A shortcut for changing temperature
A preprint describes a computational method that reuses one excited-state trajectory to produce molecular spectra at different temperatures, without recomputing that trajectory after the zero-temperature spectrum has been evaluated. The work asks whether this single-Hessian approach can make nonzero-temperature vibronic spectra less expensive to calculate.
The method combines Gaussian wavepacket dynamics with coherence thermofield dynamics, using one fixed curvature matrix, known as a Hessian, and one temperature-independent excited-state classical trajectory. In plain terms, the trajectory is held fixed while the spectrum is evaluated for different temperatures.
For harmonic potentials—idealized systems without anharmonic departures—the method is exact when the reference Hessian is taken from the final surface. The researchers then tested the approximation on Morse-potential systems under different anharmonicities and temperatures.
For the molecular examples, the on-the-fly calculations used an in-house Fortran 2018 molecular-dynamics code interfaced with Gaussian 16. Density-functional theory optimized the ground states, while time-dependent DFT treated the excited states.
The model tests found a trade-off
The first benchmark used Morse-potential model systems. Across all studied anharmonicities and temperatures, single-Hessian coherence thermofield Gaussian wavepacket dynamics outperformed the ordinary harmonic approximation.
It deteriorated slightly faster than local harmonic Gaussian wavepacket dynamics as temperature increased. The result therefore came with a trade-off: the single-Hessian method improved on the simpler harmonic approximation, but its relative performance against the local harmonic approach weakened as the model was heated.
The benchmark is best read as a result for the Morse conditions that were studied. Within those conditions, the method retained its advantage over the harmonic approximation while showing faster temperature-related deterioration than the local harmonic approach.
Four molecular tests
The workflow was then applied to four molecule-specific examples: a selenium-dioxide photoelectron system, phenyl-radical absorption, naphthalene absorption and aminocoumarin C450 absorption.
In the selenium-dioxide photoelectron comparison, the method captured the two hot bands seen in experiment—temperature-linked spectral features—and also produced additional thermally activated transitions.
The timing reported for that example separated the initial calculation from later temperature changes. The zero-temperature single-Hessian calculation took 4.8 × 10^3 seconds on 8 cores; once the trajectory had been computed, the 700 K spectrum took under one second.
In the room-temperature phenyl-radical comparison, coherence thermofield Gaussian wavepacket dynamics qualitatively captured three hot bands. That describes the recovery of the main thermal structure, rather than a claim that every spectral detail matched.
For naphthalene, the method captured temperature-dependent broadening while applying a common shift across the temperatures. The calculated spectra therefore reflected changing width without using a different shift for each temperature.
Aminocoumarin C450 provided a test of peak structure. Coherence thermofield Gaussian wavepacket dynamics preserved a three-peak progression, while broadening the zero-temperature spectrum merged the first two peaks.
The boundary is part of the result
The single-trajectory shortcut has a defined boundary. The authors state that it is not applicable when tunnelling or wavepacket splitting dominates the shape of the spectrum.
The Morse results point to a second caution: as temperature increased, the single-Hessian approximation deteriorated slightly faster than local harmonic Gaussian wavepacket dynamics. That did not remove its advantage over the ordinary harmonic approximation in the conditions tested, but it makes the studied temperature and anharmonicity ranges important to the result.
The study’s conclusion is accordingly narrower than a general claim about every molecular spectrum. Its evidence concerns the specific Morse-potential tests and four molecule-specific simulations, with the method’s usefulness assessed through those comparisons.
The preprint reports that the Python program and supporting data are openly available in Zenodo.
The work was supported by the Swiss National Science Foundation under Grant No. 10005187 and by EPFL.
The practical case for the method is clear within that scope: after the trajectory calculation is complete, additional temperatures can be obtained at very low extra cost in the selenium-dioxide example, while the aminocoumarin test shows that temperature-dependent peak structure can survive where simple zero-temperature broadening loses it.
Paper data and sources
Original title: Nonzero-temperature vibronic spectra of polyatomic molecules from a zero-temperature classical trajectory
Authors: Davide Barbiero, Jiří J. L. Vaníček
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text