Research

My research focuses on the development of electronic structure theory, computational methods, and massively parallel open source scientific software for first principles simulations of materials. I am fascinated by the elegant yet intricate mathematical structure underlying electronic structure theory and its computational methods. Building on these formalisms, my work develops new theories and methods that extend the methodological scope of first principles electronic structure calculations. I also seek to incorporate into electronic structure calculations physical conditions that are essential to experiments but difficult to accommodate in existing approaches, such as constant electrode potentials.

Along these lines, my work has led to the following developments.

Advanced electronic structure theory and methods

  • A rotationally invariant density matrix penalization method was proposed for inverse Kohn Sham DFT. The method is applicable to open shell systems and addresses the instability, low accuracy, and slow convergence often encountered in Kohn Sham matrix reconstruction for molecular and condensed phase systems in Gaussian basis sets.

  • Tensorial subspace DFT+U+J with full analytical force terms was developed in the GPW framework and implemented in CP2K. The existing Löwdin subspace DFT+U framework was also extended to DFT+U+J with full analytical force terms. For both implementations, minimum tracking linear response methods were implemented in CP2K for first principles calculations of Hubbard U and Hund J parameters.

  • The SOCP method, based on subspace occupancy constraining potentials, was proposed to provide a controllable and robust route to stabilizing target polaron configurations in first principles calculations.

Electrochemical interfaces under realistic conditions

  • Grand canonical constant potential methods were implemented in CP2K in a numerically stable manner for electrochemical systems, including constant potential SCF, planar counter charge models, improved Pulay density mixing, and potential energy surface swap first principles molecular dynamics.

  • An anisotropic interface continuum solvation method based on dielectric tensors rather than scalar dielectric functions was proposed for interfaces between solids and liquids. The corresponding CP2K module was developed, together with a finite element parallel solver based on FEniCSx for anisotropic tensorial Poisson equations and its interface to CP2K through direct memory data transfer.

  • The solvent aware SCCS continuum solvation model was implemented in CP2K to improve the definition of the solute solvent interface and to address convergence difficulties often encountered in SCCS calculations.