One of the challenges in increasing the predictive ability of Atomic scale simulations (Molecular Dynamics) is employing the right interatomic potentials. While these often take of analytical forms with parameters fit to match some macroscopic properties, a more rigorous strategy would be to use Quantum Mechanics which model explicitly electrons and nuclei and offer the most accurate description that is available to us. Unfortuantely computations at this level (e.g using DFT) are limited to a few hundreds or thousands of atoms/nuclei.
- Can we learn such interatomic potentials from a few (expensive) Quantum Mechanical simulations performed at specific positions of the nuclei?
- Can we extrapolate the form of the interatomic potential learned on a few thousand of atoms to model interactions between millions or billions of atoms?
Details about an abstracted version of the problem and information on the prize can be found here.
Additional reading, from two diametrically different perspectives:
- Albert P. Bartók, Mike C. Payne, Risi Kondor and Gábor Csányi Gaussian Approximation Potentials: The Accuracy of Quantum Mechanics, without the Electrons Physical Review Letters 104 136403 (2010)
- José Jeronimo Rodrigues, Pedro M. Q. Aguiar, João M. F. Xavier: ANSIG - An analytic signature for permutation-invariant two-dimensional shape representation. CVPR 2008
accurate discreption of matter