• No results found

Subharmonic almost periodic functions

N/A
N/A
Protected

Academic year: 2021

Share "Subharmonic almost periodic functions"

Copied!
16
0
0

Loading.... (view fulltext now)

Full text

(1)

Subharmonic almost periodic functions

A.V. Rakhnin and S.Yu. Favorov

Department ofMechanicsandMathematics,V.N.KarazinKharkov NationalUniversity

4SvobodySq.,Kharkov, 61077,Ukraine

E-mail:[email protected]

[email protected]

ReceivedJanuary21, 2005

Weprovethat almost periodicityin thesense ofdistributions coincides

with almost periodicity with respect to Stepanov's metricfor the class of

subharmonic functions in a strip fz2C : a<Imz<bg. We also prove

thatFouriercoecientsofthese functionsarecontinuousfunctions inImz.

Further,ifthelogarithmofasubharmonicalmostperiodicfunctionisa

sub-harmonicfunction,thenitisalmostperiodic.

Mathematics SubjectClassication 2000: 42A75,30B50.

Key words: subharmonic functions,almost periodic functions inthe senseof

distri-butions,Fouriercoecients.

Subharmonic almost periodic functions were introduced in [2] in connection

withinvestigationofzerodistributionofholomorphic almostperiodicfunctionsin

astrip. Inthis paperalmostperiodicitywasdened inthesense ofdistributions,

namely as almost periodicity of the convolution with a test function. However,

subharmonic functions logjf(z)j, where f(z) is a holomorphic almost periodic

function,wereconsidered much earlierinpapers [5]and[6 ],where theimportant

point was to prove almost periodicity of such functions in the sense of

distri-butions. In [4] this was extended to a subharmonic uniformly almost periodic

function whose logarithm isa subharmonic function.

Inthispaperweprovethatsubharmonicalmostperiodicinthesenseof

distri-butionsfunctionsarealmostperiodicintheclassicalsense,ifweconsiderStepanov

integral metric instead of the uniform metric. Therefore the classes of

subhar-monic almost periodic in the sense of distributions functions and subharmonic

Stepanov almostperiodicfunctions arethesame.

Now the FourierBohr coecients of such functions can be dened in the

usual way. For a horizontal strip these coecients are functions depending on

Imz. In this paperwe prove that these coecients depend continuously on Imz,

(2)

exponentialsumswithcontinuouscoecientsinStepanovmetric. Thusweprove

that subharmonic almost periodic functionsare Stepanov almost periodicin the

sense of thedenitionin[8].

In [2 ] it was proved that exp(u), where u is a subharmonic almost periodic

in the sense ofdistributions function, is also almost periodic inthe sense of

dis-tributions. Moreover, for an almost periodic function logjf(z)j, where f(z) is

aholomorphic function,jf(z)jisuniformlyalmostperiodic. Conversely,weprove

that the logarithm of a subharmonic almost periodicfunction is an almost

peri-odicfunction, provided it is a subharmonic function. Thus we obtain a stronger

than theone in[4 ],aswell astheconverse to theresultin[2].

We start withthefollowing denitions and notations(see [1, p. 51]).

Denition 1. A continuous function f(z) (z=x+iy),dened on R+iK,

where K isa compact subset of R (it isallows that K=f0g), is called uniformly

almost periodic (Bohr almost periodic), if from any sequence ft

n

gR one can

choose a subsequence ft

n 0

g such that the functions f(z+t

n 0

) converge uniformly

on R+iK.

Equivalent denitionisthe following:

For any " > 0 there exists L(") > 0 such that each interval of length L(")

contains a realnumber withthe property

sup

z2R+iK

jf(z+) f(z)j<":

Denition2. Adistributionf(z)2D 0

(S)oforder0(Sisanopenhorizontal

strip) iscalled almost periodic, iffor any testfunction '2D(S) the convolution

Z

u(z)'(z t)dxdy

is uniformly almost periodic on the real axis.

Notethat accordingto [6 ], for an almostperiodicdistribution f(z) from any

sequence fh

n

g R one can choose a subsequence fh

n 0g, such that R '(z)f(z+ h n 0

)dxdy convergeuniformlyon every set

K

=f'(z+t):t2R;'2Kg ,where

K isa compact subset ofD(S).

Anysubharmonic functionislocallyintegrable,sowecan consideritasa

dis-tribution.

A class of subharmonic almost periodic functions inan open strip S will be

denoted byWAP(S).

Furthermore,for 1<<<+1 we dene

S

[;]

(3)

ImS =fImz :z2Sg;

and for functions u, v, which are integrable on horizontal intervals in S

[;] , we denote d [;] (u;v):= sup z2S [;] 1 Z 0 ju(z+t) v(z+t)jdt:

Denition 3. A function f(z) integrable on horizontal intervals in an

open horizontal strip S is called Stepanov almost periodic, if from any sequence

fh

n

gR one can choose a subsequence fh

n 0

g anda function g(z) such that the

functionsf(z+h

n

0)converge tog(z) inthe topology denedby seminormsd

[;] ,

;2ImS.

A classof asubharmonicStepanovalmostperiodicfunctionsinanopenstrip

SwillbedenotedbyStAP(S). SincesuchfunctionsareStepanovalmostperiodic

on every liney=const,for u2StAP(S)there existsthemeanvalue

M(u;y):= lim T!1 1 2T T Z T u(x+iy)dx:

To eachsuchu we can associate FourierBohr series

u(z) X 2R a (u;y)e ix ; where a (u;y):=M(ue ix ;y):

areFourierBohr coecients.

Denition 4. A function u(z) 0 is called logarithmic subharmonic in

a domain GC, if the function logu(z) is subharmonic in this domain.

It iseasy to seethat alogarithmic subharmonic function issubharmonic.

Weprove thefollowing theorems:

Theorem 1. u(z)2WAP(S) if and onlyif u(z)2StAP(S).

Theorem 2. Let u(z) be a logarithmic subharmonic function in a strip S.

Thenlogu(z)2WAP(S) if andonlyif u(z)2WAP(S).

Theorem 3. Letu(z) be asubharmonicalmost periodicfunctioninastripS.

Thenits FourierBohr coecientsare continuous in ImS.

FromTheorem3and Besselinequalityfor FourierBohr coecientsitfollows

(4)

f2R :a

(u;y)60g)itismostcountable,whichalsofollowsfromTheorem1.12

in[6].

Theorem 4. Subharmonic function u(z) in an open horizontal strip S is

almost periodic if andonly ifthere existsa sequence of nite exponential sums

P m (z)= Nm X n=1 a (m) n (y)e i (m) n x ; (1) where n 2 R, a (m) n

(y) 2 C(ImS), which converges to the function u(z) in the

topology dened by seminorms d

[;]

,; 2ImS.

Moreover, P

m

(z), m=1;2;:::are subharmonic functions in S.

To provethe theoremsabove we usethefollowing propositions:

Proposition1. Convergence ofsubharmonicfunctionsinD 0 (G)isequivalent to the convergence in L 1 l oc (G)(see [7]).

Proposition2. Weaklimitofsubharmonicfunctionsissubharmonicfunction

(see [7]).

WedenotebyG

theGreenpotentialofameasureforthediskB(R ;0), i.e.

G (z):= Z B(R;z 0 ) log jR 2 zj R jz j d():

Lemma 1. Let measures

n

converge weakly to a measure in a

neighbor-hood of the disk B(R ;0) ,and (@B(R ;0))=0. Then for any t

1 >0,t 2 >0 such that t 2 1 +t 2 2 <R 2 , lim n!1 sup y2[ t 2 ;t 2 ] t1 Z t1 jG n (z) G (z)jdx=0; (2) where z=x+iy. Pr o of. Denote n = n . We have sup y2[ t2;t2] t 1 Z t 1 j G n (z) G (z)jdx sup y2[ t2;t2] t 1 Z t 1 Z B(R;0) log jR 2 zj R d n () dx + sup y2[ t 2 ;t 2 ] t 1 Z t 1 Z B(R;0) logjz jd n () dx: (3)

(5)

The condition (@B(R ;0)) =0 implies that the restrictions of the measures

n

to the disk B(R ;0) converge weakly to the restriction of the measure on

the disk, and the function log (jR 2 zjR 1 ) iscontinuous for jxjt 1 , jyjt 2 ,

2B(R ;0) . Thus therst term on the right-hand sideof (3) issmall. Without

lossofgenerality,wecan assumethatR<1=2,sothatforz; 2B(R ;0) wehave

logjz j<0.

Let">0beanarbitraryxednumber. Wedenotelog

"

jz j=maxflogjz

j;log"g. This function is continuous for jxj t

1 ,jyjt 2 , 2B(R ;0) and for any">0. We have sup y2[ t 2 ;t 2 ] t1 Z t1 Z B(R;0) logjz jd n () dx sup y2[ t 2 ;t 2 ] t1 Z t1 Z B(R;0) log " jz jd n () dx + sup y2[ t 2 ;t 2 ] t 1 Z t 1 Z B(R;0) jlogjz j log " jz jjdj n j()dx:

The rst term on the right-hand side of this inequality is small when n is

su-ciently large. Then

sup y2[ t 2 ;t 2 ] t 1 Z t 1 Z B(R;0) j logjz j log " jz jjdj n j()dx = sup y2[ t 2 ;t 2 ] Z B(R;0) Z [ t1;t1]\fx:jx+iy j"g (log" logjz j)dxdj n j() Z B(R;0) " Z " (log" logjxj)dxdj n j()2"j n j(B(R ;0)):

Notethat since ">0 isarbitrary,andmeasures

n

weaklyconverge to zero, one

can choosea constant C2R withj

n

j(B(R ;0))<C. Thelemma isproved.

Lemma 2. Let u

n

(z) be a sequence of subharmonic functions in a domain

GC converging to a function u 0 (z)6 1 in D 0 (G),and let sup z2G 0 u n (z)W(G 0 )<1

for any subdomain G 0

G. Then for any rectangle [a;b][;]G,

lim n!1 sup y2[;] b Z a ju n (z) u 0 (z)jdx=0: (4)

(6)

Pr oof. For every diskB(z

0

;R )Gwe have thefollowingrepresentation

u n (z) = G n R (z;z 0 )+H R (z;z 0 ;u n ) n=0;1;2:::; where n

arethe Riesz measures of the functionsu

n (z), G n R (z;z 0 ) isthe Green

potential of the measure

n inthe diskB(z 0 ;R ), and H R (z;z 0 ;u n

) arethe best

harmonicmajorantsof thefunctionsu

n

(z)inthis disk. Conditionsof thelemma

imply that

n

convergeweaklyto themeasure

0

. Without lossof generality,we

can assumethat(@B(z

0

;R ))=0,andusingLemma 1 weconclude thatfor any

t 1 ;t 2 , t 2 1 +t 2 2 <R 2 sup y 0 t 2 yy 0 +t 2 t 1 +x 0 Z t 1 +x 0 jG n R (z;z 0 ) G 0 R (z;z 0 )jdx !0;

when n! 1. From this itfollows that thefunctions H

R (z;z 0 ;u n ) converge to the function H R (z;z 0 ;u 0 ) in D 0 (B(z 0

;R )). Now using the meanvalue property,

Harnak inequality,andobviousinequality

H R (z;z 0 ;u n )W(B(z 0 ;R ))<1; n=0;1;2:::;

we obtain uniform convergence of H

R (z;z 0 ;u n ) to the function H R (z;z 0 ;u 0 ) in the rectangle [ t 1 +x 0 ;t 1 +x 0 ][ t 2 +y 0 ;t 2 +y 0

]. Covering the rectangle

[;][a;b]bya nitenumber ofsuch rectangles, we prove thelemma.

P r o o f o f T h e o r e m 1. Inclusion StAP(S) WAP(S) is obvious.

We prove the opposite inclusion. We consider arbitrary substrip S

[;]

, ; 2

ImS and a sequence fh

j

g R. Since u(z) is a subharmonic almost periodic

distribution, there exists a subsequence fh

j

k

g such that for some subharmonic

(clearly also almostperiodic)function v(z) and for any'2D(S

[;] ), uniformly in t2R, lim k!1 Z S (u(z+h j k +t) v(z+t))'(z)dxdy=0: (5)

Now we will show that the functions u(z+h

k

) converge to v(z) inthe topology

dened by seminorms d

[;]

, ; 2 ImS. Assuming the contrary, there exist

"

0

>0,;2ImS such thatfor aninnitesequence k 0 d [;] (u(z+h j k 0 );v(z))>" 0 ;

and thereforethere existsasubsequence ft

k 0g2R suchthat sup y2[;] 1 Z 0 ju(z+h j k 0 +t k 0) v(z+t k 0)jdx>" 0 : (6)

(7)

Passing to asubsequence (ifnecessary), we can assumethat u(z+h j k 0 +t k 0 )!w(z); v(z+t k 0) !w 1 (z)in D 0 (S [;] ): Lemma 2 implies sup y2[;] 1 Z 0 ju(z+h j k 0 +t k 0) w(z)jdx!0; k 0 !1; and sup y2[;] 1 Z 0 jv(z+t k 0) w 1 (z)jdx!0; k 0 !1;

and thus inequality(6)implies

sup y2[;] 1 Z 0 jw(z) w 1 (z)jdx" 0 : (7)

Onthe other hand,using(5), for anytestfunction '(z),

Z S [;] (w(z) w 1 (z))'(z)dxdy= lim k 0 !1 Z S [;] (u(z+h j k 0 +t 0 k ) v(z+t k 0))'(z)dxdy = lim k 0 !1 Z S [;] (u(z+h j k 0 ) v(z))'(z t k 0)dxdy=0; andthusw(z)=w 1

(z)almosteverywhere. Sincew(z)andw

1

(z)aresubharmonic

functions, thenw(z) w

1

(z), which contradicts (7). The theoremisproved.

Toprove Theorem 2 weneed thefollowing lemmas.

Lemma 3. Let '(t) be a function continuous in [ c;c]. Then for any ">0

there existsÆ, depending on 'and",such that forany two integrable on compact

set K functions f;g: K ![ c;c] the inequality

Z

K

jf(x) g(x)jdm<Æ

implies the inequality

Z

K

(8)

Proof. Choose >0suchthatjt 1 t 2 j< impliesj'(t 1 ) '(t 2 )j< " 2m(K) , and denote A 1 =fx2K:jf(x) g(x)j <g; A 2 =fx2K:jf(x) g(x)j g:

Noticethat m(A

2 ) 1 R A 2 jf(x) g(x)jdm, andtherefore Z K j'(f(x)) '(g(x))jdm Z A1 j'(f(x)) '(g(x))jdm+ Z A2 j'(f(x)) '(g(x))jdm m(A 1 )" 2m(K) + 2supj'(t)j Z K jf(x) g(x)jdm:

Choosingsuitable Æ,(8)follows. Thelemma isproved.

Lemma 4. Let u

n

(z) be a sequence of uniformly bounded from above

lo-garithmic subharmonic functions in a domain G C, converging to a function

u

0

(z) 60 in the sense of distributions. Then the functions logu

n

(z) converge to

the function logu

0

(z) in the sense of distributions.

Proof. Thefunctionsu

n

(z) arelogarithmic subharmonic,andinparticular

subharmonic. UsingProposition 1,u

n (z) converge tou 0 (z) inL 1 l oc (G).

Next, these functions are uniformly bounded from above by some constant

V >0, bounded from below by 0, and the function l

"

(t)= logmaxf";tg is

con-tinuous in the interval [0;V]. Lemma 3 implies that for xed " the functions

l

" (u

n

)(z) converge to the function l

" (u 0 )(z) in L 1 l oc

(G), and thus in the sense of

distributions. From Proposition 2it follows that thefunctionsl

" (u

0

)(z) are

sub-harmonicfor all",andtheir monotonelimit when"!0,i.e. thefunctionlogu

0 ,

is alsosubharmonic.

Nowwe consideradiskB(z

0

;r)G. Fromtheconvergence inL 1 (B(z 0 ;r)) ofthesequenceu n

(z)itfollowsthatthesubsequencefu

n 0

(z)gconvergesuniformly

on every xedcompact set K

1 B(z

0

;r) withpositive Lebesguemeasure. Since

thefunction logu

0

(z) issubharmonic and not identically 1 onK

1 , sup z2K 1 (logu 0 (z))C 0 ; or sup z2K 1 (u 0 (z))e C 0 :

Thus forall n 0 >n 0 sup z2K1 (u n 0(z))e C0 1 ;

(9)

and sup z2K 1 (logu n 0(z))C 0 1; 8n=0;1::: :

Sincethe functionsu

n

(z)areuniformlybounded fromabove oncompact subsets

ofG, itfollows thatthefamily flogu

n

0(z)g iscompactinD 0

(G). Thereforethere

existsasubsequencelogu

n 00 (z)whichconvergesinD 0 (G)(andalsoinL 1 l oc (G))to

some subharmonic functionv(z) inG.

Notethatfor anycompactsetKGandforany">0wehavethefollowing

inequality:

Z

K

jmaxflogu

n

0(z);log"g maxfv(z);log"gjdxdy Z

K jlogu

n

0(z) v(z)jdxdy:

Hence, the functions maxflogu

n 0

(z);log"g converge to the function

maxfv(z);log"g inL 1

l oc

(G)for any ">0.

On the other hand, as was shown above, l

" (u n )(z) converge to l " (u 0 )(z) in L 1 l oc

(G). Thus almost everywhere (and, since the functions are subharmonic,

everywhere)

maxfv(z);log"g=maxflogu

0

(z);log"g: (9)

Since a set on which a subharmonic function equals to 1 has

Lebesgue measurezero,then"!0impliesthatmes(fz 2G:v(z)<log"g)!0 ,

mes(fz 2G:logu

0

(z)<log"g)!0,andv(z)=logu

0

(z)almosteverywhere,and

henceeverywhere. Thus thesequenceof thefunctionslogu

n 0 (z)converges tothe function logu 0 (z) inD 0 (G)andinL 1 l oc (G).

If for some subsequence of the functions logu

n

j (z), "

0

> 0 and compact set

K 0 2G Z K0 jlogu n j (z) logu 0 (z)jdxdy" 0 ; (10)

then, using the above construction of the sequence u

n

j

(z), we have that some

subsequence of the sequence flogu

n j g converges to logu 0 (z) in L 1 l oc (G), which

contradicts (10). The lemmais proved.

P r o o f o f Th e o r e m2. From Proposition 3 in [2 ] it follows thatthe

inclusion logu2WAP(S) impliesthatinclusion u2WAP(S). We aregoing to

show the oppositeinclusion. Letu(z) 2WAP(S) and fh

n

gR be anarbitrary

sequence. Passing to a subsequence if necessary, we can assume that for some

subharmonic function u 0 ,uniformlyint2R, lim n!1 Z S (u(z+h n ) u 0 (z))'(z t)dxdy=0: (11)

(10)

Toprove the theoremit issucient to verifythat uniformlyint2R lim n!1 Z S logu(z+h n +t)'(z)dxdy= Z S logu 0 (z+t)'(z)dxdy: (12)

Assuming thatthis fails, forsome ">0and some sequencet

n !1, Z S logu(z+h n +t n )'(z)dxdy Z S logu 0 (z+t n )'(z)dxdy ": (13) Hereu 0

(z)isalogarithmicsubharmonicfunctionwithu

0

(z)2WAP(S). Passing

toasubsequenceandusingalmostperiodicityofthefunctionu

0 (z),wecanassume that lim n!1 Z S u 0 (z+t n )'(z)dxdy= Z S v(z)'(z)dxdy (14)

for some subharmonic in the strip S function v(z). Since the limit in (11) is

uniform int2R, (14)implies lim n!1 Z S u(z+h n +t n )'(z)dxdy= Z S v(z)'(z)dxdy:

Now Lemma 4 implies that both integrals in (13) have the same limit

R

logv(z)'(z)dxdy, when n ! 1, which is impossible. Thus (12) holds and

Theorem 2 isproved.

Proof of Theo rem3. Without lossofgenerality,wecan assumethat

S is a stripwith nite width. Let S

0

be an arbitrary substrip, S

0

S. Since

the function u(z) isalmost periodic, its Riesz measure := 1

2

u isalso almost

periodicinthesenseof distributions.

Denote K(w)= 1 2 logje w 2 1j; where 0< < max y 1 ;y 2 2ImS (y 1 y 2 ) 2 :

Notethatthe kernelK(w) isa subharmonic function whichis bounded from

aboveinS andits restriction to S

0

satisestheequation

(11)

where Æ(w) is astandard Diracmeasure. Denote

V(z)= Z

S

K(w z)'(Im w)d(w); (16)

where '0isa test function onImS such that'(y)=1 fory 2ImS

0 .

Denote P

n

= f(n 1=2;n+3=4)ImSg S. We are going to show that

V(z) isa subharmonicfunction ineveryP

n . Fixing n 0 2Z,we have Z S K(w z)'(Im w)d(w)= Z [n 0 1;n 0 +1)ImS K(w z)'(Im w)d(w) + X n2Znfn0 1;n0g Z [n;n+1)ImS K(w z)'(Im w)d(w): (17)

Every term in the right hand side of (17) is obviously a subharmonic function.

For Rew2[n;n+1), Imw2supp',z2P

n0 ,n6=n 0 ,n6=n 0 1we have e (w z) 2 =e (Rew Rez) 2 +(Imw Imz) 2 e (jn n 0 j 3=4) 2 : Thus X n2Znfn 0 1;n 0 g Z [n;n+1)ImS K(w z)'(Im w)d(w) X n2Znfn0 1;n0g sup z2Pn 0 sup w2[n;n+1)supp' 1 2 logj1 e (z w) 2 j ([n;n+1)supp'):

Since the measure is almost periodic, ([n;n+1)supp') is bounded from

above uniformlyin n(see [2]), and therefore theseries (17) converges uniformly

inz2P

n

0

and thefunctionV(z) issubharmonic inP

n

0

,and also inS.

Now we are going to show that the function V(z) is subharmonic almost

periodicinS. We consideratestfunction (z) onS andverifythatthefunction

f(t)= Z

S

V(z) (z t)dxdy

isuniformly almostperiodicon the real axis. We have

f(t)= Z S 0 @ Z S K(w z) (z)dxdy 1 A '(Im w)d(w+t):

(12)

Note thatthefunction

(w):= Z

S

K(w z) (z)dxdy

is continuous in S, because the dierence K(w) logjwj is continuous in some

neighborhoodof zero. Moreover, (w)=O(e jwj

2

)when jRewj!1.

Sincethevalues ([n;n+1]ImS

0

) areuniformlybounded inn, then

Z S '(Im w+Imz)d(w+z) 1+jwj 2 C 1 <1 (18) uniformlyinz2S 0 .

We x " > 0 and choose a test function (t), 0 (t) 1 on R, and such

that(Re w)=1on the set

w: j (w)j> " C 1 (1+jwj 2 ) :

For allt2R we have

f(t)= Z S (w)(Re w)'(Im w)d(w+t) + Z S (w)(1 (Rew))'(Imw)d(w+t):

Property(18)impliesthatthesecondintegralintheequalityisnotgreaterthan"

forallt2R. Sinceisanalmostperiodicmeasure, therstintegral isanalmost

periodic function, and if is an "-almost period, then it is a 2"-almost period

for f. Thus, thefunction V(z) isasubharmonic almostperiodic, andinaddition

(15)impliesthatV(z)=2'(y)(z) inthesenseofdistributions. Considerthe

function

H(z):=V(z) u(z):

Thisfunctionis harmonicandalmostperiodicinthesenseofdistributions inS

0 .

Let '0 be a test function in thedisk B(";0), which dependsonly on jzjand

suchthat R

'(z)dxdy=1. Sincethe convolution R H(z)'(z+)dxdy isequal to H()insome stripS 1 S 0

,thentheremarkafterDenition2 impliesuniform

almostperiodicityofthefunctionH(z) inS

1

. SoitsFourierBohr coecientsare

continuous in ImS

1

and, since " is arbitrary, in ImS

0

. Thus it is enough show

thatthe FourierBohr coecientsof V(z),

a (V;y)=M(Ve ix ;y)= lim T!1 1 2T T Z T V(x+iy)e ix dx;

(13)

arecontinuous.

Wex">0. We have

K(w)=maxfK(w); 2logNg+minfK(w)+2logN;0g=K

1

(w)+K

2 (w);

where N <1will bechosenlater. Denote

V 1 (z):= Z S K 1 (z w)'(Im w)d(w); V 2 (z):= Z S K 2 (z w)'(Im w)d(w): Since for jw 2 j<1=2 we have K(w)=1=2logj1 1+w 2 2 w 4 2! +:::j=logjwj+(w);

where (w) is a continuous function, then K

2

(w) = 0 for jwj Æ

0

> 0 and N

suciently large. Moreover, ifj(w)jlogN,thenfor all w2C and y2ImS,

T Z T K 2 (z w)dx= T Z T

minflogjx u+i(y v)j+(z w)+2logN;0gdx

1 Z 1 minflogjNx Nuj;0gdx= C N ; (19)

with some constant C, 0 < C < 1. Now using the property that

([n;n+1]supp')areboundedandthefactthatK

2

(z w)=0forjz wjÆ

0 ,

we havethat for allT

1 2T T Z T V 2 (z)e ix dx Z jRewjT+Æ0 1 2T T Z T jK 2 (z w)jdxd(w) C 2TN Z jRewjT+Æ 0 '(Imw)d(w) C 2 N "; (20) ifN is suciently large.

(14)

Further,sinceK

1

(w)=O(e jwj

2

)forjRe wj!1,thenonecanchooseatest

function 0(t)1 onR suchthat (Re w)=1 onthe set

w: jK 1 (w)j> " C 1 (1+jwj 2 ) ; where C 1

is the constantfrom (18). We have

V 1 (z)= Z S K 1 (w)(Rew)'(Im w+Imz)d(w+z) + Z S K 1

(w)(1 (Re w))'(Im w+Imz)d(w+z)=V

3

(z)+V

4 (z):

Fromthechoiceof thefunction itfollows that

jV

4

(z)j"; for z2S

0

: (21)

Sincethekernel K

1

(w)is continuousandthefamily ofshiftsofatestfunction in

ImS

0

is a compact set, then (see the remark to Denition 2) thefunction V

3 (z)

is uniformly almost periodic in S

0

and it has continuous in ImS

0

FourierBohr

coecients(see[1,p. 145]). Thus,ify

1 ;y 2 2ImS 0 andjy 1 y 2 jÆ("),then(20) and (21)imply ja (V;y 1 ) a (V;y 2 )jja (V 3 ;y 1 ) a (V 3 ;y 2 )j+ja (V 4 ;y 1 )j+ja (V 4 ;y 2 )j +ja (V 2 ;y 1 )j+ja (V 2 ;y 2 )j5": Thus a

(V;y) arecontinuous. Thetheoremis proved.

Pr o o f o f Th e or e m 4. For P

m

(z) we choose BohnerFejer sums of

the function u(z)

P m (z):= lim T!1 1 2T T Z T u(z+t) (m) (t)dt= X k (m) a (u;Imz)e iRez : Here (m)

(t)is a sequence of BohnerFejer kernels(see. [3, p. 69]), and theset

fk (m) :k (m)

6=0g is nite for every m. Note that, according to Theorem 3, the

functionsa(u;y)are continuousiny2ImS.

We aregoingto show thatP

m

(z) aresubharmonic.

Notethatthe kernels (m)

(t)are nonnegative,bounded,and

lim T!1 1 2T T Z T (m) (t)dt=1:

(15)

Alsonotethatthesubharmonicalmostperiodicfunctionu(z)isboundedfrom

above in anysubset S 0

S. Thus, using Fatou's lemma, for any m=1;2;:::,

z2S,and suciently small ,

1 2 2 Z 0 P m (z+e i' )d' lim T!1 1 2T T Z T 1 2 2 Z 0 u(z+e i' +t) (m) (t)d'dt lim T!1 1 2T T Z T u(z+t) (m) (t)dt=P m (z):

Asitisshown in[6 ],for anytest function'(z)inS and forsome (depending

only on the spectrum of the function u(z)) sequence of BohnerFejer sums, for

m!1,uniformlyint2R Z S P m (z)'(z+t)dxdy! Z S u(z)'(z+t)dxdy: (22)

Now we are going to verify that it implies the convergence in the topology

dened by seminorms d

[;]

, ; 2ImS. Indeed, if itis not true, thenfor some

sequence x

m

!1 and some; 2ImS,"

0 >0, sup y2[;] 1 Z 0 ju(x m +iy+t) P m (x m +iy+t)jdt" 0 :

Since u2StAP(S),then, passingto a subsequence ifnecessary,one can assume

thatfunctionsu(z+x

m

)converge tosomefunctionv 2StAP(S) withrespectto

metric d [;] ,and therefore sup y2[;] 1 Z 0 jP m (x m +iy+t) v(t+iy)jdt" 0 =2: (23)

Moreover, accordingto Theorem 2,

Z S u(z+x m )'(z)dxdy ! Z S v(z)'(z)dxdy: Therefore,bysettingt= x m

in(22), wehave for anytest function '(z)

lim m!1 Z S P m (z+x m )'(z)dxdy= Z S v(z)'(z)dxdy:

(16)

References

[1] C.Corduneanu, Almostperiodicfunctions. J.Wiley andSons,NewYork,1968.

[2] S.Yu. Favorov, A.Yu. Rashkovskii, and A.I. Ronkin, Almost periodic divizors in

astrip. J. Anal.Math.74 (1998),325345.

[3] B.M.Levitan, Almostperiodicfunctions. Gostehizdat, Moscow,1953.(Russian)

[4] A. Rakhnin, On one property of the subharmonic almost periodic functions.

VisnykHarkivs'kogo universytetu53(2003),No.602,2430.(Russian)

[5] L.I. Ronkin, Jessen's theorem for holomorphic almost periodic functions in tube

domains. Sibirsk.Mat. Zhurn.28(1987),199204.(Russian)

[6] L.I. Ronkin, Almost periodic distributions and divisorsin tube domains. Zap.

Nauchn.Sem. POMI247(1997),210236.(Russian)

[7] L.Hormander, Theanalysis oflinearpartial dierentialoperators.1: Distribution

theoryandFourieranalysis. SpringerVerlag,BerlinHeidelbergNewYorkTokio,

1983.

[8] O.I.Udodova, Holomorphicalmost periodicfunctionsinvariousmetrics. Visnyk

References

Related documents

Financial crises Allan M Malz allan malz@riskmetrics com RiskMetrics Group Risk 2001 Europe April 11, 2001 Overview ? What are market crises? ? Market crises and risk models ?

The feasibility of full quantification of myocardial blood flow (MBF) with per- fusion CMR has been demonstrated [20-23] and has been compared to positron emission computed tomog-

Acalypha indica leaves – Well diffusion method: In well diffusion method the water-acetone extract showed maximum inhibitory activity against the growth of

Secondly, it is assumed that competition – between students to secure the best outcomes for themselves as individuals, between individual teachers, and between schools – is

psychodynamic and behavioral activation therapy; 5) sample size of over 5 participants in the treatment group; 6) neuroimaging assessments prior to the initiation of treatment

LIST OF ABBREVIATIONS CSCW – Computer Supported Cooperative Work FX4 – Latest model of the classic London Black Cab 2006 GPS – Global Positioning Systems HCI – Computer

U1123061 Page - 77 - who related their work motivation to supportive leadership behaviours such. as “good listener”, “honest”, “making me to be creative” “pay attention

Fuzzy logic system proposes extraction of sentences using fuzzy rules and creation of a fuzzy set based on features such as sentence length, sentence location, similarity