[CPMD-list] a really weird geometry optimization problem

[Bin Pan]

Dear CPMD users,

I found a really weird problem for the following geometry 
optimization run using cpmd.
I used vanderbilt pseudopotentials with the following script on butane:

&CPMD
  OPTIMIZE GEOMETRY XYZ
  CONVERGENCE ORBITALS
    1.0e-8
  CONVERGENCE GEOMETRY
   1.0d-6
  ODIIS NO_RESET=50
   5
  MEMORY BIG
  SPLINE POINTS
   2000
  STORE
   50
&END

&DFT
  FUNCTIONAL BLYP
  GC-CUTOFF
   5.0e-6
  NO EXCHANGE CORRELATION TABLE
&END

&SYSTEM
  ANGSTROM
  SYMMETRY
   0
  CELL ABSOLUTE
   15.    15.0000    15.0000 0.0 0.0 0.0
  CHARGE
   0.0
  DUAL
   8.0
  CUTOFF
   30.0
&END

&ATOMS

ISOTOPES
   15.99940
    1.0
TORSION 4 3 2 1   81.
*C_VDB_BLYP.psp FORMATTED
   LMAX=P
4
8.716657638550 9.274163246155 8.742531776428
9.229999542236 10.000119209290 9.999937057495
10.770000457764 9.999856948853 10.000078201294
11.283608436584 11.128035545349 9.086284637451

*H_VDB_BLYP.psp FORMATTED
   LMAX=S
10
9.073311805725 8.265353202820 8.742531776428
9.073330879211 9.778561592102 7.868880271912
7.646657943726 9.274176597595 8.742531776428
8.873504638672 11.008985519409 9.999838829041
8.873167991638 9.495833396912 10.873587608337
11.126539230347 9.057957649231 9.638683319092
11.126602172852 10.157709121704 10.996479988098
10.927062988281 12.069932937622 9.447676658630
12.353608131409 11.127857208252 9.086387634277
10.927011489868 10.970179557800 8.089880943298

&END

It is so weird that starting from the second step, the geometry was 
optimized to a really strange configuration.
I had several runs using the same script with the only difference of 
changing the initial geometry and they all worked fine.
Can anyone help me? Thanks a lot in advance!

Bin.