35.11 DF-LMP2-F12 calculations with local approximations
Local variants of the DF-MP2-F12 and CCSD(T)-F12 methods are invoked by the
commands DF-LMP2-F12, DF-LCCSD-F12, DF-LCCSD(T)-F12 with ansatz 3*A(LOC) [for DF-LCCSD-F12 fixed amplitudes are used, i.e.,
the default is 3*A(LOC,FIX)]. These calculations are performed
with a different program than non-local calculations.
The (LOC) option implies that the LMP2 calculation with domain approximations is performed, and by default a local projector
as first described in
H.-J. Werner, J. Chem. Phys. 129, 101103 (2008) is used [see also
T. B. Adler, F. R. Manby, and H.-J. Werner, J. Chem. Phys. 130, 054106 (2009);
T. B. Adler and H.-J. Werner, J. Chem. Phys. 130, 241101 (2009);
T. B. Adler and H.-H. Werner, J. Chem. Phys. 135, 144117 (2011)]
This yields very similar restults as the corresponding canonical methods at much lower cost
3and the method can be applied to quite large molecules.
Special options for these local variants are (local RI works only with ansatz 3*A):
- Specifies which pairs to be treated by R12 or F12
(STRONG|CLOSE|WEAK|ALL; pairs up to the given level are included). The default is ALL. Note that even with ALL very distant pairs are neglected if these are neglected in the LMP2 as well.
- If zero, the
form of the projector is used, and local RI approximations apply to and .
If set to 1, the
form of the projector is used; any local RI approximation then applies only to the RI contribution .
- If USEPAO=1 use pair-specific local projectors instead of . This is the default if either the ansatz contains
'(LOC)' or if domain approximations are made in the LMP2 (i.e., DOMSEL<1 is explicitly specified). Otherwise the default is USEPAO=0.
USEPAO=1 automatically implies USEVRT=1, i.e. local RI approximations only affect the RI contributions .
Furthermore, if USEPAO=1 is specified and DOMSEL is not given, default domains are assumed in the LMP2. If
USEPAO=0, full domains (DOMSEL=1) will be used.
- if USEVRT=0 and FULLAO=1, local RI approximations only apply to the RI contributions .
This should give the same results as USEVRT=1 (only applies if USEPAO=0).
- Parameter for debug print
- If set to one, use local fitting. Default is no local fitting (LOCFIT_F12=0)
- Alias for LOCFIT_F12. Local fitting is not recommended in R12 calculations.
- Determine how the base fitting domains are determined (only applies if LOCFIT_F12=1):
0: Fitdomains based on united operator domains;
1: Fitdomains based in orbital domains (default);
2: Fitdomains based on united pair domains using strong pairs;
3: Fitdomains based on united pair domains using strong, close and weak pairs. Note: This is the only option implemented
in the DF-LMP2 program. Therefore, the DF-LMP2 and DF-LMP2-F12 programs might give slightly different results if default
values are used.
- Distance criterion for density fitting domain extensions in case of local fitting. The default depends on FITDOM.
- Connectivity criterion for density fitting domain extensions in case of local fitting.
- Distance criterion for RI domain extensions. Zero means full RI basis (default).
If USEPAO=1 or USEVRT=1 or FULLAO=1 a value of 5 bohr is recommended. In other cases
the local RI domains must be very large (RAODOM>12) and the use of local RI approximations is not recommended.
- Connectivity criterion for RI domain extensions. Zero means full RI basis (default).
Values greater or equal to 2 should lead to sufficiently accurate results, provided the local projector (USEPAO=1) is used.
- Screening threshold for coulomb integrals in the AO or RI basis.
- Screening threshold for F12 integrals.
- Screening threshold for half transformed integrals.
- Product screening threshold in the first half transformation.
- If set to 1, only the F12 calculation is performed, and the LMP2 is skipped. This is sometimes
useful if full domains are used, since the iterative LMP2 then causes a big overhead and needs a lot of memory.
It is then more efficient to do the a DF-MP2 calculation separately and compute the total energy as the sum of the DF-MP2 energy and the F12 energy
- Values greater than 0 invoke the local projector. PROJF=1 project vv parts only (this is formally the most accurate case but
only possible with the canonical program [ANSATZ=3*A(FIX,NOX)], PROJF=2 project vv and vo parts (default). This is unavoidable
if the local program is used [ANSATZ=3*A(LOC)].
- If , core contributions are neglected in the projector. This is a necessary approximation in order to
compute the LCCSD-F12 coupling terms efficiently (implementation not yet finished). Setting MODOMC=0 avoids the approximation.
Note that the results in J. Chem. Phys. 135, 144117 (2011) have been obtained using MODOMC=0.
Further options for density fitting are described in section 15, and further options
to choose the ansatz in section 35.7.
Typical inputs for calculations with local approximations are:
!parameters for local density fitting:
!LMP2-F12(loc) with local RI:
This would perform a local MP2 with a Boughton-Pulay domain completeness criterion of 0.985.
In the F12 part, distant pairs are not included (PAIRS=WEAK) and the local projector is used (USEPAO=1, default).
Local density fitting and local RI approximations are used.
A corresponding non-local calculation (still using localized orbitals and the diagonal ansatz) would be
ecorr_MP2=energy-energr !mp2 correlation energy
ecorr_MP2_F12=ecorr_MP2+ecorr_F12 !total correlation energy
Note: The use of local DF and RI domains is still experimental and should be use with care!