CHAMP is a quantum Monte Carlo suite of programs for electronic structure
calculations on a variety of systems (atoms, molecules, clusters, solids
and nanostructures) principally written by Cyrus Umrigar and Claudia Filippi
with major contributions by Julien Toulouse and Devrim Guclu, postdocs at Cornell
and smaller contributions by others.
|
||||
If you wish to use this program, please contact one of the principal authors: |
||||
|
||||
| ||||
CHAMP is presently a suite of programs with the following three basic capabilities:
| ||||
| ||||
In each case, the very best quantum Monte carlo algorithms have been used.
A great deal of attention has been paid to the implementation of highly efficient
algorithms, a variety of sophisticated wave functions, and tools to optimize them,
with particular emphasis on all-electron and pseudopotential finite systems. For such
systems, the code can also compute forces on the nuclei, a necessary precursor of
molecular dynamics calculations. There are both serial and parallel (MPI)
versions of the codes. Particularly noteworthy features of CHAMP are: | ||||
Recent applications of the code CHAMP | ||||
| ||||
Groups using CHAMP | ||||
| ||||
| ||||
Particularly noteworthy features of CHAMP are: | ||||
| ||||
Efficient wave function optimization by variance minimization
The wave function parameters can be optimized by variance minimization, but we rarely use this method now.
For finite systems, it is possible to optimize not only the Jastrow part but also the
determinantal part of the wave function (CI coefficients, orbital coefficients and exponents).
| ||||
Even more efficient wave function optimization by energy/variance minimization
Although naive energy minimization is far less efficient than variance minimization, in recent years
we have developed three energy minimization schemes that are very efficient and that allow one to
systematically extrapolate away the fixed-node error for many systems.
| ||||
| ||||
Effective-fluctuation-potential method to optimize the determinantal part
The "effective fluctutation potential" method is an approach originally proposed
by S. Fahy to optimize the wave function by energy minimization. It was further
developed and implemented in the code by C. Filippi, F. Schautz and A. Scemama to optimize orbital and
CI coefficients.
| ||||
| ||||
Sophisticated Jastrow factors, including e-n, e-e and e-e-n
correlations (e=electron, n=nucleus)
These include forms that are systematically improvable (within the constraint of
not using more than e-e-n correlations) and obey all three types of cusp conditions
exactly. For large systems the option exists to use Jastrow functions that go exactly
to a constant beyond some distance, thereby improving the scaling of the
computer time with system size. The earlier forms of the Jastrow factor
are described in the first three references above in:
| ||||
| ||||
Accelerated Metropolis method
It allows one to make very large moves and still have a high acceptance, resulting in very
short autocorrelation times. The gain, compared to other Metropolis methods is particularly
large when pseudopotentials are not used.
| ||||
| ||||
Very efficient diffusion Monte Carlo algorithm
It takes into account the singularities in the local energy and velocity at nodes of the
wave function and at particle coincidences.
| ||||
| ||||
Calculation of numerical forces on nuclei for finite systems
It employs correlated sampling along with a space-warp coordinate transformation that
improves the efficiency.
|