Videoconference meeting winter 2018

From OctopusWiki
Jump to: navigation, search

This meeting took take place the 13 February at 18:00 CET.

Meeting URL: https://hangouts.google.com/call/aIH7NDZjCNvRRIqhi-yHAAEI

Topics to discuss

  1. New release.
  2. Status of solids in Octopus
  3. Hybrid functionals
  4. Buildbot status.
  5. Complex scaling code

Present

Ask, David, Jose Rui, Micael, Miguel, Nicolas, Quim, Shunsuke, Xavier

Meeting Minutes

  • New release.
    • We should release Octopus 8 in the next couple of months
    • Micael and Nicolas: LDA+U should be in the new release
    • Xavier: LDA+U might need more testing before release
  • Geometry optimization
    • Xavier: there are some usability issues with the geometry optimization. These should be fixed before the release.
    • Nicolas: there are cases when the restart density is inconsistent with the last written geometry.
    • Quim: results with FIRE algorithm are consistent with other codes
    • Xavier: we should change the default GO algorithm to FIRE. All agree.
    • David: use of minimum box with GO can cause problems as atoms approach the boundaries. This should be well documented and the code should warn the user.
  • Status of solids in Octopus
    • Nicolas: Octopus is able to compute a delta-factor similar to other codes for the SG15 pseudopotential set.
    • Nicolas: Symmetries are finaly working for paramagnetic systems.
    • Shunsuke: k-point definition is not the conventional one (minus sign missing)
    • Mixed-periodicity needs to be checked.
    • Nicolas: remove experimental status for non-orthogonal cells, symmetries, SG15 pseudopotential set and file format.
    • Shusuke: there are still cases when stencil for non-orthogonal cells gives wrong results.
    • Xavier: we should try to make the Pseudo-Dojo pseudopotentials work.
    • Xavier: there should be tutorials about solids using non-orthogonal unit cells. We can start by putting some input files on the wiki.
    • Miguel: Octopus should be able to use the standard k-points paths from Curtarolo and use standard unit cells.
    • Xavier: we should write a paper on the solids implementation. Nicolas can contact the other authors.
  • Hybrid functionals
    • Nicolas: hybrids are working with k-points.
    • Nicolas: short-range screened hybrids are working.
    • Nicolas: current implementation is very slow. Not obvious how to use the approach of Lin Lin in real-space.
    • Nicolas: Lin Lin was contacted about the time-dependent case.
  • Buildbot status
    • Micael: we finally have several missing pieces of the hardware (PowerPC, GPU's) and all the compiler licenses to add builders using the PGI and IBM compilers and with CUDA/OpenCL support.
  • Complex scaling code
    • Micael and Nicolas: the complex scaling code has become more and more difficult to mantain. It should be removed unless someone is actually using it.
    • Xavier: We remove it unless someone comes up with a proposal to do something with it before Benasque.
    • Ask: not clear if there is or not a proposal to do something else with the code. Need to discuss with Angel.
  • Varia
    • Next Developers Workshop should happen this year. One option is to have it next to Benasque. Another option would be to organize it in NYC, at the Flatiron Institute. Angel needs to agree on the later option because of travel expenses.