Quick Start - Command Line

While we strongly recommend PyBigDFT as the way to drive BigDFT calculations, you may nonetheless want to run calculations from the command line. If you have installed from source, you should make sure you have setup the proper environment variables using the following command:

source install/bin/bigdftvars.sh

Input File

Input files are in the yaml format. A simple example is:

!cat cmd_work/psys.yaml

dft: {hgrids: 0.3, ixc: LDA}
kpt:
  method: mpgrid
  ngkpt: [2, 2, 2]
posinp:
  cell: [2.867, 2.867, 2.867]
  positions:
  - Fe: [0.0, 0.0, 0.0]
  units: angstroem

To use more exotic exchange and correlation potentials, you will need to:

1. Lookup the XC code on the website of libxc, and prepend a minus sign before the exchange and correlation code.

2. Copy the pseudopotentials you wish to use to the calculation directory with the name psppar.ELEMENT. We have included some in bigdft/utils/PSPfiles/.

!cat pspwork/psys.yaml

dft:
  hgrids: 0.3
  ixc: -109134 # PW91
kpt:
  method: mpgrid
  ngkpt: [2, 2, 2]
posinp:
  cell: [2.867, 2.867, 2.867]
  positions:
  - Fe: [0.0, 0.0, 0.0]
  units: angstroem

!cat pspwork/psppar.Fe

Goedecker pseudopotential for Fe
   26  16  070301 zatom,zion,pspdat
10 11  2 2 2001 0  pspcod,pspxc,lmax,lloc,mmax,r2well
     0.36000000    2     6.75678916    -0.22883251                                  rloc nloc c1 c2
    3                                                                               nnonloc
     0.27826303    2     0.62950570     7.91313242                                  rs ns hs11 hs12
                                      -10.21581002                                             hs22
     0.25138338    2    -7.93213293     7.69707888                                  rp np hp11 hp12
                                       -9.10730654                                             hp22
                         0.09786820     0.08070002                                        kp11 kp12
                                       -0.09548555                                             kp22
     0.22285578    1   -12.38579937                                                 rd nd hd11
                         0.01036288                                                       kd11

Running the Code

To run the code, use a line like this:

!cd cmd_work ; $BIGDFT_ROOT/bigdft -n psys ; cd ..

 <BigDFT> log of the run will be written in logfile: ./log-psys.yaml

Notice the -n option gives the name of the input file. If nothing is given, bigdft will look for default.yaml.

The skip command will check if the calculation has completed, and if so will immediately exit.

!cd cmd_work ; $BIGDFT_ROOT/bigdft -n psys -s yes ; cd ..

 <BigDFT> Run already performed, found final file: forces_psys.yaml
 Memory Consumption Report:
   Tot. No. of Allocations:  0
   Tot. No. of Deallocations:  0
   Remaining Memory (B):  0
   Memory occupation:
     Peak Value (MB):  0.000
     for the array: null
     in the routine: null
     Memory Peak of process: unknown
 Walltime since initialization:  00:00:00.272820000
 Max No. of dictionaries used:  1119 #( 1044 still in use)
 Number of dictionary folders allocated:  1

Results

The following files are produced by a calculation.

!ls cmd_work/*psys*

cmd_work/forces_psys.yaml  cmd_work/psys_minimal.yaml
cmd_work/log-psys.yaml     cmd_work/time-psys.yaml
cmd_work/psys.yaml

The log file contains the essential calculation information in a yaml format.

!grep "Energy (Hartree)" cmd_work/log-psys.yaml

 Energy (Hartree)                      : -2.05484837694368423E+01

The log file contains a list of all input variables and their values, not just the ones you set. They have comments beside them. This can be useful when trying to find out how to modify the calculation. For example, what was the spin of the calculation?

!grep "nspin" cmd_work/log-psys.yaml

   nspin                               : 1 #      Spin polarization treatment

In the case of geometry optimization, each step can be found in the same logfile.

!cat cmd_work/geom.yaml

dft: {hgrids: 0.35, ixc: PBE}
geopt: {"method": SQNM}
posinp:
  cell: [.inf, .inf, .inf]
  positions:
  - H: [0.0, 0.0, 0.0]
  - H: [0.0, 0.0, 0.741]
  - He: [5.292, 0.0, 0.0]
  units: angstroem

!cd cmd_work ; $BIGDFT_ROOT/bigdft -n geom -s yes ; cd ..

 <BigDFT> log of the run will be written in logfile: ./log-geom.yaml

The details of the geometry optimization procedure are also available in the data directory.

!ls cmd_work/data-geom

geopt.mon        posout_0002.yaml posout_0005.yaml posout_0008.yaml
posout_0000.yaml posout_0003.yaml posout_0006.yaml posout_0009.yaml
posout_0001.yaml posout_0004.yaml posout_0007.yaml time-geom.yaml

!cat cmd_work/data-geom/geopt.mon

  #-------------- Geopt file opened, name: ./data-geom/geopt.mon, timestamp: 2022-07-29 15:26:53.157
# COUNT  IT  GEOPT_METHOD  ENERGY                 DIFF       FMAX       FNRM      FRAC*FLUC FLUC      ADD. INFO
    0     0  GEOPT_SQNM   0.00000000000000E+00   0.00E+00  6.497E-03  9.18E-03  0.00E+00  0.00E+00   beta=1.00E+00 dim=000 maxd=0.0E+00 dsplr=0.00000E+00 dsplp=0.00000E+00
    1     1  GEOPT_SQNM  -4.05551003369715E+00  -4.06E+00  1.758E-03  2.54E-03  5.45E-04  5.45E-04   beta=1.00E+00 dim=000 maxd=6.5E-03 dsplr=9.17854E-03 dsplp=9.17854E-03
    2     2  GEOPT_SQNM  -4.05551478890277E+00  -4.76E-06  6.313E-04  7.75E-04  5.42E-04  5.42E-04   beta=1.10E+00 dim=001 maxd=2.4E-03 dsplr=1.26091E-02 dsplp=1.26091E-02
    3     3  GEOPT_SQNM  -4.05552404700316E+00  -9.26E-06  5.266E-04  8.33E-04  5.05E-04  5.05E-04   beta=1.21E+00 dim=002 maxd=1.1E-02 dsplr=2.63921E-02 dsplp=2.63921E-02
    4     4  GEOPT_SQNM  -4.05552808416222E+00  -4.04E-06  7.414E-05  1.11E-04  4.20E-04  4.20E-04   beta=1.33E+00 dim=002 maxd=2.0E-02 dsplr=5.13985E-02 dsplp=5.13985E-02
    5     5  GEOPT_SQNM  -4.05553137818127E+00  -3.29E-06  8.332E-06  1.13E-05  3.48E-04  3.48E-04   beta=1.13E+00 dim=002 maxd=3.4E-03 dsplr=5.55781E-02 dsplp=5.55781E-02
    6     6  GEOPT_SQNM  -4.05553064847653E+00   7.30E-07  6.759E-06  9.24E-06  2.91E-04  2.91E-04   beta=1.24E+00 dim=002 maxd=4.0E-04 dsplr=5.60659E-02 dsplp=5.60659E-02
    7     7  GEOPT_SQNM  -4.05553065682987E+00  -8.35E-09  7.833E-06  1.01E-05  2.46E-04  2.46E-04   beta=1.06E+00 dim=002 maxd=2.7E-04 dsplr=5.63927E-02 dsplp=5.63927E-02
    8     8  GEOPT_SQNM  -4.05553138514531E+00  -7.28E-07  3.377E-06  4.35E-06  2.09E-04  2.09E-04   beta=1.16E+00 dim=002 maxd=4.0E-04 dsplr=5.68876E-02 dsplp=5.68876E-02
    9     9  GEOPT_SQNM  -4.05553138634947E+00  -1.20E-09  3.719E-06  4.61E-06  1.80E-04  1.80E-04   beta=1.00E+00 dim=002 maxd=1.2E-04 dsplr=5.70406E-02 dsplp=5.70406E-02
SQNM converged at iteration  9. Needed bigdft calls:  9

Profiles

A number of profiles are available which can be useful for more advanced calculations. For example, the linear profile for O(N) calculations or the mixing profile can be imported for difficult to converge systems.

!cat cmd_work/*mixing*

dft: {hgrids: 0.3, ixc: LDA}
kpt:
  method: mpgrid
  ngkpt: [2, 2, 2]
posinp:
  cell: [2.867, 2.867, 2.867]
  positions:
  - Fe: [0.0, 0.0, 0.0]
  units: angstroem
import: "mixing"

!cd cmd_work ; $BIGDFT_ROOT/bigdft -n mixing -s yes ; cd ..

 <BigDFT> log of the run will be written in logfile: ./log-mixing.yaml

Full details are available in bigdft/src/input_variables_definition.yaml

PyBigDFT Compatability

All of these calculations are compatabile with the PyBigDFT approach. For example, we can build a Logfile manually:

from BigDFT.Logfiles import Logfile
from os.path import join

log = Logfile(join("cmd_work", "log-psys.yaml"))
print(log.energy)

-20.548483769436842

And even get access to the system.

from BigDFT.Systems import system_from_log
sys = system_from_log(log)
for fragid, frag in sys.items():
    print(fragid)
    for at in frag:
        print(at.sym, at.get_position())

ATOM:0
Fe [0.0, 0.0, 0.0]