Double wide sampling in free energy calculation


Click here for full thread
Clicked A Few Times
12:27:56 AM PST - Wed, Nov 27th 2013
Hi,

I think there is a little error in the example of NWChem free energy calculation. Following is the input file of the example:

  md

     system clfoh_neb

     cutoff 1.5 qmmm 1.5

     noshake solute

     isotherm

  end



  qmmm

    print low

    nsamples 1000

    ncycles 2

  end



  set qmmm:fep_geom clfoh_neb-s.xyzi clfoh_neb-e.xyzi

  set qmmm:fep_esp  clfoh_neb-s.esp clfoh_neb-e.esp

  set qmmm:fep_lambda 0.0 0.1

  set qmmm:fep_deriv .true.



  task qmmm dft fep

I am concerned is if set the double wide sampling, should we set fep_lambda as follows?

  set qmmm:fep_lambda 0.05 0.1

  set qmmm:fep_deriv .true.

Two facts support my assumption. The one is NWChem manual tell us in the fep_lambda line we set λiandλi + 1, and also say

λi − 1 = λi − (λi + 1 − λi)

So if we set qmmm:fep_lambda 0.0 0.1, then λi − 1 = − 0.1. However we want to sample from 0.0 to 0.1, therefore I guess it should be 0.05 0.1.

Another reason is the computational results. I performed 3 calculations for the example with the following settings:

First:
  set qmmm:fep_lambda 0.0 0.1

  #set qmmm:fep_deriv .true.   # without using double wide sampling

Second:
 set qmmm:fep_lambda 0.0 0.1

 set qmmm:fep_deriv .true.

Third:
 set qmmm:fep_lambda 0.05 0.1

 set qmmm:fep_deriv .true.

The total free energy of first computation is 0.63 kcal/mol, agreement with the third result 0.68, but the second result is 1.74 kcal/mol.

Am I wrong? Any suggestion is appreciated.

I have another question here. While performing free energy calculation, NWChem generate .trj file frame by frame, can we change the writing frequency? Sometimes the .trj file is too large to treat. Thanks in advance.

Jingbo