restart files

[Jack Tanner]

I have a question about restart files in xplor.  

Should a trajectory that is pieced together using restart files be 
identical to the trajectory obtained from a single continuous calculation?
I would think the answer is yes, but the following test case suggests
otherwise.

This is the energy at step 152000 in a trajectory started from a 
restart file written at step 100000:

 --------------- step=152000 at     76.00000 ps --------------------------------
 | E(kin)+E(total)=-12155.544      E(kin)=7851.728      temperature=300.286    |
 | Etotal =-20007.272 grad(E)=21.586     E(BOND)=2871.459   E(ANGL)=3679.353   |
 | E(DIHE)=1969.108   E(IMPR)=1180.962   E(VDW )=-5259.302  E(ELEC)=-22863.850 |
 | E(PVDW)=-215.554   E(PELE)=-1369.447                                        |
 -------------------------------------------------------------------------------

And here is the energy at step 152000 in a trajectory started from a 
restart file written at step 150000:

 --------------- step=152000 at     76.00000 ps --------------------------------
 | E(kin)+E(total)=-12157.636      E(kin)=7863.307      temperature=300.729    |
 | Etotal =-20020.943 grad(E)=21.600     E(BOND)=2866.401   E(ANGL)=3685.841   |
 | E(DIHE)=1969.488   E(IMPR)=1183.287   E(VDW )=-5265.852  E(ELEC)=-22872.107 |
 | E(PVDW)=-212.158   E(PELE)=-1375.843                                        |
 -------------------------------------------------------------------------------

Note that the energies at step 152000 for the two calculations are 
similar but not identical.  I used v3.851 on an alpha.

The core of the dynamcs verlet script (omitting i/o statements) was

	dynamics verlet      
		ascii=true vascii=true
		tcoupling=true tbath=300
		timestep=0.0005 nstep=50000 
		iasvel=current firsttemp=300
		iprfrq=2000 nprint=2000
	end                                       

Thanks,

Jack Tanner