Keyword MATDIA

One specifies the matrix diagonalization technique with this keyword.
Options:
DSYEV / RS / D&C / JACOBI

	DSYEV		LAPACK Householder diagonalization. This is the default in serial runs except for atomic calculations.
	RS		EISPACK Householder diagonalization. This is the default in parallel runs except for atomic calculations.
	D&C		LAPACK divide and conquer diagonalization.
	JACOBI		Jacobi diagonalization. This is the default for atomic calculations.

Description:
The fastest diagonalizer available in deMon2k is the divide and conquer diagonalization from LAPACK [283]. However, it may fail in special situations. Fail-safe diagonalizers in deMon2k are based on the Householder algorithm. The DSYEV diagonalizer is considerably faster than the RS diagonalizer. However, DSYEV is also more memory intensive. Therefore, it should be used only if enough RAM space is available. The JACOBI diagonalizer is much slower but produces very pure eigenvectors. It is used by default for atomic calculations.