Columbia University

Technology Ventures

Molecular simulation algorithm to approximate potential energy of an atom

Technology #m10-030

“Lead Inventors: Gabor Csanyi, Ph.D.; Albert Bartok-Partay, Ph.D.; Imre Risi Kondor, Ph.D.

molecular simulation models potential energy of an atom:
Molecular simulation is widely used in computational chemistry, computational biology and materials science for studying molecular systems ranging from small chemical systems to large biological molecules and material assemblies. It is a critical part of many industrial development processes in fields such as drug discovery and materials design. While the quantum mechanical equations governing molecular systems are known, accurately simulating systems consisting of more than a few dozen atoms is usually computationally expensive. While cheap approximations exist, they are not predictive. Thus there is a large gap in performance, cost and therefore applicability in the range of existing molecular dynamics simulation software.

Gaussian Approximation Potential method of molecular simulation approximates potential energy of an atom:
The Gaussian Approximation Potential (GAP) is a new method of approximating the potential energy of an atom, given the position and identity of its neighboring atoms. It allows an interatomic potential model to be generated from a database of accurate quantum mechanical calculations. It uses an interpolation algorithm to take advantage of spatial invariance and to identify target configurations. GAP does not rely on any fixed functional form, making it very flexible. With reduced complexity of calculations, this technique allows more complicated systems to be simulated with high accuracy rapidly.

Applications:
• Software for fast simulation of molecules and materials
• Research and development in nanotechnology
• Biomolecular simulation

Advantages:
• High robustness and precision interpolation in high-dimensional space
• Use available quantum mechanical data as fitting target
• Spatially invariant representations through bispectral invariants
• Reduced computational complexity
• Fast simulation

Patent Status: Patent Pending

Licensing Status: Available for Licensing and Sponsored Research Support

Publications: A. P. Bartok, M. C. Payne, R. Kondor, G. Csanyi, Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons, http://arxiv.org/abs/0910.1019

Bridging the gap: Gaussian Approximation Potential,
A. P. Bartok, Ph.D. thesis 2009

Cambridge Enterprise Licensing Opportunities