A mixed quantum-classical approach to simulate the coupled dynamics of electrons and nuclei in nanoscale molecular systems is presented. The method relies on a second order expansion of the Lagrangian in time-dependent density functional theory (TDDFT) around a suitable reference density. We show that the inclusion of the second order term renders the method a self-consistent scheme and improves the calculated optical spectra of molecules by a proper treatment of the coupled response. In the application to ion-fullerene collisions, the inclusion of self-consistency is found to be crucial for a correct description of the charge transfer between projectile and target. For a model of the photoreceptor in retinal proteins, nonadiabatic molecular dynamics simulations are performed and reveal problems of TDDFT in the prediction of intra-molecular charge transfer excitations.

An analytical formulation for the geometrical derivatives of excitation energies within the time-dependent density-functional tight-binding (TD-DFTB) method is presented. The derivation is based on the auxiliary functional approach proposed in [Furche and Ahlrichs, J Chem Phys 2002, 117, 7433]. To validate the quality of the potential energy surfaces provided by the method, adiabatic excitation energies, excited state geometries, and harmonic vibrational frequencies were calculated for a test set of molecules in excited states of different symmetry and multiplicity. According to the results, the TD-DFTB scheme surpasses the performance of configuration interaction singles and the random phase approximation but has a lower quality than ab initio time-dependent density-functional theory. As a consequence of the special form of the approximations made in TD-DFTB, the scaling exponent of the method can be reduced to three, similar to the ground state. The low scaling prefactor and the satisfactory accuracy of the method makes TD-DFTB especially suitable for molecular dynamics simulations of dozens of atoms as well as for the computation of luminescence spectra of systems containing hundreds of atoms.

Low-energy excitations and optical absorption spectrum of C(60) are computed by using time-dependent (TD) Hartree-Fock, TD-density functional theory (TD-DFT), TD DFT-based tight-binding (TD-DFT-TB), and a semiempirical Zerner intermediate neglect of diatomic differential overlap method. A detailed comparison of experiment and theory for the excitation energies, optical gap, and absorption spectrum of C(60) is presented. It is found that electron correlations and correlation of excitations play important roles in accurately assigning the spectral features of C(60), and that the TD-DFT method with nonhybrid functionals or a local spin density approximation leads to more accurate excitation energies than with hybrid functionals. The level of agreement between theory and experiment for C(60) justifies similar calculations of the excitations and optical absorption spectrum of a monomeric azafullerene cation C(59)N(+), to serve as a spectroscopy reference for the characterization of carborane anion salts. Although it is an isoelectronic analogue to C(60), C(59)N(+) exhibits distinguishing spectral features different from C(60): (1) the first singlet is dipole-allowed and the optical gap is redshifted by 1.44 eV; (2) several weaker absorption maxima occur in the visible region; (3) the transient triplet-triplet absorption at 1.60 eV (775 nm) is much broader and the decay of the triplet state is much faster. The calculated spectra of C(59)N(+) characterize and explain well the measured ultraviolet-visible (UV-vis) and transient absorption spectra of the carborane anion salt [C(59)N][Ag(CB(11)H(6)Cl(6))(2)] [Kim et al., J. Am. Chem. Soc. 125, 4024 (2003)]. For the most stable isomer of C(48)N(12), we predict that the first singlet is dipole-allowed, the optical gap is redshifted by 1.22 eV relative to that of C(60), and optical absorption maxima occur at 585, 528, 443, 363, 340, 314, and 303 nm. We point out that the characterization of the UV-vis and transient absorption spectra of C(48)N(12) isomers is helpful in distinguishing the isomer structures required for applications in molecular electronics. For C(59)N(+) and C(48)N(12) as well as C(60), TD-DFT-TB yields reasonable agreement with TD-DFT calculations at a highly reduced cost. Our study suggests that C(60), C(59)N(+), and C(48)N(12), which differ in their optical gaps, have potential applications in polymer science, biology, and medicine as single-molecule fluorescent probes, in photovoltaics as the n-type emitter and/or p-type base of a p-n junction solar cell, and in nanoelectronics as fluorescence-based sensors and switches.

Ab initio calculations in the local-density approximation have been carried out in SiC to determine the possible configurations of the isolated oxygen impurity. Equilibrium geometry and occupation levels were calculated. Substitutional oxygen in 3C-SiC is a relatively shallow effective mass like double donor on the carbon site (O C ) and a hyperdeep double donor on the Si site (O Si ). In 4H-SiC O C is still a double donor but with a more localized electron state. In 3C-SiC O C is substantially more stable under any condition than O Si or interstitial oxygen (O i ). In 4H-SiC O C is also the most stable one except for heavy n-type doping. We propose that O C is at the core of the electrically active oxygen-related defect family found by deep level transient spectroscopy in 4H-SiC. The consequences of the site preference of oxygen on the SiC/SiO 2 interface are discussed.

There are several misprints in the appendix of our original article.The Eqs. (B6) and (B7) should read