While galaxies showing both peculiar morphologies and signs of global youth have been known for decades (e.g. [1, 2, 3]), it was not until a classic study on the UBV colors of peculiar galaxies by Larson & Tinsley in 1978  that the study of starbursts in interacting galaxies began in earnest. In that seminal work the authors showed that the disturbed systems from the Arp Atlas of Peculiar Galaxies  had a larger spread in colors and significantly bluer colors than a comparison sample of normal Hubble types. The Toomres  had recently demonstrated quite convincingly that gravitational interactions provide a natural explanation for many of the types of peculiarities exhibited by the systems in Arps atlas, and Larson & Tinsley posited that such interactions induce small bursts (~ 1-5% by mass) of star formation within the host(s). They were able to explain the color distribution of the Arp systems with burst models of varying strength and age superimposed upon an underlying host with normal colors.
Much subsequent work has supported this suggestion . The consensus is that interacting galaxies as a class have mean levels of star formation that are factors of 2-5 higher than normal spirals for optically selected samples, or 2-20 times higher for samples selected on the basis of IR luminosity. This point is illustrated in Figure 1 , which compares the IR luminosities of an optically selected samples of isolated and interacting galaxies. Also shown are the values for a sample of ultraluminous infrared (ULIR 1) galaxies which are known to be on-going mergers . Similar results have been derived using other measures of star formation, finding that interactions serve primarily to concentrate moderately enhanced star formation into the central regions of galaxies rather than to globally raise the star formation rate (e.g. [10, 11, 12, 13]).
Figure 1. Histograms of LIR and LIR / LB for isolated, interacting, and infrared bright systems. The isolated and interacting samples were chosen optically, independent of their infrared properties. Arrows indicate their means. These plots show that while the brightest IR emitting systems are interacting, many interacting systems are not IR bright. From Bushouse, Lamb & Werner 1988 .
Figure 1 illustrates several other points. The first is that none of the isolated systems have LIR > 3 × 1010 L or LIR / LB > 15. At these levels almost all galaxies are interacting or merging [9, 14, 15]. In fact, IR luminosity appears to be the most efficient way to select interacting systems: while the overall peculiar fraction of optically selected samples is around 9% , the fraction of morphologically peculiar galaxies approaches 90% or higher at IR luminosities above 5 × 1011 L (see  and references therein). The second is that many systems in the interacting sample do not exhibit enhanced levels of star formation. So while the most luminous IR sources are almost invariably mergers, not all mergers are luminous IR sources.
IR luminosity is believed to be directly related to the number of hot stars present, as the dust absorbs the UV photons from these stars and reradiates them in the IR [14, 15]. It can therefore be used as a measure of the massive star formation rate (MSFR; M > 5 M) . The IR luminous systems are both gas-rich and dusty [9, 19, 20], and most of their bolometric luminosity emerges in the far IR [9, 17]. While there has been a long-standing debate over whether the most luminous systems are instead powered by an obscured AGN (see e.g. [13, 21]), this ambiguity is mostly at the highest IR luminosities, and the majority are believed to be predominantly powered by starbursts (see reviews by [7, 22]). The popular picture which has emerged is that two gas rich systems undergo a close interaction, which leads to orbital decay and eventual merging. The gas is compressed, dissipates and moves inward, stimulating a circumnuclear starburst. Dust, which is coupled to the gas, absorbs much of the UV radiation and re-radiates it in the IR. This high level of star formation quickly subsides as the burst consumes the available gas.
But do all mergers experience a strong burst of star formation? If not, what are the deciding factors? These are the issues I will discuss in this review.
1 LIR > 3 × 1011 L in this case, although other definitions are used. Back.