John A.W. Harkless
B.S.,1995, Chemistry and Mathematics, Morehouse College, Atlanta, GA
Ph.D., 2001, Theoretical Chemistry, University ofCalifornia, Berkeley, Berkeley, CA
National Research Council Postdoctoral Researcher, Computational Chemistry Group, NIST, 2001-2002
- Quantum Monte Carlo wavefunction development
- Electronic structure of metallic systems
We seek to reduce the "art" required by Quantum Monte Carlo yet retain high accuracy for "difficult" systems.
"Quantum Mechanics For Everyone”
By rewriting quantum mechanical problems in the language of statistics, we bridge the gap between experimental observations and computational predictions. Our techniques produce estimates of experimentally determined properties that are nearly indistinguishable in side-by-side comparisons. The reliability of the end result, and the similarity in presentation are a consequence of rendering quantum mechanics as a statistical and probabilistic process.
Quantum Monte Carlo (QMC) is a powerful quantum mechanical technique that is well suited for modern computing environments: it is algorithmically straightforward, embarrassingly parallel, and does not require specialized configurations of computing hardware. QMC renders the problem of electron correlation in terms of statistics and probability. As a result, practitioners are capable of achieving unprecedented accuracy by designing trial wavefunctions of unlimited variety and specificity.
"God does not play dice with the universe, but we do."
Random simulations can lead to precise, accurate predictions of deterministic systems. This is one of the simplest, central lessons to be derived from our research investigations of the behavior of electrons in materials chemistry. The equations that drive the behavior of electrons, and by extension, chemical bonding and stability, electronic properties such as conduction and color, are too complex to be solved directly by pencil and paper for all but the simplest of cases. However, by recasting these problems as probabilistic simulations subject to statistical analysis, our research group has achieved unprecedented levels of accuracy in benchmark and predictive studies. QMC holds the promise of satisfactory performance in terms of both cost and accuracy.
"Making the ‘Difficult’ Everyday”
In experimental chemistry, there are rules and exceptions, and rules for determining exceptions. One new result can create more exceptions, change the rules, or eliminate rules. It is much the same in theoretical chemistry, where there exist classes of molecules or properties that are routinely considered "difficult." These systems may be successfully interrogated through very artful application of quantum Monte Carlo (QMC) techniques, and one major research theme is the determination of the amount of necessary "art" required for successful application of QMC to difficult systems. A major algorithmic thrust of our research is development of trial wavefunctions that do not require significant expertise on the part of the user. The ultimate goal of this effort is to develop a robust procedure for the non-expert to perform QMC with an understanding of how much accuracy will be sacrificed in exchange for a lack of ‘art’ in the calculation. Deployment of such an implementation may pave the way to more widespread adoption of these techniques.