Large Magellanic Cloud Warm dust Cold dust Hydrogen

Evolution in the Dusty ISM Across the Local Group Chris Clark with Julia Roman-Duval, Karl Gordon, Caroline Bot, and Matthew Smith

Chris Clark | HotSci Talk, 3rd August 2022 Eales+ (2010); Valiante+ (2016); Smith+ (2017) Herschel observation of ~3 square degrees (~0.2% of Herschel extragalactic survey area) Importance of the Dust-to-Gas Ratio • Dust-to-Gas ratio (D/G) dictates what galaxies we can detect in ISM surveys. (Majority of galaxies with detection of cold ISM are detected only via cold dust emission.) • Variation in D/G affects reliability of using dust emission as gas mass tracer. • Galaxies with higher D/G will tend to suffer greater UV–optical extinction.

Chris Clark | HotSci Talk, 3rd August 2022 Variation in the Dust-to-Gas Ratio (D/G) De Vis+ (2019); Galliano+ (2018); De Cia+ (2016); Remy-Ruyer+ (2014) “Critical Metallicity” Transition High metallicity → higher D/G, due to more efficient dust grain growth Below “critical metallicity”, grain growth can’t keep up with dust destruction (shocks, radiation), so D/G much lower Location and behaviour of ‘critical metallicity’ transition is very poorly constrained

Chris Clark | HotSci Talk, 3rd August 2022 Variation in the Dust-to-Gas Ratio (D/G) Roman-Duval+ (2022); De Vis+ (2019); Galliano+ (2018); De Cia+ (2016); Remy-Ruyer+ (2014) ⬟ D/G from FIR: Compare dust mass from far-IR data to gas mass from HI & CO data …or… ◆ D/G from UV: Use UV absorption spectroscopy to find metal abundance in dust phase vs gas phase, then compare to H column But, more UV-based D/G measurements on the way soon from Julia Roman-Duval and Aleksandra Hamanowicz! “Critical Metallicity” Transition

Chris Clark | HotSci Talk, 3rd August 2022 Questions That Need Answering Roman-Duval+ (2022); De Vis+ (2019); Galliano+ (2018); De Cia+ (2016); Remy-Ruyer+ (2014) • Where is the ‘critical metallicity’ transition in the dust-to-gas ratio? • How much grain growth happens at different metallicities? • How much does the amount of grain growth vary between environments? • Why is there apparently dust “missing” from some FIR estimates of D/G? “Critical Metallicity” Transition

Chris Clark | HotSci Talk, 3rd August 2022 Well-Resolved Local Group Galaxies Clark+ (2021) Herschel 250 μm dust emission Large Magellanic Cloud Dist = 50 kpc M ⋆ = 3 x 109 M ⊙ Z = 0.5 Z ⊙ M31 Dist = 790 kpc M ⋆ = 1 x 1011 M ⊙ Z ~ 1.3 Z ⊙ Small Magellanic Cloud Dist = 62 kpc M ⋆ = 2 x 108 M ⊙ Z = 0.2 Z ⊙ M33 Dist = 840 kpc M ⋆ = 5 x 109 M ⊙ Z = 0.5 Z ⊙

Chris Clark | HotSci Talk, 3rd August 2022 Well-Resolved Local Group Galaxies Clark+ (2021) Herschel 250 μm dust emission H surface density from HI and CO

Chris Clark | HotSci Talk, 3rd August 2022 Clark+ (2021); Meixner+ (2013); Roman-Duval+ (2014) A Herschel extragalactic field; approximate area of Moon and Hubble Deep Field shown to left for comparison. Portion of LMC seen in Herschel 250 μm Herschel Data for Local Group Galaxies • Resolution of 7–36”, giving physical resolution better than 15pc in LMC. • 5 photometric filters (100, 160, 250, 350, and 500 μm), spanning whole dust SED peak. • Provides great constraints on SED shape, therefore dust mass, temperature, etc.

Chris Clark | HotSci Talk, 3rd August 2022 Clark+ (2021) Emission on large angular scales removed from Herschel maps during the data reduction Herschel data therefore missing lowest density ISM, whilst IRAS (& Planck) data missing highest density ISM Extended Emission Missing from Herschel IRAS 100 μm 300” resolution Herschel 100 μm 10” resolution

Chris Clark | HotSci Talk, 3rd August 2022 Combine Alllll the Data in Fourier Space… Clark+ (2021) COBE Far-infrared data, large angular scales IRAS Far-infrared data, medium angular scales Planck Submm data, large & medium angular scales COBE x IRAS FIR data, large and medium angular scales (COBE x IRAS) + Planck FIR-submm data, large & medium angular scales Herschel FIR-submm data, small angular scales ((COBE x IRAS) + Planck) x Herschel FIR-submm data, large & medium & small angular scales x → “Feathered with” + → “In concert with”

Chris Clark | HotSci Talk, 3rd August 2022 Restoring Extended Emission by Feathering Clark+ (2021) Herschel only Herschel feathered with IRAS, Planck, and COBE

Chris Clark | HotSci Talk, 3rd August 2022 SED Fitting Our New Herschel Data Clark+ (subm.); Gordon+ (2014) Every pixel’s Spectral Energy Distribution (SED) fit using a broken- emissivity modified blackbody model.

Chris Clark | HotSci Talk, 3rd August 2022 Pixel-by-Pixel Dust-to-Gas Ratio Clark+ (subm)

Chris Clark | HotSci Talk, 3rd August 2022 Strong Evolution in G/D versus Density Clark+ (subm.) Over 1 dex increases in G/D with density suggest very significant grain growth H surface density values on x-axis have been corrected for inclination. We can probe to 10x higher densities in LMC and SMC, because they are ~10x closer than M31 and M33, hence have ~10x better density resolution.

Chris Clark | HotSci Talk, 3rd August 2022 Chemical Evolution: Model vs Observed Clark+ (subm.); Asano+ (2013 Asano et al. (2013) model tracks: Parameters: Average grain size Dust temperature ISM metallicity Grain-growth episode timescale Minimum dust-to-metals ratio Column-to-volume density conversion

Chris Clark | HotSci Talk, 3rd August 2022 Clark+ (2018); Clark+ (2019) What Causes the Turnover in D/G? Possible explanations for turnover in D/G versus H: • SED fitting breaks down in some way above a certain density? • Dark gas (atomic and/or molecular) at intermediate densities, meaning D/G gets overestimated (Works for M31 & M33, but not LMC…) • Dust destruction by hot gas? • CO X-factor overestimated at higher densities, driving down D/G? (Would require very low Xco for M31, and especially M33...) • Dust mass absorption coefficient falling at higher densities? • Noise-induced anticorrelation?

Chris Clark | HotSci Talk, 3rd August 2022 Variation in the Dust-to-Gas Ratio (D/G) Roman-Duval+ (2022a); Clark+ (subm.); De Vis+ (2019); Galliano+ (2018); Remy-Ruyer+ (2014) ⬟ D/G from FIR: Compare dust mass from far-IR data to gas mass from HI & CO data …or… ◆ D/G from UV: Use UV absorption spectroscopy to find metal abundance in dust phase vs gas phase, then compare to H column

Chris Clark | HotSci Talk, 3rd August 2022 D/G from FIR versus D/G from UV Clark+ (subm.); Roman-Duval+ (2022); Roman-Duval+ (2017) Our new FIR D/G, and the UV spectroscopic D/G, now agree to within 17%

Chris Clark | HotSci Talk, 3rd August 2022 D/G from FIR versus D/G from UV Clark+ (subm.); Roman-Duval+ (2022); Roman-Duval+ (2017); Jenkins & Wallerstein (2017) Discrepancy between UV and FIR measurements reduced from 8x to 3x

Chris Clark | HotSci Talk, 3rd August 2022 D/G from FIR versus D/G from UV Clark+ (subm.); Roman-Duval+ (2022); Roman-Duval+ (2017); Jenkins & Wallerstein (2017) x2.96 offset If dust mass absorption coefficient was 2.96x smaller in SMC than assumed for the other galaxies, then the disagreement disappears Dust at SMC metallicities is known to have different physical properties to dust at higher metallicities

Chris Clark | HotSci Talk, 3rd August 2022 Variation in the Dust-to-Gas Ratio (D/G) Roman-Duval+ (2022); Clark+ (subm.); De Vis+ (2019); Galliano+ (2018); Remy-Ruyer+ (2014) ⬟ D/G from previous FIR ◆ D/G from UV ⬢ D/G from our new FIR

Large Magellanic Cloud Warm dust Cold dust Hydrogen

Chris Clark | HotSci Talk, 3rd August 2022 Conclusions & Takeaways Clark+ (subm.) • Our new Herschel maps of Local Group galaxies M31, M33, LMC, and SMC, restore the previously-removed large–scale dust emission. Data publicly available! • Combination of high resolution, and large-scale sensitivity, allows us to probe a very wide of ISM densities (up to 2 dex), and explore dust grain growth. • We find evidence for strong grain-growth, causing over 1 dex of evolution in D/G ratio. • However, D/G, and how it evolves, is not set purely by a galaxy’s mass and metallicity. • “Turnover” in D/G at higher densities, and the lingering discrepancy between UV depletion and FIR values of D/G for the SMC, hint at possible changes in dust mass absorption coefficient.

Chris Clark | HotSci Talk, 3rd August 2022 Image credit: ESA Herschel recap: • ESA mission, from 2009–2013 • 3.5m primary mirror • Photometry & spectroscopy from 52–670 μm • Six photometric filters, observed using two instruments: PACS: 70, 100, 160 μm SPIRE: 250, 350, 500 μm • Resolution 7–36 arcsec

Chris Clark | HotSci Talk, 3rd August 2022 Missing Diffuse Emission in Herschel Clark+ (2021) These patches of background have roughly the same brightness These patches of background have very different brightness Superbubbles contain significant emission Superbubbles have had almost all emission removed 100 μm observations of the Large Magellanic Cloud, from IRAS and Herschel IRAS 100 μm = 300 arcsec resolution Herschel 100 μm = 10 arcsec resolution

Chris Clark | HotSci Talk, 3rd August 2022 Power Spectra of Data Clark+ (2021)

Chris Clark | HotSci Talk, 3rd August 2022 Relative Change due to Feathering Clark+ (2021) LMC SMC M31 M33 Plots showing how much emission was restored. Contours show outlines of the galaxies. (Regions can also have flux “removed”, because the zero-point in the old Herschel maps was somewhat arbitrary. Our feathering also restored large-scale MW cirrus structure in foreground, making some regions become fainter.)

Chris Clark | HotSci Talk, 3rd August 2022 Relative Difference When Feathering Clark+ (2021) LMC SMC M31 M33

Chris Clark | HotSci Talk, 3rd August 2022 Foreground Subtraction Roman-Duval+ (2017); Clark+ (subm.)

Chris Clark | HotSci Talk, 3rd August 2022 More Emission Restored at Shorter λ Clark+ (2021) More emission needed to be restored in the shorter- wavelength Herschel bands than in the longer wavelengths Therefore, In new maps, galaxies are significantly bluer, especially in diffuse peripheries (20–30% bluer 500/160 colour) The dust temperature is main driver of FIR SED colour, so change in colour will significantly modelled dust mass

Chris Clark | HotSci Talk, 3rd August 2022 D/G Turnover Caused by Falling ⍺CO ? Clark+ (2021)

Chris Clark | HotSci Talk, 3rd August 2022 Composition vs Density in MW, LMC, SMC Clark+ (2021); Roman-Duval+ (2022a)

Chris Clark | HotSci Talk, 3rd August 2022 Clark+ (2018); Clark+ (2019); Roman-Duval+ (2022a) Dust Emissivity Changes With ISM Density? H Surface Density Dust Mass Absorption Coefficient SMC Dust Composition Recent work shows that chemical composition of dust changes with density – and does so differently in SMC vs LMC vs Milky Way (see Roman-Duval+ 2022a) Previous work found evidence for dust mass absorption coefficient falling with density. Milky Way Dust Composition

Chris Clark | HotSci Talk, 3rd August 2022 D/G Evolution vs Ionised Gas Density Clark+ (2021)

Chris Clark | HotSci Talk, 3rd August 2022 D/G Evolution with Degraded Resolution Clark+ (2021)

Chris Clark | HotSci Talk, 3rd August 2022 Dust MAC Falls at Higher Density? Clark+ (subm.); Clark+ (2019) Fainter points: Dust mass absorption coefficient is constant Bold points: Dust mass absorption coefficient falls according to power law of -0.4 above a transition density (M31 & M33 transition density = 4 M ⊙ pc-2 LMC transition density = 40 M ⊙ pc-2)

Chris Clark | HotSci Talk, 3rd August 2022 Dark Gas Candidates in the LMC? Clark+ (subm.); Clark+ (2019) 4 times less H observed from HI and CO, than predicted from dust data (when using overall LMC trends for dust vs H).

Chris Clark | HotSci Talk, 3rd August 2022 D/G of Damped Lyman-α Systems Roman-Duval+ (subm.); De Vis+ (2019); Galliano+ (2018); De Cia+ (2016); Remy-Ruyer+ (2014) D/G from emission: Compare dust mass from far-IR data to gas mass from HI & CO data …or… D/G from depletions: Use UV absorption spectroscopy to find metal abundance in dust phase vs gas phase, then compare to H column