SPECTRAL EVOLUTION OF THE EXTRAORDINARY TYPE IIn SUPERNOVA 2006gy

We present a detailed analysis of the extremely luminous and long-lasting Type IIn supernova (SN) 2006gy using spectra obtained between days 36 and 237 after explosion. We derive the temporal evolution of the effective temperature, radius, blast-wave and SN-ejecta expansion speeds, and bolometric luminosity, as well as the progenitor wind density and total swept-up mass overtaken by the shock. SN 2006gy can be interpreted in the context of shock interaction with a dense circumstellar medium (CSM), but with quite extreme values for the CSM mass of similar to 20 M-circle dot and SN explosion kinetic energy of at least 5 x 10(51) erg. A key difference between SN 2006gy and other SNe IIn is that, owing to its large amount of swept-up mass, the interaction region remained opaque much longer. At early times, Ha emission-line widths suggest that the photosphere is ahead of the shock, and photons diffuse out through the opaque CSM shell. The pivotal transition to optically thin emission occurs around day 110, when we start to see a decrease in the blackbody radius R-BB and strengthening tracers of the post-shock shell. From the evolution of pre-shock velocities, we deduce that the CSM was ejected by the progenitor star in a greater than or similar to 10(49) erg precursor event similar to 8 yr before the explosion. The large CSM mass definitively rules out models involving stars with initial masses of less than or similar to 10 M-circle dot. If the pre-SN mass budget also includes the likely SN ejecta mass of 10-20 M-circle dot and the distant >10 M-circle dot shell inferred from a light echo, then even massive M-ZAMS = 30-40 M-circle dot progenitor stars are inadequate. At roughly solar metallicity, substantial mass loss probably occurred during the star's life, so SN 2006gy's progenitor is more consistent with sequential giant luminous blue variable eruptions or pulsational pair-instability ejections in extremely massive stars with initial masses above 100 M-circle dot. This requires significant revision to current paradigms of massive-star evolution.