In press, available from the International Journal of High Speed Electronics and Systems. In continuing efforts to pull every bit of useful information out of the solid-state density functional theory calculations run for the terahertz modeling of all the molecular explosives spectra Teraview, Ltd. could get their hands on, this IJHSES paper expands greatly on the SPIE paper posted previously (available here), looking at the dependences on parameter selection in the reproductions of THz spectra for HMX and PETN. Not the assignments of the two spectra (the HMX was performed as posted below, the PETN is as posted above), but the variation in simulation quality as a function of the treatment of the electronic structure (functionals, basis sets, integration grid sizes). I'm in complete agreement with Kieron Burke (well, it's a fact, not an opinion, so what choice do I have) that "Density functional theory is a completely different, formally rigorous, way of approaching any interacting problem, by mapping it exactly to a much easier-to-solve non-interacting problem." The problem, to quantum chemists, is the empirical methodology used to develop density functionals. The fact that programs like GAMESS and Gaussian offers 30-or-so density functionals should clue the user in that the agreement between theory and experiment may be as much due to the choice of density functional as the quality of the basis set for the chemical question being addressed. Empirical methods mean never having to say you're sorry.
With the mandatory spectra figure from the paper, I also post here the timing results for running solid-state HMX and PETN on a single AMD Opteron processor. For what we do to simulate THz spectra, the DMol3 program-option "fine" grid size (0.15 Angstrom grid interval or rmaxp = 12.0 au (6.36 A); up to l = 6 if you're into the whole [Grid Points] = 0.3351[l]2 + 0.5552[l] + 3.9277 thing) is adequate for reproducing THz spectra and making assignments (grid size here referring to the quality of the mesh defined around a molecule used for matrix element evaluation).
D. G. Allis and T. M. Korter
Abstract: The analytical applications of terahertz (THz) spectroscopy for the characterization of molecular solids have been limited by the lack of information concerning the assignment of observed spectral features to specific internal (intramolecular) and external (intermolecular) atomic motions. Computational methodologies addressing the assignment of spectral data are the enabling technology for moving THz spectroscopy to the forefront of available detection methods for both imaging and spectroscopic applications. Solid-state density functional theory (DFT) studies have been performed on the high explosives cyclotetramethylenetetranitramine (HMX) and pentaerythritol tetranitrate (PETN) in order to address the dependencies of the predictions of solid-state vibrations in the terahertz (3 to 120 cm-1) region on the choice of basis set and integration grid size, building on previous work that examined this dependency on the choice of density functional. DFT THz simulations reveal that both the choice of basis set and grid size have important influences on the reproduction of spectral features. The sensitivity to basis set choice is most pronounced in the calculation of vibrational intensities, where it is found that THz absorption intensities are most accurately reproduced when derived from basis set-sensitive Mulliken atomic charges as opposed to basis set-insensitive atomic charges generated by the Hirshfeld partitioning method.