Nexus features

[jtkrogel]

A number of new features have been requested informally for inclusion in Nexus. This issue page serves as a space to list/discuss new features or new workflow support of priority.

As specific issues of priority are identified, new issues will be created for each development priority to track progress.

Requests that have been made:

• [ ] Support for k-point symmetries (twist averaging over irreducible wedge) with QMCPACK (by @lshulen)
• [ ] Support for grand canonical twist averaging with QMCPACK (by @Paul-St-Young)
• [ ] Support for chaining of workflow portions into single jobs (by @lshulen)
• [ ] Support for FDLR workflow (by @eneuscamman)
• [ ] Support for arbitrary tiling with PySCF + QMCPACK (by @Paul-St-Young)
• [ ] Support for AFQMC workflows (by @mmorale3 )
• [ ] More comprehensive type guards for PWSCF inputs (by Ron Cohen)
• [ ] Improved optimal tiling matrix estimation (by @Paul-St-Young and @jtkrogel)
• [ ] Add generation of QMCPACK wall time input (by @kayahans )
• [ ] Improvements for archiving only necessary files (by @lshulen)
• [ ] Per calculation logging (by @anbenali )
• [ ] Easier setup for new machines, via documentation or more generic support (multiple people have expressed some level of interest)
• [ ] Support for command line post-processing of QMCPACK density, momentum distribution, etc (multiple interest)
• [ ] Harden failure detection for submitted simulation runs (by Ron Cohen)
• [ ] Allow overwrite and full submission of jobs when running for real after using generate_only mode (by Ron Cohen)

See also #321 (Nexus documentation) and #158 (Nexus tests)

[lshulen]

• Full support for symmetry in nexus. Would include identifying supercell twists that are equivalent by symmetry and also knowledge of their multiplicity. Future developments should allow for symmetrizing of observables (for example the charge density or structure factor)
• It should be possible to chain parts of the workflow that are dependent on each other. This would allow for instance optimization of one and two body jastrow factors followed by optimization of three body jastrow factors in the same job submitted to the scheduler. Another example would be quantum espresso scf calculations followed by nscf ones.
• An easy facility to seed wavefunction parameters. For instance, it would be nice if it were possible to have default starting one body and three body jastrow factors for each pseudopotential.

[eneuscamman]

For excited state work, both in molecules and solids, it would be very helpful to get Nick Blunt's input system into Nexus. To start, even just the GAMESS based workflow would be fantastic.

[kylanpaa]

Addition of QMC wall time to Nexus (I guess QMCPACK has now the feature of finishing before reaching wall time, not 100% sure though). This could be set automatically based on the asked wall time in order not to loose (expensive) data.

[prckent]

This might not be the correct location, but S(K) analysis was reported as "slow". Have examples been sent?