arXiv subjects

Matthias Militzer

Publications and source records attributed to Matthias Militzer.

4 recordsLinked to original sources

arXiv2025

Scalable Data-Driven Basis Selection for Linear Machine Learning Interatomic Potentials

Machine learning interatomic potentials (MLIPs) provide an effective approach for accurately and efficiently modeling atomic interactions, expanding the capabilities of atomistic simulations to complex systems. However, a priori feature selection leads to high complexity, which can be detrimental to both computational cost and generalization, resulting in a need for hyperparameter tuning. We demonstrate the benefits of active set algorithms for automated data-driven feature selection. The proposed methods are implemented within the Atomic Cluster Expansion (ACE) framework. Computational tests conducted on a variety of benchmark datasets indicate that sparse ACE models consistently enhance computational efficiency, generalization accuracy and interpretability over dense ACE models. An added benefit of the proposed algorithms is that they produce entire paths of models with varying cost/accuracy ratio.

physics.comp-ph↗

arXiv2025

Atomistically informed phase field study of austenite grain growth

Atomistically-informed phase field simulations have been performed to investigate the effect of five common alloying elements (Nb, Ti, Mo, V, Mn) on austenite grain growth. The anisotropic simulations based on the segregation energy profiles of the solutes to four different grain boundary (GB) types from density functional theory calculations suggest a secondary role of solute drag anisotropy on grain growth. Hence, the solute trends are determined to be the same for all investigated GBs, and as a result, the $\Sigma 5(310)[001]$ GB can be considered as a representative GB for solute trend predictions. The decrease in grain growth rates due to solute additions is quantitatively described using a solute trend parameter. The following hierarchy of the solute's effectiveness to retard austenite grain growth has been determined based on the results of the presented model calculations in agreement with the experimental observations: Nb$>$Ti$>$Mo$>$V$\approx$Mn. The limitations and the strengths of the proposed approach are discussed in detail, and a potential application of this approach to steel design is proposed.

cond-mat.mtrl-sci↗

arXiv2025

Modeling solute-grain boundary interactions in a bcc Ti-Mo alloy using density functional theory

Solute segregation in alloys is a key phenomenon which affects various material characteristics such as embrittlement, grain growth and precipitation kinetics. In this work, the segregation energies of Y, Zr, and Nb to a \textgreek{S}5 grain boundary in a bcc Ti-25 at \% Mo alloy were determined using density functional theory (DFT) calculations. A systematic approach was laid out by computing the solution energy distributions in the bulk alloy using Warren-Cowley short-range order parameters to find a representative bulk-solute reference energy. Additionally, different scenarios were considered when a solute atom replaces different sites in terms of their local Ti-Mo chemistry at the GB plane to calculate the distribution of segregation energies. The solute segregation to a Mo site at the GB plane is preferred rather than to a Ti site. Further analysis shows that these segregation energy trends can be rationalized based on a primarily elastic interaction. Thus the segregation energies scale with the solute size such that Y has the largest segregation energies followed by Zr and Nb.

cond-mat.mtrl-sci↗

arXiv2013

A three-dimensional atomistic kinetic Monte Carlo study of dynamic solute-interface interaction

A three-dimensional atomistic Kinetic Monte Carlo model was developed and used to study the interaction between mobile solutes and a migrating interface. While the model was developed with a simplified energetic and topological description, it was also constructed to capture, in the absence of solute, the Burke-Turnbull model for interface migration and, in the presence of solutes, solute segregation to different types of interface sites. After parameterizing the model, simulations were performed to study the relationship between average interface velocity and imposed driving pressure for varying solute concentration and solute diffusivity. While the effect of solute concentration on solute drag pressure was found to be consistent with classical solute drag models, the effect of solute diffusivity was found to give a response not captured by either continuum or previously reported two-dimensional atomistic models. The dependence of maximum drag pressure on solute diffusivity was observed and attributed to the coupling between the structure of a migrating interface and the ability for solute to remain segregated to the interface.

cond-mat.mtrl-sci↗