MENU

0

What's New in Gaussian 03

02/03/2004

Discover what's new in Gaussian 03 - Enhanced ONIOM Method, Solvent Effects and more.

New Chemistry

Enhanced ONIOM Method

The ONIOM facility in Gaussian 03 has been significantly enhanced over that offered by Gaussian 98 [1-2]:

  • The ONIOM facility [42] now supports electronic embedding for ONIOM(MO:MM) calculations: the electrostatic properties of the MM region can be taken into account during computations on the QM region.

    • ONIOM(MO:MM) optimizations are much faster and can be reliably performed for large molecules (e.g., proteins). The algorithmic improvements include:

    • A quadratic coupled algorithm takes into account the coupling between atoms using internal coordinates (typically, those in the model system) and those in Cartesian coordinates (typically, the atoms only in the MM layer), resulting in more accurate steps.

  • MO/MM optimizations perform micro-iterations for the atoms only in the MM layer between traditional optimization steps on the real system, resulting in faster and more reliable optimizations. Electronic embedding can be combined with micro-iterations.

  • Analytic frequencies are available for ONIOM(MO:MM) calculations, and frequencies for ONIOM(MO:MO) calculations are significantly faster.

  • Gaussian 03 provides support for general molecular mechanics (MM) force fields, including read-in and modified parameters. A standalone MM optimization program is also included.

  • Support for an external program for any ONIOM model (e.g., an external MM program may be used).

Solvent Effects

The Polarizable Continuum Model (PCM) solvation method has been improved and extended [3-8]:

  • The IEFPCM model [3,9] is now the default, and analytic frequencies are now available for this SCRF method. Additional performance improvements include a new cavity generation technique [10].

  • Many additional properties can be modeled in solution (discussed later in this brochure).

  • Gaussian 03 can also produce input for Klamt's COSMO-RS program [11], which computes solvation energies, partition coefficients, vapor pressure and other bulk properties via statistical mechanics techniques.

Periodic Boundary Conditions (PBC)

Gaussian 03 offers PBC calculations for studying periodic systems: e.g., polymers, surfaces and crystals [12-15]. PBC calculations solve the Schrödinger equation subject to the boundary condition that the molecule and the wavefunction repeat indefinitely in one, two or three directions. Hartree-Fock and DFT energies and gradients are available for periodic systems.

Molecular Dynamics

Dynamics calculations can provide qualitative understanding of reaction mechanisms and quantitative details about the reaction such as product distributions. There are two main approaches to performing these calculations:

  • In Born-Oppenheimer Molecular Dynamics (BOMD), classical trajectories are calculated on a local quadratic approximation to the potential energy surface (for a review, see [16]). Our implementation [17] uses a Hessian-based algorithm for the predictor and corrector steps, an approach which results in a factor of 10 or more improvement in the step size over previous implementations. While it can make use of analytic second derivatives, BOMD is available for all theoretical methods having analytic gradients.

  • Gaussian 03 also offers Atom-Centered Density Matrix Propagation (ADMP) method [18-20] molecular dynamics (available for Hartree-Fock and DFT). Drawing on the work of Car and Parrinello [21], ADMP propagates the electronic degrees of freedom rather than solving the SCF equations at each nuclear geometry. Unlike CP, ADMP propagates the density matrix rather than the MOs. This is much more efficient if an atom-centered basis set is being used. This approach overcomes some limitations inherent in the CP implementation: e.g., there is no need to substitute D for H in order to maintain energy conservation, and both pure and hybrid DFT functionals can be used. ADMP calculations can also be performed in the presence of a solvent [22], and ADMP can be used in ONIOM(MO:MM) calculations.

Excited States

There are additions and several enhancements to excited states methods:

  • CASSCF calculations are now more efficient due to a new algorithm for evaluating the CI-vector in the full configuration interaction calculation [23]. Practical active spaces increase to about 14 orbitals for energies and gradients (they remain at about 8 orbitals for frequencies).

  • The Restricted Active Space (RAS) SCF method [24] is also available[25]. RASSCF calculations partition the molecular orbitals into five sections: the lowest lying occupieds (considered inactive in the calculation), the RAS1 space of doubly occupied MOs, the RAS2 space containing the most important orbitals for the problem, the RAS3 space of weakly occupied MOs and the remaining unoccupied orbitals (also treated as frozen by the calculation). Thus, the active space in CASSCF calculations is divided into three parts in a RAS calculations, and allowed configurations are defined by specifying the minimum number of electrons that must be present in the RAS1 space and the maximum number that must be in the RAS3 space, in addition to the total number of electrons in the three RAS spaces.

  • NBO orbitals for may be used for defining CAS and RAS active spaces. These provide good initial guesses for the required antibonding orbitals which correlate with the bonds/lone pairs of interest.

  • The Symmetry Adapted Cluster/Configuration Interaction (SAC-CI) method of Nakatsuji and coworkers is now included in Gaussian. This method has many uses: predicting very accurate excited states of organic systems, studying two-to-many electron excitation processes such as the shake-up in the ionization spectrum, and other problem types. For an overview of the SAC-CI method, see [26-27].

  • Solvent Effects: Excited states can be modeled in the presence of a solvent [28-29] using the CI-Singles and Time Dependent Hartree-Fock and DFT methods.

Molecular Properties

Gaussian 03 provides many new molecular properties:

  • Spin-spin coupling constants [31-34], which can aid in distinguishing conformations in magnetic spectra.

  • g tensors and other hyperfine spectra tensors [49-52]. Gaussian 03 can produce nuclear electric quadrupole constants, rotational constants, the quartic centrifugal distortion terms, the electronic spin rotation terms, the nuclear spin rotation terms, the dipolar hyperfine terms and Fermi contact terms. All tensors can be exported to Pickett's fitting and spectral analysis program [53].

  • Harmonic vibration-rotation coupling [43-44]: A spectroscopic property dependent on the coupling between molecules' vibrational and rotational modes. It is used to analyze detailed rotational spectra.

  • Anharmonic vibration and vibration-rotation coupling [44-48]: Using perturbation theory, these higher order terms are incorporated into frequency calculations in order to produce more accurate results.

  • Pre-resonance Raman spectra which yield information about ground state structures, connectivity, and vibrational states.

  • Optical Rotations/Optical Rotary Dispersion: Used to distinguish enantiomers of chiral systems [39-41] (this property is computed via GIAOs).

  • Electronic Circular Dichroism (ECD): This property is the differential absorption in the visible and ultraviolet regions for optically active molecules, and is used to assign absolute configurations [35-36]. Predicted spectra can also be useful in interpreting existing ECD data and peak assignments.

  • Frequency-dependent polarizabilities and hyperpolarizabilities, which can be used to study how the molecular properties of materials vary with wavelength of the incident light [37-38].

  • Magnetic susceptibilities computed with Gauge-Independent Atomic Orbitals (GIAOs) [30]. This property is the magnetic analogue to the electric polarizability, and it provides insight into the diamagnetic vs. paramagnetic character of molecules.

  • Solvent Effects: Electric and magnetic properties and the various spectra can be predicted for systems in solution as well as ones in the gas phase [54-56].

  • Properties with ONIOM: The ONIOM method may be used with these electric and magnetic properties.

Fundamental Algorithms

  • Much Better Initial Guesses: Gaussian 03 uses the Harris functional for generating initial guesses. This functional [59] is a non-iterative approximation to DFT, and it produces initial guesses which are better than those produced by Gaussian 98: for example, there are modest improvements for organic systems but very substantial improvements for compounds containing metals.

  • New SCF Convergence Algorithm: The default SCF algorithm now uses a combination of two Direct Inversion in the Iterative Subspace (DIIS) extrapolation methods EDIIS and CDIIS. EDIIS [58] uses energies for extrapolation, and it dominates the early iterations of the SCF convergence process. CDIIS, which performs extrapolation based on the commutators of the Fock and density matrices, handles the latter phases of SCF convergence. This new algorithm is very reliable, and previously troublesome SCF convergence cases now almost always converge with the default algorithm. For the few remaining pathological convergence cases, Gaussian 03 offers Fermi broadening and damping in combination with CDIIS (including automatic level shifting).

  • Density Fitting for Pure DFT Calculations: Gaussian 03 provides the density fitting approximation [60,61] for pure DFT calculations. This approach expands the density in a set of atom-centered functions when computing the Coulomb interaction instead of computing all of the two-electron integrals. It provides significant performance gains for pure DFT calculations on medium sized systems too small to take advantage of the linear scaling algorithms without a significant degradation in accuracy. Gaussian 03 can generate an appropriate fitting basis automatically from the AO basis, or you may select one of the built-in fitting sets.

  • Faster and Automated FMM: The fast multipole method (FMM) in Gaussian 98 allowed the computational cost for large DFT calculations to scale linearly with system size. In Gaussian 03, improvements to these algorithms [57] means that their performance gains can be realized for systems of more modest size as well (~100 atoms for pure DFT calculations and ~150 atoms with hybrid functionals). In addition, this feature is now fully automated: the program invokes FMM automatically when appropriate.

  • Coulomb Engine: Gaussian 03 incorporates a faster algorithm for the Coulomb operator for pure DFT calculations. The Coulomb engine produces the exact Coulomb matrix without explicitly forming four center two electron integrals. This substantially reduces the CPU time for the Coulomb problem in pure DFT calculations.

  • O(N) Exact Exchange: A new algorithm for Hartree-Fock and DFT calculations using hybrid functionals implements screening of the exact exchange contribution via the density matrix to eliminate the many zero value terms [62]. This technique results in a linear computational cost for these methods without accuracy loss.

Additional Features

  • Additional DFT Functionals:

    • OPTX exchange functional [69].

    • PBE [70-71] and B95 [72] correlation functionals.

    • VSXC [73], HCTH [74] pure functionals,

    • B1 [72] and variations, B98 [75, 83], B97-1 [76], B97-2 [77], and PBE1PBE [71] hybrid functionals.

  • High Accuracy Energy Methods:

    • G3 and variations [78,79].

    • The W1 method of Jan Martin [80-81], modified slightly to use the UCCSD method rather than ROCCSD for open shell systems (this method is denoted W1U). Gaussian 03 also includes the related W1BD method, which substitutes the BD method for coupled cluster [84]. This method is both more expensive and more accurate than CBS-QB3 and G3.

  • Douglas-Kroll-Hess scalar relativistic Hamiltonian: This feature allows all electron calculations for heavier atoms (first and second transition rows) when ECPs are not accurate enough [63-66]. For an overview, see [67-68]

  • Gaussian 03 also includes the very large universal Gaussian basis set of de Castro, Jorge and coworkers [82], which approaches the basis set limit.

© 2024 Hearne Software