[Q-e-developers] Fwd: QE-patch for unscreened hybrid vdW-DF

Nicola Marzari nicola.marzari at epfl.ch
Wed Sep 6 10:34:54 CEST 2017



Dear All,

received this from Per Hyldgaard and Jeff Neaton - hope it's acceptable, 
of interest?

			nicola



-------- Forwarded Message --------
Subject: 	QE-patch for unscreened hybrid vdW-DF
Date: 	Wed, 6 Sep 2017 06:14:21 +0200
From: 	Per Hyldgaard <hyldgaar at chalmers.se>
To: 	Nicola Marzari <nicola.marzari at epfl.ch>
CC: 	hyldgaar at chalmers.se, Jeffrey Neaton <jbneaton at lbl.gov>



Dear Nicola,

(Please redirect if we need to send this elsewhere).

As you know, we are developing both unscreened hybrid vdW-DF
and range-separated hybrid vdW-DFs. We are still working on the
paper for range-separated hybrid vdW-DFs, but the unscreened hybrid vdW-DF
is successfully launched (JCP 146, 234106 (2017); Swedish Theoretical 
Chemistry 2017,
DFT2017; a follow up on asserting the Fock mixing ratio is also coming).
I have been asked to install the unscreened hybrid vdW-DF on the set of
Swedish National Computational facilities.

The unscreened hybrid obviously requires nothing but some "funct.f90" 
rewriting
in QE -- whereas it handled in input files in VASP.  We would like 
people to try it
(and not just in VASP) and so we hope that we can submit the smallish QE 
changes
("mini-"patch is attached) for unscreened hybrids now (and come back
with rest of the range-separated coding later).

Some comments:

*) vc-relax in combination with hybrids vdW-DF is not thoroughly tested 
-- but we note
that hybrids+vc-relax is officially turned off in any case. Apart from
the issue mentioned next, there is no Ecnl-specific reason that hybrid 
vdW-DF
would not work with vc-relax (it is the Fock-exchange part that is under 
development)

*) An issue with spin vdW-DF (as opposed to spin-unpolarized vdW-DF) is that
there is not yet a code for stress-in-spin-Ecnl. That is again outside the
hybrid scope. My postdoc Yang Jiao was going to work on it (not hard,
but needs to be done), but she is now due very soon and going on
parental leave.

To not hamper development/testing of vc-relax/hybrid vdW-DF we suggest the
following (also implemented in the attached) hack-for-developers (nobody 
else will see it):
By setting esm_debug='true', one allows (and is warned that) Ecnl to be 
computed on the
total electron density and not on the spin-polarized densities (as it really
should be in spin vdW-DF but which is not in other codes). That means
that vc-relax-plus-hybrid-vdW-DF can be deveoped/tested in general even
without first writing code for stress-in-spin-Ecnl.

Of course, we have always (in haouse) had this "esm-debug option" in
place as it allows us to chart the extent that it matters to have the
consistent spin vdW-DF formulation or use the "spin-balancing"
approximation that is presently coded in VASP. One case where
avoiding the approximation clearly does matter is atoms, and hence
in cohesive energies of bulk matter, PRB*95*, 085147 (2017)... but that 
should
not stop testing/developments.

The attached patch just inserts our small "esm_debug" developers hack 
into the main code
(a few lines here and there).

Best regards
Per




-- 
----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
Director, National Centre for Competence in Research NCCR MARVEL, EPFL
http://theossrv1.epfl.ch/Main/Contact http://nccr-marvel.ch/en/project
-------------- next part --------------
A non-text attachment was scrubbed...
Name: qe-6.1_unscreened_hybrid_vdW-DF_V1.patch
Type: text/x-patch
Size: 23344 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20170906/33773834/attachment.bin>


More information about the developers mailing list