Gaussian
03 CapabilitiesGaussian has been designed with the needs of the
user in mind. All of the standard input is freeformat and mnemonic. Reasonable
defaults for input data have been provided, and the output is intended to be selfexplanatory.
Mechanisms are available for the sophisticated user to override defaults or interface
their own code to the Gaussian system. The authors hope that their efforts
will allow users to concentrate their energies on the application of the methods
to chemical problems and to the development of new methods, rather than on the
mechanics of performing the calculations. The technical capabilities of
the Gaussian 03 system are listed in the subsections below. Fundamental
Algorithms Calculation of one and twoelectron integrals over
any general contracted gaussian functions. The basis functions can either be cartesian
gaussians or pure angular momentum functions, and a variety of basis sets are
stored in the program and can be requested by name. Integrals may be stored in
memory, stored externally, or be recomputed as needed [20,21,22,23,24,25,26,27,28].
The cost of computations can be linearized using fast multipole method (FMM) and
sparse matrix techniques for certain kinds of calculations [29,30,31,32,33,34]. Transformation
of the atomic orbital (AO) integrals to the molecular orbital basis by "incore"
means (storing the AO integrals in memory), "direct" means (no integral storage
required), "semidirect" means (using some disk storage of integrals), or "conventional"
means (with all AO integrals on disk). Use of density fitting to
speed up the Coulomb part of pure DFT calculations [35,36]. Numerical
quadrature to compute DFT XC energies and their derivatives. Energies
Molecular mechanics calculations using the AMBER [37],
DREIDING [38] and UFF [39,40]
force fields. Semiempirical calculations using the CNDO [41],
INDO [42], MINDO/3 [43,44],
MNDO [43,45,46,47,48,49,50,51,52],
AM1 [43,48,49,53,54],
and PM3 [55,56] model
Hamiltonians. Selfconsistent field calculations using closedshell
(RHF) [57], unrestricted openshell (UHF) [58],
and restricted openshell (ROHF) [59] HartreeFock
wavefunctions. Correlation energy calculations using MøllerPlesset
perturbation theory [60] carried to second, third
[61], fourth [62,63],
or fifth[64] order. MP2 calculations use direct
[21,65] and semidirect
methods [23] to use efficiently however much (or
little) memory and disk are available. Correlation energy calculations
using configuration interaction (CI), using either all double excitations (CID)
or all single and double excitations (CISD) [66]. Coupled
cluster theory with double substitutions (CCD)[67],
coupled cluster theory with both single and double substitutions (CCSD) [68,69,70,71],
Quadratic Configuration Interaction using single and double substitutions (QCISD)
[72], and Brueckner Doubles Theory (BD) [73,74].
A noniterative triples contribution may also be computed (as well as quadruples
for QCISD and BD). Density functional theory [75,76,77,78,79],
including general, userconfigurable hybrid methods of HartreeFock and DFT. See
this page for a complete list of available functionals. Automated,
high accuracy energy methods: G1 theory [80,81],
G2 theory [82], G2(MP2) [83]
theory, G3 theory [84], G3(MP2) [85],
and other variants [86]; Complete Basis Set (CBS)
[87,88,89,90,91]
methods: CBS4 [91,92],
CBSq [91], CBSQ [91],
CBSQ//B3 [92,93],
and CBSQCI/APNO [90], as well as general CBS extrapolation;
the W1 method of Martin (with slight modifications) [94,95,96]. General
MCSCF, including complete active space SCF (CASSCF) [97,98,99,100],
and allowing for the optional inclusion of MP2 correlation [101].
Algorithmic improvements [102] allow up to 14 active
orbitals in Gaussian 03. The RASSCF variation is also supported
[103,104]. The
Generalized Valence BondPerfect Pairing (GVBPP) SCF method [105]. Testing
the SCF wavefunctions for stability under release of constraints, for both HartreeFock
and DFT methods [106,107]. Excited
state energies using the singleexcitation Configuration Interaction (CISingles)
method [108], the timedependent method for HF
and DFT [109,110,111],
the ZINDO semiempirical method [112,113,114,115,116,117,118,119,120],
and the Symmetry Adapted Cluster/Configuration Interaction (SACCI) method of
Nakatsuji and coworkers [121,122,123,124,125,126,127,128,129,130,131,132,133,134,135]. Gradients
and Geometry Optimizations Analytic computation of the nuclear
coordinate gradient of the RHF [136], UHF, ROHF,
GVBPP, CASSCF [137,138],
MP2 [22,23,139,140],
MP3, MP4(SDQ) [141,142],
CID [143], CISD, CCD, CCSD, QCISD, Density Functional,
and excited state CIS energies [108]. All of the
postSCF methods can take advantage of the frozencore approximation. Automated
geometry optimization to either minima or saddle points [136,144,145,146,147,148],
using internal or cartesian coordinates or a mixture of coordinates. Optimizations
are performed by default using redundant internal coordinates [149],
regardless of the input coordinate system used. Automated transition
state searching using synchronous transitguided quasiNewton methods [150]. Reaction
path following using the intrinsic reaction coordinate (IRC) [151,152]. Two
or threelayer ONIOM [153,154,155,156,157,158,159,160,161,162,163]
calculations for energies and geometry optimizations. Simultaneous
optimization of a transition state and a reaction path [164]. Conical
intersection optimization using stateaveraged CASSCF [165,166,167]. IRCMax
calculation which locates the point of maximum energy for a transition structure
along a specified reaction path [168,169,170,171,172,173,174,175,176].
Classical trajectory calculation in which the classical equations
of motion are integrated using analytical second derivatives [177,178,179,180]
using either: Born Oppenheimer molecular dynamics (BOMD) [177,178,179,180,181,182]
(see [183] for a review) [184,185,186,187,188].
This can be done using any method for which analytic gradients are available,
and can optionally make use of Hessian information. Propagation
of the electronic degrees of freedom via the Atom Centered Density Matrix Propagation
molecular dynamics model [188,189,190].
This method has similarity and differences to the related CarParrinello approach
[191]. See the discussion of the ADMP
keyword for details. This can be done using the AM1, HF, and DFT methods. Frequencies
and Second Derivatives Analytic computation of force constants
(nuclear coordinate second derivatives), polarizabilities, hyperpolarizabilities,
and dipole derivatives analytically for the RHF, UHF, DFT, RMP2, UMP2, and CASSCF
methods [25,139,192,193,194,195,196,197,198,199],
and for excited states using CIS. Numerical differentiation of
energies or gradients to produce force constants, polarizabilities, and dipole
derivatives for the MP3, MP4(SDQ), CID, CISD, CCD, and QCISD methods [143,200,201,202]. Harmonic
vibrational analysis and thermochemistry analysis using arbitrary isotopes, temperature,
and pressure. Analysis of normal modes in internal coordinates. Determination
of IR and Raman intensities for vibrational transitions [193,194,196,200,203].
Preresonance Raman intensities are also available. Harmonic vibrationrotation
coupling [204,205,206,207]. Anharmonic
vibration and vibrationrotation coupling [204,206,207,208,209,210,211,212,213,214].
Anharmonic vibrations are available for the methods for which analytic second
derivatives are available. Molecular Properties Evaluation
of various oneelectron properties using the SCF, DFT, MP2, CI, CCD and QCISD
methods, including Mulliken population analysis [215],
multipole moments, natural population analysis, electrostatic potentials, and
electrostatic potentialderived charges using the MerzKollmanSingh [216,217],
CHelp [218], or CHelpG [219]
schemes. Static and frequencydependent polarizabilities and hyperpolarizabilities
for HartreeFock and DFT methods [220,221,222,223,224,225]. 
NMR
shielding tensors and molecular susceptibilities using the SCF, DFT and MP2 methods
[226,227,228,229,230,231,232,233,234,235, 587].
Susceptibilities can now be computed using GIAOs [236,237].
Spinspin coupling constants can also be computed [238,239,240,241]
at the HartreeFock and DFT levels.

Vibrational circular dichroism
(VCD) intensities [242], and Raman optical activity [579, 583, 584].
Propagator
methods for electron affinities and ionization potentials [243,244,245,246,247,248,249]. Approximate
spin orbit coupling between two spin states can be computed during CASSCF calculations
[250,251,252,253,254].
Electronic circular dichroism [255,256,257,258,259]
(see [260] for a review). Optical
rotations and optical rotary dispersion via GIAOs [261,262,263,264,265,266,267,268,269,270,271]. Hyperfine
spectra: g tensors, nuclear electric quadrupole constants, rotational constants,
quartic centrifugal distortion terms, electronic spin rotation terms, nuclear
spin rotation terms, dipolar hyperfine terms, and Fermi contact terms [272,273,274,275,276,277,278,279].
Input can be prepared for the widely used program of H. M. Pickett [280].
Solvation
ModelsAll of these models employ a selfconsistent reaction field (SCRF)
methodology for modeling systems in solution. Onsager model (dipole
and sphere) [281,282,283,284],
including analytic first and second derivatives at the HF and DFT levels, and
singlepoint energies at the MP2, MP3, MP4(SDQ), CI, CCD, and QCISD levels. Polarized
Continuum (overlapping spheres) model (PCM) of Tomasi and coworkers [285,286,287,288,289,290,291,292,293,294,295,296,297,298,299,300,301,302,303]
for analytic HF, DFT, MP2, MP3, MP4(SDQ), QCISD, CCD, CCSD, CID, and CISD energies
and HF and DFT gradients and frequencies. Solvent effects
can be computed for excited states [298,299,300]. Many
properties can be computed in the presence of a solvent [304,305,306].
IPCM (static isodensity surface) model [307]
for energies at the HF and DFT levels. SCIPCM (selfconsistent
isodensity surface) model [307] for analytic energies
and gradients and numerical frequencies at the HF and DFT levels. 