Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Zeitsehr. f . math. Logik und Grundlagen d . Math.

Lid. 33, S. 453-456 (1987)

A NOTE ON GENTZEN’S DECISION PROCEDURE


FOR INTUITIONISTIC PROPOSITIONAL LOGIC

by KOSTADOBENin Belgrade (Yugoslavia)’)

I n his proof of the decidability of intuitionistic propositional logic in [l] (IV, 9 1.2),
GENTZEN proved a lemma which says that in his sequent-systems LK and L J every
r
provable sequent --f 0 in which no formula occurs in r, or in 0,more than 3 times
has a Cut-free proof in which all sequents are likewise restricted to sequents where no
formula occurs more than 3 times on the left, or on the right. It is clear that any pos-
itive integer rL instead of 3 would do for showing decidability, and GENTZEN seems to
have picked up 3 for the ease of the proof.
From GENTZEN’S Lemma for 3 we can very easily obtain an analogous lemma for
any n > 3. What about 1 and 21 Of course, for 1 the lemma is false, as witness the
sequent 9l I> (‘212 B), 3 --f 8,which cannot be proved without the structural rule of
Contraction. For 2 the lemma still holds. However, its proof seems to be slightly more
complicated than GENTzEN’s proof for 3, and this is probably why GENTZEN did not
choose t o prove his lemma for 2 . GENTZEN, and many other logicians who worked on
his sequent-systems, may uell have been conscious of the fact that 2 could replace 3
in GENTZEN’S Lemma. However, we were unable to confirm that indeed they were by
reference to the literature. This is why we decided to present in this note a proof of
GENTZEN’S Lemma for 2 . Establishing this lemma for 2 is useful, because it shortens
GENTZEN’Sdecision procedure for intuitionistic propositional logic.
We assume the reader is familiar with [l], whose notation we shall follow. To state
GENTZEN’S Lemma precisely, n e introduce the following terminology.
For n > 0, a sequent T + 0 is n-reduced iff no formula occurs more than n times
in the sequence of formulae and no formula occurs more than n times in the sequence
of formulae 0. It is clear that every n-reduced sequent is also n + m-reduced.
The sequent F* -+ 0* is an n-reduced version of r --f 0 iffr* -+ O* is n-reduced,
the set of formulae which occur in I‘* is equal to the set of formulae which occur in r,
and the set of formulae which occur in 0* is equal to the set of formulae which occur
in 0. With the help of the structural rules of Thinning, Contraction and Interchange
we can easily show that an n-reduced sequent is interderivable with any of its m-re-
duced versions so that in these derivations all sequents are max (n,m)-reduced.
A proof is n-reduced iff all sequents in this proof are n-reduced.
GENTZEN has demonstrated the following lemma about his sequent-systems LK and
L J , for n = 3 :
G e n t z e n ’ s L e m m a . Every proof of an n-reduced sequent S can be transformed into
an n-reduced proof of S. If the original proof is Cut-free, then the transformed proof is
C ut-free.
GUBRI~:he asked whether
I ) The writing of this note was prompted by a quest,ion of DJORDJE
GENTZEN’S Lemma holds for 2. The question was then raised in conversations with BRANISLAV
B O R I ~ALEKSANDAR
I~, KRONand PETERSCHROEDER-HEISTER.
454 K.DO~EN

Since every provable sequent has a Cut free proof, GENTZEN'SLemma amounts to
the following assertion for n = 3 :
Every provable n-reduced sequent has an n-reduced Cut-free proof.
It is very easy to establish this assertion for n > 3, once we have GENTZEN'SLemma
for n = 3. To obtain an n-reduced Cut-free proof of .a provable n-reduced sequent S ,
where n > 3, we produce first a 3-reduced Cut-free proof of a 3-reduced version S*
of S , and then we derive S from S* via n-reduced sequents with Thinning, Contraction
and Interchange.
Before getting on the proof of GENTZEN'SLemma for n = 2, we shall prove an
auxiliary lemma. For this lemma we introduce the following terminology.
Bn application of the rule for introducing implication on the left:
r+w '8,A-A
(a+)
% 3 23,F,A - + @ , A
is restricted iff % 3 23 does not occur in Ll. A proof will be called restricted iff every
application of ( 3 -+) in i t is restricted.
Now we state our auxiliary lemma about LK and LJ:
A u x i l i a r y Lemma. Every proof of a sequent S can be transformed into a restricted
proof of S. If the original proof is Cut-free, then the transformed proof is Cut-free.
P r o o f . A given proof of S is transformed in the following way. Take a lowest un-
restricted application of (=I-+), i.e. an application of (=+), not restricted, such that
every application of ( 3 -+) below it is restricted. This application looks as follows:
zl. XR
r+@,% B , A + (-
1
a % , r , d + o, A
where 3 2 23 occurs in A , and zLand zRare subproofs terminating with the left and
right premise respectively.
I d A* be the result of replacing every occurrence of % I> 23 in A by %. It is easy
to show that 23, A* -+ A is provable by using zR, 23 4 3 23, Interchange and Cut.
However, we shall transform nR into a proof n$ of 23, A* -+ A which will not involve
Cut. if nRis Cut-free. Moreover, zft will not have more applications of ( 3-+) than nR.
Occurrences of the same formula in the premises and conclusion of an application
of a rule, which correspond to one another according to the rule, are called clustered.
For example, the first formula in the sequence T i n the left premise of ( 3 +) is clustered
r
with the first formula in the sequence in the conclusion of ( 2+) (provided F is not
empty). Likewise, the three occurrences of 3 3 '$3 in the following application of
Contraction :
% 3 23, % 2 23 -+a
a223-+(x
are clustered (see [2], 5 3.41).
Consider now all the occurrences of 'ill13 23 in ?cR obtained by starting from a par-
ticular occurrence of % 3 23 in A in 23, A -+ A , and adding successively all the occur-
rences of % =I '$3 in z,, clustered with those occurrences already obtained. This way
we obtain a tree of occurrences of % 3 23 which we shall call an % 3 23-cluster. A t the
A NOTE ON GENTZEN'S DECISION PROCEDURE EOR INTUITIONISTIC PROPOSITIONAL LOGIC 45 5

top nodes of this % =I %-cluster, 2i =I 8 occurs either as the left-hand side of an


axiom :
(1 1 % 3 % . + % 3 %

or as the leftmost formula in the conclusion of on application of Thinning:


n
(2) r' .+ of
a 8, rt -+ 01

or as the leftmost formula in the conclusion of a n application of ( 3+) :

d 4
(3) r' .+ 01, %, A' -+ A'
8, r',A' -+
a n' 01,

First, we replace our 8 3 %-cluster in nR by an analogous %-cluster. Next, we


replace the figure on the left, which results from (3) after replacement by the 8-cluster,
by the figure on the right:
nt' n;; 72;;

I"' --t O', % 8,A" 4 A' 8,A" 3 A'


%, I"',A" --* Or,A' Thinning and Interchange
%, P I , A" 0') A'
Finally, if a t a top node instead of (1) we have the sequent on the left, we replace i t
by the proof on the right :
%.+%=J% %+8
Thinning
2L%-+fB (+3)
% . + % 3 %

Repeating this procedure successively for every occurrence of % 3 8 in A we obtain


our proof nt of %, A* .+ A . Let now A** be A with all occurrences of 8 3 % deleted.
We replace our lowest unrestricted application of ( ZJ -+) by:
4
B,A* - + A
ZL Interchange and Contraction
r+mx 8,A** -i n
%~!l3f,A**-+O,A
Thinning and Interchange
a3?a,r,~.+~,~
Replacing this way successively every lowest unrestricted application of ( 3-+) by a
restricted application of ( 3 +), we obtain a restricted proof of our original sequent S.
This procedure must terminate because introducing riff did not increase the number
of applications of ( 3-). The transformation procedure did not introduce any new
applications of Cut. 0 8

Now we can prove GENTZEN'SLemma for .n = 2: Suppose we are given a proof of


a 2-reduced sequent S. By the AuxiIiary Lemma, we can transform this proof into a
456 JC. DOSEN

restricted proof of S, which is Cut-free if the original proof is. This restricted proof is
next transformed as follows.
All axioms are left unchanged (these sequents are 2-reduced). I n every application
of a rule. except Contraction and Interchange, if a formula occurs more than once in
r
the sequence T , we delete from in the piremises and conclusion all but one occurrence
of this formula. So, a single occurrence ie left of each formula in r.
We proceed like-
wise with d , 0 and A , the remaining parametric schematic letters for sequences of
formulae in GENTZEN'S rules. Since we have excluded Contraction and Interchange,
and since all applications of ( 34 ) are restricted, we obtain applications of rules in
which all sequents are 2-reduced,
In applications of Contraction
2 , % 1'-
, 0 0 -+r,
3,'3
%,r+@ e-+r,%
r
from n e delete all occurrences of B, and all but one occurrence of each formula dif-
ferent from 3 ;from 0 we delete all but one occurrence of each formula in 0.
I n applications of Interchange
A,%,G,Il-+ 0 0 + 1'. 6,3 ,A
A,g,%,r+ 0 o -+ r,9, CF, A
r
from we delete all occurrences of 5Q and I&, and all but one occurrence of each formula
different from B and 6. We proceed likewise with A . From 0 we delete all but one
occurrence of each formula in 0. This way, for all rules in our proof, we are left with
applications in which all sequents are 2-reduced.
After applications of rules in our proof are transformed in that way, it might hap-
pen t h a t a sequent from the old proof occurs now in,one 2-reduced version S, as a
conclusion of a rule, and in a different 2.reduced version S, as a premise of the rule
immediately below. However, it is easy to derive A, from BC via 2-reduced sequents
with the help of Thinning, Contraction a d Interchange. Such a derivation might he
also needed to derive our original 2-reduced sequent X from a 2-reduced version of S
occurring as the conclusion of the last rule in the transformed proof. This wc
obtain a 2-reduced proof of S. The tran,cformation procedure did not introduce any
new applications of Cut. 0
This proof of GENTZEN'SLemma for n = 2 differs from GENTZEN'Sproof for 11. = 3
in having a preliminary transformation allowed by the Auxiliary Lemma, and alho in
the deltltion procedure of repeated formulae for Contraction and Tnterchange.

References
[11 GICNTZEN,G., Untersuchungen Ciber das logische SchlieWen. Math. Z. 39 (1934-35), 156-210,
405-431 (Engl. transl. in [31).
[2] GENTZEN, G., Neue Fassting des Widerspruchsfreiheitsbea eises fur dic reine Zahlcntheorie.
Forschungen zur Logik und zur Grundlegung der exakten Wissenschaften N. 8.4(1938). 19-44
(Engl. transl in [3]).
[3] S Z l B O . M. E. (ecl.),The Collected Papers of Gerhard Gentzen. North-Holland, Amsterdaiii 1969.

KoRta Doiien (Eingegangen am 18. Jnli 19%)


Matematiski Institut
Knez Milloilova 35
11000 Belgrade, Yugoslavia

You might also like