Molecular dynamics with time dependent quantum Monte Carlo

Ivan P. Christov

REVIEW 3 major objections 4 minor 2 references

Molecular dynamics with time dependent quantum Monte Carlo

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper argues that molecular dynamics can be solved ab initio by time-dependent quantum Monte Carlo, with both electrons and nuclei represented by guided walker ensembles, and demonstrates about 5 percent agreement with exact…

desk verdict A real extension of TDQMC to nuclei, but the effective-potential step is not a derivation—'ab initio' is unsupported and needs reframing. read the letter →

arxiv 2501.16132 v1 pith:YLNLLCSH submitted 2025-01-27 physics.atom-ph quant-ph

Ivan P. Christov This is my paper · ORCID

classification physics.atom-phquant-ph PACS 31.25.-v02.70.Ss

keywords time-dependentquantumMonteCarlomoleculardynamicsBohmiantrajectoriesguidingwavescorrelationsstrong-fieldionizationhydrogenmoleculeEhrenfest

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes an ab initio molecular-dynamics method, time-dependent quantum Monte Carlo (TDQMC), that treats both electrons and nuclei quantum mechanically. Each physical particle is replaced by an ensemble of Monte Carlo walkers and guiding waves in physical space, with the walkers propagated by de Broglie-Bohm velocities and the guiding waves by coupled linear Schrödinger equations. Nonlocal quantum correlations are introduced through effective potentials obtained by adaptive kernel smoothing of the walker densities, avoiding both quantum potentials and explicit exchange-correlation integrals. Applied to one- and two-dimensional $\mathrm{H}_2$ in intense ultrashort pulses, TDQMC gives ionization and internuclear-distance evolution within about 5 percent of exact Ehrenfest molecular dynamics, while time-dependent Hartree-Fock deviates by about 20 percent and does not dissociate the molecule.

What carries the argument

The central object is the walker–guiding-wave pair. For each electron $i$ and nucleus $I$, an ensemble of $M$ walkers with positions $\mathbf{r}_i^{(k)}$ and $\mathbf{R}_I^{(k)}$ moves under the de Broglie-Bohm guiding equations (8)–(9) and a random drift, while separate guiding waves $\phi_i^{(k)}$ and $\Phi_I^{(k)}$ satisfy the coupled linear Schrödinger equations (14)–(15). The exact densities in the Hartree equations are replaced by kernel density estimates (12)–(13), and the resulting effective potentials (16)–(19) encode nonlocal correlations with adaptive bandwidths $\sigma$ and $\Sigma$ set by formulas (22)–(23). The key move is that no quantum potential is computed and correlated interactions are reduced to Monte Carlo sums over walkers, which replaces the costly integrals of multi-configuration methods.

What would settle it

A decisive test would be to solve the same one-dimensional two-electron $\mathrm{H}_2$ model, with the same softened Coulomb potentials and the same 335 nm pulse at $9\times10^{14}$ W/cm$^2$, by a numerically exact time-dependent Schrödinger calculation and compare the survival probability and internuclear distance over the full pulse; if the exact curves differ from TDQMC by much more than the claimed 5 percent, the effective-potential approximation has failed. A second check is convergence in the number of walkers: if increasing $M$ systematically changes the ionization dynamics by more than the statistical error, the result is not converged and the ab initio claim is not met.

Watch

Extended reading notes

Core claim

The paper's central claim is that a fully quantum treatment of molecular dynamics is possible with TDQMC, in which both electrons and nuclei are described on equal footing by ensembles of walkers and guiding waves in physical space. The guiding waves obey coupled linear Schrödinger equations, and the walkers move by the de Broglie-Bohm guidance equations plus a random diffusive drift, so that the walker density reproduces the time-dependent probability density of each particle. Quantum correlations, including nonlocal effects, are encoded through effective potentials built from adaptive kernel density estimates of the walker distributions, with no quantum potentials and no overlap, exchange, or correlation integrals. The author shows for a one-dimensional hydrogen molecule in a strong 335 nm pulse that the TDQMC ionization and internuclear-distance evolution stay within about 5 percent of the exact Ehrenfest molecular dynamics benchmark, whereas time-dependent Hartree-Fock underestimates ionization by about 20 percent and does not dissociate the molecule.

Load-bearing premise

Everything rests on the assumption that replacing the true electron and nuclear densities in the Hartree equations by broadened, walker-sampled densities with adaptive widths faithfully encodes the quantum correlations that the simple product wavefunction leaves out; if that smoothing distorts the correlations, the claimed agreement with exact Ehrenfest dynamics would not be guaranteed and the ab initio claim would fail.

Editorial extensions

If this is right

Correlated electron–nuclear dynamics can be propagated in physical space without solving the full many-body Schrödinger equation, using explicit Coulomb potentials and no exchange-correlation integrals.
The quantum-classical limit gives a Newtonian nuclear trajectory bundle with self-consistent back-reaction, recovering previous mixed quantum-classical methods and the Ehrenfest mean-field limit as special cases.
For the one-dimensional $\mathrm{H}_2$ benchmark, TDQMC predicts final ionization only about 5 percent above exact Ehrenfest, whereas time-dependent Hartree-Fock is about 20 percent below and fails to dissociate the molecule.
The nonlocal correlation lengths $\sigma$ and $\Sigma$ grow as the molecule ionizes, providing a dynamical diagnostic of how electron–electron and electron–nuclear correlations spread during dissociation.
The two-dimensional orientation simulation indicates the method can follow molecular rotation and dissociation in an external field, not just fixed-nucleus electron dynamics.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: the same effective-potential construction could be tested on exactly solvable two-electron models with known correlation energies, such as a Hooke's-law atom, to isolate what the adaptive kernel smoothing adds beyond mean-field theory.
Editorial inference: because the method avoids overlap and exchange integrals, its scaling in particle number is likely its main advantage; a quantitative complexity study for three or more electrons would show whether the approach is practical beyond the two-electron molecule demonstrated here.
Editorial inference: the 5 percent figure is established for a softened one-dimensional model; extension to three dimensions would need to show that the adaptive bandwidth formulas remain stable when the Coulomb potentials are not softened and the walker density varies in all directions.
Editorial inference: the quantum-quantum and quantum-classical variants suggest a natural place to add trajectory surface hopping or ring-polymer corrections to restore nuclear quantum effects beyond the de Broglie-Bohm guidance, though the paper does not pursue this.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Read this one with a red pen on Section 2. The genuinely new piece is the quantum-quantum extension of your earlier TDQMC work: coupled ensembles of nuclear and electronic walkers, with guiding waves for both, plus the classical-nucleus limit. The 1D H2 test is sensible, and the ground-state energy (-1.661 a.u.) sits close to the direct Ehrenfest number (-1.665), with the time-dependent ionization within 5%. That part is fine, and the limiting-case discussion around Eqs. (24)-(25) is clear and honest.

The problem is the derivation of the effective potentials, Eqs. (16)-(19). The paper says they come from substituting the smoothed walker densities (12)-(13) into the Hartree equations (5)-(6). That is not what substitution gives. You would get a convolution of the interaction potential with the Gaussian mixture; what the paper writes is the bare potential evaluated at the walker separation, multiplied by a Gaussian in that same separation. These are different expressions, and no limit, quadrature, or other controlled approximation is offered to connect them. Since those potentials are the sole place where correlations beyond Hartree enter, the 'ab initio' claim does not follow. The benchmark itself is approximate—Ehrenfest MD keeps nuclei classical—so agreement with it can't rescue the derivation. And there are no convergence tests in the number of walkers, grid spacing, or time step, no code, and the soft-core parameters and initial widths are chosen without a sensitivity study.

That is a load-bearing flaw, not a cosmetic one. But it's a fixable framing issue as much as a mathematical one. If the author stops calling the effective potentials a consequence of the Hartree equations and instead presents them as an independent approximation—a trajectory-based, mean-field-with-nonlocal-corrections scheme—then the 5% agreement becomes a plausible demonstration of the heuristic's behavior, not a proof of ab initio status. A serious referee should ask for that reframing, plus convergence data and at least one stronger benchmark.

For whom: strong-field/attosecond chemists interested in trajectory-based methods. This deserves peer review rather than desk rejection—the extension is real, and the numerical comparison is informative—but it needs major revision before publication.

Referee Report

3 major / 4 minor

Summary. The paper proposes a time-dependent quantum Monte Carlo (TDQMC) method for molecular dynamics in which both electrons and nuclei are represented by ensembles of Bohmian walkers and guiding waves. The formalism is derived from a single Hartree product ansatz by replacing the electronic and nuclear densities in the Hartree equations with kernel-smoothed walker densities, yielding coupled Schr\"odinger equations for the guiding waves and effective nonlocal potentials. The method is applied to a one-dimensional model of H2 in an intense laser pulse, with comparisons to "exact" Ehrenfest molecular dynamics, time-dependent Hartree-Fock, and ground-state energy estimates; a two-dimensional orientation-dynamics example is also presented. The central claim is that TDQMC is an ab initio method that captures local and nonlocal quantum correlations self-consistently.

Significance. If the derivation were rigorous and the numerical evidence compelling, the method could be a significant step toward scalable correlated molecular dynamics beyond mean-field theory, with potential impact on attosecond and strong-field physics. The manuscript does present a concrete algorithm and reports agreement with an established approximate benchmark, but the central derivation is not mathematically valid, and the numerical validation is insufficient to support the "ab initio" claim. The paper contains no convergence studies, no error bars, and no independent confirmation of the effective-potential approximation, so the current contribution does not meet the standard for a methods paper in this field.

major comments (3)

[Section 2, Eqs. (12)-(19)] The claim that substituting the smoothed walker densities (12)-(13) into the Hartree equations (5)-(6) yields the effective potentials (16)-(19) is not correct. For example, inserting (12) into the electron-electron term of (6) gives a convolution of the bare potential against the Gaussian kernel, namely \int dr_j V_ee(r_i-r_j) \sum_l \exp(-(r_j-r_j^l)^2/\sigma^2)/z_j, whereas Eq. (16) evaluates the potential directly at the walker separation and multiplies by a Gaussian in that same separation. These expressions differ for any finite bandwidth, and no limit, quadrature, or other controlled approximation is provided to connect them. Since this substitution is the only mechanism by which correlations beyond the single Hartree product are introduced, the paper's central claim that the method is ab initio is unsupported.
[Section 5, benchmark comparison] The comparison benchmark, labeled "exact" Ehrenfest molecular dynamics, is not an exact solution of the full many-body Schr\"odinger equation: it treats the nuclei as classical particles and neglects nuclear quantum correlations. Agreement to within 5% for the ionization and internuclear distance therefore cannot validate the claim that TDQMC captures local and nonlocal quantum correlations for both electrons and nuclei self-consistently. The manuscript also reports no convergence study with respect to the number of walkers M or the kernel bandwidths, so it is unclear whether the observed agreement is robust or fortuitous.
[Section 2, Eqs. (22)-(23)] The adaptive bandwidths \sigma_j^k and \Sigma_J^k are introduced using the standard density-estimation formulas of Abramson and Silverman, but they are not derived from the Schr\"odinger dynamics. They therefore constitute an uncontrolled approximation inside the effective potentials, and the paper provides no evidence that the results are insensitive to the choice of pilot bandwidths \sigma and \Sigma or to the initial wavefunction widths \sigma_e and \Sigma_n. This issue is load-bearing because the effective potentials are the sole device for modeling nonlocal correlations, and the absence of a derivation or sensitivity analysis undermines the ab initio claim.

minor comments (4)

[Abstract and Section 5] There are several typographical errors, including "nuc lear" in the abstract, "ultashort" in the abstract, and "Metopolis" in Section 5 (should be "Metropolis").
[Section 2, Eq. (7)] The notation in Eq. (7) refers to an antisymmetrized product for electrons, but the later discussion of H2 in Section 5 mentions a symmetrized product for the two electronic walkers; the treatment of spin statistics in the one-dimensional model should be clarified.
[Figure captions] The captions for Figures 2-5 do not specify the color/line conventions in a way that is self-explanatory without the main text, and Figure 1's caption lacks a description of the pulse envelope that is supposedly shown.
[Section 4] The energy expression in Eq. (28) requires an estimate of the many-body quantum potential using adaptive kernel density estimation, but no specific algorithm or bandwidth choice is given for the multivariate case, making the reported ground-state energy difficult to reproduce.

Circularity Check

1 steps flagged · score 4.0 of 10

Gaussian-kernel effective potentials are adopted from the author's own prior work (ref. 26), not derived from the Hartree equations; external benchmarks give independent content, so the circularity is partial.

ansatz smuggled in via citation [Section 2, Eq. (12) and Eqs. (16)-(19)]
"One simple and efficient way to account for these quantum effects is to formally represent the particle densities in Eq. (5) and Eq. (6) by smoothed interpolation with e.g. Gaussian kernels that are centered at the positions of the Bohmian walkers (kernel density estimation) 26. ... Substituting Eq. (12) and Eq. (13) into Eq. (5) and Eq. (6), and assigning a separate guiding wave to each Bohmian walker, we transform the non-linear Hartree equations (5) and (6) to a set of coupled linear Schrödinger equations for the guiding waves."

The Gaussian-kernel smoothing is the load-bearing device that is said to account for the quantum correlations neglected by the Hartree product, and it is taken directly from the author's own ref. 26. The claimed derivation is that substituting (12)-(13) into (5)-(6) yields the effective potentials (16)-(19). But substitution would produce convolution integrals of the bare Coulomb potential against the smoothed densities, whereas (16)-(19) evaluate the bare potential at each walker separation and multiply it by a Gaussian in that separation; no limit or quadrature connects the two expressions.

full rationale

The paper's numerical results are compared with an independent Ehrenfest benchmark and with TDHF, and the reported ground-state energy (-1.661 a.u.) is not obtained by fitting a parameter to the target (-1.665 a.u.), so there is no fitted-input-called-prediction circularity. The main circularity concern is in Section 2: the effective potentials (16)-(19), which are the only mechanism for including local and nonlocal correlations beyond the single Hartree product, are introduced by invoking the author's own prior TDQMC work (ref. 26) for Gaussian kernel density estimation. The text claims they follow by substituting (12)-(13) into (5)-(6), but the substitution alone would give convolution integrals of the bare potentials with the smoothed densities, whereas the actual formulas multiply the bare potential at each walker separation by a Gaussian. This indicates that the correlation device is an ansatz adopted from a self-citation rather than derived from the stated equations. Because the Ehrenfest and TDHF comparisons provide independent numerical content for the specific predictions, the circularity is partial rather than complete; the score reflects that the method's foundational derivation is self-referential, while the benchmarked dynamics still carry independent information.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a chain of approximations: Hartree factorization, Bohmian guidance, kernel density smoothing, adaptive bandwidths, and stochastic thermalization. The soft-Coulomb and initial-width parameters are hand-chosen. No convergence proof or error estimate is provided. No new physical entities are introduced; walkers are computational devices.

free parameters (3)

soft-core Coulomb parameters a, b, c = a=1 a.u., b=1.5 a.u., c=0.5 a.u.
Hand-chosen in Eqs (31)-(33) to define the 1D model Hamiltonian; the results depend on these values.
initial guiding-wave widths sigma_e, Sigma_n = sigma_e=2 a.u., Sigma_n=0.5 a.u.
Chosen as initial conditions for the electronic and nuclear guiding waves in Section 5; not derived from the target system.
pilot constant bandwidths sigma, Sigma for adaptive kernel estimates = not specified numerically
Enter Eqs (22)-(23) through pilot density estimates; the paper does not state the constant bandwidth values used.

assumptions (6)

domain assumption Single Hartree product factorization of the total wavefunction (Eq. 4).
The paper starts from a Hartree ansatz and claims correlations are restored later via walker ensembles; this ansatz omits exchange and static correlation.
standard math Bohmian guidance equations (8)-(9) give the correct walker velocities.
Taken from de Broglie-Bohm mechanics (ref 32); standard but not derived in this paper.
ad hoc to paper Kernel-smoothed walker densities in Eqs (12)-(13) approximate the true particle densities.
The raw walker density is smoothed with Gaussians; no convergence proof links it to the exact many-body density.
ad hoc to paper Effective potentials in Eqs (16)-(19) with adaptive bandwidths represent nonlocal quantum correlations.
This is the core modeling assumption that replaces the neglected correlation terms; it is not derived from the many-body Schrodinger equation.
ad hoc to paper The random drift in Eqs (10)-(11) thermalizes walkers to |phi|^2 without biasing the dynamics.
The drift is introduced to enforce equilibrium distributions; its variance is not specified and its effect on dynamics is not analyzed.
domain assumption The 1D soft-Coulomb model with a,b,c is a valid test of molecular ionization and dissociation.
The paper uses modified 1D Coulomb potentials, which are standard models but not quantitative for real H2.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Molecular dynamics with time dependent quantum Monte Carlo." pith.science (2026). https://pith.science/paper/YLNLLCSH

@misc{pith2026250116132,
  author       = {Pith},
  title        = {Pith review of: Molecular dynamics with time dependent quantum Monte Carlo},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YLNLLCSH}},
  note         = {Machine review of arXiv:2501.16132}
}

read the original abstract

In this paper we propose an ab initio method to solve quantum many-body problems of molecular dynamics where both the electronic and the nuclear degrees are represented by ensembles of trajectories and guiding waves in physical space. Both electrons and nuclei can be treated quantum mechanically where the guiding waves obey a set of coupled Schrodinger equations (quantum-quantum description) or, alternatively, coupled Schroedinger-Newtonian equations are solved for the quantum-classical approximation. The method takes into account local and non-local quantum correlation effects in a self consistent manner. The general formalism is applied to one- and two-dimensional hydrogen molecule subjected to a strong ultashort optical pulse. Comparison is made with the results from the 'exact' Ehrenfest molecular dynamics for the molecular ionization and for the evolution of the inter-nuclear distance as the molecule dissociates.

Figures

Figures reproduced from arXiv: 2501.16132 by the authors.

**Figure 2.** (Color online) Time dependent ionization for one-dimensional hydrogen molecule in an external optical pulse with carrier frequency ω=0.137 a.u. (335 nm) and peak intensity 9 1014 W/cm2 . Black solid line- exact Ehrenfest molecular dynamics, gray (blue) line- quantum-classical TDQMC result, light-gray (red) line- quantum-quantum TDQMC result, dashed line- TDHF [PITH_FULL_IMAGE:figures/full_fig_p027_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Kapteyn, O

1 H. Kapteyn, O. Cohen, I. Christov, and M. Murnane, Science 317, 775 (2007). 2 E. Goulielmakis, V. S. Yakovlev, A. L. Cavalieri, M. Uiberacker, V. Pervak, A

work page 2007
[2]

Kienberger, U

Apolonski, R. Kienberger, U. Kleineberg, and F. Krausz, Science 317, 769 (2007). 3 K. C. Kulander, Phys. Rev. A 36, 2726 (1987). 4 S. Chelkowski, T. Zuo, O. Atabek, and A. D. Bandrauk, Phys. Rev. A 52, 2977 (1995). 5 E. Runge and E. K. U. Gross, Phys. Rev. Lett. 52, 997 (1984). 6 F. Furche and K. Burke, in Annual Reports in Computational Chemistry, edited...

work page 2007

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Kapteyn, O

1 H. Kapteyn, O. Cohen, I. Christov, and M. Murnane, Science 317, 775 (2007). 2 E. Goulielmakis, V. S. Yakovlev, A. L. Cavalieri, M. Uiberacker, V. Pervak, A

work page 2007

[2] [2]

Kienberger, U

Apolonski, R. Kienberger, U. Kleineberg, and F. Krausz, Science 317, 769 (2007). 3 K. C. Kulander, Phys. Rev. A 36, 2726 (1987). 4 S. Chelkowski, T. Zuo, O. Atabek, and A. D. Bandrauk, Phys. Rev. A 52, 2977 (1995). 5 E. Runge and E. K. U. Gross, Phys. Rev. Lett. 52, 997 (1984). 6 F. Furche and K. Burke, in Annual Reports in Computational Chemistry, edited...

work page 2007