A Robust Computational Framework for High-Throughput Density Functional Theory Calculations for Electrochemical Applications

High-throughput density functional theory (DFT) calculations are now commonly employed to screen materials for diverse applications and to populate databases of molecular properties. More recently, high-throughput DFT has been used to generate computational reaction networks for the analysis of reactive pathways. The automation of DFT calculations related to electrochemical systems has thus far been limited. Geometry optimization and SCF convergence are inherently more difficult when considering charged and open-shell molecules, coordinated metal ions, and solvated species, all of which are critically important for studies involving liquid-phase electrochemical environments such as those found in batteries. In addition, the chemical complexity of such species necessitates advanced levels of theory. We present a framework to automate the calculation of fully-optimized molecular geometries and molecular thermodynamic properties (enthalpy, entropy, and free energy) for molecules representing all the above challenges, including radical, charged, metal-coordinated, and solvated species. Our framework leverages the pymatgen,[1] fireworks,[2] custodian, and atomate[3] libraries to prepare, execute, and process electronic structure calculations using the Q-Chem DFT code.[4] Through benchmarking studies, we have identified appropriate methods for use in electrochemical environments, including an accurate density functional, basis set, and implicit solvent model. We have also implemented algorithms to correct common issues such as SCF convergence failures, geometric optimization to a non-minimal stationary point on the potential energy surface (a saddle point), and many more. In general, even when using an appropriate level of theory, 25% of calculations fail in some way without additional error correction. Our framework, using our suggested parameters and a suite of on-the-fly error handlers, converges to a minimum energy structure and calculates the molecular thermodynamic properties successfully in over 97% of cases. Individual calculations using our framework can be combined to perform complex workflows in a fully automated fashion. To date, we have applied this framework to over 25,000 unique molecules relevant to the formation of solid-electrolyte interphases in Li-ion and Mg-ion batteries. [1] S. P. Ong et al., Computational Materials Science, 68, 314–319 (2013). [2] A. Jain et al., Concurrency and Computation: Practice and Experience, 27(17), 5037-5059 (2015). [3] K. Mathew et al., Computational Materials Science, 139, 140-152 (2017). [4] Y. Shao et al., Molecular Physics, 113(2), 184-215 (2015). Figure 1