Robust and effective ab initio molecular dynamics simulations on the GPU cloud infrastructure using the Schrödinger Materials Science Suite
Authors:
Alexandr Fonari,
Garvit Agarwal,
Subodh C. Tiwari,
Casey N. Brock,
Jacob Gavartin,
Mathew D. Halls
Abstract:
Ab initio Born-Oppenheimer molecular dynamics (AIMD) is a valuable method for simulating physico-chemical processes of complex systems, including reactive systems, and for training machine learning models and force fields. Speed and stability issues on traditional hardware preclude routine AIMD simulations for larger systems and longer timescales. We postulate that any practically useful AIMD simu…
▽ More
Ab initio Born-Oppenheimer molecular dynamics (AIMD) is a valuable method for simulating physico-chemical processes of complex systems, including reactive systems, and for training machine learning models and force fields. Speed and stability issues on traditional hardware preclude routine AIMD simulations for larger systems and longer timescales. We postulate that any practically useful AIMD simulation must generate a trajectory of a minimum 1000 MD steps a day on a moderate cloud resource. In this work, we implement a computing workflow that enables routine calculations at this throughput and demonstrate results for several non-trivial atomistic dynamical systems. In particular, we have employed the GPU implementation of the Quantum ESPRESSO code which we will show increases AIMD productivity compared to the CPU version. In order to take advantage of transient servers (which are more cost and energy effective compared to the stable servers), we have implemented automatic restart/continuation of the AIMD runs within the Schrödinger Materials Science Suite. Finally, to reduce simulation size and thus reduce compute time when modeling surfaces, we have implemented a wall potential constraint. Our benchmarks using several reactive systems (lithium anode surface/solvent interface, hydrogen diffusion in an iron grain boundary) show a significant speed up when running on a GPU-enabled transient server using our updated implementation.
△ Less
Submitted 25 June, 2024;
originally announced June 2024.
How will quantum computers provide an industrially relevant computational advantage in quantum chemistry?
Authors:
V. E. Elfving,
B. W. Broer,
M. Webber,
J. Gavartin,
M. D. Halls,
K. P. Lorton,
A. Bochevarov
Abstract:
Numerous reports claim that quantum advantage, which should emerge as a direct consequence of the advent of quantum computers, will herald a new era of chemical research because it will enable scientists to perform the kinds of quantum chemical simulations that have not been possible before. Such simulations on quantum computers, promising a significantly greater accuracy and speed, are projected…
▽ More
Numerous reports claim that quantum advantage, which should emerge as a direct consequence of the advent of quantum computers, will herald a new era of chemical research because it will enable scientists to perform the kinds of quantum chemical simulations that have not been possible before. Such simulations on quantum computers, promising a significantly greater accuracy and speed, are projected to exert a great impact on the way we can probe reality, predict the outcomes of chemical experiments, and even drive design of drugs, catalysts, and materials. In this work we review the current status of quantum hardware and algorithm theory and examine whether such popular claims about quantum advantage are really going to be transformative. We go over subtle complications of quantum chemical research that tend to be overlooked in discussions involving quantum computers. We estimate quantum computer resources that will be required for performing calculations on quantum computers with chemical accuracy for several types of molecules. In particular, we directly compare the resources and timings associated with classical and quantum computers for the molecules H$_2$ for increasing basis set sizes, and Cr$_2$ for a variety of complete active spaces (CAS) within the scope of the CASCI and CASSCF methods. The results obtained for the chromium dimer enable us to estimate the size of the active space at which computations of non-dynamic correlation on a quantum computer should take less time than analogous computations on a classical computer. Using this result, we speculate on the types of chemical applications for which the use of quantum computers would be both beneficial and relevant to industrial applications in the short term.
△ Less
Submitted 25 September, 2020;
originally announced September 2020.
