Last update: 31 December 2020
TABLE OF CONTENTS
The following web page is intended as an introduction to Grenoble efforts within the deMon development community and, by extension, within the greater community of developers of software for theoretical and computational chemistry. The name "Casida" is well-known among quantum chemists and theoretical solid state physicists for pioneering efforts in time-dependent density-functional theory (TDDFT.) Not surprisingly most of the Grenoble contribution to deMon development is related to TDDFT.
In order to understand the notion of the deMon development community, it is first important to have some basic notions about the deMon family of programs. These points are covered in the next two sections.
The Grenoble development version of deMon is not intended to be a separate branch of deMon but rather to be a place to develop ideas that will eventually find their place in deMon2k. Our development effort focuses primarily on improving TDDFT in future releases of deMon2k. Nevertheless logially we are a little ahead of the deMon2k master version in our efforts. To understand our development version deMon@Grenoble, it is best to first know a bit about its predecessor deMon-DynaRho, which was one of the first (if not the first) implementation of TDDFT in a Quantum Chemistry program. That program was used as a separate post-deMon-KS program, making it unsuitable as a starting point for excited-state geometry optimizations. This, and the superiority of the deMon2k integral package over the deMon-KS integral package, was at the origin of the decision to abandon further development of deMon-DynaRho. Instead we chose to focus our efforts on implementing an improved version of TDDFT within the deMon2k code.
We welcome inquiries about trying out deMon2k and deMon@Grenoble. Please see below.
deMon is short for densité de Montréal ("density of Montreal.") It is a family of programs which grew out of Alain St-Amant's PhD thesis with Dennis R. Salahub at the Université de Montréal (UdM) in Montreal, Quebec, Canada. The main deMon program of that period came to be known as deMon-KS.
Beginning around 1995, the master deMon program began to be rewritten from the bottom up. The principal author of the current master version of deMon, known as deMon2k, is Andreas Köster. The official citation for that version is,
Gerald Geudtner is currently responsible for gathering updates from deMon developers and putting them in this master version. See an old version of the deMon2k manual (PDF)
However there are several associated programs and evolutionary spin-offs of deMon2k and of the earlier deMon-KS. Once such is the Stockholm-Berlin (StoBe) version of deMon-KS. Another is the deMon2K_KSCED spin-off of deMon2K developed in Geneva which allows for two types of beyond-Kohn-Sham methods:
The present page concerns a local development version of deMon2k which we call deMon@Grenoble. An important predecessor of deMon@Grenoble was deMon-DynaRho.
It is a bit hard to define what exactly makes deMon so special. Is it the people who haved worked on it over the years? The innovative things we have done with the program which have since spread to other quantum chemistry and physics programs? Suffice it to say that most deMon programs are characterized by the use of a double basis set. The molecular orbitals are expanded in an orbital basis set consisting of atomic orbitals represented as contractions of Gaussian-type orbitals,
ψi(r) = Σμ χμ(r) cμ,i .
The density,
ρ(r) = Σi ψi(r) n_i Σi ψi(r) = Σμ,ν χμ(r) (Σiocc cμ,i cν,i) χν(r) = Σμ,ν χμ(r) Pμ,ν χν(r) ,
is approximated by an expansion in an auxiliary basis set,
ρ'(r) = ΣI fI(r) aI ,
where the fitting coefficients are obtained by minimizing the electron repulsion "error" integral,
|| Δ ρ||2 = [ρ-ρ' | 1/r12 | ρ-ρ'] .
This auxiliary basis allows all 4-center electron repulsion integrals to be eliminated from deMon, thus permitting a formal O(N3) scaling, instead of the usual formal O(N3) (or worse) scaling of normal ab initio quantum chemistry programs. In practice scaling can be significantly better than O(N3) because of additional tricks of the trade. Another program with a similar philosphy was DGauss, developped at Cray Inc. Of course, there is much more to deMon programs than just auxiliary basis sets, but that and analytic derivatives for geometry optimizations was a key historical starting point for deMon.
Additional information about deMon may be found below and also at http://www.demon-software.com/public_html/index.html.
Many, but by no means all of the deMon developers, were at some time associated with the Salahub group at UdM. In fact the deMon logo,
, is intended to look a bit like the main building on that university campus,
Since then the deMon developers have all left UdM and spread around the world. deMon developers meetings are hosted annually by developers around the world. Here is a list of previous deMon developers meetings:
This helps us to coordinate deMon development but we have still found it useful for deMon developers to have their own versions designed to meet local research objectives, and also to provide a testing ground for new ideas before they go into the master version. deMon@Grenoble is just one example.
The post deMon-KS program, deMon-DynaRho is the immediate predecessor of Grenoble's contributions to deMon2k. It represents what was probably the first implementation of time-dependent density-functional theory (TDDFT) in a Quantum Chemistry program. (Notes in French and in Englsih for an introductory course on TDDFT may be found at http://dcm.ujf-grenoble.fr/PERSONNEL/CT/casida/Enseignement/M2/TDDFT/index.html.)
A detailed account of what are sometimes known as "Casida's equations" for linear response theory is,
The approach therein described was implemented in deMon-DynaRho,
We used deMon-DynaRho to make several contributions to the scientific community. These include a highly-cited article [CJCS98] on the artificially low onset of the TDDFT ionization threshold at -εHOMO as well as articles [CCS98,CS00] reporting a simple asymptotic correction scheme to fix the problem,
Article [CJCS98] is also interesting in so far as it represents the first treatment of an avoided crossing within a DFT-based method.
An attempt was made to reconcile TDDFT with the ΔSCF method which often gives similar results when both are applicable,
The spectra of open-shell molecules has been investigated,
We also used deMon-DynaRho to make some contributions to the problem of DFT-based calculations of NMR chemical shifts,
In short, we made a significant number of contributions to DFT and TDDFT using deMon-DynaRho. The citation of one of the last official versions of this program was,
Click here to see the deMon-DynaRho manual.
We still keep a copy of deMon-DynaRho, but it was never intended to be much more than a prototype "toy" program to investigate fundamental aspects of TDDFT on very simple systems. We learned a lot from this program and it became stable and reliable for calculations on small molecules, but we could not handle even moderately large systems such as p-nitroaminobenzene. And the post-SCF nature of deMon-DynaRho prevented us from implemented excited-state forces.
The Casida group as of 7 May 2009 (photo by Anushree KAMATH). From left to right: Miquel HUIX I ROTLLANT (Doctoral student), Mark E. CASIDA, Bhaarathi NATARAJAN (Doctoral student), Sébastien BRUNEAU (Masters student), Loïc JOUBERT DORIOL (Masters student). Far right: C. Muhavini WAWIRE (Doctoral student).
Around 2001, a decision was made to integrate TDDFT (and other good features) from deMon-DynaRho into the master version of deMon2k. We have had some success with this (which is why Andrei Ipatov and Mark E. Casida figure on the author list for deMon2k) in that we can now routinely handle larger molecules than deMon-DynaRho ever could. But we still lack many features from deMon-KS/deMon-DynaRho which we are gradually putting into our local development version, deMon2k@Grenoble. At the same time, we are using deMon2k@Grenoble to test out new creative ideas. Click here to see the deMon@Grenoble team.
Grenoble developments which may find their way into future versions of deMon2k include:
We are also working on
Relevant publications include:
This article reports our implementation of testing of the noncollinear spin-flip formalism
in deMon2k@Grenoble.
This article reports our implementation of testing of dressed TDDFT in deMon2k@Grenoble.
Here is a sample of local documentation for our development version:
If you are interested in obtaining a copy of deMon2k or of deMon2k@Grenoble, then please see below.
,
deMon is not an "open-source" "copy left" program, but it is essentially free to collaborators, who have signed and returned to us our license agreement. (Note however that certain versions of deMon, such as deMon-StoBe, do require the payment of a minor sum, which is, for example, used to help defray the fees of deMon developers meetings.) If you are interested in the the master version of deMon2k, go to http://www.demon-software.com/public_html/index.html.
Here in Grenoble, we also welcome serious collaborations with scientists wanting to help develop, apply, or just test deMon2k@Grenoble. (As of 6 April 2009, we have distributed one copy of the code to external users.) The recommended citation is,
Please contact me at Mark.Casida@univ-grenoble-alpes.fr for further information. If all goes well, you will be asked to sign our license agreement with the mention "Calculations with the deMon2k@Grenoble development version" under the heading "Research Purpose." For a few priveledged few, you may be even allowed into the inner circle of deMon2k@Grenoble developers -- those who have access to our private pages by clicking on this picture of the last vestige of Grenoble's legendary city walls (La porte de France).,
Si vous avez rencontré des difficultés avec mes pages www, veuillez me contacter à Mark.Casida@ujf-grenoble.fr.
Should you encounter difficulties with my web pages, please contact me at: Mark.Casida@ujf-grenoble.fr.