This section briefly summarizes the major differences between Gaussian 09 and Gaussian 03. See earlier sections of the manual for full details on these features (including references). A retrospective summary of the features of Gaussian versions is available on our website at www.gaussian.com/g_tech/gdiffs.pdf.
A new implementation of recent semi-empirical models, including AM1, PM3, PM3MM, PDDG and PM6, which includes analytic first and second derivatives, user-specifiable parameters, and works with the PCM solvation models.
TD-DFT gradients and numerical frequencies.
Electronic excitation energies using EOM-CCSD.
Many new DFT functionals, including the HSE, wB97, m05/m06, LC- families and the double-hybrid B2PLYP.
Empirical dispersion models are included with the corresponding functionals.
ROMP3, ROMP4, ROCCSD, and ROCCSD(T) energies.
The W1RO, W1BD, and G4 compound methods for energies.
The DFTB semi-empirical model and the DFTBA version using analytic matrix elements.
ONIOM and PCM can be combined. There are several ONIOM+PCM models.
IRCs can now be done with ONIOM, and are efficient even for molecular systems containing thousands of atoms.
A new algorithm for PCM solvation, which makes the energy a properly continuous function of the nuclear coordinates and which includes efficient solvation of all SCF properties. Geometry optimizations with PCM now converge at the same rate as the corresponding gas-phase optimizations.
State-specific self-consistent solvation can be done, to model fluorescence and other emission processes.
The SMD solvation model is available. It is parametrized to give good total solvation free energies for over a hundred solvents.
The GEDIIS geometry optimization algorithm, which is the default for minimizations. This is particularly helpful for large, floppy molecules.
Quadratic convergence ONIOM(MO:MM) optimizations, for either mechanical or electronic embedding, and for both minima and transition structures.
An input section can be read to control which atoms are frozen or unfrozen during an optimization. Atoms can be specified by atom, element, residue, or ONIOM layer.
Analytic frequency-dependent ROA intensities.
Analytic DFT hyperpolarizabilities (betas) and numerical second hyperpolarizabilities (gammas).
Electronic excitation, emission, and photoionization band shapes via Franck-Condon theory using harmonic normal modes for the two states.
Electronic excitation band shapes using Herzberg-Teller or Franck-Condon-Herzberg-Teller theory.
Normal modes can be selected for display, for anharmonic corrections, and for use in FC/HT/FCHT analysis. Selection can be by atom, element, ONIOM layer, or residue.
Information about protein secondary structure can be included in the molecular specification input and in .fchk files.
Orbital by orbital population analysis can be performed, giving contributions to an orbital by atom and by angular momentum.
Canonical UHF/UDFT orbitals can be biorthogonalized, either for visualization or for use as an initial guess for an ROHF calculation.
Natural Transition Orbital analysis of CIS and TD excitations is available.
Mulliken population analysis can be done after projection of the occupied orbitals onto a minimal basis. This gives stable populations when using extended basis sets.
The initial guess for the SCF can be generated by combining calculations on fragments, specifying the charge and spin for each fragment.
Numerical frequencies can be done using four-point differentiation instead of the default two-point, for better accuracy and numerical stability.
HF and DFT frequencies on large molecules are much faster, especially when run in parallel.
FMM and hence linear scaling Coulomb and Exchange are cluster-parallel.
ONIOM(MO:MM) frequencies on large systems are much faster, especially with electronic embedding. Frequencies with 100-200 QM atoms and 6000 MM atoms are practical.
Normal modes can be saved during large frequency calculations and reused for display or printing of modes and starting an IRC=RCFC job.
CC, BD and EOM-CCSD amplitudes can be saved on the checkpoint file and read in to later calculations, including ones using a different basis set. BD orbitals are also saved and can be read back.
Semi-empirical, HF, and DFT frequencies can be restarted in mid-calculation. See Freq=Restart for restarting numerical frequencies and the Restart keyword for restarting analytic frequencies.
CC and EOM-CC calculations can be restarted in mid-calculation.
The initial guesses for individual steps within an ONIOM calculation can be taken from separate checkpoint files. The ONIOM=OnlyInputFiles option causes the input files for each part of the ONIOM calculation to be printed, to facilitate generating wavefunctions separately.
The density fitting sets corresponding to the SVP, TZVP, and QZV basis sets are included. The /Fit keyword requests the fitting set matching the specified AO basis, or /Auto if there is no specific fitting set. Thus BVP86/SVP/Fit will use density fitting with the set defined to accompany the SVP basis, while BVP86/6-31G*/Fit is the same as BVP86/6-31G*/Auto.
The DensityFit keyword can be included in the Default.Route file to make use of fitting the default whenever a pure density functional is requested.
Density basis sets can be read in using coefficients of unnormalized primitives, density normalized primitives, or primitives normalized as though they were AOs. There are other programs which use each of these conventions, so there are published basis sets which require each of these options if they are to be read in to G09.
Single-point SCF calculations now default to full accuracy (SCF=Tight).
The default integral transformation for MP, CC, and BD calculations is Tran=IABC, which is faster than Tran=Full on most machines
The default for Freq=ROA is CPHF=RdFreq, because the frequency-dependent ROA intensities are analytic while those in the static limit are numerical and less accurate.
The default for post-SCF methods such as MP, BD and CC is Tran=IABC, which is more efficient than a full transformation on most machines.
The default SCF accuracy is SCF=Tight for all calculations, including single points.
IRCs default to a new link, L123. Unless it is requested explicitly, the old IRC link (L115) is used only for IRCMax jobs. The default algorithm in L123 is IRC=HPC, except for ONIOM(MO:MM) calculations, for which it is IRC=EulerPC. L123 can use the IRC=GS2 algorithm (the algorithm used in L115), but this is usually much more expensive than the default.
By default, IRCs report only the energy and reaction coordinate at each point on the path. Use IRC=Report to specify internal coordinates whose values should also be tabulated.
Ordinary QM frequency calculations and ONIOM(MO:MM) frequencies default to CPHF=Simultaneous, as in G03. However, semi-empirical frequencies are more efficiently done using CPHF=Separate, which is the default for these cases.
Assignment of atoms to fragments for Counterpoise and Guess=Fragment calculations or for population analysis by fragment is done in the nuclear properties part of the atom specification rather than at the end of the line, where it conflicted with ONIOM input. E.g.,
C(Fragment=3) 0.0 1.0 2.0
rather than the old format:
C 0.0 1.0 2.0 3
Isotopes are normally specified in the nuclear properties part of the atom specification lines. If they are to be read in separately, they are read in once, after the molecule specification, rather than separately by different parts of the program (IRC, Freq, etc.).
Semi-direct integral transformation is the default. This code parallelizes better than the fully direct or in-core algorithms and is similar in speed on a single processor.
The SCFVac PCM input option has been removed. If preliminary gas-phase energy is desired, do this in a separate job step before the solvated calculation.
The default SCF convergence is 108 on the density for all calculations, including single points.
The physical constants used by default are those from the 2006 CODATA tables; those use in Gaussian 03 can be requested via Constants=1998.
Stable=Opt defaults to the usual (L502) SCF procedure for the initial SCF but then uses SCF=QC for additional SCF calculations, if they are required.
The formchk utility can put additional information into the .fchk file, including user-defined MM types and other strings. This information is included when formchk is invoked with the –3 option.
The freqchk utility can recover normal modes from the .chk or .fchk file, and can save the normal modes it generates to a .chk file (but cannot save to an .fchk file).
Last update: 1 August 2013