Below are the notes from the HackMD page during the workshop.

General Questions (day 1)

VASP main page
VASP wiki manual
VASP tutorials and recordings

• Q: machine learning force-fields implemented in VASP 6.3
  • A: See this link

Hands-on Questions (day 1)

• Q: POTCAR files are missing in input directory?
  • A: You may find the POTCAR files in the /software/sse/manual/vasp/POTCARs/PBE/2015-09-21/ directory. You'll need to copy the relevant one in your running directory.
• Q: Logging in to Tetralith for the first time?
  • A: Follow the guide here: https://www.nsc.liu.se/support/getting-started/
• Q: Quick guide to Tetralith:
  • A: The quick guide link
  • A: Reservation names in quick guide are invalid, you need to skip dash in "vaspworkshop" (so usereservation vaspworkshop-2022-02-22, without dash).
  • A: Type the command usereservation vaspworkshop-2022-02-22 in the terminal to make use of the course reservation. This way your jobs will start faster. Tomorrow you'll need to: usereservation vaspworkshop-2022-02-23.
• Q: How do we know we're using the latest vasp?
  • A: before loading: check version number and date in the list given by module avail vasp.
  • A: load the module with module load VASP/... then you can check at any time what is loaded with module list.
• Q: How can I see which reservations I am allowed to use?
  • A: Type the command listreservations.
• Q: Where do I find the examples to be used in the course?
  • A: /software/sse/manual/vasp/training/ws2022/. You may just copy the directory somewhere else.
• Q: I get permission denied when I try to copy the workshop directory / POTCAR files.
  • A: Type groups to confirm that you're member of the group pg_vasp-ws2022 and vasp61. If it is not shown in the list, but you are member of the SUPR prject for the course, re-login to Tetralith. If you still have issues, write here or in the Zoom chat.
• Q: small mistake in fcc_Si example, section "1. Calculation": cpr -r should be cp -r.
  • A: Thank you for pointing it out. I'll ask Weine to correct it.
• Q: How to calculate the effective mass and optical properties?
  • A: For the effective mass, you need to do a band structure calculation first and then extract m*. There is a sample workflow here, although I have not tested it myself. Use with care.
  • A: For optical properties you may check this tutorial at the VASP wiki
  • A: github link

• Q: I get the following message after loading ASE:

  IMPORTANT INFORMATION
  =====================
  You HAVE to load a suitable Python installation together with ASE 3.19.0! e.g.:
  
  Python/3.6.3-anaconda-5.0.1-nsc1

  • A: This is expected and shown even with Python already preloaded (apparently, module loading script does not check for this dependency).
• Q: Regarding the ENCUT values, it was mentioned that the ideal is around 1 -1.5x of the ENMAX value provided. Should we do convergence testing to get the ideal ENCUT value?
  • A: Indeed, this is valuable info related to the topic on PREC
  • A: A bit longer answer - yes, one need to test in theory the ENCUT with respect to any of the properties that are calculated, but more or less they converge at the same rate. In practice, it might be rather different story especially if the you calculate properties while the cell volume changes... or for system with weak interactions (dispersion).
  • A: Please remember that the accruracy of your simulations will also depend on the k-mesh, NBANDS, and the algorithm you're using for the convergence. It's always good to do some tests and find the optimal set for the quantity that you're instrested in calculating.
• Q: How to plot the band structure with gnuplot or other software?
  • A: One example for Si in this workshop
  • A: Then you may use p4vasp or py4vasp to plot.
  • A: A video example
  • Make sure you use a fine-enough k-mesh.
  • ASE
  • sumo-bandplot thumbsup x2
• Q: How to relax the structure with VASP?
  • A: ISIF tag at VASP wiki
  • A: A related important read, energy vs volume Volume relaxations and Pulay stress
  • A: link to ISIF table
  • Always good to double check ISIF > 0 calculations with some single-point calculations (=no relaxations) or calculations with only limited degrees of freedom, especially if the structure is complicated / has few symmetries.
• Q: For convergence testing, what are the parameters that I should test for? Besides ENCUT and Kmesh?
  • A: NBANDS, convergence algorithm, EDIFF, EDIFFG,...
  • A. Also look into PREC, if it needs to be set to "Accurate". Also look into what other parameters it determines. In particular cases one might need to set those directly.
• Q: Where do I find good database for POSCAR files?
  • A: You may get the .cif from the ICSD database and then you may generate the POSCAR with cif2cell, for example.
  • A: cif2cell can be found here, it's also available as a module on Tetralith
  • A: Also good database for experimental structures, Crystallography Open Database
  • A: I think whole calculations can be found on the NOMAD database (including many for VASP), see link
  • A: Computational Materials Repository

General Questions (day 2)

• Q: How does VASP deal with i-th singlet (and triplet) excited states? What about its efficiency compared to other packages?
  • A: [PM:] I have no experience...
  • A: [Weine]: Sorry, I also don't know so much about this topic. Do you mean excited states which are fixed in a regular DFT calculation or using e.g. BSE?
• Q: How should the inputs look like when performing excited state calculations?
  • A: [PM:] I have no experience...
  • A: [William]: Set ISMEAR=-2 so the occupancy does not change during the run, and use FERWE and FERDO to set which states are occupied (your excited state). Then I think you need to start the run from the CONTCAR and WAVECAR from a completed ground state calculation. Check EIGENVAL file from ground state calculation to see how you want to change the occupancy.
  • A: [Weine]: Also, probably use the latest version of VASP, from 6.3, since there was a bug fix recently. Previous version 5.4.1 was also ok.

Hands-on Questions (day 2)

• Q: Another typo in Quick guide: command to start an interactive job should begin with interactive (missing 'c'). It is there since yesterday, but one tends to correct, submit and forget shell commands immediately afterwards, because typing in something wrong the first time is extremely common.
  • A: Thanks! Now fixed (might take a few min, and might need to reload page)
• Q: Is it good to run with single KPOINTS even when calculating DOS?
  • A: Please make difference for the KPOINTS file during the electronic structure step and the second step. Single k-point is ok only for large supercells (ideally the bands are already flat, so sampling at a single point will be enough). For the second step, you can choose to sample only at the gamma point or another k-point, although this does not sound as DOS (Density Of States) anymore but rather like band levels. For larger cells, this will lead to splits which will start to look again like DOS when they get in to a histogram plot.
• Q: PDF plot in gnuplot using the command line/script
  • A:

  gnuplot
  ...
  
  set term pdfcairo
  set out "filename.pdf"
  
  replot   # or plot with the original plot command
  
  # to flush the buffer, not necessary if you close or exit the program
  set out ""
  
  # to revert back to screen
  set term qt
  
  ...