Model Reduction for Controlled Quantum Markov Dynamics
Tommaso Grigoletto, Lorenza Viola, Francesco Ticozzi · 2025
Read Model Reduction for Controlled Quantum Markov Dynamics by Tommaso Grigoletto, Lorenza Viola, 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-2510-25546.txt.
Introduction
Tommaso Grigoletto, Lorenza Viola, and Francesco Ticozzi shrink models of quantum systems. Their 2025 paper deals with Markov type systems under outside control. The rules of motion change with time in such systems. Their method builds on Krylov spaces and algebra tools made for fixed rules. The small model tracks the key read-outs of the full one in an exact way. It also keeps the Lindblad form that such rules must have.
Catalog record
| CREATOR | Tommaso Grigoletto, Lorenza Viola, Francesco Ticozzi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-25546
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Model Reduction for Controlled Quantum Markov Dynamics Tommaso Grigoletto, Lorenza Viola, Francesco Ticozzi Abstract We consider the problem of model reduction for Markovian quantum systems whose dynamics are described by a time-dependent Lindblad generator – notably, as arising in the presence of external control. Our approach, which builds upon Krylov operator subspaces and operator-algebraic techniques introduced for time-independent generators, returns a reduced model that reproduces exactly the evolution of observables of interest and is guaranteed to be in Lindblad form. August 10, 2026 I Introduction Obtaining reduced-order models that respect structural constraints of dynamical models of interest is essential for reducing the demand of direct simulation and facilitate analysis and optimization tasks. We focus on a large class of quantum systems described by time-dependent Markovian (semigroup) dynamics [ 1 , 2 ] and leverage a series of results recently established for model reduction (MR) of time-independent generators [ 3 , 4 , 5 , 6 ] to obtain smaller-dimensional descriptions of controlled, parametric or time-dependent quantum semigroups. Our operator-algebraic approach allows for exactly reproducing the expectation values of a set of target observables and, crucially, it guarantees that the reduced model is a valid quantum model, leading to completely-positive, trace-preserving (CPTP) reduced dynamics. Physically, the kind of controlled dynamical models we consider arises in a variety of scenarios where time-local master equations are used to describe driven Markovian dynamics [ 7 , 8 , 9 ] – including the important case of periodically driven (Floquet-Markov) dynamics [ 10 ] – or where parametric families of generators are relevant – such as in studies related to Hilbert space fragmentation [ 11 , 12 ] . In the latter case, treating these parameters as the controls, an exact MR for the set can be obtained using our approach. Another natural use of the proposed technique include the study of controlled open quantum systems: for instance, in situations where a fully-controllable system of interest interacts with an uncontrollable (possibly non-Markovian) bath, our techniques allow to obtain a simplified model in terms of a smaller-dimensional bath. Other applications may be envisioned in a variety of open-loop and feedback quantum-control scenarios [ 13 , 14 ] . In particular, in the context of feedback quantum control, the proposed technique can be used in combination with existing results [ 6 ] to compute a reduced-order observer. This would lessen the computational effort associated with the real-time implementation of a filtering-based feedback scheme, potentially extending the viability of the techniques beyond what is achievable experimentally [ 15 ] . Our MR procedures are also well-suited for control strategies that entail switching between a finite number of generators [ 16 , 17 , 18 , 19 , 20 ] . The possibility of reducing a controlled dynamical system is naturally tied, and in fact complementary, to the extent the dynamical system explores the state space, and thus its controllability properties. Controllability analyses of Markovian open quantum systems have been developed under different assumptions [ 21 , 22 , 23 ] , and a general geometric approach based on Lie semigroups has been put forward [ 24 , 25 ] . While, as mentioned, our approach builds on techniques developed for time-independent generators [ 5 ] , a key difference in the present analysis relates to the construction of the operator subspace that contains all the observables of interest evolved in the Heisenberg picture. The content is structured as follows. In Sec. II we introduce relevant notation and formulate the problem of interest. Sec. III contains our main contribution – the proposed observable-based MR algorithm. In Sec. IV we illustrate the MR procedure by studying a controlled dephasing central-spin model, along with mentioning other possible examples. II Problem setting: Controlled dynamics in the Heisenberg picture We consider finite-dimensional quantum systems, with associated Hilbert space \mathcal{H}\simeq\mathbb{C}^{n} , and operator space \mathfrak{B}(\mathcal{H})\simeq\mathbb{C}^{n\times n} . The state of the system is described by a density operator \rho\in\mathfrak{D}(\mathcal{H}) , where \mathfrak{D}(\mathcal{H})\equiv\{\rho\in\mathfrak{B}(\mathcal{H})|\rho=\rho^{\dagger}\geq 0,{\rm tr}(\rho)=1\} , while physical observables are modelled as self-adjoint operators O\in\mathfrak{B}(\mathcal{H}) , O=O^{\dagger} . The expectation value of O with respect to a state \rho is given by \left<O\right>_{\rho}={\rm tr}[O\rho] . In the following, we shall drop the subscript \rho whenever the state is clear from context. We describe the evolution of the system in the Heisenberg picture, via a time-local quantum master equation of the form \dot{O}(t)=\mathcal{L}_{u(t)}[O(t)]. (1) Here, u(t):\mathbb{R}_{\geq 0}\to\mathcal{U} denotes the control input, that takes values in an admissible set \mathcal{U}\subseteq\mathbb{R}^{m} , and \mathcal{L}_{u} is a controlled Lindblad generator , namely, in units where \hbar=1 , it may be expressed in the form [ 2 , 1 ] \mathcal{L}_{u}[O]=i[H_{u},O]+\sum_{k}\mathcal{D}_{L_{u,k}}[O], (2) where H_{u}=H_{u}^{\dagger} describes the Hamiltonian (coherent) contribution to the dynamics, and the dissipative contributions have the form \mathcal{D}_{L}[O]\equiv L^{\dagger}OL-\tfrac{1}{2}\{L^{\dagger}L,O\} , with L_{u,k}\in\mathfrak{B}(\mathcal{H}) being noise (Lindblad) operators. In the case where \mathcal{L} is time-independent, super-operators of this type are the generators of CP and unital (i.e., identity-preserving) one-parameter quantum dynamical semigroups \{\Lambda_{t,t_{0}}\equiv e^{\mathcal{L}(t-t_{0})}\}_{t\geq t_{0}} , resulting in CP-divisible (Markovian) dynamics [ 26 ] . The above general setting encompasses several representative situations of interest. In particular, we highlight a few: • Open-loop coherent control, whereby the dissipator is time-independent but H(t)=H_{0}+H_{c}(t) , with H_{c}(t)\equiv\sum_{\ell}u_{\ell}(t)H_{\ell} , and H_{\ell}\in\mathcal{B}(\mathcal{H}) represent independently tunable control directions [ 27 ] . • Dissipative control, whereby the strength of (some) noise operators is modulated in time, e.g., L_{u,k}(t)\equiv u_{k}(t)\,L_{k} , possibly in conjunction with time-dependent Hamiltonian control as above. Relevant settings include open-loop or feedback switching control scenarios [ 16 , 17 , 18 , 19 , 20 ] or, in cases where the control inputs are time-periodic, Floquet-Lindblad master equations [ 10 ] . • Time-dependent perturbative scenarios, whereby the Hamiltonian and/or the noise operators are modified as H(t)\equiv H_{0}+u(t)H_{1} , L_{u,k}(t)\equiv L_{0,k}+u_{k}(t)L_{1,k} , such as in adiabatic elimination-type settings [ 28 ] . Imagine we are given a finite (or finally generated) subset of observables of interest, say, \Omega\equiv\{O_{j}\} , whose time-dependent expectation values we wish to simulate more efficiently than using the full model. Without loss of generality, we can assume that \mathbb{1}\in\Omega . Given a state \rho\in\mathfrak{D}(\mathcal{H}) and an initial observable O(t_{0})\in\Omega at time t_{0} , the solution of ( 1 ) is formally given by the time-ordered exponential \displaystyle O(t) \displaystyle\equiv\Lambda_{t,t_{0}}[O(t_{0})]\equiv\mathbb{T}e^{\int_{t_{0}}^{t}\mathcal{L}_{u(\tau)}d\tau}[O(t_{0})], (3) \displaystyle=\sum_{n=0}^{+\infty}\int_{t_{0}}^{t}dt_{1}\dots\int_{t_{0}}^{t_{n-1}}dt_{n}\,\mathcal{L}_{u(t_{1})}\dots\mathcal{L}_{u(t_{n})}[O(t_{0})], where \mathbb{T} denotes the (Dyson) time-ordering symbol and, compared to the time-independent case, the maps \{\Lambda_{t,t_{0}}\} obey the more general forward composition law \Lambda_{t,s}\circ\Lambda_{s,t_{0}}=\Lambda_{t,t_{0}} , for any t\geq s\geq t_{0} [ 7 ] . The resulting time-dependent expectation value is \left<O(t)\right>={\rm tr}[O(t)\rho] . While care is needed in ensuring that time-dependent generators respect physical consistency conditions [ 8 ] , we will assume those to be obeyed and focus on mathematical aspects of the problem in what follows. By linearity of the dynamics, the ability to simulate the observables in \Omega directly entails the ability to simulate any linear combinations of them. Hence, we can also consider \Omega to be, in general, a finite-dimensional operator subspace. In this work, we focus on the following: Problem 1 . Given a time-dependent quantum model ( 1 ) and a finite set of target observables \Omega , find a smaller (if one exists) Hilbert space \check{\mathcal{H}} on which we can