Variational Gaussian Wave-Packet Dynamics from Constrained Classical Trajectory Bundles

Taisuke Hasegawa

REVIEW 2 major objections 3 minor 29 references

Variational Gaussian Wave-Packet Dynamics from Constrained Classical Trajectory Bundles

T0 review · 2 major / 3 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper claims that variational Gaussian wave-packet dynamics is exactly the Gaussian large-N limit of a constrained bundle of classical trajectories, with the area scale calibrated to ℏ/2.

desk verdict Conditional equivalence, honestly flagged: the finite-N constrained-bundle construction is new and the formal limit is clean, but the Gaussian large-N limit is assumed rather than derived. read the letter →

arxiv 2607.07734 v3 pith:3MZZDGQ4 submitted 2026-07-07 physics.class-ph physics.chem-ph

Taisuke Hasegawa This is my paper

classification physics.class-phphysics.chem-ph

keywords Gaussianwave-packetdynamicstime-dependentvariationalprincipleclassicaltrajectorybundlesymplecticareaconstraintWignerdensitylarge-Nlimitconstrainedmolecularanharmonicsystems

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

Gaussian wave-packet dynamics is normally derived by restricting quantum evolution to a Gaussian ansatz. This paper proposes a different route: take a finite labeled bundle of classical trajectories and impose a fixed symplectic-area scale on its covariance. At finite N the bundle is the dynamical state, can be non-Gaussian, and feels the full potential without a local harmonic approximation. As N→∞, for sequences whose initial empirical distributions approach a smooth Gaussian, the residual nonlinear force drops out, the relative motion becomes linear, and with the area scale set to ℏ/2 the limiting centroid and width equations coincide with the Gaussian time-dependent variational principle; the limiting phase-space density is the corresponding Gaussian Wigner density. The paper thereby identifies the Gaussian TDVP as the Gaussian large-N limit of constrained classical-bundle dynamics, with N counting trajectories rather than orders of an ℏ expansion.

What carries the argument

The central object is the phase-space covariance Σ of a labeled bundle of N classical trajectories, constrained by the symplectic-area condition ΣJΣ = κJ. In centered variables this is the pair of conditions A = A^T (symmetric position-momentum correlation) and Π = κX^{-1} (residual momentum covariance inversely tied to position covariance). Preservation is enforced by virtual-work constraint forces Λu_i + ΩX^{-1}x_i, where u_i is the residual force beyond the linear force-displacement fit and Λ, Ω are multipliers chosen to keep R = 0 and A symmetric. In the Gaussian limit the residual force vanishes, the relative dynamics reduces to ẋ = p, ṗ = M_G x with M_G = -⟨∇²V⟩_G, and the complex widt

What would settle it

For a smooth anharmonic potential, prepare a sequence of exactly constrained initial bundles whose empirical distributions converge to the same Gaussian and integrate the finite-N equations; if the realization-averaged centroid and width do not approach the Gaussian TDVP reference as N→∞, or if the multiplier equations become singular for some initial data, the claimed large-N limit fails.

Watch

Extended reading notes

Core claim

The central claim is that the Gaussian time-dependent variational principle — the phase-space dynamics obtained by restricting quantum evolution to Gaussian wave packets — can be obtained without any wavefunction by constraining a bundle of classical trajectories. The constraint is the fixed symplectic-area relation ΣJΣ = κJ on the bundle covariance, which in centered variables is equivalent to the residual-covariance condition Π = κX^{-1} and the symmetry A = A^T. Virtual-work constraint forces with multipliers Λ and Ω preserve these conditions. In the Gaussian large-N limit the force-displacement matrix becomes the negative Gaussian-averaged Hessian, the residual force vanishes, the relati

Load-bearing premise

The load-bearing assumption is that the finite-N constrained equations are well-posed: for initial data satisfying the fixed-area conditions, multiplier matrices Λ and Ω must exist and preserve the constraints along the flow; the paper defines these forces but does not prove existence or uniqueness.

Editorial extensions

If this is right

Gaussian TDVP dynamics, including the full Gaussian Wigner density, can be produced by constrained classical trajectories without a wavefunction or quantum potential.
At finite N the method is a genuine trajectory-bundle state: non-Gaussian, coupled by collective constraint forces, with dynamics driven by the full anharmonic potential; for quadratic potentials the constraint force vanishes and the dynamics is unchanged.
The limit parameter N is a trajectory count, not an expansion order in ℏ; ℏ enters only through the calibration √κ = ℏ/2, which sets the covariance to saturate the uncertainty principle.
In the Gaussian limit the residual nonlinear force is removed by the constraint, making the relative motion linear and guaranteeing that Gaussianity is preserved at all later times.
Numerical tests on a 20-dimensional coupled anharmonic model show systematic finite-N effects of the constraint and convergence of selected observables toward the Gaussian TDVP reference as N increases.

Reading between the lines

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

A testable corollary the paper leaves open is whether the same Gaussian limit is obtained from other constraint implementations that preserve the same covariance condition; the virtual-work construction is one possible force choice, not shown to be unique.
Since the finite-N bundle retains higher-order statistics beyond its covariance, one could use it to explore whether finite-N departures carry systematic information about anharmonic quantum corrections; the paper explicitly does not claim such corrections.
The derivation suggests a route to finite-temperature or mixed-state dynamics by generalizing the fixed-area covariance condition beyond the pure-Gaussian Wigner form, though the paper does not pursue this.
The well-posedness of the finite-N multiplier equations is the main hidden assumption; probing numerically whether the consistency conditions dot R = 0 and dot A - dot A^T = 0 have solutions for generic non-Gaussian initial data would clarify the practical scope.

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

Two things to know. First, the finite-N constrained trajectory-bundle dynamics is a real conceptual novelty: a labeled bundle of classical trajectories with a fixed symplectic-area covariance, no wave function, and no local harmonic approximation. Second, the paper does not prove that Gaussian TDVP emerges as a true large-N limit—it assumes the limit of empirical averages and then derives the limiting equations. That conditional result is presented honestly, and the derivation is clean.

What's genuinely new: generalizing the 1D constant-uncertainty condition from CUMD to arbitrary d, imposing the full matrix condition ΣJΣ=κJ, and showing that in the Gaussian large-N limit with κ=ℏ²/4 the width equations combine into Ż=M_G−Z², the standard Gaussian TDVP form, with the centroid and full phase-space density matching the Wigner function. The Gaussian integration by parts and residual-force cancellation are coherent, and the authors are explicit that they do not establish general convergence of the finite-N dynamics. Credit where due: this is an honest, well-structured derivation of a conditional equivalence.

The load-bearing gap is the limit itself. The text says "empirical averages appearing below are assumed to converge to the corresponding Gaussian averages," then derives the Gaussian flow and shows Gaussianity is preserved. That is a Gaussian-ansatz consistency check, not a proof that the finite-N bundles converge to that flow. Nothing controls fluctuations; the multipliers Λ_N and Ω_N are N-dependent and their behavior as N→∞ is not analyzed. The stress-test note has this right: the claimed equivalence lives entirely in that limit, so assuming it is not a minor technicality.

The reader's well-posedness concern is also real. The finite-N dynamics is defined through consistency conditions ˙R=0 and ˙A−Ȧ^T=0, but no existence or uniqueness theorem for the multipliers is given. If those equations become singular, the finite-N system itself is undefined, and the N→∞ limit with it.

The numerics are limited but not damning: no code or data, no exact-quantum baseline, and the supplemental material is unavailable. The comparison to the TDVP reference is suggestive, but it doesn't test the convergence assumption, only behavior of two selected observables.

Proportionately: these are addressable gaps. The formal part is a legitimate conditional statement. To make the central claim a full derivation, the paper needs either a propagation-of-chaos or tightness argument, or an explicit reframing as "conditional on convergence, the limit is Gaussian TDVP." Well-posedness of the multiplier equations should be proved or at least argued with concrete conditions.

Who this is for: researchers working on trajectory-based quantum/semiclassical dynamics, Gaussian wave packets, or constant-uncertainty MD. It deserves a serious referee, and I would send it to peer review with a request for revision that addresses the convergence and well-posedness gaps and makes the numerics reproducible.

Referee Report

2 major / 3 minor

Summary. The manuscript constructs a finite-N labeled bundle of classical trajectories governed by a common Hamiltonian, subject to a collective constraint that fixes the symplectic-area scale of the bundle covariance (Sigma J Sigma = kappa J). Constraint forces are derived via virtual-work arguments to preserve the equivalent conditions A = A^T and Pi = kappa X^{-1}. In the Gaussian large-N limit, where initial empirical distributions are assumed to converge to a smooth Gaussian density and empirical averages are assumed to converge to Gaussian averages, the limiting relative motion becomes linear with force matrix M_G = -<nabla^2 V>_G, the residual force cancels, and the width equations assemble into dZ/dt = M_G - Z^2 with Z = A + i sqrt(kappa) X^{-1}. Setting kappa = hbar^2/4 reproduces the centroid and width equations of the Gaussian time-dependent variational principle, and the limiting phase-space density is identified with the corresponding Gaussian Wigner density. Numerical tests on a 20-dimensional coupled Morse model compare constrained vs. unconstrained dynamics and finite-N behavior against this Gaussian limit.

Significance. If the claimed correspondence were established as a genuine large-N convergence theorem, it would provide a new, conceptually clean trajectory-bundle interpretation of Gaussian TDVP that is distinct from Moyal/hbar expansion and from Bohmian or locally-quadratic trajectory methods. The algebraic derivation of the limiting equations is coherent, and the paper is honest about its main limitation: the convergence of the finite-N dynamics to the Gaussian flow is assumed, not proved. The construction is elegant but somewhat tailored to the answer, since the fixed-area condition at kappa = hbar^2/4 is chosen to match the covariance algebra of pure Gaussian Wigner functions. The paper ships explicit equations, a self-contained appendix, and numerical evidence of consistency. Its significance would be substantially higher if the convergence assumption were either proved or precisely qualified in the central theorem statement.

major comments (2)

[Gaussian large-N limit (Eqs. (21)-(22))] The central equivalence is conditional on an assumed convergence of empirical averages to Gaussian averages. The finite-N system is closed through N-dependent multipliers Lambda_N and Omega_N, and no control of fluctuations or tightness is provided. The Discussion concedes that the calculations 'do not establish general convergence of the finite-N dynamics.' This is load-bearing for the claim that Gaussian TDVP is the Gaussian large-N limit. Please either supply a convergence proof or a precise propagation-of-chaos statement, or reformulate the main theorem as an equivalence of limiting equations under an explicit assumption, with finite-N convergence as an open question. The current abstract overstates the result.
[Eqs. (18)-(20), Appendix B] The finite-N constrained dynamics is defined only if multiplier matrices Lambda(t) and Omega(t) exist and are unique for the consistency conditions that R-dot = 0 and A-dot - A-dot^T = 0. Appendix B gives a formal virtual-work construction but no existence or uniqueness theorem. If the multiplier system is singular for generic initial data, the finite-N trajectory flow (and hence its N to infinity limit) is undefined. Please state and prove (or at least argue) well-posedness for initial data satisfying G=0, A=A^T, R=0, or identify a regularity condition that guarantees it.

minor comments (3)

[Numerical tests] Only ten initial bundles are used for each N; the shaded bands are one standard deviation over these ten realizations, which is a small sample. The finite-N trend in X_20,20(t) is suggestive but not a quantitative convergence test. Consider adding a convergence-rate analysis or bootstrapped confidence intervals.
[Notation] In Eq. (21), the arrow M_N -> M_G should explicitly indicate the limit N to infinity; the same notation is used in Appendix C. Also, the phrase 'In taking this limit...' appears in both the main text and Appendix C; consider defining the convergence assumption once in a named condition.
[Eq. (22)] The step from U_G (I + Lambda_G)^T = 0 to (I + Lambda_G) u_G = 0 for rank-deficient U_G is justified only informally. A cleaner argument is that the covariance of (I + Lambda_G) u_G equals (I + Lambda_G) U_G (I + Lambda_G)^T = 0, so the force vanishes almost surely. The authors may wish to include this one-line justification.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the TDVP equivalence is derived from the constrained-bundle equations, not assumed; the unproven Gaussian-limit convergence is an acknowledged rigor gap, not a circular reduction.

full rationale

The derivation does not fit or assume the target TDVP equations. The finite-N constrained dynamics is defined first (Eqs. 1-20) with the fixed-area condition Eq. (4), and the constraint forces are constructed in Appendix B from virtual work, not from TDVP. The limiting section then explicitly assumes that empirical averages converge to Gaussian averages ('In taking this limit, the empirical averages appearing below are assumed to converge to the corresponding Gaussian averages'), and from that hypothesis derives the Gaussian force matrix (Eq. 21), the vanishing residual-force condition (Eq. 22), the linear relative flow (Eq. 23), the width equations (Eqs. 24-26), and the centroid equations (Eq. 27). The coincidence with Gaussian TDVP after setting κ=ℏ²/4 is therefore a derived equivalence under a stated convergence hypothesis, not a circular use of TDVP. The calibration κ=ℏ²/4 is an explicit parameter choice motivated by the Gaussian Wigner covariance algebra, and the paper's Discussion is transparent that the calculations 'do not establish general convergence of the finite-N dynamics.' That missing convergence proof is a correctness/rigor limitation, not a circular step: no equation in the chain reduces to an input by construction. The only self-citations (refs. 28-29) appear in the Discussion as a 'recovers' remark and are not load-bearing. Hence score 0.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The construction rests on standard Gaussian/symplectic identities plus three domain assumptions: Gaussian convergence of the empirical distribution, well-posedness of the multiplier-enforced constraints, and identification with literature TDVP equations. The scale κ is set by hand to ℏ/2; this choice is the main external input.

free parameters (1)

Symplectic-area scale κ = ℏ²/4 (√κ=ℏ/2)
The fixed-area condition ΣJΣ=κJ contains an arbitrary positive scale κ. The central equivalence with Gaussian TDVP holds only after setting κ=ℏ²/4, chosen by hand to match quantum uncertainty; not fitted to data but not derived from within the classical-bundle construction.

assumptions (6)

standard math Nondegenerate positive-definite phase-space covariance Σ and Williamson symplectic normal form apply.
The factorization and equivalence ΣJΣ=κJ ⇔ A=A^T, Π=κX^{-1} in Eq. (5) and Appendix A rely on X>0 and Williamson's theorem.
standard math Gaussian integration by parts identity ⟨f x^T⟩_G = -⟨∇²V⟩_G X.
Used in Eq. (21)/Appendix C to replace M_N with the Gaussian-averaged Hessian; valid for smooth potentials and nondegenerate Gaussian averages.
domain assumption Initial empirical distributions converge to a smooth Gaussian and empirical averages converge to Gaussian averages as N→∞.
Stated at the start of the Gaussian large-N limit section and Appendix C; the central theorem is conditional on this convergence.
ad hoc to paper The residual-force correlation condition R=0 is the correct compatibility condition and is preserved by the constraint forces.
R=0 is introduced from preservation of G without assuming a commutation relation (Eqs. 10-13). It is an additional dynamical closure beyond the fixed-area constraint; its necessity is argued, not proven.
domain assumption The multiplier matrices Λ and Ω exist and are unique enough to define the finite-N flow.
Eqs. (18)-(20) and Appendix B define constraint forces via consistency conditions \dot{R}=0 and \dot{A}-\dot{A}^T=0, but no existence/uniqueness theorem is given. Numerical solution is asserted.
domain assumption The referenced Gaussian TDVP width/centroid equations and pure-Gaussian Wigner covariance conditions are correct and match conventions.
The coincidence in Eq. (26)-(27) is asserted by citing refs [1-5,20-23]; the paper does not re-derive TDVP.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Variational Gaussian Wave-Packet Dynamics from Constrained Classical Trajectory Bundles." pith.science (2026). https://pith.science/paper/3MZZDGQ4

@misc{pith2026260707734,
  author       = {Pith},
  title        = {Pith review of: Variational Gaussian Wave-Packet Dynamics from Constrained Classical Trajectory Bundles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3MZZDGQ4}},
  note         = {Machine review of arXiv:2607.07734}
}

abstract

Variational Gaussian wave-packet dynamics is usually obtained by restricting quantum evolution to Gaussian wave packets. We instead construct finite-$N$ dynamics for a labeled bundle of classical trajectories under a constraint fixing the symplectic-area scale of its covariance. At finite $N$, the bundle itself is the dynamical state: it may be non-Gaussian, and the potential force on each trajectory is evaluated from the full potential without a local harmonic approximation. For sequences whose initial empirical distributions approach a smooth Gaussian density as $N \to \infty$, the limiting relative motion becomes linear and preserves Gaussianity. With the area scale set to $\hbar/2$, the limiting centroid and width equations coincide with those of the Gaussian time-dependent variational principle, and the limiting phase-space density coincides with the corresponding Gaussian Wigner density for matched initial data. Here $N$ counts trajectories, not orders in a Moyal or $\hbar$ expansion. This phase-space correspondence reveals two complementary constructions of the same Gaussian dynamics: a variational reduction of quantum evolution and the Gaussian large-$N$ limit of a constrained classical trajectory bundle.

Figures

Figures reproduced from arXiv: 2607.07734 by the authors.

**Figure 2.** FIG. 2. Finite- [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 2.** FIG. 2. Finite- [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Review history (3 revisions) →

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references

[1]

A. D. McLachlan, A variational solution of the time- dependent schr¨ odinger equation, Molecular Physics8, 39 (1964)

1964
[2]

E. J. Heller, Time dependent variational approach to semiclassical dynamics, The Journal of Chemical Physics 64, 63 (1976)

1976
[3]

R. D. Coalson and M. Karplus, Multidimensional varia- tional Gaussian wave packet dynamics with application to photodissociation spectroscopy, The Journal of Chem- ical Physics93, 3919 (1990)

1990
[4]

J. J. L. Van ´ ıˇ cek, Family of Gaussian wavepacket dy- namics methods from the perspective of a nonlinear schr¨ odinger equation, The Journal of Chemical Physics 159, 014114 (2023)

2023
[5]

R. M. Fereidani and J. J. L. Van ´ ıˇ cek, High-order geo- metric integrators for the variational Gaussian approx- 6 imation, The Journal of Chemical Physics159, 094114 (2023)

2023
[6]

E. J. Heller, Time-dependent approach to semiclassical dynamics, The Journal of Chemical Physics62, 1544 (1975)

1975
[7]

D. V. Shalashilin and M. S. Child, Locally coupled co- herent states and Herman–Kluk dynamics, The Journal of Chemical Physics118, 2061 (2003)
[8]

E. J. Heller, Cellular dynamics: A new semiclassical ap- proach to time-dependent quantum mechanics, The Jour- nal of Chemical Physics94, 2723 (1991)

1991

Show all 29 references

[9]

M. F. Herman and E. Kluk, A semiclassical justification for the use of non-spreading wavepackets in dynamics calculations, Chemical Physics91, 27 (1984)

1984
[10]

S. V. Antipov, F. Kr¨ oninger, and J. J. L. Van ´ ıˇ cek, Refined approach to cellularization: Going from Heller’s thawed Gaussian approximation to Herman–Kluk’s initial value representation, The Journal of Chemical Physics163, 154107 (2025)

2025
[11]

Garashchuk and V

S. Garashchuk and V. A. Rassolov, Quantum dynamics with Bohmian trajectories: Energy conserving approxi- mation to the quantum potential, Chemical Physics Let- ters376, 358 (2003)

2003
[12]

V. A. Rassolov and S. Garashchuk, Bohmian dynamics on subspaces using linearized quantum force, The Journal of Chemical Physics120, 6815 (2004)

2004
[13]

Garashchuk and V

S. Garashchuk and V. Rassolov, Quantum trajectory dy- namics based on local approximations to the quantum potential and force, Journal of Chemical Theory and Computation15, 3906 (2019)

2019
[14]

A. K. Pattanayak and W. C. Schieve, Gaussian wave- packet dynamics: Semiquantal and semiclassical phase- space formalism, Physical Review E50, 3601 (1994)

1994
[15]

Shigeta, H

Y. Shigeta, H. Miyachi, and K. Hirao, Quantal cumu- lant dynamics: General theory, The Journal of Chemical Physics125, 244102 (2006)

2006
[16]

Goldstein,Classical Mechanics, 2nd ed

H. Goldstein,Classical Mechanics, 2nd ed. (Addison- Wesley, Reading, MA, 1980)

1980
[17]

Ryckaert, G

J.-P. Ryckaert, G. Ciccotti, and H. J. C. Berendsen, Nu- merical integration of the Cartesian equations of mo- tion of a system with constraints: Molecular dynamics of n-alkanes, Journal of Computational Physics23, 327 (1977)

1977
[18]

velocity

H. C. Andersen, RATTLE: A “velocity” version of the SHAKE algorithm for molecular dynamics calculations, Journal of Computational Physics52, 24 (1983)

1983
[19]

Williamson, On the algebraic problem concerning the normal forms of linear dynamical systems, American Journal of Mathematics58, 141 (1936)

J. Williamson, On the algebraic problem concerning the normal forms of linear dynamical systems, American Journal of Mathematics58, 141 (1936)

1936
[20]

Dutta, N

Arvind, B. Dutta, N. Mukunda, and R. Simon, The real symplectic groups in quantum mechanics and optics, Pra- mana45, 471 (1995)

1995
[21]

Ohsawa and C

T. Ohsawa and C. Tronci, Geometry and dynamics of Gaussian wave packets and their Wigner transforms, Journal of Mathematical Physics58, 092105 (2017)

2017
[22]

H. P. Robertson, The uncertainty principle, Physical Re- view34, 163 (1929)

1929
[23]

Schr¨ odinger, Zum Heisenbergschen Unsch¨ arfeprinzip, Sitzungsberichte der Preussischen Akademie der Wis- senschaften, Physikalisch-mathematische Klasse14, 296 (1930)

E. Schr¨ odinger, Zum Heisenbergschen Unsch¨ arfeprinzip, Sitzungsberichte der Preussischen Akademie der Wis- senschaften, Physikalisch-mathematische Klasse14, 296 (1930)

1930
[24]

See Supplemental Material at [URL will be inserted by publisher] for details of the 20-dimensional coupled Morse-oscillator model, the preparation of the initial bundles, the time-integration algorithm, and step-size tests
[25]

Kubo, Wigner representation of quantum operators and its applications to electrons in a magnetic field, Jour- nal of the Physical Society of Japan19, 2127 (1964)

R. Kubo, Wigner representation of quantum operators and its applications to electrons in a magnetic field, Jour- nal of the Physical Society of Japan19, 2127 (1964)

1964
[26]

H. J. Groenewold, On the principles of elementary quan- tum mechanics, Physica12, 405 (1946)

1946
[27]

J. E. Moyal, Quantum mechanics as a statistical theory, Mathematical Proceedings of the Cambridge Philosoph- ical Society45, 99 (1949)

1949
[28]

T. Hasegawa, Communication: Constant uncertainty molecular dynamics: A simple and efficient algorithm to incorporate quantum nature into a real-time molecular dynamics simulation, The Journal of Chemical Physics 145, 171101 (2016)

2016
[29]

T. Hasegawa, Nuclear quantum dynamics of three- dimensional condensed-phase systems by constant un- certainty molecular dynamics, The Journal of Physical Chemistry Letters14, 8043 (2023). APPENDIX Appendix A: Fixed-area condition in residual-momentum variables.—From the defini...

2023

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

A. D. McLachlan, A variational solution of the time- dependent schr¨ odinger equation, Molecular Physics8, 39 (1964)

1964

[2] [2]

E. J. Heller, Time dependent variational approach to semiclassical dynamics, The Journal of Chemical Physics 64, 63 (1976)

1976

[3] [3]

R. D. Coalson and M. Karplus, Multidimensional varia- tional Gaussian wave packet dynamics with application to photodissociation spectroscopy, The Journal of Chem- ical Physics93, 3919 (1990)

1990

[4] [4]

J. J. L. Van ´ ıˇ cek, Family of Gaussian wavepacket dy- namics methods from the perspective of a nonlinear schr¨ odinger equation, The Journal of Chemical Physics 159, 014114 (2023)

2023

[5] [5]

R. M. Fereidani and J. J. L. Van ´ ıˇ cek, High-order geo- metric integrators for the variational Gaussian approx- 6 imation, The Journal of Chemical Physics159, 094114 (2023)

2023

[6] [6]

E. J. Heller, Time-dependent approach to semiclassical dynamics, The Journal of Chemical Physics62, 1544 (1975)

1975

[7] [7]

D. V. Shalashilin and M. S. Child, Locally coupled co- herent states and Herman–Kluk dynamics, The Journal of Chemical Physics118, 2061 (2003)

[8] [8]

E. J. Heller, Cellular dynamics: A new semiclassical ap- proach to time-dependent quantum mechanics, The Jour- nal of Chemical Physics94, 2723 (1991)

1991