The JWST Resolved Stellar Populations Early Release Science Program. IV. The Star Formation History of the Local Group Galaxy WLM

High-temporal resolution SFHs from CMD-fitting techniques were originally made possible because of the Hubble Space Telescope's (HST's) sensitivity and image quality, and the ability to perform precision point-spread function (PSF) photometry of resolved stars down to faint magnitudes (e.g., Dolphin 2000, 2002; Tolstoy et al. 2009). These SFHs have been instrumental in shaping our understanding of how galaxies build their stellar mass as a function of time. The scientific insights gained with the SFHs from HST data are quite diverse and, in large part, are determined by the photometric depth of a given data set.

For example, SFHs have been derived for many galaxies within the Local Group (i.e., very nearby systems) from photometric catalogs, with the majority of those catalogs reaching depths below the oMSTO. The oMSTO feature helps to break the age–metallicity degeneracy for older-age stars in a CMD, thereby facilitating robust SFH recovery with temporal resolution as fine as log(δt) = 0.05 at the oldest lookback times. These SFHs characterize the properties of satellite galaxies around the Milky Way and M31, including investigating quenching timescales and possible host-galaxy dependencies (e.g., Monelli et al. 2010; Weisz et al. 2014; Skillman et al. 2017; Weisz et al. 2019; Sacchi et al. 2021; McQuinn et al. 2023a, 2023b; Savino et al. 2023). SFHs of the Large and Small Magellanic Clouds from similarly deep HST data and ground-based data have revealed synchronized bursts of star formation that are likely interaction-driven (Harris & Zaritsky 2009; Weisz et al. 2013; Massana et al. 2022). SFHs are currently being derived from >100 HST pointings across the Clouds in an effort to quantify the radial variations in their histories (R. Cohen et al. 2023, in preparation). SFHs have also constrained the early evolution of a handful of gas-rich galaxies located within the Local Group at farther distances from the MW and M31 than the satellite systems, revealing intriguing differences on when these systems began to form their stellar mass in earnest (e.g., Cole et al. 2007; Hidalgo et al. 2011; Cole et al. 2014; Skillman et al. 2014; Albers et al. 2019).

The first in-depth look at the time-resolved history of a spiral galaxy has come from measurements of the recent and lifetime SFHs of M31, and have included tracking changes in the different structural features of the galaxy and provide a measure of the chemical evolution history of the system (Richardson et al. 2008; Lewis et al. 2015; Williams et al. 2017; Telford et al. 2019). Similar efforts are under way for the lower-mass spiral M33 (Williams et al. 2021).

Finally, SFHs have been reconstructed for numerous galaxies outside the Local Group (i.e., at slightly farther distances). The photometric depth achieved with HST imaging is more limited, reaching below the horizontal branch (HB), ∼4–5 mag below the tip of the red giant branch (TRGB), for a few galaxies (e.g., Leo P,UGC4483; McQuinn et al. 2015a; Sacchi et al. 2021), but more typically 1–3 mag below the TRGB. Given the depth of such data, SFHs can still be characterized but with significantly less temporal resolution at older lookback times (i.e., the oldest-age time bin can be as large as several Gyr). Yet, since these field galaxies are primarily gas-rich star-forming systems, the SFHs derived from the shallower imaging are useful because they robustly constrain recent (∼1–2 Gyr) star formation activity with fine temporal resolution (log(δt) = 0.10; e.g., Aloisi et al. 1999; McQuinn et al. 2010, 2015a, 2015b; Cignoni et al. 2018; Annibali & Tosi 2022; Hunter et al. 2022).

To date, almost all SFH determinations from CMD-fitting have been done using data collected with optical filters, with the notable exceptions of the SFH maps performed on near-infrared data from the ESO/VISTA survey of the Magellanic Clouds (Rubele et al. 2018; Mazzi et al. 2021). These studies demonstrate that the CMD-fitting method of SFH recovery works as well for near-infrared filters as for optical filters. A comparison between the NIR-derived and the previous optical-derived SFH maps by Harris & Zaritsky (2009) showed good quantitative agreement at all ages larger than ∼10⁸ yr. In some small areas of the Magellanic Clouds, near-infrared photometry has already shown clear advantages over optical photometry by providing sharper CMD features in regions severely affected by differential extinction and by better sampling the evolved stars (especially the age-sensitive sub-giant branch) in regions with a high total extinction. The potential advantages of near-infrared filters have been little explored outside the Magellanic Clouds, mainly because of the intrinsic difficulty to reach deep magnitudes using ground-based near-infrared cameras. This difficulty can now be superseded with the JWST, which provides a pathway for SFH work from deep near-infrared CMDs in targeted fields (which we explore in detail). Looking to the future, the large-area near-infrared observations from the Nancy Grace Roman Space Telescope will offer the opportunity to recover SFHs of galaxies residing in many different environments and to study stellar population structures that can span large physical scales, such as stellar streams and galaxy halos.

JWST's sensitivity, high resolution, and infrared capabilities enable the more efficient and robust reconstruction of SFHs of nearby galaxies, out to larger distances, and in different galactic environments than is possible with HST. JWST's larger aperture (6.5 m versus HST's 2.4 m mirror) enables imaging of resolved stars to photometric depths below the oMSTO in galaxies outside the Local Group. This capability opens up the exploration of the early history of galaxies in varied environments, including isolated environments that provide unique laboratories for probing secular evolution and testing galaxy formation models. JWST's higher resolution and wavelength coverage make morphological and color-based indicators more effective at removing background galaxies, thereby improving the purity of photometric catalogs and SFH fits (Warfield et al. 2023). Cooler stars are intrinsically brighter in the near-infrared, which increases observing efficiency when targeting evolved stars such as red giant branch (RGB) and asymptotic giant branch (AGB) stars, and (importantly for ancient SFHs) oMSTO stars. For example, a metal-poor oMSTO star (M_* = 0.76 M_⊙; 13 Gyr; [M/H] = − 1.4) has absolute magnitudes of 4.4, 3.6, 3.5, and 3.0 in the F475W, F814W, F090W, and F150W filters, respectively (based on the PARSEC stellar library). This is a gain in brightness of ∼1.4 mag moving from the F475W to the F150W filter. Finally, JWST offers potential gains in accurately and precisely recovering SFHs in metal-rich galaxies with significant dust reservoirs that have correspondingly higher extinction.

The JWST Early Release Science Resolved Stellar Populations program (ERS-1334) was designed to provide data reduction and analysis tools for resolved star science with JWST and to test JWST's ability to recover lifetime SFHs from resolved stars, paving the way for future studies. The details of the overall ERS program are presented in Weisz et al. (2023), with individual science results to be reported separately in upcoming papers from the ERS team. Here, we present the first SFH results from JWST ERS data using the well-established CMD-fitting technique as implemented in match (e.g., Dolphin 2002; McQuinn et al. 2010; Weisz et al. 2011; Skillman et al. 2017; Savino et al. 2023).

The aims of this SFH analysis are fourfold. First, we will quantify the efficacy of recovering SFHs with the F090W–F150W filter combination. These filters were chosen over the F070W and F115W filters to balance the competing needs of a wide color baseline (which improves metallicity sensitivity in the CMD), signal-to-noise requirements, and the sampling of the PSF. Second, we will test the sensitivity and accuracy of stellar evolution libraries for recovering SFHs and age-metallicity relations (AMRs; i.e., color–T_eff transformations and photometric calibrations) with the JWST near-infrared data. Third, we will compare the SFH results derived from the individual HST and JWST CMDs with a SFH derived by simultaneously fitting the JWST and HST CMDs. Finally, we will explore the stellar mass assembly of Wolf–Lundmark–Melotte (WLM) and compare with other gas-rich low-mass galaxies for which the requisite deep imaging data have been obtained.

The target selected for the JWST Near-Infrared Camera (NIRCam–based SFH is the well-studied WLM galaxy, which is a gas-rich and low-mass (M_* = 4.3 × 10⁷ M_⊙; McConnachie 2012) system located near the outskirts of the Local Group (at a distance of $\,=\,{968}_{-40}^{+41}$ kpc; Albers et al. 2019). An overview of WLM's properties is provided in Table 1.

Table 1. Properties and Observational Details of WLM

WLM Properties
MV (mag)	−14.2 (1)
rh (arcmin)	7.8 (1)
E(B − V) (mag)	0.03 (2)
Distance (kpc)	${968}_{-40}^{+41}$ (3)
μ (mag)	24.93 ± 0.09 (3)
Supergiant stars 〈[Z]〉	−0.87 ± 0.06 (4)
Gas-phase 12+log(O/H)	7.83±0.06 (5)
RGB stars 〈[Fe/H]〉	−1.14; σ = 0.33 (6)
JWST NIRCam Observations
PID	ERS-1334
R.A. (J2000)	00:01:57.429
decl. (J200)	−15:28:52.49
F090W (s)	30492
F150W (s)	23707
F090W 50% Completeness (mag)	29.23 (+4.30)
F150W 50% Completeness (mag)	27.89 (+2.96)
log(M*/M⊙), Full FoV, PARSEC	7.12 ± 0.01
log(M*/M⊙), Full FoV, BaSTI	7.13 ± 0.01
log(M*/M⊙), ACS FoV, PARSEC	6.54 ± 0.01
log(M*/M⊙), ACS FoV, BaSTI	6.58 ± 0.01
log(M*/M⊙), Joint, PARSEC	6.64 ± 0.01
log(M*/M⊙), Joint, BaSTI	6.65 ± 0.01
τ10 PARSEC; BaSTI (Gyr)	${12.61}_{-0.06}^{+0.06};$ ${13.12}_{-0.03}^{+0.02}$
τ25 PARSEC; BaSTI (Gyr)	${7.63}_{-0.03}^{+0.05};$ ${7.87}_{-0.03}^{+0.06}$
τ50 PARSEC; BaSTI (Gyr)	${5.08}_{-0.01}^{+0.05};$ ${5.17}_{-0.02}^{+0.02}$
τ90 PARSEC; BaSTI (Gyr)	${1.18}_{-0.01}^{+0.01};$ ${1.11}_{-0.01}^{+0.01}$
HST ACS Observations
PID	HST-GO-13768
R.A. (J2000)	00:01:57.319
decl. (J200)	−15:31:22.52
F475W (s)	27360
F814W (s)	34050
F475W 50% Completeness (mag)	28.28 (+3.35)
F814W 50% Completeness (mag)	27.20 (+2.27)
log(M*/M⊙), NIRCam FoV, PARSEC	6.60 ± 0.01
log(M*/M⊙), NIRCam FoV, BaSTI	6.58 ± 0.01

Note. The 50% completeness limits listed for the NIRCam filters are for the full field of view, whereas for the ACS filters they are based on the data that overlaps with the NIRCam imaging. See text for details. The listed parenthetical 50% completeness limits are in absolute magnitudes. The stellar masses reported under the JWST observations are the total mass formed in the full NIRCam field of view and overlapping footprint with ACS, as noted. Similarly, under the HST observations, it is the stellar mass formed in the overlapping footprint with NIRCam. We also report the stellar mass formed based on simultaneously fitting the NIRCam and ACS CMDs, which we list under "joint." τ₂₅, τ₅₀, and τ₉₀ refer to the lookback times when 25%, 50%, and 90% of the stellar mass was formed based on the SFH reconstructed from the full NIRCam data using the PARSEC and BaSTI stellar libraries. References: (1) McConnachie (2012); (2) Schlafly & Finkbeiner (2011); (3) Albers et al. (2019); (4) Urbaneja et al. (2008); (5) Lee et al. (2005); (6) Leaman et al. (2009).

Download table as:  ASCIITypeset image

WLM was primarily chosen because of its proximity, isolation, and the rich suite of existing data sets. Specifically, deep HST Advanced Camera for Surveys (ACS; Ford et al. 1998) imaging of the stellar disk is available that overlaps with the NIRCam field of view (see Section 2). Although the NIRCam data are deeper than the ACS data (50% completeness limits of M_F090W = +4.30 mag versus M_F475W = +3.35), the photometric depth of the ACS data still reaches below the oMSTO. Thus, the SFH derived from the ACS data facilitates a detailed comparison with the SFH derived here from the NIRCam data.

From the previously published ACS-based SFH, it appears that the central part of WLM formed ∼50% of its stellar mass within the last ∼5 Gyr, and only ∼20% of its stellar mass at early times (Albers et al. 2019). This implies that WLM experienced a period of quiescence post-reionization (z ∼ 6), which has only been noted in the two other gas-rich and isolated low-mass (M_* <10⁸ M_⊙) galaxies in the Local Group with the requisite data quality: Leo A and Aquarius (Cole et al. 2007, 2014). Such a gap in the SFH of low-mass galaxies is somewhat surprising but, given the small sample size, it is unclear whether a pause in star formation is common or atypical. Regardless, the NIRCam data can provide independent confirmation of the HST-based SFH results. Spectroscopic observations of individual stars are also available, which provide robust measurements of the stellar metallicity distribution in WLM (Leaman et al. 2009), allowing us to test the AMR derived as part of the CMD-fitting approach.

This paper is organized as follows. We briefly summarize the observations and data processing of the JWST and the HST data in Section 2. The SFHs methodology and results, including analysis of the SFHs derived from stars in the overlapping footprint between NIRCam and ACS and a comparison of the AMR with an independently measured metallicity distribution function (MDF), are presented in Sections 3 and 4. We summarize our findings in Section 5.

As shown in Figure 1, JWST observed a central region of the stellar disk in WLM with NIRCam (Rieke et al. 2023) as the primary instrument (solid lines) and an outer field with Near-Infrared Imager and Slitless Spectrograph (NIRISS; Doyon et al. 2023) in parallel (dotted lines). For NIRCam, we selected orient constraints such that one NIRCam module overlapped a region with the deepest optical HST imaging available (dashed lines, HST-GO-13678; Albers et al. 2019), while the other module overlapped a region with HST UV, optical, and IR imaging from various programs (not shown; PIDs: HST-GO-6798, HST-GO-11079, HST-GO-15275, and HST-GO-16162), and ALMA observations where CO clouds were identified (Rubio et al. 2015). This carefully chosen areal coverage with such comprehensive and multi-wavelength ancillary data enables multiple science goals, including a comparison of the SFHs derived from JWST and HST presented in this work, a study of spatial variations in the SFHs, an investigation of the evolved stars, and an exploration of the stellar populations co-spatial with the CO clouds. Dictated by the orient constraints used for the NIRCam placement, the NIRISS field covered an outer region of the galaxy to the southeast. Here, we focus on the NIRCam data. Analysis of the less-populated NIRISS imaging will be presented in a future paper.

Zoom In Zoom Out Reset image size

Table 1 lists the coordinates of our observed fields, and the total integration times per filter for NIRCam from ERS-1334. These data were obtained from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute. The specific observations that we analyzed can be accessed via doi:10.17909/t6pm-x063. We refer the reader to Weisz et al. (2023) for a detailed description of the observing strategy. Note that while the full NIRCam data includes imaging from the long-wavelength (LW) F250W and F430M filters, we only use the short-wavelength (SW) F090W and F150W filters to derive the SFH as the data are deeper and this filter pair is more sensitive to changes in stellar ages.

Figure 2 presents color images for the two NIRCam modules created using the F090W filter (blue), F150W filter (green), and the sum of the F250M and F430M filters (red). The image demonstrates the exquisite resolution and depth of the NIRCam imaging, which will be exceptional for resolved star science in nearby galaxies. Careful inspection of the fields reveals changes in the stellar populations, from small variations in the stellar density, to stellar sources still somewhat embedded in their nascent clouds, to extremely red stars that are dusty, and to AGB candidates.

Zoom In Zoom Out Reset image size

We performed PSF photometry on the NIRCam images using the photometry software package DOLPHOT ²¹ (Dolphin 2000, 2016) with the new beta-version of a JWST NIRCam-specific module (Weisz et al. 2024). As inputs, we used the F150W stage 3 drizzled i2d.fits file from the JWST data pipeline as the astrometric reference image for identifying point sources, and the cal.fits files processed with the JWST calibration pipeline version 1.9.6 and the pmap 1075 reference files to measure the flux of each identified source. The cal.fits files and the i2d.fits file were internally aligned by the JWST pipeline.

Our photometric reduction with DOLPHOT used model PSFs generated with the WebbPSF simulation tool (version 1.1.0). The WebbPSF tool adopts in-flight optical performance data and incorporates time-dependent optical path delay (OPD) maps based on the telescope configuration. We used the OPD map of the NIRCam instrument around the time our observations were taken. Despite being well matched to the telescope conditions, the WebbPSF models appear to be slightly more sharply peaked than the actual PSF of the images (i.e., the modeled PSFs have more flux in the central regions) at the ∼2% percent level, which currently limit the relative precision of our photometry (Weisz et al. 2024).

Before running photometry, we masked out bad pixels, identified saturated pixels, and applied pixel area maps using the routine nircammask included as part of the DOLPHOT distribution. The photometric parameters used to run DOLPHOT were chosen after extensive testing of the photometric recovery from NIRCam (Weisz et al. 2024). The parameters are very similar to those adopted for HST WFC3/IR imaging in the PHAT program (Williams et al. 2014). These DOLPHOT parameters were optimized for the NIRCam imaging based on the current PSF models and may need adjusting if the PSF model changes.

The photometry was filtered for well-recovered point sources. We required a signal-to-noise ratio (S/N) ≥4 in each filter, sharp ² ≤ 0.0225 in each filter, and crowd ≤ 0.5 in each filter. The sharpness parameter is a measure of how peaked or broad a source is relative to the PSF and helps to reject image artifacts and background galaxies, respectively. The crowding parameter measures how much brighter a source would have been if stars nearby on the sky had not been fit simultaneously.

We also ran a large number (2.3M) of artificial star tests (ASTs) per NIRCam module to measure the bias and completeness of the data and the photometric errors. While fewer ASTs (of order a few 100k) would have been sufficient for our SFH-recovery purposes, we ran such a large number of ASTs for data exploration and testing purposes for this first SFH fit with NIRCam data. Artificial stars were injected throughout the images with a flat spatial distribution (similar to the distribution of the stellar populations in each NIRCam module) and then recovered with DOLPHOT using the same parameters set for the photometry. We then applied the same quality cuts used for filtering the photometry and marked any star that did not meet our quality requirements as unrecovered. Based on the ASTs, the 50% completeness limits averaged over the full NIRCam stellar catalogs are m_F090W = 29.23 (M_F090W = + 4.30) mag and m_F150W = 27.89 (M_F150W = + 2.96) mag; these values are also listed in Table 1.

Figure 3 presents the NIRCam CMD with close to 600,000 well-recovered point sources; mean 1σ uncertainties per magnitude bin are shown on the right-hand side. Stellar populations with a wide range of different ages and masses are well-represented in the WLM field, creating a comprehensive CMD with various branches of stellar evolution that can be compared with existing data and with different stellar models. Specifically, the CMD includes younger stars in the upper main sequence and blue/red helium burning sequences, intermediate and older-age stars on the red giant and asymptotic giant branches and in the red clump, and older stars in the horizontal branch, sub-giant branch, and lower main sequence.

Zoom In Zoom Out Reset image size

The JWST NIRCam pointing intentionally targeted a field in WLM that overlapped with existing deep HST ACS optical imaging doi:10.17909/gs15-rb46. The region in common between the two data sets, seen in Figure 1, allows for a well-matched and detailed comparison of the resolved star photometry recovered from each telescope and the SFHs reconstructed separately from infrared versus optical data. In addition, the overlap in areal coverage also allowed us to perform a joint photometric reduction of the JWST and HST data together, and recover the SFH by simultaneously fitting the infrared and optical CMDs. Here, we briefly describe the reduction of the ACS data and the creation of stellar catalogs from the NIRCam–ACS overlapping field of view; we describe the simultaneous photometric reduction in Section 2.4.

The HST ACS observations were taken as part of HST-GO-13768, described in detail in Albers et al. (2019), and include imaging in the F475W and F814W filters. The data were processed with the same tools and methodology used on the JWST data. Specifically, the HST-only photometry was performed on the HST pipeline-processed flc.fits files using the ACS module of DOLPHOT with a F814W drizzled image. The same quality cuts used for the NIRCam data were applied to create the final ACS stellar catalog. These cuts are similar to those used on the ACS data in Albers et al. (2019), except that we use a slightly more stringent cut on sharpness (which increases the purity of the catalog at the expense of completeness). Approximately 500,000 ASTs were run on the ACS data in the region overlapping with the NIRCam data.

From the full stellar catalogs of the NIRCam Module B (the only module that overlapped with ACS) and the ACS data, we selected sources that lie within the common footprint, excluding sources that fall in any of the chip gaps. Figure 4 presents the CMDs of these stars, with representative uncertainties per magnitude bin. The photometry from both NIRCam and ACS reach below the oMSTO, although, as expected given the differences in the telescopes and observing strategies, the NIRCam data has a higher S/N at faint magnitudes. Quantitatively, the 50% completeness limits for the NIRCam data are m_F090W = 29.18 (M_F090W = +4.25) mag and m_F150W = 28.29 (M_F150W = +3.36) mag, compared with m_F475W = 28.28 (M_F475W = +3.35) mag and m_F814W = 27.20 (M_F814W = + 2.27) mag for the ACS data; the ACS values are also listed in Table 1. The completeness limit for the F150W filter in this region of WLM is somewhat fainter than that in the full NIRCam field of view, which reflects the overall lower object crowding in this region (Weisz et al. 2024). This difference in crowding has a greater impact on the F150W photometry due to the wavelength coverage and the larger PSF in this bandpass relative to the F090W.

Zoom In Zoom Out Reset image size

The total number of high-confidence stars recovered photometrically in the NIRCam data is ∼3 times higher than the number in the ACS data in the matched field of view. The higher star counts reflect the higher completeness of the NIRCam data at faint magnitudes where star counts naturally increase for any typical initial mass function (IMF). It also reflects that the lower main sequence stars are brighter in the near-infrared than in the blue F475W filter. For example, an F150W absolute magnitude of +4.3 corresponds to an oMSTO star with a mass of 0.65 M_⊙ (based on the PARSEC library), whereas in the F475W filter this same absolute magnitude corresponds to a star with 0.72 M_⊙. At brighter magnitudes, where neither data set is significantly impacted by incompleteness (e.g., F090W < 26 mag), the number of recovered stars from the two instruments are approximately equal. We compare the SFHs derived separately from these catalogs in Section 4.3.

Besides photometric depth, completeness, and star counts, there are other differences in the CMDs. Most notable are the slightly different CMD architectures due to differences in the narrower CMD features in the near-infrared. While the width of the color baselines in the F090W−F150W and F475W−F814W filter combinations are similar, the features in the NIRCam CMD become narrower as the near-infrared wavelengths approach the Rayleigh–Jeans regime of a star's spectral energy distribution for bluer sources.

We also performed simultaneous, cross-observatory PSF photometry using DOLPHOT on the NIRCam and ACS data. We used the same methodology and parameters described above, with the F090W i2d.fits file as the reference frame, the cal.fits files as the science images for NIRCam, the flc.fits files for ACS, and applied the same quality cuts to the output catalogs. For the ASTs, we generated an input list of ∼1M sources with realistic spectral energy distributions across the wavelength range covered by the F475W—F150W filters. We applied the same treatment to the ASTs as before. The stellar catalogs are similar to those generated from the individual photometry runs, but with slightly fewer stars passing the quality cuts. We recovered 177,000 stars in the NIRCam filters, compared with 202,000 in the NIRCam-only catalog, and 58,000 in the ACS filters, compared with 62,000 in the ACS-only catalog. The 50% completeness limits for the multi-observatory photometry are m_F090W = 28.77 mag, m_F150W = 28.09 mag, m_F475W = 27.74 mag, and m_F814W = 26.86 mag. These are somewhat brighter than the limits from photometry performed on each data set separately, similar to what has been noted when processing, for example, HST ACS, and WFC3/IR data simultaneously (Williams et al. 2014).

Figure 5 presents the NIRCam IR–ACS optical color–color plane with point sources that are well-recovered in all four filters from the simultaneous photometric reduction. The tightness of the stellar sequences in the color–color emphasizes the high-fidelity of the photometry. Evident in the plot are a number of features including: (i) the red clump over-density at F090W-F150W ∼ 0.8, F475W-F814W ∼ 1.3; (ii) the tight RGB sequence at F090W-F150W 0.8, F475W-F814W 1.3; (iii) the richly populated main sequence at F090W-F150W ≲0.4, F475W-F814W ≲ 0.6; and (iv) the photometric spread at fainter magnitudes at F090W-F150W ∼0.3–0.6, F475W-F814W ∼0.4–1.5. We do not see a prominent galaxy sequence in the color–color plot, which demonstrates the power of the high resolution of NIRCam and ACS imaging combined with accurate PSF photometry with DOLPHOT and carefully selected quality cuts. The shaded blue at redder optical and IR colors represents the region where, if present, the majority of foreground MW stars would lie. Given the higher Galactic latitude of WLM (b = −73°), we do not expect a significant population of foreground stars, which is confirmed by the low source counts in this region that are not coincident with the RGB sequence or the few redder sources that are likely AGB stars. Note that bluer foreground dwarfs are predominantly brighter than the magnitude ranges in our ACS and NIRCam CMDs, and are therefore not a concern.

Zoom In Zoom Out Reset image size

The SFHs and AMRs were derived using the numerical CMD-fitting software match (Dolphin 2002). match employs the well-established technique of comparing an observed CMD to a grid of synthetic simple stellar populations (SSPs) of different ages and metallicities (e.g., Tolstoy et al. 2009). The SSPs generated assume an IMF, a binary fraction, and a set of stellar evolution isochrones, and take into account photometric completeness determined from the ASTs. The modeled CMDs are linearly combined until the best-fit to the observed CMD is found based on a Poisson likelihood statistic.

The SFHs were derived using two stellar evolution libraries with solar-scaled abundance patterns updated with NIRCam in-flight filter transmission curves and Vega zero-points, namely PARSEC (Bressan et al. 2012) and BaSTI (Hidalgo et al. 2018). We assumed a Kroupa IMF (Kroupa 2001) and a binary fraction of 35% with a flat mass ratio distribution for the secondaries. The SFH solutions were fitted with an age grid of log(t) = 6.6–10.15 using a time resolution log(δt ) = 0.1 dex for ages less than log(t) = 9 and log(δt ) = 0.05 dex for older ages, and a metallicity grid [M/H] = −2.2 to −0.3 with a resolution of 0.1 dex. The upper bound on the metallicity range was chosen to be well-above the spectroscopically measured present-day metallicity of [Fe/H] ∼ − 0.8 (Bresolin et al. 2006; Urbaneja et al. 2008). We also enforced a restriction that the metallicity monotonically increases in time. For the separate SFH fits to the NIRCam data that overlap with the ACS field, we also required that the AMR from NIRCam matched the best-fitting AMR from the ACS data, thereby providing a one-to-one comparison of the recovered SFHs. We also experimented with fitting the NIRCam data without requiring a matching AMR to the ACS data. The resulting NIRCam SFHs were in excellent agreement with the results requiring a matched AMR. The independently fitted AMRs were also in good agreement with the ACS AMRs with the initial and final values of [M/H] and overall shape of the AMR a close match for each stellar library.

We limited the fitting to stars brighter than the 50% completeness limits in each filter in a given CMD, as listed in Table 1. We note that models of the AGB phase of stellar evolution are uncertain but, given the small number of AGB stars in the data, the inclusion of these stars in the catalogs does not impact the derived SFHs. Following the same procedure used in Albers et al. (2019), we adopted the distance modulus of 24.93 mag measured from the TRGB luminosity in the HST ACS data that overlap with our NIRCam field. We also assumed the same foreground extinction value of A_V = 0.10 mag based on Galactic dust maps from Schlegel et al. (1998) with recalibration by Schlafly & Finkbeiner (2011), an R_V = 3.1, and the reddening law of Fitzpatrick (1999). WLM is at a high Galactic latitude where foreground contamination is expected to be minimal, as demonstrated in Figure 5. Nonetheless, following best practices, we included a model of foreground stars in the Milky Way in the fits, generated with the TRILEGAL simulation tool (Girardi et al. 2012). Confirming that foreground stars are minimally present in our stellar catalogs, the SFH solutions with and without the inclusion of the foreground model were identical.

Statistical uncertainties due to the finite number of stars in the CMD were estimated using a hybrid Markov Chain Monte Carlo approach (Dolphin 2013). The SFH solutions derived with the two stellar libraries show the impact of different assumptions used in the stellar models, and therefore give some indication of the systematic uncertainties due to stellar evolution. However, with only two libraries, these differences likely underestimate the true systematic uncertainties of the solutions. Future work will include fits from additional stellar libraries once they have been updated with the in-flight performance of NIRCam, as well as systematic uncertainty estimates using Monte Carlo simulations that are tuned based on solutions from multiple stellar libraries (e.g., Dolphin 2012). As a consistency check for our ACS SFH solution, we compared our results to those derived from the full ACS field of view presented in the Appendix of Albers et al. (2019), which includes fits to the BaSTI and PARSEC stellar libraries used in the present work, as well as the older Padua models (Girardi et al. 2010) and the MIST models (Choi et al. 2016). We found the solutions to be in excellent agreement and well within the uncertainties in Albers et al. (2019).

Figure 6 presents the CMD fits to the stellar catalogs from the overlapping region in the NIRCam and ACS data. These include independent solutions from the NIRCam and ACS catalogs fit with the PARSEC and BaSTI stellar libraries. Each data set+library solution has four subplots, as described in the figure caption. The most useful diagnostic to interpret the quality of the fits is the residual significance shown in the lower right-hand side for each solution. Note that the scale of the significance color bar has been fixed to the same range so that differences between each data+model can be directly compared. While we only show the results for the data in the overlapping region between NIRCam and ACS, the general patterns in these residual significance plots are representative of the solutions from the full NIRCam catalogs.

Zoom In Zoom Out Reset image size

Overall, the modeled CMDs are a very good representation of the data, with the exception of a few regions. Specifically, the red clump (RC) area is not well modeled, which has been noted in numerous previous studies with ACS optical observations (e.g., Gallart et al. 2005). It is this area that sets the extrema for the ranges in the residual significance. In addition, the modeled blue helium burning (BHeB) sequences are somewhat redder (i.e., cooler) than the observed sequences. This has also been noted previously based on optical observations (e.g., McQuinn et al. 2011), and the possible causes of the disagreement have been explored theoretically (e.g., Tang et al. 2014).

Smaller differences are seen between the models and the data for RGB stars. The PARSEC and BaSTI models create a slightly redder (i.e., cooler) RGB relative to both the NIRCam and the ACS data. A similar offset was reported at the level of 0.05 mag based on the JWST ERS data on M92 (Weisz et al. 2023). It has also been previously noted for a number of stellar models in near-infrared 2MASS broadband filters across a range of metallicities in Galactic globular clusters (Cohen et al. 2015). The slight mismatches between model and data have similar trends in the JWST and HST bands, which suggests that there are no inherent biases in using the JWST over HST bands for SFH fitting, although the mismatches are somewhat more pronounced in the JWST residuals plots.

A color offset between the model and data can indicate the presence of an internal extinction that should be considered in the fits. Since internal extinction can be treated as a free parameter in match, we explored whether adding extinction to the fits would improve the solutions. Briefly, we varied extinction up to 0.50 mag, in increments of A_V = 0.05 mag, and iteratively re-fit the CMD with each stellar library. For the BaSTI models, extinction did not reduce the residual significance in the fits. For the PARSEC models, internal extinction of ∼0.30 (∼0.15) mag compensated for the offset in color between the modeled and observed RGB sequence for the NIRCam (ACS) data, with slightly lower values in the residual significance. However, while the SFH solutions with and without the extinction were consistent with each other, the inclusion of internal extinction resulted in lower stellar metallicities in the AMRs. These lower metallicities are a poor match to spectroscopic abundances (see below), are inconsistent with the expectation of minimal extinction in such a metal-poor galaxy (e.g., McQuinn et al. 2015b), and are inconsistent with the lack of internal extinction found based on the ACS data reported in Albers et al. (2019). Thus, we did not adopt any internal extinction for our final fits, although we continued to include a foreground Galactic extinction of A_V = 0.10 mag (see Section 3.1).

We also used the photometry and ASTs run simultaneously on the JWST and HST images to fit the NIRCam and ACS CMDs together. The modeled CMDs and residuals between the observed and modeled data from both libraries were qualitatively similar to the results from fitting the NIRCam and ACS data separately. We show these results in the Appendix, which can be compared to Figure 6. Quantitatively evaluating the fits is not straightforward because on the one hand using two CMDs in the fit increases the information available to derive a SFH, while on the other hand the fits are based on different numbers of stars and the more complex data can disallow some solutions. One basis for evaluation is the Poisson likelihood statistic returned by match. Using this as a measure of the fit quality, we find the likelihood statistic from modeling the 2-CMDs simultaneously was comparable to the combined values from modeling each CMD independently using the PARSEC library, but slightly worse based on BaSTI library. This would suggest that there are no obvious advantages to using both NIRCam and ACS CMDs to recover a SFH over just a single CMD. Another basis for evaluating the fits is to compare the best-fitting SFHs and AMRs from each method, which are presented in Section 4.3.

Figure 7 presents the best-fitting cumulative SFHs (top panel) and AMRs (bottom panel) from the NIRCam full field of view with solutions using the PARSEC (solid lines) and BaSTI (dashed lines) stellar libraries. The top x-axis converts lookback time to redshift based on the Planck Collaboration et al. (2020) cosmology. Shading on the SFHs represent statistical uncertainties only. The vertical gray region encompasses the approximate range of the epoch of reionization from z ∼ 10–6, corresponding to a lookback time from 13.3–12.9 Gyr.

Zoom In Zoom Out Reset image size

The SFH and AMR results from both libraries are in excellent agreement. The differences between SFH solutions give some indication of the systematic uncertainties, but additional stellar libraries with updated JWST zero-points are needed before the full range of systematic uncertainties can be estimated. We also find that the total stellar mass formed based on the SFH fits from each stellar library are in excellent agreement, which provides an additional quantitative measure of the SFH results. These values are listed in Table 1.

The SFH of WLM (in the region imaged by NIRCam) qshows that star formation occurred at early times, followed by an interruption post-reionization (z ∼ 6; lookback time of 12.9 Gyr). This pause in star formation extended over ∼3 Gyr, after which there was a marked increase in the star formation activity. To provide a quantitative measure that can be compared with other galaxies, we calculate the timescales "τ" over which different fractions of the stellar mass were formed. Focusing on just the PARSEC models for simplicity and interpolating the SFH in Figure 7, we find that ∼10% of the stellar mass formed within a lookback time τ₁₀ = 12.61 ± 0.06 Gyr, followed by a pause until ∼10 Gyr ago, and then the mass more than doubles reaching τ₂₅ at a lookback time $\,={7.63}_{-0.03}^{+0.05}$ Gyr. The uncertainties represent statistical uncertainties determined using a Markov chain Monte Carlo approach. The stellar mass then doubled again in ∼2.5 Gyr with τ₅₀ $\,={5.08}_{-0.01}^{+0.05}$ Gyr. Finally, the lookback time for 90% of the mass to form is τ₉₀ = 1.18 ± 0.01 Gyr. We provide values of τ₁₀, τ₂₅, τ₅₀, and τ₉₀ for both the PARSEC and BaSTI models in Table 1. We discuss implications of the extended pause in star formation post-reionization in Section 5.

As a check on the fits, we compare the best-fitting AMR solutions from the NIRCam data with trends based on spectroscopic abundances of stars of different ages and from gas-phase abundances measured from H ii regions in WLM. For a comparison of the AMR at intermediate ages, we use the metallicity distribution reported in Leaman et al. (2009) from 44 RGB stars that are spatially located in a similar region of the NIRCam data (labeled as the "bar" region by these authors). The authors find a mean metallicity [Fe/H] = − 1.14 ± 0.04 with an intrinsic spread of σ = 0.33 dex based on converting summed equivalent widths of calcium ii triplet lines from medium resolution (R ∼ 3400) spectroscopy to an [Fe/H] metallicity scale. This is slightly higher than the mean [Fe/H] value of −1.28, σ = 0.38 dex found for a larger sample of 126 stars in WLM that cover the stellar disk out to larger radii (Leaman et al. 2013). The mean [Fe/H] = −1.14 can be compared with the [M/H] ²² range from −1.2 to −0.8 in the AMRs at intermediate times seen in Figure 7, which suggests general agreement between the spectroscopically determined stellar metallicity distribution function and AMR trends.

While a good general comparison, the AMRs are based on the composite stellar populations, whereas the spectroscopic MDFs were created from stars in a limited CMD region, namely stars that lie on the RGB and that are within ∼1 mag below the TRGB. Thus, to compare the AMRs with the MDFs more directly, we used the best-fitting SFH and AMR solutions to generate synthetic photometry for each stellar library. We then selected stars that lie on the RGB within ∼1 mag of the TRGB and calculated the mean [M/H] values and the standard deviations. Assuming a normal distribution, we find a 〈[M/H]〉 = − 0.94; σ = 0.25 dex for the BaSTI models and 〈[M/H]〉 = − 0.92; σ = 0.22 dex for the PARSEC models, in good agreement with the 〈[Fe/H]〉 = − 1.14; σ = 0.33 dex reported in Leaman et al. (2009).

Figure 8 shows that the [M/H] distributions (top panel) and cumulative distribution functions (bottom panel) from both libraries are consistent with the [Fe/H] values for the 44 stars from Leaman et al. (2009). The largest difference is the presence of a metal-poor tail in the spectroscopically determined [Fe/H] values, which is absent in the AMR-based distribution functions. We note that the lack of a metal-poor tail in the modeled distributions is consistent with the corresponding lack of metal-poor blue horizontal branch stars in both the ACS and the NIRCam CMDs. A two-sided Kolmogorov–Smirnov (KS) test performed on the cumulative metallicity distribution functions from the Leaman et al. data and our values showed there is 26% probability of the results being drawn from the same distribution for the PARSEC models and 54% for the BaSTI models. The difference in the K-S tests is likely driven by the different peaks and widths of the distributions, which is seen most clearly in the normalized histograms in the top plot. Regardless of the subtle differences, this is quite good agreement given the somewhat different spatial regions sampled, relatively smaller number of spectroscopic measurements, and the different techniques used in the comparison.

Zoom In Zoom Out Reset image size

For a comparison of the AMR results at more recent times, we consider spectroscopic measurements of young stars and the present-day gas abundance. Specifically, the metallicity of eight young A and B supergiant stars was reported to have a mean [Fe/H] of −0.87 ± 0.06 (Urbaneja et al. 2008, see also Bresolin et al. 2006). These values are in very good agreement with the measured gas-phase oxygen abundance from H ii regions (e.g., 12+log(O/H) = 7.83 ± 0.06; Lee et al. 2005). We also note that there is evidence that WLM hosts young stars with higher metallicities based on the study of two supergiants (Venn et al. 2003) and one massive O star (Telford et al. 2021). For comparison, the AMRs at the present day show [M/H] values ranging from −0.6 to −0.8, depending on the stellar library, which is again in good agreement with the spectra-based values.

Here, we focus on a quantitative comparison of the NIRCam- and ACS-based SFHs. The fits are based solely on the stellar catalogs from the overlapping footprint between NIRCam and ACS, as seen in Figure 1, rather than the catalogs from the full field of view of either instrument.

Figure 9 presents the SFH (top panel) and AMR (bottom panel) solutions from NIRCam data (red) and ACS data (blue) using the PARSEC (solid lines) and BaSTI (dashed lines) models. Shaded regions represent statistical uncertainties on the fits. The SFH results for NIRCam and ACS data from both the PARSEC and BaSTI libraries are in excellent agreement at all lookback times. These results confirm the efficacy of the F090W–F150W filter combination and stellar evolution libraries in the near-infrared for deriving SFHs relative to the HST optical filters.

Zoom In Zoom Out Reset image size

Figure 9 also shows the SFHs from the simultaneous fits to the NIRCam and ACS CMDs (green lines). These solutions are also in excellent agreement with the SFHs and AMRs derived independently from the NIRCam data and from the ACS data (i.e., single CMD-fitting). The agreement indicates that the SFHs recovered by fitting just one CMD in the optical or in the near-infrared are not significantly biased compared to solutions recovered from simultaneously fitting two CMDs that intrinsically contain more information, and demonstrates a degree of robustness for the single CMD fit.

As a further comparison of the different SFH fits, we calculate the total stellar mass recovered from the different data sets (NIRCam, ACS, NIRCAM+ACS) and stellar libraries (PARSEC, BaSTI) in the overlapping field of view. The values are in close agreement, ranging from log(M_*/M_⊙) of 6.54 to 6.65, providing additional evidence of the consistency between the SFH recovered from the two observatories. See Table 1 for individual values.

Compared to the SFH results derived from the full NIRCam field of view (Figure 7), the SFHs from the overlapping NIRCam–ACS region (Figure 9) show a slightly slower start to star formation. This suggests a lower median stellar age at larger radii. Similarly, the AMRs in the overlapping NIRCam–ACS data have slightly lower values than the AMRs from the full NIRCam field, suggesting lower mean metallicities at larger radii.

Albers et al. (2019), who reconstructed the SFH from the full ACS data set, also report the SFH from an outer field in WLM using deep HST data from the WFC3 instrument. Compared with the SFHs from regions imaged with the ACS and NIRCam, these results suggest a strong radial age gradient. Quantitatively, the inner fields formed 50% of their stellar mass ∼5 Gyr ago compared with ∼10 Gyr in the outer field located only ∼0.4 kpc off the minor axis. Future work from our team will combine the JWST NIRCam field, the different HST fields, and the JWST NIRISS field observed as part of the ERS program to explore this in greater detail (R. Cohen et al. 2023, in preparation).

In this work, we use deep JWST NIRCam imaging of the Local Group star-forming galaxy, WLM, from the ERS-1334 program to derive the SFH and AMR of the central field in the galaxy using two sets of stellar libraries (PARSEC, BaSTI). We have leveraged existing optical imaging obtained with the HST ACS instrument on an overlapping region (Albers et al. 2019) to perform a one-to-one comparison of NIRCam-based and ACS-based SFH results. We find the SFHs and AMRs derived independently and simultaneously fitted using data from both great observatories to be in excellent agreement for the two stellar libraries. In addition, the NIRCam-based AMRs agree with spectroscopic metallicity distribution functions, providing an independent check on the recovered solutions.

This first detailed look at a SFH using NIRCam data confirms the efficacy of recovering the SFH with the F090W–F150W filter combination relative to the workhorse optical filters on HST that have been used for SFH work for the past two decades. These results also provide validation of the sensitivity and accuracy of stellar evolution libraries in the infrared relative to the optical. With JWST's greater sensitivity and resolution, detailed and robust SFHs can be recovered with greater efficiency than has been possible with HST and for galaxies at greater distances. It also validates that future SFH results from JWST can be directly compared with existing SFH studies from HST data without complications of significant biases due to observations from these different observatories and wavelength coverage.

To date, only a few gas-rich isolated dwarf irregular (dIrr) galaxies have robustly reconstructed ancient SFHs from deep CMDs in the literature. In order of decreasing mass, these include IC 1613 (Skillman et al. 2003), WLM (Albers et al. 2019, and the present work), Leo A (Cole et al. 2007), and Aquarius (Cole et al. 2014). While the most massive of these galaxies (IC1613; M_* = 1 × 10⁸ M_⊙) shows relatively constant star formation across cosmic time, the remaining three galaxies (M_* ∼2 × 10⁶ to 4 × 10⁷ M_⊙) exhibit some level of early star formation followed by a pause that occurs post-reionization. These galaxies hosted a similarly early onset of star formation but then a temporary quiescent period ensues as, presumably, the metal-poor gas present in and/or accreted onto the galaxies was prevented from cooling efficiently, with myriad heating sources (e.g., cosmological reionization, the local far-UV radiation field, heating from internal stellar feedback, etc.) potentially playing important roles.

Although this is a very small sample, the emerging pattern is intriguing, suggesting that even in the stellar mass range of 10⁶–10⁷ M_⊙, galaxies experience a post-reionization pause in star formation. Similarly deep CMDs of additional isolated dIrr galaxies are needed to confirm and further investigate this trend. JWST is currently the only facility with the capabilities to increase the sample size of isolated galaxies (i.e., galaxies outside the Local Group) with robust ancient SFHs.