Restart MSST makes box expanding

[Yiyang]

Hello, 

I need to perform a QS MSST shock simulation which has to be divided into several separated runs, because each run will take very long time.

For this purpose, I used the 'C_tersoff.inp' example (classical MD) and made a test. The script is attached.

I performed three runs:

The first run uses the attached script, without any change, to run MSST for 2000 steps, which provides state data for the 'original curve'.

The second run modifies the first two lines of the script as:

    @SET RESTART         FSLSE

    @SET NSTEP              1500

The third run modified the first two lines as:

    @SET RESTART          TRUE

    @SET NSTEP              500

The second and third runs provides state data for the restart curve.

As we can see, the restart run makes the box explode.

From the output file, the restarted variables include:

 *******************************************************************************

 *                            RESTART INFORMATION                              *

 *******************************************************************************

 *                                                                             *

 *    RESTART FILE NAME: C_tersoff-1.restart                                   *

 *                                                                             *

 * RESTARTED QUANTITIES:                                                       *

 *                       CELL                                                  *

 *                       COORDINATES                                           *

 *                       RANDOM NUMBER GENERATOR                               *

 *                       VELOCITIES                                            *

 *                       MD COUNTERS                                           *

 *                       MD AVERAGES                                           *

 *                       BAROSTAT                                              *

 *                       PARTICLE THERMOSTAT                                   *

 *                       REAL TIME PROPAGATION                                 *

 *                       PINT BEAD POSITIONS                                   *

 *                       PINT BEAD VELOCITIES                                  *

 *                       PINT NOSE THERMOSTAT                                  *

 *                       PINT GLE THERMOSTAT                                   *

 *                       HELIUM BEAD POSITIONS                                 *

 *                       HELIUM PERMUTATION STATE                              *

 *                       HELIUM FORCES ON SOLUTE                               *

 *                       HELIUM RNG STATE                                      *

 *******************************************************************************

What should I do to make a correct restart run for MSST? Specifically, what variables should I read from restart file?

Using 'RESTART_DEFAULT    .FALSE.' then read only

CELL, COORDINATES, VELOCITIES, BAROSTAT, PARTICLE THERMOSTAT

cannot solve this problem.

Thanks

Regard,

Yiyang

[Yiyang]

Hello,

I fixed a bug in my script, which is the line

      PRESSURE [bar] 1000.0

in &MSST section. This valued needs to be updated as tr(pressure_tensor)/3.0 at time step 1500,

before we performing the restart run. Doing such update, I get following results:

We can see that, if the system has evolved to a resonably stable state, then the restart run can statistically reproduce the original curve, although not an exact reproduction.

However, if the restart run starts from an unstable state, ie, at 400fs where there is huge pressure fluctuation, then there can be deviation from the original curve, as shown below:

If I completely remove the line 

      PRESSURE [bar]   ***

then the result is again erroneous.

Is it possible to improve this behavior?

Thanks

Regard,

Yiyang