Algorithmic Differentiation for Plane-Wave Density Functional Theory

Learn about algorithmic differentiation techniques for plane-wave density-functional theory in this 47-minute conference talk from IPAM's Bridging Scales from Atomistic to Continuum in Electrochemical Systems Workshop. Explore how reliable algorithmic differentiation methods can advance inverse design of materials and functionals while propagating uncertainties from functionals to DFT quantities of interest. Discover the recently developed differentiation framework for plane-wave density-functional theory that combines algorithmic differentiation (AD) and density-functional perturbation theory (DFPT) within the Density-functional ToolKit (DFTK). Examine the AD-DFPT framework that enables accurate computation of derivatives for any DFT output quantity with respect to any input parameter including geometry, density functional, or pseudopotential without manually deriving gradient expressions. Follow hands-on examples demonstrating these capabilities for inverse design of semiconductor band gaps, learning exchange-correlation functional parameters, and propagating DFT parameter uncertainties to relaxed structures. Gain insights from research conducted at École Polytechnique Fédérale de Lausanne (EPFL) on bridging computational scales in electrochemical systems through advanced differentiation techniques.