[lammps-users] A problem with lammps??

Dear Lammps users,

I have a principal question about Lammps. I have simulated a system (well relaxed previously-1600 water molecule + a peptide-the .car and .mdf files are attached) by both Lammps and discover for just 500 fs. I have printed out energies in every step in order to make a comparison. I am obtaining two very different energy profiles. Please look below,

Discover output looks as following. As it can be seen the energies fluctuate less than 20 kcal/mol

image002.png

Lammps output looks as following. As you can see in 500 fs fluctuates more than 100 kcal/mol

image008.png

Can anyone give me an idea about the reason of these differences? Is there any bug in Lammps that causes this??

I have also had problem with systems with very low pressure (like a few molecules of ideal gas in a container). The energy profile calculated by lammps for such systems is unreasonable and periodic. Please see below.

image009.png

Thanks for your help.

Fatima

image004.wmz (6.21 KB)

image006.wmz (5.71 KB)

oledata.mso (20.6 KB)

There are many options (beyond force fields)
for running dynamics in Discover and LAMMPS.
Unless you are doing the same thing in both
codes, you shouldn’t expect the answers to
be the same. Have you checked that the
initial (time = 0.0) thermodynamic quantities
are the same between the 2 codes.

Re: energy of low-pressure system with a few
molecules - why is the energy 10^6 Kcal/mole
if you just have a few molecules? Again,
have you compared a single snapshot to
see if you get the same energy/pressure as
Discover?

Steve

image009.png

image002.png

image008.png

Dear Steve,

image008.png

image002.png

image009.png

To_steven_plimpton.zip (3.17 MB)

Please post to the mail list, not to me personally.

Did the pressure also agree at time 0? If so, that
means the forces match not just the energy. If they
do, then I would run NVE for some number of steps
(in both codes) and see if you get agreement (in a statistical
sense, you won’t remain identical forever).

If so, then try NVT and NPT. If things diverge wildly
in that case, it means you are not running the same
thing in both codes. Most likely you are using bad
thermostat/barostat params in LAMMPS, esp if
your NVE system is far from equilibrium.

Steve

image008.png

image009.png

image002.png