Chapter 6
Conclusion and Prospective

6.1 What Have We Learned About ISD in the Past Decade?

We have seen in Sect. 5.3.1 that ISD studies were particular because of the inherent complexity of the dust make-up and of the low weight of evidence provided by individual observables. This sometimes leads colleagues from other fields to think that our subject is messy and stalling. If we however integrate over a large number of studies, we can delineate some clear breakthroughs. I try the exercise of listing them here ¹. This is of course a subjective account of the progress, biased by my own interests. The observations acquired by Herschel and Planck played an instrumental role in these.

6.1.1 About Dust Properties

Grain opacity. Herschel and Planck have brought invaluable information about the FIR and submm optical properties of dust grains, in the MW and nearby galaxies (cf. Sect. 3.1.3.3). Before that, the commonly-used opacities were a factor of $\simeq 2-3$ lower, biasing the dust masses one would estimate. We still do not know the exact constitution of interstellar grains and its evolution, but we know relatively well the zero level FIR-submm opacity they should have in the diffuse ISM of the MW and a few nearby galaxies.

Scaling relations. The large number of galaxies observed by Herschel and the effort to build homogeneous ancillary data samples have provided consistent estimates of the dust, stellar and gaseous properties for a large number of objects, with spatial resolution in numerous cases (cf. Sects. 4.2 – 4.3). These relations provide snapshots of galaxy evolution at different stages. They are now well sampled over most of the parameter space (metallicity, gas fraction and specific star formation rate). They are the main benchmarks dust evolution models must account for. The most important ones and the information they convey are the following.

• The dustiness-metallicity relation exhibits two distinct regimes of dust production: (i) at very low metallicity, dominated by stardust; and (ii) at high metallicity, dominated by grain growth in the ISM.
• The dustiness-gas-fraction relation links the gas depletion timescale and the evolution probed by the dustiness-metallicity relation.
• Scaling relations linking the dust content to the diffuse X-ray emission are promising tools to constrain grain sputtering timescales.

Dust properties of low-metallicity systems. I have emphasized in Sect. 4.3 that low-metallicity systems (i.e. dwarf galaxies) were crucial to constrain dust evolution models, because they sample a different grain production regime than higher metallicity objects. Herschel has provided us with the first FIR SEDs of extremely low-metallicity galaxies that were decisive in understanding the early stages of dust evolution.

The Submillimeter excess. Twenty years after its discovery the submm excess is still a mysterious epiphenomenon (cf. Sect. 3.2.2.1). In the last years: (i) several studies have hinted that it could be more prominent in diffuse regions of galaxies; and (ii) a new physical process that could explain its origin has been proposed (magnetic grains).

The AME. The AME has, for the first time, been detected in extragalactic systems (cf. Sect. 3.2.2.2). To our humble opinion, the debate about its origin (PAHs or nanosilicates) is closed, in favor of PAHs.

Stoichiometry and grain structure. Important progress has been made constraining the grain structure and stoichiometry using X-ray edge absorption. The results are sometimes difficult to understand, such as the high crystalline fraction discussed in Sect. 2.2.1.3. The technics are however promising and will revolutionize our understanding of the dust constitution, when ATHENA will be observing.

Dust models. A few dust models have been published in the last decade. The THEMIS model (cf. Sect. 2.3.2.2) is, in our opinion, the most innovative for the following reasons.

• It is designed as an evolution model. Some fundamental properties of the constitution of the grain mixture (a-C(:H) hydrogenation, size distribution, mantle thickness) can be varied to account for different observables. This way, we have a few physically-grounded parameters to empirically explore the effects of dust evolution. It is an important progress over classic, static models.
• It is laboratory-data based.
• The model can be adapted to account for polarization constraints.
• It is the only recent model accounting for the 3.4 $\mu m$ aliphatic feature.
• It is one of the only models to be consistent with the revised FIR-submm opacity that I have mentioned earlier.

Polarization. Whole-sky submm polarized emission maps have been produced by Planck, at several wavelengths (cf. Sect. 2.2.2.2). On top of permitting unprecedented studies of the magnetic field, they provide evidence that the bulk of the large grain emission is homogeneous in size and composition.

Laboratory data. Long-wavelength (FIR-submm) measures of a diversity of dust analogues have been produced (cf. Sect. 2.2.4.3). They provide the necessary data that, when consistently included in dust models, will help us to: (i) better constrain the evolution of grain mantles; and (ii) characterize more precisely the submm excess.

6.1.2 About Dust Evolution

Dust sources. Our understanding of the dust production mechanisms has considerably progressed, thanks to Herschel (cf. Sects. 4.2 – 4.3).

The modeling of scaling relations,

in particular the dustiness-metallicity relation, provides clear evidence that: (i) dust growth in the ISM is the prominent grain production regime around Solar metallicity; and (ii) condensation in SN II ejecta dominates at very low metallicity.

Observations of individual SNRs

have shown that large amounts of dust could be produced shortly after SN II explosions. It suggests that a large fraction of these freshly-formed grains must be destroyed by the reverse shock.

Evolution of the Aromatic Feature Carriers. Before the 2010s, ISO and Spitzer were crucial to understand the variations of the strength of the UIBs (cf. Sect. 3.2.1.1). In the 2010s, the detailed modeling of the FIR SED permitted by Herschel, allowed us to understand more finely how the abundance of the grains carrying these features evolves. To our mind, the important points are that: (i) their mass fraction is better correlated with metallicity than with the strength of the ISRF; and (ii) they are spatially associated with molecular clouds.

Thermal sputtering. ETGs, due to the paucity of their ISM, had been poorly studied before Herschel. We have now been able to characterize their dust content. It appears that these environments exhibit a dust deficit due to grain destruction in their permeating coronal gas. They are thus potentially interesting laboratories to constrain sputtering timescales (cf. Sect. 4.2.2.2).

Emissivity Variations in the ISM. The good coverage and sensitivity of Herschel and Planck allowed us to better characterize the way the FIR-submm emissivity evolves from the diffuse ISM to dense regions, in the MW and the Magellanic clouds (cf. Sect. 4.2). It is now clear that the increase of emissivity with ISM density, resulting from mantle accretion and grain coagulation, is a universal process.

Distant objects. The dust content of numerous galaxies at very high redshifts ($z>6$) has been constrained, thanks to ALMA. It appears that dust-rich objects existed only a few 100 Myrs after reionization, requiring fast grain build-up. In our opinion, this can be explained with rapid dust growth in the ISM (cf. Sect. 4.3).

6.2 What Are the Open Questions for the Next Decade?

I now try to delineate a few open questions that should occupy us during the next decade. The list below is as subjective as Sect. 6.1. This is not everything we should do, but rather everything we will be able to do, knowing the available observing facilities.

6.2.1 Extragalactic Dust

Studies of diffuse dust in nearby galaxies. The diffuse ISM of the MW is the only medium used to constrain dust models, because it provides simultaneous information about extinction, emission and elemental depletions (cf. Sect. 2.3). The combination of these different constraints is crucial to solve the degeneracies between emissivity and size distribution. This is why dust models are calibrated on these data, and why we do not yet have reliable dust models for the SMC, for instance. Such observations are however available, in a fragmented way, in external systems such as the Magellanic clouds and M 31. An effort should be made to produce an homogeneous data set of similar constraints in a few external galaxies and to build dust models using it. This implies several challenges.

• The extraction of the emission from the diffuse ISM is required to make sure we have homogeneous physical conditions and are not biased by the variation of the FIR opacity. The large beam of Planck renders this task difficult in external galaxies, but can be compensated by using other observations (e.g. Herschel and ground-based submm data).
• Additional observations are probably needed to cover the missing information. Facilities such as the Multi Unit Spectroscopic Explorer (MUSE;  Bacon et al., 2010, for abundances) and ALMA (for the submm continuum) provide exceptional data that have not yet been utilized to their full potential in our field.

Having different extragalactic dust models would allow us to: (i) understand how the diffuse dust properties vary as a function of metallicity; and (ii) have a more robust way to study external galaxies, by using models that take into account the effect of cosmic dust evolution on the grain mixture.

Quiescent low-metallicity galaxies. I have emphasized several times the crucial role low-metallicity systems play in constraining dust evolution. These objects are however faint and we usually observe those which are actively star-forming. We therefore suffer from a selection effect that hides the nature of low-metallicity quiescent systems. I have given an example of what such systems could bring to our understanding of the evolution of MIR features in Fig. 4.22. The question is in which quadrant of this figure they will fall? Obtaining JWST spectra of these objects would probably be a game changer.

Circumgalactic dust. Grains present in the immediate vicinity of galaxies, either in the infalling or outflowing gas, are currently poorly known. Yet, infall and outflow might be an important mechanism regulating the dustiness of galaxies. JWST and ALMA observations might be able to characterize circumgalactic grain properties beyond the nearby Universe, because of their resolving power. For local objects, NIKA2 is currently acquiring mm maps of infalls and outflows in nearby galaxies.

6.2.2 Dust Evolution Modeling

Local evolution modeling. I have emphasized that a dust model such as THEMIS provides a unique framework to model SEDs, taking into account dust evolution. Its current limitation is however that we lack a quantitative link between the evolution parameters (a-C(:H) hydrogenation, size distribution and mantle thickness) and the environmental conditions (density and ISRF intensity). This can be achieved by empirically calibrating the tuning parameters of the evolution processes discussed in Sect. 4.2. The goal would be to have a reliable dust model predicting the constitution of the grain mixture as a function of $n_{\mathrm{gas}}$ and $G_0$.

Cosmic dust evolution models. The empirical modeling of cosmic dust evolution, that was the center of Sect. 4.3, calls for several improvements that could greatly change our understanding of the matter.

1.

If we want to be able to precisely constrain the grain growth and SN II blast-wave destruction timescales, we need the most accurate possible stellar elemental and dust yields. This is an effort asked to the circumstellar community, both modelers and observers.

2.

We need to adopt a more consistent approach between the different physical elements involved (dust, gas and stellar emissions and evolutions; cf. Sect. 5.3.3.3).

3.

When modeling SEDs at scales of several tens of parsecs or larger, we need to properly account for the mixing of physical conditions. This is currently done phenomenologically (cf. Sect. 3.1.2.2). Ideally, we should move toward fitting models accounting for the statistical distribution of dust and stars, with a wide range of topologies, accounting for: (i) dust evolution in the ISM, as a function of $n_{\mathrm{gas}}$ and $G_0$, as I have mentioned earlier; and (ii) radiative transfer.

6.2.3 Dusty Epiphenomena

Long-wavelength properties. The current submm-to-cm ground-based observatories (such as NIKA2, ALMA, etc.) open windows to progress on our understanding of the submm excess (cf. Sect. 3.2.2.1) and of the AME (cf. Sect. 3.2.2.2). By combining these new observations with archival Spitzer, AKARI and Herschel data, we should be able to systematically test the different possible scenarios. In addition, this analysis should be performed with dust models including state-of-the-art submm-mm laboratory opacities of interstellar grain analogues, in order to provide a reliable baseline.

DIBs. DIBs have been extensively observed with Gaia. Since we do not know their nature, unbiased exploration of how their properties vary with all available ISM tracers should be performed, in a big data way.

6.2.4 The Need for a New FIR Observatory

I have illustrated all along this manuscript what modeling the IR SED of various regions can bring. In particular, the FIR regime is crucial to properly constrain the peak of the large equilibrium grain emission. This is the only way we can quantify the total dust mass and its excitation conditions. These quantities are necessary to interpret any other observables, such as the strength of the aromatic features, the submm excess, etc. We currently have good archival data for most nearby galaxies. We however lack: (i) continuous MIR-to-FIR spectroscopy to better constrain SED models and study the various solid-state features in emission and absorption; and (ii) deep observations of quiescent low-metallicity systems (cf. Sect. 6.2.1). After the cancellation of SPICA, our community should regroup around a new project.

6.2.5 The Public Image of Interstellar Dust

On a public relation viewpoint, we should think about the way our field is represented.

Among colleagues,

there is still a distinction between the ISM (i.e. the MW) and the ISM of other galaxies. This hierarchy between intragalactic and extragalactic ISM is becoming less and less justified. In addition, interstellar media (extragalactic ISM) provide unique constraints on ISM physics, as I have illustrated along this manuscript. The plural is justified, as they are characterized by different heating and cooling mechanisms, different grain formation processes, etc. We should motivate the new generation of astronomers to consider ISMology as a field where, depending on the studied physical process, we can use a galaxy or a Galactic region as a laboratory.

To the outside world,

“dust” physics is not very appealing. What we do is important, but we need a better name. The US environmental protection agency defines nanoparticles as having sizes roughly between 1 and 100 nm. This is very close to the range of sizes of interstellar grains (cf. Sect. 2.3.3.1). We could thus call our object of study Cosmic NanoParticles (CNP), when talking to the outside world, and keep talking about dust between us.

6.3 Current Future Projects

I finish by listing the future projects I am currently considering working on. These are motivated by the challenges I have listed in Sect. 6.2.

6.3.1 The Modelosaur Approach

It is more and more becoming a necessity to consistently model the different processes (dust, gas and stellar physics), in a large Bayesian framework. I plan to progressively include more physical processes in HerBIE to account for a wider diversity of observables:

1.

the modeling of stellar evolution;

2.

the consistency between SED modeling, and dust and chemical evolution;

3.

the account of local dust evolution and of the systematic uncertainties of the model (depending on the developers of  Jones et al., 2017);

4.

radiative transfer grids of different dust-star topologies;

5.

including the constraints from photoionization and photodissociation lines.

A first step will be the project ICED (IAS-CEA Evolution of Dust) that I have put together. It will consist in modeling the spatial distribution of the dust properties in nearby galaxies, to constrain local grain evolution. This will be done with the data from DustPedia, as well as from the NIKA2 guaranteed time project IMEGIN.

6.3.2 Out-of-the-Box Idea Bin

• One way to solve the degeneracy between size distribution and emissivity is to observe a given dust mixture illuminated by a time-varying ISRF. This way, we can see the same grains exposed to two different fields, at two different times. This is possible with the light echo of SNe (e.g. Arendt et al., 2016), but very limited. Nearby Cepheids provide periodic variable sources, that modern observatories give us access to. I am currently working on a feasibility study of the observations of their echo with the JWST and ALMA.
• I currently have an intern working on the feasibility of using Natural Language Processing (NLP) methods, in collaboration with IRIS.AI, to tackle the origin of DIBs. NLP is a machine-learning method that, when trained on a large corpus of scientific articles, can be used to find relations between concepts, formulate new research directions, etc. It is aimed at dealing with information overload. The idea of the project is to see if we can automatically find clues in the chemistry and material science literature that could be relevant to DIBs.
• I am planning to use machine-learning to accelerate the interpolation of large model grids in HerBIE. Machine-learning, trained on hydrodynamical simulations, could also be useful to simulate realistic dust-star topologies (accounting for clustering, etc.), for which we would solve the radiative transfer. These grids would be SED model building blocks.