Assumption-free real-time path-integral calculation of Caldeira-Leggett decoherence beyond the Gaussian, Ohmic, high-temperature regime

The first-principles real-time path-integral method of Nishimura and Watanabe (JHEP 09 (2025) 197) computes the reduced density matrix of the Caldeira-Leggett model without Born, Markov, or high-temperature assumptions: because the discretized action is quadratic, the unique complex saddle point is obtained from one large sparse linear system and the Gaussian fluctuations are integrated analytically. The method confirmed the master-equation decoherence scaling $\Gamma_{\text{off-diag}}(t)\sim 16\gamma t/\beta$ only inside the weak-coupling, high-temperature, Ohmic window. The open problem is to carry this assumption-free program beyond that exactly solvable window: to anharmonic system potentials and non-Gaussian initial states via the generalized thimble method (GTM), to sub-Ohmic and super-Ohmic spectral densities and lower-temperature/stronger-coupling regimes where master-equation scaling laws are untested, to the joint thermodynamic limit $N_{\mathcal{E}}\to\infty$ and long-time limit $t\to\infty$ with finite-environment recurrences under control, and to explicit calculations of environment-induced superselection (einselection). No published extension is known in the literature to date: the only Lefschetz-thimble-based decoherence calculations are the authors' own two papers, and adjacent system-bath methods (inchworm with frozen Gaussian approximation, iterative Dyson-series schemes, perturbative anharmonic master equations) do not deliver the assumption-free, contour-deformed computation requested here.

Background

Quantum decoherence is the dynamical process by which a quantum system entangled with its environment loses the capacity to exhibit interference between macroscopically distinct states; it is the standard mechanism invoked to explain the quantum-to-classical transition and, in quantum technologies, it appears as quantum noise that must be controlled. The Caldeira-Leggett (CL) model is the canonical microscopic model of decoherence and quantum Brownian motion (Caldeira and Leggett, Physica A 121 (1983) 587): a system degree of freedom is coupled linearly to a bath of $N_{\mathcal{E}}$ independent harmonic oscillators, and the reduced density matrix $\rho_{S}(x,\tilde{x};t)$ of the system is obtained by tracing out the bath from the unitary evolution of the whole.

The textbook description of the CL model is a master equation, a non-unitary evolution equation for $\rho_{S}$, derived under the Born approximation (weak system-environment entanglement), the Markov approximation (short bath memory), and a high-temperature condition, for an Ohmic environment in which the frequency-weighted coupling of bath modes scales as $\rho(\omega)C^{2}(\omega)\propto\omega^{p}$ with $p=2$; validity of the simple master equation further requires the hierarchy $\omega_{r}\ll\omega_{\mathrm{cut}}\ll\beta^{-1}$ among the renormalized system frequency $\omega_{r}$, the bath cutoff $\omega_{\mathrm{cut}}$, and the temperature $1/\beta$. In this regime the master equation predicts that off-diagonal elements of $\rho_{S}$ in the position basis (entries with $x\neq\tilde{x}$) decay exponentially at a rate proportional to $\gamma/\beta$, where $\gamma$ is the effective coupling, so that decoherence is parametrically faster than energy dissipation.

Nishimura and Watanabe (JHEP 09 (2025) 197; short version in Phys. Rev. Lett. 134 (2025) 210401) developed a first-principles alternative: evaluate the real-time path integral of the full system-plus-environment on a computer and trace out the environment numerically, with no Born, Markov, or high-temperature input. Real-time path integrals are oscillatory, which defeats naive Monte Carlo (the sign problem); the general cure is the Lefschetz thimble method, which deforms the integration contour into the complex plane along thimble manifolds on which the integrand decays, and its practical Monte Carlo implementation, the generalized thimble method (GTM), which has been demonstrated for real-time quantum mechanics including tunneling in anharmonic potentials (Nishimura, Sakai and Yosprakob, JHEP 09 (2023) 110). For the CL model with a harmonic system and Gaussian initial states, the discretized action is quadratic, so the path integral is Gaussian: the unique complex saddle point is found by solving one large sparse linear system of size $D=2N_{t}(1+N_{\mathcal{E}})+N_{\beta}N_{\mathcal{E}}$ (with $N_{t}$ and $N_{\beta}$ the numbers of real-time and Euclidean-time lattice steps), and the fluctuation integral is done analytically. For Gaussian reduced density matrices the physics is captured by two inverse widths defined through $|\rho_{S}(x,\tilde{x};t)|\simeq\exp[-\tfrac{1}{2}\Gamma_{\text{diag}}(t)((x+\tilde{x})/2)^{2}-\tfrac{1}{2}\Gamma_{\text{off-diag}}(t)((x-\tilde{x})/2)^{2}]$: growth of $\Gamma_{\text{off-diag}}(t)$ signals decoherence, while thermalization with the environment acting as a heat bath implies $\Gamma_{\text{diag}}(t)\to 2\omega_{r}\tanh(\beta\omega_{r}/2)$ and $\Gamma_{\text{off-diag}}(t)\to 2\omega_{r}\coth(\beta\omega_{r}/2)$ at late times. The calculation confirmed the master-equation prediction $\Gamma_{\text{off-diag}}(t)\sim(16\gamma/\beta)t$ in the weak-coupling, high-temperature window (parameters up to $\gamma\leq 0.4$, $\beta\geq 0.05$, $\omega_{\mathrm{cut}}\leq 16$, $N_{\mathcal{E}}\leq 512$, $t\lesssim 3$ in the paper's units), observed deviations outside it, and extrapolated to $N_{\mathcal{E}}=\infty$ at fixed moderate times.

What the demonstrated calculation cannot yet do defines the open problem. Outside the exactly-solvable setting, anharmonic system potentials or non-Gaussian initial states, the action is no longer quadratic, and the GTM Monte Carlo machinery has never been applied to an open system with an environment to trace out. The computational bottleneck is severe: the direct LU solution of the saddle-point equation costs $O(D^{3})$, which limits the attainable $N_{\mathcal{E}}$ and $t$; the paper notes that an iterative solver could in principle reduce the cost to $O(D)$ thanks to matrix sparseness, but none was implemented. Off-diagonal observables at late times show significant finite-$N_{\mathcal{E}}$ dependence attributed to recurrences of the finite environment, whose quantitative scaling is uncharacterized. Finally, environment-induced superselection (einselection), the selection of robust pointer states by the environment (Zurek, Phys. Rev. D 26 (1982) 1862), is established for the damped harmonic oscillator only within master-equation/Gaussian treatments (coherent states as pointer states; Zurek, Habib and Paz, Phys. Rev. Lett. 70 (1993) 1187), and has not been probed by the assumption-free real-time path-integral calculation across parameter regimes. Adjacent method families exist for system-bath dynamics, including inchworm Monte Carlo with frozen Gaussian approximation (Wang, Yang and Cai, Quantum 9 (2025) 1667), iterative Dyson-series schemes (Sun, Wang and Cai, J. Chem. Theory Comput. 22 (2026) 8071), and, for the purely harmonic system, the exact quantum-Brownian-motion master equation for arbitrary spectral densities (Hu, Paz and Zhang, Phys. Rev. D 45 (1992) 2843); none delivers the assumption-free, contour-deformed real-time path-integral computation beyond the Gaussian case that this problem targets.

Problem Statement

Extend the assumption-free, first-principles real-time path-integral calculation of open-system decoherence in the Caldeira-Leggett (CL) model beyond the exactly solvable Gaussian, Ohmic, high-temperature, weak-coupling window in which it has been demonstrated, and use the extended calculation to map the validity of master-equation predictions and the emergence of classicality. All four results below are required; delivering a subset does not resolve the objective. The requested results, which jointly support this single objective, are:

(a) Anharmonic/non-Gaussian extension: perform the first-principles real-time path-integral calculation of the reduced density matrix for a CL-type model with an anharmonic system potential and/or non-Gaussian initial system states, using the generalized thimble method (GTM) to overcome the non-Gaussian real-time sign problem, including the tracing-out of a large oscillator environment.

(b) Regime map: within the assumption-free formalism, explore parameter regimes outside the validated window, including sub-Ohmic ($p<2$) and super-Ohmic ($p>2$) spectral densities $\rho(\omega)C^{2}(\omega)\propto\omega^{p}$, lower temperatures, and stronger coupling, and determine where the master-equation decoherence scaling (e.g. $\Gamma_{\text{off-diag}}(t)\sim 16\gamma t/\beta$) survives, what scaling laws replace it, and what non-Markovian deviations occur. A resolving regime map must localize the boundary values of the spectral exponent $p$, the temperature, and the coupling at which the quoted master-equation scaling fails, together with the scaling law that holds on each side of that boundary.

(c) Joint thermodynamic and long-time limit: establish, with controlled discretization, the joint limit $N_{\mathcal{E}}\to\infty$ and $t\to\infty$ for the reduced density matrix, confirming (or refuting) the approach of $\Gamma_{\text{diag}}(t)$ and $\Gamma_{\text{off-diag}}(t)$ to the thermal-equilibrium predictions $2\omega_{r}\tanh(\beta\omega_{r}/2)$ and $2\omega_{r}\coth(\beta\omega_{r}/2)$, and quantify the finite-$N_{\mathcal{E}}$ recurrence effects that contaminate the off-diagonal sector at late times. This may require the enabling algorithmic step of replacing the $O(D^{3})$ direct LU solve of the sparse saddle-point system by scalable iterative or preconditioned solvers, which the source identifies as an open computational task in its own right.

(d) Einselection probe: use such explicit calculations to investigate environment-induced superselection and the quantum-to-classical transition across parameter regimes, that is, how pointer states emerge and how the preferred basis is selected by the environment, beyond the Gaussian/master-equation treatments in which this is currently established. A resolving answer must identify the emerging pointer states in a given regime and quantify their robustness, for instance by comparing the decay of coherences between candidate pointer states against that of generic superpositions, and must state in which parameter regimes the selected preferred basis changes.

The common objective binds these results: they are sequential uses of one assumption-free formalism on the same model class, with (a) establishing the beyond-Gaussian capability, (b) mapping the validity of existing predictions, (c) fixing the asymptotics that make the finite computations meaningful, and (d) applying the capability to the interpretive question. An answer must keep the calculations free of the Born, Markov, and high-temperature assumptions (any approximation introduced must be stated and controlled), and must report the observables with finite-$N_{\mathcal{E}}$, finite-lattice-spacing, and finite-time-step errors under control.

References

1. J. Nishimura and H. Watanabe, "Quantum decoherence in the Caldeira-Leggett model by the real-time path integral on a computer", JHEP 09 (2025) 197, arXiv:2503.20699, DOI: 10.1007/jhep09(2025)197, https://doi.org/10.1007/jhep09(2025)197
2. J. Nishimura and H. Watanabe, "Quantum decoherence from complex saddle points", Phys. Rev. Lett. 134 (2025) 210401, arXiv:2408.16627, DOI: 10.1103/PhysRevLett.134.210401, https://doi.org/10.1103/PhysRevLett.134.210401
3. A. O. Caldeira and A. J. Leggett, "Path integral approach to quantum Brownian motion", Physica A 121 (1983) 587, DOI: 10.1016/0378-4371(83)90013-4, https://doi.org/10.1016/0378-4371(83)90013-4
4. B. L. Hu, J. P. Paz and Y. Zhang, "Quantum Brownian motion in a general environment: Exact master equation with nonlocal dissipation and colored noise", Phys. Rev. D 45 (1992) 2843, DOI: 10.1103/PhysRevD.45.2843, https://doi.org/10.1103/PhysRevD.45.2843
5. W. H. Zurek, "Environment-induced superselection rules", Phys. Rev. D 26 (1982) 1862, DOI: 10.1103/PhysRevD.26.1862, https://doi.org/10.1103/PhysRevD.26.1862
6. W. H. Zurek, S. Habib and J. P. Paz, "Coherent states via decoherence", Phys. Rev. Lett. 70 (1993) 1187, DOI: 10.1103/PhysRevLett.70.1187, https://doi.org/10.1103/PhysRevLett.70.1187
7. J. Nishimura, K. Sakai and A. Yosprakob, "A new picture of quantum tunneling in the real-time path integral from Lefschetz thimble calculations", JHEP 09 (2023) 110, arXiv:2307.11199, DOI: 10.1007/jhep09(2023)110, https://doi.org/10.1007/jhep09(2023)110
8. G. Wang, S. Yang and Z. Cai, "Solving Caldeira-Leggett Model by Inchworm Method with Frozen Gaussian Approximation", Quantum 9 (2025) 1667, arXiv:2408.01039, DOI: 10.22331/q-2025-03-24-1667, https://doi.org/10.22331/q-2025-03-24-1667
9. Y. Sun, G. Wang and Z. Cai, "Iterative Methods for Discrete Dyson Series and Applications in Open Quantum Systems", J. Chem. Theory Comput. 22 (2026) 8071, DOI: 10.1021/acs.jctc.6c00518, https://doi.org/10.1021/acs.jctc.6c00518
10. N. Huggett, C. Käding, M. Pitschmann and J. Read, "Directly computing Wigner functions for open quantum systems", arXiv:2512.01590, https://arxiv.org/abs/2512.01590
11. S. Bhattacharjee, K. Mandal and S. Sinha, "Decoherence of a dissipative Brownian charged magneto-anharmonic oscillator: an information theoretic approach", arXiv:2510.19317, https://arxiv.org/abs/2510.19317
12. A. L. Gadelha, D. L. Nedel and C. R. Senise, "General Unitary SU(1, 1) Formulation of TFD and its Relation to Tomita-Takesaki Theory", Braz. J. Phys. 56 (2026) 191, DOI: 10.1007/s13538-026-02117-4, https://doi.org/10.1007/s13538-026-02117-4
13. Y. Liu, A. Kulkarni and S. Ryu, "Dissipative Dynamics and Symmetry Breaking in Bosonic Sachdev-Ye-Kitaev Lindbladian", arXiv:2508.04802, https://arxiv.org/abs/2508.04802