Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Periodica Mathematica H~tngarica Vol. 11 (4), (1980), pp.

271--279

A BANACH SPACE VERSION OF OKADA'S


THEOREM ON SUMMABILITY OF POWER SERIES

by
W. GAWRONSKI (Ulm), ]3. L. R. SHAWYER (London, Ont.) and
R. TRAUTNER (Ulna)

The main theorems of this paper give extensions of the classical theorem
of BOREL and OKADA on summability of power series b y means of matrix
methods. These results enable us to determine the set of summability of gener-
al concepts of summability as for example absolute summability, and we
obtain sets of equal rates of summability in the case of ordinary summability,
when the power series is summed to analytic continuation (see Section 4).

Suppose throughout that

(1) P ( z ) = .~" a k z k
o
is a power series with radius of convergence I. We denote b y A p a maximal
domain into which (1) possesses a unique analytic extension, and b y
n

~,(~) = ~ ~ z ~
0

its n-th partial sum. Furthermore we consider a matrix method


V = (c,,,~), (n, k = o, 1, 2 . . . . )

and the sequence

(2) ~n(z) = ~ cn,~ ~(z) (n = O, 1, 2 . . . . )


k=0

The research work of the second author has been supported in part by the National
Research Council of Canada.
A M S (MOS) subject classifications (1970). Primary 40D20; Secondary 40C05,
40F05, 30A14.
Key words and phrases. Okada's theorem, Borel--Okada theorem, absolute sum-
mability, analytic continuation, Banach space.
272 GAWRONSKI,
BHAWYER, TRAUTNER: SUMMABILITY OF POWER SERIES

of its transformed partial sums. Then P(z) is called compactly summable (]7)
on a set G if the sums (2) exist for z E G and the sequence of functions {Vn(Z)}
converges compactly in G, i.e., {Tn(Z)} converges uniformly on every compact
subset of G. Moreover let Z v be the largest open set where the geometric
series is compactly summable (V) to its analytic extension 1/(1 -- z). Finally
we denote by o)~p the collection of all domains Ap and we define for A, B c C
A . B = {ab : a E A , b E B}, A | B ~ (A c . Bc) c,
where A c denotes the complement of A (with respect to C).]
Now we can state the theorem of Betel and Okada on summability of
power series given in the version of ([2]; Theorem 1).

THEOREM (BoREL--OKADA). Let


(a) 0 ~ Z v,
and suppose that

lim ~ Gn,k - ~ 1
n ~ k~O

is satisfied. Then the analytic extension of P(z) is the compact g sum of (1)
in the set
(5) G -~ A v |
for every A v C o)~v.
It is the purpose of this paper to extend this theorem of ordinary summa-
bility to more general concepts of summability. The summabflity propergies
of the power series (1) are defined by properties of the sequence (vn(z)}. The
main idea consists in considering {~n(Z)} as an element of certain sequence
spaces B. Thus we get different concepts of summability by various choices
of B; for example, absolute summability is characterized by the space of all
sequences of bounded variation considered with the norm

The main task is to determine for a given matrix method V and for the power
series (1) the set of all z for which the sequence (~n(z)) belongs to B. It will
t u r n out t h a t the "function theoretic" part for the proof of the desired exten-
sion of Okada's theorem is already contained in the proof on Theorem 1 in
[2] so t h a t basically we have to deal with the "functional analytic" part.
GAWRO~SKI, SHAWYER, TRAUTNER: SUMMABLLITY OF POWER SERIES 273

Suppose that B is a Banach space being a subspace of the linear space


of all sequences. Lot Zv,s be the largest open set contained in E -- {1} for
which the geometric series is compactly summable (V) "in the sense B " , that
is for which the sequence
{~ 1 _~zk_+x]
a(z)---~ c.,~ 1 - - z J

belongs to B, and a considered as a vector-valued map a : Z v,B ~ B, is con-


tinuous (with respect to the norm in B) on Zv,n. It follows from the Heine--
Betel theorem that a(z)is uniformly continuous on every c o m p a c t subset
of Zv,s. Further we assume that the series ~ e n , k converges for every
n - = 0 , 1 , 2 , . . . and that k

(6)
k

and

(7) o gv, s
are satisfied. Finally we introduce the power series

(8) v,,(z) = c,,,k Zk (n = 0, 1, . . . ) .


k=o

We need the following auxiliary results.

L ~ 1. Suppose that condition (6) is satisfied. Then Zv, B is the largest


open set contained in C - {1} where the series (8) exists z v ( z ) = z{v~(z)} E B
and zv(z) is uniformly continuous on every compact subset of Zv, s.

PBOOF. We have the identity


1 -- z k+l 1 z
(9) . . . . v. z)
k 1 -- z 1 -- z -W 1 -- z
from which the assertion of the lemma follows b y the definition of Zv,~ and
b y (6).

LEMMA 2. Let ~ c C be a curve with finite length and K c C a compact


set. Suppose u: K . y --~ B is a continuous map. Then for every continuous func-
tion f: ~, --~ C
(10) w(z) = ~ u(zt) f(t) dt

belongs to B and ks uniformly continuous on K.

2 Periodica Math. 11 (4)


274 GAWRONSKI, SHAWYER,, TRAUTNER: 8UMMABILITY OF POW:ER SERIES

R ~ A R ~ . The integral (10) is defined by the Rieman integrals of the


components wn(z ) ---- ~ un(zt ) ](t) dt.

l>aooP of Lemma 2. The first statement follows by the standard approxi-


mation of (10) by means Of generalized Riemann sums and from the complete-
ness of B. (See e.g., [1], p. 162, prob. 1.) The uniform continuity is an i m m e -
diate consequence of the inequality

IIw(z ) - w( 2) ll _< sup l -

whore l(~) denotes the length of y.


Now we are in a position to prove

Tn:EOB~.M 1. Suppose that conditions (6) and (7) are satisfied. Then ~(z) :
= {'~n(Z)} belongs to B for every z i n
G : Ap | B
and for every A p E 9.Ip. Moreover v(z) is uniformly continuous on every compact
subset of G.

PROOF. As already mentioned at the end of Section 1 we shall use the


function theoretic part of the proof of Okada's t h e o r e m given in [2] (Satz 1).
By definition of Zv, B, we have that 1 q Zv, s and hence G _ Ap. Since
Zv, s is open, it follows from (7) that there exists r 0 E (0, 1) such that
{z: Iz] ~ re} = Zv, B. Now let K be a compact subset of G. To every z0 E K
we associate a neighbourhood {z: Iz -- Zol <~ ~} = A,(zo) say, such that (cf. (22)
in [2])
(11) A,(z0) c r 0 ~ A*(z0) if " z0 ~ 0
and
(12) A,(0) = Ar~
By compactness arguments we m a y assume without loss of generality t h a t
K = A,(zo). According to (11) and (12) we consider the two cases 0 E K and
0(~K.
(i) Suppose that 0 ~ K. :Following the proof of Theorem 1 in [2] (case 1)
we get for z E K
(13) v~(z) = l-L- P(t) (it 2 c~,k _ t v~ dr,
2gi t -- z k t z

where ~ denotes a simply closed Jordan curve of finite length which contains
K in its interior and ~' = {t: Itl (1 + re}f2 }. Clearly the first' s u m m a n d on
GAWRONS]fI, SHAWYER, TRAUTNER: SUMMABILITY OF POWEP~ SERIES 9~75

the right hand side of (13) equals P ( z ) ~ c n , k and thus is uniformly conti-
nuous oi1 K. k

and hence
f ~ v.(z/t) P(t) ,dt6 B
1 - (z/O t

and :is uniformly continuous on K, by Lemmn 2.


(ii) Suppose that a q K. Again following the proof of Theorem 1 in [2]
(case 2) we get for z 6 K

If P(t) ~ 1 1~=o~ Pi(t) z v ( t )


= t- z 2k - 2~i t - - z -t n

where ?~j are curves of finite length such that z]t belongs to a compact subset
of Zv, m when z 6 K and t 6 7/- Further the functions Pj(t) a r e continuous on
?j. Now, by Lemmas 1 and 2, the desired result follows as above.
The following statement is a direct analogue to Theorem 2 in [2]. It
follows by a simple compactness argument.

COROLLARY 1. Suppose that condition8 (6) and (7) are satisfied. The~
9 (z) -----{~.(z)} belongs to B / o r every z in

@= (J A p | Zv, B,
Ap s ~ p

Moreover ~(z) is uni/ormly continuous on every compact subset o/ @.


Theorem 1 and the associated corollary only give answers to t h e question
where summability happens at all without any regard concerning a prescribed
limit of summability and so it does not' contain Okada's theorem. :But a slight
modification of the technique used above gives the desired generalization.
In particular for the case of ordinary summability, when the given power
series is summed to its analytic extension, we get additional information on
rates of summability and related questions.
To this end we define Z~z,s to be the largest open set contained in C -- { 1 }
on which the series (8) converges (for each n -- 0, 1 . . . . ), and v(z) = {vn(z)}
belongs to B and is continuous on Zv, B; that is, the sequences
1 -- z k+l 1

are not located necessarily in B (see (9)). Further consider the condition
(7') o E Z~,B
2*
276 GAWRONSKI, SHAWYER, TRAUTNER: SUMMABIL1TY OF POWER SERIES

THEORV.M 2. Suppose that the matrix 17 has finite row sums and that con-
dition (7') is satisfied. Then { ~ . ( z ) - P(z) ~ c.,k} belongs to B for every z in
k

G' = A p o Z ~ , s

for every Ap E 9Jp. Moreover { v n ( z ) - P(z) ~ cn, k} is unqormly continuous


on every compact subset of G'. k

PROOF. The assertion follows immediately from the proof of Theorem 1


by considering (13) in the form

__1 ~ P ( t ) d t ~ , c . , k _ 1 ftvnIt ) P(t) dt


Tn'z-( ) -- 2x~i t -- z k 2~i t -- z

etc. Again as above, we have the

Coao~ 2. Suppose that the matrix V has finite row sum~ and that
condition (7') is satisfied. Then {~.(z) - - P ( z ) ~ e , ~ , k } belongs to B for every z in
k

U ApoZ~,..
A~,E ~IP

Moreover {%(z) -- P(z) ~ cn, k} is uniformly continuous on every compact


subset of ~'. k

This section is devoted to various applications of the results above.

I. Choose B to be the space e o of all null sequences considered with the


s u p r e m u m norm and assume t h a t (4) is satisfied. Then B consists of all null
sequences v(z)--~ (v.(z)} which are uniformly continuous on every compact
subset K of Z~,c.. The latter statement means t h a t

sup Ivn(z) -- vn(z')l < 8,


n~O

when lz -- z' I ~ ~t(e), which implies the uniform convergence of {vn(z)} on K.


Hence Zv, c, = Zv (cf. Section 2) and Theorem 2 reduces to the theorem of
Borel--Okada in the version cited in Section 2.
II. Let
GAWRONSK], SHAWYER, T]~AUTNER: SUMMABILITu OF P O W E R SEI~IES 277

p ~ 1, r real and F ~ (Cn,k) a matrix satisfying

.~Y c,,k + (r~ + (Cn,k -- C,~+l,~) < ~ ,


~-2-~0 n~O R~O

that is (6). Thus we obtain Okada's theorem for p-absolute summability from
Theorem 1.

III. Let {wn) be a sequence of positive numbers tending monotonically


to zero. We choose

Clearly B c Co. Hence B s consists of all null sequences vn(z ) = {vn(z)}


satisfying Vn(Z) = O(wn) (n ~ ~,~) and v(z) being uniformly continuous on
every compact subset ofZ~,~. Again as in example I this means that Vn(Z) =
= O(wn) holds uniformly on every compact subset of Z'V,B. Thus, assuming
(4), Theorem 2 gives Okada's theorem for rates of summability.
(i) For example, consider the Euler method Ep defined by the matrix

V = (cn,~) = ((~)pk(1-- p) n-k) , p > O.

Since ~ c n , k z 1, (4) is satisfied. Further we have


k
Vn(Z) = (pZ -~ 1 -- p)n.
]~ence (see example I),

and thus 0 ( Z"s~,c0, i.e. (7'). Consider P(z) = ~ z k. With Ap -~ C {1) Theorem
0
2 implies ordinary summability in the region
/
G -~ A v | ~ Z"
Taking w n ~ qn, 0 ~ q ~ 1, we get that

Z'E.,B.= {Z: Vn(Z)=O(qn)} ~- lZ: IZ-- ( 1 - - 1 } < q}

and so (observe that 0 E Z ~.B.)


+ with Ap ~_. C -- {1}, Theorem 2 shows that
a "geometric" rate of summability holds for the geometric series b y means
of the Euler method, in the concentric circles Z~,B,.
278 GAW~O~SKI, SHAWYEt~, TRAUTNER: S U ~ A B I L I T Y OF POWER SERIES

(ii) (See [2], Section 4, (ii)). Consider the Sonnenschein method Sp =


= V = (cn,k) which is defined by an entire function p(z) = ~ , p ~ z ~ with p(1) = 1
and (p(z)) n = ~ c n , kz k. I t follows t h a t ~ c n , k = 1. The choice of B = co
k=l k
(see example I) leads to
z~,.c. = {~: Ipl~)l < 1}.
For B = iNs with w n = f i n , 0 < q < l , we get

z~,.B, = {z: Ip(z)l < q}

Z' Z' z sinh z P(z) = . ~ z k and Ap


(note t h a t 0 E E~,~0, 0 E ~, s,). For p(z) = sinh I ' 0
= C -- {1} we have the following exact summability domain (for ordinary
summability)
a = Z's.,co| { 1 } ) = Z~.,c, = {z: Izl Isinh zl < s i n h 1}

which is disconnected. "Geometric" rate of summabitity holds in the region

G = Z~.,B | ( C - { 1 } ) = Z~.,s, = {z: [z[ Isinh z I < q sinh 1}

whose connected components form (for decreasing q) nested neighbourhoods


of the points z = nzli, n E Z.

IV. Let {wn} be a sequence of positive numbers tending monotonically


to infinity. We choose again, as in III,

B = Bs = {t,}: II{tn}ll = < ~ 9


n>o I w~ 1
Thus, assuming (4), Theorem 1 gives Okada's theorem for the rate of diver-
gence of summability.
In particular, choosing w, = QP, Q > 1, we get t h a t

Z'E., Bs = {Z:Vn(Z) ~--- o ( Q n ) } n (~ - {l}) = / ~ : i z_ [1 _ 1 ) < Q } n (c - {1})

and so (note ~hat 0 EZ'E~,e.) with Ap = C -- {1} again Theorem 1 shows t h a t


"geometric" rate of divergence of summability for the geometric series by
moans of the Euler method holds in the concentric circles Z'E~,B. (except
f o r z = 1).
GAWItONSKI, SHAWYER, TRAUTNER: SUM~ABILITu OF POWER SERIES 279

REFERENCES

[ 1 ] J . DIEUDOI~N]~, Foundations o/modern analysis, A c a d e m i c P r e s s , N e w Y o r k , 1960.


MR 22 @ 1107~
[2] W . GAWl~ONSlrT a n d R . TRAUTNER, Versch~irfung eines Sa~zes y o n B o r e l - - O k a d a
tiber S u r n m i e r b a r k e i t y o n P o t e n z r e i h e n , Period. Math. Hungar. 7 (1976), 2 0 1 - -
211. MR 58 @ 6809

(Receivvd July 20, 1977)

UNIVERSIT~T ULM
ABTEILUNG FOR I~ATHEMATIK
D-7900 ULN[ (DONAU)
OBERER ES]~LSBERG
FEDERAL REPUBLIC OF GER~IANY
DEPARTMENT OF ]ffATHEMATLCS
UNIVERSITY OF WESTERN ONTARIO
LONDON, ONTARIO
1~6A 5B9
CANADA

You might also like