[gmx-users] mdp file settings for PME

Justin A. Lemkul jalemkul at vt.edu
Sat Jul 10 02:15:16 CEST 2010

jojo J wrote:
> Dear experts,
> I posted this message before but I think you did not see it since I 
> noticed messages after this have been replied. I have a hydrocarbon 

Messages do not receive replies in any sort of order.  Probably no one has had 
any relevant comments yet, or they live in a time zone where they are not awake :)

> system with no charged particles. Today In one of the GROMACS tutorials 
> I saw the following setting in em.mdp file:
> fourierspacing      =  0.12
> ;fourier_nx          =   0
> ;fourier_ny          =   0
> ;fourier_nz          =   0
> ;pme_order          =   4
> ;ewald_rtol         =  1e-5
> ;optimize_fft      =  yes
> has been used for PME . Sofar I did not even include coulombtype=PME in 
> my em.mdp file. What I had was:
> constraints         =  all-bonds
> integrator          =  steep
> dt                  =  0.002    ; ps !
> nsteps              =  200
> nstlist             =  10
> ns_type             =  grid
> rlist               =  1.0
> rcoulomb            =  1.0
> rvdw                =  1.0
> ;
> ;       Energy minimizing stuff
> ;
> emtol               =  1000.0
> Can you please guide me whether it is necessary to include these lines 
> in em file for system having only partial charges like hydrocarbon 
> (alkane). Also, in md.mdp file I have only coulombtype=PME and not the 

This contradicts your statement above.  You said you had only uncharged 
particles, but clearly this is not the case.

> above Ewald settings. Do you recommend to include those settings? 

For energy minimization, the effects of plain cutoffs may or may not be 
significant.  Using plain cutoffs for actual simulation is a bad idea.  The 
artifacts are well-documented and for modern simulations any good reviewer 
should raise an immediate concern if plain cutoffs were used and not thoroughly 
justified.  A method like PME is significantly more accurate, but you can also 
achieve reasonable results using switch, shift, etc.  Read about the relevant 
mechanics of the force field you're using for what might be appropriate.

> Without the mentioned lines I am getting reasonabel results (potential 
> energy, kinetic, ,, also T coupling works perfectly). If I had better to 
> unclude them what is the best value for fourierspacing?

I don't think there's ever been any sort of systematic study on the effects of 
Fourier grid spacing.  The trade-off is between speed and accuracy.  Larger 
values speed up calculations, but make PME less accurate.  Smaller values have 
the opposite effect.  You could do a small test for yourself (0.12, 0.16, 
0.18...) to see if there are any adverse effects.


> Thank you,


Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list