[gmx-users] compressing with NPT

[Peter C. Lai]

On 2011-05-23 11:02:46AM -0500, Elisabeth wrote:
> Dear experts,
> 
> I have been considering effect of fixing bond lengths in my system (constraints= none / all-bonds) on equilibrated density using NPT runs with 4.5.4. I am showing two set of results for two ref pressures of 10 and 70 bar. in case of constraints= none I had to employ a 0.5 fs timestep otherwise system blows up. Can you please comment why equilibrated densities are different for the same reference pressure? Thanks.
> 
> 70 bar, all bonds constraint
> timestep 2fs
> 
> Statistics over 875001 steps [ 250.0000 through 2000.0000 ps ], 2 data sets
> All statistics are over 87501 points
> 
> Energy                      Average   Err.Est.       RMSD  Tot-Drift
> -------------------------------------------------------------------------------
> Pressure                    70.4993      0.092    542.944   0.222707  (bar)
> Density                     625.894       0.39    11.9075   0.110708  (kg/m^3)
> 
> 
> 70 bar, constraints= none
> timestep 0.5 fs
> 
> Statistics over 3500001 steps [ 250.0000 through 2000.0000 ps ], 2 data sets
> All statistics are over 350001 points
> 
> Energy                      Average   Err.Est.       RMSD  Tot-Drift
> -------------------------------------------------------------------------------
> Pressure                    70.6207       0.11    932.219  -0.500892  (bar)
> Density                     614.451       0.55    13.7381  -0.347349  (kg/m^3)
> 
> 
> ---
> 
> 10 bar, all bonds constraint
> timestep 2fs
> 
> Statistics over 875001 steps [ 250.0000 through 2000.0000 ps ], 2 data sets
> All statistics are over 87501 points
> 
> Energy                      Average   Err.Est.       RMSD  Tot-Drift
> -------------------------------------------------------------------------------
> Pressure                    10.4843      0.039    621.684  0.0156492  (bar)
> Density                     624.362          1    12.1354   -2.11725  (kg/m^3)
> 
> 
> 10 bar, constraints= none
> timestep 0.5 fs
> 
> Statistics over 500001 steps [ 250.0000 through 500.0000 ps ], 2 data sets
> All statistics are over 50001 points
> 
> Energy                      Average   Err.Est.       RMSD  Tot-Drift
> -------------------------------------------------------------------------------
> Temperature                  299.92       0.11    4.86352  -0.139097  (K)
> Density                     600.689          2    14.7597    11.6485  (kg/m^3)
> 

A couple of questions:

First, what is your criteria that your system has equilibrated? Have you run
longer to see if density stabilizes or plotted the g_energy to see if it has
periodicized after 2ns?

Second, in your 10bar comparisons, you have compared a 2ns run with a 500ps
run, and you can see that after 500ps the density drift over that time was 
11.6kg/m3, making it seem like it is still not in equlibrium.

Finally, are the differences in density statistically significant compared
to an expectant value? i.e. 10kg/m3 difference may or may not be significant 
at 70bar for a system of your size especially if you started at "only" 7nm 
per side and you have compressed it so much more...

Anecdotally, I have found that my own system, "compressing" to a normal 1atm
(from NVT to NPT) seems to converge faster (over the same timescale) with 
constraints = h-bond vs. constraints = all-bonds, but I do have long heavy
atom connectivities...

-- 
===============================================================
Peter C. Lai                 | University of Alabama-Birmingham
Programmer/Analyst           | BEC 257
Genetics, Div. of Research   | 1150 10th Avenue South
pcl at uab.edu                  | Birmingham AL 35294-4461
(205) 690-0808               |
===============================================================