Symmetry and sampling in reciprocal space | VASP Lecture

Martin Schlipf gives a beginner-friendly introduction to symmetry and how it affects sampling in reciprocal space. Then, he elaborates on specific considerations when using VASP, e.g., how to do a band-structure calculation, converging with respect to the number of k points, setting up k-point sampling for hybrid functionals, and more.
0:00:00 Introduction
0:01:03 Construct reciprocal lattice
0:02:50 Translational invariance
0:05:56 Computational advantage of reciprocal space
0:07:10 Outline of the lecture
0:07:40 K-point sampling: converging with respect to the number of k points
0:10:50 Surfaces/slabs, wires, isolated molecules
0:13:30 Symmetry considerations:
0:24:30 Format of the KPOINTS file
0:28:38 Density of states, type of smearning
0:41:49 Band-structure calculations
0:48:43 Hybrid functionals
0:49:28 KPOINTS_OPT
0:51:47 Q&A
0:52:08 Does it suffice to use vasp_gam to compute the band gap at the Gamma point for a bulk semiconductor?
0:53:33 For k meshes generation with the Monkhorst-Pack method, is it "saver" to use even or odd number of k points?
0:55:55 What is the difference between the KPOINTS_OPT file and the IBZKPT file?
0:59:05 How to set up the k points for a band-structure calculation using a hybrid functional?
1:02:42 What is downsampling in the context of hybrid functionals?

[@jacnexshin7251]

The VASP worldwide users need this type of interaction! Thanks, martin ! etc. KZbin is much better than a document.

[@martinschlipf5677]

Glad you liked it

[@youmu3608]

Thank you, Martin. Your lecture gave me a clear concept about the different smearing methods which I didn't understand for a long time. Again, ty for the nice talk and explanations.

[@weixie4399]

Thanks Martin!

[@HyperFocusMarshmallow]

Very useful lecture! Thanks!

[@HyperFocusMarshmallow]

To what extent would it be possible to improve the defaults so that the need for explicit choices would be reduced? What might the lowest hanging fruits be on that issue? To what extent might it be possible to improve the naming and discoverability of options so that people could learn to do more advanced computations more effectively faster?