Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.32652

Low-variance force estimators for nuclear quantum mean forces

Venkat Kapil
View original source ↗Read source document ↗Export metadata ↓

Abstract

Quantum equilibrium structure can be sampled using classical molecular dynamics with a learned effective potential for a single path-integral replica, but the primitive mean-force estimates obtained from path-integral molecular dynamics with that replica held fixed are noisy. We derive a scaled-gradient estimator with zero sampling variance for a one-dimensional harmonic oscillator and an estimated 4.7--57.5-fold reduction in constrained-sampling time for the molecular tests. The resulting potentials can be trained on these forces like standard machine-learning interatomic potentials and yield stable production simulations from which quantum expectation values of position-dependent observables are obtained as classical ensemble averages.

Explore related subjects

physics.chem-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Venkat Kapil. 2026-10-23. Low-variance force estimators for nuclear quantum mean forces. https://arxiv.org/abs/2610.32652

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Contrastive Regularization of Machine Learning Potentials

Machine learning models are increasingly trained as regressors and then deployed as generators, from language models to learned simulators. In this setting, accuracy on held-out data says little about the distribution a model produces when it is sampled. We study this problem where the correct distribution is fixed by first principles, in machine learning interatomic potentials (MLIPs) trained by minimizing the mean squared error (MSE) on density functional theory (DFT) energies and forces, then used to run molecular simulations. The MSE never constrains the Boltzmann distribution that the potential defines, so an MLIP can match held-out DFT data and still drift, when sampled, into spurious minima that corrupt its thermodynamics. Building on the variational principle of statistical physics, we introduce Contrastive Regularized MSE (CRMSE), a post-training step in which the potential acts as its own energy-based model and learns to reproduce the distribution it generates, without any new DFT data. On molecules from the MD17 benchmark, CRMSE restores physical sampling and recovers DFT free-energy profiles, while keeping energy and force errors within chemical accuracy. Our results suggest that a model meant to be sampled must be trained on the distribution it produces, not only for pointwise accuracy.

physics.chem-ph↗
arXiv2026

Resolving Peak Shifting Effects in Real-Time Time-Dependent Orbital Propagation Methods Through Nonorthogonal Expansions

Self-consistent field real-time electronic structure methods are a powerful approach for modeling ultrafast processes but suffer from systematic errors in predicted transition energies, commonly known as peak shifting, that obscure direct comparison with experiment. Although this effect has been explored in single reference methods, a general understanding that extends to multiconfigurational methods remains unresolved. Here, we develop a real-time nonorthogonal multiconfigurational self-consistent field (RT-TD NOMCSCF) formalism in which independently propagated orbital sets generate a compact, nonorthogonal wavefunction. The approach establishes a unified framework for nonlinear real-time electronic structure theories, which we use to reveal the common origin of peak shifting across single and multireference limits. Using this framework, we demonstrate that peak shifting does not arise exclusively from the effect of state averaging over the internal space basis, but also from the accessible external configurational space. By systematically expanding the internal space with independently propagated orbital sets, RT-TD NOMCSCF suppresses this aver- aging, recovers correct transition energies, and distinguishes intrinsic peak shifting from numerical peak drifting caused by propagation errors. Thus, these developments establish nonlinear real-time electronic structure methods that enables more accurate simulations of ultrafast processes.

physics.chem-ph↗
arXiv2026

Derivation of the Sample-Size Scaling of TWO-NN Intrinsic-Dimension Estimates from Molecular Dynamics Trajectories

The intrinsic dimension of a dataset is the number of independent directions needed to describe the space occupied by its data. Nearest-neighbor estimators infer it from how the probability of finding a neighbor grows around each sampled point. Because the neighbor distances $r$ decrease as the sample grows, the estimated dimension can depend strongly on the number $N$ of available points, which in molecular dynamics is set by trajectory length and saving frequency. We derive the large-sample behavior of the TWO-NN estimator for data drawn from a smooth $d$-dimensional space. The typical nearest-neighbor distance scales as $N^{-1/d}$, and smooth deviations from local uniformity produce corrections in powers of $N^{-2/d}$. Isotropic noise adds a correction proportional to $N^{+2/d}$; for a five-dimensional hyperplane in 100 noisy features, the extended expansion recovers the dimension within a few percent when noise is moderate. We test the scaling on ten independent 100 $μ$s simulations of alanine dipeptide, represented by the pairwise distances among its ten heavy atoms, whose geometric dimension is $3n_{\mathrm{at}}-6=24$. Over the investigated range, the estimate does not depend systematically on the temporal spacing between configurations, but increases from $7.6$ to $15.6$ as $N$ grows from $10^2$ to $2\times10^5$. Extrapolations that retain corrections through $r^2$, $r^4$, and $r^6$ give limiting dimensions of $24.81$, $23.47$, and $25.30$, respectively. All three estimates lie within 6% of the known dimension and bracket it, and their spread measures the uncertainty due to truncation. The derived scaling provides a practical route from finite-sample estimates to the geometric dimension.

physics.chem-ph↗