Seria I: PRACE MATEMATYCZNE X I (1968) ANNALES SOCIETATIS MATHEMATICAE POLONAE Series I: COMMENTATIONES MATHEMATICAE X I (1968)
Z. St y p i ń s k i (Poznań)
On Orlicz spaces with mixed norms (I)
Inequalities for trigonometric polynomials *
1. In the paper [4] there are given inequalities for the norm of an integral function and a trigonometric polynomial in the space IP. The purpose of the present paper is to prove analogus inequalities for trig
onometric polynomials of n variables in OrUcz spaces.
Let us denote by у (u) a convex, non-decreasing function defined and continuous for и > 0, <p(0) = 0, satisfying the following conditions:
(1) lim
«—>•0 (p(u)
и 0, lim <p{u) = o o , (p{u) > 0 for и > 0 (2) cp(uv) < c<p{u) (p (-»), e > 1, 0, v ^ 0 (condition A').
Let f ( x x, . . . , xn) be a measurable function with period 2tu in each variable. Applying the notation of [1] we write
27T
(a) ll/f e , •• • j i ? 1 x k+1? • ■• • ) ^ra)||ę> = / l/(®i, •••, xn)\dxk 0
if (p{u) = и
(b) • ? ffijc—1 J * )i ? • •• ) **'w)||ę> := sup ess 1 f { x 1, ..., xn)\
0<a5fc<27i:
if p('a) — °°
(e) № u •• •, %k— 1 1 1 ffik+ij •• * ?
C / l/Caq, ...,a?n)| \ d ,, s > 0: rp I ---Мхи < 1> otherwise.
o ' £
A function f ( x x, ..., xn) is said to belong to the space L*QM){n) if the following proceeding gives a finite value in each step. In the first step we calculate
* The results were presented April 14th 1966 in a Seminar of Dr. J. Musielak, Department of Mathematics, A. Mickiewicz University in Poznań.
2 4 8 Z. S t y p i ń s k i
||/(*, x 2, .. . , %n)\\<p by means of the formulae (a), (b) or (c) depending on the definition of cp. Here, <0, 2ти> (j ~ 2 , 3 , n) are parameters.
In the second step we calculate j|/(*, •, x s, ..., xJIJ^ = ||||/(-, x 2, . . . , x n) |]„||^
again by means of one of the formulae (a), (b), (c). After n — 1 steps we obtain ||/(«, and in the nt~h step we calculate ||/(*, ..., •)!!,,
= ||ll/(*,---, • applying (a), (b) or (c). The number ||/( • , ..., -)IU obtained in this manner is denoted by Ц/Ц,, and is called a mixed norm:
(3) ll/ll, = l l / ( - ) l l , .
Thus, L*ot2n>(n) is the Orlicz space of functions / such that ||/||ф < o©.
Moreover, we denote by T the space of finite sequences, i.e. sequences containing a finite number of elements different from zero only. The norm of the element a = {ak} e l is defined as
(4) \\a\\Tip = in f j £ > 0: J ? <p < l j .
Similarly as in case of a function one may define a mixed norm of a sequence.
The following Jensen’s inequality will be of importance in our con
siderations :
(6)
f f ( x ) d x \ j(p{f{x))dx --- 1 < - A---
mesA / mesA
We shall denote by T Vl...v (xlt ..., xn) a trigonometric polynomial of degrees vx, ..., vn with respect to variables x x, ..., xn, respectively.
We have
„ »■ ' n
1 П I 2 k SXS
(®) ^-Vx,...,Vn {XX ! •••! ®») = ^ --- Gkl---kn (i kl = ~ vl kn = ~ vn
Theorem 1.1 (a generalization of Bernstein’s inequality). Let Tn(x) be a trigonometric polynomial of degree n of one variable. There holds the following inequality:
(7) W T X ^ n W T X -
Proof. For the sake of completeness we give a simple proof of this inequality which is known in an even more general case (see [2]). By [6], Theorem 3.16, we have
2tt , 2tc
f V ^ ^ f <p(\Tn{x)\)dx.
o ' 0
Let us choose a number e > 0 such that
Then
Hence
\Tn(x)\
dx < 1.
dx.
Corollary. Let T v (хг, . . . , xn) be a trigonometric polynomial of degrees v1} ..., vn with respect to variables x x, .. . , xn. Then —dTv is
dxk a trigonometric polynomial of degree vk with respect to the variable xk and
(8) dT*h <vn
dxk < vk IItVi
<p •vn19»*
2. Theorem 2.1. Let T v{x) be a trigonometric polynomial of degree v1 and let us write h — 2n/v, xk = Ш (h = 0,1, ...), xk < xk < xk-\-rh, where r is a non-negative integer. Then
о) <р-лтт£т\ = ’Р-лттл^'Лт, « с а + г ^ м ,,
where <p_x(h) is the inverse of the function cp{h).
Proof. Let us write
In
Q{Ty) =
J
<p(\Tv(x)\)dx.Then
xk+1
e{T.) = £ j <p{\T.m)dx = f?< p к як
\<P-i(h)Tv(h)\
Hence
Thus
for
\<P-iWTv(h)\
Cs
< р ~ л т щ \! * < с \\т х
260 Z. S t y p i ń s k i
Next, applying triangle inequality for norms, substituting t — ruĄ-xk — rxk
— v and applying (2) and (5) we observe that
x k + r h
0 0 \\ J
11 x k
dt
xk+1
/
[Tvm<p-x(h)r\du xk
inf £ > 0
xk+1
/ (1Д0 / \iTv{v)]'<p_l {h)r\du 2 ' — ---;---
2к
J — ;—
X k r ^ Jq— e—
Г -But
J" <p j (\[Tv{v)]'\rh ■IL '— I d u <
\ s
2n
/ * (
\T'v{t)rh\
dt.
Hence
<P
_1W I
m \ł„ - \\T,\\f*I<
infje > 0: j
<pj1--**— )<H <
l |<
Ч.Г.Г||T„||,.
Finally,
V - l W и г р \\x k ^ У - Л ^ 1 ) I n ^ 7 иX k u r n n l f e l , | | | fc
i » ; <
< (1 +2iTir)\\Tv\\ę .
Corollary. Theorem 2.1 remaim true if we replace the trigonometric polynomial T v by an arbitrary function f continuous in <0, 2tc).
Indeed, trigonometric polynomials are dense in the space L*$My.
Hence there exists a sequence of trigonometric polynomials {Tn(x)} such that \\Тп\\,р -> ||/||v. Moreover, we may suppose that \\Tn{xk)\\T<p -> H /M Iliy Consequently,
— — 1|/<®*)||г, = lim * = M - Ш х к)\\г < lim(l +2*r)||TJ|,.
C ■ę rtr^oo c * n-+oo
Th e o r e m 2 . 2 . Let Tv%....Vn(xx, . . . y xn) . be a trigonometric polynomial of degrees rx, . . . , vn with respect to variables aq, ..., xn. Let us write
2n
hk = — (7c = О, 1, ..., n), %ik )= lh k (I = 0 , 1 , . . . ) , Vk
x {k) < x{k) < x\k)+rhk, where r is an integer. The following inequality holds:
(10)
П
max
П
<P~i(hk)\\... \\TVlk—1
,-(n)
<P
< [c{l+2nr)]n\\TVl....,J , . Proof. The variables и being fixed, the trigonometric polynomial Tv ....(aq — ux, .. . , xn — is a polynomial of degrees vx, ..., vn satisfying the equality
\\TVl,...,Vn{Xx Ux, . . . , Xn Un)\\Ty \\TVl....vn (®n • • • j МпШТу'
Hence it is sufficient to prove the inequality
■ n . - (1) ~ ( n )
П г - Л Щ - ц т . ,.... .„Igi . . . ] | > < [ с ( 1 + 2 т г г )Г ц г ,1.... „j i„.
fc=i
The proof will be performed by induction with respect to the number of variables. If n = 1, the inequality is already proved. Let us fix the variables xk+2, ..., xn and let us suppose the inequality is true for n = 1c.
Applying the previous Theorem and the Corollary we obtain [G (l+ 2ur)f\\TVl_ Vn( - , . . . , - , x k+1,...)\\4>
^ 1 1 || • • • II ^ (* ? • • • > * ? ^k+1 j • • -AO ~(k I >, Hence
[^ (1 A )] ' 11 11-^ vb ...,vn ( * 1 • • • ) * ? & k+ 1 J • • • llę>||ę>
....
Theorem 2.3. Let Tt>b >Vn(x1, . . . , x n) be a trigonometric polynomial.
Then the following inequality holds:
4 ) ... . j , « <-nn y ...
252 Z. S f c y p iń s k i
P ro o f by induction. If n — 1 we have 2л
/ cp(\Tn {xx)\)dx1 <
o ^ i 2rbxf ( \ T .l ( ^ ~ u 1)\).
h=i Hence
Thus
j
у n T ri(xx)cp_x{lf2r}i1)\o ' C ‘ *i« i
\\TV1\
< G p y gd1)Let us remark that the above inequality remains valid for every function continuous in <0,2лг>.
Now, let us suppose the theorem is true for n — 1c. If we fix the variables xk+2, .. . , xn we obtain
R em ark 1. If x[k) — xjk) (lk = 0 ,1 , ...), then the coefficient 2r at the right-hand side of inequality (11) may be replaced by 1. It is
M 2 rhi
sufficient to take in the proof integrals f in place of f .
о 0
Е е mar к 2. Writing
=<!)
...))* = сПП ™ H - ||r ”.... ' • • nr;
if x f ] = (l = 0 ,1 , ...), x\kk] = x\kk] (lk = 0 ,1 , ...), we have
n
112;,...., J , < «Г .,....< е2п(1 + 2к)п ... ..II,•
Theorem 2.4. I f <p{u) < y>(u) for и < we have
( 1 2 ) 1 И 1 т „ < I N l i y
n
(13) \\T'n ....J i , < C2n( l +2ТТ)” J 7 [?>_! (-^ ) , . 1 ( W ] 1
Proof. Let a?„ = (0, ..., ...) < {tx, ..., ...) = x. Applying mo-, notony of the norm (4) we get
\%I = Ы т у У - Л ! ) < INIz^V-iO-)- Hence
y J J A W y J - A U
4
" \MtJ ~ 4 \MtJ = 1, i-ev ll^llrr^ < INIrv •Taking into account this inequality and inequalities (10) and (11) we obtain
П
.^<с”/
7МйГ
max IT-Ł v vn"T.Ilf' —un< C” ( l +2ru)
fj
|.Л , ( 2 - j ? _ ,(* * )] ‘ ||T ^ .... , J r .Taking in the above results <p(u) — uv we obtain the respective results of [4].
R eferences
[1] J. 4 lb r y c h t, Teoria przestrzeni IP z mieszaną normą dla skończenie ad- dytywnych funkcji zbiorów, IJAM, Prace Wydziału Mat-Fiz-Chem, Poznań 1963.
[2] — A generalization of a Zygmund-Bernstein Theorem, Ann. Polon. Math.
2 (1955), pp. 64-66.
[3] M. А. К р а с н о с е л ь с к и й , Я. Б. Р у т и ц к и й , Выпуклые функции и пространства Орлича, Москва 1958.
[4] С. М. Н и к о л ь с к и й , Неравенства для целых функций конечной сте
пени и их применение в теории дифференцируемых функций многих переменных, Труды Мат. ин-та АН 38(1951), рр. 244-278.
[5] A. Z y g m u n d , Trigonometric series, Cambridge 1959, vol. II.