Software at OSC |
COLUMBUSIntroductionCOLUMBUS is one of the most flexible ab initio electronic structure program suites. It is designed for highly accurate computations and the latest version can perform multiconfiguration self-consistent field, multireference perturbation theory, multireference coupled-cluster, and multireference configuration interaction calculations. Geometry optimizations, vibrational frequencies, one-electron properties, and solution modeling are available. It performs well on open shell and excited state systems and can model relativistic effects. The COLUMBUS Home Page has additional information. VersionVarious versions are installed at OSC. AvailabilityCOLUMBUS is available on the Itanium 2 Cluster. UsageUsers are not required to complete a license agreement. Location The root of the COLUMBUS directory tree is /usr/local/Columbus/. Execution Define the COLUMBUS environment variable:
For convenience append ${COLUMBUS} onto your PATH. The above commands can be entered into the appropriate dot files. COLUMBUS's friendly mechanism for input file generation is colinp. Documentation specific to version 5.7 is web readable. colinp may require a .iargos file; this generic example for that file should reside in ones home directory. A general purpose batch script for a production job of COLUMBUS is also available in /usr/local/Columbus/current/production.pbs. This script executes argos, cidrt, tran, ciuft, cisrt, and ciudg. Users should edit their copy of this script to modify the shell variables INPUT and RESULT. INPUT specifies the path of the required input files: argosin, cidrtky, mocoef, cisrtin, and ciudgin. RESULT specifies the subdirectory of INPUT that will contain the copied output files: argosls, cidrtls, tranls, ciuftls, cisrtls, ciudgls, civout, and civfl. Examples Examples are in the EXAMPLES subdirectories. Optimal Performance COLUMBUS jobs should be submitted as single processor or single node. Requesting the whole node may improve memory and disk performance, but will be charged for usage as if all the node processors where actually consumed. In October 2001 for version 5.7 on oscbw performance of COLUMBUS on a 2 root, 8 maximum subspace size, 15,000,000 CI size calculation was approximately 1800 cpu seconds per complete iteration; walltime varies from approximately 1900 to 3000 seconds depending on node load. Bugs In version 5.7 odd electron spin-orbit ciudg calculations print the fictitious electron in the beyond last column of the wave function analysis. DocumentationSpecific documentation is available in the docs subdirectories. Version 5.7 documentation is reproduced here. |
 |
|
|
|
|
|
|
|
|
|
 |
 |
 |
 |
 |
 |
 |
|
||
|
|||||||||
