DELAYED STAR FORMATION IN ISOLATED DWARF GALAXIES: HUBBLE SPACE TELESCOPE STAR FORMATION HISTORY OF THE AQUARIUS DWARF IRREGULAR^*

The question of the early evolution of low-mass galaxies has been a vexing one for galaxy formation theory in a wide range of hierarchical merger models for more than three decades (e.g., Dekel & Silk 1986; Kauffmann et al. 1993; Somerville 2002; Benson et al. 2003; Sawala et al. 2011; Kuhlen et al. 2012). During this timespan a range of problems has been identified and addressed in successive generations of models; these problems tend to relate to the efficiency with which gas is accreted, retained, cooled, and astrated in the lowest-mass dark matter halos.

Recurring issues surrounding the over- or under-production of stars in simulations include overcooling, reionization, the role of feedback, and the effects of tidal perturbations; a recent, detailed discussion of these issues in the context of Local Group dwarf galaxies may be found in Skillman et al. (2014). A variety of recently proposed solutions invoke a range of external heating and internal feedback processes, many of which are strongly mass and/or environment dependent (e.g., Hopkins et al. 2013, and references therein); the influence of numerical resolution on the simulations is also likely to play a role (e.g., Brooks & Zolotov 2014). In addition to supernova feedback, reionization of the Universe at redshifts ≳6 has been considered as a source of heating that might inhibit star formation in the smallest galaxies (e.g., Babul & Rees 1992). In the simplest formulation, reionization may quench star formation in galaxies with a total dark halo mass ≲ 10¹⁰ M_☉ by warming gas to the point where it may be driven out of the shallow potential wells of these systems (e.g., Thoul & Weinberg 1996; Gnedin 2000; Ricotti et al. 2002); a review of the observational evidence for and against such a scenario has been given by Weisz et al. (2014b), who find that there is some evidence for negative feedback on star formation, but with large galaxy-to-galaxy variability and with some inconsistency between the quenching timescale and the epoch of reionization.

While some process (or set of processes) clearly inhibits the formation of dwarf galaxies compared to their proliferation in cold dark matter-only simulations, the nature of these processes is still an open question. Continued observational study of isolated dwarfs is necessary for at least two reasons; one is to probe the ancient star formation history (SFH) independent of local external perturbations, and another is to investigate the possible quenching or delay in star formation near the transition mass below which star formation is strongly suppressed. The most isolated galaxy previously observed with sufficient depth and resolution to distinguish the oldest main-sequence turnoff (MSTO) population is Leo A (DDO 69; Cole et al. 2007), a galaxy on the opposite side of the Local Group from Aquarius, but with similar dynamical mass. Leo A was found to be the dwarf galaxy with the youngest mean age and the smallest fraction of stars older than 10 Gyr of any Local Group member (Hidalgo et al. 2013; Weisz et al. 2014a, and references therein).

In this context, it is of great interest to know if the unusual SFH of Leo A presented in Cole et al. (2007) is an anomaly or a general feature of small, isolated galaxies. In this paper we show that the Aquarius dwarf has many similarities with Leo A, and explore some of the implications of this consistency. We present our observational program and data reductions in Section 2, and the stellar content of Aquarius is shown in Section 3. We derive the SFH of Aquarius in Section 4 and compare its initial low level of star formation and sudden increase some 6–8 Gyr ago to other Local Group dwarf galaxies in Section 5. We put these results into context with recent galaxy formation models in Section 6.

The Aquarius dwarf galaxy was discovered on Palomar sky survey prints by van den Bergh (1959) and identified by him as a low surface brightness, very blue galaxy at the limit of resolution into stars. Based on the radial velocity of neutral hydrogen, Fisher & Tully (1975) identified Aquarius as a possible Local Group member. Aquarius is extremely isolated; McConnachie (2012) lists it as one of just three Local Group galaxies with a free-fall time to the Local Group barycenter larger than a Hubble time given its radial velocity.¹¹ Currently only two other galaxies are known to lie within ≈500 kpc of Aquarius; these are SagDIG at a distance of ≈340 kpc, and NGC 6822, ≈440 kpc away. Numerical action models of Local Group dynamics indicate that Aquarius is on its first infall into the Local Group from quite a large distance, and is exceedingly unlikely to have tidally interacted with any other known galaxy (e.g., Shaya & Tully 2013).

Early CCD photometric studies found Aquarius to have an extended history of star formation continuing to very recent times, albeit at a low level (Greggio et al. 1993). Despite a high fraction of cold H i (Young et al. 2003), virtually no current star formation was detected by Hα imaging surveys (van Zee et al. 1997). The consequent lack of bright, evolved stars (e.g., Elias & Frogel 1985; Marconi et al. 1990) confirms Aquarius to be one of the lowest-luminosity gas-rich galaxies in the Local Group (and also resulted in longstanding ambiguity about its true distance). Aquarius is classified as a transition type galaxy, with properties intermediate between the dwarf irregulars and spheroidals (e.g., Mateo 1998). van den Bergh (1994) drew attention to a possible link between its low stellar mass, high gas fraction, and isolated location in the context of the Local Group density–morphology relation (Einasto et al. 1974). The prospective connection to environment was further explored by Grebel et al. (2003) in the light of proposed gas-stripping mechanisms that might produce spheroidals from low-luminosity irregulars.

The basic properties of DDO 210 are summarized in Table 1. The first tip of the red giant branch (TRGB) distance to Aquarius, 950 ± 50 kpc, was published by Lee et al. (1999), who confirmed both the Local Group membership and the unusually low luminosity of the galaxy. This distance was confirmed with Hubble Space Telescope (HST)/WFPC2 photometry by Karachentsev et al. (2002), who derived a distance of 940 ± 40 kpc; later reanalyses found distances of 968 ± 63 (Jacobs et al. 2009) and 977 ± 63 (Tully et al. 2009). McConnachie et al. (2005) found two possible locations for the TRGB and favored the more distant value of 1071 ± 39 kpc, arguing that smaller estimates could be the result of contamination of the brightest part of the red giant branch (RGB) by asymptotic giant branch (AGB) stars. We adopted the more conservative distance in planning the present observations, but in the course of color–magnitude diagram (CMD) modeling the best-fit solutions to the SFH were found to converge on a distance modulus of 24.95 ± 0.10; this corresponds to a distance of 977 ± 45 kpc and is in good agreement with our preliminary analysis of the TRGB magnitude in our data. A more complete reanalysis of the distance to Aquarius will be given in our analysis of the TRGB and variable star populations (E. D. Skillman et al., in preparation).

Table 1. Basic Properties of Aquarius

Parameter	Value	References
Galactic coordinates (ℓ, b) (deg)	341, −314	(1)
Distance (kpc)	977 ± 45	(2)
Line of sight reddening E(B − V) (mag)	0.045	(3)
Absolute magnitude MV	−10.6	(4)
Half-light radius (arcmin)	1.47 ± 0.04	(4)
Mass within r1/2 (M☉)	1.8$^{+0.9}_{-0.7}\times 10^{7}$	(5, 2)
Stellar mass (M☉)	1–2 × 106	(6, 4, 2)
MH i (M☉)	2.7 × 106	(7)
Current SFR (M☉ yr−1)	1.6 × 10−4	(8)
Stellar metallicity, range (dex)	−1.44, ± 0.35	(9)

References. Key to references: (1) Marconi et al. 1990; (2) this paper; (3) Schlafly & Finkbeiner 2011; (4) McConnachie et al. 2006; (5) Kirby et al. 2014; (6) Lee et al. 1999; (7) Begum & Chengalur 2004; (8) Lee et al. 2009; (9) Kirby et al. 2013.

Download table as:  ASCIITypeset image

The first stellar metallicities were obtained by Kirby et al. (2013), who measured spectra in the region of the NIR calcium triplet for 24 red giants, finding a mean metallicity 3.6% of solar (Z/Z_☉ = 0.036, or expressed logarithmically and referred to the iron abundance, [Fe/H] = −1.44). This value is consistent with the expectations from the dwarf galaxy mass–metallicity relation of Lee et al. (2006) and typical for a dwarf galaxy of the luminosity of Aquarius. Kirby et al. (2013) found a range of metallicities, with a 1σ spread about the mean of ±0.35 in the log, also typical for dwarfs with a prolonged history of star formation.

Kirby et al. (2014) measured the velocity dispersion of the Aquarius red giants to be σ_v = 7.9$^{+1.9}_{-1.6}$ km s⁻¹ and from this derived a dynamical mass at the half-light radius of 2.0$^{+1.0}_{-0.8}\times 10^{7}$ M_☉ and a mass to light ratio ϒ = 32$^{+16}_{-14}$. This result is in good agreement with results based on the H i rotation curve within 550 pc of the center of Aquarius, by Begum & Chengalur (2004). Under the assumption that the neutral hydrogen is distributed in a circularly symmetric disk with inclination i_H i = 27°, Begum & Chengalur (2004) found V_circ = 16 km s⁻¹, leading to an enclosed mass within 570 pc of M_dyn = 3.4 × 10⁷ M_☉.

The two dynamical mass estimates agree well within the errors, recognizing that the half-light radius used by Kirby et al. (2014) is 342 pc, about 62% of the distance to the farthest point on the H i rotation curve. Begum & Chengalur (2004) go further, and match the velocity profile to a standard Navarro, Frenk, and White (NFW) dark matter profile (Navarro et al. 1997), finding a best fit maximum velocity at the virial radius V₂₀₀ ≈ 38 km s⁻¹. This is a major extrapolation, given the virial radius of a few tens of kiloparsecs is many times farther out than the last measured data point. If taken at face value, this profile fit suggests a total mass within the virial radius on the order of 10¹⁰ M_☉, which allows at least rough comparisons to cosmological simulations.

The SFH presented here has been calculated using the code developed by Cole, which has been used extensively on other Local Group dwarfs and tested against other methods with good consistency (e.g., Skillman et al. 2003; Cole et al. 2007; Monelli et al. 2010b; Cignoni et al. 2012; Lianou & Cole 2013; Skillman et al. 2014). Here, we use the Padova-Trieste (PARSEC; Bressan et al. 2012) isochrones. These are preferred because they have the most up to date input physics, include post-helium-core-flash phases of evolution, cover a wide range of age and metallicity at user-definable grid spacings, and are tuned to the revised solar oxygen abundance (see, e.g., Bahcall et al. 2005). This specific choice plays a role in determining the precise age and duration of episodes of high or low SFR and the overall fit quality of the synthetic CMD, but the broad conclusions are robust against the choice of isochrone library owing to the depth of the photometry (Weisz et al. 2011).

The IMF is that of Chabrier (2003), which is a Salpeter-like power law for M/M_☉ ⩾ 0.8 and log-normal for lower masses. Binary stars are included as per the statistics presented in Duquennoy & Mayor (1991) and Mazeh et al. (1992), whereby 35% of the "stars" are single objects, and the rest are binary; in turn, 25% of the binaries are defined as "close" binaries in which the secondary mass is drawn from a flat IMF instead of the Chabrier IMF. The choice of IMF and binary statistics may slightly alter the mass normalization of the solutions but do not significantly affect the shape of the derived SFH, as discussed extensively in Monelli et al. (2010a) and Cignoni et al. (2012), among others. The distance and reddening were initially set to the values derived by McConnachie et al. (2006), but were allowed to vary as demanded by the fitting procedure in order to best match the observed CMD. The stability of the solutions has been confirmed by comparison to fits made independently with different choice of IMF and binary fraction, using the MATCH software package (Dolphin 2002).

The isochrone library contains stars of ages between 6.60 ⩽ log(Age/yr) ⩽ 10.12, spaced by intervals of 0.01, and metallicities from 1/150th to 1/10 solar, evenly spaced at intervals of ≈0.15 in log(Z). Multiple metallicities are allowed at a given age and we do not impose an age–metallicity relation (we do exclude metallicities above one-tenth solar from consideration). The isochrones at each metallicity are grouped into age bins in order to speed the computation; the age bins are initially taken to be evenly spaced by 0.10 in log(age), but the bins may grow or shrink depending on the noise level in the solutions that may justify finer or coarser spacing.

We adopted color and magnitude bins 0.05 and 0.10 mag wide, respectively. Because the data quality is high, we model the region with −0.5 ⩽ (m₈₁₄ − m₄₇₅) ⩽ 2.75 and 19 ⩽ m₈₁₄ ⩽ 29, but we have checked that our conclusions are not significantly altered if these limits are shifted to brighter magnitudes. The fitting procedure accepted the adopted reddening value of E(B − V) = 0.05 and the Tully et al. (2009) distance modulus (m − M)₀ = 24.95, but a good fit could not be found using the longer distance modulus from McConnachie (2012). As a check we performed a TRGB fit and found good consistency for distances ≈977 ± 45 kpc; the distance issue will be explored further in our paper presenting the variable star light curves (E. D. Skillman et al., in preparation).

The best-fit synthetic CMD is shown alongside the binned data in Figure 3. The overall fit quality is very high, as shown by the residual significance plot in the right-hand panel. The RGB color and breadth, the magnitude, color, and morphology of the red clump, the upper main sequence color, and the location and slope of the subgiant branches (SGBs) are all well-reproduced. The locations of the blue loop and red supergiant stars are indistinct due to the small number statistics and foreground contamination, but appears to be reasonable for metallicities Z ≈ Z_☉/20. The independent fits made using MATCH gave very similar solutions within the uncertainties.

Zoom In Zoom Out Reset image size

We present the lifetime SFH of Aquarius in Figure 4. The major episode of star formation in Aquarius took place ≈6–8 Gyr ago, although star formation has been continuous (on Gyr timescales) over its entire lifetime. Although star formation at the earliest times is low, it is inconsistent with zero. However, the drop in SFR between ages of 10–12 Gyr is robust and significant; the SFR in this age bin is only nonzero at the ≈2σ level. After the burst of activity at intermediate ages the SFR dropped, and has declined further over the past ≈2 Gyr.

Zoom In Zoom Out Reset image size

Aquarius is one of the least luminous gas-rich galaxies in the Local Group, but it does extend beyond the footprint of the ACS/WFC field (202''). McConnachie et al. (2006) trace the structure out to at least a radius of 4', so our field only covers ≈50% of the area over which Aquarius stars are known. This is potentially problematic for interpreting the age of Aquarius based on our SFH, given the ubiquity of age gradients and extended structures in even relatively isolated irregular galaxies—for example, the case of Sextans A and B (Dohm-Palmer et al. 1997; Bellazzini et al. 2014). It has already been noted that the young stars in Aquarius are slightly off-center from the H i contours (McConnachie et al. 2006), which may be suggestive of a possible recent disturbance.

However, there is good reason to believe that our SFH is giving a nearly complete picture of the history of Aquarius. The surface density profiles in McConnachie et al. (2006) drop exponentially, becoming indistinguishable from the foreground population at ≈4' along the major axis; the equivalent distance along the minor axis is 2'. McConnachie et al. (2006) give an exponential fit to the elliptical surface density profile of Aquarius; by integrating their profile over the area of the ACS/WFC footprint, we find that 80% of the Aquarius members brighter than V = 26 are contained in our field. The integral of our SFR(t) over the Hubble time gives a total astrated mass ≈4 × 10⁶ M_☉; based on the profiles from McConnachie et al. (2006) we may be missing ≈10⁶ M_☉. Even if all of the missed stars were formed prior to 10 Gyr ago, then at the very most the derived full-galaxy SFH would show a time-averaged SFR from 10 to 13.7 Gyr of ≲40% of the measured SFR from 6 to 8 Gyr ago. Even in this unrealistic "oldest case" scenario for the stars in the outskirts of Aquarius, the galaxy as a whole would be dominated by intermediate age stars.

As a further check for possible population gradients, and to attempt to characterize the outer structure of Aquarius, we obtained parallel imaging with the HST WFC3 simultaneously with our ACS images of the central field. The WFC3 field is centered ≈58 (1.65 kpc) from the center of the ACS/WFC field, nearly along the major axis. Despite a high degree of completeness down to a magnitude of ≈28, we were unable to identify any sign of the stellar sequences in Aquarius in these images. At most a few Aquarius stars per square arcminute could be in the WFC3 field; unless the stellar distribution is highly asymmetric and we have been very unlucky in our parallel field placement, no significant population of Aquarius stars is present at radii ≳ 5' (1.4 kpc), down to very strict limits around the old MSTO.

The metallicity evolution of Aquarius is summarized in Figure 5, and is in excellent agreement with the spectroscopic metallicities reported by Kirby et al. (2014) and with the photometric estimates in McConnachie et al. (2006). The error bars in Figure 5 show the rms of the mass-weighted metallicity in each age bin; because the isochrone models are separated by discrete intervals of ≈0.15–0.20 in log metallicity, the error bars are typically about this size, except where a very wide range of metallicities are required to reproduce the CMD, as in the oldest age bins. The wide metallicity range for ages ≳8 Gyr likely reflects both the rapid early increase in metallicity and the small number of stars in the CMD at those ages. Because of the rather coarse age resolution at the earliest times, we are unable to unambiguously assign an age to the youngest major populations with Z < 0.01 Z_☉. After the period of rapid early enrichment, ubiquitous in synthetic CMD analyses of dwarf galaxies, the metallicity remains nearly constant over time, slowly rising to a value of ≈ Z_☉/20, fairly typical for dwarf galaxies of this luminosity (e.g., van Zee et al. 1997). The mean metallicity at intermediate ages is in good agreement with the Local Group dwarf galaxy stellar mass–metallicity relationships reported by Berg et al. (2012) and by Kirby et al. (2013).

Zoom In Zoom Out Reset image size

Narrowband imaging and optical spectroscopy at Hα by van Zee et al. (1997) revealed no current massive star formation activity, even though the presence of bright blue main sequence and evolved stars proves that star formation has occurred within the past 10⁸ yr. Aquarius is one of the faintest, lowest SFR galaxies in the 11HUGS sample of gas-rich, star-forming galaxies within 11 Mpc (Kennicutt et al. 2008), with a large discrepancy between the inferred Hα and far-ultraviolet SFR owing to the flickering nature of star formation in dwarf irregulars, combined with stochastic sampling of the IMF at low SFR (Lee et al. 2009; Fumagalli et al. 2011; Weisz et al. 2012a).

A higher resolution view of the SFH over the past 1 Gyr is shown in Figure 6. Aquarius is in a period of low SFR, with seemingly random fluctuations on timescales of ≈10⁸ yr. The recent drop in activity after an upward fluctuation is in excellent agreement with the estimates of current SFR reported in Lee et al. (2009), whose UV (∼100 Myr sensitivity; dashed line) and Hα (∼10 Myr sensitivity; circled star) results are overplotted on Figure 6. The enhancement at ages of 250–300 Myr has been noted by previous authors as discussed above, and is identifiable here by the increased density of blue loop stars at m₈₁₄ ≈ 22. It is natural to interpret this increase in SFR lasting several tens of Myr as a stochastic fluctuation in SFR of the type frequently observed in dwarf galaxies, leading to the wide variety of Hα, UV, and IR star formation indicators in dwarf galaxies where the gas densities and total mass participating in star formation are small (e.g., Lee et al. 2009). The event is of shorter duration than a typical episode that would be identified as a dwarf galaxy starburst (McQuinn et al. 2010).

Zoom In Zoom Out Reset image size

In Leo A, the next most-isolated galaxy to be studied in equivalent detail, star formation also began at the oldest ages, but continued at only a very low level until ≈6 Gyr ago (Cole et al. 2007). Like Aquarius, Leo A has almost certainly never interacted with another (known) galaxy (e.g., Shaya & Tully 2013). The two galaxies are similar in their degree of isolation, total magnitude M_B, color, and gas content, so any similarities that can be identified in their SFH are of great interest.

Comparing the dynamical masses within the half-light radii, Kirby et al. (2014) gives M_1/2 = 15$^{+6}_{-5}\times 10^{6}$ M_☉ for Leo A and 20$^{+10}_{-8}\times 10^{6}$ M_☉ for Aquarius; the latter value may be reduced by 9% to account for the revised distance estimate presented in this paper (see Table 1). With that correction the two galaxies are seen to have nearly the same total mass, although they have somewhat different mass to light ratios, in part because of the large number of young stars in Leo A. Leo A also supports a higher current SFR, has a higher gas fraction (McConnachie 2012), and a slightly lower present-day metallicity (Kirby et al. 2013).

In both galaxies, the total stellar mass reported by different authors varies based on the adopted distance and IMF, and details of the integrated light photometry. Most reported stellar mass estimates for Leo A ultimately derive either from the work of Lee et al. (2003) or Vansevičius et al. (2004); correcting these to the consistent distances derived in Dolphin et al. (2002) and Cole et al. (2007) we conclude that the best estimate of the stellar mass of Leo A is ≈3 ± 1 ×10⁶ M_☉. Fewer measurements are available for Aquarius; using the method of Lee et al. (2003) to convert from integrated magnitude and B − V color to stellar mass indicates that Aquarius has ≈1/2 (using the photometry of Lee et al. 1999) to ≈1/3 (using the photometry of McConnachie et al. 2006) of the stellar mass of Leo A, ≈1–2 × 10⁶ M_☉.

The CMDs of Aquarius and Leo A, corrected for distance and reddening, are shown side by side in Figure 7, highlighting their overall similarity and differences in detail. Apart from the quite different HB/red clump morphology, the biggest difference is in the SGBs, where Leo A is much more strongly concentrated around M814₀ ≈ +1.5–2, with only a scattering of fainter stars.

Zoom In Zoom Out Reset image size

In order to quantify the similarities and differences in the SFH of Aquarius and Leo A, we rederive the Leo A SFH using the data from Cole et al. (2007) with the identical stellar evolution library and time bins used in our solution for Aquarius. The results are entirely consistent with the SFH reported in Cole et al. (2007); the two galaxies are directly compared in Figure 8. It is seen that the drought in star formation was broken earlier in Aquarius, but that the galaxy then rapidly dropped back down to nearly its lifetime average SFR. By contrast, Leo A formed stars at a very slow rate over much of its lifetime and only managed to start converting gas into stars in earnest ≈2 Gyr after Aquarius. The median age of star formation (the time by which 50% of the mass of all stars ever formed had been created) for each galaxy is shown with the vertical dotted lines; the median age of Aquarius is 6.8 Gyr, but for Leo A it is 4.2 Gyr.

Zoom In Zoom Out Reset image size

The absolute SFRs, the integral of SFR over lifetime, and metallicity evolution of the two galaxies are compared in Table 2. The two galaxies are clearly of a very similar nature, with a nearly equal total astrated mass. It is interesting that, with the two solutions run completely independently and with no enforcement of a chemical evolution law, Aquarius is found to be more metal-rich than Leo A at almost every age. For the intermediate age stars that dominate the red giant sample of Kirby et al. (2013), the metallicity difference between the two galaxies is of the same magnitude and in the same sense as observed, with Leo A about 0.1–0.2 dex more metal-poor. While Leo A is ≈1 mag brighter than Aquarius at the present day, our modeling finds that its total lifetime astrated mass has been less than Aquarius; this difference is also in the same sense, and of the same order of magnitude as the dynamical mass difference reported by Kirby et al. (2014). Presumably the difference in metallicity may be related back to the more gas-rich nature of Leo A, which is a factor of two to three more gas-rich than Aquarius (McConnachie 2012) and therefore less chemically evolved. At the oldest ages, Aquarius did not just form stars at a higher relative rate than Leo A, but had a higher absolute SFR (a fact that is reflected in its much more prominent BHB).

Table 2. Comparative SFH of Aquarius and Leo A

Lookback Time	Redshifta	Aquarius			Leo A
(Gya)	(z)	SFRb (10−4 M☉ yr−1)	∫M⋆c (106 M☉)	Z/Z☉d	SFR (10−4 M☉ yr−1)	∫M⋆ (106 M☉)	Z/Z☉
0–0.5	0–0.04	0.76 $^{+0.06}_{-0.06}$	4.03	0.065	1.82 $^{+0.10}_{-0.10}$	3.54	0.051
0.5–1	0.04–0.08	1.04 $^{+0.23}_{-0.23}$	3.99	0.065	1.73 $^{+0.14}_{-0.15}$	3.45	0.051
1–2	0.08–0.2	1.26 $^{+0.44}_{-0.37}$	3.94	0.056	3.19 $^{+0.23}_{-0.20}$	3.36	0.041
2–4	0.2–0.4	2.77 $^{+0.27}_{-0.44}$	3.81	0.053	5.70 $^{+0.53}_{-0.55}$	3.05	0.039
4–6	0.4–0.7	3.28 $^{+0.88}_{-0.27}$	3.26	0.052	6.31 $^{+0.51}_{-0.80}$	1.91	0.039
6–8	0.7–1.1	8.26 $^{+0.39}_{-1.87}$	2.60	0.034	1.33 $^{+0.14}_{-0.69}$	0.64	0.022
8–10	1.1–1.8	2.38 $^{+1.51}_{-0.26}$	0.95	0.032	0.99 $^{+0.30}_{-0.10}$	0.38	0.026
10–11.5	1.8–3	0.50 $^{+0.16}_{-0.27}$	0.48	0.020	0.12 $^{+0.26}_{-0.10}$	0.18	0.013
⩾11.5	⩾3	2.22 $^{+0.96}_{-0.97}$	0.40	0.007	0.90 $^{+0.22}_{-0.64}$	0.16	0.007

Notes. ^aAssuming Ω = 1, Ω_Λ = 0.73, H₀ = 71 km s⁻¹ Mpc⁻¹; Wright (2006). ^bAverage star formation rate over the given time period. ^c∫SFR(t)dt from formation until the end of the given time period (excludes losses due to finite stellar lifetimes). ^dAverage metallicity (mass fraction of elements heavier than helium) of stars formed during the given time period, compared to the solar metallicity Z_☉ = 0.0152.

Download table as:  ASCIITypeset image

All other galaxies with comparably deep CMDs (in absolute magnitude) are much closer to the Milky Way¹² and therefore complicated by the possibility that past interactions have strongly influenced their SFH. McConnachie (2012) shows that the irregular and transition type galaxies IC 1613, Leo T, Phoenix, and LGS-3 are all at least a factor of two closer to a massive spiral than Aquarius. The isolated systems WLM and Sagittarius DIG have been surveyed with WFPC2 (see Weisz et al. 2014a), but not to a depth sufficient to resolve the oldest MSTO; the deepest ACS data for UGC 4879 reach the level of the red clump (Jacobs et al. 2011).

The non-satellite dwarf spheroidals Tucana and Cetus are somewhat different. They differ from the gas-rich galaxies in their SFHs; this may or may not be related to a proposed history of tidal interactions (Fraternali et al. 2009; Lewis et al. 2007, for Tucana and Cetus respectively). Nevertheless while interactions may be suspected for Tucana and Cetus (e.g., Teyssier et al. 2012), the timing and identity of the main tidal perturber cannot be established with confidence. They are included here for completeness.

Most of the Local Group dwarf galaxies show either nearly constant long-term average SFR, or are dominated by much older bursts. The latter pattern is more common in the nearby satellites of the Milky Way and M31 and among the lowest-luminosity systems. The Tucana and Cetus dwarf spheroidals are nearly pure ancient populations (albeit with ages differing by ≈1 Gyr; Monelli et al. 2010b, 2010a). The dwarf irregular galaxies, which statistically tend to be found at large distances from both M31 and the Milky Way, show more uniform SFRs over cosmic time. The prototypical example with comparable ACS data could be IC 1613, a far larger galaxy (M_V = −15.2) which shows nearly constant SFR (Skillman et al. 2014). At the other end of the mass scale, Leo T (M_V = −8) is observed to show a slowly declining SFR beginning at the earliest times and continuing until ≈25 Myr ago without dramatic variation (Weisz et al. 2012b; Clementini et al. 2012).

From Figure 8, we see that there is a wide range of observed SFHs at a given galaxy mass. These galaxies are not satellites of either the Milky Way or M31, so no tidal or ram pressure quenching of star formation is expected, and indeed most of the systems show extended star forming lifetimes. Among the galaxies with a freefall time onto either the Milky Way or M31 less than a Hubble time, no examples of significantly delayed peak SFR are found. Leo A and Aquarius, the two most isolated galaxies in the sample, on the other hand, simmer for at least 3–4 Gyr after reionization before dramatically increasing their SFR. Present-day isolation cannot be the sole factor determining the early SFH, however, because the similarly distant and similarly massive dwarf spheroidals, Cetus and Tucana, have the highest fraction of stars older than 10 Gyr among the galaxies in our sample.

The comparison to other Local Group galaxies serves to reinforce the point that there is no simple relationship between current galaxy mass, isolation, and age of the main stellar population; it is physically intuitive to suppose that such a relationship would be clarified if the complete history of interaction and feedback heating for each galaxy could be known. If there are any trends, it seems that major star formation events are more evenly distributed in redshift than in lookback time, and galaxies with low baryonic mass are more prone to bursty behavior than more massive and/or gas-rich systems.