Correction Author: David A. Pearlman
Correction Date: 9/5/91

Problem: If the user specifies NTT=5 (separate solute/solvent temperature
  coupling) and there are no moving solute atoms (because a belly is being
  used and none of the solute is in the belly), the program either report
  erroneous temperatures or crash, depending on the floating exception
  handling characteristics of the host computer.

  This problem can occur in both Gibbs4 and sander.

Cause: An implicit assumption was being made that if NTT=5 was used, there
  would be at least one moving solute residue. An unprotected division
  was carried out, where the denominator could be 0.0 if there were no
  moving solute residues.

Workaround for older versions: If there are no moving solute residues because
  a belly is being used, and the belly contains no solute atoms, one should
  use NTT=1 rather than NTT=5. In this case, since there are no moving
  solute atoms, it makes no sense to separately scale the solute and solvent
  atoms.

Fix:
  (applies to both Gibbs4 and sander. 
     For gibbs4, modify routine RUNMD, which is in file giba.f:
     For sander, modify routine RUNMD, which is in file runmd.f )

     1) Add the data initialization to the subroutine RUNMD 
        The following two statements should be inserted immediately following
        the statement "EQUIVALENCE (VIR(1),ENER(19))

        SAVE SMALL
        DATA SMALL/1.0D-7/

     2) Replace the statements in RUNMD:
         
        ELSE IF (NTT.EQ.5) THEN
            SCALP = SQRT(1.0+DTTP*(EKINP0/EKGP-1.0))

        with

        ELSE IF (NTT.EQ.5) THEN
            SCALP = 1.0D0
            IF (EKGP.GT.SMALL) SCALP = SQRT(1.0+DTTP*(EKINP0/EKGP-1.0))