Equivalence classes and local asymptotic normality in system identification for quantum Markov chains (Q2339207)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Equivalence classes and local asymptotic normality in system identification for quantum Markov chains |
scientific article |
Statements
Equivalence classes and local asymptotic normality in system identification for quantum Markov chains (English)
0 references
31 March 2015
0 references
This paper treats the problem of identifying and estimating dynamical parameters of an ergodic quantum Markov chain, when only stationary output is accessible for measurements. The purpose of this paper is to characterize the statistical properties of the quantum out state of the Markov chain. Let \(\theta \mapsto V_{\theta}\) be a smooth family of isometries parameterized by the unknown parameter \(\theta \in {\mathbb R}\), and let \(K_{i \theta}\) be the corresponding Kraus operators. \(\theta = \theta_0 + u / \sqrt{n}\) with local parameter \(u\) and assume that \(V := V_{ \theta_0}\) is a primitive isometry with stationary state \(\rho_{ss}\). Let \(\Lambda_i\) denotes its eigenvalues, and denote by \(K_i = K_{ i_1 \theta_0}\) and by \(\dot{K}_i\) the derivative of \(K_{ i_1 \theta}\) with respect to \(\theta\) at \(\theta_0\). The output state corresponding to \(V_{ \theta_0 + u / \sqrt{n} }\) is denoted by \(| \Psi_{n, \varphi}(n) \rangle\), with \({\mathcal H} = {\mathbb C}^D\), \(\varphi \in {\mathcal H}\), and \({\mathcal K} = {\mathbb C}^k\). For a given initial pure state \(| \varphi \rangle\) of the system, the output state after \(n\) iterations is given by \[ \rho_{u, \varphi}(n) := \text{tr}_{ {\mathcal H}} [ | \Psi_{u, \varphi}(n) \rangle \langle \Psi_{u, \varphi} (n) | ], \] and \(\rho_u(n)\) is the output state corresponding to the stationary regime. While \(| \psi_{u, \eta, \varphi}(n) \rangle\) denotes the un-normalized conditional state. The relevant scalar products can be written in the form \[ \langle \Psi_{u, \varphi}(n) | \Psi_{v, \varphi}(n) \rangle = \langle \varphi | T_{u, v; n}^n (1) \varphi \rangle, \] and \[ \langle \psi_{u,\eta, \varphi}(n) | \psi_{v, \eta, \varphi}(n) \rangle = \text{tr} [ T_{u,v; n}^n ( | \eta \rangle \langle \eta | ) | \varphi \rangle \langle \varphi | ], \] where \(T_{u,v; n} : {\mathcal L}({\mathcal H}) \to {\mathcal L}({\mathcal H})\) is given by \[ T_{u,v; n}(X) := V_{ \theta_0 + u / \sqrt{n} }^+ ( X \otimes 1) V_{ \theta_0 + v / \sqrt{n} }. \] The interaction is described by a unitary operator \(U\) : \({\mathcal H} \otimes {\mathcal K}\) \(\to\) \({\mathcal H} \otimes {\mathcal K}\), such that after one step, the joint state of system and the first unit is \(U( | \varphi \rangle \otimes | \chi \rangle )\), while the remaining units are still in the initial state. After the interaction with the first noise unit, the system moves one step to the left and the same operation is repeated between system and the second unit, and so on. Alternatively, one can think that the system is fixed and the chain is shifted to the right. After \(n\) steps the state of the system and the \(n\) noise units is therefore \[ | \Psi(n) \rangle := U(n) | \varphi \otimes \chi^{ \otimes n} \rangle \in {\mathcal H} \otimes {\mathcal K}^{ \otimes n} \] where \(U(n)\) is the product of unitaries \(U^{(n)} \cdots U^{(1)}\) with \(U^{(i)}\) denoting the copy of \(U\) which acts on system and the \(i\)-th noise unit, counting from right to left according to the dynamics. Here is the first main theorem. Theorem. There exist \(F,a \in {\mathbb R}\) such that for each \(X \in {\mathcal B}({\mathcal H})\) and \(C > 0\) \[ \lim_{ n \to \infty} \sup_{ | u |, | v | < C} T_{u,v; n}^n (X) = \text{tr} [ \rho_{ss} X] e^{- F ( u -v)^2/8- ia (u^2 - v^2)} 1. \tag{1} \] The constant \(F\) is the limiting quantum Fisher information per sample defined by \[ F = \lim_{ n \to \infty} \frac{4}{n} \left\{ \left\| \frac{ d \Psi_{\theta}(n)}{ d \theta} \right\|^2 - \left| \left\langle \left. \frac{ d \Psi_{\theta}}{d \theta} (n) \right| \Psi_{\theta}(n) \right\rangle \right|^2 \right\}, \tag{2} \] with \(| \Psi_{\theta}(n) \rangle := | \Psi_{V_{\theta}, \varphi}(n) \rangle\), and is equal to the Markov variance of the generators \(G= i \dot{V} V^*\) and \(\tilde{G} := i \dot{U} U^*\) \[ \begin{aligned} F &= 4 \sum_{i=1}^k \left\{ \text{Tr} ( \rho_{ss} \dot{K}_i^* \dot{K}_i ) + 2 \text{Tr} \left( \text{Im} ( K_i \rho_{ss} \dot{K}_i^* ) \cdot {\mathcal R} ( \text{Im} \sum_k \dot{K}_i^* K_i ) \right) \right\} \tag{3} \\ &= 4 ( G, G)_V = 4 ( \tilde{G}, \tilde{G} )_V. \end{aligned} \] Moreover, the authors show that the output is locally asymptotically normal, in other words, it can be approximated by a simple quantum Gaussian model \({\mathcal G}\) consisting of a coherent state whose mean is related to the unknown parameter \(\theta\). Lastly, as a consistency check, they prove that a parameter related to the coordinate transformation unitaries has zero quantum Fisher information.
0 references
equivalence class
0 references
local asymptotic normality
0 references
quantum Markov chains
0 references
quantum Fisher information
0 references