Without loss of generality, we
can assume that the maximum number of essential variables among
. ,....,
the components of the form f~ (:c71 ) belongs to the component
,....,
t: (:en).
,....,
Then the assumption is equivalent to the statement that
t: (:en) fictitiously depends on a certain variable, say x i. e. t: = 2,
t:i =/~ti • Since x1 is an essential variable, at least one of the rela-
2 2
1• 2 =1= t 1• 2 andj 1• 2 =1= t 1• 2 mllSt be satisfied Let for exam-
tions t IO 00 II 0I • • • •
pie, 1~1 2 =1= f~i 2 • Then the function
2
/ I
= X1I01I, 2 v :r:1 IIII, 2
essentially depends on :c1 , as well as on all essential variables of

the subfunction 2
th
= J:. This is in contradiction with the fact
that /l has the maximum number of essential variables.
1.5.22. (1} The validity of the statement follows from the pre-
vious problem. (2} No, the statement is not valid. Hint. Consider
,...., -
I (x 3) = x 1x 3 V :r: 2 x 3 and the variables :r:1 and :r: 2 •
1,5,23. (1} ; 1, x~. (2) 1. (3} x 1 , x 2 • (4) x 2 -+ Xt, X"t ~ X2· 1.
1.5,25. 6. 1.5.26. Yes, it can. Hint. Consider the function x 1x 2x 3 V
x1x 2x~ V x 1x 3x 4 V x 2x 3x,. ,...., ,...., ,....,
1.5,27. The numbers of coordinates of the tuples o:, ~. y can be
divided into the following four groups:
A1 = {i: o:1 = ~~ = "1!=0}, A 2 = {i: o:1 = ~~· ~~ < "ld•
A 3 = {i: o: 1 < ~i• ~i = yd, A,= {i: o: 1 = ~~ = "li = 1}.
("'J ,...., ,.., - ,_ ~
Let f (o:) = f (y) = a, f (~) = u. Then if f (0) = I (1) ,...,

= a,
weputx,=x ifiEA 1 UA.,andxi=yifi EA 3 UA,.If/(0)=
"'
cr, f (1) = a, -
we put Xj = X for i E Al, X; = y for i E A 2• xi = z
fori E A a uA~. Iff (0) = a,
we put xi = :c fori EAt u Az, Xi = y
fori E .4 3 , and :ci = :: fori E A,. The essential dependence on the
obtained functions follows from Problem i.5.13.
1.5.29. ffi x2 ffi ... ffi X"n ffi a, a E {0, 1} for n;;;;;. 2, x~ 1.x~ 2 V
Xt
(J (J 11 - -
:z:2:cay
2 3 x
:C 3at , a·E{tl
, ·-1 , 3 ror. n-
, 1} , z_ -3·, x 1111
x , :c 11V x 11 f or
3 1 1 2 1 2
ll =2.
1.5.30, The function obtained from f (;,') by identifying the

variables .x1 and xi has the form :ctf!·1 i V id~oj· The condi~io.ns of
t~e _problem imply that at least one of the components f!i 1 and
1~1. J is not identically equal to zero.
302 SOLUTIONS ANSWERS A'10 fi[NTS
J 5 33 H1Pl At least one ot the funct1oD.s J and 1 becomes

equal to un1 ty on an odd nomber rd tuples
1 5 34 Hint S~ Prcblem 1 5 30
CJJAPTER T\VO
2 f
Z t 2 No Jt does not We assume {by detinltJ.on) that [AJ]
P I or any set JJJ 1D P • 1 e J n lrDd uce .such a closure opera tlon
T~en .re}atJons (1) t3) in Problem 2 t t t are obvtous1y s&tu;fied
wh1fe relat1on (~) k& not val1d
2 'l 3 (1} Yes (2) No (3} No (4) Y~s (5} Y~s (6} No (7} No
rf"'',J
2 I 4 (i) {1 J (2) {0 1 :t 1 ~1 } (3) {.r1: r 1 Y z1 z1 V :rJ V s J)

(~) l1 Z'l zl .f"¥ .l'2
.......
S'l $ .t'2 m.t'J)
2 I 5 Let 1 (.t"n) be a functl on ID a c1osed eJ ass wluch essen tJ
ally depends on all Jts varJab1es (n~ 2) Then the lunctJon
,, (y, !It y.,.. .Zt J:rt)
=== I U (VI Y:~ Yn) Z; .rn)
esscnt1alJy depe.nds on al] 2n -- f var1ables (here ("~ Y, y.,.} n
(z .;rn.} ~ 0} The funct1on I (zt .:t zn.J can then be
.su~.stJ tuted for the varJab1e g1 1n the funct1on / 1 to ohtatn a tunc
t Jon essen t1a lly depend 1ng on 3n --- 2 'aria bJ es and so on
2 f -6 IOJ ftJ lzl [0 1] 10 :rJ ft zl fx .xJ fO f .t] fO
f z .a-1
2 i 1 (t) Yes (2) No not always (3) No not ,al~ays (the only
eJ:cept1on heJng aJJ P 2 and 1ts com pleznent ,~12 a. closed empty
c}ass 0)
Z 1 8 (3) (.r-... 11)-+- g "Z$ z V g z :=::::;: z e z iB z (4) A nega
tJon I! ohta1ned from the second Iunct1on by 1denttfy1ng a1l the
varJahles (5) m (z Jl 0) = zy ~o t1) 0 = z
2 I 9 (1) K 1 E K, hut the str1ct Jnelu:sJOn K 1~ K 1 JS slso
po.ss1hle (2) K1 ~ Kt. lor example lor Jf1 = (.ty zJ and M 3 =
(z} we have K1 ~ (xyl and K 1 ~ P,,.Jll (5} See Problem Z t 9{2}
z t to (Z} Aft= [;t: 01 M, lz 11 (3} M1 ~ (x 01 M. = =
101 (5} See Problem 2 t t0(2.)
2 t t t {' } {o ; } (2} {1. :t e y } (3> {x y ~ &. y &. s l v
-(4) { ~ e 1 tJ\ (z l/ l)} SlDtt % 'Ef; 1J $ J ~ m {m (z If .s:) T4 (z y
:) m (s !I .=-)) aDd 1n the closed class im (:r 11 :)] any functJOn
~ hlt.h -ess-entl ally d e:pend i on one argument 1s equal to an 1d ~ tltll
(UD.-Cl10It
2 t 12 Let M he a preeomplete elass (1n P,) It means. that
(M I =1: P .. hut for any !unction J E P"'ft-1 we have {M U {Ill :.: : : P,
Jf IMJ i=: M
tbere e,1u:~ts a funct1on gIn {Ml,AI. such tJiat {MI ---
!M U{t}J ~ P~ We arr1ve Cit a eontradtclioll
CHAPTER TWO 303
2.1.13. If M} = Ml, the statement is obvious. Let Jll~ =1= "lJl.

We assumP. that .lll = .l/}. Since M 1 and Jl_ are closed cla~ses,
and Ml c.lf~ and M~_cM~. then ~rre~pecttve_of th~ form of
a function fm M 1 (or m .ill~), su.hstJtutmg for 1ts vapahles the
functions from Ml (= M~) dependmg on the same variable x, we
aoain obtain functions in .l/~. But this means that by adding to
lf1 an arbitrary function g in M,'""-..Jl/ 1 (=;l= 0), we get (M1 U
{g}l n Ml = .Ml =1= 1lf1, and this contradicts to the precomplete-
ness of the class llf1 in the class M.
2,1.14. (1) (0, 1, .-~], (x]. (2) (0], (1]. (4) (0, x], [x v y].
2.1.15. (1) No, it is not. (3) No, it i:: not.
2,1.16. If x EM, the statement is obvious. Let x ~ M. By
definition of the superposition operation, the substitution of an
identical function can be considered as a transformation made
only in the set M and carried out in accordance with the superpo-
sition operation. Therefore, [M U {x}] s= M U {x}. The inverse
inclusion follows from the properties of closure operation.
2.1.17. The statement follows from the fact that the function
x I y forms a complete system in P 2 •
~
2.1.18. Let f (xn) ¢canst. If by identifying all the variables

~
in the function 1 (x 11) we obtain an identical function or a nega-

tion, the statement is obvious. Let f (x, . . . , x) !E const. For
~
definiteness, we assume that f (x, . . . , x) = 0. Since f (xn) ¢

,..., ~
const, there exists a tuple a. = (a. 1 , ••• , a. 71 ) such that I (a.) = 1.

We put xi = x for a.i = 1 and xi = y for a.i = 0. Then from the
function f we obtain a function g (x, y) satisfying the following
conditions: g (0, 0) = g (1, 1) = 0, g (1, 0) = 1. If g (0, 1) = 1,
then g (x, y} = x EB y; if g (0, 1) = 0, then g (x, y) = x-+- y.
Obviously {x EB y] 3 x and [x- y] 3 x.
2.1.20. Since the set P 2 is countably infinite, the set of all its
finite subsets is 'also countably infinite. Therefore, the power of
all closed classes in P 2 which have fmite complete systems in not
more than countable.
2.1.21, Tf a closed class in P 2 differs from the sets {0 }, { 1},
{0, 1}, it contains an identical function (see Problem 2.1.18).
Consequently, it cannot be expanded to the ba~is.
2.1.22, Let SV be a complete system in M. We talw in :fJ an
arbitrary function j 1• It does not belong to any precomplete clas-
ses K1 , ••• , Kr 1 in M. Then we choose in :P a function {2 which
does not belong to at least one of the remaining precomplete clas-
ses, and so on. The number of functions in the system :P' obtained
as a result of such a process does not exceed ihe number of pre-
comple!e classes in M. The system P' is complete in ill, since
otherwise it can be expanded to a precomplete class, while it
follo\~s from the method of construction that the system :.P' is not
contai.ned completely in any of precomplete classes in M. If :P is
a basis, all the functions in :P must be used in the construction
of the corresponding system P', otherwise we would arrive at
a contradiction to tlie irreducibility of the system :P.
2.1.23. The proof can be carried out by induction on the num-
ber of occurrences of the connective-.. in the formulas representing
sr.n U'T10'\S A~SWFR~ A "'-D Hi'\1'S
the Iu nct 1ons J n the cl tJ.sed cJ a.ss {x _...... r;l If there l s one occurrence
of the conneettve ~ ID a formula the formula (accurate to nota
tton of \arJables) has vne of the !ollow1ng forms Z-flr z or :: ....... "
In th1:s casl? the sla temen t 1s ohvtous Further Jf it= ft ........ It w~
must consJder two posslble ca.ses
(a} The formula correspond1ng til the function f 2 has at 1rast
one otcurrence oJ th-e ~oDnKt~-...~...... Then 1:~ = ~J V <pt {Y"'~ and
~
-
~ -- II v I i" - yJ v ~J I v ~'!}
'b) Tht iorml1la correspondlng to th~ runct1on 12 ~3 a vat ahi~
(say v> Tben W == It V !J
2 i 24 'Ve carr} out tnd-uctilon on the number oJ occurtttnces
ot the eonnPC tJ \ e ........ II the connect1 ve ......... a ppe.ars only once we
~
hov~ z-.-. y and the s!a temenl 1S ob\ JOUs Let 2l ~ f.t- (:rn~---+-
,w
J1 (x ) an(l 1f'-\ t l1e (unction J cfJITespon drng to g: rssen t1 r. 1ly de
pend on at 1e-a :s t t.,.,. o. va tl a b} ~s i I Jt sa ttsfi e~ the eon df t rons oi the
prc.bltm then 1v 1,\ > 2.n 1 and 1N1 ' ~ ~ v1.. ' > 2" 1 Let us
no\\ suppose that }1 e.ssent1al}y depends on only one varu1ble
\\ e assurne for ctefinJte-ness that }, ~ .r1 U2 cannot be equal to
-
the PegatLon ol llte var1ahle Since .z E fx~ g]) Under such an
assumpl10n I =:. fl V xt Ohv1ously f (I ;rj_ .rn) ~ 1 II
\\ e had I (0 .xl xrt) :a 0 then f (x, xa xn,) ~ .t1 wh1ch
f"'t..
t<:Lntrad ~ets to the Iact that tne: lunet~on J {z"') es~nt,ally dtpends
on at least two \ arlab{e$ Cc.nseque-ntly 1 (fi :: 2 z8 } ~ 0
au~ \ N J \ > ?!' J \\\\a~ ~5 \{ '\ h~ i \ln~\\~n 11 d~~u~ oo ~\1 'tt..~~a
bles .t 1 zn 1nessent~ilt~ then J, z:.a t but 1n tn:s.s case J ~ 1
t e lt depends n all the varL:ahles z~ ~11- ~nessentlaUy
2 f 2 1 Let A. be a preca m ple te class. l.n P,. \Ve as~ll me th.a t
.x E. K In vtew o( Problem 2 1 16 th.e addltLOO. of the (un.ctLoa.:
to the class K does not lead to a ~ontp[ete system tn P t (s1nce
(K U (z}l :=;: K U (z} ~ z t y) ThE.s JS ut contradrct,on w1tb
the prcco.mpletencss of the class K
22
2 2 j (2) No (5) Yes 2 2 3 (t) zz V 111 (3} xy e y S 1
2 2 4 Carry out Jnduct!on on the number ot occurrence~ of
connectlves 1n the .set {0 f , & V} to the forn1ula W
,...,
2 2 & Let t--- I 1 e the Iunctton J (•71) e~.senl1ally depends
~
on lhe va rutbl~ z 1 Then there C!'·•a\t tuples a::::::: (0 a1i a f\)

~ i"'¥ if~~~./
and rt = ~ 1 e&t ttn) ~uch that 1 {~) =fr J (c: ) 81 nee t• (~

a, an):=J(O a 1 a,.) and /*(0 2
- - a - arl)~7Cl
-
a, an) \\ e have t• (f a~ an.) ~ t• (0 a~ ctn J
~
1 -e x1 1s an rssentlal vas:1.abl~ fi[ the runel\OD 1• {zn)

2 2. 8 ( t} 0 (2) 2.n 1 ( 3) t 22! {i) Y~ s ( ~} No ( 5) '\ es
2 Z.10 U ~a the decompo~HtJon I(;»)~~/ (0 z 2 .r11 ) V Zt/ (l
z2 z~)
CHAPTER TWO 30:i
2.2.11. If 1 (x 11 ) is a self-dual function, then it assumes opposite

values on any two opposite tuples of variables. Consequently, the
~
number of tuples on which the self-dual function I (x") assumes

the value 1 is equal to the number of pairs of opposite tuples (of
length n), i.e. IN 1 I = 211 - 1 •
Il-l
2.2.13. )1, (-t)i C~2211-1-i, ;;.t.
-
i =O
11
2.2.14. :rEB y EB :::, x EB y EB::: EB 1, m (x, y, z), m (.c, y, z) EB 1,

m (x, y, z) EB x EB y, m (x, y, z) EB x EB y EB 1, m (x, y, z) EB x EB z.
m (x, y, z) EB x EB z EB 1, m (:r, y, z) EB y EB z,_m (x, y, z) EB y EB
z EB 1. It can be easily verified that m (x, y, z) = m( x, y, z) EB
:rEB y, m (x, y, ~) = m (:r, y, z) EB y EB z EB 1, m (x, -y, z) =
m (x, y, z) EB 1.
2.2.15. (1) For n = 3. (2) For 11 =4m + 3, m = 0, 1, 2, ..•.
~ ~
(3) The function is not self-dual for any n;;;;:, 2 since I (0) = I (1 ).
(4) For odd n;;;;:, 3.
2.2.16. Only for odd n.
- -
.
2.2.17. (3) g (y )=y1 EB y 2 EB y 3 $ y~ EB Y., E S, Ill (x!, x 3 , /) E S
5
-
and m (x 2 , x 3 , f) E S if IE S. Substituting the function nt (x 1 ,
- - ~
x 3 , f) = xd EB x 31 EB x 1x 3 in g (y 6 ) for y 1 , the function m (x 2 , :r3 ,
I) = x 21 EB x 3 1 EB x 2 x 3 for y 2 , the variable x 2 for y 3 , and the
variables x 4 and x 5 for y 4 and y 5 respectively, obtain ;;1/ EB x 3 f EB
- - - - -
X1Xa EB x 2f EB x 31 EB x 2 x 3 EB x 2 EB x 4 EB x 5 = x1/ EB x 1 x~ EB x~f ffi
r 2X 3 EB x 4 EB x 5 •
- ~ ~
2.2.18. (1) We have I (x11 ) VI* (.•11 ) =I (x 1 , • • • , .c11 ) &

- -
f (x1 , ••• , x 11 ) = const. This constant cannot be equal to 0 since
;.,
otherwise the relation 1 (x11 )"""' 1 must be satislied. Therefore,
I (xl, ... , x11 ) & I (~, ... , ;;11 ) """'0. Proceeding from the equality
~
INtI = I N 1* I, we conclude that in each of the functions f (x 11 ),

I •"" ~
(x") and f' (x") -1 ,
= I (x x-11 ), the number of zeros is equal
••• ,
to the number of unities, i.e. I Nt I = I N 1* I = I Nl' I = 2'1-l.
Consequently, in view o[ the relation 1 & f' = 0, 1 11 ) = 0 if
~ ~ ~
0.
and only if /' (a 11 ) = 1. Therefore, 1 (x 11 ) = f* (x 11 ).
2.2.19. (3) (x t x)-+ (x EB ;) = 1.
~
?.2.20. Let I (x 11 ) be a non-self-dual function, and let all its n

variables (n;;;:, 3) be essential. From the non-self-duality of I it
~\lows that there exist two opposite tuples~ = (a 1 , . . • , a 11 ) and
I - -
a = (al, ... , a 11 ) on which the function 1 assumes identical val-
~es. Wefirstassumethatlfo") =1=1 (1"), and hence~ =!=0 11 and;: =1=
11
1 • Let us consider the function g (x, y) obtained from j {;11 ) as
a result of substituting x for any variable x·1 such that ai = 0 and
as a result of substituting y for each of th e remaining variables.
20-U63G
300 SOLUTIONS A.NS\VERS .Al'-D l:ilNTS
\~e have g (0 0) =I= g (1 f) and 6 (0 f) = g (1 0) Ohvtously

.f"¥ N ~
g (.r g} ti S L-et us no.w suppose thal f (0 11 ) :::::: I (l n) Slnce t (.zn} ¢.

con~t (bee a use- J t e.ssen1Ial1y depe-nds on at }east three var1abl~s}
Pn) such tba t I (Jl) =# f con}
"' {lllv ,..,
there ~XJSfS a tuple .tJ =,.._, (~1

Starttog \\1 th the tuple ~ l\ e construct a function h (x vl ot;:uch
......
that 1f ~ ::: 0 \\ e bUb:stJ tute :r for the var1able :ff Jn f (xn) and Jf
p - 1 ""'e subst• tu te 11 for r The lunct1on h (x y) sa tishe.s the
fc{lo\\ ~ng r~lat~ons h (0 0) =
~~ (1 1) =/=- h (0 i) Ohv\nus.l'~
Jt (r g) CtSsen ttally depends on two "ar1ables and 15 not self du.al
2 2 21 See the previoUs problem
2 2 22" 1t can he ea.sJ}y seen tba t (see Problem 2 2 14) m (.2..a;
.rF-- z') ~ ~m(a.. 6 V); If m (ct ~ y) =
0 tb~ m {P /' x') -== z
Cc.nstql. E>ntly substllUtlng m (ztt y6 : 1 ) lor :z y and s lll- tbe-
sarne {un~t.~on m {.xr£ y~ £i) v..e cbta'-n -e\ lh~r m ts y 1} tJr cn.e o(
the functLons rn (r y .z) $ x EB y rn (z JJ .z:) $ z (f) z m ();' V
.z) @ y e % If "e have one of the JJ..St three I unctions the !un~tlOD
nl. (.t v t.) c~n be eas~1y constructed For example m (z v m (.r
y t) fl) z e y} EB ;.t $ y - m (x y t) Let Us now prove that the
runctton 1n (.t g :) can be us~d to obt.Q ln a funet1on v- h1eh .ess:en U
ally depends on n (;:z: 4J var1abies tl n IS odd and n pz: 5 v. e ha \ e
~ ~
g~ (z } ==== n (..rl ~~ m (.xa .:r, .r•)} Ed+l (.t"~l 1 ) ~ g~, 1 (zt

~t x~l 1 m (.:r~, 1 ~tf zt 1 1)} It caD ca.s1ly he proved
(by Ul.ducll<ln cf l} tnat 'fa~h ol these Iunt.tlon.s t:s~ntlally dePE:ndQ
on all , t~ arguments Fct an e\en n ~ 2l ~ 4 the eo.nespond1ng
functron can be obtaJned lor example from the funct1on t 21 1 by
.........
put hng .z-2 t J :::: .x 11 \Vh1Jt prov1ng tha f I he funct1on Cn (xn) essen
trally depend$ on all Jt.s arguments 1t .should be borne u1 mmd
lhal tlw m-ed,au Jn (z y l} 1S ~ ua~ tD l only tn the- tupl-es conU11n
1ng at Jea~t two unJtJes
2 2 24 \Ve must take an arb1trary tuple (al CXt an) of
va\u~s of va..r1ables .:t1 z .rn. and compare on 1t the value.s oi
the Le fl and r1 g hand st des of th-e g~ \'en r-eI at, on \\ e sbou ld
nt-
dIS t1 ngu lSh bet ween the f o11 ow 1ng two ca.ses (I) there are at least
tv. 1 .zerog among the values a 1 (t' and a 3 and (2) there are at
{ea~t t" o unltlf'S among a 1 o:: 1 and a 3
2 t '2'5: {\) \n \ne so\utlcn to Pro"b\em Z Z 2.2 1\. "as sbt>'9.ll
hnw the med '-0. n m ( :t y s} can be construct-ed from rn (z ~ z~
~ - -
or m (.r y z.} Further m (z r .t~ :s m. (x x x~ ~ r and hence (sre
Problem 2 2 23(1}) the funct1 on z $ g $ .z can be constructed
from m (s g .;) (or m (.r ; z)) "ie1! dual functtons-; (and ;r) of one
varJable are obv1ously conttuned tn f111 (z y .:) ] (and tn fm (.t y z}l)
It follows I rom P.roblem 2 2 12 that there exzst no Bell dual funttloTr!t
u h t ch. essen tla lly depend an two v :a. tl able~ F1 na ll::,. us 1ng l h.e 'P rc
vJoUs problem we coneJude tbat. 1n order to construct any self dual
function of n ~ ~ var1ables 1t 1s surftcJent to have all gel( dual
funt h.ons or rl ..-- 1 l-:&nables {as l't ell as the functlons m (x v ~\
1
and x E9 y @ J:: t( ,.. = 3.l If. fl ~ 3 \!-.-e ~an. \.lSR. (b..e tes.u.l t 1n
l)tob1C'Jn J. 2 14 (Rtnlark Problem 2 2 14 can be soJ,ed by UsJng
CHAPTER TWO 307
-
the solution of Problem 2.2.24 and the functions x, x ED y ED z
~
and m (x, y, z).) (2) Let I (x") be an arbitrary self-dual function,

and all its variables are essemial (n;;;;. 3). We prove that if n;;;;. 4
and 1 (xu) is not a linear function (i.e. cannot be repre:.ented i~ the
form x 1 ED x., ED ... ED x 11 ED a, where a E (0, 1 }), a nonlmear
self-dual function essentially depending on three variables can be
~
obtained from f (x") by identification of variables. (If I is a linear

self-dual function, a similar statement for it is obvious.) Thus,
let 1 (x"' 11 ) be a nonlinear function essentially depending on 11;;;;. 4
variables. The nonlinearity of the function f implies that in the
Zhegalkin polynomial representing this function there exists at
least one nonlinear term. For the sake of definiteness, let this term
contain the variablex 1 • Then /=:r 1 tp 1 (x!, .. . , .rnl ED tp 0 (.r 2 , ••• , xnl
and cp 1 ==canst. We take the tuple (a~, ... , an) on which cp 1
vanishes. We have I (0, a 2 , • • • , an) = I (1, a 2 , • • • , an>· Let
us consider two cases. First we assume that a! = ... = an -= 1.
Since .r 1 is an essential variable of the function 1, there exist two
~ ~
tuples ~and y adjacent in the first component and such that 1 (~) =I=
-
~
1 (y). Obviously, these two tuples differ from the tuples 1 and
-
(0, 1, ... , 1). We substitute y for those variables x 1 which corre-
- ~
spond to zeros in the tuples 6 and y and z for all the remaining
variables except x 1 • We get the function g (x1 , y, z) satisfying the
-
relations g (0, 0, 1) = a, g (1, 0, 1) = a, and in view of the self-
duality of the function g, g (1, 1, 0) =a and g (0, 1, 0) = a. If
g (1, 1, 1) =a (and g (0, 1, 1) =a), then g (x1 , y, z) = m (x1 , y,
z) ED a. If, however, g (1, 1, 1) =a = g (0, 1, 1), then g (.r1, y, z) =
m (x 1 , y, z) ED a. Let us now consider the case when not all a 2 , ••• , an
are equal to 1. In view of the self-duality of the function f, not
~ ~
~-
1'-.J
-
all a 1 are equal to 0. Let us take the tuples 1 and 6 = (0, 1, ... , 1).
t"V
If f {1) = f (6), the above-mentioned tuples ~ and y differ both

~ ~
-
from 6 and from 1, and the problem is reduced to the first case.
~
If, however, I (1) =1= f (6), the identification of the variables in the
~
function f (x") is carried out proceeding from the tuples (0, a 2 , ••• ,
an) and (1, a 2 , • • • , an). We obtain a self-dual function
~
h (x 1 , y, z) satisfying the relations: h (1) = u, h (0, 1, 1) = u,
-
and h (0, 0, 1) = h (1, 0, 1). If h (0, 0, 1) = a, then h (x1 , y, .::) =
m (.rl, Y, .::) ED a.
If, however, h (0, 0, 1) = U, then h (x1, y, z) =
m (x1, y, z) cr. ED
Since .the superposition operation on a linear function can lead
only to lmear functions, any system :fJ complete inS must contain
a nonlinear self-dual function f ~11 ) from which, as was shown
above, the function of the form m (xa, y~. z'~'), where a, ~. and y
belong to the set {0, 1 }, can be constructed. If 1 (x, x, ... , x) = x,
-
then If (x")l contains m (x, y, .::) (see Problem 2.2.14), and hence
20•
s
~
rI (.r ) J (.see the pre VI PUS problem) Let us now suppose that
1 (s ~ z) z 1 e J preserves 0 (and 1) Then there DlUst be a
=:::$
"""" Jn 9'-' WhJrh doe! not preserve 0 11 / 1 19 a nonhnear

fun.et1on /, {xm)
function I hen 1111 :== s JJ howe' er / 1 1s a linear funcJ1on then
'I f 1} S and {f 11 } 1s a bas1s 1n S (tli-e (unet\on 1 g-eDetat~! only
the func t] ons preserVJ n g 0 \\ hl Je II generates OJl] y linear runc lt ons]
2 2 26 (1) No Jt [annot s1nce f E Sand 1 E{S (2} Yes Jt can
2 2 27 Tbe largest \a:Jue of n 1s 4
2 2 28 There are \11tO such funtllODS zl $ it'2 {B
~ hete c: t {0 1}
a;. e0 z, z,
2 2 29 (I} No (2) \ es (3) \o (4) Yes
23
2 3 2 Let f c~ ) assume oppost te values on any lwo adJacent
./"tP
tuples Let I (0) = o Then t (a) = o 1f the number ll a. ll 1! odd

Jot, ......... - .,..,
and f (a.} ::;:: 0 tf u(J. H lS f!\ en Let lrr (.tn) ~ .rl e z.l ED
"""" ~ rtW
t9
:rtl 6=1 a Th~n Nt == N 10 In v1ew of the un1queness o the repre-
sent-a tton of the Iunc.tJons by polynom1als f ~ l 0 and hence I E L
~
The op-posite Js not true (~e for examp1i! ( (.z 3 } ~ zt ED z 1}

~
2 3 6 2n Solu.tion IA:t J (~~ ~ -- ('lii e e Cn%n e

c ntl
Since f E S \\ e have c1. (s- 1 e i) e Ell c"" (tn E9 1) E& t'n.4t EB
n
1 = f (;"') Hence E i'[ = t 1 e the number of coeffietents r:'t~

~t
1. ~ 1 n Ythlch are equal to unity JS odd The number of veetot!

(.r1 en} With an odd ~ Plght tn 81"4 IS 2n 1 The coe(ficient Cnt-1
can he chose-n arb1trartly Hence the :5tatement
,.., ,.t6to'
2 3 tO If f (z~) E L and f (x 71) essen halJy depends on all 1ts

~
varJables. then I (.rYl) ~ ~~ EB e :rn E9 o where a E {0 1}

H~nce l\e eastly obtaln the necessary cond1tJOD of thr statement
~
Sufficiency Let t E L Tben there ext.st two adJacent tuples a:

,.., ~ """"'
and fS such that f (a) == J (~) (see Problem 2 3 2) Lf: t i be the
number of the coordlnate at llb.tch ~and~ dLO:er Tb.c:n f (a. 1
a: I .a %t a ..-1 a~) fiett t1ousJy depends .on x 1
2 3 11 Let P (l:""11 } he a po!ynom1a[ of degr:~ k ~ 3 \V1thout
any loss of general1ty \\e can assume that an e c A -- .:r,z 2
15 an adden() of the pol)nomral P PuttJng .r~ ::: ..rA•• .=: - Zn
.x,
we obtaan the polynorn.Jal Q 1n \\bzcb A 1s tbe only e c ol rank k
A few cases are pcss~hte here {a) All the elem-entary conJUn.ctloUs
of rank k--- 1 generated br the !et Xlf: appear 1n the poi)~Omial Q
\\ e put ZJt. 1 ~ :rh (b) Tbere ex 1st ~ j (I ~ • < J ~ k) such that
e c z1 z1 1S"1,.. 1 zit and s 1 z J 1zJ+J .rk do no~ appear
1n Q Then \\1!: put :rJ ~ ~J {t} The number of {A - 1) rank tetms
I! equal to k - 1 L~t tlie e e z 1 z:~ do not appear ~n Q
Then we Jdent~fy the varJables z 1 and .rA. In a)J the cases we ohta1n
a pol}nom1al r;f degree k - t
CHAPTER TWO 309
2.3.12. Let us consider the polynomial P representing the

function f. If the degree of the polynomial P is greater than 2, by
identification of variables we can get a polynomial Q of degree 2
from it (see Problem 2.3.11 ). Without any loss of fenerality, we
can assume that x 1x 2 appears in Q. !3Y identi_fying al t~e variables
that differ from x 1 and x 2 , we obtam a nonlmear functiOn of three
variables.
2.3.13. It follows from the solution to Problem 2.3.12 that
by identifying some variables of the function 1. it is possible to
obtain a function of three variables, representing the polynomial of
second degree. It remains for us to prove that if there are exactly
two second-degree terms in this polynomial, this number can be
reduced by unity upon identification of variables. If, for example,
x 1x 2 and x 1x 3 appear in P and x 2 x 3 does not appear in it, we put
XI = X3.
~ ~
2.3.14. If 1 (x") Ef L, there exist i and 1 such that f (x 11 ) =

xix1P 1 ED xiP 2 ED x 1P 3 ED P 4 , where Ps are functions independent
~
of x; and x 1 and P 1 ~ 0. We put i = 1 and j = 2. Let a; = (c:t 3 , • • • ,

a,) be a tuple of the values of the variables x 3 , • • • , x,. such that
~
P 1 (et)= 1. Then rp (x 1 , x 2)= f (x 1 , x 2 , et 3, . . . , c:t,.)= .r 1x 2 ED l (xlx 2),

where l (x1 , x 2 ~ is a linear function. The function rp (x 1 , x 2) becomes
equal to unity on an odd number of tuples. Hence it follows
that the face B"; 3 • • · · • 11 is the required one.
0:3 ••• o:,
~
2.3.18. Let f E L. If f forms a basis in L, then f (0) =I= 0 and

~
I (1) =1= i, and hence f essentially depends on an odd number

of variables. But then it is self-dual and cannot form a basis
in L.
2.3.19. If <D is a system of functions complete in L, there exists
in <D a function ft Ef S, i.e. a function essentially depending on
an even number of variables. Then a constant can be obtained from
it. Let a he such a constant. The system of functions ci> must also
contain the function f 2 Ef T 0 • Substituting the constant a for
the variables of the function / 2 , we get cr.
The system <D also con-
tains a function f 3 essentially depending on more than one variable.
It can be easily seen that [0, 1, f 3 ) = L. Thus, from any system
complete in L we can always isolate a complete subsystem con-
sisting of three functions. Hence the statement.
2.3.20. There are three classes {0}, {1} and {0, 1} consisting
o!llY of constants, two classes {x} and {x, ;: } consisting of func-
tions essentially depending on more than one variable, and four
classes consisting of functions of not more than one variable and
containing constants and functions differing from the constants:
(0, xl! [1., xl. {0, 1, x) and {0, 1, x, ;J, If a class of linear functions
contams a function essentially depending on more than one vari-
able we have one of the following possibilities: the class contains
or does not contain a function of an even number of variables;
the class contains or does not contain a function not preserving 0.
Hence it follows that the classes [x ffi y $ z $ 1], [x ED y ED z),
lx ffi y], [x ED y $ 1], Land only these classes are closed classes of
310 SOLVTJO~S A'\SWERS Al\D Hl'rs
llnear fu.nctl'>:OS "lnth eonta1n a Iunttl&n ~ent1:.l~y depeDdJng

on more than one 'art a hie
f"V
2 3 22 JIJnl .. If I (~ 3 ) sat1sfies the eond1t1on of the problem

and 1S
..,..,.
not se1f dualt then for any a E B 1 ,

-r'fl - ~
:and hence J
1 (e~) =I (cz)
can be spec tfied ~om p le tel y h y -s. pee t f ytng t ts values on th.e t up1-es
of the form (a 1 * a: 1 ~ 0} Beside!, cond1t1on (t) 1mphes that the
lunttton t betGmes ~qual to un1ty exaclly on t9to tuples of th13
f o.rm Tb us there e.J: 1.s t not more than s Jx non se If d ua.l fun ct 1ons
sat l s.I) 1ng condl t1 on~ {1) and (2.} It e.an be eas1l y 'tTlne-d: t ha \ all
these fune t loos are J1near
2. 3.25 For n =F 2, tb ere ex 1.st no such f unc t 1on.s All I une tt on.s
of the form ..r1xt E9 ax1 ffi l3x2 $ y (ct~ fl, y E {0, t }) and only
the.se funct1ons sat1sfy cond1t1ons (I} and (2)
2 3~26 Two {or an odd n and 0 for ~tn even n .2.3.27 In five
ways 2 3.2& I (a 1 *
- -
~ a" 1 an) ~ i 2,..3.29. For an odd n
2. 3 30 Ste Pr{)blem 2 S, i 4 2. 3 31 Se~ Frobl~m '2 3 i. 3 2 3 .. 32...
zg X V II• X l II Z J: Y
2~3 33.. ll1nt. At least one oi \be functlllns / 1 1~ and 13 1s no\
l1near
2.4.
2~4 2. (2) For even n (4) for n ;/::. 4k - t, k ~ 1, 2, •
2 4.3~ (t) For all n~ 2 (3) For n ~ 3kt k = l, 2t 2 4 4.. In
2n-:1 2.n 2n-t
seven ways 2~4 5 (2) 3 X 2 (5) 3 X 2" ' {8) {2. - 2 )/2"
11-1
{t 2~ 0 (t.S) 22 -- l 2. .. 4~ 7. (l) No l.l (.annol !.l.Dt~ x $ y ~ T1
and z-+-y E Tt
2 4 &~ ( 1} If f E T .-~ 1ts. pol )···n:om1al d ~ n \)\, c.nnta 1D. 1 as an
addend Therefore, I can be expressed Jn terms of Z!J and z y m
The equa.l1ly T-o ==== lx V Yt z $ g) follo~s from the preVIOU$ one
and from the fact that x V 11 == zy ffi x ED JJ
2 4.9 Let ell be a basls 1D T tJ Then there exislS a Junction
11 € 4l'T1 By Ident1fY.1ng all 119 varJables~ "'e obtain the con
sta.nt 0 0 also contatns a !unction fa "-'bleb l.S not monoton1c The
ftwct1on fa IS oh' 1ously :nonhnear By td rnttfylng the 1- arutble!*
and substi tut1ng 0 we cao obtatn from ft a (unction t.f 1 (¥ g)
of tbe form .:rt~ tB a. e Ty 0 "E {0 t l (see Problem 2 3 i3} Sznce
........ ~ ,.w
It. ~~ not monotonJc there e.xtst tuples a and ~ .such that " <
~ """""' ""-
p a!ld I (a) > f (~) Let A (resp B) be a li:.et of eoord1nates of the
...... ""
tuple a
(p) which are equal to unzty LE!t us cons1der the function
ffz (~ y:, .1} abu~ned ftom 12 by puttl"n.g \D ,t, ;r 1 =
x l:cr ' E. A
~ i === JJ for ' E B and ~ i ~ z for ' El
B \"\' e have (J' 1 (0 0 0} =
cr. (1 ll 0) ~ Or ~~ (I~ o,0) ~ l 11 10 thls case rrt al au
f unc tton, then e1 tber cr-:a :=::::; ..t $ !I Eil z, vr f.f s
IS a self
=
z u V z.: V y: - -
Jn both casest 1\ f' have rot ~1~ Cftl =
T ~ If ,~ IS not self dual~ \\ e
eousLder the f un-cu on <P 1 {x" y) ==
qt 1 \:t, 11 1 0} \\ e ha' e e.~ t.he.t
is = z S y or '91 ~ "ig
- r
If fila :=; z ED Yt then CF1 t "1' 2 ) -==== To
\Vhen cp 1 :::: z !I a ad cp1 ;/=- • V y, all th~ ob ta 1ned tunc t 1ons 0, 1 cr
and ~' belong to the same closed cJ a.5sl n am r Jy, the class G consJ st •
CHAPTER TWO 311
ing of all functions f such that for any two tuples a"' = (a 1 , • • • , a 11 l
"' "' -
and ~ = (~ 1 • • • • , ~ 11 ) such that I (a) = I (~). there exists i
for which ai = ~~ = 1. In view of the completeness of <1> in T 0 ,
there exists a function fa which does not belong to the class G. The
"' "'
function fa has .....,
the ,...,following
,....,
property:
.....,
there exist tuples a and ~
such that f (a) = f (~) and a; and ~ do not have unif coordinates in
common. Let A (B) be a set o! unit coordinates of the tuple "' a
~
(resp. ~). Let <p 4 (x, y) be a function obtained from fa by substitut-

ing x for x 1 if i E A, y for x 1 if z E B, and 0 for xi if i ~ A U B.
Then <p4 is either x V y or x EEl y. In both cases, we have [xy, <p~] =
T 0 • Thus, the completeness of the system (!J in T 0 implies that it
contains a subsystem containing not more than three functions.
Hence follows the statement.
2.4.10. The solution is similar to that for the previous prob-
lem. l\Iake use of the fact that a complete system contains a non-
monotonic function and one of the functions f 0 , cr E {0, 1 }, which
do not belong to classes G0 consisting of all functions f such that
,...., f"'oJ ,...., """""'
for any two tuples a and ~ with the property I (a) = f (~) = cr,
"' "'
there exists i for which the i-th coordinates of the tuples a and ~
equal to cr. An example of a basis consisting of one function is
the set (xy EEl z EEl zz}.
2.4.11. Hint. Consider the function xy EEl y EEl z.
2.4.12. (3) Let us prove the completeness of the system (D =
{xy v v XZ yz, X EEl y EEl z} in Ton s by induction. The functions
in T 0 n S which depend on not more than two variables can
be obtained by identifying the variables y of the functions in (j).
Let all the functions in T 0 n S, which clepend on less than n (n:::;. 3)
...,
variables, be superpositions of functions in <1>. Let I (x") E T 0 n S.
Then I (x1 , xlt x 3 , x 4 , • • • , x 11 ), f (x 3 , x~, xa, x~, . . . , x 11 ), and
I (x1 , x 2 , x~, x~, ... , x 11 ) depend on less tllan 11 variables. We can
now use the result of solution of Problem 2.2.24.
2.4.13. I (1, 1, 1) = 0 since f ~ T1 ; f (1, 0,1) =- 0 since f ~ M,
and f (1, 0, 1) = 1 (0, 1, 0) = 0 since f E S.
2.4.14. If IE L, there exists k;;:;. 1 such that f is congruent to
xl ffi x2 ffi ..• EEl x 2h +l EEl 1. Iff (f L, by identifying the varJUbles
we can obtain either a function <p of the form xy EEl cr (x EEl
y) ffi 1, cr E {0, 1 }, or xy EEl xz EEl yz EEl 1. In both cases, L n
s£ [<p].
2.~.18. Let us prove, for example, that T 0 n Lis a
precomplete
class m T 0 • Let f E T 0'-.,.L. By identifying the variables in /, we
can obtain a function <p either in the form xy EEl l (x, y) or xy EEl
x: ffi yz ~ l (x, y, z). In both cases, the system {x EEl y, (p} is
complete m T 0 , i.e. [(T 0 n L) U{!}) = T 0 •
2.4.19. Hint. :ry EEl z E T 0 ".. (T1 U L U S) and {1, xy EEl z} is
a system complete in P •.
2.4.21. x 1 ffi x 2 EEl .1:~~ 2.4.22. Hint. See Problem 2.3.:W.
2.4.~3. We p~ove that if I~ (T 0 n T1) US, a constant can
b e. obtamed from tt. Iff E T '-... T or f E T '-... T , then by identi-
0 1 1 0
fymg all the variables of the function 1. we obtain a constant.
3t2 'SOLUTIO 'l'S J.Jrr. S \V ER S A 'l D II INT S
Let f ~ 1$ u r1 SinCe t ti s the functtOJl I essen tia11y depends.

on .morP. than on~ varJab!e and 1 t IS pos.!ihle to 1den ttfY the varJahles
ID order to ohta1n a functiOn \\bleb c~ntial}y depends op two
- - -
v~rJablfs VIZ one oJ lb.e functions zg or z V v In both cases \\ e
ean o-hta1n a con..,tant
25
2 5 2 (I) n ;:;: 2 3 (2) For elen n (3) For odd n
2 5 4 Us~ the e.xpanston from Problem 2 5 3
2 5 .S Frw e funct l ons arnoog " h leh onr ha~ a fit t 1 t 1O:U3 'V at1
able One of tb.e functton.s ts sel[ dual ,.,
2 5 8 {t l l t r.s ncL ttut Cons1d el' f {.~=-} 55$ z _, and the tu :pies
(110} and (001)
-
2 5 t2 :E n functtons
2 a I' \Ve assume that a pr1me 1mphcant I oi the Iunctton
#oJ
f (.zn) contatns a nega tJcn of v.arJahle Let u$ suppose for exam

-
pie that I = x 1 L where L 1s an elemrntary conJunction Let us
~ ,..,
eon.s1der NL - {a: L (a} t) and an 'rh• trary tuple 0 aJ
a, 1n 1tt r.. l n view of l ts monotonzcJ ty f (1 ct1 an) === 1
It 1-cllo~s that :zL and he~t@ L ate ~!mplltan\s t~f the- Iunct'-Otr. I
Thn~ JS Jn cDntrad1t.:tioo to the tact that I u a prtme tmphc.ant
2
2..'>16 Forlr-41+ { k-;- ~~~ ~"~=i -)and lor k ~P-~
2
k-1 n- t }
( 4 ~I~ 4
2 5 18 (1} 1 es there ~l:Jsts (2) No there does not exJst
II tot A contracted d D f :tJ of the JunctiOn J baB a 1engtb eq ua) to
thr~~ Th.~ formula z• dual to ~ ~s a tonltatte-d en f of th1!
funct~on t• The contracted d D r of th~ functlDQ f !S obtaJn.ed
from .2• by ope-n1ng the parenthe.ses and by appl)•ng the rure.s
A A ~ .A A V A === A A V A B :::;::: A Cons• df"t a 11 the casE.s
V. hen the -t..On tra Cted d n f {).f tbe- {unc \~ OD j• has the le-ngth S
2 5 19 All Jov.e.r untttPS of a monotontc lunettoD are
pazr" JSe 1ncomparable Th.r maJJmum number of pa1r\\ Jse 1ncom
parable functJOD$ ID en ts { tn;l ) (see Probl~rn I f 151
n>
f'tW i'VU
2 j A1l the lunctlons f (zn} such that f (a}

21 ,.... ~
1 {or It a
n/2 and I (a) ~ 0 for Jl a Jl <:: {(n --- t)/2J are monotonic
2 5 22 (1) 2 (3) n + 2 (5) 2n --- 2
l tl 23 (I) Let I (znl c tt n s \Ve eon:udet the expanSlOD
~
--
v
1 --- .xJtJ~ .:tn/r\ SJnce I t.S selT~ual we have {fl). = ln vJeW ro
of the monolon1.c~ty of f \\'l have f 0 E \1 j~ E \f f~ V J~ = If
Thus 1f the Iuoct1oo f, JD t/ 1.s spectfied the funet1on f 1s Qpec1
fied as lit eJJ It remaJns for ns to v~r1fy that for n ~ 1 there extsts
a. fnnttlOD I E .AI" 1 such that I v
r• =I= I \Ve can take lor exam .l"tt
ple., the Iu.nel1on J (.t1 z 1 zn. 1) $ t 0 (3} Let I (~} E J/ If

CHAPTER TWO 313
the component Jh
2 is given, the function I is defined by mono-
tonicity on at least 211 - 2 tuples: if (a 3 , • • • , all) is a tuple and

1l0
2 (1, 0, a , • • • , all) = 0, then I (0, 0, a , • • • , all) = 0. If
3 3
flo2 (1, 0, a , • • • , all) = 1, then f (1, 1, ct , ••• , all) = L Thus,

3
~
3
in order to specify 1 (x 11 ) E M it is sufficient to specify the com-

ponents llo
2 , l~i 2 and additionally define the function by mono-
tonicity. Then 1 remains indeterminate on not more than 2n-2
tuples.
· 2.5.25. llle (1) = 1, llle (2) = 2, me (3) = 9, llle (4) = 114.
2.5.26. We have I MJ I= 168 (see Problem 2.5.24) and I SJ 1=
256. The statement can now be proved by induction considering
that I sn I = I sn-l 12, I ill" I < I M'l-1 12• 2.5.27. Four.
~
2.5.28. Let 1 (x11 ) E ill. For each ascending chain Z E B 11 ,

~ ~ ~
either I (a) = 0 for all a E B 11 , or there exists a tuple ~ E Z such

l"oJ ,..., - /"'J ,....,
that I (~) = 1 and 1 (a) = 0 for all a E Z, a < ~. Let a parti-

i~
1
tion be specified B 11
on ( [ : ) ascending chains. Then to spec-
21
- 11
ify the function 1 (x11 ) E ill, it is sufficient to indicate for each
~
chain in the partition whether it contains a vertex a such that

~ ~ ~
I (a) = 1, and if it does, specify the vertex ~ for which f (~) = 1

f"oJ ~ ,....., ,......
and I (a) = 0 for all a E Z such that a < ~- Since the length of
each ascending chain does not exceed n 1, there are not more +
than n + 2 possibilities for each chain.
2.5.29. Let us consider the partition of the cube B" into non-
intersecting ascending chains, considered in the solution of Prob-
lem 1.1.18. This partition has the following properties. (1) Chains
of minimum length have not more than two vertices. (2) For each
two-dimensional face containing three vertices of a chain with
k + 2 vertices, the fourth vertex of the face is contained in the
chain of the partition, which has k vertices. Property (1) implies
that on each chain of minimum length, a monotonic function can
be specified in not more than three ways. Property (2) implies
that if a monotonic (partial) function is defined on all chains
~ontatning k unities, on each chain of length k + 2 this function
1s d~fined by monotonicity everywhere except, perhaps, at two
vert.1~es. Consequently, there exist not more than three ways to
add1ttonally define it on such a chain. Considering that the number
. 1s
of chams . ( ll )
we have [llfn] & 3 ( [>:~?]- )
[n/21
' --= ·
2.5.32, (xl V .r2) (xa V x4) ... (x2n-l V Xm).
2.5.~3. Any function essentially depending on more than
one v~r1able has not more than one lower unity.
2.:1.34. Hint. See solution to Problem 1.1.36(4).
y 2.?.3~. (1) No, it is impossible. (2) No, it is impossible. (3)
es, It 1s. 2.5.44. No, it cannot. 2.5.45. No, it is impossible.
2,5,46. The statement that any function I (; 11 ) E M can be
SOL UTIO"'S AN SWEllS A ~D HINTS
repr-esent-ed 1n terms of sup~rpOS,tlo1ls generated by the fu~t~uns

of the !et tO t zy z V g} can be proved by uldUetJon on n us ng
Problem .2: 5 3 The statement then follows from the fact that
{i :EY :t V y} ~ T1 {U :zy ~ V Jl} ~ T 0 and that the funclton
xg cannot be obta1ned Irom the set {0 t z V v} and the funct1on
:r V lJ cannot be obtatned from (0 f :ry}
2 ~ 48 Any baSi& 1n }t! ton \.a1ns con!;tan ts 9;s \\ ~\1 a.~ a :funet1oo
essenttally dependJng on at least two \arJabJes On the othEr band
let <%> be a h.a s1s l [l .A/ con ta. JnJ ng more than four func t1 ons Then
at l~ast tmee of tbem d~pend on mote than ojle -var~:ab1E!' Among
these functions there extsts a function f 1 whtch 1.S not an ele
mentary conJunCtion Tben hy substitUting constant~ we obtaJn
z V y Tbere .a 1so ex 1sts a funtti on f 1 "h1eh is not an e d Itom
wh,ch we get xy Then {0 t f1 J~l sa subsystem complete In M
2 5 49 See Prob1em 2 5 48 2 5 50 (2} {1 z (g V :) } 2 5 51
See Problem 2 2 24
n
2 .5 53 L.c~ US prole lor ~ u.m JJle that J/ T -6 z.s prei!olll pl.ete
The only funct10n f E A!,T0 lS the constant 1 In JJ n T 0 there
are functton~ 0 zg z V y wh1ch togetb~r w1th t constttute a sys
tem CQmplete 1n M Let there -e:tlst one more (jlosed class Q ~h c.h
1s not con ta Jned 1n any ol the abo,. e c O.hd 1t1 ons Then there ex Jst
ft El. AI n To /, EAt n T1 / 1 ~ 2 U {0 1J and !4 ~ ;!( U
\~ -l \ lll \.h'\<~r. c:.a~ 11 ~ 1.. 11 ~ Q a\\~ -we ($b\a\\\. x y tt~m f3 h.-y
suhst• t ut1ng constants and 1den t1 f ymg var1a hles and z V y from I 4
26
2 6 1 Assume that the converse 1s true and use the completeness
crt terton
2 6 2 (4) Complete (6) Incomplete (7) CJDlplete (8) Ineom
plete
2 G 3 Prove that J ET 0 u rl u L u Af
2 6 4 (1) The s1stem can be etther llleompletP fo.r exa.mp]e
v
~
It (x3) = m (.1\ .:r, z,l e Zt m z~ fs (;a) (xl ~ r.,J z, or

complete for e-xamp1co tl (r)
,_r
= ""
m (.r 1 .r,_ x,) [r; (.z~} = .:r1 V x~.t,
(2) Complete (3) Incomplete
2 6 5 Complete Prove that / 1 E T 1 U L US U ftJ and}, 4 Tt
z 6 6 (1) {a: u e : {x m y) ~ %} {( x v 11 \ (%v 'Y> <:r: e 1.1) ~
z m (x 11 1)} (3) {0 ($ ! (.ry)} _. 'i} {z $ y (z t (.ty)} _.. I'}
{ (.r-.. y) +(g .-.w : ) (.z I (xg)} ........ .t}
_ 2 6 '1 (z ~ u} { z ~ 11 } {x S 11% fB z EB 1 J {z xy} {x z V y J
{ :t :r _... y J {1 z e 11 xg} {1 z m u z v yJ co 1 .% e u al
!l:t) {0 1 ~ -6} !/ e t Z!J} {0 i Z G) y $ I :t V y} {0 1 :t $
II Ell .; m (x 1J z)}
2 6 8 There are Se\ en such bases two of them contain one
f unc t1 on ~a th l hree eon ta 1n two (uut ll on!o ea th and t."" o more
hases conta1n ng three functions each can be Isolated from the
complete system {;ty % V J1 X fij y Z ,.._,. gJ
2 ~ 9 tt) {S) Y~ lL tan (4) No lt cannot
2 G 10 No It does not Let [1 ¢ 1 Then f\ f! AI and c1ther
1. Ef T, or f1 Ef S The functton / 2 can he deleted tn 1he former
tase and t.ht i unt t 1 on f, tn the la l ter cas~ The p ()S;i! b 1l1 t y connected
CHAPTER TWO 315
with the relation / 2 ~ 0 is analyzed in a similar way. Let us now

assume that ft == :1 and f, = 0. If fa It AI, the system {/1 , t~, fa}
is complete. Let fa E 11!. Then the function / 4 must not belong to
the class .M. If / 4 It L, the system {It• / 2 , / 4 } is complete. It re-
mains for us to analyze the case when / 4 E L'-...,AI. Since / 4 (£ S U
AI and f E L, f 4 essentially depends on an even number of varia-
bles (an~ differs from constants 0 and 1). Consequently, either
f 4 ~ T tor f 4 (£ T 0 • Therefore, one of the systems {h, fa, / 4 } or{f 2 , fa, / 4 ,}
is necessarily complete.
2.6.11. (1) P 2 • (2) 11!'-..., (T 0 U L) = 0. (3) Jlf'-..(To n T1) =
{0, 1}. (4) M n T 0 n T1 • (5) T 0 n L. (6) S. (7) L n S. (8) To.
(9) To n T1.
2.6.12. (1) K 1 c K 2 • (2) K 1 ~ K2. (3) K 1 i K 2. (4) K1 ~ K2.
(5) Kt = K 2 • (6) K 1 ~ K 2 • 2.6.13, (2) 56.
~
2.6.14. (1) Iff It T 0 U T1 US, then I=/; const and f (0) = 1,

~
while f (1) = 0. Therefore, j ~ M. If j belonged to L, in view of

the relation f (£ T 0 U T1 , the function f would essentially depend
on an odd number of variables (and have a free term equal to 1).
However, in this case f E S. (2) The number of Sheffer's functions
m the set P 2 (X 11 ) is equal to the difference between the number of
"' 11 ), satisfying the conditions j (0)
all functions j (x "' = 1 and j (1)"' =
0, and the number of self-dual functions subject to the same condi-
. h 211 2 ?11- I
twn. T us, P 2 (X ) contains 2 - - 2~
11 - 1 Sheffer's func-
tions (n ~ 2).
2.6.15. Let f (x"'11 ) be a Sheffer's function (here n ~ 3 and all
"'11
the variables are essential). Since j (x ) (£ S, there exist two
. ,..., ,...., -
oppos1te tuples ex = (ex 1 , • • • , ex 11 ) and ex' = (ex 1 , • • • , ex 11 ) on
-
"' "'
which the function f assumes the identical values j (ex) = f (ex') =
I"Y ,...., "' ,..., ,..., ,...,
a. Obviously, ex =1= 0 and ex =1= 1 (since f (0) = 1 and j (1) = 0).

Let us substitute x for variables xi corresponding to zero compo-
~
nents in the tuple ex and y for the remaining variables. We obtain

a function g (x, y) satisfying the relations g (0, 0) = 1, g (1, 1) = 0,
and .g (0, :1) = g (1, 0) = a. If a = 0, then g (x, y) = x .j, y,
and 1f a= 1, then g (x, y) = x 1 y.
2.6.16. (:1) For n = 4m +
2, m = 0, 1, 2, . . . . (2)-(5) For
no values of n. (6) For all n ~ 2.
2.6.18. Functions which possess the properties mentioned in
the .condition of the problem essentially depend on at least three
vanable!o.
2.6.21. See solution to Problem 2.2.25(2).
2.6.23. xy. x
The function V y, for example, cannot be ob-
tained_from it. Generally, the class [xy] does not contain functions
assummg the value of 1 on more than half the tuples of values
of the variables.
2.6.24. The functions satisfying the requirement in the condi-
~on of the problem essentially depend on at least four variables.
n example of such a function is x 1 EB x 2 ffi x 2x 3 EB x 3 x 4 •
2.6.25. The proof can he carried out by induction on n. The
316
btlsll of • ndul'tian a - 4 (se-e Problem 2 3 12} I ndu.ctlW slt_P

Let us assume that the ~ta tement l.S valld for n .;: ; ;: , 1
~
> 4 and Jet
OS prove 1t for n ~ s + I Let the Iunct10J1 I c~•+ 1 ) ~~sentially
~ epend on all lts arguments and not belong to the class L \\ e
rhlv lit.,/
repres.f.'nt the functJOJl I ID the fo11owlng form g {z'} e x, -+lh (r)

~
SJnce ~, ...1 JS an tssen tJal var1able of the IunetJon f h (.t1 ) ~ 0

Let us const~~.r several po.sslhl1t t~e.s (a) A ftctztJous 'ar1a hle 1s .r~
1n the funttlOl\ g and ;tJ (t * ;) \n the [unettan I~ Tben ~t u; snffict
ent to eu t ii' I .rl (b} r.. ~] t her K nor h con taln .Hell ttOU!t v~rJahles
If It E L we :iJ p p ) the 1nd uct1 ve a~umptJoO to tbe functiOn h
11 h E L but g f L \\e apply the Jnductlle .assumptlon to t
If (g h} c: L tbrn I -== ixJ $ -
$ %a) ~' +1 e r () e. rl.z4 +1 '\1- hM~
c ~ c1 E {0 t 1 and \\ e can (ot elalll iJ te tdeattfY ~, WI tb. .r, +l (c)
Tbe functton h does not contatn fictttJous vartaliles \\hi1e the
funcuon g do~s Let ~~ be one of such var1ables If k E L "e
apply the 1nrlllcll\P assumptloo to h II hE L we 1d~nl1ly z 1 'Wlth
z • ~l. (d) flna1lY., we a-c;sume that g does not contain fictL tHJUS
var1ahles and h does Let z 1 he one ol them I! g E L (and hence
h .¢i t) v. e JdentJfy z 1 y, I th the var1able on \\ b1ch h depend5 ec;sen
t~. al ty (ror l he- sake of d~nn1 ttness we ass-ume that :t t '~ sut.h a "art
a»Je-) 'V.e £>btaJn I £x1. .rt r, .r1 -i 1} ~ r!. Sl e z :s e c e
.t'8 ,.. 1/t (.z 1 Zt ~2 .x,) Obv1ously thJs tunc tlOn IS nonh.near
and z, .%'11 +! are essentJ.al variables ln It Let 1.li2 now cons~der
lne c.a~ v-.ben g E L By 1nduc.t1ve hypoth~~s th-er~ ex!st 1n g tw0
\ ar1ables (~ 1 and .x1) "hose 1dent1fica t1on 1n g Jead.s to a noo1Jnear
functiOn wh~ch rssentJalJy depends on , -- t var1ables Jf the
Iunctton /i does not degenerate to 0 as a result of the 1denttficat1on
of ~ t and 3!J.. then z, + :t ~rna lDS an essen t1al v &tla ble 1n the MW
functJon as v.ell Let us now suppose that the functioll h degener
ates to 0 as a result of the Jdent1ficatJon of .r1 and ~1 Then the two
v.a r1 abl es x" and z J arc essen t 1a1 J o h For de fin1 tene$5 \\ e assume
that z; ~ x:~ and .z1 ::::; z 3 \Ve have h {;:J) = .:r1 z 3/t 1 (x, z,) EFJ
;;~:2ll t (~., z,} e ~,h 3 (x~ z.,} and h 3 ~ h.1 $ /) 1 In tht
functJ on I we put x ~ .r a "" 1 If tn th1s case "e obtatn a function "
"h1ch essentJaJiy depends on s arguments the proof JS completed
(The non1tnearJ ty nf r.p JS obvJ ous Slnee fP (z 1 .t3 Zz ..t4 s ,) ~
g (.r i Xt =.t• :t• :z: ,) 6- L ) Howev-er ~t may happen that !or
il' r = .x, --t 1 a cer I .a 1.D v.a.rJa .D1e 1n rp bPeom~:J :fie! 1JJ ous Such .a varJ
a .ole can only he one of the \ ariahJes r\. and .x 3 (SJ:n.ee 1dent~f) 1ng
thP varJable.s .t" 2 and .r3 JD tp "tle mu.st o.Dta1n a fuoctJon g (.z'J ~~J.
.r2 .:, l,~) whlcn essentta lly dep'!nds on 1 -- 1. vatlablesl
Let ..t':t hE! a fictJ t!OU! varJable of the runellOD ~ Since I Zt (..r~.x,gl m=
z,_gt (B s-8gl $ r,) $ z,.~,g, e Xtg• (i) zJI. e g, $ x!l •I X
f.r •-l":.thJ a;. .t2ht @ ~ 3 h 3 ) wbere e-ach 1unct1on o( g1 g g, h1
li~ h3 d-ep-ends en the "'arlables r,. ~.$ and h, ~ h 1 ~ h und-er
tlle above asc;:umpt1on conl'emJng the fie t 1tJ ous nature oi t~e \'"arJ
able z 1 Jn the function q:- "e have tp -:r= f (.r1 ~~ x 1 .z:1) ~
2'1 {:t,R., -m R4) $ z,.,, ~ g~ E& x•.:tah, 1 .e Kt ~ h 1 g 1 :e h 1 and
,._
g. ~ gfJ .E5 0 Consequent}) I (zT.I) =- Zz (zl @
;1:t~_g, (B :rlg~ e X siJ @ ga @ z. i:lz,h, \V~
x,
put
-+J) (.rshl EB h~ e
~:a=- z ~ •1 'J.lll%
teads to the funct1on w=-= f (-ll za z 1 z, , z, z,) = (.ra. ID
CHAPTDR TWO 317
1) .r2 (.r 3 h1 EB h 2) EB .r1.r3g 3 EB .rig4 ill g 8 ill :r3g7 ill .r 2.r3h 3 • Since
the function f (x1 , x 2, x 2, x 4, ••• , x 3, x 2) = g (x1 , x 2, x 2 , x4, ••• .r5 )
essentially depends on all its arguments, only one of the variables
x 2 and x 3 can be fictitious for the function 'i'· But this is possible
only for hi = 0 (and under some additional conditions). Then
=
11 3 hl and f = :;: 2 (xi EB Xs+I) h 2 ill gs EB X1x3g3 EB. xig4 ill
x~g7 ill x 5 +\x3 h 2 • Smce h =/= 0, for III == 0 the tunctwn h 2 cannot
..degenerate' to 0, i.e. h, =J; 0. Consequently, the function 'P =
x 2 (xi ill 1) h 2 EB x 1x 3g 3 ill. g 8 EB x 1g 4 EB x 3g1 ill x 2x 3h 2 essentially
depend;; both on x 2 and x 3 even for 11 1 = 0. Thus, if none of the
identifications x 1 = x 3 or x 1 = x 5 +I leads to the reqmred function,
it is sufficient to put x 2 = x 5 +I· This completes the analysis of all
possibilities. The fulfilment of the inductive step is just proved.
2,6,28. (1)-(4). No, they are not true. In part (4) we can take
f (x, y) = x-+ y.
2II I <)71 I
2.6.29. Since IT~ I = IT~ 1 = 2 - , IL I = 2 +1 ,::::; 2- -
1 11
(for11
on 1 •ln l .,n I
11~2), 1Snl=2-- and !Mn1.::;;2- --+2~2-- (for n~2),
the set W is not contained completely in any dass in T0 , T I• S,
Land M.
2.6.30. An identical function is not contained in any basis
in P, (see Problem 2.1.18). Any of the remaining functwns of one
variable and of zero variables among the Boolean f1Ulctions does
not belong to any precomplE>te cla~s in P 1 • At the same time, the
hereditary system of any such function is an empty set, and hence
~
is contained in each precomplete class (in P 2 ). Let now f (.r11 ) be

a function of a certain simple base \0 and n ~ 2 (all variables
are essential). By the definition ol a prime basis, the system ~ =
(Ia"- {!}) U~l (/) is not complete in P 1 , i.e. is contained completely
mat least one precomplete class K. Iff belonged to K, the system
Ia (s [ U {!}) would not be complete.
2.6.31. (1) 1, ;: (2) 0, 1, xy, x V y, .r t y, .r I y, m (.r, y, ::),
Ill (x, y, z), m (x, y, ~). 111 (x, Y.
~. (3) ~. X EB y, .r ffi y ill 1' xy,
X v y, X-+ y, X-+ y, X EB y ill z, 111 (x, y, ::), II! (x, y. ;). (4) X e !!...·
x~ y, xy, XV y, X->- y, X-+ y, X t y, xly, X ffi y ffi ::,X illy ill z,
m (.r, y, z), m (x, y, Z), m (x, y, ;:),
m Tx, y,
~). While solving
problems (2)-(4), it is expedient to use the results formulated in
Problems 2.2.14, 2.2.20 and 2.3.12.
2.6.33. (2) Taking into account Problems 2.2.20, 2.3.12 and
2.5.3?, we c?nclude that any prime basis in P 1 consists only of
functiOns which essentially depend on not more than three vari-
ables. Let us first consider functions which essentially depend on
two arguments and are not Sheffer's !unctions: x ffi y, x- y,
x-.. y, x7 y, ;ry. and x V y. Let us lind out which of them can
be con tamed m a four-element basb in P ... It can be seen that
x Ell Y ~ TI U S U M, x - y ~ 1' 0 U S U if, .r-+ Y ~ To U S U
L U ~~ and x-+ y Et 1'1 US U L U M. Consequently, all these
ful_lchons are unsuitable for our purpose. We assume that there
e:1.1sts a four-element prime basis containing xy. Since xy E (T 0 n
ri n M)"-(L US), the remaining three functions m the basis must
e as follows: It E (T1 0 M)".T 0 , / 2 E (T 0 0 M) "-T1 and /s E
31&
{T 0 n Tl)'-. 'r Oh"V"l'()\l~ly fl ~ i and 12 ~ 0 It f:J dld U'()t htlong

to L tlie S}stem {f1 ft fa.} would be complete Consequ~ntly
fa E L and bence Is == z EB !I ED .z The case w1th the funchon
z V g lS analyzed ln a .SJml1ar way Le' us now find out l\btch
non l1 n tar funtt1 ()n~ that esse-n t la\ ll y \\ ~ p~r.d on thr-e-e at gu m~n\s
can he ton.tatned tn a four eJement pr1n1e basis If t (.l' y %} E
L US tb.en IdeD[JfYJDg variables Jll 1 we can obtaJn a non se!f
dual functton essentJal1y dependJng on two arguments (see Prob
l~m 2 2 20) But 1n lhts ca~~ J (x y z) tanno\ he con la tn~d in a: pr-liD e
four-e le men t bas1s (see the poss Lbtll ttes w1th t be (u n clLOn.s 4{ two
fl rgu men ts cons1 d ered above) l t rema 1ns for us to consJ der tb e
case \\hen I (2' 11 z) E S '.. L II I d T 0 n T1 and hence t ti l/
then for any functlon g ii S the system tJ g} 1s corop}ete tn P 1
Theref.ore 11. e ea. n assume that f E T 0 n r 1 If t 4 At then (aceu.
-
rate to red.es Lgna_tton ol var tables} f ::::::; m (.r y ;:) l f ho\\ ev-er
IE ~~ then I= m (z v z) In the former case f E (T 0 i1 1 1 n
S),(L U..nt) and hence 1n the corre~pondJng pr1me four element
bas1~ {1 I 1t ex1~t s a\ all} t be rem"" 1n 1ng three tun et1ons must sa tl sly
the candLttonsfl E (T1 n Sl,To fl € (ftr n S}'Tl and fa{; {Ton
T1)'S llowever (T1 n S)'-T 11 ~ (Trs n S)"-.T1 = ~ J e there
ex1st ne1ther / 1 nor f 1 w1th the above propert1es Hence the first
versJOD JS Jmposslhle ln the latter subcase I E {To n Tl r· s n
Al)"-L Co.usequ~ntly t~ rema1n1ng three (unc.tLons LU. the his ti
must be l1near (otherwise the funct1on 1 could he de1eted from
the .bas1 s) L1near I unc t1 ons essen tu1 JIy dPpen d1ng on an e'en
number of \ ar1ahles and dtft et1ng Jrom constants cannot belong l o
suc..h a b.a SJS Since each such Iunc t Jon IS con t:+tned ne-1 tber 1n the
set To U S U AI nor 1n tbe set T 1 U S U JI A 1tnear fuuctlon
essentially dependJng on an odd number of var1ahle.s can belong
[O such a b.as1s only 11 Jt preserves 0 and t TbereJore 1t can he
anly the (unetlon % $ y 'G3 t. Th'ls l~aves beh'-nd only eon~tants
CllAFTER THREE
31
St t Cons1d-er two cases .% -:;:. k - i and
(1) ;r +k - 1 ('2}
(7) Consider two cases z > y and ~ < 11 (8) Let :.r <y Tben ...,..z =
k - z'
f -
::r) ~ k - t - v
11 -.. z and (~ s) - (II -- x) :~= k - 1 -- z - (!1 -
=;tv y If however .z > v
"Pre have (..... .:r) -
(y ........ ~> -=-
.-.,tz (9} Consule-r two. ca!es z ~ y and % > y (10)
(11 ) The equalt tJ es arc pro, ed dttee tly h y USJ.n' tlte relat Lons. ~ ~ ===
k -- 1 - i& and ;- ~ z + 1 (12) Con.stder two cases z :::;;:: k - 2
and z i::. k - 2 {J 3) CoDSJder two case.s ~ == k - J and~
(1~) Canstder five cases (a.) :& ~ 11 == k ~ 1 (h) :r = k--- i y =F
+
k - J
lc ~ 1 (e) = 4=- k - i. y =. It - 1 (d} =r
~ ~ y, and (e) z +
*k - 1 y
k - 1 y =f= k - f z > 11 (15) Cons1der lour
+k - l,
cases (a) x = 11 == k - t (h) z ~ k _.. t. lJ ;yh k - f (e) z +
k - t, y = k - t and (d) % ::fo k - t 11 ~ k -- ! (lB) {17) Con
slder three casu (a.} s ~ k - i, (b)~ *
k - 1. z >~ and (e).:~:+
CHAPTER Tll:REE 319
k - 1, x < y. (18) Consider two cases: x = k - 1 and :r =I= k - 1.

(19) Consider two cases: x = 0 and x .=1= 0. (20) Consider three
cases: x = 0, :c = 1 and x > 2. (21) Consider four cases: x i. (22) Consider two cases: x = k - 1
and :r: =1= k - 1. (23) Consider three cases: x = k - 1, x = k - 2,
and x ~ k - 3. If x = i in the last case, there are exactly k - 2 -
i unities among the values of the functions j 0 (x), j 0 (x...:... 1), •.. ,
j 0 (x...:... (k- 3)).
3.1.2. Hint. Relations from Problem 3.1.1 can be useful for
solving this problem. (1) J 1 (x) = J 0 (max (J 0 (x), J 2 (x))).
(2) ~x = max (/ 0 (x), min (x, J 1 (:c))). (3) / 1 (x) = max (1, J 1 (.r!)),
/ 2 (x) = 1 1 (/ 1 (x)), fi (x) = Jo (x) and --; = max (j 0 (x), / 1 (x)).
(4) ; 0 (x) = (x~ - 1) 2 for k = 3 and j 0 (x) = (x~ - 1) 2 for k = 5.
(5) Let xy +x - y2 + 1 = cp (x, y). We have cp (x, x) = x + 1,
cp (x, x + 1) = 0. Further we obtain all the constants and (a)
cp (2, x) = 11 (x) for k = 3, (h) cp (0, x) = 1 - x 2 for k = 5, and
hence we can construct the functions - x 2 , x\ x
- + 4 (= x - 1)
and it (x) = 1 - (x - 1)~. (6) We put xy = 1jJ (x, y). This gives:
(a) ~ x = 1jJ (1, 1jJ (1, 1jJ (x, 1))) for k = 3, (h) ¢ 1 (x) = 1jJ (1jJ (x, 1),
1) ;-,4x = - x, ¢ 2 {x) = 1jJ (1, .r) = x + 1,¢3 (x) = 1jl 2 (1jl 2 (x)) =
x 1 2, ¢~ (x) = ¢ 3 (¢ 3 {.r)) = x + 4 and - x = ¢~ {¢ 1 {x)) = 4 -
x for k = 5. (i) x...:... x = O,j 0 (0) = 1, 3...:... 1 = 2, Jo (x...:... 1)...:...
io (x) =it (x), j 0 (x...:.... 2) -'- j 0 (x-=- 1) = j ~ (x), 1 -'- j 0 (x--'- 2) =
IJ (x), CJll (x) = (3...:... io (.r))...:... lo (x), CJl2 (x)-., cpl {x)-'- h (x), fPa (x) =
-
{(<p~ (x) --'- j 3 (x)) ...:... j 3 (x)) ...:... j 3 (x) = x. (8) - x = max (J 0 (x),
J J (x + 2), 2J1 (:c)). (9) j ~ (x) = (((J 2 ((x-'- 1) ...:...f)-'- 1)--'- 1) -=--1)...:... 1.
(10) /s (x)= J~ (x + 4). (11) h (x) = - Jh-1 (x- 2). (12)Jh-1(x) =
(· •• ((x -'-(- .r))...:... (~ x))·.:.. •.. ...:... (""" x))...:.... (~ x). (13) Jh-~ (x) =
~--------~h~2~t~·-n-lr_s______ ------
(•.. ((k - 1) -'- (x + 2)) ...:... . . . ...:... (x + 2)) ...:... (x + 2). (14)
h-1 times
.!, ..:.. 2X 1 = 0, ~ 0 = k - 1, (- x) ...:... 2 (k- 1) = j 0 (x). (15)
x = (- ({~ x)-'- 1))-'- Jh-l (x), and /h.-1 (x) can be repre-
sented by a formula generated by the set {-x, x-:- y} (see 12)).
3.1.3. If a. and k are mutually prime, there are no numbers
among a., 2a., . . . , ia., . . . , (k- 1) a. which are multiple to k
or mod k comparable. Consequently, the division of these numbers
by k give.s different remainders. Since we ~ave .k - 1 given numbers,
there ex1sts among them a number wh1ch 1s mod k comparable
a..
with L Let this number he l 0 Then we have (:r+a.)+ . . . +a.=
lo times
x + 1. To complete the proof, we note that J; (x) = Jh-1 (x +
k - 1 - l')) t l -- 0 ' 1' • • • ' k - -·
')
3.1.4. (1) Take y = m - 1. (2) Consider the function 1 -=- 2x.
3.1.5. In this case, we can use the relation x -, y = min (k -
1, Y - .:: + k - 1) with ordinary (and not mod k) addition and
subtractiOn in y - x +
k - 1. By induction in i, we can prove
that hi (x) = min (k- 1, i (k- 1 - x)), i = 1, 2, .•• , k - 1.
Indeed, h.l (x) =- x = k - 1 - x = min (k - i, k - 1 - x).
Further, 1f 11;_1 (x) =
min (k - 1, (i - 1) (k - i - x)), 1;;::. 2,
then h; (x) = min (k - i, 11;_1 (x) - x +
k - 1) = min (k - 1,
120
mtn (k -- f I --- z)) -r (k--- f -- r}} = m1n (k -- i

(t ---- 1) (k -
(k ~ 1} + ~k - l -- ;r)
:e (k - t - ~)) - m•n (J.. - 1 ~ (k-
1 - ~H Consequ:entl): h;., 1 {;r) = m1n (k - 1 (k .--... 1} \k ---
0 lf X~ A- 1
t - .r)) - { A. _ 1
:r -!= J.. ~ t
1 r
and henc:l' - i it. 1 (.x) -
J h. l (.t)
J J 6 Ohv1ousl} a compatJsOn of x 2 ~ r~ (mod l) and .r' s
z-' t rood J..) rannfJt be cDrtl~d out 1d enltt~~l) fDr an) k ~ 3 SJnce
(nr ~ :=:: k -- t \\ -e have r :;: : : (k ......._ t)' .:::::: t .r 3 :;;:::: (k - t )' !::::;
k --- 1 and x~ == (k -- f) • =- i Lrt us no 'It c Jnstder the nla t1 on

~2 s z-' (mod k) \\ e put z = 2 Th1s g1 ves 4 ~ I 0 (mod k) Tr)
~ng th~ \a}ue.s oJ k from 3 to t 5 "c see tha' tlu~ comparJson can
be- carru:d out onl) {ot k = 3 4 6 and l2 1{en-c~ Ior each of th~e
'a 1ues of k "e must flnd out \l! heth.cr the com pattson ~~ ~ L-'
(m d k) ~~ 'alHt for any \a lues of z Jn F ~ It turns out tba t for an)
k ~ 3 4 6 and 12 the compar~son .z 2 ~ .x 4 {mod k) ~~ fu]fi)led Jdrn
lJca}] y 1n E, For exam p1e for k = 6 "e }la l- e
x2 -= (u f 2 3 4 5) s: .a
4 1 0 I 4 3
3 t 7 t -nr k = l 5. and 7 tne nurober of d ~He-rent (\lncttons of
the gJ \ en form t.s k - 1 \\ h1la I o.r k -- 4 6 8 9 1[) J t 1s eq u.a I to 3
2 4 7 4 respechv[\]y For examp1~ Ior 1., ~ 4 ~e ha\~
:& === (01 2 3} z 2 = (() I 2 3 ) x., ~ {u t 2 3
0 1. 2. 1. 0 1 0 t/ 0 t H 3
antl x'l ::=. x2 .z1f+1 ==== (l ~ t) x'
3 1 S ~ 1 {x) - J If. 1 {:t) +
J 1 1 {x) {k -- 1 \etm~} +
J 1 t:t) =
J1t l ( ~ --- ~ - i) 0 ~ i ~ k ~ 2 1{ 1 tz) E Pit
1
then
g (x} ~ g (O~J"G (.:c) +
g (t} J1 (z)...L. + g (k -- t) J,. 1 (t) +
It rem a tns for us to take Inlo .account l1" (.z) = It (r} + +
11 {.t) (l terms)
3 1 9 L-Pt us constder the functJons f (x) -- c (c ~ l1 2})
It~ +
t) f (~ +
2) and 1 (1~ 11 (;z}} wh-e-re h1 ,_ (z) ~ :l 1 ~2} - +}
1t (.t} Each of them .assumes the same number of values as the
fun~ ll r>n 1{=t} has Thtt ef orP ~e tan a. ssurn e t. hat /1 t.r:} =:;
u t cz
(0 0
2)
and / 2 (.rl - ( ~ !1
;
1
) -u.here ti b1 .b 2 E {1 2.} TMn g (z} =
0 1 2 )
(0 b ~ b:z lt b1 t th~n bt. - 2 aud a _L b~. == 2 and
1 4
hence n ::;::: 0 \\ h,('.h. c.onttadtcts to the con\hllon oi the problem
(since oQ E {1 2}) If ho\\ ever br. --- 2 "e have bZJ. ::: t and a +
bz ----- t CDnsequentl) a =-= 0 and v. c agazn 8trJ \e a' a con tradJo
tJon
3 t 10 S1nre the lun:t\l on J (x) ~ c {t E 11 2 }) assume~ the
same D u.mher or 'a.l ues as the Iun.ct1on f (z) d~es the coeffte.Lent a,
can assume any 'alue ~n £ 3 \\ e can assume that ae ;e: 0 Then
/(.r} :::: a pi~ C-oll-o~ 1.n9.: three
0 2
+ a 1;t ;:o= {
0 a0
+t a aq ..,.-2a 1
) The
-ra.h(-s are assumed to he favourable
1
(a) a., + a = au+ 2aJ,
1
CHAPTER THREE 321
(b) a0 + a 1 = 0, hut a 0 + 2a 1 =f=. 0, (c) a 0 +

a 1 =I= 0, but a 0 +
2a1 = 0. In case (a), we have a 1 =
0, and a0 is arbitrary. In case
(b): a 1 =f=. 0, a 0 = - a 1 = 2a 1 • Case (c): a 1 = a 0 and 2a1 =f=. 0
(hence a 1 =f=. 0). ~nswer: a 2 is arh!trary and (1) if a 1 = 0, t~en ~ 0
is arbitrary, (2) tf a 1 =f=. 0, then etther a 0 = a 1 or a 0 = 2a 1 (t.e. tf
a 1 =f=. 0, then a 0 =f=. 0).
x
3.1.11. (1) =max (min (1, / 0 (x)), 1 1 (x)) = io (x) 2h(x). +
(2)~ x =max (/ 0 (x), min (2, / 1 (x)), min (1, / 2 (x))) = 3j 0 (x) +
2iJ (x) + j 2 (x). (3) - io (x) = J o (x) = 4io (x). (4) 2J 1 (x) =
min (4, / 1 (x)) = 4it (x). (5) J 2 (x2 +
x) = max (/1 (x), J 3 (x)) =
4it (x) + 4j 3 (x). (6) (~ x) 2 +
x = max (min (1, J 0 (x)), min {1,
/ 1 (x)), / 2 (.~;), /3 (x)) = io (x) +
h (.~;) 3j 2 (x) 3j 3 (x). + +
(7) 3j 1 (xl - j 3 (x) = max (J 1 (x), J 3 (.r)) = 3h (x) 3j3 (x}. +
(8) x + 2y =max (min (1, / 1 (x), / 0 (y)), min (/ 2 (x), / 0 (y)),
min (I 0 (x}, J 1 (y)), min (1, J 2 (x), J 1 (y)), min (1, J 0 (x), J 2 (y)),
min (/1 (x), / 2 (y))) =it (x} Jo (y) +
2j2 (x) Jo (y) 2io (.x) it (y)+ +
j 2 (x) it (y) + Jo (x) j 2 (y) +2iJ (x) j 2 (y). (9) max (x, y) = max
(min (1, / 1 (x}, / 0 (y)), min (1, / 1 (x), / 1 (y)), min (1, / 0 (x),
/ 1 (y)), min (/ 0 (x), J 2 (y)), min (/ 1 (x), J 2 (y)), min (/ 2 (x), J 2 (y)),
min (/ 2 (x), / 1 (y)), min (J'l (.:z:), lo (y)) =it (.:z:)io(y) it(.:z:)it(y) + + ·
io (x) h (y) + 2Jo (x) h (y) +
2ft (.:z:} i2 (y) +
2j2 (.:z:) i2 (y) +
2j 2 (.:z:) h (y) + 2h (.~;} io (y). (10) x-'-- y 2 = max (min (1, / 1 (x),
/ 0 (y)), min (1, / 2 (x), / 1 (y)}, min (1, / 2 (.x), 1 2 (y)), min (/ 2 (x), ·
lo (y))) =it (x) io (y) +i2 (x) it (y) h (x) i2 (y) + 2j2 (x) io (y). +
(11) ry =max (min (1, /) (x), /1 (y)), min (1, J 2 (x), /1 (y)),
min (J 1 (x), J 2 (y)), min (J 2 (.x), J 2 (y))) = h (x) it (y)
- i 2 (x) +
+
it (y) 2it (x) j 2 (y) +2j 2 (x) j 2 (y). (12) xy = max (min (1, / 1 (x),
1 0 (y)), min (1, / 3 (x), / 2 (y)), min (2, / 2 (.x), / 0 (y)), min (2, / 1 (.:z:),
1 1 (y)), min (2, / 3 (x), / 1 (y)), min (2, J 2 (x), J 2 (y)), min (/ 3 (.x),
lo (y)), min (/1 (.x), J 2 (y))) =it (x) io (y} ia (x) i2 (y) + +
2j2 (x) io (y) + 2h (x) it (y) +2j3 (x) it (y) +
2j 2 (x) i 2 (y) +
ia (x) io (y) + h (x) i2 (y).
,.., ,.., '""'11
3.1.12. Putting xn = (a1 , a 2 , • • • , an) = an, we get i (a ) =
min {max U (a),"' - J cr (a ), ,.., J cr (a ,), • • • , - J cr (a )) }. Let
r..~ t
1 a 2 n 11
a
us prove that the right-hand side of this relation coincides with
the left-hand side. If at least one ai =f=. ai, for example, for i =
'""'
i 0 , then max U (a),,.., J~ (a 1 ,), • • • , "'"'l~ (a. ), - Ia· (a.),
v1 ""i _1 lo-t to lo
"'
•
- Ia.
'o•l
(ai
o
+ 1), •.• , -
n
Ia
(an))= max (f (a),- Jcr (a 1), .•• ,
t
"' 1 a. -t(ai.-1),- 0,- /cr. (ai +1), · · ·• -Jcr (an))= k - 1.

'o 1 0 +1 ° n
- - . rJ
If, however, a= a, then max (/(a),- J cr, (a 1}, •• • • , -J cr (an))=
11
~ ~
max (f(a), - (k - 1), ••. , - (k- 1)) = 1 (a), and hence min
~
{max (/ (a), "'"' J a (a 1 ),

'
••• , ,.., J cr
n
(an))} = min (k - 1, . . .,
~ "'
k- 1, I (a), k - 1, ... , k - 1) = 1 (a).
:! l-Ot\36
322 SOLUTIONS .ANSWERS .AND HINTS
3 2
J 2 J (1) In order to s1mplLly the nolal1on we put T ({0
.l}) T and U ( {0 1 } { 2}) - 1J (a) ~ z E T and E V. (b)
(b) 11 (.t} E T n 1J {c) J, (.z) E T l1 fJ (d) z - g E T and t1. fJ
(e) ~ + y f T and E U (f) m1n (x g) E T 'Y n
(2) \Ve put T ( {1 3}) == T U ( {0 t} {2} {3}) = fl 1 fJ ( {0
3} {1 2}} = lJJ • (a) %'
aDd E 11' n 'II I (c) J 0 (%) E£ T
T E ti 1 s.nd 'E: 11 2 (b} ~ :t t T
E 1JI and E 2l' (d) z 2y E T +
I! ft 1 and Etl, (e) max (z g) E T fJ.l and. Efl 2 (f) ~~~~ T n e
Q1J 1 and. E71 t
J 2 2 (1) The subset {2} and the part1 t1on {0} U { t 2} are
suLtable (2) i - {0 2} and D = {{0 2} {f}} (3) ~ = {1} and
D ~ { {0} {I 2}} (~) if ~ {2} There IS no su1table parhtlon
5) ~ {0} and D = {{0 t } {2} {3}} (6) ~ = {0} and D --
=t
{ ) and D = {{0 {2 3)} (9) ~ = {4 and D ~ ( {0 1

1 4}
l 4} (3} ) (tO) 1S - {0 2) and D == {{0 J 4} ( 1 3 5}}
(it}~- 0 i}and D={(O 1} (2 k-ill (12)~=
i 2 k - ~}} for k ~ 4
3 2 3. {1} l£" ~ ((l ct ~ = k.~
any (unctiOn 1n P~ pt.cset\'es.
this set (for C! = '- 1t IS assumed by d-efint tLon that f (0
0) = 0) Let ~ *
0 and ~ ;:fo F 11 Then there ex1sts a E E~t
%Uth \hal a E " Ohvumsty 1 (z) !!3!!1 Q E r (b) Hence- T (~} +
P~t
(2) JIJDI It should be .first proved that different classes 1 (~)
corre.s pon-' "to iL I i1. erent proper subset3 in r h 'l'ne n umncr ot a~\
non empty subsets u1 E11 ts 2k - i
(3) If ~ = 0 then t (T (~ l) l kh Let i ~ m -- t i l ~
;::::::J
k - f Then 1(1 ("))n t ,nm k" m {s nee the \alues from~

should he assumed on ~' ;;:;$ ~ x
.... --
x"
while on the remaining
._r
-n t I nc•
ktt -- m luple! the values of the functton.s can he arb1tra.ry)
3 2 4 (I) Jf D == { {0} {I } {k -- t } } i e 1 k or
D :=.-. { { 0 1 k - t } } 1 e • = 1 then any fun ctt on 1n P11.
preserve! thus part 1ti on Let 1 ::1= t and s =:p k In the part 111 on
D ~ (~ 1 ~.} we lake a subset ' , s.uth th..tt ~ ~ t l ~ 2
and anolber subset ~ 1 dtffer1ng from ~ t Let a 1 a 1 t;: ~" and
a 1 for ~ = 61
b E ~ i ~I. us con~ui-e-r th'e {unctian f (:t) { b f ......L
1 or z -r- at
Obv1ously I (a 2) = b and henee f does not pre~erve the partLtlon D
and t t! fJ (D)
(2) llinl \Ve must first prove that rot • *
t and • =1:= k dl.fferenL
classes fJ (D) correspond to dz.Oerent part1t1ons of E~t, Fork = 3
the ~t 1 rr.s th-e fo.llow1ng pa~t~\~ons D 1 -== [ {0} {1} {2.}) D • =
{0 t) (2}} Da = f{O 2} {I)) Dt = {{0} {I 2}} D1 =
{0 1 l)} 11 (D 1) = U (D 1 ) = P1 The ren1a1nrng three classes
(Da} 1J (D 1) and 'It (D,.) d1fter trom one another and from Pa
Con s~q uen ll y there exist four d' ff er.en t c} asses oi the type 'II (D)
in P, (J f \\ e a}so take 1nto accouc t tho en tire sc t P:~ o.s v. ell) For
A. =- 4 the set 1:11 bas lbQ follow1ng partlt1on.s Dl = llD) {l)
CHAPTER THREE 323
•
{2}, {3}}, D 2 = {{0, 1}, {2}, {3}}, D 3 = {{0, 2}, {1}, {3}}, D 4 =
{{0, 3}, {1}, {2}}, D 5 = {{1, 2}, {0}, {3}}, Dl!. = {{1, 3}, {0},
{2}}, D7 ={{2, 3}, {0}, {1}}, DR= {{0, 1}, {2, 3J}, D9 = {{0, 2},
{1, 3}}, D10 = {{0, 3}, {1, 2}}, Dn = {{0}, {1, 2, 3}}, D 12 =
{{1}, {0, 2, 3}}, D 13 = {{2}, {Or 1, 3}}; D14 = {{3}, {0, 1, 2}},
D 1• = {{0, 1, 2, 3} }. Consequently, there are 14 different classes
ortype 'lt (D) in P~. In P 5 , there are 51 classes of the type
'lt (D).
(3) Let (i1 , i 2 , • • • , in) be a tuple of numbers in the set {1,
2, ... , s }. The number of such tuples is s". Arrange them in the
alphabetic order and assign them numbers from 1 to s". The tuple
(1, 1, ... , 1, 1) has the number 1, the tuple (1, 1, ... , 1, 2) the
number 2, the tuple (1, 1, ... , 1, 2, 1) the numbers+ 1, and
so on. If a tuple (i1 , i 2 , • • • , in) has a number m, the number
I ~',, I I ~'•• I ... I ~.1 ll I is denoted by dm· The number of func-
tions in the set Pk (X"), which preserve the partition D = {~ 1 ,
~ 2 , • • • , ~ 3 } is equal to ~ I ~j, ld' I ~i. ld• . . .
(j,, i •. ... , i,)
I ~i 1d,, where r = s11 and the sum is taken over all possible
r
tuples of length s71 , consisting of numbers from the set {1, 2, ... s}.
3.2.5. We denote the set ~ by ~ 1 and El< I ~ by ~ 2 • tet
I \S I = land, as in Problem 3.2.4(3), let dm stand for the number
I~; I· .. \~- I, where (i 1 , • • • , in) is the tuple of length n,
I In
consisting of numbers belonging to the set {1, 2}, where m is the
number of the tuple in the alphabetic order. Obviously, d1 = 171 •
(1) Since T ('#;)"-.'lt (D) = T ('#;)"-.(T (~) n 'lt (D)), the num-
ber of functions in P11. (X") contained in the set T (~) -.. . ._ 'lt (D),
i.e. I(T (\5)"-'lt (D))<"> I is l(T (~))(11 > I - I (T (\5) n 'lt (D))<"> I=
ln .n n "\1 d d d 1n 11 n n
l · k"" -t - LJ I ~ 1· I '· I \31· I • . . . I ~ 1· I r = l k -z -
• 1 .z r
}1=1
(j,, ... , ir)
-
.
~ .
I \3i 1d• ... I ~i ldr, where r
• r
= 271 , and the last sum
( Jh "~' Jr)
is taken over all such tuples of length 211 - 1, which consist of
numbers belonging to the set {1, 2}. The above formula is valid
for n;;, 1. If n = 0, then 1(T (\5)"-.'lt (D))(O) 1 = 0.
(2) For n >
1, we have I(<JJ (D)"-.T (\5))< 11 >I= I(<JJ (D))< 11>1-
I(T(~)n<JJ(D))<n>l= ~ l~itldll~hld2 ... l~irldr_
Up i2' .•• ,i,>
.lj l'iSJ/ti'Sizldz ... l~jzld'= ~ l'SJ/ll'SJ:/z ...
. Jl=1 . j 1 =2
(
'··
.. .•.• J,) (J" •l• ..... }. r )
l "'Cfr ld r=(k-l) 1n "\1

LJ I~J I
d2
...
•
l~J Id r (here, as before,
. 2 r
(J . . . . . . i,>
r ;2n and hE{1,2}, i=2, ... , r). For n=O, I(<Ji(D)"-.T(~))<0 >1=
102\ =k-l.
(3) I(T(~) U 'lt(D))<n>l = l(T(~))<11 'l+ I(<Ji(D)"-T(~))<n>l =
21•
324. SOLUTIONS ANSw£RS A:hD HINTS
k 1.
R.nd /1 (t} =
ao=2
+ +
4o {11 «t ....J....- «:a= 0
.ao+2al = 0
a0 +3a 1 +a 2 +aa, =-= 0
SolVJDg th1s S):stem we obtain (a) d~ = 2 a1 t a 1 = 0 4-,= {
(h) 4,o o::::.: 2 a 1 == 3 Gz ~ 0 a 1 ~ ;:;, {e) .ilo == 2 a 1 = i at:::;::: 2
«s = 3 (d) ao = 2 al == 3 a:a 2 a.J .:::::: t Consequently 2j .p{z):=
2 + 3: +
2z' + z'
x' = 2 + 3.r + 3z~ ~ 2 +:& + 2.1:= +
3.zs 2 3x+ +
(2} If the !unction 1 (z) 1n P • 1S such tb.a t f (E 1 ) s:;: {0 2J 1t can
CHAPTER THREE 325
3
be represented in the form I (x) = Lj bi2j i (x), where bi E
i=O
{0 f} i = 0 1, 2, 3. Then we must use the result of Problem
3.2.8(1). If, h~wever, I (E~)s {1, 3 }, we can consider the func-
tion g (x) = f (x) - 1.
(3) Assuming that the function f (x, y) can he represented by
a mod 4 polynomial, we have I (x, y) = a 0 o + a 01 y+ +
a 02 y2
+
a 03 y3 +
a 1 oX + ... +
a 11xy a 32 x 3y2+ a 3 ~x3y3. Considering this
relation for the pairs (0, 0), (0, 2), (2, 0) and (2, 2), we get a system
which has no solution:
aoo = 2,.
a00 + 2a 01 = 0,
a 00 + 2a 10 = 0,
a 00 + 2a 01 + 2a10 = 0.
3.2.9. Hint. It is sufficient to prove that the functions I (x) -
f (x) and 1 (x) - j3 (x) assume the values only in the set {0, 2} and
then to make use of the result of Problem 3.2.8(2).
3.2.10. If the function f (x) in P 4 can be presented by a mod
4 polynomial, it can he written in the form a 0 + a1x +
a 2x2 +
a3 x 3 , where a; E E 4 , t = 0, 1, 2, 3 (see Problem 3.2.8(1)). But
one and the same function can he represented by different poly-
nomials. Besides, each such function is represented by four differ-
ent polynomials (see, for example, the solution to Problem 3.2.8(1)).
The latter fact can he easily proved by considering that the follow-
ing relations are valid in P~: 2x3 = 2x2 = 2x and 3x3 = 2x + x 3.
The number of corresponding functions is equal to 64.
3.2.11. (1) The functionsx, x 2 and x3 in P 6 should be compared.
(2) Since 3x2 = 3x (mod 6), each function that can be repre-
sented by a mod 6 polynomial is represented by two different
polynomials. Therefore, the number of functions in P 6 which can
he represented by mod 6 polynomials is 63/2 = 108.
{3) The necessary and sufficient conditions for representing
the function (a + bJo (x))2 by a mod 6 polynomial consists in that
the function (2a + b) bJo (x) must he represented by a mod 6
polynomial. It remai11s for us to find the values c E E 6 for
w~ich the function cj 0 (x) can be represented by a mod 6 polyno-
!Dial, and after determining from each such value c the correspond-
mg pairs of values (a, b) choose the pairs for which the function
a+ bj 0 (x) cannot he represented by a mod 6 polynomial. An-
swer: either b = 2 and a = 2,5, or b = 4 and a = ·1,4.
3.2.12. (1), (2), {4) No, it cannot. (3) and {5) Yes, it can .
. 3.2.13. (1) T ({0, k - 1}). (2) T ({1, 2}). (3) 'lJ ({0, 1},
(~)' · · ., k- 1}). (4) T ({0, 1 }). (5) T ({0, k - 1}). (6) T ({0}).
T({i}). (8) T({k-1}). (9) T({1, k-1}). (10) T({O, 2}).
m)) T ({k- 2}). (12) T ({0, 1}). (13) 'lJ ({0}, {1, 2, ... , k - 1}).
k T}({1, 2, ... , k -1}). (15) T ({1, k - 2}). (16) CfJ ({0,
. - 1 , {1, ... , k - 2}). (17) T ({0, 1}).
3.2.14. (1) The set {j 0 (x), x + y} is such a subsystem (see
Problem 3.!.8 aud use the Slupecki criterion). The completeness
326 SOLUTIONS .ANSWEltS AND HINTS
of thzs s.rstem .can he proved Jn a different way \Ve have 1, (r) +

Jo (/o (~))= t s + t ; ;- + t - z 2+ (.t + (k- 2)) +
t = X + {k 1) J l {.:r) :=;;: I 0 ( ~ -- t) j 0 i z + (k -- ~))
:;:$ i :::;::: 1
2 k - 1 ,, (/Q (~) + /Q (y)) ~ Jo 0 (x v) := , , (x) /o (y}
it m (x 1/) - lrJ 0 (.r- l 11 --- m) l =
0 1 k - 1 and m. :::::
0 1 k -... 1 J 11 ) m II} ::::: 1{ m (% Y}
( l: + +
1i m (:r. Y}
(s terms) s = i 2 k- 1 k llf (x y) lS an arbltrary func
tJ on of t\\ o argum~nts then I (r v) = )J f (l m) 1{<~ m} (.t y)
il m)
wh-ere the .summat1on IS carried out over a. II pairs (l m) E E~t. X Ek
It remains lor us to take Jnto account the fact that the S).Stem of
all functtons 1n Pk wbtch depend on two arguments IS compJete
ln P~t
(2) {0 :t + y sy} c. T {{0 }) {,., (:~:} 11 (~) JJ:i 1 {.t}} c
T({O 1}) (1 2 k-l}cT({1 2 k~1})
3 2 l5 (l) At"' lt) c T (E~)'- 't)) ,: t = k- 2
(2}11 1 "-. {0 k - t Ja,(x) J._ 1 (z)} H)nl Jt (It (.r})---- 0 Jl (t}=
k - t (see also Problems 3 t 1. (t8) and (t 9))
3 2 i 6 (i} Tb~ subsyst-em 1s t.omplete 0 ::: I 0 (1) 2 J 0 (lJ} =
J (.r) .==: I 0 (max (J il (x) J t (r))) (2) The subs) stem JS not com
pfete Consider the partJtton D = f {0 t) {2}) (3) The subs) stem
JS not complete and preserves the part1t1on { {0} {1} {2 .3}}
(4) The .subsystem 1s not eo m p 1cte .and preserves the part1 t1 c n
{{0} {1. 2} {3)}
3 2 17 (1) The comp1eteness of the g•ven system 1n Sh... can he
proved as foU.ows By JnductJ()n on i U;,> 1) \\ e prove that any
functiOn g ID S,_ sa t.zsfy1ng the cond lliOn f (.z) .!$ Z for Z > i IS
genera ted hy iunctulns 1n the set {h 01 (s) k 01 {x) h 01 (x)}
Putttng then l == k ~ 1 we ohta1n th.e -requtrea. result. The 1nduc
t1 ve .step IS substantiated as foUow.s Let g (xJ ~ r for x > 1 + 1
and g 1& + 1) :;:: 1 g (I) ~ t + 1 {here J 4:- i + l and 0 ~ 1 ~ ~)
Let us take a /unction h (.r) C01nt1d1ng With g (z) everywhere e~cept
+
at s :::: l and :r :::::::; 1 f At these two values of the argument the
(unctton h ts h (l) :::: 1 and h {t +1) == l + 1 Obv1ously the
lunctilon h. (x) has aJread y be-en constructed (In v1ew of the 1nduc
t 1 ve bypotbes1s) 1\ can be eas1l y sren tba t g (.r) =
h (hou 1> (hoi (ho<.t +1) (x))})
(2) SinCe h 01 (h 1 {f +l) (hl)l (.z:)}) ~ h 0 c1 1) (.t) f 1 2
k. - 2 the gt ven s} stem generates eat h func tlon 1u Proh lem 3 2 t 7( t)
(3) Constder that hi (f ..,. 1) (z
~r=:O 1. 2 k-3
+ (k - 1)) + i == h{ 1 .. 1>( f +::~) (:r)
3 2 18 It Iollo\\s from Problem 3 2 1 (I) that the g ven sys
tem generates the set sh Using the funetiOlJ ~ +
/o (~) "e can
easLlY construct any functLOn not assumtng ex.actly one value
(1n Ett) Then g1ven that we have all the functtons assum111g not
more than t ~alues 11 ~ l ~ k - 2.) we must. show how an arb\
trary functiOn 1n P'l' whtcn does not assume .~ + i values can
be constructed Let t (x} be an arb1 trary IunctJon not a5.5um111g
cnly one v:0-l ue Then tbtte ex1s t exaet ly t ~ D va 1u~ "(}f \he atgu-
rnen t VJ z lD and 11 for winch the values .pf tile !unct1on g (x) are
equal 1 e g (lg) ~ g (l1) The elements of the set
w1ll he denoted hy l" lr.
E,.'
{lo l1}
ltt ) Let us take two funct1ons
KJ (.x) and ' ' (z) 1n the set S11. such tna t g 1 (l ,.} = r r ~ O, 1
CHAPTER THREE 327
k - 1, and ~~ (s) = ls, s = 1, 2, ..• , k - 1, while g 2 (0) =

l E Ek "g
(Ek)• We denote the function x + io (x) by h 0 (x). We
have g (x) = g 2 (h 0 (g~ (x))). Let us now take an arbitrary function
g' (x) not assuming i --r 1 values (1 :::;; i :::;; k - 2). Let one of these
values be 10• We assume that g' (b 0) = g' (b1). Let us consider
the function / 1 (x) coinciding with g' (x) everywhere except at x =
b0 , and / 1 (b 0) = 10• Obviously, the function / 1 (x) has already
been constructed (by the inductive hypothesis). Let a 11 ••• , as
be all different values assumed by the function g' (x) (here s =
k- i - 1;;;;. 1), and Eh "'- {a 1 , • • • , as} = {1 0, 11, ... , li).
We take the function / 2 (x) in the following form:
X if X =/= lo,
/ 2 (x) = { g' (b ) 1'f x = l' •
0 0
It can be easily seen that the function / 2 (x) does not assume only
one value, and hence we know how to construct it. We have g' (x) =
/2 (fi ($)).
3,2.19. (1) x...:... x = 0, J 0 (0) = k - 1, (1 0 (0)) 2 = 1. Further,
we obtain the remaining constants (for k;;;;. 4): (k - 1) ...:... 1 =
k - 2, (k - 2) ...:... 1 = k - 3, .... Then we construct max (x, y)
and min (x, y): x ...:... (x -'-- y) = min (x, y), (k - 1) ...:... x = ,..., x,
- min (- x,,..., y) = max (x, y). Thus, the initial system generates
the Rosser-Turquette system.
(2) (x -'- y) +y = max (x, y), :r: +
(k - 1) + ... + (k - 1)
+
= x 1 (here k - 1 is added k - 1 times). Consequently, we
have constructed the system {x +
1, max (x, y)} known to be
complete beforehand.
(3) If k = 21 +
1 (l;;;;. 1), we proceed as follows: ((x + 2) +
... + + = +
2) 2 x 1 (2 is added l +
1 times); x ...:... (x -'-- y) =
min (x, y), - min (- x, - y) = max (x, y). We have obtained
a Post's system. Let us now suppose that k = 21 (l;;;;. 2). Then we
construct a Rosser-Turquette system: min (x, y) and max (x, y)
are obtained in the same way as for an odd k. Then we construct
the constants x...:... x = 0, 0 + 2 = 2, 2 2 + = 4, ... , (k - 4) +
2 = k - 2, - 0 = k - 1, ,..., 2 = k - 3, •.• , - (k- 2) = 1,
and then the functions 1 1 (x) : ((k - 1) ...:.. x)...:... . , . ..!- x= J 0 (x),
•
k-1 tlmrs
lo (x + k - 2m)= 1 2 m (x), m = 1, 2, ... , l - 1, 1 2 m(- x) =
12m (k- 1 - x) = Jh-x- 2 m (x), m = 0, 1, ... , l - 1.
(4) X...:... X= 0, 1 - 02 = 1, - 1 = k - f, (k- 1) ...:... X =
~ :;. x ...:... (x...:... y) = min (x, y), ,..., min (- x, ,..., y) = max (x, y).
Usmg k - 1, 1 and x...:... y, we obtain the constants k - 2 =
(k -. 1)...:... 1, k - 3 = (k- 2) ...:... 1, .•. , 2 = 3...:.. 1. Then we
obtam ((k - 1) ...:.. x) ...:... . , . ...:... x = .[0 (x), Jh_ 1 (x) = J 0 ( - x),
k-1 times
/lt-2 (x) = 1 0((k- 2)...:... x)...:.. J 1H (x), lk_ 3 (x) = 1 0 ((k- 3)...:...
x)...:... lo ((k- 2) ...:... x) and so on. Thus, the initial system gener-
ates a Rosser-Turquette system.
(5) x+ ... + x = 0 (k terms), (- 0) ...:... 2 X 0 = k - 1,
(~ x)...:... 2 (k- 1) = j 0 (x). We obtained the system {j 0 (x),
x + Y} which is known to be complete beforehand (see Prob-
1em 3.2.14(1)). .
SOLUTIONS .ANSWERS AND BJNTS
(6} The eom~Ieteness ol thJs system can he proved by a method

iS! mJla.r to that used I or pro v 1ng the eom plet en eJS of Post s Sf$ tem
(1 e tbe system lz max {.t y)}) \Ve sbBll gtve a fragment cl the
proof conc,em1ng the eonstruet1o11 of the funrtton ~ z \\ e bave
miD (z z -f 1. z + 2 ~ +
(k - t)) ~ 0 Further ID the
-ord tne.ry \\a y we ab ta1n all the. constants and the runet tons 10 (x)
11 (r} /"-J. (.t) /1 (z) = m1n (~ + (k- ~) z (k-- 2) +
z + fk - li - t)) .z + {k - (l + i )) s + t .z)
t -- 0 t k - f Then we .construct the functJons r. t (z)
such that g,. 1 U) ~ s and g. 1 {.%') :::::;: k - 1 for z =fo 1 g~ i {z) ===
m'-n (j 1 (z) +
(k-- t) k - t - «) +
1 FLnally we get the LUk.a
sev1ch negat1on ..,..., z -== mln (g4 1 0 {x} gil i 1 (x) K~t 1 1 1 (x)
toJJ 1 (~))
(7) m1n {I ~) - t - .1: - t (z -- 1) -1 = z 2 (;r --
z -
-._-p
(k- 2)} - t = m1n (z v) - t = m10. (.¥ .v) \Ve .have con

s true ted the system I z m1n (.c 11))
{8} Ftom ~ + 11 we c\lnstruet z +
11 i + ~+ ~~ +
It - f {z 1s taken k - t tJmes) (k--- 1) (Jc -- 1) ~ 1 / 11 1 (f) ::
0, ; + 0 =; 'i
-y ~z + +
11 2 • ; 11 + +
{k -- t) :::;:
k - t} ~ It (z) We have construct the .system {] 0 (~) z + vJ

(9) 11 - z ~ /o (.r) 11 + :r = r 1• (.r) = }n (r + (k -- r)) £ :::;::
~ 'f.
1 2 k ...___ I \Ve have z +y :a::: y + 2J cr. (z} where

'l't (.tl (J 1 (l-))1 +
t == f
+
U, (z))• {' terms) 1 e f9i (J:) ~
k - t We have ohta1ned the system {/ 0 (z) :r
1--!
(r)
.vJ
'1+,
v.lueh 1s knn~ to h-e- r.am p le-te- he-f or-eba nd
(tO} z + +
z ~ 0 (k terms) J t (0) =
lr -- 1 (k - t) +
+ (k --- I) = 1 (k - I terms) f X Jl (.:r) === Jo (.1) We bave
constructed the s1stem (z + ~ r~ (.tll
(tf) (JIIw' (k - 2)) - (k- 2) :::: t - (k -- 2) :;: 0 (..., ~> 0
~ -z (,..., (~ :z}} --- ll :=. .:z - y ;t. --- (z - y) ~ mln (.z y) ~ tn)n
- =
(~ x ....., g) ~ max (.r y} ~ (k ---- 2) ~ 1 :r X 1 f ~ ; \Ve +
have constructed the Post system {.r max (.t g)} -
(i 2.) (k - 1)" ~ (k - 1) === o 01. --- x = -- z :r - (- v)
z1 + y -- (k--- f) = J \Ve have constructed the complete system
from Problem 3 2 t9(9)
(13} z - x - 0 2 X t +
z= z +
2 Then we abta! u all the
constants 0 +
2 .__ 2 2 +
2 == 4 (k-- 4.) 2 = k - 2 +
1+ 2~3 3 +
2 == 5 {k - 3) +
2 ~ k - t 'Then we ha-ve
(k ~ t) ----- it = ,...,z :r (z-- g) = m•n (x y) ~ mtu (...., z ~y):;:::
max (z y) ((k - t) - z}- - ~ ~ J, (z) Then
k- ~ tJwes
we consttw.t the re.ma1tt1ng fwtct1ons J 1 ( z) (see Problem 3 2. t 9 (4.) J
Thus the JD1t1al sy.stem generates a Rosser Turquette system
(14) I~ - t = 0 01 - .z= --z 1• - (- .r} ~ The com z
plete system {s lJUD. (::t y}} has. b~ Ci)USttutted (see mh-
Jem 3 2 f 9(6))
(15) The Junction z + z + + .rl + 2 (l --- t) l~ 2
can be eas1ly tonstruct:J from z + u + 2 Jf It ::: 2m (m. ~ 2)
2
CHAPTER THREE 32!J
. - 11-1 times
(3) (....... x) - y and x-'- y are essential functions, x ...:... x = 0,
=....
( ....... X) - 0 X, (,...... (,......x)) - y = X - y, 0 - X = - X, X -
(- y) = X + y, ,...... 0 = k - 1, X+ (k- 1) = X - 1, X - (k-
1} = -;, 1 ...:... x = j 0 (x}, j 0 (x - 1) = h (x). From x + y and io (x),
we construct x + j 0 (x); h01 (x) = (x + j 0 (x)) - h (x).
-
(4) x - y and x 2 - y are e~sential functions; (it (x))2 -
h (x) = 0, :; - 0 = ;;, :; - y = X - y, 0 - X = -x, X -
(-y) = x + y, h (;} = j 0 (x). From x + y and Jo (x), we con-
struct x + +
j 0 (x); !1 01 (x) = (x j 0 (x)) - h (x).
(5) xy is .an essential function. From :; and j 0 (x), we construct
~ll the funct10ns h (x), i = 1, 2, ... , k - 1: h (x) = j 0 (x + (k-
t)). Further we have (x + s - i) h (x) = Cs·i (x), s = 0, 1, ... ,
k - 1; i = 0, 1, ... , k - 1 (g 8 ,i (i} = s and g 3 i (x) = 0 for
x =I= i). If g (x) is an arbitrary function in P\/>,' then g (x) =
((gg(ol-t.o (x) + 1) (Cg(1)-1o 1 (x) + 1) . . . (ga(i)-1, i (x) + 1) ...
gg(l\-1)-1, k-l (x) + 1). "
(6) From x - 1, we construct, as usual, ;: (x - 1) - 1 =
X - 2, · · ., (x - (k - 2)) - 1 = x - (/c - 1\ = x + 1 = $. Let
lls denote the second function by f (x, y). Since 1 -'-- x = j 0 (x),
330 SOLUTlONS A.NSWERS AI'.D HINTS
I (z, ~) == (z: + lo (:t)} j, (y) + 1c- (:2:) (y -

11(!/)l \Ve have f (xt ~}=
z Jp (•J+Jo(.:r) + j 0{z) x - /0 (z) /1(:r) =/ 0(z), j 0 {J 0 (r}) e 0 f(x O)~
$. -r 1o ~:r)t 1 {0, :s:) == 10 (x) +
:t. . - 11 {:t) =
k 01 (;r) ThuRt we
have constructed the systelll {Z: h 01 (z) 3: +}
0 ~z)} generatzng all
{unctiOU! c.t one- -v a.ri a hle ~ n PIt: {see P 1ea td 1 t h-eot-em} It
lollows from the form of the tunetzons l {.:r 0) and I (0 ~)
that f (:t. 11) 11 an essen tull funttl an Ap plytng th.e: Stupee kt
cr• t er1 on we conel ud oe that the 1n1 hal sy.s tem ia complete
(7) Obvtously, the given functton a essential We denote lt by
tf (.:r, ~) \Ve have q> (.x, v) = j 0 (z) 11 + 10 (V) ~ ~ (zr .t') =
.....
s -
fo (.:t) :t + :zJo (z} -== io {z}t ~ (~, Jo (z)} ~ Jo (x) 1o (.1}
- - :t1oUo{~))~+
Jo (z) + (z - Jo (x)) = % = z +
1 IIJ (J 0 (.t} +
1) ~ Ot ft' (0, %) ~
z + }0 (x)~ (z + +
1) spnazv
...__ ,
T
+
1- ~ z - t Jr. (J: - 1)
I
1 (~) =}
.11:-1 tlmeo~
0 t 2 k-i)
~ ul (z}, .1:}= j 0 (/1 (z)) z+ ul (.:r) +f) 10 ,.:r)= ( 0 0 2 k-1 +
0 1 2 Jc-1) :::::: hot (.t)
(1 0 We have cons true. ted the sy~ tem
-
{z, z
0 0
+ j (I (~), hos (~)} wruch Ul complete JD Pit)
{8) The tn1\1al lunct1on IS obv1onsly essent1al We denot-e 1t
..... - - - -
by ~ (.:-. g) We have 1jl (x, .:r) ::~ :t, "¢l (.t z) :::: z/ 0 (z ..._ ~) + (J: - -
It (i)) /o (:r) + zfo (i) = 0 +
0 + Jl-1 (r) Jll.-t (Jk-~ (x)) ~ D~
~ (0( z) = /, {- z) +
0/ 9 (z) +
x X 1 = {6 (.tl + z w(.r, 0) ==-
;/, (z) + (x- lt {.:t)) X t +
OX/ 0 (z) = lo (~) + x - 11 (.r) ~
htJl (.r)
(9) Ohv1ously~ the gtven funet1on is eS!1ent1al We denote 1t by
J (z t g) We ba ve I (z, z) = ( I
n t 2
t
k---1)
f == ht {.t) t
2
h1 (hl (z)) = (
0 i 2 k --1.)
~ h, (x) ~ I (h 1 {.r) h:i (zJ) =
2 1 2 2
0 fot Jc =-= 3
h, (z} {= 311 (z) for k ~ 4 If k;;:;, 41 WEI' also can:nder the
superp ~i \1 on. hs (h 3 (.z)) '!5I 0 Further, h 1 {0) ~ l J (z 0) =z :r +
J 0 (z) I {;t 1) = z.- f (1 ~ ~} =- (0 1 2 k = h~ (l:}-1)
1 2 0 0
t 2
f (z, h,. (:t)) ~ ( 1 0 2
0 k
k -1 ~ hot (.t}
--1)
(1 0} The 1n1tial functaon JS essent1al We denote 1t by t.p (.r g)
We ba ve ql (z. z) fa (.r) + 2/1 (z) = h 1 (z) Q' (x h1 (x)) =
=
J 1 (.:t) lor k ~ 3
11.1 (:} = { ;~; + 2ft {z) f()r lf >
lr. 4k = 3 lb.~n
h 1 = Ch1 (h 1 (z))) I!!!! 0 Ift how.e"Ver, k >
4 ht (ht (z)) = ili (.r}t
/o (A1 (Jo (z))} !S 0 Further, we obtajn h 1 (0} ~ 1~ tp (0~ .::) ~
ht- 1 (x)t rp (f, x) = zf.P (z, 0) = z 1, (z) +
(11) Ohv1ousJy, the glven funCtlOD JS e.ssent1al We deuote ~~
by ~ (~ t rtl We have W<~~ .z} z::;: Zj e (z-) +
11 (3;) (.% +
Jt (.z}) +
/ 1 (z) U1 (z) - /1 (z)) == } 1 (z) +
11 (z) ....... /, (z} } 1 (.;,;) ;:::;:: It (.J")
+ +
1 Uu U~ {:t)}. 1o (:.:}) -= 1 1, t~~. 1. (1 J, (z)) ~ 0, 1, (0) ~ 1,.
CHAPTER THREE 33f
•
x
1jJ (0, x) = X, 1jJ (1, x). = h 01 (x). Fro~ we obtain x-:- 1, and then
j 1 (x) : j 0 (x - 1) = ll (x); 1jJ (1 +
11 (x), x) = x + Jo (x).
3.2.21. (1) If k =.2m + 1 (m;;:;. 1_), the system is co~plete
since (x + + ... +
2) 2= x +
1 (2 IS added here m 1 time~ +
and hence the initial system generates the complete system {x,
max (x, y) }. If, however, k = 2m (m;;:;. 2), the system is i_ncom-
plete since it preserves the subset '6" of all odd numbers m E 1n
m-1
i.e. '$ = U {2l + 1}.
1=0
(2) The system is incomplete and preserves the set {1, 2}.
(3) For k = 2m +
1 (m ;;;;;= 1), the system is complete since
it generates the complete system {x, min (x, y)} (see Problem
3.2.19(6)). The function "X can be constructed as follows:
(x - 2) - ... - 2 = x - 2m = x + 1. If k = 2m (m ;;;;.. 2), the
m times
m-1
system is incomplete since it preserves the set'$ = U {2l}.
1=0
(4) The system is incomplete and preserves the set {0, 1}.
(5) For k = 2m+ 1 (m;;;;.. 1), the system is complete since it
generates the complete system {j 0 (x), x+y} (see Problem 3.2.14(1)).
+ ... + + + + + ... +
Indeed, 2x
.--..---2x
m+t times
y = x y, (x 2) 2=
m+1 times
x+1, x+ ... +x=O, 0 +1= 1, f2...!.. x = j 0 (x). If
h times
k = 2m (m ;;;;;: 2), the system· is incomplete since it preserves
m-1
the set '$ = ~ {2l}.
1=0 .
(6) The system is incomplete and preserves a certain (k- i)-
element subset in Eh.
(7) The system is incomplete and preserves a certain one-ele-
ment subset in Elt.
{8) For odd k, the system is complete (generates, for example,
the system {x + y, j 0 (x) }). For even k, it is incomplete since it
preserves the subset of all even numbers in Eh.
(9) The system is incomplete since it preserves a certain two-
element subset in Eh.
(10) If k = 2m + 1 (m ;;;;,: 1), the system is complete: (x +
2) + ... + -
~
2 = ~ (x..:.... y) +
y = max (x, y). For k = 2m
m+ 1 times
(m ;;;;.. 2), the system is incomplete since it preserves the subset
m-1
'6 = u {2l}.
1=0
(11) For k = 2m +1 (m ;;;;;= 1), the system is incomplete since
. m
It preserves the subset '6" = U {2l }. If k = 2m (m ;;;;,: 2), the
' 1=0
system is complete. Indeed, x-=- (x ...!.. y) = min (x, y), -min(-x,
~y) =max (x, y), x...:... x = 0, ..... o = k - 1, 2j 0 (0) = 2,
(k-1)..!..2=k-3, ... , 5...!..2=3, 3...!..2=1, (k-1)...!..
332 SOLUTIONS ANSWERS Af\D HINTS
{ = k - 2 (k .____ 3) t k - 4 5 - t -- 4 J1 <~ t) - !:::
J, 1 (.r) I 1 k - 2 Jk 1 (~z)- J.o{r) Thus the Ro~er

Tnrquette s}stem ba~ been constructed
{12) Fork >
4 the sy.§tem 1s Jttcomplere For It 4 1t preserv.eq
the pa.rtttlOD D = { {0 f} {2 3}} and for k ~ 5 1t preserves
for e.Iample the part1t1on D = {{0 1} {2} {k -- 3}
{k 2 k - 1}} For k = 3 the 8).stetn JS complete Indeed
....w (J 0 ~z) J l (x)) + J 1 (.r:) J 1 ( ~z) === J 0 (x) I 0 (max (1 1 (.r}
J 1 (:r)}) J 1 {.t) 1 1 (f)- 2 ~2 = 0 ~max (.--% ~v) --
m n (~ g) 1 e "e have- constructed a Rosser Turquttte system
{f 3) For k ~ 4 the !}stem 1s lncomplPte For k = 4 the par
tltlon D =(
{0 i} {2. 3}} lS pH$erved wh\1-e: fot k > S 1t
preser,es for example the partlti()D D = {{0 f} 12 k - 3}
{k- 2 k --- 1}} fa.t k 3 the sys.tem 11 c.a.mplete ~mlil (~r.
...... y) ~ rna~ (r g) miD (2 - Jo (:.r) - 2, 1 (.i') z) ~ I, (.x)
J, ( ...... z) = J • (z) J 1 (.r) - J G (max (J 6 (.i') I, (r})} .-...0 ~ 2 Thus
\\ e have construe ted the R o=ser Tlll"q ue t t e .s ys tern
(14.) The !~ s.tem ts l ncom plete It Jlreservel\ for ex:ttm p.le the
parttt1on D = {{0 t k - 2} fk 1})
(151 The Sl stem J! complete It IS not d1fficult to construct for
examp e the followJng Rosser Turquette s}stem x - (z g)
m1n (.z y) ..... mln (~z ...... !J) - max (z g) :z - x - 0 ~o __.. .
k - I (k -- f) - (k 2) i (k - 2) - f == k - 3
3 -- t ~ 2 ({k - t) .x) ___
..___,_
A I t mes
z
...,.,.
=J 0 (x) J, (.t i)
J 0 (% If (z} ~ ~ 1. 2
{l - k - 1
1}) -
3 2 22 The funct1ons appearJng 1n each of the g1ven systems
are pol!: nomLals Consequenlly for compo$-tle k these systems a[e
IncompJete If however k IS a pr1me number not less than 3 the
completeness of a s1stem can he convellJently pro"ed by construct
tng one of the follow1ng systems (a) th.e syst~m lJo (.t z + y}
{f ~ +
11 xy comp1ete Jn the class of poJynom1als
(1) 1&. 1.s suffictent to construct the funet1ons z !I and zg +
\\ e 1ulve 1 + + 1 f x I 3 1 3 t X 1= 5 1= + + +
(k - 2) + i X t k 0 z y x X 0 - :z v (z + + + +
1)
"
+
_.
+ =1 :r --- t z
._....,.
0
:Z
% (g ~ 1) == ~g
a -!
+ +
~ t t. m~
(2) It JS :suffic ent to construct the s} stem (Jo (.:r) x + v}
% .J: +1 1 11 --- i _..... 0 X -- 0 + 1.. :::. :r + i ~ -- {y + \)+
t z - g t X zt x1 x 1.z1 -- .z 4 x"' 3.r1 xll 1 (k t
ts an e'en number SJ.nc.e k \S a prlme number not less than 3}
t - zk 1 = Jo (.:t) 0 -- % = --z z - (~y) == ~ y +
(3) We con:struct. the system {/e (z} z y} :r + + +
z 0 (k terms) 0 t - t .r' (- j) .rl zt) z' ( =
~· ..r• {-zA ~ 1 ( i) '> I) = f (k--+ +(
1 terms) t + (-.7h 1 ) - f - zh l !a (.x)
(~) \Ve conslruct the iunct1ons t z 11 and ~~~ z (k--- 1} + +
z - (k -- I) J === J + +
I (k - I) + f :.:: 0 0 t f + =
X X
0 +z - 0 !J ::= X 11 +(.r t) 11 +z I - y + 0 + +:!If + +
z +1 z + (k - i) = % - - 1 % (y - l) z + +
1 ~ :ry + t
(xy +
t) - I = ~v
CHAPTER THREE 333
(5) We construct the system {J 0 (x), x + y }. We have (x +

2y) + ... + 2y = .• - y (the number of 2y terms is (k - 1)~2),
x - x = 0 0 - x = -x, x - (-y) = x + y. (The functton
x +y can b'e obtained more rapidly: ((x 2y) 2y) +
2y = + + ... +
(h+ I )/2 tlm~s
x + y.) Further, -(k - 2) = 2, (k - 2) - 2 = k - 4, ... , 3 -

2 = 1, and from functions 1 and x - y we obtain 1 - x; 1 X x~ =
x~, x2x2 = x\ ••• , xh-3x2 = xh-1, 1 - xk-1 = Jo (x).
(6) We construct the system {Jo (x), x + y }: x - x 0, 0 - =
X= -X, X - (-y) = X + y, ,_.o = k - 1, - ( k - 1) = 1,
x2 X 1 = xJ, x2x2 = x\ •.• , xJxh-~ = xh-1, 1 - xh-1 = Jo (x).
(7) We construct the functions 1, x + y and xy. We have
+ + + +
(x 2y 1) 2z + + +
1= x 2y 2z 2,+ + (x- 2) 2y
2z+ + +
2= x 2y + + + + .. +
2z, ((x +·(2y 2z)) (2y 2z))
(2y+ + +
2z) = x y +
z (here the terms 2y 2z are contained
+
(k + +
1)/2 times), x y +
y = x + + + ... +
2y, ((x 2y) 2y)
2y = x - y (the terms 2y are contained (k - i)/2 times), x - x =
0, 0 - X = -X, X - (-y) = X + + +
y, X +
2 X 0 1 = X 1,
+
0 +
1 = 1, (x +
1) (y +
1) - (x + 1) - (y 1) = xy- 1,
(xy - 1) + 1 = xy.
(8) We construct the functions x + y and xy. We have ((x +
2y) + 2y) + ... + 2y = x - y (the number of terms 2y is
(k- 1)/2), X - X = 0, 0 - X = -x, X - (-y) = X+ y,
(x (y + 1) x + y + 1)- (x (y- 1)
2
- x + y - 1) = 4.xy +
2 -
2, ((4xy + 2) - 1) - 1 = 4xy. Further, if k = 4m + 1 (m;;;;. 1),
we consider the sum 4xy + 4xy + ... + 4xy of m terms 4xy.
It is equal to -xy. But since we have the function -x, we can easily
obtain the function xy: -(-xy) = xy. If k = 4m + 3 (m;;;;. 0},
we construct the sum 4xy + 4.xy + ... -!- 4x:~ of m -!- 1 terms.
It is equal to xy.
(9) We construct the functions 1, x + y and xy. We have
2
XX- x = 0, 0 + 0 +
1 = 1, X -j- 0 +
1 = X -j- 1, ((x +
1) + 1) + ... + f = X - 1, X+ (y - 1) + 1 = X+ y, X (x +
h-I tlffiC!>
y)- x 2 = xy.
(iQ) ((x- 2y)- 2y) - ... - 2y =X- y, X- X = 0, Q-
---:-::--~:-:--
(!<+ 1)/2 time~
X= -x, (-y) = X+ y, 0 X 0 + 0 + 1 = 1, X (y- 1) +
X-
x +
1 = xfl 1, (xy + +
1) - 1 = xy.
(11) We construct the functions 1, x + y and xy. Since k is
a prime number, xk = x, and hence we have 1 + x -
X+ XX ••• X= 1 + X and (x +
1) + 1 = X+ 2, ... , (x +
h tunes
(k- 2)) + 1 = x + (k - 1). Further, we obtain 1 + x -
lx + + x (x + 1) (x + 2) ... (x + (k - 1)) = 0, 1 + 0 = 1,
1)
+ X - y + X X y X 0 X ... X 0 = 1 + X - y, 1 + 0 -
X = 1 - x, 1 + X - (1 - y) = X + y, 1 + X - y + X X y X
1 X · .. X 1 = xy + x - y + 1, (x + 1) (y + (k - 1)) + x +
1 - (y + (k - 1)) + 1 = xy + 2, (xy + 2) + k - 2 = xy.
3.2,23, (1) Take into account the fact that {J (x), x + y} is
d. complete system and that 11 (x + ,) = 10 (x), , = 1. . • . , k - 1.
0
SOLUTJO"\S Arr..SWE.RS .Al'.D HI'tTS
(2) Cons1der Posts system (3) Add1ng, tor e~ample the functtons
,..,z and z~ we can constrUct Post~s s}:stem (4) Post s S}Stem can
eastly be constructed hy add1ng, for ex,.mp]e~ the functlODJ ~z
and i t 5} By add 1ng~ fct t xa mple the funct lOO. i ~ e tan ~ls1}y
ohta1n Po.st 1 S).stem (6) It Is sufficJent to add the functions~~
-
and s to construct Posts system (7) Apply dtreetly th@ theorem
on completeness and JDcompleteness of a polynom1al class tD P~c
3 2 24 (1) For k === 4 and 6~ \\e can constder the func.tton
f (.:r + 1 .r) {2) For k ~ 4. anal) ze the funct1on I (2t .r) and for
k = 6, the funetton 1 (4 1 ~) (3) For k = 4 • e can cons1der any
of the functlOJlS U (0 z))t~ (j (.r + 1, .:r:)) 1 and 1 - t (2t s} "hlle
lor k :=;;:: 6 the runct!OD 1 - I (4, z) should be COilSJdered (4} For
k = " any of the functums u (z 0)} 1, u <~ .r)}-' (/ (~ + 2 z}l1
or t (2~ z) -- 2 Js !JUitable Fork .=; 6 lire can ton.s1der the funct1o.n
f (4., z} - 2
3 2.25 .. (I) The .suh.system B = {k- 1 Jo (;r}t .. - v) ts a
hu1s 1.4!1. us ptove. thl.s 'Ve have 1fl {k ---- l) = 0~ J, {0) ~ 1,
(k-- 1) - i = k - 2, • 3- 1 = 2 Jo ((k - t) --- ~) =
11
J}c I (:t)~ }l-1 (t} ~ J0 {(k - 2) -- .t) - }"A; -1 (3:), lA-s (%} =
J0 ({k 3) --- .t) - 11 ({k - 2) - z).
_.-ror ~ /t (.z) =:: 10 (I ~ .t") ~
/o (2 - z) Let g (z) he an arbitrary I unction tn P1: 1 lt can be con
struc ted as foil ows At first we construe t the f U1l c ll on gIJ (.r) ~
((({k - 1) -- 10 (.x)} - lo (z)) - - I 0 (.r))~ where J., (z) Is sub
tracted from k - t sucli a number of t1mes r 0 that g 0 (0) == g (0)
i e (For g (0) = k - tt we put to (x} ::::;:
r 0 = k - 1 ---.. g (0)
k - f ) Then "e .subtract t"1 ~ k - f - g (f) t1mes the functton
/ 1 (z) from gg (s)t and sa on (In part1cnlar,. the funct1on i' can he
eanstrnc ted 1n tfll s '-ay ) In otd.er to pro\ e the eomple teness a( the
system Bt It remains for us to use the srupeckt crtteriOD (or recall
that m1n {.t:t ~) = : - (::~: - y}~ max (.z y) ~ !RI.PIDln (~$, ..... y}t
"'-'z :.::: (k- t) -- z and that the system {z- .n1a.z (~t .v)} lS complete
1n PA) Further~ we must prove that B does no~ contatJI complete
proper subs) .stems The funct1on z _. y r..a.nnot be removed (rom B
$1DCe the fUDC tl Oll3 /c - 1 .and J D (~) f SS eD tla}} Y depend OD DOt more
than one var1able Further~ we oh Vl ousl y have {J 0 (.z) z - 11 } c::
T ( {0, t }} and {k - f z ~ y} c T ( {0 k - 1J)
{2.) B= {/6 (z)~ %- y1 } Oh1llOUSly z--- y 3 Ls an -esseut1al
!unet1on~ \Ve shall show how all thE! lunctJons oi one varJable can
be const-ructed \Ve have ((/b (:t) - Jl (z)) - I~ {:t}) -- -11 {:r:) ~
h tunes
1
0, J f (0) ~ k-1~ {(k-1)- (k- f) 1)........ -- (k-1) 1 === k- t - i 0=
t tunes
t'l 2"J. ~ k- 2)" ((z- f 1 )-1 1) - --.. fl ~ z- l (l ~ f~ 2'\
l tlrnes
k-2) JJ (z) === (I1 (z- i)- 1: (:t- (t-1)))- -- J:(:r- (f-1})
1t-l 'tlmea
t:=: t, 2, , k--- 1 Let g (.r) .be an arb1trary funct1on 1n Pku It
can be construe ted as I oil ow s \\ e .fir.s t construe t the func t 1on
It (z:) if g (0) == k -1, then Itt (_z) =
k-1 If however, g (0} *
CHAPTER THREE 335
k-1, then g0 (x) = ((k-1) ...:.. Jfi (x))...:.. ... ...:.. J~ (x), where r 0 =
r 0 times
k-t-g (ll). Then we construct the function g 1 (x): if g (1)=
· k-1, then gl(x) = g0 (x), but if g (1) =/= k-1, then g 1 (x) =
Jr
(g0 (x~...:... li (x))...:... ••. ....:..
._
(x), where r1 = k-1- g (1), and so on .
r1 Urnes
On the k-th step, we obtain the function Ch-l (x): if g (k-1) =
k-1, then Cn-dx)=cn-z<x), but if g(k-1) =/= k-1, then
Ch-dx) = (Ck-dx)...:... ~~-I (x})-=- ... ...:.. JL
1 (x), where n1-1 = k-
rh_1 times
1- g (k-1). Obviously, the function Ch-l (x) coincides with the
function g (x}. In order to complete the proof of the completeness of
the system B, it is sufficient to use the Slupecki criterion. It can
be easily seen that B does not contain complete proper subsystems
since the function J 0 (x) depends only on one argument, and the
function x...:.. y2 preserves the set {0}.
Remark. For odd k, the subsystem {.r - 2, max (.r, y)} is also
a basis.
(3) B = {~J·, min (x, y), x + y}. We have_ x + (~x) =
/,· -- 1, x + k -· 1 = x-1, (x- 1) -- ... - 1 = x, ~min (~x,
----.r---'
times
h-I
""Y) = max (.r, y). We have obtained Post's system. It
should be also noted that { -x, ruin (x, y)} c T ( {0, k - 1 }),
{-x, x + y} cL and {min (x, y), x +
y} c T ({0}).
(4) If k is odd, for example, the subsystems {x + 2, max (x, y)}
and {x + 2, x...:... y} are bases. This fact can be substantiated by
using the following relations: (x + + ... +
2) 2 = x + 1 (the
term 2 is taken (k + 1)/2 times) and x...:.. (x --"--- y) = min (x, y). If
k is even (k ~ 4), the basis is the following subsystem: B = {k - 1,
x + 2, x--"--- y}. Let us prove this. We have (k- 1) + 2 = 1,
(k- 1) --"--- 1 = k - 2, ... , 3--'- 1 = 2, 1 ..:.... 1 = 0, (k- 1) -=-x=
""x, x....:.. (x...:... y) = min (x, y), ""min (~x, ~ y) = max (x, y),
((k- 1) ...:... x) -'- ... --'- x = J 0 (x), Ji (x) = J 0 (.r-'- i) -'-
h-I times
1)), i = 1, 2, ... , k - 1. Thus, the subsystem B
la (x...:.. (i -
generates the Hosser-Turquette system which is known to be com-
plete. Consequently, B is a complete system. Let us prove that B
Is a basis. Obviously, {k - 1, x ...:.. y} c T ( {0, k - 1 }), the sub-
system {k - 1, x + 2J consists only of functions which essentially
depend on not more than one variable, and for k = 2m the sub-
system {x + 2, x ...:.. y} is contained in the class T ( {0, 2, ... ,
k- 2}). Consequently, B does not contain complete proper sub-
systems.
(5) The. basis is the subsystem B = {j 0 (x}, x + y2}. In order
to prove th1s, we must take into account that j 0 (j 0 (x}) +
(j 0 (x)) 2 ==
1 and. that the function x + y~ can be used for constructing all the
unctiOD:S of one variable (by using j 0 (x), h (x), ... , ik- 1 (x))
alm o~t Ill the same way as the function x + y was used in the
so1Ut10n of Problem 3.1.8.
336 .SOLUTIO'IrrlS .Af'r.SWERS A' D HJ'-~S
3 2 26 Take J.Dto auount the fact that {0 I k-t

J, 1 (.r) wu (.z: y) m.ax <~ g)} c 'it ({0 1 k - 2}
{k - t }} mu1 (/j (z) J.,., 1.(.rl) !50 0 ~' ~ k -- 2 and
max (J 0 ( / e {..t)) J • {.t)) se k - I
3 Z Z7 l ( the var!. a hles o[ th.e (unt t ''-'US belong \.0 lhe $~ l
{z1 z 1 :&n. } the number of dt.fferen t func.llOlls an any
closed tl ass 1n P lt. ( 1n.elud tng P A as a part 1calar ease) 1s nat more
than countable Let a closed class A (G P4 ) have a fin te complete
system Then any has2s 111 1t ts fin1te Consequently the po\\er of
the set or 1 t.s bases does not exceed the power of the se ~ of a II 1t .s
fin1 te sll h.s l.S te ws w h.tch (as 1s known from the thec.ry of sets) IJ
n.ol mote th.a n couutab'U!
3 2 28 (1) Any precomplete class IS closed Consequently B
1~ a tl oM:d c~ass a. n.d the s~t A 1s also a cl ns.ed class n P A.1 ~ as an
1nterset t 1on of closed cJ asses It should he also no ted that the
clas.i B and hence the set A conta1n an 1dentJty luoclton Let w
'ons1der a closed class $ (A) cons LS trng or all the f 11nc t ans of tha
~
system P tt. wh.tch preserve the set A (l he [unc tLon f (zn} n: ~ 0

preserves the set A 1I for a oy f unc11 ons fP 1 ( x) cp,. (x) 1n A
the SUPflPMlllon J {91 (;r\ ~ 11 (.x}) lS a {untl~on lB. A) Ohvi.
ou.sJy B -c ~ (A) and $ (A) =fo PA (s1nce for example any fuDc
\.i-On. ln. P~ ,"..A 1s. no\ col.\\Aln~d 1n ~\A}) Bu\. since B IS a pte
complete cl,;1ss Jt cannot be exactly conta.tned Jn any cl{)Sed cJass
d fffer1ng from the en ttre P 4 Therefore B ::::: $ (A) The class
~{A) lS un1qu.ely detintd by t.h.e set A Ccusequ'!ntly tht r.lass B
ba v1ng the proper t1es s pec1 lied 1D the cond 1t 1ons of l he problem u
untqut
(2) It folloW'S from Problem 3 2 27{1) that the number of suth
precomplete cla$Ses 1n Pl 1s not more than the number ol all sub
~ets of the set P,., Slnce l P k J l -- .kA the po\\er of the set of all
A
subsets tn Pk ts 21t
3 2 29 (I) Let us prove that the class A1. contatns only one
(ace urate up to red est gnat 1on of va r 11 b les) f untt 1on of one var table
g (x} == .z:1 \Ve denote the (unc tton ;:;:1 !I' h y f (z y ~ \ Ve have
I (x z) - :l:t - %! I (.z:11 .x) - I (z r) :::;;:: rx1 ~ z! I (r ~'!) =
z' X' -:;;::. .zt 0 h"V' tmsly z" -=/=; z a ud z 2 -==fr. eons t
(2) \\ e d.en.ote the funet1on It (x) J 1 (y1 by q> (.r .V) \Ve have
q> (z z) ::=: 0 (4) {0 z) ~ ,-J~ 0) - fP (0 0) :.:: 0 Therefore X2
eonta1ns only one funttlon epend1ng on not more than one var1
able and this f nne tt on Js an 1d en t.t ca I zero
3 .2 30 \Ve denote the functton ~~ (.z}l, (II) by "" <~ It} Let
us eoll.SJder the general form of the .superpositiOn generated by the
set. {~ (;S: !ll} Q[ {x1 :rm lh Ynl = ~ {q(I (zl
zml '91 1 (V1 v,.J) where W1 JS e1ther a l'&riaLie or a super
pmnt1on generated by tbe set 11}:1 (.r .v)} 1 1 2 Let us prove that
If at least one W1 d 1II ers from a var1a b le the .s u perpos1 t1 on W 15
equal to an 1denttc.al zero Indeed any lunctiOil represented by the
su perpos1 t 1on genera ted by the set {~ (z yl } does not assume the
'al ue of 2 Therei ore 1f for e.1am p le W1 d1.tlers ftom a varJ able
t~n ~ ('It (£ l E~} 'JI 1 (E, E;,)) s; 'll (€ I"' {2.} E :a) ~
{0} Thus the class K 1 conta1ns only three pa1rw1se 1ncongruenf
CHAPTER FOUR 337
functions: 0, j 2 (x) and j 2 (x) j 2 (y), which essentially depend on

zero, one and two variables respectively. ,...,
3.2.31. If in the function fn (x 11 ) (n ;;::;, 2) we identify any l
~
v~riables (2~1~n), we o~tain a function co~grue~t to 1n-I+I(x 11 ~1+ 1 )

(smce j 2 (x) 12 (x) = h (x)). Further, like m. the prevwus
problem, we can prove that if at least one Wi m the superposi-
tion W = / 11 (W 1 (xu, ... , x 1m,), W2 (x21· ... , x 2 m,), ... ,
'Un (x , ••. , xm 11 )) differs from a variable, then W is equal to
7lt 11
an identical zero. Hence K~ contains, in addition to the functions
/ 1 , f.,, ... , fn, ... , only au identical zero, and any function
fn generates any function fz for l < n, but is not ~enerated by the
set {11 , / 2 , • • • , fn-d· Hence it follows that tne set of closed
classes in K~ consists of K~ itself, the class {0} and the classes [fd,
l = 1, 2, ... , and {0} ~ [ft) ~ [/ 2 ) ~ • • • ~ [fd ~ ... ~ K~.
Obviously, none of these classes is precomplete in K~ (since any
class in this chain differing from K~ is strictly contained in some
other class differing from K~).
3.2.32. Let us calculate the number of non-essential functions
I II
in Ph and subtract the obtained expression from k' . A function
( 1)
is not essential if it assumes less than k different values (in Eh)

or assumes all the k values (in Eh), but essentially depends only on
one variable. The number of the first type functions can be deter-
mined by the inclusion-exclusion method. It is cl'- 1 X
(k-1)
hn h 2
-ch- (k-2) 11 n
+ ... +(-l)iCh
h-i
(k-i)~< + ..
11 " +
hn
(-i)h- 2
?
Cj;2 + 1
(-l)h-1 Cn. The number of the second-type
functions is (k!)n.
3.2.33. Let us establish a one-to-one correspondence between
all rational numbers of the segment [0, 1) and the set of functions
{iz, /3, ... , fm, ••• }. We take an arbitrary real number y E
[0, 1) and denote by Cy the subset of all such functions in {/2 ,
/3, · .. , fm, ..• } whicli given the chosen correspondence are in
compliance with all rational numbers less than y. For y 1 < y 2 , we
have Cy, ~ Cy,· Let By = [Cvl· Obviously, By,~ By, for y 1 < Y~·
•
CHAPTER FOUR •
4.1.
4.1.1. It follows from the calculation of the type (v, x) pairs,
wh ~re v E V, x E X, and v and x are incident .
... 1.2. (2) None.
, (4.)1.5. (1) Let 0 (v) be the set of vertices adjacent to v and
0 v ~ {v} U 0 (v). By hypothesis, I 0' (v) 1 ;;::;, (n +
1)12 and
I1,1 '-? (~) I ~ (n- 1)/2. Hence it follows that pach vertex from
1
"-0 (v) Js adJacent to a certain vertex from 0' (v), and hence the
22-UGJG
338 SOLUTIONS A~SWERS .41\D HI,TS
graph 1s connected (2) It 1s poss1hle for an even n and lmposslhle

for an odd n
li i 1U Let {v u) [u;. a} h-e t1ro ~na'-n' o[ ma11mum fengtb
\\h!th have no common lertttes 1n a graph G The graph G 1s con
n cc ted Conseq uen ll y there ex1s ts a cha1 n Z ct>nn ec tin g for e:~ am
p)~ \ettJce.s " and ro L~t v1 be the )ast vertex of the chain lu u]
"h1ch HI met on the path from v to w along the cha lD Z and u 1
he the first ,ertex o[ the chain {w tl met on thts path alter v1
The "ertex t') of the cha1n lv u] dJ v1des 1t In to t \\ o parts lv v1 )
and lv 1 u) Let ( v t 1Jbe not shorter than [ L 1 u] Stmtlarly the
vtr\e-1 n. 1 d1vld~ tbe ehaln [w 1) 1nto t\\o patts Let lw w 1 }
be not shorter than (w1 tJ Then the ehaJn [u vl w1 wj ~~ longer
than each of the cha1ns lv u) and fw tJ \Ve arr1ve at a contra
dtC1lOD
4 I iS litnt Constder the complements of the graphs G and H
4 t 20 2p 4 t 22 Stx
4 1 24 (f) Let v u and w be three 'erhces of the same po'Aet of
a graph G Jn R,., \\ e assume t.hat v and u are adJacent and v and w
are nat For 'erttces uand w at least one ol the 1nclu~noos 0 (u) s
0 (w) or 0 (w) s 0 (u) 1s vahd It follov.:s hence from this and from
the equal tty ~ 0 (u} I == J 0 (w) I tha l 0 (u) ::::=: 0 (w) and there-
fore v E 0 (w) Let us now consJder tbe paJr v and w It CoHo\\~ from
what has been proved above that u and w are adJacent But the
lnGlU.SL0\1 0 (v) 5 0 (w} daes. not take place SLnce u E 0 (u)'
0 (w) The uiclus!On 0 (w) c= 0 (v) 1s not val1d e1ther since
I 0 (w) I -=== J 0 (v) J \Ve art I ve at a con tradltt1on
4 1. 25 {l) The numbe? ol edges ot the graph G IS equal to the
.sum of the numbers of edges of graphs H1 dlvJded by n - t (2) Th1.s
follol\s from 4 t 26 (I) 4 I 28 Two
4 I 34 (1) Ye~ there eXl! ts 4 t 3"41 Yes
1s not
1' l.s 4 1 39 No 1t
4 2 5 Let C he a plane 2 connected graph w1th at least two

1 n terna I {aces 1f there e.J 1s ts an 1n t ~rna I I ace separated from an
external Iace by a sJng}e stmple chain then tbe deletJon of thJs
cha1n Jeads to a 2-connected graph (pro\e thJs statement) Suppose
the. t thete \S no such a face. T h~n any lnt ema l fat-e. ha vi. ng a coro
man cha1n Wlth the external face has at lea~t one \ertex 1n common
Wltb It wh1ch does not 11-e on th1s .c:ha1n Let us prove that thJs
case IS tmposs1ble \Ve number all tbe Internal laces V.e mark
the connected p1eces of the boundary of the e~ternal !ace w h1ch
simultaneous! y heloag to an tn ternal face with the number t by
an 1nd e.x: " Then there e:t. 1st such 1nd1 ces ' and. 1 wh1eb are met
dur1ng the eJrcumventton of the boundary of the external face 1n
the wqu-en~e i 1 1 1 \Ve d~nate the eorr~pcnd1ng_ Plete'! of
the boundary by f 11 rJz r 1 fJe: Let us choose on these Pl€LCeS
p o1n ts a1 b1 a 2 b1 on the p'a.Jle (one poJ nt on each p1 ece) Then
the potnts ~A 1 and at can he connected 'by a curve whose all po1n ts
except a 1 and .:~ 1 are 1nternal poJnts of the face \\ Jth the number I
Yrh1le 'POints b1 and bs can he connected by a curve whose tnternal
part ltes 1n the face w1th the number l 1he curves Intersect at A
CHAPTER FOUR 339
certain point d which is hence an internal point of two different

faces. We arrive at a contradiction.
4.2.9. Let G be a 2-connected planar graph and G' be its plane
representation. Let n be the number of vertices, m the number of
ed(Tes and r the number of faces of the graph G' (including the
inter~al and external faces). According to Problem 4.2.6, we have
r = m - n + 2. The number of pairs of the form (v, x), where
v is a vertex incident to the edge x is equal to 2m. On the other
hand, this number is equal to the sum of the powers of the graph
vertices. Since each power by hypothesis is not less than 6, we can
write 2m ;:;;;. 6n, i.e. m ;:;;;. 3n. The boundary of each face has at
least two edges, and each edge belongs to the boundary of not more
than two faces. Hence 3r < 2m. It can be easily seen that the
system r = m - n + 2, m ;:;;;. 3n, 3r ~ 2m is incompatible.
4.2.14. We assume that the graph K 5 is planar. Let K5 be its
plane representation. Then the number r of the faces of the graph
K! is r = m - 11 + 2, where m is the number of edges, and 11 the
n~mher of vertices. As in Problem 4.2.9, we have 3r < 2m. Hence
m ~ 3n- 6. But m = iO and 11 = 5. We arrive at a contradiction.
While proving the non-planarity of the graph K 3 , 3 , we note that
each its cycle contains at least four edges. Assuming that K 3 , 3
is a planar graph, use Problem 4.2.1:1.
4.2.22. Each vertex of a 6-connected graph has a power not less
than 6 (see Problem 4.2. 9).
4.2.27. (3) X (Bn) = 2, x' (Bn) = n.
4.2.28. Two.
4.2.45. (1) Let l' (m) be the number of terminal coverings of
a chain of length m. Then -r (m) = -r (m - 2) + -r (m - 3},
• (1) = -r (2} = -r (3). The solution of recurrence relations of this
type is given in Sec. 8.3.
-
4.2.49. (i) "~< (G) is the average number of vertices which are
not adjacent to vertices of the subset U s V of power k. Conse-
quently, there exists a subset U 0 , I U 0 I = k, such that v ( U 0} <
-
vk (G). If lV is the set of vertices which are not adjacent to vertices
from U 0 , then W U U 0 is the verte.x covering of power not exceeding
k + v;; (G). (2) Let e (v, u) = 1 if the vertices v and u are not
adjacent and e (v, u) = 0 if they are. Let Sh (v) he the number of
subsets U s V composed of k vertices none of which is adjacent to
v, and let d (v) be the power of the vertex v. Then
•
~ ~
-
USV, I U I =h
v(U}=
UCV,IVI=hvEV
e(v, u)
n-d (v)
=h Sh (v}= h k < lVI . Hence
~ev vEV
1V1-1
-vh (G)~ lVI
II
i=O
--
'"*
SoLUTtO~S ANS"" £RS A!',B Ht~TS
4 3 2 The proof ran be carrJed out by 1nd uctlon on the number

of '\ettu:e-s :nr lht d\.grapb On an induttlv~ st~p 1l 1s -e,..~d'-fn\. to
de)ete one of the vertJct'S (v1 or t- 1 ) from the d1graph
4 3 5 Use 'ndut.llnn an the numh~r k
4 3 8 The statement can he proved by 1nductJ on en the number
of vert\c~ ln. th-e tol1mam~nt
4 3 10 For n = 1 the 1nequahty JS ob"VJous Let 1t he vahd
for suth n lhat 1 -s;; 71: ~ m Let us ~OOSld~r \he tournament r
with m + I vert1ces Slnce any tournament With m + f vert1c.e.s
b:a$ ex $-c t l y CZ.~ +1 .at cs there ex 'st! a v~t-e-x v0 Wllb an ou l d-egree
~[(m + l)/2] Havlng deleted the vertex v 0 tn T \\e get a touma
ment T ""b1ch bas m vert! tH and con ta lDS a ~rl S ~on Sl :s t 1ng of
at least 1 (m) of compa tJble arcs Tbe arcs emerging from va and
the arc8 belongtng to S are compatible UBlDg the lUduct1ve h)'
pothes1s we ohta1n
4. 3 12 Any -even <>rd ~r greu p eon ta1ns a. t l-east an e elemenl

Jn verse to 1tse If and d Ifi er1n g from the un 1t element of the group
If lhe gtaup of the t{}:umam~nt. T \\ere e"V~n l\ wouid t.a-nta1n an.
el e.Qlen t a. sa t1 sf y tng the a hove property S 1nce a IS not a ll n 1t
ele-m~n t th-ere ex 1~t l ~ o v etll e-es v1 and rt su'h tbat a (t 1) ~ v 1
and a {lJ1 ) --- v1 Let T contaJn the arc (v1 v,) Then th~ arc
{a. (v1) a. (v2)} !Jbould abo htlong to T \Ve amv@' at a contra
du:tiOD
' 3 15 The ptoof t-an be tanred out by lndutt\on o-n the- num
her of .strong corn pon en ts ol the tournament I l there J s only one
verte-x lD c ond ~n5a llvn oo lhere ar@ two 'e-rll c es 1 n 1t the ton den
sat1on IS hy defini t1on a transL ta ve dJgra ph
-' 3 19 carry out 1nduction on the number of vertices tn the
dJgraph On an tnductlve step one of the ,ert1c~s wtth :zero out
degree should be de1eted after es tab lishJ ng bet oreba nd tb e e.ll.sten ce
oi such a ver1e.1. tn the gtven d1graph
4 3 21 Proof can he carr1e-d out by 1nduct1on on the number
of vertJ ces 1n the d Jgraph
4: 3 25 Carry out 1ndutl1DD over the 1ength of tbe contour
4 3 29 Proof can be carraed out by 1nduct1on on the deter
uunant order
44
4 4 2 Starttng from an a rh1 tra ry verte"t of the tree u e con
struct .a ebalD add Jng a new vertex on each .step as long 4ls pos~nhle
At the moment when 1t becomes tmpossible the end vert1ce! of the
cbaJD w1ll "become pendant 'ertites of the tree The prooess of con
!itrUcttoD termtna te.s s1nce the .set of \ ~rtJce$ Js JlnJ te' and there are
n a t} eles 1n the tree
4 4 4 (f) According to Problem 4 t 26 the numbe.r of edges Jn
the gr,_ph G and 1ts conneet1veness can be reconstructed from F (G)
CHAPTCR FOUR 341
4.4.9. The induction can he carried out on the magnitude of

the tree radius.
4.4.12. The power is contmual. 4.4.14. 2. 4.4.23. Carry out
induction over the length of the vector. 4.2.28. It is true. 4.4.30.
It is true. 4.4.31. Generally, it is not true. 4.4.32. Generally, 1 t
1s not true. 4.4.33. n - 1.
4.4.34. An irreducible network does not have parallel edges
and is not p-reducihle. Hence m ~ ( ; ) - 1. Each internal
verte\. in an irreducible network has a power not lower than 3 and
poles have a power not lower than 2 for n > 2.
4.4.40. (1) (a), (b) and (c). Yes, it can. {3) (a) and (b). No, it
cannot. 4.4.43. Consider r~,, m > 3. 4.4.52. No, it is not.
4.5.
4.5.2. (1) 2( ~). (2) nJ • 4.5.3. (2) Use the inclusion-e\.clu-

m
sion relation.
4.5.5. (1) Make use of the fact that in a connected graph,
II
m~ and m ~ n ~ 1. (2) A connected graph Wl th m edges
2
has not more than m+ 1 vertices.
The number of pairs of differ-
.
ent vert1ces . hence not 1arger t h an
1s (m+
'> 1 • Hence \jJ (m ) """
_,
I -
m+1
2 . Then Stirling's formula should be used.
'
4.5.6. Note that the number of vertices in a graph does not
exceed 2m. The further line of reasoning is the same as in
Problem 4.5.5 (2).
4.5.7. See Problems 4.5.5 (2) and 4.5.6.
4.5.9. (1) The code of a tree with m edges is a dual vector of
length 2m with m unit coordinates. (2) See Problem 4.4.24.
4.5.10. Use the solution of Problem 4.5.4 and Stirling's for-
mula. 4.5.14. See Problems 8.3.18 and 8.3.19.
4.5.16. The network r (a, b) having properties I and II is
a result of substituting the networks of the type r~. k = 1, m 1 +
for the_ edges of the network q_, (a, b). The number of such net-
works IS equal to the number of arrangements of m objects in n - 1
boxes so that none of the boxes is empty, and is ( m-.;) .
n - ....
4.5.17. (2) Use the solution of Problem 4.5.14.
4~5:19. ~f a graph is not connected, the set of its vertices can
be d.1V1ded u~to two parts that there are no edges connecting the
vert1ce~ of different parts. The number of vertices in one of the
parts hes between 1 and [11'2]. The graph with numbered vertices
Is completely determined by the choice of the edges. The number
of edges v. h.u:h are forb ul den 15 equa 1 to k ( n - It) where Jc 1s the
number of vert1ces 1n one of the parts
4 5 22 A.D) subgra ph of the cube B" IS completely defined
by c:pee1fy•ng the set of 1ts vert lees The ~l cf 'ert1ces of a con
oec ted subgt3 p b. 1s d efi.ned by ~pec1 f} tng 1ts certa1 n span n ~ ng set
(t e the tree can ta1 ntng all the ".ertlces) In order to .!! pee11f l~
troo whlch 11 a subgraph of tbe cubeB" we can choose any -re-rte.1
of the cube be Ion g1 ng to the tree (there are not more than 211 ,.. a }s
of doln.g so) and a t~ ~ lth 1... ve-rt:..ees (tMre a« u-ot mote thau
"--' 1 "a)s nl do1ng so) For each .edge of a tre£~ 1ts d1rect1on as that
of an edge of the cube B n c.a.n he .s pee l.l]ed ID not .utore than n ~a ).s
Renee ~e ohta1n \.he Y~\llre-d -e-stlma\e
n
-3
2
m-3
4 526 4.5 29
3
4 5 31 Let6 (n) beasubmult1ple ol~plutor'9truthp (G)~ 1

--
JD VIeW of lDeQUahty (1) \\e have 6 (11} ~ -p (It) -+ 0 a..9 11 --t- 00
Hence ., e oh ta1 n the resu It
4 S 32. {2) Z (r (K1 J t1 t" t, t, t.J- }
2 31
{t~ + t2)(~f-r-
3ltf z + 2t.. )
4 5 33 (l) Let us coos1der an arb1trary permuta.llOD a 1a.r 4n
of numben of the set {I 2 n} and arrange parentheses 1n
1t so as to obt.a1n a subst.&tutlon \\1th a e}chc structure {J) -
(J/ I• J11 ) Ftnl v. e have 1 cycles of length t then 1t Cl e)es
o length 2 and so on (1 e the suhst1tUt1on bas the follo¥t"log form
(at) (aJ (aJ 1) (aJ 1 + 1 aJl..l..t) ) Let u.s now a.ssume that t •o
suhsll 1u t1 on.s .n1 and ~ 1 W1 t h a C) clic s true tur.P (J) are construe ted
from t~o d1fierent p.ermut.a11ons of eleme-nts of the set {t 2
nl 1n tbe wa.y d~tl~ aoo\e \Vhm d~!- ~._ eGlru:.lde. With ,z,
The co lllCld ence J.s poss1 ble for t \\ o reasons (1) 1dent ca I Cl c. Ies 1n
the su b.!i t1 t u 11 ons ""1 1 and , 1 are 1n d l fl eren t pos1 t1 ons (2) although
the C}-tles are 1d..:nt1c.al (as c)t1H of a su~\\~Ut1\)n) 1n the above
cons true t1 on they start "'"1 th the d 1fleren t IfIemen ts (.say t 23 and
231) The 6r.st reason lea.ds to a repet1t1on of the same suhst1tutJon
n n
I r J~t• tImes "hile the ~ond reason t 0 a :repe ll tiOD IJ k .r.: ll mes
1
l-1
k~t
the reasons bemg mutually Jndependent \i)
The proof can he
c.arr1ed out by JndutllOD on n USlDg the re at1ons from parts (I)
and (2) of th1s proh1em
4 5 M Use the fol)o-a.lng obVIOUS Jact a e):cle Is an even sub
.st1tUt1on I.( and only 1f 1t.s length IS odd
4 5 36 TJu.s (oil O"V.+s from the Pol ra theorem
' 5 37 Interpret the coefficieDl.5 of the bmom.Ial t z approp- +
tultely aDd use the Polya theorem
CHAPTER FIVE ~43
4.5.40. Associate each graph with a tuple of connected graphs,

i.e. the tuple of all its connective components. Then apply the
Polya theorem.
4.5.42. Any tournament is uniquely determined by its con-
densation {with numbered vertices) and the tuple of strong com-
ponents. The numbering of the vertices is required only for asso-
ciating them with the corresponding strong components of the
tournament.
4.6.
4.6.3. 14. 4.6.4. See Problem 4.1i.2. 4.6.6. This follows from
~
the fact that any function f (x 11 ) has a d.n.f. whose complexity

does not exceed nzn-1 • 4.6.16. See Problem 4.6.15.
4.6.20. This follows from the fact that the complexity of a
scheme dual to a given one is equal to the complexity of the initial
scheme, and from Problem 4.6.18.
4.6.21. If a scheme contains not more than seven contacts, it
can be plotted on a plane so than the addition of an edge between
the poles leaves it plane. Consequently, there exists a scheme dual
-
to it. The substitution in the latter scheme xcr for all contacts of
the type x 0 leads to a scheme representing a negation of the func-
tion which is represented by the initial scheme. •
4.6.22. Let us choose an arbitrary contact of a .scheme without
repetition representing the Boolean function j and consider a chain
passing through the contact. The values of variables differing from
those in the chosen chain can be fixed in such a way that the contacts
not contained in the chain will be disconnected. The obtained
scheme represents an e.c. depending on the chosen variable. Thus,
it turns out that a certain component of the function f, and hence
the function f itself, essentially depends on the chosen variable.
4.6.25. The scheme ~ in Fig. 21 has the sets {x, y }, {r, zv},
{x! r, v, ::} and {x, w, t, u} among its sections. Then if there
:x•sts a scheme without repetition ~ 1 representing the function f*.
It contains chains with conductivities xy, rw, xrvz, xwtu. \Vi thou l
any loss of generality, we can assume that the contact x adjoins the
po~e.a of ~he network ~ •. Then either contact r or contact w also
ad]oms th1s pole. In the former case, ~ 1 does not contain the chain
xrvz, and in the latter, xwtu.
~.6.26, No, it is not. l\fal>e nse of the fact that j (x 1 , • • • , xn) =
• -
f (xl, ••• , x 11 ) and the result of Problem 4.6.25.
CHAPTER FIVE
5.1.
5.1.1. Let p (v, w) ~ 2t for some v and win C. Then S!1 {v) n
S'~ (w) =/= >6. Consequently, any mapping ¢: Bn ->- C such that
S? (u) s ljl-1 (u) for any u E C is not single-valued.
SOLUTIONS ANSWERS .AND KINTS
5 I 2 (1) GeneraJ)y 1t IS not true (2) It JS true (3) Generally

J t 1s not true
...... -..
5 J 3 The .sets C0 :;::: {ct E c t~ ct U 1s even) and C1 {a: € C
i"W
il a. U1s odd ) are codes detec t1ng a s ngle error At least one of them
conta1n3 not ~~~ than half the number of words 1n C
5 t 4 (1) It detects one and corrects 0 mLstakes (l) It detects
n t and corrects (n r
1)/2} mJstakes
.5 i 7 (2) t 6 5 I 8 2" 1 5 I 9 2 5 f f 2 No there doe~ not
5 t f3 Let for a certain n > 1 there ex1st a densely packed
(n 3) code Thf'n m VJf'W of 5 t U the number t (:) for

[}
thts n 1.s a power o[ two Consequently for a certaJD k the equahty
(n + n -- 6) == 3 X 2a 1s valid Then e1ther n -t 1 lS
t) (n! -
a pov.er of two or n + 1 has the form 3 X 2 for a certaLn natural r
If n + t = 2r then n.'Z -- n + 6 = 3 X 2)l r SubstitUting n .......
2r t 1nto the last equalitY v.e obtaJn 21 3 -- 3 X 2 3 t -r
3 X 2~ r 3 For r > 3 the left hand .s~de IS an odd number exceed
1ng 3 \\rule the r1ght hand s1de 1.s e1ther an even number or 3 The
second case c.a n be cons1 d ered s1 m1larl y
~ """¥
.5,...,.t 14 Lel the vert1ees a --- {a 1 an) ~-- (~ 1 ~n)

and -y == (y1 Yn) form an {n d} code Wtthout lo55 of gener
""""' ..., .........
~
al1ty a -:::::: 0 ilnd p (a ti) == d \Ve put A a"~ = {t ~~ -- a Yt

""""
"t J o "t € {0 t l Let us ~cnst. der a vert ex 6 su-e h that .a 1 ~ 0 E<:1r
........ /'W
;tJ
~~ E A 11 and fJ -- 1 fur ' f A 11 \Ve have- p (a f.) U 6 H-

"'"¥ ~ ~
> >
l"w
l.Aot U At~ U A aa ~ l A o,1 UA 10 ( = V (~ y} d P (6 (3.) ~

,...,. .-.1
~ l"w
1 r ~~ # d p (o y) ~ H ~ n d
5 116 \\ ilhout Joss ol generaht) we assume that OE C For
...... ,..,
-
~ny tt EB_d+ 1 there e~1sts a un.1que ver'l.~x 1 E C such tha\.
f!""w' ,....
p (a y} ~ d S Lnce the we1g ht of any non 1.-e ro .code ~ o td 1s not

~,...., .,... ........ ,.... ........
less th,o.n 2d+ I -y EB~d+t Let A (y) {a a EBd+t p (a y) ~ d}
....... ,..,., ,.,
Then U A (i)~B~+I and for any 1"1 y~ 1n cnBZd+t
-,... n
-ve en n2 d+ 1
,.., rAW
we have A (1' 1) nA {'Vt) =0

5 J t 7 The <::Ode d1st.anee of any equ~d 1stan t code of pOll er
h1ghcr than 2: IS even
5 1 20 (t) In each of tbe face~ B'.} +d 01 d and B~ +d 1.! d
\\ e construct cod(\~, C0 and C1 of pov.. er m (n d) Then C0 UC1 1s
an <~+ d d) code of p<)\\er an. <rt d) (3,} R1nt 1( C Ls an
( d) COde JD BTC for (n-1}-dlmenslODa} face g the ,set Cnt IS
an (n.-. 1 d) code
5 J 23 Let n < 2d Bnd let C he an arbJtrary {n d} code o(
po\\ ~r m ~ 2 \\ e ctrns.t tut t a m a\11 x \f "hooe l nes ar-e code words
Let R be the sum oJ pa1r'\\ 1se dlt:taoc cs bet ween (unordered} t-au-s
CHAPTER FIVr:: 345
of code words. On the one hand, R~ (~)d. On the other hand,
"
-
R = '\; hi (m- h 1 ), where hi is the number of unities in the i-th
1=1
, m (m- 1)
row of the matri:\ M. Since h (m-h) ~m-/4, we have '> d~
~
nz2 2d
n
4
. Hence m ~ ·>a
- -II
•
5,1.25. (3) Let C be the ma.\.imum (n, k, d)-code. Then the set
C· = {a; ;; E C, ai = 0} is an (n - 1, k, d)-code. The number of
I ~ ~
paus of the form (1, ~) such that 1 ~ i ~ n, ~ E Ci, does not ex-
ceed 11 max 1 C; I ~ 11m {n - 1, k, d). On the other hand, each
~
vector a E C generates n - k such pairs. Hence

(n- k) m (u, k, d) ~ nrn {n - 1, k, d).
~ I'V -
e
roo, .-..J
5.1,26. Let a E 8", C s 8" and let C~ = {'\': 1' = a ~.

a:
~ ~ ~
~ E C}. Then, if Cis an (n, d)-code and p (a, ~) < d, C~ n C~ =

a: fl
~ ,....,; r>oi ,..._, ~ ,....., ,...., ~
0 . Indeed, let 'I' E C~, o E C~ and o' = oe ~. y' = 'I' ffi a.

a: fl
"' r"oJ .-..J ,....., ,..._, I'V ~ I"'>J "' -
Then p (y, o) = p (a e y', ~ e o') = 11 a e y' e ~ e o' 11 =

,..., ,...., ,..., I"V ,...., - ,...., ,..._,
II (ct e ~) e (y' EB o') 11 =1= o, since otherwise a e ~ = y' e o',

row ,....., ,..._, ,..., "' ,...., t""oJ ,...,
and hence p (a, ~) = p ('I'', o'). But p (ct, j3) < d and p (y', o') ~ d
~ ~
since y', o' E C. Hence follows the statement.

5.1.27. Th1s follows from Problem 5.1.26 if we take into account
that in any (d- !)-dimensional face of the cube B", the pairwise
distances between 'ertices do not exceed d - 1, and the number of
vertices is 2d -I.
5.2.
5.2.2. A linearly independent system is, for example, 8~. If .

s vectors in 8" are linearly independent, all their combinations of
the form (1) are pairwise different. If there e:\isted a subset con-
sisting of 11 + 1 linearly independent vectors, we would have the
equality 1 8n 1 = 211+1.
5._2.4. This follows from the fact that an (n, k)-code is a k-di-
~ens!Onal subspace of B 11 , i.e. the maximum number of linearly
mdependent vectors in it is k, and any linear combination of code
vectors belongs to the code.
5.2.5. II there is a vector \\ ith an odd weight in the code, half
t he nu~ber of code words have odd weights and the other half
even we1ghts. Otherwise, all the vectors are of an even weight. The
former statement follo\\S from the fact that the number of linear
combinations containing the given vecto~ wit!\ the odd weight is
346 SOLUTIONS Al\SWERS A 'D JIINTS
equal to the number of eamh1natloo.s whlth do not eonta1n ~t All

the combtna ti orts are dt Vl ded tnt o pa1rs !D wb 1ch ex ac tl y one com
b1nation has an odd wetght
5 2 6 A non zero vee tor Ill B"/1 ea n be chosen tn Wt--- 1 ways-
If i 11 near Iy 1nd-e~nden t v~tots s. r~ eb.os!n the c..orte!pond ~ llg t=ru b-
space ha.s 2 i ,. ec t ora Any vee tor 1n the com plernen t to this spa. ce .con
st1tutes w1th 1 chosen vectors a l1nearJy Jndependent set and any
vector 1n the subspace can be :represented as a l1near eombinatron
of the chosen \ectots Thus the (t + 1} th veetor can he th.osen ln
211 - 2l ways
S 2 7 See Problem 2 8 5 2 8 2" 1 5 2 10 It 1s true
5 2- t1 (1) m (C (H}) ~ 8 d {C {I/)) = 2 {5} m (C {H)) ~ 32
d (C (Hl) = 7
5 2 ~4 The matr1.x M (C) generat1ng the (n k) code C can b9
reduced to the form (I k P) by sulls tl t u t1ng I or rows tbe1r lln ear com
f't.l
binatlons and tran.spos1ng the columns Il a vector a 1s orthogonal to ,...., ~"fa,!
each row of the matrJx .4-/(C} then a vector ~ obtalned from ct by an

appr opna.te tra.nsposJtton of eoord1nates Is orthogonal to each row
o( the ma tr 1x. Jl (C) and v1c e versa One- of gene rat lAg m a tr LCes a f the
code C dual to the code C* generated by the matriX (l,P) has the
form (Pf I,., kl I e c• IS an (n n - k} code: Hence the code c•
dus.l to the code C J.S also an (n n -- k) code
5 2 I 5 The pi'-nblem 1.2 s1m1la t to 5 2 5
5 .2 16 The code d 1stance d 1s obY1ously not smaller than ,.., the ~
mJnl mum we1ght of the non 2ero code vector If p (et P} < d
"¥ l"'tJ
lor some code we rds. a. and ~ "e arr1 ve at s. eontrad tctton to tb.e
Attl ,...,
wnd1t1<ln of the prahlem s1nce « m ~ tS alro a code -v~tot and

f"toJ' ~ .........
p (O a e ~) < d
5 2 17 The so1utJon folJows from Problems 5 2 15 and 5 2 16
5 2 18 The number of unities 1n Jhe matr1x of the code C JS
1
equal to 1C J n On the other hand th1~ number Js not smaller
2
than d (~ C ~ - ,...,t l
5 2. 20 Let a= (t:Lj. a.-n ~ be a we lg ht d vector orth.ogona l to
the matrtX H \Ve .shall denote by hi the'
,.,_
th columo of the matriX II
It follows from the orthogonality oi a to each row of the matrt~ H
1L
that E alkl ~ 0 Th1s leads to a hnear dependence bet wren the

--l
rov..s appear In the l1Dear combJnatlOn Consequently to
h 1 wh1ch
,.,.,
each vector a of weJgbt d 1n tbe null space of the mat.r1x H there
t.orrespond d l1 n-ear1y d-epend e-1\ t eol umn s of lhts ma tr~ x ThU! 1f
each d- 1 columns of the matrrx Hare J1ne.arly 1ndep-endent the
mtn1mum we ght of the code vector Js not smaller than d and con
versely Ji there ex1.sts a set of d - 1ltnearly dependent rows there
-extsts a v~tot of we1ght s.maller than d 1n the otth.ogonat sub~pa.e~
5 2 21 The probJem Is a. corollary of Problem 5 2 20
5 2 24 (i) It folJows from 5 2 18 that g (9 S) ~ 10 \Ve can
CHAPTER FIVE 347
easily construct a linear (9, 5 )-code of power 4. Therefore,

g (9 5) E {4, 8}. Let us assume that C is n linear (9, 5)-code of
pow~r 8. Then four code vectors have an odd weight. No two of
these four vectors lie outside B~. But among three vectors in B~,
there are two the distance between which does not exceed 4.
5.3.
5.3.4. Consider the code {a, aabb, bb }.
5.3.8. See Problem 5.3. 7.
5,3,10. The number of words of length smaller than l in the
1- I
"" . kl-1 <
k-letter alphabet is LJ k• = k-t . Hence if kl -1
i=O
m (k - 1), there exists in M a word of a length not smaller than
log" (1 + m (k - 1)).
5.3.1 t. (2) For any divisible code, there exists a prefix code
with the same tuple of code word lengths (see Ref. 6, Part 5).
5.3.18. (1) Prove by induction on m.
5.3.19. This follows from 5.3.18.
5,3,22. (1) Let C = {w1 , • • • , wm} be a dual prefix code, and let
the maximum length of a code word be n. Let w = ct1 • • • ct.Mwl be a
~
word in C. Let'\'= (y1 , • • • , 1'n) be a vector with')';= cti, i=1, f... (w)
and with blanks in the remaining coordinates. Then the vector
~
y is a code of an (n- f... (w))-dimensional face of the cube B. It

follows from the prefix form of the code that faces corresponding
m
to different code words do not intersect. Hence 2J 2n- 'J..(w 1> ~ 2n.
i=l ~
(2) It follows from the completeness of the code that for any ct E B 11
~
there exists a code word w which is the prefix of ct. This means that
~
the tuple ct is contained in the face corresponding to the word

w. It follows from the prefix form of the code that the faces corre-
sponding to different code words are disjoint. Thus, the set of faces
c.OJ;responding to the words of the complete prefix code define a par-
titiOn of the cube Bn into disjoint faces. This leads to the required
equality. (3) The proof is similar to that in Problem 1.1.34.
5.3.23. (1) It is true. The statement follows from the fact that
any optimal code is a complete prefix code.
5.3,24. Prove by induction on m.
5.3.25. A prefix code with code word lengths f...h ••. , f...m
m
exists if and only if 2J 2 -i.; ~ 1 (see Problem 5.3.22). There-
m 1=!
fore, Lm =min 2J f...;, where the minimum is taken over all sets
i=l
P•h · · ., f...m} of natural numbers satisfying the condition
m -i.. m •
.~ 2 1
:;;;, 1. The minimum 2j f...i is attained on the sets {f...h ••• ,
1==1 i=l
~m} such that If...i- f...i 1~ 1, i ~ i, j ~ m. Indeed, if there exist
"'i and 'i.j such that f...; - f...j;;;: 2, after the · substitution
SOL UTIOI\S ANSWERS Al\D lli.NTS
m
of 1t-- t for .l.t and '-.j +~ for l 1 the quant' ty ~ ~
~~
does not change and the condltJon 2J 2

"" l
' ~ t rema1ns lul:fil
~1
led Let (A.i )..,.,._} he a set o[ numbers such thal ).t ~ ~ for
, t m--- r .a.,:::::: A.+ 1 for .m --t < t ~ m where ;.. IS all tnteger
m
Then 2J At:;:;:mA,+r and the condJt1on assumes the form m2 1 --
' t
r2 "' 1~ J It lollol\S from the condttJon that A> (log, m)- I
m
and hence ~ 1t ~ m.{log, m.l
' 1
5 3 27 (t} For m === 3 a:ssUDl.tng that p 1 ~ Ps ~ PJ > 0 \\e
have L (P) ::;:: P 1 +
2pt. + 2p 3 :=; I P• + +
p 1 > t The state-
rn en t then f o} 1o ws from t be fae t th.a t upon e.1 pans1 on opera t1 on the
va}ue ol tl1e code does not decrease (2) For g1 ven t > 0 aod 111 ~ 2
tOll$id~r the d1~tt1but1()n
P= ( 1-6 ~~~~i m.:t ) where 6

CllhPTER S\X
6f
6 1 1 (i) No 1\ lS not BlDte the output Slgnal at tbe 1nstant .t
dept!nds on th~ 1nput s1gnal at the subt;equent \nstant (3} Ye~ l.\ 1s
6 f 2 (3} No lf. J8 not (4) Yes Jt 1s
6 I 3 (2} Ye5 lt Js (4) Yes Jt IS
6 f 4 (1) Yes 1 t 15 (2) No 1 t IS not ( 4) No It ts not
6 t 5 ll) lt IS Itnposslhle to supplement the definJt1on to a de
term1n 1~t1-e i unc tl ()J:l (2} T be fun c.t1 on 1p ad m1 ts tbe su p~letn® t to
the detintttou af determtnlstlc funettan (4} The supp(emeut to
a d e.fin1 tJ on 1s 1m po.ss1 h le
6 f 6 (1} It JS suffic1ent to cons1der the funct1on
6 I 9 (2) The weJght of the funet~on Q' ts equal to 2 6 I 10 (f)

The vperators are equ1valent (2) 1\o they are .not 6 1 J1 {2) Yes
1t 1s For example -<pr_ = r-, ~
<p for.; ....... 1 and ~· $:;; 0 (3) "'Jo 1t 1S

):
not
6 i 12. (1} Y" lt 13 Tht!- "t\~lgh\ ts 4 (3} Yes 1t. lS 'I'ha -we1gb\
Ul 7
CHAPTER SIX 34!l
,....
6.1. f 3. (2) If r = 3, for a suitable function we can take q> (x IJ)) =
(3/4 ). .
6.1.16. For r = 3, for such a functwn we can take
y (1) = X (1),
<p: { y (t) = (:r (t) ffi X (t - 1)) y (t - 1), t ~ 2.
6 1 18. See hint to Problem 6.1.6 (1).

6:1:20. (2) In order to solve the problem, it is sufficient to con-
sider the function
,....
q> (:r (") = OOx (2) x (3) ..• x (t) . . . .
6.1.22. It is a generated operator.

6.1.24. (2) If I A I = 1 and I B I ~ 2, then I <l> A. B I = c.
II I B I = 1 and I A I ~ 1. then I <l>A. n I = 1.
6. 1.25. (1) If I A I > 1, the power of each class K 1 (s) is c .
2) The number of different classes is 1 A 1s.
A
6.1.26. (2) If I B I = 1 and I A I ~ 1, then I !ll.-~.. n I = 1.
6.2.
6.2.1. {4) The canonical equations have the following form:

- -
y (t) = F (x (t) V x (t)) q (t - 1),
q (t) = (x (t) V 'i (t)) q(t - 1),

q (0) = 0.
The input signal x (t) can be omitted since the function <p depends
on it in essentially.
6.2.2. (1) The operator definition can he supplemented so that
we obtain a h.d.-operator of weight 4, described by the following
canonical equations:
Y (t) = X
- (t) ql (t -1) q2 (t - f) ffi ql (t - f) ffi q2 (t - 1),
ql (t) = m (x (t), q1 (t- 1), q2 (t - 1)) ffi x (t} ffi q1 (t - 1),
x
q2 (t) = (t) (~ (t - 1) Vq2 (t-1)) ffi q1 (t-1) ffi q2 (t - 1},
ql (0) = q2 (0) = 0.
. 6.2.3 •. (4) The weight of the operator is 3 (the states correspond-

mg to pairs (q1, q.) = (1, 0) and {q~> q.,) = (1, 1) can be identi-
fied). - -
6.2.4•.To each function of weight w (in the indicated s~t. of
b.d.-funchons) there corresponds a canonical array contammg
11
+ 1 "i1;1put" columns (includin? the column describing the state of
t 11e f~~chon) an~ m + 1 "output' columns (taking into account the
~rans1t10n funct10n). This array contains wk" "input" tuples. At the
output" of the array, there must he some tuples of the set con-
350 $DLUTION.9 ANSWERS A.l\D HINTS
tato•ng wkm elements (output tuples} In order to .obt.aan the maJor

ant, we must assume that each 1nput tuple can be assoc1ated wtth
any "output" tup}e
6 2 8 (2) The canon1ca.l equatJons of the operator~ ha\e the
form
!I (t} =- q (t - t)
q (t) =q (t -- t l
9 (0) == 0
1 e 1.p J! an autonomous operator
6 2 9 (11 The csnan.•cal equa t1on.s ol the operator obtatned
from the opera tor 1p by 1n trod uc1 ng the feed ba.c k 1n vart a bles z"
aDd 111 have the form
]I (I) =1 _
q (I) ~ z 1 (t) ~~ (t) 4 (t - 1)
q (0) ~ 0
6 2 tO (f) The operator \\f1ght ts 2
6 2 l ( {I) T b.e folla Wlng b d operator ca. n be ta. ken for ~
-
y (t) .:::::; z 1 (t) V g (t ~ t)
cp q (t) ~ z, (t) E9 z, (t)
q (0) ~ 0
6 2 14 (l) and {2) Con.stder the operator Td{~d) ~.J) 1t I9 exped
1en t to analyze the opera tor ({)d ( tp ~ \l) where c.p -.;. ~ ts. an a petat a r
genera ted h y the .;on.stant 0
6 2 15 (3) The we1ght of the superpOSitlon Js 4 6 2 16 (2} The
operatQr 1s. autonomous {3} The operator IS Dot autonomous
6 2 19 (2) Suc.h a scheme e.x1.sts
6 Z 22. F1t.st draw aIl p oss1 b le three -v etl ex d1gra plu Wl tb
n um.b ered vert 1ces wbleb sat 1sfy cond i t1 on (c) and are such that
the out degree of each vertex JS equal to .2 The d1g1 ts 0 1 and 2
can be t.akm as the numb~n for vert1tes The v-ertex. matke-d by 0
Is convenJent to he taken as the 1Dzt1al one Then the arcs of each
of the constructed d1gr.apbs must he Mloaded 1n all possJhle \\ays
so as to ohtatn ~ioor-e' d1a.grams of rome h d operatots
63
6 3 i (l} J\.f '' a c lO'Sed class (2) 'f h-e set Af 1s n ct a e\02-~d (jlass
6 3 3 Prove hy 1nd uct1on on the number of delays that any
operator ,P 1n 1~ 0 'Pa1"' wh1ch bas one Jnput transforms the ward
I"'VU) v
0 e1t.ber mto tbe word oJ the form !I (1) !J (n 0) l 0)1.\\ or Ulto
the word r,t lhe form y (1} y ( n 0) { 1}(j} where n~ '~ the 1eng t h
of the p repe r1od (whtch depen d.s on the cho1ee of tlie operator ,., )
6 3 7 [1) The systentis Incomplete (2) The system Js complete
6 3 8 {2) '~~ (X) ~Jo(x) (X} tp~ 1 +:r, (XJ X,)}
6 3 10 The statement can be provoo a.s lollo\\s Let A-f be
a cl()sed cJass 1n Cl>o,) w h1 ch d1 ff ers from the en tJ re set tJ)W \ Ve
.assume that M 1s not a precomplete cla.ss and cons1d,fr the total1ty
of all subsets M 1n .:P (k l ' Jl \\ b 1eh sa tJ ~fy the cond 1t1on IM U
AIl'-' # <n (~) Th lS to tal~ ty ~s. na:t em p\.y We eh'01lSe 'u 1\ a snax~
CHAPTER SEVEN 351
mum chain (by inclusion) whose existence can be established either

with the Jwlp of Zermelo's axiom of choice, or by directly taking
A
into account the countability of the set eDen> (by arranging the set
CD(h)"-.,Jf in the form of a natural scale). The union of all the sets
of the chosen maximum chain will be denoted by M 0 • Then M U M 0
~
is a precomplete class in Cl>ch)· In the proof, we essentially use the

A
fact that there exists in the set !l>(h) a finite system complete with
respect to the oct of operations 0 = {0 1 , 0 2 , 0 3 , 0 4 , S}.
6.3.13. This fact can be proved almost in the same way in
which the validity oi a similar statement was established in Boolean
algebra (see Problem 2.1.16).
6,3,14. Yes, there exists such a function (see Problems 6.3.13
and 2.1.25).
6,3,15, Cf. Problem 2.1.17.
6.3.16, For k;;;:. 3, the statement directly follows from the
corresponding result in PI<. In the general case (for k ;;;:. 2}, the
subsets of autonomous operators can be used.
6.3.17. It is countable. 6.3.19. This power is continuaL
6.3.20. It is expedient to use the "power maps", i.e. to compare
the powers of the corresponding sets.
CHAPTER SEVEN
7.1
7.1.1. (1} (a) T (P) = !302!2. (b) The machine T is not appli-
cable to the word POP. (c) T (P) = 10 [01FL
7.1.2. (4) The program of one of possible machines has the form
•
q, q, ,,.
0 qoOS q2 0S q30S
1 q~!R q3 1R qrlR
7.1.3, (2) (a) 120 21q0 01; (c) [10)2 Oq0 P.

_7.1.4, (3) One of Turing's machines transforming the configu-
ration K1 into K 0 is specified by the following program:
qr0q 2 0R q~lq~lL
q1 1q1 1R q5 0q 6 0R
q20q 3 1R q 5 iq,1L
q2 1q 2 1R q6 1q70R
q 3 0q~1L q70q 0 0R
q4 0q,OL q1 1q1 1R
352
7 l S. fl) 1'tns can he ~on~ as. {nllo\\~s eaeb tDmmand of the

form q o:;q 1ps (where a .and ~belong to the external aJphahet A}
1n lh e program of the Tur ng m.achtne s replared h) A r + 1
com rna nd ~ q a.q1 t\ R 9/ifJ ii L (l runs l hroug h tfie al p:ha bet A } wh~re
qJ Js a new state (each state 1s characterJ :zed hy ts Ol\"'11 q )
7 J 7 In order to construct the mach ne Tm t ls su rficJeO l to
add m ltd d1 ttonal (new) states q Vm and .supp1crnent
l he program C) f the machtne T for e ):ample \\ 1t h the (o I low 1. n g
eomm~nds q aq aS fJm.DrJ:m.aS Ythere a JS a fixed character
ol the external alphabet
7 t 9 (t) (a} The compm!lt on rl r~ tS tnapphcable to the \\Otd
1 3 0'1 2 (b) T1 T t Is app11cable to the word t40f as a result "e obta n
1ot o-a1"
7 I I 0 (i} (a l This It era tJon 15 1na p~1Ica hie to \\ ords oJ the
form 1'1t (k ~ i) (b) It 15 Jnapphc~blr to 'A-Ords of tfle form ·f 3k
(k ~ il ertber {e) ~he 1trrn \lOJJ 15 :apphcrthle to any "ord ol the
form t~h+l! (k ~ t} As a result \\e ohta1n the word t
7 j 1 t (1) (a) T (P) ==: 10.a1 (h) T (P) = 1ElQt
7 t t 4. {3) One of th~ pos~tblt Tnt ng mach1nes 1~ ~pee,fu~J by
q10qGOR g,Oq 51L

q1 1q 20R q4 1q,1R
q:i;Og 0 fS q,Oq,OL
q~lq 3 0R q 5 tq~1L
q.,Oq,OR q1 0q1 0R
q3 1g.,tR q1 tq,1L
'1 I f 6 (3) An exa mpJe of tile poss1bJe machines ~
qlOqrlt L q3 1q,OR
q 1 tg~OR fltOg,OS
qiJ:OqgOS q,tq~OR
q2 1q30R qs0q1 1R
q3 0q 3 0S q,&tq\IS
7 1 11 (i) j (.:t) :r: 4- 1. J {x y} :z + g + 2

7 I 18 If we assume as usual that machines start to operate
n the .state q1 and the extreme Jeft unJ ty of the code ol the number
;r 1s scanne-d at lb-e 1n1t al lnstan\ \he mathlnts m~nt\oned 1n the-
problem can compute only one of the foJlo\\ Jng three funcuons
z x - i and a functlon defined nowhere
7 i 19 Ye~ \\ l s true-
7 f 20 (1) For a fixed (fin1te!) set of states there ext ~ts only
a fi.n1te number of pairw2:5e non equJvalent Tur1ng machines {w1tb
a g1v-en external alphab~t} (2} T~re ex\sts an t su-c-h that f-er any
i ~ 1 the subset of a}l luneuons of n varJables tn 'I conta n' not
more than l elements
72.
72 t (2} .:r1 +z 1 - sgn z 1
CHAPTER SEVEN 353
•
7.2.2. (1) At first, we can prove the primitive recursion of the

functions x 1 ..!... x 2 and .x2 and the~ B:P.PlY the s~perposition ope.ra-
tion. The "direct proof" of the primitive recursiOn of the function
g (x) = x~ is as follows:
g (0} = 0,
{ g (x + 1) = h 1 (x, g (x)) = x 2 + +
2x 1,
i.e. hi (x, y} = 2x + y + i;
h1 (x, 0) = 2x + 1 = Ct (x),
{ h (x,
1 y + 1) = s (h1 (x, y)} = (2x + y + 1) + 1;
g 1 (0) = 1,
{ gl (x + 1) = h 2 (x, gl (x)) = 2x + 3,
i.e. h2 (x, y} = y + 2;
h 2 (x, 0) = 2,
{ h (x, y
2 + 1) = s (h 2 (x, y)} = (y + 2) + 1.
•
7.2.5. I (x, y) = sgn x· g 1 (y) + g 2 (x .!.. 1) X sgn x X sgn Y+
g1 (x-'- 1, y ·-'- 1) X sgn x X sgn y. •
7.2. 7. (2) flx 1 ([x 1/2]) = 2xt. (4) !ln (xl ...!- x 2 ) = (x1 + x 2 ) X
sgn xh p.\• (x1 .!.. x 2) = x 1 - x 2.
7.2.8. (4) f (x1 , x 2) = x 1 (1 .!.. x 2).
7.2.9. No, it is not correct. 7.2.10, (1) No, it cannot. (2) 1 +
sgn x.
7.2.12. All these classes are countably infinite.
7.2.17. No, not always: the function / 2 (x, y) can be identically
equal to 0.
7,2,18. (1) No, not always.
7.2.19. (3) This relation is valid. Hint. This function resulting
from the minimization operation is 2x..:... 1. (4) Yes, it is. Hint.
This function resulting from the minimization operation is
7,2,20. (1) Yes, it can. (2) This statement is false for any func-
tion 11 in Kg.r"-..Kpr.r· (3) The statement is correct for some func-
tions It and / 2 in the set Kg.r "-._Kpr.r·
. 7.2.21. (1) No, they are not always satisfied. (2) Thi~ inclusion
IS f~lse for any function f (x) in Kg.r"-._Kpr.r· (3) This relation is
vahd for any !unction f (x} in Kg.r "'-.Ifpr.r·
7,2,24. (2) No, it cannot. 7.2.26. Yes, it is true.
7.3
. 7.3.1. (a) This follows from the fact that there exists a function
m Kpr.r which assumes all the values.
7.3.2. Any primitive recursive function can be associated with
an infinite set of terms reflecting the way of obtaining the function
f rom the simplest ones.
23-0636
SOLUTIONS ANS\VEHS A'D Hl!"r.TS
1. 3~:3~ If there f x 1sts a parh aU y ret:ursJ \Pt! un 1vena.J f unc t1 on

F (.r 0 : 1 , • :1' 11 }" 1t t9. defmed ~'eeyv.h'ere Then the funtllon
J (r• .r1, .:r 2
t , +
.r12 ) f 1s general recurstve and has a num
hfr y 1n the number1ng e~spond1ng to the unt\etsal fullctaon
F<n+1' But an thts case F (y., ~"t t til = F (y, y., y) ....L l
7~ 3 6. Carry aut the proof by -"dJS.gona.lt za t1 on 7, 3.1 (t) ( 4}
~ o, Jt JS Dot (5) Y est 1t Js 1 .. 3 8 .. The solulll>D 1S s1mtlar to that ol
Problem 7 a 3
7 . . 3 li. Coustder the fW1ct1on
tf h (.r) = o,
If h (~) == 1
7.3.12 Consider the sequence a~u a 1 , , .such that a 1 = 1
1( ~~ t') lS defined and a, .=;;: 0 1f {}"
1 (I) IS no' defmed, , = 0 1
7 .. 3~13 .. {1) Fo:r any autonomous .fin1te automaton~ tts output
seq u ~nee J.s qnas1 p er1od 1c
1. 34115.. For example. the 1n\ et"Slon of a word
7 3 21. (2} Tlus fol1 O\VS !rom the f acl that there cx1s t not more
• (P)
than c r dJfferent configuratiOns of Jength 'r (P) lor a certatn c
l\hJrh dep~nds on tlJ~ a1phabet of stAtes: and on the e-xternal alpha
bet of thtt machtne T ll a eonflgurat•on 1s repeated 1 4~P~tates an
1n.fi.:::u teJy long t1me
CHAPTER EIGHT
B.t.
8.1.1 (J) The number of \\~fS IS (20) 1 == 20 X f 9 X i8 The
t1ckets are not equrvalrnt The brst ticket can be dt.strJhuted In
20 ., a)~ tbe stcond m i 9 ,._a yil, and the \h1rd t.n i 8 ,. a~s (2) 201
(3} c (20t 3)
8.1.2. (1} 91 (2) C (9,. 3) X C (6 3) The first rank can be
the sen 1n C ~g '3) 'it a} s, and t b~n lhe set ond 1n C (S, 3) ?rays
U~ung the m ul t1 p1 Jcat 1on ru Itt, -a e obta 1n the re.sul t
8,1 .. 3. (1) EaCh term of a pern.Jutatton \\lth repet1l1ons can he
c.bos.en 1nde. pendently ln n \\a f!: Usl.ng the mult1 plteatlQn rute t \ \ e
A
o b tal n P ( n, r} = n." (2) Fra m. eac b. ( n., r) comb1nattoD. Wl thout

re pe t1 t1 on" \\ e ean oh ta zn rl dJfler.en t ( n. r) perm uta t1 ons, and each
(n., r) permvtat•on can he obta1ned 1n tbJ.s t-..a) Hence rl C (n 1 t) =
P (11, r} S1nee (~e Example 8) P (n r} = (n},~ Vte ha'e C (n r-} =
(n)/rt = {; ) (3} Each (n, r) cl)mbtnabon A 'ft1.lb. repe.tLhons
com posed a( elements of the set U == {a1 a 11 ) we assoc~.a te
.,
#¥
Wilh. .a vector ~ (A} of length n + r - t eonstst1ug a( r- unttle! a.a.d

n- ~ zeros and 5uch that the number of zeros between the (i-t) th
aDd 1 th un1t1es 1s equal to the number ol elements a~, 10 the tom
h1na t 1on A~ t =2 ~ n... 'P! htl e the number of z.eros he.fou lh e
fir.st. W11ty (after th.e (n.- t) th UDilY) 1s equal to the number of
CHAPTER EIGHT 355
elements a1 (resp. elements a11 ) constituting the comhina.tion. A.

There is a one-to-one correspondence between the combmatxons
and the vectors. On the other hand, the number of vectors with
n- 1 unities and r zeros is C (n +
r - 1, n - 1) since each such
vector can he put in one-to-one correspondence with a combination
of 11 + r - 1 elements taken 11 - 1 at a time. Therefore, taking
into account the result of the previous problem, we find that
.C (n, r) =
{ n+r-1)
n -1 ·
/l
••• kn. (3) •
r
8.1.5. (1) 2m11, (2) (2m) 11 ,
8.1.6.
m)r ns . The multiplication rule is used.
r r-h
8.1.7. (1) (n-IX-~lr-11-k· At first we choose the
h k
r
places for the letter a in ways , then we choose the places
h
r-h
for the letter b in ways. We still have r-h-k
k
places the first of which can he occupied by one of the remaining
n - IX - ~ letters, the second by one of 11 - IX - ~ - 1 letters,
and so on, (n-IX- ~lr-h-h ways in all.
8.1.8. (1) 4 (n - 4). The suite can he chosen in four ways, after
which the smallest value of a card can be chosen in n - 4 ways.
(2) 4n (11- i). The number of four cards can be chosen in n ways,
after which the remaining card can be chosen in 4 (n - i) ways.
(3) 12n (n- 1). (4) 4 ( ~ ) • (5) 45 (n - 4). (6) 4n (
411
;
4
) •
(7) f2 ( ~ r II+ 4( ; ) /!3,
8.1.9. (1) 147. The number which appears on the two chips can
he chosen in seven ways. It may turn out that this number appears
· on a chip twice. Then the second chip can be chosen in six ways.
Otherwise, we must choose two different numoers from the six
numbers that appear on the chips. The number of ways in which
this can be done is equal to(~). Thus, 7 X ( 6 + (~)) = 147.
(2) 26. The number of occurrences of identical faces is 6, and the
number of occurrences of pairwise different faces is ( ~} = 20.
(3) 300 + 300~ + 300
A newlyborn baby can be given one, two
3•
~r three names. The number of ways in which k names can be given
IS 3Q01t.
8.1.10, (1) { ~= ~ ). We associate n =:=each division

n1 + · · · + nit with a binary vector with k - 1 unities and
23*
SOLUTIONS ANSW£RS Al\0 UINTS
n - k zeros as follows the number of zeros preced1ng the first

un1ly 1s n 1 -- 1., the number nf 1-e~as lw:tween the ' tb. and (i + 1} \h
zero i~ ni .. 1 - t~ i 1, =, k - 2 the number of zeros after the
(A: -- i) tn unl ty ls n~r. -- t Thl s is a one to-one eorres pond enc e
The n um bPr tJ! such bma-ry "\ ect.on 13 G=!) {2) (" } t 2
Htnt. The rolutlGn should be teduced to the ptohlem af dec:nmpo

s1 tt on of n 1n to three add P11 d.s af te.r wb1 cb the resu 1t of Prob Iem 8 t 4
should he apphed (3) ! (nt
2
) + ~ (ln/2j +
1} The
representation of the DuDlhet 7" JD the form of three mult1phers
corresponds tG the repres-entat~on of n. 1n th~ form af the sum -G{ thr~
non negat1ve addends n ::=: n 1 + n 1 + n 3 Among these addends
two can he 1d en t1 c.a l In this case t t be number ol d 1.s ordered
partltlons IS rn/2] + 1 When a]l the addends are different the
number of ordered part 1t1 on.s 1s s1x t 1ntes as large as the
D'tu:abel' oJ dJsOrdt>red parti\JODs, wbtcb 1s equal to ~rm ~) { ( -
3 ( [ -; • ] + 1} ). U.smg the summa. bon rule, we obtc.un the

requ1red result
8 :1.11 (f) The ;llnswer 1s
n-- k(m--1)
k
+1) ~
Let ct= (ct 1 ,
ctn+AJ he a tuple of 2eros and un1t1est 1n v.-h1~h there ........ ,...,

are at least m
zeros bet ween two con sec u t1 ve UDI t1 es Let ~ (a) == ( P1 ~
+
1"¥
~n-A (m - t) 1) be a tup1e obtalned from et by deleting m zeros ,.,; l"'w
folloWing the ilrst., second, , (k - 1)- th ull'-t y The tuple: f\ (a.}

J!'fltl
conta1ns k Ulllttes and ts uniquely determwed from a On the other

....., f"'ffl.. I'V
hand_. the tuple a. is 1J.D Lque ly de term1ned from the tuple IJ (a)
and the number m Thus., the number of 1n1 t1al tuples 1.s equal to
the number of tuples Qf lePgth n - (m - 1} k + 1 contatntng k
u111 tH!S (2) ( n t 9
usoc1a te each n ll m ber A sa tuf ytrt g
) Wt
the cond 1t1 on of the prob1 e.m w1th a b1nary vee tor Wl th n un1 t1 e3
and ten :eros The number of un 1t1 es preced1 ng the .first zero Is
equal to the dtfferen-ce bet\\een n and the number of dtgLts of tbe
nJJmber A The number of un1t1es bet\\een the l th and (l + f) th
7.-etas, 1 ~ i ~ 8 is. equal to the number et d1g1ts ~u~ol to l 1n the
number A, and the number of unit~ es I olJ ow1ng the 111n th zero 1s
equal to the number ol nznes 1D A Then -w-e apply the result of
Enmple 4 far k = 10 {3) { "tk}
Any of the shortest walks
t:o11s1st.s cf n sout.hwatd and 1t W"e.-Stward segm-ents The l e:ngth of
a .segment u equal to the s1de oi the square Sucb a walk can be
a531gned a b1nary vector ol length n + k't JD whtch the l th coor·
dJoa te 1s eq ua I to zero 1f the t th segment 1S d tree ted south wards
and equal to t 1f 1t has a westward d1rett1on The number of such
vectou 1s { "tk)
CHAPTER EIGHT 357
8.1. 12. (1) (i a 1) (1 + +

a 2) ••• (1 an)· Each divisor +
p~' ... p~r, 0 :;:;;; ~~ :;:;;; a 1, i = 1, ... , r, can be put in a one-
to-one correspondence with a vector (~ 1 • • • . , ~rl· Then see Prob-
lem 8.1.4. (2) 212r. Each divisor pr• ...
p~r of the number n,
which cannot be divided into the square of an integer, can be put
Ill correspondence with a binary vector (~11 ••• , ~r>· (3) rr (p~h-•
r
h=l
- 1) (Ph - 1)-1 • Hint. Note that after opening the
parentheses in the expression (1 Pt p~') ... (1 + + ... + +
Pr + ... + p~r) each divisor appears exactly once as a summand.
)I ) (n)h rd (n)n-h ( 11 )
8.1. 13• <1) ( k = k! = k! (n-k)! = (n-k)! = n-k •
11
Another proof. The number is equal to the number of ways
k
in which a k-elemcnt subset can be chosen from U={a 1 , •• ,,an}·
But t:>ach k-element subset we can put in a one-to-one corre-
spondence with its complement in an (n-k)-element set U.
II k nlkl rd(n-r)l ·
2
( ) k r = kl (n- k)l rl (k- r)l = (11-- r)l rl (n- k)l (k-r)! =
n-r n-1 n-1 (n-1)1
r k-r
• 3
( ) k + k-1 = kl (n-k-1)1 +
(n-1)! (n-1)1 n n / n
(k-1)1 (n- k)! = kl (n-k)l, (n-k+k)= k , (4) k-r k
=
kl (n-k)! (k)r
(k-r)! (n-k+r)! • (5) Use the fact that
(n-k+r)r
k
n-r-1 n-r n-r-1 n-r-1
k-r
-
k-r
- k-r-1
• Then ~ k-r
-
r=O
k
n-r n-r-1
~
11
k-r
- = k . (6), (7) Directly. (8) Induction
r=O
k-r-1
k
r k+i
on 11. For n=k, we have ~ - -
-1- • Inductive
k k+1
r=k
n+1 n
r r n+1
transition 11 -+ n + 1: ~ =:6 + • By
k k k
r=k r=k
n
r
inductive hypothesis, we have ~ - Hence we ob-
k
r=h
SOLUTIONS .ANSWERS AND J{fVTS
n+t
la1t1. from tb.ls and th~ prev1aus (!qua l1ty ~
2n
{S) \Ve have (2n}t ~ ,, (ttl) 2 Tb~
nu1nbe-r {2n }~ 1S d1'Vlslbleo by
any pr1me number P1 where ,, < p 1 ~ 2n and the number (n1} 1
2n
1s not d1v~sshle by P 1 Consequently 1s d\Vl~lble by Pi
n
8 t U (1} Solut1on We put « 11- 1 -=[ (n_: f}t ] Suppose
that the toeffitlents an 1 «n 2 ct 1 1 have been. deterrn1ned
'then
._ _. [m ~aF1-I (»-1)!-an-t
a.n-(J-t-l)-
{n--- 2)1--
(1r-- i - i}l
CHAPTER EIGHT 359
The uniqueness of the representation will be proved by contra-

diction. Let two vectors a"' (m) (ct1 , ••• , ctn-1 ) and "'
= ~ (m) =
(6 1 , ••• , ~n-tl correspond to a certain m. Let j be the largest of
the subscripts in which the vectors differ from each other. Without
any loss of generality, we can assume that a1 < ~1 . We have
n-1 n-1 j-1
0= 2J ~~il- 2J ct1i! > (~J-aJ) jl - 2J ctiil
i=1 i=1 i=1
j-1
~ jl- ~ i X il > 0.
i=1
We arrive at a contradiction. (2) "' a (4) = (0, 2); "'

a (15) = (1, 1, 2);
~
a (37) = (1, 0, 2, 1). (3) f-l (0, 2, 0, 4) = 100; f-l (0, 2, 1) = 10;
!!. (1, 2, 3, 2) = 72. (4) v (2, 3, 1, 4) = 12; v (3, 5, 2, 1, 4) = 85;
v (1, 4, 3, 5, 2) = 20. (5) Let the number m be specified. We pre-
sent it in the form m = ct 1 X 11 +
ct 2 X 21 + ... +
ctn- 1 X
(n - 1)!, where a 1 :s;; i, 1 :s;; i < n. The construction of the re-
quired permutation 1t is equivalent to that of the vector (1t (1),
n (2), ... , 1t (n)). The coordinates 1t (j) will be given by inrluc-
tion. We put 1t (1) = Ct.n- 1 +
1. If the coordinates 1t (1), ... ,
n (j- 1) arc specified, we put 1t (j) = ctn-J 1 .~ + + (/),
where
s (j) is the number of 1t (k), 1 :s;; k < j, for which 1t (k) < j.
(6) ll 7 = (2, 1, 4, 3); rt 18 = (2, 3, 4, 1); 1t 28 = (2, 1, 4, 5, 3).
"'
8.1..16. (1) The coordinates of the vector ~ (m) = (~ 1 , ••• , ~It)
are defined as follows: ~ 1 is the largest integet• such that
m~{~1 ). If ~1, ••• ,~~have already been specified, then ~i+l
is the maximum integer such that m- ( ~~ ) - ( k ~ ) - ... -
1
( ~~
k-i+1 ) ......_
;;::; ( ~ 1+ 1
k-i )
. Th e ·
umqueness .
tS prove d as m .
Problem (8.1.15(1). (2)(a) f3C19)=(6, 4, 1, 0) since 19=( 1)+

(; )+n )+ ( ~ ); (b) jr (25) = (6, 3, 2); (c) tr (32) = (4, 3, 2, 1).
(3) (a} m=23 since m=( ~ )+( ~ )+( ~ )=23; (b) m=13;
(c) m= 10. (4) First we put in correspondence to the vector
-
1X=(at, ... , ctn) in BJ:
a vector "'
~=(~ 1 , ...• , ~h) in which ~i-f-1
is the number of the coordinate of the (k-j-f-1)-th unity on
the left in the tuple ~. 1 :s;; j .::;;; k. Then we put v ~) = f-l (ii} 1. +
z)+(t)+1=4.
360 SOLutJO\S AJr.SWERS .Afrr.D BI~TS
8 I 17 For n = 1 ~uaiJtr (1) can be \Ulfi.ed directly Let

the: eqnahty he provPd for a certalD n ~ t \\ e have
~+t ft+\ n
~ ('•tt) l~= 3 (( ; ) -L ( k:l )) tl= ~ ( ; )t•+
t~o l~o 1 o
n+l • •
)J { .~ 1 } ,~ = ~ { : } 'k +h ( : ) ~ft+l = u + ,).. +
•~t l-0 ~ 0
1 (1 +t)•== (1 +r)n..'
8 t !B (I) Put I ~ l1n (f) (2) Put t == -f In (f) (3) D1ffer
ent1ate (1) With feC:!Jl«l tor and putt == t (4) Dtllerent1:ate (l) tWJce
and put 1 ~ f (5} Us1ng (f) ud (3) Vi. e ob ta111
n "'
2~ "( ;)+h { :)
·-· Ja-1>
::::=: n-2" + 2" == (n +t} 2n.
{&) Intr-grat:e 1dent1ty {1) vnth ~p~t to 1 between 0 and t {;) lnt.e-.
grate 1dent 1t y (I) WJ th res peet to ., ~t-. een --1 aJJ.d 0 (8) <:any out
mduct1on on n by us1ng Problem 8 ! 13 (2) and 8 t t7 (7) (9) Com
~ the toef.fiCJePb of 1-' on the lef' and r1ght hand s1des of the
1dent1ty (t +
t),. (1 +
t}m = (I t)n•m +
(1fJ) Put m (9) k~11-M lH) Dlvldlng by ( : ) reduce the
2
" :n ,..
problem to (10} (12) \\e ban ~ h ( ; ) ( n-:::- k) - -
l-O I 0
tl tl.l R R
h { ~) ~ (lJ;k)=~ (~)2" ~ ~ {13)~ ( 1).. t{~)$

1~0 r~Q ~~o r-i
n l • A
~ (- urn ( k.;m ) = ~ c- f)lrt ( n-k~ m ) =
m-n ~-D
n.-.l
}.; {~ 1)_11....,. ( : ) (14) The problem IS reduced to (9) by subsll
·-· f'J1
luting n-1: for- /1. (15) Note that :8 (-t)t-n ( : ) ( ~ ) =

A-..:- a
--"
(~ ) ~ (-tP ( m;n)
J~a
CHAPTER EIGHT 361
•
m-1
V=O s=O k
m-1
~ e
2lli(ll!+s-r)
m
,, =Ill
"'
LJ In the latter case, we have
v=O k
used the identity

m-1 ?nin
"'l m "_ { m if 11. is multi pie of m,
-
LJ e - 0 otherwise.
\'=0
8.1.20. (1) The solution is similar to that of

m-1 2nirv 2niv
Problem 8.1.19. (3). We have ~ e- m ( 1 + a:e m r-
• V=O
m -I 2nirv 11 2nivk
-
~e-m~(~) m
-
V=O k=O
n m-1 2ni(k-r)v
~ ~) a;k ~
11
{ e- m = Ill~ (
ms+r
) a;ms+r.
k=O v=O s=O
In the latter case, we have used the fact that the sum
m-1
Lj exp {2nilv/m} is equal to 0 for 1 < l <m and is equal to m
V=O
for 1=0. (2) ~ ((t+"V3t+(1-f~l} 11 ). Use (1) for m=2, r=O,
a=JI"S. <3> ((1+V -2t+t (1+tV--2t+(t- y.::-2)n-
i (i-t{! -2t)/4t/ -2. Use (1) for m=4, r=1, a:=-tf=2·
r-1
2
(4) ( -3) ( ; ) ((1 + V -3)n-r- (1- Y -3) 11-r}/2. Using
2 k+i) =
Problem 8.1.13 (2), we obtain 'V (-1)" 3 ( n ) (
LJ 2k+1 r
k
~LtlTlONS ANSWEJ.S AND IU~TS
r-1 2~+i-r
{ :) 2!<-J)k { 2N~-;:,.) =(-l) 2 (; ) }J (-3) 2 X

~ •
{ 2.1t~-;:_r) Furtber we use (f) for m=2 a:= ]1-3 and r= f-r
8 I .ZI (1J 4 The term of the expanSJon ~ 0 2-'l 2 3f 2 a- Jt);a 1s

rat1onal1f and only 1f k lS even and 20 - k 1s a Ill ult1ple of three~
1 e for k = 2 8 14" 20 (2} 13 (3) 9 (4) 6
8 1 22 ( I) _..safi The general farm ol tbe k tb term of tbe el: pa ns1ou
has the form ( : ) A :II. where A= J (2 ~ 3t) and the upus1on
Ak contains lrom the k tb to (2k) tb powers of t The ptm.er
t"' appears u1 A~ A" A ':1 and A' w 1th tbe coellita en ts
(!)2(-3)1 {;) 2 (-3)1 {!)211(-3}' (:)2'(-3)1
3
re.speetl\ ely The .coeffic1ent of t 1 1s the sum of these quant1t1es

and 18 equal to ---536 (2) -91 (3) 13 (4) 0
S t 23 (~) \\e pul ~~ m--n 'Ihtn th! ldent1ty lS \rans-
{orruoo as follows
or
The latter relation can be proved by 1nducttoo. Assuming that

(0) 0 = 1 v, e can eastlr ver1fy the val1d 1ty of (•) for n ~ 0
\\e aSSUJU~ tbat (•) IS '\'a}Jd for a cert.a1n 11 and proVe ~hal
n+l
~ (n+ l)h = li(t + t) \\e have
£_J tn+t+2)k
A-0
n+i
~ -(n{+t~-~),.""= PJ+~+•
~~o
CHAPTER EIGHT 363
(
m+k-1 ) )
Ill
= ( m+n)
II
= ( m+n
n1
). Similarly,
8.1.24. (1)
(a)ll + (ah-t = (a )It+ k (a)l•-t = (a+ t)k •
kl (k-1)! kl k!
00
(2) The ~cries A (t) = ~ ( ~ ) th is a series of the functiou

h=O
f (t) = (1 + t)" :;iucc fU•> (0)/kl = ( ~ ) . The series A ( t) converges
for 1t 1< 1 and for all a according to the d 'Alembert criterion:
a) j(a) a-k
(kt-1 tit+! k t''=k+t 1• ~a-kt
,!~!k+1 =t<l.
Let the tet·m r,.

(t) be the rl'mainder o[ the serie~ (in Cauchy's
form). We shall demonstrate that n, (t)-. 0 as k-+ oo thus
proving the equality. We have rlt (t) = j(h+l) (9t) (1- 9)hth+ 1Jkl =
(a)h+t(1 +Bt)a-k-l(1- 9)ktk+t /kl=
k
(a-
1 ) tit (1 + St)a-1 { l - 9 ) k
1+9t .
a
Here a-t)
(k tit-+ 0 as k-. oo, the expression a (1 + 9t)a-l is
( 1 _,~~
1
bounded, and 0 < )'' < 1 for o < 9 < 1, 1 t 1 < 1. Hence
it followil that l~~ rJt (t) = n. (3) (~·a)= (- a) ( - a- 1) ••.

II
(-a-k+ 1)/kl=(-1)1< ( a+~- ). 1

(4) ~ (a -:: k) =
k=ll
n
k=O
~ ((a-k+l)-(a-
r+I
k))=(a+l)-(
r+l. r+l
a-n)
r+l ·
(S) Use
~he identity (i+t)a (1+t)b=(l+t)tt+b. (6) It follows from (5)
tf we put b= -1. (7) It follows from (3) and (5). (8)
JM SOLUTIONS ANSWERS Al\D HINTS
By usJng (5) we obtJID
(//'.}
\\e obtain ~ 2
(-1)11 { : ) ( ! r-
(1. +ill) ll v
... /-_23 (10) lt Jolla"~
from (2) that v 5 ---- ( t+ { ) tt2 ~ £..J

~ ( t/2
cJC
k. ) ( Tt )l va
2 ---
2 4
)t;;;;;;;Q.
J. 0
{ ! yJl+' = y:;-; )fa {It} See st 2.:t (I)
8 l 25 {i} {mn.)V{m.'l)" (2.) {.lj-1 ka n k.} (3) Let us

altulale { n two '9tays) tht. number An (k 1 k.} of ordenngs
or "' obJects among wlu ch there are k..1 obJects of t be first k1n d Jc t
ob)Kts of the second klnd ~te a.nd Ie, oblee\3 o! th~ a-th ~nd
F1nt method We firs' choose lt1 places among n for arrangtng the
ohJeets of the lir:st k nd Th1s can be done Ul ( lc~
Then we Jways
ehoose It 1 places for arra ng1ng tbe oh Jec ts of the second Jond Th.t.s
r:a-n be d.nne- 1n ( n ~ k, } wa)'! and so on Us og tbe mul11ph
c.auon rule we lind lhat.
Second method Let us ealculate the- number n of order1Dgs of

p.a1tw1se d 1«trent oh~t.s. \Ve !.hall assume that. th-e cb }et.ts a.u
d1 v1ded 1nto .a groups BO that the I th group con tams k, ObJett.s
(i= 1 j:) The cho1ce ol k.z places lor obJects of tbe first k1nd kt
plates for obJech of tile second Ju:p d e~ cap be made Ill
CHAPTER EIGHT 365
An (k 1, • • • , k 8 ) ways. In the group i, the objects can be arranged

in k· ways (i = 1, s). Hence A 11 (k1 , ••• , ks) k1! •.• ks! = 11!
(4) F~r s = 2, the identity follows from (1) by substituting (t.Jt 1 )
for t. Suppose that this identity has been proved for a certain
s ;;;:. 2. We shall prove that
n n
~ (~)t~+ 1 rn-k=~ (~)(t 1 +···+ts)n-ht~+l=

k=O k=O
II
'\I ( n (n-k)! k, tks

LJ k
) tk
s+l k 1 •• • ks ktl •.. ksl
t1 . . • .s =
k=O k,+ ... +ks n-k
II
- _,_..,.-.:..:n..:.l_,..--_
k I k !k!
tk' tk •th -
ks ks+t-
1· • . . •s 1 • ••
11=0 k, •• • ks
k 1 + ... +ks=n-11
8.2
A
8.2.1. (1) Proof. For n = 1, N 0 = N- Nt = S 0 - S 1 , and

formula (2) is obviously valid. Let the formula be valid for n - 1
properties, and let N.,. - be the number of objects which do
1lt•••tlk
not possess any of the properties i 1, ••• , i 11 • We have
A 11-1
No=N.,.1 .,. =N- LJ N;+ LJ Ni. i

1' · • ·• 'n-1 i=l l~i<i~rl-1
- LJ Nl.j,J<+ ... +(-1) 11 - 1N 1 , ••• , (*)
t.;;;;i<j<k,;;;;n-1
This formula is also valid for a set of objects having the prop-
erty n:
-N n-
N-t ••.• ,n-t,n- ~
LJ N·z,n
l,;;;;i,;;;;n-1
+ ... +(-1)11- 1Nt ..... n-I,n (**)
\\bere Nil the number of obJects hav1ng the prop--
1S
n-t n.
erty n and none of the propertJes it , n- I Obvlousl)
N-1 - 'I
Subtrsc tlng ('••) I rom ( •) \\- e obta.Jn lormula (2} {3) U y tleft
ni t1on formula (3) can be wr1t ten 111 the form
u-m
Npa_ = ~ (-f)ll ( m!k} 2! N.,

1
A--0 ~~ 11< <'m.•" ~n
Let us prove th.a t any obJK t pOSSesst Dg m pro periJ es " 111 he ta kfn
1nto account by th1s formula only oncet and all the other obJects
w' lln-ot he taken. 1nto ac.tount at a.ll Indeed.. 1~ elem-ents ha V1ng
1 < m. properttes are ohv1ously not taken Jttto account The ele-
:ments bav1ng 1 = m + t g1ven propert1es WJil he taken Jnto .account
(:t ~} times ID tbe .second sum But
'
=(m:t)~(-l)ll {! )~ {~
for t= 0
for t > o
Jt~o
~bU!, t be e1 e-men ta ba -v 1n g exa. t \ly m prope:J'\1 ~s ate ta k~n n1 to

account 10 (3} JUSt once wb1le the r.ema1nU:J.g OhJeets are not taken
~ A ~
1nto .account at alJ (3) It .should be noted that .limt:l = Nrn - ...Jw rn ~
then {4) can be proved hy 1nduct1<'n on m By defuu t1on, N 0 ~
~ n
N = S0 Then N 1 = Sr,- (S 4 _._ S1 + S,----.
)=== ~ (-t)lc-l Sit
h-1
Thus. (4} b.olds f{lr m --= t lA. t. ~den llt y {4} he v al~d (cr a "t.e.t tz.lD
m ~ t Then
n~m n lm+ t)
~ (~t)l-• (m-~+k )s,.•.~~= ~ (-1)~( m:k )smihA
l.~l ~~0
(~)Let us denve. lor examplet formula (5) lrom (3) By pu\\1ng

A
m n. 1n (3}, we ohta1n N,., = S n 'n ILtO:rdante w1lh ( ~) Let

$:;
rormula (5) be val1d lor all k ;a: n--v+ t. v ~ f SuhstJtUtlng

CHAP rJJR J:IUHT 36i
n-v for m (3) and using the mductivc hypothesis, we

lll
substitute the right-hand side of equality (5), in
which k is replaced by n - 'V
v
k, for Sn-\'+R for +
... 'V ( n-v+k)
anyk~1.ThisgivesSn-v=N 11 -v- LJ (-i)h II-\' , X
It= I
It II
11 _ ~+k) Nm = Nn-v -
1 •
Ill )
( 11-V ( NmX
m=n-\'+h 111=11-V+l
m-n+v
'V k (m-n+ v) _ m ) ...
LJ (-1) k - ( n-v Nm. Thus,
h=l m-n-v
formula (5) is proved. Formula (6) is proved
• •
simllary.
Let us prove (7). For this purpose, lt lS sufficient
n
to veriry that R (n, r)= ~ (-f)h-r { ; ) s,. ~ 0
k=r
for all r. Let us use formula (5). We have R(n, r)= 'l: (-l)h-r(
" k
11
;) X
h=r
n n v n
~ {~}Nv=~ Nvl} (-1)/t-r {~)(~)=2: Nv(~)x
V=k v=r k=r v=r
v n
"LJ (v-m) 'V ... { v) {v-m-1)
(-1)k-r k-m = LJ Nv m r-m-t ~ 0. (5) Let
k=r v-r
• •
us prove that 8 111 > N 111 • We put R (n, m)=S 111 -N 111 • In view
n-m
of (3), we have R(n, m)= L (-1)h+l (m~k) Sm+h =
h=l
n-m n
(according to (5))= ~ (-1)h+l('n!k) ~ ( k~m} Nr-

k=l r=k+m
1l n-m n
~ Nr h (-1)h+l ( m!k) (k~m) = ~ Nr { ~ ) X

r=m+l k=l r=m+l
r-m n
~ (-1)k+I (r-;:m) = ~ Nr (~) ~0. Let us prove that

k'=l
... r=m+l .
Nm >Sm-(m+1)Sm+l· We put R' (n, m)=N -Sm+(m+1) Sm+l·
n-m
In view of (3) we haveR' (n, m)= ~ (-i)h (m;k) Sm+h=
h=r
I
'
368 SOLUTIONS ANSWERS .AND .Hl~TS
n-m n
(aecord1ng to (5}) = ~ {~1)~ l rn;k} ~ (k~m) N,=
A:~ '1'-:;3 ~ + m:
n r m
(-.ee e'itlmate R(fl, ill})=- ~ Nr {;} ~ (~t)lf c·~m)=
-r-m+2 tt~2.
~ N~(~}cr-m~1)~U
~rn-+2
8.2 z. Too tctal number af ~~ys ~n ~ruth the hats can be re

turned i! 4t =- 24. ...Tbe prt1hab1hty.... that exactly rn persons obtaln
the1:r own bats J.s Nm/~J.
. whPre Nm IS defined by fom1ula (3) for
n = 4 We have NG ~So-- s1+ s.- s.~+ s,==: 4!-- 4X3! +
....
6 X
-'S,
21 - 4 X it
4 X 81 -
+ 1 ;;: 9., pt. ~ 3/8 N \ ~ S1 -- 2S :r. + 3S ~ --
2 X 6 X 2J + 3 X 4 - 4 X 1 = 8 p J/3
":;:::S:
1 ::::=
p, J/4, Pa = 0, P• :::: 1/.24
$:z
8 2 3 (I) The number of dJstr1bUt1oru of the obJects for ahlch

the g1ven k boxes rematn empty IS (n- k)t-. SA= ( : ) (n ~ k)'"
It t-emaln3 for us t() il))pl) formula {2) (2) Sm•"~~ ~
n-rn
(;}(n-m-1-) 7
j E(r n, m)=Nm= h (-t)ltC
11
!k}x
•~tt
n-m
Cn~k) {n- m-kV= (~) ~ (-1)" ( ~-;;m) (lt ~ m ~ k)r
~0
(3) Use formula (4) 111 Pro}llem 8 2 1 (3}
8 2.. 4. Hint Use Venn s c1reles (1) 20%, (2} 60% (3) 70°JJ
B 2.5. (1) 2 (2) 6 (3) 3
8 2.. 6. {f) Each posJtlve 1.nteger m diVJsible by n and not ex
ceedtng .z satJs.6es the Jncquaiihes t ~ nm ~ :r 1 e 0< m ~ :r./n
S1nce m JS an Jnteger t ~ m ~ lz1n} Hence l\e obtatn the state
ment t2) 4.57 Us1ng formula {t)t we obta1n S 1
~
675 S 1 = lltl
'
=
s, :::::= 9 N~ == 457 (3} 734 (4) \Ve have N :.= s. == 30m S1 ==
= N (J ~ s D --- sl + s I --- s~
Jlrj
10m.., s :1 3tn, s, = rtl 22m 0::::::::::
(5) The nurobeY p such that lfn < p ~ n 1s pr1me 1f and only 1:f
1t 1s not dlYlSable by any pt1me number smallet than y; 'Ihe
number Nf1 'flf numb-ers. 1nd1Vl$lhle by any pr1me numb-er
14
from p 11 , P;,~ IS lnlpt 1 X X p iJt. 1~
s~~, = ~ N,l ,ill

t <P"'i. II P,.lt~)in
CHAPTER EIGHT 369
for k;;;;, 1, S0 = 11. Applying formula (2), we find that the

required number is -1 ~ +
(-t)k sk =
- r- 11- 1 +
o~k~l 11
(-flh Sh. Thc> unity is subtracted since 1 is not a prime
-"'
i,;;: l'ii
h ,;:
number. (6) 25.
8.2.7. 1/3 11 • Irrespective of the distribution of other elements,
~ach element ran belong to one of the sets1 XY, XY or xY.
(:!) u 211 -I. One element in U belongs to the set XY - U XY. The -
remaining elements belong to the set XY or XY. (3) 3 11 • The equal-
ity .f U YZ = X U Y is equivalent to the equality XY U
XYZ= l!. Therefore, each eleiiwnt in U is contained in exactly
one of the ~cts XYZ, XYZ, .\'"Yz-: (4) 3" - (n~ 7n 16) 211 - 3 + + +
(112) 11 (n! -t 1) + 1. Solution. From among the total number 3 11
of !Mirs (X. }') satisfying the condition of Problem 8.2. 7 (1), we
mu~t c:-..clude the pJ.irs belonging to the set C = A 0 U A 1 U B o U
B1 UB~, where An -= {(X, Y): I X I 0}, =
A 1 = {(X, Y):
IX1=1}.B 0 =- {(X,Y):I Yl =0},B1 = {(X,Y):I Yl =1},
B~ = {(X, rl: I Y I= 2}. Obviously, A 0 A 1 = 0. Bin Bi=0 n
(0 ~ i < j ~ 2). Therefore I C I = I A 0 I + I A 1 I I Bo I + +
I B1 I+ I B! I - 1 Ao n
B 0 I - I Ao B 1 I - I Ao nB2 I - n
1 AI n Bo I - I A. n Bl 1 - I Al n B2 I (seethe figure). It can
~asily be seen that 1 A 0 1 = 2" (n elements of the set U are distri-
buted among the squares XY and XY. The squares XY and XY
are l'mpty since X -- 0). Similarly, 1 A 1 1 = n2 11 - 1 , I B 0 I = 2 11 ,
I Bl I= u2"-1 , I B~ I=(;) 2"-~. Further, we have 1A 0 n
B0 I -=- 1. 1.In il B 1 I '--= 1 A 1 n B 0 1 = II, I A 1 n B 1 I = II ( n - 1).
I Au n B! I = ( ~) . I A 1 n B! I -= II (II-;- I ) . Hence I c I =
(n + 2) 211 + 11 (u - 1) 2 11 - J -
11 11
( ~)+ 3 ) -1, and the
-
numbe; nf requirl'<l pairs is 3 11 - 1Cl. (5) 11 (:!11- 2n). From among
th~ patrs (.\", Y) satisfying the conditions of problem (2) we must
exrlurle those for which the power of a set X or Y having a lower
ll 0 11er is smaller than :!. Considering that the smaller subset is
composed o£ 11 - 1 elements, their number is 11. (6) (n 2) (2 11 - 1 - +
1). It follows frnm Problem 8.2.7 (3) that the elements of the set U
can be contained only in the squares Xl'Z, XYZ and xY.Z. Since
I ~ 1~ 1, twn cases are possible here: (a) = 0. Then the square z
XYZ is empty, and there are 2" - 2 ways in which n element in
u ~an be distributed among the squares xYz
and XYZ so that
I X I ;;;;, 1 and I Y 1;;;:,: 1, i.e. so that the squares are not empty.
(b) I Z I = 1. Then the square XYZ contains one element which
1
The expression "the element v belon!!S to the set XY'' is under-
~~oorl _here and below in the sense v E"'x (~Y), and v E XY n
" eqmvalent to !' E (U"'-.X) n (U"-.Y).
1
:!',-OGJU
•
370 S0LU110~S A~SWt:RS AND HINTS
ean In: chosen ln. n '4a's The rema\nlnlf n - 1 rlem-enls must be

d1stttbuted among tb.e ~uares xVZ and XYZ .so that ~ X ~ ~ t
and ~ Y ~ ~ f The rondJtJon ~ X 1> t has already been 5atu~hed
&JDce Ihe square X YZ IS not em ply Therefore "e must IulfiJ the
-
only cond1hon 1bat tha square XY7 1s nol ~mpty The number of
aucb ways o{ d 'st1' hull oo o{ n - 1 el~men ts 1' tn l - 1 Th:e-te(orr
1n esse (hJ we have n (2" l - j) ways and the total number of
way9 1.8 (,•+ 2) (2n 1 -- 1)
8 2 8 \Ve ta1cu\ate th-e number Nn 11 of seat1ng arrangements
lor wh1(h k g1v~n paJrs nf warr1ng knJghls are not .separated We
combane each of the g1ven patrs Jnlo a single obJect We ()htaln
211-k obJrclsv.b,cb can he arranged 10 (2n-k 1)1 ways ]Q each
o( the gt \TeU k pat~ tb.e ad v.ersarl.es. c:.a n change post t1on s Thu.s
v.tnMam v,.,Jt= 2."~2n~k-i}\ and SA= ( ~ }Nn,.. Appl)mg
f OrDI pIa (~) Wf! nnd \hat. the ?eq llired n urn her l.S
~
~ (-J)II ( =) 2A (2n-lr-t)l
A f
8 2 9 We calculate the numbe-r Nn. k of seahng arran~ements

or n marrJed eonples Jf k grv~n couples .;lre not separated \Ve com
b1n~ Ncb of suth k couplet 1nto a Slng1e obJect Then we have
2n - k places for arran;g1ng k OhJtcts the number of arran~ments
.being 2(211--k-I)A 1 (the .seats are assumed to be numbered) After
1t J.SOI.ated couples are arranged we presume the evenne~s of the
SEa \I oc.cupLed by ladtes (these s.eat!l must be aU e:1ther even or odd
otherwise two ladt~s w,JJ be .seated s1de hy s1de) The seatJngarrange
m~n t of 1ad J Ps (gent] emen) re ma J n Jng a f t~.r the arrangement of
mlrl'l'@d caupl~s can be rea.hted 1n (n - k} ~ \\ ays Us~ng the mult1
1l1ea lion ruJe \\~ ohta1n N 11 A. 2 (2n -- k --1 ),{ 1( (n - k) !)'' Furt
h"r S~:~. = ( ~ ) Nn,. It rematns for us to apply formula. (2)
83
8 3 t (1) Induction The numbers aa Q' a, 1 are ~pee
1fied by hypothes!.s 11 ali tbe terms a1 have already be-en defin~d
lor I ~ n by us1ng (9) we tben obtain a11 ,. 1 -p 1 an, -
Pslln 1 - - Pk an h .-1 (2) \Ve must pro' e tiLa.t c}..'~+li +
p1cl."+A 1 + + p4c;,;n = 0 or 1\hieh IS the :Ja.D1e tiw"n (A.h +
p 1Ah 1 + + Pk) == 0 SJnce ~ JS a root of polynomial (10)
the ex ptessl{\U 1n t be pa renthe see van1she! (3) The (a-ct. t.ha t an. ==
c1"-t + + f"JtA~ sausfies reJat1on (q} foUows from (2) and from
the /act that 1l two se uences an a.nd b, ~at1sfY reJatJ on (9) tl~e
sequence a.,. ::: aa,. + bn salJsfi:e.s thJs r~latton lor a11 a and 6
Let us na v.. pra, e t at any sequence «11 sa tls(y Ln g (9) tan be
represented 1n the form :a 11 = c1 A1 + + l"A"-l where c1 are
CHAPTER EIGHT 371
appropriate constants. In view of Proble?J 8.3.1 (1), any sequence

a satisfying (9) is completely defined by tls first terms a 0 , a 1 , • • • ,
a;_ 1
• Therefore, it remains for us to demonstrate that for any
a 0, ••• , a,._ 1 there exist r 1 , ••• , ck such that
. . ' . .,,._.. +
. ,k-1 . .
. . . +. .,,._.
C111.1 T Czl\.2 • • • Ckl\.h-1 = ah-1•
The determinant of (•) is a Vandermonde determinant (see, for

example, Richard Bellman, Introduction to Matrix Analysis).
It is equal to
I<:;;; i<i,;k
rl o. ; -
'1·) and does not vanish if
;;.. =P l..i fori =P j. Consequently, the set of equations (•) has a uni-
q~e solution. (4) In order to prove that any sequence of the form
specified in the problem satisfies (9), it is sufficient to demonstrate
that the sequence an = ,,m}..n, where A. is a root of multiplicity
r > m of polynomial (10), satisfies (9). Substituting a8 = smt.s,
s = n, ... , n + k, into the left-hand side of (9) and putting
p 0 ~ 1, we obtain L (n) = (n + k)m ).n+k + p 1 (11 -j-k-t)ml.,n+k-1 + ...
k
···+Ph 11 m).n=An ~ (n-j-k-j)m ).h-i =
i=O
k m
).fl ~ ).h-ipi ~ ( 1~ ) (k-j)i 11 m-i =
i=O i=O t
m h
'J.,Pnm ~ (. 111
• } 11 -i ~ p i}...h-i (k- j)i =
1
i=O i=O
m k
~ ( niP; (I..) , where P;(x) = ~ Pi (k- j)i xk-i. It
11 111
}, nm • )
1
i=O i=O
should be noted that P 0 (x) = P (x), where P is polynom-

ial (10). Since 1.. is an r-tuple root of the polynomial
Po, Po (x) = (x- t.)r Q0 (x), where Q0 is a polynomial. It
can be easily verified that P;
P; (x). Therefore,
+l (x) = x d~
P; = (x - t..y-i Q; (x), where Qi is a polynomial. Hence it follows
that P; (!..) = 0 for all i < r. Consequently, L (A.) = 0. Thus,
(9) is fulfilled for the sequence nm X 1.,n for m < r, and thereby
for the sequences of the form specified in the condition of the prob-
lem.
Let ~snow prove that any sequence an satisfying (9) (provided
t hat 'A; Is a root of multiplicity r;, i = 1, ..• , s, of polynomial
(10}) ha~ the form mentioned in the formulation of the problem.
Here, Without loss of generality we assume that Pk =F 0, i.e. zero
37~ SOLUTIONS ANSWERS AND HINTS
u not a root of the characte-r LSttc {} ol ynomtal (If ph ~ 0 Eq (9)

can he s1mfl16.ed ) In order to prove th1s it 1s ~ u ffi c1en t to .show (se@
solut1on o Problem 8 3 1 (3)) that for any at a1 alt. , the
!ystem
Ct 1 + + c, 1 = a..
( ~1 1 + C1 2 + + c.t .,..
1
i) l\ + + (c, 1
+ + Ca: r , 1) A.. = d1
(t 1 l + t1 .I (k _... f) + + cl ~ 1 1 (k - frl 1' "'~ 1
+ + (c; + t: 1 8 ~,(k -- t) + + c, r • 1
has. a s.o lut100. For thts purpose tt, l! suffictenL to pra ve that 1ts.
determ1nant d1ffers from zero or whJch 1s the same that the vectors
.A
... .... h ...... i t ,., t "-" ,_
0 h" 1 w ne h 1 ::::: (A- 1 ,).. 1 ""1 11r.!
,.1 I i
f~ l ~,) = 0
1c -1 aN lmearly Lndependen.t We
'
assume tba t the con verse 1s true Then there ex 1st con st.;! nts d 0
.......
dh 1 wb1cb are not all equal to zero and such tha l A =
...... __.., ht
+ dlt. 1 Ale 1 -- 0 Let Q f.:r) ~ d1 z"1 and let. h
'L-0
he an operator such that A/ (z)

dl
x d.:r \Ve put 6.hJ =
~ ? \
/:::,. (.61 lfl Then A :=;: (Q (A 1 ) 6. Q (A, 1) 6. 1 Q (A 1)
Q ('-.,) 6 ~~ -1 Q fi. 1 }) lt should ~ noted t hat Q (~ 1 ) =
= 6 r 1Q (~} for it. == 0 tf and only 1f A. 1s an
AQ (.1.•) =
r-multtple root of the polynomull Q (z) Thus A.
....
0 means that
-=
~ 1 1s a root of multipliCity r 1 l 1 Js a root of mult1phc1ty r 2 and
flnally "'-• lS a root o1 ro.ult1pl1c1tJ ?"_, oi the pDl)nounal Q ~.x) But
,.1+ '• -t- +
r, == k and Q (x) 1s a polynon:ual of a power
lowet tban k and h-e:nt e 1\ ea.nnot h.a ve: Is: ta"Ots \Ve an1 ve at a t.on
trad 1et1 on Tberef ore the set of equa t1 on.s (•) has a un1 que .sol u t1 on
8 3 2. {t) c1 +
c~ 71 Th.e chat&.ctertstl.C polynamtal xr. - 4z + 3
has the roots ).1 = 1 and .A 2 - 3 Us1ng Problem 8 3 t (3) we obtaJD
the general soluttou.c 1.k~ + c,.~ (2} c1 (-3) 11 (~ c, (-1}n (----3}n;z +
(3) t 1 (( + J'5)/2)"" + ct ((t --- Jf5)/2)n (4' (--t)" f~t + c1 n}
(5) (e1 + ~1:n) (-4) 11 +
c~ {-2.)n (6\ (-1} 11 (:t 1 c~n. c1 rt 1l + +
8 3 3 (1) 7 +
3n Tbe general solutJOn ha5 the form c. c~an +
(see Problem 8 3 2 (1}} From the JUll1al ccnd1t1Dns we ohta1n
t1 + 8c1 :::;::: tO t 1 +
9c 1 ==:: t6 Hence c1 = 1 c1 ~ t an == 7 +
+
an (2} 311 (l1--t)11 (-1/:.:t)n (3} c, (sn c~ (-2)11 + + +
where e1 = (14 - b - 4c119 c ~ (b c - 2a)/3 c, = (2b - +
e - e}It 8 (4.} eos « n 'r he roots ol the char~ct.erts tle pol yn.o,m 1&l ar.e
al 'I z: e::l:;i~ = cos" ± l SID ct The general solution 1! Dn. ==
CHAPTER EIGHT 373 .
•
c 1-n + c2M'· From the equalities a 1 = cos a:, a 2 = cos 2a: we

o~t~in c1 = c2 = 1/2. Hence an = (A.1 + 1-l:)/2 = cos a:n. (5) 1 -
(-1) 11 • (6) 3n (1 + n).
8.3.4. (1) Substituting an+ b for an in (11), we obtain
a (n +2) + +
b p (a (n + 1) + b) ++q (an b) = an + fl.
Comparing the coefficients of n on the left- and right-hand sides, as
well as the free terms, we obtain a = a:/(1 + + p q), b = ((1 +
P + q) !} - (2 + p))/(1 + + p q) 2. (2) It follows from the fact
that x = 1 is a root of the polynomial x 2 + + px q that q = -1 -
p. Substituting n (an + b) for an in the equality an +2 + pan+l -
(p +1) an = an+ !}, we find that a = a:/[2 (p + 2)); b =
(2!} (p + 2) - a: (p + 4))/(2 (p + 2) 2 ). (3) Since x = 1 is a mul-
tiple root of the polynomial x 2 + px +q, p = 2 and q = 1.
Substituting n2 (an+ b) for an in the equality an+ 2 + 2an+l +
a =an+ ~ and comparing the coefficients of na, n 2 , nand no,
,:e find that the coefficients of n3 and n 2 are equal to 0, and a = a:/6,
b = (~ - a:)/2. (4) For (1), the general solution has the form
a11 = c 1 f..~1 + +
c 2 f..~+ + a:n/(1 p q) + (~ (1 + q + p) -
a (p + 2))/(1 + +
p q) 2 , where /. 1 , 2 = (-p ± V p 2 - 4q)/2; for
(2) an = c1 (-p - 1)n + + c2 a:n/[2 (p + 2)] + (2~ (p + 2) -
a (p + 4))/2 (p + 2)2; for (3) an = n2 ((a:n/6) + - (~ 2)/2) +
Ctll + C2•
8.3.5. (1) an = 1 + ( ~ ) . The general solution of the recur-
rence relation an +1 - an = 0 is an arbitrary constant c. A partic-
ular solution of equation (1) is sought in the form a~ = n (an+ b).
Substituting this expression into (1), we obtain a~ = n (n - 1)/2.
The general solution of equation (1) has the form of an.. = a~ + c.
From the condition a1 = 1, we find that c = 1, and nence an =
1 + n (n- 1)/2. (2) an = 2 (-4)n - 3 X 2n + 5n. The general
solution of the homogeneous equation an + 2 + 2an +l - Ban = 0
has the form c1 (-4)n + c2 2n. A particular solution of equation (2)
is sought in the form a~ = d X 5n. Substituting this relation, we
obtain d = 1. The general solution of the inhomogeneous equation
(2) has the form c1 (-4)n + c2 2n +5n. From the initial conditions,
we find that c1 = 2, c2 = -3. (3) 1 n+ + 2n; (4) n (n - 5) +
(-2)n. (5) 2n-a (n 2 + 8); (6) (-3)n +2n +
n2n.
8.3.6. (1) A non-degenerate case is when either q1 ofo 0 or p 2 ofo 0.
If ql = p 2 = 0, we obviously have an = c1 pf, bn = c 2 q~. Let
ql =I= 0. Then bn = 1/q1 (an+\ - p 1an), bn+l = 1/qf (an+2 -
Ptan;~- 1 }. Substituting bn+l and bn into the second re ation, we
obtam. an+~+ (-p1 - q2) an+l + (ptq2 - p2ql) an = 0.' The pro-
blem IS reduced to Problem 8.3.1. (2) an = (5 +
2n) 211 , bn =
-(1 + 2n) 211 • (3) an = c1
(-t)n c2 + 5.5n.
+ (-1)11 +1 c2 +
5.5n, bn = c1 +
0.5 +
8.3. 7. (1) Induction on n. For n = 2, the relation F 2 +m =
F1Fm + FzFm+l = Fm + Fm+l holds for all m ~ 1. The inductive
~tep n-+ n + 1: Fn+l+m = Fn+m +Fn+m-1 = Fn-lFm +
n~m+l *Fn-2Fm + Fn-lFm+l = FnFm + Fn+lFm+I· (2) Carry
out mductlon on k. (3) If Fn+l and Fn had a common divisor d > 1,
then Fn-1 and Fn would have the same common divisor since
Fn-1 = Fntl.- Fn. It follows by induction that F 1 and F 2 should
have the dlVlsor d. (4} Method of representation. If N = 2, then
374 SOLUTIONS ANSWEitS AND HINTS
N Fl. ;:::$. +
F 1 If N > 2.. we choose the largest •• 1 such th.a t F"' 1~
Nt then the largest "• sucn that Fnz ~ N-- F 111 and so on Then t
N == Fn 1 + F,. 2 +
S1nce F11 ,.1 > F,. for n > i the repre
sen\.1 tlOO -ca nuot -con t.a.i.n two num \nts W1 th lht! sa m.-e 1D.d~1.. n _...,.. 2.
The representatiOn cannot contain twa adJaCent numbers F" and
F ra+J .s1nce F., +
F11 • 1 =;; F 11 •t and bence F n. ,_, must he chosen
at tbe step on whlch F f'-"' 1 1s chosen (5) The genera I sol ut1 on of the
recurrence .reJatton Fn:~r• =
P.,... 1 F.,. ls g1ve.n 1n ProhJem B 3 2 +
( 3) Usmg the 'n' t \a l e ond lllons~ we "Obla.ln tbe r-e-q ulred result
(6) and (7) The proof 1s carried out by 1nduct1on on n (8) Applying
tw1ce tbe 1d~nt1tY ol Problem 8 3 1 (1), v.e obta1n F,,. = J. n r:F2 n, +
F ._F1.n-+t F n-t (Pn 1Fn. -= FnF n.-t-1) + F J'. (fl +Jt,. t) -=-= +
+
n-,Fn Fn-lFnFn- 1 + F:._ ll•t (F.,..,- Fn. 1) $::;: FA.+1 + +
F~ +
F~-~ f n. +
F n +1 F n -1 (F"' --- Fa+.a) -::;: : : F:.- t FJ, + +
Jl f.Fn. - Fn•ll1 , = F~+• 11. - F~-• +
8 3.8 (f) (f ,_f)-l The ge.oeralulg funct1on far the sequence
ao-
t:Jn == 1, n == 0, t IS by d~finl ttoD the sertf!s A (t) = LJ 1n.
n-O
wh1ch lS an expansion of the funtl1on f (t) ~(I ~t)-l Into a
series for 1 t 1< 1 (2) 1 tN ~ (tN+l __ I) (t --- t) +1+ +
(3) (t- at)-1 (4) e-a.t (5) (I t)-l (6) t (I -- t) a \\ e +
(Q -001 (lq.
hale A (t} == ~ nt" -= t ~ (n +I) t" The ~r1~s Y, (n +I} J"

n~o n-~ n-O
Is obta1ned by the ternlw1se d1flerent 1at1on ol the 5cr1es
ICII:IIi
~ t• un"formly -cunverg1ng to ~ (t) -={l ~ c)-• ((lr \t \ < t

'f'l:;;;;:z(J
CD
Thus. the serJes ~ ntn converges to the funct1on f (t} = llp" (I)=
n-0
t {1- t) i. (1) 2t" {1- t) !. {8) (1-l-. t)m {9) (i -1 t)fl. {10) J {1-r-l) X
(t- t)--J (t f) t s1n a (1--- 21 cos a+ t•)-l \\ e put .:.o- (a)~ e'a: =
+
cos a t sJ P a and use lhe fact l b~ t .;a" (a) =:=$ e ifSn ::::::: cos a, T
i s1n an~ ql"" (---a:) = e- tan= cos an--' ~lD ctn Let us c:onsJdrr tbe
genera.tJng ru.octlons ..A 1 (t} = ~ 4pn (a) ,., .=: ( l - Q) (a) 1)-1
~ n-0
As t') ~ ~ ql" <--- «) t• = {i --.-. " ( - a.) &) 1
Slnc~ ~'n <1n ~
n;z:O
(IQ
(2t)-l (~ (a11) - ~ (--.an)).- we have A (t) = }J Sid ant'• =

{Zt)-1 {A. 1 (~)- A1 ('))~ f2t) 1
••-O
((1- tp {a) t)-1 ,... ~' ---tp (-- e) t}) t ~
(2,)-lt (~(cr.) - cp (-- a)) (I --- (rp (a) + cp (-a)) 1 + t J- ~ 1 1
I SID a. {1-21 cos a+ l') l {12) (i_..... I COS ct) (t -21 COS a + 1'2)-l
8 3 9. (t) el By defin1 t1on~ E {f)= 2j dnln It[ ~ ~ t,. nl
n~U n.:z::!!!!O
CHAPTER EIGHT 375
This series converges to et. (2) eat. (3) te 1• (4) t~et. (5) (I+ t)m.
co 1ll
By definition, E (t)= h (m) 11 t 71 /n! = ~ ( ';: ) t 11 = (1 + t)n.

u=O n=O
(6) el (t 2 +t).
8.3.10. {1)-{ti) Compare the coefficients of th.
8.3.11. {1) { ~1
1
) qm-np''· {2) 1. {3) ( -l)ll ( 1~2 ) •
(4) (-1)11-m { m ) • (5) ( _ J)'l-m ~ ( -1)h(r-1) { ; ) X

11-111
k
-m
( 11-m-kr ) It should be noted that (t Lt~+ . .. +Vl=
· '
em (1- er)m (1- t)-m. The coefficient of thr in tht• expansion of
(l- t')m is ( -i)k ( ; ) . The coeff1cient of t'l-m-kr in the
·
expansion of ( 1 )
- t -m 1"s ( - m
n-m-kr ) . (6) ( 1/2) 11 / 2 ( -n/2 m ) for
even n and 0 for odd 11. (7) (- 2)-n ~ 2

( :) {
1 ~ ~ 11
) • ~he coef-
h.
ficient of tk in the expansion of (1 +2t)-lf~ is ( - : ' ) 2k. The

2
coefficient of t 11 -k in the expansion of {1- t '2)-m is

1
( I-nk } ( -1t2)"-k. Therefore, the coefficient of t 11 in the expansion
I- "
of f (t) is ~ ( -~/ } 2
2k ( - 1J2)n-h ( ~~~~ ) = (-1/2) 11 X
k •
~ ( -:/2 ) (- 4)k (
k
n-=-mk ) = (- 1/2)11 ~
h
( 2: ){,;-_n~ )•
(8) ( :.~) 2 11 -~- ( ,-:;) 211-3. (9) ( -1)'1-ln-•. (ill) ( -1)11-1 '(2n-1).
t
We shall use the fact that ~ 1

~:'! =arctan t. We have (1 +
0
00
.z:~)-1 = ~ ( -1)" x2 11 • Integrating the left- and right-hand sides

11=0
00
"" ( - 1)" ~~11+1

between 0 and t, find that arctan t= LJ "-11 + 1 •
11=11
I
1X3x ... x(2n-1) 1 d.r
~0
•
(11) . Hint: ~~==::;-=arcsin t.
2x4x •.. 2n 2n+1 Jlt-x 2
376 SDLUTIO"S .ANSWERS AND Hl,TS
{i2) (- 2)1tflJ(nJ2)t tot tv~n n •nd fi tor cdd ~ {13} ( -1 )~I" t1 l +

tl! for even » and 0 for odd " (14} (-i}n ( n-1
m--l
)
8 3 t 2 (I) Let us .compare the coefftc1ent.s of tn 11n the 1denl1ly
+ +
(1 t)n (I+ t) m-2 = {1 t)Yt-m ' On the one band tlus coeffic tent
2
1~ ~ ( n~s} ( ~~""f }= :3 {-l}' 1
( ; } ( :!;) On tbl!
.
other hand, (n:_:; 2
)=(-1)n
'
1 (n:_ 1 } (2) Consadcr tbe
1dt nt1ty (1 +t) m (1- t)m = (1- t' }m and
compare the coeffjc tents 2
of t!n (3) Cons!drr the JdentJ t) ( 1 - l 1 ) 111 (I -- t} n-t;;::;:::
( ~ f)nl t-m (I -l)"' n l and coef.fieu~nts of tA m (4.) <..on~1der the
+
IdentitY ((1 tl"'+ (I- t)n))~- (f + l)~n+2 (1- tZ)n+ (t --lptn aud
coeffic•ents of tun (5) Cons1der the IdPn t1ty (( l ...,.- t) n 1 +
(1 -t}-n-1) ((I+ t) n 1 ---t!--- t}-n 1};: (i t}-.. 11 '---- (1- t)-~n 1 and +
the coefficlents of t 2 n• 1 (6) Coru1der the JdentiiY (1---t)~" (I +
2t (1-t}- 2}==- (i + t~ln and the coef.6c1ents of ~ 2 m
01) ~ 010 1;10
8 3 l3 {1) A {J) ==:: ~ Ant"- h =~ tn ~ e-s.zl\ d.z = ~ e-x)<,

n--U n~O 0 U
~ 00 ~
~
LJ nr
!'! (~t} ~ d.z = t; :a:£ {i:t) d.t (2) ~ e "-eX' dx=
1
~ t ) e u du.=
n=G 0 u l
(1- t)-t (3) For the sequence .an v..e have A (t) ~ (n) 1 l 11
J~n
E(l) = "~ _(nn]IJ ,n -==- "'

LJ
tn
(ll-J)r
Futtlu~l
D~;~n O~J~n
c:ID J'l. n oo
rl e X "'V (zl)11
LJ (n- j)l
d:&::= ~
.t... . ----In
(1'1-1)1
... r) e x~n dX ~
L.J (dJ t =A (t)
u J.:;;;O J~o o J--O
8 3 ili {1) \\e u~ the 1dellt1\y {t + t)~ ~ ~ (\ + t)n li + t)b
Companng the coefliu~nts o( ~n we tl.nU that (a~b)=
~ {n.~k} {! }=~ kl(l~k)l {a} 1_;~,{1>h Mult1piyH1.g Doth
4 h
sides hy we obtaJn the 1equ1red equaht} (2) \\ o put a -.::. It
nl
b' =:. b!h Apply1ng the Id~nt1ty proved lD 8 3 14 ( 1) to a aud b
'1'1 e .lind that (a + b ln 2; ( ; ) (a )n .r. (b ),. J\.ltdllplywg boLh
"
stde-s by 1~n v.. e arr1 ve at the 1equ1red tden t1 ty (~llltt. {a )4 J,., -
(a), -'1)
8 3 15 (f) \\ e mull1ply the cqualt ty an-- b 1 bn 1 by t" and
take the ~urn over n In the lCgiOD of lOn\ Clbrenc~
.:!II) od
tl e M"'rJlS or
2J
~~o
ant:n. and S bnt
n:JZ!!!.o
11 t the follo\\ 1ng 1drnt1 tu~~~ ~ 1e VJ-hd
CHAPTER EIGHT 377
00 eo oo oo
~ ant 11 = ~ {bn- bn-1) 111 = ~ b11 t 11 - '2j b11 -tt11 = B (t) {1- t).
1;-"u 11=0 n=i n=O
(2) We multiply the equality a 11 =bn .. 1-bn by t 11+1 and take the
:::tun over n between 0 and co. We obtain tA (t) = B (I)- b0 - tB (t).
(3) We note that b 11 =an-1-a,1• Hence B(t)= -A (t) (i-t)+a0,
where a0 = B (I). (4) l\lultiplying the equality an= nb 11 by t 11 and
00
taking the sum, we obtain A (t)= ~ nbn111 = t ~ B (t). (5) The

n=t
proof is similar to that in 8.3.15 (4). (6) Let us compare the coef-
ficients of t 11 in the equality A (t) = (1- t)-h B (t). By definition.
this coefficient on the left-hand side is an and on the right-hand
'' ll
:::ide,~ (-l)i( ~k) b

11 - j - ~ (k+:-l )bn-j=Sh(b 11 ).(7) We
i=ll )==>0
= ""
(t 1 /~)= "" b 11 t 1~. B ( -
11
t 1 /~) = ~ ( -1) 11 bnt1112 • Multiply-
have B
-
n=ll -
n=ll
iug the ~um of these :-cries by 112, we obtain+ (B (t 1 f2)+B (- tlf2))=
-
.2; b2 11 l 11 = .2; a 11 t 11 A (I). (8) Note that bn = a 11 +1-an. The equal-
n=ll u=O
ity A(I) =B (I) t (1- t)- 1 now follows from 8.3.15 (2).
8.3.16. (I) First method. Let C (t) be a generating function of
the sequence 1, 0, U, . . . . By hypothesis, C (t) =A (t) B (t).
Con..-.equently, the cqualitie::. 1=a0 b0 , ll= a 0 b1 +a 1 b0 , ••• , 0=
n n
Yo
- au-ibn ... must hold. Since a 11 =I m ) ,
\u. )1,
- { IZ-l
111
•) bi = ll
i=tl i=O
for all 11 =I, 2, . . . and ( ':: ) b0 = I for 11 = 0. Con~ecutively,

obtain b0 =1, 11 1 =-m, b~=m(m+1)'2, b3 =-m(m+i)X
\\C
(m+2l G. By induction on 11 we can easily prove that bn =
(~ ).
111
Tim:;, B(t)=(!+t)-m. Second method. A(t)=(l+t)"',
B(t)=(A(t))- 1 =(1-!-t>-m. Hence b 11 ={~,m)· (2) b0 =1, b 1 =

-a, bu=ll for 11> 1, B(t)= l+at. We have A (I)=~ (at) 11 =
n;;;:,o
<~.-aW , and 1
~incc A (t) B (t)= 1,
we obtain B (t)= 1-at.
(il) bo-1'~-1, b1'-= -:!, b11 =0, n>~. B(t)=(1-t) 2 • (4) bm=
(-1) 11 , b~n+1-0, u;>ll, B(t)=-(l+t~)- 1 • (5) b0 =..b 1 =1, B(t)=
1,•) )
1-i-t. (6) b11 = ( ; - , B(t)= l'i+t.
1
8.3.17. (I) ~lultiplying by t 11 +~ and taking the sum, we get
A(t}=alt-a0 +ptA(t)-pa 0 t+q.4(t)e~. (2) We represent A(t} in
378 SOL UTJO~S .ANSWERS A.r._D lllNTS
the form t _c 1)t,

11
+ 1
_e:z). , Ha Vlng de.""rm1ntn:4
... -..J e
1 an
d &t.~J we
1
find that A (t) ~ 1

l.1--- )-..~,
+
( 01 PlltJ+A.lllt_ " 1 Pao+l.2 a 0 } J-1av1ng
i - htC t - h. 2 '
+
determ1ned. the coe f fie Lent o( t" 'n tb.e e "< pan~Lon. of A ( () 1ntc a.
se r~es 1n 1 we cbta u1 an e1.. pre sst on f c r a 11 ( 3) Le1r us rep Nsent
n
Jl ( b I
t) Jn t e orm
.c\
1 __ 11 +(f---lt):a c"2 F
rom
cl
1 ___ 11
.e,
+(f--ltfi=
ao+fa.----..2~)1 a1 a
-~~~-- ~
( 1 -l.t) :1
w.e find tba.t c 1 =- - .. :1\.~ = + Zao., c1 = . . _L -
A
40
~
Ex pa.nd 1ng A (t' an to a se r 1es~ we obta1 n A (t}.:::: ~ (e 11.11 +

n-o
c ::1 ( n + I J A,•) t'\ an = ( a, + n ( -~• - •• ) ) )."
8 3 t.8 (l} Hint Use the 1dent,ty lU Problem 8 t t5 (3} (Z) We
mplttply eacL rela 110n of 8 3 f8 (I) by t 11+1 and take the sum over n
bet ween 0 and oa Vs•ng the lDJt1al conJttJon~. we obtain the re·
t--t
Ia tl on.s he l y, een genera t1ng func~1 ons (3) A (') =-
1
--
31
+ t21 ~
B(t)== t--5-(+t"~ ' (4) The roots of the equataon 1--3t+t ~0 1
at~ ~1~{3+ V~)f2 and l..~{l-- V5)lt w~ e.1..p1-e.~~ A (t) ,n tne

form A (t):::: t
4
-- , 1
.,
t
+ 1 --
~
h
1\.oli t
Since A (I)= ( t --1)/(1--- )..l r) X
(1 -1 1t), eq:ual1ng the r1ght hand s•dest we get a (t -.A.~t) +
b {t--- ).1 £) == t-- t lienee a:::: (A 1 - l)/(A 1 -- ).,_);::::;: ( t V'&)/(2 V 5), +
... <t>= ~ ( ~~f~! +
2 5
b=t-cz=(JI5-t)f(2 ys) Th11s,
V~-t)
. _J:,t 1 ( 1 1
SullllarJy, we ohtam B (t)=- :1- i-A.~- t~~).;e )
1 11
E.xpandmg A (t) aDd 8 (l) tnto eer1es tn t we tind that a,~
(2 JIS)-1 ((t + Jf5) A: + (J! 5-- f) .1~)t bJl =:::; ( J!S)-J (A~--- 4:) 'fak1ng
into account the Inequal1t1e:J 0 < A1 < .A 1 we obta1n li.Dl a~A; 11 -=
n ... oca
(t+ lf5)/Z }15, l1m blll-;n~ tiV 5
n--+eg.
8.3 l9 ( l) l\lult1 ply tng by I""' .and tak1ng the sum ol tbe 1mt•al
[ela.tloP~ \Ve ge\ A {t}-a,::::- tA1 (t) (2) \Ve have A (t)== (2t) 1 (t-
c)g 00
(1 ~4t)''') = (2t)- t ~ ~ ( -t)• ( ~ ) <'t)~) """'2 ~ (- tJfl-1 x

1 1 2
(
Ji..;;;;;.O n~t
01;)
( t::) (4t)P l =~ n!t {Znn.) t 11 (3) 1\fultlplytng by t 11 and

~~o
la. k1ng t. b.e- ~m. over n t Wf- o.bta1n (er the g~n.e ta llng {Wlcl lOO
CHAPTER EIGHT 379
00
A(t)= ~ a,.t" the equation .·P(t)=A{2t). We seek its :,olution

-
n=O
in the form A (t) = et:.tt. This function obviously ::tatisfies the
equation. Considering that a 1 = 1, we find that ct=- 1 whence
a,.= 1/n!. The uniqueness of the solution follows from the initial
relations.
8.3.20. (1) We number the vertices of the (n 2)-gon by +
1 2, ... , 11 + 2 in the clockwise direction. Two cases are possible
h~e. The first case. None of the diagonals passes through the ver-
tex n + 2. Then there must be a diagonal between the vertices 1
and n + 1, and the number of ways of dividing the polygon is
an_1 • The second case. There exists a diagonal emerging from the
vertex ~~ + 2. Let k be the smallest number such that the (k + 1)-th
vertex is joined to the vertex 11 + 2 by a diagonal. If k ;:;;;. 2, there
exists a diagonal of the form (1, k + 1). Then the number of
divisions of the initial (k + 1)-~on is equal to the number of di-
visions of the (k + i)-gun witn the vertices 1, 2, ... , k 1, +
multiplied by the number of divisions of the (11 - k + 2)-gon with
the vertices k + 1, k +
2, ... , n +
2, i.e. is a 11 _1an-k (2:;;;;; k:;;;;;
II
-
n). It is natural to assume that a 0 = 1. Thi,; gives a,.= '\, a11 _ 1a 11 _k.
h=l
1 211
As in Problem 8.3. 19, we find that a 11 = { ) • (2) a,. =
ll +1 ll
II T
~ 1
2
{ " ) • The problem is solved in the same way as 8.3.20 (1).
II
8.3.21. (1) n!. (2)(-1)ll-1/ (n - 1)!. (3) 12 X 32 X ... X
(11-2)~/n! for odd 11, 0 for even 11. (4) 0 for odd 11, ( -1) 11 !=2 -nf(n12)!)-1
for even 11. (5) 11~ for odd 11 and 0 for even 11. (6) an = F", where
{F11 } = 1, 1, 2, 3 is a Fibonacci's sequence. We first prove that
if {a11 } satisfies the relation
it also satisfies the relation

(u)
For n = 0 and 1, the statement is valid. Let us prove that ( •) leads
to the identity
4 •
a;I+J - anan,..:l = a;u·.l - (ln.._:!an+-i·

SOLUTIO'S ANSWERS .A~D HlNTS
8 3 22" (I) ~lultJ plyrng the recurrence relat1on by t" and taking
the 5um o'er k \\e obtain (f - t) ..An (t) ~ A 11 1 (t) (2) Su1ee
a (0 0)- 1 and a (0 k) 0 for all k > 0 A., (l) = t Catt)Jng
out 1nduc t1on on 11 and USing 8 S 22 (t} "e find that A,. (t) -
(I - t) n (3) E:rpand Jng a (rz k) 1nta a ser1es we o"bta111 a (n lr) ~
( ~" ) 1-l)'- = { n+z-f )
8 3 23 (1) 20 12) 6 ('3) 9
8 3 24 (1} A (t) (1 - t 11 ) t (I -- ~ 3 ) 1 (1 - t-') 1 Let us
pro-, e thal A (t) IS the requ1red genetatlng iuncl1on \\ e note that
(1 - J") ~ Lj t~' for k 2 3 5 1 e ,all non zero e.xpansJOn
•-0
.c oefftc1 en t,s are eq ua I 1o I The coelfie1 en t of t 11 1n the ex psn.s 1on
of the product (1 --- t 1} 1 (t - t') 1 (t ...-- ~~) I 1.! equal to tbe .sum
of products of the form ak b, c,., where a 4 Js the coe ffict en t oF t l
tn the e;rpansion of (l -- t 1} 1 bl IS the coeffi.cJent of l!l JD the
expan~non (t -- t•) 1 and em 11 the eoeffi.c1ent of t'"' .and 2c +
31 +
5m n It shou}d be noted that .al. = bl t'l ::;: 1 and
tbe number of summands JS e~actly equal to a'"' (2) A (ll ------ {I +
tl) (1 + t') (f tS) +
(3) A (I) == (I + ti + + t!P) (1 L
~3 + + J~"') (t + t!l + + t'') ~
8 3 25 (1) It should he noted that A (qt) - } J (1 + q-'• t) ===

1
-':;;;5!:1
{1 + gt) l and hence A (tJ == A (gt) (t + gt) The «ioefficJent
A (t)
oft" on the left band slde JS hy defin1tton dn and on the r1ght
hand s1de .anq"' - an 1qn By tnduet1on on n we hence obtatn
a,. ~ rz(n I)JI II" (qA - i) 1 n 1 2 a, -= 1 (2) See

~s;;;;;;o.\
solut1on to Problem 8 3 24 {t)

A
8 3.26 (t} {2) S (n k L+i)= L {- t)• v( { :!! )-

'\P~O
n+l n
( v~ 1 ) }n+t+t}k=- ~{ \1{ n~ )(-:+t)k-,- ~(-t.)"
1
t}n+l "x
'Y t ~·
( ~) (v+l}"=-S(n+t k l)+S(.r.r k l} (3) S(n k+f l)=
n n
2: l- i)"' 01
( : ) ( \1 + I) 11 l v +l) ~ { l}n v 7l ( : : ) ( \1 +
,.;;;;z 0 \' ~.a
fl
1}11 +lS {n k l) n ~ { -t)" " ( ( : \ - { n:- t} }+IS {n k U-

v:;;;;;;;O
(n + l} S (n k l) + nS (n -
It l) (4) The proof IS carried f
out by Jnduct!on on k and by us1ng 8 3 26 (3} For any Zand 11 > 0
CHAPTER EIGHT 381
n
we have S (n, 0, l) = ~ (-i)n-\' ( ': ) =0. Let the statement
v=O \
be correct for a certain any n > k and any l. Let k + 1 < rz.
k ~ 0,
From 8.3.26 (3), we have S (n, k + i, l) = (n l) S (n, k, l) + +
nS (n - i, k, l). Since k < n - 1, in view of inductive hypothesis
S (n, k, I) = S (n - i, k, l) = 0, and hence S (11, k 1, l) = +
0. (5) Induction on 11. We have S (0, 0, l) = 1° = 1 = 0!. Let
s (n, n, l) = n! for a certain n ~ 0 and any l. Using (3) and (4)
of this problem, we obtain S (n 1, 11 +
1, l) = (n + 1 + l) X +
S (n + 1, n, l) + (n + 1) S (n, n, l) = (r1 + 1) S (11, 11, l) =
(n + 1)!. (6) Induction on k. In view of 8.3.26 (5), we have
S (n, n, l) = n! > 0 for all 11 and all l. Let S (11, k, l) > 0, for
a certain k > nand any 11 and l. Using 8.3.26 (3), we get S (n, k +
1, l) = (n + l) S (n, k, l) +
nS (n - 1, k, l) > 0. (7) This
follows from parts (1), (2) and (6) of this problem. (8) This follows
from (3) and (5). (9) This can be derived from 8.3.26 (3) and (4)
hv• induction on k .
00
8.3.27. (1) In view of 8.3.26 (9), \\e have cr 1 (t) = ~ S (1, k,

h=O
O)t 11 "'
= 2J t 11 =t (1-t)-1. Further, in view of 8.3.26 (3) we have
11=1
S (n, k+ 1) = nS (n, k)+ nS (n-1, k) . .\fultiplying both sides of
this expression by and taking the sum over k, we obtain
th+l
On (t) (1-nt) = ntcr11 _ 1 (t). The statement can now be proved
by induction on n. (2) For n = 1, the right-hand sides of
formulas in (1) and (2) for crn coincide. If we prove that for
II
On (t)=t ~ (-1) 11-kk ( ~) (1-kt)-1 the recurrence relation

11=1 .
On-1 (t) = ((1-nt)fnt)cr 11 (t) is valid, the statement will follow from
it by induction. We have i-nt cr 11 (t) = 1 -nt x
nt nt
n n
t ~ (-1)11-h.k ( n) (1-kr)-1= ~ (-t)n-h. (n-1) 1-nt =
....... k ~ k-1 1-kt
h=l h=l
n n-1
Y, (- 1)ll-h. (n-1) ( 1-
LJ k-1
(n-k)
1-kt
t) = '\:1 ( - 1)
Ll
11_11+1(11-1)
k-1 X
11=1 h=l
n-1
(n-k)t '\:1
1-kt = t LJ ( -1)n-ll- 1 k
(rz-1)
k (1-kt)-1 = Gn-dt).
II= I
8.4.
' 8?4.1. (1) (0, 0, 0, 1). (2) (2, 1' 0, 0), (3) (0, 0, 2, o. 0, 0).
(..) (-, 0, 0, 0, 0, 1, 0, 0). 8.4.2. The transposition of the ele-
ments i andj will be denoted by (i, j). Then (2, 3, 4, 1) = (1, 2) X
382
(z 3) (3 4.) (4 2. 3 i) ~ 'f. 4 J (t 2.} ( l J} (3 4 5 6 f. 2.} ~

( 1 S) ( 2 6) (3 5) (4 6) (8 2 1 7 4 6 ; 5} =(t 3) (3 7) X
(4 5) (5 8) {8 7)
8 ' 3 11) There ex1st four rolatrrms (If lhP squar~ ln tht p1an~
"hu~h transform the rorm~r 10 to J tsP] f by 0<> by 90~ by 280c and
by 2 0 These rolalJons torrespond to pPrmutatulns {1 2 3 4)
(.2 3 4 1) (3 4 l 2) t-' I 2 3) The first permutation IS
-o[ ih-e (4 0 0 Ol ty~ wh,le the rema~n~ng ones are uf ll'.e
(0 0 0 1) f) pe The c) eJe Jndex bas the lorm Po (1 1 t'l t,~ ttl
(t: + 3t.. )/4 (2) In addlhon to the permutatJons mentJoned lD
8 i 3 (1) we a} so ha\e four pernrutat1ons t1 ~ 3 2) (3 2 1 4)
(2 1 4 3) (4 3 2 {) .correspond Lng to the rotatLODli O( tbe
square about the diagonals Tllo of them are of the type (2 1 0 OJ
, b1le thta olher tv. o are of tl,e I) p~ (0 0 0 t) lJ<'nce Pa ::::-
(i: + + 5t 11 2ttt )/B {3) 'I here -exu;l 12 t'6\al~ons of tbe tet'ra
hedron 1dent1cal eJ,ght rot~t1ons by J20° about the axHi p.ass•na:
throngh J!s vertex and the c~n Ire of the- oppos1 te fac~ and three
rota t1ons about an a1: JS pas~Jng throvgb thP mtd poJnts of oppos1te
ed~J llell.(e PG +
it~ 8t 1(, +3tl}li2 (-4\ Pc (t~ + Sti +
3t 2 tf)/22 (5) Pc = (I' + 8t~t 3 +
3ti)112 (6) Po (t 1 +3t 1t 1 +
2tt)/6 (7) Pa == (tf + 8tJt, + 3tj) 12 (8) From 24 rotatJons of tl1e
cube- on~ lS i d ent 1 cal '\.hlE-~ are rCJt.a l1 oos b-y 1 S0° and :sr-I are to
ta ttons by 90° about the .s tr a1 gh t l1 nes 'Pa.ss1ng through the ccn tres
of oppos1 te lace.s sJ.:r are rots. tzons by 180° about the .stra1ght hnes
pas.s1ng through the mtd poJnts oi oppos1 te edges and f?Jght are
rot&tt.ons hy 120° about axfs ec.nnett1ng op~'l-e: v~rttces The
Jden t1ta 1 perm uta t1 on g1ve.s 5.1 x cycles oi 1engt h I the three per
.m uta ll ons ..-:orr-rspon dIng to the rota t tons h v t Bot~ g1 ve two cycles
ol length 1 and l wo cyr l es of 1ength 2 l he SJ x permtJ tat 1on s corre---
li: pond 1n ~ ta the ro talL on s by 90° gt ve t.wo c "!J; cles of lenglh t aud one
CJc)c uf frngth 4 and the otbrr ~~.x permutat.1ons cortt?spondtng to

the rotauon.s b) 180° gJve three cycJes of Jength 2 The ~tght per
m uta\~ nns t onf-3 pond' ng to tbt rota \l on~ by 120° g l v-e two t..yt..les
of Jpn.gtb 3 Hence Pc ::- (t: -r- 3t'ti + 6t{t 4 + +
6'~ BI~J124
(9) \\le ha \ e five t) pe9 of rota tJOD by Oa (.sUite of rest) COtreliipondJng
to the 1dent1ty permutatJon the rotation "by 9[r> about the d1agonal
the rotat1an h\ 180° ahout the -diagonal the rotal1un by i.SOO about.
t h~ stra 1gh t Jtne passJ ng t hro tlgh the m l d poJ n ts of op p os' t e laces
and the rotat1ons by 1200 about the stra1ght Jtne paq,sJng through
the nndpmnts ol vppos1te fttcts The ldentlty permut&.\lOD .canes
pondtng to the state of rest g1 vcs the GantrLbutton to 7 G equal to
t: the .second type permutatton g1ves t}t the thJrd t}pe permu
1:
ta t1 on t j tbe I ourth type perm uta t 1on tf and the fifth type per
m uta t u>n tl E a.th ()I l hese t~ peSi cr..nespoo d~ to ' G 3 G and
8 rota tl on.s res pee tJ v~ I J- H~n ce P c
8t~)/24
t: +
(t• ..-l 6 t.., 3t lti -r 6ti ~
8 4 4 {t) \\ e prove thJs stat~ment threct1y by us1og BerusJde s

t~ mma The gro.u p G -cons~s ts ot perm uta.lluos , 1 (t 2 3 4}
.n, == (2 3 4 1) n .s ~ (3 4 1 2) :rt ~ ( 4 1 2 3) It Is
1mmedt~teJy seen that for aD) pa1r j ~ 1 1 ~ 4 there exJ~ts a per
muiat1on :n sutb that n. -- J Thus all the elen1ents are equ1~~len\
and we have one equl valence class
Let us obta1n the resu1t by usJng Berns•de s lemma \Ve have
l G I ~ 4 b1 (n1) = 4 b1 (nf) = 0 lot t = 2 3 4 lienee v (G) ~
CHAPTER DIGH'l' 383
(4 + + + 0 0 0)/4 = 1. (2) Note that ~he elemen!s 1 and 2 ?re

transformed into each other by permmatwn .:rt 3 , \\lule permutatiOn
n transforms elements 3 and 4 into each other, hut none of the
p:rmutations transforms the elements of the set {1, 2} into the
elements of the set {3, 4}. Thus, we have two orbits. Applying
Bernside's lemma, we obtain I G I = 4, b1 (n 1) = 4, b1 (n 2 ) =
b1 (.:rt 3 ) = 2, b1 (n 4 ) = 0, v (G) = (4 2 2 + + +
0)/4 = 2.
8.4.5. We mul>t prove that I G I v (G)= 2J
bt (tt), where I G I
:nEG
is the order (number of clements) of the group G, v (G) is
the number of classes of G-equivalence (orbits) on the set Z 11 , and
b1 (n) is the number of elements which do not change places
upon permutatiOn .:rt. We put Gy--,; = {.:rt EG: .:rty = x}. If 11! s Zn
is a certain orbit and x Elll, then G = lJ Gy-.x· In this case, we
yEM
obviously have G,_x nG!J->X = 0 for v =I= y. Note that I Gy-+x I=
I Gx .. x I if y ~ x, i.e if Y belongs to the same orbit as x docs.
Indeed, if aEGu-x and G.>._"C={.:rt 1, .•. , llm}, then {crtt 1, ••••
cr.:rtm} EGy .. "C where an, =I= a.:rt j for i =I= j. On the otl1er hand, if
Gy-.x={ah .. . , cr1,} and crE Gy-v thl'n {a- 1 cr~o •.• , <r 11Jh} s Gx-+x
and cr-1cr; =I= a-1ai for i =I= J. Hence it follows that l Gu-.x l =
I Gy~x I· Let now M to ••• , M, (G) be orbits and x; E 1l:fi. The
v (G) v (G)
2J bt(.:rt) = 2J IGx ... '"l= 2J 2} I Gy-x.l
1
= 2J IGI = IGiv(G).
xEG xEZn t=l yEll! i i =1
8.4.6. (1) Each permutation .:rt in S 11 of the type b = (b 1 , -

bz, ..• , b11 ) can be represented in the form of the product of
cycles so that the length of the cycles does not decrease: ;n =
(iJ) (12) ••• (ib,) (ib,+t' 'b,. 2) •••• Such t\\o notations ran gener-
ally lead to different permutations. This can be in t\~o cases:
(~) when identical cycles occupy different places in these 'nota-
tiOns, an~ (b) when the cycles are equivalent (as the cycles of a
permutatiOn) hut start with diiTereot Plements (e.g. (1 2 3) and (2
3 1)). The first reason leads to the repetition of the same permutation
n
TI
l<=l
b11! times, while the second reason, to the repetition of
fl b
TI
It=\
k It times. These reasons are independent. (2) By definition
Psn (tt. .... tn) = (nl)-1 ~

:nESn
11
~ ~ .... t~• ... t~n = (n!)-1 ~ h (b) IT (3) We have

b 11EH (b) b i=\
~r UTIO~ "=- A N'it'\ rRS .A 1\ n Ifl\ T~
)= JI et:p {t~;,rAJI.} =
"'~'
IT { ~ X~ bitt:l,r' bll (bl! n 1) The coeffic leltl of r" IS ef1 ua 1 to
k-l b~~t\
the ~;; um over -p~~ Lble 1wn 1W gall ve tn t(l g~t~ I 1 b1 ~ '.le lt
n I f
t hat b t + 2b2 + + l bn , of t crm ~ of UI e rornl 11 l:~c ~ (l 1 rA k} 1
h I
8.4 7 It should b~ not~d \hat a C:}~le n{ all ~'\tO [~n~t l\ ~~ 'n
odd permuta hon "lu ]~ -l f'l cleo ( r r n.n ud ](lnQ:t h 1-=i an (J\ ~n pt r
mu tat lOn Eac11 (l(itl p~rmu ta tton or the- t~ pl (61 blf} llJa ke!-~
to the ex press ton P 5 (1 1 t n) 1. ('ontrJ bulJ on t~ t
and to the elpres«Jon P s (tl --t (---f) ' t t ) a ccJn tr1
huh on t~ 1 t" Therefore the terms on the rJght h.and s1 de
of the equah t) to be proved \\ h~ch corre~pond tn ~11id pe-r1nuta tron.s
are ca nee11 ed out wh1} c the terms correspond 1ng to c" cn p(l'rzn u
tal \ons are dou hl-e:d B-est des '- t can oo
~a~rs. t y se-en that I l n ~ =
( Sn. l/2 == (nl)/2 Hence lol1o\l.s the .statement
8 4 8 (f) \Ve mw;t prove that the set (n ' o} \\ tth the tuurt
pl1cat1on operatton lorn•s a group that 1s
{1) thel'e ~xls'ls a un1t element
(2} rae b. element has t ts. ln verse
(3) the opera tLon 1s assoc1a tl vr
The unz t e]pm~nt J.s ohvu;u.s]y ttr> X aD v. lu~·re "lo (aq) 1' 1 nnl t
permutation tn the group G (r~~peet1ve}y1n tl1e group I}) Thr
Ln\~rse f lem~nl to n X q 1! a petrou.tat1on tt 1 ' a 1 "here n 1
and a 1 are the corre.s:pond1ng lnverse clemcn ts It r.cmn zns (() l r 1" e
I hat (n 1 ' o1' X ((JT~ '< o 2 ) '< (n, X o 3 }) ::::: ((.n 1 ' o 1) '< (!I~ "
0' 2 )) X (.tr., '< o3) Let us constder the actton oJ a pecmutaunn on
~ n '@' lem~n t v E X U Y \V 11hou t los' of genera {1ty we ca. n a o::.s\\ \n ~
that u E X Then (nl x ol) X ((n, X a'2l X (a 3 X cr3n v
:l:t (n 1n 3 ) u ((n 1 X at) '- (n 1 "'< a 2)J ' (11 3 )\ o~J l -- (lf 1 .ni) Jta
and the associa tl \ en~ss rolla\\ s from the fact that th(' Jou1tr p1u. a {Jon
ope-ra tton 1n the group G lS assocJa t1' r Thr! rdcr ~~ 1lt0 grOUJl 1~
-Gbvlously equal t.,. \ G ~ ~ If \ sult~ th~ rle:rntnls n and iJ f \ ht
permutatJon n A. a arc chosen Jndepen Jcntl\ and the tu per
muta t~ons n:1 \-.. u 1 and n:r '< a:! are ob" 1 lllsl\ d J Ucrent J ( .n 1
Ot 01 =/= 0 2 (2) J r 11 (r ) I~ the nUJUb~t lf C} clt.~ f length
n1 *f th(l
pE-Tn'lU t& t llm n 1(J) nn t hP ~e-t X ( n
t hl n t lu:~ n ':tlll) b~r { ( ' (" ~(J-8 1) f
length t o[ the Jlertnutatt.on .-,; X a LQ. tqual t(~ h +
c (~l \\ ~ h :t'
I ~ d (T1 I
Pax Jl I, I t 11 > ' G x II I ~ t\ ' t 11 = Ic r 1111
t€ i If
br (n)+c"(a) (Jt)
In
CHAPTER EIGHT 385
4 9 (1) 20 The necklace of seven beads can be co!Ot~retl in

8. • • . •)'7 _ 1'>8 wars We have a set G of seven 1.hfferent
two ~olours Ill - -n. Mtransfo~ming the necklace into itse!f. The
rotations n1, • • ., • . 18. 7 0 0 0 0 0 0) whlle any
type of identit¥ p~rmuta~on .
~thder p~rm;ta(~lon 15 0
\o 0 0 ooo
~ ~ ~ t{tf+ 6t,)h. 'A~co:di~g to Poly a's
t). The cycle
Jtnh ex IS thee nlu'nl·b·e~' of diffe:ent equivalence classes is Po (2,

2 eorem, 2 2, 2, 2, 2) = 20. (2) The eye 1e me · 1ex I~ · p (t 1, · • ·, t n ) ·-
(t\1 +'(II- 1) 111 )/11. The number of necklaces JS P (k, ... , k) -
=
(k" + (II- 1) k)/ n. . .
8 4 10 (1) The cycle mdex JS (see Prob 1em 8.·1. ' 3 3)) jJ -
( G -
(t~t· sit 't 3tll/12. According to Polya's theorem! the numbe r
of1 colo~kgs i~ Pc (2, 2, 2, 2) = 5. (2) The cycle tudex ~G ca!l
he taken from Problem 8.4.3 (!l). The number of coloun ngs !S
pG (3, 3, 3, 3, 3, 3) = 54. (3) Let M .be. the set of cube fac.es,, G
the rotation group and N the set cons1stmg of three colours. red,
blue and white. We ascribe the weight :x to red, y to bl~1e and :
to white colour. The cycle index. (;,ee Prob.lem 8/1.3 (8)) IS Pa =
(1s .L 31212 + 6tzt + 6t~ + 812)/24. Accordmg to Polya s theorem,
IS~ Pa (I 1 , I 2 , I a• I ~ ), \\·I1ere
I ~. ' I h -- .t It -1-
tl:ieI IfunctiOn
I .2
countm • •
g series .
1
yk t zh, k = 1, 2, 3, 4. Therefore Pa U1• fz, fa, /~) = ;M ((x +
y + :)6 + 3 (x + y + :)~ (x'l + y2 + ::2 ) 2 + 6 (x + Y + :.)2 X

(x' + y' + z') + 6 (r + y'! + :.::)3+ 8 (z3 + +
y3 ::.3):). The num-
ber of different colourings for which three faces arc red, two arc
blue and one is white is equal to the coefficient c3 , 2 , 1 (Po) of
:t y z in Pa U1, /21 fa, /~). We have
3 21
c3·2•1PG U1, lz, fa, /,) 1"
~c 3 ,2o1 ((x + y + z)6 + 3 (x + y + z) 2 (x 2 + y2 + :2)2) = 24
( 31 ~; 11 tsx 2x2) = 3.
8.4.11. Po (N, N, .•. , N). It follows from Polya's theorem.
8.4.12. (1) If a rooted tree has k > 1 vertices, it has 1 ~ n < k
edges incident with the root. The subtrees "planted" on edges which
are incident with the root will be referred to as branches of the
tree. If the branches are isomorphic to one anothe r, i)y transposing
them we can obtain a tree isomorphic to the initial one. The per-
mutation group of branches is a symmetry group of order n, and
the figure counting series (branches) coincides with the rooted trees
counting series T (x). The result follows from Polya 's theorem.
8.4.13, The solution is similar to that of the previous problem .
II the connectivity components of a graph are isomorphic to one
~n?~her, they can be transposed to give a graph isomorphic to the
lllltlal one. For a graph with n components, the permutation group
of components coincides with Sn. The figure counting series is
l (n). The result follows from Polya' s theorem.
8.4.14. (1), (2). Let n be a permut ation of n elements, having
k cycles. The element n in it may form a unit cycle. Then such
a permutation can be put in a one-to-one correspondence with
a permutation of n - 1 elements with k - 1 cycles (the number of
such permutations is equal to P (n- 1, k - 1)). If the elemen t n
does no~ form a unit cycle, the permut ation n can be obtained from
a certam permutation of n- 1 elements having k cycles each by
25-0636
386 50LtJfl0'\lS Al\SWE.RS Al\D Hl~rs
1nel ud u1g the e1emen t 1n to a certa Jn cycle For a .fi .1 ed perm uta t1on
of n -- 1 elements "1th It cycles., sueh an incluston can be carrJed
out Jn n - 1 n l) s (the element "' cannot be placed first sJnce n J s
th-e lnrgMt eleme-nt} II-ente 1t. IoUows that, P {n. k) = P {n - 1
k. -- t} - P (n.- t" k) {n.- t) l{ulllplyLng the e.quah.ty by t"'
and Ia 1t..J ng the sum over kt we obta1n the relat1on .P" (t) =
(t + n - 1) iPn-t (t) between the generattng functtons By JD·
duct ron tak1ng 1nto account the equal1ty P, (t) t we obta1n =
.Pn (t} == ~ (t + 1) (I+ n ~ t)
8.5.
S {i} riote th~t n ~ {i + ~} ('1-.... ~) < ({n + ;>l2)' fv:r 0 ~

B .5
l<n Hence n»Jl~ fl
0+1) {n-l)~J"JI~n.tl (t+t) ll
1~i<nJ2 t ~i<n/2
(n- t) ~ ({n + 1)/2)n (2) Note that
n
(l+ l) (2n.-t)
n
~ ""tn+1)
~
\'.}) \\t nnl" ( i+-~ r =2,{ ~} 1r~ ~~ il ·~2.!.- ~ :It <

1~6 k~b A~J
3 (4) lndut\ ~on on n for n -==-1. the tDtqu11l \ty !S -val~d
+
I I (n/3! 71 < nl then ((n 1J/.3) 11 •1 ~ (n 3)11 ((n + f }(3} ((n+ 1)/n }1J <
(see (3}} ~ (n/3)" ( n+ f) ::;;; (n +I )J (5) Note that (nr) 1 ~ (2n)! z-n
and use the result of Problem 8 5 t (2) (6) Use Cauchy s 1nequal
h;
1ty y alai alt ~ (dt +at,.... +a,.); k 11.l ;a: 0 and the fact that
the artthmetlC melin of the multlplu!rs on the left hand sule of
1i:
the 1nequatny 15
"' i1=-2~t+
n (n 2+ i) .!J "3 t- (7) Putting an=
i~i
+
(2n.- l)UX V 3.n 1 (2n!r)-1 , prove that G~+ 1 fa~ < f Further, carry
out Inductton on n {8) N()te tbat t (2n- n < n"' for 1- <:: n (9) en~
01;:1
~ n1cJJd > n'J'/n! ~10} For n~ 1 the tnequa.l\\)' \S oh"1ously

~:;g:[]
val,d \Ve e.ssume lba.l (! + «.)'~ > 1 +o.n fo-r all -1 and a
certain n;;;.: 1 Then (t + o;)n+t ==== (1 +a:) (1 +a:)»~ (1 +a:) (1 +
'a.
an)=-f+a {n+J)+" 1n ;> f+ce (n+f) (t1) Expa.od1ng both
s1des of the mequal1ty accord1Dg to Nev. tons hLnom1al theorem
"~ tompate the terms of the cxparunon Wlth the s.ama nuruhers
We have ( ; ) n-.1.1 = ( n~ :1 ) (~"~+ 1) II. for k =0 1 Furl her
the ratto ( ~ ) n-11./ ( n;

} {n+ 1)~k 1s
1
equal to (1- k/(n+
+
1)) (I -1/(n. 1))A In v1ew <>f 8 5 1 (tO) the rat to does not exceed
1 There lore, e.acb te rnt of expans !On o t (i + t 1n) n does not e.x ceed
tbe corresponding term. in the ex pa ns1on oi C1 + f I (" + j))~+l
• CHAPTER EIGHT 387
Besides in the latter expansion Wt' also have the (n+ 2)-th term
(n+ 1).!<n+1) > 0. Hence follows a strict inequality.
8.5.2. (i) We put ak = ( 2 k+
(n-k))
1
k/ (
n ) and ver1fy
·k . that
ak+ 1 fak < 1 for k ~ 1. Since a 1 = 1, it follows that the first
inequality holds. The second inequality follows from Prob-
lem 8.5.1 (9). (2) (n/k)k ~ (n)k/kl for 1 ~ k ~ n. The second
inequality can be obtained by using the ratio ak+ 1 /ak, where
ak= ( ~) kk (n-k) 11-kn-n. We have ak+IIak= (1+ (r~-k-
1)-l)ll-k-1 (1+ 1/k)-k. Since {1+ 1/m)m increases monotonically
with m (see Problem 8.5.1 (11)), a1a 1lak ~ 1 for k ~ (n -1)/2.
Considering that a 1 < 1, "e lind that the inequality ak < 1 is
valid for k ~ (n-1)/2. For k > n/2, the inequality follows from
symmetry considerations. (3) We prove the first inequality by
induction on n. For n= 1, the inequality holds. Assuming that
it is valid for a certain n ~ 1, we have
2(n+1)) = 2 2n+1
( n+i n+1
The second inequality can be proved by induction and using

the result of Problem 8.5.1 (7).
8.5.3. (1) (2n - 1)11 = (2n)l (nl)-12-n - Jf 2n (2n) (2n)2n X
e-zn(2nn)-lf2n-nenz- n- V2- n"e-nzn. (2) \I 2Tl )
= (2n)! (nl)-2 ~
11
114nn (2n)2"e-zn (2.nn)-In- 211 e211 ~ (nn)-1/ 2411 . (3) nl ([n/3]1)-2 X
((n-2 [n/31)1)-1 - (2.nn)l/ 21l11 e-n(2.n [n/3])-1 [n/3]- 2[n/ 3l e2[n/3] x
(2n (n- 2 [n/3]))-1(2 {n- 2 [n/3])-n+2(n/3] e11 - 2 [n/ 31- (2n)-l X
[n/3]-l n (n-2 [n/3])- 1/ 2 n11 [n/3]- 2[n/ 3] (n-2 [nj3J)-n+2[n/3l. We
note that {n/3)-1 < [n/3} ~ n/3, n/3 < n-2 [n/3} < n/3+2. We
have [n/3]-1 (n-2 [n/3])-1/ 2 ~ ((n/3)-1)-1 (n/3)-1/Z ~ (n/3)-3f2 x
(1 - 3/rl)-1- (n/3)-3/?; [n/3]-1 (n - 2 [n/3])-1(2 ~ (n/3)-1 ({n/3) -
2)-112 =(n/3)-3fl. (1-6/n)-1/ 2 - (n/3)-3J2. We put ct=n/3-[n/3],
0 ~a< 1. This gives [n/3] 2[n/a] (n- 2 [n/3])11 - 2£n/ 3l = ((n/3)-
2) 2[nta] ((n/3) + 2ct)11 - 2 [n/J] = (n/3)11 (1 - 3ct/n) 2[n/3] (1 +
6ct/n)"- 2 [n/ 31. Further (1-3ct/n) 2[n/ 3l=(i-3ct/n) 2«nJ3)-a) =
(1-3ct/n) 211f3 (1- 3ct/n)- 2a ~ (1 - 3a:/n)211f 3 - e- 2a. Similarly,
l1+6ct/n)"- 2[n/ 31 ~ e2ct. Hence we obtain the required asympto-
tic equality. (4) (m+1) .. · (m+n) = kl (n+m)l ~ kl
(k+1) ... (k+n) m! (n+k)l m! X
, /' 11+m kl
V n+k (n+m)"+m (n+k)-("+k) ek-m _ ml nm-kek-m (1 + I
•
mfn)n+m (1
f
+ k/n)-n-k ~ kl nm-hek-m ( 1
ml
+ m)
n
n ( 1+ k
n
) -n X
25•
388 SOLUTIO'lS .ANSWERS A.~~ ltr'lTS
(1 +-i r 11
(I+ ! r"- ~~- nln-}1; Ia the J~t case .,e ban used
s.symptot.1G cqw.hbes a.[ form ( 1 + -: )n _. e"' (l + (m/PL)),.. -

e• 1tt. ...... 1 for m' = o (n) {5) (~)U ((2.1J- i)U)-1 ~ (2n)l ((2n --1 )11)~
1
Then "'e can use St•rlmg s formula and the resul' of 8 5.2 (11
I
ss, <U rJ <1+1),.dt= n +t 1 u+t)"•lr-= 2""tl
I o n+ t.
- 1
n+ t
0
t 1 • n
Oo the otber band ~l(l+t)'"dt= ~ ~ ( : ) lJt.= h (:)X
0 0 Jl.-(1 ll~O
I n
~ ttdl= 2! .t~t ( : } (2} Us~ng the result Of Prob-w

0 A-=--0
I em 8 1 f9 (3) "e ohtam
k-t
=! 2B+ h
It should be noted tba\ 'txP{-2n:crwi}k}1 =land 1i+ exp{2.niv/k}l ==
2n:v 2nv n:v ( 1v ~v ) t
Ii+tos k +•s1n k
n\P
= 2 cos k cos k +1 sln k ~===
n:
.2 cos ,. · ~ 2 cos Hence
k k
Jl -1 2n i,.v 2n hi n
'\1'~1
{3) UM the result of Problem 8 ~ 20 (l) not1ng 1ha' 1 l +

a.e1n"l"'l<l-o:+2c:tc()S ~'II <i+a tor Oo:::v<k (5} Usmg
Problems 8 t 18 (9) and 8 5 3 (2}, "e obtam
•
CHAPTER EIGHT 389
8.5.5. (1) Yes, they are. The validity of the inequalities follows.
from the fact that
( ~ ) ak ~( ~ ) bh ~( ~ ) ch •
•
{2) No, they are not. A counterexample is a sequence {b~} such that
bk = 2 x 3-h for even k and bh= 3-k for odd k. Puttmg ~ = _i/3
and c = 2/3, we obtain 0 < ah ~ bh < ch < 1. However, m v1ew
of 8.1.20 (1),. we find that
,, Jl
~ ( -f)h ( ~ ) bk = ~ ( -i)h ( ~ ) ( ; r
h=O h=O
+ ~ ( ~~) ( ; rs =(1-1/3)1l
s
II
8.5.6. (1) We have Da = i

11
LJ'Y, (ai- a) > i
n
i=l -
ai: lai-a I ;>I
-a)~~ Ott 2•Hence Ot ~ Da/t 2 • (2) Let us consider a set A= {a 0 ,
a1, ••• , a n_ } in which a.., is the number of unities in the
2 1
~
binary vector a=(at. ... , a 11 ) having the number v. It should

be noted that
-a=2-n a v--?-- l l
Further, ~e note that the number on the left-hand side of

the inequality to be proved is equal to the number
of av for which I a,.- I ~ t a
In view of Problem 8.5.6 ]In.
(1), this number does not exceed 211Da (t yii)-3. But 211Da =
h (av-a) 2= ~(a~--a2)=-2 11az+ ~kz( ~ )=-211(ni2)2+
0~\'<2
11
O~v<2n O~k:~n
(~~*n) 211 -,2 = nzn-z. Therefore, 211Dai(t]ln) 2 = 211 -2jt2. (3) 1\Iino-
matron. "e put t=ln 11. Taking into account 8.5.6 (2), we can
II
write ~ 1
• (
11
) > ~ k-3 ( n ) :::;;,: 1 _ x
h- 1.- k LJ k ,...... (n/2+ t ]I n)2
-I k: I h-n/21 <tVti
~ ( ~ ) ;> 4
-. (211 -211Jt3)- zn+2fn2. l\Ialori-
(n-4q/ n)·
lh-nf21<t '(;j'
300 '80Ll.1Tl0NS ANSWERS A.liD BIN'TS
%1lllon We first not~ that ( ; ) <( e; }" m l'tl!w of 8 5 2 (1J
and ( k~ 1 }/ ( : }~ ~~!+ f < l/3 for k ~ n/~ 'fbert'lore,

"
h
l~l
k""' ( ~ ) c ~ {: ) +
Jt~-nJI,.
!: (:) +
~a t ~~. - -n.J2 1~,rn
(!C.
{t-c vn} -z ~ (:} ,;;;; (4e)"l' ~ 3""'+16n-~2"Jt'+

A: ll-11/21 <IYB t=D
; 1 (l +tn- 11).zn ~ n "2"+1
1
8 5 7. (1) U61Dg rormula (fJ), "e obtaw

n ) ____ Jf;\•n11(i +O(l/J\)}
( k --.. }lin h. { n. :__ k) k'-" (n ~ k )'fl.pA.
We put ~~ n}2- k Then tbe obtaJned expression can he '\\71tlen u

n: t1 +O i1/n.))2"i- 1
Further, we have
In [( t-t t' -~ 2
( i+ ~ r/H~]=( ~ -z} In { 1- ~)
+(; +z}tn (t+ ~ }=( -;~-z) (- 2
: ~{ ( ~r~ )
+(~;·+z} ( :Z--~!-(·~ )2+
(2) As m 8 51 (i),. by us1ng
Putting a= a/(4+ f), t1 = 1/(a + 1}, z = an-.. k~ "'e get

n } ,., («+ t) cJJa-z (i + 0 ({fn))
( 6ft.=- X
k J/2JUUJ (t -~/an) (I +-:t:~~n)
(a:-z/n)-an+.1: (~+z/n} -_tm.-.:c
..._ +
(4+ t)" (i O((:+ill~))
r- Y +
2~ (i- :r.fttn)()n -% (i ~~~n)CJn-~ •
•
CHAPTER EIGHT S9i
Further, we have
In ((1-zfan)an-x (1 +z/~n)l3n+x)
=(an-::) ( -::/a.n-x2/2a. 2n 2-x3/3a.3n 3- ... )
+ (~n + x) (x/~n- x2/2~2n2 + x3J3~3n3- ••. )
= x2/2n (i/a.+1/~) +O (x3jn2) = x2 (a+1)2/2an+o (1/n).
Finally, we obtain
n ) ah = (a+ 1)n+1 (1 +O ((x+i)/n)) e-(x(a+ 1)) 1 /2an

( k V2nna
x+i
where x= na (a+ 1)-1 -k. Since x= o {n /3), -'-- = o (n-1/3),
2
n
Hence it follows that
1-(::) ~ (2n)-lf2 e -x•f2 (x-1- z-3) - (2n)-lf2 x-1 e -x•tz.

f -x•/2
We put D = (xm + (h/2})- (xk- {h/2)) and '¥ (x) = e •
X
It remains for us to prove that
D- rV2ii)- 1 ('¥ (::k)- '¥ (xm»· (""")

Using (•), we obtain
D > (xm)- (xk) - (2n)-1/ 2 ('¥ (xk)- '¥ (xm)).
On the other hand, according to (•), we have
Y2ii D "-' '¥ (xk- (h/2))- '¥ (xm + (h/2))

= '¥ (xk) (I+ 0 (xmh)) -- '¥ (xm)•
For Xm - Xk ~ 1, Xk-+ oo and xhh -+ 0, we obtain '¥ (xm) =

o ('I' (x~t)) and V- 2n D ,..., 'I' (:: 11 ) ,..., 'l' (x11) - 'l' (xm). If, however,
Xm -x11 < 1, we put 11 = 11 (n, m, k) = (xm -xk) h-I. It follows
from m-k-+ oo as n-+ oo that 11 = 11 (n, m, k)-+ oo. We note
2 2 1
h
t at (xm-Xk)/2 = (xm- xk) X (xm + Xk) > h11k11.. Further
2
Y2ii D,..., 'I' (xk) (1 + 0 (xmh)) - '¥ (.xm) '¥ (xm) + !J.,
= 'l'(xk) -
where ll=O ('I' (xll) xmh) =0 ('I' (xk) x~th). But '¥ (.xh.) - '¥ (.xm);:;:.:
'Y(xh) (1 - exp {- (x~ -x~)/2}) ~ 'l' (xk) h11 Xh_. Hence it follows
that !J. = o ('I' (::11)- '¥ (xm»· Consequentlv, y21i D ~ '¥ (zit) -
.392 SOLUTIONS ANSWERS .AND HINTS
lf (l:m} Q E D Thus lt fellows Itom {•} and (•*'•} tb.at

~ { ~} fl'Y ...... (2n) 1
t' {'V (..-:h)- 'I {zm))
k~v~m
____ __ ___..
8S7 t4)'\\-eput.i'.t=~+2'Vlnn h(z)= nd~~~;;

'I hen
~ (:} av= ~ {:) av+ ~

V~h(~) k(.x}~y~.k (:r:~~)
But
~ ( ~) a~ ~ n ( :J a~•
.,">I\ (x:~}
where vt. =-== lk (.z 1} [ In. v tew of 8 5 1 (2) we hence obtaln
~ (:}a" ~ n (2nna) 1
J"' (1 +a)""' 1 exp { -.tV2}
~>Jl <x. )
=nl/ 1
(2.n:) 112 (l+a}'u·l exp {- ~ -~ Vln n-Zln n}
~ n- (2n:)-lf:l (1 + a)n""le -% / ~.
1 1 1
( •)
On the other hand It follows from 8 57 (3) that
h (~) ri''- {a+2t)" ( e-:/2 - e-;:ll }

3
(u)
h (:t}~~Ji. {Xt) vn
The requ.tred est1mate follows from (•J and (••}
8.5 8 (1) Us1ng InequalitY (f3} and the 1nequal1ty n' >
V2.n:n n"e-n follow1ng from 1t we obta1n
n } ]f21tn n11 txp (-- n + 1J(12n))
(
Art < 2n:JJ Jfl.~ (A.JJ) )..n (p.n )~'n ex p { - An- 1--1 ~}
e1/(12n)
= ,.., G(n A}
V2:rnA.p. A)..nJ-4JJ"
On the otber hand
Y2nn nn e.xp ( -n}
~ e1p {-- (1/(l2n }} (it!+ 1/~)} ~ G (n A)

-- { 2nn.Al' l.~11 tJ-I-1"
CHAPTER EIGHT 393
(2) The minorant follows from the solution of 8.5.8 (1). In order to
obtain the majorant, using inequalities (13) we find that
n) { i
(A. G (n, A.) exp 12n - 1
12A.n -
1
121-m + 360 (A.n) 3
1
+ 360 ~!-til ) 3 } •
Without loss of generality, we can assume that A.;;;:::. I-t· Then

1 1 1 1
1/(12n) < 11< 12 A.n) and 360 (A.n)a+ 360 (~-tn)3 -121-w .::;;; 180ft311 3 -
12 ~n < 0. Consequently, ( ~) < G (n, A.). (3) Using the minor-

1
ant from 8.5.8 (2), we obtain ll)
(k > G (n, A.) e - 12n>.11. But
exp {- (12nJ.!-t)- 1};;;;. exp {- (3n)- 1 } ;;;;. exp { -1/9} for 11;;;;. 3,
exp ( -1/9} > l'zt/2. For n = 2 and l..n = ftn = 1, the equality is
observed. (4) 1\lajorant
•
tl ""'
~
h=>.n
(z) < ( ~J 2}
i=O
1
( ~ 1.. r = -=n-~-:-1 ( ~~~ ) •
(5) The inequality can he easily verified when (a) 1.. = 1/2
since ~ (~) 11
<2 = ( ;
11 2
' (; r r"' 2
; (b) l..n=n-1
h~n/2
(the left-hand side is n+ 1, and the right-hand side is
( n_:i r- n); (c) 3:s;;;n.::;;;5,

1
n/2<1..n< 11-i. (This can be
checked directly.) Let now 11:;;;. 5 and n/2 < J,.n:;:;;; n-2. It fol-
lows from (4) and (1) that ~ ( ~) :;:;;; 1.. (21..-1)-I ( ~~J :;:; ; J..l/2 x
k;;;;.,.n
(21..-1)-1 (2rm (1-l..))-lf2}. -AnI-t- 11 11 • We put f (I..)=J..lf2 (21..-1)-1 X
(21tlt(1-J..))-1 / 2 • We must prove that /(1..)<1 for I..E[1/2 e, +
(n-2)/n], where E= 1/n for even 11 and 1/(2n) for odd n. Differ-
entiation with respect to 1.. reveals that the function f(l..) is convex
down on the segment under investigation. Therefore, the max-
imum values must be sought at the ends of the segment. We
have I { ~ +e) = J,/ ;

+ e (1 + 2e)-l (2nn (1/2 - e))-1/ 2 =
((1 - 4e 2} 2nn)-1J 2 ~ ((1 - 16Jn2) 2rm)-1J2 < 1 for all 11:;;;. 5. Fol·
3M SOLUTIONS ANSWERS AND RlN't.S
n > 6 we. h&ve I ((n- Z)ln) -= v'i. -Z/n {t- 4Jn)-1 (4n)-1J1 <
{1 - 4(n.)-1 (4n)...1fl < t 1! )., <: 112.. than ~ :=:: t ....... l. > t/2
{6) S1nce { ~) = { n :k ), uslDg 8 58 (S) we obtanl
1\
~ G} = ~ {~} <~~pn~-An
0 ~.It~.,. ll:-prt
~-l ~-'
8 59 (1) tn)" = nil. fi (1 - Un) =- nl. up ~ ln {1-t/n},
j~o i~l
l- t t- t QO ~ ll.-1.
z
t~1
ln(i-i/n}=- ~ ~
i~1 ~::.!Li
-!-( ~ r=- h
\J~1
v!" ~
t~i
..v
(2) We have, an the one hand, {n.)" < nl. wh1eh 1s cbv1ous On.
the other hand, (n)t ~ n
l-1
i$"0
(n - i) > (n __. 1-}Jt == 11A(f - (k/n))k ~
nlt. (t -k"Jn),.,.. (because k~ o t Vn)) ~ n'- (1-o (1)) {3) We use

k-1
tb~ fact (see Problem S 5 1.3 (2.}) that :E ,v ~ (v +i )-lJc;v+ 1 +
,~t
0 (kY) Heru:e an VteW of 8~ 8 (1) for lc-+ co and k~ Q (n) we

have
00
(n}lt = n"- e:.lp -- ~ ('VnY}-1 ((v + t) 1 k"-t 1 +0 (kY))

v~l
..,.,
==- k'H' nv + o { 1.v ( .nk-)" ) )}
n"' exp { - 2J ( ~ (v+i.)
'\l'zi!S
~ kv+t ~ k
~ nY
"
~
( k
n
}~ ~ k (
n
k)m 1
t-kln
~ 0 ( !!m.:l )
nm
'\l~fQ \'~
H !ll[je we obta.1n the reqult-ed equa b t:y (4) U:se 8 5 9 {3) i: or m ":::I 3
8 5 10. (I) In trtew of S 1 t3 (6) and. 8 5 9 (t} we have
{ ~=: )/{~)~ ~::- (: )'

(2) hiake m& of 8 59 (3)
CHAPTER EIGHT 395
8.5.11. (1) We have

k-1
(n;s) I G)= (~:;-:>k =II (i - n~i }

i=O
k-1 k-1 00
=exp{~
i=O
ln(1- n~i )}=exp {-~ ~ ~
i=O v=1
( n~i rl
k-1 k-1 k-1
=exp { -s ~ n~i-; ~ (n~i)2- ~ ~ ~ (n~J"}'

i=O t=O i=O v;;;:.3
Using 8.5.12, we obtain

k-1 k-1
i It ,, 1
h n-i =ln n-k+1 ~
i=O
(n-t)~
1=0
00
= -n7(n":-_-;-k') ( 1 + 0 ( ~) ); In n-~+1 = ~ i
CJ
( k-1 \0'.
n I '
0'=1
k-1
1
'\l (n-i)-"= ((n-k+1)-v+ 1-n-"+ 1)+0((n-k)-") ·
LJ v-i
i=O
for v > 2. Therefore
00
(n-s)j(n)-ex
k k p
{-s"' 1
~ cr
(k-i)a-
n
s2k
2n (n-k)
0'=1
00
_ ~ s" ((n-k)-v+1_n-v+ 1)}
v(v-1)
v=O
sk s (k-1)2 - ..,,_...;s....:.2k:....·~
= e:tp {- n
-
2n2 2n (n-k)
cr n vn (n-k)v-i
0'=3 V=3 .
=exp {- ~- sk2~tk (1 + 0{: ) )

s"(i+o(i))}
vn(n-k)"- 1 '
396 GOL UTJO'lS .ANSWERS AND l{I~TS
(2) It follo~s from S 5 t1 {i) {3) The maJorant IQllow:~ from the
A 1 - 1 t.:
tad <• 8 S H (t)) that II {t - n'__ ~ )~ {t~ ~ }"<e-n In
i-Q
order to obta1n the mtnorant \\e shall use the telatlOU
(see 8 5 1f (f))
clO
=exp {un (t- n~k )}=exp {-k h

v~l
! ( n.:k rl
m 1
8 512 Hint f (k) lS the upper a .ad ~ f (k) the Jo"~er
Jil:--n
tn.
m~gral sum for ~ f{:t) dx

n
PI m
B 5 13 (1) \l'e have (see 8 5 12) ~ Ink~ ~ Jn~dz +lam
t
n
llnzdz=.:dnz- ~dr=zln.zo- :& lienee~ IDk~mlnm-
m '"
m+1 +ln m On the other band (see 8 5 t2) ~ ln k~ ~ ln s
A~l I
dl: =:: m.ln m- m+ i (l)-{6} are s1m~lar to 8 5 13 (t)
B 5 14 (I} Induet1oJl on n. for n == t ''e have p 1 =: P•--
ap~ ==::: l-- a, H-ence 0 l and 0 <. p,. < i 0 < Pn+t <1 (2) It toll(.Y\\S trom lbe re-
lalton Pnu - p., = ~a:pfl< 0 (3) MaJOtant \\e "have Pll 1~- Pit;::....:
~ «p I
CHAPTER EIGHT 397
n 1
'V Ph-1-Ph ~ \' dx _ 1 (p~-fi-1).
t. Hence II=
.Ll apB """' J axB- a(~-1)
h=1 k-1 Pn
Therefore, p~-f3;;;:.t +a W-1) 11 or, which is the same,
Pn~(1+a(~-1)n)1f1-fi, (•)
1\finorant. It follows from (•) and the recurrence relation Pn=

Pn- 1 (1 - ap~: 1> that
PniPn-1 ;;;:. t - a/(1 +a(~ -1) (11 -1)). ( • *)
Further,
1 n
( 1-B 1)- \
1 Ph-1- Ph
a(~-1) Pn - - J apf •
Pn
It should be noted that, accordingto(u), Ph~ Ph- 1 (1-a) for ~
1 ~ k ~ 1 Vii [ and Ph~ Ph-1 (1- (c Yn)) fork> 1 Vli' [, where

c is a constant. Taking into account the relation (Ph-1- Ph)/
apL 1 =1 we obtain
n
2} Ph-1-Ph ~ ~ Ph-1-Ph
apll aphB -1 (1-a)
h=1 k h~]Yn[
+ h Ph-1-Ph
_ apf- 1(1-(c/Yn))
H'n [<h~n
~(a (1 -a))-1} yli' [ +n/{1- (c!lili')) = n + 0 ( yli').
Taking into account (•••), we lind that for ~>1, a(~~ 1 ) X

(p~-B- 1) ~~~ + 0 (Yn). Hence Pn ~(a (~- 1) (n +
O(yn)))l/(l-fl). Therefore, for n-+ oo Pn""' (a (~-1) n)!f(l-li>,
8.5.15. (1) We write the equation in the form x = ln t -
In x (•). Since t _,.. oo, we can assume that t > e, and hence x > 1.
Then it follows from (•) that x < 1n t, i.e. 1 < x < Int. There-
fore\ In x = f? (In In t). Thus x = In t + 0 (In In t) for t -+ oo.
Takmg logar1thms, we obtain In x = In In t + ln (1 + 0 (In In tl
In t)) = In In t + 0 (In In t/ln t). Substituting this into
' (•}, we obtain a new approximation: x = In t - In In t +
0 (In In t/ln t). Taking logarithms once again and substituting
398 SOLUTIO~S .ANSWERS A~D J:liNTS
the result 1nto the r1ght hand s1de of (•) we ohta1n the next appto:r
lmatlon provtd~ng the tequu:ed. a.t(.utacy (see. N G de BtuiJU Asym
ptot1c Methoos 1n An,alys1s North Holland. Amsterdam 1958}
9
(2) SinCe t-+ OQ, z ....... ao as well Therefore, e~ t - ln. la. t and
co
a: > In t + In (i - ln In tit) = Jn t - ~ v 1 (ln ln tit)" = Jn I --
v~l
In In tl t + 0 ((In ]n t/t) 3) Us1ng t hts rela tt on we obta 1n ex =

t - In z < t - In (ln 1 - In In tit + 0 {(In lo t/t)~)) :::::;; t --
ln In' - ln (1 -- ilnlu t/{t ln t} + 0 ({ln t) t (laIn t/t):l}) ~ t -
ln In t + 0 (ln In t/(t In t)J Taktng logarJthmst we obtam t£ <
ln (I -- ln ln I + 0 (In In tl(t Jn t))) = In t + Jn (i -- In In tit +
0 (ln lo. t/{tl In £)}) === ln. t - ln ln t(' + 0 ( (ln In tlt}~ The malor
ant and m1noran t co1ncJde mth the requ1reil accuracy
8 5 16. lllnt. Ftt.s t prove that J (t) === IJ ( t) I or t ~ og Then
t + o (t} + =
"9
the tnttlal eq u.alt t y can be wr1 t ten 1n the forzn e 'ft I)

0 (1) UsJ ng thJ s equal1l.y, prove tba t f (t) = o (1) and t.ransform
the 1ntt1al equal1ty to eif(i) ~ l + 0 {1) P1ually prove that.
/(t) = In tit + 0 { ~)
8 5 17 {1) \Ve expand A {f) into stmple lrset1ons A {t) =
A1-t
+
C1 c.
.A1-I
+ +-:;.c_~. -+B(th
m-t
where/J(t)isapoly
~\t.l lu ~td.C¢t ~ ·d:t~'I:U\\1\f- th~ t.~{\\~wu\. cl. "ll't- mu\t~p)J
-Q(t)
A (t) by A1 - l Then (A1 - t) A (t} = (t~A,)
for (t-lm)
t ~ .A1 the left hand s1de 1s equal to C1 aud the r1ght hand side
(Ax)
t a -p~Q{).t) Th
us,
ct ~ -Q (A. 1 )
p (A 1)
SJm"'' . . . rly
U1ll.
we can calcu
1
Ia te the coeffie1en ts C1 ( l ~ 2, m) The f r act 10D _ can
1 1111
OCII
be expanded Jnto the geott~ettJcal senes ( t - 1r::)- t = ::8 X

n-Ll
m -en
Ct·r We ®tsun A{t)= ~

t-1.
f: ~ {-~~
n--;0
r + B(l} H(nee
lor luge nOn- Ct

an+l
+ . . t:'
).n+l
! ' + . + ell\
~_n+l
- Cl). _,. .,..J
1
t '2. nom
{2.) Lt:\ us tltst -cona1der \be ease w~n P {') has no rOGls athe!
than .A.l and Q (tJ has a power lower than that of P (t) and ll1s not
r-1
a root of Q (') Then p (t) == ('-l 1)'• Q (t) = ~ ClJtt The e1:pan
i rQ
•
CHAPTER EIGHT 39!)
sion in powers of t p-l(t) has the form

00
p-1 {t) = { - 1)T}..1' {1-t/}..1)-r = (-- ~1 r ~ (- ~) 1 /A~.

11
n""O
Hence
00
A (t)=Q (t)'P (t)= ( _ L r .z: (

n=O i=O
Therefore
r-1
a - {-1)r}.. -(r+n) "V q· ( - r ) }.. t
n- 1 ~ • n-i I•
i=O
h.
In the general case, A (t)= A 1 {I) + ~ Qs (t) , where A 1 {I)
s=l (t-}..s)'s
is a polynomial, and Q5 (t) is a polynomial of • degree smaller
than r5 • In this case, the asymptotic value of an is determined
by the coeilicient oi !11 in the expansion of the fraction Q1 (t) '(!-
1. 1)r'. The problem is reduced to the one considered earlier.
8.5.l8. (1) 2X 311 • Use the results of Problem 8.5.17 (1). The
polynomial P 1 (t)=3t 2 --4t+t has the roots }.. 1 = 1/3 and }..~= 1.
We put Q (t) = ; (1 + t) and P (t) = t2 - ; t +; . Then
•
Q(i/3) { 1
an'""'- P' (1/3) 3
)-n-1 = - 4/9
2/3-4/3 3n+l= 2 X 311 ·
(2) an,..., ( - :
3
( ; r+ 1
) • The smaller (in magnitude) root
of the polynomial 6t 2 +5t-6 is 213, Using 8.5.17 (1), we obtain
8
the required result. (3) all ...., ( - 13 ( ; (4) a~ll = 0, r+l) .
a21l+l '""' ( -1) 11 211 +3 , (5) Hint. 6t~ -17 t 3 + 35t 2 - 221 +4 = 6 X
{t- ~)(t- ;)(t-1+il~'3)(t-1-il/3),an"" ~ 1 311 ,
(6) is 2
11
+1. (7) ( -~) (JI3-t)-1H . The roots of the. equation
2
t +2t-2=0 are }.. 1 = l/3-i and }.. 7 = -l'3-1. Representing
A (t) in the form
A (t)= at+b ,
(1-},1)2 T
400 ~or VTlO~ S ANSU ER q A ro.. D HfNTr;;
\\ ~ f1nd h) tlu~ 1nei hod o{ 1ndet tnn1 nate c oe {f~t 1~nt~ l h a\, a= 0 and
b-l us.ng 8 s 17 {l) v.e obtaJn
a 11 ={~l}').l" t (b( -;)+AtG (;:,)) ..-pf3-1} 1 ( --;)
(8.} 6n ..... ( 10 "W1 _,_1 ( ~; } Tlu• ssn !!.U(! ~t m r(l ag m l ude root <1 f
the ~~nom1nator 1s 0 7 and has a mult~p-l~c.ll)' 2 \\e oole thal

t+l 2
A {t) =- (t __ 0 ) ~ -
1
+,~
2 1 Uc;1ng 8 5 t 7 (2} we obtatn a1\ ,.,
(i 0
- r+~ ~:) (
011::1
8 5 19 (i) an,.., ( -3 (--.. 2)"A) Let A (t} == L} 6nt'"' },lult1ply1ng

n--o
both .s1des ol the relatJon by tn¥ll and tak •ng the Bum \\ e obtain
A(t) --a 1 t a0 +3t (A (t)-a~)+2t~ A (t)~O Sutce a 0 ~1 and
+ +
a1 == 2 \\ e get A (t} == (f- t}/(2t1 3J I J The roots of tbe de
nom1nator are "- 1 ::::; -!/2 and l2-==-- ~ Ussng 8 S t 7 (l) 'h-e fi.nd
that. an ,., [--- 3 (-- 2)n) {2) 1/2 As lD 8 5 19 (1) "r can wnte
A (t)~ {2 (1- t)) 1 +(2 (1-- (q- p) t)) 1 S1nce q p ~ 1 and p +
q > 0 Jq-- pJ < f the roots of the denomlnator.s are A1 ~ 1 and
"-at.:::: q--- p t.l\. t < tk~ l Ustng 8 5 17 (t) (or drrectly) we ohtaln
1 Jn-t l { .2nn 47th )
a!\ ....,
2
(3) Q:fl )FJ = ~1n - 3- · ~ sm 3 (4J an ,.... 371
(5) a 11 ri"W n!Z2n 5
8 5 .20 (I) Iltn l I r the llmlt of an. etlsts and IS equal to a
+
v. e have a== {ll bfa)/2 or IJ l/b from the rncun~te rrlat1on
sJnce a > 0 lf QQ ~ Jl b then at= (a.+b aa) 2 ;::- { l' b+ y b)/2::::
Vi 1t can he ~as~l y obtalned b) 1nd uc.ll()n that o." - V b L-et us
con~ader the ("a~ '\hen tto, > b (the case a(J 1 b tor all n ~ 0 If an> lfh
for a certa.En n ;;;:: 0 then an+ I on -- { {a 11 +bJQnJ t2) a~ ~ (b-- a!)
(2an) < 0 Thus an decreases WIth lntreasing n an .. \ -- V 6--
t
,. J.' (an. +h!on)--) b ==(an-- l/ b)%/(2an} > 0
Consequently an. decreases and has a lower bound Th1s me,ns that
there ex.1sts a IJmJt a == 1tm an As \\as sho\\n abnve thJs hmJt l'l
.,._ n-.~
equal to V b (2) As ln B 5 20 (1) 1f the JJmit an exzsts Jt 1c;

t - 3' -
equal to l b Let us consLder the case when 4o< l b As 1n 8 5 20
( i) we prove l hat tl 11 mon oton~ea 11 y 1ntre-ases anti a11 j b I t)r all <
n. H~ne~ 1l {olloVrs that t.h~re ex1~ts a lun1l fJf an. 11 l1m Gn. ~ a
1\ ....... <Je
CHAPTER EIGHT 401
then proceeding to the limit in the recurre nce relation , we obtain

a= Vb. (3) /1 + b - 1. Note tha~ a 1 - a 0• = (b- a5)/2-
a0 = (-b - b·)/2 < 0 and a 1 = (b - b~)/2 > 0, I.e. 0 < a 1 < a 0 •
Further, a 2 - a1 = (b - ai)/2 (b + +
b2 )/2 = (ai - b2 )/2 > 0,
a 2 - a 0 = (b - ai)/2 - !!!_2 -;;- -ar~2 < 0. He~ce a~ < a 2 < a 0 •
In general, an+ 2 - an - (a,. - a 11 +t)/2. By mductw n, we find
that {a 2 n} increases, a2 n < a 0 for all n.> 1, and. {a_2n+tl dec~eases,
a2n+l > a1 • Conseq uently, there ex1st the hm1ts c = hm a211
n-oo
and d = lim a 2 n+l· Proceed ing to the limit in the recurre nce re~
n-oo
lation, we obtain c = (b - d2 ) and d = (b - c2 ). Hence (c - d) X
(2 - c - d) = 0; since c < b/2 < 1/2 and d < b/2 < 1/2, we
have 2- c- d > 0; c = d. Therefo re c = V 1 + b - 1.
8.5.21. (1) Using relation s (14) and (15), we find that a 3 =
9/128 and a4 ~ 11/128. For n ;;;. 3, inequa lity (15) can be written
m the form an+ 2 ~ 17/1024 +an (323/1024 + an)· Hence 1t
follows that if an ~ 1/8, then a11 + 2 ::::;;; 1/8 for n ;;;. 3. (2) Using
the fact that an ~ 1/8, we derive from (15) a new inequa lity
)
1
an+2 ::::;;; 21l+3+n+1
471+1 8
1 ( ( 3 )n+2
+
4
n+2
211+2 +4an +
' 3 )71+2 1
::;; l 4 +2 an·
Hence, by inductio n, an::::;;; 9 ( ~ r.

(3) Using 8.5.21 (2), we
obtain from (15) a11 +2 ~ 1/271 +3 + 66an (3/4) 71 H. Using this ine-
quality, we get an= 211-1 (1 + 0 ( (3/4) 11)).
8.5.22. Inducti on on n. For n = 1, we have a1 = a1 X 1. If a 11 ::::;;;
a1n, then a 11 +1 ~an+a 1 ~a 1 (n+1).
. .
5 23. (1) Consider the relatwn a11 =
8.•
I (n k+1) n-k
I (n, k) = k+ 1 2 .
-2"
If k<Llog 2logzn }, then ah>i, and if k>[lo g2 log 2 n}, then
ak < 1. Conseq uently, the maxim um value of I (n, k) is attaine d
either for k [log 2 log 2 n}, or for k_ = [log 2 log2 n} + 1. (2) The
=
same result as in 8.5.23 (1).
8.5.24, Note that I. (n, r, k) = 1 (n, r + 1, k)/1 (n, r, k) =
2
_(k - r) r -I /(r+ 1) and that I. (n, r, k) > 1 for k;;;. r > 0. Con-
2
sequently, I (n, r, k) increas es in r. Hence max f (n, r, k) =
/(n 1 k 1 k) =(II)
k ?-h+2h
_ . f (n, r, k)-
IDin - 2 (n)r ,
' h=O k
8.5.25. (1)
11 1
2 + /n if [log2 n} > log 2 (n-log 2 n),
(2n/n}( 1+2 2 [1og. 71]-log. (7l-!og, 71 )) if log 2 (n-log 2 n)::::;;;
g (n),...,
[log2 n} ~ log 2 (n-log 2 n-log2 log 2 n),
2n-[logz7l] if [log 2 n] <log 11 (n-log 2 n-logz lOgz n).
26-0636
SOLUTIONS ANSWERS Af.,D UINTS
The funclLCn f (n k) as a functton of a real atgllment k ~t;

con\ ex down and the m1n mum as attaJned al k k.,-==-
log2 ( n~ Jog 2 11 +O ( ~ 10 2
ti } ) \\le put. k 11 = l10K.11 nl Obvwu'"l}
e1tb~r g (n) ~ f (n k 0 ) or g (n) ~ f (n k 0 ~I)
Jn order to find g (ri)
we must choose the mtntmum value between I (n k~) and
1(n kG --1) Let (a) log:z n ~ k~ > log 2 (n --log~ n) Then
I (n ko ---1) ~ 2n k.- +i + 22h• 1. ,...., 2n-tljn
J(n kt~ :.::::-- 2J1 +aziti: ~ 2" Jog. + -2Jor. Jor. n} ~ 21l+1fn
kl(l I) 11 (71
~\ (b) log 1 (n,-lerg n> > kf >log~ (n -log 1 n-1og:t log 2 :n} 'rh~n
2
J (n k,) ...., 2.11 +2~~'1 = zn (1 +22.~~-logt (ll lr.l!::s r.t)) f (11 ko 1)-
n. n '
.2n+1
= n.. Consequently g (n) ,.., I (n /t 0 ) Let {e) kd' ~ l()g~ (n -
lag, n -- l<lg 1 log 2 n} Then j (n k 0} ~ 211 ho + 22k• ~ 2n "~ flU
n 2 .8'"' '~ ~
11
2
n
10
f (n tr,~ t) # 2~"~ l<a+t > f(n k~} Hence g (1'1)=
I {n ko) == 2» ~, (2) g (n) ..... 2 73 {2e~(n) +2 where a (n) ~

rJ(n))
log 1 n - [logJ nJ 1! a (n) < f/2 and 1 (n) ,.., 2" (2t-a:{n) +
.20.('1\) 1) 1f a; (n) > t/2
Bibliography
1. C. E. Shannon and J. McCarthy, Automata Studies, Princeton

Univ. Press, Princeton, N.Y., 1956. (VI)
2. z. V. Alferova, Theory oj Algontllms (in Russian), Statistika,
Moscow, 1973. (VII) .
3. l\I. Arbib, Braws, ,lfachines and ,lfathematics, i\IcGraw-Htll,
New York, :1965. (VI, VII)
4. C. Berge, Theorie des graphes et ses applications, Paris, 1958.
(IV}
5. N. Ya. Vilenkin, Combinatorial Analysis (in Russian), Nauka,
~Ioscow, 1969. (VIII)
6. Discrete .lfathematics and .lfathematical Problems in Cyber-
netics, VoL 1 (in Russian), Nauka, Moscow, 1974. (I-V)
7. J. Kemeney, J. L. Snell, and L G. Thompson, Introduction to
Finite Jf athematics, Prentice Hall, Englewood-Cliffs, 1957.
(IV, VIII)
8. N. E. Kobrinskii and B. A. Trakhtenbrot, Introduction to
Theory of Finite Automata (in Russian), Fizmatgiz, Moscow,
1962. (VI)
9. K. A. Rybnikov (ed.), Combinatorial Analysis: Problems and
E:rerci~es (in Russian), Nauka, l\Ioscow, 1982. (IV, VIII)
10. N. Christofides, Graph Theory: An Algorithmic Approach,
London, 1975. (IV)
11. A. I. Mal'tsev, Algorithms and RecursiL•e Functions (in Rus-
sian), Nauka, Moscow, 1986. (VII)
12. M. ~Iin~ky, Computation: Finite and Infinite .lfachines, Pren-
tice Hall, Englewood-Cliffs, 1967. (VI, VII)
i3. 0. Ore, Theory of Graphs, American Math. Soc., Providence,
1962. (IV)
14. Y. Peterson and E. Weldon, Error Correcting Codes, :\I. I. T.
Press, Cambridge, ~lass., 1972. (V)
15. J. Riordan, An Introduction to Combinatorial Analysis, J. Wi-
ley, New York, !958. (IV, VIII)
16. K. A. Rybnikov, Introduction to Combinatorial Analysis (in
_ Russian), l\Ioscow University Press, Moscow, 1985. (IV, VIII)
it. V. N. Sachkov, Introduction to Combinatorial ,lfethods in
Discrete Mathematics (in Russian), Nauka, Moscow, 1982.
(IV, VIII)
1
The Roman figures at the end of a reference i•
chapters of this book to which this entry applies.
26*
BtBLIOG.RA PHY
t8 M Swamy aPd K Thulaszraman, Graplu N~twork$ and A Igor

,tfun.~, I W1l-e.y. New Yatk.r t 981 (tV}
t 9 B A Trakh tenbro t A l gor' th~rn and Comp uta tt on a.l A uto T1i at a
(lD Russ1an) Sov Radlo, l'tloscow, t974. (VI Vll}
20 B A Trak.hteubra\ and Ya M 8ar-tdln P,ntt~ A~tar114.ta.
(JD R uss1an), Nauka .Afoseow, t 970 (VI)
21 R J )VJlBon, 1 nttoducluin ta Graph Thtorv Oh, e.r and Bo) d t
Edtnburgh~ t972 (IV)
22 \V Fellerr Jntrroduchon to Probabtltty Theorv 4nd ''' .Applt
cat,onl, Vol 1 2, J Wtle) ~ New Yorkt l968 197! (Vlll)
23 F H ar.a ry.., Gro. ph. Theory, Add1sou \Ves tef...
Londan t 969 (l V)
24 ~~ Hall, Com.b,nalortal Theor}l. Bla1sdel Pub Co \Valtb.am
l!aS! , 1967 tlV ~ Vlll)
25 S V Yahlonsk y,. F un.t: ~~ on.al Co rut rue l tO 1t1 'n k Val u td Log l..C.
1n Proeeed1ngs of V A Steklov Inst of Maths VoJ 51
(1n RussJan)~ l.zd Akad Nauk SSSR, ~1oscow, 11J58 (I lV)
26 S V Yablonsky~ Jntraduct,an td Ducrtte .JJatkemat,cs ,..l1r
Moscow~ 1989 (I VJll)
21 S V Yab1Dnsky, G P Gavr1lovt and V B Kudryavtsevt
Boolean Fun.cl1oM tJnd Post 1 Claut• (tn Rus~uan} Nauka
Moscow, 1966 (I,. Il)
Notations
A •, B* 164, 174 E (t) 267

A (t}, B (t} 267, 268 Ell. 82
A (D), B (D) 120
A 3 , All) 174
am 216
F(m) (x(n)(t), q(r) (t - 1)) 189
,...., ,...,
/ (xn) 22
/* (xlt ,..., x 2 , ••• , Xn) 59
fc (xn) 155
,....,
sn(a) 11
k
G 101
c 155 G* 111
c• 161 G<Jl
G$ H
165
103
c (H) 161 GX H 103
C (n, r) 248
A G nH 128
c (n, r), c~ 248, 249 G (n, A.) 283
csk 92, 94 G(r) (x(n) (t), q(r) (t - i)) 189
g (n, d) 161
:?n, :?n. m 137
D, G, F, L, S, R 213
D*, D' 119
DA, DA, B 177, 178 H6 276
D (G) ,..., 102
Hk, n 163
DA, B (;r3) 178 hii (x) 95
Dp (n) 141 h 2 (x, y, z) 57
D (xo, X1 1 • • • , xn) 242
d (C) 155
d0 (G) 104 ,....,
d (v) 101 /1t,(x11 ) 235
d+ (v), d- (v) 118 id (v) 118
.2) 34 ir en.
ir (n) 46
J 1 (x) 82
h (x) 82
•
406 NOTA. TIOl\ S
K 42 Pc {l ta tn} 277
Kn Knl n~ 103 104 P,.. 4tl
Kc Kpr rKkr 235 pe (X'l) 51
K1 $ " $ a;. K~ 35 p 82
Kpr r Klr r 239 p•l 92 94
Kp,r r 235 .241 P (n r}....- P (n r-) 24S
Kts~ t65
Jt p (G) p (n) 14t
34
L 63 Q(t) 227
L U) 48 -146 Q 212
L,., U) 145 Q• (p C} t5?
L~ (1) L" {J) 145 rv
l U) 48 Q (' l"&) t 16
lC (j) 44 tJt 212
q{r) (t} t89
~ (P\ ilZ.
z• (m) 173 tJ 0 {r) 189
.Zc (P) f66 qT (n) 243
If-f 60 R (G) 128

Mn 55 79 R<o 229
M (X") 55 R (g h) 234
M• 59 tr (K) rr (PJ 2~3
Mf 2M
,._
Mxn. (I (%11 }) 234
M (C) 160 s 59 193
S~t 92 94
m (n d) m (n k d} 155 t56
51 (a.) tt
rn (.:c !/ z) 57
Nx 43 S (f m t 71
g(~) 233
N1 23
Nm Nm .262
Na .I
a
A
259
a\ t9t
o. 03 191 i92
o, 193
o. 205
0 (v) 0 {v) 107
0 (~ (z)) o ('$ (.:l:)) 280
a 2oa
P~ (X"') 22
.......
P (z11 ) 35 38
NOTATION S 407
v (G)
,.., 276
l' (H) 161 v (a 11 } 10
vh (x, y) 83 v (xw} 179
nh. n-h (/) 23
,.., ,..,
--22
??
p (a, ~)
p (u, v)
10
118
"' Po (u, v) 12~
x8 174
x 24, 82
~ 148
x11 , x 0 , x 1 33
x· y (mod k) 83 ~"' 189
x + y (mod k) 83 aa 12
X (t} 174
""
X(~, y~'
,.., 174
,x(ll)(t) 189
g(m) (t) 189
"" ,.., ""

a, all
~.
,.., 10
"'
an ~ ~n 11
""
,..,
all< ~~~ 11
"" ""
a ED ~ 11
""
aU~
"" 12
"" ""
an~
-"" 12
. . .' 175
a 12
""
af 23
ao (G), al (G), a 00 (G) 1i2 X (G), X' (G) 112
~o (G), ~~ (G) 112 Wr (K), Wr (P} 243

IJ(• 60
r (a, b) 124 IJ!
~'P(X)
176
rf,,, r~, 125 25
r~,(a, b), r~,(a, b) 125 IJ! [<D) 25
r"' 187 'Jl (/ (xll}) 76
A 164, 214 0 (x)

,.., ,.., 235
A (w) 164, 174 0, 1 12
NOTATIONS
,.,.,
u ann 10
ol rest (.r1 .r 2 ) 236
V K 34
i-1 l .sgn z sgn z 226 22.7
{Pim la]ff\ 2lo
&' n1 34
~ l }q
8 q:J 217
_EKt 41
k k
1 t ~
24 280
[MI 55 -.......... 2t6
(.il!l(') 201 '-'
r"'- 280
,w.:z: S2 < ~ 1t
{Vt V 2 X) 10~ EB tf 24
u i2
(a)n (.a)n h 271 n 12
{a} 179 1 & v ~ -+ I 24
n
(;) (,1 r, rtot
) .l49 _..
d a
a
4..\
{11-)r 249 dJ..l a \z.h ~ :~ %
ik
)
1l, 249
[7] 226
+
...,.
X 83
226
-83
UlLn max 83
[--::] 236 (== (- f-- 215
Subject Index
'
Absolute difference, 235 class,

alphabet of variables, 24 equivalen t, 176
arc(s), 101, 117 functional ly closed, 2v7
concordan t, 121 precomple te, 207
multiple, 117 closure, 55, 207
parallel, 117 code, 155
algorithm, opet·ator notation of, almost uniform, 171
217 alphabet, 165
alphabet, graph of, 165
external, 212 basic machine, .218
input, 176 block, 155
internal, 212 binary chaz·acleri stic
output, 176 function of, 155
automaton (s), 174ff compact, 155
without input, 192 cost of, 166
without output, 191 dual, 161
equal-weig ht, 15fi
equidistan t, 156
group, 160
Ball, 11 Ilamming, 160
basis, 55, 1ti0 lattice, 232
blank square, 212 linear, 160
Boolean difference, 41 l-multiple , 230
branching, 205 maximum , 155
optimal extension of, 167
prefix, 165
optimal, 166
Chain, 11, 102, 118 uniform, 155
ascending, 11 weight of, 156
Hamiltoni an, 104, 110, 118 of word, 155
length of, 11 code distance, 155
simple, 118 code word(s), 155
channel, coding, 155
input, 190 alphabet, l64ff
output, 190 letter, 165
character(s
• •
), 33 coefficient (s), binomial, 249
ClrCillt, 118 combinati on, 248
class, with repetition , 248
closed, 55ff, 89 without repetition , 248
of linear functions, 85 command( s), 213
precomple te, 55 completen ess, 55ff
410 SL BJECT l'D~X
co m pl ex ity diagram
sto ra ge 2~3 ~foore 187 200 202!
tJme 243 d1 gr ap h 111
c:ompos1t1on 235 ad Ja ce nc y m at rix (ff 120
co nd en sa tJo n 119
t'O U) UD C\l .OD 24 33
Cl fC Ul lle s.s 121
in et de nc e matr1x of 120
monotontc 35 s1nk of ti 9
ec.nn-ec\t-vlty 1ndex 29 source of 119
constant (s} 82 str on g co m po ne nt
c-ont~-c.t(s)
of f t9
14 ' tra ns 1 t1 ve t 21
br ea k 144 tri-v1al t t 9
m al e i44. un1la tera1 component of { t 9
to nt ro llt ng davtce 212 dtsjunct1on 24 33
co re 43 d1stance Ha xn m tn g 10
cr lte tlo n dnf
CQ rtlp letene~s 1& co nt ra ct ed 4'2
Sa lo m aa s 94 m1n1mal 43
Sl up cc ky s 94 91 sh or te st 43
Yablonsky s 94 te rm in al 43
cu be
edge of 11
fa ce of 12 Edg~{~) \n~tdeut
dJ m en sio n of 12 1.01
element of un tt delay 197
d1 rrr tJo n of t2 e.quat.10n ta -n on lta l iS S
1n co m pa ra bl e to 1n sc al ar form 189
ra nk of 12 tn ~et.tor fo tm {B 9
1nterval of 20 eq ul va len ce (s) 24
dtmen~non of 20 ba ste 3t
k th st ra tu m of fO er ro r 156
n dun~ns1onal 10 e.xtlWitve su.m 2.4
UD il t0 4
vertex{1ces) of t 0
cu t 127
m1n•mal 127 Fa ee (s)
te rm in al t27 boundary of t t 1
CJ. c1e(~) 1t t 02 ext~rnal t t 1
edge of 21 Internal t t t
Ham1Jtonlan 104 f09I fo re st -103
le ng th of 1t 275 forcn(s)
l1nearly tn de pe nd en t 1t 1 con)unctl ve normal (c: n f )
:s1mple- 102 1l 8 34
CJ: cl1c Jndex 276 co m pl ex ity of
le ng th of 34
di SJ un ct l\e normal (d n f)
34
DecttdJng 156 complexJ.ty of 34
au th -e nt lt. \tY of \5 6 le ng th of 3~
de la )e d dependence 190 first 86
de l.e tlo n 191. se co nd 86
De ~forgan s law (s) (ru les ) 3 t formula(s}
83 co m pl ex ity ol 28
d1agram de pt h of 30
of ca no nu :a l decom}X)~l t.1an dua1 60
t2 5 fq U l\a lc nt 26
SUBJEC T INDEX 411
formula(s), function(s},
Euler, 111, 114 symme tric, 28
exclusi on, 259ff transit ion, 187, 213
inclusio n, 259ff univers al, 241
Stirling 's, 281 functio nal symbol s, 25
function(s}
Ackerm ann's, 239
asympt otically equal, 280 Godel numbe ring, 242
Boolean, !Off, 22f, 48, 51 143 graph(s ), 101ft
comple xity of, 145 ' balance d, 132
minimi zation of, 42ff biparti te, 103
monoto nic, 70, 150 comple te, 104
non-mo notonic , 70 Cartesi an produc t, 103
bounde dly determ inate (b.d.) centre of, 128
176 '
chroma tic numbe r of, 112
commu tatively equiva lent, edge, 112
37 comple ment of, 103
compon ent of, 34 comple te, 103
comput able, 218, 233ff comple tely disconn ected, 103
congru ent, 55 connec ted, 102
cyclic, 49
determ inate, (d-), 174 diamet er of, 102
bounde dly, 174f correct , 112
of displac ement 213 coverin g set of vertice s
disting uishabl e ' 176 (edges), 112
dual, 59 ' cyclom atic numbe r of 111
differe nt, 137 '
elemen tary, 23
equival ent, 17!i directe d, 117ff
essenti al, 94 empty, 103
expone ntial, 267 interse ction of, 128
~enerating, 265ff
isomor phic, 102
mdistin guishab le, 176 k-facto r of, 104
k-valued, 207 labelle d, '137
linear, 63, 89f matchi ng, 104
majorit y, 57 (n, m)-, 137
monotonic, 76 non-or iented, 101
generalized 74 null, 103
nonline ar, 54' numbe red, 137
nullary , 235 oriente d, 101, 117
output, 187, 213 planar, HO
partial, 213 topolog ical, 111
pa~tial numeri cal, 218, 235 plane, 111
pnme, 77 graph(s ),
recursiv e, 233ff radius of, 128
general , 235 regular , 104
partial, 235 self-co mplem entary 108
primiti ve, 235 241f thickne ss of, 111 '
selectio n, 235 ' triangu lation, 114
self-du al, 55 union of, 103
S_heffer's, 76, 78f group, 160
stmples t, 235 dimens ionalit y of, 160
specified, 226
successor, 235
superpo sition of, 233 Homeo morphi sm, 103 0
4f2 S U B J E C T 11\DEX
l d e n l fir..atton 19 t
l m p l t c a n t 42 m o d k d 1.ffere n e e 8
3
p r me 42 m o d k p r o d u c t 83
J m p l l c a t t o D 24 83 mod k s u m 8 3
t n e q u a l 1 t y Cheb}"sh m u l t t g r a p h 101
ev
I n t e r v a l o f func:tton s 282 d1rected t f 7
43 m u lt 1 s e t( s ) 248
m a x t m a l 43
t n t r o d u c t t o n o f fee d t flerent. 248
1somorphtsm 103 dback 1 9 2 d t s o r d e r e d 248
J l e r a t t o n 235 o r d e r e d 248
J o t n t den1al 8 3
N e g a t t o n 24
L u k a s t e \ \ I C.z. s 82
k. v a l u e d l o g l c 82f Post. s 8 2
t net work (.s) tOt.fl
f u n c t t o n ( s ) of 82
elementary 82 boundary of iz.i
canon1c.al decompos1
t25 1 1on o f
H t26
L a t t t c e 229 p 125
lemma I 126
Berns1de s 216 cba1n o f ! 2 7
on nonltnrartty of c o n n e c t e d 125
63
ru nctJon strongly !25
non net,...ork(s)
o n n o n monotonic (.On t a c t ! t b e m e o f
70
funct1on d e c o m p o s a b l e 125 144
on non-self dual fun I I 12.5
l e t l e r ( s ) 17 4 c t i o n 59 p 125
loop(s) t01 ' 125
e x t e r n a l 125
H f25
1nternal 125
r .l a p K a r n a u g h 4 6 k pole t 2 3
map~Ing t 6 graph of t23
match1ng 1somorph1c t 2 3
m a x 1 m u m to.. l e n g t h o f t2 7
p e r f e c t 104 .1t 126
matnx pol@ o f 144
ehecktng !61 series p a r a II e1 i 25
code-C i 6 0 superpos1t1on o f 12
v e r t e x of 5
generatsng ! 6 0
m e d J a n 57 e q u 1 v a l e n t 127
method I n t e r n a l 127
B l a k e s 43f m1n1mal 127
cascade !52 \\Jdth of 122
F a n o s 170 number
o f J n d e t e r m t n a t e coe edge c o v e r 1 n g t t 9
38 90 ffic1en ts of v e c t o r s iO
m l n l m t z .t n g c h a r t
46 vertex eovertng ! 1
v e r t e x tndepend.Pnce 2
tt2
•
SUBJECT INDEX 413
Opera tor(s), summation , 249

autonomou s, 204 Scanning head, 212
boundedly determina te (b.d.), scheme, 191
176 connected, 144
constant, 204 contact,
determinat e (d-), 17 4 complexit y of, 144
generated, 185 minimal, 144
without input, 204 simple, 14H
dual, 149
equivalent , 144
of functional elements, 145
Path, 118 isomorphic , 144
Hamiltoni an, 11R (k, n)-, 144
Peirce's error. 24 outputs of, 144
permutatio n, 248, 275 primitive recursive, 233
cyclic, 27 5 segment, imtial, 16
with repetition, 248 sententml connective s, 24f
pole(s), input, 191 sequence(s ),
polynomia l, recurrent, 2fi5
mod-k, 89 characteri stic polynomia l
mod 2, 35
•
or, 2fi5
degree of, 35 senes,
length of, 35 figure counting, 277
Zhegalkin' s, 35, 40, 42, 54, function counting, 277
64, 72 set, 15
prefix, 164, 174 base, 15
pre-period, 184 centre of, 17
principle, duality, 59 complete, 15
program, 213 de lining, 131i
pseudograp h(s), 101 dual, 59 •
automorph ism of, 118 Post's, 94
complete, 120 Hosser-Tu rquette, 94, 96
connected, 119. self-dual, 59
strongly, 119 Sheffer's stroke, 24
unilaterall y, 119 shoot, 124
weakly, 111 state,
directed, 117 final, 212
disconnect ed, 119 initial, 212
dual, 111 terminal, 215
group of, 118 stratum, 16f
Hamiltoni an, 118 subdivisio n, 108
isomorphic , 118 complete, 109
self-dual, 111 subfunctio n, 34
strong, 119 proper, 34
unilateral, 119 subgraph, 102, 118
weak, 119 generated, 118
proper. 102
SJlaitning, 102, 118
weight of, 132
subnetwor k, 125
Rank, 33 subparti tion, 103
root, 123 subset, boundary of, 17
rule, snfftx, 164, 174
multiplica tion, 249 superposit ion, 27, 193f
sy m bo l 212 Var1able(s)
du m m y 2! 2 ap pa re nt 49f
.system es se nt ia l 49 ( t 15.
co m pl et e 55 76 fic ttt 1o us t 75
of h.eredt t ar y fu ne t 10 ns 77 ve ct or (s )
1rr ed uc i ble 55 t 22. b~nary 1-o
n un l versal 2t
Bo ol ea n 10 104
Ta bl e canon1tal t8 8
ot toeititle-nt5 ol po ly no m ia ls
38
ta pe 2f 2 vector(.s}
th eo re m l1 ne ar eo m b1 na tto n of 160
l\farkov 165 tr i v1-al 160
P1ccard 95 l tnearly de pe nd en t 160
Po ly a s 271 norm of to
Pontryag1n Kura to\\ sk1 111 \\e ig ht of 10
th eo ry vertex(Jces}
algor1thm tOt ff
212ft access1hle t t9
eod1ng t5 5f f ad Ja ce nt tOt
gr ap h 101ft <:.ut 1.04
of gr ap hs an d ne t,1 or ks 13 de gr ee of 101
Polya s 275ff 1nC1dent t t 7
to ur na m en t 1-20 "n de gr ee of 1.1S
transpos1t1on 275 tn ltJ al t t 7
tre e 103 Is ol at ed 102
w de of 24 i th. rank
gr ow in g 1 t9 t5 2
ou t de gr ee of 117
1nfin1 te f 29 pe nd an t 10
1n fti rm at 1o n 1.-'1 se pa ra ttn g 104
loaded 178 term1na 1 t t 7
m ln lm ur n sp an nt ng i 32 zero ra nk t 77
ro ot ed t2 3
pl an e 123
trunc:a te d dt ff erenc.e 83
tu pl e( s) 10 47 \V al k 102
ad Ja Ce nt 10 ch ai n 118
co m pa ra bl e tt d\ te t\e d 11.S
tn co m pa ra bl e t i le ng th of 102 17 8
O{ lpO Sl te f 0
TurLng s eo m pu ta t'i on 21 5 '" e1ght t 76
word{s) \1 4
Tu rJ ng s m ac h1 ne (s ) 212ff em pt y t6 4 114
229 235 eq ui va le nt t6 9
com pt t!tt 1on ot 2;1& Infintte t 74
co nf ig ur at io n of 214 ln pu t 176
1nLt1al 2t 5 le ng th of 164 176
te rm in al 215 ou tp ut t76
equ1 va le nt 21 6 quas1 pertodu~ 184
Iter.atton of 217 su bw or d a{ 164
zone of op er at iO n of 216
Zero (0 ) fu nc t1 on 24
Unton of funct.Lons 193 zone com p uta t 10 na 1 proee1s
un ity (t ) fu.net\Otl 24 243
TO THE READER
.Mir Publishers would be arateful for your comments on

the content, translatiOn and design of this book. We would
also be pleased to receive any other suggestions you may
wish to make.
Our address is:
Mir Publishers
2 Pervy Rizhsky Pereulok
I-110, GSP, ~[oscow, 129820
USSR
Also from M1r Publ1shers
The Theory of Cho,ce and DeCISJOn·Makrng

l Makarov~ T VJnogradskaya,
A Rubch1nsky~ and V Sokolov
The volume covers a h road spec lr um of mat hem at 1c:.a I theo r 1-e s
pert\ n~i\ t \ o \ h~ pr~n t -day u nd~r ~ ta nd1ng of thn~ t\! and d«.ts um
mak111g 1t off-ers techn1qu~s and theory and. rs full of numerous
apphca tJons of dec1s1 on rna kJng methodolOg'Je5 and thus 1s l n len
d.ed to serve both as a guEde for a wJde range or analysts and
experts closely tnvolved tn the mathematu:al madelltng of dec.tston
proc.~s~~ and a textbook for ~tudents

G.P. Gavrilov, A.A Sapozhenko: Selected Problems in

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

G.P. Gavrilov, A.A Sapozhenko: Selected Problems in

Uploaded by

Copyright:

Available Formats

G.P. Gavrilov, A.

MIR PUBLISHERS MOSCOW

Mir Publishers Moscow

FJrse pub1l5bed J 989

ISISN 5 ·03·000522-6 ~ r Jtallaast peAa11U.V.R tJluan«o- va te¥&Tn.qe~KOi.

Chapter 1. Boolean Functions: Methods of Defining and

Chapter 2. Closed Classes and Completeness 55

Chapter 3. k· Valued Logics 82

Chapter 4. Graphs and Networks 101

Chapter 5 Fundamentals of Codlng Theoey t5S

Chapter 6 Fi.o I te .Automatons 174

Chapter 7 Fundamr;ntRls of the Algorithm Theory 212

Chapler 8 Elements ol Combioat<-riaJ Analy5m 248

This collection of problems is intended as an accompani-

ehapter 1llustrate the d1ilerence betv.een k valued log1cs

problems. Solutions are given in a concise form in the

Boo\ean Functions: Meihods

t.t. Boolean Vectors

values from tJJe set ( 041 1} IS caJ led a blnary or Boo

a• or t:L The number n 15 ~ailed the length aj the

The vectors a.n are called the verttces of the cube nn

11" and IS denoted by B~ To each Boolean vector a" .

the number of the vector etn. The tuple et,. IS obviously

ming) dt.stance between the verlltes a and P of the tube

ll p (cr., ~) = 1, and oppostte 1f p (at p) = n An unor

T b.ls set.t~r.an 1s auxlliuy \V-e s.ha\1 h~ Ui>\.UI t\-n.\y pt~'h\~m!.

dered pair of adjace nt vertic es is called an edge of the cube.

The set B~ (a) ..-= {~: ,....,p (a, ~) = k} is called a sphere, ~~

while the set S~ (a) = ,..., {~: p (a, ~) ~ ,...,

of the relatio ns a 11 ~ ~n or ~ll ~ an is satisfied, a" and

if an < ~n and p (an, W') = 1. The precedence relatio n

Figure 1 shows the diagra ms of partia lly ordere d sets

a,, (notatwn: [a 0 , a 11 ]) if p (a 1_ 11 a 1) = 1 (i = 1, k).

a:n $ ~n) obtatned by extl us1-v e sum of ve-ctors a. and

a ~ector wl1ose t th c.oord1nate 1s equal to 1 1f and only 1(

a: 1 = ~~ == 1 By a. we denote a vector (opposzte to a)

th.e ~alue 1 s,_f a.t ~ 0 If a E. {0~ 1 },,., we ,....

1.1.3. F1nd the number of tuples a. E B~ sat1sfy1ng the

tond1t1on 2" '~ v (a) < 2"

{1.) $) {a..1}~ p {o. ~} + p ~, 1},

(2) P (a. l} ~ P (t:£ $ ~' l ID ~},

(3) p (a., ~) =- a + tt 2 Ct n ~ n'

(4) p (~~ ~> = tt e ~ u

les (etn, 11"), such that p (c:tn, ~n) ~ k

1.1.6. Let a and ~ be the vertices of a cube Bn,

(1) p (a, y) p (y, ~) = p (a,~);

(4) p (a, y)~ k, p (~, y)~ r.

(1) p (a, ~) > 2n/3, p (~, y) > 2n/3, p (y, a)

(2) v (a) < v (~ E9 y), v (~) < v (y E9 a), v (y) <

(3) II a II > II ~ EB y II , II ~ II > II y E9 a II, II y II >

(1) a:::;;~ is equivalent to a n~ = 0;

(6) (a y) <P y); ~

(7) a:::;; 'Y leads to the relation au(~ ny) = (au p)

1"-1 ,-v ,...., ,..., ,..., ,..., ,..., ,...,

(8) a~ y is equivalent to aU (~ n y):::;; (aU ~) n'Vi

J t i2 Let 0 ~ l < k -s;;;, n and let A (a) be a set of all

vectors 1n B C()mparable w1th a. F,Dd tl\e pow-er of the

(3) C ~ A (n) a. E B~t

of pairWISe Incomparable sets and {J a II~ k for any aEA

numbers a E A is a divisor of any of the numbers b EA

tuples 'V satisfying the relation a~ 'V ~ ~ is equal to 2h.

tex ~ satisfying the relation a;< ~ < ai+ 2 , ~ =I= Ctt+ 1

that p (a 1 ~) = k, and let C (a, ~) be a set of all verti-