LiF strange discontinuity in energy-lattice constant curve

[Casey Brock]

I am using Elk 2.3.22 to calculate the lattice constant of lithium fluoride. I have calculated the energy at different lattice constants so I can fit an equation of state, but the energies near the expected lattice constant are deviating strongly from the expected curve.

I am wondering if it is a bug in the code, as it's similar to this problem from a much earlier release:
https://sourceforge.net/p/elk/discussion/897820/thread/e8f1ea1b/?limit=25#3980

Attached is the calculated energy plotted against lattice constant. You can see that the discontinuity is near the experimental lattice constant of 7.6 bohr.

[Casey Brock]

Also, here's the input file for the point on the plot with the highest energy. The other input files are identical except for the scale.

tasks
  0

tforce
  .true.

! libxc   XC_GGA_X_PBE   XC_GGA_C_PBE
xctype
  100     101            130

lradstp
  1

sppath
 '/home/brockc/local/elk/elk_latest/species/'

scale
  3.794510

avec
  1.0  1.0  0.0
  0.0  1.0  1.0
  1.0  0.0  1.0

atoms
  2                                   : nspecies
  'Li.in'                             : spfname
  1                                   : natoms
  0.0 0.0 0.0    0.0 0.0 0.0          : atposl, bfcmt
  'F.in'                              : spfname
  1                                   : natoms
  0.5 0.5 0.5    0.0 0.0 0.0          : atposl, bfcmt

ngridk
  10 10 10

vkloff
  0.5  0.5  0.5

[J. K. Dewhurst]

Hi Casey,

Sorry for the delay. It appears that you are finding 'ghost bands' in your calculation.

I was able to get a smooth curve in two different ways:

First by increasing rgkmax:

tasks
  0

! libxc   XC_GGA_X_PBE   XC_GGA_C_PBE
xctype
  100  101  130

lradstp
  3

mixtype
  3

rgkmax
  8.5

gmaxvr
  20.0

trimvg
 .true.

sppath
 '../../species/'

scale
  3.7

avec
  1.0  1.0  0.0
  0.0  1.0  1.0
  1.0  0.0  1.0

atoms
  2                                   : nspecies
  'Li.in'                             : spfname
  1                                   : natoms; atposl below
  0.0 0.0 0.0
  'F.in'                              : spfname
  1                                   : natoms
  0.5 0.5 0.5

ngridk
  8 8 8

... or using autolinengy=.true.:

tasks
  0

! libxc   XC_GGA_X_PBE   XC_GGA_C_PBE
xctype
  100  101  130

lradstp
  3

mixtype
  3

autolinengy
 .true.

sppath
 '../../species/'

scale
  3.7

avec
  1.0  1.0  0.0
  0.0  1.0  1.0
  1.0  0.0  1.0

atoms
  2                                   : nspecies
  'Li.in'                             : spfname
  1                                   : natoms; atposl below
  0.0 0.0 0.0
  'F.in'                              : spfname
  1                                   : natoms
  0.5 0.5 0.5

ngridk
  8 8 8

You can simply set 'highq=.true.' and obtain a smooth curve.

Regards,
Kay.

[Casey Brock]

Thanks for the help, Kay.

I used the first input file you suggested and didn't see any improvements. As it turns out, I just needed to fix the muffin-tin radii and then my old input file gave me a smooth curve.

Casey