Exploring the relation between dust mass and galaxy properties using Dusty SAGE
Dian P. Triani, Manodeep Sinha, Darren J. Croton, Eli Dwek, Camilla Pacifici
MMNRAS , 1–12 (2021) Preprint 1 March 2021 Compiled using MNRAS L A TEX style file v3.0
Exploring the relation between dust mass and galaxy propertiesusing
Dusty SAGE
Dian P. Triani, , ★ Manodeep Sinha, , Darren J. Croton, , Eli Dwek, and Camilla Pacifici Centre for Astrophysics & Supercomputing, Swinburne University of Technology, Hawthorn, VIC 3122, Australia ARC Centre of Excellence for All Sky Astrophysics in 3 Dimensions (ASTRO 3D) Observational Cosmology Lab, NASA Goddard Space Flight Center, Code 665, Greenbelt, MD 20771, USA Space Telescope Science Institute, Baltimore, MD 21218, USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
We explore the relation between dust and several fundamental properties of simulated galaxiesusing the
Dusty SAGE semi-analytic model. In addition to tracing the standard galaxy proper-ties,
Dusty SAGE also tracks cold dust mass in the interstellar medium (ISM), hot dust mass inthe halo and dust mass ejected by feedback activity. Based on their ISM dust content, we divideour galaxies into two categories: ISM dust-poor and ISM dust-rich. We split the ISM dust-poorgroup into two subgroups: halo dust-rich and dust-poor (the latter contains galaxies that lackdust in both the ISM and halo). Halo dust-rich galaxies have high outflow rates of heated gasand dust and are more massive. We divide ISM dust-rich galaxies based on their specific starformation rate (sSFR) into star-forming and quenched subgroups. At redshift 𝑧 =
0, we findthat ISM dust-rich galaxies have a relatively high sSFR, low bulge-to-total (BTT) mass ratio,and high gas metallicity. The high sSFR of ISM dust-rich galaxies allows them to producedust in the stellar ejecta. Their metal-rich ISM enables dust growth via grain accretion. Theopposite is seen in the ISM dust-poor group. Furthermore, ISM dust-rich galaxies are typicallylate-types, while ISM dust-poor galaxies resemble the early-type population, and we showhow their ISM content evolves from being dust-rich to dust-poor. Finally, we investigate dustproduction from 𝑧 = 𝑧 = Key words: galaxies: ISM – ISM: dust – galaxies: formation – galaxies: evolution
The cosmic star formation history shows a rise of star formationactivity in our Universe from high-redshift till redshift 𝑧 =
2, knownas cosmic noon, and then a decrease towards 𝑧 = clouds, the raw material for stars(Kennicutt & Evans 2012). As a galaxy evolves, H clouds areconsumed to form stars, and less material is available for the nextstar formation episode. However, infalling gas can provide a freshsupply which replenishes gas that was consumed by star formationactivity (e.g Dekel & Birnboim 2006; Sánchez Almeida et al. 2014).Morphology and colour often correlate with star formationin galaxies (e.g. Roberts & Haynes 1994, for a review). Ellipticalgalaxies are dominated by an old stellar population and appearred (Thomas et al. 2005; Kormendy et al. 2009). These elliptical ★ E-mail: [email protected] galaxies have run out their star-forming gas; therefore, they have noor low star formation. At the opposite end, spiral galaxies have blueremission, showing signs for young stellar populations and active starformation (see also a review from Kennicutt 1998).Per “conventional” galaxy evolution theory, galaxies evolvefrom star-forming spiral galaxies to quenched ellipticals (Bundyet al. 2006; Skelton et al. 2012; Tojeiro et al. 2013). Dust fol-lows this evolutionary sequence, and galaxies become dust-poor atlate-times, either through dust destruction or ejection out of theISM. Galaxy formation models assume that as star formation ac-tivity decreases, mergers and disk instabilities make galaxies morespheroidal (Somerville et al. 2001; Baugh et al. 2005; Croton et al.2006, 2016). During such morphological transformations, a frac-tion of the ISM dust is ejected through SN and AGN feedback.Therefore, we see a decrease in the ISM dust mass in more ellipticalgalaxies.Since the dust content of a galaxy depends on its unique evo-lutionary history, the observed dust content in the local and high-redshift Universe provides insight into the physics of galaxy for- © a r X i v : . [ a s t r o - ph . GA ] F e b D. P. Triani et al. mation. Studies of high-redshift galaxies in the far-infrared andsubmillimetre regime reveal an abundance of star-forming galaxieswith massive dust reservoirs (Valiante et al. 2009). Dusty galaxiesare even found in the very early Universe, including A1689-zD1 at 𝑧 = . × M (cid:12) (Watson et al. 2015), andHFSL3 at 𝑧 = .
34 with a dust mass of 1 . × M (cid:12) (Riecherset al. 2013). However, observations also see star-forming galaxies atearly and late times with little dust and low metallicity (Fisher et al.2014). The dichotomy between dust-rich and dust-poor galaxiesacross redshift motivates this work.Recent galaxy surveys have measured the dust mass function(Dunne et al. 2003; Vlahakis et al. 2005; Dunne et al. 2011; Ealeset al. 2009; Clemens et al. 2013) and many scaling relations betweendust and the fundamental properties of galaxies. These include therelationship between dust mass and stellar mass (Santini et al. 2014)and the relationship between dust mass and star formation rate (SFR)(da Cunha et al. 2010; Santini et al. 2014). These scaling relationsprovide constraints for galaxy evolution models that track the dustproperties of galaxies.Historically, dust in galaxies is commonly investigated usingone zone analytical models (e.g., Dwek 1998; Zhukovska et al.2008; Valiante et al. 2009; Asano et al. 2013; Zhukovska 2014).Such models are critical to assess the various processes of dust pro-duction and destruction, especially to reproduce the dust content ofspecific galaxy populations. However, they generally use a simplis-tic approach to model gas and stellar evolution in galaxies and donot provide predictions for galaxies within a cosmological volume.Zoom-simulations coupled with self-consistent dust tracking give amore realistic accounting of how dust, metals, gas and stellar pop-ulations interact within galaxy evolution framework (Bekki 2015;Aoyama et al. 2017). But they are computationally expensive andtend to focus on reproducing the properties of individual galaxies.McKinnon et al. (2016) was one of the first works to modeldust and galaxy co-evolution within a cosmological simulation.They incorporated a detailed dust prescription to a hydrodynamicalsimulation, including stellar production, grain growth and grain de-struction by supernovae shocks. To date, a few semi-analytic galaxymodels have included self-consistent dust modelling. These modelscan reproduce a number of global trends between dust and variousgalaxy properties across cosmic time, with less detail but more ef-ficient computing cost and time compared to the hydrodynamicalsimulations (Popping et al. 2017; Vijayan et al. 2019; Triani et al.2020).In this paper, we extend the analysis of Triani et al. (2020)to investigate the different characteristics between dust-rich anddust-poor simulated galaxies using the Dusty SAGE semi-analyticmodel. Dust in the ISM forms in stellar ejecta (Dwek 1998;Zhukovska et al. 2008), grows further via grain accretion (Draine1990; Dwek 1998; Zhukovska et al. 2008; Draine 2009), and isdestroyed by supernovae shocks (Dwek & Scalo 1980; Jones et al.1994; Zhukovska et al. 2008; Slavin et al. 2015). Dusty SAGE in-cludes analytical prescriptions for these mechanisms, as well as dustlocked in stars (astration), and dust inflows from and outflows to thehalo and ejected reservoirs. In the model, dust undergoes furtherdestruction via thermal sputtering in the halo and ejected reservoirsdue to their high temperature. The primary constraints for dust mod-elled by
Dusty SAGE are the observed dust mass function and dustmass - stellar mass relation at 𝑧 = Dusty SAGE includes dusttracking in the ISM, halo, and ejected dust by feedback processes. https://github.com/dptriani/dusty-sage It provides predictions for the dust mass function and the relationbetween dust mass and stellar mass at high redshift.This paper is organised as follows: In Section 2, we provide abrief description of
Dusty SAGE . We study how the dust fraction ineach reservoir relates to the stellar mass and morphology in Section3. Then we categorise galaxies into groups based on their ISM andhalo dust content in Section 4. In Section 5, we map these groupsbased on their morphology, star formation activity and stellar mass atredshift 𝑧 =
0. We extend the relations to higher redshift, up to 𝑧 = ℎ = .
73 based onthe cosmology used of the Millennium simulation.
Here we provide only a brief overview of the galaxy and dust modelused in this work.
Dusty SAGE is built on the more generic galaxymodel
SAGE (Croton et al. 2006, 2016), but with a detailed pre-scription that tracks dust evolution. The summary of both galaxyand dust evolution mechanisms in
Dusty SAGE is presented in Fig-ure 1 in Triani et al. (2020). The model follows baryonic growthin dark matter halos taken from an N-body simulation. There arefour baryonic reservoirs: the pristine gas, the hot halo, the ISM andan ejected reservoir. In this paper, we run
Dusty SAGE on the Mil-lennium simulation (Springel et al. 2005) and select galaxies withstellar mass log 𝑀 ∗ = −
12 M (cid:12) .Mass is exchanged among the different baryonic reservoirsacross cosmic time. Pristine gas falls into the collapsed dark matterhalo and is heated to the virial temperature – the hot halo reservoir(Rees & Ostriker 1977; White & Rees 1978). A fraction of this hotgas then cools into the ISM (Sutherland & Dopita 1993). In theISM, cold hydrogen differentiates into atomic and molecular hydro-gen (Hollenbach & McKee 1979). Stars then form from molecularclouds (Kennicutt & Evans 2012). During their evolution, stars cre-ate helium and elements heavier than helium in their core, knownas ‘metals’. These metals are expelled in stellar ejecta and changethe metallicity of the ISM (ratio of metals and gas mass). Massivestars end their life as supernovae and inject a large amount of energyinto the cold ISM. This energy reheats cold ISM gas back to thehot halo reservoir, and potentially even expels it outside the halo– the ‘ejected’ reservoir (e.g., Marri & White 2003). This processis known as SN feedback. The gas accretion onto supermassiveblack holes provides another source of feedback; this energetic phe-nomenon is known as active galactic nuclei (AGN) (Croton et al.2006). The ejected gas can later be reincorporated back to the haloand galaxy.In both
Dusty SAGE and
SAGE (Croton et al. 2006, 2016),the bulge-to-total stellar mass ratio represents the morphology ofthe galaxy, which evolves during disk instabilities and mergers. Weadopt the disk stability criteria of Mo et al. (1998) and transfersufficient stellar mass from the disk to the bulge to ensure stabilityevery time an instability occurs. In a merger between a more massivecentral galaxy and a less massive satellite galaxy, bulge enrichmentdepends on the total stellar and gas mass ratio of both progenitors. Ifthe mass ratio exceeds 0.3, a ‘major’ merger has occurred: the disksof both galaxies are destroyed and all stars are placed in a bulge.Otherwise, the merger is ‘minor’ and the satellite stellar content isadded to the central bulge. Both instabilities and mergers can triggera starburst. Unlike quiescent disk star formation, stars formed in amerger-induced starburst are placed in the bulge.Besides the usual galaxy evolution processes,
Dusty SAGE
MNRAS000
MNRAS000 , 1–12 (2021) xploring the relation between dust mass and galaxy properties using
Dusty SAGE also incorporates a detailed dust evolution model. Analogous to thegas, we track dust in three reservoirs – the ISM, the hot halo and anejected reservoir. Dust is mainly processed in the ISM, then heatedto the halo and ejected reservoirs via feedback mechanisms poweredby supernovae and AGN. The total dust production rate in the ISMis described by: . 𝑀 d = . 𝑀 formd + . 𝑀 growthd − . 𝑀 destd − . 𝑀 SFd − . 𝑀 outflowd + . 𝑀 inflowd , (1)where: • . 𝑀 formd is the stellar dust formation rate: In every star formationepisode, Dusty SAGE tracks the abundance of C, N and O in AGBstars and C, O, Mg, Si, S, Ca and Fe in SN II ejecta. The condensationof these elements to form dust is given in Table 1. • . 𝑀 growthd is the grain growth rate in dense molecular clouds:Existing grains grow via metal accretion (Dwek 1998; Zhukovskaet al. 2008) where the timescale for this process depends on themetal abundance in the cold gas. • . 𝑀 destd is the destruction rate via SN shocks: We follow theprescription from Dwek & Scalo (1980); McKee (1989); Asanoet al. (2013) to compute a destruction timescale from the total coldgas mass and the supernovae rate, efficiency and swept mass. • . 𝑀 SFd is the rate for dust locked in newly formed stars, which isproportional to the star formation rate and the dust-to-gas ratio. • . 𝑀 outflowd and . 𝑀 inflowd are the dust outflow and inflow rates.SN and AGN feedback can reheat cold ISM gas and expel it to thehalo. The feedback energy, if large enough relative to the depth ofthe potential well, can even eject the gas to leave the galaxy andhost halo. In an outflow, we assume that the dust-to-gas ratio of theejected gas is equal to the ISM. Ejected gas can be reincorporatedback to the halo, while maintaining the dust-to-gas ratio of theejected reservoir. Hot gas undergoes cooling process back to thedisk, and we assume in this inflow the dust-to-gas ratio equals thatof the halo.Dust populates the halo and ejected reservoirs through out-flows from the ISM. In both reservoirs, dust grains are destroyedvia thermal sputtering on a short timescale that depends on the gasdensity and temperature. In both the halo and ejected reservoirs, weassume the virial temperature, with an isothermal density profilefor the hot gas extending to the virial radius, and a uniform densityprofile for the ejected component. In every timestep, the ejectedreservoir’s density is evaluated by dividing the gas mass with thereservoir’s volume, assuming the virial radius. However, the com-puted timescale should be taken as the upper limit for the ejectedreservoir. In reality, the ejected reservoir is a mix of the circum-galactic medium (CGM) and intergalactic medium (IGM). Gas inthe IGM might extend beyond the virial radius, resulting in a lowerdensity. The temperature might also be higher than in the model,allowing for a more efficient sputtering.Although thermal sputtering is very efficient, we still find a sig-nificant abundance of dust in the halo. Dust properties in the halodepend on the balance between the outflows and inflows from/to theISM. The depth of the galaxy’s potential well also affects the out-flow mass trapped in the halo. Galaxies with a shallow potential aremore likely to lose the outflowing materials to the ejected reservoir.We assume that the gas outflow carries dust in the same proportionas the ISM and transfers it to the halo, with no dust destruction inthe process. It may well be that a fraction of dust is destroyed whenejected out of the ISM, or the DTG ratio of the outflow differs fromthe ISM, which will alter our predictions. As we mentioned above,it is also possible that the sputtering rates are higher than our pre-diction. However, the nature of such processes is currently unclear. Future observations to quantify dust properties in the galactic windand the IGM will provide additional constraints to these processes.Figure 14 in Triani et al. (2020) shows that galaxies have moredust in their halo than in the ISM at low-redshift. This outcomeroughly agrees with the massive dust content found in the CGM(Dunne et al. 2011; Peek et al. 2015). The galaxy formation modelof Popping et al. (2017) also predicts a significant amount of dust inboth the halo and ejected reservoir. However, their dust mass densityis notably higher than our results. Dusty SAGE provides a good agreement with the galaxy stellarmass function at redshift 𝑧 =
0. It also successfully reproduces thedust mass function and various dust scaling relations over a widerange of redshifts (Triani et al. 2020). To achieve a more realisticdistribution of the stellar mass in the bulge and disc, we make slightchanges to a few of the parameters in Triani et al. (2020). These arelisted in Table 1.
In this section, we investigate the relation between galaxy morphol-ogy, represented as the bulge-to-total (BTT) mass ratio, and thefraction of dust in the ISM, halo and ejected reservoirs relative tothe total dust mass in all reservoirs. We use stellar mass in comput-ing the BTT mass ratio and assume the gas mass to be negligible.We present the median dust fraction vs BTT mass ratio in threestellar mass bins in Figure 1. The grey histogram shows the distri-bution of BTT mass ratios across all galaxies. In all panels, galaxieswith higher BTT mass ratio contain most of their dust in the halowhile those with a lower BTT mass ratio tend to keep theirs in theISM. For the low mass galaxies in the bottom panel, spiral galaxiescontain a significant fraction of their dust in the ISM and ejectedreservoir. These galaxies make up more than 50% of the overallpopulation. Conversely, the halo contains nearly 90% of the totaldust in elliptical galaxies with BTT > 0 . ∗ ( M (cid:12) ) = −
11 (central panel),the same trend occurs. Galaxies with a BTT mass ratio smaller than ∼ . ∗ ( M (cid:12) ) = .
1, Peeples et al. (2014) found a fiducial value of metals locked inCGM dust of 5 × M (cid:12) , almost twice the value for those found inthe form of ISM dust, 2 . × M (cid:12) . Ménard et al. (2010) deriveda CGM dust mass of 5 × M (cid:12) for galaxies with stellar masslog M ∗ = . (cid:12) . Although the CGM is not exactly equivalent toour halo gas component, we will use CGM dust as a proxy to testour predictions of the dust bond to, but outside the galaxy. MNRAS , 1–12 (2021)
D. P. Triani et al.
Table 1.
Fiducial
Dusty SAGE parameters used throughout this work, also compared to those from Triani et al. (2020).Parameter Description Value Triani et al. (2020) 𝑓 cosmic 𝑏 Cosmic baryon fraction 0.17 0.17 𝑧 Redshift when H II regions overlap 8.0 8.0 𝑧 𝑟 Redshift when the intergalactic medium is fully reionized 7.0 7.0 𝛼 SF Star formation efficiency from H [Myr − ] 0.005 0.005 𝑅 Instanteneous recycling fraction 0.43 0.43 𝜖 disc Mass-loading factor due to supernovae 2.0 3.0 𝜖 halo Efficiency of supernovae to unbind gas from the hot halo 0.2 0.3 𝑘 reinc Velocity scale for gas reincorporation 0.15 0.15 𝜅 R Radio mode feedback efficiency 0.09 0.08 𝜅 Q Quasar mode feedback efficiency 0.005 0.005 𝑓 BH Rate of black hole growth during quasar mode 0.015 0.015 𝑓 friction Threshold subhalo-to-baryonic mass for satellite disruption or merging 1.0 1.0 𝑓 major Threshold mass ratio for merger to be major 0.15 0.3 𝛼 burst Exponent for the powerlaw for starburst fraction in merger 0.18* 0.7 𝛽 burst Coefficient for starburst fraction in merger 0.75* 0.56 𝛿 AGBC
Condensation efficiency for AGB stars 0.2 0.2 𝛿 SNIIC
Condensation efficiency for SN II 0.15 0.15 𝜏 acc , Accretion timescale for grain growth [yr] 4 . × . × 𝑓 SN Fraction of destroyed dust to the swept dust mass by SN 0.1 0.1*value adopted from Somerville et al. (2001)
Figure 1.
The median value for dust mass fraction in each baryon reservoir versus bulge-to-total stellar mass ratio (BTT). The purple, orange and green linesrepresents hot dust mass in the halo, cold dust mass in the ISM, and dust in the ejected reservoir, respectively. The grey histogram represents the numberdistribution of galaxies in 10 BTT bins. MNRAS000
The median value for dust mass fraction in each baryon reservoir versus bulge-to-total stellar mass ratio (BTT). The purple, orange and green linesrepresents hot dust mass in the halo, cold dust mass in the ISM, and dust in the ejected reservoir, respectively. The grey histogram represents the numberdistribution of galaxies in 10 BTT bins. MNRAS000 , 1–12 (2021) xploring the relation between dust mass and galaxy properties using
Dusty SAGE Figure 2.
The cold dust mass of simulated galaxies as a function of stellarmass at redshift 𝑧 =
0. The heatmap shows the density distribution witha brighter colour representing higher density, and the black line marks themedian. The red line marks the threshold where the cold dust mass per stellarmass equals 10 − . Above this line, galaxies are classified as ISM dust-richwhile below are classified as ISM dust-poor. To investigate how the dust content of galaxies correlates with theirstellar mass, star formation activity, and morphology, we first estab-lish our simulated galaxies into groups. In the
Dusty SAGE model,the dust in galaxies is distributed in three components - the ISM,the halo and the ejected reservoir. We exclude the dust in the ejectedreservoir in our analysis since it is out of the system. Our four dustgroups are described below.
When considering dust, observers typically report the properties ofthe ISM (e.g., Eales et al. 2009; Rémy-Ruyer et al. 2014; Santiniet al. 2014; Mancini et al. 2015; da Cunha et al. 2015; Nersesian et al.2019). Although some works have extended dust measurements toinclude the CGM (Ménard et al. 2010; Peek et al. 2015), suchobservations are rare. Therefore, as a first step, we have classifiedour model galaxies based on their cold dust content in the ISM.Figure 2 shows the relation between cold dust mass in the ISM andstellar mass of our model galaxies at redshift 𝑧 =
0. We find that thedust mass in the ISM increases with the host stellar mass, but witha significant scatter.The median values of our predicted ISM dust mass versusstellar mass are in a good agreement with the observations of dustin local galaxies, such as from the DustPedia catalogue (Nersesianet al. 2019). Figure 2 shows a large scatter below the median values,marking galaxies with relatively less dust compared to the overallpopulation.We draw an arbitrary line at dust mass per stellar mass of 10 − ,marked with the red line in Figure 2 to divide the galaxy populationinto two categories: ISM dust-rich and ISM dust-poor. The ISMdust-poor group contains galaxies below the red line. Observationshave found such ISM dust-poor galaxies in both the local and high-redshift Universe (Fisher et al. 2014).Further exploration reveals that Dusty SAGE galaxies in thisISM dust-poor category vary in their halo dust content. Therefore,
Figure 3.
The distribution of the specific star-formation rate in our ISMdust-rich galaxies. The grey line at 2 . × − yr − is the Milky Way valueadopted from Licquia & Newman (2015) which we use as a threshold indefining quenched and star-forming galaxies in our model galaxies. we additionally divide the ISM dust-poor group based on theirfraction of halo dust per stellar mass described below. Several observations (Peek et al. 2015; Peeples et al. 2014; Ménardet al. 2010) extended their search for metals and dust to the CGM.Peeples et al. (2014) find that only 25% of metals created in starsstay in the ISM; a similar fate also occurs for dust. Ménard et al.(2010) found galaxies with massive CGM dust. These observationsdiscovered the existence of dust out of the ISM, which is reproducedby our model.In our model, a mix of halo and ejected dust might be a betterrepresentation of the CGM for some galaxies. However, we onlyuse halo dust to represent the observed CGM dust. We use the samethreshold as the ISM dust-mass to define the halo-dust rich galaxies:halo dust mass per stellar mass of 10 − . Halo dust-rich galaxies aredefined as ISM dust-poor galaxies with halo dust mass above thisthreshold. The population of galaxies that lack dust in both their ISM and halois classified as dust-poor. Due to the rarity of dust measurementsoutside the ISM, we found no counterpart of this category from theobservations. Their existence in our model serves as a predictionfor future surveys measuring the CGM dust.
We define the ISM dust-rich galaxies in our model as those witha fraction of ISM dust mass per stellar mass above 10 − . Dustobscures the intrinsic stellar spectra of such galaxies and re-emitsthe radiation in the infrared (Witt et al. 1992; Witt & Gordon 2000).Infrared and sub-millimeter surveys have found a substantial numberof these galaxies, especially at low redshift (e.g., Eales et al. 2009;Dunne et al. 2003; da Cunha et al. 2015; Clemens et al. 2013;Rémy-Ruyer et al. 2014).Dust accounting in the galactic ISM is the main focus of many MNRAS , 1–12 (2021)
D. P. Triani et al. current and future infrared surveys. Forthcoming galaxy survey, likethose using JWST, will measure dust in high redshift galaxies. Suchsurveys will provide additional constraints to our model predictions.Besides the relation between dust mass and stellar mass shownin Figure 2, observations have also found a relation between dustmass and SFR (da Cunha et al. 2010; Casey 2012; Santini et al.2014). Based on this, we divide our ISM dust-rich galaxies furtherinto star-forming and quenched subclasses.
Figure 3 shows the specific SFR (sSFR) distribution of the ISM dust-rich galaxies in our model. To separate star-forming from quenchedgalaxies, we use the Milky Way sSFR value as the threshold, 2 . × − yr − (Licquia & Newman 2015). Galaxies with sSFR abovethe Milky Way value are defined as the star-forming ISM dust-rich. ISM dust-rich galaxies are commonly related to star-forming spiralssince dust is initially formed in stellar ejecta and grows in molecularclouds (Valiante et al. 2009). However, we also find a population ofISM dust-rich galaxies with relatively low star formation activity inour model (see Figure 3). The quenched ISM dust-rich galaxies arethose with a sSFR below the Milky Way value of 2 . × − yr − . In the previous sections, we categorized model galaxies into fourclasses: star-forming ISM dust-rich, quenched ISM dust-rich, halodust-rich and dust-poor. We now plot the BTT mass ratio vs sSFRof each group in three mass bin in Figure 4 as indicated in thelegend, with the colour distribution representing the median gas-phase metallicity of each group. The error bars denote the 16 th and84 th percentile range.Figure 4 shows a general trend where the ISM dust-rich galax-ies in the bottom panel occupy the lower right space marking diskygalaxies with higher star formation activity. Meanwhile, ISM dust-poor galaxies in the top panel generally have lower sSFR and a largerBTT mass ratio. Since galaxies typically change from star-formingspirals to quenched ellipticals over time, this plot indicates that theISM also evolves from being dust-rich to dust-poor across h.The gas-phase metallicity in Figure 4 increase with stellarmass, as one may expect from the galaxy mass-metallicity relation(Tremonti et al. 2004). Both ISM dust-rich groups generally havehigher metallicity than the ISM dust-poor groups.Star-forming ISM dust-rich galaxies have the highest sSFR andlowest BTT mass ratio. Condensation in stellar ejecta is one of themain dust production channels in our model and is tightly correlatedwith star formation activity. Thus, we would expect ISM dust-richgalaxies to have a high sSFR. It is also possible that the high sSFRof such galaxies reflects the abundance of molecular gas in the ISM,which enables dust growth via accretion of metal grains.Quenched ISM dust-rich galaxies have much lower sSFR (onaverage) and are also located in the low BTT-ratio regime. Theirgas-phase metallicity is higher than that of the star-forming ISMdust-rich galaxies. As galaxies in this group have less sSFR thanthose in the star-forming group, their stellar dust production mustbe less efficient. However, their abundance of dust can be explainedby their high availability of gas-phase metals, which can accrete onto existing dust grains in the dense environment. The low sSFRof these galaxies also suggest that they lack massive young stars thatbecome SN, the primary destroyer of dust in the ISM. The scarcityof SN can also prevent the dust and refractory elements from beingexpelled, allowing them to preserve more dust and metals in theirISM.In contrast to the ISM dust-rich galaxies, the ISM dust-poorgroup have higher BTT mass ratios (i.e., more elliptical), loweraverage sSFRs and lower gas-phase metallicities. The halo dust-richgroup have the highest BTT mass ratio, incorporating the “bulgiest”galaxies for all mass bins. Galaxies in this group also have the lowestgas metallicities. To explore this further, we plot the distribution ofthe ratio between the ISM cold gas mass and the hot halo gas mass inFigure 5. Here, the halo dust-rich group peaks lowest of all groups,showing that galaxies in this group either have less cold gas in theirISM or more gas in their halo. The fact that these galaxies havemassive halo dust and hot gas, yet they lack ISM gas and dust,indicates an efficient mechanism to transport their ISM content outof the disk.Figure 6 shows the fraction of galaxies in each group versusoutflow rate. The halo dust-rich group dominates for outflow ratesabove 30 M (cid:12) yr − . This suggests that the efficient outflow in thesegalaxies can blow their ISM content out of the disk, while their lowsSFR does not allow them to replenish the metals and dust in theISM.Looking back to Figure 4, we see that the total percentage ofgalaxies in the most massive mass bin (log M ∗ = −
12) is 1 . Dusty SAGE , massive galaxies are likely to have deeper potentialwells, which allow them to capture most of the dust outflow from theISM in the halo. Therefore, very little dust escapes into the ejectedreservoir in feedback events. Figure 1 also shows that in the mostmassive galaxies, the majority of the dust mass is in the hot halo,and an only small fraction of their dust fraction is in the ejectedreservoir at all BTT mass ratios.In the dust-poor category, we find an anti-correlation betweenstellar mass and sSFR. The median sSFR of dwarf galaxies islog sSFR = − .
6, more than two orders of magnitude larger thanthose in the most massive bin. Galaxies will become dust-poor iftheir dust depletion rate exceeds the dust production rate. SinceFigure 2 shows a positive correlation between dust mass and stellarmass, more massive galaxies are more likely to have more dust intheir ISM. Therefore, only those galaxies with a very low produc-tion rate can become dust-poor. The anti-correlation reflects the linkbetween dust and star formation rates.Figure 4 also allows us to examine the stellar mass evolutionacross the four groups. At the massive end, galaxies with masslog M ∗ ( M (cid:12) ) = −
12 are comprised primarily of the star-formingISM dust-rich category (0.6 % of the total population), followed bythe quenched ISM dust-rich and halo dust-rich groups each contain-ing 0.3% of the total galaxies. The dust-poor group is in the minorityamongst massive galaxies. The dominance of the ISM dust-rich andhalo dust-rich groups shows that the most massive galaxies rarelydestroy or eject their dust out of the halo. Massive late-type andearly-type galaxies contain the majority of their dust in the ISM andthe halo, respectively.In the middle bin (log M ∗ ( M (cid:12) ) = − ∗ ( M (cid:12) ) = − MNRAS000
12 are comprised primarily of the star-formingISM dust-rich category (0.6 % of the total population), followed bythe quenched ISM dust-rich and halo dust-rich groups each contain-ing 0.3% of the total galaxies. The dust-poor group is in the minorityamongst massive galaxies. The dominance of the ISM dust-rich andhalo dust-rich groups shows that the most massive galaxies rarelydestroy or eject their dust out of the halo. Massive late-type andearly-type galaxies contain the majority of their dust in the ISM andthe halo, respectively.In the middle bin (log M ∗ ( M (cid:12) ) = − ∗ ( M (cid:12) ) = − MNRAS000 , 1–12 (2021) xploring the relation between dust mass and galaxy properties using
Dusty SAGE Figure 4.
Top panel shows the BTT ratio, sSFR and gas-phase metallicity of our model galaxies in the ISM dust-poor group. This group consists of twosubgroups: halo dust-rich and dust-poor. Markers with small, medium and large size represents three mass bin: log M ∗ = −
10 M (cid:12) , log M ∗ = −
11, andlog M ∗ = −
12 M (cid:12) , respectively. The points mark the median sSFR and median BTT mass ratio of each group, with the errorbars spanning the 16 th and 84 th percentiles. The colour of each point marks the gas-phase metallicity, with brighter colours representing higher value. We assume a solar metallicity of 0.02and 12 + log [ O / H ] − log [ Z / Z (cid:12) ] = .
0. The number inside parentheses in the legend represents the percentage of galaxies in each category with respect to thetotal number of galaxies with stellar mass log M ∗ = −
12 at 𝑧 =
0. Open symbols indicate equivalent observational dataset taken from DustPedia (Nersesianet al. 2019) and KINGFISH (Kennicutt et al. 2011; Rémy-Ruyer et al. 2014).
Bottom panel shows the same relation above but for the ISM dust-rich group,which is splitted into star-forming and quenched subgroups. rich still dominates (28%). However, the contributions of quenchedISM dust-rich and dust-poor groups in this mass bin are also themost significant (13.9 % and 12.8 %, respectively) compared tothose in the higher mass bins. The halo dust-rich group shares thesmallest percentage for dwarf galaxies with only 3.6%.
To add context to our theoretical predictions, we make comparisonswith two observation datasets. DustPedia (Nersesian et al. 2019) isa compilation of local galaxies with dust measurements. To createthis, the authors use the dust model
THEMIS in the SED fitting code
CIGALE to infer dust mass and galaxy properties from
Herschel multiwavelength photometry. The authors use numerical HubbleT as the morphology indicator, with lower value corresponds to alarger stellar composition in bulge relative to the disk. To converttheir Hubble T value to BTT-ratio, we use the K-band bulge-to-diskflux ratio mapping from Graham & Worley (2008). Note that theBTT ratio for our simulated galaxies is based on stellar mass, whilethe observations are based on flux.We group their data based on our definition of the star-forming ISM dust-rich, quenched ISM dust-rich and ISM dust-poor galaxies.We can not divide their ISM dust-poor galaxies further because theydo not provide a halo dust mass. We did not find galaxies classifiedas quenched ISM dust-rich in their dataset. Therefore, we plot themean values of BTT mass ratio and sSFR of their star-forming ISMdust-rich group as a pink-filled circle and their ISM dust-poor groupas a pink-filled square in Figure 4.Our second observational dataset is taken from the KINGFISHsurvey (Kennicutt et al. 2011). We group KINGFISH galaxies basedon our definitions and compute the mean dust mass, BTT mass ratio,and SFR for each group. We plot the mean dust and galaxy propertieswith markers showing the group they belong. The dust measurementfor KINGFISH galaxies is taken from Rémy-Ruyer et al. (2014).Our prediction for star-forming ISM dust-rich and ISM dust-poor galaxies are in good agreement with the observed values. Inboth the KINGFISH and DustPedia data, dust-poor galaxies arethose with the most elliptical morphology (E-type) with Hubble Tbelow -2 (Rémy-Ruyer et al. 2014; Nersesian et al. 2019). However,the map of Graham & Worley (2008) that we use to convert theHubble type to BTT ratio only extends to S0-type galaxies (HubbleT = -2). Therefore, we use the Hubble T = -2 conversion for elliptical
MNRAS , 1–12 (2021)
D. P. Triani et al.
Figure 5.
The probability distribution of the cold to hot gas mass ratio ofour simulated galaxy categories. Red, blue, green and purple lines representthe star-forming ISM dust-rich, quenched ISM dust-rich, halo dust-rich anddust-poor groups, respectively.
Figure 6.
The gas outflow rate distribution due to feedback. We assume thatthe dust-to-gas mass ratio in the outflow is the same as the general ISM. galaxies with Hubble T < -2. The converted BTT flux ratio of thesegalaxies are 0 .
15 dex lower than our median mass ratio but are stillwithin our 16 th percentile. Galaxy evolution models predict that galaxies evolve from star-forming spiral to quench elliptical (i.e., the bottom right to top leftin Figure 4) (Somerville et al. 2001; Baugh et al. 2005; Croton et al.2006, 2016). The dust content of the ISM follows this evolutionarytimeline to become dust-poor over time. There are two primary de-pletion mechanisms for ISM dust: destruction by SN shocks (Slavinet al. 2015) and ejection by SN and AGN feedback (Feldmann 2015;Popping et al. 2017). Ejected dust can be trapped in the halo if thepotential well is sufficiently deep, causing the host to end up as ahalo dust-rich galaxy. Otherwise, such galaxies end up as dust-poor. To see the evolution of the dust content in galaxies across time,we plot the relation between BTT mass ratio and sSFR of our dustgroups from redshift 𝑧 = 𝑧 = ∗ ( M (cid:12) ) = −
11) asrepresentative of the Milky Way mass range. Circle, inverted trian-gle, diamond, and squares represent the star-forming ISM dust-rich,quenched ISM dust-rich, halo dust-rich, and dust-poor categories,respectively. This plot does not show the evolution of the samegroup of galaxies. Instead, it shows us the map of galaxies belongto each group at different redshift in the BTT mass ratio and sSFRplane.Figure 7 shows how the percentage of galaxies in each groupchanges over time. The fraction of star-forming ISM dust-rich galax-ies peaks at 𝑧 = ≈
25% until 𝑧 =
0. The otherthree groups show lower fractions but continue to increase down tothe present day. This evolutionary trend reveals how star-formingISM dust-rich galaxies might convert into the other groups at lowerredshift.Across redshift, the behaviour of individual groups within theBTT mass ratio - sSFR space also changes. In the BTT mass ratiodimension, the quenched ISM dust-rich group gets diskier towards 𝑧 = . 𝑧 = 𝑧 =
0. All groups except thequenched ISM dust-rich shows a significant decrease of the mediansSFR from 𝑧 = 𝑧 =
0. The biggest sSFR decrease occurs inhalo dust-rich galaxies, by more than 2 order of magnitude, whilethe star-forming ISM dust-rich and dust-poor galaxies each undergoa single dex decrease.In the gas phase metallicity dimension, the halo dust-rich grouphave the lowest median value across redshift 𝑧 = 𝑧 =
0. Themedian metallicity of both dust-poor and star-forming ISM dust-richgalaxies increases slightly with decreasing redshift. Meanwhile, themedian metallicity of the quenched ISM dust-rich group fluctuateswith redshift; it increases ≈ . 𝑧 = 𝑧 =
2, doesnot change between 𝑧 = 𝑧 =
1, then decreases ≈ . From Figure 4 and Figure 7, we can see that although we dividegalaxies into groups solely based on their dust content, the galaxyproperties of each group are often distinct. Also, galaxy propertieswithin each group evolve with redshift. Our predictions for howthese groups change with time can help us understand the baryonicand dust physics occurring. In this section, we discussed the featuresof these groups between redshift 𝑧 = At all mass bins at redshift 𝑧 =
0, the star-forming ISM dust-rich group consistently stays at the lower right region in the BTTmass ratio - sSFR space (see Figure 4). This indicates that thesegalaxies are still actively forming stars and have an abundance ofmolecular gas in their ISM, providing a dust production channelvia condensation in stellar ejecta and grain growth in the molecularclouds. The gas-phase metallicity in this group is relatively highcompared to those in the halo dust-rich and dust-poor groups. Thisgroup dominates the galaxy population for all mass bins at redshift 𝑧 = MNRAS000
0, the star-forming ISM dust-rich group consistently stays at the lower right region in the BTTmass ratio - sSFR space (see Figure 4). This indicates that thesegalaxies are still actively forming stars and have an abundance ofmolecular gas in their ISM, providing a dust production channelvia condensation in stellar ejecta and grain growth in the molecularclouds. The gas-phase metallicity in this group is relatively highcompared to those in the halo dust-rich and dust-poor groups. Thisgroup dominates the galaxy population for all mass bins at redshift 𝑧 = MNRAS000 , 1–12 (2021) xploring the relation between dust mass and galaxy properties using
Dusty SAGE Figure 7.
The evolution of the relation shown in Figure 4 between redshift 𝑧 = 𝑧 =
3. For clarity, we focus on the middle mass bin(log 𝑀 ∗ = −
11 M (cid:12) ). Circle, inverted triangle, diamond and square represent the median value of the star-forming ISM dust-rich, quenched ISM dust-rich,halo dust-rich and dust-poor groups, respectively. lation. The median gas-phase metallicity is lower at higher redshift.This metallicity increase with decreasing redshift reflects the abilityof such galaxies always to retain an abundance of metals in theirISM. Since galaxies in the star-forming ISM dust-rich group havehigh sSFR by the selection, we expect that the ISM will be bothmetal and dust-rich. Observations of the local and high redshiftUniverse find many galaxies that belong to this group. Figure 4shows that at redshift 𝑧 =
0, both the KINGFISH (Kennicutt et al.2011; Rémy-Ruyer et al. 2014) and DustPedia (Nersesian et al.2019) datasets have galaxies with dust-to-stellar mass and sSFRclassifications similar to our star-forming ISM dust-rich group.Figure 8 shows the sSFR of our model galaxies as a functionof dust-to-stellar mass ratio from redshift 𝑧 = 𝑧 =
3. Thewhite line marks the divider between the ISM dust-rich and ISMdust-poor groups. Across redshift, many authors find star-forminggalaxies with a dust rich ISM. The general trend that the sSFR de-creases with decreasing redshift is seen both for the model and inthe observations. At 𝑧 =
0, our ISM dust-rich galaxies are in goodagreement with the observational values from DustPedia (Nersesianet al. 2019), Rémy-Ruyer et al. (2014) and Santini et al. (2014). Forthe ISM dust poor galaxies below the grey line, the observations arelower than our median value but are still within the 16 th percentile.At higher redshift, observations are yet to find ISM dust-poor galax-ies. At 𝑧 =
1, our model roughly agrees with the dust-rich data fromSantini et al. (2014), although our median value of sSFR lies 0 . − . 𝑧 = 𝑧 =
3, however, ourmodel fails to reproduce galaxies with log ( M dust / M star ) > − . Dusty SAGE and ourbase model
SAGE (Croton et al. 2016) are constrained using variousgalaxy observations at 𝑧 =
0; therefore, we still find limitationsin reproducing the high redshift population. We also note that thedataset from da Cunha et al. (2015) (red triangles) is obtained in theALMA LESS survey which is biased towards the extremely brightgalaxies.
The quenched ISM dust-rich group consists of galaxies that havean abundance of dust in their ISM despite their low star formationactivity (defined as galaxies with sSFR below the Milky Way). Onepossible scenario for this is that due to the low sSFR, these galaxiesno longer host supernovae events that destroy dust and expel metalsand dust from the ISM. This also explains the higher gas-phasemetallicity of galaxies in this group compared to those in the star-forming ISM dust-rich group. The abundance of metals in the ISMprovides material for dust growth via grain accretion, allowing suchgalaxies to accumulate dust mass. Another scenario for their dustproduction is the condensation in the ejecta from previous starformation episodes. Although they do not have a fresh supply ofstellar ejecta due to the low sSFR, it is possible that there are still
MNRAS , 1–12 (2021) D. P. Triani et al.
Figure 8.
The relation between specific star formation rate (sSFR) and the dust-to-stellar mass ratio from redshift 𝑧 = 𝑧 =
3. The heat map shows the 2Ddensity distribution of our model galaxies with brighter color representing higher density. The solid black lines mark the median while the dashed lines markedthe 16 th and 84 th percentile. The white vertical lines mark a log dust-to-stellar mass ratio = -4, where we divide the ISM dust-rich and dust-poor categories.The red circles, blue diamonds, green squares and red triangles are observational values from DustPedia: Nersesian et al. (2019), Rémy-Ruyer et al. (2014),Santini et al. (2014) and da Cunha et al. (2015), respectively. refractory elements from previous SN and AGB winds in the ISMthat condense into dust grains.Figure 4 shows that the majority of quenched ISM dust-richgalaxies can be found in the low mass population. Compared to thestar-forming ISM dust-rich group, this group is rare at low and highredshift. Figure 7 shows that the fraction of this group with stellarmass log M ∗ = −
11 at 𝑧 >
Halo dust-rich galaxies have little or no dust in their ISM but containa significant amount of dust in their halo. At redshift 𝑧 =
0, this groupincludes elliptical galaxies with the largest BTT mass ratios and lowsSFR. The median gas-phase metallicity of this group is also thelowest compared to the other groups. The small sSFR does not allowfor efficient instantaneous metal and stellar dust production in suchgalaxies. However, they should have accumulated metals and dustfrom previous star-formation episodes. The lack of metals and dustin their ISM, while their halo is dust-rich, indicates that the ISMcontent is transported into the halo (see Section 5).At higher redshift, however, the halo dust-rich group occupiesan increasingly high sSFR region with a median log sSFR > theMilky Way value. The high sSFR of these galaxies means that theyare producing metals and dust in their stellar ejecta. However, unlikegalaxies in the other groups, galaxies in this group show no increasein gas metallicity with decreasing redshift. The low abundance ofmetals and dust in their ISM is possible if the feedback is efficientlyheating and blowing out the newly formed metals and dust. Suchfeedback might be provided by AGN activity (Sarangi et al. 2019).Galaxies in a deep potential well can retain the feedback-heatedgas, dust and metals in their halo. Mass in the dark matter halo usu-ally sets the depth of the potential well. Since stellar mass correlates with halo mass, halo dust-rich galaxies should be more common inthe massive galaxies. Figure 4 shows that in the massive bin, thisgroup comprise a quarter of the population, a more significant frac-tion than the lower stellar mass bin.Ménard et al. (2010) measured dust reddening effects of back-ground quasars relative to the foreground SDSS galaxies. Theyobserved the existence of diffuse dust in halos with an amount com-parable to those in the galactic disk. For 0 . 𝐿 ∗ galaxies, the authorsestimated a dust mass of 5 × M (cid:12) in the halo. Peek et al. (2015)confirmed this halo dust mass for 0 . 𝐿 ∗ − 𝐿 ∗ low redshift galaxies.Observations of CGM dust provides evidence for our predictionthat dust is transported out of the ISM and trapped in the halo.However, such observations are still uncommon. Future CGM dustmeasurements will allow for better constraints for models such as Dusty SAGE . We define galaxies that lack dust in both the ISM and their halo asdust-poor galaxies. At redshift 𝑧 = × − and are classified as ISM dust-poor as per our defi-nition. However, both datasets do not provide measurement for dustin the halo. Therefore, we assume their halo is dust-poor, and weinclude them in the dust-poor group instead of the halo dust-richgroup.Dust depletion in the ISM occurs in two ways: destruction bySN shocks inside the galaxy disk and dust that is ejected out of theISM by feedback. If the potential well of the system is enough totrap the dust, the ejected dust will end up in halo. However, galaxieswith low stellar mass usually live in shallower potentials and thus MNRAS000
0, this groupincludes elliptical galaxies with the largest BTT mass ratios and lowsSFR. The median gas-phase metallicity of this group is also thelowest compared to the other groups. The small sSFR does not allowfor efficient instantaneous metal and stellar dust production in suchgalaxies. However, they should have accumulated metals and dustfrom previous star-formation episodes. The lack of metals and dustin their ISM, while their halo is dust-rich, indicates that the ISMcontent is transported into the halo (see Section 5).At higher redshift, however, the halo dust-rich group occupiesan increasingly high sSFR region with a median log sSFR > theMilky Way value. The high sSFR of these galaxies means that theyare producing metals and dust in their stellar ejecta. However, unlikegalaxies in the other groups, galaxies in this group show no increasein gas metallicity with decreasing redshift. The low abundance ofmetals and dust in their ISM is possible if the feedback is efficientlyheating and blowing out the newly formed metals and dust. Suchfeedback might be provided by AGN activity (Sarangi et al. 2019).Galaxies in a deep potential well can retain the feedback-heatedgas, dust and metals in their halo. Mass in the dark matter halo usu-ally sets the depth of the potential well. Since stellar mass correlates with halo mass, halo dust-rich galaxies should be more common inthe massive galaxies. Figure 4 shows that in the massive bin, thisgroup comprise a quarter of the population, a more significant frac-tion than the lower stellar mass bin.Ménard et al. (2010) measured dust reddening effects of back-ground quasars relative to the foreground SDSS galaxies. Theyobserved the existence of diffuse dust in halos with an amount com-parable to those in the galactic disk. For 0 . 𝐿 ∗ galaxies, the authorsestimated a dust mass of 5 × M (cid:12) in the halo. Peek et al. (2015)confirmed this halo dust mass for 0 . 𝐿 ∗ − 𝐿 ∗ low redshift galaxies.Observations of CGM dust provides evidence for our predictionthat dust is transported out of the ISM and trapped in the halo.However, such observations are still uncommon. Future CGM dustmeasurements will allow for better constraints for models such as Dusty SAGE . We define galaxies that lack dust in both the ISM and their halo asdust-poor galaxies. At redshift 𝑧 = × − and are classified as ISM dust-poor as per our defi-nition. However, both datasets do not provide measurement for dustin the halo. Therefore, we assume their halo is dust-poor, and weinclude them in the dust-poor group instead of the halo dust-richgroup.Dust depletion in the ISM occurs in two ways: destruction bySN shocks inside the galaxy disk and dust that is ejected out of theISM by feedback. If the potential well of the system is enough totrap the dust, the ejected dust will end up in halo. However, galaxieswith low stellar mass usually live in shallower potentials and thus MNRAS000 , 1–12 (2021) xploring the relation between dust mass and galaxy properties using
Dusty SAGE are likely to lose their dust completely through ejection. This likelyexplains why Figure 4 shows that the dust-poor group contain themost significant fraction of their galaxies in the low mass bin.In Figure 7, the dust-poor group moves to the lower BTTmass ratio, lower sSFR region with decreasing redshift while thegas metallicity increases. At high redshift, galaxies in this grouphave relatively high sSFR, but their ISM still lacks both metalsand dust. Dust formation is highly sensitive to the abundance ofmetals. In stellar ejecta, dust forms via the condensation of metalsejected from stars. In dense clouds, dust grains grow by accretingmetals. Therefore, galaxies with low metallicity can not producedust efficiently. If this low production rate can not keep up with thedepletion rate, a galaxy will end up in the dust-poor group.Fisher et al. (2014) observed the dust emission from a localdwarf galaxy, I Zwicky 18, and found it lacks dust with a dust-to-stellar mass ratio of about 10 − to 10 − . This galaxy has a SFR of0 .
05 M (cid:12) /year and a very low metallicity of 12 + log [ O / H ] = . × M (cid:12) , which makes the sSFR = . × − , consistent with the median sSFR value of the dust-poorgalaxies in the lower mass group from Dusty SAGE (see Figure 4).Although the stellar mass of I Zwicky 18 is well below our sample,we find a consistent feature where a lack of metals accompanies thelack of dust.
During galaxy’s evolution, dust physics is heavily influenced bystar formation, gas physics, outflows and the depth of the potentialwell. Therefore, the dust, stellar, and gas content of galaxies aretightly related to each other. In this paper, we use the
Dusty SAGE galaxy formation model (Triani et al. 2020) to explore the relationbetween dust content and the fundamental properties of galaxies.We have divided the galaxy population based on their dust massand observable properties. At redshift 𝑧 =
0, the properties of ourmodel galaxies are in good agreement with observations, especiallywhere the statistics are sound.Our investigation includes the comparison of morphology, spe-cific star formation rate, metallicity and stellar mass for galaxiesgrouped by their dust content from redshift 𝑧 = 𝑧 =
0. Our mainconclusions are the following: • We find that the distribution of dust in the ISM, hot halo andejected reservoir is affected by galaxies’ morphology and stellarmass. The fraction of dust in the hot halo increases with BTT massratio and stellar mass (Figure 1). • At redshift 𝑧 =
0, ISM dust-rich galaxies have the highest gas-phase metallicities. Both star-forming and quenched ISM dust-richgroups shows a low BTT mass ratio and relatively high mediansSFR. Galaxies in the halo dust-rich and dust-poor groups are metalpoor. They are distributed in the high BTT mass ratio and low sSFRregime of Figure 4. • We see an evolution in the behaviour of our grouped galaxiesin Figure 7. From 𝑧 = 𝑧 =
0: (i) the BTT mass ratio of thequenched ISM dust-rich group drops ≈ . • Across redshift, both star-forming ISM dust-rich and quenched ISM dust-rich groups maintain their high median sSFR and high gas-phase metallicity. This implies that such galaxies are still activelyproducing metals and dust via stellar production. The abundance ofmetals in the ISM also enables dust growth via metal accretion indense clouds. • The halo dust-rich group consists of galaxies with a dust-poor ISM but dust-rich halo. They have a relatively high outflowrate of heated gas, dust, and metals out of the ISM. At present,they are elliptical with very low sSFR and metallicity. The gas-phase metallicity of these galaxies does not depend significantly onredshift. At high redshift, galaxies in this group have high sSFRthat enables an efficient dust production mechanism. Their lack ofISM dust, therefore, implies effective feedback to reheat the newlyformed dust. Because these massive galaxies have a deep potentialwell, they are more likely to retain the heated dust in their halo.Figure 4 and Figure 7 shows that this group is made up of a largerfraction of massive galaxies. • Galaxies in the dust-poor group have low dust mass both intheir ISM and halo. From redshift 𝑧 = 𝑧 =
0, this group evolvesfrom the high sSFR region to the lower sSFR region. However, theyalways have low metallicity. As metals are the main ingredients fordust, this low metallicity implies that their dust production is alsosmall.Our model provides predictions for future surveys with next-generation instruments and telescopes, such as ALMA and JWST,that will measure the dust and galaxy properties in extraordinarydetail at high redshift. JWST will cover the IR spectrum in the0.6 to 28.5 micron regime, which reveals the total galactic dustcontent up to 𝑧 =
2. At 𝑧 =
3, it will only cover the dust spectrumto ≈ Dusty SAGE , leading to a better predictions and more detailedinterpretation of the observations.
ACKNOWLEDGEMENTS
We would like to thank the anonymous referee whose valuable com-ments improved the quality of this paper, and Ned Taylor for helpfulcomments during the final stages of this work. This research wassupported by the Australian Research Council Centre of Excellencefor All Sky Astro-physics in 3 Dimensions (ASTRO 3D), throughproject number CE170100013. The Semi-Analytic Galaxy Evolu-tion (SAGE) model, on which
Dusty SAGE was built, is a publiclyavailable codebase that runs on the dark matter halo trees of acosmological N-body simulation. It is available for download at https://github.com/darrencroton/sage . This research hasused python ( ), numpy (van der Waltet al. 2011) and matplotlib (Hunter 2007). DATA AVAILABILITY
The data underlying this article are available in the article. Thegalaxy formation model used to generate the data is available at https://github.com/dptriani/dusty-sage . MNRAS , 1–12 (2021) D. P. Triani et al.
REFERENCES
Aoyama S., Hou K.-C., Shimizu I., Hirashita H., Todoroki K., Choi J.-H.,Nagamine K., 2017, MNRAS, 466, 105Asano R. S., Takeuchi T. T., Hirashita H., Inoue A. K., 2013, Earth, Planets,and Space, 65, 213Baugh C. M., Lacey C. G., Frenk C. S., Granato G. L., Silva L., Bressan A.,Benson A. J., Cole S., 2005, MNRAS, 356, 1191Bekki K., 2015, MNRAS, 449, 1625Bundy K., et al., 2006, ApJ, 651, 120Casey C. M., 2012, MNRAS, 425, 3094Clemens M. S., et al., 2013, MNRAS, 433, 695Croton D. J., et al., 2006, MNRAS, 365, 11Croton D. J., et al., 2016, ApJS, 222, 22Dekel A., Birnboim Y., 2006, MNRAS, 368, 2Draine B. T., 1990, in Blitz L., ed., Astronomical Society of the PacificConference Series Vol. 12, The Evolution of the Interstellar Medium.pp 193–205Draine B. T., 2009, in Henning T., Grün E., Steinacker J., eds, AstronomicalSociety of the Pacific Conference Series Vol. 414, Cosmic Dust - Nearand Far. p. 453 ( arXiv:0903.1658 )Dunne L., Eales S. A., Edmunds M. G., 2003, MNRAS, 341, 589Dunne L., et al., 2011, MNRAS, 417, 1510Dwek E., 1998, ApJ, 501, 643Dwek E., Scalo J. M., 1980, ApJ, 239, 193Eales S., et al., 2009, ApJ, 707, 1779Feldmann R., 2015, MNRAS, 449, 3274Fisher D. B., et al., 2014, Nature, 505, 186Graham A. W., Worley C. C., 2008, MNRAS, 388, 1708Hollenbach D., McKee C. F., 1979, ApJS, 41, 555Hunter J. D., 2007, Computing in Science and Engineering, 9, 90Jones A. P., Tielens A. G. G. M., Hollenbach D. J., McKee C. F., 1994, ApJ,433, 797Kennicutt Robert C. J., 1998, ARA&A, 36, 189Kennicutt R. C., Evans N. J., 2012, ARA&A, 50, 531Kennicutt R. C., et al., 2011, PASP, 123, 1347Kormendy J., Fisher D. B., Cornell M. E., Bender R., 2009, ApJS, 182, 216Licquia T. C., Newman J. A., 2015, ApJ, 806, 96Madau P., Dickinson M., 2014, ARA&A, 52, 415Mancini M., Schneider R., Graziani L., Valiante R., Dayal P., Maio U.,Ciardi B., Hunt L. K., 2015, MNRAS, 451, L70Marri S., White S. D. M., 2003, MNRAS, 345, 561McKee C., 1989, in Allamandola L. J., Tielens A. G. G. M., eds, IAUSymposium Vol. 135, Interstellar Dust. p. 431McKinnon R., Torrey P., Vogelsberger M., 2016, MNRAS, 457, 3775Ménard B., Scranton R., Fukugita M., Richards G., 2010, MNRAS, 405,1025Mo H. J., Mao S., White S. D. M., 1998, MNRAS, 295, 319Nersesian A., et al., 2019, A&A, 624, A80Peek J. E. G., Ménard B., Corrales L., 2015, ApJ, 813, 7Peeples M. S., Werk J. K., Tumlinson J., Oppenheimer B. D., ProchaskaJ. X., Katz N., Weinberg D. H., 2014, ApJ, 786, 54Popping G., Somerville R. S., Galametz M., 2017, MNRAS, 471, 3152Rees M. J., Ostriker J. P., 1977, MNRAS, 179, 541Rémy-Ruyer A., et al., 2014, A&A, 563, A31Riechers D. A., et al., 2013, Nature, 496, 329Roberts M. S., Haynes M. P., 1994, ARA&A, 32, 115Sánchez Almeida J., Elmegreen B. G., Muñoz-Tuñón C., Elmegreen D. M.,2014, A&ARv, 22, 71Santini P., et al., 2014, A&A, 562, A30Sarangi A., Dwek E., Kazanas D., 2019, ApJ, 885, 126Skelton R. E., Bell E. F., Somerville R. S., 2012, ApJ, 753, 44Slavin J. D., Dwek E., Jones A. P., 2015, ApJ, 803, 7Somerville R. S., Primack J. R., Faber S. M., 2001, MNRAS, 320, 504Springel V., et al., 2005, Nature, 435, 629Sutherland R. S., Dopita M. A., 1993, ApJS, 88, 253Thomas D., Maraston C., Bender R., Mendes de Oliveira C., 2005, ApJ,621, 673 Tojeiro R., et al., 2013, MNRAS, 432, 359Tremonti C. A., et al., 2004, ApJ, 613, 898Triani D. P., Sinha M., Croton D. J., Pacifici C., Dwek E., 2020, MNRAS,493, 2490Valiante R., Schneider R., Bianchi S., Andersen A. C., 2009, MNRAS, 397,1661Vijayan A. P., Clay S. J., Thomas P. A., Yates R. M., Wilkins S. M., HenriquesB. M., 2019, arXiv e-prints, p. arXiv:1904.02196Vlahakis C., Dunne L., Eales S., 2005, MNRAS, 364, 1253Watson D., Christensen L., Knudsen K. K., Richard J., Gallazzi A.,Michałowski M. J., 2015, Nature, 519, 327White S. D. M., Rees M. J., 1978, MNRAS, 183, 341Witt A. N., Gordon K. D., 2000, ApJ, 528, 799Witt A. N., Thronson Jr. H. A., Capuano Jr. J. M., 1992, ApJ, 393, 611Zhukovska S., 2014, A&A, 562, A76Zhukovska S., Gail H. P., Trieloff M., 2008, A&A, 479, 453da Cunha E., Eminian C., Charlot S., Blaizot J., 2010, MNRAS, 403, 1894da Cunha E., et al., 2015, ApJ, 806, 110van der Walt S., Colbert S. C., Varoquaux G., 2011, Computing in Scienceand Engineering, 13, 22This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS000