Monday 8:30 AM

February 28, 2022

Room: 253A

Location: Anaheim Convention Center

Massively parallel supercomputers have become increasingly available for computational materials research. Hence, research codes need to be modernized or planned with this type of hardware in mind. I will discuss two examples in the context of theoretical spectroscopy for excited-electronic properties and femtosecond real-time dynamics: We integrated the modern, parallel ChASE eigensolver into an existing first-principles code for solving the Bethe-Salpeter equation for the optical polarization function, and implemented parallel reading routines. Based on numerical tests we show remarkable speedup and scaling on multi- and many-core architectures. This extends the lifetime of the code and enables studies of previously inaccessible complex materials. The second example focuses on numerical time stepping of a plane-wave expansion of the time-dependent Kohn-Sham equations. After interfacing Qb@ll with the PETSc library, we compare explicit time stepping algorithms regarding numerical error and efficiency. Machine learning is discussed as a possibility to circumvent expensive explicit time stepping.

Quantum chemistry is one of the most promising near-term applications of quantum computers. Quantum algorithms such as variational quantum eigen solver (VQE) and variational quantum deflation (VQD) algorithms have been mainly applied for molecular systems and there is a need to implement such methods for periodic solids. Using Wannier tight-binding Hamiltonian (WTBH) approaches, we demonstrate the application of VQE and VQD to accurately predict both electronic and phonon bandstructure properties of several elemental as well as multi-component solid-state materials. We apply VQE-VQD calculations for 307 spin-orbit coupling based electronic WTBHs and 933 finite-difference based phonon WTBHs. We establish a workflow for using VQD with lattice Greens function that can be used for solving dynamical mean-field theory problems. The WTBH model solvers can be used for testing other quantum algorithms also.

Charged defects have historically represented a problem for plane-wave basis DFT, as the periodic boundary conditions inherent to their design have been known to cause un-physical electrostatic effects, which are only magnified when studying low dimensional materials. This difficulty has been managed with great success by post-hoc energy corrections. However, the corrections were not performed self-consistently, which creates an issue. The charged system will have relaxed under un-physical electrostatic conditions, possibly resulting in an incorrect structure. To account for this, a newer self-consistent method was recently developed, but still requires independent verification against more well known corrections. As such, the newer method will be applied to charged and defective single and multi-layer structures of 2D materials to study how the self-consistent handling of the electrostatics affects the results of the calculations.

Molecular crystal packing can dramatically impact physical properties in pharmaceuticals and other organic materials. The ability to predict which crystal structures can occur, the thermodynamic conditions under which they occur, and their properties can be very helpful. However, such predictions are theoretically challenging due to the small 1-2 kJ/mol energy differences between crystal polymorphs. Major progress in the field has been made in recent years thanks to dispersion-corrected density functional theory (DFT) models, but the limitations of widely-used functionals are becoming increasingly apparent. We will discuss computationally affordable strategies for overcoming these limitations. Second, phonon contributions are also important when computing thermochemical properties, but doing so with DFT can be computationally demanding. An approximation for obtaining accurate phonon densities of states by combining DFT and density functional tight-binding (DFTB) models will also be presented. Together, these approaches allow more accurate prediction of the crystal structure energy landscapes.

The addition of new chemical species to an existing bulk metal significantly changes its properties, allowing for alloys to be designed to attain specific properties useful for advanced engineering systems. However, simulation of segregation behavior in metallic alloys is still computationally expensive. By analyzing atomic sites preferred for segregation as a function of the instantaneous local structure using a new descriptor framework known as Strain Functional Descriptors (SFDs), an improved model is implemented for analyzing alloy systems. Since this technique only relies upon instantaneous snapshots of the local atomic environments, computationally complex iteration over each site in the structure and/or Monte-Carlo methods can be bypassed. These descriptors also improve our understanding of specific relationships between atomic environments and their underlying physics when implemented alongside modern machine learning techniques. Extension of these techniques beyond the dilute limit will be addressed.

The diffusion of species at grain boundaries can plan an important role in numerous materials processes. Because each grain boundary can be unique and bicrystals only allow diffusion measurements of a single grain boundary at a time, this complicates the acquisition of data. We develop a method to calculate the full hydrogen diffusivity tensor for each unique grain boundary region between two grains in nickel polycrystals using molecular dynamics simulations. This method allows dozens of general-type grain boundaries to be examined in a single simulation. We examine two different grain sizes and discuss the overall effectiveness of the method.