Bolometric, blue and visual light curves are presented for a set of theoretical models for Type Ia supernovae including deflagration, detonation, delayed detonation, pulsating delayed detonation and tamped detonation scenarios. The explosions are calculated using a one-dimensional Lagrangian code consistently with a nuclear reaction network (Khokhlov 1991b,c). For the calculations of the light curves a code is used (Hoflich et al. 1992a,b) which is based on a LTE radiation transfer scheme (including both an energy equation for matter and radiation, and effects due to electron and line scattering). The transfer scheme implicitly solves the time-dependent, frequency averaged moment equations. The light curve code further contains a detailed equation of state with an elaborate treatment of the ionization balance and the ionization energies. Time-dependent expansion opacities (both Rosseland and Planck means) are used in a local approximation, which takes into account the composition structure of the explosion model. The code finally contains a Monte Carlo gamma-ray deposition scheme, which handles all relevant gamma-ray transitions and interaction processes. We find that below a temperature of about 2 10(4) K the opacity rapidly drops from a value of greater than or similar to 0.1 cm2/g by more than an order of magnitude. This drop in opacity leads to a transition from the optically thick to the optically thin regime, which in turn determines to first order the time of maximum light. The transition occurs at different times in models with different expansion rates and structures. Consequently, the absolute value of the opacity does only weakly influence the time of maximum light as long as the opacity above the transition temperature is larger than almost-equal-to 0.1 cm2/g. This is in strict contrast to light curve models using a constant opacity. The light curves of the investigated models further strongly differ in their brightness at maximum light, the width of the maximum, and their post-maximum decline rate. These differences can be understood in terms of the expansion rate of the ejecta, the total energy release, the distribution of the radioactive matter, and the total mass and density structure of the envelope. Several correlations between observable quantities and model parameters are found, which allow for a discrimination between models. For example, the maximum bolometric brightness decreases with the rise time t(bol) and increases with the total mass of radioactive Ni-56 . The photospheric velocity at t(bol) increases with L(bol). Rise times to maximum bolometric luminosity longer than almost-equal-to 15 days can hardly be provided by ''standard'' models (i.e., deflagrations, detonations or delayed detonations), but require ''non-standard'' models (i.e., pulsating delayed detonations or tamped detonations). These two classes of models can be distinguished by the time dependence of the expansion velocity at the photosphere v(ph), which is directly measurable by the Doppler shift of spectral lines. Contrary to the ''standard'' models, the ''non-standard'' models show a distinct plateau in v(ph) after maximum whose width is determined by the envelope mass. The investigated ''standard'' models all have an absolute visual magnitude of M(V) = -19.68m +/- 0.12m, while ''non-standard'' models have a systematically lower maximum brightness being as low as M(V) = -19.2m and decreasing with increasing envelope mass. Thus, if the photospheric velocity shows a plateau, the supernova should be used with care when determining the Hubble constant.
