ZORA & Memory/Hard Disk Space Problems


Just Got Here
I'm trying to calculate paramagnetic NMR shielding constants using NWChem on a small inorganic anion containing 25 atoms. Using smaller basis sets, the calculation is slow but manageable. I wanted to use a larger basis set, so I have switched to Def2-TZVP with the following input file:
echo
start nmrepr
title "Paramagnetic NMR/EPR Calculation"
memory 1500 mb

charge -3
geometry noautoz units angstrom nocenter
  symmetry c1
  load geom.xyz
end

basis "aobasis" print
  * library Def2-TZVP
end

relativistic
  zora on
  zora:cutoff_NMR 1d-8
  zora:cutoff 1d-30
end

dft
  odft
  mult 3
  xc pbe0
  iterations 100
  convergence lshift 0.75 rabuck 30
end

property
  gshift
  hyperfine
  shielding
end

task dft property

For the sake of comparison, I'd like to get the numbers from this calculation before adding diffuse functions. Unfortunately, NWChem is making *enormous* files on my hard drive and crashing my computer. Before even getting to SCF convergence, I get an error:
eaf_write: rc ne bytes -1999 bytes 524288
   IO offset    13682343936.000000
  IO error message >Write failed                   0

Checking my hard drive, NWChem has created eight nmrepr.aoints.* files (one for each processor) that are each 14 GB in size. Is there some way to avoid this? I should hope that 25 atoms is not too many to use a triple-zeta basis set.

If I leave out the relativistic block, the calculation runs with no problems. However, I can't get g or hyperfine tensors without ZORA.

Forum Vet
If you add the keyword direct to the dft input section, the code will no longer use the "*aoints*" files

dft
  odft
  mult 3
  xc pbe0
  iterations 100
  convergence lshift 0.75 rabuck 30
  direct
end


http://nwchemgit.github.io/index.php/Release66:Density_Functional_Theory_for_Molecules#DIREC...


Forum >> NWChem's corner >> General Topics