PSI4 provides a wide variety of quantum chemical methods using state-of-the-art numerical methods and algorithms. Several parts of the code feature shared-memory parallelization to run efficiently on multi-core machines (see Sec. Threading). An advanced parser written in Python allows the user input to have a very simple style for routine computations, but it can also automate very complex tasks with ease.
In this section, we provide an overview of some of the features of PSI4 along with the prerequisite steps for running calculations. Sec. Tutorial provides a brief tutorial to help new users get started. Section Psithon offers further details into the structure of PSI4 input files and how Python can be mixed with quantum chemistry directives in PSI4. Section Psithon Functions provides more detail on the Python functions provided by PSI4 and discusses some of the higher-level functions such as counterpoise correction, complete-basis-set extrapolation, and running computations on an entire database of molecules at a time. Later sections deal with the different types of computations which can be done using PSI4 (e.g., Hartree–Fock, MP2, coupled-cluster) and general procedures such as geometry optimization and vibrational frequency analysis. The Appendices include a complete description of all possible input keywords for each module, as well as tables of available basis sets and a listing of the sample input files available under psi4/samples. The user is urged to examine this directory of sample inputs, as most common types of computations are represented there. For the latest PSI4 documentation, check www.psicode.org.
The following citation should be used in any publication utilizing the PSI4 program package:
Depending on the particular modules used, the user may also wish to cite some of the following references for theoretical, algorithmic, or implementation contributions specific to PSI4 (in addition to appropriate references for the underlying theory, which are not necessarily included in the list below).
PSI has a highly optimized code for full configuration interaction and highly correlated configuration interaction, as described in
A general discussion of coupled cluster theory is given in
Implementation of frozen natural orbital (FNO) coupled cluster theory in PSI and its performance for non-covalent interactions is discussed in
Implementation of density-fitted (DF) and Cholesky decomposition (CD) coupled cluster in PSI, and its performance for non-covalent interactions and reaction energies, is discussed in
PSI4 features production-level Mukherjee-style state-specific coupled-cluster theory, including perturbative triples and also associated multi-reference perturbation theories. The theory and PSI4 implementation of these methods is discussed in the following papers.
General Mk-MRCC
Mk-MRCCSD(T)
Mk-MRCCSDT(-n)
Mk-MRPT2
PSI4 features an extremely efficient code to perform wavefunction-based Symmetry Adapted Perturbation Theory (SAPT). A good review article for this method is as follows:
PSI4 benefits enormously from the introduction of density fitting (DF) into SAPT. The theory and implementation of DF-SAPT is discussed in the following papers for various levels of SAPT.
DF-SAPT0
DF-SAPT2, DF-SAPT2+, DF-SAPT2+(3), DF-SAPT2+3
Using Natural Orbitals in SAPT
Orbital-optimized second-order perturbation theory (OMP2)
Orbital-optimized third-order perturbation theory (OMP3)
Orbital-optimized coupled electron pair approximation (OCEPA)
Orbital-optimized MP2.5 (OMP2.5)
General ADC(2) theory
Theory of “Partially-renormalized” CIS(D) and ADC(2) [PR-CIS(D) and PR-ADC(2)] and their implementation in PSI4
The majority of PSI4 was developed on Mac and Linux machines. In principle, it should work on any Unix system; however, we have not tested extensively on systems other than Mac and Linux. There is not a Windows version of PSI4.
PSI4 has been successfully compiled using Intel, GCC, and Clang compilers. For the Intel compilers, use versions 11 or 12.1 (we have had trouble with version 12.0). See Sec. Compiling and Installing for details.
PSI4 can perform ab initio computations employing basis sets of contrated Gaussian-type functions of virtually arbitrary orbital quantum number. Many parts of PSI4 can recognize and exploit the largest Abelian subgroup of the molecular point group. Table Methods displays the range of theoretical methods available in PSI4. For more details, see Tables Energy, Energy (DFT), Energy (MRCC), Gradient, and Frequency.
Method | Energy | Gradient | Reference | Parallelism |
---|---|---|---|---|
SCF (HF and DFT) | Y | Y [4] | RHF/ROHF/UHF/RKS/UKS | threaded |
DF-SCF (HF and DFT) | Y | Y [4] | RHF/ROHF/UHF/RKS/UKS | threaded |
DCFT | Y | Y | UHF | partially threaded |
MP2 | Y | Y | RHF/ROHF/UHF | threaded [3] |
DF-MP2 | Y | Y [2] | RHF/ROHF/UHF | threaded |
MP3 | Y | Y | RHF/UHF | threaded [3] |
MP2.5 | Y | Y | RHF/UHF | threaded [3] |
MP4 | Y | — | RHF | threaded [3] |
MP(n) | Y | — | RHF/ROHF | partially threaded |
ZAPT(n) | Y | — | RHF/ROHF | partially threaded |
OMP2 | Y | Y | RHF/ROHF/UHF/RKS/UKS | partially threaded |
OMP3 | Y | Y | RHF/ROHF/UHF/RKS/UKS | partially threaded |
OMP2.5 | Y | Y | RHF/ROHF/UHF/RKS/UKS | partially threaded |
OCEPA | Y | Y | RHF/ROHF/UHF/RKS/UKS | partially threaded |
CEPA(0) | Y | Y | RHF/UHF | threaded [3] |
CEPA(n), n=0,1,3 | Y | — | RHF | threaded [3] |
ACPF/AQCC | Y | — | RHF | threaded [3] |
QCISD | Y | — | RHF | threaded [3] |
QCISD(T) | Y | — | RHF | threaded [3] |
CC2 | Y | — | RHF/ROHF/UHF | threaded [3] |
CCSD | Y | Y | RHF/ROHF/UHF | threaded [3] |
DF-CCSD | Y | — | RHF | threaded [3] |
CCSD(T) | Y | Y [1] | RHF/ROHF/UHF | threaded (pthreads) |
DF-CCSD(T) | Y | — | RHF | threaded [3] |
CC3 | Y | — | RHF/ROHF/UHF | threaded (pthreads) |
Mk-MRPT2 | Y | — | RHF/ROHF/TCSCF | threaded [3] |
Mk-MRCCSD | Y | — | RHF/ROHF/TCSCF | threaded [3] |
Mk-MRCCSD(T) | Y | — | RHF/ROHF/TCSCF | threaded [3] |
CI(n) | Y | — | RHF/ROHF | partially threaded |
RAS-CI | Y | — | RHF/ROHF | partially threaded |
SAPT | Y | — | RHF | threaded |
CIS/RPA/TDHF | Y | — | ||
ADC(2) | Y | — | RHF | threaded [3] |
EOM-CCSD | Y | Y | RHF/ROHF/UHF | threaded [3] |
Geometry optimization can be performed using either analytic gradients or energy points. Likewise, vibrational frequencies can be computed by analytic second derivatives, by finite differences of analytic gradients, or by finite differences of energies. PSI4 can also compute an extensive list of one-electron properties.
The PSI4 package is distributed for free and without any guarantee of reliability, accuracy, or suitability for any particular purpose. No obligation to provide technical support is expressed or implied. As time allows, the developers will attempt to answer inquiries directed to crawdad@vt.edu or sherrill@gatech.edu. For bug reports, specific and detailed information, with example inputs, would be appreciated. Questions or comments regarding this user’s manual may be sent to sherrill@gatech.edu.
Alternatively, bug reports and comments can be submitted to the Issue tracker on GitHub . This site is viewable by all, but reporting bugs requires signing up for a free GitHub account.
Footnotes
[1] | UHF-CCSD(T) gradients only, as of beta5 |
[2] | RHF reference only. DF-MP2 is recommended as a faster alternative. |
[3] | (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18) threading through BLAS routines only |
[4] | (1, 2) DFT gradients only implemented for SCF type DF. LRC-DFT gradients not implemented yet. |