Abstract
Starting with a spectral triple, one can associate two canonical differential graded algebras (DGA) defined by Connes (Noncommutative geometry (1994) Academic Press Inc., San Diego) and Fröhlich et al. (Comm. Math. Phys. 203(1) (1999) 119–184). For the classical spectral triples associated with compact Riemannian spin manifolds, both these DGAs coincide with the de-Rham DGA. Therefore, both are candidates for the noncommutative space of differential forms. Here we compare these two DGAs in a very precise sense.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
A differential calculus on a ‘space’ means the specification of a differential graded algebra (DGA), often interpreted as space of forms. In classical geometry, the ‘space’ is a manifold and we have the de-Rham DGA, whereas in noncommutative geometry a ‘space’ is described by a triple called spectral triple. A spectral triple is a tuple \((\mathcal {A},\mathcal {H},D)\), where \(\mathcal {A}\) is an associative \(\,\star \)-algebra represented on the Hilbert space \(\mathcal {H}\) and D is a Dirac-type operator on \(\mathcal {H}\). Associated to a spectral triple, there are two canonical DGAs defined by Connes [5] and Fröhlich et al. [7]. In literature, these are denoted by \(\,\Omega _D^\bullet (\mathcal {A})\) and \(\tilde{\Omega }_D^\bullet (\mathcal {A})\) respectively, and here we call them as the Dirac DGA and the FGR DGA. Note that in [2] we have called the Dirac DGA as the Connes’ calculus. It should be noted that for the classical spectral triple associated with compact Riemannian spin manifolds, both these DGAs coincide with the de-Rham DGA ([5, 7]). Therefore, both are candidates to be declared as noncommutative space of forms. Moreover, they are same for the noncommutative torus (page 172 in [7]) but not for the \(SU_q(2)\) [3]. Hence, it is natural to ask if there is any way to compare these two DGAs so that one can declare one of them as truly noncommutative space of forms. This is important because both being generalization of the classical de-Rham forms to the noncommutative set up, any notion in noncommutative geometry involving the noncommutative space of forms, e.g. the Yang-Mills functional [5] can be defined using either the Dirac DGA or the FGR DGA. Hence, a comparison is needed to overcome the difficulty of choice between these two DGAs. This is precisely the goal of our investigation in this article and the comparison is done through explicit computation of these DGAs for a family of spectral triples. In the literature, these DGAs have been computed in very few cases, e.g. noncommutative torus, \(SU_q(2)\). This indicates that probably these are difficult to compute and we had no clue on how to compare them. Recently, the authors have identified suitable hypotheses which allow the computation of the Dirac DGA \(\,\Omega _D^\bullet \) for a class of spectral triples. This gives the first systematic computation of \(\,\Omega _D^\bullet \) for a large family of spectral triples [2]. In this article, we compute the FGR DGA \(\,\tilde{\Omega }_D^\bullet \) for the same family of spectral triples, and this leads to a comparison between these two DGAs.
To describe our computation in detail, we recall the concept of the quantum double suspension (QDS) of a \(C^*\)-algebra \(\mathcal {A}\), denoted by \(\Sigma ^2 \mathcal {A}\,\), introduced by Hong and Szymanski in [8]. Later, QDS of a spectral triple was introduced by Chakraborty and Sundar [4]. We record here a few significance of QDS.
Significance of QDS:
-
(a)
Quantum even- and odd-dimensional spheres are produced by iterating QDS to two points and the circle, respectively [8].
-
(b)
Noncommutative analogues of n-dimensional balls are obtained by repeated application of the QDS to the classical low-dimensional spaces [9].
-
(c)
If we have one spectral triple \((\mathcal {A},\mathcal {H},D)\), then by iterating QDS we produce many spectral triples. Thus, by iterating QDS on the classical cases of manifolds one produces genuine noncommutative spectral triples. Moreover, finite summability and \(\Theta \)-summability are preserved under the iteration.
-
(d)
All the torus-equivariant spectral triples on the odd-dimensional quantum spheres are obtained by iterating QDS to the spectral triple \((C^{\infty }(S^1),L^2(S^1),-i\frac{\mathrm{d}}{{\mathrm{d}}\theta })\).
-
(e)
Most importantly, QDS produces a class of examples of regular spectral triples having simple dimension spectrum [4], essential in the context of local index formula of Connes and Moskovici [6].
This article adds one more significance to the above list namely, QDS provides a comparison between the Dirac DGA and the FGR DGA and establishes the Dirac DGA as a more appropriate generalization of the classical de-Rham DGA to the noncommutative set-up. We work here under the following mild hypotheses on a spectral triple \((\mathcal {A},\mathcal {H},D)\):
-
(1)
\([D,a]F-F[D,a]\) is a compact operator for all \(a\in \mathcal {A}\), where F is the sign of the operator D,
-
(2)
\(\mathcal {H}^\infty :=\bigcap _{k\ge 1}\mathcal {D}\mathrm{om}(D^k)\) is a left \(\mathcal {A}\)-module, and \([D,\mathcal {A}]\subseteq \mathcal {A}\otimes \mathcal {E}\mathrm{nd}_{\mathcal {A}}(\mathcal {H}^\infty )\subseteq \mathcal {E}\mathrm{nd}_{\mathbb {C}}(\mathcal {H}^\infty )\).
The notable features of these hypotheses are firstly, the spectral triple associated with a first order differential operator on a manifold will always satisfy them and secondly, they are stable under the quantum double suspension. The authors have computed \(\,\Omega _D^\bullet \) for the quantum double suspended spectral triple \((\Sigma ^2 \mathcal {A},\Sigma ^2 \mathcal {H},\Sigma ^2 D)\) in [2] under these conditions. It turns out that the FGR DGA becomes almost trivial for \((\Sigma ^2 \mathcal {A},\Sigma ^2 \mathcal {H},\Sigma ^2 D)\) in the sense that it does not reflect any information about \((\mathcal {A},\mathcal {H},D)\). This phenomenon was observed in [3] for the \(SU_q(2)\). Since the torus equivariant spectral triples on the odd-dimensional quantum spheres are obtained through iterated QDS on the spectral triple \((C^{\infty }(S^1), L^2(S^1),-i\frac{\mathrm{d}}{\mathrm{d}\theta })\), this article also extends the earlier work of Chakraborty and Pal [3]. This helps us to conclude, in view of [2], that the Dirac DGA is more informative than the FGR DGA.
The organization of this paper is as follows. In section 2, we discuss Dirac DGA \(\,\Omega _D^\bullet \), the quantum double suspension and obtain a few results. Section 3 mainly deals with the computation of the FGR DGA \(\,\tilde{\Omega }_{\Sigma ^2D}^\bullet (\Sigma ^2\mathcal {A})\) for QDS, which finally leads us to the comparison between Connes DGA and FGR DGA.
2 Dirac DGA and the quantum double suspension
In this section, we recall the definition of Dirac DGA \(\,\Omega _D^\bullet \,\) from [5], and the quantum double suspension from [4, 8].
DEFINITION 2.1
A spectral triple \((\mathcal {A},\mathcal {H},D)\) over an involutive associative algebra \(\mathcal {A}\) consists of the following:
-
(1)
a \(\,\star \,\)-representation \(\pi \) of \(\mathcal {A}\) on a Hilbert space \(\mathcal {H}\),
-
(2)
an unbounded self-adjoint operator D acting on \(\mathcal {H}\),
-
(3)
D has compact resolvent and [D, a] extends to a bounded operator on \(\mathcal {H}\) for every \(\,a \in \mathcal {A}\).
We will assume that \(\mathcal {A}\) is unital and \(\pi \) is a unital representation. If \(|D|^{-p}\) is in the ideal of Dixmier traceable operators \(\mathcal {L}^{(1,\infty )}\), then we say that the spectral triple is \(\,p\)-summable. In literature, this is sometimes denoted by \(\,p^+\)-summable, \((p,\infty )\)-summable, etc. Moreover, if there is a \(\mathbb {Z}_2\)-grading \(\gamma \in \mathcal {B}(\mathcal {H})\) such that \(\gamma \) commutes with every element of \(\mathcal {A}\) and anticommutes with D, then the spectral triple \((\mathcal {A},\mathcal {H},D,\gamma )\) is said to be an even spectral triple. Associated to every spectral triple, we have the following differential graded algebra (DGA).
DEFINITION 2.2
[2, 5] Let \((\mathcal {A},\mathcal {H},D)\) be a spectral triple and \(\,\Omega ^\bullet (\mathcal {A})=\bigoplus _{k=0}^\infty \Omega ^k(\mathcal {A})\,\) be the reduced universal differential graded algebra over \(\mathcal {A}\). Here, \(\,\Omega ^k(\mathcal {A}):=\mathrm{span}\{a_0da_1\ldots da_k:a_i\in \mathcal {A}, i=1,\ldots ,k\},\,d\) being the universal differential. With the convention \((da)^*=-da^*\), we get a \(\star \,\)-representation \(\pi \) of \(\,\Omega ^\bullet (\mathcal {A})\) on \(\mathcal {Q}(\mathcal {H}):=\mathcal {B}(\mathcal {H})/\mathcal {K}(\mathcal {H})\), given by
Let \(\,J_0^{(k)}=\{\omega \in \Omega ^k:\pi (\omega )=0\}\) and \(\,J^\prime =\bigoplus J_0^{(k)}\). Since \(\,J^\prime \) fails to be a differential ideal in \(\,\Omega ^\bullet \), consider \(J^\bullet =\bigoplus J^{(k)}\), where \(\,J^{(k)}=J_0^{(k)}+dJ_0^{(k-1)}\). Then \(\,J^\bullet \) becomes a differential graded two-sided ideal in \(\,\Omega ^\bullet \) and hence, the quotient \(\,\Omega _D^\bullet =\Omega ^\bullet /J^\bullet \) becomes a differential graded algebra, called the Connes’ calculus or the Dirac DGA.
The representation \(\pi \) gives the following isomorphism:
The differential d on \(\Omega ^\bullet (\mathcal {A})\) induces a differential, denoted again by d, on the complex \(\Omega ^\bullet _D(\mathcal {A})\) so that we get a chain complex \((\,\Omega ^\bullet _D(\mathcal {A}),d\,)\) and a chain map \(\,\pi _D:\Omega ^\bullet (\mathcal {A})\longrightarrow \Omega ^\bullet _D(\mathcal {A})\) such that the following diagram
commutes. Note that \(\,\Omega _D^\bullet (\mathcal {A})\) can be defined for non-unital algebra \(\mathcal {A}\) as well as that prescribed in [2], after Remark 2.3.
Lemma 2.3
If there is a decreasing filtration
of subspaces of \(\mathcal {A},\) then \(\,\Omega _D^\bullet (\mathcal {A})\) becomes a filtered algebra.
Proof
Let \(\,J_0^{k,n}=\mathrm{ker}(\pi ^k|_{\Omega ^k(\mathcal {A}_n)})\). Then \(J_0^{k,n}\subseteq J_0^{k,n+1}\). If we let \(\,J^{k,n}=J_0^{k,n}+dJ_0^{k-1,n}\), then \(\,J^{k,n}\subseteq J^{k,n+1}\). We have
with
and
This gives a filtration on \(\,\Omega _D^\bullet (\mathcal {A})\). \(\square \)
PROPOSITION 2.4
The associated graded algebra of the filtered algebra \(\,\Omega _D^\bullet (\mathcal {A})\) is given by
Proof
By Lemma 2.3, the filtration on \(\,\Omega _D^\bullet (\mathcal {A})\) is given by \(\,\mathcal {F}_n=\bigoplus _{k\ge 0}\Omega ^k (\mathcal {A}_n)/J^{k,n}\). Hence, the associated graded algebra is given by \(\,\mathcal {G}=\bigoplus _{n\le 0}\mathcal {G}_n\), where
\(\square \)
Now we define the quantum double suspension (QDS) of \(C^*\)-algebras and spectral triples.
Notation.
-
(1)
We denote by ‘l’ the left shift operator on \(\ell ^2(\mathbb {N})\), defined on the standard orthonormal basis \(\{e_n\}\) by \(\,l(e_n)=e_{n-1}\) for \(n\ge 1\) and \(l(e_0)=0\).
-
(2)
‘N’ be the number operator on \(\ell ^2(\mathbb {N})\) defined by \(N(e_n) = ne_n\).
-
(3)
‘u’ denotes the rank one projection \(|e_0\rangle \langle e_0| := I - l^*l\,\).
-
(4)
\(\mathcal {K}\) denotes the space of compact operators on \(\ell ^2(\mathbb {N})\).
DEFINITION 2.5
[8]Let \(\mathcal {A}\) be a unital \(C^*\)-algebra. The quantum double suspension of \(\mathcal {A}\), denoted by \(\Sigma ^2{\mathcal {A}}\) is the \(C^*\)-algebra generated by \(a\otimes u\) and \(1\otimes l\) in \(\mathcal {A}\otimes \mathscr {T}\), where \(\mathscr {T}\) is the Toeplitz algebra.
There is a symbol map \(\sigma :\mathscr {T}\longrightarrow C(S^1)\) which sends l to the standard unitary generator \(\,z\) of \(C(S^1)\) and one gets the following short exact sequence
If \(\rho \) denotes the restriction of \(1\otimes \sigma \) to \(\Sigma ^2\mathcal {A}\), then one has the following short exact sequence
There is a \(\mathbb {C}\,\)-linear splitting map \(\sigma ^\prime \) from \(C(S^1)\) to \(\Sigma ^2\mathcal {A}\) which sends the standard unitary generator z of \(C(S^1)\) to \(1\otimes l\), and yields the following \(\mathbb {C}\)-vector spaces (not as algebras) isomorphism:
Notice that \(\sigma ^\prime \) is injective since it has a left inverse \(\,\rho \,\) and hence, any \(f\in C(S^1)\) can be identified with \(1\otimes \sigma ^\prime (f)\in \Sigma ^2\mathcal {A}\). For \(f=\sum _n\lambda _nz^n\in C(S^1)\), we write \(\,\sigma ^{\prime }(f):=\sum _{n\ge 0}\lambda _nl^n+ \sum _{n>0}\lambda _{-n}l^{*n}\). Now let \(\mathcal {A}\) be a dense \(\star \)-subalgebra of a \(C^*\)-algebra \(\mathbb {A}\). Define
where \(\,\mathbb {S}(\ell ^2(\mathbb {N})):=\{T=(\alpha _{ij}):\sum _{i,j}(1+i+j)^k|\alpha _{ij}|<\infty \,\,\forall \,k\ge 0\}\) is the space of Schwartz class operators on \(\ell ^2(\mathbb {N})\). Clearly, \(\,\Sigma ^2_\mathrm{alg}\mathcal {A}\) is a dense subalgebra of \(\,\Sigma ^2\mathbb {A}\) and we have the following \(\mathbb {C}\)-vector spaces (not as algebras) isomorphism at the level of subalgebra:
DEFINITION 2.6
[4]For any spectral triple \((\mathcal {A},\mathcal {H},D),\,(\Sigma ^2_\mathrm{alg}\mathcal {A},\Sigma ^2 \mathcal {H}:=\mathcal {H}\otimes \ell ^2(\mathbb {N}), \Sigma ^2D:=D\otimes I+F\otimes N)\) becomes a spectral triple, where F is the sign of the operator D and N is the number operator on \(\ell ^2(\mathbb {N})\). This is called the quantum double suspension of the spectral triple \((\mathcal {A},\mathcal {H},D)\).
It is easy to see that if \((\mathcal {A},\mathcal {H},D)\) is p-summable, then \((\Sigma ^2_\mathrm{alg}\mathcal {A},\Sigma ^2 \mathcal {H},\Sigma ^2 D)\) is a \((p+1)\)-summable spectral triple. Notice that for any \(f\in \mathbb {C}[z,z^{-1}]\,\), we have \(\,[\Sigma ^2 D,1\otimes \sigma ^\prime (f)]=F\otimes [N,f]\). The finite subalgebra \((\Sigma ^2_\mathrm{alg}\mathcal {A})_\mathrm{fin}\) is generated by \(a\otimes T\) and \(\sum _{0\le n<\infty }\lambda _nl^n+ \sum _{0<n<\infty }\lambda _{-n}l^{*n}\), where \(a\in \mathcal {A}\) and \(T\in \mathcal {B}\left( \ell ^2(\mathbb {N})\right) \) is a finitely supported matrix.
Remark 2.7
In [5], Connes represented \(\,\Omega ^\bullet (\mathcal {A})\) on \(\mathcal {B}(\mathcal {H})\) instead of on \(\mathcal {Q}(\mathcal {H})\). But the explicit computation of \(\,\Omega _{\Sigma ^2D}^\bullet ((\Sigma ^2_\mathrm{alg}\mathcal {A})_\mathrm{fin})\) is very difficult, even in the particular cases. In [2], the authors have computed \(\,\Omega _{\Sigma ^2D}^\bullet ((\Sigma ^2_\mathrm{alg}\mathcal {A})_\mathrm{fin})\) following the prescription given in Definition 2.2. The justification for this is also discussed in [2].
The computation of \(\,\Omega _{\Sigma ^2D}^\bullet ((\Sigma ^2_\mathrm{alg}\mathcal {A})_\mathrm{fin})\) has been done in [2] under the following conditions on spectral triples \((\mathcal {A},\mathcal {H},D)\).
Conditions:
-
(A)
\([D,a]F-F[D,a]\) is a compact operator for all \(a\in \mathcal {A}\,\), where \(F=\mathrm{sign}(D)\).
-
(B)
\(\mathcal {H}^\infty :=\bigcap _{k\ge 1}\mathcal {D}\mathrm{om}(D^k)\) is a left \(\mathcal {A}\)-module and \(\,[D,\mathcal {A}]\subseteq \mathcal {A}\otimes \mathcal {E}\mathrm{nd}_{\mathcal {A}}(\mathcal {H}^\infty )\subseteq \mathcal {E}\mathrm{nd}_{\mathbb {C}}(\mathcal {H}^\infty )\,\).
The notable features of these conditions are given by the following proposition.
PROPOSITION 2.8
[2]These conditions are valid for the classical case, where \(\mathcal {A}=C^\infty (\mathbb {M})\) and D is a first-order differential operator. Moreover, if a spectral triple \((\mathcal {A}\mathrm{,}\mathcal {H}\mathrm{,}D)\) satisfies these conditions, then the quantum double suspended spectral triple \((\Sigma ^2_\mathrm{alg} \mathcal {A}\mathrm{,}\Sigma ^2\mathcal {H}\mathrm{,}\Sigma ^2D)\) also satisfies them.
Notation.
-
(1)
In this article, we will work with \((\Sigma ^2_\mathrm{alg}\mathcal {A})_\mathrm{fin}\) and denote it by \(\,\Sigma ^2\mathcal {A}\,\) for notational brevity.
-
(2)
For all \(f\in \mathbb {C}[z,z^{-1}],\) we denote \(\,[N,f]\) by \(\,f^\prime \,\) for notational brevity.
-
(3)
‘\(\mathcal {S}\)’ denotes the space of finitely supported matrices in \(\mathcal {B}(\ell ^2(\mathbb {N}))\,\).
-
(4)
\((e_{ij})\) will denote infinite matrix with 1 at the ij-th place and zero elsewhere. We call it an elementary matrix.
The notion of unitary equivalence of spectral triples forms a category of spectral triples. That is, we have the following definition.
DEFINITION 2.9
The objects of the category \(\mathcal {S}\mathrm{pec}\) are spectral triples \((\mathcal {A},\mathcal {H},D)\). A morphism between two such objects \((\mathcal {A}_i, \mathcal {H}_i,D_i)\), \(i=1,2\) is a tuple \((\phi ,\Phi )\), where \(\phi :\mathcal {A}_1\rightarrow \mathcal {A}_2\) is a unital algebra morphism and \(\Phi : \mathcal {H}_1\rightarrow \mathcal {H}_2\) is a unitary which intertwines the algebra representations and the Dirac operators \(D_1,D_2\).
PROPOSITION 2.10
The association \(\mathcal {F}:(\mathcal {A},\mathcal {H},D)\longmapsto \Omega _D^\bullet (\mathcal {A})\) gives a covariant functor from \(\mathcal {S}\mathrm{pec}\) to \(\mathrm{DGA}\), the category of differential graded algebras over \(\mathbb {C}\).
Proof
Consider two objects \((\mathcal {A}_1,\mathcal {H}_1,D_1),(\mathcal {A}_2,\mathcal {H}_2,D_2)\in \mathcal {O}\mathrm{b}(\mathcal {S}\mathrm{pec})\) and suppose there is a morphism \((\phi \,,\Phi ):(\mathcal {A}_1,\mathcal {H}_1,D_1)\longrightarrow (\mathcal {A}_2,\mathcal {H}_2,D_2)\). Define
for all \(\,a_j\in \mathcal {A}_1,\,n\ge 0\,\). To show \(\Psi \) is well-defined, we must show that \(\Psi (\pi (d_1J_0^m))\subseteq \pi (d_2J_0^m)\) for all \(m\ge 1\), where \(d_1,d_2\) are the universal differentials for \(\Omega ^\bullet (\mathcal {A}_1),\Omega ^\bullet (\mathcal {A}_2)\) respectively. Observe that
Consider an arbitrary element \(\,\xi \in \pi (d_1J_0^n)\). By definition, \(\,\xi =\sum \prod _{i=0}^n[D_1,a_i]\in \pi (d_1J_0^n)\) such that \(\sum a_0 \prod _{i=1}^n[D_1,a_i]=0\). Now, using equation (2.2) and \(\Phi \) is a unitary (surjectivity is enough), we have
This shows well-definedness of \(\Psi \). Now it is easy to check that \(\Psi \) is a DGA morphism. \(\square \)
Remark 2.11
One can weaken the definition of morphism of spectral triples by demanding the map \(\Phi \) to be only linear. This was defined in [1]. But for Proposition 2.10 to hold, one requires surjectivity of \(\Phi \). However, the reason why we have assumed \(\Phi \) to be unitary will be justified in the next section.
Lemma 2.12
The quantum double suspension of a spectral triple is a covariant functor \(\Sigma ^2\) on the category \(\,\mathcal {S}\mathrm{pec}\).
Proof
Easy to verify. \(\square \)
PROPOSITION 2.13
The functor \(\Sigma ^2\) gives an equivalence \(\Sigma ^2(\mathcal {S}\mathrm{pec})\cong \mathcal {S}\mathrm{pec}\,\) of categories, and hence \(\Sigma ^2\) is not a constant functor.
Proof
Recall that as a linear space \(\Sigma ^2\mathcal {A}=\mathcal {A}\otimes \mathcal {S}\bigoplus \mathbb {C}[z,z^{-1}]\) and \(\Sigma ^2D=D\otimes 1+ F\otimes N\). Suppose \((\phi ,\Phi ):(\Sigma ^2\mathcal {A}_1,\Sigma ^2\mathcal {H}_1,\Sigma ^2D_1)\rightarrow (\Sigma ^2\mathcal {A}_2, \Sigma ^2\mathcal {H}_2,\Sigma ^2D_2)\) is an isomorphism in the sense of Definition 2.9. One can replace N by \(N+g(N)\) for a suitable function g such that \((\Sigma ^2\mathcal {A},\Sigma ^2\mathcal {H},D\otimes I+F\otimes (N+g(N)))\) remains an honest spectral triple, and computations done in [2] does not get affected. It is possible to choose such a function g so that the following map
becomes one-to-one. This is possible since \(D_1\) has discrete spectrum. This will imply that any unitary \(\tilde{\Phi }:\mathcal {H}_1\otimes \ell ^2 (\mathbb {N})\rightarrow \mathcal {H}_2\otimes \ell ^2(\mathbb {N})\) is of the form \(\Phi \otimes 1\), where \(\Phi :\mathcal {H}_1\rightarrow \mathcal {H}_2\) is a unitary. This will assure that algebra isomorphism \(\tilde{\phi }:\Sigma ^2\mathcal {A}_1\rightarrow \Sigma ^2\mathcal {A}_2\) is of the form \(\phi \otimes 1\bigoplus 1\), where \(\phi :\mathcal {A}_1\rightarrow \mathcal {A}_2\) is an algebra isomorphism. This shows that
The other implication ‘\(\Leftarrow \)’ is obvious. \(\square \)
Recall Theorem (3.22) from [2].
Theorem 2.14
[2]. For the spectral triple \((\Sigma ^2\mathcal {A}\mathrm{,}\Sigma ^2\mathcal {H}\mathrm{,}\Sigma ^2D)\), we have
-
(1)
\(\Omega _{\Sigma ^2 D}^1(\Sigma ^2 \mathcal {A})\cong \Omega _D^1(\mathcal {A})\otimes \mathcal {S}\bigoplus \Sigma ^2\mathcal {A}\,\).
-
(2)
\(\Omega _{\Sigma ^2 D}^n(\Sigma ^2 \mathcal {A})\cong \Omega _D^n(\mathcal {A})\otimes \mathcal {S}\), for all \(n\ge 2\).
-
(3)
The differential \(\,\,\delta ^0:\Sigma ^2 \mathcal {A}\longrightarrow \Omega _{\Sigma ^2 D}^1(\Sigma ^2\mathcal {A})\,\) is given by
$$\begin{aligned} a\otimes T+f\longmapsto [D\mathrm{,}a]\otimes T\bigoplus \left( a\otimes [N\mathrm{,}T]+f^\prime \right) . \end{aligned}$$ -
(4)
The differential \(\,\,\delta ^1:\Omega _{\Sigma ^2 D}^1(\Sigma ^2\mathcal {A})\longrightarrow \Omega _{\Sigma ^2 D}^2 (\Sigma ^2\mathcal {A})\,\) is given by
$$\begin{aligned} \delta ^1|_{\Omega _D^1(\mathcal {A})\otimes \mathcal {S}}=d^1\otimes 1 \quad \hbox { and }\quad \delta ^1|_{\Sigma ^2 \mathcal {A}}=0. \end{aligned}$$ -
(5)
The differential \(\,\,\delta ^n:\Omega _{\Sigma ^2 D}^n(\Sigma ^2 \mathcal {A})\longrightarrow \Omega _{\Sigma ^2 D}^{n+1} (\Sigma ^2\mathcal {A})\) is given by \(\,\delta ^n=d^n\otimes 1\) for all \(n\ge 2\).
Here, \(d:\Omega _D^\bullet (\mathcal {A})\longrightarrow \Omega _D^{\bullet +1}(\mathcal {A})\) is the differential of the Dirac DGA.
Remark 2.15
The DGA \(\Omega _{\Sigma ^2 D}^\bullet (\Sigma ^2\mathcal {A})\) can be described alternatively as follows. Notice that for the (graded) algebra \(\Omega _D^\bullet (\mathcal {A})\), one can consider \(\Sigma ^2(\Omega _D^\bullet (\mathcal {A}))=\Omega _D^\bullet (\mathcal {A})\otimes \mathcal {S}\bigoplus \mathbb {C}[z,z^{-1}]\). This is a graded algebra whose degree zero term is \(\mathcal {A}\otimes \mathcal {S} \bigoplus \mathbb {C}[z,z^{-1}]=\Sigma ^2\mathcal {A}\) and the degree n term is \(\,\Omega _D^n(\mathcal {A})\otimes \mathcal {S}\) for \(n\ge 1\). That is,
as a graded algebra. Then, as a graded algebra \(\,\Omega _{\Sigma ^2 D}^\bullet (\Sigma ^2 \mathcal {A})=\Sigma ^2 (\Omega _D^\bullet (\mathcal {A}))\bigoplus \Sigma ^2\mathcal {A}\), where \(\Sigma ^2\mathcal {A}\) sits in the degree 1 term.
COROLLARY 2.16
The cohomology of \((\Omega _{\Sigma ^2 D}^\bullet (\Sigma ^2\mathcal {A}),\delta ^\bullet \) is given by
-
(1)
\(H^0(\Sigma ^2\mathcal {A})=H^0(\mathcal {A})\otimes \mathcal {S}_\mathrm{diag}\bigoplus \mathbb {C}\),
-
(2)
\(H^1(\Sigma ^2\mathcal {A})=H^1(\mathcal {A})\otimes \mathcal {S}_\mathrm{diag}\bigoplus \mathcal {A}\otimes \mathcal {S}_\mathrm{diag}\bigoplus \mathrm{Ker}(d^1)\otimes \mathcal {S}_\mathrm{off}\bigoplus \mathbb {C}\),
-
(3)
\(H^n(\Sigma ^2\mathcal {A})=H^n(\mathcal {A})\otimes \mathcal {S}\), for all \(n\ge 2\),
where \(H^\bullet (\mathcal {A})\) denotes the cohomology of \((\,\Omega _D^\bullet (\mathcal {A}),d^\bullet )\) and \(\,\mathcal {S}_{\mathrm{diag}}\,\mathrm{,}\mathcal {S}_\mathrm{off}\) denote the spaces of finitely supported diagonal and off-diagonal matrices respectively.
Proof
We have \(H^0(\Sigma ^2\mathcal {A})=\mathrm{Ker}(\delta ^0)\). A general element of \(\mathcal {A}\otimes \mathcal {S}\) can be written in terms of elementary matrices of the form \(a\otimes T=\sum _{i,j}a_{ij}\otimes e_{ij}\). We now have the following:
This proves part (1). For part (2), observe that
and \(\mathrm{Im}(\delta ^0)=\mathrm{Im}(\delta ^0|_{\mathcal {A}\otimes \mathcal {S}})\bigoplus \mathbb {C}[z,z^{-1}]/\mathbb {C}\). Hence, we need to determine \(\frac{\mathrm{Ker}(d^1)\otimes \mathcal {S}\bigoplus \mathcal {A}\otimes \mathcal {S}}{\mathrm{Im}(\delta ^0|_{\mathcal {A}\otimes \mathcal {S}})}\). Now
Hence \(\delta ^0=\delta ^0_1\oplus \delta ^0_2\). Observe that \(\mathrm{Im}(\delta ^0_2)=\mathrm{Im}(d^0)\otimes \mathcal {S}_\mathrm{diag}\). Now
is a well-defined linear isomorphism. Hence,
This proves part (2), and part (3) is easy to verify. \(\square \)
If \(\mathcal {A}\) comes with a decreasing filtration
then the algebra \(\Sigma ^2\mathcal {A}\) has the induced filtration. By Lemma (2.3), \(\,\Omega _{\Sigma ^2 D}^\bullet (\Sigma ^2\mathcal {A})\) then becomes a filtered algebra.
PROPOSITION 2.17
The associated graded algebra of the filtered algebra \(\,\Omega _{\Sigma ^2 D}^\bullet (\Sigma ^2\mathcal {A})\) is
Hence, \(\mathcal {G}(\Sigma ^2\mathcal {A})\) depends only on the filtration of \(\mathcal {A}\).
Proof
By Lemma 2.4, the associated graded algebra is
For \(p=0\),
and for \(p\ge 2\),
by Propositions (3.8) and (3.10) in [2]. Finally, for \(p=1\),
by part (1) of Theorem 3.20 in [2]. Hence, our claim follows. \(\square \)
3 FGR DGA for the quantum double suspension
In this section, our objective is to compute the DGA of Fröhlich et al. for the quantum double suspension. We first recall its definition from [7].
DEFINITION 3.1
For any p-summable spectral triple \((\mathcal {A},\mathcal {H},D)\), consider the following functional:
Let
Then,
is a differential graded algebra called the FGR DGA.
Remark 3.2
-
(1)
Note that for a p-summable spectral triple \((\mathcal {A},\mathcal {H},D)\),
$$\begin{aligned} \mathrm{lim}_{\lambda \rightarrow \infty }\left( \frac{1}{\lambda }\mathrm{Tr}(T\mathrm{e}^{-\lambda ^{-2/p}D^2})\right) =\Gamma \Bigg (\frac{p}{2}+1\Bigg )\mathrm{Tr}_\omega (T|D|^{-p}) \end{aligned}$$for all \(T\in \mathcal {B}(\mathcal {H})\) ([5], page 563). Hence, the functional considered in equation (3.3) is nothing but the Dixmier trace \(\mathrm{Tr}_\omega \) up to a positive constant.
-
(2)
Since, for any compact operator \(K\in \mathcal {K}(\mathcal {H}),\,\mathrm{Tr}_\omega (K|D|^{-p})=0\), the functional in (3.3) is well-defined on \(\,\pi (\Omega ^\bullet (\mathcal {A}))\subseteq \mathcal {B}(\mathcal {H})/\mathcal {K}(\mathcal {H})\).
-
(3)
For the classical case of manifolds and the noncommutative torus, \(K^n=J_0^n\) (Definition 2.2). Hence, the FGR DGA coincides with the Dirac DGA in these cases [7].
Lemma 3.3
The association \(\mathcal {G}:(\mathcal {A},\mathcal {H},D)\longmapsto \tilde{\Omega }_D^\bullet (\mathcal {A})\) gives a covariant functor from \(\mathcal {S}\mathrm{pec}\) to DGA, the category of differential graded algebras over \(\mathbb {C}\).
Proof
Consider two objects \((\mathcal {A}_1,\mathcal {H}_1,D_1),(\mathcal {A}_2,\mathcal {H}_2,D_2)\in \mathcal {O}\mathrm{b}(\mathcal {S}\mathrm{pec})\) and suppose there is a morphism \((\phi \,,\Phi ):(\mathcal {A}_1,\mathcal {H}_1,D_1)\longrightarrow (\mathcal {A}_2,\mathcal {H}_2,D_2)\). Define
for all \(\,a_j\in \mathcal {A}_1,\,n\ge 0\). To show \(\Psi \) is well-defined, we must show that \(\Psi (\pi _1(K_1^m))\subseteq \pi _2(K_2^m)\) for all \(m\ge 0\). Observe that
Now, \(\,\Phi D_1=D_2\Phi \) will imply that \(\,\Phi \mathrm{e}^{-tD_1^2}\Phi ^*=\mathrm{e}^{-tD_2^2}\). Let us denote
Now,
and \(\mathrm{Tr}(\mathrm{e}^{-tD_1^2})=\mathrm{Tr}(\mathrm{e}^{-tD_2^2})\). This proves that \(\Psi (\pi _1(K_1^m))=\pi _2(K_2^m)\), i.e. \(\Psi \) is well-defined, and one can check that it is a DGA morphism. \(\square \)
Remark 3.4
-
(1)
Although, surjectivity of \(\Phi \) is enough to ensure that Dirac DGA is a functor, it fails in this case of FGR DGA. This is the reason we have chosen \(\Phi \) to be unitary. Unless \(\Phi \) is both one-to-one and onto, it is not guaranteed that \(\Psi (\pi _1(K_1^m))\subseteq \pi _2(K_2^m)\).
-
(2)
One may come up with a different definition of the category \(\,\mathcal {S}\mathrm{pec}\) of spectral triples which allows larger set of morphisms than ours; such that both the Dirac DGA and FGR DGA become functors. Here we stress the point that it will not contradict our main result in this article as we shall see. Because of this reason, we have chosen the simplest possible definition for the category \(\mathcal {S}\mathrm{pec}\).
To make the computation possible, we need to use the functional in (3.3) in a different disguise, namely
The well-definedness of this functional follows from the next lemma.
Lemma 3.5
Let \((\mathcal {A},\mathcal {H},D)\) be a p-summable spectral triple. Then the functional \(\oint \) is equal to the Dixmier trace up to a positive constant (which depends only on p).
Proof
Recall the following equality
proved in [10] for any \(B\in \mathcal {B}(\mathcal {H})\). Now take \(q=1/p\) and \(A=|D|^{-p}\). \(\square \)
COROLLARY 3.6
For any \(\,T_1\otimes T_2\in \mathcal {B}(\mathcal {H}\otimes \ell ^2(\mathbb {N}))\),
Remark 3.7
It is this corollary which makes the computation in this section possible. Moreover, since both the functionals \(\int \) and \(\oint \) become equal up to a constant, and we are interested in the spaces \(K^n\) in Definition 3.3, it is absolutely permissible to choose \(\oint \) over \(\int \).
Lemma 3.8
\(K^0(\Sigma ^2\mathcal {A})=\mathcal {A}\otimes \mathcal {S}\).
Proof
Choose arbitrary element \(\sum _ka_k\otimes T_k\) of \(\mathcal {A}\otimes \mathcal {S}\). In terms of elementary matrices, we can write \(\,T_k=\sum _{i,j} \,\alpha _{ij}^{(k)}e_{ij}\,\). Then
Hence \(\mathcal {A}\otimes \mathcal {S}\subseteq K^0(\Sigma ^2\mathcal {A})\). Now, for arbitrary \(\,\sum _ka_k\otimes T_k+f\in K^0(\Sigma ^2\mathcal {A})\),
because the same calculation as above proves that both \(\oint f(\sum _k a_k\otimes T_k)^*\) and \(\oint (\sum _k a_k\otimes T_k)f^*\) are zero. For any \(f\in \mathbb {C}[z,z^{-1}]\), \(\oint ff^*\) is just the integration of the function \(\,ff^*\equiv |f|^2\,\) against the Haar measure on \(S^1\). This shows that \(f=0\), i.e. \(K^0(\Sigma ^2 \mathcal {A})\subseteq \mathcal {A}\otimes \mathcal {S}\). \(\square \)
Remark 3.9
In Assumption 2.13, page 131 in [7], the authors have assumed that \(K^0=\{0\}\). Lemma 3.8 shows that this is never true in the case of quantum double suspension.
Lemma 3.10
\(\oint (F\otimes 1)\pi (\omega )=0\) for any \(\,\omega \in \Omega ^1(\mathcal {A}\otimes \mathcal {S})\).
Proof
Let
Then, using elementary matrices \((e_{ij})\), we have
and this concludes the proof. \(\square \)
Lemma 3.11
\(\pi (K^1(\Sigma ^2\mathcal {A}))=\pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\bigoplus \pi (K^1(\mathbb {C}[z,z^{-1}]))\).
Proof
We first prove that \(\pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S})\subseteq \pi (K^1(\Sigma ^2\mathcal {A}))\). The arbitrary element of \(\pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\) looks like \(\,\pi (\omega )=\sum _k(a_{0k}\otimes T_{0k})[\Sigma ^2D,a_{1k} \otimes T_{1k}]\). Then, using elementary matrices \((e_{ij})\), we get
and
Let \(T_{kjiq}=a_{0kij}[D,a_{1kjq}]+Fa_{0kij}a_{1kjq}(j-q)\). Now
Hence, \(\,\pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\bigoplus \pi (K^1(\mathbb {C}[z,z^{-1}]))\subseteq \pi (K^1(\Sigma ^2 \mathcal {A}))\).
To show the converse, choose an arbitrary element \(\pi (\omega )=\sum _k(a_{0k}\otimes T_{0k}+f_{0k})[\Sigma ^2D,a_{1k}\otimes T_{1k}+f_{1k}]\,\) in \(\pi \left( K^1(\Sigma ^2\mathcal {A})\right) \). Then
where \(\pi (\tilde{\omega })\in \pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\). Hence \(\,\pi (\omega )^*=\pi (\tilde{\omega })^*+ \sum _k F\otimes (f_{0k}f_{1k}^\prime )^*\). Since \(\,\pi (\omega )\in \pi \left( K^1(\varSigma ^2 \mathcal {A})\right) \), we have
This shows that \(\oint (\sum _k f_{0k}f_{1k}^\prime )^*(\sum _k f_{0k}f_{1k}^\prime )=0\) (using Lemma 3.10). That is, \(\sum _kF\otimes f_{0k}f_{1k}^\prime \in \pi (K^1(\mathbb {C}[z,z^{-1}]))\). Hence \(\,\pi (K^1(\Sigma ^2\mathcal {A})) \subseteq \pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\bigoplus \pi (K^1(\mathbb {C}[z,z^{-1}]))\). \(\square \)
PROPOSITION 3.12
\(\tilde{\Omega }_{\Sigma ^2 D}^1(\Sigma ^2 \mathcal {A})\cong \mathbb {C}[z,z^{-1}]\,\) as \(\Sigma ^2\mathcal {A}\,\)-bimodule.
Proof
We have \(\tilde{\Omega }_{\Sigma ^2 D}^1(\Sigma ^2\mathcal {A})\cong \pi (\Omega ^1(\Sigma ^2\mathcal {A}))/(\pi (K^1 (\Sigma ^2\mathcal {A}))+\pi (\mathrm{d}K^0(\Sigma ^2\mathcal {A})))\). But \(\pi (\mathrm{d}K^0(\Sigma ^2\mathcal {A}))\subseteq \pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\) and \(K^0(\mathbb {C}[z,z^{-1}])=\{0\}\). This says that
Here, the first isomorphism follows from the fact that (see Proposition 3.8 in [2])
and we refer [3] for the following fact:
\(\square \)
Remark 3.13
Recall that \(\Sigma ^2\mathcal {A}\cong \mathcal {A}\otimes \mathcal {S}\bigoplus \mathbb {C}[z,z^{-1}]\,\) as \(\,\mathbb {C}\,\)-vector spaces, where \(\mathbb {C}[z,z^{-1}]\) is identified with the quotient \(\Sigma ^2\mathcal {A}/\mathcal {A}\otimes \mathcal {S}\). These direct sum and isomorphism are also as \(\Sigma ^2\mathcal {A}\)-bimodule. Hence, \(\tilde{\Omega }_{\Sigma ^2 D}^1\left( \Sigma ^2\mathcal {A}\right) \,\) is always a finitely generated projective \(\Sigma ^2\mathcal {A}\)-bimodule (compare with Assumption 2.13 in [7], page 131).
Lemma 3.14
\(\oint \pi (\omega )=0\) for any \(\,\omega \in \Omega ^n(\mathcal {A}\otimes \mathcal {S})\) and for all \(n\ge 2\).
Proof
Recall Lemma 3.15 from [2] which says that
Hence, for \(\omega \in \Omega ^n(\mathcal {A}\otimes \mathcal {S})\), we have \(\,\pi (\omega )=\sum _{r=0}^n\sum _kF^r\pi (v_{r,k})\otimes T_{r,k}\,\) with \(v_{r,k}\in \Omega ^{n-r}(\mathcal {A})\). Writing each \(T_{r,k}\) in terms of elementary matrices \((e_{ij})\), we get that
Then
and we are done. \(\square \)
Lemma 3.15
\(\pi (K^n(\Sigma ^2 \mathcal {A}))=\pi (\Omega ^n(\mathcal {A}\otimes \mathcal {S}))\bigoplus \pi (K^n(\mathbb {C}[z,z^{-1}]) ),\) for all \(n\ge 2\).
Proof
Note that for any algebra \(\mathcal {A}\), we have
Lemma 3.11 proves that \(\pi (\Omega ^1(\mathcal {A}\otimes \mathcal {S}))\subseteq \pi (K^1(\Sigma ^2 \mathcal {A}) )\). Since
we get that
because \(K^\bullet \) is a graded ideal in \(\,\Omega ^\bullet \). Hence, we have the inclusion ‘\(\supseteq \)’. Now, recall Proposition 3.8 from [2], which says that
Since \(K^n\subseteq \Omega ^n\), using Lemma 3.14, we get the inclusion ‘\(\subseteq \)’ and this completes the proof. \(\square \)
Theorem 3.16
For \((\Sigma ^2\mathcal {A},\Sigma ^2 \mathcal {H},\Sigma ^2 D)\),
-
(1)
\(\tilde{\Omega }_{\Sigma ^2 D}^n(\Sigma ^2\mathcal {A})=\mathbb {C}[z,z^{-1}]\) for \(n=0,1;\)
-
(2)
\(\tilde{\Omega }_{\Sigma ^2 D}^n(\Sigma ^2\mathcal {A})=0\) for all \(n\ge 2\).
Proof
Part (1) follows from Lemma 3.8 and Proposition 3.12. Now, Lemmas 3.11 and 3.15 shows that for all \(n\ge 1\),
But \(J_0^n(\Sigma ^2 \mathcal {A})\subseteq K^n(\Sigma ^2 \mathcal {A})\). Hence, equation (3.6) reduces to
So, for all \(n\ge 1\),
and consequently for all \(\,n\ge 1\),
Recall Propositions 3.8 and 3.10 from [2], which say that
and
Hence, equation (3.9) turns out to be
Finally, using equations (3.7), (3.10) and (3.12), we have for all \(n\ge 2\),
Now, the facts that \(\pi _N(\Omega ^n(\mathbb {C}[z,z^{-1}]))=\mathbb {C}[z,z^{-1}]\) and \(\pi _N(\mathrm{d}J_0^{n-1}(\mathbb {C}[z,z^{-1}]))= \mathbb {C}[z,z^{-1}]\) for all \(n\ge 2\) (see Lemmas 3.11 and 3.12 in [2]) completes part (2). \(\square \)
In view of Theorems 2.14 and 3.16, the conclusion of this article comes as the following final theorem.
Theorem 3.17
There is a category \(\mathcal {S}\mathrm{pec}\) of spectral triples such that the Dirac DGA, the FGR DGA and the quantum double suspension, denoted by \(\,\mathcal {F},\mathcal {G},\Sigma ^2\) respectively, become covariant functors. Let \(\,\mathcal {C}\) be the subcategory of commutative spectral triples. Restricted to \(\mathcal {C}\), both the functor \(\mathcal {F}\) and \(\mathcal {G}\) are equal to the de-Rham DGA. Unlike \(\mathcal {F}\circ \Sigma ^2\), the functor \(\mathcal {G}\circ \Sigma ^2\) becomes a constant functor on \(\mathcal {S}\mathrm{pec}\).
References
Bertozzini P, Conti R and Lewkeeratiyutkul W, A category of spectral triples and discrete groups with length function, Osaka J. Math. 43(2) (2006) 327–350
Chakraborty P S and Guin S, Connes’ calculus for the quantum double suspension, J. Geom. Phys. 88 (2015) 16–29
Chakraborty P S and Pal A, Spectral triples and associated Connes–de Rham complex for the quantum SU(2) and the quantum sphere, Comm. Math. Phys. 240(3) (2003) 447–456
Chakraborty P S and Sundar S, Quantum double suspension and spectral triples, J. Funct. Anal. 260(9) (2011) 2716–2741
Connes A, Noncommutative geometry (1994) (San Diego, CA: Academic Press Inc.)
Connes A and Moscovici H, The local index formula in noncommutative geometry, Geom. Funct. Anal. 5(2) (1995) 174–243
Fröhlich J, Grandjean O and Recknagel A, Supersymmetric quantum theory and non-commutative geometry, Comm. Math. Phys. 203(1) (1999) 119–184
Hong J H and Szymański W, Quantum spheres and projective spaces as graph algebra, Comm. Math. Phys. 232(1) (2002) 157–188
Hong J H and Szymański W, Noncommutative balls and mirror quantum spheres, J. London Math. Soc. 77(3) (2008) 607–626
Sukochev F and Zanin D, \(\zeta \)-function and heat kernel formulae, J. Funct. Anal. 260(8) (2011) 2451–2482
Acknowledgements
The first author (PSC) acknowledges support of DST, India through Swarnajayanti Fellowship Award Project No. DST/SJF/MSA-01/2012-13 and the second author (SG) acknowledges support of DST, India through INSPIRE Faculty Award (Award No. DST/INSPIRE/04/2015/000901).
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by B V Rajarama Bhat.
Rights and permissions
About this article
Cite this article
Chakraborty, P.S., Guin, S. Comparison between two differential graded algebras in noncommutative geometry. Proc Math Sci 129, 29 (2019). https://doi.org/10.1007/s12044-019-0467-y
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s12044-019-0467-y
Keywords
- Dirac differential graded algebra
- Connes’ calculus
- FGR differential graded algebra
- spectral triple
- quantum double suspension