35.7 Choosing the ansatz and the level of approximation

The Ansatz can be chosen using the ANSATZ option and/or by options on the command line.

3A
Ansatz 3A;
3*A
Ansatz 3A with EBC approximation
3B
Ansatz 3B
3*B
Ansatz 3B with EBC approximation
3C
Ansatz 3C
3*C
Ansatz 3C with EBC approximation

The ansatz can be further detailed by appending options in parenthesis, e.g.

ANSATZ=3B(D)

These options can be one of

D
Use diagonal ansatz
FIX
Use diagonal ansatz with fixed amplitudes (orbital invariant)
FIXC
Use diagonal ansatz with fixed amplitudes and canonical orbitals
DX
Use diagonal ansatz and assume X-matrix to be diagonal (only for Ansatz 3A)
GBC
Use GBC approximation (only for 3B, default in 3A)
EBC
Use EBC approximation (same as *)
HX1
Use HX1 approximation (only for 3B).
HY1
Use HY1 approximation (only for 3B and 3C).
HY2
Use HY2 approximation (only for 3B and 3C).
HY
Default hybrid approximation. HX1 and HY2 approximation in 3B, HY2 in 3C.
NOZ
Neglect Z terms (only for 3B and 3C).
NOX
Neglect X terms (only for 3A and 3B).

Several options separated by commas can be given. For instance

ANSATZ=3C(FIX,HY1)

uses the diagonal ansatz 3C(D) with fixed coefficients and the hybrid (HY1) approximation.

Alternatively or in addition, the following options can be given on the command line:

DIAG=1
Use diagonal ansatz.
DIAGX=1
Use diagonal ansatz and assume X-matrix to be diagonal.
GBC=1
Use GBC approximation (only for 3B, default in 3A).
EBC=1
Use EBC approximation (same as *).
HYPRID=n
Use HYn approximation.
HYPRIDX=1
Use HX1 approximation.
NOZ=1
Neglect Z terms (only 3B and 3C).
NOX=1
Neglect X terms (only 3A and 3B).
FIX=1
Use diagonal ansatz with fixed coefficient approximation (orbital invariant).
FIX=2
Use diagonal ansatz with fixed coefficient approximation. Evaluate only first order energy expression, not the Hylleraas functional. Very fast but less accurate and reliable!
FIXCAN=1
Use diagonal ansatz with fixed coefficient approximation and canonical orbitals. A non-iterative method is used to evaluate the energy. This is equivalent to FIX=1,CANONICAL=1 and is most efficient.
FIXCAN=-1
As FIXCAN=1, but equations are solved iteratively (test purpose only).
CABS=1
Use CABS (default). If CABS=0 is given, CABS is disabled. However, if RI_BASIS=OPTRI, the orbital and OPTRI basis sets are automatically merged, and then exactly the same results as with CABS=1 are obtained.
ORTHO_CABS=1
Construct CABS basis from orthogonal MOs and ABS basis rather than AO and RI basis.
THRABS=thrabs
Threshold for smallest eigenvalue of S in auxiliary ABS (only used with ORTHO_CABS=1; default=THRCABS).
THRCABS=thrcabs
Threshold for smallest eigenvalue of S in CABS (default 1.d-8).
THRCABS_REL=thrcabs_rel
Relative CABS threshold (default 1.d-9). The actual threshold is max(thrcabs,eigmax*thrcabs_rel, where eigmax is the largest eigenvalue of the overlap matrix.
PRINT=level
Print parameter. PRINT=1 give information about all computed integrals and the iterations.
DEBUG=level
Can be used to obtain extended debug print.
SOLVE=0
Use a most efficient pair-specific fully iterative method (default).
SOLVE=1
Use simple fully iterative method.
SOLVE=2
Use pair specific iterative method (more expensive).
SOLVE=3
Use pair specific non iterative method (most expensive, only with canonical orbitals).
CANONICAL=1
Use canonical orbitals and full domains.
DOMSEL=1
Use full domains and localized orbitals (unless CANONICAL=1 is given).
SCALE_TRIP=1
Scale triples energy as explained in section 35.10.
SINGLES
If set to one, include CABS singles correction (default=1)
CORE_SINGLES
If set to one, include CABS singles correction for core orbitals (default=0)
EXTGEN
For open-shell systems: If 1 (default,recommended), include all occupied valence orbital pairs for $mn$ in $T^{ij}_{mn}$, independent of spin (as described in J. Chem. Phys. 130, 054104 (2009), section II.E). If 0, use only pairs mn where the spins of $i$ and $m$, and $j$ and $n$ are equal.

For instance

ANSATZ=3C,fix=1,hybrid=1,canonical=1

implies a canonical 3C calculation with diagonal ansatz 3C, using fixed coefficient and hynrid approximations. The combination of the options fix=1 and canonical=1 implies a non-iterative calculation of the energy and is recommended. The above is equivalent to all of the following:

ANSATZ=3C(FIXC,HY1)
ANSATZ=3C(D,FIXC,HY1)
ANSATZ=3C(D,HY1,FIX),canonical=1

Note that the HF convergence threshold should be rather strict to obtain accurate results (use ACCU,14 in the HF).

Numerous further options are for specialist use only and not described here. See explicit.registry for a full list.

molpro@molpro.net 2018-10-21