[CPMD-list] Discrepancy between MORTENSEN and TUCKERMAN

[Eung-Gun Kim]

Dear List Subscribers,

I was comparing results from three different methods, namely, PBC, TUCKERMAN,
and MORTENSEN, for calculating an isolated molecule and noticed a significant
deviation with the choice of MORTENSEN.  Test conditions and observations are
summarized below:

  1) CPMD 3.9.1 on IBM SP4
  2) box size: 25 x 25 x 25 A for all three methods
       spatial dimension of the molecule ~ 10 A
  3) Geometry optimization was performed first under PBC, and the resulting
     geometry was used as input for TUCKERMAN and MORTENSEN.
  4) Geometry optimization finished in one step with TUCKERMAN while it took
     additional *160* geometry optimization steps with MORTENSEN.
  5) The optimized geometry from MORTENSEN is different--but not unphysical--
     from either PBC or TUCKERMAN.
  6) The total energy is also quite different for MORTENSEN:

            -158.02613142: PBC
            -158.02596806: TUCKERMAN
            -158.08337274: MORTENSEN

I've attached a portion of my input for MORTENSEN below for your reference.  For
TUCKERMAN, I simply replaced the keyword MORTENSEN with TUCKERMAN.  Thanks very
much for reading.

Best regards,

EG Kim

###############################################################################

&CPMD
  RESTART LATEST COORDINATES HESSIAN WAVEFUNCTION
  OPTIMIZE GEOMETRY
  LBFGS
  ODIIS
    5
  CONVERGENCE ORBITALS
    1.D-6
  ELECTROSTATIC POTENTIAL
  STORE
    50
  RESTFILE
    2
&END

&DFT
  FUNCTIONAL BLYP
&END

&SYSTEM
  ANGSTROM
  POINT GROUP
    AUTO
  SYMMETRY
    ISOLATED
  POISSON SOLVER MORTENSEN
  CELL ABSOLUTE DEGREE
    25.0  25.0  25.0   90.0 90.0 90.0
  CUTOFF
    70
  CHARGE
    0
&END