arXiv subjects

Fabian L. Thiemann

Publications and source records attributed to Fabian L. Thiemann.

14 recordsLinked to original sources

arXiv2026

Microscopic origin of polytype-dependent melting in SiC revealed by machine-learning molecular dynamics

Predicting how crystal structure influences high-temperature stability remains a key challenge in materials modelling and design. Silicon carbide (SiC), one of the most thermally and chemically stable materials known, provides an ideal system for studying this problem because its many polytypes preserve similar local tetrahedral bonding while differing in long-range stacking geometry. Here, we combine phase-coexistence machine-learning molecular dynamics with finite-temperature phonon analysis, enabled by a fine-tuned MACE interatomic potential that accurately describes crystalline, high-temperature, and disordered configurations across multiple SiC polytypes. We identify a clear relative stability ordering, 3C > 2H > 9R, reflected consistently in structural disordering, interlayer sliding, and finite-temperature phonon spectra. Across all polytypes, melting initiates through the formation of short C-C contacts and carbon-rich local regions, followed by a progressive loss of tetrahedral Si-C connectivity. The reduced stability of the long-period 9R polytype is traced to low-frequency transverse-acoustic shear modes associated with relative bilayer sliding, which are already present in the 0 K phonon spectra and soften further at high temperature. These modes generate larger lateral bilayer displacements, linking enhanced interlayer sliding to local chemical disordering and ultimately melting. More broadly, our results show that high-temperature stability in polytypic covalent materials is governed not only by local bond strength, but also by stacking-dependent transverse dynamics.

cond-mat.mtrl-sci↗

arXiv2026

Strain-Dependent Wetting of Graphene

Understanding how water wets graphene is critical for predicting and controlling its behaviour in nanofluidic, sensing, and energy applications. A key measure of wetting is the contact angle made by a liquid droplet against the surface, yet experimental measurements for graphene span a wide range, with no consensus for free-standing graphene. Here, we use a machine learning potential with ab initio accuracy to provide an atomistic first-principles prediction for this unsolved problem, finding a weakly hydrophilic contact angle of $72.1 \pm 1.5 °$. More importantly, we unveil that graphene's wetting properties are highly sensitive to mechanical strain: tensile strain makes graphene significantly less hydrophilic, while compressive strain induces coherent ripples around the droplet, resulting in pronounced anisotropic wetting and contact angle hysteresis. We show that there is a strong coupling between the three-phase contact line and the intrinsic thermal ripples of free-standing graphene, which contributes to this strain sensitivity. Our results demonstrate that the wettability of 2D membranes are governed not only by their chemistry but also by their dynamic morphology, introducing a new class of wetting behaviour unique to atomically thin materials that offers an additional explanation for variability in experimental measurements. These findings reveal that mechanical strain may be a practical route to controlling wetting in 2D nanomaterials-based technologies, with promising consequences for nanofluidic and nano-filtration applications.

cond-mat.mes-hall↗

arXiv2026

Better without U: Impact of Selective Hubbard U Correction on Foundational MLIPs

The training of foundational machine learning interatomic potentials (fMLIPs) relies on diverse databases with energies and forces calculated using ab initio methods. We show that fMLIPs trained on large datasets such as MPtrj, Alexandria, and OMat24 encode inconsistencies from the Materials Project's selective use of the Hubbard U correction, which is applied to certain transition metals only if O or F atoms are present in the simulation cell. This inconsistent use of +U creates two incompatible potential-energy surfaces (PES): a lower-energy GGA surface and a higher-energy GGA+U one. When trained on both, MLIPs interpolate between them, leading to systematic underbinding, or even spurious repulsion, between U-corrected metals and oxygen- or fluorine-containing species. Models such as MACE-OMAT and -MPA exhibit repulsion between U-corrected metals and their oxides, limiting their value for studying catalysis and oxidation. We link the severity of this pathology to the oxygen number density in U-corrected training configurations. This explains why OMAT-trained models are most affected and suggests the issue might worsen as expanding future datasets increasingly include configurations with low oxygen content, such as those generated through combinatorial exploration of multi-element or defect-containing systems. Our simple per-U-corrected-atom shift aligns PBE+U and PBE energies for identical structures, yielding a smoother PES compared to existing correction schemes, which target phase diagram accuracy. As a result, models trained on datasets with our shift applied exhibit smaller mean absolute errors for the adsorption energies of oxygen on U-corrected elemental slabs. Since datasets omitting +U entirely (e.g. MatPES, MP-ALOE) avoid these pathologies, we recommend excluding +U in future fMLIP datasets. For existing datasets, our post-hoc correction provides a low-cost improvement.

physics.chem-ph↗

arXiv2026

Equivariant Interatomic Potentials without Tensor Products

Foundational machine-learned interatomic potentials have emerged as powerful tools for atomistic simulations, promising near first-principles accuracy across diverse chemical spaces at a fraction of the cost of quantum-mechanical calculations. However, the most accurate equivariant architectures rely on Clebsch-Gordan tensor products whose computational cost scales steeply with angular resolution, creating a trade-off between model expressiveness and inference speed that ultimately limits practical applications. Here we introduce Geodite, an equivariant message-passing architecture that replaces tensor products while incorporating physical priors to ensure smooth, well-behaved potential energy surfaces. Trained on the Materials Project trajectories dataset of inorganic crystals, Geodite-MP achieves accuracy competitive with leading methods on benchmarks for materials stability prediction, thermal conductivity, phonon-derived properties, and nanosecond-scale molecular dynamics, while running $3\text{--}5\times$ faster than models performing similarly. By combining predictive accuracy, computational efficiency, and physicality, Geodite enables faster large-scale atomistic simulations and high-throughput screening that would otherwise be computationally prohibitive.

physics.comp-ph↗

arXiv2025

Computational tuning of the elastic properties of low- and high-entropy ultra-high temperature ceramics

Ultra-high temperature ceramics (UHTCs) represent a class of crystalline materials for extreme environments. They can withstand extremely high temperatures but are mechanically difficult to work with due to their inherent brittleness. Mixture compounds, in particular high-entropy mixtures, offer a pathway to tune the physical properties of UHTCs such as their elastic constants. Here we fine-tune the MACE-MPA-0 universal machine-learning potential on rocksalt carbide UHTCs containing group IV-V metals and demonstrate that not only do the elastic constants deviate from the rule of mixtures approximation in the high-entropy limit, but also in the low-entropy limit of binary and ternary mixtures. We find that this is caused by distortion imposed by the lattice mismatch, enabling the tuning of the physical properties of UHTC mixtures in both low- and high-entropy compounds. We identify a three-component mixture compound, HfCVCZrC, as the best balance between synthesizability and toughness, and apply our developed MACE-UHTC model to identify a range of non-equimolar candidate compositions of this compound which may enable the synthesis of a mixture UHTC with a Young's modulus up to 40 GPa below that of ZrC.

cond-mat.mtrl-sci↗

arXiv2025

Global properties of the energy landscape: a testing and training arena for machine learned potentials

Machine learning interatomic potentials (MLIPs) have achieved remarkable accuracy on standard benchmarks, yet their ability to reproduce molecular kinetics -- critical for reaction rate calculations -- remains largely unexplored. We introduce Landscape17, a dataset of complete kinetic transition networks (KTNs) for the molecules of the MD17 dataset, computed using hybrid-level density functional theory. Each KTN contains minima, transition states, and approximate steepest-descent paths, along with energies, forces, and Hessian eigenspectra at stationary points. We develop a comprehensive test suite to evaluate the MLIP ability to reproduce these reference landscapes and apply it to a number of state-of-the-art architectures. Our results reveal limitations in current MLIPs: all the models considered miss over half of the DFT transition states and generate stable unphysical structures throughout the potential energy surface. Data augmentation with pathway configurations improves reproduction of DFT potential energy surfaces, resulting in significant improvement in the global kinetics. However, these models still produce many spurious stable structures, indicating that current MLIP architectures face underlying challenges in capturing the topology of molecular potential energy surfaces. The Landscape17 benchmark provides a straightforward but demanding test of MLIPs for kinetic applications, requiring only up to a few hours of compute time. We propose this test for validation of next-generation MLIPs targeting reaction discovery and rate prediction.

physics.chem-ph↗

arXiv2025

Force-Free Molecular Dynamics Through Autoregressive Equivariant Networks

Molecular dynamics (MD) simulations play a crucial role in scientific research. Yet their computational cost often limits the timescales and system sizes that can be explored. Most data-driven efforts have been focused on reducing the computational cost of accurate interatomic forces required for solving the equations of motion. Despite their success, however, these machine learning interatomic potentials (MLIPs) are still bound to small time-steps. In this work, we introduce TrajCast, a transferable and data-efficient framework based on autoregressive equivariant message passing networks that directly updates atomic positions and velocities lifting the constraints imposed by traditional numerical integration. We benchmark our framework across various systems, including a small molecule, crystalline material, and bulk liquid, demonstrating excellent agreement with reference MD simulations for structural, dynamical, and energetic properties. Depending on the system, TrajCast allows for forecast intervals up to $30\times$ larger than traditional MD time-steps, generating over 15 ns of trajectory data per day for a solid with more than 4,000 atoms. By enabling efficient large-scale simulations over extended timescales, TrajCast can accelerate materials discovery and explore physical phenomena beyond the reach of traditional simulations and experiments. An open-source implementation of TrajCast is accessible under https://github.com/IBM/trajcast.

physics.comp-ph↗

arXiv2025

Defects induce phase transition from dynamic to static rippling in graphene

Two-dimensional (2D) materials display nanoscale dynamic ripples that significantly impact their properties. Defects within the crystal lattice are the elementary building blocks to tailor the material's morphology. While some studies have explored the link between defective structures and rippling dynamics in 2D materials, a comprehensive understanding of this relationship has yet to be achieved. Here, we address this using machine learning-driven molecular dynamics simulations. Specifically, we find that above a critical concentration of defects, free-standing graphene sheets undergo a dynamic transition from freely propagating to static ripples. Our computational approach captures the dynamics with atomic resolution, and reveals that the transition is driven by elastic interactions between defects. The strength of these interactions is found to vary across defect types and we identify a unifying set of principles driving the dynamic-to-static transition in 2D materials. Our work not only rationalises puzzling experimental results for defective 2D materials, but also paves the way to design two-dimensional devices with tailored rippling dynamics. These insights could lay the foundations for a new class of disorder-based catalytic and interfacial materials.

cond-mat.mtrl-sci↗

arXiv2024

Introduction to machine learning potentials for atomistic simulations

Machine learning potentials have revolutionised the field of atomistic simulations in recent years and are becoming a mainstay in the toolbox of computational scientists. This paper aims to provide an overview and introduction into machine learning potentials and their practical application to scientific problems. We provide a systematic guide for developing machine learning potentials, reviewing chemical descriptors, regression models, data generation and validation approaches. We begin with an emphasis on the earlier generation of models, such as high-dimensional neural network potentials (HD-NNPs) and Gaussian approximation potential (GAP), to provide historical perspective and guide the reader towards the understanding of recent developments, which are discussed in detail thereafter. Furthermore, we refer to relevant expert reviews, open-source software, and practical examples - further lowering the barrier to exploring these methods. The paper ends with selected showcase examples, highlighting the capabilities of machine learning potentials and how they can be applied to push the boundaries in atomistic simulations.

physics.chem-ph↗

arXiv2024

On the increase of the melting temperature of water confined in one-dimensional nano-cavities

Water confined in nanoscale cavities plays a crucial role in everyday phenomena in geology and biology, as well as technological applications at the water-energy nexus. However, even understanding the basic properties of nano-confined water is extremely challenging for theory, simulations, and experiments. In particular, determining the melting temperature of quasi-one-dimensional ice polymorphs confined in carbon nanotubes has proven to be an exceptionally difficult task, with previous experimental and classical simulations approaches report values ranging from $\sim 180 \text{ K}$ up to $\sim 450 \text{ K}$ at ambient pressure. In this work, we use a machine learning potential that delivers first principles accuracy to study the phase diagram of water for confinement diameters $ 9.5 < d < 12.5 \text{ Å}$. We find that several distinct ice polymorphs melt in a surprisingly narrow range between $\sim 280 \text{ K}$ and $\sim 310 \text{ K}$, with a melting mechanism that depends on the nanotube diameter. These results shed new light on the melting of ice in one-dimension and have implications for the operating conditions of carbon-based filtration and desalination devices.

cond-mat.mtrl-sci↗

arXiv2023

Classical quantum friction at water-carbon interfaces

Friction at water-carbon interfaces remains a major puzzle with theories and simulations unable to explain experimental trends in nanoscale waterflow. A recent theoretical framework -- quantum friction (QF)-- proposes to resolve these experimental observations by considering nonadiabatic coupling between dielectric fluctuations in water and graphitic surfaces. Here, using a classical model that enables fine-tuning of the solid's dielectric spectrum, we provide evidence from simulations in general support of QF. In particular, as features in the solid's dielectric spectrum begin to overlap with water's librational and Debye modes, we find an increase in friction in line with that proposed by QF. At the microscopic level, we find that this contribution to friction manifests more distinctly in the dynamics of the solid's charge density than that of water. Our findings suggest that experimental signatures of QF may be more pronounced in the solid's response rather than liquid water's.

cond-mat.soft↗

arXiv2021

Defect-Dependent Corrugation in Graphene

Graphene's intrinsically corrugated and wrinkled topology fundamentally influences its electronic, mechanical, and chemical properties. Experimental techniques allow the manipulation of pristine graphene and the controlled production of defects which allows to control the atomic out-of-plane fluctuations and, thus, tune graphene's properties. Here, we perform large scale machine learning-driven molecular dynamics simulations to understand the impact of defects on the structure of graphene. We find that defects cause significantly higher corrugation leading to a strongly wrinkled surface. The magnitude of this structural transformation strongly depends on the defect concentration and specific type of defect. Analysing the atomic neighborhood of the defects reveals that the extent of these morphological changes depends on the preferred geometrical orientation and the interactions between defects. While our work highlights that defects can strongly affect graphene's morphology, it also emphasises the differences between distinct types by linking the global structure to the local environment of the defects.

cond-mat.mtrl-sci↗

arXiv2021

Machine learning potentials for complex aqueous systems made simple

Simulation techniques based on accurate and efficient representations of potential energy surfaces are urgently needed for the understanding of complex aqueous systems such as solid-liquid interfaces. Here, we present a machine learning framework that enables the efficient development and validation of models for complex aqueous systems. Instead of trying to deliver a globally-optimal machine learning potential, we propose to develop models applicable to specific thermodynamic state points in a simple and user-friendly process. After an initial ab initio simulation, a machine learning potential is constructed with minimum human effort through a data-driven active learning protocol. Such models can afterwards be applied in exhaustive simulations to provide reliable answers for the scientific question at hand. We showcase this methodology on a diverse set of aqueous systems with increasing degrees of complexity. The systems chosen here comprise bulk water with different ions in solution, water on a titanium dioxide surface, as well as water confined in nanotubes and between molybdenum disulfide sheets. Highlighting the accuracy of our approach with respect to the underlying ab initio reference, the resulting models are evaluated in detail with an automated validation protocol that includes structural and dynamical properties and the precision of the force prediction of the models. Finally, we demonstrate the capabilities of our approach for the description of water on the rutile titanium dioxide (110) surface to analyze the structure and mobility of water on this surface. Such machine learning models provide a straightforward and uncomplicated but accurate extension of simulation time and length scales for complex systems.

physics.chem-ph↗

arXiv2020

Machine Learning Potential for Hexagonal Boron Nitride Applied to Thermally and Mechanically Induced Rippling

We introduce an interatomic potential for hexagonal boron nitride (hBN) based on the Gaussian approximation potential (GAP) machine learning methodology. The potential is based on a training set of configurations collected from density functional theory (DFT) simulations and is capable of treating bulk and multilayer hBN as well as nanotubes of arbitrary chirality. The developed force field faithfully reproduces the potential energy surface predicted by DFT while improving the efficiency by several orders of magnitude. We test our potential by comparing formation energies, geometrical properties, phonon dispersion spectra and mechanical properties with respect to benchmark DFT calculations and experiments. In addition, we use our model and a recently developed graphene-GAP to analyse and compare thermally and mechanically induced rippling in large scale two-dimensional (2D) hBN and graphene. Both materials show almost identical scaling behaviour with an exponent of $η\approx 0.85$ for the height fluctuations agreeing well with the theory of flexible membranes. Based on its lower resistance to bending, however, hBN experiences slightly larger out-of-plane deviations both at zero and finite applied external strain. Upon compression a phase transition from incoherent ripple motion to soliton-ripples is observed for both materials. Our potential is freely available online at [http://www.libatoms.org].

cond-mat.mtrl-sci↗