In this talk I will present an algorithm that allows one to use multi Slater wavefunction as a trial state in the AFQMC algorithm. The algorithm ensures that the cost of the calculation "decreases" (instead of increases) when increasing the number of determinants up to fairly large determinant sizes. I will also demonstrate that the highly accurate trial wavefunction allows one to calculate properties (other than energies) with reasonable accuracy.
Towards the later half of the talk I will present ways of evaluating the exchange matrix while performing self consistent field calculations on large systems. Exchange is needed in hybrid DFT and Hatree Fock calculations and is usually the bottleneck in these calculations.
Back to Workshop IV: Monte Carlo and Machine Learning Approaches in Quantum Mechanics