Infrared SEDs of Seyfert galaxies: Starbursts and the nature of the obscuring medium Chris O'Dea, Jacob Noel-Storr, Catherine Buchanan (RIT), Jack Gallimore (Bucknell), Stefi Baum, David Axon, Andrew Robinson (RIT), Moshe Elitzur (Kentucky), Martin Elvis (CfA)
OUTLINE ● The Sample: Extended 12μm Seyfert ● The Observations: Spitzer IRAC, IRS SL & LL, MIPS SED ● The Diagnostics: Spectral Shapes, PAH, Molecular Gas, Si 9.7 μm, forbidden lines [NeV], [SIII]. (Preliminary Results) ● The Issues: AGN-Starburst Connection, Constraints on the Obscuring Torus ● Conclusions
The Extended 12 μm Sample of Seyfert galaxies Extended 12 μ m sample selected in mid-IR (IRAS 12 μ m) to minimize selection effects (Rush, Malkan, Spinoglio 1993) Our sample: All 87 Seyfert galaxies with cz < 10,000 kms -1 At z = 0.03, scale is 0.6 kpc/arcsec Initially classified: 37 Sy 1's and 50 Sy 2's Optical spectroscopy (and classifications) are heterogeneous and a few classifications have changed and a few more may change as better data is obtained Sy 1’s and 2’s have similar redshift and IRAS 12 μ m flux density distribution (Top) distribution of redshift. (Bottom) distribution of 12 μm flux density. Sy 1’s and 2’s have a similar distribution. The hatched objects are the 51 sources studied by Buchanan et al. (2006).
Spitzer Observations Our Spitzer program : ~4 – 100 μ m SEDS IRAC imaging 3.6 – 8 μ m IRS low resolution spectra in mapping mode 5 – 38 μ m MIPS spectral energy distributions 55 – 100 μ m This talk focuses on the IRS spectra. R ~ 60 to 125, slit width ~ 3.7” (SL) and 10.7” (LL). The data were processed using the Spitzer pipeline (S15.3.0; March 2007), To register the nuclear spectrum, the (background-subtracted) spectral images were interpolated using cubic splines, We used SPICE to extract the nuclear spectrum from the registered image. Jumps between orders removed by multiplicative scaling of SL orders to match LL
Spitzer Observations Our Spitzer program : ~4 – 100 μ m SEDS IRAC imaging 3.6 – 8 μ m IRS low resolution spectra in mapping mode 5 – 38 μ m MIPS spectral energy distributions 55 – 100 μ m This talk focuses on the IRS spectra. data were processed using the Spitzer pipeline (S15.3.0; March 2007), Background subtraction employed on-source data for the other spectral order (e.g., the SL2 on-source pointing was used as a background for SL1). These effective backgrounds were median combined and subtracted from each on-source pointing. To register the nuclear spectrum, the (background-subtracted) spectral images were interpolated using cubic splines, fitting surface brightness vs. slit position perpendicular to the slit. We used SPICE to extract the nuclear spectrum from the registered image. Note that the SPICE extractions assume point source structure, but in fact there is evidence for extended emission. The slit width between the SL and LL modules doubles, from ~ 5" to 10". Where extended emission is present, the result is a step up between SL and LL (~14 μm). The presence of extended emission at shorter wavelengths is verified on IRAC images, particularly at 8 microns.
Range of Spectral Shapes in the 12 μm Sample Left two panels have strong PAH. In right panel, last 3 are 2 with strong Si absorption and one misc. From a sample of 51 objects studied by Buchanan et al (2006).
Spectral shapes Strong PAH & Si 9.7 µm abs, red continuum, e.g., Mrk 938 (47%) Strong Si 9.7 µm absorption, e.g., NGC1194 (4%) Warm dust, flat µm slope, e.g., NGC4151 (31%) Weak PAH & Si, Red continuum, e.g., NGC3516 (16%) We find four distinct 5-35 µm spectral shapes in the sample 51 objects studied by Buchanan et al. (2006)
Mid-IR Diagnostics using PAHFIT Smith & Draine’s PAHFIT (slightly modified) used to fit the IRS Spectra Continuum approximated by a series of grey bodies at a range of Temp. Comprehensive set of PAH, Si, forbidden lines 69 sources IRS spectra are rich in diagnostic lines/features (e.g., Voit 1992, Spinoglio & Malkan 1993, Genzel et al. 1998, Sturm et al. 2002). (Top) NGC3516, S1.5, weak PAH, slightly red µm continuum. (Bottom) NGC3511, S1, strong PAH, red µm continuum.
[NeV] 14/24 μm vs. [SIII]18/33 μm ● Both line ratios seem to scatter around their low density values (e.g., Sturm et al. 2002). – [NeV] n e < 10 3 – [SIII] n e < few 10 2 ● The data are consistent with Sy 1s and 2s having the same distribution of the two ratios. ● Ratios in the lower left corner would be consistent with some extinction, but uncertainties are too high to confirm. (Left) (Top) Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. [NeV] is a high ionization line (97.1 eV) and should only be detected in sources with an AGN. [SIII] is a lower ionization line (23.3 eV) and can be produced in starbursts. These ratios are a diagnostic of density and/or extinction (Voit 1992; Alexander et al. 1999). The Mann-Whitney U test and KS test show that the Sy 1’s and 2s have a similar distribution of both ratios. There is no correlation between the two ratios. (Bottom) Same plot with errors shown based on the uncertainties in the fits. (Right) Density-sensitive ratios of fine-structure transitions from the ground-state triplets of [Ne V], [Ne III], [S III], and [O III]. Solid lines show an electron temperature of 10,000 K; dotted lines show 20,000 K. Horizontal dashed lines are results for NGC4151 (Alexander et al. 1999).
6.2 μm PAH EQW ● The Sy 2 show a tail to higher values of PAH EQ. ● The distributions of PAH for Sy 1s and 2s are significantly different. ● Thus, the Sy 2s tend to include galaxies with more star formation than the typical Sy 1. The 6.2μm PAH feature is thought to be relatively unconfused (Peeters et al 2004) and so should be a good indicator of star formation. The Mann-Whitney U test shows that the Sy 2s have higher PAH EQW than the Sy 1 with the probability of no difference
6.2 μm PAH EQW vs μm Spectral Index ● There is a correlation between PAH EQW and slope of µm continuum – We see a vertical band of Sy 1s and 2s with alpha ~ 0 and low PAH. – At EQW > 0.5, There is a horizontal band of sources which includes the Starbursts and Liners at high PAH EQW and a range of (mostly red) MIR spectral index. ● There are a few Sy 1s with red MIR spectra but relatively low PAH EQW. These may be: – star forming objects in which the unshielded PAH have been destroyed by UV photons from the AGN. – sources in which the AGN heats the cool dust (e.g., Siebenmorgen et al. 1997) ● In red Seyferts, Mid-IR is dominated by cold dust heated by star formation. Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. The 6.2μm PAH feature is thought to be relatively unconfused (Peeters et al 2004) and so should be a good indicator of star formation. 20 μm is just redward of the “knee” in the SED of some Sy 1s and 2s. Thus, the μm spectral index characterizes the contribution of warm dust to the mid-IR (Buchanan et al. 2006, Brandl et al. 2006, Cleary et al. 2007). A spectral index of zero indicates a large contribution from warm dust. Positive spectral index is a red SED. The probability of a non-correlation is small (Kendall Tau 1.2E-5; Spearman Rank 4.5E-6).
6.2 μm PAH EQW vs. MIPS Spectral Index ● There is a weak correlation between PAH EQW and slope of FIR continuum ● Star formation is common in Seyferts ● In red Seyferts, FIR is dominated by cold dust heated by star formation. ● Sy 1 and 2 have similar distribution of MIPS slope, but different distributions of PAH EQW. Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. The 6.3 μm PAH feature is thought to be relatively unconfused (Peeters et al 2004) and so should be a good indicator of star formation. The spectral index of the MIPS SED (50-90 µm) indicates the relative contributions of warm and cold dust to the SED. Positive spectral index is a red SED. The correlation is 3σ for the Sy 1, and about 2σ for the Sy 2 (Kendall Tau). For the combined Sy 1 and 2 sample the probability of a non- correlation is quite small (Kendall Tau 9.9E-5, Spearman Rank 5E-5).
6.2 μm PAH EQW vs. H 2 S(1) 17 μm EQW ● The H2 S(1) 17 μm line is common in both types of Seyfert Galaxies ● There is a correlation between PAH EQW and Molecular Hydrogen EQW (consistent with the QUEST QSOs Schweitzer et al. 2006). – Presumably the excitation mechanisms are different, so the correlation may indicate the mass of gas undergoing star formation. – Alternately, could be due to UV flux disassociating H 2 and destroying PAH ● The starburst and liners lie to the upper right in the diagram. ● Sy 1 and 2 have similar distribution in the plane suggesting they have a similar distribution of star formation properties. Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. The 6.3 μm PAH feature is thought to be relatively unconfused (Peeters et al 2004) and so should be a good indicator of star formation. The probability of a non-correlation is exceedingly small (Kendall Tau consistent with zero, Spearman Rank 8.8E-10).
20-30 μm Spectral Index vs. [NeV] 24 μm EQW ● The data are consistent with Sy 1s and 2s having the same distribution of [NeV] EQW. ● Sources with low [NeV] EQW show a broad range in mid-IR slope (from flat to very red). ● Sources with high [NeV] EQW tend to show a flatter mid-IR slope. ● This gives an anti-correlation between a star formation indicator (red mid-IR slope) and an AGN indicator (high [NeV] EQW). – This may be partially due to dilution of the [NeV] EQW by the additional red continuum. ● This suggests that in general, star formation, not the AGN, is responsible for heating the cooler dust. Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. [NeV] has a lower ionzation potential of 97.1 eV and should only be seen in objects with an AGN. The anti correlation between slope and [NeV] is 3σ for Sy 1, but not formally significant for Sy 2. For the combined sample of Sy 1 and 2, the probability of a non correlation is about for both Kendall Tau and Spearman Rank.
6.2 μm PAH EQW vs. [NeV] 24 μm EQW ● The Seyferts scatter on this diagram with no correlation. ● This gives a non correlation between a star formation indicator (high PAH) and an AGN indicator (high [NeV] EQW). – A correlation could be masked if the line strengths are proportional to the strength of the continuum. Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. [NeV] has a lower ionzation potential of 97.1 eV and should only be seen in objects with an AGN.
6.2 μm PAH vs. [NeV] 24 μm (Luminosity) ● There is a correlation between the luminosity of the 6.2 PAH μm feature and the luminosity of the [NeV] 24 μm line. There is a lot of scatter at high luminosity. ● This suggests that generally the AGN and Starburst components scale together. ● Thus, there may be a physical AGN-Starburst connection. ● Consistent with the higher luminosity QSOs (Schweitzer et al. 2006) Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. [NeV] has a lower ionzation potential of 97.1 eV and should only be seen in objects with an AGN. There probability of a non correlation is 0.01 (Spearman Rank) and (Kendall Tau).
20-30 μm Spectral Index vs. H2 S(1) 17 μm EQW There is a subset of objects with flat slope but a range of H2 line EQW. There is a second group with red slope and high H2 EQW. Thus, we see a trend for sources with redder continuum slope to have a larger molecular line EQW. This is consistent with both a red mid- IR continuum and the molecular line being signatures of star formation. Sy 1 and 2 have similar distribution in the plane suggesting they have a similar distribution of star formation properties. Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. 20 μm is just redward of the “knee” in the SED of some Sy 1s and 2s. Thus, the μm spectral index characterizes the contribution of warm dust to the mid-IR (Buchanan et al. 2006). A spectral index of zero indicates a large contribution from warm dust. Positive spectral index is a red SED. The probability of a non correlation is very small (Kendall Tau 7.1E-7, Spearman Rank 1.4E-7).
Principal Component Analysis ● Identifies components ● PCA separates shapes well ● PC1 is similar to a starburst Red = Spectrum of M82 Blue = PC1 (Buchanan et al. 2006)
Relation of PC1 to Seyfert type ● The first eigenvector contribution is correlated with the amount of extended IR emission in the galaxy (Buchanan et al. 2006). ● The first eigenvector contribution differs for Sy 1s and Sy 2s. – Sy 2s have more starburst contribution on average (consistent with Maiolino et al ) – selection effect? (Top) Distribution of the relative contribution of eigenvector 1 for all objects, Sy 1’s, and Sy 2’s. The hatched objects are those reclassified as non-Seyferts. (Bottom) Flux deficit = (IRAS-Spitzer)/IRAS. Spitzer slit widths are 3.5 to 10”, while IRAS is arcminutes. Blue square = Sy 1, Red circle = Sy 2, Green triangles = liners, Yellow stars = Starburst. (Buchanan et al. 2006)
Constraints on the Obscuring Torus Annotated by M. Voit Strength of the Si 9.7 µm feature. Anisotropy ratio in mid-IR. Caveats Dust heated by star formation. Emission from dust in the NLR (e.g., Mason et al. 2006) Disk winds (e.g., Kartje et al. 1999).
Introduction to Si 9.7 μm Strength ● S 9.7 = ln (F 9.7 /F cont ) ● Many Sy 1s and 2s have very weak Si strength (e.g., Roche et al. 1991, Shi et al. 2006, Hao et al. 2007). ● Si strength is a constraint on torus models (e.g., Pier & Krolik 1992; Laor & Draine 1993). (Left). Average spectra of QSOs (red), Seyfert 1s (green), Seyfert 2s (blue), and ULIRGs (black).. The 10 μm silicate strengths measured from the average spectra are: 0.20, -0.21, -0.54, and for quasars, Seyfert 1s, Seyfert 2s, and ULIRGs, respectively. (Right) Distribution of the 10 μm silicate strength, S10, for QSOs, Seyfert 1s, Seyfert 2s, and ULIRGs. Silicate absorption increases to the right. In QSOs, Seyfert 1s, and Seyfert 2s, the distributions of sources that are also in the ULIRG sample are shaded. The averages of the 10 μm silicate strengths are 0.20 for quasars, for Seyfert 1s, for Seyfert 2s, and for ULIRGs. (Hao et al. (2007).
6.2 μm PAH EQW vs. Si 9.7 μm Strength ● Our results are consistent with Spoon et al (2007). – Many Sy 1s and 2s have very weak Si strength – The Liners and Starburst galaxies tend to show Si in absorption. ● A subset of the Sy 2 have substantial Si absorption. The distributions of Si strength are significantly different in Sy 1s and 2s. ● There is no correlation between PAH and Si among the Sy 2s. ● Sy 1s show a trend for higher PAH EQW to be associated with Si absorption. (Top) Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. The Kendall Tau test shows that the probability of a non correlation of PAH EQW with Si strength for the Sy 1s is (Bottom) Spoon et al (2007). The two dotted black lines are mixing lines between the spectrum of the deeply obscured nucleus of NGC 4418 and the starburst nuclei of M82 and NGC 7714, respectively. Galaxy types are distinguished by their plotting symbol: Filled circles: ULIRGs and HyLIRGs. Filled triangles: Starburst galaxies. Filled squares: Seyfert galaxies and QSOs. Filled diamonds: Other infrared galaxies. There are two branches: a lower branch possibly with clumpy dust, and a diagonal branch extending to much strong Si absorption which may include smoother dust.
[NeV] 24 μm EQW vs. Si 9.7 μm Strength ● The distributions of [NeV] EQW and [NeV] 14/24 are similar in Sy 1 and 2. ● There is no correlation of [NeV] EQW and [NeV] 14/24 with Si strength. – [NeV] does not seem to be a strong orientation indicator. – [NeV] may be produced predominantly outside the torus, e.g. in the NLR Purple squares = Sy 1, red crosses = Sy 2, green diamond = Sy 3/Liner, blue triangle = Starburst. For the Si feature, positive means emission. [NeV] is a high ionization line and should only be detected in sources with an AGN. There is no correlation between [NeV] EQW or [NeV] 14/24 ratio and Si strength (Kendall Tau). The distributions of [NeV] EQW or [NeV] 14/24 are similar in Sy 1s and 2s (Mann-Whitney U and KS).
8 GHz Radio Nuclear Emission is an Orientation Independent Calorimeter of AGN Power ● Radio emission is unobscured and presumably orientation independent ● Radio luminosity is proportional to the ionizing luminosity of the accretion disk (e.g., Baum & Heckman 1989, Rawlings & Saunders 1991) ● High resolution 8 GHz VLA observations resolve out the contribution from the more diffuse starburst component (Thean et al 2000) ● The 12 μm sample Sy 1 and 2 have similar distributions of nuclear radio flux density (Buchanan et al. 2006). (Top) Radio Power vs. [OIII] emission line luminosity (Xu, Livio, Baum 1999). (Bottom) Distribution of nuclear 8 GHz radio emission (Thean et al 2000) for the 12µm sample. (Top) Sy 1. (Bottom) Sy 2. Hatched objects are those in Buchanan et al (2006).
IR/radio ratio higher in Sy 1’s ● Using a heterogeneous sample of Seyferts, Heckman (1995) found that the ratio 10.6 μm/radio was a factor of ~4 higher in Sy 1’s than Sy 2’s. We repeat that check with our data and find the ratio is about a factor of 5 higher in Sy 1’s than 2’s. ● We also compare the median IR/radio ratio for Sy 1’s and 2’s over the IRS spectra and find a range of ratio which trends to ~2 longer than 15 μm. ● The torus emission is anisotropic by ~2 in the μm region. (Top) Ratio of 10 μm/8 GHz flux density for Sy 1’s (top) and 2’s (bottom). (Middle) Median ratio of IRS/ 8GHz flux density for Sy 1’s and 2’s. (Bottom) Ratio of the two curves from the middle plot showing the relative difference in the ratios for Sy 1s and 2’s. All figures from Buchanan et al. (2006).
Implications for Torus Models ● Smooth density torus models predict large changes in relative brightness as a function of inclination to the torus axis (e.g., Rowan- Robinson & Crawford 1989; Pier & Krolik 1993; Siebenmorgan et al. 2004). ● Clumpy torus models predict much smaller changes in relative brightness as a function of inclination (Nenkova et al. 2002). (Top). Pier & Krolik 1993 ApJ , constant-density model fitted to NGC1068 for various inclinations. The parameters used are: a/h=0.3, where a = inner radius of torus, h = height of torus, vertical optical depth of 1, radial optical depth of 1, and T_eff (inner torus temperature) of 800K. (Bottom). Clumpy torus model from Nenkova et al ApJ 570 L9. The particular model used is TAB-N10q2-sig45p2-D.tv10, i.e.: no direct AGN emission, 10 clumps along the line of side, radial density distribution of clumps going as r -2, 45 ◦ opening angle of the torus, Gaussian vertical distribution of clumps, and optical depth (V) of 10 for individual clumps.
Selection Effects ● The Sy 2 AGN are fainter in the mid-IR presumably due to obscuration from the torus. ● Selection at 12 μm implies – Sy 2s possibly underrepresented in the 12 μm sample relative to Sy Is. – The Sy 2s which are included are likely to have some additional source of IR flux, e.g., star formation. – Eigenvector 1 and 6.2 μm PAH EQW are higher in the Sy 2, but other star formation indicators (e.g., H 2 and μm slope, MIPS slope) are not significantly different. ● Need large unbiased (radio selected?) sample
Conclusions from Spitzer Study of 12 µm Sample ● Starbursts are common and co-exist with both Seyfert types. ● The luminosity of starburst (PAH) and AGN ([NeV]) tracers is correlated. – This suggests there is a “real” AGN-Starburst Connection. ● The relative starburst contribution to the SED can vary greatly between objects but has a similar range in Sy1s and 2s. There are mixed indications as to whether Sy 2s have more star formation than Sy 1s. ● The cooler dust is most likely heated by star formation rather than the AGN. ● SED Shapes are related to Sy type but the variation in the population is dominated by the starburst contribution. – Makes finding obscured AGN tricky, but there are some useful diagnostics including [NeV] and µm slope ● Torus probably complicated (e.g., clumpy) and optically thick out to at least 35 µm – Selection effects likely in the 12μm sample
The End