2. COSMIC MOLECULES AND LOCAL INTERSTELLAR MEDIUM

The properties of the interstellar gas in various environments share many common features between different galaxies. Most of these properties are well examplified in the Milky Way, and in particular in local interstellar clouds. Therefore, the detailed information that we can get from local interstellar molecules is essential to understand the behaviour of the molecular medium through the Universe. Indeed, the techniques are the same, both for observations, mainly millimetre wave radio astronomy, and for modelling line formation as well as physical and chemical properties.

2.1. Physics and various components of the local interstellar molecular gas

Molecules are mostly found in the `molecular gas' which is one of the four or five major components with characteristic properties of temperature and density, and thus ionization and chemical composition, that one traditionally distinguishes in the ISM of a galaxy like the Milky Way (see e.g. Table 1 of Lequeux 2005 and Table 1.1 of Tielens 2005): hot coronal intercloud gas, warm ionized and neutral media, cold neutral atomic medium, molecular clouds. Molecules can hardly survive to photodissociation in the tenuous (n_H ~ 1-100 cm^-3), warm (T_K ~ 100-1000 K) atomic gas permeated by the UV radiation from massive stars. They are practically completely absent in the ionized gas (T_K ~ 10⁴ K) of the diffused ionized medium and the dense HII regions around massive stars; as well as in the coronal gas (T_K ~ 10⁶ K).

The abundances of the various species, atoms, ions, molecules and dust grains, found in the interstellar medium are controlled first by the typical "cosmic" abundances of atomic elements, reproduced in Table 1 and characterized by the overwhelming abundance of H and He, and the relatively large abundance of O, C and N.

As described e.g. in Tielens (2005), Lequeux (2005) and Dyson & Williams (1980), the molecular gas must be dense enough (n_H 10²-10³ cm^-3) so that its external layers (A_v 1, N_H 2 × 10²¹ cm^-2) efficiently shield the interior from UV photodissociation. Most of the molecular gas in the Milky Way is distributed in Giant Molecular Clouds (GMCs), with average densities slightly above 100 cm^-3, typical sizes of 30-50 pc, typical masses of a few 10⁵ M. While the external layers of GMCs have a substantial warmer HI component possibly with comparable mass, most of their interior is very cold, with temperature not much exceeding 10 K, as the result of the balance between heating by cosmic rays and cooling by CO rotation lines. Their gross structure is well traced by CO lines (see Sections 3 & 4), possibly complemented by extinction and ray emission generated by cosmic rays. Except for helium, most of their inner gas is molecular hydrogen, with about 2% in mass in sub-micron dust grains, large aromatic molecules (PAH), other molecules and a few atoms (mainly O). The degree of ionization maintained by cosmic rays is extremely low, e.g. ~ 10^-7. However, these molecular clouds are complex structures, with clumpiness at various scales and strong turbulence as attested by the width of several km/s of the molecular lines such as CO, as well as enhanced magnetic field of a few tens µG, roughly proportional to n^0.5 (see e.g. Fig. 2.6 of Lequeux 2005). They are generally self-gravitating, but remain stable for several 10⁷ years with the balance of magnetic and turbulent pressure and gravity. Their dense condensations (~ 10³ - 10⁵ cm^-3, parsec size and masses ~ 10-10³ M) are particularly interesting as related to the process of star formation and the site of a peculiar chemistry. In the absence of stars already formed, such `dark clouds' are even colder than the ambient molecular medium. However, the presence of young luminous stars may heat the gas of neighbouring condensations up to temperatures ~ 100 K, changing the chemical processes at work.

The molecular clouds are in constant interaction with the other phases of the interstellar medium and the massive stars that they form. They have thus boundary layers where the molecules are more or less photodissociated by the external UV radiation. Such `Photo-Dissociation Regions' (PDR) may be particularly active in the vicinity of OB associations, generating special physical and chemical conditions. Because of their larger inertia, massive molecular clouds are less permeated by interstellar shocks than the more diffuse phases of the intersellar medium. However, the molecules may survive and even be specially synthesized in the compressed, overdense, hot regions of interstellar shocks propagating in less dense regions, where their lines, e.g. rovibration lines of H₂, may provide basic diagnostic of shocks. Decades of active research on the local ISM have very well documented the physical and chemical properties of molecular clouds in our various Milky Way environment, their rich composition, complex structure and dynamics and their symbiosis with young stars. Milky Way molecular clouds provide us with the best templates to understand the molecular ISM in other galaxies, despite the frequent occurrence of more extreme conditions encountered in earlier and more violent stages of galaxy evolution where extragalactic molecules are currently observed.

2.2. Observational techniques

Because the molecular ISM is cold and opaque to the visible and UV radiation, our best information comes from millimetre radio astronomy and molecular rotational transitions, where the photon energy is of the order of kT_K. Indeed, as reminded in Section 1.3, the development of interstellar molecular astrophysics has closely followed and often motivated the progress of millimetre techniques. The size of the antennae is limited by the need of high surface accuracy and precise pointing. The first studies were performed in the 1970s by 1-10 m single dishes such as Columbia 1.2 m, Texas 5 m, Bell Lab. 7 m, NRAO 12 m, FCRAO 14 m, etc. They were superseded in the 1980-1990s by the Nobeyama 45 m and IRAM 30 m larger dishes, and the millimetre interferometers (BIMA 9 x 6 m, Nobeyama 6 x 10 m, OVRO 6 x 10.4 m, IRAM 6 x 15 m).

Millimetre astronomy is strongly affected by atmospheric absorption (and emission), mostly from H₂O bands, in a large part of the spectrum, especially at high frequency. However, it benefits from a number of excellent transmission windows. The latter allow multi-transition studies of practically all interesting molecules except hydrids with a single heavy atom. The 3 mm window, ~ 72-116 GHz, which includes at least one rotation transition of most molecules, has even produced important results in ordinary, low elevation sites especially in the first stages of millimetre radio astronomy. It is a chance for studying the Galactic and extra-galactic molecular medium, that the three first transitions of CO, the most important interstellar molecule, occur at frequencies with good atmospheric transmission, J = 1-0 at 115.271 GHz, J = 2-1 at 230.538 GHz, J = 3-2 at 345.796 GHz. However, the advantage of high altitude, dry sites has become essential, even in the 1.3 mm window which includes the 2-1 CO line which turns out to be often significantly more sensitive than the 1-0 line. Such excellent sites are mandatory for extending molecular observations to the sub-millimetre range, as shown by the pioneer work of CSO, JCMT and SMA at Mauna Kea. This has justified the choice of the 5000 m Chajnantor site in Chile for the worldwide mm-submm project ALMA (Section 12).

Millimetre studies fully benefit from the fundamental advantages of radio astronomy for high velocity resolution with the heterodyne techniques, and high angular resolution with multi-dish interferometers. Both are of course essential for studying structure and dynamics of galaxies. They will be fully implemented in ALMA with its 54 x 12-m (+ 12 x 7 m) dishes, its six initial frequency bands and its baseline up to 14.5 km. The development of millimetre astronomy was long impeded by the difficulty of making high sensibility detectors. However, the progress has been constant in this field so that the ALMA receivers, with supra-conducting junctions (SIS) and Hot Electron Transistors (HET), will approach the quantum noise limit in broad-band receivers. In parallel, the high resolution of the `back-end' spectrometers and the transport and processing of the interferometer signals have fully benefited from the progress of high speed information technology and computers, culminating in the giant ALMA correlator (under construction).

Other wavelength ranges are complementary to the millimetre one for investigating the Galactic and extra-galactic molecular gas for more specific, but important, goals. There is a natural extension to lower frequencies of the centimetre and decimetre radio range, for three main fields: 1) these frequency ranges include a few very important fine structure lines, mostly of abundant hydrids without important millimetre lines, OH, H₂O and NH₃, with strong masers for OH and H₂O; 2) at very high redshift 3 mm lines, such as CO(1-0), are shifted to cm wavelengths; 3) the first rotation transitions of very heavy molecules, such as HC_nN, are located at cm wavelengths; however, their low abundance will not allow much development in extragalactic studies. The Green Bank Telescope (GBT) already studies extragalactic mega-masers and CO(1-0) at very high redshift. The extension of the VLA (EVLA) will allow many more similar studies, waiting for the tremendous sensitivity and VLBI capability of the Square Kilometre Array (SKA) which will make a breakthrough in this field. In the infrared, there is also some extension to the far-infrared (100-200 µm) for high-J rotational transitions, mostly H₂O. The mid-infrared is important for PAH and dust molecular features (Section 8), while the whole infrared range is essential for H₂ lines, although there are highly forbidden (see Section 10). The most important ranges for H₂ at zero-redshift are: i) ~ 5-30 µm for rotation lines emitted by the warm molecular medium; such studies will certainly have important developments with future infrared space projects such as JWST, SPICA, SAFIR, etc; and ii) ~ 2-5 µm for rovibration lines emitted in shocks and regions with strong UV radiation, and observable from the ground with very large telescopes and eventually adaptive optics. The allowed UV strong absorption lines of H₂ (and other molecules) are much more powerful to trace small column densities of H₂ as currently observed by FUSE in the local diffuse gas, and with large telescopes such as VLT, when they are redshifted. However, the number of such detections in extra-galactic lines of sight has remained limited (see Section 7.1).

In conclusion, studies of interstellar molecules belong mainly to the realm of millimetre radio astronomy. They benefit from the outstanding technical progress in this field and the power of image synthesis and of high-resolution heterodyne spectroscopy with large radio arrays, which will culminate with ALMA. However, millimetre techniques will eventually be complemented in the exploration of the molecular world of galaxies by other impressively large facilities in wavelength ranges from radio to infrared and visible.

2.3. Modelling (millimetre) molecular line formation, and diagnostic of physical conditions and molecular abundances in the ISM

Most of our information on the molecular interstellar medium comes from the detected intensities and profiles of (millimetre) molecular lines. With a relatively simple modelling, one infers estimates of the main physical parameters, temperature, density, velocity distribution, and the chemical abundances of the observed molecules. Of course in practice, all these quantities are some kind of average over the volume traced by the radiotelescope beam. As the main cooling of the molecular medium is achieved through lines of the most abundant species, mostly CO, such a modelling is also the basis of the evaluation of the cooling rates (see Section 2).

2.3.1 Equations of statistical equilibrium and radiative transfer.     The main goal is to determine the populations of the rotation levels and the line radiation intensities which are coupled through the equations of statistical equilibrium and radiative transfer. As usual, the most simple case is that of local thermodynamical equilibrium (LTE) when the collision transition rates dominate over radiative transitions. The rotational populations have then a Boltzman distribution determined by a single excitation temperature equal to the gas kinetic temperature. However, LTE is never fully realized with the low density of the interstellar medium. For a given rotational transition, LTE is only approached when the density n_H2 exceeds a critical value A/C where A is the Einstein coefficient for spontaneous emission and C is the collisional rate. The A coefficient scales as µ^{2 3} where µ is the dipole moment and the line frequency. In practice the decay rate is lowered by line trapping for optically thick lines so that the critical density is decreased. Typical values thus range from 300 cm^-3 for CO(1-0), because of the low value of the CO dipole, to ~ 10⁵ cm^-3 for the first transitions of molecules with usual, large dipoles, such as HCN, CS or HCO⁺. For an optically thick line at LTE, the line intensity J is just equal to the Planck function B(T_K).

In the general case, one must solve the coupled equations of statistical equilibrium and radiative transfer. For all usual molecules the Einstein radiation coefficients are well known. Values for the collisional transition rates rotational levels (or cross-sections, mainly for H₂, see e.g. Flower & Launay 1985) are provided either by full quantum calculations for the most important simple systems, or, for the others, by various approximations accurate enough to match the uncertainties of the radiative transfer geometry. In most cases, the treatment of radiative transfer is much simplified by the use of the Large Velocity Gradient (LVG) approximation based on local photon trapping and the escape probability method (Sobolev 1960, 1963, Goldreich and Kwan 1974, Scoville and Solomon 1974). It is equivalent to replace the Einstein A coefficient by A where is the `escape probability'. For optically thick lines with line optical depth _l >> 1, is just proportional to _l^-1. The line optical depth _l is generally estimated from a Doppler gaussian profile with the observed line width v. It is worth noting that the Lorentz wings of millimetre molecular lines are always negligible because of the very small values of the Einstein A coefficients.

Despite its success in representing a basic feature of radiative transfer with Doppler broadening, the LVG approximation is clearly limited to account for the complexity of radiative transfer in actual media. The development of computing power allows a more and more generalized use of Monte Carlo methods for modelling radiative transfer, with a degree of complexity for describing the structure of the interstellar medium adapted to the available information (see e.g. Gonçalves et al. 2004 and references therein). Such methods are particularly well adapted for computing the emerging intensities and complex line profiles resulting from actual inhomogeneous distributions of densities, temperatures, abundances, velocity fields, and the resulting populations of the rotational levels.

2.3.2. Radiotelescope signals: flux density and antenna temperature     In brief, through such modelling, rotational millimetre lines are a powerful tool for tracing molecular abundances, including isotope varieties, and physical conditions in molecular clouds such as density, temperature and velocity fields. However, it is clear that the derived information is some kind of average over the molecular gas included in the radiotelescope beam. This is a serious drawback for extragalactic studies with the large beams of single dish millimetre telescopes. For the 10-30m telescopes discussed in Section 2.2, in most cases the beam diameter ranges from 10" to 1'. For the nearest galaxies at a distance of ~ 1 Mpc, this corresponds to scales of ~ 50-300 pc, i.e. at best the typical diameter of a giant molecular cloud. At any redshift z 0.5, the corresponding beam-encompassed distances (varying little with z) are rather ~ 50-300 kpc, so that there is no hope to derive any other information than global values for a galaxy. Therefore, extragalactic molecular studies really need the increased angular resolution of millimetre interferometers. Current best facilities, such as the IRAM interferometer, (Fig. 7a), already reach ~ 0.3-0.5" resolution. This provides nice details on nearby galaxies, but, for high z galaxies, hardly allows to distinguish molecular emission of their core from the possible one of their most outer regions. However, ALMA, with better than ~ 0.1" resolution and high sensitivity, will provide detailed information on sub-kpc molecular structures at practically any redshift. Comparable or even higher resolution are already achieved by cm/dm very long baseline interferometry (VLBI) with the very strong intensities of extragalactic mega-masers (see below). Such capabilities will be greatly extended by SKA, including for thermal low-J, high-z molecular lines. The strongest ones may already be angularly resolved with the current VLA angular resolution (~ 0.15", see e.g. Walter et al. 2004, Riechers et al. 2006c).

For optically thin lines, the line flux density S detected at a given frequency by a radiotelescope (generally expressed in Jansky, 1 Jy = 10^-26 Wm^-2Hz^-1) is proportional to the total number of molecules in the upper level u of the transition at the corresponding radial velocity v within the telescope beam. The line profile is just given by the velocity distribution of such molecules, and the line velocity integrated flux density Sv = S(v) dv ¹ is proportional to the total number of molecules in level u within the beam (or more exactly the convolution of the molecule distribution with the beam). If the population distribution among the rotational level may be estimated, e.g. with an approximate rotational excitation temperature, the total number _i of the considered molecules i within the beam may be directly inferred. However, the corresponding abundance of molecules i, _i = N_i / N_H2, also requires the knowledge of N_H2 which can only be indirectly derived.

When the line ul is not optically thin, the emission from level u at each position in the cloud must be weighted by the absorption factor exp[-_iul(v)], where _iul(v) is the optical depth up to the boundary of the cloud, and depends on the frequency and hence on v. This gives a larger weight to the emission by the outer layers of the cloud, and also modifies the line profile. For very thick lines, the specific intensity J is just B(T_K) at LTE, and it generalizes as J = B (T_exc) where (E_u - E_l) / kT_exc = -log(n_u g_l / n_l g_u) is the line excitation temperature. Thus, for thick lines, as often CO lines are, the flux density directly reflects the excitation temperature through S = J = B(T_exc), where is the solid angle subtended by the source (more exactly, the solid angle of the source convolved with the telescope beam) ². If T_exc varies within the cloud, the line intensity is determined by the excitation temperature of the outer layers; in particular CO lines often form at LTE and reflect the kinetic temperature of the boundary layers of clouds.

2.3.3. Interstellar masers.     In extreme cases, the divergence from LTE may even lead to cases of `population inversion' where the population n<sub>u</sub> / g<sub>u</sub> of the upper level is larger than the lower one n<sub>l</sub> / g<sub>l</sub>. The excitation temperature T<sub>exc</sub> is then negative, as well as the line optical depth, leading to an exponential amplification exp(|<sub>ul</sub>|) of the line intensity. The resulting very powerful maser lines have been observed since the beginning of molecular radio astronomy in a few transitions of OH ( = 18 cm with four hyperfine lines) and H<sub>2</sub>O ( = 1.35 cm) in Galactic massive star forming regions (see e.g. Lo 2005 for a list of general references about interstellar masers). The highly non linear amplification favours emission in the directions where the velocity configuration leads to the largest amplification, with very small solid angles <sub>s</sub> for the emission regions, which are measurable by VLBI with milli-arcsec resolution. From the large observed line flux densities, S, one infers extremely high brightness temperatures T<sub>b</sub> = S / <sub>s</sub> × <sup>2</sup> / 2k, ranging up to 10<sup>14</sup> K. Much more luminous OH and H<sub>2</sub>O maser emission (`mega-masers') was discovered in the nuclear regions of external galaxies with line luminosities ~ 10² - 10⁴ L, 10⁶ times more luminous than typical Galactic masers (see Lo 2005 and Sections 4 & 10).

Conditions leading to such powerful Galactic and extragalactic maser emission are extremely complex. Population inversion is a necessary condition, together with velocity coherence along the line of sight allowing a significant gain. The pumping mechanisms achieving population inversion are not always fully understood. They result from complex combinations of collisional and IR-line radiative effects whose net result favours the excitation of the upper level of the maser transition, either through cascades from upper levels or because spontaneous IR emission from the lower maser level is faster than from the upper one. It is certainly not just a chance that the two main molecules with strong interstellar maser emission, OH and H₂O, are abundant and have a complex rotational level structure. For extragalactic mega-masers, it is generally agreed that the intense mid-infrared radiation in extreme starburst regions plays a major role in OH pumping, while the energy required to achieve the collisional pumping of H₂O in the central parsec of AGN comes from the AGN (see Lo 2005 and Section 11).

2.4 Uniqueness of basic processes of interstellar chemistry

2.4.1. Known interstellar molecules.     Immediately after the beginning of the harvest of discoveries of interstellar molecules by millimetre radio astronomy, it was realized that the interstellar chemistry is extraordinary and unique, through the number of detected exotic molecules and their peculiar abundances which were derived. There are now about 150 interstellar molecules identified in the local ISM of the Milky Way (see updated list in http://astrochemistry.net/, see also http://www.cv.nrao.edu/~awootten/allmols.html, and http://physics.nist.gov/cgi-bin/micro/table5/start.pl by F.J. Lovas for NIST recommended frequencies for observed interstellar molecular microwave transitions). Formed mainly from H, C, O & N atoms, they can also include S, Si, metals, etc. They belong to two main classes: expected common stable species with up to ~ 10 atoms; and a large number of exotic unstable species, very uncommon and rare because of their instability in normal laboratory conditions, including radicals (such as CH_n with n = 2 to 6), ions (such as HCO⁺ and N₂H⁺), long polyyne chains (HC_2n+1N with n = 1 to 3), small cycles (such as C₃H₂), or isomers (such as HNC). Most of the latter are extremely unstable and hardly observed in the laboratory, except as very transitory intermediate reaction products. A number of them were not even known before their discovery in interstellar space. Clearly, the existence of such exotic compounds is directly linked to the incredibly low interstellar densities, which allow them to survive for 10⁵-10⁶ yr and to be observable provided they are efficiently formed. They are obviously completely out of equilibrium with respect to very low interstellar temperatures. Three-body processes are practically excluded in the interstellar gas phase. Grain processes had long been proposed for molecule synthesis, but they cannot account for most of such unstable species. On the other hand, the latter are logically explained by reaction chains initiated by energetic particles such as UV photons and cosmic rays. Only exothermal reactions with practically no activation barrier are generally possible, implying radical and mostly ion reactions. However, the creation of the initial chemical bonds remains difficult without a third body. It practically needs dust grain reactions, at least for the formation of H₂. The interpretation of interstellar chemistry, combining gas phase and grain processes, has been excellently reviewed a number of times in the last three decades (see e.g. Solomon & Klemperer 1972, Herbst & Klemperer 1973, Watson 1976, Tielens 2005, Lis, Blake & Herbst 2005, and references therein). The basic features have not much changed during this period, with however the addition of deeper insights on various particular points, including new information from observations, laboratory and theory. Current views may be summarized as follows.

2.4.2 Gas phase chemistry.     Although the importance of grain processes is not to be underestimated, the main success in detailed modelling has come from gas phase reactions. They are simpler and pretty well understood, especially for charged species. The most direct possible analogy with other out-of-equilibrium chemistry we are used to in dilute gases, is probably photochemistry in the upper earth atmosphere. Ultraviolet radiation from massive stars is ubiquitous in the interstellar medium, except in molecular clouds. It is certainly even stronger in many other, younger, galaxies than the Milky Way, with more UV photons generated from star formation, or less shielding by dust. Shielding from UV radiation is crucial for the survival of interstellar molecules, and there are practically no UV photons deep in molecular clouds. However, photochemistry is essential in all regions more or less permeated by interstellar UV radiation, either ambient or coming from a local source: boundary regions of molecular medium; translucent, mostly atomic, clouds; outer circumstellar shells, etc. The most spectacular photochemistry occurs in the photodissociation regions (PDRs), and such photo-processes are among the best understood ones in interstellar chemistry. Their main effect is the destruction of molecules, explaining the low abundances of molecules in the diffuse interstellar medium, and even their quasi-absence in high-redshift optical absorption systems (see Section 7.1). However, the radicals and ions they generate may rival with, and even supersede, cosmic rays for initiating reaction chains, forming various species. The physics and the chemistry of PDRs are entirely dominated by the UV field which is many orders of magnitude larger than the standard interstellar field. In addition to the enhancement of photochemistry itself, an important feature of PDRs is the high temperature which allows opening new reaction channels, such as C⁺, O or OH with H₂, overcoming small activation energies (Tielens & Hollenbach 1985, Tielens 2005, Lequeux 2005, and references therein). Note that the UV radiation also achieves the excitation and dominates the life cycle of PAHs, especially in PDRs (see Section 8).

As the interior of molecular clouds is deprived from UV photons, it is agreed that the chemistry leading to the formation of the abundant unstable species is mainly initiated by cosmic rays which easily propagate inside (X-rays may play a similar role in some cases, especially AGN, see Section 11). The energy reservoir in cosmic rays is quite significant, and indeed comparable to the other forms of interstellar energies which are all roughly in equipartition, the gas thermal energy in particular. With typical energies ~ 0.1-10 GeV, they dominate the heating and ionization of molecular clouds and, from the ionization of H₂ and He, they trigger their gas phase chemistry with molecular ions. Ion reactions are particularly efficient because of the long range r^-4 charge-induced dipole interaction which leads to very large pseudo-capture cross-sections and most often to the absence of activation energy for exothermal reactions. Many ion reactions thus proceed with very fast rates, close to the universal Langevin pseudo-capture rate, ~ 2 × 10^-9 cm^-3s^-1 (e.g. Su & Bowers 1979). Basic modelling (Herbst & Klemperer 1973, Watson 1973, 1976) shows that they well explain many key properties of interstellar chemistry including the presence and the central role of molecular ions such as H₃⁺ and HCO⁺, the high abundance of CO, the presence of key radicals such as OH and CH, and isomers such as HNC/HCN, and the high abundance of deuterated species.

Standard models of gas phase interstellar chemistry also include (see publicly available codes from e.g. http://www.physics.ohio-state.edu/~eric/research.html, and the UMIST reaction rate database http://www.udfa.net/): radical reactions without activation barrier whose rates are generally more uncertain than those of ion reactions; universal destruction of molecular ions by dissociative recombination; radiative association generally very efficient for very large molecules, but slow and uncertain for small species; etc. This yields large networks of reactions, with up to hundreds of species and thousands of reactions. However, many reaction rates can only be guessed. The large time constants involved, ~ 10⁵-10⁶ yr, make better to search for time dependent solutions than just the equilibrium state that may never be achieved. Despite success on various particular points, detailed quantitative predictions remain overall difficult for various reasons: uncertain rates for most important processes; difficult modelling of grain processes; lack of information about the physics, structure and evolution of the clouds; coupling between chemical and physical evolution (even able to lead to bistable equilibrium); etc. Therefore, the most notable advances in the last decades have rather addressed particular questions often triggered by observational results, such as the PAH chemistry, or various extensions especially to the warm gas: hot molecular cores, shocks, PDRs, AGN, turbulent heating, etc.

2.4.3. Grain processes     are fundamental because they are by far the main source of new chemical bonds and also because grains are one of the main repositories of heavy elements and a potential source for complex species including PAHs. However, despite significant advances, they remain the most difficult part of interstellar chemistry because of the complexity of grain structure and chemical evolution and of the interactions with the gas. Surface reactions are the only efficient mechanism to form H₂, and hence generate most interstellar molecular bonds by subsequent gas phase reactions, at a sufficient rate to compensate destruction processes. We have a basic logical scheme for this formation since the beginning of modern interstellar chemistry (Hollenbach & Salpeter 1971, Watson & Salpeter 1972), which has little changed since (see e.g. Tielens 2005 for current views): available H atoms in the gas hit grains every ~ 10⁶ × [300 cm^-3/n] yr, have to stick on the surface through physisorption - i.e. weak binding through Van der Waals forces - with a high probability and easily migrate so that they almost surely find another H previously fixed on a chemisorption defect site, and react to form an H₂ molecule which is immediately ejected from the grain because its adsorption energy is too weak. Such a scheme is well supported by the general knowledge of surface physics and a few laboratory results with some materials. However, it is highly dependent on the unknown actual detailed surface state so that quantitative estimates remain out of reach and entirely rely indeed on the knowledge of the interstellar H₂ abundance and destruction rate.

For all other gaseous molecules which could significantly form in grains, the rates are even qualitatively highly uncertain, because they depend on complex processes allowing the return to the gas of compounds much more tightly bound to the grains. However, there are a number of facts well established about chemical grain processes (see e.g. Tielens 2005 and references therein). Any gas species periodically hits a grain and may stick with a probability which exponentially depends on the temperature and the physisorption energy. All species, except H₂ and He, thus eventually stick on the coldest grains of dense molecular cores. Through subsequent surface reactions, grains therefore accrete mantles of compounds more or less volatile depending on their temperature, whose composition may be traced by infrared spectroscopy. Ice mantles, primary products of O + H chemistry are thus commonly found in dense clouds. Sensitive infrared spectroscopy reveals that ice is mixed with various impurities, expected products of accretion plus surface reactions in the H rich interstellar context (Tielens 2005 and references therein): CH₃OH, CO₂, NH₃, CH₄, CO, OCS, etc. In the cloud boundary layers such mantles may be exposed to UV photochemistry leading to more stable mantles of organic polymers with radical defects. It is natural to think that such a rich chemistry may be at the origin of the synthesis of a number of interstellar molecules, especially those which are not easily formed by gas reactions. In particular, abundant deuterated species should be synthesized through the strong isotopic fractionation associated with the very low grain temperature. However, a main problem for the assessment of the importance of these processes remains the way that such mantles may be desorbed with injection of complex molecules into the gas.

It is well established that grain mantles are submitted to a harsh processing when they leave the interior of molecular clouds. The volatile mantles are first sublimated when their temperature rises and they are exposed to UV radiation. More stable polymer mantles and even refractory cores of silicate or amorphous carbon may eventually be destroyed by sputtering or grain shattering when they are periodically submitted to interstellar shocks or hot gas (Draine & Salpeter 1979a, b). However, such processes are probably too violent to preserve most molecular bonds. Milder UV photodesorption is probably efficient for keeping grain surface clean of physically adsorbed molecules such as H₂O at the edge of molecular clouds. However, its efficiency is more uncertain for heavier molecules more or less chemically bound to the grains. Intermediate desorption processes have also been proposed, such as cosmic ray driven mantle explosions (see e.g. Léger et al. 1985, Tielens 2005). This key question of grain desorption is currently addressed by several experimental and modelling studies with significant recent progress (e.g. see Collings et al. 2004, Garrod et al. 2007).

Products of a warm chemistry, associated with grain desorption products, is observed in interstellar shocks and hot cores surrounding many protostars. The latter are warm (> 100 K) and dense (> 10⁶ cm^-3) regions where abundant hydrogenated molecules such as H₂, NH₃, CH₃OH are found, similar to those of interstellar ices. Their origin from desorption of molecular mantles from heated grains is also supported by the very large abundance of deuterated species reflecting the important isotopic fractionation at the very low temperature of the grains. However, the multitude of more complex organic species also observed in the gas probably requires active gas reactions from the primary desorbed products (see e.g. Caselli et al. 1993, Tielens 2005). Similar, but more extreme processes, occur in interstellar shocks (see e.g. Hollenbach & McKee 1979, 1989, McKee & Hollenbach 1980, Flower et al. 1995, Flower & Pineau des Forêts 1995). Post-shock temperatures higher than 1000 K allow reactions with endothermal or activation energies such as the formation of OH and H₂O from O + H₂. The gas composition suggests not only the grain desorption of icy mantles, but also sputtering of silicates when abundant SiO is observed.

2.4.4. Isotope fractionation     is the general result of the dependence of the zero-point vibration energies on the atomic masses at the very low interstellar temperatures. The energy difference for deuterium, D, versus hydrogen, H, E, is very large, several hundreds K, i.e. much larger than kT in cold media, allowing enormous overabundance of deuterated compounds (Watson 1974, Roueff et al. 2000, 2005 and references therein). It is still significant for heavy elements such as ¹³C versus ¹²C, a few tens K, and thus for the determination of element isotope ratios such as ¹³C / ¹²C reflecting nucleosynthesis processes. However, the ratios of molecular isotope varieties are rarely just determined by the exponential factor exp(-E / kT) of thermal equilibrium. In particular for deuterated molecules such as HD / H₂, DCO⁺ / HCO⁺, DCN / HCN, D₂CO / H₂CO, ND₃ / NH₃, they depend in a complex way on the chemical processes and the thermal history of the molecular material. Their observational determination may thus provide precious information, such as the importance of cold grain chemistry in warm media. A recent breakthrough in this field is the discovery of poly-deuterated molecules (Lis et al. 2002, van der Tak et al. 2002), such as ND₃ (see references in Roueff et al. 2005).

The ubiquitous interstellar polycyclic aromatic hydrocarbons (PAHs) (Section 8) are in some respects intermediate between usual interstellar molecules and grains for their chemistry as well as for their physical properties (see e.g. Tielens 2005 and references therein, Omont 1986, Puget & Léger 1989). Among significant processes are: 1) ionization processes: single and multiple ionization, Coulomb explosions, electron recombination and attachment; 2) photochemistry and other chemical reactions proceeding through unimolecular reactions in intermediate complexes: photo-dehydrogenation and loss of side groups, eventual photo-destruction, accretion of H and other gas atoms, ions and small molecules, etc.; 3) interactions with grains: accretion onto grains, PAH injection into the gas from grain desorption or shattering.

2.5. Relation between interstellar and other cosmic molecules

Most cosmic media displaying significant amounts of molecules share some kind of similarity, especially for the element abundances, and to a lesser degree for the physical conditions such as temperature, density and UV radiation. It is thus not surprising that the other cosmic molecules present many similarities with interstellar molecules, such as the overwhelming importance of H₂, but also large differences. The latter are particularly obvious when there are large differences in density or chemical composition.

In the atmosphere of cool stars (red giants, supergiants and dwarfs) and brown dwarfs, the relatively high density leads to local thermodynamical equilibrium (LTE) for the molecular abundances which are just determined by the laws of chemical equilibrium at temperatures of a few thousand Kelvin. With `normal' element abundances (C/O < 1), one finds CO and H₂O as dominant molecules (after H₂), similarly to the ISM. However, the gas is much less rich in complex and C-bearing molecules. Another striking difference is the absence of dust (except in brown dwarfs and some M dwarfs: in fact below 200K dust plays a major role in the opacities of cool stars). The absence of dust makes metals free to efficiently form diatomic oxides, which, for a few of them, such as TiO, display spectacular spectral bands in the visible. In the special case of the atmospheres and circumstellar envelopes of carbon-rich giants (C/O > 1), most oxygen is locked in CO; the dominant other molecules include then CN in the atmosphere and C₂H₂ and HCN in the colder circumstellar envelope.

The stellar media which are the closest to the interstellar medium, are those of the extended `circumstellar envelopes' and `planetary nebulae' ejected at the end of the AGB red giant phase. The conditions in the cold, UV-shielded, circumstellar envelopes may be very similar to molecular clouds, but with a peculiar chemistry characterised by short time constants, clear-cut photochemistry and cases with C/O > 1 (see e.g. Glassgold 1996). Planetary nebulae are more similar to interstellar HII regions, with molecules and peculiar photodissociation regions in the youngest ones (Huggins et al. 2005).

The atmospheres of planets are close to LTE at low temperature, with abundant CH₄ and NH₃ in jovian planets, an absence of primitive H₂ and peculiar composition in telluric ones. Photochemistry in the upper layers may have some broad similarities with that in the ISM.

However, the most obvious and fundamental connection is between interstellar molecules and comets. The molecules detected in comas are essentially a subset of those identified in the ISM. Their presence and abundance are very well explained from the action of photochemistry on parent molecules directly ejected from the comet nuclei (see e.g. Bockelée-Morvan et al. 2000). The latter are pretty similar to the most abundant interstellar molecules, and more precisely to those of the dust grain mantles.

Indeed, comet nuclei have a direct relationship with interstellar dust and its UV-processed mantles accreted from interstellar gas. They are in some respects samples of the interstellar medium and their composition partially reflects that of both interstellar dust and gas (see e.g. Ehrenfreund & Charnley 2000). The detailed analysis of comet samples will thus be fundamental, not only to understand the process of comet formation, but also as a unique information about properties and composition of the interstellar dust and even gas. In particular, they can cast some light on the existence and formation of molecules in the context of formation of the Solar System, and maybe in the general interstellar medium. A good example is that of amino-acids. Their detection is not yet confirmed by radio astronomy in the interstellar medium. Even if they are marginally detected in the future, the detectability of molecules with this type of complexity will remain at the limit allowed by the spectral confusion of the accumulation of all other weak interstellar lines. On the other hand, amino-acids have been detected in meteorites (see e.g. Cronin & Pizzarello 1997, Ehrenfreund & Charnley 2000).

Anyway, the rich variety of interstellar molecules just slightly less complex than amino-acids, detected by millimetre radio astronomy, shows that the special processes of interstellar chemistry may be very efficient in synthesizing the building blocks of pre-biotic molecules. The information one can expect from the combination of interstellar molecules, comet nuclei, asteroids and meteorites, is at the heart of astro-biology and certainly fundamental for eventually progressing in understanding the origin of life.

¹ v is the line full width at half-maximum (FWHM) and S is approximately the peak line flux density Back.

² In the Rayleigh-Jeans regim B(T) = 2kT / ², and the flux density S is proportional to T_exc. For an extended source broader than the beam, is the beam solid angle ² / A_a where A_a is the antenna surface area. For a source of brightness temperature T, the specific power picked up by the antenna is P = A_a S = 2kT (kT per polarization, Nyquist theorem).

The simplicity of this relation has induced radioastronomers to use temperature as unit to measure the signal of radiotelescopes. However, for millimetre and sub-millimetre wavelengths and typical values of T_exc, the Rayleigh-Jeans approximation kT/h >> 1 is generally not valid (kT/h = 0.7 [ / 1 mm][T/10 K]). Nevertheless, antenna temperatures are still in use with the same formal definition T_A^* = (² / 2k) I A_eff S / 2k, where the effective antenna area A_eff takes into account the antenna imperfection. But T_A^* is no longer equal to the source brightness temperature T_b in the case of an extended source and perfect antenna, but to T_b^* = [h / k] / [exp(h / kT_b) - 1]; and the added * aims at reminding that. See radio astronomy textbooks (e.g. Kraus 1986, and also Lequeux 2005) for complete technical definitions. Back.