Spitzer IRS spectra of a large sample of Seyfert galaxies:: A variety of infrared spectral energy distributions in the local active galactic nucleus population

We are conducting a large observing program with the Spitzer Space Telescope to determine the mid- to far-IR spectral energy distributions of a well-defined sample of 87 nearby, 12 mu m - selected Seyfert galaxies. In this paper we present the results of Spitzer IRS low-resolution spectroscopy of a statistically representative subsample of 51 of the galaxies (59%), with an analysis of the continuum shapes and a comparison of the Seyfert types. We find that the spectra clearly divide into groups based on their continuum shapes and spectral features. The largest group (47% of the sample of 51) shows a very red continuum suggestive of cool dust and strong emission features attributed to polycyclic aromatic hydrocarbons. Sixteen objects (31%) have a power-law continuum with spectral indices of alpha(5-20) (mu m) = -2.3 to -0.9 that flatten to alpha(20-35) (mu m) = -1:1 to 0.0 at similar to 20 mu m. Clear silicate emission features at 10 and 18 mu m are found in two of these objects (Mrk 6 and Mrk 335). A further 16% of the sample show power-law continua with unchanging slopes of alpha 5-35 mu m = -1:7 to -1.1. Two objects are dominated by a broad silicate absorption feature. One object in the sample shows an unusual spectrum dominated by emission features, which is unlike any of the other spectra. Some spectral features are clearly related to a starburst contribution to the IR spectrum, while the mechanisms producing observed power-law continuum shapes, attributed to an active galactic nucleus (AGN) component, may be dust or nonthermal emission. The IR spectral types appear to be related to the Seyfert types. Principal component analysis results suggest that the relative contribution of starburst emission may be the dominant cause of variance in the observed spectra. The derived starburst component of each spectrum, however, contributes < 40% of the total flux density. We compare the IR emission with the optically thin radio emission associated with the AGN and find that Seyfert 1 galaxies have higher ratios of IR to radio emission than Seyfert 2 galaxies, as predicted by the unified model if the torus is optically thick in the mid- IR. However, smooth-density torus models predict a much larger difference between Seyfert types 1 and 2 than the factor of 2 difference observed in our sample; the observed factor of similar to 2 difference between the type 1 and type 2 galaxies in their IR-to-radio ratios above 15 mu m requires the standard smooth-density torus models to be optically thin at these wavelengths. However, the resulting low torus opacity requires that the high observed columns detected in X-ray absorption be produced in gas with a very low dust-to-gas ratio (perhaps within the dust sublimation region). On the other hand, our observations may be consistent with clumpy torus models containing a steep radial distribution of optically thick dense clumps. The selection of our sample at 12 mu m, where the torus may be optically thick, implies that there may be orientation-dependent biases in the sample; however, we do not find that the sample is biased toward Seyfert 2 galaxies with more luminous central engines, as would be expected. We find that the Seyfert 2 galaxies typically show stronger starburst contributions than the Seyfert 1 galaxies in the sample, contrary to what is expected based on the unified scheme for AGNs. This may be due to the selection effect that only those Seyfert 2 galaxies with strong starburst contributions had high enough integrated 12 mu m flux densities to fall abov the flux limit of the sample.