[lammps-users] Shake Determinant 0.0

Hi, Steve. Thanks for your response.
The data and log file I am sending you ended with the error message “ERROR on proc 0: Shake determinant = 0.0”
The step number at which the message appears and the program crashes depends on the timestep I choose for the simulation. I have used values of 1-0.5-0.1 fs. I also tried modifing the drag value but it had no effect (values of 2, that I know is huge).
Using the same data file and only modifing the tstep and Tsatrt and Tstop in the nvt fix, my results were the following:
Tstart(K), Tstop(K), drag, tstep(fs), # steps
300, 300, 1, 0.2, 200- Error: Shake Det. 0.0
300, 300, 0.5, 0.2, 1100- Error: Shake Det. 0.0
300, 300, 0.1, 0.2, 1400- Error: Shake Det. 0.0
100, 100, 1, 0.2, 500- Error: Shake Det. 0.0
100, 100, 0.5, 0.2, 1500- Error: Shake Det. 0.0
100, 100, 0.1, 0.2, 35900- Error: Shake Det. 0.0
100, 100, 0.1, 2, 4800- Error: Shake Det. 0.0 (DIFFERENT DRAG VALUE)

I thank you in advance for your help,
Estefania

2009/7/22 Steve Plimpton <[email protected]>

data-slab-canal-14-relajado-monocapa (8.25 KB)

log.lammps (29.5 KB)

Hi, Steve. Thanks for your response.
The data and log file I am sending you ended with the error message “ERROR on proc 0: Shake determinant = 0.0”
The step number at which the message appears and the program crashes depends on the timestep I choose for the simulation. I have used values of 1-0.5-0.1 fs. I also tried modifing the drag value but it had no effect (values of 2, that I know is huge).
Using the same data file and only modifing the tstep and Tsatrt and Tstop in the nvt fix, my results were the following:
Tstart(K), Tstop(K), drag, tstep(fs), # steps
300, 300, 1, 0.2, 200- Error: Shake Det. 0.0
300, 300, 0.5, 0.2, 1100- Error: Shake Det. 0.0
300, 300, 0.1, 0.2, 1400- Error: Shake Det. 0.0
100, 100, 1, 0.2, 500- Error: Shake Det. 0.0
100, 100, 0.5, 0.2, 1500- Error: Shake Det. 0.0
100, 100, 0.1, 0.2, 35900- Error: Shake Det. 0.0
100, 100, 0.1, 2, 4800- Error: Shake Det. 0.0 (DIFFERENT DRAG VALUE)

I thank you in advance for your help,
Estefania

2009/7/22 Steve Plimpton <[email protected]>

data-slab-canal-14-relajado-monocapa (8.25 KB)

log.lammps (29.5 KB)

Dear LAMMPS users:

I posted this question last week but I’m having problems with my e-mail and it seems it wasn’t correctly posted.
Here it goes again, sorry if I’m doubling the info and the messages.

I am simulating a channel of rutile (110) with a monolayer of SPC/E water molecules, using NVT and SHAKE fixes. The atoms in the solid phase are fixed via the wiggle fix with a 0,0 amplitude. I’m only SHAKEing the bond of the water molecules (or so I think).
The data file consists on the list of atoms, with full style. 180 atoms in total of which the first 156 correspond to the rutile channel and 157-180 to the water molecules. Similarly, molecule id=1 corresponds to the rutile channel and mol id>=2 to the water molecules adsorbed to the surface. I only list the bonds and angles corresponding to the water molecules.

I have been running nvt simulations that ended in the error “Shake determinant = 0.0.” I have checked all the parameters, I’ ve modified the tstep from 1 to 0,1 (real units) but I still get the same error.

Additionally, simulating the rutile channel only for about 0.5 ps without water did not give any problems.

For further details, I attach the data log file that ended with the error message “ERROR on proc 0: Shake determinant = 0.0”
The step number at which the message appears and the program crashes depends on the timestep I choose for the simulation. I have used values of 1-0.5-0.1 fs. I also tried modifing the drag value but it had no effect (values of 2, that I know is huge).
Using the same data file and only modifing the tstep and Tstart and Tstop in the nvt fix, my results were the following:

Tstart(K), Tstop(K), drag, tstep(fs), # steps
300, 300, 1, 0.2, 200- Error: Shake Det. 0.0
300, 300, 0.5, 0.2, 1100- Error: Shake Det. 0.0
300, 300, 0.1, 0.2, 1400- Error: Shake Det. 0.0
100, 100, 1, 0.2, 500- Error: Shake Det. 0.0
100, 100, 0.5, 0.2, 1500- Error: Shake Det. 0.0
100, 100, 0.1, 0.2, 35900- Error: Shake Det. 0.0
100, 100, 0.1, 2, 4800- Error: Shake Det. 0.0 (DIFFERENT DRAG VALUE)

I thank you in advance for your help, and I apologise again for the double-shooting.
Best regards,
Estefania

data-slab-canal-14-relajado-monocapa (8.25 KB)

log.lammps (29.5 KB)

I ran your problem. It has several issues. First, you are
running an older version of LAMMPS, since a couple
of your input command throw errors. Please upgrade.

Second, you set the Coulomb cutoff for Ewald to 1.5 Angs,
which is very short. This means Ewald (or PPPM) has
to work very hard, and it is slow. Just set the Coulomb
and LJ cutoff the same, as in the script I attach.

Third, your system is not running in a stable manner.
There are wild temperature fluctuations, even in the
log file you sent. When I ran it, I eventually get NaNs
in the thermo output b/c it blows up. I never saw
the SHAKE determinant error you mention. But
if the system is crazy, then SHAKE can have
problems as well. I.e. it can't constain a water
that has blown apart.

I suggest you viz your system and get it to
run in a stable manner w/out SHAKE before you turn that fix
on. You should be able to do that w/ a 1 fmsec
timestep. Then use a 2 fmsec timestep with
SHAKE. If the temperature is wild, then you are
doing something wrong. Although you have a tiny
system so that could also affect the temperature.

My modified input script is attached.

Steve

in.shake.bug (2.62 KB)