ebook img

Compact composition operators on Hardy-Orlicz and weighted Bergman-Orlicz spaces on the ball PDF

0.16 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Compact composition operators on Hardy-Orlicz and weighted Bergman-Orlicz spaces on the ball

COMPACT COMPOSITION OPERATORS ON HARDY-ORLICZ AND WEIGHTED BERGMAN-ORLICZ SPACES ON THE BALL STÉPHANECHARPENTIER ABSTRACT. UsingrecentcharacterizationsofthecompactnessofcompositionoperatorsonHardy- Orlicz andBergman-Orliczspacesonthe ball([2, 3]), we first showthata compositionoperator ¥ which is compact on every Hardy-Orlicz (or Bergman-Orlicz) space has to be compact on H . Then, although it is well-known that a map whose range is contained in some nice Korányiap- proachregioninducesacompactcompositionoperatoronHp(BN)oronAap(BN),weprovethat, 1 foreachKorányiregionG ,thereexistsamapf :BN G suchthat,Cf isnotcompactonHy (BN), 1 wheny growsfast. Finally,weextend(andsimplify→theproofof)aresultbyK.Zhuforclassical 0 weightedBergmanspaces,byshowingthat,underreasonableconditions,acompositionoperator 2 Cf iscompactontheweightedBergman-OrliczspaceAay (BN),ifandonlyif n a y 1 1/(1 f (z))N(a ) − J lim −| | =0. 9 z 1 y (cid:16)1 1/(1 z)N(a ) (cid:17) ||→ − −| | 1 In particular,we deducethatthe compactn(cid:16)essofcomposit(cid:17)ionoperatorsonAya (BN)doesnotde- ] pendona anymorewhentheOrliczfunctiony growsfast. A F . h at 1. INTRODUCTION m LetB = z=(z ,...z ) CN,(cid:229) N z 2 <1 denotetheopenunitballofCN. Givenaholo- [ N 1 N ∈ i=1| i| 1 morphicmapnf :BN BN,thecompositionoperaotorCf ofsymbolf isdefinedbyCf (f)= f f , v for f holomorphicon→B . Compositionoperators havebeen extensivelystudiedoncommon◦Ba- N 7 nachspacesofanalyticfunctions,inparticularontheHardyspacesHp(B )andontheBergman 7 N 6 spaces Ap(BN), 1 p<¥ . The continuity and compactness of these operators have been char- ≤ 3 acterized in terms of Carleson measures ([5]). In dimension one, the boundedness ofCf for any . 1 f : D D is a consequence of the Littlewood subordination principle ([16]). In CN, N > 1, 0 it is w→ell-known that there exists some map f : B B such that the associated composition 1 N N 1 operator is not bounded on Hp(B ). Whatever the d→imension, it appears that both boundedness N v: and compactness of Cf on Hp(BN) (resp. Ap(BN)) are independent of p. On the other hand, Xi everycompositionoperatoris obviouslyboundedon H¥ and itis not difficultto check thatCf is compactonH¥ ifand onlyif f <1. Thusthereisa“break”between H¥ andHp(B )(resp. r ¥ N a Ap(B )), forthecompactnesskinkdimensionone, andevenfortheboundedness,whenN >1. N These observations first motivated P. Lefèvre, D. Li, H. Queffélec and L. Rodríguez-Piazza to study composition operators on Hardy-Orlicz spaces Hy (D) (resp. Bergman-Orlicz spaces Ay (D)) of the disc ([7, 10, 8, 9]), and then the author of [2, 3] to look at these questions in CN. These spaces both provide an intermediate scale of spaces between H¥ and Hp(B ) (resp. N Ap(B ))andgeneralizethelatter. Inparticular,in[10],theauthorswereinterestedinthequestion N of whether there are some Hardy-Orlicz spaces on which the compactness ofCf is equivalentto ¥ that on H . In fact, they answer this question in the negative, by proving ([10, Theorem 4.1]) that, for every Hardy-Orlicz space Hy (D), one can construct a surjectivemap f :D D which → 2000MathematicsSubjectClassification. Primary:47B33-Secondary:32C22;46E15. Keywordsandphrases. Carlesonmeasure-Compositionoperator-Bergman-Orliczspace-Hardy-Orliczspace -angularderivative. 1 2 COMPACTCOMPOSITIONOPERATORSONHARDY-ORLICZANDBERGMAN-ORLICZSPACESONBN induces a compact composition operatorCf on Hy (D). This result extends that obtained by B. MacCluer and J. Shapiro for Hp(D) ([13, Example 3.12]). The same problem in the Bergman- Orlicz case has not yet been completely solved. In several variables, the situation is much more surprizing, as we show in [2, 3] that there exist some Hardy-Orlicz and Bergman-Orlicz spaces, ¥ “close”enoughtoH , onwhichevery compositionoperatorisbounded. In this paper, we are mainly interested in the possibility to extend some known results about compactnessofcompositionoperators on classicalHardy orBergman spaces, tothecorrespond- ing Orlicz spaces. We think that this study may outlinesome interesting phenomena and precise thelinkbetween thebehaviorofCf and that off . Firstofall,wecomebacktothe“break”betweenH¥ andHp,1 p<¥ ,forthecompactness of Cf . There is no difference between being compact for Cf on≤one Hp(BN) and on every Hp(B ), while this property clearly depends on the Orlicz function y in Hy (B ). Therefore, N N we can wonder if the above question answered by [10] was the good one; indeed, the study of Cf on Hardy-Orlicz spaces arises the following question: what can we say about a composition operator which is compact on every Hardy-Orlicz space? It turns out that such an operator has ¥ to be compact on H , which seems to us to be a positive result, because it confirms that Hardy- Orlicz spaces covers well the “gap” between every Hp and H¥ . This result also stands when we replaceHardy-Orliczspaces by Bergman-Orliczspaces. Moreover,inHardyorBergmanspaces,compactness(andboundedness)ofcompositionoper- ators is handled in terms of geometric conditions, emphasizing the importance of the manner in whichthesymbolf approachestheboundaryofB . Tobeprecise,ifwedenotebyG (z ,a) B , N N forz S and a>1, theKorányiapproach region ⊂ N ∈ a G (z ,a)= z B , 1 z,z < 1 z 2 , N ∈ | −h i| 2 −| | n (cid:16) (cid:17)o it is known ([12]) that if f takes the unit ball into a Korányi region G (z ,a) with a small enough angular opening a, then Cf is compact on Hp(BN) and on Aap(BN). When N = 1, the Korányi regions are just non-tangential approach regions. In this paper, we show that this result does not holdanymoreforHardy-OrliczspacesonB ;forBergman-Orliczspaces,weobtainsucharesult N indimensiononeonly. In [13], the authors related the compactness of the composition operator Cf on Hp(D) or Aap(D) to the existence of angular derivative for f at the boundary. We say that the angular derivativeoff existsat apointz T ifthereexistsw Tsuch that ∈ ∈ f (z) w − z z − has a finite limit as z tends non-tangentially to z through D. The Julia-Caratheodory Theorem thenasserts thatthenon-existenceofan angularderivativeforf at somez Tisequivalentto ∈ 1 z (1.1) lim −| | =0. z z 1 f (z) → −| | ShapiroandTaylor[17]pointedoutthatifCf istobecompactonHp(D),thenf cannothavean angularderivativeat evenasinglepointin T, whichmay bewritten: 1 z (1.2) lim −| | =0. z 11 f (z) | |→ −| | In[13],itisproventhat(1.1)isnotsufficienttothecompactnessofCf onHardyspacesoftheunit disc in general, yet it is when f is univalent. However, this condition is necessary and sufficient forCf to be compact on every weighted Bergman spaces of the disc. The last main goal of this STÉPHANECHARPENTIER 3 paper is to extend some of these results to Hardy-Orlicz and Bergman-Orlicz spaces of the unit ball. In severalvariables, wecan also define theangularderivativeoff :B B at a pointin the N N unitsphereS andtheJulia-CaratheodoryTheoremalsoholdsinB ([15,T→heorem8.5.6]). Here, N N aswealreadysaid,thesituationiscomplicatedbythefactthatsomecompositionoperatorsarenot bounded on Hardy or Bergman spaces, and the fact that even the boundedness ofCf on Aap (BN) dependsona . In[18],K.ZhuprovesthatCf iscompactonAap(BN)ifandonlyifCondition(1.2) is satisfied, wheneverCf is bounded on some Abp(BN), for some 1<b <a . This assumption − is somehowjustified by the aboveobservation and by [13, Section 6], in which theauthors show that, for any a > 1 and any 0< p<¥ , there existsf :B B with no angular derivativeat N N any point of SN, s−uch thatCf is bounded on Aap(BN) but not→compact. There even exists such a map f such thatCf is not bounded on Aap (BN). In the present paper, we generalize Zhu’s result toweightedBergman-Orliczspacesontheball,byusingrecentcharacterizationsofboundedness and compactnessofcompositionoperators on thesespaces ([2]). Weshow that,ifCf is bounded y y onsomeAb (BN), 1<b <a , then itiscompact onAa (BN)ifand onlyif − y 1 1/(1 f (z) )N(a ) − −| | (1.3) lim =0, z 1 y (cid:16)1 1/(1 z )N(a ) (cid:17) | |→ − −| | (cid:16) (cid:17) where N(a )=N+a +1, under a mild and usual regularity condition on the Orlicz function y . Our proof is quite simple, while that of K. Zhu uses a Schur test in H2(B ) and the fact that the N compactness of composition operators on Hp(B ) does not depend on p. Combining this result N y with the automatic boundedness of every composition operator on Aa (BN) when y satisfies the y D 2-Condition, we get that the compactness on such Aa (BN) does not depend on a anymore. To y beprecise,Cf iscompact onAa (BN) ifandonlyif y 1(1/(1 f (z) )) − lim −| | =0, z 1 y −1(1/(1 z )) | |→ −| | whenevery satisfies theD 2-Condition. We have to mention that Condition (1.3) is, in any case, necessary. Moreover, the authors of [7] obtained such a result in dimension one, as announced in [11]. However, their proof uses the characterizationofthecompactnessofcompositionoperatorsintermsoftheNevanlinnacounting functionandis morecomplicated. We organize our paper as follows: a first preliminary part is devoted to the definitions and the statementsofthealreadyknownresultsweneed. Themainpartcontainsthethreemostimportant resultsmentionnedabove. Notation. Throughout this paper, we will denote by ds the normalized invariant measure on N a the unit sphere SN = ¶ BN, and by dva = ca 1 z 2 dv, a > 1, the normalized weighted −| | − Lebesguemeasureon theball. (cid:16) (cid:17) Given two points z,w CN, the euclidean inner product of z and w will be denoted by z,w , ∈ h i that is z,w = (cid:229) N z w; the notation will stand for the associated norm, as well as for the h i i=1 i i |·| modulusofacomplexnumber. Ifa > 1 is areal number,wewilldenotebyN(a ) thequantityN+a +1. − 4 COMPACTCOMPOSITIONOPERATORSONHARDY-ORLICZANDBERGMAN-ORLICZSPACESONBN 2. PRELIMINARIES 2.1. Hardy-Orlicz and Bergman-Orlicz spaces - Definitions. A strictly convex function y : y (x) R R is called an Orlicz function if y (0)=0, y is continuous at 0 and +¥ . + + → x −x−−+−→¥ If (W ,P) is a probability space, the Orlicz space Ly (W ) associated to the Orlicz func→tion y on (W ,P)isthesetofall(equivalenceclassesof)measurablefunctions f onW suchthatthereexists someC>0,suchthat W y |Cf| dPisfinite. Ly (W )isavectorspace,whichcanbenormedwith theso-called LuxembuRrgno(cid:16)rm (cid:17)defined by f f =inf C>0, y | | dP 1 . y k k W C ≤ (cid:26) Z (cid:18) (cid:19) (cid:27) Itis well-knownthat Ly (W ), isa Banach space(see [14]). y k·k Taking W = S an(cid:16)d dP = ds , t(cid:17)he Hardy-Orlicz space Hy (B ) on B is the Banach space N N N N of analytic functions f : BN C such that f Hy := sup0<r<1 fr y < ¥ , where fr Ly (SN) is defined by f (z)= f (rz). →Every functionkf kHy (B ) admitska rkadial boundary lim∈it f such r N ∗ that f =sup f < ¥ ([3, Section∈1.3]). For simplicity, we will sometimes denote ∗ y 0<r<1 r y by k y kthenormonHky (BkN),emphasizingthatHy (BN)can beseenas asubspaceofLy (SN). k·k WithW =BN anddP=dva ,a > 1,theweightedBergman-OrliczspaceAay (BN)isLy (BN) − y ∩ H(BN), where H(BN) stands forthevectorspace ofanalyticfunctionson theunitball. Aa (BN) isaBanach space. From thedefinitions,it iseasy toverify thatthefollowinginclusionshold: H¥ Hy (BN) Hp(BN) and H¥ Aya (BN) Aap(BN) ⊂ ⊂ ⊂ ⊂ foreveryOrliczfunctiony andany1 p<¥ . Moreover,ify (x)=xp,forsome1 p<¥ and foreveryx 0, thenHy (BN)=Hp(B≤N) and Aya (BN)=Aap (BN). ≤ ≥ 2.2. Four classes of Orlicz functions. Let y be an Orlicz function. In order to distinguish the Orlicz spaces and to get a significant scale of intermediate spaces between L¥ and Lp(W ), we definefourclasses ofOrliczfunctions. – Thetwofirstconditionsareregularityconditions: wesaythaty satisfiesthe(cid:209) -conditionif 0 itsatisfiesoneofthefollowingtwoequivalentconditions: y (Bx) y (C By) B (i) For any B>1, there exists some constantC 1, such that for any B ≥ y (x) ≤ y (y) x ylargeenough; ≤ y (Bx)n y (C y) n (ii) Foranyn>0,thereexistsC >0suchthat foranyx ylargeenough. n y (x)n ≤ y (By) ≤ Let us notice that (2) (1) is obvious, while an easy induction allows to prove (1) (2); the ⇒ ⇒ detailsareleft tothereader. IftheconstantC canbechosenindependentlyofB,theny satisfiestheuniform(cid:209) -Condition. B 0 – The(cid:209) -classconsistsofthoseOrliczfunctionsy suchthatthereexistsomeb >1andsome 2 x >0, suchthat y (b x) 2by (x), forx x . 0 0 ≥ ≥ – The third one is a condition of moderate growth: y satisfies the D -Condition if there exist 2 x >0 and aconstantK >1, suchthat y (2x) Ky (x) forany x x . 0 0 ≤ ≥ – Thefourthconditionisaconditionoffast growth: y satisfies theD 2-Conditionifitsatisfies oneofthefollowingequivalentconditions: (i) ThereexistC>0 andx >0, suchthaty (x)2 y (Cx) foreveryx x ; 0 0 ≤ ≥ STÉPHANECHARPENTIER 5 (ii) Thereexistb>1,C>0 andx >0 suchthaty (x)b y (Cx), foreveryx x ; 0 0 ≤ ≥ (iii) For every b > 1, there exist C > 0 and x > 0 such that y (x)b y (C x), for every b 0,b b ≤ x x . 0,b ≥ Finally,wementionthat theseconditionsare notindependent(see[7, Proposition4.7]): Proposition2.1. Let y beanOrliczfunction. (1) Ify satisfiestheuniform(cid:209) -Condition,thenitsatisfiesthe(cid:209) -Condition; 0 2 (2) Ify satisfiestheD 2-Condition,then itsatisfiestheuniform(cid:209) -Condition. 0 For any 1 < p <¥ , every function x xp is an Orlicz function which satisfies the uniform (cid:209) -Condition,(so(cid:209) and(cid:209) -conditions7−t→oo)and theD -Condition. Attheoppositeside, forany 0 2 0 2 a>0andb 1,x eaxb 1belongstotheD 2-Class(andthentotheuniform(cid:209) -Class),yetnot 0 ≥ 7−→ − totheD -one. Inaddition,theOrliczfunctionswhichcanbewrittenx exp a(ln(x+1))b 1 2 → − fora>0 and b 1, satisfythe(cid:209) and (cid:209) -Conditions,butdonot belongtot(cid:16)heD 2-Class. (cid:17) 2 0 ≥ For a complete study of Orlicz spaces, we refer to [6] and [14]. We can also find precise and usefulinformationin[7],suchasotherclassesofOrliczfunctionsandtheirlinkswitheachother. 2.3. Background results. All the results of the present paper are based on characterizations of the boundedness and compactness of composition operators on Hardy-Orlicz and Bergman- Orliczspaces([2,3]). AsIalreadysaid,thesecharacterizationsessentiallydependonthemanner inwhichtheOrliczfunctiongrows. The characterizations of the boundedness and compactness of Cf involve adapted Carleson measures, and then geometric notions. For z S and 0 <h <1, let us denote by S(z ,h) and N ∈ S (z ,h)thenon-isotropic“balls”, respectivelyin B and B , defined by N N S(z ,h)= z B , 1 z,z <h and S (z ,h)= z B , 1 z,z <h . N N { ∈ | −h i| } ∈ | −h i| We say that a finite positive Borel measure m on BN is a(cid:8)y -Carleson measure, y (cid:9)an Orlicz function,if 1 m (S (z ,h))=O , h→0(cid:18)y (Ay −1(1/hN))(cid:19) uniformly in z S and for some constant A > 0. m is a vanishing y -Carleson measure if the N ∈ aboveconditionis satisfied forevery A>0 and withthebig-Oh conditionreplaced by alittle-oh condition. A finitepositiveBorel measure m on B isa(y ,a )-Bergman-Carleson measureif N 1 m (S(z ,h))=O , h→0 y Ay −1 1/hN(a ) ! uniformly in z SN and for some constant A >(cid:0)0. m is(cid:0)a vanish(cid:1)in(cid:1)g (y ,a )-Bergman-Carleson ∈ measureiftheaboveconditionissatisfiedforeveryA>0andwiththebig-Ohconditionreplaced byalittle-ohcondition. When y satisfies the D -Condition, a (vanishing) y -Carleson measure (resp. (vanishing) 2 (y ,a )-Bergman-Carlesonmeasure)isa(vanishing)Carlesonmeasure(resp. (vanishing)Bergman- Carleson measure)(see[2, 3,Sections 3]). For f :BN BN, we denote by m f the pull-back measure of s N by the boundary limit f ∗ of → f ,and by m f ,a thatofdva byf . To beprecise, foranyE BN (resp. E BN), ⊂ ⊂ m f (E)=s N (f ∗)−1(E) and m f ,a (E)=va f −1(E) . (cid:16) (cid:17) (cid:0) (cid:1) 6 COMPACTCOMPOSITIONOPERATORSONHARDY-ORLICZANDBERGMAN-ORLICZSPACESONBN 2.3.1. Resultsfor Hardy-Orliczspaces(see[3, Section 3]). Themaintheoremis thefollowing: Theorem 2.2. Lety beanOrliczfunctionwhichsatisfiesthe(cid:209) -Conditionandletf :B B 2 N N → beholomorphic. (1) If y satisfies the uniform (cid:209) 0-Condition, then Cf is bounded from Hy (BN) into itself if andonlyif m f isa y -Carleson measure. (2) Ify satisfiesthe(cid:209) 0-Condition,thenCf iscompactfromHy (BN)intoitselfifandonlyif m f isa vanishingy -Carleson measure. (3) If y satisfies the D 2-Condition, then Cf is bounded (resp. compact) from Hy (BN) into itselfif andonlyif m f isa Carlesonmeasure(resp. a vanishingCarlesonmeasure). (4) If y satisfiestheD 2-Condition,thenCf isboundedon Hy (BN). The first two points are contained in [3, Theorem 3.2]; according to [5, Theorem 3.35], the third point means that, if y satisfies the D 2-Condition, then Cf is bounded (resp. compact) on Hy (B ) if and only if it is on Hp(B ) (see [3, Corollary 3.4]). The last point is [3, Theorem N N 3.7]. Due to the non-separability of small Hardy-Orlicz spaces, [3, Theorem 3.2] is not a direct consequence of Carleson-type embedding theorems obtained in [3, Section 2]; however, if we follow the proofs of these embedding theorems directly for composition operators, by working on spheres of radius 0<r <1, then we get the followingcharacterizations of both boundedness andcompactnessofcompositionoperators: Theorem 2.3. Let y be an Orlicz function satisfying the (cid:209) -Condition and let f :B B be 2 N N → holomorphic. (1) If y satisfies the uniform (cid:209) 0-Condition, then Cf is bounded on Hy (BN) if and only if thereexistssomeA>0 suchthat 1 (2.1) 0<sur<p1m f r(S (z ,h))=Oh→0(cid:18)y (Ay −1(1/hN))(cid:19) uniformlyin z S . N (2) If y satisfies th∈e (cid:209) 0-Condition, thenCf is compact on Hy (BN) if and only if, for every A>0, 1 (2.2) 0<sur<p1m f r(S (z ,h))=Oh→0(cid:18)y (Ay −1(1/hN))(cid:19) uniformlyin z S . N ∈ In the further, we will see how these two characterizations are useful depending on the situa- tions. 2.3.2. Results for Bergman-Orlicz spaces (see [2, Section 3]). The main result is similar to that statedinthepreviousparagraph: Theorem 2.4. Let y beanOrliczfunction,leta > 1 and letf :B B beholomorphic. N N − → y (1) If y satisfies the uniform (cid:209) 0-Condition, then Cf is bounded from Aa (BN) into itself if andonlyif m f isa (y ,a )-Bergman-Carlesonmeasure. y (2) If y satisfiesthe (cid:209) 0-Condition, thenCf is compact from Aa (BN) into itself if and only if m f isa vanishing(y ,a )-Bergman-Carlesonmeasure. y (3) If y satisfies the D 2-Condition, then Cf is bounded (resp. compact) from Aa (BN) into itself if and only if m f is a Bergman-Carleson measure (resp. a vanishing Bergman- Carlesonmeasure). y (4) If y satisfiestheD 2-Condition,thenCf isboundedon Aa (BN). STÉPHANECHARPENTIER 7 According to [5, Theorem 3.37], the third point means that, if y satisfies the D -Condition, 2 y thenCf is bounded (resp. compact) on Aa (BN), ifand only ifCf is bounded (resp. compact)on Aap(BN). 3. MAIN RESULTS 3.1. Compactness of Cf on every Hardy-Orlicz or Bergman-Orlicz spaces. The following theoremcompletesbothTheorem2.2 and Theorem2.4. Theorem 3.1. Letf :B B beaholomorphicmap. Thefollowingassertionsareequivalent: N N → (1) Cf iscompact onHy (BN), foreveryOrliczfunctiony ; y (2) Forsomea > 1,Cf is compacton Aa (BN), forevery Orliczfunctiony ; (3) Cf iscompact−onH¥ (BN); (4) f <1. ¥ k k Proof. Itiswell-knownthatCf iscompactonH¥ (BN)ifandonlyif f ¥ <1. Usingthefactthat a compositionoperator is compact on Hy (BN) (resp. Aya (BN)) if akndkonly if for every bounded sequence (fn)n ⊂ Hy (BN), kfnky ≤ 1, (resp. (fn)n ⊂ Aya (BN), kfnkAay ≤ 1) which tends to 0 uniformlyoneverycompactsubsetofBN,thenkfn◦f ky −n−−→¥ 0(resp. kfn◦f kAya −n−−→¥ 0),itis notdifficulttoshowthatif f ¥ <1,thenCf is compacton→Hy (BN)(resp. Aya (BN)→.) k k Itremainstoprove(1) (4)and(2) (4). Wefirstdealwiththeproofof(1) (4). Wewilluse thenecessarypartofThe⇒orem2.3. Let⇒usassumethatf inducesacompactcom⇒positionoperator oneveryHardy-Orliczspace. According to(2.2), thismeansthat 1 0<sur<p1zsuSpN m f r(S (z ,h)) =oh→0(cid:18)y (Ay −1(1/hN))(cid:19), ∈ (cid:0) (cid:1) foreveryA>0 and everyOrliczfunctiony , which inturn implies 1 (3.1) sup sup m f (S (z ,h)) , 0<r<1z SN r ≤ y (Ay −1(1/hN)) ∈ (cid:0) (cid:1) foreveryA>0, foreveryOrliczfunctiony andforh sufficientlysmall. Weintendtoshowthat sup sup m f (S (z ,h)) =0, r 0<r<1z S N ∈ (cid:0) (cid:1) forall0<h≤h0,h0 ∈(0,1). Bycontradiction,wesupposethat sup sup m f r(S (z ,h)) 6=0 0<r<1z S N foreveryh>0, since ∈ (cid:0) (cid:1) h sup sup m f (S (z ,h)) 7−→0<r<1z S r N ∈ (cid:0) (cid:1) is an increasing function on (0,1). A straightforward computation shows that inequality (3.1) is satisfied for every A >0, for every Orlicz function y and for h small enough, if and only if we have,by puttingx=1/h, y 1 xN 1 − (3.2) , ≤ A y 1 1/ sup sup (cid:0)m f (cid:1)(S (z ,1/x)) − r 0<r<1z S ! N ∈ (cid:0) (cid:1) foreveryA>0,foreveryOrliczfunctiony andforxlargeenough. Thefollowinglemmaensures thatthiscannotoccur: 8 COMPACTCOMPOSITIONOPERATORSONHARDY-ORLICZANDBERGMAN-ORLICZSPACESONBN Lemma 3.2. Let f,g:[0,+¥ [ [0,+¥ [ be two increasing functions which tend to +¥ at +¥ . There exist d > 0 and a conti→nuous increasing concave function n : [0,+¥ [ [0,+¥ [, with n (f (x)) → lim n (x)=+¥ ,suchthat d >0,for everyxlargeenough. x +¥ n (g(x)) ≥ → We assume for a while that this lemma has been proven, and we finish the proof of Theorem 3.1. Withthenotationsofthelemma,weput f (x)=xN and g(x)=1/ sup sup m f (S (z ,1/x)) . r ( 0<r<1z S N ∈ (cid:0) (cid:1) It is clear that limx +¥ f (x)=+¥ ; sinceCf is supposed to be compact on every Hy (BN), it is in particular bound→ed on Hp(B ) ([3, Corollary 3.5]), then we have g(x) +¥ (Theorem N −x−−+−→¥ 2.2, 3). Now, the above lemma provides a constant d > 0 and a continuous→increasing concave functionn , tendingtoinfinityat infinity,suchthat 1 n xN /n d >0,  sup m f Sf (x ,1/x) ≥ (cid:0) (cid:1) x SN   ∈ (cid:0) (cid:0) (cid:1)(cid:1) for every x large enough. It is not difficult to check that n can be constructed such that y =n 1 − x isanOrliczfunction,i.e. suchthat +¥ (thatiswhatwearedoingintheproofofthe n (x) −x−−+−→¥ → lemmabelow). Therefore, weget acontradictionwithCondition(3.2), sowe musthave sup sup m f r Sf (x ,h) =0, 0<r<1x S N ∈ (cid:0) (cid:0) (cid:1)(cid:1) foreveryh>0 smallenough. It followsthatthere existssome0<r <1such that 0 (3.3) sup m f r(C (r0,1))=0, 0<r<1 where C (r ,1) = z B ,r < z <1 . We intend to show that f 1(C (r ,1)) = 0/, which 0 N 0 − 0 { ∈ | | } shouldgivetheresult. Let0<r<1and let uslookat theset f 1(C (r ,1)) S = z S ,f (z ) C (r ,1) . r− 0 N N r 0 ∩ { ∈ ∈ } Condition(3.3)implies s f 1(C (r ,1)) S =0. N r− 0 N ∩ Since f r is continuous on BN, f r−1((cid:0)C (r0,1)) SN must b(cid:1)e an open subset of SN and then must ∩ beempty. So wehaveproventhat, foranyr (0,1), ∈ z S ,f (z ) C (r ,1) =f 1(C (r ,1)) rS =0/, N r 0 − 0 N { ∈ ∈ } ∩ whererS = z B , z =r , hence N N ∈ | | (cid:8) f 1(C(cid:9)(r ,1))= f 1(C (r ,1)) rS =0/. − 0 − 0 N ∩ 0<r<1 [ (cid:0) (cid:1) The proof in the Bergman-Orlicz context is much easier. Proceeding as above and using the necessarypartofthesecondpointofTheorem2.4, wegetthatCondition m f (C (r0,1))=0must hold, for some 0 < r < 1. By continuity of the map f on B , f 1(C (r ,1)) cannot be but 0 N − 0 empty. To finishtheproof, wehaveto dothatofLemma3.2: STÉPHANECHARPENTIER 9 ProofofLemma 3.2. The proof will be constructive. Let f and g be given as in the statement of the lemma. We are going to build by induction a sequence (a ) which will be of interest in the n n construction of the desired function n . We put a =0, a =1, and we deduce a from a and 0 1 n+2 n a in thefollowingway: wedefine n+1 b =sup g(x), f (x) a n+2 n+1 { ≤ } and a =max b ,a +(a a ) . n+2 n+2 n+1 n+1 n { − } Weobservethat: (1) If f (x) a , then g(x) a ; n+1 n+2 ≤ ≤ (2) a a a a 1. n+2 n+1 n+1 n − ≥ − ≥ Wenowconstructtheconcavefunctionn asacontinuousaffineone,whosederivativeisequalto 1 e = ontheinterval(a ,a ),andwithn (0)=0. Ofcoursen isincreasingand n n n+1 √n(a a ) n+1 n then maps [0,+−¥ [ into itself. Since e is decreasing, because of 2) above, n is concave. In order n tocheck that n tendstoinfinityat infinity,wecomputen (a ): n 1 n (a )=n (a )+e (a a )=n (a )+ . n+1 n n n+1 n n − √n Thereforen (an+1)=(cid:229) nk=+11 √1k whichshowsthatlimx→+¥ n (x)=+¥ , sincean →+¥ . n f (x) Wenowcheck that ◦ isboundedbelowby someconstantd >0, when xisbig enough. n g(x) Let x [0,+¥ [, and let n◦ be an integer such that a f (x) a ; we have n (f (x)) n (a ). n n+1 n Using∈thefirstpropertyofthesequence(a ) above,≤weget n≤(g(x)) n (a ). Thisy≥ields,for n n n+2 ≤ n 1, ≥ n (f (x)) n (an) = (cid:229) nk=1 √1k d >0, n (g(x)) ≥ n (a ) (cid:229) n+2 1 ≥ n+2 k=1 √k hencetheresult. (cid:3) (cid:3) 3.2. Korányi regions and compactness ofCf on Hardy-Orlicz and Bergman-Orlicz spaces. Forz S anda>1,werecallthattheKorányiapproachregionG (z ,a)ofangularopeningais N ∈ defined by a G (z ,a)= z B , 1 z,z < 1 z 2 . N ∈ | −h i| 2 −| | n (cid:16) (cid:17)o [5, Theorem 6.4]and thethirdpartofTheorem 2.2yieldsthefollowingresult: Theorem3.3. Lety beanOrliczfunctionsatisfyingtheD (cid:209) -Condition. Letalsof :B B 2 2 N N ∩ → beholomorphic. WeassumethatN >1and wefixb =(cos(p /(2N))) 1. N − (1) Iff (BN) G (z ,bN),thenCf is boundedonHy (BN); (2) If f (BN)⊂ G (z ,b), for some z SN and for some 1 < b <bN, then Cf is compact on Hy (B ).⊂ ∈ N (3) Both of the above results are sharp in the following sense: given c > b , there exists N f with f (BN) G (z ,c), for some z SN, andCf not bounded on Hy (BN); there also existssomef w⊂ithf (BN) G (z ,bN),∈forsomez SN,andCf notcompactonHy (BN). ⊂ ∈ 10 COMPACTCOMPOSITIONOPERATORSONHARDY-ORLICZANDBERGMAN-ORLICZSPACESONBN Remark 3.4. (1) If N =1, the two first point of the previous theorem are true if we put b =+¥ 1 and G (z ,+¥ ) = D. Indeed, the first point is nothing but the continuity of every composition operator on the disc, and the second one is contained in [5, Proposition 3.25] which says that, whenever f (D) is contained in some nontangential approach region in D, then Cf is Hilbert- SchmidtinH2(D), hencecompacton everyHp(D), 1 p<¥ . ≤ (2) Followingthe proof of[4, Theorem 3.3], it is not difficult to showthat theboundedness or compactness ofCf on Hy (BN) implies that on Aya (BN), for any a > 1, as soon as the Orlicz function y satisfies the D -Condition. Thus, the first two points of th−e previous theorem also 2 holdsfor Bergman-Orliczspaces. ThefollowingresultshowsthatTheorem3.3doesnotholdassoonastheOrliczfunctiongrows fast. Theorem 3.5. Let y be an Orlicz function satisfying the D 2-Condition. Then, for every z S N and every b > 1, there exists a holomorphic self-map f taking BN into G (z ,b), such thatC∈f is notcompacton Hy (B ). N Remark 3.6. Observethat thereisno assumptiononN. ProofofTheorem 3.5. Theproof will usethenecessary part ofthe second point ofTheorem 2.2. First ofall, werecall thatD 2-Conditionimplies(cid:209) -Condition(see Paragraph 2.2). We denoteby 2 e thevector(1,0...0)inCN. Itisclearlyenoughtoprovethetheoremforz =e . Foranyb>1, 1 1 weset 2cos 1(1/b) (3.4) b = − p in(0,1). We need alemmawhoseproofisincludedin thatof[5, Theorem6.4]: Lemma 3.7. Let b>1 and let b be defined by (3.4). There exists a holomorphicmap f :B N B ,with f (B ) G (e ,b),suchthat → N N 1 ⊂ (3.5) s f 1(S (e ,h)) Ch1/b , N − 1 ≥ forsomeconstantC>0 dependingonlyonf andb. Proof. Withoutgoingintodetails,webrieflygivetheideasoftheproof. Itusesthedeep Alexan- drov’sresultwhichgivestheexistenceofnon-constantinnerfunctionsinB ([1]). Therefore,we N considerafunctionf which can bewritten f = k j ,0 , ′ ◦ where 0 is the (n 1)-tuple (0,...,0), j is (cid:0)an inner f(cid:1)unction with j (0) =0, and where k is a ′ biholomorphic ma−p from D onto the non-tangential approach region G (1,b) in the disc, defined by b G (1,b)= z D, 1 z < 1 z 2 . ∈ | − | 2 −| | (cid:26) (cid:27) (cid:16) (cid:17) Onecanshowthatthelower-estimate(3.7)holdsforthismapf ,usingthefactthatinnerfunctions j aremeasurepreserving mapsofS intoT(see[15, p. 405])inthefollowingsense: N s N (j ∗)−1(E) =s 1(E), (cid:16) (cid:17) foranyBorel set E inT. (cid:3)

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.