Annu. Rev. Astron. Astrophys. 1996. 34: 749-792 Copyright © 1996 by Annual Reviews. All rights reserved

4. PROPERTIES OF LUMINOUS INFRARED GALAXIES

Substantial multiwavelength observations now exist for large samples of IRAS galaxies; the most extensive and highest spatial resolution observations are of objects in the BGS. Data for LIGs are summarized below by wavelength region. Wherever possible, emphasis is placed on trying to understand the variation of the properties of LIGs as a function of infrared luminosity.

4.1 Optical and Near-Infrared Imaging

Early optical imaging studies of IRAS galaxies are largely split into two groups: morphological classifications of relatively bright infrared sources that had previously been cataloged in optical surveys and observations of the most luminous infrared sources or sources selected for their extreme infrared properties (typically high infrared-to-blue or extreme color temperature). The former captured, almost exclusively, relatively nearby objects with L_ir < 10¹¹ L and can be summarized by the results from Rieke & Lebofsky (1986), who found that nearly all E's and most S0s have L_ir < 10⁹ L, with most spirals having higher infrared luminosities; for L_ir = 10¹⁰-10¹¹ L, nearly all of the sources were Sb or Sc galaxies. It also appeared that ~ 12-25% of LIGs were peculiar or interacting systems (e.g. Soifer et al. 1984a). A much higher proportion of interacting and disturbed systems was reported from samples selected on the basis of extreme infrared properties. The most luminous sources (L_ir 3 x 10¹² L) in the IRAS database (Kleinmann & Keel 1987; Sanders et al. 1987b, 1988b; Hutchings & Neff 1987; Vader & Simon 1987a) were universally classified as strong interactions/mergers. Sources with high L_ir / L_B ratios (e.g. van den Broek 1990; Klaas & Elsässer 1991, 1993), or warm colors, either f₆₀ / f₁₀₀ or f₂₅ / f₆₀ colors (e.g. Armus et al. 1987, 1990; Sanders et al. 1988b; Heisler & Vader 1994), were predominantly ( 70%) strongly interacting/peculiar systems, with the remaining objects often being amorphous or elliptical-like in appearance.

More recent data for the complete BGS now shows that the fraction of objects in that are interacting/merger systems appears to increase systematically with increasing infrared luminosity. Images of objects in the BGS (Sanders et al. 1988a, 1996a; Melnick & Mirabel 1990) show that the fraction of strongly interacting/merger systems increases from ~ 10% at L_ir / L = 10.5-11 to ~ 100% at L_ir / L > 12. Figure 4 shows images of the complete sample of 10 ULIGs in the original BGS (Sanders et al. 1988a). Other studies of ULIGs have generally reached a similar conclusion that 95% are merger systems (Kim 1995, Murphy et al. 1996, Clements et al. 1996; although see Lawrence et al. 1989, Leech et al. 1994).

Figure 4. Optical (r-band) CCD images of the complete sample of ten ULIGs from the original BGS (Sanders et al. 1988a). Tick marks are at 20" intervals.

The improved angular resolution and lower optical depth in the near-infrared as compared to the optical has proved to be particularly useful for disentangling nuclear morphology not seen in the optical data (e.g. Carico et al. 1990, Graham et al. 1990, Eales et al. 1990). The mean and range of projected nuclear separations for ULIGs in the BGS are ~ 2 kpc and < 0.3 to 10 kpc respectively (Sanders 1992). More recent K-band imaging of larger samples of ULIGs (Murphy et al. 1996, Kim 1995) generally confirms these results, although a few systems appear to have nuclear separations as large as 20-40 kpc.

Despite their extreme infrared luminosities, photometry of ULIGs confirms that most are only moderately luminous in the optical and near-infrared. For the 10 ULIGs in the original BGS the median blue absolute magnitude is bar M_B = -20.7 (Jensen et al. 1996) [compared to the mean value < M_B > = -20.2 reported by Armus et al. (1990) for their more distant sample of ``Arp 220-like'' objects], bar M_r ~ -21.6 (Murphy et al. 1996; corrected by +0.44 mag by J Surace, private communication), and bar M_K' = -25.2 (Jensen et al. 1996). Compared to an L* galaxy² the median total luminosities for ULIGs are ~ 2.5 L_B*, ~ 2.7 L_r*, and ~ 2.5 L_K'*, where the range around the median (excluding Seyfert 1 objects) is -0.9 to +1.6 mag in M_r and -1.0 to +2.2 mag in M_K'. Most ULIGs contain compact nuclei, with typically one quarter of the total K'-band luminosity originating in the inner 1" radius, except for the few Seyfert 1 galaxies (e.g. Mrk 231) where the pointlike nuclear source can be as much as a factor of ~ 5 times stronger than the surrounding galaxy. The typical host galaxies of ULIGs, therefore, appear to be ~ 2L* at K'.

4.2 Optical and Near-Infrared Spectroscopy

Although extensive optical redshift surveys have been carried out to identify IRAS galaxies, much of these data either have spectral resolution that is too low or are too limited in wavelength coverage to be of use for anything more than simple redshift determinations. Higher-resolution observations (typically = 3-5 Å for ~ 3800-8000 Å) are now available for most of the IRAS galaxies in the imaging studies discussed above. Elston et al. (1985) classified the majority of IRAS minisurvey objects as ``starbursts taking place in dusty galaxies''. On the other hand, ~ 50% of the objects with ``warm'' colors, f₂₅ / f₆₀, from the sample of de Grijp et al. (1985) were classified as Seyferts, with an additional ~ 20% classified as LINERS (Osterbrock & De Robertis 1985). Observations of ``warm'' objects at fainter flux levels in the IRAS database produced the first two infrared selected QSOs (Beichman et al. 1986, Vader & Simon 1987a) plus several of the most intrinsically luminous infrared sources currently known, all of which are classified as Seyfert 2 in direct emission (e.g. Kleinman & Keel 1987, Hill et al. 1987, Frogel et al. 1989, Cutri et al. 1994), but have been shown to contain hidden broad-line regions in polarized light (Hines 1991, Hines & Wills 1993, Hines et al. 1995). Vader et al. (1993) also report that ~ 60% of their ``warm'' sample of ``60 µm Peakers'' are Seyferts.

Heckman et al. (1987) and Armus et al. (1989, 1990) have used long-slit spectroscopy (4500-8000 Å) and narrow-band (H + [NII]) imaging to show that Arp 220-like ``tepid'' LIGs appear to contain huge ( 10 kpc), powerful emission-line nebulae often characterized by spectacular loops and bubbles, which they interpret as a starburst-driven superwind (see also Section 6.4). The visible spectrum of these objects appears to be dominated by young stars, with ~ 20% showing evidence for a substantial intermediate-age population (few x 10<sup>8</sup> years) and another ~ 20% showing strong Wolf-Rayet lines, indicating a very young starburst ( 10<sup>7</sup> years). The total H+N[II] luminosity (corrected for extinction) is typically a factor of ~ 30 larger than that for isolated spiral galaxies (Kennicutt & Kent 1983). As a class, these objects are similar to LINERs; about half are intermediate between LINERs and low-excitation H II regions. Steep Balmer decrements imply that they are being viewed through substantial amounts of obscuring dust. These data have been used to suggest that the ``Arp 220-phase'' may be characterized by an ongoing powerful circumnuclear starburst that may be rapidly clearing out the obscuring gas and dust in the inner few kiloparsecs of these objects.

More recently, the fraction of LIGs of different spectral type has been determined from an analysis of long-slit spectroscopy of complete samples of objects using several diagnostic emission-line ratios (e.g. Veilleux & Osterbrock 1987). Figure 5 shows the results from an analysis of a complete subsample of objects in the BGS (Kim et al. 1995, Veilleux et al. 1995), supplemented by a larger sample of ULIGs (Kim et al. 1996). The percentage of Seyfert galaxies increases systematically from ~ 4% at L_ir / L = 10-11, to ~ 45% at L_ir / L > 12.3, whereas the percentage of LINERs remains relatively constant (~ 33%) at all infrared luminosities L_ir / L > 10.

Figure 5. The optical spectral classification of infrared galaxies versus infrared luminosity (Kim 1995).

The amount of published near-infrared spectroscopy for IRAS galaxies is small compared to the available optical data. Rieke et al. (1985) presented a detailed analysis of relatively low-dispersion infrared data at K and L bands for NGC 6240 and Arp 220, claiming that the former could be powered entirely by a very luminous starburst whereas as much as half of the luminosity in Arp 220 appeared to be due to a Seyfert nucleus. Most of the early low-dispersion data for LIGs have been superseded by spectra covering the K-band window at resolving powers / ~ 300-800 using infrared CCD arrays. Goldader (1995) and Goldader et al. (1995) find that LIGs with Seyfert-like optical classifications also show evidence for dominant non thermal emission in the K-band, and most if not all ULIGs with Seyfert 2 optical spectra show evidence for Seyfert 1 linewidths in Pa (Goodrich et al. 1994) or Pa (Veilleux et al. 1996); however, for ``cool'' LIGs, no new broad-line regions are discovered in the near-infrared that were not already seen at optical wavelengths. Perhaps the most intriguing new result is the systematic low value of the Br / H₂S(1) line ratio in ULIGs as compared with lower luminosity objects, a result that suggests that the dominant luminosity source in ULIGs is still highly obscured even at near-infrared wavelengths (Goldader et al. 1995). A more detailed analysis of Arp 220 using several infrared lines (Armus et al. 1995a, b) also suggests that as much as 80-90% of the total luminosity could be powered by an obscured AGN.

For the few hyperluminous infrared galaxies (HyLIGs: L_ir > 10¹³ L) that have been discovered in the IRAS database, all are at z 1, so that near-infrared spectra provide the only means for observing several of the most prominent diagnostic lines (e.g. H and H). All of these objects have rest-frame optical emission-line ratios characteristic of Seyferts [e.g. IRAS 15307+3252 (Soifer et al. 1995, Evans et al. 1996a); IRAS 10214+4724 (Soifer et al. 1992, 1995; Elston et al. 1994)].

4.3 Mid- and Far-Infrared (Post-IRAS) Observations

Ground-based observations in the mid-infrared, and far-infrared measurements with the Kuiper Airborne Observatory (KAO) have been carried out in an attempt to set meaningful constraints on the size of the infrared emitting region in a few LIGs. Becklin & Wynn-Williams (1987) reported that the 20-µm size of Arp 220 was smaller than 1.5" (500 pc), and they estimated a visual extinction of at least 50 mag based on the depth of a deep silicate absorption feature at 10 µm. Dudley & Wynn-Williams (1996) use the depth of the silicate absorption feature to estimate 10-µm sizes of only a few parsecs for Arp 220 and the warm ULIG IRAS 08572+3915. Matthews et al. (1987) used slit scans to show that the 10-µm source in Mrk 231 was smaller than 1" (800 pc). Miles et al. (1996) have obtainet 10-µm maps for 10 LIGs and find that a large fraction (~ 65-100%) of the 10-µm emission in ULIGs and warm LIGs originates in an unresolved ( 0.6") core.

Observations with the KAO using drift scans at 50-100 µm (Joy et al. 1986, 1989; Lester et al. 1987) have shown that the emission regions in a few sources (e.g. Arp 220, Arp 299, NGC 1068) contain dominant compact ( 10") components at these wavelengths; however, larger telescopes or interferometers are clearly needed before more meaningful constraints can be set on the size of the far-infrared sources responsible for the bulk of the far-infrared luminosity in LIGs.

4.4 Submillimeter Continuum

Ground-based measurements in the submillimeter continuum (~ 350-860 µm) have been obtained for a few of the brightest LIGs. Emerson et al. (1984) found that the far-infrared/submillimeter emission from Arp 220 could be fit by a single temperature dust model with T_dust ~ 60 K, and they derived an optical depth of ~ 1 at 180 µm (!) for an assumed source size of 4". More recently, Rigopoulu et al. (1996a) have interpreted their submillimeter measurements for all ULIGs in the BGS as being consistent with thermal dust emission.

Submillimeter observations of LIGs have also been reported by Eales et al. (1989) and Clements et al. (1993), with the general result that the far infrared/submillimeter continuum in all of the objects can be reasonably fit by a single temperature dust model (assuming a ^-2 emissivity law) with dust temperatures of 30-50 K. There is no obvious evidence for large amounts of cooler dust, although large quantities of sufficiently cool dust (i.e. T_dust 20 K) cannot be ruled out (e.g. Devereux & Young 1991).

4.5 Radio Continuum

A ``tight correlation'' between the flux in the infrared and the radio continuum has been found in several studies of normal, starburst, and Seyfert galaxies (van der Kruit 1971, Rieke & Low 1972, Dickey & Salpeter 1984, Helou et al. 1985, Sanders & Mirabel 1985, Wunderlich et al. 1987). Figure 6a shows that the logarithmic ratio of far-infrared and radio continuum flux densities, q = log {[F_fir / (3.75 x 10¹² Hz)]/[f(1.49 GHz)]}, is relatively constant, q ~ 2.35, for most LIGs in the BGS, with a rather small dispersion at a given L_ir ( ~ 0.2). This relationship appears to hold for sources covering several orders of magnitude in L_ir, from quiescent disk-like spirals like M31 to the powerful nuclear sources in ULIGs like Arp 220, although there is evidence that the mean changes slightly at both low and high infrared luminosities. At lower infrared luminosities, < q > appears to increase slightly as galaxies transit from heating dominated by young stars to heating by an old stellar population (Condon et al. 1991a, Xu et al. 1994), whereas at the highest infrared luminosities < q > increases, apparently due mainly to optical depth effects in the radio (Condon et al. 1991b). More recently, Bicay et al. (1995) have shown that < q > for Markarian starbursts is enhanced by a factor of ~ 3 relative to Markarian Seyferts, but the reason for this increase is not immediately clear.

Figure 6. (a) q versus Lir. Solid and open circles represent LIGs and lower luminosity BGS galaxies respectively. Open triangles refer to the nearby starburst galaxies M82 and NGC 253, and open diamonds represent the Milky Way and three ``normal'' nearby spiral galaxies (IC 342, NGC 6946, NGC 891). Asterisks refer to optically selected QSOs that have been detected in CO (I Zw1, Mrk 1014, 3C 48), and ``+'' signs represent PRGs detected by IRAS. (b) The ratio of molecular (H2) to atomic (H I) gas versus the infrared-to-blue luminosity ratio in IRAS BGS galaxies (Mirabel & Sanders 1989). The arrows are due to lower and upper limits in the measurement of the H I fluxes. Most galaxies with Lir / LB 20 exhibit H I absorption and OH megamaser emission with velocity extents of several hundred km s-1. (c) Lir / L'CO versus Lir. The dashed lines represent mean values for nearby ``normal'' spirals (Lir/L'CO ~ 18), nearby starburst galaxies (Lir/L'CO ~ 50), and the most extreme star-forming GMC cores in the Milky Way (Lir/L'CO ~ 100). (d) Correlation of the central concentration of molecular gas with the Lir/L'CO ratio for LIGs in the IRAS BGS (Scoville et al. 1991, Bryant 1996). Small black and larger grey circles represent objects where the spatial resolution was sufficient to resolve circumnuclear regions of < 1 kpc and 1-2 kpc diameter respectively.

Most radio galaxies and radio-loud QSOs have q-values much lower than 2.35 (typically by factors of ~ 2-4 in the log), usually due to the presence of strong, very-compact radio cores combined with extended radio lobes/jets that are apparently decoupled from the infrared emission. However, evidence exists that radio galaxies still show a correlation in their far-infrared and radio fluxes, but with a value < q > ~ -0.65 (e.g. Golombek et al. 1988, Knapp et al. 1990, Impey & Gregorini 1993).

The ``radio-infrared correlation'' has been used in an attempt to overcome the poor angular resolution of far-infrared instruments. High-resolution radio surveys of LIGs in the BGS were made at 1.49 GHz (Condon et al. 1990) and at 8.44 GHz (Condon et al. 1991b). The VLA maps show that nearly all galaxies with L_fir 10¹¹ L are dominated by extended, diffuse radio emission, whereas most ULIGs are dominated by compact, sub-arcsec radio sources. Condon et al. (1991b) concluded that most LIGs in the BGS - with the exception of the Seyfert 1 galaxy Mrk 231, which is dominated by a variable ultracompact radio source ( 1 pc) - can be modeled by ultraluminous nuclear starbursts (see also Crawford et al. 1996). These starburst regions would be so dense that they are optically thick even to free-free absorption at = 1.49 GHz and to dust absorption at 25 µm ! If this is true, then X rays, infrared, and radio waves may not be able to probe the dense cores of ULIGs; the far-infrared luminosity is then at best a good calorimeter. Due to Compton scattering in such dense clouds, even a compact source of hard X rays will be hidden from the observer !

As a further probe of the size of the radio sources in LIGs, Lonsdale et al. (1993) carried out a sensitive VLBI survey of 31 objects and found that typically ~ 12% of the radio flux arises in cores only 5-150 milliarcsec in size (which rules out a single supernova interpretation of the compact radio cores). These compact VLBI cores are comparable in power to the total radio power of typical Seyfert galaxies (Ulvestad & Wilson 1989) and radio-quiet QSOs (Kellermann et al. 1989).

4.6 Gas Content

Single-dish observations of millimeter-wave emission from the rotational transitions of CO (_FWHM ~ 20-60") and the 21-cm line of H I (_FWHM ~ 3-10') now exist for the majority of objects in the BGS. These data have shown that the total neutral gas content, in particular the total mass of molecular gas, appears to play a critical role in the genesis of LIGs. More recently, interferometer maps of a few dozen LIGs and lower luminosity infrared-selected objects have provided dramatic pictures of the redistribution of the H I and H₂ gas that occurs during interactions and mergers.

4.6.1 ATOMIC GAS (H I)   The first observations of LIGs at centimeter wavelengths revealed somewhat perplexing properties. At = 21 cm, the most luminous infrared galaxies showed very broad H I absorption lines, indicating rotation plus large amounts of unusually turbulent neutral gas (Mirabel 1982). The H I profiles typically show absorption features with widths of a few hundred to up to 1000 km s^-1, with total column densities 10^21-22 atoms cm^-2. VLA observations of the H I absorption had suggested that most of the absorbing H I is located in the inner few hundred parsecs of these galaxies (e.g. Baan et al. 1987), in front of the nuclear radio continuum sources. The total masses of H I in a complete sample of galaxies with L_fir 2 x 10¹⁰ L are in the range of 5 x 10⁸ to 3 x 10¹⁰ M, with only a weak correlation between M(H I) and L_fir (Mirabel & Sanders 1988).

Mirabel & Sanders (1988) carried out a statistical analysis of the difference between the radial velocities of the H I 21-cm line absorptions and the optical redshifts. Although the discrepancy between the radio and optical velocities is usually smaller than the velocity width of the H I absorptions, there is a clear trend for the radio redshifts to be greater than the optical redshifts. Among 18 galaxies with H I absorption, 15 were found with V_{H Iabs} > V_opt and only 3 with V_{H Iabs} V_opt. The mean value of V_{H Iabs} - V_opt is 90 km s^-1. From VLA observations Dickey (1986) found a similar trend, from which he estimated an accretion rate of ~ 1 M year^-1 into the nuclear regions.

The optical redshifts may be affected by systematic errors and there are some caveats to the interpretation of the statistical discrepancy between optical and radio velocities as due entirely to infall of H I. The optical redshifts are often determined from emission lines in low dispersion spectra, which due to large scale superwinds (see Section 6.4) are usually asymmetric with extended blue wings. In this context, some of the statistical discrepancy between the radio and optical redshifts could be due to optical line-emitting gas mixed with dust that is moving radially, probably outward. However, from a careful analysis of the available data, Martin et al. (1991) concluded that most of the H I seen in absorption in is indeed infalling toward the central source.

4.6.2 MOLECULAR GAS (H₂)   Substantial information on the total molecular gas content of LIGs has been obtained from single-dish observations of millimeter-wave CO emission for large samples of IRAS galaxies. An important discovery has been that all LIGs appear to be extremely rich in molecular gas. Early CO observations of infrared selected galaxies (most with L_ir = 10¹⁰-10¹¹ L found a rough correlation between CO and far-infrared luminosity (Young et al. 1984, 1986a, b; Sanders & Mirabel 1985). Assuming a constant conversion factor between CO luminosity and H₂ mass, M(H₂) / L'_CO = 4.6 [M (K km s^-1 pc²)^-1] (e.g. Scoville & Sanders 1987), total H₂ masses were in the range ~ 1-30 x 10⁹ M, or approximately 0.7 to 20 times the molecular gas content of the Milky Way. Mirabel & Sanders (1989) found that the ratio of total H₂ to H I mass is typically > 1 with some evidence that M_T(H₂) / M_T(H I) increases with increasing L_ir / L_B (Figure 6b). Multitransition CO measurements (e.g. Sanders et al. 1990, Radford et al. 1991, Braine et al. 1993, Devereux et al. 1994, Rigopoulu et al. 1996b) and detection of strong emission from dense gas tracers such as HCN (Solomon et al. 1992a), CS, and HCO⁺ (see reviews by Mauersberger & Henkel 1993, Radford 1994, Gao 1996) indicate that the mean molecular gas temperatures and densities in the central regions of LIGs are hot, T_kin = 60-90 K, and dense, n(H₂) ~ 10⁵-10⁷ cm^-3, similar to the conditions in massive Galactic giant molecular cloud (GMC) cores.

Significant improvements in detector performance in the late 1980s resulted in a dramatic increase in the number of extragalactic infrared sources detected in CO. The first objects detected at z 0.03 were ULIGs from the BGS; these included the first detection of CO emission from a Seyfert 1 galaxy (Mrk 231: Sanders et al. 1987b) and two of the most intrinsically luminous CO sources currently known (VII Zw 31: Sage & Solomon 1987; IRAS 14348-1447: Sanders et al. 1988d). Detections of objects at z > 0.1 soon followed, including the first CO detections of UV-excess QSOs (Mrk 1014: Sanders et al. 1988c; I Zw 1: Barvainis et al. 1989), an infrared selected QSO (IRAS 07598+6508: Sanders et al. 1989b), a PRG (4C 12.50: Mirabel et al. 1989), and a radio-loud QSO (3C 48: Scoville et al. 1993).

Figure 6c includes data from several single-dish CO(1->0) surveys of IRAS galaxies (Sanders et al. 1991, Mirabel et al. 1990, Tinney et al. 1990, Downes et al. 1993, Mazzarella et al. 1993, Young et al. 1995, Elfhag et al. 1996, Solomon et al. 1996, Evans 1996) illustrating both the general trend of increasing L_ir/L'_CO ratio with increasing L_ir and the fact that this ratio can vary by nearly a factor of 30 at a given L_ir. Assuming M(H₂) = 4.6 L_CO, Figure 6c shows that the total molecular gas mass in ULIGs - typically M(H₂) 10¹⁰ M - is on average more infrared luminous than any of the most active star-forming Galactic GMC cores [which have diameters typically of ~ 2-5 pc and M(H₂) = 10³-10⁴ M].

Millimeter-wave interferometer measurements of CO emission have been obtained for approximately two dozen LIGs (e.g. Scoville et al. 1991; Okumura et al. 1991; Bryant 1996; Yun & Hibbard, private communication). Figure 6d shows that for these objects, nearly all of which are advanced mergers, ~ 40-100% of the total CO luminosity, or M(H₂) = 1-3 x 10¹⁰ M assuming the standard Milky Way conversion factor, is contained within the central r < 0.5-1 kpc. The mean surface density in the central 1 kpc regions of ULIGs is typically in the range < (H₂) > = 1.5-7 x 10⁴ M pc^-2, although the H₂ masses in these extreme regions may be overestimated by a factor of 2-3 (Downes et al. 1993, Solomon et al. 1996). Even allowing for such a decrease in the conversion factor, these values are ~ 50-250 times larger than the mean gas surface density in the central 1 kpc of the Milky Way and would seem to imply enormous optical depths (A_V ~ 200-1000 mag) along an average line of sight toward the nucleus of these objects. Such high values are consistent with the implied high optical depths in the nuclear regions of ``cool'' ULIGs at K-band (see Section 4.2) and optical depths near unity at ~ 200 µm implied by the far-infrared/submillimeter measurements for Arp 220 (see Section 4.4).

4.6.3 OH MEGAMASERS The high-density molecular gas in the central regions of LIGs is the site of the most luminous cosmic maser sources known. The amplified main OH lines at 1667 and 1665 MHz correspond to transitions between the hyperfine splitting of a -doubling level at 18 cm. The isotropic luminosities in the OH lines can be as strong as ~ 10⁴ L, almost a million times that of the most luminous OH maser sources in the Galaxy, hence the name megamaser. Since the first detection of OH maser emission from the ULIG Arp 220 (Baan et al. 1982), it became evident that this type of emission could be detected out to redshifts of z ~ 0.5, and therefore, that it could be used to probe the circumnuclear high-density interstellar gas in distant infrared galaxies (Baan 1985, Stavely-Smith et al. 1987, Martin et al. 1991, Kazès & Baan 1991). About 50 OH megamasers have now been identified (Baan 1993). The OH megamaser spectra usually exhibit broad linewidths and extended velocity wings that could be due to the rapid rotation of circumnuclear molecular disks (e.g. Montgomery & Cohen 1992), large-scale outflow motions (Mirabel & Sanders 1987, Baan 1989), and/or distinct components arising in the interactive system (e.g. Baan et al. 1992).

The isotropic OH 1667 MHz luminosity is proportional to (L_ir)² (Martin et al. 1988, Baan 1989). This has been interpreted by Baan & Haschick (1984) as low-gain amplification of the nuclear continuum radio source by intervening OH that is being pumped by far-infrared radiation from dust. Inversion of the OH population requires a source of steep-spectrum emission in the far-infrared such as that expected from warm dust emission. Usually an OH megamaser galaxy requires a strong nuclear radio continuum source, a column density of gas along the line of sight > 10²² cm^-2, a ratio f₆₀ / f₁₀₀ 0.7, and L_ir > 10¹¹ L (Mirabel & Sanders 1987).

The location of megamasers in the L_ir versus M(H₂) plane shows that they occur in objects with the largest L_fir / M(H₂) ratios (Mirabel & Sanders 1989). At present we consider it an open question whether the origin of the far-infrared luminosity in megamasers is entirely due to star formation. A considerable fraction of the energy could ultimately come from a compact nonthermal source at the dynamical center of these galaxies. Future high-resolution VLBI observations of the OH emission may be a way to answer this question.

Figure 7. OH megamaser emission, H I 21-cm line emission, and CO(1->) line emission from the high Lir/LB object III Zw 35 (Mirabel & Sanders 1987).

There is increasing evidence that OH megamasers can eventually be used to successfully probe the molecular gas kinematics and the dynamic masses at the very centers of starburst galaxies and AGNs. Furthermore, given the observed L_OH L²_ir relationship in megamaser galaxies, the spatial distribution of the far-infrared emitting region can be inferred with unprecedented angular resolution. Single-dish observations have in the past been interpreted as showing evidence for spatial extents of a few hundred parsecs in the OH emission from nearby systems (Baan 1993). However, MERLIN observations of III Zw 35 have shown that most of the OH emission arises from a region only 100 pc x 60 pc in size, which is more compact than the radio continuum source (Montgomery & Cohen 1992). The velocity of the OH emission appears to trace out the characteristic signature of a rotating disk, which would imply a dynamical mass of ~ 10⁹ M inside a region of ~ 15 pc in radius.

It is interesting that one of the most compact megamasers mapped so far with MERLIN, III Zw 35, is also among those megamasers with the largest implied infrared pumping efficiencies (Mirabel & Sanders 1987). Figure 7 shows OH emission from III Zw 35 up to 450 km s^-1 below the systemic velocity, which raises the interesting possibility of extreme circular velocities in a compact molecular torus. Similar high velocity wings, perhaps due to rapidly rotating disks, have been observed in Mrk 231 and Mrk 273 (Stavely-Smith et al. 1987). In addition, recent VLBI observations of Arp 220 strongly suggest that the OH peak emission originates in a structure 1 pc across and that nearly all of the OH emission comes from a region 10 pc in size (Lonsdale et al. 1994). This would imply that the maser is physically 10-100 times smaller than previously thought (Baan 1993) and that much of the far-infrared emission from Arp 220 arises in a very small region, possibly a torus surrounding a quasar nucleus. These new results suggest that OH megamasers may eventually be used to probe the central engines in AGNs, much as the H₂O megamasers have been used in NGC 4258 (Miyoshi et al. 1995).

4.7 High-Energy Observations

A strong correlation has been found between the X-ray luminosity in the Einstein 0.2-4.5 keV energy band and the far-infrared and blue luminosities for 51 galaxies in the BGS (David et al. 1992). The best fit for the X-ray luminosity, L_X = 9.9 x 10^-5 L_B + 9.3 x 10^-5 L_fir, has two components, one proportional to L_B due to the old population (Type I supernovae and low mass X-ray binaries) and a second component proportional to L_fir due to young objects (Type II supernovae, O stars, high-mass binaries). LIGs with large L_ir / L_B ratios are dominated by the second component and can have a significant excess of X-ray flux relative to the mean (up to a factor of 10 in the case of Arp 220). It has been proposed that LIGs could make a significant contribution (up to 40%) to the X-ray background (Griffiths & Padovani 1990).

The cross-correlation between the IRAS PSC and the ROSAT All-Sky Survey (0.1-2.4 keV) allowed the identification of 244 IRAS galaxies that appeared to be positionally coincident with ROSAT X-ray sources (Boller et al. 1992). This sample is dominated by galaxies with active nuclei. An unexpected result was the discovery of a dozen spiral galaxies with X-ray luminosities up to 10⁴³ erg s^-1, well above those found by the Einstein satellite. These objects have steep spectra with typical photon indexes of 2.5-3.6, which would make them preferentially detected by ROSAT because of its sensitivity in the soft X rays. Optical spectroscopy revealed active nuclei in half of these objects (Boller et al. 1993). In the galaxy IRAS 15564+6359 a compact component can be disentangled from that of any extended component, since in the 0.1-2.4 keV energy range flux variations with a doubling time of 1500 sec (corresponding to a source size of 2 x 10^-5 pc) have been discovered (Boller et al. 1994). The non-detection of optical AGN activity in this galaxy is intriguing.

ROSAT observations of Arp 220 reveal extended emission with T_kin ~ 10⁷ K (Heckman et al. 1996), presumably due to gas ejected from the active nuclear region (see also Section 6.4). The morphology is similar to, but much more powerful than, the extended X-ray components observed in the classic starburst galaxies NGC 253 and M82 (Fabbiano 1988) and the edge-on spiral NGC 3628 (Fabbiano et al. 1990). ROSAT observations of NGC 4038/39 (``The Antennae'') have shown that about 55% of the X-ray emission can be identified with the two galaxy nuclei plus a large ring of H II regions; the remainder appears to be diffuse emission from gas at ~ 4 x 10⁶ K that extends up to 30 kpc away from the merging disks (Read et al. 1995).

To test whether it is possible to use X rays to search for obscured AGN in ULIGs, as suggested by Rieke (1988), it is instructive to consider the case of NGC 1068. Spectropolarimetry of NGC 1068 by Antonucci & Miller (1985) revealed a Seyfert 1 nucleus hidden in a dense torus of dust and gas. Millimeter-wave observations of HCN imply ~ 1.6 x 10⁸ M of molecular gas within a region 34 pc in radius (Tacconi et al. 1994). Assuming a spherical distribution implies a column density 5 x 10²⁴ atoms cm^-2, which - given the expected high metalicity - would produce an almost total absorption of X rays by Compton scattering. In fact, ASCA observations revealed that the direct component in NGC 1068 is totally absorbed even at energies of 20 keV and that the observed spectrum below 20 keV is light scattered from the AGN by electrons over the absorption torus (Ueno et al. 1994). Because the central gas densities in ULIGs are likely to be even greater than in NGC 1068, hard X rays most likely cannot be used to probe the central engines in these galaxies.

Gamma rays from extragalactic systems have so far mostly been detected from AGNs, either as quasi-isotropic emission from unobscured Seyferts or as beamed emission from highly variable Blazars. As expected, NGC 1068 was not detected at > 20 keV (Dermer & Gehrels 1995). Observations of the nearby lower-luminosity starburst galaxies NGC 253 and M82, over the energy range 0.05-10 MeV, yelded a 4 detection of NGC 253 in continuum emission up to 165 keV, with an estimated luminosity of ~ 10⁴⁰ erg s^-1. No significant flux at high energies was detected from M82 (Bhattacharya et al. 1994).