arXiv subjects

Hubert Beck

Publications and source records attributed to Hubert Beck.

5 recordsLinked to original sources

arXiv2026

More converged, less accurate? Reassessing standard choices for ab initio water using machine learning potentials

Accurately simulating the properties of liquid water remains a central challenge in molecular simulations. In this work, we use machine learning potentials to investigate how the convergence settings of electronic structure calculations impact the predicted structural and dynamical properties of simulated water and ice. We evaluate the true performance of several reference methods in classical and path-integral molecular dynamics. When we compare a popular, computationally pragmatic revPBE0-D3 setup against a highly converged one, our results reveal that its widely reported experimental agreement degrades. Applying the same highly converged settings to the $\mathrm{\omega}$B97X-rV functional, we find an improved agreement with experimental results. MP2 with a triple-$\zeta$ basis set commonly used for liquid water shows poor performance, which is indicative of insufficient convergence. These findings underscore the need for fully converged reference calculations when evaluating the fundamental accuracy of electronic structure methods and developing reliable models for aqueous systems.

physics.chem-ph↗

arXiv2025

Multi-head committees enable direct uncertainty prediction for atomistic foundation models

Machine learning potentials have become a standard tool for atomistic materials modelling. While models continue to become more generalisable, an open challenge relates to efficient uncertainty predictions for active learning and robust error analysis. In this work, we utilise MACE and its multi-head mechanism to implement a committee neural network potential for message-passing architectures, where the committee comprises multiple output modules attached to the same atomic environment descriptors. As with traditional committees of independent networks, the standard deviation of the predictions functions as an estimate of the model's uncertainty. We show for a range of datasets in custom-build models that the uncertainty of the force predictions correlates well with the true errors. We subsequently apply this concept to foundation models, specifically MACE-MP-0, where we train only the newly attached output heads while keeping the remaining part of the model fixed. We use this approach in an active learning workflow to condense the training set of the foundation model to just 5\% of its original size. The foundation model multi-head committee trained on the condensed training set enables reliable uncertainty estimation without any substantial decrease in prediction accuracy.

physics.chem-ph↗

arXiv2024

Elucidating the Nature of $\pi$-hydrogen Bonding in Liquid Water and Ammonia

Aromatic compounds form an unusual kind of hydrogen bond with water and ammonia molecules, known as the $\pi$-hydrogen bond. In this work, we report ab initio path integral molecular dynamics simulations enhanced by machine-learning potentials to study the structural, dynamical, and spectroscopic properties of solutions of benzene in liquid water and ammonia. Specifically, we model the spatial distribution functions of the solvents around the benzene molecule, establish the $\pi$-hydrogen bonding interaction as a prominent structural motive, and set up existence criteria to distinguish the $\pi$-hydrogen bonded configurations. These serve as a structural basis to calculate binding affinities of the solvent molecules in $\pi$hydrogen bonds, identify an anticooperativity effect across the aromatic ring in water (but not ammonia), and estimate $\pi$-hydrogen bond lifetimes in both solvents. Finally, we model hydration-shell-resolved vibrational spectra to clearly identify the vibrational signature of this structural motif in our simulations. These decomposed spectra corroborate previous experimental findings for benzene in water, offer additional insights, and further emphasize the contrast between $\pi$-hydrogen bonds in water and in ammonia. Our simulations provide a comprehensive picture of the studied phenomenon and, at the same time, serve as a meaningful \textit{ab initio} reference for an accurate description of $\pi$-hydrogen bonding using empirical force fields in more complex situations, such as the hydration of biological interfaces.

physics.chem-ph↗

arXiv2023

Reducing the cost of neural network potential generation for reactive molecular systems

Although machine-learning potentials have recently had substantial impact on molecular simulations, the construction of a robust training set can still become a limiting factor, especially due to the requirement of a reference ab initio simulation that covers all the relevant geometries of the system. Recognizing that this can be prohibitive for certain systems, we develop the method of transition tube sampling that mitigates the computational cost of training set and model generation. In this approach, we generate classical or quantum thermal geometries around a transition path describing a conformational change or a chemical reaction using only a sparse set of local normal mode expansions along this path and select from these geometries by an active learning protocol. This yields a training set with geometries that characterize the whole transition without the need for a costly reference trajectory. The performance of the method is evaluated on different molecular systems with the complexity of the potential energy landscape increasing from a single minimum to a double proton-transfer reaction with high barriers. Our results show that the method leads to training sets that give rise to models applicable in classical and path integral simulations alike that are on par with those based directly on ab initio calculations while providing the computational speed-up we have come to expect from machine-learning potentials.

physics.chem-ph↗

arXiv2019

Structure and Binding in Halide Perovskites: Analysis of Static and Dynamic Effects from Dispersion-Corrected Density Functional Theory

We investigate the impact of various levels of approximation in density functional theory calculations for the structural and binding properties of the prototypical halide perovskite MAPbI$_3$. Specifically, we test how the inclusion of different correction schemes for including dispersive interactions, and how in addition using hybrid density functional theory, affects the results for pertinent structural observables by means of comparison to experimental data. In particular, the impact of finite temperature on the lattice constants and bulk modulus, and the role of dispersive interactions in calculating them, is examined by using molecular dynamics based on density functional theory. Our findings confirm previous theoretical work showing that including dispersive corrections is crucial for accurate calculation of structural and binding properties of MAPbI$_3$. They furthermore highlight that using a computationally much more expensive hybrid density functional has only minor consequences for these observables. This allows for suggesting the use of semilocal density functional theory, augmented by pairwise dispersive corrections, as a reasonable choice for structurally more complicated calculations of halide perovskites. Using this method, we perform molecular dynamics calculations and discuss the dynamic effect of molecular rotation on the structure of and binding in MAPbI$_3$, which allowed for rationalizing microscopically the simultaneous occurrence of cubic octahedral symmetry and MA disorder.

cond-mat.mtrl-sci↗