arXiv subjects

Markus Meuwly

Publications and source records attributed to Markus Meuwly.

At least 37 records · Page 2Linked to original sources

arXiv2025

Feshbach Resonances in Cold Collisions: Benchmarking State of the Art ab initio Potential Energy Surfaces

High-quality potential energy surfaces (PES) are a prerequisite for quantitative atomistic simulations, with both quantum and classical dynamics approaches. The ultimate test for the validity of a PES are comparisons with judiciously chosen experimental observables. Here we ask whether cold collision measurements are sufficiently informative to validate and distinguish between high-level, state-of-the art PESs for the strongly interacting Ne-H$_2^+$ system. We show that measurement of the final state distributions for a process that involves only several metastable intermediate states is sufficient to identify the PES that captures the long-range interactions properly. Furthermore, we show that a modest increase in the experimental energy resolution will allow for resolving individual Feshbach resonances and enable a quantitative probe of the interactions at short and intermediate range.

physics.chem-ph↗

arXiv2025

End-to-End Photodissociation Dynamics of Energized H$_2$COO

The end-to-end dynamics of the smallest energized Criegee intermediate, H$_2$COO, was characterized for vibrational excitation close to and a few kcal/mol above the barrier for hydrogen transfer. From an aggregate of at least 5 $μ$s of molecular dynamics simulations using a neural network-representation of CASPT2/aug-cc-pVTZ reference data, the branching ratios into molecular products HCO+OH, CO$_2$+H$_2$, or H$_2$O+CO was quantitatively determined. Consistent with earlier calculations and recent experiments, decay into HCO+OH was found to be rare $(\sim 2 \%)$ whereas the other two molecular product channels are accessed with fractions of $\sim 30 \%$ and $\sim 20 \%$, respectively. On the 1 ns time scale, which was the length of an individual MD simulation, more than 40 \% of the systems remain in the reactant state due to partial intramolecular vibrational redistribution (IVR). Formation of CO$_2$+H$_2$ occurs through a bifurcating pathway, one of which passes through formic acid whereas the more probable route connects the di-radical OCH$_2$O with the product through a low-lying transition state. Notably, none of the intermediates along the pathway accumulate and their maximum concentration always remains well below 5 \%. This work demonstrates that atomistic simulations with global reactive machine-learned energy functions provide a quantitative understanding of the chemistry and reaction dynamics for atmospheric reactions in the gas phase.

physics.chem-ph↗

arXiv2025

Towards Large-Scale Condensed Phase Simulations using Machine Learned Energy Functions

Accurate, yet computationally efficient energy functions are essential for state-of-the art molecular dynamics (MD) studies of condensed phase systems. Here, a generic workflow based on a combination of machine learning-based and empirical representations of intra- and intermolecular interactions is presented. The total energy is decomposed into internal contributions, and electrostatic and van der Waals interactions between monomers. The monomer potential energy surface is described using a neural network, whereas for the electrostatics the flexible minimally distributed charge model is employed. Remaining contributions between reference energies from electronic structure calculations and the model are fitted to standard Lennard-Jones (12-6) terms. For water as a topical example, reference energies for the monomers are determined from CCSD(T)-F12 calculations whereas for an ensemble of cluster structures containing $[2,60]$ and $[2,4]$ monomers DFT and CCSD(T) energies, respectively, were used to best match the van der Waals contributions. Based on the bulk liquid density and heat of vaporization, the best-performing set of LJ(12-6) parameters was selected and a wide range of condensed phase properties were determined and compared with experiment. MD Simulations on the multiple-nanosecond time scale were carried out for water boxes containing 2000 to 8000 monomers, depending on the property considered. The performance of such a generic ML-inspired parametrization scheme is very promising and future improvements and extensions are discussed, also in view of recent advances for water in particular in the literature.

physics.chem-ph↗

arXiv2025

High-Energy Reaction Dynamics of O$_3$

The high-temperature atom exchange and dissociation reaction dynamics of the O($^3$P) + O$_2(^3Σ_g^{-} )$ system are investigated based on a new reproducing kernel-based representation of high-level multi-reference configuration interaction energies. Quasi-classical trajectory (QCT) simulations find the experimentally measured negative tempe-rature-dependence of the rate for the exchange reaction and describe the experiments within error bars. Similarly, QCT simulations for a recent potential energy surface (PES) at a comparable level of quantum chemical theory reproduce the negative $T-$dependence. Interestingly, both PESs feature a ``reef" structure near dissociation which has been implicated to be responsible for a positive $T-$dependence of the rate inconsistent with experiments. For the dissociation reaction the $T-$dependence correctly captures that known from experiments but underestimates the absolute rates by two orders of magnitude. Accounting for an increased number of accessible electronic states reduces this to one order of magnitude. A neural network-based state-to-distribution model is constructed for both PESs and shows good performance in predicting final translational, vibrational, and rotational product state distributions. Such models are valuable for future and more coarse-grained simulations of reactive hypersonic gas flow.

physics.chem-ph↗

arXiv2025

Reaction Dynamics for the [NNO] System from State-Resolved and Coarse-Grained Models

The dynamics for the NO($X^2 Π$) + N($^4$S) $\leftrightarrow$ N$_{2}(X^{1}Σ_{g}^{+}$) + O($^{3}$P) reaction was followed in the $^3$A' electronic state using state-to-state (STS) and Arrhenius-based rates from two different high-level potential energy surfaces represented as a reproducing kernel (RKHS) and permutationally invariant polynomials (PIPs). Despite the different number of bound states supported by the RKHS- and PIP-PESs the ignition points from STS and Arrhenius rates are at $\sim 10^{-6}$ s whether or not reverse rates are from assuming microreversibility or explicitly given. Conversion from NO to N$_2$ is incomplete if Arrhenius-rates are used but complete turnover is observed if STS-information is used. This is due to non-equilibrium energy flow and state dynamics which requires a state-based description. Including full dissociation leads asymptotically to the correct 2:1 [N]:[O] concentration with little differences for the species' dynamics depending on the PES used for the STS-information. In conclusion, concentration profiles from coarse-grained simulations are consistent over 14 orders of magnitude in time using STS-information based on two different high-level PESs.

physics.chem-ph↗

arXiv2025

Augmenting chemical databases for atomistic machine learning by sampling conformational space

Machine learning (ML) has become a standard tool for the exploration of chemical space. Much of the performance of such models depends on the chosen database for a given task. Here, this aspect is investigated for "chemical tasks" including the prediction of hybridization, oxidation, substituent effects, and aromaticity, starting from an initial "restricted" database (iRD). Choosing molecules for augmenting this iRD, including increasing numbers of conformations generated at different temperatures, and retraining the models can improve predictions of the models on the selected "tasks". Addition of a small percentage of conformers (1 % ) obtained at 300 K improves the performance in almost all cases. On the other hand, and in line with previous studies, redundancy and highly deformed structures in the augmentation set compromise prediction quality. Energy and bond distributions were evaluated by means of Kullback-Leibler ($D_{\rm KL}$) and Jensen-Shannon ($D_{\rm JS}$) divergence and Wasserstein distance ($W_{1}$). The findings of this work provide a baseline for the rational augmentation of chemical databases or the creation of synthetic databases.

physics.chem-ph↗

arXiv2025

Reaction Dynamics of the H + HeH$^+$ $\rightarrow$ He + H$_2^+$ System

The reaction dynamics for the H + HeH$^+$ $\rightarrow$ He + H$_2^+$ reaction in its electronic ground state is investigated using two different representations of the potential energy surface (PES). The first uses a combined kernel and neural network representation of UCCSD(T) reference data whereas the second is a corrected PES (cR-PES) that eliminates an artificial barrier in the entrance channel appearing in its initial expansion based on full configuration interaction reference data. Despite the differences between the two PESs, both yield $k_{v=0,j=0} \approx 2 \times 10^{-9}$ cm$^3$/molecule/s at $T = 10$ K which is consistent with a $T-$independent Langevin rate $k_{\rm L} = 2.1 \times 10^{-9}$ cm$^3$/molecule/s but considerably larger than the only experimentally reported value $k_{\rm ICR} = (9.1 \pm 2.5) \times 10^{-10}$ cm$^3$/molecule/s from ion cyclotron resonance experiments. Similarly, branching ratios for the reaction outcomes are comparable for the two PESs. However, when analysing less averaged properties such as initial state-selected $T-$dependent rate coefficients and final vibrational states of the H$_2^+$ product for low temperatures, the differences in the two PESs manifest themselves in the observables. Thus, depending on the property analyzed, accurate and globally valid representations of the PES are required, whereas more approximate and empirical construction schemes can be followed for state-averaged observables.

physics.chem-ph↗

arXiv2025

Structure and Dynamics of Deep Eutectic Systems from Cluster-Optimized Energy Functions

Generating energy functions for heterogeneous systems suitable for quantitative and predictive atomistic simulations is a challenging undertaking. The present work combines a cluster-based approach with electronic structure calculations at the density functional theory level and machine learning-based energy functions for a spectroscopic reporter for eutectic mixtures consisting of water, acetamide and KSCN. Two water models are considered: TIP3P which is consistent with the CGenFF energy function and TIP4P which - as a water model - is superior to TIP4P. Both fitted models, {\bf M2$^{\rm TIP3P}$} and {\bf M2$^{\rm TIP4P}$}, yield favourable thermodynamic, structural, spectroscopic and transport properties from extensive molecular dynamics simulations. In particular, the slow and fast decay times from 2-dimensional infrared spectroscopy and the viscosity for water-rich mixtures are described realistically and consistent with experiments. On the other hand, including the co-solvent (acetamide) in the present case is expected to further improve the computed viscosity for low-water content. It is concluded that such a cluster-based approach is a promising and generalizable route for routine parametrization of heterogeneous, electrostatically dominated systems.

physics.chem-ph↗

arXiv2024

The Bigger the Better? Accurate Molecular Potential Energy Surfaces from Minimalist Neural Networks

Atomistic simulations are a powerful tool for studying the dynamics of molecules, proteins, and materials on wide time and length scales. Their reliability and predictiveness, however, depend directly on the accuracy of the underlying potential energy surface (PES). Guided by the principle of parsimony this work introduces KerNN, a combined kernel/neural network-based approach to represent molecular PESs. Compared to state-of-the-art neural network PESs the number of learnable parameters of KerNN is significantly reduced. This speeds up training and evaluation times by several orders of magnitude while retaining high prediction accuracy. Importantly, using kernels as the features also improves the extrapolation capabilities of KerNN far beyond the coverage provided by the training data which solves a general problem of NN-based PESs. KerNN applied to spectroscopy and reaction dynamics shows excellent performance on test set statistics and observables including vibrational bands computed from classical and quantum simulations.

physics.chem-ph↗

arXiv2024

Machine Learning-Based Enhancements of Empirical Energy Functions: Structure, Dynamics and Spectroscopy of Modified Benzenes

The effect of replacing individual contributions to an empirical energy function are assessed for halogenated benzenes (X-Bz, X = H, F, Cl, Br) and chlorinated phenols (Cl-PhOH). Introducing electrostatic models based on distributed charges (MDCM) instead of usual atom-centered point charges yields overestimated hydration free energies unless the van der Waals parameters are reparametrized. Scaling van der Waals ranges by 10 \% to 20 \% for three Cl-PhOH and most X-Bz yield results within experimental error bars, which is encouraging, whereas for benzene (H-Bz) point charge-based models are sufficient. Replacing the bonded terms by a neural network-trained energy function with either fluctuating charges or MDCM electrostatics also yields qualitatively correct hydration free energies which still require adaptation of the van der Waals parameters. The infrared spectroscopy of Cl-PhOH is rather well predicted by all models although the ML-based energy function performs somewhat better in the region of the framework modes. It is concluded that refinements of empirical energy functions for targeted applications is a meaningful way towards more quantitative simulations.

physics.chem-ph↗

arXiv2024

Force Fields for Deep Eutectic Mixtures: Application to Structure and 2D-Infrared Spectroscopy

Parametrizing energy functions for ionic systems can be challenging. Here, the total energy function for an eutectic system consisting of water, SCN$^-$, K$^+$ and acetamide is improved vis-a-vis experimentally measured properties. Given the importance of electrostatic interactions, two different types of models are considered: the first (model M0) uses atom-centered multipole whereas the other two (models M1 and M2) are based on fluctuating minimal distributed charges (fMDCM) that respond to geometrical changes of SCN$^-$. The Lennard-Jones parameters of the anion are adjusted to best reproduce experimentally known hydration free energies and densities which are matched to within a few percent for the final models irrespective of the electrostatic model. Molecular dynamics simulations of the eutectic mixtures with varying water content (between 0% and 100%) yield radial distribution functions and frequency correlation functions for the CN-stretch vibration. Comparison with experiments indicate that models based on fMDCM are considerably more consistent that those using multipoles. Computed viscosities from models M1 and M2 are within 30% of measured values and their change with increasing water content is consistent with experiments. This is not the case for model M0.

physics.chem-ph↗

arXiv2024

Energy Relaxation of N$_2$O in Gaseous, Supercritical and Liquid Xenon and SF$_6$

Rotational and vibrational energy relaxation (RER and VER) of N$_2$O embedded in xenon and SF$_6$ environments ranging from the gas phase to the liquid, including the supercritical regime, is studied at a molecular level. Calibrated intermolecular interactions from high-level electronic structure calculations, validated against experiments for the pure solvents were used to carry out classical molecular dynamics simulations corresponding to experimental state points for near-critical isotherms. Computed RER rates in low-density solvent of $k_{\rm rot}^{\rm Xe} = (3.67\pm0.25)\cdot10^{10}$ s$^{-1}$M$^{-1}$ and $k_{\rm rot}^{\rm SF_6} = (1.25\pm0.12)\cdot10^{11}$ s$^{-1}$M$^{-1}$ compare well with rates determined by analysis of 2-dimensional infrared experiments. Simulations find that an isolated binary collision (IBC) description is successful up to solvent concentrations of $\sim 4$ M. For higher densities, including the supercritical regime, the simulations do not correctly describe RER, probably due to neglect of solvent-solute coupling in the analysis of the rotational motion. For VER, the near-quantitative agreement between simulations and pump-probe experiments captures the solvent density-dependent trends.

physics.chem-ph↗

arXiv2024

Diffusion and Spectroscopy of H$_2$ in Myoglobin

The diffusional dynamics and vibrational spectroscopy of molecular hydrogen (H$_2$) in myoglobin (Mb) is characterized. Hydrogen has been implicated in a number of physiologically relevant processes, including cellular aging or inflammation. Here, the internal diffusion through the protein matrix was characterized and the vibrational spectroscopy was investigated using conventional empirical energy functions and improved models able to describe higher-order electrostatic moments of the ligand. H$_2$ can occupy the same internal defects as already found for Xe or CO (Xe1 to Xe4 and B-state). Furthermore, 4 additional sites were found, some of which had been discovered in earlier simulation studies. The vibrational spectra using the most refined energy function indicate that depending on the docking site the spectroscopy of H$_2$ differs. The maxima of the absorption spectra cover $\sim 20$ cm$^{-1}$ which are indicative of a pronounced effect of the surrounding protein matrix on the vibrational spectroscopy of the ligand. Electronic structure calculations show that H$_2$ forms a stable complex with the heme-iron (stabilized by $\sim -12$ kcal/mol) but splitting of H$_2$ is unlikely due to a high activation energy ($\sim 50$ kcal/mol).

physics.chem-ph↗

arXiv2024

Accurate Tunneling Splittings for Ever-Larger Molecules from Transfer-Learned, CCSD(T) Quality Energy Functions

This work combines state-of-the-art machine learning techniques with highest-level electronic structure calculations and full-dimensional quantum tunneling calculations to obtain a quantitative characterization of tunneling splittings for system sizes that are currently out of reach using traditional approaches. For intramolecular hydrogen transfer in tropolone, the best computed splitting including perturbative corrections in the ring-polymer instanton calculations is 0.94 cm$^{-1}$ and compares with 0.974 cm$^{-1}$ from experiments. On the other hand, for intermolecular double hydrogen transfer in the (propiolic acid)-(formic acid) dimer, the computations yield 0.0147 cm$^{-1}$ which is larger by 40 % compared with experiment (0.0097 cm$^{-1}$) but still in much better agreement than previous attempts (0.63 cm$^{-1}$). The strategy pursued in the present work is applicable to yet larger systems and other properties of interest and provides a rational route for highest-accuracy energy functions for prediction and benchmarking electronic structure methods vis-a-vis experiments.

physics.chem-ph↗

arXiv2024

${\it Asparagus}$: A Toolkit for Autonomous, User-Guided Construction of Machine-Learned Potential Energy Surfaces

With the establishment of machine learning (ML) techniques in the scientific community, the construction of ML potential energy surfaces (ML-PES) has become a standard process in physics and chemistry. So far, improvements in the construction of ML-PES models have been conducted independently, creating an initial hurdle for new users to overcome and complicating the reproducibility of results. Aiming to reduce the bar for the extensive use of ML-PES, we introduce ${\it Asparagus}$, a software package encompassing the different parts into one coherent implementation that allows an autonomous, user-guided construction of ML-PES models. ${\it Asparagus}$ combines capabilities of initial data sampling with interfaces to ${\it ab initio}$ calculation programs, ML model training, as well as model evaluation and its application within other codes such as ASE or CHARMM. The functionalities of the code are illustrated in different examples, including the dynamics of small molecules, the representation of reactive potentials in organometallic compounds, and atom diffusion on periodic surface structures. The modular framework of ${\it Asparagus}$ is designed to allow simple implementations of further ML-related methods and models to provide constant user-friendly access to state-of-the-art ML techniques.

physics.chem-ph↗

arXiv2024

Kernel-based Minimal Distributed Charges: A Conformationally Dependent ESP-Model for Molecular Simulations

A kernel-based method (kernelized minimal distributed charge model - kMDCM) to represent the molecular electrostatic potential (ESP) in terms of off-center point charges whose positions adapts to the molecular geometry. Using Gaussian kernels and atom-atom distances as the features, the ESP for water and methanol is shown to improve by at least a factor of two compared with point charge models fit to an ensemble of structures. Combining kMDCM for the electrostatics and reproducing kernels for the bonded terms allows energy-conserving simulation of 2000 water molecules with periodic boundary conditions on the nanosecond time scale.

physics.chem-ph↗

arXiv2024

High-Energy Reaction Dynamics of N$_{3}$

The atom-exchange and atomization dissociation dynamics for the N($^4$S) + N$_2(^1 Σ_{\rm g}^+)$ reaction is studied using a reproducing kernel Hilbert space (RKHS)-based, global potential energy surface (PES) at the MRCI-F12/aug-cc-pVTZ-F12 level of theory. For the atom exchange reaction $({\rm N_A N_B} + {\rm N_C} \rightarrow {\rm N_A N_C} + {\rm N_B}$), computed thermal rates and their temperature dependence from quasi-classical trajectory (QCT) simulations agree to within error bars with the available experiments. Companion QCT simulations using a recently published CASPT2-based PES confirm these findings. For the atomization reaction, leading to three N$(^4{\rm S})$ atoms, the computed rates from the RKHS-PES overestimate the experimentally reported rates by one order of magnitude whereas those from the PIP-PES agree favourably, and the $T$-dependence of both computations is consistent with experiment. These differences can be traced back to the different methods and basis sets used. The lifetime of the metastable N$_3$ molecule is estimated to be $\sim 200$ fs depending on the initial state of the reactants. Finally, neural network-based exhaustive state-to-distribution models are presented using both PESs for the atom exchange reaction. These models will be instrumental for a broader exploration of the reaction dynamics of air.

physics.chem-ph↗

arXiv2024

SCN as a Local Probe of Protein Structural Dynamics

The dynamics of lysozyme is probed by attaching -SCN to all alanine-residues. The 1-dimensional infrared spectra exhibit frequency shifts in the position of the maximum absorption by 4 cm$^{-1}$ which is consistent with experiments in different solvents and indicates moderately strong interactions of the vibrational probe with its environment. Isotopic substitution $^{12}$C $\rightarrow ^{13}$C leads to a red-shift by $-47$ cm$^{-1}$ which is consistent with experiments with results on CN-substituted copper complexes in solution. The low-frequency, far-infrared part of the protein spectra contain label-specific information in the difference spectra when compared with the wild type protein. Depending on the positioning of the labels, local structural changes are observed. For example, introducing the -SCN label at Ala129 leads to breaking of the $α-$helical structure with concomitant change in the far-infrared spectrum. Finally, changes in the local hydration of SCN-labelled Alanine residues as a function of time can be related to angular reorientation of the label. It is concluded that -SCN is potentially useful for probing protein dynamics, both in the high-frequency (CN-stretch) and far-infrared part of the spectrum.

physics.chem-ph↗