Lecture Notes in Mathematics 2021 Editors: J.-M.Morel,Cachan B.Teissier,Paris Forfurthervolumes: http://www.springer.com/series/304 • Andreas Defant Classical Summation in Commutative and Noncommutative L -Spaces p 123 AndreasDefant CarlvonOssietzkyUniversityOldenburg DepartmentofMathematics CarlvonOssietzkyStrasse7-9 26111Oldenburg Germany [email protected] ISBN978-3-642-20437-1 e-ISBN978-3-642-20438-8 DOI10.1007/978-3-642-20438-8 SpringerHeidelbergDordrechtLondonNewYork LectureNotesinMathematicsISSNprintedition:0075-8434 ISSNelectronicedition:1617-9692 LibraryofCongressControlNumber:2011931784 MathematicsSubjectClassification(2011):46-XX;47-XX (cid:2)c Springer-VerlagBerlinHeidelberg2011 Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialis concerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation,broadcasting, reproductiononmicrofilmorinanyotherway,andstorageindatabanks.Duplicationofthispublication orpartsthereofispermittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9, 1965,initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer.Violations areliabletoprosecutionundertheGermanCopyrightLaw. Theuseofgeneral descriptive names,registered names, trademarks, etc. inthis publication does not imply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevantprotective lawsandregulationsandthereforefreeforgeneraluse. Coverdesign:deblik,Berlin Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) Preface In the theory of orthogonal series the most important coefficient test for almost everywhere convergence of orthonormal series is the fundamental theorem of MenchoffandRademacher.Itstatesthatwheneverasequence(α)ofcoefficients k satisfies the “test” ∑ |α logk|2 < ∞, then for every orthonormal series ∑ αx k k k k k in L (μ ) we have that ∑ α x converges μ -almost everywhere. The aim of this 2 k k k researchistodevelopasystematicschemewhichallowsustotransformimportant parts of the now classical theory of almost everywhere summation of general orthonormal series in L (μ ) into a similar theory for series in noncommutative 2 L -spaces L (M,ϕ) or even symmetric spaces E(M,ϕ) constructed over a p p noncommutativemeasure space (M,ϕ), a von Neumannalgebra M of operators actingonaHilbertspaceH togetherwithafaithfulnormalstateϕonthisalgebra. In Chap.2 we present a new and modern understandingof the classical theory on pointwise convergenceof orthonormalseries in the Hilbert spaces L (μ ), and 2 show that large parts of the classical theory transfer to a theory on pointwise convergenceofunconditionallyconvergentseriesinspacesL (μ ,X)ofμ -integrable p functionswithvaluesinBanachspacesX,ormoregenerallyBanachfunctionspaces E(μ ,X) of X-valued μ -integrablefunctions. Here our tools are strongly based on Grothendieck’smetric theoryof tensor productsandin particularon his the´ore`me fondamental. In Chap.3 this force turns out to be even strong enough to extend ourschemetothesettingofsymmetricspacesE(M,ϕ)ofoperatorsandHaagerup L -spacesL (M,ϕ).Incomparisonwiththeoldclassicalcommutativesettingthe p p newnoncommutationsettinghighlightsnewphenomena,andourtheoryasawhole unifies,completesandextendsvariousresults,bothin the commutativeandin the noncommutativeworld. Oldenburg,Germany AndreasDefant v Contents 1 Introduction .................................................................. 1 1.1 Step1.................................................................... 2 1.2 Step2.................................................................... 4 1.3 Step3.................................................................... 6 1.3.1 CoefficientTestsintheAlgebraItself ......................... 7 1.3.2 Tests in the Hilbert Spaces L (M,ϕ) 2 andMoreGenerallyHaagerupL ’s............................ 9 p 1.3.3 TestsinSymmetricSpacesE(M,τ),τaTrace............... 11 1.4 Preliminaries ............................................................ 13 2 CommutativeTheory ....................................................... 15 2.1 MaximizingMatrices................................................... 15 2.1.1 SummationofScalarSeries.................................... 16 2.1.2 MaximalInequalitiesinBanachFunctionSpaces ............ 19 2.1.3 (p,q)-MaximizingMatrices.................................... 21 2.1.4 MaximizingMatricesandOrthonormalSeries................ 24 2.1.5 MaximizingMatricesandSummation:TheCaseq<p...... 25 2.1.6 BanachOperatorIdeals:ARepetitorium...................... 28 2.1.7 MaximizingMatricesandSummation:TheCaseq≥p...... 33 2.1.8 AlmostEverywhereSummation............................... 37 2.2 BasicExamplesofMaximizingMatrices.............................. 40 2.2.1 TheSumMatrix................................................. 40 2.2.2 RieszMatrices .................................................. 46 2.2.3 Cesa`roMatrices................................................. 50 2.2.4 KroneckerMatrices............................................. 55 2.2.5 AbelMatrices................................................... 61 2.2.6 SchurMultipliers ............................................... 62 vii viii Contents 2.3 LimitTheoremsinBanachFunctionSpaces .......................... 65 2.3.1 CoefficientTestsinBanachFunctionSpaces ................. 65 2.3.2 LawsofLargeNumbersinBanachFunctionSpaces......... 74 3 NoncommutativeTheory ................................................... 79 3.1 TheTracialCase ........................................................ 79 3.1.1 SymmetricSpacesofOperators................................ 80 3.1.2 MaximalInequalitiesinSymmetricSpacesofOperators..... 86 3.1.3 TracialExtensionsofMaximizingMatrices................... 95 3.1.4 TheRow+ColumnMaximalTheorem........................ 100 3.1.5 AlmostUniformConvergence ................................. 109 3.1.6 CoefficientTestsinSymmetricOperatorSpaces.............. 111 3.1.7 LawsofLargeNumbersinSymmetricOperatorSpaces ..... 116 3.1.8 ACounterexample.............................................. 118 3.2 TheNontracialCase .................................................... 123 3.2.1 HaagerupL -Spaces............................................ 123 p 3.2.2 MaximalInequalitiesinHaagerupL -Spaces................. 126 p 3.2.3 NontracialExtensionsofMaximizingMatrices............... 128 3.2.4 AlmostSureConvergence...................................... 129 3.2.5 CoefficientTestsinHaagerupL -Spaces...................... 134 p 3.2.6 LawsofLargeNumbersinHaagerupL -Spaces ............. 139 p 3.2.7 CoefficientTestsintheAlgebraItself ......................... 141 3.2.8 LawsofLargeNumbersintheAlgebraItself................. 148 3.2.9 MaximalInequalitiesintheAlgebraItself .................... 150 References ........................................................................ 159 Symbols ........................................................................... 165 AuthorIndex ..................................................................... 167 SubjectIndex .................................................................... 169 Chapter 1 Introduction In the theory of orthogonal series the most important coefficient test for almost everywhereconvergenceoforthonormalseriesisthefundamentaltheoremofMen- choffandRademacherwhichwasobtainedindependentlyin[60]and[81].Itstates thatwheneverasequence(α)ofcoefficientssatisfiesthe“test”∑ |α logk|2<∞, k k k thenforeveryorthonormalseries∑ αx inL (μ )wehavethat k k k 2 ∑α x convergesμ -almosteverywhere. (1.1) k k k The aim of this research is to develop a systematic scheme which allows us to transform important parts of the now classical theory of almost everywhere summation of orthonormalseries in L (μ ) into a similar theory for series in non- 2 commutativeL -spacesL (M,ϕ)orevensymmetricspacesE(M,ϕ)buildupover p p anoncommutativemeasurespace(M,ϕ),avonNeumannalgebraM ofoperators actingonaHilbertspaceH togetherwithafaithfulnormalstateϕonthisalgebra. The theory of noncommutative L -spaces – the analogs of ordinary Legesgue p spaces L (μ ) with a noncommutative von Neumann algebra playing the role of p L∞(μ ) – for semifinite von Neumann algebras was laid out in the early 1950s by Dixmier [10], Kunze [49], Segal [84], and Nelson [66]. Later this theory was extended in various directions. In a series of papers P.G. Doods, T.K. Doods and dePagter[11,12,13,14],Ovchinnikov[69,70],ChilinandSukochev[88],Kalton andSukochev[45],Sukochev[85,86,87,88]andXu[96]developedamethodfor constructing symmetric Banach spaces of measurable operators, and Haagerup in [24]presentedatheoryofL -spacesassociatedwithnotnecessarilysemifinitevon p Neumannnalgebras.See[80]forabeautifulsurveyonthewholesubject. Inthe1980stheschoolofŁo´dz´ startedasystematicstudy,mainlyduetoHensz and Jajte, of Menchoff-Rademacher type theorems within the Gelfand-Naimark constructionL (M,ϕ),thenaturalHilbertspaceconstructedoveravonNeumann 2 algebra M and a faithful normal state ϕ on M, and most of their results were latercollectedinthetwoSpringerLectureNotes[37,38].Basedonnewtechniques mainlyduetoJunge,Pisier,andXu(seee.g.[39,42,79])someoftheseresultswere A.Defant,ClassicalSummationinCommutativeandNoncommutativeLp-Spaces, 1 LectureNotesinMathematics2021,DOI10.1007/978-3-642-20438-8 1, ©Springer-VerlagBerlinHeidelberg2011 2 1 Introduction generalizedin[7],andweplantousesimilartoolstocomplementandextendthis workinvariousdirections. In a first step we give a new and modern understandingof the classical theory on pointwise convergence of general orthonormal series in the Hilbert spaces L (μ ). Then in a second step we show that large parts of the classical theory 2 transfertoatheoryonpointwiseconvergenceofunconditionallyconvergentseries in spaces L (μ ,X) of μ -integrable functions with values in Banach spaces X, p or more generally Banach function spaces E(μ ,X) of X-valued μ -integrable functions. Here our tools are strongly based on Grothendieck’s metric theory of tensor products and in particular on his the´ore`me fondamental – this force in a third step is evenstrong enoughto extendourscheme to the setting of symmetric spacesE(M,ϕ)ofoperatorsandHaagerupL -spacesL (M,ϕ). Itturnsoutthat p p incomparisonwiththeoldclassicalcommutativesettingthenewnoncommutative settinghighlightsnewphenomena,andourtheoryasawholeunifies,completesand extends the original theory, both in the commutative and in the noncommutative setting: Step1 Step2 Step3 classical summationof summationof summationof unconditionally unconditionally =⇒ =⇒ orthonormal convergentseriesin convergentseriesin seriesinL2(μ ) Banachfunctionspaces operatoralgebras 1.1 Step 1 Let us start by describing more carefully some of the prototypical results of the classicaltheoryofalmosteverywheresummationofgeneralorthogonalserieswhich form the historical background of these notes; as standard references we use the monographs[1]and[47].SeeagaintheMenchoff-Rademachertheoremfrom(1.1) which is the most fundamental one. As quite often happens in Fourier analysis, this fact comes with an at least formally stronger result on maximal functions. Kantorovitchin [46] provedthat the maximalfunction (of the sequence of partial sums)ofeachorthonormalseries∑ αx inL (μ )satisfying∑ |α logk|2<∞,is k k k 2 k k square-integrable, (cid:2) (cid:2) (cid:2) j (cid:2) sup(cid:2)∑αx (cid:2)∈L (μ ), (1.2) k k 2 j k=0 or in other terms: For each choice of finitely many orthonormal functions x ,...,x ∈L (μ )andeachchoiceoffinitelymanyscalarsα,...α wehave 0 n 2 0 n 1.1 Step1 3 (cid:3)(cid:3)(cid:3)sup(cid:2)(cid:2)∑j αx (cid:2)(cid:2)(cid:3)(cid:3)(cid:3) ≤C(cid:4)∑n |α logk|2(cid:5)1/2, (1.3) k k k j k=0 2 k=0 whereCisanabsoluteconstant.ThisinequalityisalsoknownastheKantorovitch- Menchoff-Rademacherinequality. Results ofthistypecanandshouldbeseeninthecontextoflimittheoremsfor random variables. To illustrate this by an example, we mention that by a simple lemma usually attributed to Kronecker, the Menchoff-Rademacher theorem (1.1) impliesthefollowingstronglawoflargenumbers:Givenasequenceofuncorrelated randomvariablesX onaprobabilityspace(Ω ,P)wehavethat k (cid:4) (cid:5) 1 ∑j X −E(X ) convergesP-almosteverywhere, (1.4) j+1k=0 k k j provided ∑log2kVar(X )<∞. k2 k k There is a long list of analogs of the Menchoff-Rademacher theorem for various classicalsummationmethods,ase.g.Cesa`ro,Riesz,orAbelsummation.Recallthat asummationmethodformallyisascalarmatrixS=(s )withpositiveentriessuch jk thatforeachconvergentseriess=∑ x ofscalarswehave n n k s=lim ∑sjk∑x(cid:2). (1.5) j k (cid:2)=0 Letusexplainwhatwemeanbyacoefficienttestforalmosteverywhereconvergence of orthonormal series with respect to a summation method S = (s ) and an kj increasing and unboundedsequence (ω) of positive scalars. We mean a theorem k whichassuresthat k ∑αkxk=lim ∑sjk∑α(cid:2)x(cid:2) μ -a.e. (1.6) k j k (cid:2)=0 whenever ∑ αx is an orthonormal series in some L (μ ) with coefficients (α) k k k 2 k satisfyingthetest∑ |αω|2<∞;intheliteraturesuchasequenceωisthenoften k k k calledaWeylsequence. Forexample,thecoefficienttest∑ |α loglogk|2<∞assuresthatallorthonor- k k malseries∑ αx arealmosteverywhereCesa`ro-summable,i.e. k k k ∑αkxk= j+11 ∑j ∑k α(cid:2)x(cid:2) μ -a.e., (1.7) k k=0(cid:2)=0