Lie-Hopf algebras and their Hopf cyclic cohomology Bahram Rangipour Texas A&M University, College Station May 1, 2014 . . . . . . Local Index Formula in NCG Theorem (Connes-Moscovici, 1995) For ∫ an (odd) spectral triple (A, H, D) such that the residues − T := Ress=0 Tr(s ℓ T |D|−2s ), T ∈ {A, [D, c], |D|−z ; z ∈ C} ℓ make sense, one has: [(φn )n=1,3,... ] is a cocycle in the (b, B)-bicomplex of A, 0 n φn (a , . . . , a ) = ∑ k,ℓ ∫ cn,k,ℓ − a0 [D, a1 ](k1 ) . . . [D, an ](kn ) |D|−n−2|k| ℓ ∇(T ) = [D 2 , a], T (k) = ∇k (T ), |k| = k 1 + . . . + k n , ( ) (−1)|k| Γ(ℓ) |k| + n2 cn,k,ℓ = . k1 ! . . . kn !(k1 + 1) . . . (k1 + · · · + kn + n) . . . . . . Theorem (Connes-Moscovici 1998) I For any n ≥ 1 and any oriented flat manifold M n , there is a canonical Hopf algebra Hn acting on the algebra An := Cc∞ (FM) >▹ Diff(M). I There is a canonical cyclic cohomology theory, associated to Hn , canonically isomorphic to the Gelfand Fuks cohomology of the Lie algebra of formal vector fields on Rn . I There is a characteristic map from the mentioned cyclic cohomology of Hn to the cyclic cohomology of the algebra An such that the index cocycle is trapped in its image. . . . . . . Modular pair in involution Let H be a Hopf algebra. I An algebra map δ : H → C is called a character I An element σ ∈ H is called a group-like if ∆(σ) = σ ⊗ σ I The pair (δ, σ) is called modular pair in involution (MPI) if δ(σ) = 1 and e 2 (h) = σhσ −1 S δ eδ (h) = ∑ δ(h(1) )S(h(2) ) Here S . . . . . . Hopf cyclic cohomology, HC (H, σ Cδ ) We define operators b and B, b := q+1 ∑ H ⊗q o ( (−1)i di , B := i=0 q ∑ b / H ⊗q+1 B ) sq−1 (1 − t). (−1)qi t i i=0 where d0 (h1 ⊗ . . . ⊗ hq ) = 1 ⊗ h1 ⊗ . . . ⊗ hq , ∑ di (h1 ⊗ . . . ⊗ hq ) = h1 ⊗ . . . ⊗ ∆(hi ) ⊗ . . . ⊗ hq , dq+1 (h1 ⊗ . . . ⊗ hq ) = h1 ⊗ . . . ⊗ hq ⊗ σ, sj (h1 ⊗ . . . ⊗ hq ) = h1 ⊗ . . . ⊗ ε(hj+1 ) ⊗ . . . ⊗ hq , ∑ eδ (h1(2) ) · (h2 ⊗ . . . ⊗ hq ⊗ σ), t(h1 ⊗ . . . ⊗ hq ) = S where H acts on H ⊗q diagonally. . . . . . . Action of Hopf algebras on algebras The same way that groups or Lie algebras act on algebras by automorphisms or derivations respectively, we want Hopf algebras act on algebras. In this case we say an algebra A is H-module algebra if h ◃ (ab) = ∑ (h(1) ◃ a)(h(2) ◃ b), h ◃ 1A = ε(h)1A . . . . . . Characteristic map Suppose a Hopf algebra H acts on an algebra A. Provided for a MPI (δ, σ) the algebra A obtains a δ-invariant σ-trace, τ : A → C, i.e ( ) τ h(a) = δ(h)τ (a), ( ) τ (ba) = τ aσ(b) there is a characteristic map χτ : HC • (H, σ Cδ ) → HC • (A) χτ (h1 ⊗ . . . ⊗ hn )(a0 ⊗ . . . ⊗ an )) = τ (a0 h1 (a1 ) · · · hn (an )) . . . . . . Crossed product algebra Let M be a manifold and Γ ≤ Diff M. One defines the left action of Γ on C ∞ (M) by ϕ ◃ f = f ◦ ϕ−1 Define the crossed product algebra C ∞ (M) >▹ Γ ∑ A typical element is a finite sum of i f i Uϕ∗i where fUϕ∗ stands for f >▹ ϕ−1 . Multiplication reads as 2 ∗ f 1 Uϕ∗1 f 2 Uϕ∗2 = f 1 (ϕ−1 2 ◃ f )Uϕ2 ϕ1 . . . . . . The Hopf algebra Hn I Let M = Rn . Vector fields on FM are generated by Yji := yjk I ∂ , ∂yki Xk = yki ∂ ∂x i We lift them on An := Cc∞ (FM) >▹ Γ by Xk (fUφ∗ ) = Xk (f )Uφ∗ , Yji (fUφ∗ ) = Yji (f )Uφ∗ I We have Yji (ab) = Yji (a)b + aYji (b), I However, for Xk we have i Xk (ab) = Xk (a)b + aXk (b) + δj,k (a)Yij (b) , I i (fU ∗ ) = γ i (φ)fU ∗ , and Here δjk φ φ jk [ ]i i γjk (φ)(x, y ) = y −1 · φ′ (x)−1 · ∂µ φ′ (x) · y j ykµ . . . . . . continued ... I One defines the higher order operators δj,k|l by 1 ,...,lm I I δj,k|l = [Xlm , δj,k|l ] 1 ,...,lm 1 ,...,lm−1 i i i δµ − δi δµ − δjl|k = δµl They satisfy the Bianchi identities δjk|l µk jl jk Definition As an algebra Hn is the subalgebra of L(An ) generated by Xk , Yji I and all δj,k|l . The comultiplication of Hn is obtained by the 1 ,...,lm Leibniz rule h(ab) = h(1) (a) h(2) (b), h ∈ Hn , a, b ∈ An . So we see that i ⊗ Yij ∆(Xl ) = Xl ⊗ 1 + 1 ⊗ Xl + δj,l ∆(Yji ) = Yji ⊗ 1 + 1 ⊗ Yji i i i ∆(δj,k ) = δj,k ⊗ 1 + 1 ⊗ δj,k . . . . . . Actions of Hn on AΓ (FM) I Hn acts on AΓ (FM) I AΓ (FM) possesses the trace ∫ f ϖFM ∗ FM τ (fUφ ) = 0 I if φ = id otherwise The character on Hn defined by δ(Yji ) = δji , δ(Xl ) = 0, δ(δji0 j1 |j2 ,...,jm ) = 0 . . . . . . Structure of Hn I Xk , Yji form a representation of g := g ℓaffine n I i ∗ δj,k|ℓ generates F := U(a+ n) 1 ,...,ℓm I ( ) U(g), F forms a matched pair of Hopf algebras ( )cop Hn = U(g) I▹ F . . . . . . Past progress ([Moscovici-R]) C • (H, Cδ ) χτ / C • (AΓ ) O O I • (a) Ctop ELH / C •,• (g∗ , F) c−w Φ / C • (Ω) Bott Θ . . . . . . •,• Cc−w (g.∗ , F) .. .O .. O ∂c−w .. .O ∂c−w ∧2O g∗ bc−w ∂c−w / (∧2 g∗ ⊗ ∧2 F)F O ∂c−w bc−w ∂c−w gO ∗ bc−w ∂c−w C bc−w / ... ∂c−w bc−w / (g∗ ⊗ ∧2 F)F O ∂c−w bc−w / (∧2 g∗ ⊗ ∧3 F)F O / (g∗ ⊗ ∧3 F)F O bc−w / ... bc−w / ... ∂c−w bc−w / (C ⊗ ∧2 F)F / (C ⊗ ∧3 F)F , bc−w (α ⊗ f 0 ∧ · · · ∧ f q ) = α ⊗ 1 ∧ f 0 ∧ · · · ∧ f q , ∂c−w (α ⊗ f 0 ∧ · · · ∧ f q ) = ∑ ∂α ⊗ f 0 ∧ · · · ∧ f q − θi ∧ α ⊗ Xi ◃ (f 0 ∧ · · · ∧ f q ). i . . . . . . We now define a multiplication on the bicompplex p,q Ccoinv (g∗ , F) := (∧p g∗ ⊗ F ⊗q+1 )F . p,q r ,s p+r ,q+s Ccoinv ⊗ Ccoinv → Ccoinv , (ω1 ⊗ f 0 ⊗ . . . ⊗ f q ) ∗ (ω2 ⊗ g 0 ⊗ . . . ⊗ g s ) = ω1 ∧ ω2 ⊗ f 0 ⊗ . . . ⊗ f q−1 ⊗ f q g 0 ⊗ g 1 ⊗ . . . ⊗ g s Let us now define the graded multiplication (ω1 ⊗ f˜) · (ω2 ⊗ g̃ ) = (−1)qr (ω1 ⊗ f˜) ∗ (ω2 ⊗ g̃ ) and eventually we use the canonical projection π : F ⊗(q+1) → ∧q+1 F to define the multiplication on •,• (g∗ , F), by which it becomes a commutative DG algebra. Cc−w . . . . . . Hopf version of universal connection and curvature 1,0 i ωji := θji ⊗ 1 + θk ⊗ ηj,k ∈ Cc−w (g∗ , F) 1,1 i Ωij := θk ⊗ 1 ∧ ηj,k ∈ Cc−w (g∗ , F) . . . . . . i i ∂T (Ωij ) = −θl ∧ θlk ⊗ 1 ∧ ηj,k − θl ∧ θk ⊗ 1 ∧ ηj,k|l i s − θjr ∧ θk ⊗ 1 ∧ ηri ,k + θkr ∧ θk ⊗ 1 ∧ ηj,r + θsi ∧ θk ⊗ 1 ∧ ηj,k . ωki · Ωkj k i k i k = θki ∧ θp ⊗ 1 ∧ ηj,p + θl ∧ θp ⊗ ηk,l ⊗ ηj,p − θl ∧ θp ⊗ ηk,l ηj,p ⊗ 1. Ωik · ωjk i i k i k = −θl ∧ θjk ⊗ 1 ∧ ηk,l − θl ∧ θp ⊗ 1 ⊗ ηk,l ηj,p + θl ∧ θp ⊗ ηk,l ⊗ ηj,p . In other words ∂T (Ωij ) = Ωik · ωjk − ωki · Ωkj . . . . . . Similarly we have, i i ∂T (ωji ) = −θjk ∧ θki ⊗ 1 − θℓ ∧ θℓk ⊗ ηjk − θℓ ∧ θk ⊗ ηj,k|ℓ q i i i − θjp ∧ θk ⊗ ηp,k + θqi ∧ θk ⊗ ηj,k − θkp ∧ θk ⊗ ηj,p − θk ⊗ 1 ∧ ηj,k On the other hand we have ωki · ωjk = i k i k + θℓ ∧ θp ⊗ ηk,l ηj,p θki ∧ θjk ⊗ 1 + θℓ ∧ θjk ⊗ ηk,ℓ + θki ∧ θp ⊗ ηj,p We see that ∂T (ωji ) = −Ωij + ωki · ωjk . . . . . . Weil Algebra The truncated Weil algebra Ŵ := ⊕p,q≥0 Ŵ p,2n is recalled as follows. q Ŵ p,2q = Ap (gln ) ⊗ S2n (gln ) It is the commutative DG algebra generated by the connection elements Tji of degree 1 and the curvature elements Rji of degree 2. dTji = −1 ⊗ Rji + Tki ∧ Tjk ⊗ 1, dRji = Tjk ⊗ Rki − Tki ⊗ Rjk . . . . . . Hopf-Weil basis By the universal property of Ŵ (g ℓn ) we define the following DG algebra map r +k,k ∗ L : Ŵ (r ,2k) (g ℓn ) → Cc−w (g , F) (1) L(Tji ) = ωji , (2) L(Rji ) = Ωij Theorem r +k,k ∗ The map L : Ŵ (r ,2k) (g ℓn ) → Cc−w (g , F) defined above is a quasi-isomorphism. . . . . . . Lie-Hopf Algebras We first introduce the setting. I F is a commutative Hopf algebra. I g is a finite dimensional Lie algebra. I g acts on F by derivations, : g ⊗ F → F. I F coacts on g, H : g → g ⊗ F. For a fixed basis {X1 , · · · , XN } of g, we write the coaction as H : Xi 7→ Xj ⊗ fi j . . . . . . . Lie-Hopf Algebras j For the element fi,k := Xk fi j , we say the coaction H : g → g ⊗ F satisfies the structure identity of g if fj,ik − fi,jk = ∑ s,r k r s Cs,r fi fj + ∑ l k Ci,j fl l Finally we introduce an action of g on F ⊗2 as X • (f ⊗ g ) := X <0> f ⊗ X <1> g + f ⊗ X g . . . . . . . Lie-Hopf Algebras Definition We say F is a g−Hopf algebra if (A) Coaction H : g → g ⊗ F satisfies the structure identity of g, (B) Coalgebra structure of F is g−equivariant, i.e. ∆(X f ) = X • ∆(f ), ε(X f ) = 0. and Theorem F is a g−Hopf algebra if and only if (F, U(g)) is a matched pair of Hopf algebras meach means F I▹ U(g) is a Hopf algebra. . . . . . . Coinvariant Lie subalgebra and Hopf algebra Thanks to (A) { } g0 := gF := X ∈ g | H(X ) = X ⊗ 1 , is a Lie algebra. And because of (B) F0 := Fg := F ⟨g ◃ F⟩ is a Hopf algebra . . . . . . Cartan calculus for Lie-Hopf algebras For any Y ∈ g0 we define the contraction ιY and the Lie derivative LY on C •,• (g∗ , F). ιY (ω⊗ f˜) = ιY (ω)⊗ f˜, LY (ω⊗ f˜) = LY (ω)⊗ f˜ +ω⊗Y ◃ f˜. (3) (a) The contraction ιY is a derivation of degree −1. (b) The Lie derivative LY is a derivation of degree 0. (c) LX = ∂T ιY + ιY ∂T (d) [∂T , LY ] = 0 (e) L[Y1 ,Y2 ] = [LY1 , LY2 ] (f) [ιY1 , LY2 ] = ι[Y1 ,Y2 ] . . . . . . Weil homomorphism Let α : g∗0 → g∗ be a (algebraic) Cartan connection i.e. α ◦ adY = adY ◦α, ιY (α(ω)) = ω(Y ). 1,0 We extend α to α : g∗0 → Cc−w (g∗ , F) by α(ω) = (α(ω))<0> ⊗ (α(ω))<−1> . •,• The extension of α defines a Cartan connection on Cc−w (g∗ , F). As a result we get a map of g0 -DG algebras •,• Cα : ∧• g∗0 ⊗ S • (g∗0 )[2q] → Cc−w (g∗ , F) Here q = dim g − dim g0 . . . . . . . Classical geometries Geometry 1. General 2. Vol preserving Automorphism group Structure group Diff(Rn ) Diff Vol GLn (Rn ) SLn 3. Symplectic Diff Sp (R2n ) Sp2n 4. Contact Diff Cn (R2n+1 ) Cn2n+1 . . . . . . Geometries and the corresponding Hopf algebras ( ) CΓ∞ M) := Cc∞ (M) >▹ Γ Geometry Lie alg. Hopf alg. Iso. Lie alg. Algebra General an Hn g ℓn ( ) CΓ∞ F + (Rn ) Vol pres. san SHn sℓn ( ) CΓ∞ Fs+ (Rn ) Sympl. sp2n SpH2n sp2n ( + 2n ) CΓ∞ Fsp (R ) Contact Fn2n+1 CnH2n+1 cn2n+1 ( + 2n+1 ) CΓ∞ Fcn (R ) . . . . . . The quantum symmetry of AΓ := Cc∞ (Rn ) o Γ ∂ , ∂x i Xk (fUϕ∗ ) = Xk (f )Uϕ∗ , I Xk = I For Xk we have Xk (ab) = Xk (a)b + I Here σki (fUϕ∗ ) = ∑ σki (a)Xi (b) ∂ϕi ∗ fU ∂x k ϕ and higher derivatives σji1 ,...,jk (fUϕ∗ ) = ∂ k ϕi fU ∗ ∂x j1 · · · ∂x jk ϕ . . . . . . The Hopf algebra Kn We let the Hopf algebra Kn be the subalgebra of L(AΓ generated by Xℓ , σji1 ,...,jm , σ −1 here σ = det[σji ] is the Jacobi automorphism of AΓ . Its Hopf algebra structure is ∆(Xl ) = Xl ⊗ 1 + σlk ⊗ Xk , ∆(σji ) = σjk ⊗ σki , ∆(σ) = σ ⊗ σ, ∆(σ −1 ) = σ −1 ⊗ σ −1 , ∆(σji1 ,...,jk ) = [∆(Xjk ), ∆(σji1 ,...,jk−1 )], ε(σ) = ε(σ −1 ) = 1, ε(σji ) = δji , ε(Xl ) = ε(σji1 ,...,jk ) = 0. . . . . . . Action of Kn on AΓ I Kn acts on AΓ I AΓ possesses the trace τ (fUφ∗ ) −1 I σ C I I = ∫ f ϖRn if φ = id 0 otherwise Rn defines a MPI module on Kn τ is a σ −1 trace χτ : HC • (Kn ,σ −1 C) → HC • (AΓ ) . . . . . . −1 HC (Kn ,σ C) There is a map of algebras •,• •,• Cc−w (V ∗ , FK ) → Cc−w (g ∗ , g ℓn , FK ) That induces HP • (Kn ;σ −1 C) ∼ = HP • (Hn , g ℓn ; Cδ ) Theorem Hopf cyclic cohomology of Kn with coefficients in all Chern classes. . . σ −1 C . consists of . . . 1,1 i Ω̂ij = θk ⊗ 1 ∧ S(γjp )γp,k ∈ Cc−w (V ∗ , FK ) ( •,• ) HP m Cc−w (V ∗ , FK ) = ⊕ ⟨Tr(Ω̂i1 ) · · · Tr(Ω̂ik )⟩ i1 +···+ik =m . . . . . . For n = 1 we have the following b + B cocycles in HC 1 (K1 ,σ C0 := 1 ⊗ σ −1 X1 ∈ 1 C1 := 1 ⊗ σ −2 σ1,1 ∈ σ −1 −1 C) C ⊗ K1 , σ −1 C ⊗ K1 . . . . . . For n = 2 we have the following b + B cocycles in HC 2 (K2 ,σ −1 C) C0 := 1 ⊗ σ −1 σ2i X1 ⊗ σ −1 Xi − σ −1 σ1i X2 ⊗ σ −1 Xi , s s C1 := 1 ⊗ σ −1 σ2i S(σsj )σj,1 ⊗ σ −1 Xi − σ −1 σ1i S(σsj )σj,2 ⊗ σ −1 Xi , s k ⊗ σ −1 S(σkr )σs,t ⊗ C2 := 1 ⊗ σ −1 S(σri )σ1t σi,2 s k − σ −1 S(σri )σ2t σi,1 ⊗ σ −1 S(σkr )σs,t , s (C1 )2 := 1 ⊗ σ −1 S(σsi )σ1t σi,2 ⊗ σ −1 S(σkr )σrk,t s − σ −1 S(σsi )σ2t σi,1 ⊗ σ −1 S(σkr )σrk,t . . . . . . . A surprise χτ : HC • (Kn ,σ −1 C) → HC • (AΓ ) is not injective. For n=1 χτ (C1 ) = 0 For n=2 χτ (C1 ) = χτ (C12 ) = 0 . . . . . .