Quantum model reduction for continuous-time quantum filters

T.Tommaso Grigoletto, C.Clément Pellegrini, F.Francesco Ticozzi · 2025

Read Quantum model reduction for continuous-time quantum filters by T.Tommaso Grigoletto, C.Clément Pellegrini, F.Francesco Ticozzi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-13885.txt.

Introduction

Tommaso Grigoletto, Clément Pellegrini, and Francesco Ticozzi shrink quantum filters without losing exact answers. Their paper appeared in two thousand twenty-five. Quantum filters track a system under continuous watching. They feed simulation, feedback control, and estimation tasks. Often only a few quantities from the filter truly matter. A smaller filter could compute just those.

Their method finds reduced filters that match the full one exactly on the wanted values. Older reduction tricks give only near matches. The reduced model keeps the Belavkin filtering form, so it stays physically readable. Tools come from minimal realization theory and noncommutative conditional expectations. Tests on prototype examples show the procedure works. Exact small models could ease both simulation and hardware control.

Catalog record

CREATORT.Tommaso Grigoletto, C.Clément Pellegrini, F.Francesco Ticozzi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE147 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-13885

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Quantum model reduction for continuous-time quantum filters T.Tommaso Grigoletto, C.Clément Pellegrini, F.Francesco Ticozzi Abstract The use of quantum stochastic models is widespread in dynamical reduction, simulation of open systems, feedback control and adaptive estimation. In many applications only part of the information contained in the filter’s state is actually needed to reconstruct the target observable quantities; thus, filters of smaller dimensions could be in principle implemented to perform the same task. In this work, we propose a systematic method to find, when possible, reduced-order quantum filters that are capable of exactly reproducing the evolution of expectation values of interest. In contrast with existing reduction techniques, the reduced model we obtain is exact and in the form of a Belavkin filtering equation, ensuring physical interpretability. This is attained by leveraging tools from the theory of both minimal realization and non-commutative conditional expectations. The proposed procedure is tested on prototypical examples, laying the groundwork for applications in quantum trajectory simulation and quantum feedback control. , and ?? Department of Information Engineering, University of Padova, Italy, ??; ?? ?? Institut de Mathématiques de Toulouse, UMR5219, Université de Toulouse, CNRS, UPS IMT, F-31062 Toulouse Cedex 9, France, ?? ?? Department of Physics and Astronomy, Dartmouth College, Hanover, New Hampshire 03755, USA 1 Introduction Despite the comforting unitarity of quantum dynamics as prescribed by Schrödinger’s equation [ 62 ] , its stochastic extensions have emerged as natural candidates to model quantum measurement processes as dynamical systems [ 51 , 2 ] , and to introduce spontaneous localization mechanisms in quantum theory [ 34 , 33 ] . Probabilistic behavior can be introduced in Schrödinger’s equation as a stochastic fluctuation of the Hamiltonian operator, forcing one to add some correction terms in order to maintain its state as a valid state vector [ 10 ] . Independently, quantum stochastic evolutions of the same form have been derived in the 1970’s as the result of the dynamical interaction of quantum systems with an infinite-dimensional environment, modeled as a quantum field in the framework of quantum probability [ 49 ] . In the pioneering work of Belavkin, stochastic models of this type emerge as the quantum equivalent of a Kushner-Stratonovich equation [ 44 , 57 , 25 ] , i.e. the dynamical model for a quantum system undergoing indirect continuous observation [ 16 , 23 , 24 , 61 ] . Later, similar models emerged in quantum optics, and have been used to model different types of measurements and their fluctuations [ 28 , 64 , 65 ] . For a review of quantum optical models from a mathematical perspective and derivations of the models that avoid the need for noncommutative operator-valued processes, see [ 10 ] . The potential of stochastic quantum evolutions as quantum filtering equations, which provide state estimation based on measurements, to support control algorithms have been already proposed in Belavkin’s work [ 15 ] , and have been developed into a subfield of quantum control [ 3 ] . State-based feedback control based on stochastic models has been experimentally implemented on different platforms [ 56 , 22 ] . In any application of these models that requires numerical integration of the resulting SDEs, increasing the size of the system introduces a major hurdle. When integration has to be performed in real time, as in feedback protocols, this issue practically limits applicability to extremely small systems. In this work we aim to construct smaller models that are able to exactly reproduce the output of interest for the model, which shall be assumed to be some linear functional of the state of the system. This is done by projecting the dynamics onto subspaces or algebra that contain the full trajectories of the observables of interest in Heisenberg’s picture - a Krylov subspace [ 43 ] that is called observable subspace in linear system theory [ 42 ] , as well as the space generated by the measurement outcome. The idea has been introduced for deterministic dynamics in [ 37 , 35 ] and for discrete time processes in [ 38 ] . The works [ 45 , 40 ] are similar in spirit but are limited to the classical case. The quantum continuous-time scenario, as we shall see, presents peculiar challenges. The main other approach that has been proposed to limit the size of the model to be integrated is the so-called quantum projection filter , by introducing parametrization of the state and constructing reductions to the corresponding manifolds approximations [ 54 , 48 , 30 , 32 , 39 , 31 ] . With respect to our case, however, the resulting models are approximate and do not guarantee to preserve the form of the dynamics, limiting their physical interpretability. System and measurements: We consider a finite-dimensional quantum system, \mathcal{H}\simeq\mathbb{C}^{n} , whose state is described, for t\geq 0 , by the density operator \rho_{t}\in\mathfrak{D}(\mathcal{H}) and subject to p continuous measurement of homodyne type [ 10 ] and q measurements of counting type [ 9 ] . For each measurement of homodyne type, labeled with j=1,\dots,p , the output signal is a scalar stochastic process (Y_{t}^{j})_{t\geq 0} , whose dynamics obeys the stochastic differential equation dY_{t}^{j}={\rm tr}[(D_{j}+D_{j}^{*})\rho_{t}]dt+dW_{t}^{j}, (1) where \{(W_{t}^{j})_{t\geq 0}\}_{j=1}^{p} are independent Wiener processes and D_{j}\in\mathfrak{B}(\mathcal{H}) are operators that describe the effects of the measurement on the system. For each measurement of counting type, labeled with j=1,\dots,q , the output is a scalar counting process (N_{t}^{j})_{t\geq 0} of stochastic intensity \int_{0}^{t}{\rm tr}[C_{j}\rho_{s}C_{j}^{*}]ds , where C_{j}\in\mathfrak{B}(\mathcal{H}) are operators that describe the effect of measurement on the system. Stochastic master equation. The evolution of the state (\rho_{t})_{t\geq 0} under these assumptions is known as a quantum trajectory and is modeled by a jump-diffusion stochastic differential equation (SDE), known in the literature as stochastic master equation (SME) [ 10 ] or quantum filtering equation [ 16 , 24 ] . Given an Hermitian operator H , a set of arbitrary operators \{L_{j}\}_{j=1}^{m} , and the sets of operators that describe the relation between the state \rho_{t} and the measurement outcomes \{D_{j}\}_{j=1}^{p},\{C_{j}\}_{j=1}^{q} the state \rho_{t} satisfies the following stochastic differential equation: \begin{split}d\rho_{t}=&\,\mathcal{L}(\rho_{t-})dt\\ &+\sum_{j=1}^{p}\left[D_{j}\rho_{t-}+\rho_{t-}D_{j}^{*}-{\rm tr}[D_{j}\rho_{t-}+\rho_{t-}D_{j}^{*}]\rho_{t-}\right]\left(dY_{t}^{j}-{\rm tr}[D_{j}\rho_{t-}+\rho_{t-}D_{j}^{*}]dt\right)\\ &+\sum_{j=1}^{q}\left[\frac{C_{j}\rho_{t-}C_{j}^{*}}{{\rm tr}[C_{j}\rho_{t-}C_{j}^{*}]}-\rho_{t-}\right]\left(dN_{t}^{j}-{\rm tr}[C_{j}\rho_{t-}C_{j}^{*}]dt\right),\end{split} (2) where the operator \mathcal{L} is the so called Lindblad operator and is defined as \begin{split}\mathcal{L}(\rho)\equiv&-i[H,\rho]+\sum_{j=1}^{m}L_{j}\rho L_{j}^{*}-\frac{1}{2}\{L_{j}^{*}L_{j},\rho\}+\sum_{j=1}^{p}D_{j}\rho D_{j}^{*}-\frac{1}{2}\{D_{j}^{*}D_{j},\rho\}+\sum_{j=1}^{q}C_{j}\rho C_{j}^{*}-\frac{1}{2}\{C_{j}^{*}C_{j},\rho\}.\end{split} (3) If \rho_{0}\in\mathfrak{D}(\mathcal{H}) , then the solution (\rho_{t})_{t\geq 0} of Eq ( 2 ) is valued in \mathfrak{D}(\mathcal{H}) [ 50 ] . Note that, in Equation ( 2 ) one can include parameters that characterize the measurements efficiencies [ 5 ] . We here decided to not include them in order to lighten the notations. Nonetheless the introduction of efficiency parameters does not change the results we derive next as those are simply scalar coefficients that are not affected by the involved reductions. Ouput functionals . In many cases of practical interest, we are not actually interested in all the information contained in (\rho_{t})_{t\geq 0} . In particular, quantum filtering equations are often used to estimate the state \rho_{t} which is then used to to compute estimates of linear functionals of the state. Relevant cases include: • In simulation of quantum trajectories aimed to study the evolution of observables of interest, e.g. {\rm tr}[O\rho_{t}] , subject to continuous measurement [ 59 ]

Read the complete text