As already mentioned in section 2.3, some galaxies have MIR colors that resemble those found in nearby HII regions (e.g. compare [Dale et al. 1999], to [Contursi et al. 1997]). This indicates that the MIR-FIR properties of galaxies are useful tools to monitor star formation.
3.1. Spectral energy distributions of starburst galaxies
A first effect that can easily be understood is that the SED of a starburst galaxy is globally shifted toward higher dust temperatures. For instance, Krügel et al. (1998) found that the starburst galaxies they studied did not require a cold (10K) dust phase in their SED. This is also confirmed by Klaas et al. (1999), or in NGC 6090 ([Acosta-Pulido et al. 1999]): the coldest dust phase required for these galaxies is 30-50K, in sharp contrast with what is obtained on normal spirals (see sec. 2.1). One should however note that these temperature decompositions depend highly on the exponent of the dust emissivity assumed ( = 1-2): Klaas et al. (1999) note that if = 2 were used for their sample, instead of = 1, there would be room for a colder dust phase. Therefore, rather than attempting to determine precise temperature values, one should remember that in starburst galaxies, the peak of the SED shifts from 100-200 µm to the 60-100 µm range.
The shift in the FIR is obviously reflected in the MIR. As the heating intensity rises, the small grain emission gradually shifts towards short wavelengths in the MIR window (see section 1.2), producing a steeply rising continuum that can start anywhere in the 4-20 µm range. It is important to realize that the wavelength at which the small grain continuum dominates over the infrared bands can vary and is in fact most of the time beyond 12 µm. This can lead to some starburst galaxies being erroneously classified as normal from their ISOPHOT-S spectrum (e.g [Lu et al. 1999]), which only extends to 12 µm.
Observations of known starbursts in the MIR also reveal that the infrared bands are rarely suppressed: all the star-forming regions of the Antennae show significant, if not dominant, emission from the infrared bands ([Vigroux et al. 1996]); the sample of galaxies with warm IRAS colors selected by Mouri et al. (1998) shows well-defined infrared bands but no sign yet of a rising continuum. Finally, the template starburst spectrum used by Lutz et al. (1998) (again from ISOPHOT-S) is not markedly different from that of NGC 891 ([Mattila et al. 1999]). Therefore, the clear signature of a starburst-powered galaxy in the MIR spectrum is more the presence of a strong continuum longward of ~ 12 µm than the absence of infrared bands (see e.g. [Crowther et al. 1999] for a nice example on NGC 5253)
Finally, an important point for starburst galaxies is the amount of extinction present on the line of sight. Because infrared bands are located on both sides of the 9.7 µm silicate feature, a band-dominated extinction-free MIR spectrum will still show a depression around 10 µm. Optical depth measurements are thus better made with line ratios ([Lutz et al. 1996]) and, in starburst galaxies, span a very wide range, AV ~ 10-100 ([Genzel et al. 1998]). More recently an attempt to measure the extinction by its imprint on the infrared band shapes has met with some success ([Rigopoulou et al. 2000]) and confirmed the range mentioned above.