10. INFRARED H₂ EMISSION, TRACER OF WARM MOLECULAR GAS, SHOCKS AND PHOTODISSOCIATION REGIONS

10.1. Basic features and physics of H₂ emission, and Milky Way observations

In normal conditions, H₂, the overwhelming compound of the molecular interstellar medium, is not easy to directly detect (Shull & Beckwith 1982, Combes & Pineau des Forets 2000). UV absorption (see Section 7.1) is inefficient to probe deeper than the outskirts of molecular clouds with A_v 1 (however, see Keel 2006). Rovibration infrared lines are limited to weak quadrupole lines. They include: i) vibration lines with v = 1,2... - S(J_l), Q(J_l), O(J_l) with J = +2, 0 and -2, respectively - with wavelengths in the good 2 µm atmospheric window for the most important lines with v = 1; ii) pure rotation lines S(J_l) (J = 2, which do not connect the rotation levels corresponding to different nuclear spin configuration (ortho-H₂ with I = 1 and J odd and para-H₂ with I=0 and J even) which span the whole mid-IR range: S(0) 28.22 µm, S(1) 17.03 µm, S(2) 12.28 µm, S(3) 9.66 µm, S(4) 8.03 µm, S(5) 6.91 µm, S(6) 6.11 µm, S(7) 5.51 µm, etc. All these lines have high excitation energy, from 510 K for S(0) to 7200 K for S(7) pure rotation lines, and 5600 K and 10000 K for v = 1-0 and v = 2-1 lines respectively.

Such IR lines are practically useless in absorption (however, see Lacy et al. 1994, Usuda & Goto 2005) because of the extremely large values for the column density N_H2 they require to be detectable: typical values of N_H2 needed to achieve an optical depth = 0.01 are a few 10²³ cm^-3 (A_v ~ a few 10²) for v = 1-0 2 µm ro-vibration lines and a few 10²⁴ cm^-3 (A_v ~ a few 10³) for pure rotation lines (see e.g. Shibai et al. 2001).

On the other hand, despite the high excitation energies, near-IR and mid-IR emission lines of H₂ have proved to be widespread and important diagnosis tools in all molecular media harbouring energetic processes, related to massive star formation, shocks or AGN, able to achieve upper level excitation (see e.g. the introduction of Roussel et al. 2007 for a review of previous references on extragalactic H₂ infrared lines). The physics of the different excitation processes is various; but the most important ones are well understood for a molecule as simple as H₂. This allows very precise modelling when the astrophysical conditions are well defined. Thermal excitation is generally dominating for the pure-rotational lines (Section 10.3). An excitation temperature ~ 200 K is enough to ensure a substantial excitation of the S(0) 28.22µm line, ~ 300-400 K for S(1), but ~ 1400 K for an upper line such as S(7) (see e.g. Rosenthal et al. 2000, Higdon et al. 2006a). Pure rotational lines of H₂ have been detected in various Galactic sources with ISO (see e.g. the Galactic Center by Lutz et al. 1996, Orion by Rosenthal et al. 2000) and Spitzer.

The case of the higher energy required to excite vibration lines is more complex. Following the first detections of such lines thirty years ago, their excitation has been the object of elaborated modelling which allows a precise use of such lines for diagnostic of the conditions in the emitting medium. The two main excitation routes are UV fluorescence and collisions in the hot gas of shocks, although it has been proposed that rovibrational excitation in the process of H₂ formation may play a significant role, in particular in recombination on the grain surfaces (see e.g. references quoted in Dalgarno 2001). In many cases, the properties of the H₂ near-IR emission are compatible with thermal collisional excitation in a hot gas (~ 2000 K) such as found in strong shocks like prominent Galactic sources (see e.g. in the close vicinity of the Galactic Center, Gatley et al. 1984) or Orion, Beckwith 1981). In such shocked regions, the H₂ rovibration populations in the ground electronic state are usually well thermalised by collisions (Shull & Beckwith 1982; Black & van Dishoeck 1987 and references therein). A direct indication of dominent collisional shock excitation is a very small value for the ratio of similar v = 2-1 and v = 1-0 lines (~ 0.13 at 2000 K).

Alternatively, in the vicinity of strong radiation sources, such as photodissociation regions, ultraviolet pumping may dominate the vibrational excitation of H₂. There is then no energy limitation for populating the higher v levels. The accurate knowledge of the various ultraviolet and infrared transition probabilities allows a precise prediction of the resulting intensities of the emission infrared lines. As discussed e.g. by Dalgarno (2001) and references therein, particularly useful is the ratio of the S(1)(2-1) and S(1)(1-0) intensities. It is found to be close to the predicted value, 0.54, in various Galactic reflection nebulae including the best studied one NGC 2023. UV excitation by photons from OB stars is also believed to dominate the excitation of the wide H₂ emission observed in the inner 400 pc region of the Galaxy, where the ratio of the H₂ to far-IR luminosity agrees with that in starburst galaxies (Pak et al. 1996).

The multiplicity of the rovibrational lines may span a broad range of excitation energy (e.g. from 500 K to 17000 K in Orion, Rosenthal et al. 2000). The accuracy of the models then allows various other diagnostics, in particular of their selective extinction. For high gas densities, collisions and fluorescence may compete, and the relative line intensities can be used to infer densities and radiation fields (Sternberg & Dalgarno 1989). In addition, other heating mechanisms than shocks may be important, such as X-ray illumination where hard X-ray photons are capable of penetrating deeply into molecular clouds and heating large amounts of gas (e.g. Maloney et al. 1996). Even heating by strong UV in photodissociation regions may be the main excitation mechanism of the first pure-rotational lines.

10.2. 2 µm H₂ emission in galaxies

The studies of the H₂ near-infrared lines in galaxies were initiated in the late 1970s by Gautier et al. (1976), at the begining of infrared spectroscopy. Since this time they have followed all the progress of the telescopes and instrumentation of near-infrared astronomy. This was already a mature field in the 1980s and 1990s. It remains at the forefront of major developments in progress such as adaptive optics studies, JWST and extremely large telescopes.

As it could be inferred from the results in the Galaxy and especially in the Galactic Center region, detectable near-IR emission from external galaxies comes from large concentrations of hot molecular gas found in extended regions of strong photodissociation or shocks, especially in the central regions hosting AGN or major starbursts.

Various H₂ rovibration transitions were observed in a number of objects in the Magellanic Clouds (see Israel & Koorneef 1991, Pak et al. 1998, and references therein). Consistent conclusions are that UV radiative excitation is the energetically dominant mechanism.

The 2 µm lines of H₂ are relatively easy to detect in the central regions of nearby galaxies especially when they present some AGN or strong starburst. Many, more or less detailed, spectroscopic studies have thus been carried out. However, the analysis of the early results was often not obvious and controversial (see e.g. the reviews by Mouri 1994; Goldader et al. 1997). Indeed, more comprehensive results very often reveal a mixture of various excitation mechanisms from UV, shocks and even AGN X-rays, at various spatial scales. The most fruitful observations have best succeeded in disentangling such a complexity by high angular resolution and velocity-resolved spectroscopy.

The most widespread extragalactic H₂ emission is related to starbursts. In the central regions of starbursts galaxies, such as the archetypes M 82 or NGC 253, the great majority of the 2 µm line emission arises from energy states excited by ultraviolet fluorescence (Pak et al. 2004). It is also found that the ratio of the H₂ v = 1-0 S(1) line to FIR continuum luminosity is constant over a broad range of galaxy luminosities, as well as in normal late-type galaxies (including the Galactic center) as in nearby starburst galaxies, and especially in LIRGs (Goldader et al. 1997, Pak et al. 2004). This is consistent with a common origin of FIR and H₂ emission from the UV radiation from photodissociation regions (PDRs) illuminated by recently formed OB stars. The spatial distribution of the H₂ emission also correlates well with the submillimetre continuum emission and the CO emission (Pak et al. 2004).

Even a good part of the H₂ line emission showing evidence of thermalisation in starburst nuclei and especially ULIRGs (Davies et al. 2003, Pak et al. 2004) should come from the densest parts of PDRs (n_H2 10⁴ cm^-3). While the v = 1 levels are thermalised at ~ 1000 K, UV-pumped gas is needed to account for the higher levels. A similar conclusion applies for the extended H₂ emission within a few hundred parsec around many AGN (Davies et al. 2006 and references therein). However, Quillen et al. (1999) argued that shock excitation may also be dominant in a fraction of the Seyfert galaxies they studied (see also Rodríguez-Ardila et al. 2005). Another case of shock dominated emission is the relatively nearby, double nucleus, late merging, luminous LIRG NGC 6240. We group the discussion of its extraordinarily strong H₂ rovibration lines, discovered by Joseph et al. (1984), with that of the equally strong rotation lines in Section 10.3.

Modern instrumentation on 8-10 m telescopes gives new opportunities for near-infrared H₂ studies of the central regions of starburst and AGN (see e.g. observations with the Integral Field Unit of the Gemini Near-Infrared Spectrograph reported by Riffel et al. 2006a). The availability of efficient adaptive optics instrumentation, such as integral-field spectroscopy with SINFONI at VLT, opens the possibility to directly study up to size scales ( 10 pc) comparable to those on which models predict the molecular torus around AGN should exist (e.g. most recently Schartmann et al. 2005), as shown by the first results (Davies et al. 2006, Mueller-Sanchez et al. 2006, Zuther et al. 2006, Neumayer et al. 2007, Reunanen et al. 2007).

Very extended near-IR H₂ emission have also been detected up to 20 kpc, in the extended regions of the central galaxies of several tens of cooling-flow clusters (Donahue et al. 2000, Edge et al. 2002, Jaffe et al. 2005, Johnstone et al. 2007). Warm H₂ (~ 1000-2500 K) seems present wherever there is ionization in the cores of cooling flows, and in most cases it also coincides with CO emission. The relative strentghs of the P line to the H₂ lines might indicate a source of UV excitation hotter than 10⁵ K, whose nature is still unknown.

The prospects appear extremely rich for exploiting the ubiquitous emission of near-IR lines of H₂ in strong starbursts and shocks, with the expected new capabilities of infrared astronomy: both from the ground for local galaxies, with the developments in instrumentation, adaptive optics, extremely large telescopes and interferometry; and from space, especially at high redshift, with the jump in performance expected from JWST (and even SPICA), with respect to Spitzer and ISO.

10.3. Mid-IR H₂ pure-rotational lines in galaxies

The emission of rotational lines of H₂ by galaxies pertakes many features with that of rovibrational lines. In particular, the main emitters are again the various types of starbursts and the AGN. However, there are two main differences: their excitation energies are a factor 5-10 smaller which makes their collisional excitation much easier, so that they probe the more widespread moderately warm molecular gas of only a few hundred Kelvin; they are extremely difficult to observe from the ground so that practically all the results have come from the space missions ISO and Spitzer.

The results of ISO on molecular hydrogen and warm molecular gas in galaxies have been recently reviewed by Verma et al. (2005) (see also Habart et al. 2005). In addition to the detections that they quote in the spectra of half-a-dozen active galaxies, the most comprehensive work is the analysis of a sample of 21 starburst and Seyfert galaxies presenting pure rotational lines from S(0) to S(7) by Rigopoulou et al. (2002). A multi-line analysis, including S(0), in this sample yields a temperature around ~ 150 K for the bulk of the warm gas, both for starbursts and Seyferts. The mass of this warm gas is about 10% of the total mass of molecular gas probed by CO in starbursts, and a larger fraction in Seyferts. Such a temperature and mass of warm gas are compatible with PDR heating in starbursts, and with additional heating of a larger mass fraction by X-rays in Seyferts. However, low velocity shocks may also contribute.

Spitzer has detected H₂ in much more galaxies. However, published results and their analysis are still very incomplete. Higdon et al. (2006a) have reported multi-line detections of a large sample of ~ 60 ULIRGs. The results extend those of Rigopoulou et al. (2002). However, the lack of sensitivity for the S(0) line tends to favour the emission of smaller masses of warmer gas (~ 300 K) in the S(1)-S(3) lines. Hotter gas is also revealed in a small fraction of the sample by the detection of the S(7) line.

A major Spitzer result is the determination of the properties of warm H₂ in the central regions of normal galaxies derived by Roussel et al. (2007) from measurements of rotational lines in the Legacy Program SINGS (Kennicutt et al. 2003). This study extends previous extragalactic surveys of emission lines of H₂, to fainter and more common systems (L_FIR = 10⁷ to 6 × 10¹⁰ L) of all morphological and nuclear types. It has securely detected the 17 µm S(1) transition in about 45 galaxies, probing the range 100-1000 K. The derived column densities amount to a significant fraction of column densities of the total molecular hydrogen, between 1% and more than 30%. The H₂ line intensities scale tightly with the emission in the PAH bands which can be understood from a dominant origin in photodissociation regions. However, many sources classified as AGN strongly depart from the rest of the sample, in having warmer H₂, smaller mass fractions of warm gas, and an excess of power emitted in H₂ with respect to PAHs, favouring shock excitation. In many star-forming sources, deviations from an apparent ortho to para ratio of three are detected, consistent with the effects of pumping by far-UV photons combined with incomplete ortho-para thermalization by collisions.

Most of these ISO and Spitzer studies were focussed on the central regions of galaxies. However, see Valentijn & van der Werf (1999); and among the most remarkable works are a few observations of extended regions in strongly interacting /merging galaxies. With ISO, Lutz et al. (2003) found very strong mid-IR H₂ lines, from S(0) to S(11), in the late merging, double active nucleus, system NGC 6240. This luminous LIRG was known to display complex, high-velocity H₂ 2µm emission extending over ~ 5 kpc and peaking between the two nuclei (van der Werf et al. 1993, Tecza et al. 2000). Shocks due to the turbulent central velocity field and the superwind created in a nuclear starburst are likely to dominate these extraordinary levels of emission. From a re-analysis of archival ISOCAM-CVF data of the early merger the Antennae, Haas et al. (2005) have found that the strongest H₂ emission is displaced from the regions of active star formation (Fig. 3). This indicates that the bulk of excited H₂ gas is shocked by the collision itself in the region where the two galaxies overlap.

Even more exceptionally strong H₂ lines have been reported by Egami et al. (2006) in the infrared-luminous brightest galaxy of the cluster Zwicky 3146 (z = 0.29). The line luminosities and inferred warm H₂ gas mass (10¹⁰ M) are six times larger than those of NGC 6240. Strong H₂ pure-rotational emission lines are also seen in cooling-flow clusters (Johnstone et al. 2007), and in some mid-IR weak radio galaxies, where these lines are excited by shocks induced by galaxy mergers or by AGN jets in the interstellar medium (Ogle et al. 2006). One particularly spectacular result is the detection by Appleton et al. (2006) of a powerful high-velocity H₂ emission associated with an intergalactic shock wave in the Stephan's Quintet group of galaxies. The molecular emission extends over 24 kpc along the X-ray emitting shock-front, and seems to be generated by the shock wave caused when a high-velocity intruder galaxy collides with filaments of gas in the galaxy group. The S(1) and S(2) lines of warm molecular hydrogen (~ 400 K) were also recently detected with Spitzer by Higdon et al.(2006b) in two tidal dwarfs galaxies formed in the tidal trails of NGC 5291.

Many other important results may be expected from Spitzer, including serendipiteous ones, thanks to the remarkable capabilities of the Infra-Red Spectrometer (IRS) and the large amount of time it has devoted to spectra of galaxies.

10.4. Prospects for detecting H₂ in forming galaxies

10.4.1. Warm molecular gas at various redshifts

There is no doubt that the role of infrared H₂ lines to probe the warm molecular gas in various violent contexts in galaxies will further develop in the future with the expected increase of ground and space capabilities. The expectations for the near-IR rovibrational lines with ELTs, adaptive optics and JWST have been discussed in Section 10.2. JWST will bring a similar breakthrough for pure-rotational lines at moderate redshifts, with the tremendous gain in capabilities expected for its mid-IR instrument MIRI with respect to Spitzer IRS (Section 12.3.1). Space missions more oriented toward far-IR, such as SPICA (Matsumoto 2005) and the SAFIR project (Benford et al. 2004), or dedicated to H₂ lines such as the H2EX project (Falgarone et al. 2005b) will allow an extensive exploration of the bulk of the warm H₂ gas in the low-z universe and in more or less deep extensions at high z.

10.4.2. Importance of H₂ lines in the physics of galaxy formation

The presence of molecules is thought to be essential for various steps occuring in the process of galaxy formation. However, outside the marginal cases of tidal dwarfs and other mergers discussed above, true cases of galaxies in formation, especially primordial ones at very high z, are rather out of reach of current observations of molecular lines. Their detection is an important goal for JWST and eventually ALMA. Similarly to star formation, molecules are thought to be essential for cooling the gas below ~ 10⁴ K, which is a prerequesite for full gravitational collapse of proto-galaxies and most proto-stars. In primordial gas, before the formation of heavy elements in the first stars, the species available for achieving such a cooling from ~ 10⁴ K to ~ 10² K, are very limited, only H₂ and HD indeed. However, in the absence of the extremely efficient formation of H₂ on grains, the common belief is that molecules may exist only as very small amounts, despite the fact that the gas thermodynamical conditions would be highly in favour of converting all hydrogen atoms into H₂. However, even small amounts of H₂ or HD may be enough for reaching the required cooling efficiency of condensations of primordial gas of various sizes from proto-galaxies to massive proto-stars.

The first question to address is then the expected amount of H₂ and HD in the post-Big Bang primordial gas issued from the recombination process, until it forms the first dark-matter and baryonic condensations leading to the first stars and galaxies. Despite the limited observational information, this is certainly one of the part of the Universe where the chemistry should be the best understood because of its extremely simple element abundances, well determined physical parameters and the absence of structure (until the first collapses of density peaks into dark-matter halos). Indeed, the exquisite measurements of the CMB features, such as the acoustic peaks (e.g. Spergel et al. 2006), confirm how amazingly well are understood the details of the physics of hydrogen recombination (see e.g. Seager et al. 2000 for a detailed description of this physics).

The derivation of molecular abundances in this primordial gas is indeed a splendid academic problem which might have born fundamental implications for galaxy formation and even CMB anisotropies. It is interesting to realize that the calculations of the primordial H₂ abundance have extended over more than 40 years, practically since the discovery of the CMB. The main processes forming H₂ in such conditions were immediately identified, using H₂⁺ as an intermediate product (Saslaw & Zipoy 1967) or more importantly H^- (Peebles & Dicke 1968). Later refined works (e.g. Lepp & Shull 1984, Puy et al. 1993, Palla et al. 1995, Galli & Palla 1998, Stancil, Lepp & Dalgarno 1996, 1998, the review by Lepp, Stancil & Dalgarno 2002, Puy & Pfenniger 2005, Núñez-López et al. 2006) included other intermediates such as HeH⁺ (also negligible with respect to H^-), calculations of deuterium and lithium chemistry, and updated cosmological parameters and reaction rates. Some of them brought substantial revisions to the important abundances of HD and LiH. Even the recent work by Hirata & Padmanabham (2006) have brought a non negligible decrease of the H₂ abundance from 2.6 × 10^-6 to 6 × 10^-7, by taking into account the effects of the nonthermal background produced by cosmic hydrogen recombination.

However, it is now well agreed that the abundance of H₂ in the non-condensed primordial gas, ~ 10^-6 (e.g. Hirata & Padmanabhan; as well as that of HD ~ 10^-10 e.g. Núñez-López et al. 2006), is far too small for cooling of early halos. Indeed, while it is widely believed that H₂ (and HD) cooling is essential in the final collapses forming the first stars and galaxies, it is recognized that the only H₂ and HD important for this are formed in the enhanced density gas of already partially collapsed halos (e.g. Tegmark et al. 1997). However, the basic chemical processes responsible for the formation of H₂ in this denser gas are mostly the same as in the non-condensed primordial gas, i.e. mainly the H^- channel. Similarly, the abundance of primordial LiH is now found to be so small (Lepp et al. 2002) that its possible role in scattering CMB photons seems negligible, and is anyway much smaller than that of Li.

Indeed, the question of the effect of H₂ and HD in galaxy formation is made more difficult because their abundance in the stages where they may be important, much depends on the highly complex and nonlinear processes involved in the formation of galaxies. In this context, eventually important molecular processes at galactic scales are strong shocks which should unavoidly form large amounts of molecules in the postshock dense cooling gas. Given the importance of H₂ emission in local galactic-size shocks, there is no doubt that H₂ (and HD) lines should be the main coolants of such shocks in primordial gas below ~ 10⁴ K. The rovibrational lines of H₂ emitted in the strongest shocks should be detectable by JWST (e.g. Ciardi & Ferrara 2001), and some of them perhaps by SPICA as well as some pure-rotational lines (Mizusawa et al. 2005, see also Omukai & Kitayama 2003 for the more ambitious SAFIR project). As discussed above, primordial HD should also be negligible in galaxy formation. However, various studies (e.g. Uehara & Inutsuka 2000, Flower & Pineau des Forêts 2001, Galli & Palla 2002, Núñez-López et al. 2006, Johnson & Bromm 2006, Greif et al. 2007) underline the substantial contribution of newly formed HD to gas cooling below ~ 500 K during the collapse of primordial clouds. The possibility of marginally detecting HD rotational lines with ALMA, especially the first one, 1-0, with rest wavelength 112 µm, was discussed in particular by Núñez-López et al. (2006) (see also Mizusawa et al. 2005). Despite the uncertainty about the abundance of HD, there is a marginal chance that this line may be detected with ALMA, in massive starburst galaxies, when their redshift is large enough (z 6.3) to move the line into a good atmospheric window.

The above conclusions about H₂ lines may apply to various objects displaying strong shocks in the era of formation of the first stars and galaxies and reionization (e.g. Johnson & Bromm 2006, Alvarez et al. 2006, Wise & Abel 2007). More generally, modelling the behaviour of H₂ and its emission lines is an essential part of the current intense simulation activity of this crucial era (see e.g. Yoshida et al. 2003, Ricotti & Ostriker 2004, Susa & Umemura 2004, Reed et al. 2005 and references therein). With the sensitivity of JWST and maybe of SPICA, H₂ lines will be an essential diagnostic tool for the most massive objects of this yet unexplored epoch.