Alex Granovsky
gran@classic.chem.msu.su
the ISPIN filter only works if the number of alpha electrons is equal
to the number of beta electrons and so unfortunately it cannot be used
for doublets.
Try increasing nhgss, nstgss, and mxxpan in $det/$cidet groups.
NHGSS = dimension of the Hamiltonian submatrix which is diagonalized to obtain the initial guess eigenvectors. The determinants forming the submatrix are chosen on the basis of a low diagonal energy, or if needed to complete a spin eigenfunction. The default is 300. NSTGSS = Number of eigenvectors from the initial guess Hamiltonian to be included in the Davidson's iterative scheme. It is seldom necessary to include extra states to obtain convergence to the desired states. The default is the value for NSTATE. MXXPAN = Maximum number of expansion basis vectors in the iterative subspace during the Davidson iterations before the expansion basis is truncated. The default is the larger of 10 or NSTGSS+1. If this value is decreased, convergence is hurt but the memory required is decreased.
Hope this helps.
Kind regards,
Alex Granovsky
On Wed May 15 '13 5:05pm, lello wrote
-------------------------------------
>Hi all,
>I have read on the forum that it is possible to improve the convergence of SA-MCSCF using ISPIN keyword to generate a better selection of initial states (for ALDET method).
>ISPIN=0 filters all odd spin numbers and ISPIN=1
>filters all even spin numbers.
>What about doublet states? Is is possible to improve the initial
>guess for doublets?
>lello