Recent Advances in First-Principles Based Molecular Dynamics
Dynamics
DOI:
10.1021/acs.accounts.1c00503
Publication Date:
2022-01-14T00:41:12Z
AUTHORS (4)
ABSTRACT
ConspectusFirst-principles molecular dynamics (FPMD) and its quantum mechanical-molecular mechanical (QM/MM) extensions are powerful tools to follow the real-time of a broad variety systems in their ground as well electronically excited states. The continued advances computational power have enabled simulations QM regions larger sizes for more extended time scales. In addition, development parallel algorithms has boosted performance QM/MM methods even on existing computer architectures. case density functional-based FPMD, several hundreds thousands atoms can now be customarily simulated tens picoseconds. spite this progress, scale limitations remain severe, especially when high-rung exchange-correlation functionals or high-level wave function based used. To ameliorate this, large number enhanced sampling been introduced but most approaches that developed increase efficiency FPMD sacrifice favor enhancing sampling. Here, we present some recent boosting while keeping full dynamic information. These include highly efficient implementation FPMD-based not only enables fully flexible combinations different electronic structure force fields via communication library, it also exploits parallelism both classical descriptions. second type acceleration discuss is family specially devised multiple-time-step make use suitable breakups total nuclear forces into fast components calculated lower level slowly varying correction evaluated with method at long intervals. gain scheme mostly depends cost difference between two advantageous yield speedups without compromising accuracy method. And finally, third class presented here machine learning models accelerated techniques. combination all substantial orders magnitude preserving accuracy.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (78)
CITATIONS (46)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....