DUST IN THE DIFFUSE EMISSION OF THE GALACTIC PLANE: THE HERSCHEL/SPITZER SPECTRAL ENERGY DISTRIBUTION FITTING^*

The first Herschel (Pilbratt et al. 2010) images of the galactic plane obtained during the Science Demonstration Phase (SDP) reveal the far-infrared diffuse emission of the interstellar medium (ISM) with unprecedented beauty and detail. The Hi-Gal project (Molinari et al. 2010) provides an unbiased photometric survey of the inner Galactic plane emission between 70 and 500 μm where the spectral energy distribution (SED) is dominated by the largest dust grains (a ∼ 0.1 μm). Combined with the Spitzer GLIMPSE (Churchwell et al. 2009) and MIPSGAL (Carey et al. 2009) surveys that cover the dust SED from 3.6 to 24 μm, light mostly emitted by the smallest grains (a ≲10 nm), we will for the first time reliably sample the full dust SED of the diffuse emission at all spatial scales down to ∼40'' and over a large fraction of the Galactic plane.

At 70 and 160 μm, many of the observed structures show a close spatial correlation with some of the classical infrared tracers of small dust particles, like the emission measured at 8 and 24 μm, and therefore, one expects a partially similar origin and/or a similar dependence on ISM physical properties. To properly understand the origin of this emission is the first step of the data analysis.

Dust is heated by stellar ultraviolet (UV)–visible photons and re-radiates the absorbed energy in the infrared. Assuming some dust properties, this dust emission is widely used to trace crucial quantities like cloud masses or star-forming activity. However, to accurately derive these quantities requires taking into account the dust properties' evolution, how it affects its emission, and how it reflects the ISM properties. On the other hand, dust plays a crucial role for the ISM physics and chemistry. Since dust and the rest of ISM are tightly interlocked, changes on the dust properties affect those of the gas phase. It is then necessary to characterize the physical processes responsible for the dust evolution to assess their role within the ISM life cycle and to be able to use dust as a reliable tracer of the ISM. Accessing the full dust SED over a large fraction of the sky allows us to follow the behavior of every dust component over a broad range of physical conditions which is invaluable information for the study of dust evolution.

In this Letter, we present the first analysis of the Hi-Gal data combined with that of GLIMPSE and MIPSGAL to obtain the full dust SED from 8 to 500 μm. In Section 2, we describe the observations and the fitting method that makes use of a physical dust model, DustEM⁶ (Compiègne et al. 2010). In Section 3, we present and discuss the results and how our analysis method allows us to better interpret the behavior of extended emission observed with the Herschel/Spitzer photometers. We conclude in Section 4.

Two fields were observed (at l ∼ 30° and l ∼ 59°) during the SDP as part of the Hi-Gal program with PACS 70 and 160 μm and all of the SPIRE channels (Molinari et al. 2010). For this study, to satisfy the assumptions made in the model we describe below, we focus on a subfield of the l ∼ 59° field that shows no obvious H ii regions or young star cluster and that is relatively diffuse regarding the rest of the l ∼ 59° field and the l ∼ 30° field. We also avoid the higher latitudes (|b| ≳ 08) that display weak intensity where Herschel data could be less reliable at this early stage of data processing. The maps were obtained with the ROMAGAL pipeline (Traficante et al. 2010). For PACS 160 μm, SPIRE 250, 350, and 500 μm, we applied the calibration described in Bernard et al. (2010). The gain uncertainty for these data is taken to be 20%. The zero level was corrected by cross-calibration with Planck data with absolute uncertainties of ±19.8, 14.6, 7.5, and 3.0 MJy sr^-1 at 160, 250, 350, and 500 μm, respectively, which represent about half of the gain uncertainty for the faintest pixels of the studied region. We cross-calibrate the PACS 70 μm data with the MIPS 70 μm data from the MIPSGAL survey (R. Paladini et al. 2010, in preparation). We then apply the 15% gain uncertainty of MIPS 70 μm. We also use the IRAC 8 μm data from the GLIMPSE survey and the MIPS 24 μm data from the MIPSGAL survey. The gain uncertainty for these two data sets is taken to be 10%. Point sources are subtracted from the IRAC 8 μm data.

The zodiacal light contribution, important at shorter infrared wavelengths (Kelsall et al. 1998) has been removed from all the Spitzer data (on average, 1.1, 16.5, and 4.5 MJy sr⁻¹ at 8, 24, and 70 μm, respectively) and then from the PACS 70 μm following its cross-calibration on MIPS 70 μm. We do not perform any subtraction of the zodiacal emission at longer wavelengths where its contribution is negligible (<1 MJy sr⁻¹). We bring every map to the lowest resolution of the SPIRE 500 μm one (FWHM∼37''). We do this assuming a Gaussian point-spread function of appropriate width. We then project all maps into the SPIRE 500 μm grid (pixel field of view ∼115). Figure 1 displays the IRAC  8 μm, PACS 70 μm, and SPIRE 500 μm maps.

Zoom In Zoom Out Reset image size

We use the DustEM dust model described in Compiègne et al. (2010) to analyze the data. We consolidate the standard four grain populations into three: the polycyclic aromatic hydrocarbons (PAHs), small amorphous carbons representing the very small grains (VSGs), and we merged large amorphous carbons and silicates into a single big grains (BGs) population. Using the mpfit (Markwardt 2009) IDL minimization routine,⁷ we choose to adjust the following four parameters to fit the observed SED for each pixel: the (1) PAH and (2) VSG abundances relative to BGs, Y_PAH and Y_VSG, (3) the BG opacity, τ_BG, and (4) U_MMP83 a scaling factor of the "solar neighborhood" Mathis et al. (1983) (hereafter MMP83) exciting radiation field. Self-extinction along the line of sight can be significant at 8 μm and is accounted for assuming $I_\lambda \,=\,I_{0,\lambda }\,\frac{1-e^{-\tau _\lambda }}{\tau _\lambda }$, where I_0,λ is the integrated emissivity and τ_λ the total dust opacity that is computed from the dust model using Y_PAH, Y_VSG and τ_BG. Since in this first analysis we do not perform any separation of the components along the line of sight, the derived parameters result from the spatial mixing of different physical conditions. Fortunately, at l ∼ 59°, we expect only contributions from the Vulpecula star formation region (d ∼ 2 kpc) and the Perseus arm (d ∼ 8.5 kpc).

We focus on the behavior of the smallest particles (PAH and VSG) and hence, we do not assume any variation of the BG properties (i.e., τ_BG/N_H =  constant). In that case and to ease the following discussion, we can convert τ_BG into the hydrogen column density, N_H, and assume the relative abundances of PAHs and VSGs, Y_PAH and Y_VSG to be their abundance relative to hydrogen. Indeed, emissivity and/or abundance of the BG is known to evolve (e.g., Stepnik et al. 2003; Désert et al. 2008) and can be studied using the Herschel data (see Paradis et al. 2010) but to take such variations into account is beyond the scope of this Letter.

Figure 2 illustrates the obtained fitted spectra for two observed SEDs. The photometric points computed from a modeled spectrum take into account the color corrections. Schematically, the 160, 250, 350, and 500 μm photometric points give constraints on U_MMP83 and τ_BG (N_H) through the BG emission. The shape of the BG SED depends on U_MMP83 as this grain population is at thermal equilibrium, in contrast with the stochastically heated grains (PAHs and VSGs) whose SED shape is invariant regarding the radiation field intensity (see Draine & Li 2007; Compiègne et al. 2010). Hence, given U_MMP83 the absolute level of the BG SED constrains N_H. Finally, the 8 and 24 μm constrain the abundances of the two stochastically heated populations since the intensity of their emission scales as Y × N_H × U_MMP83.

Zoom In Zoom Out Reset image size

Figure 1 shows maps of the Y_PAH, Y_VSG, N_H, and U_MMP83 parameters and a map of the product Y_PAH × N_H × U_MMP83. χ²_reduced ≲ 2 for all pixels over these maps. N_H is given in unit of 10²⁰ H cm⁻² and U_MMP83 is dimensionless. Y_PAH and Y_VSG are given relative to the value for the diffuse high galactic latitude medium (DHGL; |b| ≳ 15°), M_PAH/M_H =  7.8 × 10⁻⁴ and M_VSG/M_H =  1.65 × 10⁻⁴ (see Compiègne et al. 2010). Note that M_BG/M_H =  9.25 × 10⁻³. The uncertainties on these parameters are given as computed in mpfit from the covariance matrix.

The column density lies between ∼1.5 × 10²² H cm⁻² for the most diffuse part and ∼1.1 × 10²³ H cm⁻² toward the dense filamentary structures, with a relative uncertainty of ∼18%. The U_MMP83 ranges between ∼0.5 and ∼1.5 with a relative uncertainty of ∼22%, which is consistent with the absence of young stellar cluster over our field. In agreement with the fact that the radiation field is shielded, we see that U_MMP83 decreases toward the highest column density regions. These U_MMP83 and N_H values are in agreement with previous estimates (see Bernard et al. 2010). Y_PAH goes from ∼1.0 to ∼2.0 with a relative uncertainty of ∼16%. The obtained Y_VSG spans a wider range than Y_PAH, going from ∼0.5 to ∼3.0 but with a relative uncertainty going from ∼40% for the highest values to ∼60% for the smallest values. The variations of these parameters over the map are therefore significant.

Figure 2 shows the averaged SED over the two boxes seen in Figure 1 and illustrate the wide Y_VSG variations. For the top (bottom) panel spectrum, we have N_H =  5.1 ± 0.9 × 10²² H cm⁻² (2.5 ± 0.5 × 10²² H cm⁻²), U_MMP83 =  1.2 ± 0.2 (0.9 ± 0.2), Y_PAH =  1.7 ± 0.2 (1.4 ± 0.2), and Y_VSG =  0.8 ± 0.5 (2.7 ± 1.0).

Y_PAH/Y_VSG seems to vary at large spatial scales (decreases from b ∼ 02 to b ∼ −07) and also at the edge of some dense filaments (near the "chimney" region around coordinate 58.4, +0.55). Both Y_PAH and Y_VSG decrease toward some of the densest filamentary structures but it appears not to be systematic and could be biased by the too simplistic assumption made on the self-absorption. A decrease of the smallest dust abundance toward dense regions regarding the biggest grains was reported by previous work and was interpreted as the coagulation of these smallest particles together with the bigger ones (e.g., Stepnik et al. 2003; Flagey et al. 2009). Previous studies also reported a lack of correlation or even an anti-correlation between the aromatic infrared bands and the mid-IR continuum emission (interpreted as the evolution of small dust properties) at the illuminated ridge of molecular clouds (e.g., Abergel et al. 2002; Berné et al. 2007; Compiègne et al. 2008), toward high Galactic cirrus at the interface between atomic and molecular material (Miville-Deschênes et al. 2002) or at Galactic scale in the Large Magellanic Cloud (Paradis et al. 2009).

The spatial structure of SPIRE 500 μm intensity, which is dominated by the BG contribution, is well correlated with N_H. Indeed, U_MMP83, and subsequently the BG temperature, are quite stable in our field so that the intensity variations at these wavelengths (Rayleigh tail of the blackbody-like emission) are dominated by the column density variations (assuming no BG emissivity variations). Comparing the 8 μm and Y_PAH maps, it is striking that the aromatic infrared band (AIB) intensity does not directly trace the PAH abundance. PAHs are stochastically heated and the observed AIB intensity (i.e., IRAC 8 μm) then scales linearly with the product Y_PAH × N_H × U_MMP83 as seen in Figure 1. The only difference is related to the extinction along the line of sight that can be seen in the IRAC 8 μm (as dark filamentary structures correlated with the N_H map) and not in the Y_PAH × N_H × U_MMP83 map. The PACS 70 μm image shows a close spatial correlation with the classical infrared tracers of small dust particles, like the emission measured at 8 μm (see Figure 1). The emission at 70 μm is not due to a single dust component. As seen in Figure 2, in the framework of our model both the VSGs that are stochastically heated and the BGs that are at thermal equilibrium contribute to the emission at this wavelength. For U_MMP83 ≲ 100, the BG emission at 70 μm fall in the Wien part of the blackbody like emission that makes it more sensitive to U_MMP83 than at 500 μm. Therefore, the two emission components (from VSGs and BGs) at 70 μm are sensitive to both N_H and U_MMP83 which better explains correlation of this map with the 8 μm map (also sensitive to both N_H and U_MMP83) as opposed to the 500 μm map (more sensitive to N_H).

The model allows us to compute the relative contribution of the three dust populations in the different photometric bands. The VSG contribution to PACS 70 μm increases if the VSG abundance relative to BG increases and/or if U_MMP83 decreases (shifting the BG emission toward longer wavelengths). In the studied field, the VSG contribution to the PACS 70 μm intensity goes from ∼10% up to ∼50% with a median value of ∼27%. The maximum contribution of VSGs to PACS 100 μm and PACS 160 μm is ∼17% and ∼7% (the median is 9% and 3%), respectively. For the top (bottom) panel spectrum of Figure 2, the contribution is 12%, 4%, 2% (35%, 13%, 5%) for PACS 70, 100 and 160 μm, respectively. This result strongly suggests that the proper analysis of Herschel spectrum of diffuse emission including 70 μm may require to account for stochastically heated grains.

We have presented the first analysis of the diffuse emission of the Galactic plane as observed by Herschel combining the Hi-Gal data with the GLIMPSE/MIPSGAL Spitzer data. Toward a subfield of the l ∼59° Hi-Gal SDP field, we performed a pixel-by-pixel fitting of the full dust SED between 8 and 500 μm using a physical dust emission model, DustEM (Compiègne et al. 2010).

Assuming that the BG properties remain constant, the unique wavelength coverage provided by the Spitzer and Herschel photometric observations allows us to derive the following parameters for our dust model: the PAH and VSG abundances, Y_PAH and Y_VSG, the column density, N_H, and the intensity of the exciting radiation field, U_MMP83. To our knowledge, this is the first time PAH and VSG abundance maps are derived using such a pixel-by-pixel SED fitting at resolution <1' and over such an extended field. These abundances, as well as N_H and U_MMP83, vary significantly over the field. As already reported by previous studies, Y_PAH and Y_VSG appear not to be positively correlated.

Although it was already theoretically known, our analysis method provide a firm demonstration that IRAC 8 μm does not trace the PAH abundance but the product Y_PAH × N_H × U_MMP83. We also showed that at 70 μm, the modeled emission is due to both the BGs and the VSGs. At these wavelengths, the BG emission is sensitive to both the intensity of the exciting radiation field and the column density is likewise sensitive to the VSG, explaining the similar spatial structure seen in PACS 70 μm maps regarding shorter wavelengths. Using our model, we derived the VSG contribution to the PACS channels that can be up to ∼50%, ∼17%, and ∼7% at 70, 100, and 160 μm, respectively. We conclude that the interpretation of Herschel spectrum of the diffuse emission down to 70 μm may require us to take into account the stochastically heated population (VSG).

Finally, our analysis allows for a better understanding of the first Herschel images of the Galactic plane diffuse emission by disentangling the different dust population contributions and also revealing the great potential of the Herschel/Spitzer synergy combined with a physical dust model for the study of dust evolution (see also Abergel et al. 2010).