Normand Mousseau
Titulaire de la Chaire de recherche du Canada
en physique numérique de matériaux complexes


   
Université de Montréal Département de physique RQMP groFNUM Séminaires Visiteurs


Livres

Maîtriser notre avenir énergétique

Entrevues et Médias

Le défi des ressources minières

Entrevues et Médias Présentations et conférences

La révolution des gaz de schiste

Entrevues et Médias Présentations et Conférences

L'avenir du Québec passe par l'indépendance énergétique

Entrevues et Médias

Au bout du pétrole

Entrevues et Médias

Où va notre argent?

Ne manquez pas de lire Où va notre argent? Une fiscalité pour les riches, du collectif Attac-Québec dont je fais partie.


Recherche

Logiciels

Version for empirical potentials

We have two versions of the activation-relaxation technique (ART nouveau) code developed by Barkema and Mousseau (PRL 1996) and modified significantly by Malek and Mousseau (PRB 2000). A version in C and one in Fortran 90. Both programs implement ART nouveau as described by Malek and Mousseau, PRB 2000.

You can download either or both of them below. The "manual" is written for the C version but can easily be applied to the F90 version. If you need any more information, do no hesistate to contact me.

  • Very succinct user manuel for ART. The manual was written for the C version
  • ART version C. This is closer to the original version so the code is a bit less clear.

  • ART nouveau, version 3.0 This code has been revised extensively. It incorporates the advances presented in Machado et al PRE (2012) paper as well as suggestions by Cances et al. JCP (2009), making it significantly faster than previous versions.

  • ART nouveau for DFT, version 3.0 This is an implementation of ART nouveau that can run both standard empirical potentials and DFT forces. An example is provided with SIESTA. (July 2014).

While this version implements an empirical potential, it is conceived to interface rapidly with ab initio codes as it has the option of saving configurations at regular intervals. (Eduardo Machado-Charry, CEA-Grenoble; Laurent Karim Béland and Normand Mousseau, Université de Montréal; 3 April 2012.)

Previous versions

  • ART nouveau, "Version 2.3" A double counting in the local_move was removed. (Corrected by Eduardo Machado-Charry, CEA Grenoble, and Normand Mousseau, U. de Montréal; 16 June 2010).
  • ART nouveau, "Version 2.2" A bug in DIIS was corrected. In the previous version, the first direction selected was not incorporated properly in the conversion scheme. (Corrected by Eduardo Machado-Charry, CEA Grenoble, and Peter Brommer, U. de Montréal; 23 April 2010).
  • ART nouveau, "Version 2.1" I have corrected the routine saddle_converged.f90 to allow ART to be run without DIIS (21 December 2009).
  • ART nouveau, "Version 2.0" This version now includes a faster method for converging to the saddle when the system is already near this, the "Direct inversion in iterative subspace" which was proposed by Pulay (Chem. Phys. Lett. 73, 393 (1980)). Thanks to Damien Caliste and Pascal Pochet for pointing this method to me! (17 March 2008).
  • ART nouveau, "Version 1.2". It seems that I used the wrong art01.f90 when I prepared the previous package. Here it is corrected (by M.-A. Malouin (Montréal). I have also added the possibility to reject all events by using a negative temperature (27 October 2007).
  • ART nouveau, version 1.100. Corrections proposed by Cosmin Marinica (CEA, Saclay) and implemented by M.-A. Malouin (Montréal). These corrections are mostly in the activation routine, making it more efficient. (26 June 2007)
  • ART nouveau, version 1.003. Corrects a bug in the push from the saddle point and to allow the acceptation of new events. (Updated on February 19, 2007)
  • ART nouveau, version 2006. Written expressely to be distributed, this version is recommended.
    Version 1.002 Corrects a bug in reading the parameters (01 December 2006)
    Version 1.001 Corrects a bug in defining the number of events (11 April 2006)
    New version updated on 17 February 2006! (Université de Montréal)

Version for SIESTA

It is also possible to use ART with ab-initio programs such as SIESTA. You can find an example of such an application in El-Mellouhi, Mousseau and Ordejon (PRB 2004).

The version of ART nouveau given in the program below saves the configuration at every single step in the activation process since it is costly to start over again. Because of the significant overhead associated with this constant writing to the disk, it is not recommended for empirical potential (use the version above).

Note: ART has not yet been interfaced with the server of SIESTA 2.X but should work with it. SIESTA is not provided here.

History

  • 16 June 2010 (version 2.3) A bug in DIIS was corrected. In the previous version, the first direction selected was not incorporated properly in the conversion scheme. (Corrected by Eduardo Machado-Charry, CEA Grenoble, and Normand Mousseau, U. de Montréal; 16 June 2010). Download version 2.3
  • 23 April 2010 (version 2.2) A bug in DIIS was corrected. In the previous version, the first direction selected was not incorporated properly in the conversion scheme. (Corrected by Eduardo Machado-Charry, CEA Grenoble, and Peter Brommer, U. de Montréal; 23 April 2010). Download version 2.2
  • "21 December 2009" (version 2.1) Corrected the call to DIIS to allow ART nouveau to work even without the final converge to the saddle point with DIIS. Also added the proper environment variables for DIIS in the example file siestart.sh. Download version 2.1
  • "15 November 2008" (version 2.0) Includes a relaxation with DIIS as well as an example to try with SIESTA. Download version 2.0
  • 5 March 2008 (version 1.5) Includes the possibility of converging towards the saddle point and relax to a new minimum. Also it is now possible to deform the perfect crystal lattice to ease activation. (Corrections made by Kevin Levasseur-Smith). Download version 1.5
  • 26 June 2007 (version 1.100) Corrects the same bugs as above (art2006_1.100).
    Download version 1.100
  • 22 June 2006 (version 1.010) Corrects bugs with the output of events.list.
  • 17 June 2006 (version 1.009) Allow negative temperature to avoir moving to new conformation; corrects bugs with energy jumps and the output of events.list. Version 1.009 optimised for SIESTA (17 June 2006).
  • 16 May 2006. (version 1.007) Corrected various bugs; now moves the log.file if it exists; handles better multiple atomic types.
  • 22 March 2006. Corrected a bug with respect to the creation of the xmol-type of type. To get it right, it is now essential to at least give the type of one atom in the siestart.sh file.
  • 17 March 2006. Polished the previous version but did not correct any new bug. Introduced version numbers. This one is 1.005.
  • 8 March 2006. Corrected bugs for the calculation of the force (it did not compute the right force before). Also modified the push over the saddle point (now uses the eigenvector).
  • 17 February 2006. Added some control on the size step for Lanczos and for the parameters of SIESTA.
  • 16 February 2006. Corrected a bug in the subroutine for local moves and added a new option to select an atom for the initial move.
  • 12 February 2006. Initial version of the code
Page modifiée le 10-07-2014 à 07:06