Probing the dusty stellar populations of the Local Volume Galaxies with JWST/MIRI

TL;DR

This paper demonstrates how JWST/MIRI can be used to identify and classify various stellar populations in Local Volume galaxies through color-color diagrams and filter selection strategies, leveraging existing infrared data.

Contribution

It introduces new MIRI color classification schemes and evaluates optimal filter combinations for stellar population studies in nearby galaxies.

Findings

MIRI colors effectively distinguish young stellar objects and evolved stars.

Identified optimal filter combinations for population selection.

Provided a framework for future JWST observations of Local Volume galaxies.

Abstract

The Mid-Infrared Instrument (MIRI) for the {\em James Webb Space Telescope} (JWST) will revolutionize our understanding of infrared stellar populations in the Local Volume. Using the rich {\em Spitzer}-IRS spectroscopic data-set and spectral classifications from the Surveying the Agents of Galaxy Evolution (SAGE)-Spectroscopic survey of over a thousand objects in the Magellanic Clouds, the Grid of Red supergiant and Asymptotic giant branch star ModelS ({\sc grams}), and the grid of YSO models by Robitaille et al. (2006), we calculate the expected flux-densities and colors in the MIRI broadband filters for prominent infrared stellar populations. We use these fluxes to explore the {\em JWST}/MIRI colours and magnitudes for composite stellar population studies of Local Volume galaxies. MIRI colour classification schemes are presented; these diagrams provide a powerful means of identifying young stellar objects, evolved stars and extragalactic background galaxies in Local Volume galaxies with a high degree of confidence. Finally, we examine which filter combinations are best for selecting populations of sources based on their JWST colours.

Tables

Table 1. Table 1 Classification types used by SAGE-Spec (Woods et al., 2011a ) , which we adopt throughout this paper.

Code	Object type	Number
C-AGB	Carbon-rich AGB stars	145
C-PAGB	Carbon-rich post-AGB stars	119
C-PN	Carbon-rich planetary nebulae	113
O-AGB	Oxygen-rich AGB stars	173
O-PAGB	Oxygen-rich post-AGB stars	123
O-PN	Oxygen-rich planetary nebulae	127
RSG	Red Supergiants	174
STAR	Stellar photospheres	130
YSO-1	Embedded Young Stellar Objects	153
YSO-2	Young Stellar Objects	114
YSO-3	Evolved Young Stellar Objects	177
YSO-4	HAeBe Young Stellar Objects	121
HII	H ii regions	134
GAL	Galaxy	136

Table 2. Table 2 Estimated effective wavelengths and zero points for the MIRI filters.

Filter	Calc. ${\lambda }_{\mathrm{eff}}$	$\Delta \lambda$	Est. Zero Point
	$\mu$m	$\mu$m	Jy
F560W	$-$	1.2	116.60
F770W	7.62	2.2	68.10
F1000W	9.94	2.0	40.71
F1130W	11.31	0.7	31.82
F1280W	12.79	2.4	24.76
F1500W	15.04	3.0	17.99
F1800W	17.96	3.0	12.46
F2100W	20.75	5.0	9.13
F2550W	25.32	4.0	6.18

Table 3. Table 3 MIRI fluxes (mJy) for our sample, calculated according to equation  3 .

Class	[F770W]	[F1000W]	[F1130W]	[F1280W]	[F1500W]	[F1800W]	[F2100W]	[F2550W]
GAL	4.7 $\pm$ 0.3	10.0 $\pm$ 0.4	3.0 $\pm$ 0.3	7.0 $\pm$ 0.4	9.8 $\pm$ 0.6	7.2 $\pm$ 0.8	10.5 $\pm$ 0.5	14.4 $\pm$ 1.2
C-AGB	26.4 $\pm$ 1.0	27.6 $\pm$ 0.9	28.9 $\pm$ 0.8	18.8 $\pm$ 1.2	14.1 $\pm$ 0.7	12.3 $\pm$ 0.9	10.5 $\pm$ 0.5	8.6 $\pm$ 0.5
RSG	24.3 $\pm$ 1.3	18.3 $\pm$ 0.7	16.5 $\pm$ 0.7	12.9 $\pm$ 1.1	10.2 $\pm$ 0.7	10.1 $\pm$ 0.6	8.7 $\pm$ 0.4	7.0 $\pm$ 0.5
YSO-2	56.6 $\pm$ 2.1	45.7 $\pm$ 1.9	59.5 $\pm$ 2.3	67.8 $\pm$ 2.3	69.5 $\pm$ 2.8	72.5 $\pm$ 2.9	77.5 $\pm$ 2.7	97.6 $\pm$ 5.7
O-AGB	22.4 $\pm$ 1.2	55.4 $\pm$ 0.9	51.1 $\pm$ 0.8	33.1 $\pm$ 1.1	18.6 $\pm$ 0.6	15.2 $\pm$ 0.7	12.7 $\pm$ 0.5	8.9 $\pm$ 0.4

Table 4. Table 4 Estimated MIRI Magnitudes and uncertanties for our LMC and background galaxy sample.

Class	[F770W]	[F1000W]	[F1130W]	[F1280W]	[F1500W]	[F1800W]	[F2100W]	[F2550W]
GAL	10.37 $\pm$ 0.06	9.05 $\pm$ 0.04	10.05 $\pm$ 0.09	8.82 $\pm$ 0.05	8.15 $\pm$ 0.06	8.07 $\pm$ 0.11	7.35 $\pm$ 0.06	6.60 $\pm$ 0.04
C-AGB	8.52 $\pm$ 0.04	7.92 $\pm$ 0.04	7.60 $\pm$ 0.03	7.79 $\pm$ 0.07	7.74 $\pm$ 0.05	7.51 $\pm$ 0.07	7.35 $\pm$ 0.04	7.16 $\pm$ 0.06
RSG	8.62 $\pm$ 0.06	8.37 $\pm$ 0.04	8.21 $\pm$ 0.05	8.20 $\pm$ 0.09	8.10 $\pm$ 0.07	7.72 $\pm$ 0.06	7.56 $\pm$ 0.05	7.38 $\pm$ 0.09
YSO2	7.68 $\pm$ 0.04	7.37 $\pm$ 0.04	6.81 $\pm$ 0.04	6.39 $\pm$ 0.04	6.00 $\pm$ 0.04	5.58 $\pm$ 0.04	5.18 $\pm$ 0.04	4.52 $\pm$ 0.04
O-AGB	8.69 $\pm$ 0.05	7.16 $\pm$ 0.04	6.98 $\pm$ 0.04	7.18 $\pm$ 0.04	7.45 $\pm$ 0.04	7.28 $\pm$ 0.04	7.15 $\pm$ 0.04	7.11 $\pm$ 0.07

Table 5. Table 5 The 10 σ 𝜎 \sigma 10000s magnitude limits for the MIRI filters, estimated from the zero points given in Table  2 .

Filter	Magnitude
F560W	22.15
F770W	21.09
F1000W	19.69
F1130W	18.43
F1280W	18.67
F1500W	17.78
F1800W	16.39
F2100W	15.27
F2550W	13.43

Equations

$$f_{\nu }^{\mathrm{MIRI}}({\lambda }_{\mathrm{eff}})=F_{\nu }({\lambda }_{\mathrm{eff}})\times K,$$

$$K=\frac{\int (F_{\nu }/F_{{\nu }_o})(\nu /{\nu }_{\mathrm{eff}}{)}^{-1}\text{leavevmode}{R}_{\gamma }\text{leavevmode}d\nu }{\int (\nu /{\nu }_{\mathrm{eff}}{)}^{-2}\text{leavevmode}{R}_{\gamma }\text{leavevmode}d\nu }\text{leavevmode},$$

$$⟨F_{\nu }⟩=\frac{\int F_{\nu }\text{leavevmode}{\nu }^{-1}\text{leavevmode}{R}_{\gamma }\text{leavevmode}d\nu }{\int {\nu }^{-1}\text{leavevmode}{R}_{\gamma }\text{leavevmode}d\nu }.$$

$${\lambda }_{\mathrm{eff}}=\frac{\int {\lambda }^2{R}_{\gamma }\text{leavevmode}d\lambda }{\int \lambda R_{\gamma }\text{leavevmode}d\lambda }\text{leavevmode}.$$

$$[M_i]=-2.5\mathrm{lo}{\mathrm{g}}_{10}(f_{\nu }^{\mathrm{MIRI},i}/f_{i,\mathrm{zero}}).$$

Full text

1. Introduction

The James Webb Space Telescope (JWST) will revolutionize our knowledge of the stellar populations of galaxies out to and beyond the Virgo cluster. The large (6.5 m) aperture of JWST combined with its subarcsecond spatial resolution and broad wavelength coverage (0.6–28.5 $\mu$m) will provide unprecedented opportunities to study the properties of resolved dusty stellar populations at moderate and large distances ($\sim$4 Mpc). Notably, JWST will enable the accurate measurement of the star formation histories of nearby galaxies, determine the age of the oldest stellar populations and characterize metal enrichment in a wide variety of environments.

Infrared observations of stellar populations are critical for studies of chemical enrichment, as both the early and late stages of a star’s life are enshrouded in dust and molecular gas. However, the ability to correctly interpret infrared observations (especially for galaxies at high-redshift) relies on accurate constraints of the stellar populations. Currently, resolved mid-infrared stellar population studies of galaxies with multiple populations are limited to nearby galaxies such as the Magellanic Clouds. With JWST this will be possible for well over a hundred galaxies within the Local Volume.

The dusty stellar populations of the Large and Small Magellanic Clouds (D = 50 and 60 kpc; Ngeow & Kanbur 2008; Szewczyk et al. 2009; Feast 2013), with metallicities of 0.5 and 0.2 $Z_{\odot}$ (Russell & Dopita, 1992), have been well characterized in the IR from 3.6–500 $\mu$m using the Spitzer and Herschel Space Telescopes (Meixner et al., 2006, 2013; Gordon et al., 2011). Over 10.5 million point sources were photometrically detected, and key populations of sources throughout the Magellanic Clouds identified in the colour magnitude diagrams (CMDs). Furthermore, over 1250 of these sources were observed with Spitzer’s InfraRed Spectrograph (IRS), revealing the mineralogy and evolutionary status of these sources (Woods et al., 2011a; Ruffle et al., 2015, Jones et al. in press.). This large spectroscopic data-set and corresponding point source classification is an ideal empirical library of template spectrum, which we can use to predict the flux densities and magnitudes that we will detect with the JWST Mid-Infrared Instrument (MIRI; Rieke et al. 2015).

The Mid-Infrared Instrument (MIRI; Rieke et al. 2015) on board JWST will provide broad-band imaging, coronography and spectroscopy over the 5–28.5 $\mu$m wavelength range (Rieke et al., 2015; Wright et al., 2015; Bouchet et al., 2015). MIRI imaging will be a factor of 50 more sensitive than Spitzer’s IRAC and MIPS instruments with a seven fold improvement in resolution; the field-of-view for the MIRI imager (MIRIM) is 74” $\times$ 113”, with a plate scale of 0.11 arcsec/pixel and a full-width half maximum of $\sim$0.7” (Bouchet et al., 2015). MIRIM provides continuous mid-IR wavelength coverage for broad-band imaging in nine filters from 5.6–25.5 $\mu$m; these filters have been optimized for the detection of the most astronomically relevant molecular and dust species. For instance, the F1130W and F1000W filters are sensitive to PAH’s and silicate dust respectively.

The MIRI imager is significantly more sensitive than the Medium Resolution Spectrograph (MRS), with a larger field-of-view, which means that studies of stellar population on a galactic scale can be done in a reasonable time. The down side is that it is more difficult to correctly identify the nature of the observed object. In this paper we will evaluate potential photometric classifications with JWST/MIRI using the stellar populations of the Large Magellanic Cloud (LMC) as a template. In Section 2 we describe our method for calculating the MIRI synthetic photometry. Section 3 presents the results of these calculations and shows some illustrative MIRI colour-magnitude diagrams for the stellar populations of the Magellanic Clouds. Finally in Section 4 we discuss how to apply this study to more distant galaxies.

2. Method

The nine JWST/MIRI broadband imaging filters will generate a wealth of photometric data for mid-IR stellar populations. In the following section we use the MIRI filter curves from Glasse et al. (2015) to develop colour selection criteria that can isolate and differentiate IR stellar populations. This analysis will provide guidance for planing JWST observing programs and determining point-source classifications across a range of metallicities.

2.1. Mid-Infrared Spectroscopic Templates

2.1.1 Observational data

In order to predict fluxes and colours through the MIRI filters we have made use of Spitzer-IRS spectra (Houck et al., 2004), from the SAGE-Spec legacy survey of the LMC (Kemper et al., 2010). The IRS spectra cover a wavelength range from 5.3 to 38 $\mu$m with spectral resolutions, R$=\lambda$/$\Delta\lambda$ $\sim$60–600. The distance to the LMC is well known (see Ngeow & Kanbur, 2008; Szewczyk et al., 2009; Feast, 2013), the galaxy’s inclination angle is nearly face on ($\sim$30 degrees; van der Marel & Cioni 2001) and all the stellar populations are at essentially the same distance, thus, the LMC Spitzer-IRS spectra can be used to predict fluxes and colours for any photometric filter within this range. As there is no spectral coverage by the Spitzer-IRS at $\lambda<5.3$ $\mu$m we cannot predict fluxes for the F560W filter. For all sources in the LMC we adopt a distance modulus of 18.49 $\pm$ 0.05 (Pietrzyński et al., 2013).

Over 1000 point sources in the LMC have been observed with the Spitzer IRS and homogeneously reduced by the SAGE-Spec Spitzer legacy program (Kemper et al., 2010; Woods et al., 2011b), yielding a coherent spectral catlaogue. The reduced Spitzer IRS data can be obtained from the NASA/IPAC Infrared Science Archive.111http://irsa.ipac.caltech.edu/data/SPITZER/SAGE These sources cover the range in luminosities and colours parametrized by IRAC, MIPS, and 2MASS magnitudes found in the SAGE-LMC photometric survey (Meixner et al., 2006). They have been classified in a uniform manner with respect to their base properties according to their mid- and far-infrared spectral characteristics, SED shape, pulsation period, and bolometric magnitude by Woods et al. (2011a); Ruffle et al. (2015) and Jones et al. (in press) using a classification flow chart (fig. 3 in Ruffle et al. 2015).

The point sources observed by the SAGE-Spec survey have been separated according to their evolutionary stage (young stellar objects, main sequence star, asymptotic giant branch star, post-asymptotic giant branch and planetary nebula), chemistry (Oxygen- or Carbon-rich) and by mass in the case of red supergiants. The Young Stellar Object (YSO) class has been further subdivided on the bases of their spectral features, i.e. embedded YSOs (YSO-1) have silicate and ice absorption features at 10 $\mu$m and 15 $\mu$m; conversely late-type massive YSOs (YSO-3) have silicate features in emission. As these stars become hotter and excites its environment (they develop a compact H ii region) the spectra exhibit Polycyclic Aromatic Hydrocarbons (PAHs) emission features and atomic emission lines, indicative of a UV radiation field. The intermediate-mass YSO-4 class corresponds to candidate Herbig AeBe (HAeBe) stars. The spectral features of YSOs can be used as a proxy for evolutionary stage: evolving from heavily embedded, YSO-1s to compact H ii regions.

A limited number of sources classified as YSO-3 may be confirmed as compact H ii regions with JWST. This confusion arises because evolved YSOs and (ultra)compact H ii region have a continuum of similar properties, and the spatial resolution probed by the Spitzer-IRS in the LMC is at best $\sim 1pc\leavevmode\nobreak\ (4.2^{\prime\prime})$. This makes it challenging to resolve compact H ii regions and separate the contributions of the YSO from its environment. Readers interested in evolved-YSOs should therefore consider both the YSO-3 and H ii categories.

Several mid-IR spectroscopic studies of these spectral-classes have been published in the literature. Buchanan et al. (2006, 2009) obtained and classified the spectra of 123 of the 250 most luminous 8 $\mu$m sources in the LMC. Whilst, Kraemer et al. 2017 classify the bright mid-infrared population of the SMC. Objects specifically targeted were various intermediate-to-high-mass post-main sequence stars or stars which are undergoing significant mass loss.

Evolved stars were the focus of several Spitzer-IRS programs (PID: 200, 1094, 3591, 3505, 3583, 50147 and 50167); their selection was predominantly based on near- and mid-IR colour classification schemes (e.g. Egan et al. 2001 for GTO and cycle 1–3 programs or the SAGE evolved star photometric cuts by Blum et al. 2006, for later cycles), and in some instances includes variables identified in ground-based programs. The spectral properties of the carbon stars observed in the Magellanic Clouds was published by Sloan et al. (2016); the sample of O-rich AGB stars and RSGs was published by (Jones et al., 2012). A large subset (145) of the dust-producing AGB and RSG spectroscopic sample has independently been characterized by Groenewegen et al. (2009).

Post-AGB stars and PNe were targeted by programs: 103, 20443, 30788, 50092 and 50338, and their spectra published by Stanghellini et al. (2007); Bernard-Salas et al. (2008, 2009); Volk et al. (2011); Gielen et al. (2011); Matsuura et al. (2014); Sloan et al. (2014). The spectra of the PNe were obtained early on in the Spitzer mission; the majority of these sources had been previously observed with HST. Post-AGB stars were only explicitly targeted late on in Spitzers cryogenic mission. Candidate post-AGB stars were identified using a combination of flux limits and mid-IR colours to exclude YSOs and supergiants. The post-AGB candidates were then cross-matched with the literature to ensure a clean selection.

There are $\sim$300 YSO and Hii spectra observed in the LMC, these are predominantly from program 40650 (PI: Looney) and their spectra published by Seale et al. (2009). YSO spectroscopic targets were selected from the Gruendl & Chu (2009) catalogues of YSO candidates. These catalogues of high-mass YSOs in the LMC, were produced using aperture photometry on the SAGE images.

The LMC spectroscopy available in the Spitzer archive is biased towards the science goals of the PI and the selection effects of the original programs. Spectroscopic studies performed in cycles 1–3 were skewed towards the brightest sources in the LMC as target selection was limited by the Midcourse Space Experiment (MSX; Egan et al. 2001) sensitivity limits (7.5 mag at 8 $\mu$m) or was biased towards categories of object known prior the launch of Spitzer. The reader is referred to table 2 of Kemper et al. (2010) for a comprehensive description of all the Spitzer programs which have targeted objects in the LMC with the IRS.

To eliminate biases in the LMC spectroscopic sample, the Spitzer SAGE-Spec program (PID: 40159) targeted unexplored and underrepresented region of IRAC/MIPS colour-magnitude space for a range of object classes. Spectroscopic candidates were carefully selected to cover the full range in luminosities and colour found in the SAGE data, whilst simultaneously sampling the key phases of stellar evolution. The sensitivity limit of the SAGE-Spec survey is $[8.0]=7.78+0.98\times([8.0]-[24])$ and corresponds to a bolometric magnitude of M${}_{\rm bol}<-3.75$. It should also be noted that our spectral catalogue is not very sensitive to low-mass ($<$5 M*⊙*) sources and in crowded star-forming regions some of the IRS sources may be small unresolved clusters (Ward et al., 2017) which would be resolved into multiple components with JWST.

The Spitzer-IRS spectra sample the complete range of object classes found in the infrared stellar populations of the Magellanic Clouds. Table 1 gives a summary of the spectral classification groups used by Woods et al. (2011a); Ruffle et al. (2015) and Jones et al. (in press). This spectral inventory can act as an empirical template for analyzing and interpreting future infrared data on stellar populations for more distance galaxies in both the local and high-redshift universe.

As a check on the flux calibration of the spectra, we compare the IRS spectra to the SAGE photometry at 8.0 and 24 $\mu$m. The absolute flux calibration of the SAGE photometry has a higher fidelity than the Spitzer IRS spectra and is accurate to 3% (Rieke et al., 2004; Reach et al., 2005; Engelbracht et al., 2007; Bohlin et al., 2011). Spectrophotometry for each source was synthesized from the Spitzer spectra at the effective central wavelength in each filter and checked against the observed IRAC and MIPS observations at 8 and 24 $\mu$m. About 17% of the spectral observations are offset from the Spitzer SAGE photometry by over 10%. In some cases this is due to stellar variability, as both AGB stars and YSOs can show strong variations in brightness. To alleviate disparities in the flux, we exclude sources which show deviations which cannot be accounted for by pulsations. These sources are not included in Table 1. Any remaining systematic effects are minimal since we are looking at a large population of stars.

In addition to stellar sources, we also incorporate into our sample Spitzer-IRS spectra for a broad range of galaxy morphologies, including ellipticals, spirals, merging galaxies, blue compact dwarfs, and luminous infrared galaxies. The galaxies in our sample are within $z<0.05$ and their properties have been summarized by Brown et al. (2014). Low resolution IRS spectra for these sources covering the 5.2–38 $\mu$m wavelength range were obtained from version seven of the Cornell Atlas of Spitzer IRS Sources (Lebouteiller et al., 2011), using the tapered column extraction that is optimized for extended sources. The inclusion of these galaxies (denoted as GAL) in our sample allows us to determine the level of confusion between the stellar population of the host galaxy and unresolved background contaminating point sources.

2.1.2 Data from radiative transfer models

To complement the observed Spitzer-IRS spectra we have also obtained model data from the Grid of Red supergiant and Asymptotic giant branch star ModelS (grams; Sargent et al. 2011; Srinivasan et al. 2011), and the grid of YSO models developed by Robitaille et al. (2006). These model grids have the advantage over observed sources in regard to lack of noise in the data and the fundamental knowledge about their stellar properties. This allows us to trace the evolution of a source using physical quantities in colour-magnitude space. It also enables us to probe different mass regimes to what the Spitzer-IRS sample is sensitive too. However, the models suffer from some limitations; they do not account for molecular nuances and the complex dust emission from various species present in the spectra, for instance the YSO models do not include contributions from PAH emission or atomic emission lines. They also do not include multiple stellar sources, which can influence the near- and mid-IR emission. As both the observed spectra and the models have their strengths an weaknesses we use both methods to investigate MIRI colours.

The grams models of carbon- and oxygen-rich AGB stars and RSGs samples a large range of stellar and spherical dust shell parameters relevant to evolved stars which are undergoing mass loss. The grams models were produced using the 2Dust radiative transfer code (Ueta & Meixner, 2003) and were constructed around model photospheres computed by Kučinskas et al. (2005) and Aringer et al. (2009) for the M and C stars, respectively. The oxygen-rich models were computed by Sargent et al. (2011) using oxygen-deficient silicate grains (i.e. silicates that have not reached their stochiometric proportions) from Ossenkopf et al. (1992), while the carbonaceous dust (Srinivasan et al., 2011) is composed of a mixture of 90% amorphous carbon (Zubko et al., 1996) and 10% silicon carbide (SiC; Pegourie 1988). The grams models do not distinguish O-rich AGB stars from RSGs. For both the carbon and oxygen rich grids, we select models with a stellar effective temperature of Teff = 2100–4700 K, a dust shell inner radius of Rin = 3 or 7 Rstar and log(g) = -0.5. These parameters were chosen as they are representative of the range of values expected for AGB stars, and together with mass-loss they have greatest influence upon the model output.

For the YSOs we use the precomputed 2D radiative transfer model grid developed by Robitaille et al. (2006). These models cover a range of stellar masses from 0.1–50 M*⊙* and assume a young central source (0.001–10 Myr) with a rotationally flattened infalling envelope, bipolar cavities and a flared accretion disk. Each model SED is computed at 10 inclination angles. The dust in the YSO models is represented by a mixture of astronomical silicates and graphite at solar abundances from the optical constants of Laor & Draine (1993). The grain size distribution varies with location in the disk and envelope. The models are divided into three stages, ranging from the early envelope infall to when the young star is dispersing its protoplanetary disk.

Stage I YSOs have substantial envelopes, and the central source accretes dust and gas at rates of $\dot{M}_{\rm env}/{M_{*}}>10^{-6}\,{M_{\odot}}$ yr*-1*. Stage II YSOs have a dispersed protostellar envelope and optically thick disks $\dot{M}_{\rm env}/{M_{*}}<10^{-6}\,{M_{\odot}}$ yr*-1* and $M_{\rm{disk}}/M_{\star}>10^{-6}$. Finally, stage III YSOs have optically thin disks $\dot{M}_{\rm env}/{M_{*}}<10^{-6}\,{M_{\odot}}$ yr*-1* and $M_{\rm{disk}}/M_{\star}<10^{-6}$.

The Woods et al. (2011a) YSO-1 and YSO-2 objects are synonymous with the Stage I sources from the Robitaille et al. (2006) models; whilst the YSO-3 and YSO-4 class is the analogue of Stage II objects from the high and intermediate mass regimes, respectively.

As the relative importance of the various parameters in the Robitaille et al. (2006) models are dependant on the given evolutionary stage, we do not limit a parameters range to follow a particular star throughout its evolution, instead we focus on YSOs in three stellar mass regimes: 0.2, 2.0 and 20 $\pm 2.5\%$ $M_{\odot}$ at a disk inclination of 48.5*∘*. These values are representative of the minimum YSO mass a JWST program may detect in a reasonable integration time for the Milky Way, the Magellanic Clouds and more distant local group galaxies.

2.2. Predicted MIRI Photometry

In order to calculate predicted MIRI broadband flux densities and colours for mid-IR stellar populations we use the same convention outlined by earlier infrared space telescopes such as IRAS, ISO, Spitzer and WISE (e.g. Beichman et al., 1988; Blommaert et al., 2003; Reach et al., 2005; Wright et al., 2010). Here photometry for the desired photometric band $f_{\nu}^{\rm{MIRI}}(\lambda_{\rm{eff}})$ is synthesised from a convolution of the template source spectra $F_{\nu}$ with the MIRI filter functions from Glasse et al. (2015).

The MIRI relative system response function $R_{\gamma}$, defined as the fraction of detected electrons per photon crossing the focal plane of the MIRI Imager were created by Glasse et al. (2015) from two independent measurements. The response function for the MIRI filters are in units of electrons per photon. This can be converted to a photon-counting response function $R$ in units of electrons per unit energy, by re-normalizing the product of $\lambda R_{\gamma}$ (e.g.  Bessell, 2000).

Fluxes in the MIRI wavebands were calculated by integrating the Spitzer/model spectra ($F_{\nu}$) of each source over the MIRI spectral response according to the following equations:

[TABLE]

where $\lambda_{\rm{eff}}$ is the effective wavelength of the filter, $F_{\nu}(\lambda_{\rm{eff}})$ is the flux density of the input Spitzer spectrum at $\lambda_{\rm{eff}}$ and $f_{\nu}^{\rm{MIRI}}(\lambda_{\rm{eff}})$ is the monochromatic flux density that would be given if the source was observed by the MIRI imager. The dimensionless quantity $K$ is defined in terms of frequency units by

[TABLE]

where $F_{\nu_{o}}$ is a reference spectrum, this is assumed to be a $\nu F_{\nu}$ = constant flux spectrum (see e.g. Blommaert et al., 2003; Reach et al., 2005; Hora et al., 2008; Bohlin et al., 2011).

Alternatively, the flux can be calculated in terms of the photon weighted mean flux over the bandpass. This effective flux is defined in frequency units as

[TABLE]

This method is traditionally used by HST and optical telescopes; it does not involve colour corrections or effective wavelengths, prevalent in flux calculations for IR space-based telescopes.

The choice of methodology between equations 1 and 2, and equation 3 can return a flux for the MIRI filters ($f_{\nu}^{\rm{MIRI}}$) which differs by a few percent (typically $\lesssim 3\%$). However the divergence can be more severe if the shape of source spectrum significantly differs from that of the reference spectrum. By convention the Jansky systems assumes a $\nu F_{\nu}$ = constant reference system.

When calculating uncertainties associated with the MIRI fluxes; we first consider the flux uncertainties of the observed Spitzer IRS sample and the absolute flux calibration errors. The formal uncertainties in the mean ($\sigma/\sqrt{N}$) and systematic uncertainties from the absolute flux calibration (Sloan et al., 2015); the presence of other sources in the slit; and pointing errors, which introduce discontinuities between the segments, need to be assessed. The flux uncertainties and calibration are described in detail by Woods et al. (2011b); Lebouteiller et al. (2011), and are summarized in the appendix.

To estimate the total uncertainty of a flux we use the quadratic sum of these measured and systematic errors. For the grams and YSO models we assume a total uncertainty of 5%, to reflect the uncertainty in the stellar atmosphere models.

The final uncertainties in the MIRI fluxes were computed by propagating the uncertainties from the spectra and models.

In Figure 1 we present the Spitzer-IRS spectra of a representative subsample of evolved stars, YSOs and galaxies and the MIRI broadband flux densities these sources would have if they were observed with JWST. These sources are representative of their stellar class and can be characterized by their spectral appearance in the mid-IR.

2.3. Effective wavelength

As noted earlier the flux density of a source is evaluated at the effective wavelength of the filter. The optimal choice for the effective wavelength is the filter wavelength which is least sensitive to the spectral shape of the source i.e. the weighted average wavelength, defined as

[TABLE]

This is related to the effective-frequency via $\nu_{\rm eff}=c/\lambda_{\rm eff}$. Table 2 gives the effective wavelengths calculated using the current spectral response curves ($R_{\gamma}$) for the MIRI filters (Glasse et al., 2015).

2.4. Magnitudes

We estimate zero-magnitude flux densities for the MIRI bands to determine magnitudes of our template sources from their predicted flux densities. We define the magnitude system such that

[TABLE]

The zero-points magnitudes are estimated relative to an ‘ideal’ Vega photospheric reference spectrum, and were determined by integrating a single-temperature Kurucz model spectrum of Vega (Cohen et al., 1992) over the MIRI bands, using the method outlined in equations 1 and 2. This ideal Vega spectrum had a uniform effective temperature across the stellar surface and no infared excess.

By using the Vega magnitude system we eliminate the dependence on spectral shape, this follows the methodology outlined by Reach et al. (2005) to define the Spitzer/IRAC Magnitude System and should be comparable to other magnitudes relative to Vega in optical and infrared systems. Table 2 gives the resulting zero-magnitude flux densities for the MIRI channels. These zero points do not account for in-orbit telescope performance and will need to be corrected once the total system throughput is verified post-launch.

3. Results

We derive MIRI synthetic photometry for over 1000 Spitzer-IRS spectra with known object types. This photometry is analogous to the flux density (in Janskys) that would be obtained if a source was observed with the MIRI instrument. Table 3 presents the MIRI flux density calculated according to equation 3. The source type and predicted MIRI magnitudes calculated according to equation 5 are given in Table 4; all reported magnitudes are in the Vega System.

From this library of stellar classes, MIRI fluxes and magnitudes presented in Tables 3 and 4 it is straightforward to compute the expected fluxes and colours for the stellar populations of more distant galaxies. The flux listed in Table 3 scales with $1/(d/50)^{2}$ where d is in kpc and the magnitudes can simply be adjusted to the required distance modulus from that of the LMC, $M-m=$ 18.49 $\pm$ 0.05 mag (Pietrzyński et al., 2013).

The types of objects in our sample are representative of the IR stellar populations of Local Group galaxies and includes both YSOs and evolved stars. One of the major contaminants of stellar populations studies are unresolved background galaxies. In order to explore the background galaxy population we group all the spectra of individual galaxies into one class (GAL) and do not distinguish between galaxies based on their morphological type. It is important to note that the flux of the background galaxies should not be scaled with distance.

These results can be used to guide the choice of filter with which to observe specific object classes with JWST and to identify key populations in colour-colour and colour magnitude diagrams once JWST observations become available. The colour is independent of distance.

3.1. MIRI Colours

We present example colour-magnitude (CMD) and colour-colour diagrams (CCD) to aid the identification of the various IR stellar and extragalactic populations which will be observed with JWST. We estimate that the 3$\sigma$ uncertainty in the colour is up to $\pm$ 0.6 mag. By employing different filter combinations it is possible to isolate specific populations of objects with minimal confusion. Figures 3 and 5 show example MIRI CMD and CCDs constructed from Spitzer-IRS spectra with the regions occupied by various types of objects illustrated. Whilst Figures 6 and 7 show the results from the evolved star and YSO models.222Due to the large number of JWST bands and filter combinations available, we do not show all possible CMD and CCDs, however the interested reader can use the results from Tables 3 and 4 to generate CMD and CCDs for any potential filter combinations and science goals. This can be accomplished by either by selecting objects of interest e.g. just the AGB stars or the population as a whole and then scaling the fluxes or magnitudes to the required distance. The CMDs and CCDs (Figures 3 and 5) from the spectral sample, provide an accurate representation of the brightest IR population one may expect in a galaxy (including the variations due to various dust compositions). The model CMDs (Figures 6 and 7) enable us to assess low-mass sources and rare sources classes which are not evident in the observed sample, and hence provide completeness. Below we describe the most favourable MIRI filters to identify specific populations of objects with minimal confusion.

3.1.1 Main sequence stars

Stars with a stellar photosphere, but no additional dust or gas features have MIRI colors that fall around zero. The majority of main-sequence and sub-giant stars fall within this class and can only be distinguished on the basis of their short wavelength colours. In MIRI CCDs stars tend to cluster in the upper left side of the diagram (e.g. Figure 5, panel A).

In some instances main sequence OB stars can illuminate and heat patches of dense interstellar medium (ISM) surrounding the star (Sheets et al., 2013; Adams et al., 2013). Stars of this nature have colours typical of stellar photospheres at ($\lambda<8$ $\mu$m) but show a strong infrared excess at longer wavelengths ($\lambda>20$ $\mu$m) indicative of warm dust associated with cirrus hotspots.

3.1.2 Evolved stars

Evolved stars occupy a wide range of MIRI colour-space. Their precise location depends on both the abundance and the chemistry of the dust in their circumstellar envelope. Evolved stars have either oxygen-rich or carbon-rich circumstellar envelopes depending on the C/O ratio in their atmosphere. When C/O $<1$, the CO molecule ties up the carbon, resulting in oxygen-rich molecules and dust grains. Conversely, when C/O $>1$ all the oxygen in locked up in CO resulting in carbon-rich molecules and dust. The [F1130W] band centred on the prominent 11.3 $\mu$m PAH feature and the peak of the Silicon Carbide (SiC) feature, is a powerful diagnostic tool for separating evolved carbon rich sources from oxygen rich sources. The SiC feature at $\sim$ 11.3 $\mu$m weakens with increasing metallicity, thus this filter may not be as an effective discriminant in metal-rich environments. Carbon-rich post-AGB stars and PNe may show a combination of SiC and PAHs; for these objects the [F1130W] band is a powerful diagnostic.

3.1.3 Oxygen-rich AGB Stars

Stars on the early AGB are characterized by a slight IR excess, molecular absorption features and photometric variability. They do not show any prominent dust features and have MIRI colours slightly to the red of main-sequence stars. Around oxygen-rich stars, the fundamental vibrational mode of SiO causes a slight absorption feature in the spectrum at $\sim$8 $\mu$m, which will affect the [F770W] channel.

Simple metal oxides are the first dust species to form in low-density winds, with amorphous silicates becoming increasing prominent as the star evolves. In O-rich stars with weak dust emission, a broad, low-contrast alumina (Al2O3) feature which peaks at $\sim$11 microns is often found to co-exist with amorphous silicates. Therefore it maybe possible to trace changes in O-rich dust evolution using the four MIRI bands that cover the 10 $\mu$m region.

Amorphous silicates are the dominant dust component in oxygen-rich AGB stars, RSGs and post-AGB stars, with broad features at 10 and 20 $\mu$m. These features show considerable variation in shape, width and in the peak wavelength as the star evolves (e.g. Sloan & Price, 1998; Cami et al., 1998; Speck et al., 2000; Jones et al., 2014). This can cause a spread in the MIRI colours. Oxygen-rich AGB stars are best identified using the [F1000W], [F1280W], [F1500W] and [F2100W] filters, e.g. Fig. 3, panels B, E and F.

Fig. 6 shows the evolutionary tracks from the oxygen-rich and carbon-rich grams models. At low dust mass-loss rates, the O-rich grams models form a narrow sequence in the [F1000W] vs. [F1000W]–[F1500W] CMD, this reaches the highest magnitudes at the bluest colours. As the infrared excess increases the colours of the O-AGB stars start to diverge. They become redder and have fainter magnitudes. A similar dispersion in the mid-IR colour is not see in the carbon-stars with the onset of mass-loss; instead C-AGB stars tend to occupy a distinctive wedge of less than 0.5 mag in IR colour space.

The colours and evolutionary tracks of the AGB stars from the grams models shows good agreement with the JWST/MIRI colours derived from IRS observation (c.f. Fig. 3B). The major difference between the two results is the density of O-AGB sources with [F1000W]–[F1500W] $>$ 1; this is a consequence of the sampling in the models. The grams models have not been weighted by an initial mass function, thus this region of colour space is over populated for a given stellar population.

O-AGB stars and C-AGB stars show the cleanest separation from each other in the [F1500W]–[F2100W] colour. In this diagram the grams O-AGB stars have a higher source density at [F1500W]–[F2100W] $\sim$ 1.2 where the evolutionary tracks turn back on themselves; this is due to the source becoming optically thick and hence the silicate emission features changing to absorption features.

In some cases if the O-AGB star is undergoing an intense superwind and the dust shell is optically thick, the 10 $\mu$m silicate feature may be in absorption or self-absorption. These objects are rare, especially in low metallicity environments (Sloan et al., 2008; Jones et al., 2012, 2014) and thus may blend in with other objects e.g. YSOs in colour space. For certain colour combinations, such as the [F1500W]–[F2100W] colour, an O-AGB star may have progressively redder colours until the source becomes optically thick. At this point, the colour now becomes bluer as mass-loss increases, and the star traverse back on itself in colour space. This effect is seen in the left panel of Fig. 6.

3.1.4 Oxygen-rich post-AGB stars

Depending on the binary fraction of the star or the inclination angle of the disk, the SED of post-AGB stars can be either single peaked or double-peaked (e.g.  Ueta & Meixner, 2003; van Aarle et al., 2011). In post-AGB stars with a double-peaked SED; one peak is due to stellar emission from the hot central star and the other due to an outward moving circumstellar dust shell (van Winckel, 2003). As post-AGB stars age, the detached, expanding dust shell cools and becomes fainter in the [F560W] and [F770W] bands, which are sensitive to warm dust (e.g. Min et al., 2013).

O-PAGB stars form a diagonal branch in the [F1000W]–[F1500W] vs. [F1000W] CMD as seen in Fig. 3B), where they occupy the less-dense region between C-rich AGB stars and YSOs. As they transition from the AGB region to the PN region of colour-space, O-rich PAGB stars can have similar mid-IR properties to evolved YSOs. To conclusively identify O-PAGB stars a long colour-baseline is required. The two colours should cover the hot stellar emission (with the NIRCam filters) and the second should be sensitive to the oxygen-rich dust.

Post-AGB stars with a binary companion can have a high crystalline silicate fraction and large dust grains indicative of a circumbinary dusty disc (Gielen et al., 2011). These post-AGB stars have a significant near-IR excess, which is superimposed on top of the stellar emission in the SED, rather than as double peak (de Ruyter et al., 2006). In cases where the SED is singled peaked, O-rich post-AGB stars may be indistinguishable (in the MIRI filters) from their less evolved AGB counterparts. These ‘Disc’ sources would be classified as AGB stars with a large dust excess by Woods et al. (2011a).

3.1.5 Red Supergiants

Red supergiants (RSGs) have very similar dust characteristics to O-AGB stars and are almost indistinguishable in colour space. Due to their higher mass (8–25 $M_{\odot}$) RSGs are in general more luminous than AGB stars, and may fall on a sequence that extends to bright magnitudes than the AGB population, however, care needs to be taken as this region is also inhabited by foreground stars.

Previous colour classification schemes for separating RSG from O-AGB stars rely on near-IR photometry (e.g. Boyer et al., 2011). RSG (with little to no dust) are slightly bluer than the O-AGB stars in the $J-K_{s}$ versus $K_{s}$ and $J-$[3.6] CMDs due to their warmer effective temperatures. As RSG become enshrouded by dust, their $J-K_{s}$ colour grows redder and it becomes impossible to separate dusty AGB from RSG using the current set of photmetric bands.

Figure 3 panel B shows that in the mid-IR RSGs are discernible by their [F1000W]-[F1500W] colours. In this CMD O-rich AGB stars form a left-leaning vertical sequence, which is slightly to the blue of a near-vertical sequence at [F1000W]-[F1500W] $\sim$ 0.2 which extends to bright magnitudes, formed of RSGs. Finally slightly to the red of this RSG track is a right-leaning vertical-sequence composed of C-AGB stars; all three sequences merge at lower luminosities where the photosphere dominate the colours in systems.

3.1.6 Carbon-Rich Stars

In carbon-rich AGB stars, molecular absorption bands due to acetylene (C2H2) and numerous other carbon chain molecules form. Carbon rich molecules produce strong absorption features in the 4–8.5 and 13–14 $\mu$m wavelength intervals (Matsuura et al., 2006), the strength of these bands increases at low-metallicity which may affect the [F560W], [F770W] and [F1280W] MIRI bands in metal poor carbon stars. At solar metallicity, HCN may also be an important opacity source in carbon stars up to Teff $\sim$ 2800 K (Eriksson et al., 1984; Aoki et al., 1999; Harris et al., 2002), affecting the [F560W] and [F1500W] MIRI fluxes.

Amorphous carbon and graphite are the dominant dust species in carbon stars (see Groenewegen et al., 2009; Sloan et al., 2016, and references therein), they produce a dust-dominated continuum in the mid-IR, however they do not have a clear spectroscopic signature. Instead dust features due to Silicon Carbide (SiC) at 11.3 $\mu$m and the broad 26–30 $\mu$m feature, which is often attributed to MgS dust (Goebel & Moseley, 1985; Hony et al., 2002), are the characteristic features of carbon-rich sources.

C-AGB stars are easily separated from all other types of objects using MIRI fluxes. The [F1500W]–[F2100W] colour (Fig. 3F) is especially good for this as it separates carbon- and oxygen-rich evolved stars, and evolved stars from YSOs. Colours in the $\lambda=10-20$ $\mu$m region enable us to separate these evolved stars as both carbon-rich and oxygen-rich AGB stars have prominent dust features in this region due to Silicon Carbide (SiC) and amorphous silicates, respectively.

3.1.7 Carbon-rich post-AGB stars

Carbon-rich post-AGB stars have UV and optically excited PAH features, triangular SiC features at 11.3 $\mu$m, a prominent feature at 30 $\mu$m possibly due to MgS (Sloan et al., 2007; Smolders et al., 2010; Matsuura et al., 2014), although alternate carriers have been suggested (Zhang et al., 2009) and in some instances a ‘unidentified’ 21-$\mu$m emission feature which is unique to this class of stars (Kwok et al., 1989; Hrivnak et al., 2009). The strongest observed PAH features occur at wavelengths of 5.7, 6.2, 7.7, 8.6, and 11.3 $\mu$m.

C-PAGB stars occupy the relatively isolated region of colour-space between AGB stars and YSOs, thus they can be identified in MIRI colour-colour space using a wide variety of colours (see Figure 5, panel C). By using a combination of filters pollution of the colour-selected C-PAGB stars by YSOs and extremely-red carbon stars is limited.

3.1.8 Planetary nebulae

Planetary nebulae are emission line sources that can have a dust continuum that peaks in the mid-IR between $\lambda\sim 20-40$ $\mu$m, they may also have dust features due to either silicate or carbonaceous material (i.e. SiC or MgS). All the PNe in our sample have a thermal IR continuum which rises toward longer wavelengths, however the strength of the dust continuum compared to the emission lines varies considerably (Stanghellini et al., 2007). In addition to forbidden line emission from collisionally excited atomic species, most carbon-rich PNe (C-PN) have PAH emission lines in their spectra, with features at 6.2, 7.7 and 11.2 $\mu$m typical (Bernard-Salas et al., 2009). PAH emission is often used as a discriminant between C-PN and O-PN as not all objects show solid state features.

Oxygen-rich PNe have spectral energy distributions that rapidly rise toward the far-IR. There is some overlap in colour with massive YSOs and background galaxies, however the O-PNe tend to be fainter than the vast majority of YSOs. Both the [F1130W]–[F1800W] and the [F1130W]–[F1500W] colours (see Figures 3D and 3G) provide a clean separation between O-PN, YSOs and other emission line sources. The CMDs for both colours are almost identical, however the [F1130W]–[F1800W] colour has a slightly larger spread due to the longer baseline, and is thus preferred.

Due to their atomic emission lines, PAH features and rising dust continuum, carbon-rich PNe have colours which are more or less indistinguishable from the bulk of the YSO population. It is difficult to separate PNe and YSOs using only one or two colour combinations, however multiple CCD composed of four different MIRI bands may help in their identification (c.f. Fig. 5D).

3.1.9 Young Stellar Objects: YSOs

As YSO’s evolve their envelopes become hotter and less dense. Thus, YSOs of different evolutionary stages span a range in IR colours. We illustrate this by comparing two independent but complementary YSO classification schemes; one from observed spectra (Woods et al., 2011a) and one from models (Robitaille et al., 2006). The spectral data provide the most accurate information for massive (M $>$ 8 M*⊙) YSOs, whilst the Robitaille et al. (2006) models provide the complete range of YSO masses down to 0.2 M⊙*. Both these schemes portray a evolutionary sequence (1–3) from least to most evolved.

In the mid-IR, YSOs are characterized by a superposition of ice, PAH and oxygen-rich dust features on a very red cold dust continuum. As discussed in Section 2.1.1, we have divided the YSO spectra into four groups to reflect changes in their IR properties as they evolve towards the main sequence. The YSO spectral groups 1–3 represent an evolutionary sequence for massive YSOs: from deeply embedded sources in the early stages of formation, to stars surrounded by a dusty envelope which is being progressively dispersed. The majority of these sources would be classed as Stage I or Stage II YSOs on the colour-colour diagram of Robitaille et al. (2006).

YSOs are prominent in all the MIRI bands. They occupy the reddest regions of colour space and are clearly separated from the evolved stars. Embedded protostellar objects (YSO-1) have the reddest colours, due to their dense cool envelopes. In principal as you go to bluer colours the more evolved a YSO will be. However, clear subdivisions between the YSO 1-3 categories is not possible, as YSOs inhabit regions with complex backgrounds. Evolved intermediate-mass YSOs (YSO-4) have a flat or declining spectrum with silicate emission, they occupy a slightly different region of colour space compared to the YSO 1–3 classes. This region of colour space is relatively sparse, typically populated by sources with dusty disks.

The best way to isolate YSOs in colour space is to select two MIRI filters (e.g. [F770W]–[F2500W] or [F1000W]–[F2100W]; Fig. 3, panels A and C and Fig. 5, panel C) with a wide baseline in wavelength that have $\lambda>12$ $\mu$m. This mitigates contamination from AGB stars, which unlike YSOs only have moderate amounts of cold dust (Jones et al., 2015b). Consequently, the YSOs have a rising SED at $\lambda>20$ $\mu$m, whereas AGB stars and post-AGB stars generally have a falling spectra after 20$\mu$m.

Shown in Figure 7 are model sequences for YSOs with a stellar mass of 0.2, 2.0 and 20 $\pm 2.5\%$ $M_{\odot}$ at a disk inclination of 48.5*∘*. These models sample a variety of the SEDs in the Robitaille et al. (2006) model grid for low, intermediate and high mass stars. We identify in the MIRI CMDs and CCDs the models which correspond to each of the three evolutionary stages: Stage I, II and III defined by Robitaille et al. (2006). At earlier stages the disk geometry and the inclination has a significant effect on the SED, since this alters the optical depth along the line of sight. Although, globally the inclination angle is not as important as it averages out over the stellar population. Stage I sources have the reddest colours and correspond to the youngest embedded YSOs; here the envelope dominates the mid-IR flux. As a source of a given mass evolves the envelope and disk disperse; the YSO becomes brighter with bluer colours due to the lower-extinction. An evolutionary sequence is seen in the colour distribution with Stage III, II, and I sources moving from lower left to upper right. Figure 7 shows that the evolutionary stages of YSOs are best separated with a long baseline with one colour $\gtrsim 18$ $\mu$m. This long baseline also provides the greatest separation between YSOs and non-YSOs.

It can be clearly seen in Figure 7 that the YSOs models have a large spread in luminosity, this corresponds to the initial mass of the YSO. In principal if you have a nearby field-of-view with little contamination you could cleanly separate the stages, and potentially detect YSOs with sub-solar masses. This would require substantial integrations times, even for the Magellanic Clouds.

3.1.10 H ii regions

Compact H ii regions form around young hot stars and are typically embedded in giant molecular clouds. They have a steeply rising infrared continuum and can be differentiated from PNe using the long-wavelength shape of the SED. Furthermore, H ii regions are typically brighter than PNe and occupy the upper-right region of the MIRI CMDs. H ii regions are among the reddest sources in the [F1000W]–[F1500W] CMD (Fig. 3B) due to their sharply rising continuum.

The [F1280W]–[F1800W] colour (Figs. 3 and 5, panels E and B, respectively) is one of the best filter selections for identifying H ii regions. The limited overlap with a small number of YSOs in this colour space is probably due to the limited angular resolutions of Spitzer’s IRS which caused ambiguity in the spectral classification between (ultra)compact H ii regions and evolved YSOs; consequently (ultra)compact H ii regions were classified as YSOs by Woods et al. (2011a).

3.1.11 Background Galaxies

Background Galaxies are major contaminants to studies of IR stellar populations in the Local Group, particularly at low flux levels. In the JWST era this problem will be further exacerbated as we push resolved stellar population studied to greater distances. One way of excluding galaxies to examine each individual source in the image to see if it is marginally resolved, this can be done by in a quantitative manner by fitting an elliptical shape to the source.

The constituents and morphologies of the individual galaxies in the background sample are described by Brown et al. (2014). These galaxies cover a wide ranges in classes and have spectra which are composed of the composite stellar population for that galaxy. Quiescent galaxies tend to have blue colours and occupy a similar region to dust-free stars in MIRI CMDs. Conversely, galaxies undergoing active star-formation populate a large range of MIRI colour space; due to their strong 6.2 and 7.7 $\mu$m PAH features and their steeply rising SEDs. As such star-forming galaxies have overlapping colours with YSOs.

The MIRI CMDs and CCDs highlights the difficulty in distinguishing between YSOs and unresolved background sources in the same color-magnitude space; nonetheless, a careful choice of filter combinations can be used to separate galaxies from other stellar populations. Many classes of extragalatic objects will be quite red in the MIRI bands, and we find that the [F770W]–[F1000W] versus [F1500W]–[F2550W] CCD (Fig. 5, panel D) is good discriminant between galactic and non-galactic sources. This CCD also highlights the diversity in galaxy colours.

4. Applicability to Local Volume galaxies

JWST/MIRI’s unparalleled sensitivity and spatial resolution at 5–28 $\mu$m will enable resolved stellar population studies out to $\sim$4 Mpc; over 100 galaxies fall within this volume. Going beyond the Local Group with detailed studies of resolved stellar populations will enable us to measure star formation histories, probe active star formation regions, ascertain the age of the oldest stellar populations and determine the chemical and dust enrichment of galaxies with properties very different to our own.

Local volume galaxies span a wide range in metallicity ($-2.72$  $<$  [Fe/H] $<$ 0.5). This large metallicity baseline provides a foundation for understanding which types of objects produce dust, and the significance of their dust production as galaxies evolve. Insights into the connection between resolved stellar populations and galaxy evolution can be used to study the early universe. For instance, it is unclear if intermediate- and high-mass stars can account for the substantial dust abundances observed at $z\gtrsim 6.4$ (Beelen et al., 2006; Valiante et al., 2009; Gall et al., 2011).

JWST will be able to obtain moderately-deep multi-band stellar photometry for the M81 group at 3.6 Mpc and the Sculptor filament at 3.9 Mpc. In Table 5 we list the 10$\sigma$ magnitude limits for long-duration observations of faint sources with the MIRI filters. In Figure 8 we have scaled the MIRI photometry of the LMC to that of a galaxy at 3.6 Mpc and over-plotted the MIRI sensitivity limits for a 10$\sigma$ detection in 10,000s on the [F1000W] vs. [F1000W]–[F1280W] CMD. At this distance only the brighter red objects, such as RSGs, AGB stars, PNe, massive YSOs (M $\gtrsim$ 10 M*⊙*) and H ii regions can be detected. The depth/detection limit of JWST observations of galaxies at 3 Mpc would be comparable to those of the Wide-Field Infrared Survey Explorer (WISE) for the LMC (e.g. Wright et al., 2010; Nikutta et al., 2014). From these observations we would be able to study the formation of new stars, characterize the largest dust producers and probe substructures within a galaxy.

IR surveys similar to the Spitzer SAGE (Meixner et al., 2006; Gordon et al., 2011) legacy programs which conducted census of all objects in the Magellanic Clouds brighter than 15 mag at 8 $\mu$m are achievable (in 10,000s per filter, per pointing) with JWST out to $\sim$450 kpc. The decline in source density towards fainter magnitudes in Figure 8 is a limitation of the Spitzer spectroscopic sample, rather than an observation effect which would be due to JWST/MIRI. Colours of fainter populations e.g. low-mass YSOs and evolved stars can be inferred from scaling the models presented in Section 3 to the required distance and comparing these to the sensitivity limits derived in Table 5.

MIRI observations would allow detailed studies of individual YSOs and H ii regions. YSOs can be identified with only MIRI two colours; however multi-band photometry is required to constrain the physical processes in the protostar, circumstellar disk, and collapsing outer envelope via radiative transfer modelling of its SED. By observing red-luminous populations we can also determine star formation rates for a much larger volume ($\gtrsim$100) of galaxies than is currently possible with Spitzer (e.g Whitney et al., 2008; Sewiło et al., 2013). These studies extrapolate the mass function derived from star counts and SED fitting with a standard IMF, to derive an empirical limit for the current star formation rate of a galaxy. Furthermore, MIRI observation of YSOs in conjunction with H i, H$\alpha$, and CO gas tracers will reveal the the initial conditions and process of star-formation across a range of galactic environments.

The IR emission from evolved stellar populations of local volume galaxies measured by JWST can be compared to the ISM dust masses obtained from global flux measurements (Draine et al., 2007) by the Spitzer Infrared Nearby Galaxies Survey (SINGS; Kennicutt et al., 2003). Determining the nature of interstellar dust as galaxies evolve in metallicity is essential, as its origin and rate of destruction by supernova shocks (e.g Temim et al., 2015) is uncertain. Dust production rates of evolved stars can be estimated from mid-IR colours (Matsuura et al., 2009; Jones et al., 2015a) or via SED fitting with model grids (Riebel et al., 2012; Srinivasan et al., 2016). This can constrain the luminosities, dust chemistry and the current dust-production rate of each star in the sample. The rate of dust injection by the evolved stellar population can be compared to the current ISM dust mass to determine the time-scale at which stellar sources replenish ISM dust. This will directly constrain dust models (e.g. Dwek & Cherchneff, 2011) and provide the foundations for deciphering JWST observations of more distant galaxies in the high-redshift universe.

To ensure representation, samples selected based on photometric data needs to be confirmed with spectroscopic observations. With JWST we will be able to conduct detailed spectroscopic studies of point sources within the Local Group. Spectroscopic surveys will provide critical information about the dust compositions, spectral types and ages of each star. By linking the MIRI colours of LMC objects to their dust characteristics and infrared spectral type (see Section 3.1) we can effectively distinguish between sources of similar colour and efficiently select candidates for JWST spectroscopic observations through the careful use of the MIRI CMDs. This will ensure that there is a high success rate of JWST spectroscopically observing targets with the desired object class and chemistry. Furthermore, spectroscopic observations with JWST can evaluate and refine our proposed photometric classification scheme, to ensure clean subdivisions between stellar classes.

5. Summary

JWST is expected to revolutionize our understanding of composite stellar populations in the Local Volume. In this paper we have used over 1250 Spitzer-IRS spectra from the SAGE-Spec legacy program and model spectra from the grams and YSO grids to calculate synthetic MIRI fluxes and magnitudes for IR-bright stellar populations and background galaxies. These results can be used to select JWST filters which will provide good photometric sampling of SEDs, used to measure the stellar and circumstellar properties of the stars.

We have developed colour-colour and colour-magnitude classification schemes for JWST/MIRI to identify and select samples of similar objects for future studies. These results can easily be adapted for use in other galaxies, e.g. the M81 group, to probe composite stellar populations across a range of galaxy types, metallicities and star formation environments, beyond the boundaries of the Local Group. In Section 3.1 we assessed the best filters to target specific stellar populations throughout a galaxy, and discussed how to remove foreground and background contamination based on their JWST broadband colours. We highlight the F1000W and F2100W MIRI filters as they provide a clean separation between evolved stars, young stellar objects and background galaxies.

Jones and Meixner acknowledge support from NASA grant, NNX14AN06G, for this work. Justtanont is partly supported by the Swedish National Space Board.

Appendix A Flux calibration of the Spitzer spectra

Prior to joining the different modules of the Spitzer IRS spectra together, flux uncertainties in the Spitzer sample are carefully propagated using formal uncertainties in the mean ($\sigma/\sqrt{N}$). To photometrically calibrate the spectra, the mean flux density is converted to Jy using IRS observations of the standard stars HR 6348 (K0 III) for the short-low data, and HR 6348 and HD 173511 (K5 III) for the long-low (Sloan et al., 2015). Spectra observed with the short-high and long-high modules were extracted using a full-slit extraction and calibrated using $\xi$ Dra (K2 III) as a standard.

The flux calibration was applied to each individual spectral nod and order. Here the spectroscopic (i.e. the point-to-point) uncertainty in the flux is better than 0.5% at most wavelengths (Sloan et al., 2015). When combining the spectra from the nod positions, the larger of the propagated uncertainties from the two nods or the uncertainty in mean was adopted. This combination introduces several systematic errors in the flux (Lebouteiller et al., 2011), which depends on source geometry (G. Sloan 2016, private communication). The error in the absolute flux calibration is generally around $\sim$5%, however this can be considerably more for extended sources or bad pointings.

To produce the final spectrum, a scalar multiplicative correction was applied to each spectral segment to remove discontinuities which arise from pointing errors. Segments are normalised upwards (using the wavelengths where they overlapped), to align with the best-centered segment. This correction is typically less than 10 % and has no dependence on wavelength. Data at the ends of the segment that could not be calibrated reliably were trimmed from the spectra. As with the low-resolution modules, short-high data was stitched to the long-high data, but no orders were adjusted relative to other orders within the same module, as they were obtained at the same time with identical telescope pointings. For more details on this procedure see Woods et al. (2011b); Lebouteiller et al. (2011); Sloan et al. (2015).