Next Generation Methods for Advanced Condensed Phase Simulations in Q-Chem

点击翻译按钮获取中文摘要

PROJECT SUMMARY Next Generation Methods for Advanced Condensed Phase Simulations in Q-Chem Biophysical systems exist in the condensed phase, and that is the environment in which their properties should be computer-modeled. The correct theory to describe the electrons is using ab initio (AI) quantum mechanics (QM), whilst nuclear motion requires molecular dynamics (MD). The combination, AIMD, is thus the appropriate tool for biophysical simulations. While use of AIMD is vastly more expensive than MD with empirical potentials, it is nonetheless the standard to aspire to. AIMD enables correct treatment of bond-breaking for reactive processes, as well an accurate description of the non-bonded interactions that determine solvation and conformational preferences. This Phase II proposal has the objective of bringing a production level AIMD code to the Q-Chem software package. The key justification for the proposed work, and the potential value of the resulting product is that it will bring together capabilities that are not found jointly in any other AIMD code. The valuable synergy between the density functional theory implementation for periodic boundary conditions (DFT-PBC), and advanced algorithms for efficiently and accurate propagating the MD is the core innovation of this project. With regard to DFT-PBC (the first specific aim), the focus is on implementing high precision, high efficiency algorithms for the critical components of DFT with advanced functionals. Our code will support the latest meta-generalized gradient approximations (mGGAs), with inclusion of non-local van der Waals density functionals, that are not available in DFT-PBC codes to date. We will addition- ally provide support for range-separated exact exchange, with high efficiency. These capabilities will come with energies and gradients. Our software framework can also permit all-electron calculations as needed e.g. for NMR properties that depend on the electron density at the nucleus. Our modular code will support efficient on-node parallelism. To propagate MD efficiently and stably (the second specific aim), we employ two innovative statis- tical mechanics (SM) algorithms that have been proven in conventional MD, but are not yet available in any production AIMD code. First, we are extending the inertial extended Lagrangian self-consistent field (iEL/SCF) method to work robustly and efficiently with AIMD, building upon promising Phase I results, by combining it with a stochastic-isokinetic integration (SII) scheme to enable a single but larger MD time step. Second, we will explore the combination of iEL/SCF-SII with a multiple time- stepping method in which will explore whether different components of the QM force can be updated on different timescales in the AIMD. In final Aim 3 we test the combined DFT-PBC and iEL/SCF-SII capabilities on biophysical appli- cations including zwitterionic glycine and valine peptides in aqueous solution and molecular crystals. 1