(729a) A Twenty-Five Year Perspective on Differences and Possibilities in the Practical Adoption of Molecular Simulations Vs Computational Quantum Chemistry

Computational quantum chemistry has become a practical tool, widely for generating thermochemistry and electronic properties of materials. Molecular simulation has been used selectively in industry, largely for qualitative insights.

I will explore and contrast applications over the past 25 years since AIChE hosted a plenary session for six invited industrial speakers that Paul Mathias, Ken Cox, and I organized for the 1994 San Francisco Annual Meeting (see this link for more pre-CoMSEF details). Additional applications and comparisons are available in the 2002 report International Comparative Study on Applying Molecular and Materials Modeling, the 2009 report International Assessment of Research and Development in Simulation-Based Engineering and Science, and presentations at this meeting in the CoMSEF session "Industrial Applications of Computational Chemistry and Molecular Simulation."

These experiences suggest that limited routine use of molecular simulations can be understood better through the three elements of cyberinfrastructure - hardware, software, and people:

• Many-electron, many-atom, many-molecule, and many-mer modeling efforts have benefitted greatly from hardware advances in computing speed, in memory and storage, and in visualization.
• Software packages for computational quantum chemistry have been deployed widely and have high ease-of-use; most are commercial codes although some are open-source. In molecular simulations, excellent software and force fields have long been publicly available.
• Perhaps the weakest aspect has been in people. As Ken Cox once noted, the people goals for software have shifted from trying just to be "foolproof and user-friendly" to needing to be "fool-friendly and user-proof." Many people want deliverables quickly but resist learning enough to generate them. Use of molecular simulation still seems limited to experts, and results are often qualitative rather than quantitative. In contrast, computational quantum chemistry has become an easily used tool for generating certain properties accurately with useful precision.

For molecular simulations, achieving broader use will require software capabilities that easily and reliably deliver quantitative, needed results of specific value. I will remark on several trends that promise to help, including advances in raw and specialized computing power; computational material science; Reactive Molecular Dynamics; Energy-Minimization Multi-Scale methods; and data science and machine learning.