The Polycyclic Aromatic Hydrocarbon Mass Fraction on a 10 pc Scale in the Magellanic Clouds

We combine observations from the Spitzer Space Telescope (Werner et al. 2004) and the Herschel Space Observatory (Pilbratt et al. 2010) to cover the MIR through FIR spectral energy distribution (SED). In the MIR, we use observations from the Spitzer Legacy program SAGE (Surveying the Agents of the Galaxy), which observed both the LMC (SAGE-LMC; Meixner et al. 2006) and the SMC (SAGE-SMC; Gordon et al. 2011). The final images produced by SAGE-SMC include deeper measurements in the main star-forming regions of the SMC, from the S³MC program (Bolatto et al. 2007). In the FIR and submillimeter range, the Herschel Key Project HERITAGE (the Herschel Inventory of the Agents of Galaxy Evolution; Meixner et al. 2013, 2015) surveyed both clouds. This leads to a total of 11 photometric bands included in our study, at 3.6, 4.5, 5.8, 8.0 (Spitzer, IRAC; Fazio et al. 2004), 24, 70 (Spitzer, MIPS; Rieke et al. 2004), 100, 160¹¹ (Herschel, PACS; Poglitsch et al. 2010), 250, 350, and 500 μm (Herschel, SPIRE; Griffin et al. 2010).

The goal of this study is to model dust emission from the MIR to the submillimeter range. However, there is a significant contribution from stars in the Spitzer IRAC bands. For the average stellar populations in a pixel, we can account for stellar emission with a simple assumption of a 5000 K blackbody (Draine et al. 2007, and see Section 3), but due to the proximity of the MCs, occasionally a resolution element is dominated by a very bright source that is not well modeled by a blackbody. In particular, this can be young stellar objects or evolved stars (Woods et al. 2011; Jones et al. 2017). To avoid contamination by these sources, we mask out the bright point sources that show up in the short-wavelength bands. We perform our own masking of the brightest sources in these bands in two steps. A first set of sources is simply chosen by looking at the images, and selecting the brightest stars. Second, to remove as much of the contaminating point sources as possible, we perform a fit with the unmasked images. Using these fit results, we mask stars where there is an evident bias in the parameter maps due to a point source in the image. We mask all the selected sources by replacing the value within a small radius from the source with the averaged value of the local diffuse emission.

Prior to fitting we perform several additional steps of image processing. We correct the IRAC and SPIRE images with extended source factors, as suggested by the IRAC Instrument Handbook¹² and the KINGFISH User Guide.¹³ Then, all observations are convolved to the SPIRE 500 resolution (∼36'') using the Aniano et al. (2011) convolution kernels. Even though the clouds are at relatively high Galactic latitude, the observations still suffer from a non-negligible emission from the MW cirrus. We follow Gordon et al. (2014) and Chastenet et al. (2017) to remove this foreground emission. We convert the foreground H i MW column density map to a dust column density using coefficients 1.47 × 10⁻³, 9.32 × 10⁻⁴, 1.34 × 10⁻², 4.28 × 10⁻², 2.53 × 10⁻², 2.63 × 10⁻¹, 1.36, 1.07, 1.85, 1.20, 0.62, and 0.25 MJy sr⁻¹ (10²⁰ H i cm^–2)⁻¹ from 3.6 to 500 μm respectively, and subtract the foreground cirrus. We then perform a background subtraction to get rid of residual emission from the cosmic IR background, zodiacal light, and mosaicing offsets. Background regions are selected by eye: we visually identify portions of the images where we can avoid contamination from dust emission from the target galaxies. In the LMC, these are chosen to be at the edges of the images. In the SMC, we avoid the SMC bar and wing to select the background pixels. We fit and subtract a tilted plane, which removes the gradient across the background. All images are then projected with the final pixel grid sampling the point-spread function with approximately independent pixels, that is l_pixel ∼ 42'', which corresponds to a pixel size of ∼12 pc in the SMC and ∼10 pc in the LMC. After the final projection, a background covariance matrix ${{ \mathcal C }}_{\mathrm{bkg}}$ is constructed from the background pixels, to quantify the correlations between noise in different photometric bands (Gordon et al. 2014).

In this study, we are interested in possible correlations between the fitted dust properties and other components of the ISM. We use the SHASSA survey (the Southern H-Alpha Sky Survey Atlas; Gaustad et al. 2001) to investigate the spatial distribution of the ionized gas from Hα emission.¹⁴ We use the smoothed maps of the LMC (field 013), and SMC (field 510), at 4' angular resolution. The maps are projected onto the final pixel grid of our data set, thus oversampling the point-spread function of the Hα emission data.

We use ¹²CO(J = 1 − 0) maps from the NANTEN (Fukui et al. 1991) survey of the SMC (Mizuno et al. 2001) and the LMC (Fukui et al. 2008) to trace the spatial distribution of the molecular gas (∼3' resolution). In Section 5.2, we will use a 3σ detection threshold in CO integrated intensity to define the "molecular gas phase." We determine this threshold by computing the standard deviation σ in a region where we do not find any detections by eye. We will consider the molecular gas phase as every pixel above 3σ. In the SMC, we find a 3σ value of 0.3 K km s⁻¹, and 0.75 K km s⁻¹ in the LMC. Assuming ${\alpha }_{\mathrm{CO}}^{\mathrm{SMC}}=76\,{M}_{\odot }\,{\mathrm{pc}}^{-2}\ {({\rm{K}}\mathrm{km}{{\rm{s}}}^{-1})}^{-1}$, and ${\alpha }_{\mathrm{CO}}^{\mathrm{LMC}}=10\,{M}_{\odot }\,{\mathrm{pc}}^{-2}\ {({\rm{K}}\mathrm{km}{{\rm{s}}}^{-1})}^{-1}$ (Jameson et al. 2016), this corresponds to a molecular gas surface density of 22.8 M_⊙ pc⁻² in the SMC, and 7.5 M_⊙ pc⁻² in the LMC.

In Figure 1, we show two examples of the data in the SMC and the LMC fit by the best model (i.e., maximum likelihood), in two pixels of the diffuse ISM. The error bars at short wavelengths show that the errors are small enough to strongly constrain ${q}_{{\rm{PAH}}}$. Residuals show that the fits are good in both galaxies at short wavelengths, especially 8 μm. At long wavelengths, the fractional residuals are mostly negative, indicating that the model generally overestimates the data (residuals peak at less than 10%).

Zoom In Zoom Out Reset image size

The fitting is done with the DustBFF tool (Gordon et al. 2014), which determines the n-dimensional a posteriori likelihood distribution of the parameters. This is a Bayesian fitting tool that uses flat priors on the parameter distributions, and a covariance matrix, built from background pixels and instrument errors, to propagate uncertainties. We use the same calibration errors on Spitzer and Herschel instruments as those in Gordon et al. (2014) and Chastenet et al. (2017). The diagonal elements of the covariance matrix include both uncorrelated (or statistical) and correlated errors, and the non-diagonal elements measure only correlated errors between the photometric bands. We refer the reader to Gordon et al. (2014) for further details on the covariance matrix and the DustBFF fitting.

We use the five-dimensional (for five fitting parameters) likelihood function to build realizations of the parameter maps: a realization is a sample of the five-dimensional grid, weighted by the likelihood function. The realizations render noise properties more accurately than finding only the maximum of the likelihood function. To measure values like $\langle {q}_{{\rm{PAH}}}\rangle$ and their associated 16th and 84th percentiles, we prefer the realization method over the expectation value (Gordon et al. 2014; Chiang et al. 2018; Utomo et al. 2019), since it draws samples without marginalizing the likelihood over all but one parameter dimension. Since we compute weighted values with both ${q}_{{\rm{PAH}}}$ and Σ_d, it is better to use the full-dimension likelihood distribution to avoid losing information. We build a large number of realization maps for each parameter simultaneously (both the fitted parameters and the calculated parameters like $\overline{U}$ and ${f}_{\mathrm{PDR}}$), and use them to calculate the mean in each pixel. These are presented in Figures 2 and 3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

From the realizations of a single pixel or group of pixels, we can determine their statistical error by measuring the standard deviation from a large number of realizations. When combining many pixels together we can propagate these uncertainties to calculate the error on the mean values. To represent the intrinsic scatter of a parameter within a specific region of the galaxy (e.g., Figures 7 or 8), we use the 16th and 84th percentiles of the dust mass- (or dust luminosity for ${f}_{\mathrm{PDR}}$)-weighted distribution of said parameter. For all averaged parameters, e.g., $\langle {q}_{{\rm{PAH}}}\rangle$, quoted in the rest of the paper, the statistical uncertainties are generally very small, due to the large numbers of pixels begin averaged together. We instead list the ± the 16th and 84th percentiles of the dust mass- or luminosity-weighted distributions.

Figures 2 and 3 show the results from fitting the pixels of the SMC and the LMC maps, respectively, as well as the computed values of $\overline{U}$ and ${f}_{\mathrm{PDR}}$, as described in Section 3. The overall distribution of Σ_d shifts toward higher values in the LMC than in the SMC, as expected because of the larger dust-to-gas ratio of the former (Gordon et al. 2014). A significant difference is also seen in the ${q}_{{\rm{PAH}}}$ parameter, which is clearly higher in the LMC than in the SMC (a larger version of the ${q}_{{\rm{PAH}}}$ map is shown in Figure 4). We find dust-mass-weighted averages for each galaxy of

$$\begin{eqnarray*}\begin{array}{rcl}\langle {q}_{\mathrm{PAH}}^{\mathrm{SMC}}\rangle & = & {1.0}_{-0.3}^{+0.3} \% ,\ \mathrm{and}\\ \langle {q}_{\mathrm{PAH}}^{\mathrm{LMC}}\rangle & = & {3.3}_{-1.3}^{+1.4} \% .\end{array}\end{eqnarray*}$$

Both of these values fall within the range seen by Draine et al. (2007) in the SINGS sample at the relevant metallicities. By studying a sample of low-metallicity galaxies, Rémy-Ruyer et al. (2015) found a power-law relation between the metallicity and the PAH fraction (see their Equation (5)). If we apply this relation to the MCs' metallicities (SMC: 12 + log(O/H) ∼ 8.1; LMC: 12 + log(O/H) ∼ 8.3; Russell & Dopita 1992), we find

$$\begin{eqnarray*}\begin{array}{rcl}0.69 \% & \leqslant & {q}_{\mathrm{PAH}}^{\mathrm{RR}15}(\mathrm{SMC})\leqslant 3.47 \% ,\,\mathrm{and}\\ 1.25 \% & \leqslant & {q}_{\mathrm{PAH}}^{\mathrm{RR}15}(\mathrm{LMC})\leqslant 6.31 \% .\end{array}\end{eqnarray*}$$

Our average PAH fractions found in this work fall well within these ranges as well.

Zoom In Zoom Out Reset image size

In both galaxies over most of the area, ${U}_{\min }$ ∼ $\overline{U}$. Thus we only show the fitting parameter ${U}_{\min }$ in Figures 2 and 3. We calculate the dust-mass-weighted average of $\overline{U}$ and find $\langle \overline{U}\rangle$ = ${1.2}_{-0.4}^{+0.4}$ in the SMC and $\langle \overline{U}\rangle$ = ${1.6}_{-1.0}^{+1.6}$ in the LMC. Utomo et al. (2019) recently studied the distributions of mass and temperature in four nearby galaxies, including the SMC and the LMC, using a single-temperature modified blackbody model. They found that the distribution of dust mass as a function of radiation field intensity peaks at values of U_peak = 1.1 in the SMC and U_peak = 1.8 in the LMC, consistent with our results.

Sandstrom et al. (2010) found a dust-mass-weighted PAH fraction of ∼0.6% in the SMC (compared to the 1.0% in this study). One possible explanation for our lower fraction is the limited coverage of the dust emission SED in Sandstrom et al. (2010) compared to this paper: here, the longest wavelength is 500 μm while it is 160 μm in Sandstrom et al. With the addition of the Herschel bands, we are able to better constrain the total dust mass and temperature, particularly in regions with colder dust, affecting the PAH fraction. In addition, the extent of the S³MC Spitzer maps used in that paper did not allow as accurate a MW foreground removal as enabled by the expanded SAGE-SMC coverage. This may have resulted in an oversubtraction of actual SMC emission in the MIR bands, decreasing the Sandstrom et al. (2010) ${q}_{{\rm{PAH}}}$ value.

In the LMC, Paradis et al. (2009) found an enhanced PAH fraction with respect to the large grain abundance in the stellar bar. Our results do not show an increased ${q}_{{\rm{PAH}}}$ in this region, as other regions of the LMC show the same PAH fraction as in the optical bar. This difference could be the result of including the Herschel bands, and thereby more accurately measuring ${q}_{{\rm{PAH}}}$. It may also be the result of using a different dust model (e.g., the Desert et al. 1990 model in the paper by Paradis et al. 2009), or a different stellar continuum model.

We find values of the global PDR fraction (i.e., the fraction of the dust luminosity produced by regions where U ≥ 10²):

$$\begin{eqnarray*}\begin{array}{rcl}\langle {f}_{\mathrm{PDR}}^{\mathrm{SMC}}\rangle & = & {3.29}_{-0.02}^{+0.01} \% ,\,\mathrm{and}\\ \langle {f}_{\mathrm{PDR}}^{\mathrm{LMC}}\rangle & = & {9.17}_{-0.04}^{+0.04} \% .\end{array}\end{eqnarray*}$$

These values are below the average values found in NGC 628 and NGC 6946, two nearby, resolved galaxies, by Aniano et al. (2012; ∼11%). However, this is not surprising as the spatial scales in their study are coarser than for the MCs: due to limited resolution, the luminosity-weighted ${f}_{\mathrm{PDR}}$ is biased toward high-luminosity values. The resulting weighted average is then more sensitive to these high-luminosity regions due to the blending of signal. This is one of the key results of the recent study by Utomo et al. (2019). We do find high ${f}_{\mathrm{PDR}}$ values in star-forming regions, up to ∼60% in 30 Dor (LMC), and ∼50% in N66 (SMC).

The spatial distributions of Ω_* trace the regions with high stellar density (Zaritsky et al. 2002, 2004). They show the old stellar spheroid population of the SMC, as well as the optical bar in the LMC.

In Figure 4, we show the maps of the ${q}_{{\rm{PAH}}}$ parameter in the LMC (top) and the SMC (bottom). The contours are Hα emission from the SHASSA survey (Gaustad et al. 2001) at level of ∼1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻² (∼300 dR; solid line; the reason for choosing this value is discussed later in this section). The labeled circles are H ii regions as identified in Lopez et al. (2014); the radii correspond to those given in their Table 1, and are defined as the limit where they enclose 90% of the Hα emission of the source. There are other known H ii regions in the MCs; however, identification of H ii regions and their boundaries is not straightforward, and we use only those in Lopez et al. (2014) for the sake of homogeneity. From these maps we draw several conclusions which we discuss in the following (sub)sections: (1) ${q}_{{\rm{PAH}}}$ varies dramatically between the SMC and LMC, (2) ${q}_{{\rm{PAH}}}$ shows variations within each galaxy, and (3) a primary driver for the variation in ${q}_{{\rm{PAH}}}$ within each galaxy appears to be correlated with the presence of ionized gas as traced by Hα.

To investigate the variations of ${q}_{{\rm{PAH}}}$ within and between the galaxies, we make four separations with respect to the ISM gas phase. In Table 2, we report $\langle {q}_{{\rm{PAH}}}\rangle$ in these four phases. There is no overlap between these values as we use only pixels that do not fall in two gas phase definitions. In short, we observe the following:

$$\begin{eqnarray}&&{q}_{\mathrm{PAH}}^{\mathrm{mol}}\sim {q}_{\mathrm{PAH}}^{\mathrm{diff}.\,\mathrm{neutr}.}\gt {q}_{\mathrm{PAH}}^{\mathrm{out}-{\rm{H}}\ {\rm{II}}}\gt {q}_{\mathrm{PAH}}^{{\rm{H}}\ {\rm{II}}}.\end{eqnarray} \tag{ 7 }$$

The implications of these observations for the PAH life-cycle will be discussed further in Section 5.2. We define the four phases as follows.

• 1.  
  Ionized gas toward H ii regions. This is simply defined by the pixels falling within the radii of H ii regions from Lopez et al. (2014). In the LMC, there is a clear drop in the dust-mass-weighted PAH fraction, and $\langle {q}_{\mathrm{PAH}}^{{\rm{H}}\,{\rm{II}}}\rangle$ reaches only $\sim {1.8}_{-1.3}^{+1.1} \%$, i.e., slightly more than half of the galaxy average. In the SMC, the dust-mass-weighted PAH fraction reaches $\sim {0.8}_{-0.5}^{+0.3} \%$. We note that the harder radiation field in and near H ii regions, which is not taken into account in our fitting, would lead us to overestimate ${q}_{{\rm{PAH}}}$ in these regions. The values found here are therefore conservative.
• 2.  
  Non-H ii region ionized gas. We distinguish the ionized gas inside and outside H ii regions, by selecting pixels whose Hα surface brightness is above I_Hα ∼ 1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻² but not in identified H ii regions. Although this ionized gas is in a more diffuse phase than the gas in H ii regions, we avoid identifying it as "diffuse ionized gas (DIG)" due to the specific ways that DIG is defined in nearby galaxies (for reviews, see Mathis 2000; Haffner et al. 2009, and references therein). We discuss this further in Section 5.3. We find $\langle {q}_{\mathrm{PAH}}^{\mathrm{out}-{\rm{H}}\,{\rm{II}}}\rangle ={2.9}_{-1.2}^{+1.1} \%$ in the LMC and $\langle {q}_{\mathrm{PAH}}^{\mathrm{out}-{\rm{H}}\,{\rm{II}}}\rangle ={0.9}_{-0.3}^{+0.3} \%$ in the SMC.
• 3.  
  Molecular gas. We use ¹²CO $(J=1-0)$ maps (Section 2) to trace the molecular gas. We define this phase with every pixel above the 3σ detection threshold, and I_Hα ≲ 1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻² . We find $\langle {q}_{\mathrm{PAH}}^{\mathrm{mol}.}\rangle \sim {4.3}_{-0.9}^{+1.3} \%$ in the LMC, and $\langle {q}_{\mathrm{PAH}}^{\mathrm{mol}.}\rangle \sim {1.1}_{-0.2}^{+0.1} \%$ in the SMC, similar to the values in the diffuse neutral medium.
• 4.  
  Diffuse neutral gas. This is defined with the regions that fall into none of the above categories: with Hα emission lower than ∼1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻², which also means outside of an H ii region, and below the 3σ CO detection. We use the dust-mass-weighted PAH fraction in the diffuse neutral medium as a reference value for each cloud. We find $\langle {q}_{\mathrm{PAH}}^{\mathrm{ref}}\rangle ={1.1}_{-0.3}^{+0.2} \%$ in the SMC, and $\langle {q}_{\mathrm{PAH}}^{\mathrm{ref}}\rangle ={4.1}_{-0.8}^{+0.6} \%$ in the LMC.

Table 2.  $\langle {q}_{{\rm{PAH}}}\rangle$ (in %) in Different Gas Phases of Each Galaxy

	SMC	LMC
Global Average	${1.0}_{-0.3}^{+0.3}$	${3.3}_{-1.3}^{+1.4}$
H ii regions	${0.8}_{-0.5}^{+0.3}$	${1.8}_{-1.3}^{+1.1}$
Non-H ii region ionized gasa	${0.9}_{-0.3}^{+0.3}$	${2.9}_{-1.2}^{+1.1}$
Diffuse neutralb	${1.1}_{-0.3}^{+0.2}$	${4.1}_{-0.8}^{+0.6}$
Molecular gasc	${1.1}_{-0.2}^{+0.1}$	${4.3}_{-0.9}^{+1.3}$

Notes.

^aAll pixels above the lower limit in Hα, excluding H ii regions. ^bAll pixels below the CO detection, the limit in Hα, and outside of H ii regions. ^cAll pixels with CO detection not overlapping with pixels with I_Hα ≳ 1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻².

Download table as:  ASCIITypeset image

The H ii regions, as identified in Lopez et al. (2014), appear as minima in the ${q}_{{\rm{PAH}}}$ map, and suggest that PAHs are destroyed inside H ii regions. In Figure 4, we can see that most H ii regions, marked by the black labeled circles, are indeed low in PAHs, with respect to the abundance in the other parts of each galaxy.

We are interested in understanding how H ii regions affect the PAH fraction. In Figure 5, we show the radial profiles of $\langle {q}_{{\rm{PAH}}}\rangle$ for each H ii region in the SMC (bottom) and the LMC (top). We see that $\langle {q}_{{\rm{PAH}}}\rangle$ drops to very low values only in a handful of cases (notably, 0% in 30 Dor). This is expected, as the diffuse neutral gas projected along the line of sight contaminates our measurement of the ${q}_{{\rm{PAH}}}$ inside the H ii region. The dust emission along the line of sight reflects the full column through the ISM of the galaxy, not just the H ii region, so we do not expect the observed ${q}_{{\rm{PAH}}}$ to be 0%. In the particular case of 30 Dor, a result consistent with 0% PAH fraction could mean that the actual H ii region dominates the entire line of sight through the LMC.

Zoom In Zoom Out Reset image size

For each H ii region, we measure the radius at which $\langle {q}_{{\rm{PAH}}}\rangle$ goes back to the global average (horizontal gray line in Figure 5). In this figure, we see that $\langle {q}_{{\rm{PAH}}}\rangle$ in each H ii region goes back to the galaxy average at different radii. We treat these radii as the "spheres of influence" of each H ii region on the surrounding PAH fraction. In Figure 6, we plot these radii against the total Hα luminosity from within that radius. We find that these radii correlate very well with Hα luminosity with a power-law coefficient between 0.35 and 0.4 in both galaxies. If the Hα surface brightness were constant, growing the radius would lead to a dependence of L_Hα on r². We see in Figure 6 that is not the case. It rather suggests that the sphere of influence on the PAH fraction of the ionizing stars within an H ii region scales as one would expect for the Strömgren sphere (Strömgren 1939). In this particular case, assuming a constant gas density the radius of the Strömgren sphere grows with the ionizing photon production rate (as does the Hα luminosity) with a power-law coefficient of one-third. If PAHs are destroyed in ionized gas by photodestruction or sputtering, then one would expect the region with a deficit of PAHs relative to the galaxy average to grow as the size of the Strömgren sphere. In Binder & Povich (2018), the authors found a relation between the population of stars inside star-forming regions of the MW, and the PAH fraction. They show that a single O6 star is less effective at destroying PAHs than a population of stars that extends to O2/O3, and Wolf-Rayet stars. Glatzle et al. (2019) showed that the growth of H ii regions is related to the dust content within, including PAH abundance, by impacting the ionization fronts. A more detailed study of H ii regions would be interesting to possibly link the initial $\langle {q}_{{\rm{PAH}}}\rangle$ value at small radius in Figure 5, the properties of the ionizing star(s) within, and the expansion of H ii regions.

Zoom In Zoom Out Reset image size

Even outside of the H ii regions, we find a clear relation between increasing Hα emission and decreasing $\langle {q}_{{\rm{PAH}}}\rangle$. Figure 7 shows $\langle {q}_{{\rm{PAH}}}\rangle$ binned as a function of Hα surface brightness for the SMC (blue squares) and the LMC (orange circles). Both correlations tend to reach $\langle {q}_{\mathrm{PAH}}\rangle \sim 0 \%$ at high I_Hα values, independently of the average value of each galaxy. This agrees with the scenario where PAHs are destroyed in H ii regions. In Figure 7, we identify the value I_Hα ∼ 1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻² (or ∼300 dR) as the Hα emission level at which $\langle {q}_{{\rm{PAH}}}\rangle$ starts to rapidly decrease in the LMC (dashed gray line). We use this value to plot the solid contours on Figure 4. This marks a limit where the PAH fraction drops more steeply with Hα surface brightness. We use this value as a separation between the diffuse neutral gas and ionized gas. Figure 7 shows that the PAH fraction decreases even at low values of Hα surface brightness, but that it drops rapidly only after it reaches the vertical dashed line (∼1.5 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻², or 300 dR). Before that value, the PAH fraction decreases only mildly. To first order the Hα emission traces the surface density of ionized gas. Given that the distribution of H i is fairly flat across both galaxies (e.g., Stanimirovic et al. 2000), it is possible that the turn-over at 300 dR occurs at a point where ionized gas starts to make a large contribution to the total surface density. The limit marked by the vertical dashed line is lower than the Hα surface brightness in H ii regions, implying that PAHs may undergo destruction even in ionized gas outside the H ii regions.¹⁵

Zoom In Zoom Out Reset image size

A decrease of PAH fraction in DIG has been suggested by previous works on the MW by Dobler et al. (2009) and Dong & Draine (2011). In Dong & Draine, the authors studied the Hα-to-free–free emission in the MW DIG, and found that this ratio corresponds to a lower temperature than they were able to produce with their model using the MW diffuse ISM PAH abundance. With a model consisting of ionized gas, recombining gas in the process of cooling, and cool neutral gas, they managed to reproduce the observations by lowering (by a factor of ∼3) the PAH fraction in photoionized regions with respect to that of the global average in the ISM. Our results, constrained from observed SEDs, agree with a scenario where the PAH fraction is lower in ionized gas.

A high intensity of the radiation field is often quoted as a cause for enhanced destruction of PAHs. Here, we can test that scenario by looking at $\langle {q}_{{\rm{PAH}}}\rangle$ as a function of ${U}_{\min }$,or $\overline{U}$. In our work, we only adjust for the radiation field intensity and not the hardness. In the top panel of Figure 8, we show the variations of $\langle {q}_{{\rm{PAH}}}\rangle$ in bins of $\overline{U}$, in the diffuse neutral medium (filled symbols), and the non-H ii region ionized gas (above the limit in Hα emission and outside of H ii regions; empty symbols). We choose to compare the diffuse neutral medium and the non-H ii region ionized gas because ionized-gas-related destruction processes should not occur in the former. In the diffuse neutral medium of the SMC, $\langle {q}_{{\rm{PAH}}}\rangle$ seems to increase slightly with $\overline{U}$ although with increasing uncertainty. This uptick in $\langle {q}_{{\rm{PAH}}}\rangle$ may be the result of an overestimation due to the fact that we do not take into account a change in radiation field hardness (see Section 5.3.1) as $\overline{U}$ increases. It is possible that, in low-metallicity environments, changes in the radiation field hardness are more important, hence the observed increase of $\langle {q}_{{\rm{PAH}}}\rangle$ in the SMC. In the diffuse neutral medium in the LMC, however, we notice a decrease of $\langle {q}_{{\rm{PAH}}}\rangle$ at $\overline{U}$ ∼ 3. Given the scatter, there are only minor variations of $\langle {q}_{{\rm{PAH}}}\rangle$ up to log($\overline{U}$) = 0.5 ($\overline{U}$ ∼ 3). At this point, $\langle {q}_{{\rm{PAH}}}\rangle$ decreases noticeably. Below this, the intensity of the radiation field does not seem to affect the PAH fraction. As expected, in the ionized gas outside of H ii regions, there seems to be a decrease of $\langle {q}_{{\rm{PAH}}}\rangle$ with $\overline{U}$ even for $\overline{U}$ < 1. This is not surprising because, in defining this phase, we selected the pixels where $\langle {q}_{{\rm{PAH}}}\rangle$ decreases in Figure 7. When we control for the intensity of the radiation field, we can see that the gas phase has an impact on the PAH fraction. For an identical $\overline{U}$, the PAH fraction does not decrease as quickly, whether it is in the diffuse neutral or the non-H ii region ionized gas. We also point out that, even at the lowest $\overline{U}$, $\langle {q}_{{\rm{PAH}}}\rangle$ in the ionized medium never reaches that of the neutral medium.

Zoom In Zoom Out Reset image size

The bottom panel of Figure 8 shows the ratio of PAH fraction in different gas phases $\langle {q}_{\mathrm{PAH}}^{\mathrm{out}-{\rm{H}}\,{\rm{II}}}\rangle /\langle {q}_{\mathrm{PAH}}^{\mathrm{diff}.\ \mathrm{neut}.}\rangle$ (same color code as the top panel). It is interesting to notice the increasing offset between the PAH fraction in the ionized gas and the neutral medium as $\overline{U}$ increases, and that this offset is the same in both galaxies. We discuss these results in the context of PAH destruction in Section 5.2.

The difference between the global averages $\langle {q}_{\mathrm{PAH}}^{\mathrm{SMC}}\rangle$ and $\langle {q}_{\mathrm{PAH}}^{\mathrm{LMC}}\rangle$ is in good agreement with the power-law dependence of ${q}_{{\rm{PAH}}}$ with metallicity found by Rémy-Ruyer et al. (2015).¹⁶ Our values also sit well within the scatter of the Draine et al. (2007) results, where a steep drop of ${q}_{{\rm{PAH}}}$ occurs at a metallicity threshold of 12 + log(O/H) ∼ 8.1.

The degree to which our results agree with either the threshold or power-law dependence, however, depends on the gas phase. In the MW, studies have found a diffuse neutral medium ${q}_{{\rm{PAH}}}$ value of ∼4.6% (Li & Draine 2001; Weingartner & Draine 2001a). For a fair comparison, this should be compared to our selection of the diffuse neutral medium in the LMC and SMC listed in Table 2. In this case, the power-law boundaries found by Rémy-Ruyer et al. (2015) underestimate the PAH fraction in the LMC. Indeed, the LMC's diffuse neutral gas ${q}_{{\rm{PAH}}}$ is very similar to the MW's, while the SMC falls a factor of ∼4 lower. If we treat the diffuse neutral gas as a reference value, where most destruction processes are not operating, our results strengthen the interpretation that there is a threshold in metallicity where ${q}_{{\rm{PAH}}}$ decreases significantly, and that it lies between SMC and LMC metallicities.

We use the separation of the gas phases in the MCs and their associated PAH fraction to investigate the life-cycle of PAHs in the ISM. We first go over some of the possible scenarios for PAH formation and how they might affect $\langle {q}_{{\rm{PAH}}}\rangle$ in each phase.

Theoretical studies have found that the atmospheres of post-asymptotic giant branch (AGB) stars could be a favored site for the formation of PAHs (Cherchneff et al. 1992; Sloan et al. 2007). PAHs produced by AGB stars should primarily input into the diffuse ISM, since the old stellar distribution is not closely related to the current location of star-forming regions or dense gas (e.g., see Sandstrom et al. 2010 for a comparison between the SMC old stellar distribution and the ISM). If PAHs are primarily formed by AGB stars and no additional destruction or production mechanisms are operating in the diffuse neutral gas, one could argue that the $\langle {q}_{\mathrm{PAH}}^{\mathrm{ref}}\rangle$ we have identified should reflect the efficiency of PAH formation by AGB stars. Boyer et al. (2011, 2012) studied the stellar population of both MCs, and found that the carbon-rich AGB population has similar dust production rates in the SMC and in the LMC, despite their different metallicities. Although other evolved stars such as red supergiants seem to be more effective at producing dust in the LMC, they do not play a major role in the overall dust production by evolved stars (e.g., Boyer et al. 2011). Therefore, without some additional destruction mechanism in the diffuse neutral phase, we cannot explain the different $\langle {q}_{\mathrm{PAH}}^{\mathrm{ref}}\rangle$ between the SMC and the LMC solely with more effective formation of PAHs in AGB atmospheres in the LMC.¹⁷

Another scenario for PAH formation suggested by theoretical studies (e.g., Jones et al. 1996) is shattering of large carbon grains, leading to the production of smaller carbon grain fragments. Jones et al. (1996) argued that the shattering of carbon grains can occur at velocities as low as 1 km s⁻¹, and is the prevalent process affecting dust grains at shock velocities ≥100 km s⁻¹. Their study focused on the grain–grain interaction in the so-called warm intercloud medium, similar to our definition of diffuse neutral medium. There, they found that the redistribution of carbon-rich dust mass from large grains into small grains, and even PAH-like fragments, is significant. In Slavin et al. (2015), the authors found that the redistribution of dust mass in smaller grains is important in supernova remnants shocks up to ∼200 km s⁻¹. A critical question, however, is whether the hot shocked gas is able to quickly sputter and destroy any PAH fragments that are created by shattering. Theoretical calculations by Micelotta et al. (2010b) suggest that PAHs are quickly destroyed by collisions with energetic particles at shock velocities higher than 100 km s⁻¹. This work leads to the conclusion that shattering of dust grains, in <100 km s⁻¹ shocks, could be a significant source of PAH-like grains. Because the shattering process is directly linked to the available carbon-rich dust mass, one would expect a higher abundance of small grain fragments if there is a higher dust-to-gas ratio. Given the observed dust-to-gas ratio in the MCs (∼0.003, ∼0.0008 in the LMC and SMC, respectively; Leroy et al. 2011; Gordon et al. 2014; Roman-Duval et al. 2014), this would be consistent with our finding of higher $\langle {q}_{\mathrm{PAH}}^{\mathrm{ref}}\rangle$ in the LMC, compared to that in the SMC. In the former, the higher $\langle {q}_{\mathrm{PAH}}^{\mathrm{ref}}\rangle$, assuming destruction of PAHs, is not significant in the diffuse neutral medium, and could be the result of more efficient shattering of large carbonaceous grains.

Another hypothesis for PAH formation is growth in the molecular phase of the ISM (e.g., Sandstrom et al. 2010, 2012; Zhukovska et al. 2016). In this scenario, the lower $\langle {q}_{{\rm{PAH}}}\rangle$ in the SMC would be related to less efficient growth processes in the molecular gas, possibly due to the lower abundance of metals to accrete onto existing grains to form PAHs. We do see that the $\langle {q}_{\mathrm{PAH}}^{\mathrm{mol}.}\rangle$ is lower in the SMC than in the LMC. Previous work by Sandstrom et al. (2010) found enhanced ${q}_{{\rm{PAH}}}$ in dense regions of the SMC compared to the diffuse gas phases, which was interpreted as evidence for growth in the molecular gas and destruction operating in the diffuse ISM. We do find a higher PAH fraction in the molecular gas phase of each galaxy with respect to the global averages, (see Table 2; LMC: 4.3%; SMC: 1.1%), but the $\langle {q}_{{\rm{PAH}}}\rangle$ is similar in the molecular and diffuse neutral phases of each galaxy. The observation of the same $\langle {q}_{{\rm{PAH}}}\rangle$ in the molecular and diffuse neutral gas is consistent either with formation of PAHs in the diffuse ISM and incorporation into molecular clouds, or its inverse, assuming no destruction processes operate differentially in these two phases. In the same molecular gas, accretion and coagulation of small grains onto big grains (e.g., Stepnik et al. 2003; Köhler et al. 2012) could also lead to a decrease in the observed PAH fraction. In our case, observational evidence for that mechanism would be a lower PAH fraction in the molecular regions. Since we do not see such variation, we do not favor this possibility.

Observational studies have shown that the PAH size distribution and properties are sensitive to the local radiation field and gas ionization. This is seen in the vicinity of MW H ii regions and photodissociation regions (e.g., Berné et al. 2007; Compiègne et al. 2007; Arab et al. 2012; Peeters et al. 2017) as well as in nearby galaxies (Gordon et al. 2008; Paradis et al. 2011; Relaño et al. 2016). PAH destruction can be accomplished in several ways: mediated by interaction with energetic photons, sputtering by particles in hot gas, or chemical reactions. A long-standing hypothesis to explain the PAH deficit at low metallicity is that PAHs are more readily destroyed in such conditions (e.g., Madden et al. 2006; Galliano et al. 2008). A possible scenario for enhanced destruction of PAHs at low metallicity is photodestruction by the radiation field, either due to increased intensity or hardness. There are a number of observational studies that have found correlations that agree with such a scenario. For example, Madden et al. (2006) and Gordon et al. (2008) found that the PAH features disappear as the radiation field hardness increases in H ii regions (measured from MIR neon and sulfur line ratios). Theoretical studies have also shown that PAHs are subject to sputtering and fragmentation in ionized gas (Micelotta et al. 2010a; Bocchio et al. 2012) because of electronic and/or atomic interactions. There, projectile particles can reach a high velocity with respect to that of the grains, leading to catastrophic collisions. In general, the regions of the galaxy with the hardest radiation fields will also be those where ionized gas exists and the overall intensity of the radiation field is higher. Therefore, to distinguish between the potential destruction mechanisms, we need to attempt to separate the effects of these quantities. We do so using Hα emission as a tracer of ionized gas, and our determination of $\overline{U}$ as a tracer of the intensity of the radiation field, as presented in Section 4.5 and Figure 8. We do not have a direct tracer for the hardness of the radiation field covering the full extent of the galaxies. We note the similar trends of the ratios $\langle {q}_{\mathrm{PAH}}^{\mathrm{out}-{\rm{H}}\,{\rm{II}}}\rangle /\langle {q}_{\mathrm{PAH}}^{\mathrm{diff}.\ \mathrm{neut}.}\rangle$ of the SMC (blue) and the LMC (orange), in the bottom panel of Figure 8. They suggest that metallicity does not have an impact on the relative efficiencies of the destruction processes between the diffuse neutral medium and the ionized gas outside of H ii regions (in the same bin of the radiation field). We do not test, however, the impact of metallicity on the global amount of each gas phase in a galaxy, which might have an impact on the overall destruction of PAHs.

PAHs in a more intense radiation field could be more easily destroyed because they are more fragile, due to their ionization state. Ionized PAHs are less stable, and more prone to losing, for example, H atoms (Montillaud et al. 2013). Weingartner & Draine (2001b) showed that the charge of small carbonaceous grains varies depending on a parameter $G\sqrt{T}$/n_e, where G is the radiation field intensity, T the temperature of the gas, and n_e the electron density. To test whether PAH ionization may lead to higher destruction rates, the different behavior of the neutral and ionized gas phases at fixed $\overline{U}$ from Figure 8 is of interest. In a bin of $\overline{U}$, we control for the radiation field intensity, and it is the same in both the diffuse neutral and the non-H ii DIG. Weingartner & Draine (2001b) showed that the gas temperature does not significantly affect the ionization state of PAHs. The only parameter remaining then is n_e. If the electron density were lower in the non-H ii ionized gas compared to the diffuse neutral gas, then PAHs could be more ionized in this medium, facilitating their destruction. While this seems counter-intuitive, given that the non-H ii region ionized gas is defined by its ionization state, the difference in overall density between the ionized and neutral phases and the low, but non-negligible, fractional ionization in the neutral medium could lead to the situation where n_e is lower for the ionized gas. Li & Draine (2001) showed that PAHs larger than 7 Å are more ionized in the MW's warm neutral medium than in the warm ionized medium. In that case, the lower $\langle {q}_{\mathrm{PAH}}^{\mathrm{out}-{\rm{H}}\,{\rm{II}}}\rangle$ compared to that of the diffuse neutral medium could not be explained by an easier destruction of PAHs because of their ionization.

Based on the comparison between the gas phases in the two galaxies, we could explain the offset in $\langle {q}_{{\rm{PAH}}}\rangle$ between the LMC and SMC by: (i) fragmentation of large grains in the diffuse medium leading to more PAHs in the LMC due to its higher dust content; (ii) formation of PAHs in the molecular ISM and their injection in the diffuse gas, assuming there is no preferential destruction in either molecular or diffuse gas. The formation of PAHs in AGB stars cannot explain the difference in PAH fraction between the SMC and the LMC. The destruction mechanisms (e.g., photodestruction by the radiation field), suggested to be more effective in lower-metallicity galaxies, do not differ significantly between the SMC and the LMC in this work (see the bottom panel of Figure 8).

5.3.1. Impact of the Modeled Radiation Field

5.3.2. Diffuse Ionized Gas

5.3.3. Metallicity Variations across the MCs