• Nie Znaleziono Wyników

On inclusions between Orlicz spaces

N/A
N/A
Protected

Academic year: 2021

Share "On inclusions between Orlicz spaces"

Copied!
13
0
0

Pełen tekst

(1)

ROCZN1KI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Séria I: PRACE MATEMATYCZNE XXVI (1986)

Ma r ia n No w a k

(Poznan)

On inclusions between Orlicz spaces

Abstract. Let <p, ф be ^-functions and (E, I , ц) be a measure space with a positive measure. Let L*4’, L0,p -denote the Orlicz space and the space of finite elements, respectively. The main aim of the paper is to show that if there hold the equalities: L0* = (J L*v, L*v — f) L0^,

<peT &eS

where for every <pe T (феЕ) the proper inclusion L*41 ç L°* holds, then the set T (resp. S) must be uncountable.

In [2] and [7] there are two equalities proved:

( + ) L°* = (J

L*9,

( + + )

L** = C\L°*,

4> * Ф

where the union (intersection) is taken over all ф-functions q> (resp. ф) such that for all c > 0: ф{си)/(р(и)-^> 0 as u-+ oo in the case where E is a compact subset of Rm with the usual Lebesgue measure or ф(си)/(р(и)-► 0 as u-> oo and и-* 0 in the case where E = Rm.

Equality ( + + ) is used in studying some linear topologies in Orlicz spaces (see [8], the proof of Theorem 5.1, [10], p. 45). In [1] the equality L1 = (J W is proved, where I f denotes the

<P

Orlicz class, E is a compact subset of Rm and the union is taken over all iV-functions q>.

From Theorem 3.1 of this paper it follows that in equalities ( + ) and ( + + ) the proper inclusions L*q> ç L0^ hold and hence, by Theorem 4.3 (Theorem 5.3), the above union (intersection) is taken over uncountable set of (^-functions </> (resp. ф).

PRELIMINARIES

0.1. ^-functions

0.1.1. It is said that a function <p\ [0, oo)-» [0, oo) is a (p-function if it is continuous, non-decreasing and such that q>(0) = 0, q>(u)> 0 for u > 0 and (/>(u)-> oo for u — > oo ([4]).

0.1.2. Let ф and (p be ^-functions. We shall denote ф -kq>

{ф ■< (p, ф -k <f>) if there exists a constant c > 0 such that lim inf (р(си)/ф(и) > 0 as u-* oo and u~* 0 (resp. as w—► oo, as n —► 0). In other words, the relation ф A (p (ф -< (p, ф -к (p) holds if there exist constants c, d > 0 such that ф(си) ^ dq>(u) for и ^ 0 (resp. for и ^ u0 ^ 0, for и ^ u0, u0 > 0) ([4], p. 123).

0.1.3. Let ф and q> be (^-functions. We shall denote ф <1 cp

(Ф < q>, ф < (p) if for an arbitrary c > 0: lim sup ф(си)/(р(и) < oo as u-+ oo

(2)

and и -+

0

(resp. as и -*

go

, as и -+

0 ) .

In other words, the relation ф

q>

<

<p, ф

< q>)

holds if for an arbitrary c

>

0 there exists a constant d

>

0 such that ф(си) ^ d<p(u) for и ^ 0 (resp. for и ^ u0 ^ 0, for w < u0,

u

0 > 0) ([8], p. 72).

0.2. Orlicz spaces

0.2.1. Let

q>

be «^-function and let

( E , I , p )

be an arbitrary measure space with a cr-additive and positive measure p on a-algebra I . For a real valued, /i-measurable function x defined on

E

we write

Qv (x) = J <p(\x{t)\)dp.

E

oo

It is clear that if p is a purely atomic measure, i.e., E = (J e^, ea m being m= 1

different atoms, then ^-measurable functions x in E are constant in each of the sets eam. Then we have

M x) = Z m= 1 where x (t)= c m for fe e ".

By L*(p we denote the linear space of those functions

x

for which Q^iXx)

<

o o

for some Я > 0, by L0<p the linear space of those functions

x

for which

qv

(Лх)

< o o

for all Л > 0 and by Lv the set of those

x

for which

qv(x) < o o .

The linear spaces L*<p and L0<p are called the Orlicz space and the space of finite elements, respectively. The set L? is called the Orlicz class. Moreover, it

is known that L0<p is the maximal linear subspace of L** which is contained in V ([6]).

Let (E, I , p) be a finite (infinite) measure space. Then by B(E) (B0(E)) we will denote the set of all functions defined and /{-measurable on E, which are bounded (are bounded and vanish outside of some set of finite measure).

0.2.2. The inclusion L*v <= L** holds if and only if

(a) ф -kcp (ф <p) if (E, I , p) is an infinite (finite) atomless measure space ([4], 2.4),

00

(b) ф -к. (p if E = (J ea m with 0 < inf p(ea m) < sup p{ea m) <

o o

([3]).

m=

1

m

m

0.23. A sequence (x„) in L*v is called (p-modular convergent to x e L **, in symbols х„Д x, if there exists a constant X > 0 such that ^ (A(x„ — x)) — > 0 ([6])-

0.2.4. In L*v, an F-norm can be defined as follows: ||x||v

= inf {e > 0: Q^ix/e) ^ ej. Moreover, ||x„||v -*■ 0 implies x„ 0 for a sequence

(x„) in L*v ([6]).

(3)

R E S U L T S

1. In this section we shall give necessary and sufficient conditions for the inclusion L*v c= L0^.

1.1.

Th e o r e m.

The inclusion L** a L0* holds and simultaneously x„-^0 implies ||x„|j^ -*• 0 for every sequence (x„) in L*<p if

(a) ф <1 (p {ф < (p) if (£, Г, p) is a measure space (a finite measure space), 00

(b) ф < (p if E = (J eam, where ea m are different atoms with inf p(e^) > 0.

m — 1 m

P roof, (a) Since ф < (p (ф < (p), for every natural number m there exists a constant dm > 1 such that ф{и) < dm (pm(u) for и ^ 0 (resp. for и > um ^ 0), where (pm{u) = (p(u/m). Hence by ([4], 2.23), we have immediately

GO

L*9 = (J l f m с Lf and since L°* is the maximal linear subspace of £**

m = 1

which is contained in Lf, we get L*< p c: L0*.

From ([8], Lemma 3.1) and ([5], 3.32), we have that x„-^0 implies

\\хп\\ф -* 0 for every sequence (x„) in £*ф.

(b) First, we shall prove that the inclusion L*v с L°* holds. Obviously, it sufficies to show that L** c i f . Indeed, let x e L * <p. We have x(t) = cm for tee°m and denote c = inf p{ea m) > 0. Hence there exists a number A0 > 0 such that

m

00 X

(1) с X <p(A0 \cm\)< X

4

> M c m\)p{ea m) = ^(A 0x )< oo.

m - 1 m = 1

Since ф <3 (p there exists a constant d > 1 and a number u0 > 0 such that

(2) ф(и) ^ d(p(À0u) for и < u0.

From (1) it follows that lim <p(A0|cJ) = 0 and hence there exists a natural

m->oo

number N such that (p(À0\cm\) < (p{À0u0) for N. Hence |c j <

m

0 for w ^ N. Therefore from (2) and (1) we obtain

вф(х)= X Ф (\cm\) p(ea J = X Ф(\Ст\)р(еа т)+ X Ф(\Ст\)РЮ

m = 1 m = 1 m= N

N - 1 oo

< X + d X <p(lo\cm\)p{ea m) <

00

.

m = 1 m = N

Thus x e L f.

Now, we shall show that 0 implies ||хп||^-> 0 for every sequence (x„) in L*v. Indeed, let x„-^ 0. We have x„(t) = c” for te e a m. Then there exists a constant A0 > 0 such that

= f <PttoK\)p(ea J - ^ 0 .

m = 1

(

3

)

(4)

We shall show that j|xj|^ -> 0. In fact, let г be an arbitrary positive number.

Since ф < (p, there exist a constant d > 1 and a number u0 > 0 such that

(4) ф(и/£) ^ d(p(À0u) for u ^ u 0.

Denote E„ = { te E : |x„(OI < «о} and x'n = xn-xEn, x'„' = -X

e

-

e

„- We have Qv^ox^) ^ (p{X0u0) ii ( E - E n) and hence by (3) there exists a natural number IVj such that p{E — E„) ^ (Я0*„)/<?(Я0uo) < c for N x.

Therefore p{E — En) = 0 for n ^ IV j and hence

(5) x" = 0 for n ^ N i .

On the other hand, from (3) it follows that there exists a natural number N 2 such that

CO

(6) = X И Л )к«1)М О <еА * for n ^ N 2.

m= 1

Then from (5), (4) and (6) for n ^ max(N lt N 2) we get

* 00

e*(xje) = Q^ix'Je) = £ ф (\c^\/e)p (C n En) m= 1

00 g

£ <pC*0ld ) ju ( O ^ d*- = £,

m = 1 d

i-e-, 1Ы* ^ 8-

1.2.

Th e o r e m.

Suppose the inclusion L <=. L0* holds. Then

(a) ф <\ (p (ф < (p) if (E , Z, p) is an infinite (finite) atomless measure space,

CO

(b) ф < q> if E = (J ea m with 0 < inf/ 4 0 ^ sup / 4 0 < oo.

m = 1 m m

P ro o f, (a) Since L** <= L0^, we have L** = (J l f m c i f , where (pm(u)

m= 1

= (p{ujm). Hence by ([4], 2.23) for every natural number m there exists a constant dm> 0 such that ф(и) ^ dm(pm(u) for и ^ 0 (u ^ um ^ 0) and this means that ф <i <p (resp. ф < </>).

(b) Suppose L*v c L0^ and let us assume that ф < q> does not hold. Let

<5 be a number such that 0 < à < 1. Then there exists a constant c0 > 1 and a sequence (un) of positive numbers such that

(1) (p(un) ^ 3/(2nd) and ф(с0un) > 2"<p(u„), where d = sup / 4 0 -

m

Then there exists a sequence {EJ of pairwise disjoint sets in E such that

(2)

d

< / 4 0 ^

2 > (

m

„) <

р

Ы

(5)

Define

un for te E n, n = 1 ,2 ,...,

CO

0 for t e E — IJ E„.

Then

00

M * ) = Z Z < p W /(2> M = i >

and hence

x g

L*v.

On the other hand, from (2), (3) and (1) we have

00 00

M c‘o *)= Z *Mco un)p(E n) ^ £ ф (co un)((1/2"(p{u„)) — d)

GO

> Z (1-<5) = 00>

and hence хфЬ0ф. Thus we are led to a contradiction.

1.3.

Th e o r e m.

Suppose xn-^

0

implies

llxJI,/,-* 0

for every sequence

( x n)

in B0(E) (B(E) if p(E) < oo). Then there hold conditions (a) and (b) from Theorem 1.2.

P ro o f, (a) It follows from ([11], Theorem 2.4).

(b) Suppose x„-^0 implies ||xn||^,->0 for every sequence (xn) in B0(E) and let us assume that the relation ф < tp does not hold. Then conditions (1) and (2) of the proof of the previous theorem are fulfilled. We shall show that there exists a sequence (xn) in B0(E) such that хпЛ 0 and HxJI^-y> 0.

Then xne B 0(E). We have e„(x„) < g (м„)/(2п <p (n„)) = 1/2" and hence x„-^0.

On the other hand, we have

<P*(c0 *„) = Ф(c0 un)p (£„) ^ ф(с0ип) - ( \ - 2"dtp(M„))/(2" <p(u„))

> l- < 5 > 0 and hence ||х„|Ц -f 0.

2. Now, we shall give a necessary condition for the inclusion Ь0ф c L*<p.

Namely, the following holds:

2.1.

Th e o r e m.

There exists х е Ь 0ф such that хфЬ*4* if

(a) lim inf ф (cu)/q> (u) = 0 for all c > 0 as и -> оо or u~* 0 (и -> oo ) if {E, I , p) is an infinite {resp. finite) atomless measure space,

Define

for te E n,

for t$E„.

(6)

(b) lim inf ф(си)/ср(и) = 0 for all o O if E = (J ea m with

и -*0 m —

1

0 < inf ц{е^) < sup ц(е„) < oo.

m m

(c) lim \l/(cu)/(p(u) = 0 (lim ф(си)/(р(и) = 0) for all c > 0 if E = U ea m

u-*0 и

— *oo

m=

1

with inf /i(e“) > 0 (inf ju(e“) = 0).

m m

P ro o f, (a) Under the assumption for every pair of natural numbers (i,j) there exists a number ui} > 0 such that

ф(21+]ии) < (p(uu/i)/i2

(resp. ф (Т+]ии) < (p(uu/i)/i2 and ф{2*'ч ии) ^ n 2/2).

Now, let Еи (i , j = 1, 2 ,...) be pairwise disjoint /z-measurable sets in E such that

И(Еи) = l/(/22 W +4 ) )

(resp. Ц (Eij) = Ц (E)/(i2 2J ф (2i+j u0))).

for teE ij, i , j = 1 , 2 ,..., for t e E - U Ü Еи .

«= l j=l

First, we shall show that x e L i.e., ^(A\x(t)\)dpi < oo for all A > 0.

E

Indeed, let Я > 0 be given and choose a natural number N such that Я < 2N.

Then we have Define

*(0

oo oo

^(A \x (t)\)d n = £ ( Z Ф№ij)v(Eij))

E i = 1 j = 1

Then

^ X (X ф(Аии)ц{Еи)) + £ (X Ф(2Nuij)и(Ец))-

i = 1 j = 1 i = N j = I

00 00 00 00

I d Ф(2"щ,)1л(Е„))*; £ d ^ (2 'tg /( ;22 w ' +' “«)))

00 00 00

^ Z ( Z VO’2^) = Z V*2 < °o

i

=

n

i i = iv

oo 00 00

( I d ^(2"иц)лх{£у)) ^ д(£) I 1/i2 < oo).

i = JV j = 1 i = N

(7)

Moreover, for every natural number i'0 with 1 ^ i 0 ^ N — l there exists a natural number j 0 such that Я < 2l° +Jo. Then we have

oo J 0 1 oo

Z ^(ÂulV)At(£/o/) < Z Z

J= 1 J= 1 •/'=

jo

ф{2‘0+1ои,*

oj

M M

J0~ 1

s: X 1ИЛ«,0, ) М М +

j - i Z ^(2 ,0+' 0«,ы)/(.а2 ^ (2 '0+^ 1Ы))

« I + P £ 1 / 2 ' < с о .

j=l *0 7=J0

oo -/'о- 1 u / £ \ oo

(I ♦ f c . / l l i l i v X I + S 1 /2 '< CO).

7=i

j

- 1 *

j=Jo

Hence

/V— 1 00

Z ( Z •И Я и .о у Ы ^ Н oo.

i = l J = 1

Now, we shall show that хфЬ*4*, i.e., § (p(2\x(t)\)dfi = oo for all Я > 0.

E

Indeed, let Я be an arbitrary positive number and let M be a natural number such that Я > 1/M. We have

{<р(Я|х(Г)|)4и= X ( Z Ф(Яму)^(£у))

7 = 1 i = l

00 00

^ Z ( Z Ww.j/O^ (£«■;))

7 = 1 i — M

00 00

^ Z ( Z i2 Ф (2‘+j “y)/(*2 2j Ф (2,+J «y))) 7= 1 i= A#

oo x x

= 1 ( 1 1/2') = Z 1 = °°>

7* = 1 i = M i = M

x *

(resp. |

ф

(Я|

х

(0 | ) ^ ^ Z M£) = oo).

£ i=M

(b) Let Ô be a number such that 0 < <5 < 1. Since for all c > 0:

lim inf ф(си)/ф{и) = 0, for every pair of natural numbers ( i,/) there exists

и

-*0

a number ui} > 0 such that

Ф (2i+j ии) ^ S/(di2 2j) and ф (2 +j uu) < (p(uu/i)/i2,

where d = sup ц(еа т).

(8)

Then there exist pairwise disjoint sets

cz

E such that

i2 2) ф (2‘+j ии) d ^ р(Ей) <

122 ф (21+)ии) {i,j = 1, 2, ...).

Define

*(0 = 0

for te E u, i , j = 1, 2, .., 00 00 for te JE - u

U

Eu.

i = l 7 = 1

Similarly as in the proof of (a) we show that and хфЬ*<р.

(c) Since for all c > 0: lim ф (cu)/(p (u) = 0 (lim ф(си)/<р{и) — 0), for

и-*0 u->oo

every pair of natural numbers (i, j) there exists a number vu > 0 such that ф(2'+]и) < <p(u/i)/i2 for и ^ i’и (resp. for и ^ vu).

Since inf p(e„) > 0 (resp. inf ju(e^) = 0) there exist pairwise disjoint sets

m m

Eu in E such that

n(Eij) ^ l/( r 2Jф(2‘+jvu)) (resp. p(Eu) < 1 /(/2 2J‘ф( 2 +jvu)))

for every pair (i, ;). Hence for every t>fj there exists a positive number му ^ i?y (resp. Uij^Vij) such that p (£y) — l/(i2 2j ф (2i+j ии)) and simultaneously, ф{2'+]ии) ^ (p{uu/i)/i2.

Define

for teEfj, i , j — 1, 2, ...,

for Û Ü %

i = l J = 1

Similarly as in the proof of (a) we show that x e L 0* and хфЬ**.

From Theorem 2.1, Definition 0.1.2 and Theorem 0.2.2 we get

2.2.

Th e o r e m.

I f a measure space (E , I , p) is of the type (a) (b), (c), where (a) (E , I , p) is an inifnite atomless measure space,

(b) (E , I , p) is a finite atomless measure space, 00

(c) E = (J ea m with 0 < inf p(e“) < sup p(e„) < oo,

m = 1 m m

then the following conditions are equivalent : (i) L** cz L**,

(ii) L°* cz L**,

(iii) (р-кф ((р-кф, <р-кф) if (E, Z, p) is of the type (a) (resp. (b), (c)).

R em ark . The above theorem follows from Theorem 10 of [9] in the

case where E is the set of all natural numbers.

(9)

3. We shall present here a criterion for the proper inclusion L** L0ф.

3.1.

Th e o r e m.

The proper inclusion L**

Ç

Ь0ф holds if and only if (a) ф < (p and lim inf ф (cu)/<p(u) = 0 for all c > 0 as и -> oo or и -*0 if (.E, Г, p) is an infinite atomless measure space,

(b) ф < <p and lim inf ф(си)/(р(и) — 0 for all c > 0 if (E, I , p) is a finite

u -►ao

atomless measure space,

00

(c) ф < q> and lim inf ф(си)/(р(и) — 0 for all c > 0 if E = У ea m with

и ->0 m = 1

0 < inf jU (0 ^ sup d(em) < 00 •

m m

P ro o f. Necessity follows from Theorem 1.2, Theorem 2.2 and Theorem 0.2.2. Sufficiency follows from Theorem 1.1 and Theorem 2.1.

4. Now, using the proof of Theorem 2.1, we prove a generalization of Theorem 2.24 of [4].

4.1.

Th e o r e m.

Let ф and (pm be (p-functions for m — 1 ,2 ,... The inclusion 00

Т0ф cz (J i f (Pm holds if and only if there exist a natural number m0 and

m - 1

constants c, d > 0 such that the inequality

( + ) <Pm0 (u) ^ d ф (cm)

holds

(a) for и ^ 0 if (E, I , p) is an infinite atomless measure space, (b) for и ^ u0 ^ 0 if (E, I , p) is a finite atomless measure space,

00

(c) for и ^ U

q

,

u

0 > 0 if E = U eH m with 0 < inf p(ea m) ^ sup p{ea J < со.

m = 1 m m

P ro o f. Sufficiency being obvious, we prove the necessity. Suppose 00

Ь0ф cz (J l?* ”1 and assume that ( + ) does not hold. It means by Definition

m = 1

0.1.2 that for j

=

1, 2, ...: lim inf i

j /

(cu)/(Pj(u) = 0 for all c > 0 as u-> oo or и -> 0 in case (a) (resp. as и —► oo in case (b), as и -> 0 in case (c)).

Then for every pair of natural numbers (i,j) there exists a number uu > 0 such that

<М2НЧ ) < (pj(Uij/i)/i2 in case (a),

»// (2‘+J Uij) < <Pj(ipj/iyi2 and ф(2,+]ии) ^ n2/2 in case (b), ф(2i +j My) < (pj(Uij/i)/i2 and ф(2i+jму) ^ S/(di2 2j),

where d = sup p{e^) and <5 is a number such that 0 < <5 < 1 in case (c).

(10)

for

t e E i j ,

i , j = 1, 2, for fe JE - û Ü Еф

i = i j= i

where sets are as in the proof of Theorem 2.1.

Similarly, as in the proof of Theorem 2.1 we show that х е Ь 0ф.

Now, we shall prove that х ф Ь for к = 1, 2, ... Indeed, let к be a fixed natural number. Let Я > 0 be given and choose a natural number N such that Я > 1/iV. Then in case (a) we have

j (pk \x(t)\)dp = £ ( Z <Pk (Лии) n(Eij))

E

j = 1 « = 1

^ Z (Z Р*(м«У0/*(£«/•))2* f <p*(w.-*/0/*(£,-*)

j= i ;=/v j= v

£ (2

Ф

(2 i + ft Mj*)/(i2 2*

ф (2i+k uik)) ~

£ 1/2* = 00.

1 = ЛГ 1 = ЛГ

Hence хф Ь*^. Similarly we show that x£L*n in cases (b) and (c). Thus condition ( + ) holds.

00

4.2.

Th e o r e m.

Suppose the equality Ь0ф — (J L*<Pm holds. Then there

m = 1

exists a natural number m0 such that Ьоф — L m° and ф satisfies the A2- condition

(a) for all и if (E. I , p) is an infinite atomless measure space, (b) for large и if (E, I , p) is a finite atomless measure space,

CO

(c) for small и if E = (J ea m with 0 < inf p{e™) ^ sup p(ea m,) < oo.

m = 1 m m

P ro o f. By Theorem 4.1, there exists a natural number m0 such that

<Pm0 ^ Ф (<Pm0 <Pm0 -к Ф) in case (a) (resp. (b), (c)). Hence from Theorem 0.2.2 we have Ь0ф с: Ь*ф cz L <Pm°. On the other hand, since L*<Pmо с: Ь0ф we have Ь0ф = L*Vm° and by Theorem 1.2, it follows ф <i (pMQ (resp. ф < q>mQ, ф < (Pm0). Hence ф <i ф {ф < ф, ф < ф) and this means that ф satisfies the d 2-condition for all и (resp. for large u, for small u).

From the above Theorem we obtain:

4.3.

Th e o r e m.

Let (E, I , p) satisfy the assumptions as in previous theorems o f this section. Suppose the equality Ь0ф = (J L*v holds, where T is

<peT

In all cases define

x(t) =

(11)

an arbitrary set of (p-functions and there holds the proper inclusion L*4* Ç L°*

for every (peT.

Then the set T is uncountable.

4.4. T

heorem

. Suppose the equality Ь0ф = (J L*v holds and suppose ф

<peT

does not satisfy the A2-condition of types (a), (b), (c) of Theorem 4.2. Then the set T is uncountable.

5. Finally, we prove a generalization of Theorem 2.21 of [4].

5.1. T

heorem

. Let <p and фт be (p-functions for m = 1, 2, ... The inclusion L*4*

zd

f] L 4>m holds if and only if there exists a natural number m0 and

m = 1

constants c, d > 0 such that the inequality

( + ) (p{u )^d sup(фх{си), ..., фто(си)) holds for и satisfying (a), (b), and (c) from Theorem 4.1.

P ro o f. Sufficiency. Denote фт(и) = supfi/^(м), ..., фт(и)). Then ( + )' means that (p к фто (<P ^ ^ $m0) in case (a) (resp. (b), (c)). Hence from Theorem 0.2.2 we have Ь°Фт cz L*v. Since фто{и) ^ ф1(и) + ... + фто(и), we have f| L°*m cz

L * m ° ,

and hence f)

Е ° Фт

cz

L * <p.

m = 1 oo m = 1

Necessity. Suppose L*<p

zd

f) Ь Фт and assume that ( + ) does not hold.

m= 1

Then there exists a sequence (um) of positive numbers such that:

(p{um) > 2m\J/m{rn2um) in case (a),

(p(um) > 2тфт(т2ит), мт | oo and ^ 1(n1) > 1 in case (b), (p (

m

J > 2m фт (m2 um), um j 0 and фт (m2 um) ^ 0/{2m d), where d = sup p{e^) and Ô is a number such that 0 < <5 < 1 in case (c).

m

Then there exists a sequence (Em) of pairwise disjoint ju-measurable sets in £ such that:

n(Em) = 1/(2 тфт(т2ит)) p(Em) = p (£)/(2m фт (m2 um)) 1

2m фт(т2 um) In all cases define

— d ^ p(Em) ^ 1 2тфт(т2и,

in case (a), in case (b), in case (c).

x(t) = rnum 0

for te E m, m — 1, 2, ..., 00

for t e E — (J Em.

m= 1 We shall show that x e fj Е°Фт and x$ L * <p.

m= 1

(> - Prace Matematyczne 26.2

(12)

Indeed, let n be an arbitrary natural number and let X > 0. Choose a natural number m0 such that m0 ^ max(n, X). Since the sequence фт(и) is non-decreasing for и ^ 0, in the case (a) we have

00

Q^n (Лх) = Z (^mwj/(2m фт (m2 n J)

m = 1

™0~1 сю

< Z ФЛтоит)/(2тфт(т2ит))+ £ фт{т2ит)/(2тфт(т2ит))< со.

m = 1 m = mg

oo Since ф„(и) ^ ^„(«) (n = 1, 2, ...) we have xe f]

и = 1 00

Similarly, we prove that x e

П

L0^" in cases (b) and (c). On the other

n = 1

hand, for an arbitrary X > 0 choose a natural number mj such that X > 1 /m l . Then in case (a) we get

oo 00

Z (Xmum)/(2m фт(т2 um)) ^ £ (p(um)/(2m фт(m2 u j)

m = 1 m= mj

oo

> Z 1 = 00 •

m =

Similarly, in case (b) and (c) we have

qv

х

) = oo for all X > 0. Thus хфЬ**.

Thus we get a contradiction and hence ( + ) holds.

5.2.

Th e o r e m.

Suppose the equality L*v = f] L Vm holds. Then there

m = 1 m 0

exists a natural number m0 such that L*<p = f] Ь Фт and q> satisfies the A2-

m — 1

condition of types (a), (b), (c) of Theorem 4.2.

00

P ro o f. Let L*v

= П

L Vm. By Theorem 5.1, there exists a natural

m = 1

number m0 such that (p ^ фто {(p < фто, (p фто) in (a) (resp. (b), (c)), where Фт0(и) — sup(^i (n), ..., фт0(и)) for и ^ 0. Hence from Theorem 0.2.2 we have L 0 c L*<p. It is seen that L 0 = f] Ь Фт and therefore f] Ь Фт

m = 1 m= 1

— L*v. Since L*v cz L ^m°, from Theorem 1.2 we get фто<кр (фто< <p, Фт0 < Ф) in case (a) (resp. (b), (c)). Hence q> <\ q> (q> < (p, (p < tp) and this means that q> satisfies the A2-condition for all и (resp. for large n, for small w).

The following theorems follow directly from the above theorem.

5.3.

Th e o r e m.

Let (E , I , fi) satisfy the assumptions as in this section and

(13)

let T be a set of (p-functions ф such that: ф1, ..., фте Т implies L*v L°^w.

Suppose the equality L*v = 0 L0^ holds. Then the set T is uncountable.

феТ

5.4.

Th e o r e m.

Suppose the equality L*,p = f) L0* holds and (p does not феТ

satisfy the A2-condition of types (a), (b) (c) of Theorem 4.2. Then the set T is uncountable.

References

[1] M. A. K r a s n o s e l’s k ii and Ya. B. R u tic k ii, Convex functions and Orlicz spaces, Groningen 1961.

[2] R. L e s n ie w ic z , On two equalities for Orlicz spaces, Bull. Acad. Polon. Sci. 27 (1979), 557-560.

[3] W. A. J. L u x e m b u r g , Banach functions spaces, Thesis, Delft 1955.

[4] W. M a tu s z e w s k a , Przestrzenie funkcji cp-calkowalnych I, Prace Mat. 5 (1961), 121-139 (in Polish).

[5] —, Przestrzenie funckji (p-calkowalnych II, ibidem 6 (1961), 149-164 (in Polish).

[6] —, On generalized Orlicz spaces, Bull. Acad. Polon. Sci. 8 (1960), 349-353.

[7] M. N o w a k , On two equalities for Orlicz spaces, Funct. Approx. 10 (1980), 69-81.

[8] —, On two linear topologies on Orlicz spaces L*q>, Comment. Math. 23 (1983), 71-84.

[9] I. V. Sr a g in , Uslovia vlozeni klassov posledovatelnostei i sledstvia iz nih, Mat. Zam. 20 (1976), 681-692 (in Russian).

[10] Ph. T u r p in , Convexités dans les espaces vectoriels topologiques généraux, Dissert. Math.

131 (1976).

[11] M. W isla , Continuity o f the Identity Embedding of Some Orlicz Spaces II, Bull. Acad.

Polon. Sci. 31 (1983), 143-150.

INSTITUTE OF MATHEMATICS, A. MICKIEWICZ UNIVERSITY, Poznan INSTYTUT MATEMATYKI, UNIWERSYTET im. A. MICKIEWICZA, Poznan

Cytaty

Powiązane dokumenty

Moreover, in this case the Mackey topology is semimetrizable and it is a unique semimetrizable locally convex topology between the weak topology o f X and the

MICKIEWICZ UNIVERSITY, Poznan 1NSTYTUT MATEMATYKI, UNIWERSYTET

Let E be a finite-dimensional Euclidean space with the usual Lebcsgue measure... Some equalities for Orlicz

Complex extreme point, complex strict convexity and complex uniform convexity of complex Banach space are natural generalizations of extreme point, strict

ANNALES SOCIETAT1S MATHEMATICAE POLONAE Series I: COMMENTATIONES MATHEMATICAE XXVII (1987) ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO.. Séria I: PRACE MATEMATYCZNE

ROCZNIKI PQLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Séria I: PRACE MATEMATYCZNE XXIV (1984)M. Hence in this case p is

In [3] there are given necessary and sufficient conditions for relative а(Ьф, Lr )-compactness of a subset of an Orlicz space Ьф... We indirectly prove that В is a weakly

The norm generated by the function 99 is defined in the third chapter in which also one concerns the mutual relations of various kinds of convergence of the