Wiki

:::bash
~/$DIR/src/liquid.sims/print.close.to.exp.fragments.sh file.par E.comb.dat dimer.ljf

Input arguments

:::text
file.par: CHARMM-compatible parameter file
E.comb.dat: ab-initio and force-field energies
dimer.ljf: geometrical information of ab-initio snapshots

The following files must be present in the working directory:

density.exp
vapor.exp

where the two files contain only the target experimental values, e.g., a target density of 0.878 would yield a density.exp file:

:::text
0.878

The script only works with up to two optimized atom types.

Output example

:::text
Looking for parameter sets with:
     1.3904 < rho  <  1.5996
     9.9045 < delH < 11.3955
-0.4200 2.1000 11.2454 1.5399
RMSE: 0.5206 kcal/mol
Within 35% of best RMSE:

-0.3600 2.1000 10.7199 1.5300
RMSE:  0.5235 kcal/mol; Error on RMSE: -0.0055; Relative error on experimental data:  0.0243

-0.3800 2.1000 10.9502 1.5389
RMSE:  0.5223 kcal/mol; Error on RMSE: -0.0032; Relative error on experimental data:  0.0406

-0.4000 2.1000 11.1332 1.5435
RMSE:  0.5213 kcal/mol; Error on RMSE: -0.0013; Relative error on experimental data:  0.0557

-0.4200 2.1000 11.2454 1.5399
RMSE:  0.5206 kcal/mol; Error on RMSE:  0.0000; Relative error on experimental data:  0.0634

Best paramters:
-0.3600 2.1000 10.7199 1.5300
          RMSE:  0.5235 kcal/mol
       Density:  1.5300 g/mol
Heat of vapor.: 10.7199 kcal/mol

where the target density was 1.495 and the target heat of vaporization was 10.65. The first few lines provide the range of acceptable values for the computed density and heat of vaporization, which are +/- 5% of the target experimental value. Then, the script prints all sets that have acceptable thermodynamic observables, within 35% of the lowest RMSE among the set. For instance, the line

:::text
-0.4200 2.1000 11.2454 1.5399

describes epsilon=0.42, Rmin_2=2.10, DeltaH=11.2454 (heat of vaporization), and density=1.5399.

Finally, the best parameter is identified as the one with the smallest deviation from experiment. Note, however, that stochastic errors of the simulations and measurement errors of the experimental data may thwart this analysis. In this case, we recommend the user to consider different sets and validate them with another observable (e.g., free energy of hydration).