From large \(N\) matrices to the noncommutative torus (Q5940466)

By using the realization of the \(C^*\)-algebra \(A_\theta\), which is the norm closure of the algebra of smooth functions \({\mathcal A}_\theta\) of the noncommutative two-torus \({\mathbf T}^2_\theta\), as a subalgebra of approximately finite dimensional \(C^*\)-algebra by \textit{M. Pimsner} and \textit{D. Voiculescu} [J. Oper. Theory 4, 201-210 (1980; Zbl 0525.46031)], approximations of correlation functions for field theories on the noncommutative torus in terms of expectation values constructed matrices acting on a finite dimensional vector space (Section 3), and connections of bundles on the noncommutative torus in terms of connections of matrix models (Section 4) are given. \({\mathcal A}_\theta\) is the unital *-algebra generated by two unitary elements \(U_1\), \(U_2\) with the relation \(U_1U_2=e^{2\pi i\theta} U_2U_1\). If \(\theta=M/N\) is a rational number, \({\mathcal A}_{M/N}\) is a twisted matrix bundle over \(C^\infty ({\mathbf T}^2)\) of topological charge \(M\) whose fibres are \(N\times N\) complex matrix algebras. Using this fact and continued fraction expansion of irrational numbers, Pimsner-Voiculescu constructed a tower of matrix algebras \(A_n\), \(A_n= \mathbb{M}_{q_n} (\mathbb{C})\oplus \mathbb{M}_{q_{n-1}} (\mathbb{C})\), \(q_n\) and \(q_{n-1}\) are determined by the continued fraction expansion of \(\theta\), with the embeddings \(\rho_n: A_{n-1}\to A_n\), and showed there is a unital injective *-morphism \(\rho: A_\theta \to A_\infty\), \(A_\infty= \overline {\cup^\infty_{n=0} A_n}\). \(A_\infty\) can be visualized as the algebra of infinite matrices and the Hilbert space \({\mathcal H}_\infty\) on which \(A_\infty\) is represented as the algebra of bounded operators is also constructed. To approximate field theory and gauge theory on the noncommutative torus by matrix models, realization of \(A_\theta\) as the subalgebra of \(A_\infty\) for rational \(\theta\) is also necessary. So in this paper, Pimsner-Voiculescu realization is precisely explained including the case when \(\theta\) is a rational number (Section 2). In Section 3, approximation of correlation function \(C(\theta)= \langle\psi_n', {\mathcal G}\psi\rangle\), \({\mathcal G}\) is an operator in \(A_\theta\) and \(\psi',\psi\) are states in \({\mathcal H}_\infty\), by \(C_n(\theta_n)= \langle\psi_n', {\mathcal G}_n\psi_N \rangle_{{\mathcal H}_n}\), where \({\mathcal G}_n\) is a matrix obtained replacing \(U_a\) by direct sum of clock or shift operators in \(\mathbb{M}_{q_n}\) and \(\mathbb{M}_{q_{n-1}}\), is given (Proposition 3). A connection \(\nabla_a\) on a vector bundle \({\mathcal H}\) over the noncommutative torus is a Hermitian operator acting on \({\mathcal H}\) such that \([\nabla_a, U_b]=2\pi \delta_{ab}U_b\), \(a,b=1,2\). The authors say to approximate this property by finite dimensional matrices is impossible, in general. So an exponential version is considered, and it is proved the existence of matrices \(\nabla_a^{(n)}\) such that \(e^{i\nabla_a^{(n)}}\in A_n\) norm converges to \(e^{i \nabla a}\) (Proposition 3). Arguments from \(K\)-theory and Morita equivalence in the approximation of the noncommutative torus by the tower of matrices are also explained in Section 2. The paper is concluded remarking that there remains some difficulty to generalize the results of this paper to higher dimensional noncommutative tori.