Without data, you’re just another person with an opinion.
(Attributed to W. Edwards DEMING)
In this chapter, we review how the models presented in Sect. 2.3 are used to derive the grain properties of nearby galaxies. The term “dust properties” is vague. In the literature, it often indistinctively encompasses the three following categories (Galliano, Galametz, & Jones, 2018).
SED modeling is one of the main methods to empirically derive the dust properties of a region or a galaxy (cf. e.g. Galliano et al., 2018, for a review). The inherent complexity of astrophysical sources requires to account for the diversity of physical conditions within the studied region. The treatment of radiative transfer, even in an extremely approximated fashion, is thus necessary.
Radiative transfer is the method solving the propagation of multiple rays of light, emitted by one or several sources, through a macroscopic heterogeneous medium. It accounts for the scattering, absorption and emission, at each point and along each direction, in the studied region.
Radiative transfer deals with several quantities that are often improperly defined or mixed together in the literature: intensity, flux, emissivity, brightness, etc. We have already seen some of these quantities in Sect. 1.2.4. We now define them and explicit their differences (cf. Chap. 1 of Rybicky & Lightman, 1979, for a complete review). In what follows, we assume stationary systems. The time variable, , is used only to denote constant rates.
The moments of the specific intensity. The primary radiative transfer quantity is the specific intensity or brightness (cf. Fig. 3.1.a):
(3.1) |
The specific intensity is the electromagnetic energy, , per unit time, , area, , solid angle, , and frequency, . It therefore quantifies the infinitesimal power carried by a monochromatic light ray. This quantity depends on the position in the region,, and on the direction of propagation, . We adopt the spherical coordinate conventions used in physics (Fig. 3.1; cf. Appendix C.1):
(3.2) |
The first moments of the specific intensity, relative to its angular distribution, are physically meaningful.
(3.3) |
It is the specific intensity averaged over all directions. It is often used to quantify the ISRF, accounting for rays coming from all directions. In the isotropic case, we have , .
(3.4) |
It represents the monochromatic power per unit area passing through a surface element perpendicular to . The factor is there to account for the reduction of the density of rays that are not perpendicular to the surface. In the isotropic case, , because there is the same amount of flux passing through the area in both directions.
(3.5) |
It is the momentum flux carried by the photons, as is the momentum of a single photon. The term comes from two sources: (i) one comes from the reduced fraction of inclined rays, similar to the flux; (ii) the second factor comes from the fact the pressure is the momentum vector component that is perpendicular to the surface.
Specific energy density. The radiative energy within a volume element at a given time is the specific energy density:
(3.6) |
The second equality comes from the fact that , being a volume element (cf. Fig. 3.1.d). The specific intensity is a power per unit area, whereas the energy density is an energy per unit volume. Both quantities are linked, combining Eq. () and Eq. ():
(3.7) |
If we integrate Eq. () over all directions, we get, using Eq. ():
(3.8) |
Emission coefficient and emissivity. The monochromatic emission coefficient is the power radiated in a given direction, per unit volume and frequency:
(3.9) |
We have also seen, in Sect. 1.2.4, the emissivity:
(3.10) |
where is the mass density of the ISM, and , its mass element. The factor , in Eq. (), makes the solid angle fraction of monochromatic power emitted in a given direction, per unit mass. In Sect. 1.2.4, we were considering the volume and mass of a grain, whereas, here, we are considering the volume and mass of the ISM. Combining Eq. () and Eq. (), we get:
(3.11) |
Extinction coefficient and opacity. The amount of specific intensity absorbed and scattered along an infinitesimal path length, , in the direction , is the extinction coefficient, , defined such that:
(3.12) |
Similarly to the convention we have adopted for in Sect. 1.2.4, we pose to distinguish absorption and scattering.. Eq. () can be expressed with microscopic quantities, assuming the absorbers and scatterers have a cross-section and a density :
(3.13) |
If the composition of the ISM is homogeneous, then . The opacity, that we have seen in Sect. 1.2.4, is related to by:
(3.14) |
Mean free path. The mean free path of a photon with frequency, , at position can be defined as:
(3.15) |
It is the average length a photon will be able to travel before being absorbed or scattered. We can make the same remark as for the emissivity: in Sect. 1.2.4, we were considering the cross-section of dust particles per mass of grain, whereas Eq. () gives the cross-section of the whole ISM per mass of ISM. Table 3.1 gives typical values of in a homogeneous medium, assuming the dust constitution of the THEMIS model (cf. Sect. 2.3.2.2).
In the diffuse ISM (WNM; ; Table 3.6), the mean free path of a photon in the visible range is of the order of a kiloparsec.
| HIM | WNM | CNM | Molecular clouds
| |
| |||||
| 139 kpc | 1.39 kpc | 13.9 pc | 0.0417 pc | 86.1 a.u. |
| 177 kpc | 1.77 kpc | 17.7 pc | 0.0532 pc | 110 a.u. |
| 223 kpc | 2.23 kpc | 22.3 pc | 0.0669 pc | 138 a.u. |
| 275 kpc | 2.75 kpc | 27.5 pc | 0.0824 pc | 170 a.u. |
| 358 pc | 3.58 kpc | 35.8 pc | 0.107 pc | 222 a.u. |
| 691 kpc | 6.91 kpc | 69.1 pc | 0.207 pc | 427 a.u. |
| 1021 kpc | 10.2 kpc | 102 pc | 0.306 pc | 632 a.u. |
| 1734 kpc | 17.3 kpc | 173 pc | 0.52 pc | 1073 a.u. |
|
The radiative transfer equation accounts for the variation of the specific intensity under the effects of absorption, scattering and emission (cf. Steinacker et al., 2013, for a review in the case of a dusty medium). It is schematically represented in Fig. 3.2. This equation can be written:
(3.16) |
Eq. () can be rewritten, using the optical depth, , defined such that , or:
(3.17) |
Eq. () implies that . At a given wavelength, a medium is: (i) optically thin or transparent, if ; and (ii) optically thick or opaque, if . Replacing by in Eq. (), we obtain:
(3.18) |
where the source function, , includes all the terms added to the specific intensity:
(3.19) |
We will discuss exact numerical solutions to this equation in Sect. 3.1.1.4. For now, we discuss trivial solutions, when some processes are assumed negligible.
Propagation in vacuum. Let’s assume we have a star of radius, , and surface temperature, , located at . The flux at has to be integrated only over the hemisphere where a surface element of the star emits: . If there is no ISM around the star, Eq. () simply becomes . The solution is thus along the directions coming from the star and 0 in all other directions. At an arbitrary distance, , from the star, the solid angle it occupies is . The flux is thus , which is the classic dilution of the flux.
Emission only. Let’s assume we are observing, at submm wavelengths, a molecular cloud constituted of equilibrium grains at K, with opacity . At these wavelengths, the extinction is negligible. Eq. () is therefore simply within the cloud. The solution is thus , where is the position of the edge of the cloud along the direction of the sightline. If the density is constant, we get , which can be simplified as . This expression of the brightness is often used in radio-astronomy.
Absorption only. Let’s assume we are observing a background star through a cold molecular cloud, in the MIR. At these wavelengths, the albedo is close to 0 (Sect. 1.2.2.3). If we make the assumption that the background star is much brighter than the thermal emission of the cloud, Eq. () simply becomes . The solution is therefore along the direction coming from the star and 0 in all other directions.
Emission and absorption. We can merge the two previous cases. It could correspond to a hot molecular cloud, observed at MIR wavelengths. Its thermal emission is absorbed by the cloud itself. The solution of Eq. (), in this case, is:
(3.20) |
If the cloud contains a single grain species at temperature , the solution becomes:
(3.21) |
If we look in a direction away from the background star, we get the classical self-absorption formula:
(3.22) |
This solution is displayed in Fig. 3.3. Eq. () has the following two limit regimes.
Scattering and absorption with central illumination. Let’s assume that we have a homogeneous spherical cloud, of radius , and a central isotropically illuminating source, with monochromatic luminosity, . If is the monochromatic luminosity escaping the cloud, the escape fraction can be defined as:
(3.23) |
The second equality defines as the effective optical depth of the medium.
(3.24) |
(3.25) |
This formula simply subtracts among the few photons that may have interacted with the grains those which have been scattered. Contrary to Eq. (), it is valid for any value of .
Városi & Dwek (1999, hereafter VD99) have proposed an empirical approximation to interpolate the two regimes of Eq. () and Eq. (), resulting in the following escape fraction:
(3.26) |
with:
(3.27) |
They have benchmarked this approximation with a Monte-Carlo radiative transfer model (cf. Sect. 3.1.1.4). Both are in very good agreement in most of the astrophysically relevant parameter space. This solution is displayed in Fig. 3.4.a.
Scattering and absorption with uniform illumination. VD99 have also proposed an approximation in the case where the internal illumination of the cloud is uniform. They start from the escape probability of a homogeneous sphere with uniform illumination, without scattering, given by Osterbrock (1989):
(3.28) |
A demonstration of this formula is given in Appendix C of VD99. VD99 find, with a recursive argument, that the following expression is in relatively good agreement with a Monte-Carlo radiative transfer model:
(3.29) |
VD99 show that Eq. () perfectly agrees with the exact solution for a particular value of . The largest discrepancies, of about , are found at high optical depth, for close to 0. Eq. () is demonstrated in Fig. 3.4.b. We can see that, at a given , this geometry has an overall larger escape fraction than central illumination.
Scattering and absorption with external illumination. There are also expressions in the case of a homogeneous spherical cloud, externally illuminated by an isotropic radiation field. VD99 derive the following approximation for the absorbed fraction:
(3.30) |
The absorbed fraction is indeed more relevant in the case of external illumination, as it would be difficult to observationally separate the escaping radiation from the cloud and the ambient ISRF. On the contrary the absorbed fraction is meaningful if one wants to evaluate the heating of the cloud. This approximation is demonstrated in Fig. 3.4.c.
The three formulae for spherical clouds, given in Eq. (), Eq. () and Eq. (), do not allow us to model the internal heating of the cloud. These expressions are indeed global escape and absorbed fractions, but they do not account for the gradient of illumination within the cloud that would lead to a gradient of heating rate.
The ISM is a highly heterogeneous medium, with contrast densities of several orders of magnitude. A useful approximation is the clumpy medium, composed of: (i) a diffuse, uniform InterClump Medium (ICM), characterized by its density, ; and (ii) dense, spherical clumps, with density, , radius, , and volume filling factor, .
Effective optical depth of a clumpy medium. Let’s assume that we are looking at a background star through a cloud, and that the albedo of the grains is negligible. Along a given sightline, we have seen in Sect. 3.1.1.2 that the brightness in the direction of the star is . If the clumpy structure of the cloud is unresolved, the brightness measured in the telescope beam can be written as the sum of sightlines, some passing through the ICM, others through the clumps:
(3.31) |
This is the general expression. In the case of an homogeneous medium, Eq. () simplifies: , where is the optical depth of the homogeneous medium. In order to have the same dust mass and opacity as in the clumpy medium, we need to have:
(3.32) |
In the homogeneous medium, is indeed the mass surface density of a cloud of depth . If there is a statistical distribution of clumps in the beam, it must be identical in the clumpy medium. Thus, the brightness of the homogeneous medium is:
(3.33) |
Invoking the arithmetic-mean/geometric-mean inequality 1 (e.g. page 456 of Cauchy, 1821), we conclude that: . An important consequence of this result is that, from an observational point of view, we can miss a large mass of dust hidden in clumps. VD99 discuss this result in more detail.
The effective optical depth of a clumpy medium is always lower than that of a homogeneous medium with the same dust constitution and dust mass.
The mega-grains approximation. Neufeld (1991) proposed a simple approach to explain the leakage of Ly- photons by galaxies. He treated dusty gas clumps, in an empty ICM, as large grains with their own albedo and asymmetry parameter. This idea was then further developed by Hobson & Padman (1993), in cases where the ICM is non empty. They named it the mega-grains approximation, as clumps are treated as grains, although they have macroscopic sizes. They applied this approach to a clumpy infinite slab, externally illuminated on one side, and compared the results to a Monte-Carlo radiative transfer model. VD99 then refined some of the expressions of Hobson & Padman (1993) and applied them to the case of a spherical clumpy cloud, with the three types of illuminations we have discussed in Sect. 3.1.1.2: (i) central; (ii) uniform; and (iii) external. VD99 systematically benchmarked their results with a Monte-Carlo radiative transfer code. We briefly review their results in the rest of this section.
Escape fractions for a clumpy medium. VD99 derived a series of expressions, based on Eq. (), Eq. () and Eq. (), but replacing grain properties by effective mega-grains properties. These analytical approximations are all summarized in Sect. 5 of VD99. We demonstrate these analytical expressions for the three types of illuminations in Fig. 3.5. Overall, they provide a good agreement with Monte-Carlo radiative transfer calculations. They are also very easy to compute. The weakest point concerns the treatment of the grain heating. The mega-grains formalism allows us to separate the absorbed fractions in the clumps and in the ICM. It thus provides different heating rates in the two phases. In Fig. 3.5, we can clearly see that the clump emission (red) is significantly colder than that of the ICM (magenta). It however does not allow us to estimate the gradient of radiation field within the ICM and within clumps. This is the most dramatic in the case of central illumination. In order to obtain a more realistic SED for this particular case, VD99 used an ad hoc prescription, assuming a power-law distribution of equilibrium grain temperatures controlled by several tuning parameters depending on the grain type.
The radiative transfer equation (Eq. ) can be solved numerically. There are two main classes of methods (Steinacker et al., 2013, for a review).
MCRT is by far the most popular method. In this section, we briefly review its principle and apply it to an example.
Setting the model. To solve the radiative transfer equation (Eq. ), we need to specify the following physical ingredients.
The principle of Monte-Carlo radiative transfer. To compute a MCRT model, we need to draw a large number of photons (typically per wavelength bin), and execute the following steps. The procedure is schematically represented on Fig. 3.6.
These steps constitute the most basic implementation of MCRT. Numerous optimizations have however been proposed in the last fifty years (e.g. Witt, 1977a,b,c; Witt & Oshel, 1977; Yusef-Zadeh et al., 1984; Whitney & Hartmann, 1992; Wood, 1997; Wood & Jones, 1997; Városi & Dwek, 1999; Baes & Dejonghe, 2001; Gordon et al., 2001; Misselt et al., 2001; Steinacker et al., 2002, 2006; Wood et al., 2008; Baes et al., 2011; Camps & Baes, 2015; Siebenmorgen et al., 2015; Natale et al., 2015; Juvela, 2019). Improvements and optimizations include: (i) massive parallelization, and the use of GPU; (ii) the production of synthetic photometric images; (iii) the treatment of polarization by scattering.
Numerical method to randomly draw photons. To simulate the random walk of a photon, there are two sets of random variables to draw: (i) random interaction events; and (ii) scattering angles.
(3.34) |
For simplicity, corresponds to the path length along the direction of the photon, starting at . Drawing a variable from this distribution can be achieved the following way (e.g. Városi & Dwek, 1999). It uses the rejection method (cf. Appendix C.2.3.1). These steps are represented on Fig. 3.7.
(3.35) |
where is a uniform random variable between 0 and 1.
Example of a clumpy medium. For this manuscript, we have developed a MCRT model, following the procedure previously described. We have applied it to a spherical clumpy cloud, similar to those discussed in Sect. 3.1.1.3. The radius is pc, with an ICM density, cm. The pc clumps have a density, cm, with a filling factor, . We assume THEMIS grain constitution, and assume all grains are at thermal equilibrium. This cloud is centrally illuminated by a K star, with .
Despite its intensive numerical requirements, MCRT is thus the most flexible and accurate way to compute the SED of an interstellar cloud. It is however important to make sure that the spatial resolution of the density grid is fine enough to resolve the shortest mean free path of photons. Otherwise, we would be smearing out a potential sub-grid temperature gradient. For the present model, we have 0.01 pc cells for pc.
To accurately compute a radiative transfer model, it is necessary to resolve scale-lengths of the order of the mean free path of UV photons.
Ideally, each time we study a Galactic region or a galaxy, we should solve the radiative transfer equation (Eq. ). This is however, most of the time, impossible, because of the lack of constraints on the actual 3D structure of the region. Even if we have a collection of high-angular resolution multiwavelength images of our object, the matter and stellar distributions along the sightline have to be inferred. This inference is possible when the large-scale geometry of the object is quite regular, for instance: a protostellar disk and its jet, or a galactic disk and its bulge, etc. We will discuss the MCRT modeling of disk galaxies in Sect. 3.1.3.2. Otherwise, most often, we need to adopt empirical approaches that allows us to constrain the dust properties, despite our uncertainty of the spatial structure of the region.
The MBB (cf. Sect. 1.2.4.1) is historically the most widely-used dust SED model. It is controlled by the three following parameters (Eq. and Eq. ).
Its physical assumptions are simplistic: (i) the IR emission is optically thin; (ii) the dust is made of a single species of grains at thermal equilibrium with the ISRF; and (iii) the opacity is a power-law. By inferring both and , this model is designed to constrain both the dust excitation and its optical properties. This model was popularized by Hildebrand (1983), in the IRAS days (cf. Sect. 2.1.2). At the time, it was well adapted, being a simple, but still physical model, with only three parameters to fit four broadbands (the four IRAS bands at 12, 25, 60 and 100 ). It has however several limitations that are often disregarded in the literature.
The emissivity index derived from a single MBB fit is always lower than its true, intrinsic value.
A MBB fit infers parameters whose physical meaning is difficult to assess.
Dust models provide useful frameworks to model SEDs (Sect. 2.3). Without the possibility to compute the radiative transfer, we are however facing the problem of the mixing of physical conditions. A prescription, proposed by Dale et al. (2001), has proven to be a powerful solution to this issue. It consists in assuming that the dust mass is distributed in regions with different starlight intensities, , following a power-law:
(3.36) |
The idea is that the shape of the observed SED is used to constrain this distribution of ISRFs, assuming a dust mixture constitution. By lack of a better term, we call this approach the composite model. The free parameters are:
It thus provides a way to account for the potential complexity of the region without having to model the radiative transfer. The model SED is then simply:
(3.37) |
where is the monochromatic emissivity of the dust model, exposed to a single starlight intensity, (Fig. 2.24). Fig. 3.13.a shows an example of a SED fit, using Eq. (). We have a added a free-scaling black body, at K, to account for the stellar continuum that may be contaminating the MIR photometric bands. The composite approach is flexible enough to be usable in a diversity of environments. Dale et al. (2001) lists several simple geometries for which Eq. () is actually the solution. It is also adapted to more complex ISM topologies. For instance, Fig. 3.13.b shows the dust mass distribution as a function of , for each cell in the MCRT simulation of Sect. 3.1.1.4. Despite the complex, clumpy structure of this cloud, it can be reasonably well-approximated by a power-law (shown in red).
The average starlight intensity. The parameters of Eq. () do not have a very clear physical meaning. Besides, they are often degenerate: the uncertainties on the three parameters , and are strongly correlated. It can be more efficient to quote the average of the distribution:
(3.38) |
This parameter quantifies the average starlight intensity, heating the bulk of the dust mass. It is the equivalent of the equilibrium temperature of a MBB, as it controls the peak emission wavelength, except that it accounts for the mixing of physical conditions and the stochastic heating of small grains. The Total InfraRed (TIR) luminosity, , can be expressed:
(3.39) |
where the constant is the bolometric emissivity of the dust model. For the THEMIS mixture, heated by the Mathis et al. (1983) ISRF, it is (Table 2.7).
Constraining the dust properties. The composite approach of Eq. () allows us to constrain parameters that are not extremely sensitive to radiative transfer effects (i.e. to variations of the local intensity and spectral shape of the ISRF; cf. e.g. Galliano et al., 2018, 2021, for a discussion).
Eq. () provides acceptable estimates of , and (or ).
Limitations of the composite approach. Using Eq. () to model typical broadband SEDs of Galactic regions and nearby galaxies is an efficient method. However, as any approximation, it has some limitations.
An alternative distribution. The starlight distribution of Eq. () is not the only possible one. Draine & Li (2007) proposed the following:
(3.40) |
which simply is the linear combination of the power-law distribution of Eq. (), with a Dirac distribution centered at , fixing and . The power-law distribution is supposed to account for star-forming regions, with a large gradient of starlight intensity, and the Dirac represents the diffuse ISM, uniformly illuminated. The parameter controls the relative weight of these two components. This distribution was elaborated when Spitzer data were being analyzed. There was no coverage beyond . There was thus no constraint on the Rayleigh-Jeans regime of the SED. The role of the Dirac component was to fit the SED peak, avoiding the dust mass to explode by lack of constraint on the cold dust distribution. This parametrization however became problematic when Herschel data arrived. The FIR-submm slope of the observed SED can indeed not be fitted as well with this model (Eq. ) as with the composite approach (Eq. ). This is because the model’s slope is the intrinsic slope of the grain mixture. This is demonstrated in Fig. 3.16.a. In Fig. 3.16.b, we see that the starlight intensity distribution fit can not go down as low as the composite model. The Dirac component fits the FIR peak with a compromise . This is reminiscent of the discussion we had about MBB fits, in Sect. 3.1.2.1. This was demonstrated by G21, who compared different approaches by fitting the SEDs of about 800 galaxies of the DustPedia project (Davies et al., 2017) and Dwarf Galaxy Sample (DGS; Madden et al., 2013). Fig. 3.17 compares the composite approach, as a reference, to the following models (see the complete discussion in Galliano et al., 2021).
Several codes in the literature model the panchromatic 3 SED of galaxies, with a simplified treatment of the radiative transfer (e.g. Silva et al., 1998; Charlot & Fall, 2000; Galliano et al., 2008a; da Cunha et al., 2008; Boquien et al., 2019; Fioc & Rocca-Volmerange, 2019). They include the emission by stellar populations, in addition to dust SEDs.
Stellar SEDs. Stars of different masses have different spectra and lifetimes. This is shown in Fig. 3.18, in the form of a Hertzsprung-Russell diagram. Massive stars have: (i) the highest effective temperatures, their SEDs peaking in the UV; (ii) the highest luminosities; and (iii) the shortest lifetimes, of only a few million years. This is the opposite for low-mass stars: their SED peaks in the NIR, and they live longer than several hundred million years. These characteristics have a profound impact on the variation of stellar SEDs with time. The intrinsic emission of a stellar population can be simulated using evolutionary synthesis (e.g. Fioc & Rocca-Volmerange, 1997). This approach follows, at each time step, the formation and evolution of stars with different masses, . When these stars are born, all populations contribute to the SED. It is shown as the magenta curve in Fig. 3.19. This initial SED is dominated by massive stars and peaks in the UV. Then, as the most massive stars, which also have the shortest lifetime, die, their contribution to the SED is suppressed. Consequently, when these stars disappear, the UV-side of the SED decreases. After several hundred million years, the SED peaks in the NIR, as it originates only in low-mass stars (orange and red curves in Fig. 3.19). There is also a drastic evolution of the emissivity as a function of time, as low-mass stars are significantly less luminous (cf. Fig. 3.18). To compute such synthetic spectra, the following quantities need to be defined.
We will more extensively discuss these quantities in Sect. 4.3, when modeling the chemical evolution of galaxies. We have summarized in Table 3.2 the main properties of the different stellar classes.
| Effective temperature, T | Initial mass, | Initial luminosity, L | Lifetime, |
O | K | Myr | ||
B | K | 12 Myr1 Gyr | ||
A | K | Gyr | ||
F | K | Gyr | ||
G | K | Gyr | ||
K | K | … | ||
M | K | … | ||
|
Putting everything together. Empirical panchromatic SED models usually include the following physical ingredients.
We illustrate this approach with the model of Galliano et al. (2008a), applied to two galaxies, in Fig. 3.20. This particular model separates the emission from H II regions, which are powered by massive ionizing stars (cyan). The magenta curve shows the escaping radiation from H II regions. It includes the dust emission and the free-free continuum. The escaping radiation from H II regions, as well as non-ionizing field stars (yellow), heat the neutral ISM (red). The degeneracy between both dust components was solved by using radio observations to constrain the level of the free-free emission. The gas density in H II regions, impacting the equilibrium temperature of large grains, is constrained by the MIR continuum. The remaining emission is thus assumed to come from the neutral ISM. We will discuss in more detail the results of this model, in Sect. 4.3.
Limitations of empirical panchromatic models. The approach we have just described has several advantages. In a single fit, it provides estimates of the SFR, the age of the stellar populations, the stellar mass, and the dust properties. Its major limitation however resides in the sensitivity of its results to the assumed ISM topology. The ISM indeed has a fractal structure, with several orders of magnitude of contrast density (e.g. Combes, 2000). The optical depth from the model thus probably underestimates the total dust column density (cf. Sect. 3.1.1.3). In addition, the extinction and emission properties of these models are usually not consistent. The modeling of the microscopic grain constitution and of their macroscopic spatial distribution can differ from one side of the electromagnetic spectrum to the other. For instance, assuming a Calzetti et al. (1994) attenuation law and taking a mean radiation field to account for dust heating is virtually equivalent to decoupling the extinction and emission. Finally, the consistency brought by modeling all multi-wavelength tracers at once, which is a priori positive, leads to propagating the systematic uncertainties, due to arbitrary choices of ISM topologies, to parameters that could have been considered independent of these effects, if they had been modeled separately (, , ; Sect. 3.1.2.2).
The main limitation of dust studies lies in the near impossibility to constrain at the same time the microscopic dust properties and their sub-pixel macroscopic spatial distribution. All the approaches we have discussed in this chapter face this problem. It can be illustrated with what Galliano et al. (2018) called the matryoshka effect. This empirical effect originates in the impact of the spatial resolution of the observations on the constrained parameters.
Demonstration on the LMC. Fig. 3.21 demonstrates the effect with the modeling of the dust mass in a strip covering one fourth of the LMC, by Galliano et al. (2011). The different images on top show the spatial resolution that is used. The first image is one single large pixel. The second one is divided in four pixels, and so on. The curve in the bottom panel of Fig. 3.21 shows the dust mass in the strip derived by summing all the pixels, at each resolution. We see that this mass increases with spatial resolution, until reaching a plateau around pc. For this particular region, the discrepancy with the global mass is about . Galliano et al. (2011) interpreted this effect by noticing that the cold dust, which accounts for most of the mass and, at the same time, is the least luminous, is diluted into the warm dust emission when we sum all the regions together. With spatial resolution however, we can better separate the bright and cold regions. It is thus reasonable to assume that the most accurate estimate of the dust mass is the one obtained at the highest resolution. This assumption is confirmed by the fact that the length-scale at the growth curve plateau ( pc) corresponds roughly to the mean free-path of a U-band photon at a density cm (Table 3.1; the LMC has a half-Solar metallicity; Pagel, 2003). This is the typical density of the CNM and diffuse H2 phase (cf. Table 3.6). It is possible that, if we could increase the resolution, we would see another plateau around cm ( pc), corresponding to dense H2 clouds.
Generalization. This effect has been independently confirmed by Galametz et al. (2012), Roman-Duval et al. (2014) and Aniano et al. (2020), although it is less important if the maximum spatial resolution is not as high as ours.
The dust mass derived at high spatial resolution is always larger than its global estimate.
We now review the application of SED models to observations of nearby galaxies, aimed at constraining the grain properties in different environments. We illustrate the different aspects using our own projects and collaborations.
There is a wide diversity of galaxy types. Several systems of classification have been developed, through the years. In particular, the Hubble-de Vaucouleurs system, although outdated, is still used nowadays.
The outdated galaxy morphological classification. The most famous observational system of morphological classification is the Hubble tuning fork or Hubble-de Vaucouleurs diagram, represented in Fig. 3.22. It was originally developed by Hubble (1936), and refined by de Vaucouleurs (1959). It is based on the morphological characteristics of galaxies in the visible range: presence and thickness of spiral arms, bars, rings, etc. There are essentially three classes of galaxies (left, center and right parts of Fig. 3.22): (i) ellipticals, also called Early-Type Galaxies (ETG); (ii) spirals, also called Late-Type Galaxies (LTG); and (iii) irregulars and dwarf spheroidals. There are sub-categories with abstruse notations (SAa, E2, etc.) that would be a waste of time to describe here. Overall, this is a mid-XX-century empirical classification, based on stellar dynamics, that does not take into account the IR information (especially the SF activity), nor the radio properties (presence of an AGN). Recently, with the DustPedia collaboration (Davies et al., 2017), we explored the sensitivity of dust properties to galaxy types (e.g. Davies et al., 2019; Bianchi et al., 2018; Nersesian et al., 2019). We did not find any clear systematic differences between adjacent sub-categories in Fig. 3.22, and we found a large scatter within each class. There are overall trends between the three main categories, because they are linked to the gas fraction, metallicity and stellar populations. We will discuss those throughout this manuscript. The terminology “late/early-type” was introduced to denote the evolution of galaxies along the sequence. We now know that the sequence is reversed: early-type galaxies are the oldest objects, and late-types, the youngest ones. In addition, the most numerous galaxies in the local Universe, dwarf galaxies are under-represented in this diagram. They are part of the irregulars, which is a category by default. Finally, this Hubble-de Vaucouleurs classification was based on the local Universe, while galaxies at high redshift can exhibit different morphologies, such as clumpy chains and tadpoles (e.g. Elmegreen, 2015). We have developed an alternate, non-parametric classification, taking into account IR morphology (Baes et al., 2020). It emphasizes the clumpy nature of the dust distribution in local galaxies.
Stellar morphology is not particularly relevant to ISM studies.
Galactic properties that matter to ISM studies. A few global quantities, such as the metallicity, the SFR or the gas fraction are more relevant to assess the general properties of the ISM of a galaxy. The three main categories of the Hubble-de Vaucouleurs diagram can be characterized the following way.
(a) Irregular, dwarf | (b) Spiral | (c) Elliptical |
(I Zw 18) | (M 33) | (CentaurusA) |
Low metallicity () | Solar metallicity () | High metallicity () |
The dust spatial distribution, that is the value of various dust parameters in different regions or pixels of a galaxy, can be determined in the MW and nearby galaxies. This determination however requires good quality, homogenized multi-wavelength images of the studied objects. This is therefore significantly more complex than modeling the SED of a point source.
Homogeneous multi-wavelength data sets. To accurately model SEDs of galaxies, the observed fluxes must originate, at each wavelength, in the same region, and must trace only the emission we are modeling (i.e. the dust emission, the escaping stellar emission, etc.). The following artifacts can be encountered.
(3.41) |
(), where and are the flux and uncertainty coming from the data reduction pipeline. The random variable, , is independent, normal with mean 0 and standard deviation 1.
Calibration uncertainties can be computed afterwards, as they are proportional to the flux. These uncertainties are fully correlated between pixels and partially correlated between wavelengths (e.g. Galliano et al., 2021).
This technical data preparation can be tedious, but it is crucial as dust model results directly rely on it. A significant effort has been put into providing homogenized databases of nearby galaxies. Among them, the most important surveys are the following.
Properties of individual galaxies. Numerous studies have presented the SED modeling of nearby galaxies, and their derived dust properties, either globally or spatially-resolved. We have participated to several such projects (e.g. Whaley et al., 2009; Galametz et al., 2009; Boselli et al., 2010; Galametz et al., 2010; O’Halloran et al., 2010; Eales et al., 2010; Cortese et al., 2010; Sauvage et al., 2010; Bendo et al., 2010; Roussel et al., 2010; Gordon et al., 2010; Davies et al., 2010; Boquien et al., 2010; Skibba et al., 2012; De Looze et al., 2012a; Galametz et al., 2013; Ciesla et al., 2014; Gordon et al., 2014; Galametz et al., 2016; Bianchi et al., 2018; Nersesian et al., 2019). Reviewing them would be unwieldy, here. In general, these studies provide dust parameters (mass, starlight intensity, PAH mass fraction, etc.) of different objects, which can be used in combination with other tracers to refine our understanding of the studied source. They also provide scaling relations and calibrations of various diagnostics such as SFR tracers. These results can also be used to train machine-learning models that could predict the SED of a poorly-observed galaxy (e.g. Dobbels et al., 2020).
Identifying dust heating sources. A particular question, that has been tackled by several studies, is the identification of the sources responsible for dust heating within galaxies. In the MW, 3D reconstruction of the ISM distribution showed that the heating by young, O/B stars (Table 3.2) was prominent in molecular regions, whereas the atomic phase was mainly heated by lower-mass stars (e.g. Sodroski et al., 1997; Paladini et al., 2007). In nearby galaxies, this depends on the SF activity of the galaxy. For instance, we showed that PAHs were essentially heated by field stars in the quiescent galaxy NGC 2403. These molecules are however heated by the escaping radiation from H II regions in the more actively star-forming object, M 83 (Jones et al., 2015). More generally, with the DustPedia sample, we found that dust in ETGs was mainly heated by old stars (Nersesian et al., 2019). It is only when considering more gas-rich galaxies that the contribution of young stars becomes more important. It can account for up to of the dust luminosity in extreme late-type galaxies (Sm–Irr, Fig. 3.22; Nersesian et al., 2019). These different heating sources have an impact on the global escape fraction (i.e. the fraction of stellar radiation leaving the galaxy unattenuated). Massive stars being embedded in molecular cocoons, they have a lower escape fraction than Low- and Intermediate-Mass Stars (LIMS) which occupy lower density regions. In the DustPedia sample, we showed that the escape fraction was on average , with mild variations across galaxy types (Bianchi et al., 2018). It is slightly lower in LTGs (). We emphasize that this nearby galaxy sample lacks the deeply enshrouded star formation of LIRGs and ULIRGs, where the global escape fraction can drop down to (e.g. Clements et al., 1996). The question of the dust heating contribution can now be tackled with more accuracy using 3D MCRT models.
Large-scale radiative transfer models of galaxies. Applying a 3D MCRT model to reproduce the spatial flux distribution of galaxies, in all wavebands, is not straightforward. Indeed, the observations provide only 2D projected constraints. This is why most studies favor edge-on galaxies, as the images of such objects provide constraints on both the radial and azimuthal distributions, assuming axisymmetry (Fig. 3.25). Several studies have modeled the effect of extinction on the optical data of disk galaxies using such codes (e.g. Xilouris et al., 1999; Alton et al., 2004; Bianchi, 2007). They were able to answer the recurring question about the optical thickness of disk galaxies (Disney et al., 1989). In particular, Xilouris et al. (1999) found that the face-on optical depth of typical spiral galaxies is less than one, in all optical bands. Concerning dust heating, recent progress has been made, especially by the DustPedia collaboration, using the MCRT code SKIRT (Baes & Camps, 2015).
A model such as SKIRT can also be used to model the radiative transfer in simulations of galaxies (e.g. Trčka et al., 2020). Finally, these models account for the energy balance between the escaping UV-visible light and the re-emitted IR-submm radiation. Interestingly enough, several studies report a deficit of modeled FIR emission by a factor , compared to the observations (Alton et al., 2000, 2004; Dasyra et al., 2005; De Looze et al., 2012a,b). This discrepancy is thought to result from a lack of detail in modeling the geometry. In particular, the presence of young stars, deeply embedded in molecular clouds, at sub-grid resolutions, could compensate for this deficit without significantly altering the extinction (cf. e.g. Baes et al., 2010).
Dust masses derived from SED fits directly depend on the assumed grain opacity. Using the MBB parametrization (Eq. ), both the scaling, , and the emissivity index, , are important. There are particular situations, where we can reverse the process and constrain these two parameters:
Studies of the emissivity index. There are numerous publications presenting MBB fits of nearby galaxies. However, as discussed in Sect. 3.1.2.1, the derived emissivity index, , is degenerate with temperature mixing. The best constraints on the intrinsic are obtained in the submm regime, where only massive amounts of very cold dust ( K) could bias the value. Table 3.3 lists effective emissivity indices, , for several objects, obtained with Planck, with constraints up to . It appears that all the values are lower than 2, and that low-metallicity systems have a lower than higher metallicity galaxies. Boselli et al. (2012), studying a volume-limited sample with Herschel (up to ), also found an average , and hinted that low-metallicity objects tend to have . In M 33, derived from Herschel observations is around 2 in the center and decreases down to 1.3 in the outer parts (Tabatabaei et al., 2014). On the other hand, the outer regions of M 31 exhibit a steeper slope () than in its center (Draine et al., 2014). This contradictory behaviour does not appear to originate in fit biases, as both increasing and decreasing trends of with radius are found in the sample of Hunt et al. (2015).
Grain opacity in the LMC. An important result, that has often been misunderstood, concerns the grain opacity in the LMC. Galliano et al. (2011), modeling a strip covering one fourth of the LMC (cf. Fig. 3.21), with the composite approach (Sect. 3.1.2.2), found that the dustiness distribution in this galaxy was most of the time larger than the maximum value it could in principle reach. This maximum value is set by the elemental abundances. The fraction of elements locked-up in grains can indeed not be larger than the amount available in the ISM. The metallicity of the LMC is (Pagel, 2003). The maximum dustiness of the LMC is thus . The dust mixture that was used is an update of the Zubko et al. (2004, BARE-GR-S). It is essentially based on the Draine & Li (2007) optical properties, a pre-Herschel model. This discrepancy is shown in Fig. 3.26 (the red histogram). The only explanation is that this grain mixture is not emissive enough to account for the observed FIR-submm emission. We thus proposed an alternate dust model, simply replacing graphite by amorphous carbons (the ACAR sample of Zubko et al., 1996), without altering the total carbon fraction. This simple modification boosts the emissivity by a factor 6 . With this new model, most of the dustiness distribution is centered around its expected value, and is clear from the forbidden range (yellow). It was called the AC model (blue histogram in Fig. 3.26). The tail of the distribution in the forbidden zone originates in cold regions, where the uncertainty is large. The conclusion of this modeling was that LMC grains must be a factor more emissive than the Draine & Li (2007) model. We actually presented a preliminary version of this result during Herschel’s science demonstration phase (Meixner et al., 2010).
Confirmation in other systems. The conclusion of the Galliano et al. (2011) study was that grains in the LMC had to be more emissive than in the MW, because the Draine & Li (2007) model was at the time consistent with the MW. This last statement however happened to be inexact. Planck Collaboration et al. (2016c) modeled the all sky dust emission, using also the Draine & Li (2007) model. The estimated along the sightlines of Quasi-Stellar Objects (QSO) was systematically lower than their dust-emission-derived . Their comparison of emission and extinction thus indicates that the Galactic opacity is in fact also a factor of higher than previously assumed. In addition, in M 31, Dalcanton et al. (2015) derived a high spatial resolution map of . As in the Galaxy, the emission-derived map (Draine et al., 2014) was found to be a factor of higher. We emphasize that, although each of these studies found evidence of local variations of the emissivity as a function of the density (cf. Sect. 4.2.1.1), the overall opacity seems to be scaled up compared to Draine & Li (2007). In other words, in all the environments where enough data is available to constrain , it is found a factor of higher than the original Draine & Li (2007) properties. Dust models therefore need to use an opacity consistent with these constraints. This is the case of the THEMIS model. Its FIR-submm opacity is very close to our AC model (cf. Fig. 4 of Galliano et al., 2018).
It is reasonable to adopt the THEMIS grain opacity (cf. Sect. 2.3.3), when modeling galaxies.
The opacity in nearby galaxies. The DustPedia collaboration conducted several studies aimed at constraining the grain opacity in nearby galaxies. First, Bianchi et al. (2019) studied the actual emissivity of 204 late-type galaxies, that is the ratio of IR emission to H column density. We found an emissivity MJy/sr/( H/cm), consistent with the MW, except for the hottest sources. These estimates were derived using global fluxes, integrated over the whole galaxy. In parallel, Clark et al. (2019) modeled in details the two face-on galaxies, M 74 and M 83. We could map the grain opacity. This was done by converting metallicity maps into oxygen depletion maps, and comparing those to the dust mass. The derived opacities were quoted at : m/kg in M 74, and m/kg in M 83. These values are consistent with the Herschel-Planck-revised opacities of the THEMIS model ( m/kg; cf. Fig. 2.26).
We have seen in Sect. 3.1.2.2 that there is a degeneracy between the grain size and starlight intensity distributions. This degeneracy arises from the fact that it is observationally difficult to distinguish the MIR emission of a hot region from the MIR emission of small grains. In the early 2000s, we did not know it was impossible, so we did it (Galliano et al., 2003, 2005). We modeled the SED of the following four BCDs: NGC 1569 (), II Zw 40 (), He 2-10 ()and NGC 1140 (), to infer their size distribution. We interpreted these results in light of shock processing.
Grain processing by shock waves. Shock waves from SNe process dust grains, while sweeping the ISM.
Fig. 3.27 shows the model of Jones et al. (1996) for graphite and silicates. In both panels, the grey curve is the initial, MRN size distribution. The color curves show the size distribution obtained after the mixture has been swept by a shock of velocity, . The main effect of the blast wave is to fragment and shatter grains, turning large grains into smaller grains. The qualitative effect is similar for both compositions. This is best seen at low velocity (blue curves). Large grains are depleted and there is an excess of small grains. At higher velocity, the distribution tends toward a log-normal centered around nm (magenta curve). Vaporization also leads to a net loss of dust mass. This model has since then been refined by Bocchio et al. (2014), who applied it to THEMIS grains. We will discuss more extensively dust processing by SN blast waves in Chap. 4.
The modeling strategy. Galliano et al. (2003, 2005) modeled the UV-to-mm global SED of the four BCDs, using the DBP90 dust model, the stellar evolutionary synthesis code PÉGASE (Fioc & Rocca-Volmerange, 1997, Fig. 3.19), and the photoionization code, CLOUDY (Ferland et al., 1998). The modeling scheme was the following.
This model is self-consistent per se. It however avoids the degeneracy between size and ISRF distributions by assuming a simple geometry (the shell). The dust is thus uniformly illuminated in this model, which is an unrealistic assumption for a star-forming galaxy. It is therefore likely that some of the emission we have attributed to small grains originates in hot regions. The results Galliano et al. (2003, 2005) obtained, that we will discuss in the following paragraphs, are nevertheless qualitatively consistent with the properties we expect in these environments. It is possible that only a fraction of the MIR emission originates in compact H II regions, not excluding an overabundance of small grains.
The grain size distribution in four dwarf galaxies. The inferred size distribution, in the four BCDs, are shown in Fig. 3.28. We display the three components of the DBP90 model: PAHs, VSGs and BGs (cf. Sect. 2.1.2.5). The most striking features, common to the four objects, are the following.
Two studies used a similar approach, assuming uniform illumination, and fitting the SED varying the size distribution of the DBP90 model, in NGC 1569 (Lisenfeld et al., 2002) and in the LMC (Paradis et al., 2009). They also concluded to an overabundance of small grains.
Consequence on the extinction curves. We have briefly mentioned in Sect. 2.1.2.2 that the extinction curves in the Magellanic clouds were systematically different from the MW. Fig. 3.29.a compares the extinction towards different sightlines in the LMC and SMC, to the range of extinction curves in the MW. We see that the LMC () is on average (red curve) similar to the MW. However, toward the massive star-forming region 30 Doradus (LMC2 supershell; green curve), it is steeper, with a weaker 2175 Å bump. When we go to the SMC (; blue curve), the difference is more pronounced: the extinction curve has a very steep UV-rise and lacks the 2175 Å bump. The origin of these variations are still debated. Our four BCDs brought an interesting perspective on this open question. The extinction curves derived from the size distributions of Fig. 3.28 are shown in Fig. 3.29.b. We can see that they are systematically steeper than the average MW (). They also have a 2175 Å that is either weaker (NGC 1569, He 2-10 and NGC 1140) or similar (II Zw 40) to the MW 7. In other words, these extinction curves lie between the LMC and SMC, consistent with the metallicity range of these BCDs. The modeling of Galliano et al. (2003, 2005) therefore provides a coherent view of the grain properties in these environments, where shock waves have an instrumental role in shaping the grain sizes.
Sect. 3.1 was devoted to modeling the whole IR SED, which is necessary to estimate the total dust content of galaxies. There are however several other important properties that can be self-consistently studied by focussing on a particular wavelength range.
Mid-IR spectra have been extensively observed since the first light of ISO. Spitzer and AKARI have extended our knowledge of this spectral range and the JWST will likely revolutionize it.
Until now, we have discussed PAHs from a general point of view, and how to estimate their mass fraction. We now focus on the information that the analysis of their detailed MIR emission spectrum can bring. In what follows, we interchangeably use the terms UIBs, aromatic features and PAH bands. The only case where these terms are not equivalent is when discussing the UIBs around 3 , coming from a mixture of aromatic and aliphatic features, that can not be solely attributed to PAHs.
MIR spectra of galaxies. Fig. 3.30 illustrates the diversity of MIR spectra encountered in different environments. It shows three extreme galaxies from the Hu et al., in prep. sample.
Laboratory and theoretical PAH physics. Although we do not know the exact composition of the interstellar carbon grain mixture responsible for the aromatic feature emission, the brightest bands have been attributed to the main vibrational modes of PAHs. There are still some debates about the origin of the weakest features (cf. e.g. Allamandola et al., 1999; Verstraete et al., 2001; Tielens, 2008; Boersma et al., 2010; Jones et al., 2013). In Fig. 3.30, we have labeled the different bands with a given mode. These modes are schematically represented in Fig. 3.31.a.
The charge and size of the PAHs are the main parameters controlling their emission spectrum.
To study variations of the aromatic feature spectrum with environmental conditions, one needs to measure the intensity of each band. This task is not as simple as it appears. There are indeed several challenges.
MIR spectral decomposition methods are therefore an essential tool to properly study PAH features.
Calibrating feature properties. Since the properties of the different UIBs are not a priori known, we need to empirically infer them. In our work, we parametrize the band spectral profile with a split-Lorentz function (Hu et al., in prep.):
(3.42) |
where is the central frequency of the feature, and and are its widths on the short- and long-wavelength sides. Having an asymmetric feature is indeed necessary to accurately fit good quality spectra. This asymmetry may originate in the anharmonicity of the transition responsible for the band, or may be due to unresolved blended features. The parameters characterizing each individual features, , and , could be derived from each fit. However, most of them would be quite uncertain, using an average Spitzer spectrum. For that reason, we have calibrated these parameters (i.e. inferred their reference value), using high-resolution, high-signal-to-noise-ratio spectra of Galactic regions. This calibration is demonstrated in Fig. 3.33.
Table 3.4 gives the resulting band parameters. With these parameters fixed, we can fit even low-signal-to-noise-ratio spectra varying only the intensity of each band (parameter in Eq. ).
|
|
| Type |
3.291 | 0.020 | 0.019 | Main |
3.399 | 0.011 | 0.024 | Main |
3.499 | 0.077 | 0.071 | Small |
5.239 | 0.025 | 0.058 | Small |
5.644 | 0.040 | 0.080 | Small |
5.749 | 0.040 | 0.080 | Small |
6.011 | 0.040 | 0.067 | Small |
6.203 | 0.031 | 0.060 | Main |
6.267 | 0.037 | 0.116 | Main |
6.627 | 0.120 | 0.120 | Small |
6.855 | 0.080 | 0.080 | Small |
7.079 | 0.080 | 0.080 | Small |
7.600 | 0.480 | 0.502 | Plateau |
7.617 | 0.119 | 0.145 | Main |
7.870 | 0.170 | 0.245 | Main |
8.362 | 0.016 | 0.016 | Small |
8.620 | 0.183 | 0.133 | Main |
9.525 | 0.107 | 0.600 | Small |
10.707 | 0.100 | 0.100 | Small |
11.038 | 0.027 | 0.073 | Small |
11.238 | 0.053 | 0.153 | Main |
11.400 | 0.720 | 0.637 | Plateau |
11.796 | 0.021 | 0.021 | Small |
11.950 | 0.080 | 0.222 | Small |
12.627 | 0.200 | 0.095 | Main |
12.761 | 0.081 | 0.140 | Main |
13.559 | 0.160 | 0.161 | Small |
14.257 | 0.152 | 0.059 | Small |
15.893 | 0.178 | 0.200 | Small |
16.483 | 0.100 | 0.059 | Small |
17.083 | 0.496 | 0.562 | Plateau |
17.428 | 0.100 | 0.100 | Small |
17.771 | 0.031 | 0.075 | Small |
18.925 | 0.037 | 0.116 | Small |
|
Fitting every feature at once. The total MIR spectrum contains the emission from different physical processes that have to be separated in order to accurately measure UIB intensities (cf. Fig. 3.30). Several models have tackled this problem since the ISO days (e.g. Verstraete et al., 1996; Boulanger et al., 1998; Laurent et al., 2000; Madden et al., 2006; Smith et al., 2007; Galliano et al., 2008b; Mori et al., 2012; Lai et al., 2020).
Fig. 3.34.a demonstrates such a fitting method on the total spectrum of M 82 (Galliano et al., 2008b). It is labeled as the Lorentzian method, as the UIBs are modeled with Lorentz profiles. This earlier model did not use all the bands given in Table 3.4, that we are now using.
Central wavelength | Species | Transition | Type |
4.052 | H I | Brackett | recombination |
5.511 | H2 | 0–0 S(7) | ro-vibrational |
5.908 | H I | Humphreys | recombination |
6.109 | H2 | 0–0 S(6) | ro-vibrational |
6.910 | H2 | 0–0 S(5) | ro-vibrational |
6.985 | Ar II |
2P–2P | forbidden |
7.460 | H I | Pfund | recombination |
7.502 | H I | Humphreys | recombination |
8.025 | H2 | 0–0 S(4) | ro-vibrational |
8.991 | Ar III |
3P–3P | forbidden |
9.665 | H2 | 0–0 S(3) | ro-vibrational |
10.511 | S IV |
2P–2P | forbidden |
12.279 | H2 | 0–0 S(2) | ro-vibrational |
12.369 | H I | Humphreys | recombination |
12.814 | Ne II |
2P–2P | forbidden |
15.555 | Ne III |
3P–3P | forbidden |
17.035 | H2 | 0–0 S(1) | ro-vibrational |
18.713 | S III |
3P–3P | forbidden |
21.829 | Ar III |
3P-3P | forbidden |
25.890 | O IV |
2P–2P | forbidden |
28.219 | H2 | 0–0 S(0) | ro-vibrational |
33.481 | S III |
3P–3P | forbidden |
34.815 | Si II |
2P–2P | forbidden |
35.349 | Fe II |
6D–6D | forbidden |
36.014 | Ne III |
3P–3P | forbidden |
|
Alternative methods. Several other MIR spectral fitting methods have been dicussed in the literature. The two following ones are worth mentioning.
The advantages of this method are that: (i) fits are not extremely degenerate, even at low signal-to-noise ratios, because of the small number of free parameters, compared to the Lorentzian method; (ii) the four different classes are physically meaningful, providing a clear interpretation of the results. However, its lack of flexibility prevents accurate fits, that could overlook new information present in the observations.
Applying a spectral decomposition method to a set of MIR spectra allows us to study the variations of several band ratios that contain physical information about the small carbon grain properties.
Observed band ratios in galaxies. Fig. 3.36 demonstrates the diversity of spectra among galaxies (panel a) or within one (panel b). This figure emphasizes the differences in terms of aromatic band intensity. All these spectra are normalized by the 11.3- feature intensity. Yet, they exhibit large variations of their 6-to-9- features. This can be more precisely quantified by studying the correlation between specific band ratios, such as in Fig. 3.37 (Galliano et al., 2008b). The quantity is simply the intensity of the feature centered at . The bottom two panels show the results for integrated galaxies and Galactic regions, whereas the bottom two panels show the results for a few spatially-resolved sources. Overall the trends are similar for both types. It means that this is a multiscale relation, valid at sub-pc scales (within the Orion bar or M 17) and kpc scales (among integrated galaxies). These three displayed band ratios span about an order of magnitude, and are linearly correlated with each other. It implies that the 6.2, 7.7 and 8.6 features are tied together, while the 11.3 can vary independently. This is what was illustrated in Fig. 3.36. The only parameter that can explain such a variation is the charge of the PAHs (cf. Fig. 3.32).
The variation of the PAH charge can explain most of the UIB variations observed in the nearby Universe.
Effects of ionization and size. Although ionization is the main driver of the UIB relative variations in galaxies, other effects can play a role in specific environments. The most important one of these secondary effects is the PAH size distribution. Fig. 3.38 shows numerical simulations of several key UIB ratios (Hu et al., in prep.). We have varied: (i) the minimum PAH size expressed in number of C atoms, , highlighted in Fig. 3.38.a; (ii) the PAH charge fraction, , highlighted in Fig. 3.38.b; (iii) the ISRF intensity and hardness. To illustrate the last effect, we have computed the model grid for the Solar neighborhood ISRF (Mathis et al., 1983). This values are the bright grid points. We have also computed the grid for a hot star spectrum, with (the faint grid points). The effect of the ISRF is non negligible, but it is less drastic than that of the charge and size. Most astrophysically relevant ISRFs will be intermediate between the two extreme cases we have displayed. Fig. 3.38 illustrates the following points (see also Rigopoulou et al., 2021, for a more complete calculation using Density Functional Theory).
Such band ratio diagrams have been used to demonstrate systematic variations of the PAH size distribution in various environments. We studied the spatial variations of in NGC 1097 (Fig. 3.30) and showed it was systematically lower in the central region, close to the AGN (Wu et al., 2018b). The most likely explanation is that the hard radiation field from the central engine is selectively destroying the smallest PAHs. This effect was also shown by Smith et al. (2007) and Sales et al. (2010), on global scales.
The particular case of low-metallicity environments. In low-metallicity systems, band ratio variations can be more difficult to probe, as the band equivalent widths are lower, and thus more uncertain (cf. Sect. 3.2.1.4). In the LMC, Mori et al. (2012) found different trends in neutral and ionized sightlines. Toward the latter, there are evidences that PAHs have a lower charge (as a consequence of the higher recombination rate) and are on average larger (due to the destruction of the smallest PAHs). In contrast, in the SMC, Sandstrom et al. (2012) found very weak ratios and weak 8.6 and 17 bands, implying small weakly ionized PAHs. This last point is consistent with the trend of with found by Smith et al. (2007). This was also noted by Galliano et al. (2008b), who found that low-metallicity systems tends to lie on average toward the lower left corner of Fig. 3.37, whereas the upper left corner is essentially populated by Solar-metallicity sources. However, Hunt et al. (2010) argued that BCDs exhibit a deficit of small PAHs. If there is a smooth variation of PAH size distribution with metallicity, these results are in contradiction. Sandstrom et al. (2012) noted that these BCDs are more extreme environments than the SMC, and that photodestruction could dominate the PAH processing (cf. Sect. 4.2.2.1). We note that the solution to this apparent controversy might alternatively reside in the difference in studied spatial scales. In the Magellanic Clouds, Spitzer spectroscopy gives a spatial resolution of a few parsecs, compared to a few hundred in nearby BCDs. The fact is that the LMC and SMC exhibit strong spatial variations of their UIB spectrum. Whelan et al. (2013) showed a diversity of MIR spectral properties in the SMC. In this study, we demonstrated that the PAH emission in a region like N66 is dominated by its diffuse component, and not by its bright clumps, where PAHs are destroyed. At the other extreme, the molecular cloud SMC-B1#1 shows unusually high UIB equivalent widths (Reach et al., 2000). Also, the ratio indicates that PAHs are more compact in 30 Doradus and more irregular outside (Vermeij et al., 2002). All these elements suggest that there is a complex balance of processes shaping the MIR spectra throughout low-metallicity environments.
UIBs as diagnostics of the physical conditions. The fact that ionization dominates the UIB variation in galaxies opens the possibility to use specific observed band ratios to quantify the physical conditions. The charge of an ensemble of molecules is indeed the balance between: (i) the ionizing photon rate; and (ii) the electronic recombination rate. The first quantity is usually quantified by the variable , defined as the integral of the ISRF in the FUV (e.g. Hollenbach & Tielens, 1997):
(3.43) |
The recombination rate is roughly proportional to ( being the electron density, and , the gas temperature; e.g. de Jong, 1977). The ratio of these two rates, often called the photoionization parameter, therefore quantifies this equilibrium (e.g. Chap. 5 of Tielens, 2005):
(3.44) |
The electron density can be related to the total H density by considering that most electrons in the neutral gas come from the photoionization of C. We thus have (cf. Sect. 2.2.3.1). Galliano et al. (2008b) measured the ratio in Galactic regions where , and had been reliably estimated (Fig. 3.39.a). It allowed us to propose an empirical relation between and :
(3.45) |
In other words, measuring provides an estimate of . With such a relation, the diagrams of Fig. 3.37 can now be turned into the diagnostics of Fig. 3.40. This relation has, since then, been refined by several studies, especially Boersma et al. (2016). Although unique, the diagnostics of Eq. () has however the following limitations.
Finally, the calibration of Eq. () can also be estimated theoretically. Fig. 3.39.b shows the theoretical ratio as a function of and (Galliano, 2009). It has been derived by computing the stochastic emission of PAHs within the PDR model of Kaufman et al. (2006). Such model indeed computes the charge balance of PAHs at each point within the cloud, where and are known. One grid point value corresponds to a whole cloud.
Other properties. We end this section by briefly reviewing other properties that can be studied with observations of UIBs.
The evolution of the shape of the UIB spectrum, probed by studying band ratio variations, is not the only diagnostics of the small carbon grain properties. The overall aromatic feature strength, relative to the continuum (i.e. to the emission of the rest of the dust populations) shows drastic variations across environments (cf. Fig. 3.30). These variations trace the evolution of the mass fraction of their carriers – PAH or small a-C(:H).
Effect of ISRF hardness. PAH and small a-C(:H) are known to be sensitive to hard an intense radiation fields. They tend to evaporate near massive stars, and can be assumed to be fully depleted in H II regions (e.g. Cesarsky et al., 1996b; Galametz et al., 2013; Galliano et al., 2018). This effect can be quantified by studying the variation of the aromatic feature strength with a tracer of the ISRF hardness.
Such a trend is shown on Fig. 3.41, with the Madden et al. (2006) results 8. It clearly indicates that PAHs are less abundant in regions permeated with a hard ISRF, at all spatial scales down to pc. Note however that both tracers do not come from the same physical region: (i) PAH being destroyed in H II regions are tracing the neutral gas; (ii) [Ne III] and [Ne II] obviously come from the ionized gas. The correlation therefore reflects the mixing of phases within the beam.
Effect of metallicity. The effects of ISRF and metallicity are often degenerate, as at low metallicity: (i) stars of a given mass have a systematically higher effective temperature, because of line blanketing effects; (ii) the ISM is less opaque, because of the lower dustiness, and thus more permeated by UV radiation. The highest [Ne III]/[Ne II] ratios are found in BCDs, as well as the lowest I(PAH)/I(cont). Fig. 3.42.a shows the evolution of I(PAH)/I(cont) as a function of metallicity (Madden et al., 2006). There is a paucity of PAHs in low-metallicity environments. The questions is whether this is the result of their increased destruction, or if they have not been produced. We will come back to this question, when discussing dust evolution in Sect. 4.2.2.1.
PAHs are under-abundant in low-metallicity environments.
The absence of metallicity threshold. The relation of Madden et al. (2006) was the first spectroscopy-based demonstration of the effect. Shortly before, Engelbracht et al. (2005) showed the broadband correlation of as a function of metallicity. They showed both quantities were clearly correlated. They however argued there were essentially two populations, below and above . Galliano et al. (2008a) showed that this was a bias due to the saturation of IRAC as a PAH tracer at low metallicity. When the aromatic feature strength becomes indeed low, is not anymore a measure of I(PAH)/I(cont), but is a measure of the temperature of the continuum. This is illustrated on Fig. 3.42.b. We show that when the actual mass fraction of small a-C(:H) drops below , becomes insensitive to its value.
At long wavelengths, in the submm-to-cm range, dust emission does not always behave as the extrapolation from the Rayleigh-Jeans law (). Two peculiar phenomena have been widely discussed in the literature: (i) the submillimeter excess; and (ii) the Anomalous Microwave Emission (AME).
An excess emission above the modeled dust continuum is often observed, longward . The most significant reports of this submm excess can not be accounted for by: (i) thermal dust emission; (ii) free-free and synchrotron continua; and (iii) molecular line emission (cf. Fig. 3.43; Galliano et al., 2003).
Studying this excess is important, as: (i) it could bias dust mass estimates; (ii) it potentially contains untapped physical information about the ISM.
Possible explanations. The origin of the submm excess is currently debated. The following explanations have been proposed. These different scenarios are not exclusive.
Empirical properties of the excess. Since the origin of the submm excess is still unknown, most studies focus on characterizing its observed properties.
The submm excess is more prominent in low-metallicity environments, and in diffuse regions.
Reality of the phenomenon. The reality of the submm excess has been questioned for the two following reasons. We try to address these criticisms in order to support its likeliness.
As we have seen in Sect. 2.2.2.3, the AME is a centimeter continuum excess that can not be accounted for by: (i) the extrapolation of dust thermal emission; (ii) molecular line emission; and (iii) free-free and synchrotron continua (Fig. 2.10). Its SED looks like a bump peaking around cm (cf. Fig. 2.10). It is commonly attributed to the dipole emission of fastly rotating, small dust grains. The faster grains rotate, the shorter the frequency of the emission peak is.
The AME in extragalactic environments. In nearby galaxies, the first unambiguous detection of an AME has been obtained in an outer region of NGC 6946 (Murphy et al., 2010; Scaife et al., 2010). Follow up observations showed evidence for AME in eight regions of this galaxy (Hensley et al., 2015). This study showed that the spectral shape of this AME is consistent with spinning dust, but with a stronger AME-to-PAH-surface-density ratio, hinting that other grains could be the carriers. Overall, the AME fraction is highly variable, in nearby galaxies. Peel et al. (2011) put upper limits on the AME in M 82, NGC 253 and NGC 4945. These upper limits suggest that AME/100 is lower than in the Milky Way, in these objects. In M 31, Planck Collaboration et al. (2015a) report a measurement of the AME, consistent with the Galactic properties. Finally, Bot et al. (2010), fitting the NIR-to-radio SED of the LMC and SMC, tentatively explained the submm/mm excess with the help of spinning dust, in combination with a modified submm dust emissivity (cf. Sect. 3.2.2.1). They concluded that, if spinning grains are responsible for this excess, their emission must peak at 139 GHz (LMC) and 160 GHz (SMC), implying large ISRF intensities and densities. Draine & Hensley (2012) argued that such fastly rotating grains would need a PDR phase with a total luminosity more than two orders of magnitude brighter than the SMC.
Controversy about the carriers of the AME. Although PAHs have been considered the most likely carriers of the AME, Hensley & Draine (2017) argued that nanosilicates could be a viable alternative. They showed that nanosilicates can indeed account for the AME, without violating the other observational constraints (depletions, emission, extinction; cf. Sect. 2.2). This claim relies on the earlier findings of Hensley et al. (2016), showing that AME correlates better with dust radiance, 10, in the MW. Hensley et al. (2016) also found some fluctuations in the PAH emission relative to the AME intensity, traced by WISE. The fact is that there is no observational evidence of nanosilicates in the ISM. In particular, these grains would emit the 9.8 and 18 features, as they would be stochastically heated. These bands would eventually remain diluted in the aromatic feature emission, provided the abundance of nanosilicates is low enough. We can however note that in H II regions, where PAHs are destroyed, we can see the underlying continuum, and silicate features in emission are rarely present (e.g. Peeters et al., 2002b). They can be seen only in the most compact H II regions, and their 9.8-to-18 ratio indicates that they originate in large equilibrium grains. A weak correlation does not always indicate an absence of causality. This issue might reside in the fact that Hensley et al. (2016) used the WISE band as a tracer of PAH intensity, whereas this broad photometric band also encompasses a significant fraction of the underlying continuum, and is biased towards neutral PAHs. We have tried to address this controversy by refining the constraints on the PAH emission.
Correlation with charged PAHs. Bell et al. (2019) focussed on the 10-wide Galactic molecular ring surrounding the O-type star, -Orionis (cf. Fig. 3.45). We chose this region, because the Planck data indicate there is a large gradient of AME intensity. We fitted the spatially-resolved SED, at angular sampling, using the AKARI 9 and IRAS 12 bands to constrain the PAH abundance, and longer wavelength bands for the rest of the emission. We used the dust SED code HerBIE (HiERarchical Bayesian Inference for dust Emission; Galliano, 2018, cf. Sect. 5.3.3). We were able to address the controversy, thanks to the combination of: (i) rigorous SED fitting, allowing us to account for all the available information, not only a few broadbands; (ii) better observational constraint on the PAH emission (9 and 12 ); and (iii) focussing on a clean region rather than the whole sky. Our results are shown in Fig. 3.46. We have found very good pixel-to-pixel correlations between the AME intensity, derived by Planck Collaboration et al. (2016a), and the dust and PAHs surface densities from SED fitting (e.g. cf. Fig. 3.46.a). Our analysis however show that, if the dust mass per pixel is very well correlated with the intensity of AME per , (), the correlation is better with the mass of PAHs, and even better with the mass of ionized PAHs (; cf. Fig. 3.46.b). Our Bayesian results also show that there is a 0 probability that the total dust could correlate better with AME than with PAHs. Our impression is that the study of Hensley et al. (2016) may have too quickly interpreted a poor correlation as an absence of causality. The scatter in WISE was likely not the result of an intrinsic scatter of the PAH/AME correlation, but rather a combination of observational artefacts: (i) noise; (ii) contribution of the continuum and ionic lines. Our SED model and our better MIR coverage allowed us to more accurately recover the true PAH column density.
AME correlates better with charged PAHs.
We end this chapter with a short introduction to ISMology, since the knowledge of the ISM as a whole is crucial to understanding ISD. We discuss a few of our results in this domain and illustrate the way dust can be used to better understand the gas. The video lectures and accompanying slides of the 2021 “ISM of galaxies” summer school (35 hours of lecture in total), that we have organized, can provide a good introduction to this subject. Otherwise, the books of Spitzer (1978), Tielens (2005) and Draine (2011) are references.
The ISM is intrinsically a multi-phase medium, with large contrast densities and differences in temperatures. The order of magnitude of its average density is cm. We list the physical characteristics of the main phases in Table 3.6. We will discuss each phase individually in the rest of this section.
The cooling function. The way the gas cools across phases has a decisive impact on the multiphase structure of the ISM. It is possible to estimate its cooling rate as a function of temperature, or, in other words, how the thermal energy of the gas is converted into radiation. This quantity is called the cooling function. It is represented on Fig. 3.47. We have highlighted the dominant processes in the different temperature regimes.
The exact shape of this cooling function can vary sensibly. It depends on: (i) metallicity, as this parameter impacts the relative abundances of the different species; and (ii) the radiation field, which impacts the ionization state of the gas.
The neutral atomic gas is the most abundant phase in the MW: it accounts for of the total ISM mass, and of the total baryonic mass (stars and ISM). It fills up about of the MW volume. The neutral gas is far from thermal equilibrium, but it is roughly in pressure equilibrium, with:
(3.46) |
The photoelectric heating. In neutral regions, the direct heating of the gas by absorption of stellar UV photons is not efficient, because only a small fraction of these photons can be absorbed through the different available atomic lines. In these regions, the heating of the gas is indirect. Dust grains absorb much more efficiently UV photons, with their spectrally continuous cross-section. The absorption of an energetic photon (of a few eV) can lead to the ejection of an electron, via the photoelectric effect. This electron will then collide with the gas and heat it. This is the photoelectric heating of the gas (de Jong, 1977; Draine, 1978; Bakes & Tielens, 1994; Weingartner & Draine, 2001b; Kimura, 2016). We have schematically represented this process in Fig. 3.48. The overall efficiency of this process is related to the integrated surface of dust grains. According to Table 2.3, this surface is dominated by small grains. The smallest grains, especially PAHs, are therefore responsible for most of this heating. Wolfire et al. (1995) give an expression for the photoelectric heating rate:
(3.47) |
where is the efficiency of conversion of FUV energy into gas heating (it is a few percents), and and have been defined in Eqs. (-).
The two stable neutral atomic phases. For simplicity, let’s assume the heating of the neutral ISM is assured only by photoelectric heating. Let’s also use Eq. () with a fixed value of the photoelectric heating efficiency, , and . The equilibrium is obtained when cooling and heating are balanced:
(3.48) |
which becomes, using Eq. ():
(3.49) |
This is the black line we have represented in Fig. 3.49. There are three pressure equilibrium positions (the three dots), at the value of Eq. (). We have hatched in grey the regime corresponding to unstable solutions. In this regime, the pressure indeed decreases with density. Thus, a small pressure increase, above the green dot, will decrease the density. It will thus make the gas expand and its temperature increase, moving further away from the green dot. On the opposite, a small decrease of the pressure, below the green dot, will increase the gas density, and make its temperature decrease, at the same time. The gas will thus collapse and move further away from the green dot. The two only stable solutions correspond to the two main neutral atomic phases of the ISM.
Both of these phases are observed with H Iline, and UV-optical absorption lines.
There are two distinct neutral atomic phases in pressure equilibrium: the WNM and the CNM.
The ionized gas accounts for a moderate mass fraction of the ISM in the MW, but fills up a large volume. It is of the ISM mass and of the total baryonic mass (stars and ISM).
The Hot Ionized Medium (HIM). The HIM is a coronal gas (, K; Table 3.6) filling up of the MW volume. It is at pressure equilibrium with the WNM and CNM. McKee & Ostriker (1977) showed that this phase was heated by SN shockwaves. It is composed of overlapping SN bubbles, and is sometimes referred to as the intercloud medium. This gas, which pervades everywhere, can be found significantly above the Galactic disk, contrary to the other phase, which are contained to the disk. It is observed through FUV absorption lines of ions, and diffuse soft X-ray and synchrotron emissions.
H II regions. H II regions are short-lived dense ionized regions around OB star associations. The gas is ionized by photons from the massive stars. These regions are not in equilibrium, they are expanding. Strömgren (1939) has estimated the radius of an homogeneous sphere of ionized gas, around a star of H-ionizing photon rate, . Such a sphere is called a Strömgren sphere. Its radius, , can be estimated by balancing photoionization and recombination:
(3.50) |
where is the case B recombination rate. The product is the electronic recombination rate to any H level :
(3.51) |
Recombination down to will indeed produce an ionizing photon that will be absorbed by another H atom. Rearranging Eq. (), Strömgren’s radius is thus:
(3.52) |
For an O5 star, with , pc. This estimate can be refined, accounting for He and other atomic species, as well as dust screening (e.g. Osterbrock, 1989, for an extensive lecture).
The Warm Ionized Medium (WIM). The WIM is a diffuse phase (, K; Table 3.6) filling up about of the MW volume. It is roughly in pressure equilibrium with the other thermal phases, although it can be found expanding in some regions. This gas is photoionized by photons from OB star associations, escaping from H II regions. The electrons ejected by this photoionization provide the main heating source. It is observed through optical lines, essentially H, as well as some fine-structure lines and free-free emission.
The molecular gas constitutes only a moderate fraction of the ISM mass and a small fraction of its volume. It is however crucial, as it is where stars are formed. In the MW, of the ISM mass is molecular, which corresponds to of the total baryonic mass (stars and ISM).
Molecular hydrogen formation. The formation of the most abundant molecule of the Universe, H2, is inefficient in the gas phase. This is because the freshly formed molecule needs to evacuate 4.5 eV to remain stable. Yet, due to its symmetry, H2 does not have rotational transitions that would allow it to radiate at these energy levels. H2 formation thus takes place on grain surfaces (e.g. Bron et al., 2014, for a review). We have represented the two main processes on Fig. 3.50: the so-called Langmuir-Hinshelwood (physisorption) and Eley-Rideal (chemisorption) mechanisms. In both cases, the grain serves as a catalyst and helps evacuate the formation energy when the newly formed molecule is released in the gas phase. Similarly to the photoelectric heating, this process happening on grain surfaces, it takes place primarily on small grains. Temperature fluctuations therefore play an important role in its efficiency (e.g. Le Bourlot et al., 2012).
The diffuse molecular gas. Molecular gas can be observed at moderate densities (; Table 3.6). This is often associated with the CNM with large enough column densities to allow H2 to be self-shielded (i.e. its UV electronic lines are optically thick). It is also heated by photoelectric effect and cosmic rays. It cools primarily via [C II] and can be observed through UV absorption lines.
Photodissociation regions. Same as H II regions, PDRs are a phase defined by the presence of massive stars in their vicinity. They are not a stable phase of the ISM, but they are very important since they are the interface between the H II region and the molecular cloud (e.g. Hollenbach & Tielens, 1997, for a review). Because of their high density and their high , they absorb a significant fraction of the FUV radiation by OB stars and reradiate it in the IR, as dust thermal emission and fine structure lines. Since they are the regions where molecules are being dissociated by FUV photons and subsequently recombine, they are the place of many important chemical reactions. Fig. 3.51 shows the structure of a typical PDR. We have performed an isobaric run with the Meudon PDR code (Le Petit et al., 2006), for and . We show the variation of the abundances of the main species, and we represent in the upper part the H I/H2 and C II/C I/CO transitions. This figure demonstrates that H2, being efficiently self-shielded, exists at lower A(V), whereas CO appears deeper into the cloud.
Dense molecular clouds. Dense molecular clouds (; Table 3.6) contain most of the molecular gas, concentrated in a small volume. These molecular clouds are gravitationally bound. Their structure is clumpy and filamentary. The gas motions are controlled by turbulence. They can be arranged in molecular complexes of sizes up to pc (e.g. Solomon et al., 1987). The densest cores are collapsing and will lead to star formation. One of the most challenging issue of ISMology is the difficulty to measure the mass of molecular clouds. As we have mentioned earlier, H2 is a symmetric molecule. It thus does not have any dipole moment allowing rotational transitions. Its first transitions are its rovibrational levels (H extsubscript2,0-0, H extsubscript2,0-0 and so on) that need temperatures of a few hundred K to be pumped. Cold molecular clouds are thus primarily traced by the second most abundant molecule, CO, which is asymmetric. CO rotational lines, CO(J10) and CO(J21), are the most commonly observed transitions. The conversion of an observed CO(J10) line intensity, , to a H2 column density requires the knowledge of a CO-to-H2 conversion factor, , such that (e.g. Bolatto et al., 2013):
(3.53) |
The factor has been calibrated on Galactic molecular clouds: . This value is however metallicity dependent, as we will see in Sect. 3.3.2.3.
We have already discussed the potential of dust as a tracer of the physical conditions in the ISM, especially in Sect. 3.2.1.3. We review here a few examples where dust tracers were used to refine the results of studies dedicated to star formation or gas physics.
Star formation rates. Dust-related SFR tracers rely on the fact that young stars are extremely luminous and are enshrouded with dust. If the clouds are optically thick and if their covering factor is unity, the OB star luminosity is: . Contrary to a common misconception, this is independent of dust properties. Assuming a typical IMF, burst age and metallicity, can be converted to: (e.g. Kennicutt, 1998a). The contribution of old stars to is negligible for high enough SFRs. Alternatively, monochromatic SFR indicators have been proposed. Calzetti et al. (2007) and Li et al. (2010) found that the 24 and 70 monochromatic luminosities were good local SFR indicators (on spatial scales of kpc): and . We have discussed the use of aromatic features as SFR tracers in Sect. 3.2.1.3. Finally, several composite indicators have been calibrated (Hao et al., 2011). By combining FUV or H measurements with the 24 or TIR indicators, they account for the fact that star-forming regions are not completely opaque.
Resolving star formation. Hony et al. (2015) performed a comparison of different SFR estimators with the actual stellar content of the star-forming region N66, in the SMC. In this study, we derived the stellar surface density, , based on individual star counts from HST photometry. When compared to the dust mass surface density, , derived from SED modeling, we found a significant scatter, at pc linear resolution. The SFRs derived from or H underestimate the more reliable, -derived SFR, by up to a factor of . This is likely due to ionizing photons escaping the region 11. Finally, converting our map to a gas mass surface density map, , we found that the pixels of our region are lying above the Schmidt-Kennicutt relation 12. This might be due to low density gas, inefficient at forming stars, that is included in global Schmidt-Kennicutt relations, but absent when zooming on star-forming regions.
PDR Properties. We have participated to numerous studies aiming at constraining the PDR properties in low-metallicity environments (Cormier et al., 2010, 2012; Lebouteiller et al., 2012; Cormier et al., 2015; Chevance et al., 2016; Lebouteiller et al., 2017; Wu et al., 2018a; Cormier et al., 2019; Lebouteiller et al., 2019). The common point of these studies is their use of numerous fine structure lines observed by Herschel and Spitzer, as well as the dust emission, to constrain the main parameters of a PDR model (, , filling factor). The challenge lies in the multiple degeneracies, due to the fact that a given line can trace several phases. For instance, [C II] comes from the ionized gas, the neutral gas and from an important fraction of molecular clouds (Fig. 3.51). Such a degeneracy can be solved by using additional lines to constrain the properties of these different phases. Dust tracers are also useful, either as a constraint or as a self-consistency test. As an example, Chevance et al. (2016) modeled the gas properties in 30 Doradus and derived the typical depth of PDRs, in terms of visual extinction magnitude, . Fig. 3.52 shows the comparison of to the value of this parameter, derived from SED modeling, . We can see that both quantities are in good agreement, validating the PDR results.
(a) Three color image of 30 Doradus | (b) estimates |
Photoelectric heating. Assuming that [C II] is the main gas coolant, the photoelectric efficiency, that we already discussed in Sect. 3.3.1.1, , can be approximated by the gas-to-dust cooling ratio: . Detailed studies usually add other lines to the gas cooling rate, like [O I], to have a more complete estimate (e.g. Lebouteiller et al., 2012; Cormier et al., 2015; Lebouteiller et al., 2019). Overall, Smith et al. (2017) found that , with an average of , in a sample of 54 nearby galaxies. It appears that is lower when the dust temperature is higher (Rubin et al., 2009; Croxall et al., 2012). This is not likely the result of the destruction of the UIB carriers, as their intensity correlates the best with the [C II] emission (e.g. Helou et al., 2001). It is rather conjectured to be due to the saturation of grain charging in UV-bright regions. The shape of the ISRF also has a consequence on the accuracy with which represents the true UV, photoelectric-efficient flux. Indeed, Kapala et al. (2017) showed that the variations of in M 31 could be explained by the contribution of old stars to . Finally, one of the most puzzling features is that is higher at low metallicity (Poglitsch et al., 1995; Madden et al., 1997; Cormier et al., 2015; Smith et al., 2017; Madden et al., 2020), while the UIB strength drops in these systems (Sect. 4.2.2.1). This is currently poorly understood. However, in the extreme case of I Zw 18 (), where no UIB is detected (e.g. Wu et al., 2006) and the photoelectric heating is estimated to be negligible, the gas-cooling-to-TIR ratio is still (Lebouteiller et al., 2017). In this instance, we have shown the gas could be heated by X-rays.
The dark gas. We have discussed in Sect. 3.3.1.3 that the photodissociation of H2 and CO at different depths into molecular clouds leads to biases in gas mass estimates. H2 is indeed self-shielded. It therefore exists at column densities roughly independent of metallicity. On the contrary, CO, which is significantly less abundant, is not self-shielded (i.e. its electronic lines remain optically thin at large column densities). Consequently, CO needs to be shielded by dust to survive. It thus exists only deeper into molecular clouds. The H2 gas that exists in regions where CO is photodissociated is often referred to as the CO-dark molecular gas. Other tracers can be used to constrain this dark gas: (i) dust emission (e.g. Israel, 1997; Leroy et al., 2011); (ii) [C II] (e.g. Madden et al., 1997); (iii) and -rays (e.g. Grenier et al., 2005). Using to derive with Eq. () can therefore be biased if the dark gas fraction is significantly larger than in the MW, where has been calibrated. This is what happens in low-metallicity systems, where the dustiness is lower, because of the lower abundance of heavy elements (cf. Sect. 4.3.1). This is represented in Fig. 3.53. We see that in the low-metallicity cloud (on the right), CO cores are much smaller, because of the increased photodissociation of this molecule, compared to the left cloud (Solar metallicity). It results that the mass of CO-dark H2 is significantly larger at low metallicity. We have investigated this ISM component in the LMC, using dust mass surface density, concluding it could account between and of the total molecular gas mass (Galliano et al., 2011). Recently, we studied dark gas in the star-forming region N11 of the LMC, modeling the full set of IR emission lines (Lebouteiller et al., 2019). We showed that most of the molecular gas in this region is CO-dark and that [C II] traces mostly this component. We extended this analysis to a sample of nearby dwarf galaxies (Madden et al., 2020). We found that of the molecular gas mass is not traced by CO(J10).
The molecular gas content of low-metallicity systems is dominated by dark gas.
Pressure and radiation field. The pressure in molecular clouds can be significantly larger than in the pressure equilibrium phases of the ISM: in the Orion bar (Goicoechea et al., 2016); compared to in the HIM, WIM and CNM (Eq. ). We have studied the physical conditions of the molecular gas in the central region of the starbursting galaxy, M 83 (Wu et al., 2015). We used the CO Spectral Line Energy Distribution (SLED) observed by Herschel to estimate its column density and pressure, and , in different regions. We have also performed SED modeling to estimate the mean starlight intensity heating the grains, (Eq. ). This allowed us to show that both quantities are correlated (cf. Fig. 3.54.a). We also noted that the pressure gradient was oriented along a chain of radio sources, corresponding to a radio jet (Fig. 3.54.b). We derived a similar correlation between the ISRF strength and the gas thermal pressure in the Carina nebula (Wu et al., 2018a). Such a correlation was also found by Joblin et al. (2018) in the Orion bar. They argue that the photoevaporation of the PDR can explain this relation.
(a) Pressure/ISRF relation | (b) SFR map of M 83 |
1.Stating that , provided that , .
2.VD99 solved this issue by assuming a power-law distribution of temperatures that we have discussed in Sect. 3.1.1.3. This is however an ad hoc solution that needs to be calibrated for each dust species, by using an actual MCRT model. Our goal being to demonstrate what a MCRT model brings, we chose to compare it only to the SED that the mega-grains approximation allows us to derive, by itself.
3.Literally for all wavelengths, from the X-rays to the radio.
4.Attenuation is not equivalent to extinction. The extinction is the sum of scattering and absorption along the sightline toward a point source, whereas the attenuation is a global account of the reddening of an ensemble of stars, potentially mixed with the dust. The extinction is directly linked to dust properties and is independent of geometry. The attenuation is a synthetic quantity depending both on the dust properties and on the ISM topology (e.g. Buat et al., 2019, for a recent discussion).
5.The Holmberg radius is the radius within which the B-band surface brightness of the galaxy is 26.5 magnitudes per squared arcsecond.
6.This factor is not precise, as the slope of the opacity is also changed. The difference in emissivity is thus not a sole scaling.
7.We note that the strength of the bump is controlled by the carbon-to-silicate grain ratio, a parameter that is poorly constrained by the SED fit.
8.The results of Madden et al. (2006) consisted in a first spectral decomposition of ISO/CAM spectra, that we have refined in Galliano et al. (2008b), and added Spitzer/IRS spectra to the sample.
9.-FeO is the notation to design ferromagnetic FeO, called maghemite. It must be distinguished from its non-magnetic form, noted -FeO, called hematite. Both have distinct crystalline structures: cubic lattice for maghemite; trigonal crystals for hematite (e.g. Lȩcznar, 1977).
10.The radiance is the spectral integral of the specific intensity: .
11.We saw a similar discrepancy in the center of M 83, where the [N II] line was significantly more extended than other SFR tracers (Wu et al., 2015).
12.The Schmitt-Kennicutt relation is the empirical correlation between and for a wide diversity of galaxies (Schmidt, 1959; Kennicutt, 1998b).