Hello all,

I'm performing many geometry optimizations on a SuSE cluster with Academic CASTEP 6.01 via MPI and noticed a strange bug.  The cluster has a walltime of 24 hours, so I often have to resume calculations.  In this particular case, the calculation resumes but incorrectly converges after the initial step.  If this isn't an appropriate avenue to report this bug (may be in MS CASTEP as well), please let me know where to report it.  I can also provide the checkfile (1.1GB) and other data if necessary.

Here is the relevant portion of the CASTEP output

BFGS: Restarting previous geometry optimization.

BFGS: finished iteration     0 with enthalpy= -4.71581581E+004 eV

+-----------+-----------------+-----------------+------------+-----+ <-- BFGS

| Parameter |      value      |    tolerance    |    units   | OK? | <-- BFGS

+-----------+-----------------+-----------------+------------+-----+ <-- BFGS

|  dE/ion   |   0.000000E+000 |   5.000000E-006 |         eV | Yes | <-- BFGS

|  |F|max   |   7.840047E-003 |   1.000000E-002 |       eV/A | Yes | <-- BFGS

|  |dR|max  |   0.000000E+000 |   5.000000E-004 |          A | Yes | <-- BFGS

|   Smax    |   1.633129E-002 |   2.000000E-002 |        GPa | Yes | <-- BFGS

+-----------+-----------------+-----------------+------------+-----+ <-- BFGS

BFGS: Geometry optimization completed successfully.

As you can see, the program incorrectly thinks the dE/ion and |dR|max converge when they haven't been calculated.

Thanks,

Steve Monaco