CHARMM Development Project Forums User Discussion & Questions Molecular Dynamics Erro when running Drude with CHARMM FF

I was trying to run dimethyl ether (DME) liquid simulation using CHARMM Drude FF files provided by Mackerell Lab: http://mackerell.umaryland.edu/charmm_drude_ff.shtml
The initial structure is from a 10 ns DME liquid simulation using additive CHARMM FF (all properties match what was reported in the previous paper)
I added Drude particles and lone pairs, then only relaxed Drude particles.
But when I start the liquid simulation I kept getting the error message saying: ** ERROR IN HardWallDrude ** The drude is too far away from atom 16 d = 189.9591
I also tried not relaxing Drude (so the Drude particles initially overlap with their real atoms) but I would get the same error message on a different atom: ** ERROR IN HardWallDrude ** The drude is too far away from atom 1 d = 0.7066
I also tried using several completely different initial configurations, sometimes the simulation won't crash but the volume would increase to an unreasonable value.
Previously, I also encountered the incorrect density problem when I was running Drude w/o lone pairs, but that was caused by too few images and was fixed after I made sure I had 26 images.
But this time I have the correct number of images and my system is still blowing up and most of the time the simulation crashed right in the beginning because the error in hardwalldrude
The setting for the simulation is the same as my previous thread: https://www.charmm.org/ubbthreads/ubbthreads.php?ubb=showflat&Number=37883#Post37883
Is this only caused by my configuration (but it's pre-equilibrated using additive FF) or am I doing something wrong?

Thanks in advance!

Perhaps there was a problem with your conversion from additive to Drude for PSF and COOR files.

This may not be the problem du jour, but I noticed your FFT grid settings are a bit small; 48 would be better than 32 for a ca. 46 Å box

In a test simulation I ran a while ago, I used the setting

DrudeHardWall L_WALL 0.2

In more recent tests with OpenMM I used 0.25 Å (0.025 nm)

Thank you for your reply.
I used the following script to convert my nonpolar equilibrated structure to Drude:

open unit 91 card read name equil.pdb
read sequ pdb unit 91
set residue DMET

generate @residue first none last none setup warn drude dmass 0.4 !show

rewind unit 91
read coor pdb unit 91
coor print
set seed3 1 C1 1 O2 1 C3

ic param
ic seed @seed3
ic build
!ic print
coor sdrude
coor shake
coor print

open write unit 10 card name dme.crd
write coor card unit 10

open write unit 10 card name dme.psf
write psf card unit 10

open write unit 10 card name dme.pdb
write coor pdb unit 10


I view the crd file in vmd and the position of lone Paris and Drude particles seem normal.
I just tried set L_WALL to 0.25 still raised the same error on the same atom: ** ERROR IN HardWallDrude ** The drude is too far away from atom 1 d = 0.7066
Is there an example simulation of drude of small organic molecules? The only example I can find is for swm4-water (inside c31test folder) but it seems it's for an older version.

Maybe there's a problem with the post-conversion minimization, then. The Drude should not be that far away from the atom center.

This error is when I tried not minimized Drude particles at all so initially, they overlap with their real atoms, but I still get this error message saying the atom is too far away.

The scripts I was given all have the following sequence as part of the post-conversion dynamics start script:

cons harm force 100000.0 sele .not. type D* end
mini SD nstep 50
cons harm force 0.0 sele .not. type D* end
coor copy comp

SHAKE bonh param tolerance 1.0e-9 -
fast water swm4 -
select .not. type D* end -
select .not. type D* end

mini SD nstep 50

tpcontrol nther 2 ...

Thank you. I tried more different initial configurations with this setting, one of them worked. By the look of the density, the system seems to be working properly.