Infrared Spectral Energy Distributions of Seyfert Galaxies: Spitzer Space Telescope Observations of the 12 µm Sample of Active Galaxies J. F. Gallimore1,2, A. Yzaguirre1,3, J. Jakoboski1,4, M. J. Stevenosky1,5, D. J. Axon6,7, S. A. Baum8, C. L. Buchanan9, M. Elitzur10, M. Elvis11, C. P. O’Dea6, A. Robinson6 draft:January 27, 2010 0 1 0 ABSTRACT 2 n a The mid-infrared spectral energy distributions (SEDs) of 83 active galaxies, mostly J Seyfertgalaxies, selectedfromtheextended12µmsamplearepresented. Thedatawere 7 2 collected using all three instruments, IRAC, IRS, and MIPS, aboard the Spitzer Space Telescope. The IRS data were obtained in spectral mapping mode, and the photometric ] O datafromIRACandIRSwereextractedfrommatched,20(cid:48)(cid:48) diametercircularapertures. C TheMIPSdatawereobtainedinSEDmode, providingverylowresolutionspectroscopy h. (R ∼ 20)between∼ 55and90µminalarger, 20(cid:48)(cid:48) ×30(cid:48)(cid:48) syntheticaperture. Wefurther p presentthedatafromaspectraldecompositionoftheSEDs,includingequivalentwidths - o and fluxes of key emission lines; silicate 10 µm and 18 µm emission and absorption r t strengths; IRACmagnitudes; andmid-farinfraredspectralindices. Finally, weexamine s a the SEDs averaged within optical classifications of activity. We find that the infrared [ SEDs of Seyfert 1s and Seyfert 2s with hidden broad line regions (HBLR, as revealed 1 by spectropolarimetry or other technique) are qualitatively similar, except that Seyfert v 4 1s show silicate emission and HBLR Seyfert 2s show silicate absorption. The infrared 7 SEDs of other classes within the 12 µm sample, including Seyfert 1.8-1.9, non-HBLR 9 4 Seyfert 2 (not yet shown to hide a type 1 nucleus), LINER, and HII galaxies, appear . 1 0 0 1Department of Physics and Astronomy, Bucknell University, Lewisburg, PA 17837 1 : 2Currently on leave at NRAO, 520 Edgemont Rd., Charlottesville, VA 22903 v i 3Department of Physics, California State University, Fullerton, P.O. Box 6866, Fullerton, CA 92834-6866 X r 4Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91190 a 5Franklin Pierce Law Center, Two White Street, Concord, NH 03301 6Department of Physics, Rochester Institute of Technology, 84 Lomb Memorial Drive, Rochester, NY 14623 7School of Mathematical and Physical Sciences, University of Sussex, Falmer, Brighton, BN1 9QH, UK 8Chester F. Carlson Center for Imaging Science, Rochester Institute of Technology, 54 Lomb Memorial Drive, Rochester, NY 14623 9School of Physics, University of Melbourne, Parkville, Victoria, 3010 Australia 10Department of Physics and Astronomy, University of Kentucky, Lexington, KY 40506 11Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138 – 2 – to be dominated by star-formation, as evidenced by blue IRAC colors, strong PAH emission, andstrongfar-infraredcontinuumemission, measuredrelativetomid-infrared continuum emission. Subject headings: galaxies: active — galaxies: Seyfert — galaxies: spiral — infrared: galaxies 1. Introduction Threecomponentsdominatethemid-infraredspectrumofanactivegalaxy(Genzel&Cesarsky 2000): (1) thermal radiation from a dusty, compact medium that surrounds the active nucleus (AGN) and can obscure direct sight-lines to it; (2) PAH features and thermal dust continuum associated with star-formation or perhaps a powerful starburst; and (3) line features arising from molecular, atomic, and ionic species. The dusty medium surrounding the AGN is commonly referred to as “the dusty torus.” Its presence is inferred for many AGNs based, for example, on spectropolarimetry (Antonucci & Miller 1985; Antonucci 1993; Urry & Padovani 1995; Smith et al. 2004), X-ray spectroscopy (Risaliti et al. 2007, 2002), and infrared aperture synthesis measurements (Jaffe et al. 2004; Tristram et al. 2007). Observations further indicate that the dusty medium must be axisymmetric, permitting low extinction sight-lines to type 1 AGNs, where the broad-line region (BLR) is apparent in total intensity (Stokes I) spectra (the “pole-on” view), but high-extinction sight-lines to type 2 AGNs, which show suppressed broad line emission and AGN continuum (the “edge-on” view). Sincethedustytorusre-radiatesincidentAGNcontinuuminthemid-infrared,itsSEDprovides an indirect measure of the AGN luminosity (Nenkova et al. 2008b), important especially for heavily obscured AGNs where other indirect diagnostics may not be available. Mid-infrared fine-structure lines can also constrain the intrinsic shape of the AGN SED, because particular line ratios are sensitive to the shape of the SED but less sensitive to extinction compared to optical/UV lines of species with similar ionization energies (Alexander et al. 2000). The torus SED depends on the geometry and clumpiness of the torus, among other properties (Nenkova et al. 2002, 2008b). For example, smooth, cylindrical tori produce very strong 10 µm silicate (Sil) features, whether in emission when viewed more nearly pole-on, or deep absorption when viewed edge-on (Pier & Krolik 1992). Clumpy tori, somewhat independent of the assumed geometry, instead produce much weaker Sil features, because inter-clump sight-lines can provide a view of hotter dust on the far side of the torus; our view of clumpy tori includes a mix of cold and hot clump surfaces that dilutes the Sil features (Nenkova et al. 2008b). Individual clumps are heated from outside and cannot produce the strong absorption that a centrally heated dust shell can have (Sirocky et al. 2008; Nenkova et al. 2008a; Levenson et al. 2007). The mid-infrared SED is also an important diagnostic of star-formation. Young star clusters – 3 – embedded in GMCs are predicted to produce weaker PAH equivalent width in comparison to older clusters, owing to photodestruction and continuum dilution by hot dust grains (Efstathiou et al. 2000). In global models, PAH features are sensitive to the fractional luminosity of OB associations and the gas density of their surrounding ISM (Siebenmorgen & Kru¨gel 2007). Since the dusty medium re-radiates incident stellar radiation, the mid-far infrared luminosity constrains the luminous contribution of star-formation. The mid-infrared SED of active galaxies therefore contains diagnostics of the AGN, its sur- rounding dusty torus, and any surrounding star-formation. The described tracers can, in principle, be disentangled by spectral decomposition of the global SED (Farrah et al. 2003; Marshall et al. 2007). At the present, available infrared instruments, at least in their default mode of use, suffer frommismatchedaperturesleadingtodiscontinuitiesintheobservedSEDsormismatchedcoverage of spectral lines that bias line ratios. To remedy this problem of aperture matching, we have used the Spitzer Space Telescope to ob- serve a sample of active galaxies, mostly Seyferts, in synthetically matched, 20(cid:48)(cid:48) diameter apertures spanning λ3.6-36 µm and a larger aperture, ∼ 20(cid:48)(cid:48)×30(cid:48)(cid:48), covering λ55-90 µm. These moderate to low resolution SEDs are well-suited for spectral decomposition, studies of the PAH and Sil features, as well as global constraints on the AGN and star-formation contributions to the luminosity. The observations and data obtained for this survey are presented in this work. Section 2 describes the sample selection. Section 3 provides a detailed account of the observations and data reduction, with attention to artifact correction, synthetic aperture matching, and corrections for extendedemission. Section4summarizestheextractionofspectralfeaturesusingamodifiedversion of the PAHFIT tool (Smith et al. 2007b) and measurements based on the line-subtracted SEDs. We conclude in Section 5 with a summary of the properties of the sample and avenues for future analysis. 2. Sample Selection Thesample,listedinTable1,comprisesasubsetoftheextended12µmsampleofAGNs(Rush et al. 1993). This parent sample was defined by an IRAS detection at 12 µm, F (IRAS) > 0.22 Jy, 12 and color selection F (IRAS) > 0.5F (IRAS) or F (IRAS) > F (IRAS) to remove stars but 60 12 100 12 few galaxies. AGNs were identified based on prior (usually optical) classification. We restricted the sample to include (1) only those objects categorized by Rush et al. (1993) as Seyferts or LINERs, and (2) only those sources with cz < 10,000 km s−1. Three sources were removed subsequent to observations owing either to pointing errors or saturation in the Spitzer observations: NGC 1068, M-3-34-64, and NGC 4922. Our sample ultimately includes 83 Seyfert and LINER nuclei. The main advantage of this sample over other AGN catalogs is its large collection of published and archival multiwavelength observations (Rush et al. 1996; Hunt & Malkan 1999; Hunt et al. 1999; Thean et al. 2001, 2000; Gallimore et al. 2006; Spinoglio et al. 2002). In addition, there are – 4 – comparable numbers of Seyfert 1 (S1) and Seyfert 2 (S2) nuclei, and their redshift distributions are statistically indistinguishable (Thean et al. 2001, 2000). TheoriginalsurveypaperofRushetal.(1993)broadlyassignedopticalclassificationsoftype1 (broad-lineAGN)andtype2(narrow-lineAGN),butTran(2003)pointedoutthat,evengivensuch coarse binning of activity type, there were many misclassifications. To aid in a more sophisticated comparison of infrared properties to optical classifications, we endeavored to collect updated and more precise classifications from the literature. The revised classifications are included in Table 1 and illustrated in Figure 1. We find that 20% (7 / 35) of the Rush et al. Type 1s are re-classified as hidden broad line region (HBLR) Seyfert 2s (S1h or S1i), LINER, or HII (star-forming galaxy), and 28% (13 / 46) of the Rush et al. Type 2s are re-classified as S1.n, LINER, or HII. Note that HBLRs have been sought in all but three of the 20 S2s in our survey: NGC 1125, E33-G2, and NGC 4968 (Tran 2003). 3. Observations and Data Reduction The sample galaxies were observed using all of the instruments of the Spitzer Space Telescope (Program ID 3269, Gallimore, P.I.): the four broadband channels (3.6 µm, 4.5 µm, 5.8 µm, and 8.0 µm) of the Infrared Array Camera (IRAC; Fazio et al. 2004); the low resolution gratings of the Infrared Spectrograph (IRS; Houck et al. 2004), operating in spectral mapping mode; and the Multiband Imaging Photometer for Spitzer (MIPS; Rieke et al. 2004), operating in SED mode. The resulting SEDs are provided in Figure 2. Several sample galaxies were observed as part of other Spitzer programs that made use of different observing strategies; for example, in some cases only single-pointing (“Staring Mode”) IRS spectra are available. We summarize the observations and data reduction techniques both for our observing program and archival Spitzer data. 3.1. Spitzer IRAC Observations IRAC observations were centered on the NED coordinates of the sample galaxies based on catalog names listed in Rush et al. (1993). A beamsplitter and supporting optics centers the target on two detectors simultaneously (Fazio et al. 2004), either at 3.6 and 5.8 µm or 4.5 and 8.0 µm, and so each observation consisted of two pointings at a common orientation of the focal plane relative to sky. These observations were taken as snapshots with no attempt to mosaic or dither. To guard against saturation, we used the high-dynamic range (HDR) mode which provides 0.6 s and 12 s integrations at each pointing. The data were initially processed and calibrated by the IRAC basic calibration data (BCD) pipeline, version S14.0. The pipeline performs basic processing tasks, including bias and dark – 5 – current subtraction; response linearization for pixels near saturation; flat-fielding based on obser- vationsofhigh-zodiacalbackgroundregions;saturationandcosmic-rayflagging,thelatterindicated by signal detections more compact than the PSF; and finally flux calibration based on observations of standard stars1. The nominal photometric stability for compact sources is better than 3% in all detectors (Reach et al. 2005). Corrections for extended sources2 are accurate to ∼ 10%, but the contribution of extended emission and the resulting correction is small for most of the sam- ple galaxies. Color-corrections are typically much greater, especially in the 8.0 µm channel where in-band PAH emission can result in factors of two or greater corrections. The data were further processed for remaining artifacts, including cosmic-rays and detector artifacts not corrected by the BCD pipeline. The steps performed for artifact removal and photo- metric extraction are detailed below in Sections 3.1.1–3.1.4. Figure 3 illustrates the effect of our artifact removal techniques. The IRAC photometry is listed in Table 2. Detailed below, the photometry is presented in the IRAC magnitude system with zero-point flux densities 280.9, 179.7, 115.0, and 64.13 Jy for the 3.6, 4.5, 5.8, and 8.0 µm channels, respectively (Reach et al. 2005). The photometry includes corrections for extended emission and color-corrections. 3.1.1. Cosmic-ray and bandwidth-effect mitigation All of the images were affected to varying degrees by cosmic-ray and solar proton hits. More- over, 5.8 and 8.0 µm images near saturation suffered from the bandwidth effect, which manifests as a row-wise trail of fading source images repeating every four pixels from the affected source. These multiple images do not conserve flux but artificially add signal to the sources; for our purposes, these bandwidth effect artifacts behave like cosmic ray hits near our target galaxies. TheBCDpipelineattemptstoidentifycosmic-rayhitsbylocatingdetectionsthatarenarrower than the PSF. Extended cosmic ray tracks are however missed by this procedure. The v. S14.0 BCD pipeline further has no means of mitigating the bandwidth effect. We corrected these artifacts in three, post-BCD steps. Firstly, we generated difference maps between the long and short exposure images. Difference signal that exceeded 5σ on the long exposure images was flagged as an artifact, and these artifacts were replaced with signal from the short exposure image. The matching mask and uncertainty images were updated accordingly. This techniquewasparticularlysuccessfulatremovingbandwidthartifactswithlittleimagedegradation; the affected image regions are at relatively high signal-to-noise on the short exposure images. WenextappliedvanDokkum’s(2001)algorithmtoflagcosmicraysnotpickedupondifference 1Details are provided in the IRAC Data Handbook, available at http://ssc.spitzer.caltech.edu/irac/dh/. 2http://ssc.spitzer.caltech.edu/irac/calib – 6 – images. ImagesareconvolvedwithaLaplacianfilter, whichenhancessharpedgesonimagefeatures and effectively identifies cosmic ray tracks. The IRAC point response function (PRF) is however undersampled in all four detectors, and real, compact sources such as field stars or the AGN appear as false positives. These false positives can be filtered by measuring the asymmetry of the detected source, parameter f in van Dokkum’s notation. Cosmic ray hits tend to be more asymmetric lim than the PRF, corresponding to a larger value of f . Using a few sample images and trial-and- lim error, we determined lower threshold values for f that flag obvious cosmic rays but pass field lim stars and galaxy images; specifically, we found f = 8 worked well for the 4.5, 5.8, and 8.0 µm lim images, and f = 10 for the 3.6 µm images. lim Finally, the few remaining cosmic rays were flagged interactively by inspection of four-color images. Residual cosmic rays appear as color-saturated pixels in this representation and so were easily identified, flagged, and replaced by bilinear interpolation of neighboring pixels. 3.1.2. Bias artifacts Residual bias artifacts affect the pipeline-processed data. Software is currently available at the Spitzer Science Center to mitigate these artifacts where the image comprises mainly compact sources, but the algorithm breaks down in the presence of extended or diffuse emission. The host galaxy is detected for many of the sample AGNs, and so we had to develop new techniques to eliminate these bias artifacts. The 3.6 and 4.5 µm images show the effects of column pull-down and multiplexer bleed (“muxbleed”). Column pull-down is evident as a depressed bias level along columns that run through pixels near saturation. The bias adjustment is nearly constant along a column, but there may be slightly different bias offsets above and below saturated pixels. One approach is to evaluate the bias depression in source-free regions, but, for some sources, the presence of extended emission over a majority fraction of the array reduces or eliminates valid background regions. We suppressed the diffuse emission by performing row-wise median filtering across pull-down columns and used the median difference to determine the bias correction. Muxbleedalsoaffectsrowscontainingpixelsnearsaturation,mostevidentasarowpull-up,but also impacting the bias level on neighboring rows. The images are read as four separate readout channels that are interlaced every four columns. The bias on each readout channel is affected differently, resulting in a vertical pinstripe pattern over some region of the array near saturated pixels. Tomitigatethemuxbleedartifactinthepresenceofextendedemission,wefirstmedian-filtered the image using an 8 × 8 kernel to smooth out the pinstripe pattern over two readout cycles, subtracted the median-filtered image to remove diffuse emission, and generated residual images of each readout channel. Muxbleed appears on each channel readout image as decaying, horizontal – 7 – stripes along rows containing pixels near saturation with surrounding bands of weaker, constant bias offset. We determined a final muxbleed model by fitting the brighter muxbleed stripes with a cubic polynomial and determining the median DC-level offset in the surrounding bands. The5.8µmimageswereaffectedbyresidualdarkcurrent,appearingasaslowlyvaryingsurface brightness gradient of the background. This “first frame” artifact results from the sensitivity of the dark current to properties of the previous observation and the time elapsed since that observation. Toremovethisdarkcurrentgradient,wefirstmaskedbrightstarsanddiffuseemissionfromgalaxies and then fit a bilinear surface brightness model to the background. Several 5.8 and 8.0 µm images were also affected by banding, which is a decaying signal along rowsorcolumnscontainingsaturatedpixels. Thesebandswerereducedbyfittingseparaterow-wise and column-wise polynomials to the surface brightness of off-source regions of the array. 3.1.3. Saturation and Distortion Final corrections were performed using custom IDL scripts and the MOPEX software package3 (Makovoz & Khan 2005; Makovoz et al. 2006). The short exposure images were used to replace saturated pixels on the long-exposure images. The images were then corrected for distortion and registered to a 1(cid:48).(cid:48)22 × 1(cid:48).(cid:48)22 grid. The resulting data products are the science image, calibrated in surface brightness units MJy sr−1; an uncertainty image based on the propagation of statistical uncertaintiesthroughthepipelineandpost-BCDprocessing, butwhichdoesnotincludesystematic uncertaintiesassociatedwithcalibration;andacoverageimage,which,forthesesnapshotexposures, marks good pixels as “1,” bad pixels (i.e., known bad pixels or cosmic ray hits) as “0,” and intermediate values indicating that good and bad pixels were used in the distortion correction for that pixel. 3.1.4. Photometric extraction We extracted flux density measurements using a synthetic, 20(cid:48)(cid:48) diameter circular aperture centered on the brightest infrared source associated with the active galaxy. The exception is NGC 1097, which has an off-nucleus star-forming region that is brighter than the central point source; in that case, the aperture was centered on the central point source. To determine the point- source contribution to the aperture, each image was convolved with a 2-D, rotationally symmetric Ricker wavelet (Gonz´alez-Nuevo et al. 2006). The width of the central peak of the wavelet was tuned to match the width of the nominal IRAC PRF, effectively subtracting extended emission and enhancing point sources. We used field stars to calibrate the point source response of the 3http://ssc.spitzer.caltech.edu/postbcd/mopex.html – 8 – wavelet-convolved image; this calibration includes a correction for aperture losses. We next applied extended source corrections4 to the residual signal (total − point source). Finally, the resulting photometry was color-corrected following the prescription described in the IRAC Data Handbook. The uncorrected calibration provides a flux density measurement at a nominal wavelength assuming a nominal spectral shape νF = constant over the broadband ν response. The color-correction adjusts the calibration based on the true shape of the spectrum. The result is the true flux density at a nominal wavelength rather than a broadband average; for example, the color-corrected flux reported for the 8 µm camera will be closer to peak of the 7.7 µm PAH feature plus any underlying continuum at that wavelength rather than bandpass-weighted average. For reference, the nominal wavelengths for the IRAC cameras are 3.550, 4.493, 5.731, and 7.872 µm. Color-correction involves integrating the infrared spectrum weighted by the IRAC filter band- passes. Toperformthisintegration,weusedourIRSspectra(Section3.3,below)withextrapolation to shorter wavelengths assuming a power law slope matching the observed, uncorrected 3.6 µm flux density. Note that this color-correction technique does not force a match between the IRS and IRAC photometry; the flux scale of the spectrum is normalized so that only the shape of the IRS spectrum influences the color correction. Photometric corrections for extended sources are not ac- curately known for the IRS spectra, and so the accuracy of the IRS spectral extractions precluded separating the color corrections for extended and point source contributions for a given source; rather, the color correction was determined based on the integrated signal in the aperture. 3.2. Archival Spitzer IRAC Observations IRAC observations of 16 sample galaxies were obtained through public release to the Spitzer archive. Data processing largely followed the techniques described above for our observations, except that, where our observations comprise snapshots, the archival observations all employed dithering or mosaicking techniques. The archival data were initially mosaicked using standard procedures with MOPEX, including background matching based on overlaps. Artifacts, particularly banding, seriously affected the matching of mosaic overlap regions and so we developed a modified mosaicking procedure building on the algorithm of Regan & Gruendl (1995). We modified their least-squares approach to employ least absolute deviations as a more robust measure of the quality of overlap matching, and we further masked bright source and arti- fact regions prior to overlap matching. The final images were assembled using a median stack of astrometrically alignedimages, backgroundsubtracted, and finallysub-imaged toroughly 5(cid:48) square to match our survey observations. 4http://ssc.spitzer.caltech.edu/irac/calib/ – 9 – Residual bias artifacts were removed by computing column or row biases where sufficient backgroundwasavailableforbiasdetermination. Cosmicrayswerelargelymitigatedbythemedian stack, butresidualbadpixelswereidentifiedandflaggedinteractively, asdescribedinSection3.1.1. 3.3. Spitzer IRS Spectral Mapping We observed the sample galaxies using the Spitzer IRS in spectral mapping mode and using the Short-Low (SL) and Long-Low (LL) modules. The modules cover a wavelength range of ∼ 5 – 36 µm with resolution R = λ/∆λ ranging from 64 to 128. The integration times were 6 seconds per slit pointing. The spectral maps were constructed to span > 20(cid:48)(cid:48) in the cross-slit direction and centered on the target source. The observations were stepped perpendicular to the slit by half slit-width spacings. For the SL data, the mapping involved 13 observations stepped by 1(cid:48).(cid:48)8 perpendicular to the slit, and, for the LL data, 5 observations stepped by 5(cid:48).(cid:48)25 perpendicular to the slit. The resultingspectralcubesspanroughly25(cid:48).(cid:48)2×54(cid:48).(cid:48)6×(5.3–14.2µm)fortheSLdataand29(cid:48).(cid:48)1×151(cid:48)(cid:48)× (14.2 – 36 µm) for the LL data. The raw data were processed through the Spitzer BCD pipeline, version S15.3.0. The pipeline handled primary processing tasks including identification of saturated pixels; detection of cosmic ray hits; correction for “droop,” in which charge stored in an individual pixel is affected by the total flux received by the detector array; dark current subtraction; and flat-fielding and response linearization. Details are provided in the IRS Data Handbook5. Skysubtractionusedoff-sourceorders. InSLandLLobservations,thesourceiscenteredinthe first or second order separately, with the “off-order” observing the sky at a position offset parallel to the slit: 79(cid:48)(cid:48) away for SL and 192(cid:48)(cid:48) away for LL. Sky frames were constructed using median combinations of the off-source data and subtracted from the on-source data of matching order. No detectable contamination of the sky frame appeared in our observations based on inspection of the sky frame data. In addition, a portion of the first order slit spectrum appears on the second order observation and is used as an additional check of spectral features that appear near the first / second order spectral boundary. Data cubes were constructed by first registering the individual, single-slit spectra onto a uni- form grid (slit-position vs. wavelength). The IRS slits do not align with the detector grid, and the detector pixels undersample the spatial resolution. Interpolating the undersampled slit onto a uniform grid produces resampling noise (Allington-Smith et al. 1989), which significantly impacts the spectra of unresolved sources. We instead used the pixel re-gridding algorithm described by Smith et al. (2007a) that minimizes resampling noise. Pixels from the original image are effectively 5http://ssc.spitzer.caltech.edu/mips/dh – 10 – placed atop the new, registered pixel grid. Uniform surface brightness is assumed across the origi- nal pixel, and that original signal is weighted and distributed based on the fractional overlap with pixels on the new grid. After registering the single-slit frames, the data were re-gridded by bilinear interpolation, one wavelength plane at a time, into the final data cube. Flux uncertainty images were similarly processed, with modifications to accommodate variance propagation, to produce an uncertainty cube. The spectra were extracted from synthetic, 20(cid:48)(cid:48) diameter circular apertures centered on the brightest, compact IR source nearest the target coordinates. Fractional pixels at the edge of the aperture were accounted for by assuming uniform surface brightness across the pixel and weighting by the area of intersection between the pixel and aperture mask. The cube-extracted spectra were optimally weighted based on a three-dimensional modifica- tion of Horne’s (1986) two-dimensional slit extraction algorithm. Successful application of this algorithm requires estimation of spatial profiles P , the probability that a detected photon falls xλ in a given pixel x rather than some other pixel on the spectral map at wavelength λ. Optimal extraction requires that the fractional uncertainty of the estimator for P is less than the frac- xλ tional uncertainty of original spectral image. Generation of P therefore requires some smoothing xλ along the wavelength axis. To help preserve real variations of P with wavelength, we employed xλ Savitzky-Golay (1964) polynomial smoothing. In this smoothing scheme, P is calculated by a xλ polynomial fit to the spectrum at pixel x over the fitting window δλ either centered on λ or limited by the ends of the spectrum. We performed trial-and-error smoothing experiments on CGCG381- 051, which shows relatively weak continuum at short wavelengths but high eqw PAH features, to decideonthepolynomialorderandwindowsmoothingparameters;ultimatelyweselectedquadratic polynomials and 5-pixel smoothing windows to balance improved signal-to-noise and the tracking of real variations of P with wavelength. xλ The effect of optimal extraction is illustrated in Figures 4 and 5, which compare the optimally- extracted and non-weighted extraction of the IRS SL spectrum of NGC 3079 and F01475-0740, and Table 3, which compares the measurements derived from these extractions (see Section 4.1 for a description of the measurement technique). NGC 3079 presents a challenging case because it is edge-on and shows bright, extended PAH emission. The spatial profile is therefore a complex function of wavelength, and coarse smoothing in the wavelength direction could potentially affect the measurement of the PAH fluxes and equivalent widths. We find however that the fractional difference between the optimally-weighted spectrum and the unweighted spectrum is typically only a few %, comparable to the statistical uncertainties in the unweighted spectrum. Furthermore, the measured fluxes and equivalent widths agree to within the measurement uncertainties; in fact, the measurement uncertainties are dominated by the systematics of the measurement technique. Compared to NGC 3079, F01475-0740 is a compact source at relatively low signal-to-noise. Figure 5 clearly illustrates the advantage of optimum weighting. The continuum shape and PAH spectral features are preserved, but the formal statistical uncertainties are reduced by factor of