Monday 8:30 AM

February 28, 2022

Room: 255A

Location: Anaheim Convention Center

Ferroelectric materials are of considerable interest for memory and logic device applications such as ferroelectric random-access memory (FRAM), ferroelectric field-effect transistors (FeFET) and negative-capacitance field-effect transistors (NC-FET). While much effort has been devoted to ferroelectric perovskites, it is increasingly recognized that a ferroelectric orthorhombic form of HfO

Atomic cluster expansion of sp- and d-element systems. The atomic cluster expansion (ACE) [1] provides a generalisation of the lattice cluster expansion [2] to arbitrary degrees of freedom, including continuous variables[3]. As the expansion is general, it is applicable to different materials and I will give examples for sp-valent semiconductors and d-valent transition metals, including magnetic iron. I will further discuss accuracy and computational expense of ACE and compare to other methods[4]. The efficiency of ACE means that phase diagrams can be obtained and I will present automated workflows for phase diagram computation. Finally, I will show that ACE can be fitted without user interference, so that parameterisations for many different materials can be obtained quickly. [1] Drautz, Phys. Rev. B 99 (2019) 014104 [2] Sanchez, Ducastelle, Gratias, Physica A 128(1984)334 [3] Drautz, Phys. Rev. B 102(2020)024104 [4] Lysogorskiy et al., npj Comput. Mater. 7(2021)97

Cluster expansion effective Hamiltonians provide a computationally efficient link between electronic structure calculations and phenomenological descriptions of crystalline materials. The open source statistical mechanics software package CASM enables construction, parameterization, and application of cluster expansion effective Hamiltonians of mixed discrete and continuous degrees of freedom. This allows for (kinetic) Monte Carlo calculations to predict finite temperature thermodynamic and kinetic properties of multi-component alloys, including treatment of strain, vibrational, and magnetic effects. Here we will review the formalism CASM uses to construct cluster expansion effective Hamiltonians of mixed discrete and continuous degrees of freedom, describe the methods implemented by CASM to parameterize and apply them, and survey its application for property prediction in a wide variety of materials systems.

Magnetic exchange hardening, a phenomenon that increases a material's resistance to demagnetization, typically occurs in two-phase mixtures consisting of a ferromagnetic phase and an antiferromagnetic phase. We demonstrate the unique case of exchange hardening in a single, chemically-disordered full-Heusler alloy: Mn(1-x)Fe(x)Ru2Sn. Using nanoscale ab initio calculations, we parameterized a chemomagnetic cluster expansion Hamiltonian suitable for Monte Carlo simulations containing thousands of interacting sites and capable of describing bulk magnetic and chemical behavior. In our simulations, we find that nanoscale fluctuations in composition (as expected in any disordered phase) result in antiferromagnetic pockets in a bulk ferromagnetic system, complete with separate Neel and Curie temperatures of (anti)ferromagnetic ordering that match experimental observations. I will also briefly discuss how similarly simple ab initio calculations on small molecular systems can be used to parameterize accurate force fields, which can be used in molecular dynamics simulations to link the nanoscale and the mesoscale.

Interatomic Potentials have long been used for atomistic modeling where the interesting questions are out of reach by first-principles approaches. Traditional empirical potentials are typically fitted to experimental data. They typically have poor general accuracy but are physically well-behaved. On the other hand, machine-learned interatomic potentials are far more expressive than physically motivated interatomic potentials like Lennard-Jones, Stillinger-Weber, Embedded Atom Potentials, etc., but they are also more likely to be completely wrong outside of the training domain, are more difficult to train reliably, and are computationally expensive. We have developed MLIPs for the Hf-Ni-Ti shape memory alloy. We share cautionary tales, best practices for generating training sets, and demonstrate how community tools make for "easy entry" to realistic thermodynamic modeling with these potentials.