Matrix product states, random matrix theory and the principle of maximum entropy (Q2376337): Difference between revisions

The principle of maximum entropy postulates that among all possible probability distributions compatible with the prior information the best choice is the one which maximizes the Shannon entropy. This principle can be extended to more general situations. In this paper, the case of 1D spin states is analyzed. Ground states of quantum spin chains are well represented by matrix product states with the polynomial bond dimension. Let a bulk state of a chain with boundary effects in exponentially small regions of size \(b\) at both ends, homogeneity in the bulk and experimental access to an exponentially small central region of size \(l\) is given by \[ \rho = \sum^d_{i_{b+1}, \dots, i_{N-b}, \atop j_{b+1}, \dots, j_{N-b} = 1 } \text{tr}(LA_{i_{b+1}} \dots A_{i_{N-b}} R A^{\dagger}_{j_{N-b}} \dots A^{\dagger}_{j_{b + 1}})|i_{b+1} \dots i_{N-b}\rangle\langle j_{b+1} \dots j_{N-b}|, \] where \(A_i, L \geq 0\) and \(R \geq 0\) are \(D \times D\) matrices with \(D = \text{poly}(N)\). Then it is proved that the reduced density matrices \(\rho_l\) taken from the ensemble introduced with \(D \geq N^5\) have generically maximum entropy. The proof is based on the Weingarten graphical calculus [\textit{B. Collins} and \textit{I. Nechita}, Commun. Math. Phys. 297, No. 2, 345--370 (2010; Zbl 1191.81050)] and a new estimate of the Weingarten function.