1. Tristan Bereau Avatar Tristan Bereau
  2. Untitled project
  3. lj-fit.pc-mtp

Wiki

Clone wiki

lj-fit.pc-mtp / Monte-Carlo parameter variation

View History 

:::bash
~/$DIR/src/fit.lj/fit.LJ.monte-carlo.py file.par \
  -e E.comb1.dat [E.comb2.dat [...]] -f dimer1.ljf [dimer2.ljf [...]] \
  [-t temperature]
Input arguments 
:::text
file.par: CHARMM-compatible parameter file
E.comb*.dat: Raw ab-initio and force-field energies for any number of compounds
dimer*.ljf: Geometrical information for ab-initio calculations
temperature: effective temperature used in the Metropolis criterion
Make sure the list of E.comb*.dat and dimer*.ljf files is correctly ordered.

The script produces a Monte Carlo variation of LJ parameters. For each line outputted, the script will provide the root-mean squared error against the ab-initio data and the set of LJ parameters involved 

:::textx
#Energy  O=C             F               OR              C=O             HOR             CR              HCMM             
25.1045  0.1934  1.5000  0.1341  1.5000  0.1500  1.5000  0.1500  1.5000  0.1372  1.5000  0.1500  1.5000  0.1500  1.5000  
24.6141  0.1934  1.5000  0.1341  1.5000  0.1500  1.3840  0.1500  1.8599  0.1372  1.5000  0.1712  1.5000  0.1500  1.5000  
23.3553  0.1694  1.8599  0.1510  1.5000  0.1500  1.5000  0.1934  1.3107  0.1237  1.5000  0.1712  1.5000  0.1500  1.3840  
11.1947  0.1694  1.8599  0.1499  1.5413  0.1533  1.5000  0.1774  1.3107  0.1237  1.3130  0.1712  1.5000  0.1403  1.3840  
11.2018  0.1937  1.4267  0.1728  1.5413  0.1403  1.3840  0.1579  1.8599  0.1237  1.3130  0.1828  1.5000  0.1533  1.5000  
11.1599  0.1994  1.4267  0.1728  1.3594  0.1403  1.3656  0.1579  1.8855  0.1237  1.3130  0.1828  1.5000  0.1533  1.4586  
 4.2608  0.1579  1.8855  0.1883  1.3594  0.1296  1.4586  0.1994  1.4267  0.1237  1.0880  0.1795  1.5000  0.1403  1.3656  
 4.2304  0.1744  1.8855  0.1883  1.3594  0.1296  1.3052  0.2041  1.5040  0.1237  1.0880  0.1795  1.3468  0.1327  1.3656
Rather than a fit (see, fitting script), this provides an ensemble of parameter sets. The distribution of RMS errors will critically depend on the effective temperature used. A value of 0.1 will tend to give the lowest RMSE results, while higher temperatures will provide a larger variety of parameters. Note that in light of reproducing properties, the best parameter set is not necessarily the one with the lowest RMSE.

Updated