Correction Author: David A. Pearlman
Correction Date: 9/10/91

Programs: gibbs,  sander

Severity: Minor


Problems: 1) When temperature scaling option NTT<0 was chosen 
             (velocities randomly reassigned every so often) *and* 
             a belly was being used, the velocities for the non-moving 
             (non belly atoms) were not being zeroed out after the 
             assignment.

          2) If NTT <0, *and* a belly was being used, *and* NSEL was >0
             (only a specified number of atoms had their velocities
             changed very so often), non-moving non-belly atoms could
             be chosen to be among the NSEL atoms.

          
           Problem (1) could occur in both SANDER and GIBBS.
           Problem (2) could only occur in GIBBS.

Affects: The temperatures reported for a simulation where NTT<0 and IBELLY>0
         would appear wrong (too high). The effect would become greater
         as the percentage of the system considered non-moving increased.

         The actual MD trajectory, free energies, etc. would not be affected
         (would be correct).

Fix: 1) When a belly simulation is being carried out zero out the velocities
        of all non-moving atoms immediately after velocity reassignment.

     2) Before counting any atom as among the group of NSEL, determine if it
        can move, when a belly run is being performed.

Additional Changes: The version of NEWVEL used in program SANDER has been
     updated so that it is now the same as that used in GIBBS.


Routines affected:

     GIBBS:   Routines RANVEL and NEWVEL in file veloc.f
              Routine RUNMD in file runmd.f

     SANDER:  Routine NEWVEL in file newvel.f
              Routine RANVEL added to file veloc.f
              Routine RUNMD in file runmd.f