Respected Sir,
Some of the problems I am facing in reproducing results of research paper "Lee and Drummond PRB B 83, 245114 (2011)" are following:
1) I had been looking into CASINO manual for what value of dtvmc should be used. But could not find any hint about it.
2) In finding pair correlation function, how can we obtain a smooth curve? I have tried using greater number of nsteps but that could bring better smoothening.
About dtvmc and PCF
-
Neil Drummond
- Posts: 95
- Joined: Fri May 31, 2013 10:42 am
- Location: Lancaster
- Contact:
Re: About dtvmc and PCF
Dear Ankush,
Usually the VMC time step is automatically optimised by choosing it such that the VMC move-acceptance probability is 50%. This is the default behaviour, unless you have set opt_dtvmc to 0 in the input file. If you want to check that your VMC time step is OK, check that the overall acceptance ratio is close to 50%. (This information can be found in the out file, just after the end of VMC equilibration.)
The pair correlation function (PCF) is obtained by binning the sampled interlayer distances. To make the sampled PCF smoother, either gather more data (as you suggested) or make the bins larger. The latter can be done post facto using the plot_expval utility. There will always be some noise in the Monte Carlo data, however, unless you go to the limit of an infinite number of samples. If you really need a smooth PCF then you will have to fit an appropriate curve to the Monte Carlo sampled PCF.
Best wishes,
Neil.
Usually the VMC time step is automatically optimised by choosing it such that the VMC move-acceptance probability is 50%. This is the default behaviour, unless you have set opt_dtvmc to 0 in the input file. If you want to check that your VMC time step is OK, check that the overall acceptance ratio is close to 50%. (This information can be found in the out file, just after the end of VMC equilibration.)
The pair correlation function (PCF) is obtained by binning the sampled interlayer distances. To make the sampled PCF smoother, either gather more data (as you suggested) or make the bins larger. The latter can be done post facto using the plot_expval utility. There will always be some noise in the Monte Carlo data, however, unless you go to the limit of an infinite number of samples. If you really need a smooth PCF then you will have to fit an appropriate curve to the Monte Carlo sampled PCF.
Best wishes,
Neil.