Noncommutative geometry and motives (a quoi servent les endomotifs?)
aa r X i v : . [ m a t h . QA ] N ov Noncommutative geometry and motives(a quoi servent les endomotifs?)
Caterina Consani ∗ Mathematics Department, The Johns Hopkins University3400 N. Charles Street, Baltimore MD 21218 USAemail: [email protected]
Abstract.
This paper gives a short and historical survey on the theory of pure motivesin algebraic geometry and reviews some of the recent developments of this theory innoncommutative geometry. The second part of the paper outlines the new theory ofendomotives and some of its relevant applications in number-theory.
This paper is based on three lectures I gave at the Conference on “Renormalizationand Galois theories” that was held in Luminy, at the Centre International de Ren-contres Math´ematiques (CIRM), on March 2006. The purpose of these talks wasto give an elementary overview on classical motives (pure motives) and to surveyon some of the recent developments of this theory in noncommutative geometry,especially following the introduction of the notion of an endomotive .It is likely to expect that the reader acquainted with the literature on motivestheory will not fail to notice the allusion, in this title, to the paper [17] in whichP. Deligne states that in spite of the lack of essential progresses on the problem ofconstructing “relevant” algebraic cycles, the techniques supplied by the theory ofmotives remain a powerful tool in algebraic geometry and arithmetic.The assertion on the lack of relevant progresses on algebraic cycles seems, unfor-tunately, still to apply at the present time, fifteen years after Deligne wrote hispaper. Despite the general failure of testing the Standard Conjectures, it is alsotrue that in these recent years the knowledge on motives has been substantiallyimproved by several new results and also by some unexpected developments.Motives were introduced by A. Grothendieck with the aim to supply an intrinsicexplanation for the analogies occurring among various cohomological theories inalgebraic geometry. They are expected to play the role of a universal cohomological ∗ Work partially supported by NSF-FRG grant DMS-0652431. The author wishes to thank theorganizers of the Conference for their kind invitation to speak and the CIRM in Luminy-Marseillefor the pleasant atmosphere and their support.
Caterina Consani theory by also furnishing a linearization of the theory of algebraic varieties and inthe original understanding they were expected to provide the correct frameworkfor a successful approach to the Weil’s Conjectures on the zeta-function of a varietyover a finite field.Even though the Weil’s Conjectures have been proved by Deligne without appeal-ing to the theory of motives, an enlarged and in part still conjectural theory ofmixed motives has in the meanwhile proved its usefulness in explaining conceptu-ally, some intriguing phenomena arising in several areas of pure mathematics, suchas Hodge theory, K -theory, algebraic cycles, polylogarithms, L -functions, Galoisrepresentations etc.Very recently, some new developments of the theory of motives to number-theory and quantum field theory have been found or are about to be developed,with the support of techniques supplied by noncommutative geometry and thetheory of operator algebras.In number-theory, a conceptual understanding of the main result of [10] on theinterpretation proposed by A. Connes of the Weil explicit formulae as a Lefschetztrace formula over the noncommutative space of ad`ele classes, requires the intro-duction of a generalized category of motives inclusive of spaces which are highlysingular from a classical viewpoint.The problem of finding a suitable enlargement of the category of (smooth projec-tive) algebraic varieties is combined with the even more compelling one of the def-inition of a generalized notion of correspondences. Several questions arise alreadywhen one considers special types of zero-dimensional noncommutative spaces, suchas the space underlying the quantum statistical dynamical system defined by J. B.Bost and Connes in [6] (the BC-system). This space is a simplified version of thead`eles class space of [10] and it encodes in its group of symmetries, the arithmeticof the maximal abelian extension of Q .In this paper I give an overview on the theory of endomotives (algebraic andanalytic). This theory has been originally developed in the joint paper [7] withA. Connes and M. Marcolli and has been applied already in our subsequent work[11]. The category of endomotives is the minimal one that makes it possible tounderstand conceptually the role played by the absolute Galois group in severaldynamical systems that have been recently introduced in noncommutative geome-try as generalizations of the BC-system, which was our motivating and prototypeexample.The category of endomotives is a natural enlargement of the category of Artinmotives: the objects are noncommutative spaces defined by semigroup actionson projective limits of Artin motives. The morphisms generalize the notion ofalgebraic correspondences and are defined by means of ´etale groupoids to accountfor the presence of the semigroup actions.Endomotives carry a natural Galois action which is inherited from the Artin mo-tives and they have both an algebraic and an analytic description. The latter isparticularly useful as it provides the data of a quantum statistical dynamical sys-tem, via the implementation of a canonical time evolution (a one-parameter familyof automorphisms) which is associated by the theory of M. Tomita ( cf. [36]) toan initial state (probability measure) assigned on an analytic endomotive. This oncommutative geometry and motives C ∗ -algebra and in particular to the algebra of the BC-system.The implication in number-theory is striking: the time evolution implements onthe dual system a scaling action which combines with the action of the Galoisgroup to determine on the cyclic homology of a suitable noncommutative motiveassociated to the original endomotive, a characteristic zero analog of the actionof the Weil group on the ´etale cohomology of an algebraic variety. When thesetechniques are applied to the endomotive of the BC-system or to the endomotiveof the ad`eles class space, the main implication is the spectral realization of thezeroes of the corresponding L -functions.These results supply a first answer to the question I raised in the title of thispaper (a quoi servent les endomotifs?). An open and interesting problem is con-nected to the definition of a higher dimensional theory of noncommutative motivesand in particular the introduction of a theory of noncommutative elliptic motivesand modular forms. A related problem is of course connected to the definitionof a higher dimensional theory of geometric correspondences. The comparisonbetween algebraic correspondences for motives and geometric correspondences fornoncommutative spaces is particularly easy in the zero-dimensional case, becausethe equivalence relations play no role. In noncommutative geometry, algebraic cy-cles are naturally replaced by bi-modules, or by classes in equivariant KK -theory.Naturally, the original problem of finding “interesting” cycles pops-up again inthis topological framework: a satisfactory solution to this question seems to beone of the main steps to undertake for a further development of these ideas. The theory of motives in algebraic geometry was established by A. Grothendieckin the 1960s: 1963-69 ( cf. [35],[21]). The foundations are documented in theunpublished manuscript [19] and were discussed in a seminar at the Institut desHautes ´Etudes Scientifiques, in 1967. This theory was conceived as a fundamentalmachine to develop Grothendieck’s “long-run program” focused on the theme ofthe connections between geometry and arithmetic.At the heart of the philosophy of motives sit Grothendieck’s speculations on theexistence of a universal cohomological theory for algebraic varieties defined overa base field k and taking values into an abelian, tensor category. The study ofthis problem originated as the consequence of a general dissatisfaction connectedto an extensive use of topological methods in algebraic geometry, with the re-sult of producing several but insufficiently related cohomological theories (Betti,de-Rham, ´etale etc.). The typical example is furnished by a family of homomor-phisms H i et ( X, Q ℓ ) → H i et ( Y, Q ℓ ) connecting the groups of ´etale cohomology oftwo (smooth, projective) varieties, as the prime number ℓ varies, which are notconnected, in general, by any sort of (canonical) relation. Caterina Consani
The definition of a contravariant functor (functor of motivic cohomology) h : V k → M k ( V k ) , X h ( X )from the category V k of projective, smooth, irreducible algebraic varieties over k to a semi-simple abelian category of pure motives M k ( V k ) is also tied up with thedefinition of a universal cohomological theory through which every other classical,cohomology H · (here understood as contravariant functor) should factor by meansof the introduction of a fiber functor (realization ⊗ -functor) ω connecting M k ( V k )to the abelian category of (graded) vector-spaces over Q V k H · ✲ gr V ect Q M k ( V k ) ω ✻ h ✲ Following Grothendieck’s original viewpoint, the functor h should implement thesought for mechanism of compatibilities (in ´etale cohomology) and at the sametime it should also describe a universal linearization of the theory of algebraicvarieties.The definition of the category M k ( V k ) arised from a classical construction in al-gebraic geometry which is based on the idea of extending the collection of algebraicmorphisms in V k by including the (algebraic) correspondences . A correspondencebetween two objects X and Y in V k is a multi-valued map which connects them.An algebraic correspondence is defined by means of an algebraic cycle in the carte-sian product X × Y . The concept of (algebraic) correspondence in geometry ismuch older than that of a motive: it is in fact already present in several works ofthe Italian school in algebraic geometry ( cf. Severi’s theory of correspondences onalgebraic curves).Grothendieck’s new intuition was that the whole philosophy of motives is regulatedby the theory of (algebraic) correspondences:“ ...J’appelle motif sur k quelque chose comme un groupe de cohomologie ℓ -adique d’un schema alg´ebrique sur k , mais consid´er´ee comme ind´ependant de ℓ , etavec sa structure enti`ere, ou disons pour l’instant sur Q , d´eduite de la th´eorie descycles alg´ebriques... ” ( cf. [15], Lettre 16.8.1964).Motives were envisioned with the hope to explain the intrinsic relations betweenintegrals of algebraic functions in one or more complex variables. Their ultimategoal was to supply a machine that would guarantee a generalization of the mainresults of Galois theory to systems of polynomials equations in several variables.Here, we refer in particular to a higher-dimensional analog of the well-known resultwhich describes the linearization of the Galois-Grothendieck correspondence forthe category V ok of ´etale, finite k -schemes V ok ∼ → { finite sets with Gal(¯ k/k )-action } , X X (¯ k ) oncommutative geometry and motives A classical problem in algebraic geometry is that of computing the solutions of afinite set of polynomial equations f ( X , . . . , X m ) = 0 , . . . , f r ( X , . . . , X m ) = 0with coefficients in a finite field F q . This study is naturally formalized by intro-ducing the generating series ζ ( X, t ) = exp( X m ≥ ν m t m m ) (2.1)which is associated to the algebraic variety X = V ( f , . . . , f m ) that is defined asthe set of the common zeroes of the polynomials f , . . . , f r .Under the assumption that X is smooth and projective, the series (2.1) encodes thecomplete information on the number of the rational points of the algebraic variety,through the coefficients ν m = | X ( F q m ) | . The integers ν m supply the cardinality ofthe set of the rational points of X , computed in successive finite field extensions F q m of the base field F q .Intersection theory furnishes a general way to determine the number ν m as inter-section number of two algebraic cycles on the cartesian product X × X : namely thediagonal ∆ X and the graph Γ F r m of the m -th iterated composite of the Frobeniusmorphism on the scheme ( X, O X ): F r : X → X ; F r ( P ) = P, f ( x ) f ( x q ) , ∀ f ∈ O X ( U ) , ∀ U ⊂ X open set . The Frobenius endomorphism is in fact an interesting example of correspondence ,perhaps the most interesting one, for algebraic varieties defined over finite fields.As a correspondence it induces a commutative diagram H ∗ ( X × X ) −∩ Γ Fr −−−−−→ H ∗ ( X × X ) x p ∗ y ( p ) ∗ H ∗ ( X ) F r ∗ −−−−→ H ∗ ( X ) (2.2)in ´etale cohomology. Through the commutativity of the above diagram one gets away to express the action of the induced homomorphism in cohomology, by meansof the formula F r ∗ ( c ) = ( p ) ∗ ( p ∗ ( c ) ∩ Γ F r ) , ∀ c ∈ H ∗ ( X ) , Caterina Consani where p i : X × X → X denote the two projection maps. Here, ‘algebraic’ refersto the algebraic cycle Γ F r ⊂ X × X that performs such a correspondence.For particularly simple algebraic varieties, such as projective spaces P n , the com-putation of the integers ν m can be done by applying an elementary combina-torial argument based on the set-theoretical description of the space P n ( k ) = k n +1 \ { } /k × ( k = any field). This has the effect to produce the interestingdescription | P n ( F q m ) | = q m ( n +1) − q m − q m + q m + · · · + q mn . (2.3)This decomposition of the set of the rational points of a projective space wascertainly a first source of inspiration in the process of formalizing the foundationsof the theory of motives. In fact, one is naturally led to wonder on the casualityof the decomposition (2.3), possibly ascribing such a result to the presence of acellular decomposition on the projective space which induces a related break-up onthe set of the rational points. Remarkably, A. Weil proved that a similar formulaholds also in the more general case of a smooth, projective algebraic curve C / F q of genus g ≥
0. In this case one shows that | C ( F q m ) | = 1 − g X i =1 ω mi + q m ; ω i ∈ ¯ Q , | ω i | = q / . (2.4)These results suggest that (2.3) and (2.4) are the manifestation of a deep andintrinsic structure that governs the geometry of algebraic varieties.The development of the theory of motives has in fact shown to us that this struc-ture reveals itself in several contexts: topologically, manifests its presence in thedecomposition of the cohomology H ∗ ( X ) = ⊕ i ≥ H i ( X ), whereas arithmeticallyit turns out that it is the same structure that controls the decomposition of theseries (2.1) as a rational function of t : ζ ( X, t ) = Q i ≥ det(1 − tF r ∗ | H i +1 et ( X )) Q i ≥ det(1 − tF r ∗ | H iet ( X )) . This is in fact a consequence of the description of the integers ν m supplied by theLefschetz-Grothendieck trace formula ( cf. [22]) | X ( F q m ) | = X i ≥ ( − i tr(( F r m ) ∗ | H iet ( X )) . Originally, Grothendieck proposed a general framework for a so called categoryof numerically effective motives M ( k ) Q over a field k and with rational coeffi-cients. This category is defined by enlarging the category V k of smooth, projectivealgebraic varieties over k (and algebraic morphisms) by following the so-called pro-cedure of pseudo-abelian envelope. This construction is performed in two steps:at first one enlarges the set of morphisms of V k by including (rational) algebraic oncommutative geometry and motives X, Y ∈ Obj( V k ), oneworks with correspondences f : X Y which are elements of codimension equalto dim X in the rational graded algebra A ∗ ( X × Y ) = C ∗ ( X × Y ) ⊗ Q / ∼ num of algebraic cycles modulo numerical equivalence . We recall that two algebraiccycles on an algebraic variety X are said to be numerically equivalent Z ∼ num W ,if deg( Z · T ) = deg( W · T ) , (2.5)for any algebraic cycle T on X . Here, by deg( V ) we mean the degree of thealgebraic cycle V = P finite m α V α ∈ C ∗ ( X ).The degree defines a homomorphism from the free abelian group of algebraic cycles C ∗ ( X ) = ⊕ i C i ( X ) to the integers. On the components C i ( X ), the map is definedas follows deg : C i ( X ) → Z , deg( V ) = (P α m α if i = dim X, i < dim X. The symbol ‘ · ’ in (2.5) refers to the intersection product structure on C ∗ ( X ),which is well-defined under the assumption of proper intersection. If Z ∩ T is proper( i.e. codim( Z ∩ T ) = codim( Z ) + codim( T )), then intersection theory supplies thedefinition of an intersection cycle Z · T ∈ C ∗ ( X ). Moreover, the intersectionproduct is commutative and associative whenever is defined.Passing from the free abelian group C ∗ ( X ) to the quotient C ∗ ( X ) / ∼ , moduloa suitable equivalence relation on cycles, allows one to use classical results ofalgebraic geometry (so called Moving Lemmas) which lead to the definition of aring structure. One then defines intersection cycle classes in general, even whencycles do not intersect properly, by intersecting equivalent cycles which fulfill therequired geometric property of proper intersection.It is natural to guess that the use of the numerical equivalence in the original defi-nition of the category of motives was motivated by the study of classical construc-tions in enumerative geometry, such as for example the computation of the numberof the rational points of an algebraic variety defined over a finite field. One of themain original goals was to show that for a suitable definition of an equivalencerelation on algebraic cycles, the corresponding category of motives is semi-simple .This means that the objects M in the category decompose, following the rules of a theory of weights ( cf. section 2.3), into direct factors M = ⊕ i M i ( X ), with M i ( X )simple ( i.e. indecomposable) motives associated to smooth, projective algebraicvarieties. The importance of achieving such a result is quite evident if one seeks,for example, to understand categorically the decomposition H ∗ ( X ) = ⊕ i H i ( X ) incohomology, or if one wants to recognize the role of motives in the factorization ofzeta-functions of algebraic varieties. Caterina Consani
Grothendieck concentrated his efforts on the numerical equivalence relation whichis the coarsest among the equivalence relations on algebraic cycles. So doing, heattacked the problem of the semi-simplicity of the category of motives from theeasiest side. However, despite a promising departing point, the statement on thesemi-simplicity escaped all his efforts. In fact, the result he was able to reachat that time was dependent on the assumption of the
Standard Conjectures , twostrong topological statements on algebraic cycles. The proof of the semi-simplicityof the category of motives for numerical equivalence (the only equivalence relationproducing this result) was achieved only much later on in the development of thetheory ( cf. [24]). The proof found by U. Jannsen uses a fairly elementary butingenious idea which mysteriously eluded Grothendieck’s intuition as well as allthe mental grasps of several mathematicians after him.By looking at the construction of (pure) motives in perspective, one immedi-ately recognizes the predominant role played by the morphisms over the objects,in the category M ( k ) Q . This was certainly a great intuition of Grothendieck.This idea led to a systematic study of the properties of algebraic cycles and theirdecomposition by means of algebraic projectors , that is algebraic cycles classes p ∈ A dim X ( X × X ) satisfying the property p = p ◦ p = p. Notice that in order to make sense of the notion of a projector and more in general,in order to define a law of composition ‘ ◦ ’ on algebraic correspondences, one needsto use the ring structure on the graded algebra A ∗ . The operation ‘ ◦ ’ is defined asfollows. Let us assume for simplicity, that the algebraic varieties are connected (thegeneral case can be easily deduced from this). Then, two algebraic correspondences f ∈ A dim X + i ( X × X ) (of degree i ) and f ∈ A dim X + j ( X × X ) (of degree j )compose accordingly to the following rule (bi-linear, associative) A dim X + i ( X × X ) × A dim X + j ( X × X ) → A dim X + i + j ( X × X )( f , f ) f ◦ f = ( p ) ∗ (( p ∗ ( f ) · ( p ) ∗ ( f ))) . In the particular case of projectors p : X X , one is restricted, in order tomake a sense of the condition p ◦ p = p , to use only particular types of algebraiccorrespondences: namely those of degree zero. These are the elements of theabelian group A dim X ( X × X ).The objects of the category M ( k ) Q are then pairs ( X, p ), with X ∈ Obj( V k )and p a projector. This way, one attains the notion of a Q -linear, pseudo-abelian,monoidal category (the ⊗ -monoidal structure is deduced from the cartesian prod-uct of algebraic varieties), together with the definition of a contravariant functor h : V k → M ( k ) Q , X h ( X ) = ( X, id) . Here ( X, id) denotes the motive associated to X and id means the trivial ( i.e. identity) projector associated to the diagonal ∆ X . More in general, ( X, p ) refersto the motive ph ( X ) that is cut-off on h ( X ) by the (range of the) projector p : X X . Notice that images of projectors are formally included among theobjects of M ( k ) Q , by the procedure of the pseudo-abelian envelope. The cut-off oncommutative geometry and motives p on the space determines a corresponding operation incohomology (for any classical Weil theory), by singling out the sub-vector space pH ∗ ( X ) ⊂ H ∗ ( X ).The category M ( k ) Q has two important basic objects: and L . is the unitmotive = (Spec( k ) , id) = h (Spec( k )) . This is defined by the zero-dimensional algebraic variety associated to a point,whereas L = ( P , π ) , π = P × { P } , P ∈ P ( k )is the so-called Lefschetz motive . This motive determines, jointly with , a decom-position of the motive associated to the projective line P h ( P ) = ⊕ L . (2.6)One can show that the algebraic cycles P × { P } and { P } × P on P × P donot depend on the choice of the rational point P ∈ P ( k ) and that their sum isa cycle equivalent to the diagonal. This fact implies that the decomposition (2.6)is canonical. More in general, it follows from the K¨unneth decomposition of thediagonal ∆ in P n × P n by algebraic cycles ∆ = π + · · · + π n , ( cf. [31] and [18]for the details) that the motive of a projective space P n decomposes into pieces(simple motives) h ( P n ) = h ( P n ) ⊕ h ( P n ) · · · ⊕ h n ( P n ) (2.7)where h i ( P n ) = ( P n , π i ) = ( h ( P n )) ⊗ i , ∀ i >
0. It is precisely this decompositionwhich implies the decomposition (2.3) on the rational points, when k = F q !For (irreducible) curves, and in the presence of a rational point x ∈ C ( k ), oneobtains a similar decomposition (non canonical) h ( C ) = h ( C ) ⊕ h ( C ) ⊕ h ( C )with h ( C ) = ( C, π = { x } × C ), h ( C ) = ( C, π = C × { x } ) and h ( C ) =( C, − π − π ). This decomposition is responsible for the formula (2.4).In fact, one can prove that these decompositions partially generalize to anyobject X ∈ Obj( V k ). In the presence of a rational point, or more in general bychoosing a positive zero-cycle Z = P α m α Z α ∈ C dim X ( X × X ) (here X is assumedirreducible for simplicity and dim X = d ), one constructs two rational algebraiccycles π = 1 m ( Z × X ) , π d = 1 m ( X × Z ); m = deg( Z ) = X α m α > π , π d in the Chow group CH d ( X × X ) ⊗ Q ofrational algebraic cycles modulo rational equivalence. The corresponding classesin A d ( X × X ) (if not zero) determine two motives ( X, π ) ≃ h ( X ) and ( X, π d ) ≃ h d ( X ) ( cf. e.g. [34]).For the applications, it is convenient to enlarge the category of effective motivesby formally adding the tensor product inverse L − of the Lefschetz motive: oneusually refers to it as to the Tate motive . It corresponds, from the more refined0
Caterina Consani point of view of Galois theory, to the cyclotomic characters. This enlargementof M ( k ) Q by the so-called “virtual motives” produces an abelian, semi-simplecategory M k ( V k ) Q of pure motives for numerical equivalence. The objects of thiscategory are now triples ( X, p, m ), with m ∈ Z . Effective motives are of courseobjects of this category and they are described by triples ( X, p, L = (Spec( k ) , id , − L − = (Spec( k ) , id ,
1) and is therefore reminissent of (infact induces) the notion of Tate structure Q (1) in Hodge theory.In the category M k ( V k ) Q , the set of morphisms connecting two motives ( X, p, m )and (
Y, q, n ) is defined byHom((
X, p, m ) , ( Y, q, n )) = q ◦ A dim X − m + n ( X × Y ) ◦ p. In particular, ∀ f = f ∈ End((
X, p, m )), one defines the two motivesIm( f ) = ( X, p ◦ f ◦ p, m ) , Ker( f ) = ( X, p − f, m ) . These determine a canonical decomposition of any virtual motive as Im( f ) ⊕ Im(1 − f ) ∼ → ( X, p, m ), where the direct sum of two motives as the above ones is definedby the formula (
X, p, m ) ⊕ ( Y, q, m ) = ( X a Y, p + q, m ) . The general definition of the direct sum of two motives requires a bit more offormalism which escapes this short overview: we refer to op.cit. for the details.The tensor structure(
X, p, m ) ⊗ ( Y, q, n ) = ( X × Y, p ⊗ q, m + n )and the involution ( i.e. auto-duality) which is defined, for X irreducible anddim X = d by the functor ∨ : ( M k ( V k ) Q ) op → M k ( V k ) Q , ( X, p, m ) ∨ = ( X, p t , d − m )(the general case follows from this by applying additivity), determine the structureof a rigid ⊗ - category on M k ( V k ) Q . Here, p t denotes the transpose correspondenceassociated to p ( i.e. the transpose of the graph). One finds, for example, that L ∨ = L − . In the particular case of the effective motive h ( X ), with X irreducibleand dim X = d , this involution determines the notion of Poincar´e duality h ( X ) ∨ = h ( X ) ⊗ L ⊗ ( − d ) that is an auto-duality which induces the Poincar´e duality isomorphism in anyclassical cohomological theory. A category of pure motives over a field k and with coefficients in a field K (ofcharacteristic zero) is supposed to satisfy, to be satisfactory, several basic proper-ties and to be endowed with a few fundamental structures. In the previous sectionwe have described the historically-first example of a category of pure motives and oncommutative geometry and motives K = Q ). One natu-rally wonders about the description of others categories of motives associated tofiner (than the numerical) equivalence relations on algebraic cycles: namely thecategories of motives for homological or rational or algebraic equivalence relations.However, if one seeks to work with a semi-simple category, the afore mentionedresult of Jannsen tells us that the numerical equivalence is the only adequate re-lation. The semi-simplicity property is also attaint if one assumes Grothendieck’sStandard Conjectures. Following the report of Grothendieck in [20], these con-jectures arose from the hope to understand (read prove) the conjectures of Weilon the zeta-function of an algebraic variety defined over a finite field. It waswell-known to Grothendieck that the Standard Conjectures imply the Weil’s Con-jectures. These latter statements became a theorem in the early seventies (1974),only a few years later the time when Grothendieck stated the Standard Conjec-tures (1968-69). The proof by Deligne of the Weil’s conjectures, however, doesnot make any use of the Standard Conjectures, these latter questions remain stillunanswered at the present time. The moral lesson seems to be that geometrictopology and the theory of algebraic cycles govern in many central aspects thefoundations of algebraic geometry.The Standard Conjectures of “Lefschetz type” and of “Hodge type” are statedin terms of algebraic cycles modulo homological equivalence ( cf. [26]). They implytwo further important conjectures. One of these states the equality of the homo-logical and the numerical equivalence relations, the other one, of “K¨unneth type”claims that the K¨unneth decomposition of the diagonal in cohomology, can bedescribed by means of (rational) algebraic cycles. It is generally accepted, nowa-days, to refer to the full set of the four conjectures, when one quotes the StandardConjectures. In view of their expected consequences, one is naturally led to studya category of pure motives for homological equivalence . In fact, there are severalcandidates for this category since the definition depends upon the choice of a Weilcohomological theory ( i.e. Betti, ´etale, de-Rham, crystalline, etc) with coefficientsin a field K of characteristic zero.Let us fix a cohomological theory X H ∗ ( X ) = H ∗ ( X, K ) for algebraicvarieties in the category V k . Then, the construction of the corresponding categoryof motives M homk ( V k ) K for homological equivalence is given following a proceduresimilar to that we have explained earlier on in this paper for the category of motivesfor numerical equivalence and with rational coefficients. The only difference is thatnow morphisms in the category M homk ( V k ) K are defined by means of algebraiccorrespondences modulo homological equivalence. At this point, one makes explicituse of the axiom “cycle map” that characterizes (together with finiteness, Poincar´eduality, K¨unneth formula, cycle map, weak and strong Lefschetz theorems) anyWeil cohomological theory ( cf. [26]).The set of algebraic morphisms connecting objects in the category V k is enlarged byincluding multi-valued maps X Y that are defined as a K -linear combinationof elements of the vector spaces C ∗ ( X × Y ) ⊗ Z F/ ∼ hom , Caterina Consani where F ⊂ K is a subfield. Two cycles Z, W ∈ C ∗ ( X × Y ) ⊗ F are homologicallyequivalent Z ∼ hom W if their image, by means of the cycle class map γ : C ∗ ( X × Y ) ⊗ F → H ∗ ( X × Y )is the same. This leads naturally to the definition of a subvector-space A ∗ hom ( X × Y ) ⊂ H ∗ ( X × Y ) generated by the image of the cycle class map γ . Thesespaces define the correspondences in the category M homk ( V k ) K . If X is purely d -dimensional, then Corr r ( X, Y ) := A d + rhom ( X × Y ) . In general, if X decomposes into several connected components X = ` i X i , onesets Corr r ( X, Y ) = ⊕ i Corr r ( X i , Y ). In direct analogy to the construction ofcorrespondences for numerical equivalence, the ring structure (“cap-product”) incohomology determines a composition law ‘ ◦ ’ among correspondences.The category M homk ( V k ) K is then defined as follows: the objects are triples M = ( X, p, m ), where X ∈ Obj( V k ), m ∈ Z and p = p ∈ Corr ( X, X ) is anidempotent. The collection of morphisms between two motives M = ( X, p, m ), N = ( Y, q, n ) is given by the setHom(
M, N ) = q ◦ Corr n − m ( X, Y ) ◦ p. This procedure determines a pseudo-abelian, K -linear tensor category. The tensorlaw is given by the formula( X, p, m ) ⊗ ( Y, q, n ) = ( X × Y, p × q, m + n ) . The commutativity and associativity constraints are induced by the obvious iso-morphisms X × Y ∼ → Y × X , X × ( Y × Z ) ∼ → ( X × Y ) × Z . The unit object in thecategory is given by = (Spec( k ) , id , M homk ( V k ) K is a rigid category, as it is endowed with an auto-duality functor ∨ : M homk ( V k ) K → ( M homk ( V k ) K ) op . For any object M , the functor − ⊗ M ∨ is left-adjoint to − ⊗ M and M ∨ ⊗ − isright-adjoint to M ⊗ − . In the case of an irreducible variety X , the internal Homis defined by the motiveHom(( X, p, m ) , ( Y, q, n )) = ( X × Y, p t × q, dim( X ) − m + n ) . The Standard Conjecture of K¨unneth type (which is assumed from now onin this section) implies that the K¨unneth components of the diagonal π iX ∈ Corr ( X, X ) determine a complete system of orthogonal, central idempotents.This important statement implies that the motive h ( X ) ∈ M homk ( V k ) K has theexpected direct sum decomposition (unique) h ( X ) = X M i =0 h i ( X ) , h i ( X ) = π iX h ( X ) . The cohomology functor X H ∗ ( X ) factors through the projection h ( X ) → h i ( X ). More in general, one shows that every motive M = ( X, p, m ) gets this way oncommutative geometry and motives Z - grading structure by setting( X, p, m ) r = ( X, p ◦ π r +2 m , r ) . (2.8)This grading is respected by all morphisms in the category and defines the structureof a graduation by weights on the objects. On a motive M = ( X, p, m ) = ph ( X ) ⊗ L ⊗ ( − m ) = ph ( X )( m ) in the category, one sets M = ⊕ i Gr wi ( M ) , Gr wi ( M ) = ph m + i ( X )( m )where Gr wi ( M ) is a pure motive of weight i . One finds for example, that hasweight zero, L = (Spec( k ) , , −
1) has weight 2 and that L − = (Spec( k ) , ,
1) hasweight −
2. More in general, the motive M = ( X, p, m ) = ph ( X )( m ) has weight − m .In order to achieve further important properties, one needs to modify the nat-ural commutativity constraint ψ = ⊕ r,s ψ r,s , ψ r,s : M r ⊗ N s ∼ → N s ⊗ M r , bydefining ψ new = ⊕ r,s ( − rs ψ r,s . (2.9)We shall denote by ˜ M homk ( V k ) K the category of motives for homological equiva-lence in which one has implemented the modification (2.9) on the tensor productstructure.An important structure on a category of pure motives (for homological equiv-alence) is given by assigning to an object X ∈ Obj( M homk ( V k ) K ) a motivic co-homology H imot ( X ). H imot ( X ) is a pure motive of weight i . This way, one viewspure motives as a universal cohomological theory for algebraic varieties. The mainproperty of the motivic cohomology is that it defines a universal realization ofany given Weil cohomology theory H ∗ . Candidates for these motivic cohomologytheories have been proposed by A. Beilinson [3], in terms of eigenspaces of Adamsoperations in algebraic K -theory i.e. H j − nmot ( X, Q ( j )) = K n ( X ) ( j ) and by S. Bloch[5], in terms of higher Chow groups i.e. H j − nmot ( X, Q ( j )) = CH j ( X, n ) ⊗ Q .The assignment of a Weil cohomological theory with coefficients in a field K which contains an assigned field F is equivalent to the definition of an exact real-ization ⊗ -functor of ˜ M homk ( V k ) K in the category of K -vector spaces r H ∗ : ˜ M homk ( V k ) K → Vect K , r H ∗ ( H imot ( X )) ≃ H i ( X ) . (2.10)In particular, one obtains the realization r H ∗ ( L − ) = H ( P ) which defines thenotion of the Tate twist in cohomology. More precisely:– in ´etale cohomology: H ( P ) = Q ℓ ( − Q ℓ (1) := lim ←− m µ ℓ m is a Q ℓ -vectorspace of dimension one endowed with the cyclotomic action of the absolute Galoisgroup G k = Gal(¯ k/k ). The “twist” (or torsion) ( r ) in ´etale cohomology corre-sponds to the torsion in Galois theory defined by the r -th power of the cyclotomiccharacter (Tate twist)– in de-Rham theory: H DR ( P ) = k , with the Hodge filtration defined by F ≤ = 0, F > = k . Here, the effect of the torsion ( r ) is that of shifting the Hodge filtrationof − r -steps (to the right)4 Caterina Consani – in Betti theory: H ( P ) = Q ( −
1) := (2 πi ) − Q . The bi-graduation on H ( P ) ⊗ C ≃ C is purely of type (1 , r ) is here identified with the compositeof a homothety given by a multiplication by (2 πi ) − r followed by a shifting by( − r, − r ) of the Hodge bi-graduation.Using the structure of rigid tensor-category one introduces the notion of rank associated to a motive M = ( X, p, m ) in ˜ M homk ( V k ) K . The rank of M is definedas the trace of id M i.e. the trace of the morphism ǫ ◦ ψ new ◦ η ∈ End( ), where ǫ : M ⊗ M ∨ → , η : → M ∨ ⊗ M are resp. the evaluation and co-evaluation morphisms satisfying ǫ ⊗ id M ◦ id M ⊗ η =id M , id M ∨ ⊗ ǫ ◦ η ⊗ id M ∨ = id M ∨ . In general, one setsrk( X, p, m ) = X i ≥ dim pH i ( X ) ≥ . Under the assumption of the Standard Conjectures (more precisely under the as-sumption that homological and numerical equivalence relations coincide) and thatEnd( ) = F (char( F ) = 0), the tannakian formalism invented by Grothendieckand developed by Saavedra [33], and Deligne [16] implies that the abelian, rigid,semi-simple tensor category ˜ M homk ( V k ) K is endowed with an exact, faithful ⊗ -fibrefunctor to the category of graded K -vector spaces ω : ˜ M homk ( V k ) K → VectGr K , ω ( H ∗ mot ( X )) = H ∗ ( X ) (2.11)which is compatible with the realization functor. This formalism defines a tan-nakian ( neutral if K = F ) structure on the category of motives. One then intro-duces the tannakian group G = Aut ⊗ ( ω )as a K -scheme in affine groups. Through the tannakian formalism one shows thatthe fibre functor ω realizes an equivalence of rigid tensor categories ω : ˜ M homk ( V k ) K ∼ → Rep F ( G ) , where Rep F ( G ) denotes the rigid tensor category of finite dimensional, F repre-sentations of the tannakian group G . This way, one establishes a quite useful dic-tionary between categorical ⊗ -properties and properties of the associated groups.Because we have assumed all along the Standard Conjectures, the semi-simplicityof the category ˜ M homk ( V k ) K implies that G is an algebraic, pro-reductive group,that is G is the projective limit of reductive F -algebraic groups.The tannakian theory is a linear analog of the theory of finite, ´etale coverings ofa given connected scheme. This theory was developed by Grothendieck in SGA1(theory of the pro-finite π ). For this reason the group G is usually referred to asthe motivic Galois group associated to V k and H ∗ . In the case of algebraic varietiesof dimension zero ( i.e. for Artin motives) the tannakian group G is nothing butthe (absolute) Galois group Gal(¯ k/k ).In any reasonable cohomological theory the functors X H ∗ ( X ) are deducedby applying standard methods of homological algebra to the related derived func-tors X R Γ( X ) which associate to an object in V k a bounded complex of k -vector oncommutative geometry and motives D ( k ) of complexes of modules over k ,whose heart is the category of motives. This is the definition of cohomology as H i ( X ) = H i R Γ( X ) . Under the assumption that the functors R Γ are realizations of corresponding mo-tivic functors i.e. R Γ = r H ∗ R Γ mot , one expects the existence of a (non-canonical)isomorphism in D ( k ) R Γ mot ( X ) ≃ ⊕ i H imot ( X )[ − i ] . (2.12)Moreover, the introduction of the motivic derived functors R Γ mot suggests thedefinition of the following groups of absolute cohomology H iabs ( X ) = Hom D ( k ) ( , R Γ mot ( X )[ i ]) . For a general motive M = ( X, p, m ), one defines H iabs ( M ) = Hom D ( k ) ( , M [ i ]) = Ext i ( , M ) . (2.13)The motives H imot ( X ) and the groups of absolute motivic cohomology are relatedby a spectral sequence E p,q = H pabs ( H qmot ( X )) ⇒ H p + qabs ( X ) . The first interesting examples of pure motives arise by considering the category V ok of ´etale, finite k -schemes. An object in this category is a scheme X = Spec( k ′ ),where k ′ is a commutative k -algebra of finite dimension which satisfies the followingproperties. Let ¯ k denote a fixed separable closure of k k ′ ⊗ ¯ k ≃ ¯ k [ k ′ : k ] k ′ ≃ Q k α , for k α /k finite, separable field extensions3. | X (¯ k ) | = [ k ′ : k ].The corresponding rigid, tensor-category of motives with coefficients in a field K is usually referred to as the category of Artin motives : CV o ( k ) K .The definition of this category is independent of the choice of the equivalence rela-tion on cycles as the objects of V ok are smooth, projective k -varieties of dimensionzero. One also sees that passing from V ok to CV o ( k ) K requires adding new objectsin order to attain the property that the category of motives is abelian. One canverify this already for k = K = Q , by considering the real quadratic extension k ′ = Q ( √
2) and the one-dimensional non-trivial representation of G Q = Gal( ¯ Q / Q )that factors through the character of order two of Gal( k ′ / Q ). This representationdoes not correspond to any object in V o Q , but can be obtained as the image ofthe projector p = (1 − σ ), where σ is the generator of Gal( k ′ / Q ). Therefore,image( p ) ∈ Obj( CV o ( Q ) Q ) is a new object.6 Caterina Consani
The category of Artin motives is a semi-simple, K -linear, monoidal ⊗ -category.When char( K ) = 0, the commutative diagram of functors V ok GG −−−−→ ∼ { sets with Gal(¯ k/k )-continuous action } y h y l CV o ( k ) K ( ∗ ) −−−−→ ∼ { finite dim. K -v.spaces with linear Gal(¯ k/k )-continuous action } where l is the contravariant functor of linearization S K S , ( g ( f ))( s ) = f ( g − ( s )) , ∀ g ∈ Gal(¯ k/k )determines a linearization of the Galois-Grothendieck correspondence (GG) bymeans of the equivalence of categories ( ∗ ). This is provided by the fiber functor ω : X → H ( X ¯ k , K ) = K X (¯ k ) and by applying the tannakian formalism. It follows that CV o ( k ) K is ⊗ -equivalentto the category Rep K Gal(¯ k/k ) of representations of the absolute Galois group G k = Gal(¯ k/k ).These results were the departing point for Grothendieck’s speculations on the def-inition of higher dimensional Galois theories ( i.e. Galois theories associated to sys-tem of polynomials in several variables) and for the definition of the correspondingmotivic Galois groups.
The notion of an endomotive in noncommutative geometry ( cf. [7]) is the naturalgeneralization of the classical concept of an Artin motive for the noncommuta-tive spaces which are defined by semigroup actions on projective limits of zero-dimensional algebraic varieties, endowed with an action of the absolute Galoisgroup. This notion applies quite naturally for instance, to the study of several ex-amples of quantum statistical dynamical systems whose time evolution describesimportant number-theoretic properties of a given field k ( cf. op.cit , [14]).There are two distinct definitions of an endomotive: one speaks of algebraic oranalytic endomotives depending upon the context and the applications.When k is a number field, there is a functor connecting the two related cate-gories. Moreover, the abelian category of Artin motives embeds naturally as a fullsubcategory in the category of algebraic endomotives ( cf. Theorem 5.3) and thisresult motivates the statement that the theory of endomotives defines a naturalgeneralization of the classical theory of (zero-dimensional) Artin motives.In noncommutative geometry, where the properties of a space (frequently highlysingular from a classical viewpoint) are analyzed in terms of the properties of theassociated noncommutative algebra and its (space of) irreducible representations,it is quite natural to look for a suitable abelian category which enlarges the original, oncommutative geometry and motives cf. (2.12)), the construction ofa universal (co)homological theory representing in this context the absolute mo-tivic cohomology ( cf. (2.13)) and possibly also the set-up of a noncommutativetannakian formalism to motivate in rigorous mathematical terms the presence ofcertain universal groups of symmetries associated to renormalizable quantum fieldtheories ( cf. e.g. [12]).A way to attack these problems is that of enlarging the original category of al-gebras and morphisms by introducing a “derived” category of modules enrichedwith a suitable notion of correspondences connecting the objects that should alsoaccount for the structure of Morita equivalence which represents the noncommu-tative generalization of the notion of isomorphism for commutative algebras.
The sought for enlargement of the category
Alg k of (unital) k -algebras and (unital)algebra homomorphisms is defined by introducing a new category Λ k of cyclic k (Λ)-modules. The objects of this category are modules over the cyclic category Λ. This latter has the same objects as the simplicial category ∆ (Λ contains ∆ assub-category). We recall that an object in ∆ is a totally ordered set[ n ] = { < < . . . < n } for each n ∈ N , and a morphism f : [ n ] → [ m ]is described by an order-preserving map of sets f : { , , . . . , n } → { , , . . . , m } .The set of morphisms in ∆ is generated by faces δ i : [ n − → [ n ] (the injectionthat misses i ) and degeneracies σ j : [ n + 1] → [ n ] (the surjection which identifies j with j + 1) which satisfy several standard simplicial identities ( cf. e.g. [9]). Theset of morphisms in Λ is enriched by introducing a new collection of morphisms:the cyclic morphisms . For each n ∈ N , one sets τ n : [ n ] → [ n ]fulfilling the relations τ n δ i = δ i − τ n − ≤ i ≤ n, τ n δ = δ n τ n σ i = σ i − τ n +1 ≤ i ≤ n, τ n σ = σ n τ n +1 τ n +1 n = 1 n . (3.1)8 Caterina Consani
The objects of the category Λ k are k -modules over Λ ( i.e. k (Λ)-modules). Incategorical language this means functorsΛ op → M od k ( M od k = category of k -modules). Morphisms of k (Λ)-modules are therefore nat-ural transformations between the corresponding functors.It is evident that Λ k is an abelian category, because of the interpretation of amorphism in Λ k as a collection of k -linear maps of k -modules A n → B n ( A n , B n ∈ Obj(
M od k )) compatible with faces, degeneracies and cyclic operators. Kernelsand cokernels of these morphisms define objects of the category Λ k , since theirdefinition is given point-wise.To an algebra A over a field k , one associates the k (Λ)-module A ♮ . For each n ≥ A ♮n = A ⊗ A ⊗ · · · ⊗ A | {z } (n+1)-times . The cyclic morphisms on A ♮ correspond to the cyclic permutations of the tensors,while the face and the degeneracy maps correspond to the algebra product ofconsecutive tensors and the insertion of the unit. This construction determines afunctor ♮ : Alg k → Λ k Traces ϕ : A → k give rise naturally to morphisms ϕ ♮ : A ♮ → k ♮ , ϕ ♮ ( a ⊗ · · · ⊗ a n ) = ϕ ( a · · · a n )in Λ k . The main result of this construction is the following canonical descriptionof the cyclic cohomology of an algebra A over a field k as the derived functor ofthe functor which assigns to a k (Λ)-module its space of traces HC n ( A ) = Ext n ( A ♮ , k ♮ ) (3.2)( cf. [9],[29]). This formula is the analog of (2.13), that describes the absolutemotivic cohomology group of a classical motive as an Ext-group computed in atriangulated category of motives D ( k ). In the present context, on the other hand,the derived groups Ext n are taken in the abelian category of Λ k -modules.The description of the cyclic cohomology as a derived functor in the cyclic categorydetermines a useful procedure to embed the nonadditive category of algebras andalgebra homomorphisms in the “derived” abelian category of k (Λ)-modules. Thisconstruction provides a natural framework for the definition of the objects of acategory of noncommutative motives.Likewise in the construction of the category of motives, one is faced with theproblem of finding the “motivated maps” connecting cyclic modules. The nat-ural strategy is that of enlarging the collection of cyclic morphisms which arefunctorially induced by homomorphisms between (noncommutative) algebras, byimplementing an adequate definition of (noncommutative) correspondences. Thenotion of an algebraic correspondence in algebraic geometry, as a multi-valuedmap defined by an algebraic cycle modulo a suitable equivalence relation, has here oncommutative geometry and motives KK -theory ( cf. [25]). Likewise in classical motive theory, one may prefer to work with(compare) several versions of correspondences. One may decide to retain the fullinformation supplied by a group action on a given algebra ( i.e. a noncommuta-tive space) rather than partially loosing this information by moding out with theequivalence relation (homotopy in KK -theory). KK -theory There is a natural way to associate a cyclic morphism to a (virtual) correspon-dence and hence to a class in KK -theory. Starting with the category of separable C ∗ -algebras and ∗ -homomorphisms, one enlarges the collection of morphisms con-necting two unital algebras A and B , by including correspondences defined byelements of Kasparov’s bivariant K -theoryHom( A , B ) = KK ( A , B )([25], cf. also § § E = E ( A , B ) = ( E, φ, F )which satisfy the following conditions:– E is a countably generated Hilbert module over B – φ is a ∗ -homomorphism of A to bounded linear operators on E ( i.e. φ gives E the structure of an A - B bimodule)– F is a bounded linear operator on E such that the operators [ F, φ ( a )], ( F − φ ( a ), and ( F ∗ − F ) φ ( a ) are compact for all a ∈ A .A Hilbert module E over B is a right B -module with a positive , B -valued innerproduct which satisfies h x, yb i = h x, y i b , ∀ x, y ∈ E and ∀ b ∈ B , and with respectto which E is complete ( i.e. complete in the norm k x k = p k h x, x i k ).Notice that Kasparov bimodules are Morita-type of correspondences. They gener-alize ∗ -homomorphisms of C ∗ -algebras since the latter ones may be re-interpretedas Kasparov bimodules of the form ( B , φ, E = E ( A , B ), that is a A - B Hilbert bimodule E as defined above, one associates, under the assumption that E is a projective B -module of finite type ( cf. [7], Lemma 2.1), a cyclic morphism E ♮ ∈ Hom( A ♮ , B ♮ ) . This result allows one to define an enlargement of the collection of cyclic mor-phisms in the category Λ k of k (Λ)-modules, by considering Kasparov’s projectivebimodules of finite type, as correspondences.One then implements the homotopy equivalence relation on the collection of Kas-parov’s bimodules. Two Kasparov’s modules are said to be homotopy equivalent( E , φ , F ) ∼ h ( E , φ , F ) if there is an element( E, φ, F ) ∈ E ( A , I B ) , I B = { f : [0 , → B | f continuous } Caterina Consani which performs a unitary homotopy deformation between the two modules. Thismeans that ( E ˆ ⊗ f i B , f i ◦ φ, f i ( F )) is unitarily equivalent to ( E i , φ i , F i ) or equiva-lently re-phrased, that there is a unitary in bounded operators from E ˆ ⊗ f i B to E i intertwining the morphisms f i ◦ φ and φ i and the operators f i ( F ) and F i . Here f i : I B → B is the evaluation at the endpoints.There is a binary operation on the set of all Kasparov A - B bimodules, givenby the direct sum. By definition, the group of Kasparov’s bivariant K -theory isthe set of homotopy equivalence classes c ( E ( A , B )) ∈ KK ( A , B ) of Kasparov’smodules E ( A , B ). This set has a natural structure of abelian group with additioninduced by direct sum.This bivariant version of K -theory is reacher than both K -theory and K -homology, as it carries an intersection product. There is a natural bi-linear, asso-ciative composition (intersection) product ⊗ B : KK ( A , B ) × KK ( B , C ) → KK ( A , C )for all A , B and C separable C ∗ -algebras. This product is compatible with compo-sition of morphisms of C ∗ -algebras. KK -theory is also endowed with a bi-linear, associative exterior product ⊗ : KK ( A , B ) ⊗ KK ( C , D ) → KK ( A ⊗ C , B ⊗ D ) , which is defined in terms on the composition product by c ⊗ c = ( c ⊗ C ) ⊗ B⊗ C ( c ⊗ B ) . A slightly different formulation of KK -theory, which simplifies the definition ofthis external tensor product is obtained by replacing in the data ( E, φ, F ) theoperator F by an unbounded , regular self-adjoint operator D . The corresponding F is then given by D (1 + D ) − / ( cf. [1]).The above construction which produces an enlargement of the category of sep-arable C ∗ -algebras by introducing correspondences as morphisms determines anadditive, although non abelian category KK ( cf. [4] § cf. [32]) and this result is in agreementwith the construction of the triangulated category D ( k ) in motives theory, whoseheart is expected to be the category of (mixed) motives ( cf. section 2.3 and [17]).A more refined analysis based on the analogy with the construction of a categoryof motives suggests that one should probably perform a further enlargement bypassing to the pseudo-abelian envelope of KK , that is by formally including amongthe objects also ranges of idempotents in KK -theory.In section 5 we will review the category of analytic endomotives where maps aregiven in terms of ´etale correspondences described by spaces Z arising from locallycompact ´etale groupoids G = G ( X α , S, µ ) associated to zero-dimensional, singularquotient spaces X (¯ k ) /S with associated C ∗ -algebras C ∗ ( G ). In view of what wehave said in this section, it would be also possible to define a category wheremorphisms are given by classes c ( Z ) ∈ KK ( C ∗ ( G ) , C ∗ ( G ′ )) which describe sets ofequivalent triples ( E, φ, F ), where (
E, φ ) is given in terms of a bimodule M Z withthe trivial grading γ = 1 and the zero endomorphism F = 0. The definition of oncommutative geometry and motives KK -theory) is particularly easy in the zero-dimensional case because the equiv-alence relations play no role. Of course, it would be quite interesting to investigatethe higher dimensional cases, in view of a unified framework for motives and non-commutative spaces which is suggested for example, by the recent results on theLefschetz trace formula for archimedean local factors of L -functions of motives ( cf. [7], Section 7).A way to attack this problem is by comparing the notion of a correspondencegiven by an algebraic cycle with the notion of a geometric correspondence usedin topology ( cf. [2], [13]). For example, it is easy to see that the definition of analgebraic correspondence can be reformulated as a particular case of the topological(geometrical) correspondence and it is also known that one may associate to thelatter a class in KK -theory. In the following two sections, we shall review andcomment on these ideas. In geometric topology, given two smooth manifolds X and Y (it is enough toassume that X is a locally compact parameter space), a topological (geometric)correspondence is given by the datum X f X ← ( Z, E ) g Y → Y where:– Z is a smooth manifold– E is a complex vector bundle over Z – f X : Z → X and g Y : Z → Y are continuous maps, with f X proper and g Y K -oriented (orientation in K -homology).Unlike in the definition of an algebraic correspondence ( cf. section 2.3) one doesnot require that Z is a subset of the cartesian product X × Y . This flexibility isbalanced by the implementation of the extra piece of datum given by the vectorbundle E . To any such correspondence ( Z, E, f X , g Y ) one associates a class inKasparov’s K-theory c ( Z, E, f X , g Y ) = ( f X ) ∗ (( E ) ⊗ Z ( g Y )!) ∈ KK ( X, Y ) . (3.3)( E ) denotes the class of E in KK ( Z, Z ) and ( g Y )! is the element in KK -theorywhich fulfills the Grothendieck Riemann-Roch formula.We recall that given two smooth manifolds X and X and a continuous orientedmap f : X → X , the element f ! ∈ KK ( X , X ) determines the GrothendieckRiemann–Roch formula ch( F ⊗ f !) = f ! (Td( f ) ∪ ch( F )) , (3.4)2 Caterina Consani for all F ∈ K ∗ ( X ), with Td( f ) the Todd genusTd( f ) = Td( T X ) / Td( f ∗ T X ) . (3.5)The composition of two correspondences ( Z , E , f X , g Y ) and ( Z , E , f Y , g W ) isgiven by taking the fibered product Z = Z × Y Z and the bundle E = π ∗ E × π ∗ E ,with π i : Z → Z i the projections. This determines the composite correspondence( Z, E, f X , g W ). In fact, one also needs to assume a transversality condition onthe maps g Y and f Y in order to ensure that the fibered product Z is a smoothmanifold. The homotopy invariance of both g Y ! and ( f X ) ∗ show however that theassumption of transversality is ‘generically’ satisfied.Theorem 3.2 of [13] shows that Kasparov product in KK -theory ‘ ⊗ ’ agrees withthe composition of correspondences, namely c ( Z , E , f X , g Y ) ⊗ Y c ( Z , E , f Y , g W ) = c ( Z, E, f X , g W ) ∈ KK ( X, W ) . (3.6) K -theory In algebraic geometry, the notion of correspondence that comes closest to thedefinition of a geometric correspondence (as an element in KK -theory) is obtainedby considering classes of algebraic cycles in algebraic K -theory ( cf. [30]).Given two smooth and projective algebraic varieties X and Y , we denote by p X and p Y the projections of X × Y onto X and Y respectively and we assumethat they are proper . Let Z ∈ C ∗ ( X × Y ) be an algebraic cycle. For simplicity,we shall assume that Z is irreducible (the general case follows by linearity). Wedenote by f X = p X | Z and g Y = p Y | Z the restrictions of p X and p Y to Z .To the irreducible subvariety T i ֒ → Y one naturally associates the coherent O Y -module i ∗ O T . For simplicity of notation we write it as O T . We use a similarnotation for the coherent sheaf O Z , associated to the irreducible subvariety Z ֒ → X × Y . Then, the sheaf pullback p ∗ Y O T = p − Y O T ⊗ p − Y O Y O X × Y (3.7)has a natural structure of O X × Y -module. The map on sheaves that correspondsto the cap product by Z on cocycles is given by Z : O T ( p X ) ∗ (cid:0) p ∗ Y O T ⊗ O X × Y O Z (cid:1) . (3.8)Since p X is proper, the resulting sheaf is coherent. Using (3.7), we can writeequivalently Z : O T ( p X ) ∗ (cid:16) p − Y O T ⊗ p − Y O Y O Z (cid:17) . (3.9)We recall that the functor f ! is the right adjoint to f ∗ ( i.e. f ∗ f ! = id ) and that f ! satisfies the Grothendieck Riemann–Roch formulach( f ! ( F )) = f ! (Td( f ) ∪ ch( F )) . (3.10) oncommutative geometry and motives O T ⊗ O Y ( p Y ) ! O Z (3.11)and then applying p ∗ Y . Using (3.10) and (3.4) we know that we can replace (3.11)by O T ⊗ O Y ( O Z ⊗ ( p Y )!) with the same effect in K -theory.Thus, to a correspondence in the sense of (3.8) that is defined by the image in K -theory of an algebraic cycle Z ∈ C ∗ ( X × Y ) we associate the geometric class F ( Z ) = c ( Z, E, f X , g Y ) ∈ KK ( X, Y ) , with f X = p X | Z , g Y = p Y | Z and with the bundle E = O Z .The composition of correspondences is given in terms of the intersection productof the associated cycles. Given three smooth projective varieties X , Y and W and (virtual) correspondences U = P a i Z i ∈ C ∗ ( X × Y ) and V = P c j Z ′ j ∈ C ∗ ( Y × W ), with Z i ⊂ X × Y and Z ′ j ⊂ Y × W closed reduced irreduciblesubschemes, one defines U ◦ V = ( π ) ∗ (( π ) ∗ U • ( π ) ∗ V ) . (3.12) π : X × Y × W → X × Y , π : X × Y × W → Y × W , and π : X × Y × W → X × W denote, as usual, the projection maps.Under the assumption of ‘general position’ which is the algebraic analog of thetransversality requirement in topology, we obtain the following result Proposition 3.1 ( cf. [7] Proposition 6.1). Suppose given three smooth projectivevarieties X , Y , and W and algebraic correspondences U given by Z ⊂ X × Y and V given by Z ⊂ Y × W . Assume that ( π ) ∗ Z and ( π ) ∗ Z are in generalposition in X × Y × W . Then assigning to a cycle Z the topological correspondence F ( Z ) = ( Z, E, f X , g Y ) satisfies F ( Z ◦ Z ) = F ( Z ) ⊗ Y F ( Z ) , (3.13) where Z ◦ Z is the product of algebraic cycles and F ( Z ) ⊗ Y F ( Z ) is the Kasparovproduct of the topological correspondences. Notice that, while in the topological (smooth) setting transversality can always beachieved by a small deformation ( cf. § III, [13]), in the algebro-geometric frameworkone needs to modify the above construction if the cycles are not in general position.In this case the formula [ O T ] ⊗ [ O T ] = [ O T ◦ T ]which describes the product in K -theory in terms of the intersection product ofalgebraic cycles must be modified by implementing Tor-classes and one works witha product defined by the formula ([30], Theorem 2.7)[ O T ] ⊗ [ O T ] = n X i =0 ( − i h Tor O X i ( O T , O T ) i . (3.14)4 Caterina Consani
To define the category of algebraic endomotives one replaces the category V ok ofreduced, finite-dimensional commutative algebras (and algebras homomorphisms)over a field k by the category of noncommutative algebras (and algebras homo-morphisms) of the form A k = A ⋊ S.A denotes a unital algebra which is an inductive limit of commutative algebras A α ∈ Obj( V ok ). S is a unital, abelian semigroup of algebra endomorphisms ρ : A → A. Moreover, one imposes the condition that for ρ ∈ S , e = ρ (1) ∈ A is an idempotent of the algebra and that ρ is an isomorphism of A with the compressed algebra eAe .The crossed product algebra A k is defined by formally adjoining to A new gener-ators U ρ and U ∗ ρ , for ρ ∈ S , satisfying the algebraic rules U ∗ ρ U ρ = 1 , U ρ U ∗ ρ = ρ (1) , ∀ ρ ∈ S, (4.1) U ρ ρ = U ρ U ρ , U ∗ ρ ρ = U ∗ ρ U ∗ ρ , ∀ ρ j ∈ S,U ρ x = ρ ( x ) U ρ , xU ∗ ρ = U ∗ ρ ρ ( x ) , ∀ ρ ∈ S, ∀ x ∈ A. Since S is abelian, these rules suffice to show that A k is the linear span of themonomials U ∗ ρ aU ρ , for a ∈ A and ρ j ∈ S .Because A = lim −→ α A α , with A α reduced, finite-dimensional commutative algebrasover k , the construction of A k is in fact determined by assigning a projective system { X α } α ∈ I of varieties in V ok ( I is a countable indexing set), with ξ β,α : X β → X α morphisms in V ok and with a suitably defined action of S . Here, we have implicitlyused the equivalence between the category of finite dimensional commutative k -algebras and the category of affine algebraic varieties over k .The graphs Γ ξ β,α of the connecting morphisms of the system define G k = Gal(¯ k/k )-invariant subsets of X β (¯ k ) × X α (¯ k ) which in turn describe ξ β,α as algebraic corre-spondences. We denote by X = lim ←− α X α , ξ α : X → X α the associated pro-variety . The compressed algebra eAe associated to the idem-potent e = ρ (1) determines a subvariety X e ⊂ X which is in fact isomorphic to X , via the induced morphism ˜ ρ : X → X e .The noncommutative space defined by A k is the quotient of X (¯ k ) by the action of S , i.e. of the action of the ˜ ρ ’s.The Galois group G k acts on X (¯ k ) by composition. By identifying the elementsof X (¯ k ) with characters, i.e. with k -algebra homomorphisms χ : A → ¯ k , we writethe action of G k on A as α ( χ ) = α ◦ χ : A → ¯ k, ∀ α ∈ G k , ∀ χ ∈ X (¯ k ) . (4.2) oncommutative geometry and motives ρ , i.e. ( α ◦ χ ) ◦ ρ = α ◦ ( χ ◦ ρ ). Thus thewhole construction of the system ( X α , S ) is G k -equivariant. This fact does notmean however, that G k acts by automorphisms on A k !Moreover, notice that the algebraic construction of the crossed-product algebra A k endowed with the actions of G k and S on X (¯ k ) makes sense also when char( k ) > k ) = 0, one defines the set of correspondences M ( A k , B k ) by using thenotion of Kasparov’s bimodules E ( A k , B k ) which are projective and finite as rightmodules. This way, one obtains a first realization of the resulting category of non-commutative zero-dimensional motives in the abelian category of k (Λ)-modules.In general, given ( X α , S ), with { X α } α ∈ I a projective system of Artin motivesand S a semigroup of endomorphisms of X = lim ←− α X α as above, the datum of thesemigroup action is encoded naturally by the algebraic groupoid G = X ⋊ S. This is defined in the following way. One considers the Grothendieck group ˜ S ofthe abelian semigroup S . By using the injectivity of the partial action of S , onemay also assume that S embeds in ˜ S . Then, the action of S on X extends todefine a partial action of ˜ S . More precisely, for s = ρ − ρ ∈ ˜ S the two projections E ( s ) = ρ − ( ρ (1) ρ (1)) , F ( s ) = ρ − ( ρ (1) ρ (1))only depend on s and the map s : A E ( s ) → A F ( s ) defines an isomorphism ofreduced algebras. It is immediate to verify that E ( s − ) = F ( s ) = s ( E ( s )) andthat F ( ss ′ ) ≥ F ( s ) s ( F ( s ′ )). The algebraic groupoid G is defined as the disjointunion G = a s ∈ ˜ S X F ( s ) which corresponds to the commutative direct-sum of reduced algebras M s ∈ ˜ S A F ( s ) . The range and the source maps in G are given resp. by the natural projection from G to X and by its composition with the antipode S which is defined, at the algebralevel, by S ( a ) s = s ( a s − ) , ∀ s ∈ ˜ S . The composition in the groupoid correspondsto the product of monomials aU s bU t = as ( b ) U st .Given two systems ( X α , S ) and ( X ′ α ′ , S ′ ), with associated crossed-product al-gebras A k and B k and groupoids G = G ( X α , S ) and G ′ = G ( X ′ α ′ , S ′ ) a geometriccorrespondence is given by a ( G , G ′ )-space Z = Spec( C ), endowed with a rightaction of G ′ which fulfills the following ´etale condition. Given a space such as G ′ ,that is a disjoint union of zero-dimensional pro-varieties over k , a right action of G ′ on Z is given by a map g : Z → X ′ and a collection of partial isomorphisms z ∈ g − ( F ( s )) z · s ∈ g − ( E ( s )) (4.3)fulfilling the following rules for partial action of the abelian group ˜ Sg ( z · s ) = g ( z ) · s, z · ( ss ′ ) = ( z · s ) · s ′ on g − ( F ( s ) ∩ s ( F ( s ′ ))) . (4.4)6 Caterina Consani
Here x x · s denotes the partial action of ˜ S on X ′ . One checks that such anaction gives to the k -linear space C a structure of right module over B k . Theaction of G ′ on Z is ´etale if the corresponding module C is finite and projective over B k .Given two systems ( X α , S ) and ( X ′ α ′ , S ′ ) as above, an ´etale correspondence istherefore a ( G ( X α , S ) , G ( X ′ α ′ , S ′ ))-space Z such that the right action of G ( X ′ α ′ , S ′ )is ´etale.The Q -linear space of (virtual) correspondences Corr(( X α , S ) , ( X ′ α ′ , S ′ ))is the rational vector space of formal linear combinations U = P i a i Z i of ´etalecorrespondences Z i , modulo the relations arising from isomorphisms and equiva-lences: Z ` Z ′ ∼ Z + Z ′ . The composition of correspondences is given by the fiberproduct over a groupoid. Namely, for three systems ( X α , S ), ( X ′ α ′ , S ′ ), ( X ′′ α ′′ , S ′′ )joined by correspondences( X α , S ) ← Z → ( X ′ α ′ , S ′ ) , ( X ′ α ′ , S ′ ) ← W → ( X ′′ α ′′ , S ′′ ) , their composition is given by the rule Z ◦ W = Z × G ′ W (4.5)that is the fiber product over the groupoid G ′ = G ( X ′ α ′ , S ′ ).Finally, a system ( X α , S ) as above is said to be uniform if the normalizedcounting measures µ α on X α satisfy ξ β,α µ α = µ β . Definition 4.1.
The category EV o ( k ) K of algebraic endomotives with coefficientsin a fixed extension K of Q is the (pseudo)abelian category generated by the fol-lowing objects and morphisms. The objects are uniform systems M = ( X α , S ) ofArtin motives over k , as above. The set of morphisms in the category connectingtwo objects M = ( X α , S ) and M ′ = ( X ′ α ′ , S ′ ) is defined asHom ( M, M ′ ) = Corr (( X α , S ) , ( X ′ α ′ , S ′ )) ⊗ Q K. The category CV o ( k ) K of Artin motives embeds as a full sub-category in thecategory of algebraic endomotives ι : CV o ( k ) K → EV o ( k ) K . The functor ι maps an Artin motive M = X to the system ( X α , S ) with X α = X , ∀ α and S = { id } . The category of algebraic endomotives is inclusive of a large and general class ofexamples of noncommutative spaces A k = A ⋊ S which are described by semigroupactions on projective systems of Artin motives.One may consider, for instance a pointed algebraic variety ( Y, y ) over a field k and a countable, unital, abelian semigroup S of finite endomorphisms of ( Y, y ), oncommutative geometry and motives unramified over y ∈ Y . Then, there is a system ( X s , S ) of Artin motives over k which is constructed from these data. More precisely, for s ∈ S , one sets X s = { y ∈ Y : s ( y ) = y } . (4.6)For a pair s, s ′ ∈ S , with s ′ = sr , the connecting map ξ s,s ′ : X sr → X s is definedby X sr ∋ y r ( y ) ∈ X s . (4.7)This is an example of a system indexed by the semigroup S itself, with partialorder given by divisibility. One sets X = lim ←− s X s .Since s ( y ) = y , the base point y defines a component Z s of X s for all s ∈ S .The pre-image ξ − s,s ′ ( Z s ) in X s ′ is a union of components of X s ′ . This defines aprojection e s onto an open and closed subset X e s of the projective limit X .It is easy to see that the semigroup S acts on the projective limit X by partialisomorphisms β s : X → X e s defined by the property β s : X → X e s , ξ su ( β s ( x )) = ξ u ( x ) , ∀ u ∈ S, ∀ x ∈ X. (4.8)The map β s is well-defined since the set { su : u ∈ S } is cofinal and ξ u ( x ) ∈ X su ,with suξ u ( x ) = s ( y ) = y . The image of β s is in X e s , since by definition of β s : ξ s ( β s ( x )) = ξ ( x ) = y . For x ∈ X e s , we have ξ su ( x ) ∈ X u . This shows that β s defines an isomorphism of X with X e s , whose inverse map is given by ξ u ( β − s ( x )) = ξ su ( x ) , ∀ x ∈ X e s , ∀ u ∈ S. (4.9)The corresponding algebra morphisms ρ s are then given by ρ s ( f )( x ) = f ( β − s ( x )) , ∀ x ∈ X e s , ρ s ( f )( x ) = 0 , ∀ x / ∈ X e s . (4.10)This class of examples also fulfill an equidistribution property , making the uni-form normalized counting measures µ s on X s compatible with the projective sys-tem and inducing a probability measure on the limit X . Namely, one has ξ s ′ ,s µ s = µ s ′ , ∀ s, s ′ ∈ S. (4.11) In this section we assume that k is a number field. We fix an embedding σ : k ֒ → C and we denote by ¯ k an algebraic closure of σ ( k ) ⊂ C in C .When taking points over ¯ k , algebraic endomotives yield 0-dimensional singularquotient spaces X (¯ k ) /S , which can be described by means of locally compact ´etalegroupoids G (¯ k ) and the associated crossed product C ∗ -algebras C ( X (¯ k )) ⋊ S . Thisconstruction gives rise to the category of analytic endomotives .One starts off by considering a uniform system ( A α , S ) of Artin motives over k and the algebras A C = A ⊗ k C = lim −→ α A α ⊗ k C , A C = A k ⊗ k C = A C ⋊ S. (5.1)8 Caterina Consani
The assignment a ∈ A → ˆ a, ˆ a ( χ ) = χ ( a ) ∀ χ ∈ X = lim ←− α X α (5.2)defines an involutive embedding of algebras A C ⊂ C ( X ). The C ∗ -completion C ( X ) of A C is an abelian AF C ∗ -algebra. One sets¯ A C = C ( X ) ⋊ S. This is the C ∗ -completion of the algebraic crossed product A C ⋊ S . It is defined bythe algebraic relations (4.1) with the involution which is apparent in the formulae( cf. [27],[28]).In the applications that require to work with cyclic (co)homology, it is im-portant to be able to restrict from C ∗ -algebras such as ¯ A C to canonical densesubalgebras A C = C ∞ ( X ) ⋊ alg S ⊂ ¯ A C (5.3)where C ∞ ( X ) ⊂ C ( X ) is the algebra of locally constant functions. It is to this cat-egory of smooth algebras (rather than to that of C ∗ -algebras) that cyclic homologyapplies.The following result plays an important role in the theory of endomotives andtheir applications to examples arising from the study of the thermodynamicalproperties of certain quantum statistical dynamical systems. We shall refer tothe following proposition, in section 5.1 of this paper for the description of theproperties of the “BC-system”. The BC-system is a particularly relevant quantumstatistical dynamical system which has been the prototype and the motivatingexample for the introduction of the notion of an endomotive. We refer to [7], § Proposition 5.1 ([7] Proposition 3.1).
1) The action (4.2) of G k on X (¯ k ) definesa canonical action of G k by automorphisms of the C ∗ -algebra ¯ A C = C ( X ) ⋊ S ,preserving globally C ( X ) and such that, for any pure state ϕ of C ( X ) , α ϕ ( a ) = ϕ ( α − ( a )) , ∀ a ∈ A , α ∈ G k . (5.4)
2) When the Artin motives A α are abelian and normal, the subalgebras A ⊂ C ( X ) and A k ⊂ ¯ A C are globally invariant under the action of G k and the states ϕ of ¯ A C induced by pure states of C ( X ) fulfill α ϕ ( a ) = ϕ ( θ ( α )( a )) , ∀ a ∈ A k , θ ( α ) = α − , ∀ α ∈ G abk = G k / [ G k , G k ](5.5)On the totally disconnected compact space X , the abelian semigroup S ofhomeomorphisms acts, producing closed and open subsets X s ֒ → X , x x · s .The normalized counting measures µ α on X α define a probability measure on X with the property that the Radon–Nikodym derivatives ds ∗ µdµ (5.6) oncommutative geometry and motives X . One lets G = X ⋊ S be the corresponding´etale locally compact groupoid. The crossed product C ∗ -algebra C ( X (¯ k )) ⋊ S coincides with the C ∗ -algebra C ∗ ( G ) of the groupoid G .The notion of right (or left) action of G on a totally disconnected locally com-pact space Z is defined as in the algebraic case by (4.3) and (4.4). A right action of G on Z gives on the space C c ( Z ) of continuous functions with compact support on Z a structure of right module over C c ( G ). When the fibers of the map g : Z → X are discrete (countable) subsets of Z one can define on C c ( Z ) an inner productwith values in C c ( G ) by h ξ, η i ( x, s ) = X z ∈ g − ( x ) ¯ ξ ( z ) η ( z ◦ s ) (5.7)A right action of G on Z is ´etale if and only if the fibers of the map g arediscrete and the identity is a compact operator in the right C ∗ -module E Z over C ∗ ( G ) given by (5.7).An ´etale correspondence is a ( G ( X α , S ) , G ( X ′ α ′ , S ′ ))-space Z such that the rightaction of G ( X ′ α ′ , S ′ ) is ´etale.The Q -vector space Corr(( X, S, µ ) , ( X ′ , S ′ , µ ′ ))of linear combinations of ´etale correspondences Z modulo the equivalence relation Z ∪ Z ′ = Z + Z ′ for disjoint unions, defines the space of (virtual) correspondences.For M = ( X, S, µ ), M ′ = ( X ′ , S ′ , µ ′ ), and M ′′ = ( X ′′ , S ′′ , µ ′′ ), the compositionof correspondencesCorr( M, M ′ ) × Corr( M ′ , M ′′ ) → Corr(
M, M ′′ ) , ( Z, W ) Z ◦ W is given following the same rule as for the algebraic case (4.5), that is by the fiberproduct over the groupoid G ′ . A correspondence gives rise to a bimodule M Z overthe algebras C ( X ) ⋊ S and C ( X ′ ) ⋊ S ′ and the composition of correspondencestranslates into the tensor product of bimodules. Definition 5.2.
The category C ∗ V oK of analytic endomotives is the (pseudo)abeliancategory generated by objects of the form M = ( X, S, µ ) with the properties listedabove and morphisms given as follows. For M = ( X, S, µ ) and M ′ = ( X ′ , S ′ , µ ′ ) objects in the category, one sets Hom C ∗ V K ( M, M ′ ) = Corr( M, M ′ ) ⊗ Q K. (5.8)The following result establishes a precise relation between the categories ofArtin motives and (noncommutative) endomotives. Theorem 5.3 ([7], Theorem 3.13).
The categories of Artin motives and algebraicand analytic endomotives are related as follows. (1)
The map
G 7→ G (¯ k ) determines a tensor functor F : EV o ( k ) K → C ∗ V oK , F ( X α , S ) = ( X (¯ k ) , S, µ ) from algebraic to analytic endomotives. Caterina Consani (2)
The Galois group G k = Gal(¯ k/k ) acts by natural transformations of F . (3) The category CV o ( k ) K of Artin motives embeds as a full subcategory of EV o ( k ) K . (4) The composite functor c ◦ F : EV o ( k ) K → KK ⊗ K (5.9) maps the full subcategory CV o ( k ) K of Artin motives faithfully to the category KK G k ⊗ K of G k -equivariant KK -theory with coefficients in K . Given two Artin motives X = Spec( A ) and X ′ = Spec( B ) and a component Z = Spec( C ) of the cartesian product X × X ′ , the two projections turn C into a( A, B )-bimodule c ( Z ). If U = P a i χ Z i ∈ Hom CV o ( k ) K ( X, X ′ ), c ( U ) = P a i c ( Z i )defines a sum of bimodules in KK ⊗ K . The composition of correspondences in CV o ( k ) K translates into the tensor product of bimodules in KK G k ⊗ Kc ( U ) ⊗ B c ( L ) ≃ c ( U ◦ L ) . One composes the functor c with the natural functor A → A C which associates toa ( A, B )-bimodule E the ( A C , B C )-bimodule E C . The resulting functor c ◦ F ◦ ι : CV o ( k ) K → KK G k ⊗ K is faithful since a correspondence such as U is uniquely determined by the corre-sponding map of K -theory K ( A C ) ⊗ K → K ( B C ) ⊗ K . The prototype example of the data which define an analytic endomotive is thesystem introduced by Bost and Connes in [6]. The evolution of this C ∗ -dynamicalsystem encodes in its group of symmetries the arithmetic of the maximal abelianextension of k = Q .This quantum statistical dynamical system is described by the datum given by anoncommutative C ∗ -algebra of observables ¯ A C = C ∗ ( Q / Z ) ⋊ α N × and by the timeevolution which is assigned in terms of a one-parameter family of automorphisms σ t of the algebra. The action of the (multiplicative) semigroup S = N × on thecommutative algebra C ∗ ( Q / Z ) ≃ C (ˆ Z ) is defined by α n ( f )( ρ ) = ( f ( n − ρ ) , if ρ ∈ n ˆ Z , otherwise , ρ ∈ ˆ Z = lim ←− n Z /n Z . For the definition of the associate endomotive, one considers the projective system { X n } n ∈ N of zero-dimensional algebraic varieties X n = Spec( A n ), where A n = Q [ Z /n Z ] is the group ring of the abelian group Z /n Z . The inductive limit A =lim −→ n A n = Q [ Q / Z ] is the group ring of Q / Z . The endomorphism ρ n : A → A associated to an element n ∈ S is given on the canonical basis e r ∈ Q [ Q / Z ], oncommutative geometry and motives r ∈ Q / Z , by ρ n ( e r ) = 1 n X ns = r e s . (5.10)The Artin motives X n are normal and abelian, so that Proposition 5.1 applies.The action of the Galois group G Q = Gal( ¯ Q / Q ) on X n = Spec( A n ) is obtained bycomposing a character χ : A n → ¯ Q with the action of an element g ∈ G Q . Since χ is determined by the n -th root of unity χ ( e /n ), this implies that the action of G Q factorizes through the cyclotomic action and coincides with the symmetry group ofthe BC-system. The subalgebra A Q ⊂ ¯ A C = C ( X ) ⋊ S coincides with the rationalsubalgebra defined in [6].There is an interesting description of this system in terms of a pointed algebraicvariety ( Y, y ) ( cf. section 4.1) on which the abelian semigroup S acts by finiteendomorphisms. One considers the pointed affine group scheme ( G m ,
1) (the mul-tiplicative group) and lets S be the semigroup of non-zero endomorphisms of G m .These endomorphisms correspond to maps of the form u u n , for some n ∈ N .Then, the general construction outlined in section 4.1 determines on ( G m ( Q ) , S = N × acting on G m ( Q ) as specified above. Itfollows from the definition (4.6) that X n = Spec( A n ) where A n = Q [ u ± n ] / ( u nn − . For n | m the connecting morphism ξ m,n : X m → X n is defined by the algebrahomomorphism A n → A m , u ± n u ± am with a = m/n . Thus, one obtains anisomorphism of Q -algebras ι : A = lim −→ n A n ∼ → Q [ Q / Z ] , ι ( u n ) = e n . (5.11)The partial isomorphisms ρ n : Q [ Q / Z ] → Q [ Q / Z ] of the group ring as described bythe formula (5.10) correspond under the isomorphism ι , to those given by (4.8) on X = lim ←− n X n . One identifies X with its space of characters Q [ Q / Z ] → ¯ Q . Then, theprojection ξ m ( x ) is given by the restriction of (the character associated to) x ∈ X to the subalgebra A m . The projection of the composite of the endomorphism ρ n of (5.10) with x ∈ X is given by x ( ρ n ( e r )) = 1 n X ns = r x ( e s ) . This projection is non-zero if and only if the restriction x | A n is the trivial character,that is if and only if ξ n ( x ) = 1. Moreover, in that case one has x ( ρ n ( e r )) = x ( e s ) , ∀ s , ns = r , and in particular x ( ρ n ( e k )) = x ( e nk ) . (5.12)2 Caterina Consani
For k | m the inclusion of algebraic spaces X k ⊂ X m is given at the algebra levelby the surjective homomorphism j k,m : A m → A k , j k,m ( u m ) = u k . Thus, one can rewrite (5.12) as x ◦ ρ n ◦ j k,nk = x | A nk . (5.13)This means that ξ nk ( x ) = ξ k ( x ◦ ρ n ) . By using the formula (4.9), one obtains the desired equality of the ρ ’s of (5.10)and (4.10).This construction continues to make sense for the affine algebraic variety G m ( k )for any field k , including the case of a field of positive characteristic. In this caseone obtains new systems, different from the BC system. The functor F : EV o ( k ) K → C ∗ V oK which connects the categories of algebraic and analytic endomotives establishes asignificant bridge between the commutative world of Artin motives and that of non-commutative geometry. When one moves from commutative to noncommutativealgebras, important new tools of thermodynamical nature become available. Oneof the most relevant techniques (for number-theoretical applications) is suppliedby the theory of Tomita and Takesaki for von Neumann algebras ([36]) which as-sociates to a suitable state ϕ ( i.e. a faithful weight) on a von Neumann algebra M ,a one-parameter group of automorphisms of M ( i.e. the modular automorphismsgroup) σ ϕt : R → Aut( M ) , σ ϕt ( x ) = ∆ itϕ x ∆ − itϕ . ∆ ϕ is the modular operator which acts on the completion L ( M, ϕ ) of { x ∈ M : ϕ ( x ∗ x ) < ∞} , for the scalar product h x, y i = ϕ ( y ∗ x ).This general theory applies in particular to the unital involutive algebras A = C ∞ ( X ) ⋊ alg S of (5.3) and to the related C ∗ -algebras ¯ A C which are naturallyassociated to an endomotive.A remarkable result proved by Connes in the theory and classifications of factors([8]) states that, modulo inner automorphisms of M , the one-parameter family σ ϕt is independent of the choice of the state ϕ . This way, one obtains a canonicallydefined one parameter group of automorphisms classes δ : R → Out( M ) = Aut( M ) / Inn( M ) . oncommutative geometry and motives M = M ⋊ σ ϕt R and the dual scaling action θ λ : R ∗ + → Aut( ˆ M ) (6.1)are independent of the choice of (the weight) ϕ .When these results are applied to the analytic endomotive F ( X α , S ) associated toan algebraic endomotive M = ( X α , S ), the above dual representation of R ∗ + com-bines with the representation of the absolute Galois group G k . In the particularcase of the endomotive associated to the BC-system ( cf. section 5.1), the resultingrepresentation of G Q × R ∗ + on the cyclic homology HC of a suitable Q (Λ)-module D ( A , ϕ ) associated to the thermodynamical dynamics of the system ( A , σ ϕt ) de-termines the spectral realization of the zeroes of the Riemann zeta-function andof the Artin L -functions for abelian characters of G k ( cf. [7], Theorem 4.16).The action of the group W = G Q × R ∗ + on the cyclic homology HC ( D ( A , ϕ )) ofthe noncommutative motive D ( A , ϕ ) is analogous to the action of the Weil groupon the ´etale cohomology of an algebraic variety. In particular, the action of R ∗ + isthe ‘characteristic zero’ analog of the action of the (geometric) Frobenius on ´etalecohomology. This construction determines a functor ω : EV o ( k ) K → Rep C ( W )from the category of endomotives to the category of (infinite-dimensional) repre-sentations of the group W .The analogy with the Tannakian formalism of classical motive theory is striking. Itis also important to underline the fact that the whole thermodynamical construc-tion is non-trivial and relevant for number-theoretic applications only because ofthe particular nature of the factor M (type III ) associated to the original datum( A , ϕ ) of the BC-system.It is tempting to compare the original choice of the state ϕ (weight) on the algebra A which singles out (via the Gelfand-Naimark-Segal construction) the factor M defined as the weak closure of the action of ¯ A = C ( X ) ⋊ S in the Hilbert space H ϕ = L ( M, ϕ ), with the assignment of a factor (
X, p, m ) r ( r ∈ Z ) on a puremotive M = ( X, p, m ): cf. (2.8). In classical motive theory, one knows thatthe assignment of a Z -grading is canonical only for homological equivalence orunder the assumption of the Standard Conjecture of K¨unneth type. In fact, thedefinition of a weight structure depends upon the definition of a complete systemof orthogonal central idempotents π iX .Passing from the factor M to the canonical dual representation (6.1) carries alsothe advantage to work in a setting where projectors are classified by their real di-mension ( ˆ M is of type II ), namely in a noncommutative framework of continuousgeometry which generalizes and yet still retains some relevant properties of thealgebraic correspondences ( i.e. degree or dimension).The process of dualization is in fact subsequent to a thermodynamical “coolingprocedure” in order to work with a system whose algebra approaches and becomesin the limit, a commutative algebra ( i.e. I ∞ ). Finally, one has also to implement a4 Caterina Consani further step in which one filters ( i.e. “distils”) the relevant noncommutative motive D ( A , ϕ ) within the derived framework of cyclic modules ( cf. section 3.1). Thisprocedure is somewhat reminissent of the construction of the vanishing cohomologyin algebraic geometry ( cf. [23]).When the algebra of the BC-system gets replaced by the noncommutative algebraof coordinates A = S ( A k ) ⋊ k ∗ of the ad`ele class space X k = A k /k ∗ of a numberfield k , the cooling procedure is described by a restriction morphism of (rapidlydecaying) functions on X k to functions on the “cooled down” subspace C k of id`eleclasses ( cf. [7], Section 5). In this context, the representation of C k on the cyclichomology HC ( H k, C ) of a suitable noncommutative motive H k, C produces thespectral realization of the zeroes of Hecke L -functions ( cf. op.cit , Theorem 5.6).The whole construction describes also a natural way to associate to a noncommu-tative space a canonical set of “classical points” which represents the analogue incharacteristic zero, of the geometric points C ( ¯ F q ) of a smooth, projective curve C / F q . [1] S. Baaj, P. Julg, Th´eorie bivariante de Kasparov et op´erateurs non born´es dans les C ∗ -modules hilbertiens, C. R. Acad. Sci. Paris Ser. I Math.
296 no. 21 (1983), 875–878.[2] P. Baum, Cycles, cocycles and K -theory (unpublished manuscript).[3] A. A. Beilinson, R. MacPherson, V. Schechtman, Notes on motivic cohomology, DukeMath. Journal
54 (2) (1987), 679–710.[4] B. Blackadar, K -theory for operator algebras , Second edition, MRSI Publications(1998).[5] S. Bloch, Algebraic cycles and higher K -theory, Advances in Math.
61 (1986), 267–304.[6] J. B. Bost, A. Connes, Hecke algebras, type III factors and phase transitions withspontaneous symmetry breaking in number theory,
Selecta Math. (N.S.)
Advances in Math. (to appear).[8] A. Connes, Une classification des facteurs de type III,
Ann. Sci. ´Ecole Norm. Sup. (4) 6 (1973), 133–252.[9] A. Connes, Cohomologie cyclique et foncteurs Ext n , C. R. Acad. Sci. Paris S´er. IMath.
296 no. 23, (1983), 953–958.[10] A. Connes, Trace formula in noncommutative geometry and zeros of the Riemannzeta function,
Selecta Math. (N.S.) [11] A. Connes, C. Consani, M. Marcolli, The Weil proof and the geometry of the adelesclass space, math.NT/0703392.[12] A. Connes, M. Marcolli, Renormalization and motivic Galois theory, IMRN
Publ. RIMSKyoto Univ.
20 (1984) 1139–1183.[14] C. Consani, M. Marcolli, Quantum statistical mechanics over function fields,
J. ofNumber Theory
123 (2007), 487–528.[15] P. Colmez, J. P. Serre,
Correspondence Grothendieck-Serre , DocumentsMath´ematiques, Soc. Math. de France (2001).[16] P. Deligne, Cat´egories tannakiennes in
Grothendieck Festschrift vol. 2, Progress inMath. 87, Birkh¨auser, 1990, 111–198.[17] P. Deligne, `A quoi servent les motifs?, in
Motives (ed. by Jannsen et al.), Proc.Sympos. Pure Math. 55 Part 1, American Math. Soc., Providence RI (1994), 143–161.[18] M. Demazure, Motifs des vari´et´es alg´ebriques,
S´em. Bourbaki 1969-70 , expos´e 365,Lect. Notes in Math. 180, Springer-Verlag (1971).[19] A. Grothendieck, Motifs, (manuscript).[20] A. Grothendieck, Standard conjectures on algebraic cycles, in
Algebraic Geometry ,Bombay Colloquium 1968, Oxford, 1969, 193–199.[21] A. Grothendieck, R´ecoltes et semailles: r´eflexions et t´emoignage sur un pass´e demath´ematicien, Universit´e des Sciences et Techniques du Languedoc, Montpellier etCNRS.[22] A. Grothendieck et al.,
Th´eorie des Topos et Cohomologie ´Etale des Sch´emas , Lec-ture Notes in Math. 269, 270, 305, Springer-Verlag, Heidelberg (1972-73).[23] A. Grothendieck et al.,
Groupes de Monodromie en G´eom´etrie Alg´ebrique , LectureNotes in Math. 288, 340, Springer-Verlag, Heidelberg (1972-73).[24] U. Jannsen, Motives, numerical equivalence, and semi-simplicity,
Invent. Math. K -functor and extensions of C ∗ -algebras, Izv. Akad.Nauk SSSR Ser. Mat.
44 no. 3 (1980), 571–636, 719.[26] S. L. Kleiman, The standard conjectures, in
Motives , (ed. by Jannsen et al.), Proc.Sympos. Pure Math. 55 Part 1, American Math. Soc., Providence RI (1994), 3–20.[27] M. Laca, I. Raeburn Semigroup crossed products and the Toeplitz algebras of non-abelian groups,
J. Funct. Anal.
139 (1996), 415–440.[28] M. Laca, From endomorphisms to automorphisms and back: dilations and full cor-ners,
J. London Math. Soc. (2) 61 no. 3 (2000), 893–904.[29] J. L.Loday,
Cyclic homology , Grundlehren der Mathematischen Wissenschaften 301.Springer-Verlag, Berlin, 1998. xx+513 pp.[30] Yu. I. Manin, Lectures on the K -functor in algebraic geometry, Uspehi Mat. Nauk
24 no. 5 (1969), 3–86.[31] Yu. I. Manin, Correspondences, motifs, and monoidal transformations, Mat. Sb.(N.S.) 77 (1968), 475–507. Caterina Consani[32] R. Meyer, R. Nest, The Baum-Connes conjecture via localization of categories,
Lett.Math. Phys.
69 (2004), 237–263.[33] N. Saavedra
Cat´egories Tannakiennes in Lecture Notes in Math. 265, Springer-Verlag, Berlin and New-York, 1972.[34] A. J. Scholl, Classical motives, in
Motives , (ed. by Jannsen et al.), Proc. Sympos.Pure Math. 55 Part 1, American Math. Soc., Providence RI (1994), 163–187.[35] J. P. Serre, Motifs, in
Ast´erisque