betaFeedback

WorksMathematics

Dimension-free Ergodicity of Path Integral Molecular Dynamics

Abstract

The quantum thermal average plays a central role in describing the thermodynamic properties of a quantum system. Path integral molecular dynamics (PIMD) is a prevailing approach for computing quantum thermal averages by approximating the quantum partition function as a classical isomorphism on an augmented space, enabling efficient classical sampling, but the theoretical knowledge of the ergodicity of the sampling is lacking. Parallel to the standard PIMD with NN ring polymer beads, we also study the Matsubara mode PIMD, where the ring polymer is replaced by a continuous loop composed of NN Matsubara modes. Utilizing the generalized ΓΓ calculus, we prove that both the Matsubara mode PIMD and the standard PIMD have uniform-in-NN ergodicity, i.e., the convergence rate towards the invariant distribution does not depend on the number of modes or beads NN.

AMS subject classifications: 37A30, 82B31, 81S40

Introduction

Calculation of the quantum thermal average plays an important role in quantum physics and quantum chemistry, not only because it fully characterizes the canonical ensemble of the quantum system, but also because of its wide applications in describing the thermal properties of complex quantum systems, including the idea quantum gases [1], chemical reaction rates [2, 3], density of states of crystals [4], quantum phase transitions [5], etc. However, since the computational cost of direct discretization methods (finite difference, pseudospectral methods, etc.) grows exponentially with the spatial dimension [6, 7], the exact calculation of the quantum thermal average is hardly affordable in high dimensions. In the past decades, there have been numerous methods committed to compute the quantum thermal average approximately, and the path integral molecular dynamics (PIMD) is among the most prevailing ones.

The PIMD is a computational framework to obtain accurate quantum thermal averages. Using the imaginary time-slicing approach in Feynman path integral [8], the PIMD maps the quantum system in Rd\mathbb{R}^{d} to a ring polymer of NN beads in RdN\mathbb{R}^{dN}, where each bead represents a classical duplicate of the original quantum system, and adjacent beads are connected by a harmonic spring potential. When the number of beads NN is large enough, the classical Boltzmann distribution of the ring polymer in RdN\mathbb{R}^{dN} is expected to yield an accurate approximation of the quantum thermal average. Since the development of the PIMD in 1970s, this framework has been widely used in the calculations in the chemical reaction rates [9, 10, 11], transition state theory [12] and tunneling splittings [13, 14]. Several variants of the PIMD, the ring polymer molecular dynamics (RPMD) [15, 16], the centroid molecular dynamics (CMD) [17, 18] and the Matsubara dynamics [19, 20], have been employed to compute the quantum correlation function. Recently, there have been fruitful studies on designing efficient and accurate algorithms to enhance the numerical performance of the PIMD [21, 22, 23, 24]. The PIMD can also be applied in multi-electronic-state quantum systems, see [25, 26] for instance. The general procedure to compute the quantum thermal average in the PIMD is demonstrated as follows:

  1. Ring polymer approximation. Pick a positive integer NN and map the quantum system to a ring polymer system of NN beads, where each bead is a classical duplicate of the original system.

  1. Construction of sampling. Equip the ring polymer system with a stochastic thermostat (e.g., Langevin, Andersen and Nosé–Hoover) so that the resulting stochastic process samples the Boltzmann distribution of the ring polymer.

  1. Stochastic simulation. Evolve the stochastic process with a proper time discretization method, and approximate the quantum thermal average by the time average in the long-time simulation.

The PIMD framework based on the ring polymer beads is referred to as the standard PIMD in this paper.

Although the standard PIMD has become a mature framework to compute the quantum thermal average, a rigorous mathematical understanding of its convergence is still sorely lacking. A fundamental question is the dimension-free ergodicity, i.e., does the stochastic process sampling the Boltzmann distribution has a convergence rate which does not depend on the number of beads NN? The property of uniform-in-NN ergodicity has been partially validated when the potential function is quadratic [27, 28], but a rigorous justification in the general case is still to be investigated.

An important goal of this paper is to justify the dimension-free ergodicity of the standard PIMD. Nevertheless, our proof mainly relies on an alternative PIMD framework, the Matsubara mode PIMD, which is more convenient for analysis. As its name indicates, this framework is based on the Matsubara modes [19, 29] rather than the discrete beads of the ring polymer. In the Matsubara mode PIMD, the ring polymer is expressed as the superpositions of NN Matsubara modes, resulting in a continuous imaginary-time ring polymer loop rather than a ragged one (see Figure 3.2 of [19] for example). The Matsubara mode PIMD has a close relation to the standard PIMD, because the Matsubara mode PIMD with suitable discretization is equivalent to the standard PIMD [29, 30], except for the minor difference in the mode frequencies. In fact, the standard PIMD utilizes the normal mode frequencies [19], whose continuum limit give rise to the Matsubara frequencies. It is also pointed in [29, 30] that the standard PIMD has better performance than the Matsubara mode PIMD in computing the quantum thermal average.

The main contribution of this paper is that we prove the uniform-in-NN ergodicity of the Matsubara mode PIMD and the standard PIMD, namely, the convergence rate towards the invariant distribution does not depend on NN, which stands for the number of modes in the Matsubara mode PIMD and the number of beads in the standard PIMD. The authors believe that this is the first dimension-free ergodicity result of the PIMD in the form of the underdamped Langevin dynamics. Our proof strategies include the Bakry–Émery calculus [31] and the recently developed generalized Γ\Gamma calculus [32, 33]. It should be stressed that the convergence rate can still depend on the potential function V(x)V(x) and the temperature, and in particular the convergence rate is exponentially small in the low temperature limit. This a a common feature of the Bakry–Émery calculus with a non-convex potential function, referring to Proposition 5.1.6 of [31].

The paper is organized as follows. In § 2 we derive the Matsubara mode PIMD from the Feynman path integral representation of the quantum thermal average. In Section 3 we prove the uniform-in-NN ergodicity the Matsubara mode PIMD and the standard PIMD. In § 8 we implement the numerical tests to show the performance of the Matsubara mode PIMD and the standard PIMD in computing the quantum thermal average and validate the dimension-free ergodicity.

Derivation of the Matsubara mode PIMD

In this section we derive the Matsubara mode PIMD from the Feynman path integral representation of the quantum thermal average. First, we show that the quantum equilibrium can be equivalently interpreted as the classical Boltzmann distribution of a continuous loop representing the ring polymer in the position space. Second, we represent the energy function E(ξ)\mathcal{E}(\xi) of the ring polymer loop in terms of the Matsubara coordinates ξ\xi. Finally, we construct an infinite-dimensional underdamped Langevin dynamics to sample the distribution exp⁡{−E(ξ)}\exp\{-\mathcal{E}(\xi)\}, formulating the Matsubara mode PIMD.

Note that the our derivation does not rely on the normal mode transform, and is thus different from the traditional ways to generate the Matsubara modes [19].

Quantum equilibrium as the Boltzmann distribution of a continuous loop

We consider the quantum system in Rd\mathbb{R}^{d} given by the Hamiltonian

H^=p^22+V(q^),(1)\hat{H} = \frac{\hat{p}^{2}}{2} + V(\hat{q}), \tag*{(1)}

where q^\hat{q} and p^\hat{p} are the position and momentum operators in Rd\mathbb{R}^{d}, and V(⋅)V(\cdot) is a real-valued potential function in Rd\mathbb{R}^{d}. When the quantum system is at a constant temperature T=1/βT=1/\beta, the state of the system can be described by the canonical ensemble with the density operator e−βH^e^{-\beta\hat{H}}, and thus the partition function Z=Tr⁡[e−βH^]\mathcal{Z}=\operatorname{Tr}[e^{-\beta\hat{H}}]. The quantum thermal average of the system means the canonical average of an observable operator O(q^)O(\hat{q}), namely

⟨O(q^)⟩β=Tr⁡[e−βH^O(q^)]Tr⁡[e−βH^],(2)\langle O(\hat{q})\rangle_{\beta}=\frac{\operatorname{Tr}[e^{-\beta\hat{H}}O(\hat{q})]}{\operatorname{Tr}[e^{-\beta\hat{H}}]}, \tag*{(2)}

Here, we assume the observable operator O(q^)O(\hat{q}) depends only on the position operator q^\hat{q}, where O(⋅)O(\cdot) is a real-valued function in Rd\mathbb{R}^{d}.

Utilizing the Feynman path integral [8, 34], the partition function can be interpreted as the integral with respect to a continuous loop x(τ)x(\tau) in Rd\mathbb{R}^{d} parameterized by τ∈[0,β]\tau\in[0,\beta]:

Z=Tr⁡[e−βH^]=∫exp⁡{−E(x)}D[x],\mathcal{Z}=\operatorname{Tr}[e^{-\beta\hat{H}}]=\int\exp\{-\mathcal{E}(x)\}\mathcal{D}[x],

where D[x]\mathcal{D}[x] is the formal Lebesgue measure of the continuous loop in Rd\mathbb{R}^{d}, and the energy functional E(x)\mathcal{E}(x) is defined by

E(x)=12∫0β∣x′(τ)∣2dτ+∫0βV(x(τ))dτ.(3)\mathcal{E}(x)=\frac{1}{2}\int_{0}^{\beta}|x'(\tau)|^{2}\mathrm{d}\tau+\int_{0}^{\beta}V(x(\tau))\mathrm{d}\tau. \tag*{(3)}

As a consequence, the quantum canonical ensemble e−βH^e^{-\beta\hat{H}} is exactly mapped to a classical Boltzmann distribution of the continuous loop x(τ)x(\tau) with the energy E(x)\mathcal{E}(x). Finally, the quantum thermal average ⟨O(q^)⟩β\langle O(\hat{q})\rangle_{\beta} can be formally written as

⟨O(q^)⟩β=1Z∫[1β∫0βO(x(τ))dτ]exp⁡{−E(x)}D[x].(4)\langle O(\hat{q})\rangle_{\beta}=\frac{1}{\mathcal{Z}}\int\left[\frac{1}{\beta}\int_{0}^{\beta}O(x(\tau))\mathrm{d}\tau\right]\exp\{-\mathcal{E}(x)\}\mathcal{D}[x]. \tag*{(4)}

Represent the energy function with Matsubara modes

Consider the following eigenvalue problem with periodic boundary conditions:

−c¨k(τ)=ωk2ck(τ),τ∈[0,β],k=0,1,2,…,-\ddot{c}_{k}(\tau)=\omega_{k}^{2}c_{k}(\tau),\quad\tau\in[0,\beta],\quad k=0,1,2,\ldots,

where the eigenvalues and eigenfunctions are explicitly given by

ω0=0,c0(τ)=1β;\omega_0=0,\qquad c_0(\tau)=\sqrt{\frac{1}{\beta}};
ω2k−1=2kπβ,c2k−1(τ)=2βsin⁡(2kπτβ),k=1,2,⋯ ;\omega_{2k-1}=\frac{2k\pi}{\beta},\qquad c_{2k-1}(\tau)=\sqrt{\frac{2}{\beta}}\sin\left(\frac{2k\pi\tau}{\beta}\right),\quad k=1,2,\cdots;
ω2k=2kπβ,c2k(τ)=2βcos⁡(2kπτβ),k=1,2,⋯ .\omega_{2k}=\frac{2k\pi}{\beta},\qquad c_{2k}(\tau)=\sqrt{\frac{2}{\beta}}\cos\left(\frac{2k\pi\tau}{\beta}\right),\quad k=1,2,\cdots.

The eigenfunctions {ck(τ)}k=0∞\{c_k(\tau)\}_{k=0}^{\infty} form the orthonormal Fourier basis of L2([0,β];R)L^2([0,\beta];\mathbb{R}), and thus any continuous loop x(⋅)x(\cdot) can be uniquely represented as

x(τ)=∑k=0∞ξkck(τ),τ∈[0,β],x(\tau)=\sum_{k=0}^{\infty}\xi_kc_k(\tau),\quad\tau\in[0,\beta],

where {ξk}k=0∞\{\xi_k\}_{k=0}^{\infty} in Rd\mathbb{R}^d are the coordinates of x(⋅)x(\cdot) in different Matsubara modes. Conversely, for a given continuous loop x(⋅)x(\cdot), the Matsubara coordinates are calculated from

ξk=∫0βx(τ)ck(τ) dτ,k=0,1,2,⋯ .\xi_k=\int_0^\beta x(\tau)c_k(\tau)\,\mathrm{d}\tau,\quad k=0,1,2,\cdots.

Utilizing the Matsubara coordinates, the energy function E(x)\mathcal{E}(x) in (3) is written as

E(ξ)=12∑k=0∞ωk2∣ξk∣2+V(ξ),with V(ξ):=∫0βV(∑k=0∞ξkck(τ))dτ,(5)\mathcal{E}(\xi)=\frac{1}{2}\sum_{k=0}^{\infty}\omega_k^2|\xi_k|^2+\mathcal{V}(\xi),\quad\text{with }\mathcal{V}(\xi):=\int_0^\beta V\left(\sum_{k=0}^{\infty}\xi_kc_k(\tau)\right)\mathrm{d}\tau, \tag*{(5)}

and the target Boltzmann distribution is formally given by exp⁡{−E(ξ)}\exp\{-\mathcal{E}(\xi)\}.

Formulation of the Matsubara mode PIMD

In the following, we construct an infinite-dimensional Langevin dynamics to sample the Boltzmann distribution exp⁡{−E(ξ)}\exp\{-\mathcal{E}(\xi)\}. Without any preconditioning, the vanilla underdamped Langevin dynamics for sampling exp⁡{−E(ξ)}\exp\{-\mathcal{E}(\xi)\} reads

ξ˙k=ηkη˙k=−ωk2ξk−∫0β∇V(x(τ))ck(τ) dτ−γηk+2γB˙k,(6)\begin{aligned} \dot{\xi}_k&=\eta_k\\ \dot{\eta}_k&=-\omega_k^2\xi_k-\int_0^\beta\nabla V(x(\tau))c_k(\tau)\,\mathrm{d}\tau-\gamma\eta_k+\sqrt{2\gamma}\dot{B}_k, \tag*{(6)} \end{aligned}

where γ>0\gamma>0 is the damping rate on each mode, {ηk}k=0∞\{\eta_k\}_{k=0}^{\infty} are the auxiliary velocity variables in Rd\mathbb{R}^d, and {Bk}k=0∞\{B_k\}_{k=0}^{\infty} are independent Brownian motions in Rd\mathbb{R}^d. However, the high-frequency modes in the energy function E(ξ)\mathcal{E}(\xi) poses the infamous stiffness problem [27, 35] in the time discretization of (6): the time step needs to be extremely small to stabilize of the high-frequency part of the dynamics.

In this paper, we employ the preconditioning [27, 36] to scale the internal frequencies of the Matsubara modes. Introduce a>0a>0 and rewrite the energy function E(ξ)\mathcal{E}(\boldsymbol{\xi}) as

E(ξ)=12∑k=0∞(ωk2+a)∣ξk∣2+Va(ξ),with Va(ξ):=∫0βVa(∑k=0∞ξkck(τ)) dτ,(7)\mathcal{E}(\boldsymbol{\xi})=\frac{1}{2}\sum_{k=0}^{\infty}(\omega_k^2+a)|\xi_k|^2+\mathcal{V}^a(\boldsymbol{\xi}),\quad\text{with }\mathcal{V}^a(\boldsymbol{\xi}):=\int_0^\beta V^a\left(\sum_{k=0}^{\infty}\xi_kc_k(\tau)\right)\,\mathrm{d}\tau, \tag*{(7)}

References

  1. [1]Kerson Huang. Statistical mechanics. John Wiley & Sons, 2008.
  2. [2]Donald A. McQuarrie. Statistical mechanics. Sterling Publishing Company, 2000.
  3. [3]Peter W. Atkins, Julio De Paula, and James Keeler. Atkins’ physical chemistry. Oxford university press, 2023.DOI
  4. [4]Neil W. Ashcroft and N. David Mermin. Solid state physics. Cengage Learning, 2022.
  5. [5]Subir Sachdev. Quantum phase transitions. Physics world, 12(4):33, 1999.
  6. [6]Steven A. Orszag. Comparison of pseudospectral and spectral approximation. Studies in Applied Mathematics, 51(3):253–259, 1972.DOI
  7. [7]Jie Shen, Tao Tang, and Li-Lian Wang. Spectral methods: algorithms, analysis and applications. Springer, 4011, 2011.DOI
  8. [8]Richard P. Feynman, Albert R. Hibbs, and Daniel F. Styer. Quantum mechanics and path integrals. Courier Corporation, 2010.
  9. [9]William H. Miller. Path integral representation of the reaction rate constant in quantum mechanical transition state theory. The Journal of Chemical Physics, 63(3):1166–1172, 1975.DOI
  10. [10]Ian R. Craig and David E. Manolopoulos. Chemical reaction rates from ring polymer molecular dynamics. The Journal of chemical physics, 122(8):084106, 2005.pubmed.ncbi.nlm.nih.gov/15836019
  11. [11]Xuecheng Tao, Philip Shushkov, and Thomas F. Miller. Microcanonical rates from ring-polymer molecular dynamics: Direct-shooting, stationary-phase, and maximum-entropy approaches. The Journal of Chemical Physics, 152(12), 2020.DOI
  12. [12]Gregory A. Voth. Feynman path integral formulation of quantum mechanical transition-state theory. The Journal of Physical Chemistry, 97(32):8365–8377, 1993.
  13. [13]Edit Mátyus, David J. Wales, and Stuart C. Althorpe. Quantum tunneling splittings from path-integral molecular dynamics. The Journal of chemical physics, 144(11):114108, 2016.DOI
  14. [14]Christophe L. Vaillant, David J. Wales, and Stuart C. Althorpe. Tunneling splittings from path-integral molecular dynamics using a Langevin thermostat. The Journal of chemical physics, 148(23):234102, 2018.
  15. [15]Scott Habershon, David E. Manolopoulos, Thomas E. Markland, and Thomas F. Miller III. Ring-polymer molecular dynamics: Quantum effects in chemical dynamics from classical trajectories in an extended phase space. Annual review of physical chemistry, 64:387–413, 2013.DOI
  16. [16]Ian R. Craig and David E. Manolopoulos. Quantum statistics and classical mechanics: Real time correlation functions from ring polymer molecular dynamics. The Journal of chemical physics, 121(8):3368–3373, 2004.pubmed.ncbi.nlm.nih.gov/15303899
  17. [17]Seogjoo Jang and Gregory A. Voth. A derivation of centroid molecular dynamics and other approximate time evolution methods for path integral centroid variables. The Journal of chemical physics, 111(6):2371–2384, 1999.DOI
  18. [18]Jianshu Cao and Gregory A. Voth. The formulation of quantum statistical mechanics based on the Feynman path centroid density. III. Phase space formalism and analysis of centroid molecular dynamics. The Journal of chemical physics, 101(7):6157–6167, 1994.DOI
  19. [19]Michael J. Willatt. Matsubara dynamics and its practical implementation. PhD thesis, Cambridge University, 2017.DOI
  20. [20]Timothy J. H. Hele, Michael J. Willatt, Andrea Muolo, and Stuart C. Althorpe. Boltzmann-conserving classical dynamics in quantum time-correlation functions: “Matsubara dynamics”. The Journal of Chemical Physics, 142(13):134103, 2015.DOI
  21. [21]Jianshu Cao and Glenn J. Martyna. Adiabatic path integral molecular dynamics methods. II. Algorithms. The Journal of chemical physics, 104(5):2028–2035, 1996.DOI
  22. [22]Mark E. Tuckerman, Dominik Marx, Michael L. Klein, and Michele Parrinello. Efficient and general algorithms for path integral Car–Parrinello molecular dynamics. The Journal of chemical physics, 104(14):5579–5588, 1996.DOI
  23. [23]Thomas E. Markland and David E. Manolopoulos. An efficient ring polymer contraction scheme for imaginary time path integral simulations. The Journal of chemical physics, 129(2):024105, 2008.DOI
  24. [24]Bruce J. Berne and D. Thirumalai. On the simulation of quantum systems: path integral methods. Annual Review of Physical Chemistry, 37(1):401–424, 1986.DOI
  25. [25]Jianfeng Lu and Zhennan Zhou. Path integral molecular dynamics with surface hopping for thermal equilibrium sampling of nonadiabatic systems. The Journal of Chemical Physics, 146(15), 2017.DOI
  26. [26]Xinzijian Liu and Jian Liu. Path integral molecular dynamics for exact quantum statistics of multi-electronic-state systems. The Journal of Chemical Physics, 148(10), 2018.DOI
  27. [27]Jianfeng Lu, Yulong Lu, and Zhennan Zhou. Continuum limit and preconditioned Langevin sampling of the path integral molecular dynamics. Journal of Computational Physics, 423:109788, 2020.
  28. [28]Roman Korol, Nawaf Bou-Rabee, and Thomas F. Miller. Cayley modification for strongly stable path-integral and ring-polymer molecular dynamics. The Journal of chemical physics, 151(12), 2019.DOI
  29. [29]Rob D. Coalson. On the connection between Fourier coefficient and Discretized Cartesian path integration. The Journal of chemical physics, 85(2):926–936, 1986.DOI
  30. [30]Charusita Chakravarty, Maria C. Gordillo, and David M. Ceperley. A comparison of the efficiency of Fourier-and discrete time-path integral Monte Carlo. The Journal of chemical physics, 109(6):2123–2134, 1998.DOI
  31. [31]Dominique Bakry, Ivan Gentil, Michel Ledoux, et al. Analysis and geometry of Markov diffusion operators. Springer, 103, 2014.DOI
  32. [32]Pierre Monmarché. Hypocoercivity in metastable settings and kinetic simulated annealing. Probability Theory and Related Fields, 172(3–4):1215–1248, 2018.arxiv.org/abs/1502.07263
  33. [33]Pierre Monmarché. Generalized Γ calculus and application to interacting particles on a graph. Potential Analysis, 50(3):439–466, 2019.
  34. [34]James Glimm and Arthur Jaffe. Quantum physics: a functional integral point of view. Springer Science & Business Media, 2012.
  35. [35]Xuda Ye and Zhennan Zhou. Efficient sampling of thermal averages of interacting quantum particle systems with random batches. The Journal of Chemical Physics, 154(20), 2021.DOI
  36. [36]Nawaf Bou-Rabee and Andreas Eberle. Two-scale coupling for preconditioned Hamiltonian Monte Carlo in infinite dimensions. Stochastics and Partial Differential Equations: Analysis and Computations, 9:207–242, 2021.
  37. [37]Jian Liu, Dezhang Li, and Xinzijian Liu. A simple and accurate algorithm for path integral molecular dynamics with the Langevin thermostat. The Journal of chemical physics, 145(2):024103, 2016.DOI
  38. [38]Yu Cao, Jianfeng Lu, and Lihan Wang. On explicit L²-convergence rate estimate for underdamped Langevin dynamics. arXiv preprint arXiv:1908.04746, 2019.DOI
  39. [39]Cédric Villani. Hypocoercivity. Number 949-951. American Mathematical Soc., 2009.
  40. [40]Ben Leimkuhler and Charles Matthews. Molecular dynamics. Interdisciplinary applied mathematics, 39:443, 2015.
  41. [41]Marco Lauricella, Letizia Chiodo, Fabio Bonaccorso, Mihir Durve, Andrea Montessori, Adriano Tiribocchi, Alessandro Loppini, Simonetta Filippi, and Sauro Succi. Multiscale Hybrid Modeling of Proteins in Solvent: SARS-CoV2 Spike Protein as test case for Lattice Boltzmann–All Atom Molecular Dynamics Coupling. arXiv preprint arXiv:2305.05025, 2023.DOI

Paper details

vlnc:W_dmpw8yq9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents