The deepest image of the Universe at a wavelength of 15 microns
Rosalind Hopwood, Stephen Serjeant, Mattia Negrello, Chris Pearson, Eiichi Egami, Myungshin Im, Jean-Paul Kneib, Jongwan Ko, Ian Smail
aa r X i v : . [ a s t r o - ph . C O ] D ec **FULL TITLE**ASP Conference Series, Vol. **VOLUME**, **YEAR OF PUBLICATION****NAMES OF EDITORS** The Deepest Image of the Universe at aWavelength of 15 microns
Rosalind Hopwood, Stephen Serjeant, Mattia Negrello, ChrisPearson, , , Eiichi Egami, Myungshin Im, Jean-Paul Kneib, Jongwan Ko, Ian Smail Department of Physics & Astronomy, The Open University, UK Space Science & Technology Department, CCLRC RutherfordAppleton Laboratory, UK Department of Physics, University of Lethbridge, Canada Department of Astronomy, The University of Arizona, USA Department of Physics & Astronomy, FPRD, Seoul NationalUniversity, Korea OAMP, Laboratoire d’Astrophysique de Marseille, France Institute for Computational Cosmology, Durham University, UK
Abstract.
We present photometry, photometric redshifts and extra galacticnumber counts for ultra deep 15 micron mapping of the gravitational lensingcluster Abell 2218 (A2218), which is the deepest image taken by any facilityat this wavelength. This data resolves the cosmic infrared background (CIRB)beyond the 80% that blank field
AKARI surveys aim to achieve. To gain anunderstanding of galaxy formation and evolution over the age of the Universe anecessary step is to fully resolve the CIRB, which represents the dust-shroudedcosmic star formation history. Observing through A2218 gives magnifications ofup to a factor of 10, thus allowing the sampling of a more representative spreadof high redshift galaxies, which comprise the bulk of the CIRB. 19 pointedobservations were taken by
AKARI ’s IRC MIR-L channel, and a final combinedimage with an area of 122.3 square arcminutes and effective integration time of8460 seconds was achieved. The 5 σ sensitivity limit is estimated at 41.7 µ Jy.An initial 5 σ catalogue of 565 sources was extracted giving 39 beams per source,which shows the image is confusion limited. Our 15 micron number countsshow strong evolution consistent with galaxy evolution models that incorporatedownsizing in star formation.
1. Data Reduction
Our data consists of 19 pointed observation of A2218, taken with the 15 micronfilter of the IRC-L (Onaka et al. 2007) aboard
AKARI (Murakami et al. 2007).The data was reduced using the standard IRC pipeline, version 20070912. Thepipeline’s median sky subtraction was utilized but resulted in dark areas, signif-icantly around the brighter sources. To deal with this issue, and the remnantsof scattered light persisting post-pipeline, a further median sky subtraction ofthe background areas was performed. Hot pixels were masked and removed, andthe remaining bad pixels were addressed using an IDL sigma filtering routine.Figure 1 shows four corresponding postage stamp sections taken from frame 191
Hopwood et al. illustrating the post-pipeline output, a median filtered mask, the median sub-tracted result and the sigma filtered result. The resulting 19 images were alignedwith Aladin and IDL’s HASTROM, and then average combined to give the finalL15 image.
Figure 1. Postage stamp image sections showing, from left to right, anexample of the post pipeline output, median filtered mask, median subtractedimage and sigma filtered image.
A 5 σ source extraction was performed with DAOFIND (Stetson 1987). Theresulting catalogue was eyeballed to eliminate any spurious detections, giving afinal number of 565 sources. A Monte Carlo completeness test was performedusing an IDL routine written to convolve a normalized empirical PSF with thefinal L15 image, to create artificial sources at random positions. The artificialsources were placed sufficiently apart from one another and known sources toavoid self-confusion. The test was performed for 80 flux bins, covering the rangeof flux densities for detected sources within the L15 image. In order to reducestatistical errors the test was repeated until an effective 18452 sources per binwas achieved. The completeness test results show the L15 image is 50% completeto 30.7 µ Jy and 80% complete to 39.4 µ Jy.
Aperture photometry was taken at random positions on the final image, exclud-ing the edges and the brightest source, and the results were plotted as a his-togram of flux density against number, see Figure 2. The resulting distributionhas an asymmetric tail that signifies the contributions from bright sources. Thecombined confusion and detector noise can be represented by fitting a Gaussianto the histogram, with a standard deviation of 8.33 µ Jy giving a 5 σ sensitivityestimate of 41.67 µ Jy. The mean rms per pixel of the L15 noise map is 2.20 µ Jy.
Figure 2. Distribution of random aperture photometry of the L15 image,fitted with a Gaussian of standard deviation 8.33 µ Jy. eepest image at a wavelength of 15 microns
2. Band-merged catalogue
Multi waveband data of A2218 was used to identify counterparts of the L155 σ source catalogue. HST WFP2 F450, F606 and F814, Palomar 200inch Hale U, v, b, i and INGRID WFC Ks and J images of A2218 were provided byIan Smail. Spitzer IRAC Ch 1 to 4 data was obtained via Leopard and com-bined with Mopex. An AKARI
S11 image was provided by Myungshin Im andJongwan Ko, and a Spitzer MIPS 24 µ m image was provided by Eiichi Egami.The counterparting procedure identified an extra 368 sources and 3 spurious 5 σ detections, giving a combined counterpart catalogue of 930 sources.
3. Photometry
Aperture photometry of the 5 σ source catalogue was taken with PHOT (Stetson1987) using an aperture of 5.96 ′′ and a sky annulus of radii 19.07 ′′ and 31.0 ′′ . Anaperture correction of 1.30, derived using a growth curve correction method, wasapplied and the IRC data user’s manual version 1.4 (Lorente et al. 2008) conver-sion factor of 1.691 was used to convert from ADU to µ Jy. For the HST images,photometry was obtained from the published catalogue of Smail et al. (2001) viathe NED database. The remaining images were subject to a growth curve cor-rection method to obtain corrected aperture photometry, using a routine writtenin IDL. For each image one or two mean growth curves were empirically con-structed. These curves were used to calculate aperture radius and correction foreach remaining source. IDL’s APER was used to take the subsequent aperturephotometry. Our Ks band photometry was compared, where available, to previ-ously published Ks band photometry Smail et al. (2001) and showed a less than2% difference.
4. Photometric Redshifts
We used EaZy (Brammer, van Dokkum & Coppi 2008) to gain photometric red-shift estimates for our 5 σ source catalogue. EaZy utilizes a minimum χ SEDfitting method, which is suitable for data sets with no available spectroscopicredshifts (Zspec) or a biased set of Zspec. In our case the majority of the smallZspec available are biased at the cluster distance. The theoretical SED tem-plates, used by EaZy, are based on semi-analytical models, and a linear combi-nation of templates can be fitted simultaneously. EaZy gives the option to applypriors, aimed at breaking the template colour degeneracies seen with increasingredshift. A comparison of the resulting redshift estimates for our sources withknown Zspec shows that applying priors gives an improved correlation of ap-proximately 10%. Our spectra were also fitted using the photometric redshiftcode illustrated in Negrello et al. (2009), Photz from here on. This code isuniquely optimised for fitting mid-to-far-infrared PAH and silicate features seenin starburst SEDs. Starburst template (Takagi et al. 2003) and AGN template(Efstathiou & Rowan-Robinson 1995) components were simultaneously fitted byphotz. A comparison of the Photz and EaZy redshifts estimates for sources withprominent mid-to-far-infrared features shows a correlation of around 0.8.
Hopwood et al.
5. Number Counts
Differential number counts (dN/dS), corrected for incompleteness, were takenfor our 5 σ catalogue and normalized to a Euclidean slope. These counts werecorrected for flux amplification by applying magnification factors obtained viaLENSTOOL (Jullo et al. 2007). Figure 3 compares our corrected and uncor-rected counts with previously published differential number counts. Our countscorrected for lensing show an upturn around 2 mJy and peak around 0.4 mJy, inagreement with previous counts (e.g., Elbaz et al. 1999) and the Pearson et al.(2007); Pearson, C. (2009) model. Figure 3. Euclidean normalized differential number count comparison.
Acknowledgments.
We thank the Great Britain Sasakawa Foundation forsupport with grant number 3108. This research is based on observations with
AKARI , a JAXA project with the participation of ESA.