My Community

Dear all,
I came with some problems while calculating large unitcell structures.
I got to calculate the bandstructure and DOS of the structure with a large unitcell which contains 300 atoms or more. Here are the INCAR and KPOINTS:
INCAR:
ENCUT = 400
ISTART = 1
ICHARG = 11
PREC = Accurate
NBANDS = 704
ISMEAR = 0
SIGMA = 0.05
GGA = PE
ALGO = Fast
LPLANE = .TRUE.
LVDW=.TRUE.
LREAL = Auto
LORBIT=11

KPOINTS:
k-points along high symmetry lines
10
Line-mode
Reciprocal
0.500 0.000 0.000 1.0 ! M
0.000 0.000 0.000 1.0 ! Gamma

0.000 0.000 0.000 1.0 ! Gamma
0.500 0.500 0.000 1.0 ! K

0.500 0.500 0.000 1.0 ! K
0.000 0.500 0.000 1.0 ! N

0.000 0.500 0.000 1.0 ! N
0.000 0.000 0.000 1.0 ! Gamma

while the calculation always got stuck here and never went on:
serial 3D FFT for wavefunctions
parallel 3D FFT for charge:
minimum data exchange during FFTs selected (reduces bandwidth)

After searching on the internet, somebody said this may be caused by lack of memory space. But I am pretty sure the memory of the machine is enough since I'm running on a cluster computer system and 128GB is allocated. Then I tried to recompile VASP adding -heap-arrays 64 into the makefile, but this same problem still remains.

I would really appreciated any hint to understand how to solve this problem.

Thanks,
Hanming

Hi Hanming,
Try to do each segment of k-path at once. For example try this KPOINTS:
k-points along high symmetry lines
10
Line-mode
Reciprocal
0.500 0.000 0.000 1.0 ! M
0.000 0.000 0.000 1.0 ! Gamma

hatdau wrote:Hi Hanming,
Try to do each segment of k-path at once. For example try this KPOINTS:
k-points along high symmetry lines
10
Line-mode
Reciprocal
0.500 0.000 0.000 1.0 ! M
0.000 0.000 0.000 1.0 ! Gamma

Dear hatdau
First thanks a lot for your reply. I have tried this method and with 300 atoms it does reduce some memory cost and get the consequence. But now I want to simulate the system with 1000+ atoms, this problem comes again.
I don't know why VASP has a limit on the memory cost of simulation. Or did I do something wrong while compiling or simulating?

Best
Hanming