An Overview of the Dwarf Galaxy Survey

One avenue to help us understand the star formation activity and the evolution of dwarf galaxies is studying the dust emission properties. In addition to being the culprit in obscuring the starlight in galaxies, potentially biasing our view of stellar activity, dust plays a more positive role as a major actor in the thermal equilibrium of galaxies. Dust absorbs starlight of a wide range of energies, emitting in the MIR to submm/mm wavelength regime. In spite of their metal-poor ISM, dwarf galaxies have been known, even since IRAS observations, to possess non-negligible amounts of dust with some low metallicity galaxies globally harboring warmer dust on average than their dust-rich counterparts. The more active blue compact dwarf galaxies (BCDs), for example, revealed higher IRAS 60/100 μm flux ratios (i.e., warmer dust) compared to those of disk-type metal-rich galaxies while the elevated 25/12 μm fluxes highlighted the presence of warm stochastically-heated MIR-emitting dust and the dearth of the smallest grains/large molecules, polycyclic aromatic hydrocarbons (PAHs) (e.g., Helou 1986; Hunter et al. 1989; Sauvage et al. 1990; Melisse & Israel 1994). The dramatic improvement in sensitivity of Spitzer accentuated these findings of galaxy-wide warmer dust in star forming dwarf galaxies (e.g., Jackson et al. 2006; Rosenberg et al. 2006; Cannon et al. 2006; Walter et al. 2007), with an extreme case being one of the lowest metallicity galaxies known in the Universe, SBS0335-052 (Houck et al. 2004). ISO and Spitzer spectroscopy confirmed the relatively weak PAH bands in dwarf galaxies (e.g., Madden et al. 2006; O'Halloran et al. 2006; Wu et al. 2006; Hunter & Kaufman 2007; Smith et al. 2007; Hunt et al. 2010) and are sometimes completely absent as metallicities drop to a threshold level of about 12 + log(O/H) = 8.2—about 1/5 to 1/3 Z_⊙ (Engelbracht et al. 2005).

While observations have been modeled with sufficient MIR to submm coverage in only a limited number of local Universe metal-poor galaxies with IRAS, ISO, Spitzer, and ground-based telescopes, their SEDs were already exhibiting evidence for dust properties with other notable differences from those of our Galaxy as well as other metal-rich starburst galaxies. Excess emission in the submm/mm wavelength range over the expected Rayleigh-Jeans behavior was noted in the COBE observations of our Galaxy (Reach et al. 1995) and has been detected in low-metallicity dwarf galaxies using ground-based telescopes observing 850/870 μm with SCUBA on the JCMT and with LABOCA on APEX (e.g., Galliano et al. 2003, 2005; Bendo et al. 2006; Galametz et al. 2009; Zhu et al. 2009; Galametz et al. 2011). In the LMC, where Herschel brings 10 pc resolution at 500 μm, Galliano et al. (2011) have highlighted locations of submm excess ranging from 15% to 40% compared to that expected from dust SED models. The intensity of the excess is anticorrelated with the dust mass surface density. Possible explanations put forth to explain the submm excess include: (1) very cold dust component (e.g., Galliano et al. 2005; Galametz et al. 2011); (2) unusual dust emissivity properties (e.g., Lisenfeld et al. 2002; Meny et al. 2007; Galliano et al. 2011; Paradis et al. 2011); (3) anomalous spinning dust which has explained excess emission appearing at radio wavelengths (e.g., Draine & Lazarian 1998; Hoang et al. 2011; Ysard et al. 2011); and (4) magnetic dipole emission from magnetic grains (Draine & Hensley 2012). No single explanation put forth to date as the origin of the submm excess is completely satisfactory for all cases observed. Thus, the origin of the submm excess remains enigmatic. With more sensitive submm wavelength coverage, Herschel is confirming the flatter submm slope indicative of the submm excess present in some dwarf galaxies in the DGS (i.e., Rémy-Ruyer et al. 2013).

The global MIR SEDs of dwarf galaxies, often dominated by prominent ionized gas lines and weak PAH emission, resemble HII regions (e.g., Madden et al. 2006), or even, to some extent, ULIRGs (Fig. 1). Galaxy-wide, star forming dwarf galaxies often have a larger proportion of warm dust than spiral galaxies, with steeply rising MIR to FIR slopes and overall SEDs peaking at wavelengths shortward of ∼60 μm (Fig. 1) (e.g., Galametz et al. 2009, 2011). Starbursting dwarf galaxies have a higher relative abundance of small grains compared to our Galaxy (e.g., Lisenfeld et al. 2002; Galliano et al. 2003, 2005). In one of the most metal-poor galaxies detected to date, SBS 0335-052 (1/40 Z_⊙), the MIR to FIR SED is unlike that of any other galaxy observed so far, with silicate absorption bands superimposed on a featureless continuum peaking at an unusually short wavelengths, ∼30 μm (Thuan et al. 1999b; Houck et al. 2004). The smallest grains in galaxies (the most numerous in low metallicity galaxies being the very small grains, not the PAHs which are abundant in disk galaxies) are normally playing the most important role in gas heating, commonly via the photoelectric effect (Bakes & Tielens 1994). Thus, how the dust size distribution and abundance varies within galaxies is critical to understanding the heating and cooling processes. Given these observed dust properties, the heating and cooling processes will, undoubtedly, be altered in dust-poor galaxies relative to dust-rich galaxies. The structure of the ISM in terms of relative filling factors of star clusters, atomic, ionized, and molecular gas, looks to be different from those in our Galaxy or other metal-rich starburst galaxies: the ionized regions and the galaxy-wide hard radiation field play an important role in shaping the observed SED and the gas properties (Lebouteiller et al. 2012; Cormier et al. 2012). Simply assuming Galactic dust properties in the analyses of SEDs of low metallicity galaxies may lead to erroneous results in the determination of dust mass, extinction properties, gas mass, star formation properties, etc.

Zoom In Zoom Out Reset image size

The choice of dust composition to use when interpreting the observations complicates the problem. For example, assuming the often-used graphite dust analog, which, along with silicate and PAH components, well-describes the dust composition of the Galaxy (e.g., Zubko et al. 2004; Draine & Li 2007) can lead to inexplicably large dust masses in low metallicity galaxies such as the LMC, as shown by Galliano et al. (2011). While either amorphous carbon or graphite and silicate and PAHs in the SED model can successfully fit the observations, the resulting larger dust mass using graphite is sometimes difficult to reconcile with the metallicity, the observed gas mass, and the expected D/G (§ 2.4). Possessing higher emissivity properties, an amorphous carbon component (Zubko et al. 1996) can result in a lower dust mass that agrees better in some cases with the measured gas mass and the expected D/G. A recent reevaluation of the optical properties of (hydrogenated) amorphous carbon may lead to yet a different view of the variations in the observed dust properties in galaxies (Jones 2012a, 2012b, 2012c).

Many of the conclusions on the dust properties of low metallicity ISM rest on the need for higher spatial resolution and longer wavelength coverage into the submm, as provided by Herschel. While Draine et al. (2007) concluded that there is little effect overall on the dust masses of galaxies with or without submm observations, Galametz et al. (2011) found that including submm constraints in the SED modelling process can affect the dust masses estimated. To complicate interpretation, the single temperature modified black body, often used to describe the dust FIR-submm properties, can lead to different dust masses, for example, compared to more realistic dust SED models (e.g., Galliano et al. 2011; Dale et al. 2012; Galametz et al. 2012). This is a consequence of the fact that 70–500 μm emission from galaxies can originate from different heating sources and a single modified blackbody fitting these data can overestimate the average dust temperature and thus underestimate the dust mass within the galaxies (e.g., Bendo et al. 2010; Boquien et al. 2011; Bendo et al. 2012b; Dale et al. 2012). Herschel is well-positioned to lend clarity to these issues.

The Kennicutt-Schmidt law demonstrates the empirical relationship between total gas column density and star formation rate surface density. The observed slope of ∼1.4 indicates that the efficiency in forming stars increases with the gas surface density (Kennicutt 1998). Evidence is growing for molecular gas surface density, in particular, as the regulator for star formation in galaxies of the local Universe as well as at high-redshift, and not necessarily the total gas density (e.g., Kennicutt et al. 2007; Bigiel et al. 2008; Genzel et al. 2012). While mostly more-massive galaxies have been used to derive this relationship, does the same recipe hold for star formation in the wide range of low mass dwarf galaxies or, for that matter, high- z low metallicity dwarf galaxies?

We take for granted that CO ( J = 1 - 0) is the best tracer of molecular gas in local and distant galaxies, considering the observational difficulties in detecting directly the most abundant molecule, H₂. While HI halos are often very extended in dwarf galaxies, CO has been notoriously difficult to detect. Traversing from CO ( J = 1 - 0) observations to the total H₂ gas mass in dwarf galaxies is a perplexing issue, one that has been hampering studies of low metallicity ISM for decades. This CO ( J = 1 - 0) deficit is often difficult to reconcile with the molecular gas necessary to fuel star formation and a valiant effort has been made in determining a recipe for the CO ( J = 1 - 0)-to-H₂ conversion factor in low metallicity galaxies (e.g., Leroy et al. 2011; Schruba et al. 2012; Bigiel et al. 2011, and references within). On the other hand, surprisingly prominent FIR fine structure cooling lines, such as the 158 μm [CII] line, often interpreted as the best tracer of photodissociation regions (PDRs) around molecular clouds illuminated by massive stars, have been detected in low metallicity dwarf galaxies with the Kuiper Airborne Observatory (KAO) and the ISO LWS (Stacey et al. 1991; Poglitsch et al. 1995; Madden et al. 1997; Madden 2000; Bergvall et al. 2000; Hunter et al. 2001; Vermeij et al. 2002; Brauher et al. 2008; Israel & Maloney 2011). Thus, exceptionally bright 158 μm [CII] with exceptionally faint CO ( J = 1 - 0) renders the ISM of low metallicity galaxies and the conditions for star formation rather elusive.

L_[CII] has been hailed as an important tracer of star formation activity (Stacey et al. 1991; Boselli et al. 2002; de Looze et al. 2011). For example, values of L_[CII]/L_CO(J = 1 - 0) of more active normal starburst galaxies have ratios of ∼4,000 while those of more quiescent spiral galaxies are lower by a factor of 2 (Stacey et al. 1991). In dwarf galaxies, however, values of L_[CII]/L_CO(J = 1 - 0) can range from about 4,000 to as high as ∼80,000 (Madden 2000)—much higher than the most active metal-rich galaxies (Fig. 2). What is the explanation behind such strikingly high L_[CII]/L_CO(J=1-0)?

Zoom In Zoom Out Reset image size

Toward regions of the Local Group galaxy, IC10, for example, [CII] excitation from collisions with H atoms could not explain the surprisingly high observed [CII] intensity. Assuming the [CII] excitation originated in PDRs, comparison of the observed HI column density required an additional excitation source, namely H₂ molecules, to reconcile the cooling rate with the photoelectric UV heating rate. This led to the suggestion that up to 100 times more molecular hydrogen gas may be present compared to the H₂ inferred from the relatively faint CO ( J = 1 - 0) lines (Madden et al. 1997). These early KAO detections of the 158 μm [CII] line in low metallicity galaxies, namely the LMC (Poglitsch et al. 1995; Israel et al. 1996; Israel & Maloney 2011) and IC10 (Madden et al. 1997) were the first observational evidence for the presence of a reservoir of H₂ gas not traced by CO ( J = 1 - 0), but residing in the C⁺-emitting region—the "CO-dark" molecular gas.

The presence of the CO-dark molecular gas is regulated by conditions present in low metallicity environments. Molecules which are not effective at self-shielding, such as CO, will be photo-dissociated before H₂ in the PDRs. The combined effect of the strong FUV field from the nearby young star clusters and the decreased dust abundance is to shift the atomic-molecular gas transition deeper into the cloud, reducing the relative size of the CO core and increasing the volume of the PDR. In this scenario, a potentially significant reservoir of H₂ can be present in a C⁺-emitting PDR, outside the smaller CO core. This component, not traced by CO nor HI, has also been uncovered in our Galaxy via γ ray observations (Grenier et al. 2005) and more recently with Planck (Ade et al. 2011; Paradis et al. 2012). This effect of the CO-dark molecular gas in the PDR envelope is shown theoretically to be controlled strongly by A_V and should increase in importance as the metallicity decreases (Wolfire et al. 2010; Krumholz & Gnedin 2011; Glover & Clark 2012). More recent detailed studies comparing the dust and observed gas properties in the LMC and other nearby galaxies with Herschel and Planck have likewise highlighted the presence of this otherwise undetected gas phase (§ 2.4), which could also be in the form of optically thick HI.

In addition to the widely-used [CII] line, the DGS spectroscopy programme traces the other key FIR ISM cooling lines such as [OIII], [OI], [NIII] and [NII] (Table 1). These important cooling lines are diagnostics to probe the FUV flux, the gas density (n) and temperature (T), and the filling factor of the ionized gas and photodissociation regions (e.g., Wolfire et al. 1990; Kaufman et al. 2006). While the new NASA-DFG airborne facility, SOFIA, will continue observing the FIR fine-structure lines beyond Herschel's lifetime, it must contend with the residual effects of the Earth's atmosphere. Hence, Herschel allows us the best sensitivity and spatial resolution to probe a variety of emission lines from these intrinsically faint dwarf galaxies.

Since the ionization potential of C⁰ is 11.3 eV (Table 1)—less than that required to ionize atomic hydrogen—[CII] can be found in the ionized as well as the neutral phases. One way to determine how much of the observed [CII] may be excited by collisions with electrons in low density ionized regions, for example, is to use the [NII] lines at 122 and 205 μm (e.g., Oberst et al. 2011). Since N⁰ requires an ionization potential of 14.5 eV, these lines arise unambiguously from a completely ionized phase. Their relatively low critical densities ( n_crit) for collisions with electrons (48–200 cm^-3) make them useful to associate any observed [CII] emission possibly arising from the diffuse ionized gas, and thus provide a means to extract the PDR-only origin component of the [CII] emission. The [NII] 122 μm line is detected with PACS only in the brightest regions of the DGS galaxies. The PACS [NII] 205 μm line was not observable in the dwarf galaxies: leakage from other orders of the grating renders the flux calibration of the [NII] 205 μm line unreliable and the sensitivity of the instrument beyond 190 μm is degraded. The [NII] 205 μm line is, however, covered by the SPIRE FTS which has observed 4 DGS galaxies. The [OIII] 88 μm line has a similarly low n_crit for collisions with electrons (Table 1) and can also be associated with the low density ionized medium. It has the highest ionization energy of these FIR lines and traces the hard photons traversing the diffuse ionized medium. Thus, with Herschel, along with the existing ancillary data, we can more accurately model the diffuse and dense atomic, ionized, and molecular phases of the low-metallicity ISM.

The sensitivity provided by Herschel, along with the wide range of MIR lines that have been observed for the target galaxies with the Spitzer IRS spectrograph, provides ∼20 MIR and FIR fine structure lines in a wide variety of low metallicity galaxies in the DGS. Detailed studies of the photodissociated and photoionized gas can characterize the physical properties of the ISM as well as the properties of the energetic sources in low metallicity environments (see Cormier et al. [2012] for an example of such complex extragalactic modeling). Additionally, these observations will be able to shed light on quantifying the molecular gas reservoir in dwarf galaxies.

MIR H₂ lines, possible tracers or warm gas, can sometimes be present in Spitzer IRS spectra of dwarf galaxies, most often at low levels (Wu et al. 2006; Hunt et al. 2010), but are not normally thought to account for a substantial reservoir of the molecular gas in dwarf galaxies because most of this gas lies at lower T (< 100 K).

The SPIRE FTS, covering 194–670 μm, provides simultaneous coverage of CO rotational lines ranging from J = 4 - 3 through J = 13 - 12, a range within which the spectral line energy distribution (SLED) of CO lines has been found to peak for AGNs, star forming galaxies, and for high- z QSOs unveiling, often, a previously unknown warm molecular component (e.g. Panuzzo et al. 2010; Rangwala et al. 2011; Kamenetzky et al. 2012; Spinoglio et al. 2012). The warm molecular gas component of the relatively bright, well-studied nearby dwarf galaxies, NGC 4214, IC10, He2–10, and 30 Doradus, the most massive star-forming region in our neighboring LMC, are observed with the SPIRE FTS. 30 Doradus, He2–10, and IC10 have already been detected from the ground in CO transitions as high as J = 7 - 6 (Bayet et al. 2004, 2006; Pineda et al. 2012). While CO J = 1 - 0 can be excited at low densities (10² to 10³cm^-3) and low temperatures (∼10 K–50 K), higher- J CO lines effectively trace the denser and warmer gas connected to ongoing star-formation. The high J ground-based CO lines (up to J = 7 - 6) have been touted to be the best available SFR indicators accessible with ground-based instruments (Bayet et al. 2009). The FTS observations, aided by PDR and chemical models, will lead to the determination of the physical parameters (including density, temperature, filling factor and mass) of the CO-emitting molecular reservoir in low metallicity galaxies. This warm component is essential to helping provide a complete inventory of all of the gas in these star-forming low metallicity galaxies, much of it possibly residing in the warmest reservoirs more reliably traced by the high- J CO lines from the FTS observations.

Limits in sensitivity, spatial resolution, and wavelength coverage in the past have left an ambiguity in the behavior of the D/G at low metallicities (e.g., Galametz et al. 2011). The DGS is finally extending studies on D/G in galaxies to the lowest metallicities in the local Universe (Rémy-Ruyer et al. 2013).

While the proportion of heavy elements in the gas and dust scales with the metallicity in the disk of our Galaxy (Dwek 1998), it is not completely clear that this also holds for low metallicity dwarf galaxies. For a sample of low-metallicity galaxies, Lisenfeld & Ferrara (1998) and Issa et al. (1990) found that the dependence of metallicity on D/G can deviate from the expected linear relationship, which implies that the fraction of metals in the dust phase varies over the history of a galaxy. These studies, however, were performed for dust masses determined from IRAS observations, therefore not tracing cooler dust beyond 100 μm. James et al. (2002), who did a similar study using some ground-based 850 μm observations, include a few low metallicity galaxies in their large sample and find a flatter trend of the D/G at lower metallicities that needs to be tested for a larger range of metallicities, particularly for lower metallicities. In a study conducted before Herschel, Galametz et al. (2011) demonstrated that submm wavelength observations were indispensable to constraining the total dust masses, leaving some ambiguity in the measured D/G even without considering a submm excess. On the other hand, Draine et al. (2007) concluded that the D/G of the SINGS sample of galaxies was well-determined with Spitzer observations up to 160 μm with an additional cirrus component to explain the 850 μm observations of the metal-rich galaxies in the sample. More recently, Dale et al. (2012) results extending the SINGS sample to Herschel wavelengths with the KINGFISH survey (Kennicutt et al. 2011) note a submm excess present in their low Z galaxies. Herschel is beginning to uncover more examples of possible candidates to study the submm excess as shown, for example, in the dwarf galaxies in the Herschel Virgo Cluster Survey (Grossi et al. 2010). Likewise, a metallicity dependence of the SPIRE band ratios in the Herschel Reference Survey (Boselli et al. 2012) suggests the presence of a submm excess. More complete wavelength sampling by Herschel will help resolve the submm behavior and refine our extragalactic dust SED models.

Bot et al. (2004) modelled the dust in the diffuse ISM of the SMC ( Z = 1/10), and concluded that the D/G is 30 times lower than that of the Galaxy—3 times lower than the value a linear scaling of the metallicity would suggest. The SMC result supports the idea of depletion of heavy elements in the dust, such as occurs through supernova events. The large sample of DGS galaxies surveyed from MIR to the submm, having widely varying metallicities, allows us to rigorously test various chemical evolution models (e.g., Edmunds 2001; Dwek 1998; Galliano et al. 2008). Tighter constraints were put on the metallicity dependence of the D/G by Galametz et al. (2011) but the behavior below Z = 12 + log(O/H) ∼ 8 has been ambiguous due to lack of low metallicity observations. Investigating the behavior of D/G with metallicity in the DGS will help to decipher the origin of the dust, for example, from supernovae versus evolved low mass stars.

One popular way to get at possible CO-dark molecular gas is via the dust mass. Using the dust mass determined from observations, along with an assumed D/G (most often used as a linear or near-linear relation with metallicity), can give the total gas mass which can be compared to that observed in HI and CO. This has been shown to reveal gas reservoirs not emitting in CO or HI (e.g., Leroy et al. 2011, and references within). Measuring dust mass and using an expected D/G has also been used to determine the CO-to-H₂ conversion factor for high- z galaxies (Magdis et al. 2012; Magnelli et al. 2012). Such methods to get at any possible dark gas, however, require confidence in the dust mass determination as well as understanding metallicity variations and the D/G variations throughout a galaxy.

Possible uncertainties in the total dust mass, as well as the total molecular gas reservoir, again highlights the difficulty in understanding the ISM of low metallicity galaxies and the local conditions for star formation. With the new Herschel data and self-consistent modeling of the gas and dust together we can begin to unravel the mysteries of the ISM physics in dwarf galaxies.

As 2 of the 3 PACS bands (Poglitsch et al. 2010) can be observed simultaneously (either 70 μm and 160 μm or 100 μm and 160 μm), 2 sets of observations were required to obtain all of the 3 PACS bands. Mapping the largest galaxies of the Local Group with the 3 PACS bands required ∼10 hr of observing time per galaxy. For each band, we obtained two sets of maps with scanning angles separated by 90° to reduce the possibility of striping in the maps. All of the sources were observed in the PACS scan-map mode with map sizes varying from 4^' × 4^' to 30^' × 30^' using medium scan speed.

The PACS data were reduced using the Herschel Interactive Processing Environment (HIPE) version 7.0. The details of the data reduction and the various map reconstruction techniques and tests are described in Rémy-Ruyer et al. (2013).

Three different map-making methods, PhotProject, MADmap, and Scanamorphos (Roussel 2012), were performed on all of the DGS PACS data and resulting aperture photometry values were compared to determine the preferred method. The PACS bolometer thermal drifts, in the form of 1/f noise, leave imprints on the maps which manifest in large scale spatial structures. These are removed in PhotProject via high-pass filtering, having the tendency to suppress extended, large scale emission. The MAD map method creates maximum likelihood maps from the time ordered data, assuming the detector noise is Gaussian, which is not the case for the 1/f noise. Extended features tend to be reproduced somewhat better than in PhotProject maps. Scanamorphos takes advantage of the redundancy in the maps and accounts for the correlated and uncorrelated, nonthermal noise. From Level 1 data treatment in HIPE, the data were exported into Scanamorphos and maps reconstructed using the default options. After detailed tests comparing all methods (Rémy-Ruyer et al. 2013), PhotProject was in most cases found to be more well-adapted for point sources while Scanamorphos was chosen for the extended galaxies.

The FWHM beam sizes for PACS bands are 5.2'', 7.7'' and 12'' for the 70, 100 and 160 μm observations while final pixel sizes are 2'', 2'' and 4'', respectively. The sensitivities we achieve in the PACS observations are normally a 1σ value ranging ∼1.2–8.4 MJy sr^-1 for 70 and 100 μm and 0.6–5.6 MJy sr^-1 for 160 μm (Table 3). Calibration uncertainties are 3%, 3%, and 5% for the 70, 100 and 160 μm PACS bands. We obtain a signal-to-noise ratio (S/N) of at least 10 for the brightest galaxies or the brightest regions in galaxies, and 3–5σ detections typically for the fainter (lowest metallicity) galaxies.

The improvement in spatial resolution of Herschel over Spitzer can be immediately seen in Figure 5, presenting the nearby (5 Mpc) galaxy, NGC 1705, in the DGS (O'Halloran et al. 2010). Two clusters that are blended at FIR wavelengths with Spitzer are resolved with PACS on Herschel, allowing for detailed studies of the cooler dust properties for the first time.

Zoom In Zoom Out Reset image size

The PACS spectrometer footprint consists of 5 × 5 pixels of 9.4'' each, with a total field of view of 47'' × 47''. Each spatial pixel covers 16 spectral elements. In only a few cases were all of the brightest 7 FIR fine structure lines observed (Table 4): 158 μm [CII], 63 and 145 μm [OI], 88 μm [OIII], 57 μm [NIII], and the 122 and 205 μm [NII] lines. In all but a few of the faintest galaxies, at least the [CII] and [OIII] were observed (Fig. 4). The spectral resolutions for the [CII], [OIII] and [OI] lines are: ∼1250 (240 km s^-1), ∼2400 (125 km s^-1), and ∼3300 (90 km s^-1), respectively. The chop/nod line observing mode was used for the compact sources which fit within the spectrometer footprint. For sources too extended for chopping (maximum chop of ± 3^') we made raster maps using the unchopped grating scan mode, which replaced the decomissioned wavelength switching mode, used only for LMC/N11B (Lebouteiller et al. 2012). The data have been reduced to Level 1 within HIPE and then input into PACSman, an IDL package to handle line-fitting and map projection for PACS observations (Lebouteiller et al. 2012). Maps are projected on a subpixel grid with with each final pixel size about 3'' × 3'', roughly 1/3 of the original spatial pixel. Details of the PACS data processing of the survey will be presented in Cormier et al. (2013, in preparation).

The size of the spectroscopic maps vary depending on source and line (Table 4). With the FWHM of 9.5'' at 60–90 μm and 11.5'' at 150 μm, the spatial resolution of Herschel allows us to explore the morphology of the galaxies and determine the physical properties traced by the FIR cooling lines for the first time within dwarf galaxies, not just around the brightest star forming regions. We achieved a dynamic range in flux values greater than or equal to 10 in [CII], [OI] 63 μm, and [OIII] observations—in some cases as high as 100 in the brightest sources, thus providing the sensitivity to extract physical information from the more diffuse ISM. In some Local Group galaxies, the S/N can be as high as 100 for some lines in the extended regions. For the more compact sources (often more distant), the median S/N is around 13 for [CII], 18 for [OI] 63 μm, 38 for [OIII], 6 for [NII] 122 μm, and 15 for [OI] 145 μm. As an example of the kind of sensitivity brought to these studies by PACS, Madden et al. (1997) detected low surface brightness extended [CII] emission in IC 10 at the level of 1 × 10^-15 W m^-2 in a 55'' beam of FIFI on the KAO. If such emission were spread uniformly over the beam, the expected signal at the PACS spatial resolution would be ∼4 × 10^-17 W m^-2. In comparison, the sensitivity of PACS after single spectroscopic integration (effective on-source time of 150 s) for this line, in an equivalent KAO beam, is 2.9 × 10^-18 W m^-2, and thus was rapidly detected with PACS with a S/N ranging from 10 to 100 across IC10.

SPIRE (Griffin et al. 2010) observed 250, 350 and 500 μm bands simultaneously, with FWHM beam sizes of 18.1'', 25.2'' and 36.6'', respectively. The details of the data reduction are described in Rémy-Ruyer et al. (2013). Sixteen galaxies, including those with sizes larger than 1.5', were observed in the large scan-map mode, while the remaining galaxies were observed in small scan-map mode. Observations were performed as cross-scans, thus removing 1/f noise and at nominal speed (30'' s^-1). SPIRE data were processed mainly through HIPE, using the wavelet Deglitcher. The residual baseline subtraction applies the median of the bolometer timelines for the whole observation, not only a single scan leg (BriGadE method, Smith et al. in preparation). Calibration uncertainties are estimated to be ∼7% for all SPIRE bands. At 250 μm, 8 sources are detected and unresolved and 11 are not detected. At 500 μm, 5 sources are unresolved and 19 are not detected (Fig. 3). The 1σ sensitivities we achieved with SPIRE range from ∼0.4 –8.8 MJy sr^-1, 0.3–4.8 MJy sr^-1 and 0.1–1.8 MJy sr^-1, for 250, 350 and 500 μm, respectively (Table 3) and final pixel sizes are 6'', 8'' and 12'', respectively. Mapping the 11 larger galaxies takes about 1 hr at most per galaxy to reach the desired sensitivities with SPIRE while the small scan mode required about 10 minutes per galaxy.

The SPIRE Fourier Transform Spectrometer (FTS) covers 194–671 μm (447–1550 GHz), providing simultaneous coverage, for example, of CO rotational lines ranging from J = 4 - 3 through J = 13 - 12, as well as other molecules and atomic fines structure lines, such as [NII] 205 μm and [CI] lines at 492 and 809 GHz. The spectrometer array can be operated in three spectral resolutions, low (δν = 25 GHz), medium (δν = 7 GHz), and high (δν = 1.2 GHz), and is split into the Spectrometer Long Wave array (SLW: 303–671 μm) and the Spectrometer Short Wave array (SSW: 194–313 μm). The SLW and SSW consist of 7 and 17 unvignetted bolometers arranged in a hexagon pattern covering a field of view of about 3'' × 3''. The beam ranges from 17'' to 40'' across the FTS wavelength range. Depending on the image sampling mode, the bolometer arrays are moved around the requested pointing within the FOV in 1-point (sparse), 4-point (intermediate), or, 16-point (full) jiggles, resulting in a map with beam spacings of 50.5'', 25.3'', or 12.7'' for the SLW, and 32.5'', 16.3'', or 8.1'' for the SSW.

NGC 4214 and IC10 are observed with the FTS in high spectral resolution using the sparse sampling mode for a total of 2.7 and 3.7 hr, respectively. He2–10 and 30 Doradus, the most massive star forming region in our neighboring LMC, are observed in intermediate sampling mode for 4.6 and 2.8 hr, respectively, and also with high spectral resolution. The SPIRE FTS data is first processed using the extended source calibration of the Herschel Science Center pipeline which assumes the source is uniformly distributed within the FOV. Since this assumption is not necessarily accurate, we correct for the source distribution using the photometry maps and the method introduced in Wu et al. (2013, in preparation). Calibration uncertainties in the SPIRE FTS range from about 5% at the lower wavelength to ∼10% toward the long wavelength end.

The DGS galaxies have been observed with numerous ground-based and space-based observatories (Fig. 4). Having the full range of UV to radio observations available opens the door to some very powerful multiphase modeling of the dust and gas, and to relate this to a coherent picture of chemical evolution of galaxies. An important goal of the studies to come out of the DGS is to characterize the ISM as a function of metallicity and make the link to the ISM and star formation processes in truly primordial galaxies. With the vast multi-wavelength Herschel data set, along with the complementary data that exists or is being collected, the physical characteristics of the nebular gas, the photodissociation regions, the neutral gas and stellar properties can be characterized.

The first step in exploiting the Herschel observations has been the construction of the detailed database of existing observations from UV to radio wavelengths. In addition to the new PACS and SPIRE photometry and spectroscopy, the database includes the NIR to mm bands necessary to best characterize the dust properties using existing IRAS data, Spitzer MIR to FIR observations, as well as the critical ground-based submm observations: 850 μm, 870 μm and 1.2 mm data from SCUBA/JCMT or SCUBA-2/JCMT, LABOCA/APEX, MAMBO/IRAM and, complementing the longer submm to radio wavelengths, Planck data. Information on the atomic, ionized and molecular gas phases includes the Spitzer IRS spectroscopy and is being complemented by data from the literature. A campaign of follow up observations to complete the molecular, atomic and dust submm/mm database is currently underway (Cormier et al. 2013, in preparation; Rémy-Ruyer et al. 2013). To characterize the stellar activity, 2MASS and GALEX data have been collected for the database.

The ancillary data compiled for the DGS sample to date include the following:

• 1.  
  GALEX maps in FUV and NUV (1528 and 2271 Å) observations are collected from the STScI MAST archive (http://galex.stsci.edu/GR6/). These data give us a view into the star formation histories of the galaxies by capturing the direct UV emission from young stars to the most recent star formation activity (e.g., Hunter et al. 2010). The dust absorption properties in the UV provide additional constraints on the SED modeling of the dust emission properties in the MIR to FIR. GALEX observations exist for all but 9 of the DGS sample.
• 2.  
  2MASS observations have been compiled from IRSA (http://irsa.ipac.caltech.edu/) and exist for 34 galaxies in the DGS. The J, H and K bands from 2MASS provide direct access to the emission from the oldest stellar population.
• 3.  
  Spitzer has observed all 50 targets of the DGS with various combinations of IRAC bands at 3.6, 4.5, 5.8, and 8.0 μm (beam sizes : 1.7'', 1.7'', 1.9'', 2.0'') and MIPS bands at 24, 70, and 160 μm (beam sizes : 6'', 18'', 38''). The IRAC and MIPS micron data consist of a combination of archival data and the cycle 5 program: Dust Evolution in Low-Metallicity Environments (P.I. F. Galliano; ID: 50550). The MIPS data are reduced and compiled in Bendo et al. (2012b) and recently publicly released. Spitzer IRAC bands cover the emission from the MIR-emitting small grains and the PAHs, the most important species in galaxies for gas heating via the photoelectric effect (Bakes & Tielens 1994). The MIR and FIR range of the dust SEDs and the evolved stellar population are well constrained with the IRAC and MIPS bands. Furthermore, MIPS shares two bands with PACS at 70 and 160 μm. It is important to compare the two instruments and assure that they correspond and to understand cases of disagreement (Rémy-Ruyer et al. 2013). Spitzer IRS spectroscopy, covering 5–35 μm, exists for all but 5 of the DGS galaxies and have been recently reprocessed (Rémy-Ruyer et al. 2013). The IRS spectroscopy covers the PAHs bands as well as many ionized lines which are relatively bright in dwarf galaxies and are thus important in constraining the properties of the ionizing sources and the nebular properties (e.g., Cormier et al. 2012; Lebouteiller et al. 2012). For example, MIR nebular lines, such as [NeII], [NeIII], [SIV], and [SIII], accessible with the Spitzer IRS, have relatively high critical densities (>1000 cm^-3), complementing the PACS low density tracers, [NII], [NIII], and [OIII] lines (Table 1). Also, the 33 μm [SiII] line, which requires an ionization energy of 8.1 eV, complements the PACS neutral PDR gas tracers such as [CII] and [OI].
• 4.  
  The sensitivity of IRAS at 12, 25, 60 and 100 μm (beam sizes: 10.8^' × 4.5^' to 3.0^' × 5.0^') provides data for about one half of the DGS sample (data from the IRSA site). The 12 μm observations provide an additional valuable MIR constraint useful for modeling the global SEDs of the DGS galaxies.
• 5.  
  The DGS galaxies have also been detected by WISE at 3.4, 4.6, 12, and 22 μm with spatial resolutions of 6.1'', 6.4'', 6.5'', and 12.0'', respectively, for all but one galaxy (data from the IRSA site). The 22 μm observations, especially, will give the necessary constraint on the MIR wavelength range of the SED when neither IRAS nor Spitzer 24 μm are available.
• 6.  
  Submm observations from ground-based telescopes are crucial to determine the presence and characterization of the submm excess. Usually this excess may begin to appear around 450–500 μm, compelling observations longward of 500 μm. About 30% of the DGS sample have already been observed in the submm range with either the MAMBO 1.2 mm (IRAM—30 m telescope), LABOCA (APEX), SCUBA (JCMT) or SCUBA-2 (JCMT). New observations for DGS sources are currently underway with SCUBA-2 as well as with LABOCA on APEX to augment the submm observations to characterize and study the submm excess (Rémy-Ruyer et al. 2013).
• 7.  
  Planck has observed the sky with bands from 30 to 857 GHz. Due to the relatively large Planck beams (4.5'–33') and limited sensitivity, only a handful of galaxies are included in the Planck Early Release Catalog. Further Planck data processing in the next data-release in 2013 is expected to produce more results for the DGS targets.

A summary of the available ancillary data currently included in the DGS database can be found in Table 5. Figure 4 gives a view of range of existing Herschel and ancillary for the DGS sample, and how many galaxies were detected for the different data sets.

An example of the detailed FIR and submm dust distribution that Herschel reveals is shown in Figure 6 for NGC 4214, a nearby ( D = 2.9 Mpc) star-forming dwarf galaxy with metallicity 12 + log(O/H) = 8.2 (Kobulnicky & Skillman 1996). The PACS and SPIRE images from the DGS are the highest spatial resolution data at these wavelengths. The distribution of the dust traced by the 3 FIR PACS bands, where the peak of the total luminosity exists, resembles, at first glance, that of the longer wavelength SPIRE bands, which are mostly tracing the overall colder dust component. There are two particularly prominent star-forming regions seen at Spitzer and Herschel images: the more northern of these two prominent peaks is called central (C), while the peak just to the southeast of this is called SE. The quiescent extended "neck" region to the northwest is called NW. The C and SE sites containing clusters of widely varying ages and characteristics (Úeda et al. 2007a; Fanelli et al. 1997) are very prominent at all PACS bands and at least up to 350 μm and are sitting in a more extended, diffuse dust component. These 2 Herschel peaks are seen prominently in Hα (Fig. 6): the massive central HII region harboring a range of clusters overall ∼3 to 4 Myr (MacKenty et al. 2000) and also a super star cluster, the SE site being the second most prominent Hα region configured in a series of knots of some of the youngest clusters in the galaxy, ∼2 Myr (Úeda et al. 2007b). The NW region is characterized by a more quiescent cluster of Hα knots, also apparent in Herschel observations. The contrast between the bright star forming regions and the extended low surface brightness emission decreases toward longer wavelength, where, for example, at 500 μm the emission is rather dominated by the more extended cooler dust emission.

Zoom In Zoom Out Reset image size

At other wavelengths there are some obvious similarities to the Herschel data. In Figure 6 we include maps from the rich ancillary dataset available for NGC 4214: GALEX tracing the youngest stellar components, Hα showing the distribution of the ionized gas, IRAC 3.6 μm shows the evolved stellar component, IRAC 8 μm the PAHs and/or hot dust component, MIPS 24 μm (Bendo et al. 2012b) tracing the warm dust component, and 21 cm HI, from the THINGS survey (Walter et al. 2008). Note how extended the HI emission is, while the star formation activity is primarily concentrated toward the central 2 star forming regions (NGC4214-C and NGC4214-SE; see Fig. 6 caption).

Four PACS lines are very bright throughout NGC 4214: [OIII] 88 μm, [CII] 158 μm, and the [OI] 63 and 145 μm lines (Fig. 7). The [OIII] 88 μm line is the brightest everywhere, not the [CII] 158 μm line, which is most often the brightest line throughout normal metallicity galaxies. The very porous ISM in NGC 4214, typical of low metallicity star forming galaxies, is very evident in the extent of the wide-scale distribution of [OIII] 88 μm, requiring the presence of 35 eV photons, consistent with the deficit of PAHs found on galaxy-wide scales, presumably destroyed by the permeating unattenuated UV photons in dwarf galaxies (e.g., Engelbracht et al. 2005; Madden et al. 2006). The porosity of NGC 4214 and other dwarf galaxies is also shown by studies concluding the importance of the escaping ionizing radiation on the dust SED (Hermelo et al. 2013; Cormier et al. 2012). The [NII] 122 and 205 μm lines are also detected toward the peak star forming regions, but are one or two orders of magnitudes fainter than the other lines. All of the lines peak toward the two main star forming regions (Fig. 8): the [OIII] peak coincides with the location of the super star cluster in region C, while the [CII] peak is shifted off of the super star cluster and coincides with the peak of the neutral PDR tracers, the 2 [OI] lines. The S/N of the [CII] map is remarkable—more than 10 throughout the region mapped as illustrated in Figure 9, allowing for detailed spatial studies of the heating and cooling of the near and beyond the low metallicity star forming regions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To illustrate the kind of spectral and spatial studies that can be carried out on the gas and dust properties, we present the combined gas and dust SEDs of the three main HII regions, created from Spitzer IRAC, MIPS, IRS, and Herschel observations (Fig. 10). The data treatment and dust modelling details are described in Rémy-Ruyer et al. (2013). The shapes of the dust SEDs are representative of typical compact HII regions. These SEDs include the MIR and FIR lines from the PACS spectroscopic maps to present the full gas plus dust SEDs. Note the relatively wide diversity of behavior of the high excitation vs. low excitation lines and the variation in the line to FIR continuum levels. For example, the NW region shows a much more quiescent dust SED as well as low levels of high excitation ionic lines, which are in contrast to the those of the two bright star forming regions at the origin of the peaks of the warm dust emission (C and SE). Notice that toward the central and southeastern star forming regions, the PDR tracers, [CII] 158 μm and [OI] 63 μm lines, as well as the ionized gas tracer, [OIII] 88 μm, dominate the infrared line emission and account for 2–4% of the total L_TIR (Cormier et al. 2010). The rich MIR wavelength range shows prominent high excitation ionic lines, such as [NeII] 12.8 μm, [NeIII] 15.5 μm, [OIV] 25.9 μm and [SIV] 10.5 μm, arising from the rather dense HII regions as well as the PAH bands, emitting from 3 to 17 μm. Cormier et al. (2012) demonstrate the level of complexity possible now in modeling ∼20 MIR and FIR fine structure lines self-consistently with the dust continuum.

Zoom In Zoom Out Reset image size

In Figure 11 we present examples of modelled SEDs for the new Herschel data as well as with the ancillary dataset available for VIIZw403 and NGC 1569. From these modeled SEDs we will have access to new parameters for the galaxies, such as the total dust mass, total luminosity, new SFR and dust composition and dust size distribution (Rémy-Ruyer et al. 2013). Comparison of the observed SEDs with modeled SEDs in dwarf galaxies often points to an as-yet inexplicable excess (§ 2) which begins to become apparent near ∼500 μm, the longest wavelength observed by Herschel, or sometimes becomes more evident at longer submm wavelengths, in which case it may be detectable with ground-based submm telescopes, such as JCMT (SCUBA2), APEX (LABOCA), IRAM (GISMO) or ALMA, for example. VIIZw403 already shows a submm excess beginning at ∼500 μm (Fig. 11). In NGC 1569, on the other hand, there is no hint of submm excess throughout the Herschel SPIRE bands, including ∼500 μm. However, a small excess beginning at ∼SCUBA 850 μm observations is present, growing at mm wavelengths (Galliano et al. 2003). In this case, the submm excess would not be detectable without wavelengths beyond Herschel. The Herschel data are crucial to properly constraining the submm dust emission to be able to accurately extrapolate and interpret the SED to wavelengths >500 μm. Without the Herschel data, the Rayleigh-Jeans side of the dust SED is poorly constrained, making it more difficult to properly assess whether the submillimetre excess is present. Accurate quantification of the submm excess at SPIRE wavelengths must take into account careful consideration of the most recent calibration and beam sizes in the aperture photometry.

Zoom In Zoom Out Reset image size

Note that these are global SEDs, and spatial SED modelling throughout well-resolved galaxies may uncover regions where the presence of this submm excess may be apparent even at Herschel wavelengths, as has been detected with SPIRE within the LMC, where the presence of this excess is anticorrelated with the dust surface density (Galliano et al. 2011).