Benchmark problems
From QMC
It would be useful to formulate benchmark problems that can be used to evaluate the relative merits of various methods. Formulate and benchmark away to your heart's content.
[edit] Record setting wavefunction results.
Since one of the main problems in QMC is the trial wavefunction, it would be useful to have a central repository of the variational energy and variance and the fixed-node energy of different types of trial wavefunctions on "standard systems." A definition of standard systems might be ones that 2 or more groups are interested in and are easy to describe; all electron atoms, the electron gas, simple diatomics come to mind. In order to be useful, enough information has to be given to fully characterize the system and wavefunction. In my opinion (DMC), if someone posts an energy, they should be willing to divulge all the wavefunction parameters, since someone may want to verify the calculation or debug their own code. Literature references should also be given. At the minimum, if someone posts a result, they should identify themselves. (the scribe field)
In case the exact energy is known (say for all-electron atoms) one can also give the error in the VMC energy and FN-DMC energy. The error bars also need to be specified, and, in principle, should include any systematic errors such as timestep, lattice, or population bias.
For periodic systems, one should give the results before size extrapolation.
Other quantities of interest, but less well defined, are CPU time/step, code used to perform the calculation and method used to do the optimization.
Benchmarks:
- All-electron diatomics
- Other all-electron molecules
