ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Séria I: PRACE MATEMATYCZNE X IX (1976)
J
ôzefA
ntoni(Bratislava)
O n modular spaces oî strongly summable sequences
Modular spaces of strongly summable sequences are investigated in papers [1] and [2]. Also some generalizations of some results of above papers are included.
Let T, T b, T 0, T f denote the space of all real sequences, the space of all bounded real sequences, the space of all real sequences which con
verge to zero and the space of all real finite sequences, i.e. sequences x = {tv}, where % — 0 except a finite number of v. Let A = {amn) be such a regular summability method for which amn > 0 and no column consists of zeros only. Let Ф = {(pv} denote a sequence of 97 -function (non
decreasing continuous function 9 ? defined for w > 0 for which 99 ( 0 ) = 0 ,
<p(u) > 0 when u > 0 and lin ^ w ) = оо). The sequence will be denoted
U->OQ
b y x — {£„}, у = {$„}, в = (0), xn = (tlt t2, tn, 0, 0, ...) etc. We adopt th e following notations (as in [ 2 ]):
OO
~ ^ , ^'mv(Pv{\%\)}
v = l
T ° = { x : <r*(a?)->0 as m ->00 and super^(<B)< 00},
m
Т*ф = {x: there exists a
A > 0such th a t
A a ? e T °},Т ф = {x: kceT0 for all A > 0}.
The sets T% and Т ф are linear spaces as it is easy to see. The functiona sup o*(x) for xeT°,
m
00
for xeT% — T°t
е*ф{х) =
can be defined on Т ф. This functional о*ф is satisfying the conditions:
A l.
q%(
x) = 0 if and only if x = 0.
A2. й (-а > ) = q I ( ho ).
A3. Q^ax + by) < £*ф{я) -j- Q%{y), where a, 0 and a + b = 1 .
B l. Q%{kc) 0 as A — >0.
2 J. A n t o n i
We shall be interested in spaces Т ф. On Т ф we can define a F-norm generated by modular дф = дФ/тф as follows:
||
я| 1
ф= i n f > 0: q (Û
< s; •
If all functions <pv are s-convex (0 < s < 1 ), then a norm can be defined apcording to the following formula:
INI® = inf je > 0 : вФ\^1 <
The relation between the modular and the norm convergence is the following: ||ж и||ф'— >0 if and only if дФ(лхп) -> 0 for all Я > 0.
T
heorem1 . Т ф is a complete linear space with respect to the F-norm || • ||ф.
P ro o f. Let be a Cauchy sequence, i.e. \\xp — <юй\\ф -*»0 as_p, q ^o o . The above is equivalent to the condition @ф(Л(хр — xQ)) -*0 for all Я > 0.
I t follows th at for a given &> 0 there is a positive integer n 0 such th a t
OO
SUP Z amv<PA^\tf~ta v\) < Ф for p, q ^ n0. The condition Q<p(h(xp — О
m v= l
implies th a t the sequence {l„} converges for all v and therefore there exists a sequence x — {tv}, where % — lim<“. Let m, n be positive integers.
n->co
Then for all m, n and for p , q > ?i0 the formula П
) < * / 2
is valid. From the last inequality we get
П П
lim £ a mv<pv(X\t?-t?\) == ] ? a mv<pv(X \t? -tv\) ^ s { 2 .
q->oo j r=i
00
As the last formula is true for all n, it follows th a t £ amv<pv{h \tf — $,[)
OO t > = l
< e/2 for all m. From this we conclude th a t sup ^ amv(pv(X\tf — tv\) < s.
m v= l
We proved also th a t the sequence {xn} converges to æ in norm. We are going to show th a t а>еТф:
OO OO
£ amv<Pv(*\tv\) < £ amvcpv{h 1C —^1 +Я1С1)
v = l V —1
OO
2 Я |С -У ) + ^(2Д|С1))-
v = l
As {t™}eT0, £ a mv(pv(2X |C!) ->0 as m -> oo and for all X > 0. Since
V = l 00
the sequence {xn} converges to x in norm, we have JT 1 am„<j5,,(2AjC —<„ 1 ) ->0
and х е Т ф. *’=1
R e m a rk . In the proof we have taken into account th at if {xn} is a Cauchy sequence, then the sequence {Ç} is a Cauchy sequence, too.
This statement is a consequence of regularity of the m atrix A and of the supposition about the columns of the matrix.
T
heorem2. Let х е Т ф. Then j|& — хп\\ф-> 0.
P ro o f. Let x = {£„}, xn = (txi t2, ..., tn, 0, 0, ...). Since х е Т ф, a^{Xx) -> 0 as m ->
oo,i.e. for a given
e> 0 there is a positive integer m 0 such
OO
th at o*{xls)< e /2 for m > m 0. Since the series ^ a mv%{\tv\Je) converges, there is a positive integer v0 such th a t v==1
Then we get
for m = 1 , 2 , . . . , m0.
9 ф
x — xn e
< sup V (—) + sup V am vç>v (—)< e
' £ / m>mo ^ \ e / for all n > v0 and the proof is complete.
C
orollary. The space Т ф is separable with respect to the norm ||*||ф.
The proof follows immediately from the inclusion T f с Т ф for any sequence Ф and Theorem 2.
Papers [1] and [ 2 ] show th a t Т фп Т ь = T v n T b if sequences Ф and W are stationary. In case of arbitrary sequences this equality does not hold as we can see from the following example.
E
xample. Let us put <pn(u) — un and y>n(u) = ulln. The sequence x = {1 /п }еТ ф, because
lim
m -*oo
= о
for all X > 0 ((amn) is a regular matrix). The sequence {1 jn }4 T v , because lim
m->oo
у X— 1 '
= 1 .v=l
4 J. A n t o n i
This example shows th a t T 0 cannot be a subset of Т ф for any sequence Ф of ^-functions.
Now we are going to give a necessary and sufficient condition to have T 0n T b m T wn T b. The theorem containing th a t condition will be preceded by Lemma 1 To express Lemma 1, for every subset В of the set Ж of all positive integers we define
OO , 0 0
ЪЛ{В) = limsup ] ? a mvxB(v), ô A(B) = lim J ? amvxB(v)
m ->oo v==i m -+ co Vss.j
if exists and where %B is a characteristic function of th e set B.
L
emma1 . Let x = {%}, tv > 0 be a bounded sequence. Then lim £ amvtv
— 0 i f and only i f for every e > 0 , $A ({v: tv^ e}) = 0. w^°° V=1 P ro o f. Assume th a t for every e > 0 , SA ({v: tv^ e}) = 0 . Let us write B e = {v : tv^ e}. Let tv^ K (v = 1, 2, ...). Then we have
OO
limsup amvtv < limsup J ? u,w^ + limsup ^ amvtv
m —*oo V=1 m->oo veBe m -+ oo Vej v - S s
< КдА (Ве) + едА( Ж - В Е) = s .
OO
Since s > 0 is arbitrary, lim fff amvtv = 0 .
m—>oo v = i
Now we are going to prove the inverse implication. Suppose th a t there exists a number e > 0 such th a t 3A (Be) > 0 for some sequence {£„}.
Then we have
limsup amvtv > limsup amvtv > eôA{Be) > 0
rn-уоо у==1 m -* со veBe
and thus lim ) fa mvtv Ф 0 .
m-> oo v = l
oo oo oo oo
Let X =
/ 7 ( - oo, oo), Y =/7
<0 ,
oo)and Z =
UП < ~ к , К ) .
n = l n —l K = 1 n —1
A basis of neighbourhoods of point zero is to be introduced in the spaces X and Y. The basis depends on the m atrix A and consists of all sets of the form:
0 eB = {(h, h , \tp\ < e for veB }, 0 eB = {(^, t2, ...) € Y : tr <i s for v € B f
for all e > 0 and for all В с Ж such th at ôA (B) = 1. I t is easy to show th a t the above mentioned system forms a basis of neighbourhoods of zero. The function Ф: A->Y can be associated to every sequence Ф — {y>n}
of «^-functions in such a way th a t for every projection nn we have
тгпоФ = <pnQ7in, i.e. Ф((^, t2, . . . ) ) = (<Pi(\h\), <P*{\h\), •• •)•
The following characterization of the space Т ф can be given by means of the function Ф. The sequence {Q еТф if and only if the sequence {' Ч^!}€П Thi s assertion can be easily proved. According to
e>0
В
Lemma 1 the statement {tv} eТ ф is equivalent to the statement bA({v:
(pv{X\tv\) < e}) = 1 for all À > 0 and s > 0. The last relation gives {A|(,I}«U®_1(Oæ ) = {ft}: VÀ l<,l)<« for ve B)
В
for every s > 0 and all Я > 0 .
T
heorem3. Let {(pn} and {ipn} be two sequences of <p-functions. Then T wn T b c= Т фп Т ь i f and only i f
n u ^ ) n ^ c п и ф -1(ОеВ) п г .
e>0
В
«>0В
The function 4? is analogously defined by the sequence {yjn} as the function Ф by the sequence {qn}.
P ro o f. Let {tv} e T уГ\Тъ. Then
c= n U ф~1{оеВ) ^
e>0
В
e> 0в
for all Я > 0 and therefore {tr}eT 0 n T b.
The inverse implication can be proved by an equivalent relation.
The inclusion T ^ n T b с Т фс\Тъ is equivalent to
Уф У У дл({”- V vU ftlX *}) = 1 ~ à A({v. % w t,)< e } ) = 1 .
e>0 Л>0 '
Putting Я = 1, we have th a t y>v(\tv\) < e}) = 1 => bA[{v:
<Pr(\K\) < «}) = 1 for all {tv}eT b and every e > 0. Let 0„}еП U r _ 1 (0 £jB)n Z .
e>0
В
That means th a t for a given e > 0 there is a set В such th a t ôA(B) — 1 and %(\tv\)< e for veB . Therefore we have bA ({v: %{\t„\) < e}) = 1 . By supposition it holds ôA ({v: (pv{\tv\) < e}) = 1 . This means th a t {tv}e Г){JФ ~1{OeB) n Z and the proof of theorem is complete.
e>0 В
In the case (pn = <p and ipn = xp the condition of Theorem 2 is satis
fied. I t can be easily verified in the following way. Let à* = inf and bf = inf {y)_ 1 (e)}. We show that
Ф~ЧО,в ) = V - \ O w > ) and Ф -'(0 < )В ) - r - \ O lS), ф-ЧО.в) = {ft}: f ( M ) < « for re B ) = {ft}: \t,\ < S' for r« B ).
r ' { 0 ^ ) = {ft}: V (ftl)< V(%) t o = {ft}: \t„\ < »1 for v * B ).
Similarly we can obtain
<t> = { M : 14 < 4 ' for re B], r - ' ( 0 . B) = {{*.}: 141 < 4 * for veB }.
Since d* and ôJ converge to zero as s -> 0 , the equation
n U *-l(Ow)nz = nu r n o ^ n z
e>0
В
e>0В
is true. Thus, we have obtained the equality Т фслТъ — T Pr\T b.
Specifying the sequence of ^-functions as well as the matrix A we can obtain the more convenient conditions. For example, if we choose as (pn = uPn and A as the matrix of Cesàro method, then the theorems are true. ( I A = XVn.)
T
heorem4. Lei
{p
n }and {qn} be two convergent sequences of real number such that p n ^ a > 0, qn ^ a > 0 (n — 1 , 2, ...). Then Tp^c\Tb = T Qnn T b.
The proof is easy and we omit it.
E e m a rk . If p n ^ p (pn > p), then TPn => Tp {TPn c Tp). We can verify this assertion in the following way. Let {tv} eTp. Let us put for Я > 0, t’v = tv if À\tv\ < 1 and t' = 0 for other v and t" — tv — t'v for all v.
Then {t’9}eTPn because TPn<~\Tb = Tp r\T b. Since
then {t'v'}eTPn, and thus {tv} e TPn.
T
heorem5. Let p n ^ a > 0, p n ->p. Then TPn = Tp if and only i f
\pn - p \ lnw = 0 ( 1 ).
Let us put p n = p + en. To prove the theorem, we will distinguish the following cases:
1 ° еп > 0 for infinite many n and en < 0 for finite many n.
2° sn > 0 for finite many n and sn < 0 for infinite many n.
3° en > 0 and en < 0 for infinite many n.
4°
еп> 0 and en < 0 only for finite number of n.
I t is obvious th a t in case 4° we have Tp = Tp . We are going to give the complete proof only in case 1 °; the proofs in other cases are similar.
I t is obvious th a t T„ <= Tn. We wish to show th a t T n a T n . Assume th a t
En \ n n— 0(1). Let К be so chosen th a t en ln w < K . Let {tv}eTp . Since {tv}eTp , then the necessary condition for (0,1) summability, i.e.
m CO
(M tj)1
n 0 is fulfilled. I t follows th a t {X\tn\)p < n for sufficiently large n or psn Inn < sn Inn <
K .Let en > 0 for n ^ n 0. Given s > 0, there exists a positive integer m 0 > n 0 such th a t
1
m Y w k
<m n.(j+l
«о ЛЛ
2 (K + l) p
<
for all m > m0. Therefore we get
1 m
m
«о
I У л т ) р* - ( т ) р m I A-i
V =1
n0< —-I Y W m I AmJ *>=1
d* ' - ( m r
< «.
+ — I У ( m i ) * - (m n *
m I«o+i
+ — \ Y w t,\)piW t,\Yn- i )
m ! A j '
«■ 0 + 1
Thus we have proved th a t {tn}eTPn.
Next we prove the inverse implication. Assume th a t limsup |sjln w
n
= oo. The proof will be finshed if we show th a t there exists a sequence {tn} eTp such th a t {tn}4Tp . One can choose a sequence of integers which satisfies the conditions:
(a) limeftfclnw.fc =
oo,Jc-> oo
(b) nk > (пк_г)Рпь,р, (c) lim inf--- — b > Пу, 1 .
k~><x> W 'k—1
Then we can choose a non-decreasing positive function defined on the set of all positive integers such th a t
(а) П т / ( 7 г ) = oo,
А-» oo
Ф) /(&) < — ln% for all nk where a = supp n
(t
8 J. A n t o n i
We put f(k ) = — Innk for all h if the last is inequality satisfied nk
CL
Ъу putting — ln n fc for ffk ) m it. Let us put tv = 0 for v Ф nk and
n kQj
lk - 1
, Пк П,
*“* = U W ~ / ( * —i)
1/p
The sequence {tv}eTp since for nk ^ . m < nk+1 we have
m m J (k )
But the sequence {\tv\Pv} does not satisfy the necessary condition for ( 0 , 1 ) summabiliby, because there holds:
\K\Pn ^ / nk
lim su p ---> limsup — ---
n-* oo Г1- k-*oо Uk
п Ъ 1а
— limsup ih n. 1
fc-J-00
ГЬк \f{h) f( k 1 )
1+ en klp
= limsup
f c - > OO
J \ h) > 1 .
Therefore {tv}4Tp and the proof is finished.
B ern a rk . In the case when p n oo, qn ->
ooas n
- » ooTheorem
3is not valid.
The following theorem is a generalization of Theorem 2.4 of [ 2 ].
T
heorem6. I f Т ф с T w and x {cT 0for all i, then \\х^\\т -> 0 i f ||а?г-||ф -> 0.
We get the theorem from the closed graph theorem and from the coordinate convergence which is a consequence of the convergence in norm.
Next we investigate spaces of strongly summable sequences.
D
efin itio n. The sequence x = {tv} will be called Ф-strongly summable to number t — t(x) if {x — t}eT 0 . We shall denote the set of all Ф-strongly summable sequences by Т(Ф).
D
e fin itio n. The sequence Ф = {<pn} of ^-functions is called lower - regular at the point zero if there exists a ^-function xp and a fixed neigh
bourhood U of the point zero such th a t xp(u) < inî<pn(u) for all ue U.
n
T
heorem7. Let Ф
=pn be lower-regular and equiconiinuous at the point zero. Then Т(Ф) provided with F -norm ||*||ф is a complete, normed linear space, and the coordinates are continuous functionals on Т(Ф).
First we show th a t the modular q 0 satisfies condition B1 for every
хеТ(Ф ). Suppose хеТ(Ф ), x = {tP}. Then there is a pumber t such th a t
{tv — t}e T 0 . We have
CO o o o o
X amv9v(M%\)< ]?^< р Л Ы \К ~ Ц ) + £ amv(pv(2X\t\)
V — l y = l J>=1
for all m, and thus
oo
9 ф {Л\%)< ( 2 Я
(я?— *)) -h snp Y amv<pv{2Â\t\).
m V = 1
Since {x — t)e T 0 , Q0(À{æ — t)) ->0 as Я ^ О , it follows from the equi- continuity of the sequence Ф a t 0 th a t to a given s > 0 there is a ô > 0 such th a t <pn(u) < s/2M for u < ô and for all n (M is such a constant th a t JT \amv\ < M for m = 1 ,2 , 3, ...)• If 2X\t\ < ô, then sup ]?amv<pv(22.\t\)
v = l oo m v = l
< e. Thus we get sup ^ a mvcp(2X\t\) ->0 as Я ->0. Thus we proved th at q 0.
m v = l
satisfies condition B1 for all хеТ(Ф). I t remains to show th a t Т(Ф) is complete. Assuming oon = {%}еТ(Ф) and дФ[Х{оор — xa)) -» 0 as p, q -> oo,.
for every Я > 0 , we get tn -> t as n -> oo for every v. We show th a t {æn}
CO
converges to хеТ{Ф) in norm. Given e > 0, then sup amv<pv{X\t? —
m v = i
< e/2 for sufficiently large p and q. Therefore we have
OO
amv<pv{Â\t? -t? \) < e/ 2 , V =1
r
amv(pv(X\t?-t?\) < e /2
v — 1
for all positive integers m and r. Using the last inequality we get
r r
lim ] ? a mv<pv{À\t?-t«I) = ^ a mv<pv( À \t f - t v\)< e/2.
Qr-*OOv — i i>= 1
But the last inequality is true for all m and r, therefore we have^
OO
sup £ amv(pv(A\tf — tv\) < e, i.e. the sequence {,xn} converges to oo in norm.
m v = i
Take e > 0 and let tn = t(oon). Let щ Le such ^-function th a t щ (и)
< inf<pn{u) for every ue U, where Ü = < 0 , a); putting y{u) = y ^ u ) for П
ue U and гр(и) = lim щ(и) for и > a we obtain a function which satisfies.
u-*a~ ц->а~
yi(u) < <pn(n) for all и 3 s 0 and for all n. Therefore we have
10 J. A n t o n i
Multiplying by amv and summing we have
ww oo
I tP ~ t q\ X amv(<Pv { \t? - tp\)+<Pp(№-%\)) + ^ amv<Pv ( W ~ tq\) • V =1
Since
V =1
oo
lim ^ a mvy { l\tp - t q\) = v (l\tp - t q\)
m— >oo v—i
oo
lim ^ am(pv(\tf ~ t q\) = 0 ,
7П-+00 v=i
OO
lim £ amv<Pv( Itq - tq\) = 0 , J V =1
v= 1
^nd, the sequence {#„} is a Cauchy sequence, i.e. sup £ amv<pv{\tf —19\) < s
m v = i
io r sufficiently large p and q, thus we have yj(^ \ip — t9\) < e for sufficiently large p, q. Hence {tn} is a Cauchy sequence. Let tn ->t and X > 0. Then
we get
OO oo
У < У « т,9>ЛЗЛ |е-П ) +
OO 00
V = 1 V = 1
oo oo
< sup У ^mvtyv (ЗА\tv ^|)H“ ^mvtyv (ЗЯ 1 1
' v=l
oo
у »«,?>.( ЗЯ 1 С - П ) . V = 1
oo
Since {a?n} converges to я?, then sup £ a mv(pv(3X \iv — tf\) converges to :zero as p->oo. Suppose m v==1
CO
дф(ЗХ(х — œp)) = sup J^<V9?v(3A|t,,— if|) < e/3
for p > p i. For a given e > 0 there exists a number ô > 0 such th a t q>v(u)
< e/3 for all u < ô and all v. Assume 3X\tp — t\ < 6 for p ^ p z. Then we have
OO
lim ^ amv(pv(X\tv — t\) < e/3 + e/3 + 0 < e.
m -> oo j,_.2
Hence the last inequality is true for arbitrary e > 0 , thus we have
4>еТ(Ф).
T
heorem8. The generalized limit t(œ) is defined uniquely and it is a modular continuous functional on Т(Ф).
P ro o f. Assume th a t there exists a sequence sc = {%}еТ(Ф) for which t(sc) = rx, t(oc) = r2 and rx ф г г. Then we have
< 9VU \ К - Гл\)+%{МК-Гь\).
Multiplying by amv and summing we get
CO
lim ^mv У*
m->oo
v=l
X
~2
\r1- r 2\ < lim
oo
CO
^am v< PvW v- r i|) +
j>=1
+ l i m V
amv<pv{ l \ t v — r 2\)
= 0 .ПГП—i/Y»
Thus rx = r2, and this contradiction proves the assertion.
Now suppose сспеТ(Ф), осеТ(Ф), and, дФ[Х(ссп — ж)) -> 0 f o r a fixed A > 0 . We shall show t(scn) -+t(sc). Let q[X{œn — sc)) < e for n ^ n 0. Then we have
Multiplying by amv and summing we get
(
Я \ 00— \tn — t\ К lim ] ? а тР(срфА\Р-%\)+<рг{Х —$|)) +
O J m->oQ
00
+ sup y a mv(pv{ X \t^ -tv\).
m „ = 1
Therefore it holds y)(^X\tn — t\) < e for n ^ n Q and consequently t{a>n) — > t(sc).
References