Spectral action on noncommutative torus (Q2475511)

The authors compute the spectral action on noncommutative tori with the help of zeta functions. We briefly recall the important notion of spectral action in noncommutative geometry. Roughly speaking, for a given spectral triple \(( \mathcal{A},\mathcal{H},\mathcal{D}) \) where \( \mathcal{A}\) is an algebra acting on a Hilbert space \(\mathcal{H}\) and \( \mathcal{D}\) is a Dirac-like operator, the spectral action \(\mathcal{S} ( \mathcal{D}_{A},\Phi ,\Lambda ) :=\text{Tr}( \Phi ( \mathcal{D}_{A}/\Lambda ) ) \) of the covariant Dirac operator \( \mathcal{D}_{A}:=\mathcal{D}+A+\varepsilon JAJ^{-1}\) for a 1-form \( A=\sum_{i}a_{i}[ \mathcal{D},b_{i}] \) with \(a_{i},b_{i}\in \mathcal{A}\), an even positive cut-off function \(\Phi \), and the mass scale \( \Lambda \), counts the spectral values of \(| \mathcal{D}_{A}| \) less than \(\Lambda \), where \(J\) is a real structure with \(J\mathcal{D} =\varepsilon \mathcal{D}J\) for \(\varepsilon \in \{ \pm 1\} \). The smooth noncommutative \(n\)-torus \(C^{\infty }( \mathbb{T}_{\Theta }^{n}) \) determined by a nonzero skew-symmetric matrix \(\Theta \in M_{n}( \mathbb{R}) \) consists of \(\sum_{k\in \mathbb{Z}^{n}}a_{k}U_{k}\) with a trace \(\tau :\sum_{k\in \mathbb{Z}^{n}}a_{k}U_{k}\mapsto a_{0}\), where \( k\mapsto a_{k}\) is a Schwartz function on \(\mathbb{Z}^{n}\) and \(U_{k}:=e^{- \frac{i}{2}k\cdot \chi k}u_{1}^{k_{1}}\cdots u_{n}^{k_{n}}\) for unitary generators \(u_{i}\), \(1\leq i\leq n\), satisfying \(u_{i}u_{j}=e^{i\Theta _{ij}}u_{j}u_{i}\) and the upper triangular restriction \(\chi \) of \(\Theta \). A matrix \(\Theta \in M_{n}( \mathbb{R}) \) is called diophantine if \( v:=^{t}\Theta ( u) \) is a diophantine vector for some \(u\in \mathbb{Z }^{n}\), i.e., there are \(\delta ,c>0\) such that \(| w\cdot v-m| \geq c| w| ^{-\delta }\) for all \(0\neq w\in \mathbb{Z}^{n}\) and \( m\in \mathbb{Z}\). For \(\mathcal{A}=C^{\infty }(\mathbb{T}_{\Theta}^{n})\) with \(\frac{1}{2\pi}\Theta\) real skew-symmetric diophantine and a selfadjoint 1-form \(A\), the authors show that \[ \begin{aligned} \mathcal{S}(\mathcal{D}_{A},\Phi,\Lambda) &= \begin{cases} 4\pi\Phi_{2}\Lambda^{2}+\mathcal{O}(\Lambda^{-2}) \quad&\text{for }n=2,\\ 8\pi^{2}\Phi_{4}\Lambda^{4}-\frac{4\pi ^{2}}{3}\Phi(0) \tau(F_{\mu\nu}F^{\mu\nu})+\mathcal{O}(\Lambda^{-2}) \quad&\text{for }n=4,\end{cases}\\ \text{and in general,} &=\sum_{k=0}^{n}\Phi _{n-k}c_{n-k}( A) \Lambda ^{n-k}+\mathcal{O}( \Lambda ^{-1}) \\ &\qquad\qquad\text{with }c_{n-2}( A) =0=c_{n-k}( A) \text{ for }k\text{ odd}, \\ &\qquad\qquad\qquad\text{where }\Phi _{k}:=\frac{1}{2} \int_{0}^{\infty }\Phi ( t) t^{\frac{k}{2}-1}\,dt.\end{aligned} \] Furthermore, when \(n\) is odd, \(\zeta _{D_{A}}( 0) -\zeta _{D}( 0) =0 \) for \(D:=\mathcal{D}+P_{0}\) and \(D_{A}:=\mathcal{D}_{A}+P_{A}\), where \(P_{0} \) and \(P_{A}\) are the projections onto \(\ker ( \mathcal{D}) \) and \(\ker ( \mathcal{D}_{A}) \), respectively, and \(\zeta_{X}( s) :=\text{Tr}(| X|^{-s})\).