ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Séria I: PRACE MATEMATYCZNE XXV (1985)
M a g d a len a J a r o sz e w sk a (Poznan)
On interpolation inequalities in the space H™,P(Q ) with mixed norm
1. In this paper we prove some interpolation inequalities for functions / from the space Щ р(й) with mixed norm. At first, we shall give some lemmas for the functions from H m,p(Q). The results contain some inequalities from [4] for n = 1 and also from [3], [10] for = 0.
2. The index i runs through 1, n unless otherwise stated. Det R be the set of real numbers and /с, > 0 an integer, 1 < p, < go , Я; ^ 0. In the following we shall use vector notations, i.e., x = ( x lf x„), p = (pl5 . .. , p„) etc. In this paper we assume that Q, is a cube, because the interpolation theorem, proved for a cube, can be proven also for open, bounded Qt having the restricted cone property^). We can observe this fact for mixed norm studying, for instance, the proof of Theorem 1, [6]. So let Qt denote an open,
k- _
bounded cube, subset of the real Euclidean space R '. Let Q = P Üh Q
i = 1
n " к П
= P Üh R n = P R k\ У ki — N. The measure means always Lebesgue
«=1 i= l « = 1
measure. To simplify the notation we shall write for example:
$ \ f(x )\ d x = J ... J |/(x)|dx! . . . d x n,
Q Qn Q i
J \ f W d x = l l / C (ei = S [■■■ S (S
bQ Qn Q i Q l
J I \f\’ d x = j [ . . . ( ) 1 \ М х Г Ч х 1р 1рЧ х 2 . . У ’'1’’- Ч Х ' .
bQ i = 1 Qn Q i i=1
1 P tJP n - 1
Let
HI = I HI. HI = I l D 'f(x)
i = 1
S = 1a1'1/) * )
a«i,}... гх1%
i1) For definitions of the cone property and the restricted cone property see among others
Ш, [6].
if |/| > 0 , Dlf (x) = / (x ) if / = 0. Let & m denote the set of polynomials P (x) of degree ^ m.
We shall denote by C m(Q) the space of functions defined in Q which have continuous partial derivatives of order up to m ^ 0.
We shall denote by H m p(Q), p ^ 1, m integer ^ 1 the space of functions which is the completion of C °°(0) with respect to the norm
( 1 ) ! X V r / C J 11"--
П We denote by f Q the mean integral value of the function / on Q = P
i= 1
/ e = p ц - ' т i f(x)dx.
i ~
1Q
We denote by L p£ (Q), x; ^ 0, the subspace of U (Q ) of functions / for which
(2) Ш/НкР-^я, = SUP f [ 1 ц т Г 1‘Р""“Р‘ inf I \ f-P \ pd x }llp" < CO,
Qi 1 Pz& m b Q
where Q( is the subcube of Ц .
The space L P,*(Q) is a Banach space under the norm (3) II/II l £ a(«) = II/II lp №) + III/III l £a№)-
This space was studied in [2], [6], [7] and else. We know that
R e m a r k 1. The space L $ X(Q) is isomorphic to M P,X(Q) if 0 ^ Я,- < k t and to C 0,a(Q) if < Я , k, -f pf, af = (Я, — /c^/p,-, for definition of M P,X(Q) and C°'a(Q) see [9].
We shall denote by Щ ,р(й) the space of functions which is the completion of С °°(0) with respect to the norm
(5) \\Л\н™’р(а> = \\f\\H m’P{ + I lll^ / IIL g -W To simplify the notation we set
\jyf\ LP(Q) { I ll£>7ïl
i*i =j
Pn LP(Q)
VPn
W f l L M « ) = I W f \ \ L P A 0 V
In this paper we use the definition of the space with mixed norm from [2].
The notations and some parts of the proofs are analogous as in [3 ] ; for the sake of completeness we outline these results here.
3. We know from (17) [6 ]: If f e H 1 ,p{Q), then
< j \f\” d x } ' " ’" « 0 { J X I D ' j f l pd x } ‘ ,p’ , j = 1...
bQ '
( 6 ) n,
where r-} is the length of edge of Qjt j — 1, n, and hence j f |/|'’d x}1,p” « c 1(p)cJ. £ ! J | D » / | •’ d x }'1"".
bQ M =1 bQ
Multiplying the above inequalities for j = 1, n side by side and applying for f - f Q we get
( 7 ) ! J \ f - f Qi " d x } llPn < c(n, p ) П г/'" I ! j \D' f\’ d x } ' " ’\
bQ *'=i l«l = i bQ
L emma 1. Let Q{ be a cube o f the edge rt and let / е Я ш,р( 0 ; then fo r every system o f numbers [aa}|a| = m there exists a polynomial T (x)e # (x ) such that
(8) j \ f - 3 r ( x ) - U - ^ m Q \ P d x ^ c 2(n,p) П r", J ’ £ f m - a l i ’ dx.
bQ j = 1 |s| = m bQ
P ro o f. Lemma 1 holds for m = 1. Fixing N of numbers a lt . . . , a N, it is sufficient to assume
(9) ^ { x ) = Y , a i x i .
i= 1 From (8), really, we have
J \ f - ^ ( x ) - U ~ . r ( x n Q\ " d x ^ c 2 (n, p ) П A Jn L I J I D J - a i ’ dx.
bQ J=1 1=1 bQ
If m > 1, then the thesis follows by induction. Let us assume that the lemma holds for f e H m~i,p(Q), fixed numbers [aa}|a| = m. Then there exist N of polynomials ^~i (x )e i such that
(10) £ J ID J - F M - l D J - r ^ d x
« = i bQ
« с 2(и,р) n L f
j = 1 |s| = m bQ
N xi
For ,T (x ) = £ [ J ^ i ( x u . . . , t, xf + 1, . .. , x J d t + lD J - J T fx ^ Q - X i]
i= 1 0 by (7) we have
(11) f | / - ^ ( х ) - [ / - ^ ( х ) ] в | 'Л с
f \ D , U - ^ m pdx.
J=1 i=i bQ
From (10) and (11) we get (8).
R e m a r k 2. We observe, from the above, that ^~(x) is of the form
^ ~ (x )= £ as xs + R (x), R ( x ) e ^ m- 1.
|s| = m
T heorem 1. I f Qi is a cube o f the edge rh then fo r every f e H 2,p{Q) (12) i IIA / -(A / W l1P(ei ^ c3 \D2f\]ym ||/-/e ||{&,,
i = 1
where c3 does not depend on r{.
P ro o f. Let v e H 2,P(Q); then we know from Theorem 1, [6],
l ° ‘ v\m Q ) « { l » 2 < 4 > l l < 4 , + £ " 1 IW I
u® } :
c4 does not depend on r; .
Studying the proof of the generalization of Theorem 3.2, [1], for the scalar p and studying the proof of Theorem 1, [6], we write
а з ) in 1 »\Lm> « <5 n < 4 1+ '-~ ‘ 'LP(Q)
where r = m axri5 i = 1, . .. , n.
Let = Y, aiXi + a о be a polynomial such that i = 1
(14) J If - P l \p d x = inf J \ f-P \ pdx.
bQ Pe* i bQ
Let us write (13) for v = / —P i, taking into account that r ■ЧП ri lln
i = 1
( I 5)
I m - a A \ ^ c 6 № 2/ ii/ 4 ) n / - p 1iii/ 4 ) + n '■r1/- w f - P A r J -
i « i =1 i=i
Applying Lemma 1 and Remark 2 we get
П n 11" р)\°гл]!г1 LP(Q) ‘
(16)
J= 1 By 1.10, [3], we know
WV ~ VÀ\
lP(Q)
< 2in f lle - c ll^ e ,, taf ll»--P|l1, (e) « inf ll»-c|lu,B) « ll«’- ‘’o l U e,- From (15), (16) and the above two inequalities there holds (12).
T heorem 2. L et Q{ be a cube o f the edge r(; then fo r every f e H 2,p(Q) we
have
(17) £ inf | 1 А '/ -Л и е ) « с 7 £ inf|U>,Dj/-cH$ inf Н/-Л1[йв)-
i = l P e & i i,j- 1
where c7 does not depend on r,.
P ro o f. Let P 2 be a polynomial from such that I I / - E l l i n a = “ rf HZ-f’ lU g , and P 2 (x) = £ üij xf Xj + R (x), R (x) e ^ .
»j= i
By (13) for v = f —P l , we conclude (18) X iia / - d , p 2|
i= 1 LP(Q)
r f ‘'”1 1 /- P A w + {( I IIB. D jf -null L PQ )m I I /- A ll^ g ,} '
i = 1 i j = 1
Let be a complex number such that
Il A D J - b,i\ = inf II D, D j f - c|| i . i = 1... ».
ceC
We know by Lemma 1 that there exists a polynomial ^ " ( x ) e ^ 2 W such that (19) \ \ f - ^ ^ K m < c 9 W f ln E I I A V - b y l
f = l i,j= 1 LP(Q)-
By Remark 1 , 3 T { X ) = bij X,- Xj + R (x), R (x) e ^ . Applying Lemma 1, [7],
»'j=i
to the polynomial P 2 {x) — ^ { x ) , we get for every i, j
(20) f = l П r r VP' Ifly-iyl « c 10 П '•rl',l| I A -^ W I I lP(01. t= 1 t- 1 i HI = 2.
By (19), (20) and from the fact that ]^[ r,-2/n) ij ^ 1, we have
i = 1
( 21 )
Let us write
N
п ^ к{1р' к - ь и1 ^ с п x i i A z y - а д
t=i •J=l LP(Q) •
(22) { £ I I A V - a y l l ^ e . ) U=1
< c 12 {( I m D J - h j W ^ J 11+ ( I Iny-byl П
ij — 1 ij= l < — 1
By (21) and (22) we get
(23) ! I \ \ D , D j f - a , j \ \ L P W ) } 112 i>j= 1
< Ci3( Z IID iD jf-b ,
i j = 1
I - Г '
'LP(Q)f
Moreover, we have
(24) П T ш r - ^2llt p(CI < П ' Г I I / - p
i= 1 i = 1 Ii& )
Inequalities (19) and (24) yield
(25) П ' T 1," l l / - P 2 l l l p(ai < c 9 | | /-P
i= 1 11,2 ( I I I A V - ^ L o , )
1/2
'rP(Q) U=i From (18), (22), (25) we get (17).
Following Campanato we rewrite from [3] the next two lemmas:
L emma 2. I f B 0, B lf B 2 and a are non-negative numbers, then the next inequalities are equivalent:
(26) B 1 ^ c l0 (eB2 + £ aB 0 + B 0), £ > 0 , (27) B x ^ с 11(вВ2 + Е~а B 0), 0 < e < l , (28) B x ^ c 12( B f 1+a)B j /(1+a> + B 0),
where c 10, c n , c 12 are positive constants which do not depend on B 0, B\, B 2, £.
L emma 3. I f B 0, B x, B m(m > 1) are non-negative numbers and fo r every j, 1 < j ^ m — 1 satisfy the inequality
(29) B j ^ c ^ d U s B j ^ + e - ' B j ^ ) , 0 < e < 1, then there exists a constant c lA( m ,j) such that
(30) Bj < c i4 (m, j)(£m~j B m + £~j B 0), 0 < e < 1.
T heorem 3. F or every / e H f ,p((2) there exists a constant c(n, p) such that
(31) I I I I A / l l l i f . W M x ll|D,VIII*.fî1«i.),,2|ll/llli^<ffl-
I = 1 i,j = 1
P r o o f. For every cube Qt <= we write inequality (17). Dividing both
sides of this inequality by ]~J Lu (Qd] A,/k'P' we get i = 1
X 1 № , ) У ^ кт inf \\DJ-P\\m o .
i= 1 Pei?!
« C l ! x Ы < Ш ~ 4 " ‘m inf II Д 0 / -с | | ^ в)! ‘ '2 X
i , J = 1 C6C
x l l î [ р ( Ш Г 2,'"“'’' inf Н/-Л11Л0)) 1/2-
i = 1 Pe;3»2
Taking into account (2) we get (31).
T heorem 4. F or every function f e H j,p(Q) n L$,P(Q) the follow ing inequali
ty holds:
(32) £ III A / lllif« ,« « c7 (n, p) ( £ III A Djf\\\L^ ( a ) 1/2
i= 1 i.j= 1
1
/2
.L% Ш)’
where & = ( i i + Jui)/2.
П
P ro o f. Analogously as before we divide (17) by Yl [^(6/)] (2, + k,)/2fc,p‘
i — 1 and by (2), we get (32).
T heorem 5. F or every function f e H f,p(Q) the following inequality holds:
(33) X HA/ILg^m, «
1= 1
P ro o f. We get (33) repeating the proof of Theorem 1, applying (12) instead of (17) and taking into account (3) and (4).
T
h e o r e m6 . F or every function / е Я ? , р ( 0 ) , 0 ^ л , < /с; + р ь there exists a constant c 15(n, p) such that
(34) \ D l f \ L P 0>\n) ^ c l5 (n, P ) {\ D 2f \ l^ \ a ) II/ llLg’A№) + l l / l l L g ’W -
P ro o f. We take the thesis of Theorem 1, [6], for the function f - P , then
Л