Diffusion Monte Carlo (DMC)¶
Diffusion Monte Carlo (DMC) is a projector method that extracts the ground state energy from a trial wavefunction by evolving it in imaginary time.
Overview¶
DMC solves the Schrödinger equation in imaginary time \(\tau = it\):
\[ -\frac{\partial \Psi}{\partial \tau} = (\hat{H} - E_T) \Psi \]
As \(\tau \to \infty\), the wavefunction projects onto the ground state \(\Phi_0\), provided the trial wavefunction \(\Psi_T\) has a non-zero overlap with \(\Phi_0\).
Key Features¶
- Fixed-Node Approximation: To maintain the fermionic nature of the wavefunction (antisymmetry), the nodal surface is fixed to that of the trial wavefunction \(\Psi_T\). This provides a variational upper bound to the ground state energy.
- Importance Sampling: The random walk is guided by the trial wavefunction to improve efficiency.
- Time Step Error: The simulation uses a finite time step \(\tau\). Results should be extrapolated to \(\tau \to 0\).
Running DMC¶
DMC calculations are controlled by the blocking_dmc and dmc modules.
%module blocking_dmc
dmc_nstep 100
dmc_nblk 200
%endmodule
%module dmc
tau 0.01
etrial -15.8
%endmodule
See Input Keywords for details.