SPARSE

SPARSE
	Appendix A. Keyword Reference

Name

SPARSE — Perform sparse matrix calculation using the specified neglect threshold.

Synopsis

				SPARSE=TIGHT

				SPARSE=MEDIUM

				SPARSE=LOOSE

				SPARSE=n.nnn

Description

This keyword turns on a sparse matrix calculation (otherwise the equivalent full matrix calculation is performed). If a PSOLVE method is not specified, then by default the purification of the density matrix (PDM) method will be used.

SPARSE=TIGHT

Perform a sparse matrix calculation neglecting matrix elements smaller than 1 × 10^-8. The resulting energy should reproduce the equivalent full matrix result to within nano-hartree accuracy.

SPARSE=MEDIUM

Perform a sparse matrix calculation neglecting matrix elements smaller than 1 × 10^-6. The resulting energy should reproduce the equivalent full matrix result to within micro-hartree accuracy.

SPARSE=LOOSE

Perform a sparse matrix calculation neglecting matrix elements smaller than 1 × 10^-5. The resulting energy should reproduce the equivalent full matrix result to within milli-hartree accuracy.

SPARSE=`n.nnn`

Perform a sparse matrix calculation neglecting matrix elements smaller than n.nnn. A larger value will eliminate more values thus saving CPU time and lowering memory and disk space requirements but will correspondingly increase the error in the overall calculation. A value much larger than 1 × 10^-5 may not yield stable results and result in job failure.

Abbreviation:

none

Requires: