Abstract
We propose a new solvable class of multidimensional quantum harmonic oscillators for a linear diffusive particle and a quadratic energy absorbing well associated with a semi-definite positive matrix force. Under natural and easily checked controllability conditions, for non necessarily reversible models with possibly transient free particle diffusions, the ground state and the zero-point energy are explicitly computed in terms of the unique positive fixed point of a continuous time algebraic Riccati matrix equation. We also present an explicit solution of normalised and time dependent Feynman–Kac measures on path spaces in terms of a time varying linear dynamical system coupled with a differential Riccati matrix equation. A refined non asymptotic analysis of the stability of these models is developed based on a recently developed Floquet-type representation of time varying exponential semigroups of Riccati matrices. We provide explicit and non asymptotic estimates of the exponential decays to equilibrium of Feynman–Kac semigroups in terms of Wasserstein distances or Boltzmann-relative entropy. For reversible models we develop a series of functional inequalities including de Bruijn identity, Fisher’s information decays, log-Sobolev inequalities, and entropy contraction estimates. In this context, we also provide a complete and explicit description of all the spectrum and the excited states of the Hamiltonian, yielding what seems to be the first result of this type for this class of models. We illustrate these formulae with the traditional harmonic oscillator associated with real time Brownian particles and Mehler’s formula. The analysis developed in this article can also be extended to solve time dependent Schrodinger equations equipped with time varying linear diffusions and quadratic potential functions.
Similar content being viewed by others
References
Aguiar, V., Guedes, I.: Fisher information of quantum damped harmonic oscillators. Phys. Scr. 90(4), 045207 (2015)
Ando, T., van Hemmen, J.L.: An inequality for trace ideals. Commun. Math. Phys. 76, 143–148 (1980)
Arnaudon, M., Del Moral, P.: A duality formula and a particle Gibbs sampler for continuous time Feynman–Kac measures on path spaces. Electron. J. Probab. 25, 1–54 (2020)
Baker, H.F.: Note on the integration of linear differential equations. Proc. Lond. Math. Soc. 2(1), 293–296 (1905)
Bateman, H.: Partial Differential Equations of Mathematical Physics. Dover Publications, New York (1944)
Bennett, C.H., Wiesner, S.J.: Communication via one- and two-particle operators on Einstein–Podolsky–Rosen states. Phys. Rev. Lett. 69, 2881 (1992)
Buonaura, M.C., Sorella, S.: Numerical study of the two-dimensional Heisenberg model using a Green function Monte Carlo technique with a fixed number of walkers. Phys. Rev. B 57(18), 11446 (1998)
Anderson, B.D.O., Bishop, A.N., Del Moral, P., Palmier, C.: Backward nonlinear smoothing diffusions. Teor. Veroyatnost. i Primenen. 66(2), 305–326 (2021)
Bishop, A.N., Del Moral, P.: On the stability of Kalman–Bucy diffusion processes. SIAM J. Control Optim. 55(6), 4015–4047 (2017). arXiv:1610.04686
Bishop, A.N., Del Moral, P.: On the stability of matrix-valued Riccati diffusions. SIAM J. Control Optim. 55(6), 4015–4047 (2017)
Bishop, A.N., Del Moral, P., Kamatani, K., Rémillard, R.: On one-dimensional Riccati diffusions. arXiv:1711.10065 (2017)
Bishop, A.N., Del Moral, P.: An explicit Floquet-type representation of Riccati aperiodic exponential semigroups. Int. J. Control 94(1), 258–266 (2021)
Bittanti, S., Laub, A.J., Willems, J.C. (eds.): The Riccati Equation. Springer, New York (1991)
Bouchaud, J.P.: Quantum mechanics with a nonzero quantum correlation time. Phys. Rev. A 96(5), 052116 (2017)
Blanes, S., Casas, F., Oteo, J.A., Ros, J.: The Magnus expansion and some of its applications. Phys. Rep. 470(5–6), 151–238 (2009)
Brockett, R.W.: Finite Dimensional Linear Systems. Wiley, New York (1970)
Caffarel, M., Claverie, P.: Treatment of the Schrödinger equation through a Monte Carlo method based upon the generalized Feynman–Kac formula. J. Stat. Phys. 43(5), 797–801 (1986)
Caffarel, M., Claverie, P.: Development of a pure diffusion quantum Monte Carlo method using a full generalized Feynman–Kac formula. I. Formalism & II Applications to simple systems. J. Chem. Phys. 88(2), 1088–1099, 1100–1109 (1988)
Caffarel, M.: Stochastic methods in quantum mechanics. In: Numerical Determination of the Electronic Structure of Atoms, Diatomic and Polyatomic Molecules, pp. 85–105. Springer, Dordrecht (1989)
Caffarel, M., Assaraf, R.: A pedagogical introduction to quantum Monte Carlo. In: Defranceschi, M., Le Bris, C. (eds.) Mathematical Models and Methods for Ab Initio Quantum Chemistry in Lecture Notes in Chemistry, p. 45. Springer, New York (2000)
Cancès, E., Jourdain, B., Lelièvre, T.: Quantum Monte Carlo simulations of fermions. A mathematical analysis of the fixed-node approximation. Math. Mod. Methods Appl. Sci. 16(9), 1403–1440 (2006)
Champagnat, N., Villemonais, D.: Exponential convergence to quasi-stationary distribution and \(h\)-process. Probab. Theory Relat. Fields 164(1–2), 243–283 (2016)
Champagnat, N., Villemonais, D.: General criteria for the study of quasi-stationarity. arXiv preprint arXiv:1712.08092 (2017)
Champagnat, N., Villemonais, D.: Lyapunov criteria for uniform convergence of conditional distributions of absorbed Markov processes. Stoch. Process. Appl. 135, 51–74 (2021)
Champagnat, N., Villemonais, D.: Uniform convergence to the \( Q \)-process. Electr. Commun. Probab. 22 (2017)
Cook Horne, C.: Analyzing the Dynamics of Coupled Quantum Harmonic Oscillators. Dissertation Brigham Young University (2017)
Coppel, W.A.: Dichotomies in Stability Theory. Springer, New York (1978)
Cordero-Soto, R., Suazo, E., Suslov, S.K.: Models of damped oscillators in quantum mechanics. J. Phys. Math. 1, 1–16 (2009)
Cordero-Soto, R., Suazo, E., Suslov, S.K.: Quantum integrals of motion for variable quadratic Hamiltonians. Ann. Phys. 325(9), 1884–1912 (2010)
Davidova, A.L., Bordely, S., Neda, Z.: Collective behavior of coupled quantum mechanical oscillators. Stud. Univ. Babes-Bolyai Phys. 60(1) (2015)
Davydov, A.S.: Quantum Mechanics. Pergamon Press, Oxford (1965)
Dekker, H.: Classical and quantum mechanics of the damped harmonic oscillator. Phys. Rep. 80(1), 1–110 (1981)
Del Moral, P.: Feynman–Kac Formulae. Springer, New York (2004)
Del Moral, P., Miclo, L.: A Moran particle system approximation of Feynman–Kac formulae. Stoch. Process. Appl. 86(2), 193–216 (2000)
Del Moral, P.: Mean Field Simulation for Monte Carlo Integration. CRC Press, Boca Raton (2013)
Del Moral, P., Miclo, L.: Branching and interacting particle systems approximations of Feynman-Kac formulae with applications to non-linear filtering. In Séminaire de Probabilités, XXXIV, Volume 1729, Lecture Notes in Mathematics, pp. 1–145. Springer, Berlin (2000)
Del Moral, P., Penev, S.: Stochastic Processes: From Applications to Theory. CRC Press, Boca Raton (2017)
Del Moral, P., Miclo, L.: Particle approximations of Lyapunov exponents connected to Schrödinger operators and Feynman–Kac semigroups. ESAIM: Probability and Statistics, no. 7, pp. 171–208 (2003)
Del Moral, P., Miclo, L.: On the stability of nonlinear Feynman-Kac semigroups. Annales de la Faculté des sciences de Toulouse: Mathématiques 11(2) (2002)
Del Moral, P., Tugaut, J.: On the stability and the uniform propagation of chaos properties of ensemble Kalman–Bucy filters. Ann. Appl. Probab. 28(2), 790–850 (2018)
Del Moral, P., Kurtzmann, A., Tugaut, J.: On the stability and the uniform propagation of chaos of a class of extended ensemble Kalman–Bucy filters. SIAM J. Control Optim. 55(1), 119–155 (2017)
Delor, M., et al.: Directing the path of light-induced electron transfer at a molecular fork using vibrational excitation. Nat. Chem. https://doi.org/10.1038/nchem.2793 (2017)
Dimassi, M., Sjöstrand, J.: Spectral Asymptotics in the Semi-Classical Limit. Cambridge University Press, Cambridge (1999)
Ekert, A.K.: Quantum cryptography based on Bell’s theorem. Phys. Rev. Lett. 67, 661 (1991)
Fano, U.: Description of states in quantum mechanics by density matrix and operator techniques. Rev. Mod. Phys 29, 7493 (1957)
Gobet, E., She, Q.: Perturbation of Ornstein–Uhlenbeck stationary distributions: expansion and simulation. HAL Id:hal-01345926 (2016)
Evensen, G.: The ensemble Kalman filter: theoretical formulation and practical implementation. Ocean Dyn. 53(4), 343–367 (2003)
Frazer, R.A., Duncan, W.J., Collar, A.R.: Elementary Matrices and Some Applications to Dynamics and Differential Equations. Cambridge University Press, Cambridge (1938)
Flugge, S.: Practical Quantum Mechanics. Springer, Berlin (1999)
Grigis, A., Sjöstrand, J.: Microlocal Analysis for Differential Operators: An Introduction. Cambridge University Press, Cambridge (1994)
Hetherington, J.H.: Observations on the statistical iteration of matrices. Phys. Rev. A 30(5), 2713 (1984)
Higham, N.J.: Functions of Matrices: Theory and Computation. SIAM, Philadelphia (2008)
Ince, E.L.: Ordinary Differential Equations. Dover, New York (1956)
Ismail, M.E.H., Ruiming, Z.: A review of multivariate orthogonal polynomials. J. Egypt. Math. Soc. 25(2), 91–110 (2017)
Ikeda, S., Fillaux, F.: Incoherent elastic-neutron-scattering study of the vibrational dynamics and spin-related symmetry of protons in the KHCO3 crystal. Phys. Rev. B 59, 41344145 (1999)
Ivrii, V.: Microlocal Analysis and Precise Spectral Asymptotics. Springer, New York (2013)
Kustura, K., Rusconi, C.C., Romero-Isart, O.: Quadratic quantum Hamiltonians: general canonical transformation to a normal form. Phys. Rev. A 99(2), 022130 (2019)
Landau, L.D., Lifshitz, E.M.: Mechanics. Pergamon Press, Oxford (1976)
Landau, L.D., Lifshitz, E.M.: Quantum Mechanics: Nonrelativistic Theory. Pergamon Press, Oxford (1977)
Lancaster, P., Rodman, L.: Algebraic Riccati Equations. Oxford University Press, Oxford (1995)
Lelièvre, T., Rousset, M., Stoltz, G.: Free Energy Computations: A Mathematical Perspective. Imperial College Press, London (2010)
Lelièvre, T., Rousset, M., Stoltz, G.: Computation of free energy differences through non-equilibrium stochastic dynamics: the reaction coordinate case. J. Comput. Phys. 222(2), 624–643 (2007)
Makarov, D.N.: Coupled harmonic oscillators and their quantum entanglement. Phys. Rev. E 97(4), 042203 (2018)
Martinez, A.: An Introduction to Semiclassical and Microlocal Analysis. Springer, New York (2002)
McDermott, R.M., Redmount, I.H.: Coupled classical and quantum oscillators. arXiv preprint arXiv:quant-ph/0403184 (2004)
Magnus, W.: On the exponential solution of differential equations for a linear operator. Commun. Pure Appl. Math. VII 4, 649–673 (1954)
Park, D.: Dynamics of entanglement in three coupled harmonic oscillator system with arbitrary time-dependent frequency and coupling constants. Quantum Inf. Process. 18(9), 1–23 (2019)
Paz, J.P., Roncaglia, A.J.: Dynamics of the entanglement between two oscillators in the same environment. Phys. Rev. Lett. 100, 220401 (2008)
Peano, G.: Intégration par séries des équations différentielles linéaires. Math. Ann. 32(3), 450–456 (1888)
Pollock, E.L., Ceperley, D.M.: Simulation of quantum many-body systems by path-integral methods. Phys. Rev. B 30(5), 2555 (1984)
Prach, A., Tekinalp, O., Bernstein, D.S.: Infinite-horizon linear-quadratic control by forward propagation of the differential Riccati equation. IEEE Control Syst. 35(2), 78–93 (2015)
Prach, A.: Faux Riccati equation techniques for feedback control of nonlinear and time-varying systems. Ph.D. thesis. School of Natural and Applied Sciences. Aerospace Engineering. Middle East Technical University (2015)
Romero, E., et al.: Quantum coherence in photosynthesis for efficient solar-energy conversion. Nat. Phys. 10, 676–682 (2014)
Rousset, M.: On the control of an interacting particle approximation of Schrödinger ground states. SIAM J. Math. Anal. 38(3), 824–844 (2006)
Rousset, M., Stoltz, G.: Equilibrium sampling from non equilibrium dynamics. J. Stat. Phys. 123(6), 1251–1272 (2006)
Halpin, A., et al.: Two-dimensional spectroscopy of a molecular dimer unveils the effects of vibronic coupling on exciton coherences. Nat. Chem. 6, 196201 (2014)
Rasmussen, C.E., Williams, C.K.I.: Gaussian Processes for Machine Learning. Adaptive Computation and Machine Learning. MIT Press, Cambridge (2006)
Samuel, L., Braunstein, H., Kimble, J.: Teleportation of continuous quantum variables. Phys. Rev. Lett. 80, 869–872 (1998)
Schiff, L.I.: Quantum Mechanics, 3rd edn. McGraw-Hill, New York (1968)
Sjöstrand, J.: Microlocal Analysis for the Periodic Magnetic Schrödinger Equation and Related Questions. Microlocal Analysis and Applications, pp. 237–332. Springer, Berlin (1991)
Tao, Y., Mehta, P.G., Meyn, S.P.: Feedback particle filter. IEEE Trans. Autom. Control 58(10), 2465–2480 (2013)
Tao, Y., Laugesen, R.S., Mehta, P.G., Meyn, S.P.: Multivariable feedback particle filter. Automatica 71, 10–23 (2016)
Taghvaei, A., Mehta, P.G.: Gain function approximation in the feedback particle filter. In: 2016 IEEE 55th Conference on Decision and Control (CDC), pp. 5446–5452. IEEE (2016)
Taghvaei, A., Mehta, P.G.: An optimal transport formulation of the linear feedback particle filter. In: 2016 American Control Conference (ACC). IEEE (2016)
Tao, Y., Blom, H.A.P, Mehta, P.G.: The continuous-discrete time feedback particle filter. In: American Control Conference. IEEE (2014)
van der Spek, R.: From the Quantum Harmonic Oscillator to the Ornstein-Uhlenbeck Process and Back. http://resolver.tudelft.nl/uuid:08ad4a94-483d-46db-8840-6f73c3e48a70Thesis. TU Delft, Applied Mathematics and Applied Physics (2019)
Yosida, K.: Brownian motion on the surface of the 3-sphere. Ann. Math. Stat. 20(2), 292–296 (1949)
Zhang, Y., Liu, W., Chen, Z., Li, K., Wang, J.: On the Properties of Kullback–Leibler Divergence Between Gaussians. arXiv preprint arXiv:2102.05485 (2021)
Zworski, M.: Semiclassical Analysis. AMS, Providence (2012)
Acknowledgements
The authors would like to thank two anonymous referees, whose comments helped to shape the presentation of this article.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no competing interests to declare that are relevant to the content of this article.
Additional information
Communicated by S. Chatterjee.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
A Riccati Matrix Differential Equations
In this section, we discuss some of the theory behind Riccati matrix differential equations and present some results that are of use throughout the paper. This section is mainly taken from [12].
1.1 A.1 Gians fixed point formulae
Recall the the positive and negative fixed points \((P^-_{\infty },P_{\infty })\) of the Riccati equation (28). The difference between the two
is defined via the Gramian matrices
Consider now the linear matrix functional
Rearranging and using (98) implies that
Recall that \(Q_{\infty }\) is defined as \(P_{\infty }\) by replacing (A, R, S) by \((A^{\prime },S,R)\). In the same vein, \(\Delta _t^h\) is defined as \(\Delta _t\) by replacing (A, R, S) and \(P_{\infty }\) by \((A^{\prime },S,R)\) and \(Q_{\infty }\). Thus, by symmetry arguments and (30), we also have
with the Gramian matrices
The following lemma proves the second equality on the right hand side of (39).
Lemma A.1
For any (A, R, S) satisfying the rank condition (4), we have
Proof
The Gramian \(\Delta _{\infty }\) satisfies the Sylvester equations given by
It then follows that
where we have used (97) to obtain the final equality. From this we obtain
In the same vein, we have
Combining the last assertion with (103) we conclude that
as required. \(\square \)
1.2 A.2 A Floquet-type representation
For any \(P\in {{\mathcal {S}} }^0_r\) and \(\delta >0\) set
The next theorem provides an explicit description of \( {{\mathcal {E}} }_t(P)\), defined in (33), in terms of the matrices \((A,S,P_{\infty })\). In what follows, \(\Vert \cdot \Vert \) stands for the spectral norm of matrices.
Theorem A.2
[Floquet-type representation [12]]. For any time horizon \(t\ge 0\) and any \(P\in {{\mathcal {S}} }_r^0\) we have Riccati exponential semigroup formula
where \(\mathbb {F}_t(P)\) was defined in (99). For any \(t\ge \delta >0\) we have the uniform estimates
In addition, for any \(t\ge 0\) we have the exponential estimates
for some positive matrices \(\Pi _{-,\delta },\Pi _{+,\delta }\) and some \(\alpha ,\beta >0\), all of which depend on the model parameters (A, R, S). Finally, we have the bounds
From this theorem, we can develop some useful identities. Indeed, using the decomposition
for \(Q_1, Q_2 \in {{\mathcal {S}} }_r^0\), applying (104) we have the closed form Lipschitz type matrix formula
Applying (109) with \(Q_2=P_{\infty }\) and using (100), we recover the Bernstein-Prach-Tekinalp formula [71, 72] given by
1.3 A.3 Lipschitz inequalities
Combining Theorem A.2, (106) and (109) we easily obtain the following result.
Theorem A.3
For any time horizon \(t\ge \delta >0\) and any \(Q_1,Q_2\in {{\mathcal {S}} }_r^0\) we have the Lipschitz estimate
with the parameters \((\alpha ,\beta , \chi _{\delta })\) defined in (106) and Theorem A.2. In addition, for any \(t\ge 0\) we have the local Lipschitz estimate
with the parameters \(\chi (P_i)\) defined in Theorem A.2.
Noting that
where \(\Delta _t\) was defined in (98), we also obtain the following corollary.
Corollary A.4
For any time horizon \(t\ge \delta >0\) and any \(Q_1,Q_2\in {{\mathcal {S}} }_r^0\) we have the Lipschitz estimate
with the parameters \((\alpha ,\beta , \chi _{\delta })\) defined in (106) and Theorem A.2. In addition, for any \(t\ge 0\) we have local Lipschitz estimate
with the parameter \(\chi (Q_i)\) defined in Theorem A.2.
The first coordinate of the evolution semigroup (32) can be written as
Using the decomposition
we readily check the following theorem.
Theorem A.5
For any time horizon \(t\ge \delta >0\) and any \(Q_1,Q_2\in {{\mathcal {S}} }_r^0\) we have the estimate
with the parameters \((\alpha ,\beta , \chi _{\delta })\) defined in (106) and Theorem A.2. In addition, for any \(t\ge 0\) we have the estimate
with the parameter \(\chi (Q_i)\) defined in Theorem A.2.
B Path Integral Formulations
In this section, we discuss in further detail the h-process and its link to the absorbed particle process \((X_t^c)_{t \ge 0}\). We also discuss how one may extend the methodology used in this article to non-linear diffusions.
1.1 B.1 Backward h-processes
For a fixed time horizon \(t\ge 0\), we let \(\overline{X}_t\) be a random sample from \( {{\mathcal {N}} }(\widehat{X}_t,P_t)\). We also denote by \(X^h_{t,s}(x)\), with \(s\in [0,t]\), be the backward diffusion defined by
starting at \(X^h_{t,t}(x)=x\) at time \(s=t\). In the above display, \(P_s\) stands for the solution of the Riccati matrix differential equation defined in (27). The backward semigroup property is given for any \(u\le s \in [0,t]\) by the mapping composition formulae
We assume that \(\overline{X}_t\) and \((W_s)_{s\le t}\) are independent.
Rewritten in terms of the density \(g_s\) of the Gaussian distribution \( {{\mathcal {N}} }(\widehat{X}_s,P_s)\), we have
The following theorem, taken from [8] links the non-absorbed particle process with the above backward diffision
Theorem B.1
[ [8]]. Assume that \(X_0\sim {{\mathcal {N}} }(\widehat{X}_0,P_0)\). In this situation, for any \(t\ge 0\) we have the backward formulation of the Feynman–Kac path integral
The random state \(\overline{X}_{t,s}^h\) is a Gaussian variable with a mean \(\widehat{X}^h_{t,s}\) and covariance matrix \(P^h_{t,s}\) satisfying the backward equations
with the terminal condition \((\widehat{X}^h_{t,t},P_{t,t}^h)=(\widehat{X}_t,P_t)\), where \((\widehat{X}_s,P_s)\) is the solution to the forward equations described in (27).
1.2 B.2 Extensions to non-linear diffusions
The h-process methodology can be extended to more general generators \( {{\mathcal {L}} }\) and other choices of the potential function V. We now assume that \( {{\mathcal {L}} }\) is the generator of the diffusion equation
for some drift function A(x) and some diffusion matrix valued function B(x) with appropriate dimensions. We also assume there exists some ground state \(h_0\) associated with some energy \(\lambda _0\); that is, we have that
In this situation, the h-process \(X^{h}_{t}\) is a diffusion with generator defined by
with the carré-du-champ operator
where we have defined \(R(x):=B(x)B(x)^{\prime }\). Equivalently, the h-process is defined by the diffusion
Let \(\overline{X}_t\) a random sample from the Feynman–Kac probability measures \(\eta _t\) defined as in (21) for some potential function V.
Whenever it exists, let \(g_s\) be the density of the normalised or unnormalised Feynman–Kac measures \(\eta _s\) or \(\gamma _s\). In this situation, following the analysis developed in [8], the assertion of Theorem B.1 remains valid with the backward diffusion
with the terminal condition \(\overline{X}_{t,t}^h=\overline{X}_{t}\) and the R-divergence m-column vector operator with j-th entry given by the formula
C McKean–Vlasov Interpretations
In this section, we discuss several classes of McKean-Vlasov interpretations of the distribution of a non-absorbed particle. These probabilistic models and their mean field simulation are defined in terms of a nonlinear Markov process that depends on the distribution of the random states so that the flow of distributions of all random states coincides with the conditional distribution of a non-absorbed particle. To the best of our knowledge, these particle models have not been applied to the models considered in this article, despite being applicable in a wide variety of situations. In particular, these methods can be applied in high dimension and even when A is unstable.
1.1 C.1 Interacting jump processes
Let \(\overline{X}_t\) be a nonlinear jump diffusion process with generator
The process starts at \(\overline{X}_0=X_0\). Recall that the second order differential kinetic energy operator \( {{\mathcal {L}} }\) defined in (2) coincides with the generator of the process \(X_t\) arising in the Feynman–Kac representation (8). Between the jumps the process \(\overline{X}_t\) evolves as \(X_t\). At rate \(V(\overline{X}_t)\) the process jumps onto a new location randomly selected according to the distribution \(\overline{\eta }_t\). Observe that
This shows that \(\overline{\eta }_t\) satisfies the same evolution equation as the one satisfied by \(\eta _t\) given in (26). Thus, for any choice of the generator \( {{\mathcal {L}} }\) and any choice of the potential function V we have that
The mean field particle interpretation of the nonlinear process \(\overline{X}_t\) is defined by a system of N walkers, \(\xi ^i_t\), evolving independently as \(X_t\) with a spatial jump rate \(V_t\), for \(1\le i\le N\). At each jump time, the particle \(\xi ^{i}_t\) jumps onto a particle uniformly chosen in the pool. The occupation measure of system is given by the empirical measure
Mimicking (23) we also define the normalising constant approximations
Observe that the N ancestral lines \(\zeta _t^i:=(\xi ^i_{s,t})_{0\le s\le t}\) of length t of the above genetic-type process can also be seen as a system of N path-valued particles evolving independently as the historical process \(Y_t=(X_{s})_{0\le s\le t}\) of \(X_t\), with a jump rate \(V_t(X_t)\) that only depends on the terminal state \(X_t\) of the ancestral line \(Y_t\).
The interacting particle system discussed above belongs to the class of diffusion Monte Carlo algorithms, see for instance the series of articles [17,18,19,20], as well as the articles [21, 38, 61, 62, 74, 75] in the context of ground state calculations and the series of articles [3, 34, 36,37,38] in the context of a general class of continuous time Feyman-Kac path integral formulae. The discrete time version of the interacting jump processes discussed above coincides with the genetic-type samplers and the Monte Carlo reconfiguration/selection methodologies discussed in [33, 36, 38], see also the pioneering articles in the mid-1980s by Hetherington [51] and further extended by Caffarel and his co-authors in the series of articles [17,18,19], see also Buonaura-Sorella [7] as well as the pedagogical introduction to quantum Monte Carlo by Caffarel-Assaraf [20].
1.2 C.2 Interacting diffusions
For any probability measure \(\eta \) on \(\mathbb {R}^r\) we let \( {{\mathcal {P}} }_{\eta }\) denote the \(\eta \)-covariance
where \(e(x):=x\) is the identity function and \(\eta (f)\) is a column vector whose i-th entry is given by \(\eta (f^i)\) for some measurable function \(f:\mathbb {R}^r\rightarrow \mathbb {R}^r\).
We now consider three different nonlinear McKean-Vlasov-type diffusion process,
for any skew symmetric matrix \(M^\prime _t=-M_t\) that may also depend \(\overline{\eta }_t\). In all three cases \(( {{\mathcal {W}} }_t,\overline{ {{\mathcal {W}} }}_t)\) are independent copies of \(W_t\); and \(\overline{X}_0\) is an independent copies of \(X_0\). We also assume that \(( {{\mathcal {W}} }_t, {{\mathcal {W}} }_t,\overline{X}_0)\) are independent. In all three cases in (117), \(\overline{\eta }_t \) stands for the probability distribution of \(\overline{X}_t\); that is, we have that
Observe that, in all three cases the stochastic processes discussed above depend in some nonlinear fashion on the law of the diffusion process itself.
Theorem C.1
In all the three cases presented in (117), for any \(t\ge 0\) we have the Gaussian preserving property
Proof
Let \(\overline{X}_t\) be the process defined as in (1) by replacing \( {{\mathcal {P}} }_{\overline{\eta }_t}\) by \(P_t\). In this case, we have
Applying Ito’s formula and taking expectations we obtain
This yields the linear system
We conclude that \(\overline{X}_t\) is a linear diffusion with mean \(\widehat{X}_t\) and covariance matrix \(P_t\). The proof for the other two cases follows the same lines of arguments, thus we leave the details to the reader. \(\square \)
The mean-field particle interpretation of the first nonlinear diffusion process in (117) is given by the Mckean–Vlasov type interacting diffusion process
where \(( {{\mathcal {W}} }_{t}^i,\overline{ {{\mathcal {W}} }}^i_t,\xi ^i_0)_{1\le i\le N}\) are N independent copies of \(( {{\mathcal {W}} }_{t},\overline{ {{\mathcal {W}} }}_t\overline{X}_0)\). In the above display, the \(P^N_t\) are the rescaled empirical covariance matrices given by the formulae
with the empirical measures
Note that (119) is a set of N stochastic differential equations coupled via the empirical covariance matrix \(P_t^N\). The mean-field particle interpretation of the second and third nonlinear diffusion processes in (117) are defined as above by replacing \( {{\mathcal {P}} }_{\eta _t}\) by the sample covariance matrices \(P^N_t\). The quasi-invariant measure \(\eta _{\infty }\) and the parameter \(\lambda _0\) are computed using the limiting formulae (114) and (115).
The interacting diffusions discussed above belong to the class of Ensemble Kalman filters, see for instance the pioneering article by Evensen [47], the series of articles [9,10,11], as well as [40, 41] and the references therein.
In contrast with the interacting jump process discussed in “Appendix C.1” none of the nonlinear diffusions discussed in (117) can be extended to more general generators \( {{\mathcal {L}} }\) and other choices of the potential function V.
We end this section with an application of the seminal feedback particle filter methodology recently developed by Mehta and Meyn and their co-authors [81,82,83,84,85] to Feynman–Kac models. Consider the diffusion
where \(U_t(x)\) is the solution of the Poisson equation
In the above display \(g_t(x)\) stands for the density of the distribution \(\overline{\eta }_t\) of the random state \(\overline{X}_t\). The generator \( {{\mathcal {L}} }_{\overline{\eta }_t}\) of the above time varying diffusion satisfies the equation
Integrating by part the last term we obtain the formula
from which we conclude that
This shows that \(\overline{\eta }_t=\hbox { Law}(\overline{X}_t)=\eta _t\) coincides with the normalised Feynman–Kac measures.
For linear-Gaussian models we have \(\eta _t= {{\mathcal {N}} }(\widehat{X}_t,P_t)\). Thus, the Poisson equation resumes to the formula
The solution of the above equation is clearly given by
The resulting diffusion coincides with the second case in (117).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Del Moral, P., Horton, E. Coupled Quantum Harmonic Oscillators and Feynman–Kac path integrals for Linear Diffusive Particles. Commun. Math. Phys. 402, 2079–2127 (2023). https://doi.org/10.1007/s00220-023-04772-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-023-04772-z