Molecule explodes when running quantum (Born-Oppenheimer) molecular dynamics in parallel


Clicked A Few Times
Hi,

I tried unsuccesfully to run a DFT QMD (quantum molecular dynamics) job of CO2 using the following input. The job ran perfectly in serial mode, but when I tried to run it in parallel the molecule exploded after a few hundred cycles. The molecule just starts to move weird and then explodes and the program returns an error due to lack of convergence of the electronic structure.
I tried OpenMPI versions 1.4.3 and 1.4.2 and got the same results.
I ran this job on my Athlon II X2 laptop, on an Intel Core i7 and a Core 2 Quad with the same results.
The comand for running was

mpirun -np 4 nwchem input.nw
It's certainly not the timestep, as I checked it with a very small dt.
Any ideas about why this happens?

Thank you very much

Diego

_______________________________________________________
Title "Dinamica moleccular DFT de CO2 en vacio"

permanent_dir ./co2
scratch_dir ./scratch

Start CO2

echo

charge 0

geometry noautosym noautoz units angstrom
C     0.00000     0.00000     0.00000
O 1.50000 0.00000 0.00000
O -1.50000 0.00000 0.00000
end

ecce_print ecce.out

basis "ao basis" cartesian print
 O library "3-21G*"
C library "3-21G*"
end

dft
 mult 1
XC b3lyp
mulliken
print low
end

md
 system CO2
cutoff 5.0
noshake solute
equil 1000 data 1000 step 0.002
isotherm 1.0 300.0 anneal 0.0 1.0
record rest 1000 scoor 1 prop 1 svelo 1
print step 1 stat 1000 energies 1
update motion 1000
end

task dft dynamics
__________________________________________________________


Forum >> NWChem's corner >> Running NWChem