Magnetic fields in starburst galaxies and the origin of the FIR-radio correlation

We estimate minimum energy magnetic fields ( B-min) for a sample of galaxies with measured gas surface densities, spanning more than four orders of magnitude in surface density, from normal spirals to luminous starbursts. We show that the ratio of the minimum energy magnetic pressure to the total pressure in the ISM decreases substantially with increasing surface density. For the ultraluminous infrared galaxy Arp 220, this ratio is similar to 10(-4). Therefore, if the minimum energy estimate is applicable, magnetic fields in starbursts are dynamically weak compared to gravity, in contrast to normal star-forming spiral galaxies. We argue, however, that rapid cooling of relativistic electrons in starbursts invalidates the minimum energy estimate. We assess a number of independent constraints on the magnetic field strength in starburst galaxies. In particular, we argue that the existence of the FIR-radio correlation implies that the synchrotron cooling timescale for cosmic-ray electrons is much shorter than their escape time from the galactic disk; this in turn implies that the true magnetic field in starbursts is significantly larger than B-min. The strongest argument against such large fields is that one might expect starbursts to have steep radio spectra indicative of strong synchrotron cooling, which is not observed. However, we show that ionization and bremsstrahlung losses can flatten the nonthermal spectra of starburst galaxies even in the presence of rapid cooling, providing much better agreement with observed spectra. We further demonstrate that ionization and bremsstrahlung losses are likely to be important in shaping the radio spectra of most starbursts at GHz frequencies, thereby preserving the linearity of the FIR-radio correlation. We thus conclude that magnetic fields in starbursts are significantly larger than B-min. We highlight several observations that can test this conclusion.