arXiv subjects

Sergey V. Sukhomlinov

Publications and source records attributed to Sergey V. Sukhomlinov.

6 recordsLinked to original sources

arXiv2026

FlashCart: Fast Cartesian Tensor Products for Equivariant Interatomic Potentials

Machine-learned interatomic potentials extend atomistic simulations beyond the length- and timescales accessible to electronic-structure methods. However, the computational cost of equivariant architectures limits the local correlations they can represent in practice and therefore their achievable accuracy. Here we introduce FlashCart, which makes higher-order correlations affordable by combining generated GPU kernels with an architecture that recursively builds equivariant features and compresses them to a fixed width at each step. We express tensor products in independent Cartesian components and symbolically simplify them and their derivatives, producing fused kernels that often outperform optimized spherical counterparts. We then show that increasing correlation order improves accuracy more efficiently than increasing width, depth, or tensor rank. On SPICE-MACE-OFF, FlashCart models advance the measured accuracy-efficiency frontier: a model with $5.6$ million parameters achieves lower energy and force errors and $10\times$ faster inference than a transformer with $189$ million parameters.

cs.LG↗

arXiv2026

Granite sliding on granite: friction, wear rates, surface topography, and the scale-dependence of rate-state effects

We study tribological granite-granite contacts as a model for tectonic faulting, combining experiments, theory, and molecular dynamics simulations. The high friction in this system is not dominated by particulate wear or plowing, as frequently assumed, but by cold welding within plastically deformed asperity junctions. We base this conclusion on the observation that wear is repeatedly high after cleaning contacts but decreases as gouge accumulates, while friction shows the opposite trend. Moreover, adding water reduces wear by a factor of ten but barely decreases friction. Thermal and rate-dependent effects-central to most earthquake models-are negligible: friction remains unchanged between -40°C and 20°C, across abrupt velocity steps, and after hours of stationary contact. The absence of rate-state effects in our macroscopic samples is rationalized by the scale-dependence of pre-slip. The evolution of surface topography shows that quartz grains become locally smooth, with height spectra isotropic for wavelength below 10 microns but anisotropic at longer wavelengths, similar to natural faults. The resulting gouge particles have the usual characteristic sizes near 100 nm. Molecular dynamics simulations of a rigid, amorphous silica tip sliding on α-quartz reproduce not only similar friction coefficients near unity but also other experimentally observed features, including stress-introduced transitions to phases observed in post-mortem faults, as well as theoretical estimates of local flash temperatures. Additionally, they reveal a marked decrease of interfacial shear strength above 600°C.

physics.geo-ph↗

arXiv2025

Advanced Langevin thermostats: Properties, extensions to rheology, and a lean momentum-conserving approach

The Langevin equation accounts for unresolved bath degrees of freedom driving the system toward the bath temperature. Because of this, numerical solutions of the Langevin equation have a long history. Here, we recapitulate, combine, and extend existing Langevin-equation based thermostats, scrutinize their properties and demonstrate their superiority over global kinetic-energy controls. Our work includes compact, asymptotic-analysis based derivations of stochastic thermostats, including the highly accurate Grønbech-Jensen scheme. Proposed extensions include a precise, colored and a lean, momentum-conserving thermostat.

cond-mat.stat-mech↗

arXiv2025

Cold self-lubrication of sliding ice

The low kinetic friction between ice and numerous counterbodies is commonly attributed to an interfacial water layer, which is believed to originate from pre-existing surface water or from melt water induced by high contact pressures or frictional heat. However, even the currently leading theory of frictional melting appears to defy direct experimental verification. Here we present molecular simulations of ice interfaces that reveal that ice surfaces liquefy without melting thermodynamically but predominantly by cold, displacement-driven amorphization. Despite effective self-lubrication, very small ice friction is found to require water to slip past a hydrophobic counterface -- or an excess amount of water, produced by, e.g., extreme sliding velocities.

cond-mat.soft↗

arXiv2024

Brittleness of metallic glasses dictated by their state at the fragile-to-strong transition temperature

The effect of cooling on the brittleness of glasses in general, and bulk metallic glasses (BMGs) in particular, is usually studied with continuously varying cooling rates; slower cooling rates lead to stiffer, harder, and more brittle glasses than higher cooling rates. These protocols obscure any potential discontinuity that a glass might experience depending on whether its microstructure resembles that of a fragile or a strong glass-forming liquid. Here, we use large-scale molecular dynamics to simulate the nanoindentation behavior of model BMGs (Zr$_{0.6}$Cu$_{0.3}$Al$_{0.1}$) obtained by rapidly quenching equilibrium melts from temperatures above and below the fragile-to-strong transition temperature $T_\textrm{fst}$, leading to fragile and strong glasses, respectively. While the contact modulus deduced from the indentation simulation evolves smoothly with the temperature $T_\text{q}$ from which the melt is quenched, the plastic response changes quasi-discontinuously as $T_\text{q}$ passes through $T_\textrm{fst}$. In particular, strong glasses develop highly asymmetric flow profiles with mature shear bands, in contrast to fragile glasses. Quantitative differences reveal themselves not only through a formal von Mises localization parameter analysis but also through image analysis of flow patterns using pre-trained artificial intelligence models. Moreover, seemingly erratic flow profiles for our indentation geometry produced surprisingly reproducible and, thus, deterministic features. It remains to be determined to what extent other classes of glass formers follow our observation that the degree of brittleness is significantly influenced by whether the melt is fragile or strong when it falls out of equilibrium at the glass transition temperature.

cond-mat.mtrl-sci↗

arXiv2022

Interatomic potentials: Achievements and challenges

Interatomic potentials approximate the potential energy of atoms as a function of their coordinates. Their main application is the effective simulation of many-atom systems. Here, we review empirical interatomic potentials designed to reproduce elastic properties, defect energies, bond breaking, bond formation, and even redox reactions. We discuss popular two-body potentials, embedded-atom models for metals, bond-order potentials for covalently bonded systems, polarizable potentials including charge-transfer approaches for ionic systems and quantum-Drude oscillator models mimicking higher-order and many-body dispersion. Particular emphasis is laid on the question what constraints ensue from the functional form of a potential, e.g., in what way Cauchy relations for elastic tensor elements can be violated and what this entails for the ratio of defect and cohesive energies, or why the ratio of boiling to melting temperature tends to be large for potentials describing metals but small for short-ranged pair potentials. The review is meant to be pedagogical rather than encyclopedic. This is why we highlight potentials with functional forms sufficiently simple to remain amenable to analytical treatments. Our main objective is to provide a stimulus for how existing approaches can be advanced or meaningfully combined to extent the scope of simulations based on empirical potentials.

cond-mat.mtrl-sci↗