A N N A L E S S O C I E T A T I S M A T H E M A T I C A E P O L O N A E S e r i e s I : C O M M E N T A T I O N E S M A T H E M A T I C A E X V I I ( 1 9 7 3 )
R O C Z N I K I P O L S K I E G O T O W A R Z Y S T W A M A T E M A T Y C Z N E G O S é r i a I : P R A C E M A T E M A T Y C Z N E X V I I ( 1 9 7 3 )
M agdalena J aroszewska (Poznan)
A generalization oî the Riesz-Thorin Theorem
1. Introduction, notation. The aim of this paper is to prove a gener
alization of the Riesz-Thorin Theorem on interpolation of linear operators {for the Riesz-Thorin Theorem see [7]), where T is an operator from LP(Q) to L (q,X)(û) (both spaces with mixed norm). We also formulate here a theorem which gives an application of this generalization. The results contain those of Campanato and Murthy [4], and Benedek and Panzone [
1].
The index i =
1, n everywhere, unless otherwise stated. Let R be the set of real numbers, m{ >
0, integer,
1< pt- < со, 1 < < oo, p 1 > ... ^ p n, q 1 > ... > qn, 0 < < mi Jr p i . In the following we shall apply vector notations, i.e. p = (px, ..., p n), x = (x1} ..., xn) etc. Let Q{ be open, connected, bounded subsets of the real Euclidean space Rmi. By d(Qi) = $ we denote the diameter of Q{ and we write Ü = Pi=i &i, & = P rl=lQi . Let I (Xe-, q {) c= Rmi be the ball with centre at x°te Q{ and radius q ,-> 0 and let Sf = T(x\, gt) д(х0, g) = P?=
1I(®?, &'), S = I (x0, g)nQ.
We suppose that the boundary дй{ of is of Lebesgue measure zero in Rmi and that satisfy condition (A) (see [5]). To simplify the notations, we shall write, for example:
\f(x)\dx = J . .. J I f(x)\dx1 . . . d x n,
/ \f(x)\»dx = H/|ÇÎ(S) = / 1 ... ( / I/ ( *)[”' .
Ъ $> Ьп
In this paper we use the definitions of L iP,X)(Q) and M^P,X)(Q) spaces from [5] and [
6]. The definitions are slightly modified so that numbers Аг- related e.g. to M (P,X)(Q) should be substituted, in the definition and elsewhere, by numbers Яг-— P Again, if we refer to results of those papers,
Рг p n
we accept this modification. Then e.g. condition A* > —- mi should be substituted by Яг- > etc.
/
2. Definition. We shall denote by L^P,^(Q) the space of functions / locally integrable in Q, for which there exists a positive constant M depending on f such that for every x0, q with x\ eQ{) 0 < q { < there holds the inequality
( 1 ) Then
inf f Jf ( x ) - c \ pd x < M ( f ) П f= i
( 2 )
П
\\\f\\\L(p>*)(0) = sup { / 7 QihPn,n inf / \ f{x)-c\ pdccy/Pn = in f[iif(/)]1/J?«
*=1 ceRbS
0 <6{<еЧ
is a seminorm in l t p,^(Q). It is easy to show the fact: if / e L iP,^(Q)f then for every x 0 , q with e , 0 < q { < q *- there exists only one constant о depending on x0, g , f such that
(3) inf j \f { x ) - c \ pdx = J \ f ( x ) - c { x 0, Q,f,p)\pdx.
ctRbs ь«
The existence of the only one constant c follows from the uniform con
vexity of the space I^(S). It follows from (2) and (3) that the seminorm in L M (D) can be expressed as:
(4) lll/llli(P,V) = su? { f j e î hPnlPi f I f { x ) - c { x Q, e J , p ) \ pd x y IPn.
x V e â i i=
1
bSo
The space L (P,*](Q) is a normed space with three equivalent norms (6), (6) and (7):
(б) ||/|1ьМ,0, = { ll / f ô ( 0) + lll/lll2h>.4(fi)}lft’",
( 6 ) = Шта) + ШЯ 1 Ы ы,
(7) п
= \]f\\mQ) + sup { f ] Qï*ipnlpi J \f(x) - c { x ü, Q J , 2)1 pdx}llPn.
xi eûi i==1 0< Q i«ï\
The equivalence of norms (5) and (6) is obvious, but the equivalence of norms (6) and (7) we shall prove in Lemma 1, separately. Let %{x0, q ) be the characteristic function of the set S. Then we observe that X'C(x0, Q,f, 2) is the projection of %- f , f e L 2 (Q) on the line {%•<?: ceR}
c L 2 (Q) (where L 2 (Q) denotes the space LP(Q) with p = (p 19 . . . , p n)t р { = 2 ) and then, among others the correspondence f X'G(xoi
is linear and continuous from L 2 (Q) to {%'C\ ce R}.
Biesz-Thorin theorem 89
L emma 1. I f e L {p,^{Q), 2 < < oo and condition (A) is satisfied, then
(8) ll/llz,(p>A)(fl) II/II i ,( p * a )(/3) ^ cill/z,iP>A)(«)ll •
P roof. It is sufficient to prove the right-hand side of inequalities (
8), because the left-hand side one is obvious. Let us fix x\e û i , 0 < < q
Let f e L iP^{Q), pi ^ 2 . It is easy to see from the Holder’s inequality that L (P,X){Q) <=. L(
2,A)(i3). It is clear that
И^о, Q’f ’ P ) - c (xo? 2 )12
< 2 2|c(a?0, Q , f , p ) - f ( s o ) \ 2 -\-2*\c(x0, 6 , f , 2 )- f{x) \K Integrating both sides of this inequality with respect to x{ on St, succes
sively, for i — 1 , n, applying condition (A) we get (9) 1 ф 0 , в , / , р ) - с ( ! 1 ; 1 „ в , / , 2 ) \ ^ 2 [ ] к Г 112 еТт* 12 У~
г'= 1
x{( J \c(x0, e , f , p ) - f ( x ) \ 2dx)ill2 + ( J \ f { x ) - C( x 0, Q,f,
2)|
2<L?)1/2}
bs . b$
П
2 / 7 K 112 Qimi'2 {( f \o{æ0, Q , f , p ) - f ( x ) \ 2d x y 2+ .
*=i
+ ( / \f(oo)-c{xQ, e , / , p ) |
2<&r)1/2}
ft®
22f j K m Qrmil2{ j \Фо, Qj,p)-f(v)\Pdæ}llPn-fl[p(Si)fpi-2^
i —l bS i = l
n
C z f l
Next, it is also clear, that
\ f ( x ) - c ( x o, Q, f, 2)\Pl < 2pi \ f ( x ) - c ( x 0, Q, f , p ) l Pl +
+ 2р1 |с(я?о, Q, f , P ) - c ( x o , e , f , 2 ) Г . Integrating both sides of this inequality with respect to x{ on suc
cessively, for i =
1, ..., n and rising to the suitable powers, we obtain (10) { / \ f ( x ) - c { x 0, g , f , 2 ) \ pd x y IPn
bs
< 2 { f l f ( x ) - c ( x 0, f , p ) \ pdx}llPn +
b$ П
+ 2 \c(x0, Q , f , p ) - c { x 0, Q, f, 2 ) \ [ J [ p { S i ) f lpi.
From (9) and (10) we get
{ / 1 f ( x ) - c ( x 0, e J , W d x } llp« ^ C s f l & ' Vil \ m LiP’*Ha>,
foS i=
1hence
n
sup { П QÏhPnlPi / If ( x ) ~ G(xo, Q,f, 2)1 pdxy,Pn < C 3 \\\f\\\L(p,x){Q)
i= 1 bS
о<в{<е\
and hence the thesis follows. _
We shall introduce the product space Ca(0) and we shall prove the isomorphism of L^P,X)(Q) and Ca{Q) under some assumptions, applying the method of Campanato [3].
D efinition . We shall denote by C(Q) the space of continuous func
tions / in Q.
It is a Banach space with the norm ll/llc(5j =* sup //(я?) I.
xeSi
D efinition . We shall denote by Ca(Q) the space of functions / which, for every , yi e Qi , satisfy the Holder’s condition in Q with the exponent a = K , ..., a j
П
\ f ( v) - f ( y) \ < K [ ] \xi - V i \ ai,
0< аг <
1. г=1
The space Ca{Q) is a Banach space with the norm
(
11
)The term
ll/llo»(5) sup \f{%)\ + sup xeü xi>VieSii
xi*Vi
\fix) - f { y ) \ n
П \xi-Vi\ai
is
11 l/l I lca(fi) sup xi,Vi‘Qi
\f(x ) - f ( y ) i n
П \xi-Vi\ai
г= 1
a seminorm in Ca{Q). It follows from Theorem 2 [5], that
_ } , __ /kb) ,
L (P,X){Ü) c= Ca{Q), where щ = —--- -.
Pi
T heorem 1. If condition (A) is satisfied, mt < A* < Щ + Ри
■at = (Аг- —т г-)/рг-, then spaces L(V,X)(Q) and Ca(Q) are isomorphic.
Biesz-Thorin theorem 91
Proof, (a) Let х г , у { е й { , let f e l S p,^( Q) . It is easily to find from (23) [5] that
sup 11 l/l 11ьОМ)(я) •
xi>Viepi /7 \x. — y.\ai 1 1 i=l
We have from (16) [
6]
П
N®, eJ) < см\и»мт -[J ep-m‘)№ <
г=
1n и гг-
^5ll/lli(P>A)(fl) j f j &i Pi ^ ll/llz,(P>A)(0)-*
Applying Lemma 3 [5] and Theorem 1 [5] we get Sup|/(a?)| Cfe||/Hi (p,A)(lî) ctiid W
6deduce next
ZcQ
\ ~f ( qq \ _f {'ll} I
snp I/(x) I + sup_ —---< C 7 (||/||х(р,А)(Я) + M l/l I\ l № \ q )) xeQ х%уф U \ xi - y i \ ai
an<i 11/11^“ (^) < 11/11ь(г»>А)(Я) •
(b) Let fe Ca( ü ), х{, у{ e $ t- c It is clear that П
{inf j \ f ( x ) - c \ pdx y iPn < J
11l/l I lca(«) • ] 7 l ^ -
2/ilai
СеЛЬ«
i = 1n n
< c.111Л ! /7 f>? / to < c10 ■ • 111/II ic.(Bj • /7 е.-л
» = l 51S i = l
hence
I I l/l I li(P»A)(fi) ^ ^10 I I l/l I ^ “(Й) and
||/|1х(г>.А)(0) ^ ^п11/11са(й)- Prom (a) and (b) the thesis follows.
3. We shall denote by I a linear operation defined for all simple functions / on X — Pi= 1 Rli into the class of measurable functions on Г = Р » =1ЕЧ Let 8 = Р г?=1#г- = Р£=
1[
1(жг-, д{) п й {] and let (7) denote the norm in the space L(P,X)(Q).
T heorem
2. Let the linear operation T be continuous simultaneously
from L p 1 (Q) to l } ,l>’V){Q), where f = (p[, = («}, •••? £«)•
V = (А{, A^), j = 1 , 2 , such that for every fe L p7 (O) there holds the ine
quality
hen for every 0 < t < 1, T is a linear operation from LP(Q) to where p , q, A are defined in the following way :
1 l - t t
— H--- 2 » % = 1,
Pi p\ Pi
1 l —t t
—1 H--- if. г = 1,
9% 9i 9i
Af l - t t „
A + — AJ, i = 1,
«< 9 a 9i
and there holds the inequality (13)
Proof. We introduce the notations
Pj
=(Pi, --- j PL)
=P
=(Pi, --',Pm)
=' 1 i \
9j = (9i, ■ / 1
" V r
J 5
ii
' i i \ II ч / 1
■ , • ■- , — , ~ ( л ,
, «1 /
i =
1,
2. We define the functions:
at (z) = ( l —z)ai+za*j i = l , . . . , m , Pi(%) = ( 1 - * ) $ + * $ , i = 1, .. . , w,
? 4 {t)pi{t) =
( 1—t)A}# + zA?/Sj, i = 1 , z = Ç + iy.
We have, in particular : a^tf) = аг-, /1г-(/) = /Зг-, а*(0) = aj, /?г-(0) = $•
аг(1) = < $ , Ш = fi , АЛО)А(О) = А}Д, A, (l)^(l) = AJ/?J.
Let ns fix 0 < t < 1. At first, we suppose that qi > 1. Let «я/c ,Q,
• j / — measurable subset. We denote by the vector space of simple
functions on s/, which are finite, linear combinations of characteristic
functions of measurable subsets of We know from [1] that SF{sé)
is dense in Lp{s4) for p { > 1. It is sufficient to prove (13) for / е ^ ( ^ )
Let ns take z in the strip 0 < £ < 1. Let f e ^ r(Q), g € ^( S) and \\f\\jj>ia) = 1*
Biesz-Thorin theorem 93
ilflUii-fys) — 1* We define, as in [1], the functions:
Vi = Vi(xi+ 17 •••:> Я'то)
...7® to )II£(^ .... ÿ (pi
<pj = v A V i + i , ••• , У п )
/ i ! » l - f y + i(*) 1-^(в)
ii i— i FT T| ■,Iz^/(p|=1sfc)] 1-^+l 1-^'
1 < г < га — 1, 1 < j < n — 1,
ai+i(*) ai(z)
1 ° i+ l ai ,
“i W
Fz{xi, . . . , x m) = |/| “i signf - f j ip{,
i — 1 n — 1
1 - W
GziVi, - = \g\ -signg f j v j .
We see that for z = i!: Ft = / , ^ = g and we have (see [
1])
= K ,!l (
4-....,
ато(°))(Д)
1n j nLV(0)|Я|ато(°)/ато =
1’
= ii ? 1+,-, ii ( * ....
1\
“ to O)/^) = =
1and ||Gy _L_ = 1 , ll^l+ij - i - =
1.
l ' - p H s ) Let ns consider the function
»(*,
0,
2) = / / f T F , - G , d y + f [ I F , - c ( x , e, TFS, 2)]G,dy.
i = l S S
Let ns fix x , q . The function h(x, q , z ) satisfies, in the strip
0< £ <
1,
* = £ + assumptions of the three-lines theorem (see [7], p. 93). If /€#"(£), TFZ is of the form
S т о -1
T F Z - V [ J é ÿ ^ I i x , ) , 7 = 1 »'=1
where ci;- are constants ^
0, li} are linear expressions in z, Xj are character
istic functions of some parallelepipeds. Simultaneously
S
то—1Ф , e , t f „
2) = у f f 4 №-с(х, e , t x„
2),
7 = 1 i = l
7
= 1 г=1
Substituting this in h(x, g, z ), we see that A (a?, g, г) is a finite linear
combination, with constant coefficients, of exponentials az with a >
0.
Then we apply the Holder’s inequality to the function h(x, q , z ), when
0