POLYCYCLIC AROMATIC HYDROCARBONS IN GALAXIES AT z ∼ 0.1: THE EFFECT OF STAR FORMATION AND ACTIVE GALACTIC NUCLEI

The full details of the SSGSS sample selection, observations, and data reduction are reported in M. J. O'Dowd et al. (2009, in preparation). To summarize:

SSGSS is a Spitzer spectroscopic survey of 101 normal, star-forming galaxies from the Lockman hole region. This ∼10 square degree field of low Galactic H i/cirrus is extensively surveyed in multiple wave bands. SSGSS galaxies were selected to have coverage by SDSS, Galaxy Evolution Explorer (GALEX), and IRAC (four channel) and MIPS (24 μm, 70 μm). We applied a surface brightness limit of I_5.8 μm>0.75 MJy sr⁻¹, and a flux limit of F_24 μm>1.5 mJy. These criteria yielded 154 galaxies, from which we selected 101 galaxies to span the range of normal galaxy properties by uniformly sampling D_n(4000) versus NUV-K space. This sample has a redshift range of 0.03 < z < 0.2 with median redshift z_med = 0.08, and a total infrared luminosity range of 3.7 × 10⁹ L_☉ < L_TIR < 3.2 × 10¹¹ L_☉, with median 3.9 × 10¹⁰ L_☉.

Spectroscopy was obtained in staring mode using the "low-res" short-low (SL) and long-low (LL) IRS modules for the entire sample, providing 5.2–38.0 μm coverage with resolving power of ∼60–125. "hi-res" short-high (SH) spectroscopy was obtained for a subsample of the brightest 33 galaxies, providing 9.9–19.6 μm coverage at R ≈ 600. For the bright subsample, exposure times were 8 minute for the SL and LL modules, and 16 minute for SH. For the remaining sample, exposure times were 8 minute for SL and 16 minute for LL. All data were taken during Spitzer IRS campaign 37.

We used the two-dimensional Spitzer data products processed by the Spitzer Pipeline version S15.3.0, which performed standard IRS calibration (ramp fitting, dark subtraction, droop, linearity correction, and distortion correction, flat fielding, masking and interpolation, and wavelength calibration). Sky subtraction was performed manually, with sky frames constructed from the two-dimensional data frames, utilizing the shift in galaxy spectrum position between orders to obtain clean sky regions. IRSCLEAN (v.1.9) was used to clean bad and rogue pixels. SPICE was used to extract one-dimensional spectra, and these spectra were finally combined and stitched manually by weighted mean.

Due to problems with some observations, the sample studied in this paper consists of 92 galaxies observed with the low-res modules and a subset of 32 observed at hi-res. In addition, for subset of sources spectrophotometric calibration failed for the SL second-order module, resulting in fluxes a factor of ∼2 lower than expected based on the SL first-order module spectrum and IRAC photometry. We eliminate these sources from analyses involving the affected spectral region, as described in Section 2.2.3.

Figure 1 shows the composite low-res spectrum for this sample. PAH features and complexes at 6.2, 7.7, 8.6, 11.3, 12, 12.6, and 17 μm are prominent, as are a number of atomic emission lines. A thermal dust continuum component dominates redward of ∼13 μm. These composite spectra are normalized between 20 μm and 24 μm.

Zoom In Zoom Out Reset image size

2.1.1. Aperture Effects

To measure the strengths of the PAH features, we use the PAHFIT spectral decomposition code of (Smith et al. 2007, S07 hereafter). This code performs χ² fitting of multiple spectral components, including PAH features modeled as Drude profiles, the thermal dust continuum, starlight, prominent emission lines, and dust extinction. For ease of comparison with the results of S07, we use identical temperatures for the thermal continuum components and identical central wavelengths and widths for the PAH features.

We report a sample of the fitted fluxes of the most prominent PAH features and PAH feature complexes in Table 1. These include the 6.2 and 8.3 μm bands, which are discrete PAH features, and the 7.7, 11.3, and 17 μm complexes, which are blends of three, two, and four subfeatures, respectively. Henceforth, we refer to individual PAH features and complexes of multiple features simply as "features". In Table 1, we also report the fitted dust extinction optical depth, τ_9.7 and the integrated line strengths for [Ne ii]_12.8 μm and [Ne iii]_15.6 μm.

Table 1. A Sample of the Fitted Fluxes of the Most Prominent PAH Features and PAH Feature Complexes

SSGSS	R.A.	Decl.	z	F6.2	F7.7	F8.6	F11.3	F17	$F_{[{\rm Ne}\,{\scriptstyle {{\rm II}}}]}$	$F_{[{\rm Ne}\,{\scriptstyle {{\rm III}}}]}$	τ9.7
1	160.34398	58.89201	0.066	1.35E-16	4.35E-16	8.89E-17	1.09E-16	4.72E-17	1.08E-17	1.17E-18	...
2	159.86748	58.79165	0.045	7.04E-17	2.14E-16	4.88E-17	5.43E-17	4.21E-17	3.51E-18	4.58E-18	...
11	162.41000	59.58426	0.047	4.56E-17	1.58E-16	3.32E-17	5.37E-17	3.40E-17	2.67E-18	2.02E-19	...
12	162.36443	59.54812	0.072	1.55E-16	6.61E-16	1.37E-16	1.99E-16	1.40E-16	3.41E-17	3.92E-18	1.47
14	162.52991	59.54828	0.153	8.67E-17	3.11E-16	6.74E-17	8.93E-17	4.97E-17	1.06E-17	1.93E-18	...
15	161.78737	59.63707	0.153	1.60E-17	5.81E-17	2.67E-17	3.88E-17	1.41E-17	8.37E-18	...	...
16	161.48123	59.15443	0.072	8.38E-17	3.10E-16	6.03E-17	8.49E-17	6.12E-17	1.35E-17	1.42E-18	...
17	161.59111	59.73368	0.047	5.68E-16	2.02E-09	4.11E-16	4.74E-16	2.81E-16	4.71E-17	1.11E-17	0.94
27	161.11412	59.74155	0.072	1.32E-16	1.19E-09	2.67E-16	3.29E-16	1.97E-16	6.58E-17	1.38E-17	...
28	161.71980	56.25187	0.103	1.98E-18	1.95E-16	4.68E-17	8.54E-17	2.67E-17	7.68E-18	4.23E-18	...
30	162.26756	56.22390	0.046	1.13E-16	4.02E-16	7.74E-17	1.00E-16	3.01E-17	5.10E-18	4.63E-18	...
32	163.00845	56.55043	0.117	7.61E-17	2.64E-16	5.72E-17	6.91E-17	1.75E-17	9.13E-18	4.76E-18	...
33	161.92709	56.31395	0.185	2.99E-17	1.61E-16	2.68E-17	5.48E-17	3.89E-17	7.02E-18	2.23E-18	0.66
34	161.75783	56.30670	0.046	9.03E-17	2.76E-16	6.15E-17	7.29E-17	3.47E-17	1.01E-17	3.14E-18	0.20
39	162.04231	56.38041	0.074	5.47E-17	1.97E-16	4.41E-17	5.54E-17	1.49E-17	8.42E-18	1.47E-18	...
45	161.76901	56.34029	0.113	2.41E-17	2.01E-16	4.64E-17	6.85E-17	3.07E-17	1.73E-17	1.02E-18	...
47	163.39658	56.74202	0.102	7.47E-17	2.48E-16	5.14E-17	6.64E-17	2.89E-17	1.46E-17	2.18E-18	...
48	163.44330	56.73859	0.200	7.44E-17	2.92E-16	5.86E-17	7.52E-17	1.88E-17	7.51E-18	1.98E-18	...
54	163.26968	56.55812	0.115	2.12E-16	7.50E-16	1.54E-16	1.96E-16	1.30E-16	1.41E-17	1.62E-18	0.82
61	163.19810	56.48840	0.073	3.14E-17	1.23E-16	2.00E-17	4.14E-17	4.56E-17	3.36E-17	3.98E-18	...
62	163.09050	56.50836	0.133	1.03E-16	3.82E-16	7.03E-17	9.18E-17	5.97E-17	2.15E-17	...	...
64	163.53931	56.82104	0.073	2.62E-16	9.32E-16	1.85E-16	2.04E-16	1.69E-16	4.45E-17	6.49E-18	1.14
65	158.22482	58.10917	0.118	1.82E-16	6.69E-16	1.26E-16	1.74E-16	1.00E-16	2.72E-17	4.82E-18	0.29
69	159.04880	57.72258	0.076	8.83E-17	2.46E-16	5.92E-17	7.23E-17	4.54E-17	1.38E-17	1.91E-18	...
70	159.34668	57.52069	0.090	4.72E-17	2.17E-16	4.75E-17	6.48E-17	2.57E-17	8.37E-18	2.46E-18	...
73	158.91122	57.59536	0.080	1.88E-17	9.81E-17	2.06E-17	2.47E-17	2.50E-17	7.12E-18	...	...
77	158.91344	57.71219	0.044	3.93E-17	1.62E-16	3.22E-17	3.56E-17	1.60E-17	1.12E-17	5.46E-18	...
83	159.73558	57.26361	0.119	3.13E-19	3.06E-16	6.76E-17	8.36E-17	4.79E-17	4.69E-18	3.78E-18	...
94	159.63510	57.40035	0.061	7.79E-17	2.77E-16	5.33E-17	6.52E-17	4.08E-17	1.44E-17	5.61E-19	...
95	161.48724	57.45520	0.115	7.48E-17	2.66E-16	5.84E-17	8.49E-17	2.52E-17	1.33E-17	1.04E-19	0.79
98	160.29099	56.93161	0.050	6.40E-17	7.98E-16	2.18E-16	2.64E-16	1.37E-16	1.81E-17	1.00E-17	...
99	160.30701	57.08246	0.077	1.62E-16	5.30E-16	1.11E-16	1.23E-16	8.50E-17	2.09E-17	1.41E-17	...

Notes. Integrated fluxes in W m⁻² for PAH features and complexes and for $[{\rm Ne}\,{\scriptstyle {{\rm II}}}]_{12.8 \,\mu {\rm m}}$ and $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]_{15.6 \,\mu {\rm m}}$ emission lines derived from PAHFIT spectral decomposition of the subsample of SSGSS galaxies with hi-res spectroscopy. Also reported are the redshift of the galaxy, z, and dust extinction given as τ_9.7, the optical depth at 9.7 μm. PAH strengths and absorption come from fitting of low-res data, while emission lines come from fitting of hi-res data.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Figure 2 shows examples of PAHFIT decompositions of the SSGSS spectra.

Zoom In Zoom Out Reset image size

2.2.1. Dust Extinction

2.2.2. Spline Fitting

2.2.3. Problematic 6.2 μm Feature Fits

2.2.4. The Relative Dominance of Star Formation and AGNs

Throughout this paper, we color code to indicate the location in log $[{\rm O}\,{\scriptstyle {{\rm III}}}]_{5007}/{\rm H}\beta$ versus log $[{\rm N}\,{\scriptstyle {{\rm II}}}]_{6583}/{\rm H}\alpha$ space: the so-called BPT diagram of Baldwin et al. (1981; Figure 3). Red points indicate more powerful AGNs satisfying the Kewley et al. (2001) designation, green points indicate composite sources with weaker active components between the Kewley and the Kauffmann et al. (2003b) designation, while black points indicate galaxies dominated by stellar light, below the Kauffmann designation.

Zoom In Zoom Out Reset image size

BPT designation must be interpreted carefully. Rather than an absolute discriminator between star-forming galaxies, composite sources, and AGNs, for this study we instead interpret it as a measure of the relative dominance of star formation and AGNs with the Sloan aperture. The similarity of the Sloan and Spitzer SL apertures means that BPT designation measures this relative dominance within roughly the same galactic region responsible for the observed PAH emission. The regions probed in this study are larger than those probed in earlier, local studies. This is an important difference as it means that we observe a more global stellar population, where any AGN component is less overwhelming.

2.2.5. Correlation of PAH Ratios and Comparison to Other Studies

Ratios of the integrated fluxes of the fitted Drude profiles were used to determine PAH luminosity ratios, designated L_λ1/L_λ2, where λ is the central wavelength of the feature in micrometers. Figure 4 shows the tight correlations between different short-to-long wavelength PAH ratios. L_6.2/L_11.3 versus L_7.7/L_11.3 follows a tight locus with ∼0.25 dex scatter. L_8.6/L_11.3 versus L_7.7/L_11.3 shows a similar scatter, but broadens toward low ratio values. The tight correlations suggest that lower wavelength PAH features are strongly coupled and are less strongly tied to the 11.3 μm feature.

Zoom In Zoom Out Reset image size

In this figure, we also compare our results to two other spectroscopic studies of PAH intensities.

S07 present spectral decomposition of IRS observations of 59 SINGS galaxies, and we use an identical spectral decomposition technique (see Section 2.2). An important difference between S07 and our survey is, with their local targets and shorter exposure times, they are dominated by the highest surface brightness regions near the nucleus, while our observations sample more of the extended structure. Their reported PAH ratios are very similar to ours, with the exception that the S07 sample have, on average, higher L_7.7/L_11.3, and lack the small number of galaxies with L_6.2/L_11.3 < 0.3 that we observe.

Galliano et al. (2008, G08 hereafter) present spectral decomposition of the ISO and Spitzer IRS observations of Galactic H ii regions and dwarf spiral and starburst galaxies. We take the results of their most comparable decomposition technique, in which they fit Lorenzian profiles to the PAH bands (as opposed to the Drude profiles used by PAHFIT). While their samples show very similar trends in the 7.7 μm and 8.6 μm features versus the 11.3 μm feature, they find that the 6.2 μm feature is on average stronger by a factor of 1.3—2 than both our results and those of S07. This discrepancy is likely due to the differences in the fitting techniques.

Models of the stochastic heating of dust grains (Tielens 2005; Draine & Li 2007a; Schutte et al. 1993) show that the relative power emitted in a given PAH band is strongly dependent on the distribution of grain sizes. In general, smaller dust grains emit more power in the shorter wavelength bands, and larger grains dominate longer wavelengths bands. In the regime of grain sizes expected in interstellar dust (tens to hundreds of carbon atoms), rapid drops are expected in the 6.2 μm and 7.7 μm bands relative to longer wavelength features, and in 6.2 μm relative to 7.7 μm, with the increasing grain size.

It is also expected that the ionization state of a PAH molecule will have a dramatic effect on its spectrum (Draine & Li 2007a). In particular, carbon–carbon vibrational modes are known to be significantly more intense in ionized PAH molecules (Tielens 2005). As the 6.2 μm and 7.7 μm bands result from radiative relaxation of CC stretching modes, the ratios of these bands to those arising from carbon–hydrogen modes such as the 11.3 μm feature are expected to drop by an order of magnitude between completely neutral and completely ionized PAH clouds. At the same time, the 6.2 μm feature should change little relative to the 7.7 μm feature as the ionization fraction changes.

Comparison of the PAH band power ratios L_6.2/L_7.7 to L_11.3/L_7.7 is useful in extricating these two effects, with the former being sensitive to the PAH grain size but relatively unaffected by ionization as both bands result from the similar CC modes, while the latter is very sensitive to ionization, and somewhat less so to the grain size. Figure 5 shows these ratios for our sample plotted over the Draine & Li (2007a) models. Most of the sample is restricted to a tight locus between 0.2 and 0.4 in both L_6.2/L_7.7 and L_11.3/L_7.7, although within this locus there is a weak trend in the direction expected for constant ionization fraction with changing grain size, albeit with significant scatter. We discuss the distribution of active galaxies in this plot in Section 3.3.

Zoom In Zoom Out Reset image size

We compare the PAH spectra resulting from our PAHFIT models to SDSS optical diagnostics of star formation history and AGN activity. Figure 6 shows the ratio of short-to-long wavelength PAH bands (L_6.2/L_7.7, L_6.2/L_11.3, and L_7.7/L_11.3) versus the star formation diagnostics D_n(4000), specific star formation rate (SSFR), and Hα equivalent width (Hα EW). D_n(4000) (Kauffmann et al. 2003a) measures the 4000Å break over a narrower bandpass than the standard D(4000) (Bruzual et al. 1983) and provides a measure of the luminosity-weighted stellar age. The SSFR (Brinchmann et al. 2004) is the star formation rate (SFR) as determined by population synthesis models per unit stellar mass. The Hα line provides a direct measure of recent star formation, and in particular the presence of short-lived OB stars.

Zoom In Zoom Out Reset image size

These star formation diagnostics are relatively free of the effects of dust extinction: D_n(4000) and Hα EW because of the narrow bandpasses over which these diagnostics are calculated, and the SSFR because it has been corrected for extinction (Brinchmann et al. 2004).

The most striking correlations are seen relative to L_7.7/L_11.3, with the short wavelength PAH band becoming more dominant with increasing recent star formation. The ratios of the 6.2 μm feature to longer wavelength features follow similar, albeit weaker trends to L_7.7/L_11.3.

Table 2 shows the results of the regression analyses of PAH ratios to optical properties. We use Kendall's correlation to determine the significance of the trends, given as probabilities of obtaining the observed data given the null hypothesis of no correlation. We obtain a significance of <5% for L_6.2/L_11.3 and L_7.7/L_11.3 versus all star formation diagnostics, while L_6.2/L_7.7 shows no significant correlations.

Table 2. Results of Linear Fits and Regression Analysis for PAH Ratios Versus SDSS Optical Properties

Diagnostic	Sample	$\displaystyle \frac{L_{6.2}}{L_{7.7}}$	$\displaystyle \frac{L_{6.2}}{L_{11.3}}$	$\displaystyle \frac{L_{7.7}}{L_{11.3}}$
	Full sample	...	−0.07 (1.6%)	−0.26 (<0.01%)
Dn(4000)	Star forming	...	...	−0.39 (3%)
	AGN/composite	...	...	...
	Full sample	...	0.2 (0.17%)	0.21 (<0.01%)
SSFR	Star forming	...	...	0.13 (0.2%)
	AGN/composite	...	...	...
	Full sample	...	0.16 (0.01%)	0.19 (<0.01%)
Hα EW	Star forming	...	0.1 (0.5%)	0.16 (0.01%)
	AGN/composite	...	...	0.19 (2%)
	Full sample	...	−0.5 (0.1%)	−0.53 (<0.01%)
$\displaystyle \frac{[{\rm N}\,{\scriptstyle {{\rm II}}}]}{{\rm H}\alpha }$	Star forming	...	...	...
	AGN/composite	...	...	...
	Full sample	...	...	−0.20 (1.2%)
$\displaystyle \frac{[{\rm O}\,{\scriptstyle {{\rm III}}}]}{{\rm H}\beta }$	Star forming	...	...	−0.07 (4%)
	AGN/composite	...	...	−0.25 (5%)

Notes. The numbers given are the slopes of the best linear fits, followed in brackets by the significance of the trend according to Kendall's rank correlation. Dashes indicate that the correlation is not significant at the 5% level or better.

Download table as:  ASCIITypeset image

$[{\rm O}\,{\scriptstyle {{\rm III}}}]/{\rm H}\beta$ provides a measure of the ionization parameter weighted by the SFR, and hence of the hardness of the radiation field present in the galaxy (Kewley et al. 2001). From Figure 7 (left) it can be seen that there is a global trend between this line ratio and L_7.7/L_11.3, and no strong trends with ratios involving the 6.2 μm feature.

Zoom In Zoom Out Reset image size

The emission-line ratio $[{\rm N}\,{\scriptstyle {{\rm II}}}]/{\rm H}\alpha$ provides a diagnostic of gas-phase metallicity, although it is also affected by radiation field hardness. This ratio strongly correlates with the age of the stellar population. As expected, $[{\rm N}\,{\scriptstyle {{\rm II}}}]/{\rm H}\alpha$ follows the trends observed with star formation diagnostics (Figure 7, middle). The best global correlation is again with L_7.7/L_11.3, with the ratios of short-to-long wavelength PAH bands decreasing with the increasing emission-line ratio.

Table 2 also shows the results of the correlation analysis for PAH ratios versus emission-line ratios.

A more reliable measure of gas-phase metallicity, derived from a range of SDSS optical nebular lines (Tremonti et al. 2004) is available for a subset of the sample, although contains no BPT-designated AGNs or composite sources (Figure 7, right). The correlations of PAH ratios with this metallicity measure are slightly weaker than that for $[{\rm N}\,{\scriptstyle {{\rm II}}}]/{\rm H}\alpha$, suggesting that the trends observed in $[{\rm N}\,{\scriptstyle {{\rm II}}}]/{\rm H}\alpha$ may be dominated by the trend with radiation hardness. However, there is a notable locus of high short-to-long wavelength PAH ratios for lower metallicities. Galaxies in these upper left loci have the lowest metallicities (12 + log(O/H) < 9) and are among the youngest in the sample (D_n(4000) < 1.3), and exhibit the hardest radiation fields of the star-forming galaxies ($[{\rm O}\,{\scriptstyle {{\rm III}}}]/{\rm H}\beta > 0.4$).

It is known that PAH emission around 8 μm is stronger in higher metallicity galaxies (Engelbracht et al. 2005). The dominance of short wavelength PAH features in the lowest metallicity galaxies as seen in Figure 7 (right) may result from a difficulty in building large PAH molecules in an under-rich medium. It may also result from other age-dependent effects, given the close link between metallicity and age. However, it is not expected to arise through the link between metallicity and radiation field hardness; a hard radiation field will either preferentially destroy small PAH molecules, decreasing this ratio, or ionize molecules, which should not produce the observed change in L_6.2/L_7.7 with metallicity, as seen in Section 3.1.

From Figures 6 and 7, it can be seen that galaxies with dominant AGN components (those above the Kauffmann divide on the BPT diagram) have, on average, weaker short-to-long wavelength PAH bands. There is a strong link between AGN incidence and the star formation history of a galaxy; most notably, AGNs are not apparent in the most strongly star-forming galaxies, because of swamping of diagnostic lines, dust obscuration, or for reasons related to the life cycle of the AGN. The link between age and BPT designation is apparent from Figure 6. This makes it difficult to determine whether AGN activity, star formation history, or a combination of the two are responsible for the trends observed between PAH ratios and optical properties.

3.3.1. The Effect of Star Formation

3.3.2. The Effect of AGNs

The split-sample regression analysis for L_7.7/L_11.3 versus $[{\rm O}\,{\scriptstyle {{\rm III}}}]/{\rm H}\beta$ reveals that the correlation still holds for the AGN+composite subsample with a similar slope to the full sample. While there is a formal correlation for the star-forming subsample, the best-fit slope is so shallow as to suggest that the correlation is not meaningful. This suggests that there is a correlation between the hardness of the radiation field originating from an AGN and the 7.7 μm-to-11.3 μm ratio.

In Figure 8 (left), we show our sample on the BPT diagram, color-coded as in Figure 3. For better visual clarity, the symbol size is proportional to the ratio of long-to-short wavelength PAH band luminosity (upper: L_7.7/L_6.2, lower: L_11.3/L_7.7), and hence to the average grain size and/or neutral PAH fraction. The histograms (Figure 8, right) show the distribution of PAH ratios for different BPT designations. The strongest trend is found for L_7.7/L_11.3, for which the active galaxies show significantly lower values than star-forming galaxies (with <0.1% significance by a KS test). More dominant AGNs satisfying the Kewley designation have only marginally lower L_7.7/L_11.3 than composite sources satisfying the Kauffmann designation (10% significance). For L_6.2/L_7.7, the differences in the subsamples are not highly significant; there is a 10% significance for the difference in this ratio between active and star-forming galaxies.

Zoom In Zoom Out Reset image size

Although there is only a tentative difference in the L_7.7/L_11.3 distributions between the BPT-designated full AGNs and composite sources, these galaxies have indistinguishable distributions of D_n(4000) and Hα EW by a KS test. Thus, any difference in PAH ratios is likely due to AGN dominance and not due to a link between AGNs and star formation properties.

To better establish the independence of this relationship to star formation, we analyze a controlled subsample restricted to a narrow age range, with 1.3 < D_n(4000) < 1.5. Figure 9 shows the histogram of L_7.7/L_11.3 for galaxies in this range, split between galaxies with dominant AGN activity (either composite or full AGN by BPT designation), versus galaxies which dominated by star formation. In this range, both subsamples have indistinguishable distributions of both D_n(4000) and Hα EW. A KS test of these PAH ratios shows that they are different, with a significance level of 2%. It is therefore likely that AGNs play some role in reducing the relative 7.7 μm-to-11.3 μm PAH emission, and that the trend is not solely due to the link between AGN status and the age of the stellar population.

Zoom In Zoom Out Reset image size

In order to better separate the role of an AGN and star formation, we look at the relationship between hardness of the radiation field as measured by $[{\rm O}\,{\scriptstyle {{\rm III}}}]/{\rm H}\beta$ with stellar population age as measured by D_n(4000) (Figure 10). As a sensitive age diagnostic, D_n(4000) is highly correlated with [N ii])/Hα, and so this plot is analogous to the BPT plot, but with a clearer separation between young and older stellar populations. In this plot, the symbol size is again proportional to the long-to-short wavelength PAH ratio for the purpose of visual clarity. Here we see that, for hard radiation fields as measured by $[{\rm O}\,{\scriptstyle {{\rm III}}}]/{\rm H}\beta$, there is a striking difference in the PAH ratio between younger and older stellar populations. Younger stellar populations (D_n(4000) < 1.3) with hard radiation fields ([O iii]/Hβ>0.3) have stronger short wavelength PAH bands than older galaxies with similar $[{\rm O}\,{\scriptstyle {{\rm III}}}]/{\rm H}\beta$ (significance: 0.05% for L_7.7/L_11.3, 3% for L_6.2/L_7.7); they also have stronger short wavelength bands than galaxies of all ages with softer radiation fields ([O iii]/Hβ < 0.3, significance: 2% for L_7.7/L_11.3, 1% for L_6.2/L_7.7). Older populations with hard ionizing fields (which are also predominantly AGN-dominated) have stronger long wavelength bands than other populations (significance: <0.01% for L_7.7/L_11.3, 8% for L_6.2/L_7.7). The differences are illustrated in the histograms in the right panel of Figure 10.

Zoom In Zoom Out Reset image size

Thus, it seems that the behavior of PAH molecules in the presence of a hard radiation field differs depending on the age of the stellar population, and so it may depend on whether the radiation comes predominately from starburst activity or an AGN. These trends may also arise from a link between metallicity and both radiation field hardness and stellar population. Below, we discuss the factors that drive the evolution of short-to-long wavelength PAH ratios.

The youngest galaxies begin their evolution with low D_n(4000), low metallicity, a moderately hard radiation field resulting from starburst activity, and high short-to-long wavelength PAH ratios. The increase in metallicity with age is expected to allow the larger grains that dominate the longer wavelength feature in each ratio to be assembled more efficiently, resulting in a drop in this ratio as seen in Section 3.2.

Additionally, as age increases the starburst radiation field also softens. This may result in a decrease in the ionization fraction of the PAH molecules, which, according to the Draine & Li (2007a) models (Figure 5), may contribute to the drop in 7.7 μm-to-11.3 μm ratio between the youngest galaxies in this sample and older galaxies.

Even after these youngest, starburst stages, aging galaxies continue to exhibit a decrease in short-to-long wavelength PAH ratios, and in particular in L_7.7/L_11.3 (see Figure 6). Some of these galaxies experience AGN activity, and so are subject to radiation fields that are often much harder than in starburst galaxies. If the initial decrease in the PAH ratio with the age is due to the increasing neutral fraction of PAH molecules, then AGNs do not seem to re-ionize these molecules; indeed, the trend of diminishing the PAH ratio with the galaxy age continues through an increasing incidence of AGNs.

In Section 3.3.2, we saw that AGNs seemed to cause a decrease in the PAH ratio, particularly L_7.7/L_11.3, independently of star formation diagnostics, and that L_7.7/L_11.3 also dropped with AGN-sourced radiation hardness (Figure 7, lower left). This may be due to preferential destruction of small PAH grains by X-rays and/or shocks from the AGNs. It may also be due to heating of PAH molecules by the AGN, as ionization fraction decreases with increasing gas temperature (Tielens 2005). Thus, if the AGN is capable of heating PAH to high temperatures in dense regions, compared to the more diffuse heating by stars, then the net effect may be to reduce the ionization fraction.

3.4.1. [Ne ii] and [Ne iii] Measurements

This differing effect of an AGN- or starburst-sourced hard radiation field on the PAH spectrum is consistent with the result of S07, who plot L_7.7/L_11.3 versus $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]_{15.6 \,\mu {\rm m}}/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]_{12.8 \,\mu {\rm m}}$ for their sample. $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]$ provides a near-infrared (NIR) measure of the hardness of the radiation field. For a subsample of galaxies with significant AGN components (Seyferts and LINERs), they find a strong anti-correlation between the two ratios: the harder the ionizing field in the AGN, the weaker the short wavelength PAH band. However, the hardness of the radiation field in star-forming galaxies—radiation that is primarily stellar in origin—shows no correlation with the PAH ratio. Brandl et al. (2006) find a similar absence of correlation for 22 starburst nuclei over a similarly broad range of $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]$.

We have measured $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]$ for the subsample of 32 galaxies with the IRS hi-res spectrograph. In Figure 11, we compare our results with S07. Our results are consistent; however, the SSGSS sample does not include $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]$ values as high as those found by S07, and so we do not see the full trend between these ratios. The likely reason for the difference in measured $[{\rm Ne}\,{\scriptstyle {{\rm III}}}]/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]$ is that for the local galaxies studied by S07, the IRS slit samples a smaller, more central, and hence more AGN-dominated region than for our more distant galaxies.

Zoom In Zoom Out Reset image size

The 17 μm PAH band is well separated from the denser blueward bands, and hence provides one of the cleanest measures of PAH strength. This band may arise from C–C–C bending (Draine & Li 2007b); however, the true vibrational mode or modes responsible for this emission are not established. Nonetheless, it is expected that the dominant contributing grain size will be larger than lower wavelength bands. This band falls in the LL IRS module, which has a significantly larger slit width than both the SL module and the SDSS aperture, at 105 versus ∼37 and 3'', respectively. As a result, the sampling of different regions of the galaxy introduces unquantifiable uncertainties, and so we limit the analysis of this feature to qualitative comparisons.

In Figure 12, we plot the ratio of the 7.7 μm bands to the 17 μm band versus SDSS optical properties. With respect to most optical diagnostics, L_7.7/L₁₇ displays similar correlations to L_7.7/L_11.3, with similar dispersions. This ratio shows a significantly better correlation with respect to the SSFR, and a worse correlation with respect to $[{\rm N}\,{\scriptstyle {{\rm II}}}]/{\rm H}\alpha$. As with L_7.7/L₁₇, no correlation is observed with respect to ${\rm Ne}\,{\scriptstyle {{\rm III}}}]/[{\rm Ne}\,{\scriptstyle {{\rm II}}}]$. Dividing the sample between BPT-designated AGN/composite sources and star-forming galaxies, the same correlations are observed with all star formation diagnostics for star-forming galaxies. The correlations are not significant for other diagnostics or for the subsample of galaxies with AGN components.

Zoom In Zoom Out Reset image size

In Figure 13, we compare the histograms of the PAH ratios L_6.2/L_7.7, L_7.7/L_11.3, L_7.7/L₁₇, and L_11.3/L₁₇ divided into subsamples of varying of AGN activity. As was seen in Figure 8, L_6.2/L_7.7 reveals a slight relative increase in the strength of the longer wavelength PAH feature for galaxies with an AGN component, while L_7.7/L_11.3 shows a much stronger increase. There is also a significant increase in the strength of the 17 μm feature relative to the 7.7 μm feature with increasing AGN incidence and power. There is an increase in the 17 μm feature relative to the 11.3 μm feature between star-forming galaxies and full AGNs, but not between star-forming galaxies and composite sources.

Zoom In Zoom Out Reset image size

This may indicate a threshold in AGN power needed to destroy the larger PAH grains that dominate the 11.3 μm feature; only full AGNs are capable of destroying such grains with enough efficiency to register a difference in the L_11.3/L₁₇ ratio. Conversely, the average molecule sizes that dominate both the 6.2 and 7.7 μm features are destroyed with similar efficiency even for composite sources, resulting in a little change in L_6.2/L_7.7 with AGNs' status.

In Figure 14, we study the 17 μm feature in comparison to shorter wavelength features. L_6.2/L_7.7 versus L_7.7/L₁₇ shows a very narrow range of the former ratio over an order of magnitude variation in the latter. L_7.7/L₁₇ correlates well with L_7.7/L_11.3. Interestingly, an even tighter correlation is observed between L_7.7/L₁₇ and L_11.3/L₁₇, suggesting that the 7.7 μm and 11.3 μm features are more closely tied than the 11.3 μm and 17 μm features.

Zoom In Zoom Out Reset image size