We show by constructing explicit homotopy operators that the Hochschild (co)homology of an A∞-algebra of Stasheff admits a differential calculus structure. As an application, we reproduce a result of Tradler which says that the Hochschild cohomology of a cyclic A∞-algebra admits a Batalin-Vilkovisky algebra structure.
Citation: Youming Chen, Weiguo Lyu, Song Yang. A note on the differential calculus of Hochschild theory for A∞-algebras[J]. Electronic Research Archive, 2022, 30(9): 3211-3237. doi: 10.3934/era.2022163
[1] | Yizheng Li, Dingguo Wang . Lie algebras with differential operators of any weights. Electronic Research Archive, 2023, 31(3): 1195-1211. doi: 10.3934/era.2023061 |
[2] | Noosheza Rani, Arran Fernandez . An operational calculus formulation of fractional calculus with general analytic kernels. Electronic Research Archive, 2022, 30(12): 4238-4255. doi: 10.3934/era.2022216 |
[3] | Yongge Tian . Characterizations of matrix equalities involving the sums and products of multiple matrices and their generalized inverse. Electronic Research Archive, 2023, 31(9): 5866-5893. doi: 10.3934/era.2023298 |
[4] | Yongjie Wang, Nan Gao . Some properties for almost cellular algebras. Electronic Research Archive, 2021, 29(1): 1681-1689. doi: 10.3934/era.2020086 |
[5] | A. A. Heliel, A. Ballester-Bolinches, M. M. Al-Shomrani, R. A. Al-Obidy . On σ-subnormal subgroups and products of finite groups. Electronic Research Archive, 2023, 31(2): 770-775. doi: 10.3934/era.2023038 |
[6] | Ming Ding, Zhiqi Chen, Jifu Li . The properties on F-manifold color algebras and pre-F-manifold color algebras. Electronic Research Archive, 2025, 33(1): 87-101. doi: 10.3934/era.2025005 |
[7] | Agustín Moreno Cañadas, Pedro Fernando Fernández Espinosa, José Gregorio Rodríguez-Nieto, Odette M Mendez, Ricardo Hugo Arteaga-Bastidas . Extended Brauer analysis of some Dynkin and Euclidean diagrams. Electronic Research Archive, 2024, 32(10): 5752-5782. doi: 10.3934/era.2024266 |
[8] | J. Vanterler da C. Sousa, Kishor D. Kucche, E. Capelas de Oliveira . Stability of mild solutions of the fractional nonlinear abstract Cauchy problem. Electronic Research Archive, 2022, 30(1): 272-288. doi: 10.3934/era.2022015 |
[9] | Doston Jumaniyozov, Ivan Kaygorodov, Abror Khudoyberdiyev . The algebraic classification of nilpotent commutative algebras. Electronic Research Archive, 2021, 29(6): 3909-3993. doi: 10.3934/era.2021068 |
[10] | Quanguo Chen, Yong Deng . Hopf algebra structures on generalized quaternion algebras. Electronic Research Archive, 2024, 32(5): 3334-3362. doi: 10.3934/era.2024154 |
We show by constructing explicit homotopy operators that the Hochschild (co)homology of an A∞-algebra of Stasheff admits a differential calculus structure. As an application, we reproduce a result of Tradler which says that the Hochschild cohomology of a cyclic A∞-algebra admits a Batalin-Vilkovisky algebra structure.
In differential geometry, on a smooth manifold M, we have the following classical structures (cf. Kobayashi and Nomizu [1,Proposition 3.10]):
(a) the space of polyvector fields, under the wedge product and Schouten bracket, forms a Gerstenhaber (also called super-Poisson) algebra;
(b) the space of differential forms, together with the exterior differential and wedge product, forms a commutative differential graded algebra; and
(c) vector fields act on differential forms by Lie derivative and by contraction, which satisfies the following two identities:
LX=d∘ιX+ιX∘d,[ιX,LY]=ι[X,Y], | (1.1) |
where X,Y are vector fields on M, LX is the Lie derivative and ιX is the contraction.
There are analogous statements in the holomorphic, symplectic and even in the non-commutative geometry.
For instance, given an associative algebra A, which is viewed as a non-commutative "space", the Hochschild cohomology HH∙(A,A) and Hochschild homology HH∙(A,A) play the roles of polyvector fields and differential forms on this space, and the Connes cyclic operator on HH∙(A,A) substitutes the de Rham differential. One may similarly define a version of contraction and Lie derivative as in the smooth manifolds case, which satisfy (1.1). This was first obtained by Daletskii-Gelfand-Tsygan [2], and summarized by Tamarkin-Tsygan in [3]. According to Tamarkin-Tsygan, a pair of spaces satisfying the above (a), (b) and (c) form a structure of differential calculus, a notion introduced in the same paper. In this note, we first show a similar result:
Theorem 1.1. Let A be an A∞-algebra over a field K. Then the Hochschild cohomology and homology of A,
(HH∙(A,A),HH∙(A,A),∪,[−,−],∩,B) |
is a differential calculus, where ∪ is the cup product, ∩ is the cap product, [−,−] is the Gerstenhaber Lie bracket and B is the Connes differential.
This result is known to experts Dolgushev-Tamarkin-Tsygan [4] and has been essentially laid out by Kontsevich in his article Formal (non)commutative symplectic geometry, and was also explained in Section 7 of his work Notes on A∞-algebras, A∞-categories and noncommutative geometry [5], joint with Soibelman. We here give all necessary calculations required to prove (1.1), which seems to be rarely found in the literature.
Another motivation for us to show the above result is that it is related to the study of Calabi-Yau algebras, a notion introduced by Ginzburg in [6], where he also showed that, for a Calabi-Yau algebra, say A, there is a Batalin-Vilkovisky algebra structure on its Hochschild cohomology. The proof is heavily based on the differential calculus structure on the Hochschild (co)homology of A (see also [7]).
On the other hand, for a Calabi-Yau algebra, if it is Koszul (see Priddy [8]), then its Koszul dual algebra is a cyclic associative algebra (that is, an associative algebra with a cyclically invariant non-degenerate pairing). Tradler showed in [9] that there is also a Batalin-Vilkovisky algebra structure on the Hochschild cohomology of such cyclic associative algebra. Recently in [10], Chen, the third author and Zhou proved that for a Koszul Calabi-Yau algebra the Batalin-Vilkovisky algebras on the Hochschild cohomology of A and on that of its Koszul dual are isomorphic.
To understand the Batalin-Vilkovisky algebra structure on Calabi-Yau algebra in a more general setting, such as N-Koszul Calabi-Yau algebras in the sense of Berger [11], or even more generally, exact complete Calabi-Yau algebras in the sense of Van den Bergh [12], one is led to understand the differential calculus structure on cyclic A∞-algebras (that is, A∞-algebras with a cyclically invariant non-degenerate pairing), since in both of these two cases, the "Koszul dual" of these types of Calabi-Yau algebras are cyclic A∞-algebras.
Corollary 1.2 (Tradler [9]). If A is a cyclic A∞-algebra, that is, A is a finite dimensional A∞-algebrawith a cyclically invariant non-degenerate pairing, then the Hochschild cohomology HH∙(A,A) has a Batalin-Vilkovisky algebra structure.
This corollary is originally due to Tradler [9,Theorem 2]. Here we give an alternative proof, which is in the same spirit of Menichi [13] for finite dimensional symmetric algebras from the differential calculus point of view.
Throughout the note, we work over a ground field K. All algebras are associative algebras over K with unit. All vector spaces, their tensors and morphisms etc. are over K.
In this section, we recall the definitions of differential calculus and some of its applications. Let us start with Gerstenhaber algebras.
Definition 2.1 (Gerstenhaber [14]). A Gerstenhaber algebra is a quaternion (H∙,∪,[−,−],1), where H∙ is a N-graded vector space and 1∈H0, such that:
(1) (H∙,∪) is a graded commutative algebra with unit 1∈H0;
(2) (H∙,[−,−]) is a graded Lie algebra of degree −1, i.e.,
[f,g]=−(−1)(|f|−1)(|g|−1)[g,f] |
and a graded Jacobi identity
(−1)(|f|−1)(|h|−1)[[f,g],h]+(−1)(|g|−1)(|f|−1)[[g,h],f]+(−1)(|h|−1)(|g|−1)[[h,f],g]=0; |
(3) The Lie bracket [−,−] is a derivation with respect to the product ∪, i.e.,
[f,g∪h]=[f,g]∪h+(−1)|g|(|f|−1)g∪[f,h], |
for arbitrary homogenous elements f,g,h∈H∙, where |f| is the degree of the homogenous element f.
Definition 2.2 (Tamarkin-Tsygan [3]). Let H∙ be a N-graded vector space and H∙ be Z-graded vector space. A differential calculus is the data
(H∙,H∙,∪,[−,−],1,∩,B) |
such that:
(1) (H∙,∪,[−,−],1) is a Gerstenhaber algebra;
(2) H∙ is a graded module over (H∙,∪) through the map
∩:Hm⊗Hn→Hn−m,f⊗μ↦f∩μ, |
for μ∈Hn and f∈Hm, i.e., if we define ιf(μ):=f∩μ, then ιf∪g=ιfιg;
(3) there exists a map B:H∙→H∙+1 such that B2=0, and
[ιf,Lg]gr=ι[f,g], |
where Lg:=[B,ιg]gr=Bιg−(−1)|g|ιgB, for f,g homogenous elements of H∙.
Hochschild [15] introduced the cohomology theory of associative algebras. But the Hochschild cohomology ring of a K-algebra is a Gerstenhaber algebra, which was first discovered by Gerstenhaber in [14]. Given a K-algebra A, its Hochschild cohomology groups are defined as HHn(A,A)≅ExtnAe(A,A) for n≥0, where Ae=A⊗KAop is the enveloping algebra of A. There exists a projective resolution of A as Ae-module, the so called normalized bar resolution ¯Bar∙(A) which is given by ¯Barr(A)=A⊗¯A⊗r⊗A, where ¯A=A/(K⋅1A), that is,
¯Bar∙(A):⋯→A⊗¯A⊗r⊗Adr→A⊗¯A⊗r−1⊗A→⋯→A⊗¯A⊗Ad1→A⊗2(μ→A), |
where the map μ:A⊗A→A is the multiplication of A, and the differential dr is given by
dr(a0⊗¯a1⊗⋯⊗¯ar⊗ar+1)=a0a1⊗¯a2⊗⋯⊗¯ar⊗ar+1+r−1∑i=1(−1)ia0⊗¯a1⊗⋯⊗¯ai−1⊗¯aiai+1⊗¯ai+2⊗⋯⊗¯ar⊗ar+1+(−1)ra0⊗¯a1⊗⋯⊗¯ar−1⊗arar+1. |
The Hochschild cohomology complex is C∙(A,A)=HomAe(¯Bar∙(A),A). Note that Cr(A,A)=HomAe(A⊗¯A⊗r⊗A,A)≅HomK(¯A⊗r,A) for each r≥0. We also identify C0(A,A) with A. Thus C∙(A,A) has the following form:
C∙(A,A):Aδ0→HomK(¯A,A)→⋯→HomK(¯A⊗r,A)δr→HomK(¯A⊗(r+1),A)→⋯. |
It is not difficult to give the definition of δ∙, in fact, for any f in HomK(¯A⊗r,A), the map δr(f) is defined by
δr(f)(¯a1⊗⋯⊗¯ar+1)=(−1)r−1a1⋅f(¯a2⊗⋯⊗¯ar+1)+r∑i=1(−1)i+r−1f(¯a1⊗⋯⊗¯ai−1⊗¯aiai+1⊗¯ai+2⊗⋯⊗¯ar+1)+f(¯a1⊗⋯⊗¯ar)⋅ar+1. |
Moreover, the cup product f∪g∈Cm+n(A,A)=HomK(¯A⊗(m+n),A) for f∈Cm(A,A) and g∈Cn(A,A) is given by
(f∪g)(¯a1⊗⋯⊗¯am+n):=g(¯a1⊗⋯⊗¯an)⋅f(¯an+1⊗⋯⊗¯am+n). |
One can prove that this cup product induces a well-defined product in Hochschild cohomology
∪:HHm(A,A)×HHn(A,A)⟶HHm+n(A,A). |
As a consequence, the graded K-vector space HH∙(A,A)=⨁n≥0HHn(A,A) is a graded commutative algebra with unit 1A [14,Corollary 1].
Furthermore, the Lie bracket is defined as follows. Let f∈Cm(A,A) and g∈Cn(A,A). If m,n≥1, then for 1≤i≤m, define f∘ig∈Cm+n−1(A,A) by
(f∘ig)(¯a1⊗⋯⊗¯am+n−1):=f(¯a1⊗⋯⊗¯ai−1⊗¯g(¯ai⊗⋯⊗¯ai+n−1)⊗¯ai+n⊗⋯⊗¯am+n−1), |
if m≥1 and n=0, then g∈A and define
(f∘ig)(¯a1⊗⋯⊗¯am−1):=f(¯a1⊗⋯⊗¯ai−1⊗¯g⊗¯ai⊗⋯⊗¯am−1), |
for any other case, set f∘ig to be zero. Now we define
fˉ∘g:=m∑i=1(−1)(i−1)(n−1)f∘ig |
and
[f,g]:=fˉ∘g−(−1)(m−1)(n−1)gˉ∘f. |
Such bracket [ , ] induces a well-defined Lie bracket in Hochschild cohomology
[ , ]:HHm(A,A)×HHn(A,A)⟶HHm+n−1(A,A). |
It is well known that (HH∙(A,A),∪,[ , ],1A) is a Gerstenhaber algebra [14,Page 267].
Meanwhile, the Hochschild chain complex is defined by C∙(A,A):=A⊗Ae¯Bar∙(A). Note that Cr(A,A)=A⊗Ae(A⊗¯A⊗r⊗A)≅A⊗¯A⊗r, and the differential is given by
b(a0⊗¯a1⊗⋯⊗¯ar)=r−1∑i=0(−1)ia0⊗⋯⊗¯ai−1⊗¯aiai+1⊗¯ai+2⊗⋯⊗¯ar+(−1)rara0⊗¯a1⊗⋯⊗¯ar−1. |
For f∈Cm(A,A) and a0⊗¯a1⊗⋯⊗¯an∈Cn(A,A), the cap product is
f∩(a0⊗¯a1⊗⋯⊗¯an)=a0f(¯a1⊗⋯⊗¯am)⊗¯am+1⊗⋯⊗¯an, |
while the Connes differential is defined by
B(a0⊗¯a1⊗⋯⊗¯ar)=r∑i=0(−1)ir1⊗¯ai⊗⋯⊗¯ar⊗¯a0⊗⋯⊗¯ai−1. |
Originally, the differential calculus on Hochschild cohomology and homology of associative algebras was obtained by Daletskii-Gelfand-Tsygan [2]; see also Tamarkin-Tsygan [3].
Theorem 2.3 (Daletskii-Gelfand-Tsygan, [2]). Let A be an associative algebra.Denote by HH∙(A,A) and HH∙(A,A)the Hochschild cohomology and homology of A respectively. Then
(HH∙(A,A),HH∙(A,A),∪,[−,−],∩,B) |
is a differential calculus, where ∪ is the cup product, ∩ is the cap product, [−,−] is the Gerstenhaber Lie bracket and B is the Connes differential.
Let (H∙,H∙,∪,[−,−],1,∩,B) be a differential calculus. Consider H′∙:=HomK(H−∙,K) the graded dual space of H∙. Then we can define the following two operations:
κf:H′∙⟶H′∙−|f|,B′:H′∙⟶H′∙+1, |
κf(Ω)(μ):=(−1)|f||Ω|Ω(ιfμ),B′(Ω)(μ):=(−1)|Ω|Ω(Bμ) |
for any arbitrary homogenous elements f∈H∙, μ∈H∙ and Ω∈H′∙ defined as well as the map ∩′:Hm⊗H′n→H′n−m by f∩′Ω:=κf(Ω). Then we have the following proposition.
Proposition 2.4. The data (H∙,H′∙,∪,[−,−],1,∩′,B′) is a differential calculus.
Proof. Note that (H∙,∪,[−,−],1) is a Gerstenhaber algebra and B′2=0. First, we have
(κfκg)(Ω)(μ)=κf(κg(Ω))(μ)=(−1)|f|(|Ω|−|g|)κg(Ω)(ιfμ)=(−1)−|f||g|+|Ω|(|f|+|g|)Ω(ιgιfμ)=(−1)|f||g|+|Ω|(|f|+|g|)Ω(ιg∪fμ)=(−1)|f||g|κg∪f(Ω)(μ)=κf∪g(Ω)(μ), |
for arbitrary homogenous elements f,g∈H∙, μ∈H∙ and Ω∈H′∙.
Next, we verify the condition (3) of the Definition 2.2. Let L′g:=[B′,κg]=B′κg−(−1)|g|κgB′, then we have
L′g(Ω)(μ)=(B′κg−(−1)|g|κgB′)(Ω)(μ)=(−1)|Ω|−|g|κg(Ω)(Bμ)−(−1)|g|+|g|(|Ω|+1)B′(Ω)(ιgμ)=(−1)|Ω|−|g|+|g||Ω|Ω(ιg(Bμ))−(−1)|g||Ω|+|Ω|Ω(B(ιgμ))=−(−1)|Ω|+|g||Ω|Ω(Lgμ), |
and from this we get that
[κf,L′g](Ω)(μ)=(κfL′g−(−1)(|g|+1)|f|L′gκf)(Ω)(μ)=(−1)|f|(Ω|−|g|+1)L′g(Ω)(ιfμ)+(−1)|Ω|(1+|g|)κf(Ω)(Lgμ)=(−1)(|f|+|g|+1)|Ω|+|f|(|g|+1)+1Ω(Lgιfμ)+(−1)(|f|+|g|+1)|Ω|Ω(ιfLgμ)=(−1)(|f|+|g|+1)|Ω|Ω([ιf,Lg]μ)=(−1)(|f|+|g|+1)|Ω|Ω(ι[f,g]μ)=κ[f,g](Ω)(μ). |
The proposition now follows.
For finite dimensional associative algebras, we have the following result.
Corollary 2.5 (Menichi [16], Remark 17). Let A be a finite dimensional algebra, and denote by HH∙(A,A′) the Hochschild cohomology of A with value in A′:=HomK(A,K).Then the data (HH∙(A,A), HH∙(A,A′), ∪, [−,−], 1A, ∩′, B′) is a differential calculus.
Proof. For finite dimensional algebra A, we have HH∙(A,A′)≅HH∙(A,A)′. Thus by Proposition 2.4, this corollary holds.
In this section we consider a refined version of differential calculus which is called differential calculus with duality.
Definition 3.1 (Lambre [17]). A differential calculus (H∙,H∙,∪,[−,−],1,∩,B) is called a differential calculus with duality if there exists an element (called volume form) η∈Hd for some integer d such that B(η)=0 and the map
∂(−):=−∩η:H∙→Hd−∙ |
is an H∙-module isomorphism. In this situation, the map ∂ is called the Van den Bergh-Poincaré duality.
This structure is strongly related to the so-called Batalin-Vilkovisky algebras.
Definition 3.2. A Batalin-Vilkovisky algebra is a Gerstenhaber algebra (H∙,∪,[−,−],1) with a linear map Δ:H∙→H∙−1 such that Δ2=0, Δ(1)=0 and
[f,g]=(−1)|f|(Δ(f∪g)−Δ(f)∪g−(−1)|f|f∪Δ(g)), |
for arbitrary homogeneous elements f,g∈H∙.
Given a differential calculus with duality (H∙,H∙,∪,[−,−],1,∩,B,η), the following commutative diagram
![]() |
defines an operator Δ:=−∂−1∘B∘∂, which is called the Batalin-Vilkovisky operator. In particular, we have the following result due to Lambre which is important in constructing the Batalin-Vilkovisky algebra from differential calculus structures (cf. Lambre [17,Theorem 1.6]).
Theorem 3.3 (Lambre [17]). Let (H∙,H∙,∪,[−,−],1,∩,B,η) be a differential calculus with duality. Then the quintuple (H∙,∪,[ , ],1,Δ) is a Batalin-Vilkovisky algebra.
Proof. Take Δ:=−∂−1B∂, and let f∈Hp, g∈Hq, z∈Hn.
Claim_:
[f,g]∪∂−1(z)=(−1)p(q−1)Δ(g∪(f∪∂−1(z)))+(−1)qf∪(g∪Δ∂−1(z))−f∪Δ(g∪∂−1(z))+(−1)(p−1)(q−1)g∪Δ(f∪∂−1(z)). | (3.1) |
Indeed, by the equation [ιg,Lf]gr=ι[g,f], we can obtain the equation:
[f,g]∩z=(−1)p−1B(f∩(g∩z))+f∩(B(g∩z))−(−1)(q−1)(p−1)g∩B(f∩z)−(−1)(p−1)qg∩(f∩B(z)). | (3.2) |
Since ∂ is an H∙-module isomorphism and apply ∂−1 to Eq (3.2), we can obtain this claim.
Let z=η, then ∂−1(z)=∂−1(η)=1∈H0, so
Δ(1)=−∂−1B∂∂−1(η)=−∂−1B(η)=0, |
and
Δ∘Δ=∂−1B∂∂−1B∂=∂−1BB∂=0. |
By Eq (3.1), we have that
[f,g]=(−1)p(q−1)Δ(g∪f)−f∪Δ(g)+(−1)(p−1)(q−1)g∪Δ(f)=(−1)p(q−1)+pqΔ(f∪g)−f∪Δ(g)+(−1)(p−1)(q−1)+q(p−1)Δ(f)∪g=(−1)p(Δ(f∪g)−Δ(f)∪g−(−1)pf∪Δ(g)). |
Thus we have the theorem.
Note that not all associative algebras admit the structure of differential calculus with duality on its Hochschild (co)homology. In the literature, there are two main classes of associative algebras having this property: Calabi-Yau algebras and finite dimensional symmetric algebras. The notion of Calabi-Yau algebras is introduced by Ginzburg [6]. More precisely, an algebra A is called Calabi-Yau algebra of dimension d if A has a finite length resolution of finitely generated projective Ae-modules, and there is an isomorphism RHomAe(A,Ae)≅A[−d] in the derived category of Ae-modules. The following result is due to Ginzburg [6,Theorem 3.4.3]; see also Lambre [17].
Theorem 3.4 (Ginzburg [6]). Let A be a Calabi-Yau algebra of dimension d.Then
(HH∙(A,A),HH∙(A,A),∪,[ , ],1A,∩,B) |
is differential calculus with duality, and thereforethere is a Batalin-Vilkovisky algebra on HH∙(A,A).
Another version of differential calculus with duality is defined on the Hochschild (co)homology of symmetric algebras. Recall a finite dimensional algebra A is symmetric if there exists a nondegenerate bilinear form ⟨−,−⟩:A⊗A→K such that ⟨ab,c⟩=⟨a,bc⟩ and ⟨a,b⟩=⟨b,a⟩ for arbitrary elements a,b,c∈A.
Theorem 3.5 (Tradler [9], Menichi [13]). Let A be a symmetric algebra.Then
(HH∙(A,A),HH∙(A,A′),∪,[−,−],1A,∩′,B′) |
is differential calculus with duality, and thereforethere is a Batalin-Vilkovisky algebra on HH∙(A,A).
Following this line, there are some interesting relevant works. Indeed, there is a "twisted" version of Theorem 3.4 and Theorem 3.5, which are obtained recently by Kowalzig and Krähmer [18] and Lambre, Zhou and Zimmermann [19] respectively. More generally, Menichi [13] considered the algebras over a cyclic operad with multiplication and showed their cohomology gives rise to a Batalin-Vilkovisky algebra structure.
In this section, we review the definitions of Hochschild (co)homology for A∞-algebras, and give a proof of the Gerstenhaber algebra structure on their Hochschild cohomology. Let us start with the definition of A∞-algebras.
Definition 4.1 (Stasheff [20]). An A∞-algebra over K is a graded vector space A:=⨁i∈ZAi with K-linear maps mn:A⊗n→A of degree n−2 for each n≥1, called the A∞-operators, satisfying the following A∞-relations
∑n=j+k+l≥1j,l≥0,k≥1(−1)j+klmj+1+l(id⊗j⊗mk⊗id⊗l)=0, |
that is,
n∑k=1n−k∑j=0(−1)kηj+kj+j+klmj+1+l(a1,⋯,aj,mk(aj+1,⋯,aj+k),aj+k+1,⋯,an)=0, | (4.1) |
where ηj=|a1|+⋯+|aj|+j and n=j+k+l.
Recall that the bar construction B(A) of A is the tensor coalgebra ∞⨁n=0(sA)⊗n with the coproduct
Δ[a1,⋯,an]:=n∑i=0[a1,⋯,ai]⊗[ai+1,⋯,an], |
where [a1,⋯,an] denotes the element (sa1)⊗⋯⊗(san)∈(sA)⊗n, and s is the suspension with degree |s|=1. For simplicity, we also write [a1,n]:=(sa1)⊗⋯⊗(san) and a1,n:=a1⊗⋯⊗an with some abuse of notation.
Given an A∞-algebra (A,{mn}n≥1), we denote by C∙(A,A):=Hom(B(A),A) the Hochchild cochain of A. Notice that we consider the graded-version Hom and total degree. The standard algebraic structures on Hochchild cochain complex (and induces on Hochchild cohomology) of A∞-algebra may obtain from its brace algebra (see Gerstenhaber-Voronov [21]). Let us recall that the braces are the maps
C∙(A,A)×⋯×C∙(A,A)→C∙(A,A),(f,f1,⋯,fk)↦f{f1,⋯,fk} |
given as follows: for any homogeneous elements [a1,⋯,an]∈B(A), f{f1,⋯,fk}[a1,⋯,an] is given by
∑(−1)k∑l=1ηil(|fl|+1)f[a1,⋯,f1[ai1+1,⋯,aj1],⋯,fk[aik+1,⋯,ajk],⋯,an], |
where the sum run over all 0≤i1≤j1≤⋯≤ik≤jk≤n, and ηi=∑is=1|as|+i. Clearly, the degree of f{f1,⋯,fk} satisfies |f{f1,⋯,fk}|=|f|+|f1|+⋯+|fk|+k. If k=1, then we denote that fˉ∘g:=f{g}, actually, it is just the pre-Lie operator introduced by Gerstenhaber [14]:
f{g}[a1,n]=∑0≤i≤j≤n(−1)ηi(|g|+1)f[a1,i,g[ai+1,j],aj+1,n]. |
For any homogeneous elements f,g∈C∙(A,A), the Gerstenhaber Lie bracket of f,g is given by
[f,g]:=fˉ∘g−(−1)(|f|+1)(|g|+1)gˉ∘f. |
By [Getzler [22], Lemma 1.2], with the formula
(fˉ∘g)ˉ∘h−fˉ∘(gˉ∘h)=f{g,h}+(−1)(|g|+1)(|h|+1)f{h,g}, | (4.2) |
we know that (C∙(A,A),[−,−]) is a graded Lie algebra of degree 1. Note that the space of coderivations Coder(B(A)) is a graded Lie algebra with bracket the graded commutator, and there is an isomorphisms of graded Lie algebra between sC∙(A,A) and Coder(B(A)) (see [Getzler-Jones [23], Proposition 1.2]). Since B(A) is cofree, the coderivation is determined by its corestriction to degree 1, and we have the following equivalent definition:
Definition 4.2 (Stasheff [20]). An A∞-algebra A is a graded vector space A equipped with a codifferential
D:B(A)→B(A) |
(i.e., a coderivation of degree |D|=−1 with D∘D=0) and D(1)=0.
In the following, if (A,{mn}n≥1) is an A∞-algebra in Definition 4.1, then we denote by (A,m) its associated A∞-algebra in Definition 4.2 with |m|=−2. In fact, we consider the following composition: Coder(B(A))≅sC∙(A,A)s−1⟶C∙(A,A), a codifferential D∈Coder(B(A)) of degree −1 corresponds to a Hochschild cochain m∈C∙(A,A) of degree −2. The condition D2=0 corresponds to mˉ∘m=0 which can be translated to Eq (4.1) in the definition of A∞-algebras. The cup product on C∙(A,A) is given by
f∪g:=(−1)|g|(|f|+1)m{g,f}. |
for any homogeneous elements f,g∈C∙(A,A). Clearly, |f∪g|=|f|+|g|.
Definition 4.3. Let (A,m) be an A∞-algebra. An A∞-bimodule M of A is a graded vector space with operations
bi,j:A⊗i⊗M⊗A⊗j⟶M, i,j≥0 |
of degree i+j−1 such that for any integers k and l, any homogeneous element ω∈M,
0=k∑i=1k−i+1∑j=1(−1)θ1bk−i+1,l(a1,j−1,mi(aj,i+j−1),⋯,w,⋯,ak+l)+k∑i=0l∑j=0(−1)θ2bk−i,l−j(a1,k−i,bi,j(ak−i+1,⋯,w,⋯,ak+j),⋯,ak+l)+l∑i=1l−i+1∑j=1(−1)θ3bk,l−i+1(a1,⋯,w,⋯,ak+j−1,mi(ak+j,k+i+j−1),⋯,ak+l), |
where
θ1=iεj−1+j−1+i(k+l−i−j),θ2=(i+j−1)εk−i+k−i+(l−j)(i+j−1),θ3=i(εk+j−1+|w|)+k−j+i(l−i−j+1) |
and εi:=i∑l=1|al|. We denote by (M,b) the A∞-bimodule M.
Remark 4.4. An A∞-algebra A itself is naturally an A∞-bimodule of A. Meanwhile, for an A∞-bimodule M of A, its dual M′ of M is still an A∞-bimodule of A.
Now we recall the definition of the Hochschild cohomology of A∞-algebras with value in an A∞-bimodules.
Definition 4.5. Let (A,m) be an A∞-algebra and (M,b) be an A∞-bimodule over (A,m). Then the Hochschild cohomology HH∙(A,M) of A with value in M is given by the cohomology of the Hochschild cochain complex
(C∙(A,M),δ) |
where C∙(A,M):=Hom(B(A),M), the differential δ(f):=bˉ∘f−(−1)|f|+1fˉ∘m, for f∈Hom(B(A),M) and
b∘f[a1,⋯,an]:=∑0≤i≤j≤n(−1)ηi(|f|+1)bi,n−j[a1,⋯,f[ai+1,⋯,aj],⋯,an],f∘m[a1,⋯,an]:=∑0≤i≤j≤n(−1)ηif[a1,⋯,m[ai+1,⋯,aj],⋯,an], |
where ηi=i∑r=1(|ar|+1). Especially, when M=A, δ(f)=[m,f]=mˉ∘f−(−1)|f|+1fˉ∘m.
An important feature of the Hochschild cochain complex C∙(A,A) of an A∞-algebra A is that it also admits an A∞-algebra structure (see [Getzler [22], Proposition 1.7]). From this A∞-algebra structure we have the following equations:
m{[m,f],g}+(−1)|f|+1m{f,[m,g]}+[m,m{f,g}]=0, | (4.3) |
m{[m,f],g,h}+(−1)|f|+1m{f,[m,g],h}+(−1)|f|+|g|m{f,g,[m,h]}+m{m{f,g},h}+(−1)|f|+1m{f,m{g,h}}+[m,m{f,g,h}]=0, | (4.4) |
for any homogenous elements f,g,h∈C∙(A,A). In particular, we have
Proposition 4.6. Let (A,m) be an A∞-algebra, then (C∙(A,A),[ , ],δ) is a differential graded Lie algebra.
Proof. By Eq (4.2), (C∙(A,A),[ , ]) is a graded Lie algebra. We only need to show that
δ[f,g]=[δf,g]+(−1)|f|+1[f,δg], |
for any f,g∈C∙(A,A). Equivalently, we need to prove
[m,[f,g]]=[[m,f],g]+(−1)|f|+1[f,[m,g]]. |
Then the proposition follows by the graded Jacobi identity.
Lemma 4.7. Let f,g∈C∙(A,A), then we have that
δfˉ∘g−δ(fˉ∘g)+(−1)|f|+1fˉ∘δg=m{f,g}+(−1)(|f|+1)(|g|+1)m{g,f}. |
Proof. Using Eq (4.2), we have that
LHS=[m,f]ˉ∘g−[m,fˉ∘g]+(−1)|f|+1fˉ∘[m,g]=(mˉ∘f)ˉ∘g−(−1)|f|+1(fˉ∘m)ˉ∘g−mˉ∘(fˉ∘g)+(−1)|f|+|g|(fˉ∘g)ˉ∘m−(−1)|f|fˉ∘(mˉ∘g)−(−1)|f|+|g|fˉ∘(gˉ∘m)=[(mˉ∘f)ˉ∘g−mˉ∘(fˉ∘g)]+(−1)|f|[(fˉ∘m)ˉ∘g−fˉ∘(mˉ∘g)]+(−1)|f|+|g|[(fˉ∘g)ˉ∘m−fˉ∘(gˉ∘m)]=m{f,g}+(−1)(|f|+1)(|g|+1)m{g,f}+(−1)|f|[f{m,g}+(−1)|g|+1f{g,m}]+(−1)|f|+|g|[f{g,m}+(−1)|g|+1f{m,g}]=RHS. |
Hence we have the lemma.
Corollary 4.8. Let f,g∈C∙(A,A), then we have that
δfˉ∘g−δ(fˉ∘g)+(−1)|f|+1fˉ∘δg=(−1)|f|+1(f∪g−(−1)|f||g|g∪f). |
Proof. By the Lemma 4.7, we have that
LHS=(−1)|f|(|g|+1)g∪f−(−1)|f|f∪g=RHS |
and thus we have the corollary.
Proposition 4.9. Let (A,m) be an A∞-algebra, then (C∙(A,A),∪,δ) is a differential graded algebra which is commutative up to homotopy.
Proof. By Eq (4.3), we have that
δ(f∪g)=δf∪g+(−1)|f|f∪δg, |
for any f,g∈C∙(A,A). According to Eq (4.4), we can obtain that
(f∪g)∪h−f∪(g∪h)=(−1)α(δ(m{h,g,f})+m{δh,g,f}+(−1)|h|−1m{h,δg,f}+(−1)|g|+|h|m{h,g,δf}), |
where α=(−1)|h|(|g|−1)+(|g|+|h|)(|f|−1). By the Corollary 4.8, the cup product is commutative up to homotopy and hence we have the proposition.
In [14], Gerstenhaber proved that there is a Gerstenhaber algebra on Hochschid cohomology of an associative algebra. Analogously, there is a similar Gerstenhaber algebra on Hochschid cohomology of an A∞ algebra, which was first observed by Getzler-Jones in [24].
Lemma 4.10. Let (A,m) be an A∞-algebra, then we have that
f{f1,f2}ˉ∘f3=f{f1,f2,f3}+(−1)(|f2|−1)(|f3|−1)f{f1,f3,f2}+(−1)(|f1|+|f2|)(|f3|−1)f{f3,f1,f2}+(−1)(|f2|−1)(|f3|−1)f{f1ˉ∘f3,f2}+f{f1,f2ˉ∘f3} | (4.5) |
(fˉ∘f1){f2,f3}=f{f1,f2,f3}+(−1)(|f1|−1)(|f2|−1)f{f2,f1,f3}+(−1)(|f2|+|f3|)(|f1|−1)f{f2,f3,f1}+f{f1ˉ∘f2,f3}+(−1)(|f1|−1)(|f2|−1)f{f2,f1ˉ∘f3}+fˉ∘(f1{f2,f3}) | (4.6) |
for any f,f1,f2,f3∈C∙(A,A).
Proof. It follows by straight-forward computation.
Theorem 4.11 ([24], [21]). Let (A,m) be an A∞-algebra, then (HH∙(A,A),[ , ],∪) is a Gerstenhaber algebra.
Now let us recall the definition of Hochschild homology of an A∞-algebra and the Connes differential; for more details, we refer to Getzler-Jones [23].
Definition 4.12. Let (A,m) be an A∞-algebra, and set C∙(A,A):=A⊗B(A). Then the Hochschild homology HH∙(A,A) of A is the homology of the Hochschild chain complex (C∙(A,A),b) where the differential b is given by
b(a0[a1,⋯,an]):=∑0≤j≤i≤n(−1)ηi(ηn−ηi)m[ai+1,⋯,an,a0,⋯,aj][aj+1,⋯,ai]+∑0≤i≤j≤n(−1)ηia0[a1,⋯,ai,m[ai+1,⋯,aj],aj+1,⋯,an], |
where ηi=∑is=0(|as|+1).
An element e∈A0 is called a strict unit if mn[a1,⋯,ai,e,ai+1,⋯,an−1]=0, for n≠2; and m2[e,a]=(−1)|a|m2[a,e]=a.
Definition 4.13. Suppose that (A,m) is an A∞-algebra with a strict unit. Then the Connes differential
B:C∙(A,A)→C∙(A,A), |
is given by
B(a0[a1,⋯,an])=n∑i=0(−1)ηi(ηn−ηi)e[ai+1,⋯,an,a0,a1,⋯,ai], |
for any a0[a1,⋯,an]∈C∙(A,A), where ηi=∑is=0(|as|+1).
It is not difficult to see B2=0 up to homotopy and Bb+bB=0.
Remark 4.14. If we define
t(a0[a1,⋯,an])=(−1)η0(ηn−η0)a1[a2,⋯,an,a0], |
and
T(e[a0,⋯,an,e,e])=(−1)η0(ηn−η0)e[a1,⋯,an,e,e,a0], |
then B2=0 up to homotopy, that is,
Bn+1Bn=bn+3sn+sn−1bn, |
where
sn(a0[a1,⋯,an])=n+1∑j=0Tj(1⊗n∑i=0ti⊗1⊗1)(e[a0,⋯,an,e,e])=n+1∑j=0Tj(n∑i=0(−1)ηi−1(ηn−ηi−1)e[ai,⋯,an,a0,⋯,ai−1,e,e])=n∑i=0(−1)ηi−1(ηn−ηi−1){e[ai,⋯an,a0,⋯,ai−1,e,e]+n−i∑j=1(−1)(ηn−1)(ηi+j−1−ηi−1)e[ai+j,⋯an,a0,⋯,ai−1,e,e,ai,⋯,ai+j−1]+(−1)ηn−1(|an|+1)e[a0,⋯,ai−1,e,e,ai⋯,an]+n+1∑j=n−i+2(−1)ηi+j−n−2(ηn−ηi+j−n−2)e[ai+j−n−1,⋯,ai−1,e,e,ai,⋯,an,a0,⋯,ai+j−n−1]}. |
In the last part of previous section, we recalled the definition of the Connes differential. Now we give another two differential calculus operators: the contraction (or cap product) and the Lie derivative.
Given an A∞-algebra (A,m). For any homogeneous elements x=a0[a1,…,at]∈C∙(A,A) and f1,…,fk∈C∙(A,A), the contraction {f1,⋯,fk}∩x is defined by
∑(−1)ηt(ηn−ηt)+k∑r=1(ηn−ηt+ηir−1)(|fr|+1)m[at+1,⋯,an,a0,a1,⋯,f1[ai1+1,⋯,aj1],⋯,fk[aik+1,⋯,ajk],⋯,as][as+1,⋯,at], |
where the sum runs over all 0≤s≤t≤n and 0≤i1≤j1≤⋯≤ik≤jk≤n.
Definition 5.1. Assume that (A,m) is an A∞-algebra. For any homogeneous element f∈C∙(A,A), the cap product ι is given by
ιf(x):=f∩x, |
for any homogeneous elements x∈C∙(A,A).
In fact, the cap product is well-defined in homology level by the following lemma.
Lemma 5.2. Assume that (A,m) is an A∞-algebra.Then we have
ιδf=[b,ιf]gr:=b∘ιf−(−1)|f|ιf∘b, |
for any homogenous element f∈C∙(A,A).
Proof. Denote
θ=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)|f|,ξ=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|f|+1). |
Let x:=a0[a1,k]. Then we have that
ιδfx=ι[m,f]x=∑(−1)θm[at+1,k,a0,⋯,[m,f][ai1+1,j1],⋯,as][as+1,t]=∑(−1)θ+(ηi2−ηi1)(|f|−1)m[at+1,k,a0,⋯,m[ai1+1,⋯,f[ai2+1,j2],⋯,aj1],⋯,as][as+1,t]+∑(−1)θ+ηi2−ηi1+|f|m[at+1,k,a0,⋯,f[ai1+1,⋯,m[ai2+1,j2],⋯,aj1],⋯,as][as+1,t]. |
We also have that
bιfx=b(∑(−1)ξm[at+1,k,a0,⋯,f[ai1+1,j1],⋯,as][as+1,t])=∑(−1)ξ+ηn−ηt+ηi2+|f|m[at+1,k,a0,⋯,f[ai1+1,j1],⋯,as][as+1,⋯,m[ai2+1,j2],⋯,at]+∑(−1)ξ+(ηn−ηt+ηi2+|f|)(ηt−ηi2)m[ai2+1,t,m[at+1,k,a0,⋯,f[ai1+1,j1],⋯,as],⋯,aj2][aj2+1,i2] |
and
ιfbx=ιf(∑(−1)(ηk−ηi1)ηi1m[ai1+1,k,a0,j1][aj1+1,i1]+∑(−1)ηi1a0[a1,i1,m[ai1+1,j1],aj1+1,k])=∑(−1)(ηk−ηi1)ηi1+(ηi1−ηt)ηn+(ηk−ηt+ηi2)(|f|−1)m[at+1,⋯,m[ai1+1,k,a0,⋯,aj1],⋯,f[ai2+1,j2],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|f|−1)m[at+1,k,a0,⋯,f[ai2+1,j2],⋯,m[ai1+1,j1],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|f|−1)m[at+1,k,a0,⋯,f[ai2+1,⋯,m[ai1+1,j1],⋯,aj2],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2)(|f|−1)m[at+1,k,a0,⋯,m[ai1+1,j1],⋯,f[ai2+1,j2],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|f|−1)m[at+1,k,a0,⋯,f[ai2+1,j2],⋯,as][as+1,⋯,m[ai1+1,j1],⋯,at]+∑(−1)ηi1+(ηk−ηt−1)ηt+(ηk−ηt+ηi2)(|f|−1)m[at+1,⋯,m[ai1+1,j1],⋯,ak,a0,⋯,f[ai2+1,j2],⋯,as][as+1,t]. |
Then we can obtain that
ιδf−bιf+(−1)|f|ιfb=(−1)ξ(mˉ∘m)[at+1,k,a0,⋯,f[ai1+1,j1],⋯,as][as+1,t]=0. |
Hence we prove this lemma.
Proposition 5.3. Let (A,m) be an A∞-algebra, then the (HH∙(A,A), ∩) is a graded module over (HH∙(A,A), ∪), that is to say, there exists a linear map
∩:HHp(A,A)⊗HHn(A,A)⟶HHn−p(A,A)φ⊗x↦φ∩x:=ιφ(x), |
satisfies
ιφ∪ψ=ιφιψ, |
for φ∈HH∙(A,A), ψ∈HH∙(A,A) and x:=a0[a1,n]∈HH∙(A,A).
Proof. It suffices to verify the identity
ιφ∪ψ−ιφιψ=(−1)|φ||ψ|[ι{ψ,δφ}−(−1)|ψ|ι{φ,δψ}+(−1)|φ|ι{ψ,φ}b+(−1)|ψ|bι{ψ,φ}]. | (5.1) |
We first compute the terms in Eq (5.1) one by one. Denote
ξ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|φ|+|ψ|+1). |
Let x:=a0[a1,k]. Then we first have
ιφ∪ψx=∑(−1)ξm[at+1,k,a0,⋯,φ∪ψ[ai1+1,j1],⋯,as][as+1,t]=∑(−1)ξ+(ηi2−ηi1)(|ψ|−1)+(ηi3−ηi1)(|φ|−1)+ψ(φ−1)m[at+1,k,a0,⋯,m[ai1+1,⋯,ψ[ai2+1,j2],⋯,ψ[ai3+1,j3],⋯,aj1],⋯,as][as+1,t]. |
Meanwhile, let
θ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|φ|−1). |
Then we have that
ιφιψx=ιφ(∑(−1)θm[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,as][as+1,t])=∑(−1)θ+(ηt−ηi2)(ηk−ηt+ηi2+|ψ|)+(ηk−ηi2+ηi3+|ψ|−1)(|φ|−1)m[ai2+1,t,m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,as],⋯,φ[ai3+1,j3],⋯,aj2][aj2+1,i2]. |
Furthermore, we take
ζ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|ψ|−1)+(ηk−ηt+ηi2−1)|φ|, |
and we have that
ι{ψ,δφ}x={ψ,δφ}∩x=∑(−1)ζm[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,δφ[ai2+1,j2],⋯,as][as+1,t]=∑(−1)ζ+(ηi3−ηi2)(|φ|−1)m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,m[ai2+1,⋯,φ[ai3+1,j3],⋯,aj2],⋯,as][as+1,t]+∑(−1)ζ+ηi3−ηi2+|φ|m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,φ[ai2+1,⋯,m[ai3+1,j3],⋯,aj2],⋯,as][as+1,t]. |
Now we denote
τ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)|ψ|+(ηk−ηt+ηi2−1)(|φ|−1), |
and we have that
ι{δψ,φ}x={δψ,φ}∩x=∑(−1)τm[at+1,k,a0,⋯,δψ[ai1+1,j1],⋯,φ[ai2+1,j2],⋯,as][as+1,t]=∑(−1)τ+(ηi3−ηi1)(|ψ|−1)m[at+1,k,a0,⋯,m[ai1+1,⋯,ψ[ai3+1,⋯,aj3],⋯,aj1],⋯,φ[ai2+1,j2],⋯,as][as+1,t]+∑(−1)τ+ηi3−ηi1+|ψ|m[at+1,k,a0,⋯,ψ[ai1+1,⋯,m[ai3+1,⋯,aj3],⋯,aj1],⋯,φ[ai2+1,j2],⋯,as][as+1,t]. |
We also have that
ι{ψ,φ}bx=ι{ψ,φ}(∑(−1)(ηk−ηi1)ηi1m[ai1+1,k,a0,j1][aj1+1,i1]+∑(−1)ηi1a0[a1,i1,m[ai1+1,j1],aj1+1,k])=∑(−1)(ηk−ηi1)ηi1+(ηi1−ηt)ηn+(ηk−ηt+ηi2)(|ψ|−1)+(ηk−ηt+ηi3)(|φ|−1)m[at+1,⋯,m[ai1+1,k,a0,⋯,aj1],⋯,ψ[ai2+1,j2],⋯,φ[ai3+1,j3],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|ψ|−1)+(ηk−ηt+ηi3−1)(|φ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,j2],⋯,φ[ai3+1,j3],⋯,m[ai1+1,j1],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|ψ|−1)+(ηk−ηt+ηi3−1)(|φ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,j2],⋯,φ[ai3+1,⋯,m[ai1+1,j1],⋯,aj3],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2)(|ψ|−1)+(ηk−ηt+ηi3)(|φ|−1)m[at+1,k,a0,⋯,m[ai1+1,j1],⋯,ψ[ai2+1,j2],⋯,φ[ai3+1,j3],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|ψ|−1)+(ηk−ηt+ηi3−1)(|φ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,j2],⋯,φ[ai3+1,j3],⋯,as][as+1,⋯,m[ai1+1,j1],⋯,at]+∑(−1)ηi1+(ηk−ηt−1)ηt+(ηk−ηt+ηi2)(|ψ|−1)+(ηk−ηt+ηi3)(|φ|−1)m[at+1,⋯,m[ai1+1,j1],⋯,ak,a0,⋯,ψ[ai2+1,j2],⋯,φ[ai3+1,j3],⋯,as][as+1,t].+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|ψ|−1)+(ηk−ηt+ηi3)(|φ|−1)m[at+1,k,a0,⋯,φ[ai2+1,⋯,m[ai1+1,j1],⋯,aj2],⋯,φ[ai3+1,j3],⋯,as][as+1,t]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+(ηk−ηt+ηi2−1)(|ψ|−1)+(ηk−ηt+ηi3)(|φ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,j2],⋯,m[ai1+1,j1],⋯,φ[ai3+1,j3],⋯,as][as+1,t]. |
Finally, we denote
ρ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|ψ|−1)+(ηk−ηt+ηi2−1)(|φ|−1), |
and we can also compute that
bι{ψ,φ}x=b(∑(−1)ρm[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,φ[ai2+1,j2],⋯,as][as+1,t])=∑(−1)ρ+ηk−ηt+ηi3+|φ|+|ψ|−1m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,φ[ai2+1,j2],⋯,as][as+1,⋯,m[ai3+1,j3],⋯,at]+∑(−1)ρ+(ηk−ηt+ηi3+|φ|+|ψ|−1)(ηk+|φ|+|ψ|)m[ai3+1,t,m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,φ[ai2+1,j2],⋯,as],⋯,aj3][aj3+1,i3]. |
Hence we can obtain that
ιφ∪ψ−ιφιψ=(−1)|φ||ψ|[ι{ψ,δφ}−(−1)|ψ|ι{φ,δψ}+(−1)|φ|ι{ψ,φ}b+(−1)|ψ|bι{ψ,φ}]+∑(−1)(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|ψ|−1)+(ηk−ηt+ηi2−1)(|φ|−1)+|ψ|(|φ|−1)+1mˉ∘m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,φ[ai2+1,j2],⋯,as][as+1,t]=(−1)|φ||ψ|[ι{ψ,δφ}−(−1)|ψ|ι{φ,δψ}+(−1)|φ|ι{ψ,φ}b+(−1)|ψ|bι{ψ,φ}]. |
This finishes the proof of the proposition.
Next, we give the definition of Lie derivative acting on Hochschild chain complex of A∞-algebras.
Definition 5.4. Let (A,m) be an A∞-algebra. The Lie derivative is given by
Lf(a0[a1,⋯,an]):=∑0≤i≤j≤n(−1)(ηi−1)(|f|+1)a0[a1,⋯,ai,f[ai+1,⋯,aj],⋯,an]+∑0≤j≤i≤n(−1)ηi(ηn−ηi)+|f|−1f[ai+1,⋯,an,a0,⋯,aj][aj+1,⋯,ai], |
for any homogenous elements f∈C∙(A,A) and x=a0[a1,…,an]∈C∙(A,A). In particular, taking f=m, then Lf=−b.
Proposition 5.5. Let (A,m) be an A∞-algebra, then we have that
Lφ=[B,ιφ]gr:=Bιφ−(−1)|φ|ιφB, |
for any φ∈HH∙(A,A).
Proof. We only need to prove that
Lφx−Bιφx+(−1)|φ|ιφBx=bSφx−Sδφx−(−1)|φ|Sφbx, | (5.2) |
for x:=a0[a1,k]∈HH∙(A,A), where
Sφx=∑(−1)ξe[at+1,k,a0,⋯,φ[ai1+1,j1],⋯,at], |
the sum runs over 0≤i1≤j1≤t≤k, and
ξ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|φ|−1). |
We compute the terms in Eq (5.2) one by one. Firstly,
Bιφx=B(∑(−1)ξm[at+1,k,a0,⋯,φ[ai1+1,j1],⋯,as][as+1,t])=∑(−1)ξ+(ηk−ηt+ηi2+|φ|)(ηt−ηi2)e[ai2+1,t,m[at+1,k,a0,⋯,φ[ai1+1,j1],⋯,as],⋯,ai2][ai2+1,j2]. |
Secondly,
ιφBx=ιφ(∑(−1)(ηk−ηi)ηie[ai+1,k,a0,i])=∑(−1)(ηk−ηi)ηiφ[ai+1,k,a0,j][aj+1,i]+∑(−1)(ηk−ηi)ηiφ[ai+1,j][aj+1,k,a0,i]. |
Thirdly, denote
ζ:=(ηk−ηt+ηi2)(|φ|−1). |
Then we have that
Sφbx=Sφ(∑(−1)(ηk−ηi1)ηi1m[ai1+1,k,a0,⋯,aj1][aj1+1,i1]+(−1)ηi1a0[a1,i1,m[ai1+1,j1],⋯,ak])=∑(−1)(ηk−ηi1)ηi1+(ηk−ηi1+ηt−1)(ηi1−ηt)+ζe[at+1,i1,m[ai1+1,k,a0,⋯,aj1],⋯,φ[ai2+1,j2],⋯,at]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+ζ+|φ|−1e[at+1,k,a0,⋯,φ[ai2+1,j2],⋯,m[ai1+1,j1],⋯,at]+∑(−1)ηi1+(ηk−ηt)(ηt−1)+ζ+|φ|−1e[at+1,k,a0,⋯,φ[ai2+1,⋯,m[ai1+1,j1],⋯aj2],⋯,at]+∑(−1)ηi1+(ηk−ηt−1)(ηt−1)+ζe[at+1,k,a0,⋯,m[ai1+1,j1],⋯,φ[ai2+1,j2],⋯,at]+∑(−1)ηi1+(ηk−ηt−1)ηt+ζe[at+1,⋯,m[ai1+1,j1],⋯,ak,a0,⋯,φ[ai2+1,j2],⋯,at]. |
We set
τ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|φ|−1), |
and then we have that
bSφx=∑(−1)τb(e[at+1,k,a0,⋯,φ[ai1+1,j1],⋯,at])=∑(−1)τ+(ηk+|φ|−1)(ηk−ηj1+ηi1−1)m[φ[ai1+1,j1],e][aj1+1,k,a0,i1]+∑(−1)τm[e,a0][a1,⋯,φ[ai1+1,j1],⋯,ak]+∑(−1)τ+ηi2−ηt−1e[at+1,⋯,m[ai2+1,j2],⋯,ak,a0,⋯,φ[ai1+1,j1],⋯,at]+∑(−1)τ+ηi2−ηt−1e[at+1,⋯,m[ai2+1,k,a0,j2],⋯,φ[ai1+1,j1],⋯,at]+∑(−1)τ+ηi2−ηt−1e[at+1,⋯,m[ai2+1,k,a0,⋯,φ[ai1+1,j1],⋯,aj2],⋯,at]+∑(−1)τ+ηk−ηt+ηi2−1e[at+1,k,a0,⋯,m[ai2+1,j2],⋯,φ[ai1+1,j1],⋯,at]+∑(−1)τ+ηk−ηt+ηi2−1e[at+1,k,a0,⋯,m[ai2+1,⋯,φ[ai1+1,j1,⋯,aj2],⋯,at]+∑(−1)τ+ηk−ηt+ηi2+|φ|e[at+1,k,a0,⋯,φ[ai1+1,j1],⋯,m[ai2+1,j2],⋯,at]. |
Lastly, we take
ρ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)|φ|, |
and we have that
Sδφx=∑(−1)ρm[at+1,k,a0,⋯,δφ[ai1+1,j1],⋯,at]=∑(−1)ρm[at+1,k,a0,⋯,m{φ}[ai1+1,j1],⋯,at]+∑(−1)ρ+|φ|m[at+1,k,a0,⋯,φ{m}[ai1+1,j1],⋯,at]=∑(−1)ρ+(ηi2−ηi1)(|φ|−1)m[at+1,k,a0,⋯,m[ai1+1,⋯,φ[ai2+1,j2],⋯,aj1],⋯,at]+∑(−1)ρ+ηi2−ηi1+|φ|m[at+1,k,a0,⋯,φ[ai1+1,⋯,m[ai2+1,j2],⋯,aj1],⋯,at]. |
It is not difficult to check Eq (5.2) according to the above computation.
We are now in a position to prove Theorem 1.1.
Proof of Theorem 1.1. By Theorem 4.11, Proposition 5.3 and Proposition 5.5, it is sufficient to show that the identity
[Lφ,ιψ]gr=(−1)|φ|−1ι[φ,ψ], | (5.3) |
holds for any φ,ψ∈HH∙(A,A).
We only need to prove that
Lφιψx−(−1)(|φ|−1)|ψ|ιψLφx−(−1)|φ|−1ι[φ,ψ]=bHφ,ψx−(−1)|φ|+|ψ|Hφ,ψbx−Hδφ,ψx−(−1)|φ|Hφ,δψx, | (5.4) |
for x:=a0[a1,k]∈HH∙(A,A), where
Hφ,ψx=∑(−1)ζφ[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]. |
Here the sum runs over 0≤i1≤j1≤j2≤i2≤k, and
ζ:=(ηk−ηi2)ηi2+(ηk−ηi2+ηi1−1)(|ψ|−1). |
We compute the terms in Eq (5.4) one by one. Firstly, denote
θ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|ψ|−1), |
and then we have that
Lφιψx=Lφ(∑(−1)θm[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,as][as+1,t])=∑(−1)θ+(ηk−ηt+ηi2+|ψ|−1)(|φ|−1)m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,as][as+1,⋯,φ[ai2+1,j2],⋯,at]+∑(−1)θ+(ηk−ηt+ηi2+|ψ|)(ηt−ηi2)+|φ|−1φ[ai2+1,t,m[at+1,k,a0,⋯,ψ[ai1+1,j1],⋯,as],⋯,aj2][aj2+1,i2]. |
Secondly, we have that
ιψLφx=ιψ(∑(−1)(ηi1−1)(|φ|−1)a0[a1,i1,φ[ai1+1,j1],⋯,ak]+(−1)(ηk−ηi1)ηi1+|φ|−1φ[ai1+1,k,a0,⋯,aj1][aj1+1,i1])=∑(−1)(ηi1−1)(|φ|−1)+(ηk−ηt)(ηt+|φ|−1)+(ηk−ηt+ηi2−1)(|ψ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,j2],⋯,as][as+1,⋯,φ[ai1+1,j1],⋯,at]+∑(−1)(ηi1−1)(|φ|−1)+(ηk−ηt)(ηt+|φ|−1)+(ηk−ηt+ηi2−1)(|ψ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,⋯,φ[ai1+1,j1],⋯,aj2],⋯,as][as+1,t]+∑(−1)(ηi1−1)(|φ|−1)+(ηk−ηt)(ηt+|φ|−1)+(ηk−ηt+ηi2+|φ|)(|ψ|−1)m[at+1,k,a0,⋯,φ[ai1+1,j1],⋯,ψ[ai2+1,j2],⋯,as][as+1,t]+∑(−1)(ηi1−1)(|φ|−1)+(ηk−ηt)(ηt+|φ|−1)+(ηk−ηt+ηi2−1)(|ψ|−1)m[at+1,k,a0,⋯,ψ[ai2+1,j2],⋯,φ[ai1+1,j1],⋯,as][as+1,t]+∑(−1)(ηi1−1)(|φ|−1)+(ηk−ηt+|φ|−1)(ηt+|φ|−1)+(ηk−ηt+ηi2+|φ|)(|ψ|−1)m[at+1,⋯,φ[ai1+1,j1],⋯,ak,a0,⋯,ψ[ai2+1,j2],⋯,as][as+1,t]+∑(−1)(ηk−ηi1)ηi1+|φ|−1+(ηk−ηi1+ηt+|φ|−1)(ηi1−ηt)+(ηk−ηt+ηi2+|φ|)(|ψ|−1)m[at+1,i1,⋯,φ[ai1+1,k,a0,j1],⋯,ψ[ai2+1,j2],⋯,as][as+1,t]. |
Thirdly, denote
ξ:=(ηk−ηt)ηt+(ηk−ηt+ηi1−1)(|φ|+|ψ|), |
and then we have that
ι[φ,ψ]x=∑(−1)ξm[at+1,k,a0,⋯,[φ,ψ][ai1+1,j1],⋯,as][as+1,t]=∑(−1)ξ+(ηi2−ηi1)(|ψ|−1)m[at+1,k,a0,⋯,φ[ai1+1,⋯,ψ[ai2+1,j2],⋯,aj1],⋯,as][as+1,t]+∑(−1)ξ+(ηi2−ηi1)(|φ|−1)+(|φ|−1)(|ψ|−1)+1m[at+1,k,a0,⋯,ψ[ai1+1,⋯,φ[ai2+1,j2],⋯,aj1],⋯,as][as+1,t]. |
Next, we have that
bHφ,ψx=∑(−1)ζ+(ηi2−ηi3)(ηk−ηi2+ηi3+|φ|+|ψ|)m[ai3+1,i2,φ[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj2],⋯,aj3][aj3+1,i3]+∑(−1)ζ+ηk−ηi2+ηi3+|φ|+|ψ|φ[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1⋯,m[aj3+1,i3],⋯,ai2]. |
Continually, we have that
Hφ,ψbx=Hφ,ψ(∑(−1)(ηk−ηi1)ηi1m[ai1+1,k,a0,⋯,aj1][aj1+1,i1]+(−1)ηi1a0[a1,i1,m[ai1+1,j1],⋯,ak])=∑(−1)(ηk−ηi1)ηi1+(ηi1−ηi2)(ηk−ηi1+ηi2−1)+(ηk−ηi1+ηi3)(|ψ|−1)φ[ai2+1,i1m[ai1+1,k,a0,⋯,aj1],⋯,ψ[ai3+1,j3],⋯,aj2][aj2+1,i2]+∑(−1)ηi1+(ηk−ηi2)(ηi2−1)+(ηk−ηi2+ηi3−1)(|ψ|−1)φ[ai2+1,k,a0,⋯,ψ[ai3+1,j3],⋯,aj2][aj2+1,⋯,m[ai1+1,j1],⋯,ai2]+∑(−1)ηi1+(ηk−ηi2)(ηi2−1)+(ηk−ηi2+ηi3−1)(|ψ|−1)φ[ai2+1,k,a0,⋯,ψ[ai3+1,j3],⋯,m[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ηi1+(ηk−ηi2)(ηi2−1)+(ηk−ηi2+ηi3−1)(|ψ|−1)φ[ai2+1,k,a0,⋯,ψ[ai3+1,⋯,,m[ai1+1,j1],⋯,aj3],⋯,aj2][aj2+1,i2]+∑(−1)ηi1+(ηk−ηi2)(ηi2−1)+(ηk−ηi2+ηi3)(|ψ|−1)φ[ai2+1,k,a0,⋯,m[ai1+1,j1],⋯,ψ[ai3+1,j3],⋯,aj2][aj2+1,i2]+∑(−1)ηi1+(ηk−ηi2−1)ηi2+(ηk−ηi2+ηi3)(|ψ|−1)φ[ai2+1,m[ai1+1,j1],⋯,ak,a0⋯,ψ[ai3+1,j3],⋯,aj2][aj2+1,i2]. |
We can also obtain that
Hδφ,ψx=∑(−1)ζδφ[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]=∑(−1)ζm{φ}[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+|φ|φ{m}[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]=∑(−1)ζ+(ηi3−ηi2)(|φ|−1)m[ai2+1,⋯,φ[ai3+1,j3],⋯,ak,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+(ηi3−ηi2)(|φ|−1)m[ai2+1,⋯,φ[ai3+1,k,a0,⋯,aj3],⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+(ηi3−ηi2)(|φ|−1)m[ai2+1,⋯,φ[ai3+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj3],⋯,aj2][aj2+1,i2]+∑(−1)ζ+(ηk−ηi2+ηi3+|ψ|−1)(|φ|−1)m[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,φ[ai3+1,j3],⋯,aj2][aj2+1,i2]+∑(−1)ζ+(ηk−ηi2+ηi3)(|φ|−1)m[ai2+1,k,a0,⋯,φ[ai3+1,j3],⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+(ηk−ηi2+ηi3)(|φ|−1)m[ai2+1,k,a0,⋯,φ[ai3+1,⋯,ψ[ai1+1,j1],⋯,aj3],⋯,aj2][aj2+1,i2]+∑(−1)ζ+|φ|+ηi3−ηi2φ[ai2+1,⋯,m[ai3+1,j3],⋯,ak,a0,⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+|φ|+ηi3−ηi2φ[ai2+1,⋯,m[ai3+1,k,a0,⋯,aj3],⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+|φ|+ηi3−ηi2φ[ai2+1,⋯,m[ai3+1,k,a0,⋯,ψ[ai1+1,j1],⋯,aj3],⋯,aj2][aj2+1,i2]+∑(−1)ζ+|φ+ηk−ηi2+ηi3+|ψ|−1φ[ai2+1,k,a0,⋯,ψ[ai1+1,j1],⋯,m[ai3+1,j3],⋯,aj2][aj2+1,i2]+∑(−1)ζ+ηk−ηi2+ηi3φ[ai2+1,k,a0,⋯,m[ai3+1,j3],⋯,ψ[ai1+1,j1],⋯,aj2][aj2+1,i2]+∑(−1)ζ+ηk−ηi2+ηi3φ[ai2+1,k,a0,⋯,m[ai3+1,⋯,ψ[ai1+1,j1],⋯,aj3],⋯,aj2][aj2+1,i2]. |
Lastly, denote
ζ′:=(ηk−ηi2)ηi2+(ηk−ηi2+ηi1−1)|ψ|, |
we compute that
Hφ,δψx=∑(−1)ζ′φ[ai2+1,k,a0,⋯,δψ[ai1+1,j1],⋯,aj2][aj2+1,i2]=∑(−1)ζ′+(ηi3−ηi1(|ψ|−1)φ[ai2+1,k,a0,⋯,m[ai1+1,⋯,ψ[ai3+1,j3],⋯,aj1],⋯,aj2][aj2+1,i2]+∑(−1)ζ′+ηi3−ηi1+|ψ|φ[ai2+1,k,a0,⋯,ψ[ai1+1,⋯,m[ai3+1,j3],⋯,aj1],⋯,aj2][aj2+1,i2]. |
Through comparing the two sides in Eq (5.4) according the above computations, we have done.
Corollary 5.6. For any homogenous elements φ,ψ∈HH∙(A,A), [B,Lφ]gr=0 and [Lφ,Lψ]gr=L[φ,ψ].
Proof. By Proposition 5.5, we obtain B∘Lφ=(−1)|φ|+1B∘ιφ∘B=(−1)|φ|+1Lφ∘B, and then by Eq (5.3), we get [Lφ,Lψ]gr=[[B,ιφ]gr,Lψ]gr=(−1)|φ|(|ψ|+1)[[B,Lψ]gr,Lφ]gr+[B,[ιφ,Lψ]gr]gr=[B,ι[φ,ψ]]gr=L[φ,ψ].
We start with the definition of cyclic A∞-algebras.
Definition 6.1. Let (A,m) be a finite dimensional A∞-algebra with strict unit. An A∞-cyclic structure of degree d on A is a non-degenerate bilinear form
⟨−,−⟩:A[1]⊗A[1]→K |
of degree −d (i.e., |a|+|b|=d−2, if ⟨sa,sb⟩≠0), such that
⟨sa,sb⟩=−(−1)(|a|+1)(|b|+1)⟨sb,sa⟩, |
and
⟨sm[a1,⋯,an],san+1⟩=(−1)(|a1|+1)n+1∑i=2(|ai|+1)⟨sm[a2,⋯,an+1],sa1⟩ |
for any homogenous elements ai∈A and integer n≥0.
Remark 6.2. There exists a non-shifted version of cyclic A∞-algebra, see [12,Section 11]. An A∞-cyclic structure of degree d on A is a non-degenerate bilinear form
⟨−,−⟩′:A⊗A→K |
of degree −d, such that
⟨a,b⟩′=(−1)|a||b|⟨b,a⟩′, |
and
⟨m[a1,⋯,an],an+1⟩′=(−1)n+|a1|⋅n+1∑i=2(|ai|)⟨m[a2,⋯,an+1],a1⟩′ |
for any homogenous elements ai∈A and integer n≥0. In fact, in the Definition 6.1, if we take ⟨a,b⟩′:=(−1)|a|⟨sa,sb⟩, then we can obtain the non-shifted version. In this note, we adopt the shifted version since the sign rules in this case are just the Koszul sign convention.
Proposition 6.3. Let A be a cyclic A∞-algebra of degree d.Then the Hochschild data
(HH∙(A,A),HH∙(A,A′),∪,∩′,[−,−],B′,Ω) |
is a differential calculus with duality, where HH∙(A,A′) is the Hochschild cohomology of A with value in A∞-bimodule A′, and Ω∈HHd(A,A′) is a volume form.
Proof. By Theorem 1.1 and Proposition 2.4, we obtain that (HH∙(A,A), HH∙(A,A′), ∪, ∩′, [−,−], B′) is a differential calculus. The only thing left is to show the existence of the duality. That is:
Claim 6.4. There exists an element Ω∈HHd(A,A′), such that
HH∙(A,A)→HHd−∙(A,A′), f↦κf(Ω) |
is an isomorphism.
In fact, the cyclic structure of A induces an isomorphism Φ:A⟶A′ given by
Φ(a)(b)=(−1)|a|⟨sa,sb⟩, |
and an isomorphism of complexes φ:C∙(A,A)→C∙(A,A′) given by φ(f):=Φ∘f, and a duality C∙(A,A′)≅(C∙(A,A))′ given by φ(f)(x):=(−1)|a0|(|f|+1)⟨sa0,sf[a1,⋯,an]⟩. Now we show that
φ(g∪f)=(−1)|f|(|g|+d)f∩′φ(g). |
Given any x=a0[a1⋯,an]∈C∙(A,A) and f,g∈C∙(A,A), we have
φ(g∪f)(x)=∑(−1)|a0|(|f|+|g|+1)+|f|(|g|−1)+η′i(|f|+1)+η′s(|g|+1)⟨sa0,sm[a1,⋯,f[ai+1,⋯,aj],⋯,g[as+1,⋯,at],⋯⋯,an]⟩=∑(−1)|g|(|f|−1)+ηi(|f|+1)+ηs(|g|+1)+|a0|⟨a0,sm[a1,⋯,f[ai+1,⋯,aj],⋯,sg[as+1,⋯,at],⋯,an]⟩=∑(−1)|g|(|f|−1)+ηi(|f|+1)+ηs(|g|+1)+(|a0|+1)(ηn+|f|+|g|)⟨sm[a1,⋯,f[ai+1,⋯,aj],⋯,g[as+1,⋯,at],⋯⋯,an],sa0⟩ |
where η′i=i∑i=1(|ai|+1), and
(f∩′φ(g))(x)=(−1)|f|(|g|+d)φ(g)(ιfx)=(−1)|f|(|g|+d)+(ηn−ηt)ηt+(ηn−ηt+ηi−1)(|f|+1)⋅φ(g)(∑m[at+1,⋯,an,a0,⋯,f[ai+1,⋯,aj],⋯,as][as+1,⋯,at])=∑(−1)|f|(|g|+d)+(ηn−ηt)ηt+(ηn−ηt+ηi−1)(|f|+1)+(|g|+1)(ηn−ηt+ηs+|f|+1)⋅⟨sm[at+1,⋯,an,a0,⋯,f[ai+1,⋯,aj],⋯,as],sg[as+1,⋯,at]⟩=∑(−1)|f|(|g|+d)+(ηn−ηt)ηt+(ηn−ηt+ηi−1)(|f|+1)+(|g|+1)(ηn−ηt+ηs+|f|+1)+(|at+1|+1)(ηn+|f|+|g|+1)⋅⟨sm[at,⋯,an,a0,⋯,f[ai+1,⋯,aj],⋯,g[as+1,⋯,at]],sat+1⟩=∑(−1)|f|(|g|+d)+(ηn−ηt)ηt+(ηn−ηt+ηi−1)(|f|+1)+(|g|+1)(ηn−ηt+ηs+|f|+1)+(ηn−ηt+|a0|+1)(ηn+|f|+|g|+1)⋅⟨sm[a1,⋯,f[ai+1,⋯,aj],⋯,g[as+1,⋯,at],⋯,an],sa0⟩=(−1)|f|(|g|+d)φ(g∪f)(x). |
Picking the map id:k→ke lying in C0(A,A), where e is the strict unit, then we have
φ(f)=(−1)|f|df∩′Ω=(−1)|f|dκf(Ω), |
where Ω denotes the element φ(id) lying in Cd(A,A′). This proves the claim, and Proposition 6.3 follows.
Theorem 6.5 (Tradler [9]). If A is a cyclic A∞-algebra, that is, A is a finite dimensional A∞-algebrawith a cyclically invariant non-degenerate pairing, then the Hochschild cohomology HH∙(A,A) has a Batalin-Vilkovisky algebra structure.
Proof. This theorem is direct from Theorem 3.3 and Proposition 6.3.
During the preparation of this paper, the first author was partially supported by the Natural Science Foundation of Chongqing (grant No. CSTC2020JCYJ-MSXMX0160), the second author was supported by NSFC (No.11301186) and by STCSM (No. 13dz2260400) and the Fundamental Research Funds for the Central Universities and the third author was partially supported by the Natural Science Foundation of Tianjin (Grant No. 20JCQNJC02000). The second author thanks his Ph.D. supervisor Guodong Zhou for his valuable suggestions and a variety of assistances in this note. We thanks to the anonymous referees for reading the manuscript carefully. Based on their comments, we added Remark 4.4 and 6.2 and modified some signs in Section 6 to make the note more rigorous and readable.
The authors declare there is no conflicts of interest.
[1] | S. Kobayashi, K. Nomizu, Foundations of differential geometry, 1. Interscience Publishers, a Division of John Wiley & Sons, New York-London, 1963. |
[2] | Y. L. Daletskii, I. M. Gelfand, B. L. Tsygan, On a variant of noncommutative differential geometry, Soviet Math. Dokl., 40 (1990), 422–426. |
[3] | D. Tamarkin, B. Tsygan, The ring of differential operators on forms in noncommutative calculus, Graphs and patterns in mathematics and theoretical physics, Proc. Sympos. Pure Math., Amer. Math. Soc., Providence, RI, 73 (2005), 105–131. https://doi.org/10.1090/pspum/073/2131013 |
[4] | V. A. Dolgushev, D. E. Tamarkin, B. L. Tsygan, Noncommutative calculus and the Gauss-Manin connection, Higher structures in geometry and physics, Progr. Math., 287, Birkhäuser/Springer, New York, (2011), 139–158. https://doi.org/10.1007/978-0-8176-4735-3_7 |
[5] | M. Kontsevich, Y. Soibelman, Notes on A∞-algebras, A∞-categories and non-commutative geometry. I, Homological mirror symmetry, 153-219, Lecture Notes in Phys., 757, Springer, Berlin, 2009. https://doi.org/10.1007/978-3-540-68030-7_6 |
[6] | V. Ginzburg, Calabi-Yau algebras, arXiv preprint math/0612139, 2006. |
[7] |
L. de Thanhoffer de Völcsey, M. Van den Bergh, Calabi-Yau deformations and negative cyclic homology, J. Noncommut. Geom., 12 (2018), 1255–1291. https://doi.org/10.4171/JNCG/304 doi: 10.4171/JNCG/304
![]() |
[8] |
S. B. Priddy, Koszul resolutions, Trans. Amer. Math. Soc., 152 (1970), 39–60. https://doi.org/10.1090/S0002-9947-1970-0265437-8 doi: 10.1090/S0002-9947-1970-0265437-8
![]() |
[9] |
T. Tradler, The Batalin-Vilkovisky algebra on Hochschild cohomology induced by infinity inner products, Ann. Inst. Fourier (Grenoble), 58 (2008), 2351–2379. https://doi.org/10.5802/aif.2417 doi: 10.5802/aif.2417
![]() |
[10] |
X. Chen, S. Yang, G. Zhou, Batalin-Vilkovisky algebras and the noncommutative Poincaré duality of Koszul Calabi-Yau algebras, J. Pure Appl. Algebra, 220 (2016), 2500–2532. https://doi.org/10.1016/j.jpaa.2015.11.016 doi: 10.1016/j.jpaa.2015.11.016
![]() |
[11] |
R. Berger, Koszulity for nonquadratic algebras, J. Algebra, 239 (2001), 705–734. https://doi.org/10.1006/jabr.2000.8703 doi: 10.1006/jabr.2000.8703
![]() |
[12] |
M. Van den Bergh, Calabi-Yau algebras and superpotentials, Sel. Math. New Ser., 21 (2015), 555–603. https://doi.org/10.1007/s00029-014-0166-6 doi: 10.1007/s00029-014-0166-6
![]() |
[13] |
L. Menichi, Batalin-Vilkovisky algebras and cyclic cohomology of Hopf algebras, K-Theory, 3 (2004), 231–251. https://doi.org/10.1007/s10977-004-0480-4 doi: 10.1007/s10977-004-0480-4
![]() |
[14] |
M. Gerstenhaber, The cohomology structure of an associative ring, Ann. Math., 78 (1963), 267–288. https://doi.org/10.2307/1970343 doi: 10.2307/1970343
![]() |
[15] |
G. Hochschild, On the cohomology groups of an associative algebra, Ann. Math., 46 (1945), 58–67. https://doi.org/10.2307/1969145 doi: 10.2307/1969145
![]() |
[16] |
L. Menichi, Batalin-Vilkovisky algebra structures on Hochschild cohomology, Bull. Soc. Math. France, 137 (2009), 277–295. https://doi.org/10.24033/bsmf.2576 doi: 10.24033/bsmf.2576
![]() |
[17] |
T. Lambre, Dualité de Van den Bergh et Structure de Batalin-Vilkovisky sur les algèbres de Calabi-Yau, J. Noncommut. Geom., 4 (2010), 441–457. https://doi.org/10.4171/JNCG/62 doi: 10.4171/JNCG/62
![]() |
[18] |
N. Kowalzig, U. Krähmer, Batalin-Vilkovisky structures on Ext and Tor, J. Reine Angew. Math., 697 (2014), 159–219. https://doi.org/10.1515/crelle-2012-0086 doi: 10.1515/crelle-2012-0086
![]() |
[19] |
T. Lambre, G. Zhou, A. Zimmermann, The Hochschild cohomology ring of a Frobenius algebra with semisimple Nakayama automorphism is a Batalin-Vilkovisky algebra, J. Algebra, 446 (2016), 103–131. https://doi.org/10.1016/j.jalgebra.2015.09.018 doi: 10.1016/j.jalgebra.2015.09.018
![]() |
[20] |
J. D. Stasheff, Homotopy associativity of H-spaces, Ⅱ, Trans. Amer. Math. Soc., 108 (1963), 293–312. https://doi.org/10.1090/S0002-9947-1963-0158400-5 doi: 10.1090/S0002-9947-1963-0158400-5
![]() |
[21] |
M. Gerstenhaber, A. A. Voronov, Homotopy G-Algebras and Moduli Space Operad, Int. Math. Res. Not., 3 (1995), 141–153. https://doi.org/10.1155/S1073792895000110 doi: 10.1155/S1073792895000110
![]() |
[22] | E. Getzler, Cartan homotopy formulas and the Gauss-Manin connection in cyclic homology, Israel Math. Conf. Proc., 7 (1993), 65–78. |
[23] |
E. Getzler, J. D. S. Jones, A∞-algebras and the cyclic bar complex, Illinois J. Math., 34 (1990), 256–283. https://doi.org/10.1215/ijm/1255988267 doi: 10.1215/ijm/1255988267
![]() |
[24] | E. Getzler, J. D. S. Jones, Operads, homotopy algebra and iterated integrals for double loop spaces, arXiv preprint, hep-th/9403055, 1994. |
[25] | B. L. Tsygan, Noncommutative Calculus and Operads, Topics in Noncommutative Geometry, 16 (2012), 19–66. |