2.7. Luminous Infrared Galaxies
In carrying to the extreme the concept of a starburst, we find the powerful luminous infrared galaxies (LIGs). At luminosities above 1011 L, LIGs (LFIR > 1011 L) are the dominant extragalactic objects in the local universe (z < 0.3) with such high luminosities. Some, having LFIR > 1012 L, are the most luminous local objects (see  for a review). These galaxies possess very large amounts of molecular material (e.g., [329, 330, 331, 332, 333]). They present large CO luminosities, but also a high value for the ratio LFIR / LCO, both being about one order of magnitude greater than for spirals, implying a higher star formation rate per solar mass of gas. (27) LIGs are generally regarded as recent galaxy mergers in which much of the gas of the colliding objects has fallen into a common center (typically less than 1 kpc in extent), triggering a huge starburst phenomenon . There is evidence for the existence of even more extreme enviroments within LIGs (see, e.g., ): These, larger than giant molecular clouds but with densities found only in small cloud cores, appear to be the most outstanding star formation places in the universe. They are well traced by HCN emission, i.e., they produce a substantial fraction of the whole HCN emission observed for the whole galaxy , see also [334, 335]. The CR enhancement factor in these small but massive regions can well exceed the average value for the galaxy. In Arp 220, for instance, two such regions were discovered to contain about 2 × 109 M . If the CR enhancement in these regions is significantly larger than the starburst average, these extreme environments could be the main origin for any -ray emission observed from this galaxy .
2.7.2. Propagation and further studies
Using the PSCz catalogue , Smialkowski et al.  constructed all-sky maps of UHE proton intensities plausibly originating in LIGs, taking into account effects of particle propagation through the extragalactic medium and the possible influence of the regular galactic magnetic field. The PSCz catalogue consists of almost 15000 IR galaxies with known redshifts, covering 84% of the sky. Finding correlations with such an overdistributed sample might be thought as a risky business. There are, however, some phenomenological reasons - apart of LIGs being super-starbursts - to search for a possible UHECR origin in LIGs.
Arp 299, one of the the brightest infrared source within 70 Mpc and a system of colliding galaxies showing intense starburst, appeared earlier as VV 118 in the list of candidates for the AGASA triplet presented by in Ref. [124, 63, 127]. Indeed, for Arp 299, even when a hidden AGN was observed, it can not account for the whole FIR luminosity  and a strong starburst activity is required to explain it. It might be considered as a prime candidate for the origin of the triplet , if such is not a statistical fluctuation.
Equatorial maps of the expected intensity of the UHE protons originating in LIGs for the energy region 40-80 EeV with sky coverage and declination dependent exposure for the AGASA experiment were presented in Ref. . Expected proton intensities for 40-80 EeV appear to show good correlation with the distribution of the experimental events from AGASA, especially, from the region of the sky near Arp 299 and the AGASA triplet of events (RA 170°, 60°). No correlation of the data events above 80 EeV with the expected CR intensities was reported.
The latter result was confirmed in Ref. , where application of the Kolmogorov-Smirnov (KS) to the 11 highest AGASA events above 1020 eV yields a KS probability of < 0.5%, rejecting the possible association at > 99.5% significance level. This was used to argue that the existing CR events above 1020 eV do not owe their origin to long burst GRBs, rapidly rotating magnetars, or any other events associated with core collapse supernovae (although it would yet be too early to completely rule out such a possibility, based only in a 2 deviation). (28)
A more sensitive approach to study a possible LIG origin of UHECRs, perhaps, is not to look for possible correlations between the whole PSCz calatog and UHECRs, but rather to select a priori which LIGs are the most likely to be detectable by their possible UHECR emission. There are several LIGs for which reasonable values of CR enhancements, comparable to, or lower than, the ratio between their SFR and the Milky Way's, can provide a -ray flux above GLAST sensitivity, and if the CR spectrum is sufficiently hard, also above the sensitivities of the new Cerenkov telescopes. These LIGs are then most likely to appear as new point-like -ray sources. Even when it is natural to expect that a LIG will emit -rays, only the more gaseous, nearby, and CR-enhanced galaxies are the ones which could be detected as -ray point sources .
Out of the HCN just presented in Refs. [334, 335] (29) and the larger Pico dos Dias Survey (PDS, ) (30), the most likely -ray sources are listed in Table 1, together with EGRET fluxes upper limits, obtained in . In our opinion, this group, as well as the HCN and the PDS should be separately taken into account when searching for possible UHECR correlations with new sets of data. (31)
|Name||DL||R||log(LFIR / L)||log M(H2 / M)||<k>||F> 100 MeVEGRET|
|[Mpc]||[pc]||[10-8 photons cm-2 s-1]|
|NGC4038 / 9||22||0.49||10.65||9.07||412||<3.7|
27 For stars to form, a large mass fraction at high density, dM / d (log n), is required. The Milky Way has an order of magnitude more gas at less than 300 cm-3 than what it has at 104 cm-3, contrary to LIGs, which have nearly equal mass per decade of density between 102 and 104 cm-3 (see, e.g., Ref. [332, 333]) Back.
28 Following , the core collapse supernova rate per galaxy, summed over all galaxy types, is sn 0.011 SN per galaxy-year , which yields a volume averaged rate of supernovae of sn ng 2.2 × 10-4 SN/Mpc3-year. The supernova rate for galaxies with far infrared emission is snfir = 2.5 × 10-4 L10 SN/(FIR galaxy)-year . Integrating this supernova rate over the FIR galaxy luminosity function yields snfir nfirg 0.7 × 10-4Kobsc(L10max / 300) SN/Mpc3-year. Kobsc, the correction for supernovae missed in the existing optical and near infrared supernova detection surveys, might be as large as 10, and probably is at least as large as 3 . Thus, the luminous infrared galaxies contribute at least 28% (Kobsc = 1) of the total supernova rate, with a total space density only 25% that of all normal galaxies. Furthermore, since the brightest infrared galaxies (LFIR > 1012 L) dominate the contribution from all FIR galaxies, and these are quite rare, with space density ~ 4 × 10-8 Mpc-3 , correlations of UHECR arrival directions with the sky positions of the luminous IRAS galaxies offers a promising opportunity to test the hypothesis that UHECR acceleration has something to do with core collapse supernovae, as is implied by the GRB shock and magnetar unipolar inductor models for the acceleration sites. Back.
29 This survey is a systematic observation (essentially, all galaxies with strong CO and IR emission were chosen for HCN survey observations) of 53 IR-bright galaxies, including 20 LIGs with LFIR > 1011 L, 7 with LFIR > 1012 L, and more than a dozen of the nearest normal spiral galaxies. It also includes a literature compilation of data for another 9 IR-bright objects. This is the largest and most sensitive HCN survey (and thus of dense interstellar mass) of galaxies to date. Back.
30 The PDS survey consists of relatively nearby and luminous starbursts galaxies selected in the FIR. PDS galaxies have a lower mean IR luminosity log(LIR / L) = 10.3± 0.5, redshifts smaller than 0.1, and form a complete sample limited in flux in the FIR at 2 × 10-10 erg cm-2 s-1. Back.
31 A stacking procedure with EGRET data is to be reported by Cillis et al. . Back.