Quantum Monte Carlo (QMC) method is becoming increasingly popular in calculations of molecular properties because of its exclusive description of the electron correlation, better scalability with increasing system size, less dependence on basis sets and ease of parallelization. However, the properties and behavior of a molecule are highly dependent on the geometrical arrangement of its atoms, which QMC cannot predict and must take as input from alternative (sometimes inferior) methods.
This project aims to develop a way performing geometry optimizations in QMC , and implementing this new capability within CMQMC. This is a challenge due to the stochastic nature of QMC and the consequent statistical noise associated with QMC observables (such as energies and molecular forces). Calculation of an optimised molecular structure amounts to finding the global minimum on a noisy multidimensional surface, and so statistical methods are being investigated that take advantage of this noise to efficiently identify the global minimum on a potential energy surface.
The improvement in the accuracy of calculated physical and chemical properties, provided by optimised molecular geometries, has the potential to impact numerous fields of research. Examples include the development of high performance materials, improvement in the efficiency of chemical reactions, or understanding biological processes at the molecular level.
For more information, contact the Project Leader, Dr Deidre Cleland.