Heisenberg formulation of adiabatic elimination for open quantum systems with two time-scales

François-Marie Le Régent and Pierre Rouchon

arXiv: 2303.17308v2 · PDF · 7 September 2023
Categories: quant-ph, math.OC
Licence: CC BY 4.0 The site shows the text.

Abstract

Consider an open quantum system governed by a Gorini, Kossakowski, Sudarshan, Lindblad (GKSL) master equation with two times-scales: a fast one, exponentially converging towards a linear subspace of quasi-equilibria; a slow one resulting from perturbations (small arbitrary decoherence and Hamiltonian dynamics). Usually adiabatic elimination is performed in the Schrödinger picture. We propose here an Heisenberg formulation where the invariant operators attached to the fast decay dynamics towards the quasi-equilibria subspace play a key role. Based on geometric singular perturbations, asympotic expansions of the Heisenberg slow dynamics and of the fast invariant linear subspaces are proposed. They exploit Carr’s approximation lemma from center-manifold and bifurcation theory. Second-order expansions are detailed and shown to ensure preservation, up to second-order terms, of the trace and complete positivity for the slow dynamics on a slow time-scale. Such expansions can be exploited numerically.

Items 6

Lemma

Lemma 1

PL4C

Take the slow-fast dynamics (?). For \epsilon small enough, it admits a unique invariant slow manifold of dimension \bar d with local real coordinates (x_1, \ldots, x_{\bar d}) and analytic versus \epsilon

\rho = \sum_{d=1}^{\bar d} x_{d} \left(\sum_{n=0}^{+\infty} \epsilon^n \widehat{S}_d^{(n)}\right)

Moreover, the evolution on this invariant slow linear subspace is governed by the following linear system, analytic versus \epsilon

\frac{d}{dt} x(t) =\left(\sum_{n=1}^{\infty} \epsilon^n F^{(n)}\right) x(t)

where F^{(n)} is the matrix of real entries F^{(n)}_{d,d'}. Here \widehat{S}_d^{(n)} and F^{(n)}_{d,d'} are given recursively by (?) and (?) starting with \widehat{S}_d^ {(0)}=\widehat{S}_d and F^{(0)}_{d,d'}=0.

Lemma

Lemma 2

ITJS

For the slow/fast system (?) and \epsilon small enough, exists \bar d linearly independent Hermitian operators \widehat{J}_d(\epsilon), analytic versus \epsilon such that the linear subset of Hermitian operators

\Big\{ \rho ~\big| ~\forall d\in\{1,\ldots,\bar d\}, \rm{Tr}\left(\widehat{J}_d(\epsilon) \rho\right) = 0 \Big\}

is invariant. Any trajectory of (?) starting in this subset converges exponentially to zero with a strictly positive rate independent of \epsilon. Moreover by construction, for all \forall d,d'\in\{1,\ldots,\bar d\}, \rm{Tr}\left(\widehat{J}_d(\epsilon) \widehat{S}_{d'}\right)=\delta_{d,d'}.

Lemma

Lemma 3

5V1Q

For \epsilon small enough, exist \gamma >0 and M>0 such that for any T >0 and any x(0)\in\mathbb{R}^{\bar d} we have

\sum_{d}\left| \rm{Tr}\left(\widehat{S}_{d} \mathcal{K}_{\epsilon,T}(\rho_0) \right) - z_{\epsilon,d}(T) \right| \leq M e^{-\gamma T} \sqrt{\rm{Tr}\left(\rho^2_0\right)}

where z_{\epsilon}(T)= e^{ T F(\epsilon) } E^{-1}(\epsilon) x(0) and \rho_0= \sum_d x_d(0) \widehat{S}_d.

Used by Proof

Proof of Lemma 3

It is based on the following arguments inspired by [carr-book, theorem 2, point b, page 4] and by the notion of shadow trajectories around hyperbolic invariant manifold (see, e.g.,[Hirsh_shadowing_1993] and (https://spartacus-idh.com/liseuse/116/35page/142)). Since \overline{\mathcal{K}}_0^*(\widehat{S}_d)=\widehat{J}_d one has

\rm{Tr}\left(\widehat{S}_{d} \mathcal{K}_{\epsilon,T}(\rho_0) \right) = \rm{Tr}\left(\widehat{J}_d e^{T(\mathcal{L}_0+\epsilon\mathcal{L}_1)}(\rho_0)\right)

The initial state on the invariant manifold \rho_{\epsilon,0}=\sum_d z_{\epsilon,d}(0) \widehat{S}_d(\epsilon) and \rho_0 are such that \rho_0-\rho_{\epsilon,0} belongs to the invariant fast manifold. Thus exist \gamma >0 and M>0 such that

\left\|e^{T(\mathcal{L}_0+\epsilon \mathcal{L}_1)} (\rho_0-\rho_{\epsilon,0})\right\| \leq M e^{- \gamma T} \left\|\rho_0-\rho_{\epsilon,0}\right\| .

We conclude with e^{T(\mathcal{L}_0+\epsilon \mathcal{L}_1)} (\rho_{\epsilon,0})= \sum_d z_{\epsilon,d}(T) \widehat{S}_d(\epsilon) and \rm{Tr}\left(\widehat{J}_d \widehat{S}_{d'}(\epsilon)\right) = \delta_{d,d'}.

Lemma

Lemma 4

9X8Q

Consider the slow/fast system (?), the second-order approximation of the slow dynamics

\frac{d}{dt} x = \big(\epsilon F^{(1)} + \epsilon^2 F^{(2)}\big) x

and its propagator matrix \mathcal{G}^{(2)}_{\frac{\bar T}{\epsilon}}= e^{\frac{\bar T}{\epsilon }(\epsilon F^{(1)} + \epsilon^2 F^{(2)})} over the slow time scale \bar{T}/\epsilon. For any \bar T >0, exists M_{\bar T} >0 such that for any x(0)\in\mathbb{R}^{\bar d} we have

\left\| \mathcal{K}_{\epsilon,\tfrac{\bar T}{\epsilon }}(\rho_0) - \rho(\tfrac{\bar T}{\epsilon }) \right\| \leq M_{\bar T} \epsilon^2 \|x(0)\|

where \rho_0=\sum_d x_d(0) \widehat{S}_d and \rho(\tfrac{\bar T}{\epsilon })=\sum_{d} x_d(\tfrac{\bar T}{\epsilon }) \widehat{S}_d with x(\tfrac{\bar T}{\epsilon })= \mathcal{G}^{(2)}_{\frac{\bar T}{\epsilon}} x(0).

Proof

NILT

It combines lemma 3, the fact that \mathcal{G}^{(2)}_{\frac{\bar T}{\epsilon}}= e^{\frac{\bar T}{\epsilon } F(\epsilon) } + O(\epsilon^2) since \frac{\bar T}{\epsilon } F(\epsilon) = \bar T (F^{(1)} + \epsilon F^{(2)}) + O(\epsilon^2) and that E^{-1}(\epsilon) x(0) = x(0)+ O(\epsilon^2) since E^{-1}(\epsilon)= I_{\bar d} + O(\epsilon^2).

Uses Lemma 3

Source and licence

The text on this page is from arXiv:2303.17308v2 by François-Marie Le Régent and Pierre Rouchon, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2303/2303.17308.jsonl.