LCAO complex without the phase factor

[anbenali]

This PR enables the complex build of LCAO branch with QMC_COMPLEX.
The Phase factor is however not applied yet as we do not know exactly how to handle it for now. However, if the phase factor is known to be 1 (the Gamma point case), results are correct.
This mainly enables the use of Kpoints and promitive cells with supertwists from Pyscf.

-In terms of changes:

Converter:

• Since the tiling is done inside the Pyscf converter, carrying the twist number in the naming of the input files does not make sens any more and was removed.
• shape of NAO and NMO was fixed to close bug Cannot reproduce PySCF diamondC_Xx1x1-Gaussian-pp Workflow #1422
• Coordinate of supertwist passed to HDF5 and compared against twist tag in xml Input
• Data stored under SuperTwist instead of Kpoint0 (backwards compatibility for molecules preserved)
• Remove abort in converter if Wavefunction is complex. Instead puts IsComplex Flag to True and stores Imaginary part of the WF in a Real array.

QMCPACK:

• Reading Real Coeff Matrix and Complex Coeff Matrix from two function that should be templated into one function with specialization.

• Passing parameters (PBCimages, twist) to evalV and evalVGH, in prevision of the application of the phase factor.

• Creating a fiunction in the builder evaluating exp(ik.g) where k is the supertwist and g are the translation vctors (PBCImages). Not called yet as it is yet to be confirmed.

TEST CASES:

• Update solids H5 files with new Nomenclature (storage of all data under SuperTwist instead of KPOINT0). Since I am the only one who used production wise this branch of the code it was easier to just update examples. data is however the same.

• Change completely the diamondC_2x1x1-Gaussian_pp_kpts built using a primitive cell and 2 kpoints with a complex wavefunction with null imaginary part. Trial wavefunction comes from HF which allows a direct comparison of accuracy using VMC and No Jastrows.
  Test works with the real code and complex code.

This PR cancelled the preious WIP PR.

[qmc-robot]

Can one of the maintainers verify this patch?

[prckent]

OK to test

[ye-luo]

I'm going to try template the LoadCMatrix after you address my request. In this case, serialization works better.

[prckent]

Unless objections are posted, will merge first thing Monday. Needed due to workshop schedule.

[anbenali]

You mean the 1e-6? It is just slightly below machine precision in single precision. To specify it outside we might need to have it defined at compile time but we could indeed specify it somewhere.

…

On Sat, May 4, 2019, 15:44 Paul R. C. Kent ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In src/QMCWaveFunctions/lcao/LCAOrbitalBuilder.cpp <#1578 (comment)>: > aAttrib.put(cur); if (cur != NULL) aAttrib.put(cur); + if (std::abs(twist[0]-0.0) > 1e-6 && std::abs(twist[1]-0.0) > 1e-6 && Better to #define the magic number as well — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#1578 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF4XVQ2RZLPSPWWXP2JZFWTPTXYUFANCNFSM4HKXEXCA> .

[prckent]

Yes. Good to define all constants with a meaningful name vs burying them in the source.

[ye-luo]

@anbenali in the kpt test folder "C_Diamond-211.wfnoj.xml" is missing.

[ye-luo]

Q: what is the difference between diamondC_2x1x1-Gaussian_pp and diamondC_2x1x1-Gaussian_pp_kpts?

[anbenali]

Super cell using gamma and real wavefunction with DFT trial wavefunction VS primitive cell with kpoints and a complex wavefunction with a ~ 0 imaginary part.

…

On Sun, May 5, 2019, 11:16 Ye Luo ***@***.***> wrote: Q: what is the difference between diamondC_2x1x1-Gaussian_pp and diamondC_2x1x1-Gaussian_pp_kpts? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1578 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF4XVQZXXBU3LNLCTSBE7TDPT4B4DANCNFSM4HKXEXCA> .

[prckent]

@ye-luo OK?

[ye-luo]

@anbenali Could you add C_Diamond-211.wfnoj.xml in kpt folders?
@prckent This is the only thing left.

[anbenali]

Sorry for the delay. They are added now.

…

On Mon, May 6, 2019 at 9:10 AM Ye Luo ***@***.***> wrote: @anbenali <https://github.com/anbenali> Could you add C_Diamond-211.wfnoj.xml in kpt folders? @prckent <https://github.com/prckent> This is the only thing left. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1578 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF4XVQ5FVYQU6KXHKOA2R73PUA35RANCNFSM4HKXEXCA> .

-- ----------------------- Anouar Benali, PhD Leadership Computing Facility Argonne National Laboratory Building 240 Office - 2127 9700 S Cass Av., Argonne Il, 60439 (630) 252-0058