*CI INPUT  

Purpose:

Options for a CI calculation.

.CIDENSITY  

Calculate CI one-electron density matrix and natural orbital  occupations after convergence.

.CINO  

Generate CI   natural orbitals  for CI root number ISTACI, clear the "f21" file and write the orbitals with label "NEWORB ". The ".STATE  " option must be specified.

.CIROOTS  

Default: NROOCI = 1
READ (LUINP,*) NROOCI
Converge the lowest NROOCI CI roots  to threshold.

.DISKH2  

Active two-electron MO integrals on disk (see comments below).

.MAX ITERATIONS  

READ (LUINP,*) MXCIMA
Max iterations in iterative diagonalization of CI matrix (default = 20).

.STATE  

Default: not specified
READ (LUINP,*) ISTACI
Alternative to ".CIROOTS  ". Converge root number ISTACI to threshold, converge all lower roots only to THQMIN (from the "*OPTIMIZATION  " input group, see p. ).

.THRESHOLD  

Default = 1.D-05
READ (LUINP,*) THRCI
Threshold for CI energy gradient. The CI energy will be converged to approximately the square of this number.

.THRPWF  

Default = 0.05D0
READ (LUINP,*) THRPWF
Only CI coefficients greater than threshold are printed (PWF: print wave function).

.WEIGHTED RESIDUALS  

Use energy weighted residuals  (see comments below).

.ZEROELEMENTS  

Zero small elements in CI trial vectors (see comments below).

Comments:

DISKH2: By default the CI module will attempt to place the two-electron integrals with four active indices in memory for more efficient calculation of CI sigma vectors, if memory is insufficient for this the integrals will automatically be placed on disk. The DISKH2 keyword forces the integrals always to be on disk.

WEIGHTED RESIDUALS: Normally the CI states will be converged to a residual less than the specified threshold, and this will give approximately the squared number of decimal places in the energy. Depending on the value of the energy, the eigenvectors will be converged to different accuracy. If the eigenvectors are wanted with, e.g., at least 6 decimal places for property calculations, specify a threshold of 1.0D-6 and weighted residuals.

ZEROELEMENTS: an experimental option to perhaps save time (if the CI module can use sparseness) by zeroing all elements less than 1.0D-3 times the largest element in the CI trial vector before orthonormalization against previous trial vectors.

See also ".SYM CHECK" under "*OPTIMIZATION" (p. ).