The Gauss-Bonnet theorem for noncommutative two tori with a general conformal structure (Q446421)

Let \(\Sigma\) be a closed oriented 2-dimensional manifold equipped with a Riemannian metric \(g\), \(\zeta(s)\) the spectral zeta function of \(\Delta_g= d^*d\). Then the Gauss-Bonnet theorem says \[ \zeta(0)= \dim\ker\Delta_g= {1\over 6}\chi(\Sigma). \] Extension of this theorem for the noncommutative two torus was presented in [\textit{A. Connes} and \textit{P. Tretkoff}, The Gauss-Bonnet theorem for the noncommutative two torus. Noncommutative geometry, arithmetic, and related topics. Baltimore, MD: Johns Hopkins University Press. 141--158 (2011; Zbl 1251.46037)] (hereafter referred as [1]). Let \(U\), \(V\) be the generators of the noncommutative two torus \(\mathbb{T}_\theta\); \(VU= e^{2\pi i\theta}UV\), \(\theta\) is an irrational number, and \(A^\infty_\theta\) is the algebra of smooth functions on \(\mathbb{T}_\theta\). Then introducing derivations \(\delta_1\), \(\delta_2\); \[ \delta_1(U)= U,\;\delta_1(V)=0,\quad \delta_2(U)= 0,\;\delta_2(V)= V, \] a complex structure \(\partial\), \(\partial^*\) is defined by \[ \partial=\delta_1+ i\delta_2,\quad\partial^*= \delta_1- i\delta_2 \] in [1]. These are reviewed in \S2 of this paper. Let \(\Delta\) be \(\partial^*\partial\), \(\zeta(s)\) its spectral \(\zeta\) function, then showing \(\zeta(0)+ 1= 0\), [1] proves that the value of the zeta function of the operator \(\Delta'= k\Delta k\sim\Delta\) is independent of \(k\). Here \(k\) is an invertible positive element of \(A^\infty_\theta\). Since \(\chi(\mathbb{T}^2)= 0\), Connes-Tretkoff say this is the Gauss-Bonnet theorem for the noncommutative 2-dimensional torus. In this paper, taking \[ \partial= \delta_1+ \tau\delta_2,\quad \partial^*= \delta_1+ \overline\tau\delta_2, \] \(\tau=\tau_1+ i\tau_2\), \(\tau_2> 0\), that is considering 2-dimensional noncommutative torus corresponding to the arbitrary 2-dimensional torus, this result is extended for the spectral \(\zeta\)-function of \(\Delta= \partial^*\partial\) (\S5, Th. 5.1). The authors remark this is already indicated in [1]. Proof also follows [1]. Since \[ \zeta(s)= {1\over\Gamma(s)} \int^\infty_0 \text{Trace}^+(e^{-t\Delta'})\,t^{s-1}\,dt, \] by using symbol expansion \(b_0(\xi,\lambda)+ b_1(\xi, \lambda)+\cdots\) of the inverse of \((\Delta'-\lambda)\), we have \[ \zeta(0)+ 1=- \int \tau_0(b_2(\xi, -1))\,d\xi, \] where \(\tau_0(\sum_{m,n} a_{m,n} U^m V^n)= a_{0,0}\) (\S3). For arbitrary \(\tau\), computation of \(b_2\) is impossible without using a computer. The authors say it would occupy eight full pages, which is not presented in this paper. They try to abridge it by using a computer (\S4). We can not check validity of these abridgments (one of them occupies four pages!), without using a computer. But finally, as in [1], relying symbolic calculus \[ \begin{multlined} \zeta(0)+ 1= {2\pi\over\tau_2} \phi(f(\Delta)(\delta_1(k)) \delta_2(k))+ {2\pi|\tau|^2\over\tau_2} \phi(f(\Delta) (\delta_2(k))\delta_2(k))+\\ {2\tau_1\over\tau_2} \phi(f(\Delta)(\delta_1(k))\delta_2(k))+ {2\pi\tau_1\over\tau_2} \phi(f(\Delta)(\delta_2(k))\delta_1(k)),\end{multlined} \] is derived. This is much more complicated than the case \(\tau=i\) treated in [1] (Lemma 5.1. cf. Lemma 3.2 of [1]). But showing \[ \phi(f(\Delta)(\delta_i(k)) \delta_j(k))= -\phi(f(\Delta)(\delta_j(k)) \delta_i(k), \] \(\zeta(0)+ 1=0\) is derived from this formula. Hence \(\zeta(0)\) is independent of \(k\) for arbitrary \(\tau\) (\S5, the last section).