Notes on Options

Note

The options referred to in the Theoretical Methods: SCF to FCI section below and indexed in Keywords by Module are placed in set blocks as described in Job Control Keywords, not as arguments to a Python function (like energy()).

Note

All PSI4 keyword names and values are insensitive to case, both those that are placed in set blocks and as Python function arguments. The few exceptions are documented for the database() function, where case structure must match the database file.

Note

Boolean options can be specified by yes, on, true, or 1 for affirmative and no, off, false, or 0 for negative, all insensitive to case.

Note

Certain convergence and tolerance keywords, of type double (real numbers), may be specified using either a real number or an integer; and integer X is then treated as the number of converged decimal digits required. For example, to request as energy converged to \(10^{-6} E_h\), the user may set the e_convergence keyword to 0.000001, 1.0e-6, or 6.

Notes on Psivars

Note

Starting in 1.6, there are three standard ways to access an excited state property. We give examples below, but the method name and property name may change. * method ROOT 0 -> ROOT m property to get root m.

• method ROOT 0 -> ROOT m property - h TRANSITION to get root m and

  independently specify that the total transition symmetry is h, e.g., A2.

• method ROOT 0 (h) -> ROOT m (i) property to get the transition between two roots, specifying the symmetry of both states and the index of the target roots among states of their own symmetry.

For example, to target the second excited-state, which is also the lowest energy state of its irrep, the first two calls will take m = 2, while the last takes m = 0. Methods that use this interface are: TD-fctl. Note that numberings are associated with the calculation much more strongly than with the molecular system. Changing the number of roots sought, the symmetry subspace or the symmetry apportionment of roots under which the computation is run, or the excited state method are all likely to scramble root numberings.

Alternate Implementations

Depending on the reference (RHF, UHF, ROHF) and the integral treatment (conventional CONV, density-fitted DF, and Cholesky-decomposed CD), computational methods are sometimes implemented by multiple coders or even multiple times. PSI4 transparently selects the most efficient implementation, so one generally needn’t consult this table. However, to understand the details of what combinations are accessible or what alternate implementations are available, read on.

Below, “Y” means method available in module, “D” means module is default for that method, and “” means method not available. HF, DFT, and MP2 default to density-fitted integrals, while all higher methods default to conventional integrals. Therefore, for a closed-shell molecule:

• runs MP2 with default DF with default implementation DFMP2

  energy('mp2')
  

• runs MP2 with CONV with default implementation OCC

  set mp2_type conv
  energy('mp2')
  

• runs MP2 with default DF with implementation OCC

  set qc_module occ
  energy('mp2')
  

Overlapping capabilities of PSI4. “Y” is available; “D” is default.

name _ type select 1	QC_MODULE REFERENCE _	energy()									gradient()
		RHF			UHF			ROHF			RHF			UHF			ROHF
		CV	DF	CD	CV	DF	CD	CV	DF	CD	CV	DF	CD	CV	DF	CD	CV	DF	CD
mp2 MP2_TYPE	CCENERGY
	DETCI	Y						Y
	DFMP2		D 2			D 2			D			D
	FNOCC	Y
	OCC	D	Y	D	D	Y	Y	D	Y	D	D	Y		D	D
mp3 MP_TYPE	CCENERGY
	DETCI	Y						Y 3
	DFMP2
	FNOCC	Y
	OCC	D	D	D	D	D	D				D	D		D	D
mp2.5 MP_TYPE	CCENERGY
	DETCI
	DFMP2
	FNOCC
	OCC	D	D	D	D	D	D				D	D		D	D
mp4 MP_TYPE	CCENERGY
	DETCI	Y						Y 3
	DFMP2
	FNOCC	D
	OCC
lccd CC_TYPE	CCENERGY
	DETCI
	DFMP2
	FNOCC	D
	OCC	Y	D	D	D	D	D				D	D		D	D
lccsd, cepa(0) CC_TYPE	CCENERGY
	DETCI
	DFMP2
	FNOCC	D
	OCC
ccsd CC_TYPE	CCENERGY	D			D			D			D			D			D
	DETCI
	DFMP2
	FNOCC	Y	D	D
	OCC		Y	Y								D
ccsd(t) CC_TYPE	CCENERGY	D			D			D			D			D
	DETCI
	DFMP2
	FNOCC	Y	D	D
	OCC		Y	Y
a-ccsd(t) 4 CC_TYPE	CCENERGY	D
	DETCI
	DFMP2
	FNOCC
	OCC		D	D
cisd CI_TYPE	CCENERGY
	DETCI	Y						D
	DFMP2
	FNOCC	D
	OCC
adc(2) MP_TYPE	ADCC	Y			D
	BUILTIN	D

Footnotes

1

Algorithm type selection keyword below. Values to the right: conventional CONV (here abbreviated CV), density-fitted DF, and Cholesky-decomposed CD.

2(1,2)

Also available for KS reference.

3(1,2)

Conditions have no default module (computationally inefficient) and can only be accessed by specifying QC_MODULE.

4

Also known as CCSD(AT), Lambda-CCSD(T), CCSD(T)_L