Adding new basis sets to the ECCE basis set tool
Gets Around
9:47:01 PM PDT - Thu, Jun 20th 2013
Gary and ECCE users,
I couldn't find an officially recommended way of adding basis sets to ECCE, and so ended up writing my own set of scripts to do so.

In case anyone else is interested in converting nwchem basis sets to the ECCE format, I've put the scripts at http://sourceforge.net/projects/nwbas2ecce/, together with ECCE basis set files for def2-svp, def2-svpd, def2-tzvp etc.

It's not elegant, but it works.

Gets Around
9:26:08 AM PDT - Wed, Jun 26th 2013
Very cool. Are you interested in contributing back the new basis sets to the open source distribution of ECCE? I'm planning a new ECCE release later in July and could make them part of the release. I haven't read through your documentation for what you did, but I'm assuming you must have figured out the basic process. On our end we use a script named gbsDavConverter script along with an executable named load_tgbs (in the src/apps/davutils directory). I don't think we make this gbsDavConverter script part of the open source distribution of ECCE, but I don't see why we couldn't do that--I'm guessing your nwbas2ecce script does some if not most of the same stuff as our gbsDavConverter. It's been so many years since the basis sets have been updated, as you discovered, that I can't remember myself how this all works. I do know that by running load_tgbs we get the binary "WebDAV formatted" version of the basis set files and we tar up those files and include them in the distribution. Then when ECCE is installed those files are extracted as part of the data server bundle. So getting your basis sets into the upcoming ECCE release may just be a matter of getting the *.BAS* and *.POT* files and I can run load_tgbs to create the binary .tar file that is part of the ECCE distribution.

Gary

Gets Around
12:27:33 AM PDT - Thu, Jun 27th 2013
If I can at all help with the development with ECCE I'm more than happy to do so.

There are two scripts:
nwbas2ecce generates the .POT and .BAS files. For a file like lanl2dz_ecp (as supplied in nwchem) the script recognizes the presence of both basis/end and ecp/end statements in the file and generates both the .POT and the .BAS, while for e.g. 3-21g it only generates .BAS and for def2-ecp it only gives a .POT file.
It also generates a file with the name of the basis set, the file the basis set is contained in, and the elements included -- that file can be used to expand the category files (e.g. pople, ECPOrbital etc.).
Finally, it also gives a rudimentary .meta file with the references -- it's far from perfect though.

eccepag generates the binary .pag file (i.e. reverse-engineered webDAV format) which goes in the .DAV directory. This is generated based on specified textual input i.e. it's separate from the files generated by nwbas2ecce and the accuracy of the content of the .pag file is up to the person generating it (e.g. whether a basis set is segmented, spherical, belongs to ECPOrbital or pople etc.).

I'd need someone to check the latter (and maybe general quality control) since I'm not a theoretical or computational chemist, and actually have a shockingly poor knowledge of these things.

But I'm happy to generate all the files that are needed. I also suspect that some of the current .BAS and .POT files in ECCE need to be regenerated due to errors being fixed etc.

I've put example files that I've generated for the def2-series in http://sourceforge.net/p/nwbas2ecce/code/ci/master/tree/example/

[deleted comment about something I've since fixed]

Gets Around
1:09:26 AM PDT - Thu, Jun 27th 2013
Gary,
I've had a quick look at things again. I think this is very do-able. However, since I'm starting teaching in one month, I'd suggest we include one series of basis sets e.g. the def2- series simple because that's what I use on a regular basis -- there may be better cases for other ones.

That buys us (me?) a bit of time to do a bigger overhaul i.e. regenerating the existing basis sets (e.g. STUTTGART_RSC97.POT doesn't have La and Lu, whereas stuttgart_rsc_1997_ecp does) and also perhaps come up with a way of validating them (unless there's already a tool for doing so).

And I would definitely need help with classifying the basis sets properly (for e.g. the .pag file and category files).

Gets Around
12:40:01 AM PDT - Fri, Jun 28th 2013
Gary,
I've started to check my basis set files vs the nwchem ones by comparing nwchem calculations where the basis sets are specified as
  • library "def2-svp"
with computations where they are given with exponents and coefficients via ECCE.
I won't be able to check everything between H and Rn, but I'm picking out a couple of elements from different locations in the periodic table and comparing the energies.

Is there any specific time-frame I should work towards? And what would be your preferred way of obtaining the files?

/Andy


Forum >> ECCE: Extensible Computational Chemistry Environment >> General ECCE Topics