arXiv subjects

Mie Andersen

Publications and source records attributed to Mie Andersen.

17 recordsLinked to original sources

arXiv2026

Fine-tuning universal machine learning potentials for transition state search in surface catalysis

Determining transition states (TSs) of surface reactions is central to understanding and designing heterogeneous catalysts but remains computationally prohibitive with density functional theory (DFT). While machine learning potentials (MLPs) offer significant speedups, task-specific models have limited transferability across catalytic systems, and universal MLPs (uMLPs) lack the accuracy needed for reactive configurations. Here, we present a workflow based on active learning to iteratively fine-tune uMLPs for DFT-quality TS search. Using 250 TSs from the CO2 hydrogenation reaction network on metal and single-atom alloy surfaces, we first benchmark TS search algorithms, identifying the Sella algorithm as most robust, and propose a modification (Bond-Aware Sella) that substantially improves its success rate. We then explore sequential and batch active-learning strategies for fine-tuning and show that DFT-quality TS structures can be found using only 8 DFT single-point calculations on average per structure. This demonstrates the viability of fine-tuned uMLPs for high-throughput catalyst screening.

cond-mat.mtrl-sci↗

arXiv2026

Machine learning exploration of binding energy distributions of H2O at astrochemically relevant dust grain surfaces

Binding energies (BEs) of adsorbates on interstellar dust grains critically control adsorption, desorption, diffusion, and surface reactivity, and therefore strongly influence astrochemical models of star- and planet-forming regions. While recent computational studies increasingly report full distributions of BEs rather than single representative values, these distributions are typically derived for either bare grain surfaces or thick water-ice mantles. In this work, we bridge these regimes by systematically investigating the BE distributions of water on partially and fully ice-covered dust grain surfaces. We employ machine-learning interatomic potentials (MLIPs) based on graph neural networks to model water adsorption on graphene and on the Mg-terminated (010) surface of forsterite, representing carbonaceous and silicate grains, respectively. The models enable extensive sampling of adsorption sites on water clusters, monolayers, and bilayers generated under both crystalline (thermally processed) and amorphous (low-temperature) growth conditions. At submonolayer coverage, the chemical nature of the underlying grain strongly affects both ice morphology and binding energies, with Mg-O interactions on silicate surfaces producing particularly deep binding sites. From monolayer coverage onward, adsorption on both substrates is dominated by hydrogen bonding within the ice, reducing the influence of the grain material. Across all coverages, amorphous ice structures systematically shift the BE distributions toward stronger binding compared to crystalline ice, introducing highly stable defect and pocket sites. These results demonstrate that BE distributions in the submonolayer to few-layer ice regime are broad and highly surface dependent, and they provide physically motivated input for next-generation astrochemical models incorporating surface heterogeneity.

physics.chem-ph↗

arXiv2025

How accurate are foundational machine learning interatomic potentials for heterogeneous catalysis?

Foundational machine learning interatomic potentials (MLIPs) are being developed at a rapid pace, promising closer and closer approximation to ab initio accuracy. This unlocks the possibility to simulate much larger length and time scales. However, benchmarks for these MLIPs are usually limited to ordered, crystalline and bulk materials. Hence, reported performance does not necessarily accurately reflect MLIP performance in real applications such as heterogeneous catalysis. Here, we systematically analyze zero-shot performance of 80 different MLIPs, evaluating tasks typical for heterogeneous catalysis across a range of different data sets, including adsorption and reaction on surfaces of alloyed metals, oxides, and metal-oxide interfacial systems. We demonstrate that current-generation foundational MLIPs can already perform at high accuracy for applications such as predicting vacancy formation energies of perovskite oxides or zero-point energies of supported nanoclusters. However, limitations also exist. We find that many MLIPs catastrophically fail when applied to magnetic materials, and structure relaxation in the MLIP generally increases the energy prediction error compared to single-point evaluation of a previously optimized structure. Comparing low-cost task-specific models to foundational MLIPs, we highlight some core differences between these model approaches and show that -- if considering only accuracy -- these models can compete with the current generation of best-performing MLIPs. Furthermore, we show that no single MLIP universally performs best, requiring users to investigate MLIP suitability for their desired application.

cond-mat.mtrl-sci↗

arXiv2025

Interpretable machine learned predictions of adsorption energies at the metal--oxide interface

The conversion of $\mathrm{CO_2}$ to value-added compounds is an important part of the effort to store and reuse atmospheric $\mathrm{CO_2}$ emissions. Here we focus on $\mathrm{CO_2}$ hydrogenation over so-called inverse catalysts: transition metal oxide clusters supported on metal surfaces. The conventional approach for computational screening of such candidate catalyst materials involves a reliance on density functional theory (DFT) to obtain accurate adsorption energies at a significant computational cost. Here we present a machine learning (ML)-accelerated workflow for obtaining adsorption energies at the metal--oxide interface. We enumerate possible binding sites at the clusters and use DFT to sample a subset of these with diverse local adsorbate environments. The data set is used to explore interpretable and black-box ML models with the aim to reveal the electronic and structural factors controlling adsorption at metal--oxide interfaces. Furthermore, the explored ML models can be used for low-cost prediction of adsorption energies on structures outside of the original training data set. The workflow presented here, along with the insights into trends in adsorption energies at metal--oxide interfaces, will be useful for identifying active sites, predicting parameters required for microkinetic modeling of reactions on complex catalyst materials, and accelerating data-driven catalyst design.

cond-mat.mtrl-sci↗

arXiv2025

Transition States Energies from Machine Learning: An Application to Reverse Water-Gas Shift on Single-Atom Alloys

Obtaining accurate transition state (TS) energies is a bottleneck in computational screening of complex materials and reaction networks due to the high cost of TS search methods and first-principles methods such as density functional theory (DFT). Here we propose a machine learning (ML) model for predicting TS energies based on Gaussian process regression with the Wasserstein Weisfeiler-Lehman graph kernel (WWL-GPR). Applying the model to predict adsorption and TS energies for the reverse water-gas shift (RWGS) reaction on single-atom alloy (SAA) catalysts, we show that it can significantly improve the accuracy compared to traditional approaches based on scaling relations or ML models without a graph representation. Further benefitting from the low cost of model training, we train an ensemble of WWL-GPR models to obtain uncertainties through subsampling of the training data and show how these uncertainties propagate to turnover frequency (TOF) predictions through the construction of an ensemble of microkinetic models. Comparing the errors in model-based vs DFT-based TOF predictions, we show that the WWL-GPR model reduces errors by almost an order of magnitude compared to scaling relations. This demonstrates the critical impact of accurate energy predictions on catalytic activity estimation. Finally, we apply our model to screen new materials, identifying promising catalysts for RWGS. This work highlights the power of combining advanced ML techniques with DFT and microkinetic modeling for screening catalysts for complex reactions like RWGS, providing a robust framework for future catalyst design.

cond-mat.mtrl-sci↗

arXiv2025

Breaking scaling relations with inverse catalysts: a machine learning exploration of trends in $\mathrm{CO_2}$ hydrogenation energy barriers

The conversion of $\mathrm{CO_2}$ into useful products such as methanol is a key strategy for abating climate change and our dependence on fossil fuels. Developing new catalysts for this process is costly and time-consuming and can thus benefit from computational exploration of possible active sites. However, this is complicated by the complexity of the materials and reaction networks. Here, we present a workflow for exploring transition states of elementary reaction steps at inverse catalysts, which is based on the training of a neural network-based machine learning interatomic potential. We focus on the crucial formate intermediate and its formation over nanoclusters of indium oxide supported on Cu(111). The speedup compared to an approach purely based on density functional theory allows us to probe a wide variety of active sites found at nanoclusters of different sizes and stoichiometries. Analysis of the obtained set of transition state geometries reveals different structure--activity trends at the edge or interior of the nanoclusters. Furthermore, the identified geometries allow for the breaking of linear scaling relations, which could be a key underlying reason for the excellent catalytic performance of inverse catalysts observed in experiments.

cond-mat.mtrl-sci↗

arXiv2024

LG-Sleep: Local and Global Temporal Dependencies for Mice Sleep Scoring

Efficiently identifying sleep stages is crucial for unraveling the intricacies of sleep in both preclinical and clinical research. The labor-intensive nature of manual sleep scoring, demanding substantial expertise, has prompted a surge of interest in automated alternatives. Sleep studies in mice play a significant role in understanding sleep patterns and disorders and underscore the need for robust scoring methodologies. In response, this study introduces LG-Sleep, a novel subject-independent deep neural network architecture designed for mice sleep scoring through electroencephalogram (EEG) signals. LG-Sleep extracts local and global temporal transitions within EEG signals to categorize sleep data into three stages: wake, rapid eye movement (REM) sleep, and non-rapid eye movement (NREM) sleep. The model leverages local and global temporal information by employing time-distributed convolutional neural networks to discern local temporal transitions in EEG data. Subsequently, features derived from the convolutional filters traverse long short-term memory blocks, capturing global transitions over extended periods. Crucially, the model is optimized in an autoencoder-decoder fashion, facilitating generalization across distinct subjects and adapting to limited training samples. Experimental findings demonstrate superior performance of LG-Sleep compared to conventional deep neural networks. Moreover, the model exhibits good performance across different sleep stages even when tasked with scoring based on limited training samples.

cs.LG↗

arXiv2022

Predicting binding energies of astrochemically relevant molecules via machine learning

The behaviour of molecules in space is to a large extent governed by where they freeze out or sublimate. The molecular binding energy is thus an important parameter for many astrochemical studies. This parameter is usually determined with time-consuming experiments, computationally expensive quantum chemical calculations, or the inexpensive, but inaccurate, linear addition method. In this work we propose a new method based on machine learning for predicting binding energies that is accurate, yet computationally inexpensive. A machine learning model based on Gaussian Process Regression is created and trained on a database of binding energies of molecules collected from laboratory experiments presented in the literature. The molecules in the database are categorized by their features, such as mono- or multilayer coverage, binding surface, functional groups, valence electrons, and H-bond acceptors and donors. The performance of the model is assessed with five-fold and leave-one-molecule-out cross validation. Predictions are generally accurate, with differences between predicted and literature binding energies values of less than $\pm$20\%. The validated model is used to predict the binding energies of twenty one molecules that have recently been detected in the interstellar medium, but for which binding energy values are not known. A simplified model is used to visualize where the snowlines of these molecules would be located in a protoplanetary disk. This work demonstrates that machine learning can be employed to accurately and rapidly predict binding energies of molecules. Machine learning complements current laboratory experiments and quantum chemical computational studies. The predicted binding energies will find use in the modelling of astrochemical and planet-forming environments.

astro-ph.GA↗

arXiv2022

Selectivity Trends and Role of Adsorbate-Adsorbate Interactions in CO Hydrogenation on Rhodium Catalysts

Predictive-quality computational modeling of heterogeneously catalyzed reactions has emerged as an important tool for the analysis and assessment of activity and activity trends. In contrast, more subtle selectivities and selectivity trends still pose a significant challenge to prevalent microkinetic modeling approaches that typically employ a mean-field approximation (MFA). Here, we focus on CO hydrogenation on Rh catalysts with the possible products methane, acetaldehyde, ethanol and water. This reaction has already been subject to a number of experimental and theoretical studies with conflicting views on the factors controlling activity and selectivity towards the more valuable higher oxygenates. Using accelerated first-principles kinetic Monte Carlo (KMC) simulations and explicitly and systematically accounting for adsorbate-adsorbate interactions through a cluster expansion approach, we model the reaction on the low-index Rh(111) and stepped Rh(211) surfaces. We find that the Rh(111) facet is selective towards methane, while the Rh(211) facet exhibits a similar selectivity towards methane and acetaldehyde. This is consistent with the experimental selectivity observed for larger, predominantly (111)-exposing Rh nanoparticles and resolves the discrepancy to earlier first-principles MFA microkinetic work that found the Rh(111) facet to be selective towards acetaldehyde. While the latter work tried to approximately account for lateral interactions through coverage-dependent rate expressions, our analysis demonstrates that this fails to sufficiently capture concomitant correlations among the adsorbed reaction intermediates that crucially determine the overall selectivity.

physics.chem-ph↗

arXiv2022

Predicting binding motifs of complex adsorbates using machine learning with a physics-inspired graph representation

Computational screening in heterogeneous catalysis relies increasingly on machine learning models for predicting key input parameters due to the high cost of computing these directly using first-principles methods. This becomes especially relevant when considering complex materials spaces, e.g. alloys, or complex reaction mechanisms with adsorbates that may exhibit bi- or higher-dentate adsorption motifs. Here we present a data-efficient approach to the prediction of binding motifs and associated adsorption enthalpies of complex adsorbates at transition metals (TMs) and their alloys based on a customized Wasserstein Weisfeiler-Lehman graph kernel and Gaussian Process Regression. The model shows good predictive performance, not only for the elemental TMs on which it was trained, but also for an alloy based on these TMs. Furthermore, incorporation of minimal new training data allows for predicting an out-of-domain TM. We believe the model may be useful in active learning approaches, for which we present an ensemble uncertainty estimation approach.

physics.chem-ph↗

arXiv2021

Interface between graphene and liquid Cu from molecular dynamics simulations

Controllable synthesis of defect-free graphene is crucial for applications since the properties of graphene are highly sensitive to any deviations from the crystalline lattice. We focus here on the emerging use of liquid Cu catalysts, which has high potential for fast and efficient industrial-scale production of high-quality graphene. The interface between graphene and liquid Cu is studied using force field and ab initio molecular dynamics, revealing a complete or partial embedding of finite-sized flakes. By analyzing flakes of different sizes we find that the size-dependence of the embedding can be rationalized based on the energy cost of embedding versus bending the graphene flake. The embedding itself is driven by the formation of covalent bonds between the under-coordinated edge C atoms and the liquid Cu surface, which is accompanied by a significant charge transfer. In contrast, the central flake atoms are located around or slightly above 3 {\AA} from the liquid Cu surface and exhibit weak vdW-bonding and much lower charge transfer. The structural and electronic properties of the embedded state revealed in our work provides the atomic-scale information needed to develop effective models to explain the special growth observed in experiments where various interesting phenomena such as flake self-assembly and rotational alignment, high growth speeds and low defect densities in the final graphene product have been observed.

cond-mat.mtrl-sci↗

arXiv2019

Ab initio thermodynamics of hydrocarbons relevant to graphene growth at solid and liquid Cu surfaces

Using ab initio thermodynamics, the stability of a wide range of hydrocarbon adsorbates under various chemical vapor deposition (CVD) conditions (temperature, methane and hydrogen pressures) used in experimental graphene growth protocols at solid and liquid Cu surfaces has been explored. At the employed high growth temperatures around the melting point of Cu, we find that commonly used thermodynamic models such as the harmonic oscillator model may no longer be accurate. Instead, we account for the translational and rotational mobility of adsorbates using a recently developed hindered translator and rotator model or a two-dimensional ideal gas model. The thermodynamic considerations turn out to be crucial for explaining experimental results and allow us to improve and extend the findings of earlier theoretical studies regarding the role of hydrogen and hydrocarbon species in CVD. In particular, we find that smaller hydrocarbons will completely dehydrogenate under most CVD conditions. For larger clusters our results show that metal-terminated and hydrogen-terminated edges have very similar stabilities. While both cluster types might thus form during the experiment, we show that the low binding strength of clusters with hydrogen-terminated edges could result in instability towards desorption.

cond-mat.mtrl-sci↗

arXiv2019

A Practical Guide to Surface Kinetic Monte Carlo Simulations

This review article is intended as a practical guide for newcomers to the field of kinetic Monte Carlo (KMC) simulations, and specifically to lattice KMC simulations as prevalently used for surface and interface applications. We will provide worked out examples using the kmos code, where we highlight the central approximations made in implementing a KMC model as well as possible pitfalls. This includes the mapping of the problem onto a lattice and the derivation of rate constant expressions for various elementary processes. Example KMC models will be presented within the application areas surface diffusion, crystal growth and heterogeneous catalysis, covering both transient and steady-state kinetics as well as the preparation of various initial states of the system. We highlight the sensitivity of KMC models to the elementary processes included, as well as to possible errors in the rate constants. For catalysis models in particular, a recurrent challenge is the occurrence of processes at very different timescales, e.g. fast diffusion processes and slow chemical reactions. We demonstrate how to overcome this timescale disparity problem using recently developed acceleration algorithms. Finally, we will discuss how to account for lateral interactions between the species adsorbed to the lattice, which can play an important role in all application areas covered here.

physics.comp-ph↗

arXiv2019

Beyond scaling relations for the description of catalytic materials

Computational screening for new and improved catalyst materials relies on accurate and low-cost predictions of key parameters such as adsorption energies. Here, we use recently developed compressed sensing methods to identify descriptors whose predictive power extends over a wide range of adsorbates, multi-metallic transition metal surfaces and facets. The descriptors are expressed as non-linear functions of intrinsic properties of the clean catalyst surface, e.g. coordination numbers, d-band moments and density of states at the Fermi level. From a single density-functional theory calculation of these properties, we predict adsorption energies at all potential surface sites, and thereby also the most stable geometry. Compared to previous approaches such as scaling relations, we find our approach to be both more general and more accurate for the prediction of adsorption energies on alloys with mixed-metal surfaces, already when based on training data including only pure metals. This accuracy can be systematically improved by adding also alloy adsorption energies to the training data.

cond-mat.mtrl-sci↗

arXiv2015

In Situ Detection of Active Edge Sites in Single-Layer MoS$_2$ Catalysts

MoS2 nanoparticles are proven catalysts for processes such as hydrodesulphurization and hydrogen evolution, but unravelling their atomic-scale structure under catalytic working conditions has remained significantly challenging. Ambient pressure X-ray Photoelectron Spectroscopy (AP-XPS) allows us to follow in-situ the formation of the catalytically relevant MoS2 edge sites in their active state. The XPS fingerprint is described by independent contributions to the Mo3d core level spectrum whose relative intensity is sensitive to the thermodynamic conditions. Density Functional Theory (DFT) is used to model the triangular MoS2 particles on Au(111) and identify the particular sulphidation state of the edge sites. A consistent picture emerges in which the core level shifts for the edge Mo atoms evolve counter-intuitively towards higher binding energies when the active edges are reduced. The shift is explained by a surprising alteration in the metallic character of the edge sites, which is a distinct spectroscopic signature of the MoS2 edges under working conditions.

cond-mat.mtrl-sci↗

arXiv2014

High Crystallinity and Decoupling of Graphene on a Metal: Reduced Coulomb Screening and Tunable pn-Junctions

High quality epitaxial graphene films can be applied as templates for tailoring graphene-substrate interfaces that allow for precise control of the charge carrier behavior in graphene through doping and many-body effects. By combining scanning tunneling microscopy, angle-resolved photoemission spectroscopy and density functional theory we demonstrate that oxygen intercalated epitaxial graphene on Ir(111) has high structural quality, is quasi free-standing, and shows signatures of many-body interactions. Using this system as a template, we show that tunable pn-junctions can be patterned by adsorption and intercalation of rubidium, and that the n-doped graphene regions exhibit a reduced Coulomb screening via enhanced electron-plasmon coupling. These findings are central for understanding and tailoring the properties of graphene-metal contacts e.g. for realizing quantum tunneling devices.

cond-mat.mtrl-sci↗

arXiv2011

Structure and stability of small H clusters on graphene

The structure and stability of small hydrogen clusters adsorbed on graphene is studied by means of Density Functional Theory (DFT) calculations. Clusters containing up to six H atoms are investigated systematically -- the clusters having either all H atoms on one side of the graphene sheet (\textit{cis}-clusters) or having the H atoms on both sides in an alternating manner (\textit{trans}-cluster). The most stable cis-clusters found have H atoms in ortho- and para-positions with respect to each other (two H's on neighboring or diagonally opposite carbon positions within one carbon hexagon) while the most stable trans-clusters found have H atoms in ortho-trans-positions with respect to each other (two H's on neighboring carbon positions, but on opposite sides of the graphene). Very stable trans-clusters with 13-22 H atoms were identified by optimizing the number of H atoms in ortho-trans-positions and thereby the number of closed, H-covered carbon hexagons. For the cis-clusters, the associative H$_2$ desorption was investigated. Generally, the desorption with the lowest activation energy proceeds via para-cis-dimer states, i.e.\ involving somewhere in the H clusters two H atoms that are positioned on opposite sites within one carbon hexagon. H$_2$ desorption from clusters lacking such H pairs is calculated to occur via hydrogen diffusion causing the formation of para-cis-dimer states. Studying the diffusion events showed a strong dependence of the diffusion energy barriers on the reaction energies and a general odd-even dependence on the number of H atoms in the cis-clusters.

cond-mat.mes-hall↗