1 Introduction

AGN (active galactic nuclei) have been detected so far only in about a third of the nearby ULIRGs (Genzel et al. 1998), in the remaining two thirds they may have escaped detection due to enormous extinction (Sanders 1999). Since most of the known AGN are pronounced MIR emitters, the extinction required to hide an AGN must be very high, but so far no technique is known to probe the entire MIR extinction. When the amount of extinction is determined exclusively from data, which are themselves affected by extinction - like the 9.7 and 18 $\mu$m silicate absorption features or the [SIII] 18.7 $\mu$m/33.5 $\mu$m line ratio -, the result can be biased toward too low values, representing only the properties of a shallow surface, but not the entire dust column. Yet, a robust technique is required which provides the order of extinction and clues to whether the absorbing dust is located in a foreground screen or whether it is mixed with the emitters.

Extinction estimates of ULIRGs have also been derived from longer wavelengths, for example the shape of the spectral energy distribution from the moderately affected far-infrared (FIR) to the optically thin sub-mm regimes. These spectra can be fitted by one modified blackbody with a single temperature. This approach, however, ignores the multitude of dust temperatures suggested from the patchy optical morphology, and the formally derived extinction values up to $A_V \approx 1000$ are questionable. Refinements using several modified blackbodies rely on the choice of the dust emissivity $\beta$ and the decomposition into the various temperature components (Klaas et al. 1997; Lisenfeld et al. 2000; Klaas et al. 2001), hence the opacities cannot unambigously be determined.