Density Functional Theory for High-Throughput Screening of Chemicals and Materials
Recent developments in physical and computer sciences enable quantitative predictions of chemical reactions and thermodynamic data from first principles by multiscale modeling. The hierarchical approach integrates different theoretical frameworks ranging from those describing phenomena at the electronic length and time scales to those pertinent to complex biomolecular systems and macroscopic phase transitions, promising broad applications to problems of practical concern. Whereas multiscale modeling has been emerging as a popular computational tool for engineering applications, the connection between calculations at different scales is far from being coherent, and the multiple choices of quantum/classical methods at each scale renders numerous combinations that have been rarely calibrated against extensive experimental data.
During my PhD studying, on one hand, we try to build a reliable multiscale procedure for predicting the solvation free energies of a large set of small molecules in liquid water at ambient conditions. Using the experimental data for the hydration free energies as the benchmark, we find that the theoretical results are sensitive to the selection of quantum-mechanical methods for determining atomic charges and solute configurations, the assignment of the force-field parameters in particular the atomic partial charges, and approximations in the statistical-mechanical calculations.
On the other hand, we have investigated four representative versions of non-local density functionals for predicting gas adsorption in both slit pore model and a large library of metal-organic frameworks (MOFs) under a broad range of temperatures and pressures, compared with Monte Carlo simulation data. Overall all four versions of DFT are reasonably accurate in comparison with the simulation results, but for each specific interested condition, there is a best candidate. Besides, DFT is able to generate density profiles revealing microscopic details such as favorable adsorption sites and self-diffusion coefficient by entropy scaling method.
From a computational perspective, the DFT calculation is at least one order of magnitude faster than conventional simulation methods, rendering it as a versatile and promising tool for large-scale screening of chemicals and nanostructured materials.