ab initio md of lithium fluoride

[Michael Mitchell]

Hello all:

I'm a newbie from the classical MD world. Recently, I've been attempting to run QS dynamics on the rock salt structure of LiF at 300K to get my feet wet. My simulations are on the 2x2x2 supercell of the cubic crystal structure. The simulation runs and the thermodynamic properties (temp, energy etc) are all converged after an initial equilibration, however the system melts, transforming from the rock salt structure into an amorphous blend after only 1ps or so of dynamics. Since this structure has been simulated before with QS, I must be doing something wrong. Can anybody see anything obviously wrong with the following input?

Best,

Mike

============================================================================================

@SET MD_SAVE_FREQ  1

@SET LA 8.2

@SET LB 8.2

@SET LC 8.2

@SET RTEMP 298

@SET PREFIX LiF.2x2x2

@SET NSTEPS 10000

&GLOBAL

 PRINT_LEVEL medium

 PROJECT ${PREFIX}.${RTEMP}

 RUN_TYPE MD

 WALLTIME 36000000

&END GLOBAL

&MOTION

 &CELL_OPT

  EXTERNAL_PRESSURE 1.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0

  KEEP_ANGLES TRUE

  OPTIMIZER BFGS

  TYPE MD

  &PRINT

   &CELL

    FILENAME ./${PREFIX}.${RTEMP}K

    &EACH

     CELL_OPT ${MD_SAVE_FREQ}

    &END EACH

   &END CELL

  &END PRINT

 &END CELL_OPT

 &PRINT

  &VELOCITIES

   FILENAME ./${PREFIX}.${RTEMP}K

   FORMAT XYZ

   &EACH

     MD ${MD_SAVE_FREQ}

   &END EACH

  &END VELOCITIES

  &TRAJECTORY

   FILENAME ./${PREFIX}.${RTEMP}K

   FORMAT XYZ

   &EACH

     MD ${MD_SAVE_FREQ}

   &END EACH

  &END TRAJECTORY

 &END PRINT

 &MD

  ENSEMBLE NVT

  STEPS ${NSTEPS}

  TIMESTEP 0.5

  TEMPERATURE ${RTEMP}

  &LANGEVIN

   GAMMA 0.0001

  &END LANGEVIN

  &THERMOSTAT

   REGION MOLECULE

   TYPE NOSE

   &NOSE

    LENGTH 3

    YOSHIDA 3

    TIMECON 100

   &END NOSE

  &END THERMOSTAT

 &END MD

&END MOTION

&FORCE_EVAL

 METHOD Quickstep

 &DFT

  CHARGE 0

  MULTIPLICITY 1

  WFN_RESTART_FILE_NAME ${PREFIX}.${RTEMP}.md.wfn.restart

  BASIS_SET_FILE_NAME /global/home/users/mike/codes/cp2k/tests/QS/BASIS_SET

  POTENTIAL_FILE_NAME /global/home/users/mike/codes/cp2k/tests/QS/GTH_POTENTIALS

  &MGRID

   CUTOFF [Ry] 320

  &END MGRID

  &PRINT

   &MULLIKEN

    FILENAME ./${PREFIX}.${RTEMP}

    &EACH

     MD 1

    &END EACH

   &END MULLIKEN

  &END PRINT

  &QS

   EPS_DEFAULT 1.0E-12

   EXTRAPOLATION ASPC

   EXTRAPOLATION_ORDER 3

   MAP_CONSISTENT

   METHOD GPW

  &END QS

  &SCF

   EPS_SCF 1.0E-5

   MAX_SCF 50

   SCF_GUESS ATOMIC

   &DIAGONALIZATION

    ALGORITHM DAVIDSON

   &END DIAGONALIZATION

   &OUTER_SCF

    EPS_SCF 1.0E-5

    MAX_SCF 20

   &END OUTER_SCF

   #&OT

   # MINIMIZER DIIS

   # PRECONDITIONER FULL_ALL

   #&END OT

  &END SCF

  &XC

   &XC_FUNCTIONAL PADE

   &END XC_FUNCTIONAL

  &END XC

  &POISSON

   PERIODIC XYZ

   POISSON_SOLVER PERIODIC

  &END POISSON

 &END DFT

 &SUBSYS

  &CELL

   ABC ${LA} ${LB} ${LC}

   PERIODIC XYZ

  &END CELL

  &COORD

   @INCLUDE ${PREFIX}.cp2k.xyz

   SCALED .TRUE.

  &END COORD

  &KIND Li

   BASIS_SET DZV-GTH-PADE

   POTENTIAL GTH-PADE-q3

   #BASIS_SET DZVP-GTH-PADE

   #POTENTIAL GTH-PADE-q3

  &END KIND

  &KIND F

   BASIS_SET DZVP-GTH-PADE

   POTENTIAL GTH-PADE-q7

  &END KIND

 &END SUBSYS

&END FORCE_EVAL

[Matt W]

Hi Mike,

I can't see anything obviously wrong with you setup - especially if it runs fine for a ps or so it is hard to spot with what you've provided.

How do the energy and temperature behave later on?

The only things I can suggest are -  could the nose thermostat can be a little unstable for equilibriation, you could try the CSVR thermostat with a short time constant? Perhaps your system is too small for a gamma point only calculation?

Matt