loading

Logout succeed

Logout succeed. See you again!

ebook img

Universal covering space of the noncommutative torus PDF

file size0.08 MB

Preview Universal covering space of the noncommutative torus

Universal covering space of the noncommutative 4 torus 1 0 2 n a January 28, 2014 J 7 2 PetrR.Ivankov* ] e-mail: *[email protected] A O . h t a m [ 1 Abstract v 8 4 Gelfand-Na˘imarktheoremsuppliescontravariantfunctorfromacategoryofcom- 7 mutative C∗− algebras to a category of locally compact Hausdorff spaces. Therefore 6 any commutative C∗− algebra is an alternative representation of a topological space. . 1 Similarlyacategoryof(noncommutative) C∗−algebrascanberegardedasacategory 0 ofgeneralized(noncommutative)locallycompactHausdorffspaces. Generalizationsof 4 topological invariants may be defined by algebraicmethods. Forexample SerreSwan 1 : theorem states that complex topological K - theory coincides with K - theory of C∗ - v algebras. However the algebraic topology have a rich set of invariants. Some invari- i X ants do not have noncommutative generalizations yet. This article contains a sample of noncommutative universal covering. General theory of noncommutative universal r a coverings is being developed by the author of this article. However this sample has independentinterest,itisveryeasytounderstandanddoesnotrequireknowledgeof Hopf-Galoisextensions. Contents 1 Introduction 2 2 Algebraicconstructionofknownuniversalcoveringspaces 3 2.1 Algebraicconstruction oftheR →S1 covering . . . . . . . . . . . . . . . . . . 3 2.2 Algebraicconstruction oftheR2 → S1×S1 covering . . . . . . . . . . . . . . 4 1 3 Universalcoveringofanoncommutativetorus 5 4 Discussion 6 5 Acknowledgment 6 1 Introduction FollowingGelfand-Na˘imarktheorem[2]statesthatcategoryoflocallycompactHausdorff topologicalspacesisequivalenttoacategoryofcommutative C∗−algebras. Theorem1.1. LetHausbeacategoryoflocallycompactHausdorffspaceswithcontinuousproper mapsas morphisms. And, let C∗Comm be the category of commutativeC -algebras with proper *-homomorphisms(sendapproximateunitsintoapproximateunits)asmorphisms. Thereisacon- travariantfunctorC : Haus→ C∗CommwhichsendseachlocallycompactHausdorffspaceXto thecommutativeC∗ -algebra C (X) (C(X) if X iscompact). Conversely, thereisa contravariant 0 functor Ω : C∗Comm → Haus which sends each commutative C∗ -algebra A to the space of characterson A(withtheGelfandtopology). ThefunctorsCandΩ areanequivalenceofcategories. Soany(noncommutative)C∗−algebramayberegardedasgeneralized(noncommutative) locallycompactHausdorfftopologicalspace. Wemaysummarizeseveralpropertiesofthe GelfandNa˘imarkcofunctorwiththe followingdictionary. TOPOLOGY ALGEBRA Locallycompactspace C∗ -algebra Compactspace UnitalC∗ -algebra Continuous map *-homomorpfism Minimalcompactification Unitization Maximalcompactification Algebraifmultpicators Closedsubset Ideal Disjoint unionoftopological spaces(∐X) Directsumof(pro)- C∗ algebras(⊕A) ι ι Principalfibration Hopf-Galoisextension. Universalcovering ? Thisarticleassumeselementaryknowledge offollowingsubjects. 1. Algebraictopology [3]. 2. C∗−algebrasandoperatortheory[1],[2], Weusefollowing notation. 2 Symbol Meaning N monoid ofnaturalnumbers Z ringofintegers R (resp. C) Fieldofreal(resp. complex)numbers Q Fieldofrationalnumbers H Hilbertspace B(H) AlgebraofboundedoperatorsonHilbertspace H K(H) or K AlgebraofcompactoperatorsonHilbertspace H U(H)⊂B(H) Group ofunitaryoperatorsonHilbertspace H U(A)∈ A Group ofunitaryoperatorsofalgebra A A+ C∗−algebra A withadjointedidentity M(A) Amultiplier algebraof C∗-algebra A C(X) C∗ -algebraofcontinuous complexvalued functions ontopologicalspace X C (X) C∗ -algebraofcontinuous complexvalued 0 functionsontopologicalspacewhichtendsto0atinfinity C (X) Algebraofcontinuous functionswithcompactsupport c sp(a) Spectrumof elementof C∗-algebraa ∈ A If X˜ → X is an universal covering then there is a natural ∗-homomorphism f : C (X) → 0 M(C (X˜)). Homomorphism f can be defined by a faithful representations π : C (X) → 0 0 B(H),π∗ :C (X˜) → B(H)suchthat 0 π∗(f(x)x˜) =π(x)π∗(x˜), π∗(x˜f(x))= π∗(x˜)π(x), x ∈ C (X), x˜ ∈ C (X˜). (1) 0 0 Wewouldlike constructananalogueof 1foranoncommutative torus. 2 Algebraic construction of known universal covering spaces 2.1 Algebraic construction of the R → S1 covering Ourconstruction containstwoingredients: 1. Algebraicanalogueof n -listedcoveringprojection f : S1 →S1 ; n 2. AlgebraicanalogueofR →S1; 2.1. Constructionofn-listedcoveringprojection. Itiswellknown thatC(S1)isa C∗-algebra which is generated by a single unitary element u ∈ U(C(S1)). The C(S1) algebra can be faithfully represented, i.e. there is an inclusion C(S1) → B(H). Let sp(u) ∈ C be the spectrum of the u (it is known that sp(u) = {z ∈ C | |z| = 1}), φ ∈ B∞(sp(u)) is a Borel-measurablefunctionsuchthat (φ(z))n =z (∀z ∈sp(u)). (2) 3 According to spectral theorem [2] there exist v = φ(u) ∈ U(B(H)) and vn = u. Let C(u) → B(H) (resp. C(v) → B(H)) be a C∗- algebra generated by u (resp. v), then we have an inclusion C(u) ⊂ C(v) which corresponds to an n - listed covering projection f : S1 →S1. n 2.2. ConstructionofR → S1. A circle S1 can be parameterizedby an angle parameter θ ∈ [−π,π]. Any function g ∈ C([−1,1]) such that g(−1) = g(1) corresponds to ϕ ∈ C(S1) g suchthatφ (θ)= g(θ/π). Letusfixasequenceofoperatorsu =u,u ,u ,...∈ B(H)such g 0 1 2 that u2 = u , wehavea sequence of inclusions C(u) = C(u ) ⊂ C(u ) ⊂ C(u ) ⊂ ...→ n+1 n 0 1 2 B(H). Wewouldliketoprovethatthissequenceandanyrepresentationπ : C(u)→ B(H) naturallydefinesa representation π∗ : C (R) → B(H). Let C (R) ⊂ C (R) be analgebra 0 c 0 offunctionswithcompactsupport. Thereisanaturalinclusioni : C (R)→ B(H)defined c byfollowing way. If f ∈ C (R) thenthereis n ∈ N suchthatsupportof f iscontainedin c [−2n,2n]. If f∗ ∈ C([−1,1])issuchthat f(x) = f∗(2nx)then f∗(−1)= f∗(1)=0,andwe can define φf∗ ∈ C(S1). If πn : C(S1) ≈ C(un) → B(H) then we set i(f) = πn(φf∗). It is clearthatthisdefinitiondoesnotdependonn. SinceC (R)isdenseinC (R),aninclusion c 0 i : C (R) → B(H) canbe continuously extendedto a representation π∗ : C (R) → B(H). c 0 Representationsπ andπ∗ satisfy(1). 2.2 Algebraic construction of the R2 → S1 ×S1 covering 2.3. ConstructionofC∗-algebra. TheC(S1×S1)isgeneratedbytwounitaryelementsu,v∈ U(C(S1×S1)). ThereisafaithfulrepresentationC(S1×S1) → B(H). Thisrepresentation inducestwofaithfulrepresentationsπ :C(u) → B(H), π : C(v)→ B(H). Construction u v fromsection2.1suppliesfollowingtworepresentations: 1. π∗ :C (R) → B(H), u 0 2. π∗ : C (R) → B(H). v 0 ItisnaturallytosupposethatC (R2)isisomorphictothenormcompletionofsubalgebra 0 of B(H)generatedbyoperatorsoffollowing type: π∗(f )π∗(f ), π∗(f )π∗(f ); (f , f ∈ C (R)). (3) u 1 v 2 v 1 u 2 1 2 0 Howeveritisnot alwaystrue. Thisconstruction isnot unique because therearedifferent Borel-measurable functions which satisfy (2). This algebra is not always commutative, because one can select element u ∈ B(H) such that u = u2 and u v = −vu . However 1 1 1 1 anyalgebraconstructedby(3)isrepresentativeofanunique Moritaequivalenceclass. 2.4. Morita equivalence. Although constructed above algebra is not always isomorphic to C (R2)itisstronglyMoritaequivalenttoit[1]. Asitisprovenin[4]aσ-unitalC∗-algebra 0 A is strongly Morita equivalent to a σ-unital C∗-algebra B if there is a ∗-isomorphism A⊗K ≈ B⊗K. I find that good noncommutative theory of universal coverings should beinvariantwithrespecttoMoritaequivalence. ThistheorycanreplaceC∗-algebraswith theirstabilizations(recallthatthestabilizationofa C∗ algebra A isa C∗-algebra A⊗K). 4 Definition2.5. Let A bea C∗-algebra, A → B(H)isafaithfulrepresentation,u∈U(A+), v ∈ U(B(H)), is such that vn = u and vi ∈/ U(A+), (i = 1,...,n−1). A generated by v algebraisaminimal subalgebraof B(H)whichcontainsfollowingoperators: 1. via; (a ∈ A, i =0,...,n−1) 2. avi. Denoteby A{v} ageneratedby v algebra. Lemma 2.6. Let A be a C∗-algebra, A → B(H) is a faithful representation, u ∈ U(A+) is an unitary element such that sp(u) = {z ∈ C | |z| = 1}, ξ,η ∈ B∞(sp(u)) are Borel measured functionssuchthatξ(z)n =η(z)n =z(∀z ∈sp(u)). Thenthereisanisomorphism A{ξ(u)}⊗K → A{η(u)}⊗K (4) whichisalsoaleft A-moduleisomorphism. Theisomorphismisgivenby ξ(u)⊗x 7→ η(u)⊗ξη−1(u)x; (x ∈ K). (5) Proof. Follows fromthe equalityξ(u) =ξη−1(η(u)). Let u ∈ U(C(S1×S1)) be an unitary such that u 6= vn for any n > 1,v ∈ U(C(S1×S1)). One can construct different sequences x = u,x ,x ,... ∈ B(H), y = u,y ,y ,... ∈ B(H), 0 1 2 0 1 2 suchthat x = x2, y = y2 butC∗ -algebras n+1 n n+1 n A{x } ⊂ A{x } ⊂... 1 2 arenotisomorphic to C∗ -algebras A{y }⊂ A{y } ⊂.... 1 2 Howeverfollowingsequences A{x }⊗K ⊂ A{x }⊗K ⊂... 1 2 A{y }⊗K ⊂ A{y }⊗K ⊂... 1 2 contain isomorphic algebras. If A is the norm completion of an algebra generated by (3) then A⊗K ≈ C (R2)⊗K, i.e. A isstronglyMoritaequivalenttoC (R2). 0 0 3 Universal covering of a noncommutative torus A noncommmutative torus [5] A is a C∗-algebra generated by two unitary elements θ (u,v∈U(A ))suchthat θ uv =e2πiθvu, (θ ∈R). If θ ∈ Q then noncommutative torus is strongly Morita equivalent to commutative one, i.e. A ⊗K ≈ C(S1×S1)⊗K, and our construction is the same as in the section 2.2. A θ 5 case θ ∈/ Q ismore interesting. Howeveraconstruction universalcoveringfullycoincides with the considered in section 2.2 one. Let A → B(H) be a faithful representation. This θ representationinduces two representations π : C(u) → B(H), π : C(v) → B(H). These u v representations induce representations π∗(C (R)) → B(H), π∗(C (R)) → B(H). The u 0 v 0 universalalgebraofnoncommutative torusisanorm completionof analgebragenerated byoperatorsoffollowing type π∗(f )π∗(f ), π∗(f )π∗(f ); (f , f ∈ C (R)). u 1 v 2 v 1 u 2 1 2 0 ThisalgebraisnotuniquebutitisarepresentativeoftheuniquestrongMoritaequivalence class. 4 Discussion It is known the Maxwell’sequations of classical electrodynamic aremore important than Maxwell’sproof. NowIamoccupiedbygeneraltheoryofnoncommutativeuniversalcov- erings, which uses theory of Hopf C∗-algebras and Hopf-Galois extensions [6]. However I obtained a new algebra which can be regarded as a locally compact spectral triple [5]. Maybe this result is more interesting than a general theory. This algebra is also interest- ing because it contains almost commutative sector, i.e. noncommutativity parameter θ is infinitesimal. Itmeansthatthereisasequenceof algebras A → A →... θ θ/2 such that noncommutativity parameter tends to 0. Maybe this algebra has a physical sense. I found an analogy of this algebra with [7] Kaluza-Klein theory. In Kaluza-Klein theorywedonotobservecompactdimensions ofthe Universebecausetheyarecompact. Weobserveflatquotientspace,whichcorrespondstoasubalgebraoftheUniverse. Maybe weobservealmostcommutativesubalgebraoftheUniverse,becausewecannotobservea noncommutative algebra. 5 Acknowledgment Iwouldliketoacknowledgea"Non-commutativegeometryandtopology"seminar,orga- nizedby: 1. Prof. AlexanderMishchenko, 2. Prof. IvanBabenko, 3. Prof. EvgenijTroitsky, 4. Prof. VladimirManuilov, 5. Dr. AnvarIrmatov fordiscussionofmywork. 6 References [1] B. Blackadar. K-theory for Operator Algebras, Second edition. Cambridge University Press1998. [2] G.J.Murhpy.C∗-AlgebrasandOperatorTheory.AcademicPress1990. [3] E.H.Spanier.AlgebraicTopology.McGraw-Hill.NewYork 1966. [4] LawrenceG. Brown, Philip Green, and Marc A. Rieffel. Stable isomorphismand strong Morita equivalence of C -algebras. Source: Pacific J. Math. Volume 71, Number 2 , 349- 363,1977 [5] J.C.Várilly.AnIntroductiontoNoncommutativeGeometry.EMS2006. [6] Lecturenotesonnoncommutativegeometryandquantumgroups,EditedbyPiotrM.Hajac [7] J.M.Overduin,P.S.Wesson, Kaluza-KleinGravity,arXiv:gr-qc/9805018,1998. 7

See more

The list of books you might like