Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

The Journal of Chemical Physics

Tensor propagator for iterative quantum time evolution of reduced density matrices. II. Numerical methodology

Nancy Makri, Dmitrii E. Makarov

Published Mar 15, 1995• 488 citations
Publisher

Abstract

In a recent Letter [Chem. Phys. Lett. 221, 482 (1994)], we demonstrated that the dynamics of reduced density matrices for systems in contact with dissipative harmonic environments can be obtained in an iterative fashion by multiplication of a propagator tensor. The feasibility of iterative procedures in reduced dimension spaces arises from intrinsic features of the dissipative influence functional in Feynman’s path integral formulation of quantum dynamics. Specifically, the continuum of frequencies characteristic of broad condensed phase spectra disrupts phase coherence to a large extent, such that the dynamics of an augmented reduced density tensor becomes Markovian. In a preceding article [J. Chem. Phys. 102, 4600 (1995)] we examined in detail the formal properties of the tensor propagator. In the present paper we show that the tensor propagator can be further decomposed into a product of small rank tensors, resulting in an extremely simple and efficient numerical scheme that scales almost linearly with the dimension of the augmented reduced density tensor. Numerical application to a model electron transfer reaction is presented.

DOI
10.1063/1.469509

Related papers

Tensor propagator for iterative quantum time evolution of reduced density matrices. I. Theory

Nancy Makri, Dmitrii E. Makarov

For common condensed phase problems described by a low-dimensional system coupled to a harmonic bath, Feynman’s path integral formulation of time-dependent quantum mechanics leads to expressions for the reduced density matrix of the system where the effects of the harmonic environment enter through an influence functional that is nonlocal in time. In a recent Letter [Chem. Phys. Lett. 221, 482 (1994)], we demonstrated that the range of the nonlocal interactions is finite even at zero temperature, such that the nonlocal kernel extends over only a few time steps if the path integral is expressed in terms of accurate quasiadiabatic propagators. This feature arises from disruption of phase coherence in macroscopic environments and leads to Markovian dynamics for an augmented reduced density tensor, permitting iterative time evolution schemes. In the present paper we analyze the structure and properties of the relevant tensor propagator. Specifically, we show that the tensor multiplication scheme rigorously conserves the trace of the reduced density matrix, and that in cases of short-range nonlocality it leads to Redfield-type equations which are correct to all orders in perturbation theory and which take into account memory effects. We also argue that a simple eigenvector analysis reveals (without actual iteration) the nature of the dynamics and of the equilibrium state, and directly yields quantum reaction or relaxation rates.

The Journal of Chemical Physics•Mar 15, 1995•636 citations

Path integral renormalization for quantum dissipative dynamics with multiple timescales

Nancy Makri

The reduced density matrix of a system coupled to a harmonic dissipative bath can be obtained accurately using an iterative path integral methodology that involves a propagator tensor, whose rank is roughly equal to the number of time steps that span the bath-induced memory. The present work shows that even in situations where the bath is characterized by high-frequency modes (which necessitate a small time step) and slow oscillators (which give rise to very long memory), efficient propagation that takes into account the memory on all time scales is feasible by renormalizing the propagator matrix to include successively larger-length influence functional correlations. At its crudest level, the renormalized propagator is an ordinary matrix, thus the computational demands of the method are very modest, similar to those for propagating the bare system. Yet, test applications on dissipative two-level systems show that the method produces accurate results in demanding situations. By including memory on successively increasing time scales, the method is capable of capturing nonexponential kinetics induced by sluggish solvents.

Molecular Physics•May 10, 2012•30 citations

Quantum dynamics for a system coupled to slow baths: On-the-fly filtered propagator method

Eunji Sim

An on-the-fly filtered propagator functional path integral scheme is introduced as an efficient way of calculating the iterative dynamics of complex condensed systems. Time evolution of the reduced density matrix of a dissipative quantum system is evaluated iteratively by filtering the negligible propagator elements at each propagation step. This on-the-fly filtering along with the finiteness of the bath memory dramatically reduces the configuration space to be integrated without losing numerical accuracy of the results. The required computational storage space for the configuration and the weight of the survived path segments increases linearly with the bath memory length. Up to the bath memory time, it is found that a strikingly small fraction of the configurations survives the on-the-fly filtering process and the number of surviving configurations increases algebraically with the propagation time. At times longer than the bath memory time, the number of surviving configurations required for numerically accurate results is essentially saturated and is less than 0.1% of the total number of the configurations. This new scheme is extremely useful for the problems of low-frequency solvents for which the bath memory spans many time steps of the propagation.

The Journal of Chemical Physics•Sep 8, 2001•92 citations

Efficient memory equation algorithm for reduced dynamics in spin-boson models

Andrei A. Golosov, Richard A. Friesner, Philip Pechukas

The dynamics of a one-dimensional quantum system coupled to a harmonic bath can be expressed through Feynman’s path integral expression for the reduced density matrix. In this expression the influence of the environment is seen in correlations between positions of the system that are nonlocal in time. Makri and Makarov [J. Chem. Phys. 102, 4600 (1995)] showed that for many practical problems correlations over only a few time steps, Δkmax, need to be taken into account, which led to an efficient iterative scheme. However, this algorithm scales as the size of the system to the power of 2(Δkmax+1), which restricts the size of the system that can be studied with this method. In this work we present an efficient algorithm which scales linearly with Δkmax. In our method the reduced density matrix is written as a convolution of its past values with an integral equation kernel. The calculation of that kernel is based on a perturbative expansion of the discretized quasiadiabatic path integral expression for the reduced density matrix. The expansion ignores certain types of correlations.

The Journal of Chemical Physics•Jan 1, 1999•60 citations

Small matrix disentanglement of the path integral: Overcoming the exponential tensor scaling with memory length

Nancy Makri

The discretized path integral expression for the reduced density matrix (RDM) of a system interacting with a dissipative harmonic bath is fully entangled because of influence functional terms that couple the variables at different time points. The iterative decomposition of the path integral, which exploits the finite length of influence functional memory, involves a tensor propagator whose size grows exponentially with the memory length. The present Communication disentangles the path integral by recursively spreading the temporal entanglement over longer path segments, while decreasing its contribution. Eventually, the entangled term becomes sufficiently small and may be neglected, leading to iterative propagation of the RDM through simple multiplication of matrices whose size is equal to that of the bare system. It is found that the temporal entanglement length is practically equal to the bath-induced memory length. The small matrix decomposition of the path integral (SMatPI) is stable and very efficient, extending the applicability of numerically exact real-time path integral methods to multi-state systems.

The Journal of Chemical Physics•Jan 27, 2020•101 citations

Iterative blip-summed path integral for quantum dynamics in strongly dissipative environments

Nancy Makri

The iterative decomposition of the blip-summed path integral [N. Makri, J. Chem. Phys. 141, 134117 (2014)] is described. The starting point is the expression of the reduced density matrix for a quantum system interacting with a harmonic dissipative bath in the form of a forward-backward path sum, where the effects of the bath enter through the Feynman-Vernon influence functional. The path sum is evaluated iteratively in time by propagating an array that stores blip configurations within the memory interval. Convergence with respect to the number of blips and the memory length yields numerically exact results which are free of statistical error. In situations of strongly dissipative, sluggish baths, the algorithm leads to a dramatic reduction of computational effort in comparison with iterative path integral methods that do not implement the blip decomposition. This gain in efficiency arises from (i) the rapid convergence of the blip series and (ii) circumventing the explicit enumeration of between-blip path segments, whose number grows exponentially with the memory length. Application to an asymmetric dissipative two-level system illustrates the rapid convergence of the algorithm even when the bath memory is extremely long.

The Journal of Chemical Physics•Apr 3, 2017•49 citations
In a recent Letter [Chem. Phys. Lett. 221, 482 (1994)], we demonstrated that the dynamics of reduced density matrices for systems in contact with dissipative harmonic environments can be obtained in an iterative fashion by multiplication of a propagator tensor. The feasibility of iterative procedures in reduced dimension spaces arises from intrinsic features of the dissipative influence functional in Feynman’s path integral formulation of quantum dynamics. Specifically, the continuum of frequencies characteristic of broad condensed phase spectra disrupts phase coherence to a large extent, such that the dynamics of an augmented reduced density tensor becomes Markovian. In a preceding article [J. Chem. Phys. 102, 4600 (1995)] we examined in detail the formal properties of the tensor propagator. In the present paper we show that the tensor propagator can be further decomposed into a product of small rank tensors, resulting in an extremely simple and efficient numerical scheme that scales almost linearly with the dimension of the augmented reduced density tensor. Numerical application to a model electron transfer reaction is presented.
For common condensed phase problems described by a low-dimensional system coupled to a harmonic bath, Feynman’s path integral formulation of time-dependent quantum mechanics leads to expressions for the reduced density matrix of the system where the effects of the harmonic environment enter through an influence functional that is nonlocal in time. In a recent Letter [Chem. Phys. Lett. 221, 482 (1994)], we demonstrated that the range of the nonlocal interactions is finite even at zero temperature, such that the nonlocal kernel extends over only a few time steps if the path integral is expressed in terms of accurate quasiadiabatic propagators. This feature arises from disruption of phase coherence in macroscopic environments and leads to Markovian dynamics for an augmented reduced density tensor, permitting iterative time evolution schemes. In the present paper we analyze the structure and properties of the relevant tensor propagator. Specifically, we show that the tensor multiplication scheme rigorously conserves the trace of the reduced density matrix, and that in cases of short-range nonlocality it leads to Redfield-type equations which are correct to all orders in perturbation theory and which take into account memory effects. We also argue that a simple eigenvector analysis reveals (without actual iteration) the nature of the dynamics and of the equilibrium state, and directly yields quantum reaction or relaxation rates.
The reduced density matrix of a system coupled to a harmonic dissipative bath can be obtained accurately using an iterative path integral methodology that involves a propagator tensor, whose rank is roughly equal to the number of time steps that span the bath-induced memory. The present work shows that even in situations where the bath is characterized by high-frequency modes (which necessitate a small time step) and slow oscillators (which give rise to very long memory), efficient propagation that takes into account the memory on all time scales is feasible by renormalizing the propagator matrix to include successively larger-length influence functional correlations. At its crudest level, the renormalized propagator is an ordinary matrix, thus the computational demands of the method are very modest, similar to those for propagating the bare system. Yet, test applications on dissipative two-level systems show that the method produces accurate results in demanding situations. By including memory on successively increasing time scales, the method is capable of capturing nonexponential kinetics induced by sluggish solvents.
An on-the-fly filtered propagator functional path integral scheme is introduced as an efficient way of calculating the iterative dynamics of complex condensed systems. Time evolution of the reduced density matrix of a dissipative quantum system is evaluated iteratively by filtering the negligible propagator elements at each propagation step. This on-the-fly filtering along with the finiteness of the bath memory dramatically reduces the configuration space to be integrated without losing numerical accuracy of the results. The required computational storage space for the configuration and the weight of the survived path segments increases linearly with the bath memory length. Up to the bath memory time, it is found that a strikingly small fraction of the configurations survives the on-the-fly filtering process and the number of surviving configurations increases algebraically with the propagation time. At times longer than the bath memory time, the number of surviving configurations required for numerically accurate results is essentially saturated and is less than 0.1% of the total number of the configurations. This new scheme is extremely useful for the problems of low-frequency solvents for which the bath memory spans many time steps of the propagation.
The dynamics of a one-dimensional quantum system coupled to a harmonic bath can be expressed through Feynman’s path integral expression for the reduced density matrix. In this expression the influence of the environment is seen in correlations between positions of the system that are nonlocal in time. Makri and Makarov [J. Chem. Phys. 102, 4600 (1995)] showed that for many practical problems correlations over only a few time steps, Δkmax, need to be taken into account, which led to an efficient iterative scheme. However, this algorithm scales as the size of the system to the power of 2(Δkmax+1), which restricts the size of the system that can be studied with this method. In this work we present an efficient algorithm which scales linearly with Δkmax. In our method the reduced density matrix is written as a convolution of its past values with an integral equation kernel. The calculation of that kernel is based on a perturbative expansion of the discretized quasiadiabatic path integral expression for the reduced density matrix. The expansion ignores certain types of correlations.
The discretized path integral expression for the reduced density matrix (RDM) of a system interacting with a dissipative harmonic bath is fully entangled because of influence functional terms that couple the variables at different time points. The iterative decomposition of the path integral, which exploits the finite length of influence functional memory, involves a tensor propagator whose size grows exponentially with the memory length. The present Communication disentangles the path integral by recursively spreading the temporal entanglement over longer path segments, while decreasing its contribution. Eventually, the entangled term becomes sufficiently small and may be neglected, leading to iterative propagation of the RDM through simple multiplication of matrices whose size is equal to that of the bare system. It is found that the temporal entanglement length is practically equal to the bath-induced memory length. The small matrix decomposition of the path integral (SMatPI) is stable and very efficient, extending the applicability of numerically exact real-time path integral methods to multi-state systems.
The iterative decomposition of the blip-summed path integral [N. Makri, J. Chem. Phys. 141, 134117 (2014)] is described. The starting point is the expression of the reduced density matrix for a quantum system interacting with a harmonic dissipative bath in the form of a forward-backward path sum, where the effects of the bath enter through the Feynman-Vernon influence functional. The path sum is evaluated iteratively in time by propagating an array that stores blip configurations within the memory interval. Convergence with respect to the number of blips and the memory length yields numerically exact results which are free of statistical error. In situations of strongly dissipative, sluggish baths, the algorithm leads to a dramatic reduction of computational effort in comparison with iterative path integral methods that do not implement the blip decomposition. This gain in efficiency arises from (i) the rapid convergence of the blip series and (ii) circumventing the explicit enumeration of between-blip path segments, whose number grows exponentially with the memory length. Application to an asymmetric dissipative two-level system illustrates the rapid convergence of the algorithm even when the bath memory is extremely long.