AKARI Far-Infrared Source Counts in the Lockman Hole
Shuji Matsuura, Mai Shirahata, Mitsunobu Kawada, Yasuo Doi, Takao Nakagawa, Hiroshi Shibai, Chris P. Pearson, Toshinobu Takagi, Woong-Seob Jeong, Shinki Oyabu, Hideo Matsuhara
aa r X i v : . [ a s t r o - ph ] J un AKARI Far-Infrared Source Counts in the LockmanHole
Shuji
Matsuura Mai
Shirahata Mitsunobu
Kawada Yasuo
Doi Takao
Nakagawa Hiroshi
Shibai Chris P.
Pearson Toshinobu
Takagi Woong-Seob
Jeong Shinki
Oyabu and Hideo Matsuhara Department of Infrared Astrophysics, Institute of Space and Astronautical Science (ISAS), JapanAerospace Exploration Agency (JAXA) 3-1-1, Yoshinodai, Sagamihara, Kanagawa 229-8510, [email protected] Graduate School of Sciences, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan Department of General System Studies, Graduate School of Arts and Sciences, The University ofTokyo, 3-8-1 Komaba, Meguro-ku, Tokyo 153-8902, Japan (Received 2007 April 11; accepted 2007 0)
Abstract
We report initial results of far-infrared observations of the Lockman hole withFar-Infrared Surveyor (FIS) onboard the AKARI infrared satellite. On the basis ofslow scan observations of a 0.6 deg × µ m down to 77, 26 and 194 mJy (3 σ ), respectively.The counts at 65 and 140 µ m show good agreement with the Spitzer results. However,our 90 µ m counts are clearly lower than the predicted counts by recent evolutionarymodels that fit the Spitzer counts in all the MIPS bands. Our 90 µ m counts above26 mJy account for about 7% of the cosmic background. These results provide strongconstraints on the evolutionary scenario and suggest that the current models mayrequire modifications. Key words: cosmology:observation - galaxies:evolution - galaxies:starburst - in-frared:galaxies
1. Introduction
One of the main scientific objectives of AKARI (Murakami et al. 2007) is to investigatethe history of galaxy evolution by measuring thermal emissions from interstellar dust heated bythe UV light from stars. Far-Infrared Surveyor (FIS) onboard AKARI with four photometricbands of 65, 90, 140 and 160 µ m (Kawada et al. 2007) is designed to detect such dusty objectsat wavelengths near the peak of the dust emission. Mapping observations with AKARI in1he slow scan mode provide 1–2 orders of magnitude higher sensitivity than that in the allsky survey mode with fast scan, and they are suitable for probing distant luminous infraredgalaxies.The IRAS mission has discovered the presence of luminous infrared galaxies in localuniverse (Neugebauer et al. 1984), and their number counts show an excess over the predictedcounts for the non-evolution scenario (Hacking & Soifer 1991). Far-infrared deep surveys withInfrared Space Observatory (ISO) have confirmed the findings by IRAS at much deeper fluxlevels; faint galaxy counts at 90 and 170 µ m show steep increase of the excess counts as fluxlevels become fainter (Kawara et al. 1998, Puget et al. 1999, Efstathiou et al. 2000, Linden-Vornle et al. 2000, Matsuhara et al. 2000, Dole et al. 2001, Rodighiero et al. 2003, Kawara etal. 2004). These ISO results have inspired us with insight for the link between local luminousinfrared galaxies and galaxy formation in early epoch, and they have provided useful constraintson the galaxy evolution models. The Spitzer Space Telescope (Werner et al. 2004) with theMultiband Imaging Photometer for Spitzer (MIPS) (Rieke et al. 2004) provided the numbercounts at 24, 70 and 160 µ m down to 1–2 orders of magnitude deeper fluxes than previouslyreached and strongly constrained the galaxy evolution models (Papovich et al. 2004, Dole etal. 2004, Frayer et al. 2006a). The Spitzer results in all MIPS bands are well fitted withphenomenological models that predict most of the faint galaxies lie at redshifts between 0.7and 0.9 (Dole et al. 2003, Lagache et al. 2003, 2004). Their models reproduce a broad shapeof the measured redshift distribution for the Spitzer 24 µ m samples, but the models still showdiscrepancies in some details at high redshifts z > . µ m, where IRAS and ISO data at faint fluxlevels below 100 mJy are not accurate enough to provide tight constraints on the evolutionarymodels.Recent advances in observational study in wide wavelength range from optical tomillimeter-wave have shown that luminous infrared galaxies are minority in local universe butmajor building blocks of the high redshift universe in terms of the total radiation energy release(Blain et al. 2002). Unresolved faint galaxies at high redshifts would form the Cosmic InfraredBackground (CIB). In fact, CIB in the far-infrared regime measured with COBE (COsmicBackground Explorer) accounts about half of the total energy of optical/infrared background(Puget et al. 1996, Hauser et al. 1998, Lagache et al. 1999, Hauser & Dwek 2001, Finkbeineret al. 2002). That tells us the importance of exploration of far-infrared sources which areresponsible for the energy release in the cosmic history. Recent Spitzer results of ultra-deep2alaxy counts at 70 µ m (Frayer et al. 2006b) and of the stacking analysis for faint sources at70 and 160 µ m accompanied with the 24 µ m sources (Dole et al. 2006) show that more than ∼
70% of CIB has been resolved into point sources. However, previous galaxy counts at thepeak of the measured CIB ( ∼ µ m) in between the MIPS bands have not reached to suchdeep levels.In this work, we performed far-infrared deep surveys with AKARI in the Lockman hole,in order to measure the source counts at previously unexplored sensitivity near the photometricbands covering the peak wavelength of CIB. This survey has been done in the performanceverification (PV) phase of AKARI mission to demonstrate and evaluate performance of theslow scan observation. The survey area was, therefore, limited to 0.7 deg , and east patchof the Lockman hole (LHEX field), where many observations including follow-up observationsfrom the ground for the ISO survey (Oyabu et al. 2005, Rodighiero et al. 2005) have been done,was selected for the observation field so that the cross calibration between missions could bedone. In this paper, we present the initial results of the number counts for the sources detectedwith AKARI.
2. Observations
The observations of the Lockman hole were carried out in the PV phase in 2006 May,to check the performance of the FIS instrument for observing faint objects and to demonstratethe slow-scan observations for wide-area mapping. The survey covers a 0.6 deg × ) region centered at RA(J2000) = 10 h m s and DEC(J2000) = +57 ◦ ′ ′′ , the east sideof the lowest cirrus region corresponding to the ISO deep survey.The data were taken using the FIS-02 slow-scan observational mode of the astronomicalobservation template (AOT) with 15-arcsec/s scan rate and 2-s period reset (Kawada et al.2007). The FIS-02 observation provides a map data with double sightings of each source in a0.13-deg wide, 1.25-deg long strip area by a single turn-around scan. The mapping observationwas carried out with 11 contiguous scans; 12 scans were actually done but one scan data werelost with trouble of the ground station. To secure sufficient redundancy and to improve thesensitivity, one third of the width of each single scan strip was ovelapped with the next strip forproducing mosaic image. It took 2 hours of telescope time in total to cover the entire surveyfield. The data in the four photometric bands, centered at 65 µ m, 90 µ m, 140 µ m and 160 µ m,were simultaneously taken, and the same field was surveyed in all the bands except for a smallmargin of the survey area corresponding to the sight difference between field-of-view (FOV) ofeach band.AOT includes the calibration sequence for each observation as follows. The dark mea-surements by closing the cold shutter and responsivity check with an internal calibrator arecarried out during maneuver to change the satellite attitude from the all-sky survey mode tothe source targeting mode. After the maneuver the shutter is opened, the sky signal is mon-3tored during a settling time for fine control of the satellite attitude, and then the slow-scanobservation starts. At turning point of the round-trip scan, the shutter is shortly closed, andresponsivity check with the internal calibrator is done. The total exposure time for the round-trip scan is 10 min. The same calibration sequence as above is repeated during the maneuverfor returning to the all-sky survey mode. In addition to the calibration sequence by continuouslight illumination, stimulator flashes every 1 min to check the responsivity drift is operated asis done in the all-sky survey mode. Such highly redundant calibration data set enabled us tocorrect the responsivity change refering the astronomical calibration data taken by separateobservations with the same calibration sequence.
3. Data reduction and calibration
The raw data were initially reduced using a part of the FIS All-sky survey pipeline tool forthe ADU to volt conversion, flagging of bad data (cosmic-ray events and other discontinuities)and the correction of non-linear integration ramps. The reduced data were processed with anofficial data analysis tool specified for the FIS slow-scan observation (SS-tool) to produce thebasic calibrated data products and the final co-added map.Initial process of SS-tool is the slope calculation of the ramps removing the cosmic-rayevents identified in each ramp and correcting the after effect of the calibration lamp (stimuator)flashes. The reset interval of the integration ramp for the Lockman hole observations is set tobe 2 seconds. The 2-s ramps are comprised of 52 and 35 samples per pixel for 65/90 µ m and140/160 µ m detectors, respectively. The slope calculation processing reduces the samplingrate to one data per ramp per array pixel to maximize signal-to-noise and to avoid a periodicstructure arising from incompleteness of the non-linear ramp correction. For 15-arcsec/s scan,the reduced data provides only one data sample per pixel for a source crossing time, but theNyquist sampling condition is resultantly satisfied in real space domain after co-addition of thedata using all array pixels.Glitches and subsequent tails induced by cosmic-ray hits affect the data severely becauseof slow transient behavior of the Ge:Ga detectors. The integration ramps affected by the glitcheswere flagged out, and the tails with a 20 s or a longer time constant were suppressed by medianfiltering in the background subtraction processing as described later. The tail with a shorttime constant is not filtered out in this stage, but the affected data are removed by the sigmaclipping in the co-addition process. The stimlator flashes also causes after effects similar to thetails by cosmic-ray hits, but they were repeatable in amplitude and successfully corrected usinga template of their average time profile with an accuracy limited by the random noise.4 .2. Flat fielding The second step of the SS-tool processing is to produce basic calibrated data for eacharray pixel, i.e., the flat fielding. This process is based on the observations of known diffusesources; zodiacal emissions and interstellar dust emissions, which are expected to be smoothwithin the field of view (FOV). The diffuse-source observations were used only for the flatfielding, and the absolute flux calibration for point-sources was separately done (see Section 4),because it might be different from that for diffuse-source due to complex aperture effects.Zodiacal emissions are expected to be an almost perfect flat source; their anisotropy inarcmin scales are smaller than 1% (Abraham et al. 1998). At 65 and 90 µ m, the sky brightnessof the Lockman hole is dominated by the zodiacal emissions, and the contributions of Galacticcirrus (interstellar dust) emissions, bright point sources and CIB are expected to be negligible.Hence, observed sky itself could be used for the flat fielding. The responsivity distribution ofthe detector arrays is derived from the average of time series data during the slow scan, wherethe average was calculated after removing the data that exceeds the 3-sigma noise level foreach pixel. The flat was built from the data during the first half of the single turn-aroundscan of each pointed observation, because the latter part of the data was affected by the straylight as described in the following section. The flat fielding is done by dividing the data by theresponsivity distribution. With this ”self” flat fielding method, any stripes in the image causedby the flat field error were buried under the random noise.At 140 and 160 µ m, detector signals are dominated by ’offset’ light with a constantintensity from the internal light source. The offset light is implemented to improve responsespeed of the stressed Ge:Ga detector specified for fast scan of the all sky survey mode (morethan 10 times faster than the slow scan used for the Lockman hole survey). Although Galacticcirrus emissions at high latitudes could be a flat source with moderate smoothness ( ∼ µ m is less than ∼
2% of the offset light, and its fluctuation is also negligible. Hence, the selfflat fielding method with the average sky signal including the offset light was applied to correctthe responsivity variation in the detector arrays, as was done for the 65 and 90 µ m bands. In order to emboss faint point-sources on the co-added map, subtraction of the skybackground is required unless its brightness is constant in the field. Unexpectedly, there existsa stray light of earth limb emission onto the focal plane under a condition of small earthavoidance angles of the telescope, which varies with attitude and orbit of the satellite. Thestray light shows not only a long-term variation depending on seasonal change of the orbitalinclination but also short-term variation during a single observation due to change of the earth5voidance angle at every moment. Brightness of the stray light has the maximum (2–3 MJy/srat 90 µ m) at the beginning and the end of the slow scan, and the minimum and plateau regionappear around the mid time.To subtract such slowly varying components as the stray light, we applied the medianfiltering to the data in time domain with a window size of 20 data samples corresponding to10 arcmin ( ∼
15 times FWHM). Then, the smoothed component was subtracted from the data.As a result of this filtering, extended structure larger than roughly a half of the window sizedisappears from the map.
The final step of the SS-tool processing is the co-addition of the calibrated time seriesdata onto a sky grid. The sky position of each data point was derived from the telescopeboresight according to the satellite attitude and from the array pixel map on the focal plane.The grid pixel sizes for all the wave bands were set to 30 arcseconds to secure redundancy,i.e., sufficient number of data per pixel ( > In Fig. 1, we show the final co-added maps of the Lockman hole field at all wavelengths inequatorial coordinates with a pixel size of 30 arcseconds. The image after Gaussian smoothingwith a window size of 1 arcminute is shown as linear contours in the unit of surface brightness.The pixel noise of each map was computed and summarized in the last column of Table 1 insurface brightness. At 65, 140 and 160 µ m, the pixel noises are dominated by instrumentalnoise. At 90 µ m, the pixel noise is contributed by the source noise, which can be recognizedas a positive tail exceeding the standard deviation in the histogram of pixel distribution. Thenoise difference between the 65 and 90 µ m data arises from differences in optical efficiency andfilter bandwidth, while the detectors for these two bands are identical. The noise values for the140 and 160 µ m bands are different for the same reason. Consequently, the 90 µ m band is the6ost sensitive in all the photometric bands.
4. Photometry and point-source calibration
To extract point sources from the final co-added image, we used photometry tools, FIND,GCNTRD and APER, in the IDL Astronomy User’s Library at NASA/GSFC (Landsman 1993).The tools search center positions of point sources with a Gaussian window function approxi-mating the Point Spread Function (PSF) and measure the fluxes by aperture photometry. PSFin each wavelength band is derived from many observations in the PV phase and for calibration.The aperture radius was set to the full width half maximum (FWHM) of PSF. The aperturephotometry was carried out without the sky subtraction, which was already done in the medianfiltering process. This sky subtraction method is helpful to avoid losing sources near a defectpixel caused by some artifacts.
Photometric calibration used in this paper is based on point-source observations in thePV phase in various fields, because of lack of well-calibrated source in the Lockman hole.According to the measured PSF, the photometric signals were compared with the expectedfluxes of the calibration sources converted to the ’flat’ spectra, ν × F ν = constant, and theirlinear correlation factors were derived. The expected fluxes of the calibration sources rangefrom 2 Jy to 200 Jy at 90 µ m, and deviation from the linear relation had no flux dependencein such a wide flux range. The calibration accuracy estimated from the standard deviation forvarious measurements was 17%, 16%, 12% and 22% at 65, 90, 140 and 160 µ m, respectively.The point-source noise in flux unit, i.e. the 1-sigma detection limit for point source, is derivedfrom the pixel noise in surface brightness and the aperture correction factor, as summarized inTable 1. As the flux calibration for AKARI refers to the flat spectrum, the real flux is obtained bythe color correction depending on Spectral Energy Distribution (SED) of the source. However,it is difficult to measure the infrared color for all the detected sources especially at faint fluxlevels, because most of the sources were detected at 90 µ m only. Hence, for producing thesource counts we assume that SED of all sources have the flat spectrum, and we estimate theuncertainties of fluxes for various SEDs.The color correction was available only for bright sources detected in multiple wavelengthbands of AKARI. In Fig. 2, color-corrected (open symbols) and uncorrected (filled symbols)spectra of three bright sources in the Lockman hole are shown. The data are compared withmodified graybody spectra with different color temperatures; combinations of a graybody spec-7rum with an emissivity β , F ∼ ν β × B ( T ), and natural extension from the graybody towardsshorter wavelengths by a power-law spectrum with an index α , F ∼ ν − α , (e.g., Blain et al. 2002).The error bars are the point-source noise in Table 1. These sources have been detected by theISO survey and optically identified, and their redshifts have been measured by spectroscopicobservations (Oyabu et al. 2005). Two of the sources, ID76 and ID87, are nearby star-forminggalaxies at z = 0.08 and 0.09, and the color correction is up to 10% in all bands. The modifiedgraybody models ( T =27 K, β =1.5 and α =2.4 for ID76, T =23 K, β =1.5 and α =2.4 for ID87)give good fits to the measured spectra even for uncorrected data. Another source (ID104) isULIRG with a moderate redshift of z = 0.362. A graybody model with T =25 K (effectively T =18 K corresponding to the redshift), β =1 and no power-law component provides reasonablefit to the data. The color correction for the 65 µ m data is 29%, which is an extreme case inthe parameter range of color temperature. If the assumed color for ID104 is correct, the real65 µ m flux is expected to be below the detection limit, but the observed flux shows a finitevalue at 65 µ m. The high flux at 65 µ m may be due to noise-induced flux boosting, which issimilar to Eddington bias or Malmquist bias and usually seen for sources near the detectionlimit (Heraudeau et al. 2004).Frayer et al. (2006a) reported that the Spitzer sources detected at 24, 70 and 160 µ min the xFLS field have a typical dust temperature of 30 K with an emissivity of β =1.5 anda power-law index of α =2.4. This color temperature is consistent with 15–25 K for β =2derived from the European Large Area ISO Survey (ELAIS) (Heraudeau et al. 2004) and fromthe ISO deep survey in Lockman hole (Oyabu et al. 2005). According to these results, mostof far-infrared sources have color temperatures ranging from 15–40K with β =1–2 and of thepower-law indices of α =1–2.5. Uncertainty of the flux measurement for the sources with colorsin this parameter range is estimated to +4/ − − −
4% and +2/ −
1% at 65, 90,140 and 160 µ m, respectively. In most cases for high temperature sources, the color correctionuncertainties are smaller than the flux calibration errors. For nearby cold dust componentsor high redshift sources, the uncorrected flux at 65 µ m assuming the flat spectrum could begreatly overestimated.
5. Source counts µ m counts The final catalogs were first produced at 90 µ m, at which the deepest data were obtained.Sources with signal-to-noise ratios of S/N > ∼ µ m, 85 sources were found in the entire survey field down to 26mJy. The surface density of sources derived from the beam size, Ω = 4 . × − sr, is ∼
70 beamsper source. This is far above 22 beams per source that is required for the 90% completeness8ithout source confusion estimated by Helou & Beichman (1990) and also much greater than27 beams per source as a criterion for the 3-sigma detection in case of the Euclidian counts(Franceschini 2000).In Fig. 3 the integral counts at 90 µ m are plotted as filled circles. The data are rawcounts not corrected for incompleteness. Error bars along the count axis are 1-sigma Poissonuncertainty, and the flux errors correspond to total uncertainties including both absolute cal-ibration and color correction. Our results are compared with those from the ISO surveys at90 µ m towards the Lockman hole (Linden-Vornle et al. 2000, Kawara et al. 2004, Rodighieroet al. 2003) and the ELAIS field (Heraudeau et al. 2004). A no-evolution model by Pearson(2007) is also shown as a reference.Our source counts in the Lockman hole reaching to ∼
25 mJy at S/N > ∼
100 mJy, our counts show good agreement with all the ISO results except for Kawara etal. (2004). A small discrepancy at the bright end could be due to large statistical errors of ourcounts.At fainter flux levels, our integral counts seem to agree with Rodighiero et al. (2003) at60 and 30 mJy. However, individual sources in the counts listed in Rodighiero et al. (2003) arenot always consistent with our measurements in both flux and position, while relative fluxesof all the sources listed in Kawara et al. (2004) are well aligned with our data, as describedin Appendix. Our counts do not show the strong excess reported in Kawara et al. (2004) andalso in related papers (Kawara et al. 1998, Matsuhara et al. 2000). This discrepancy is partlyexplained by different flux calibration by a factor of ∼ µ m data; they have suffered from glitches and tails induced by cosmicrays (Franceschini et al. 2001).In order to check the effect of the field variance to the counts, we divided our observedfield into two at DEC = 57.28 deg as number of the detected sources for the two sub-samplesare equal to each other, and then we produced the source counts for each sub-samples. InFig. 3 the source counts in the fields of DEC < > µ m counts are shown in differential form dN/dS normalized tothe Euclidean law N ∼ S − . with error bars of 1-sigma Poisson uncertainty. Total uncertaintyof absolute calibration and color correction is indicated with a cross symbol at lower left (at20 mJy) of the figure. The results are summarized in Table 2 together with the integral countdata. Our differential counts are compared with previous results from the same references as theintegral counts. Again, our counts are slightly lower than that for the previous observations, butboth the ISO results except for Kawara et al. (2004) and our results show a general tendencyof flat counts with no steep rise or drop in the measured flux range. At wavelengths other than 90 µ m, the sources accompanied with the 90- µ m sourceswere selectively extracted. In terms of real source extraction against any artifacts, the catalogproduced in such a way is more reliable than that individually produced at each wavelength.The 90- µ m selected catalogs are possibly biased as to miss exceptionally hot or cold/high-zpopulations. However, such biasing effects are expected to be small. Because, the detectionlimit at 90 µ m is much better than that at the other wavelengths, and any detected sourcehaving ordinary temperatures of T =15–40K with β =1–2 should be identified at 90 µ m.To extract the commonly detected sources, a centroid near a 90- µ m source was searchedat each wavelength allowing a small position difference corresponding to the beam size. Onlysignals associated with the 90 µ m sources having S/N > µ m, the signal-to-noise threshold for the source extraction was lowered to S/N > µ m is set to S/N > µ msources compared with the real sources. As a result, 11 and 6 sources were found at 65 and 140 µ m, respectively. It is noteworthy that all the 6 sources at 140 µ m were identified at 65 µ m.At 160 µ m, no signal matches to the criteria for source extraction.In Fig. 5 and Fig. 6, differential source counts at 65 and 140 µ m are plotted with filledcircles. Error bars for the counts are 1-sigma Poisson uncertainty. The flux error correspondingto both absolute calibration error and color correction uncertainty are shown by a cross symbolin each figure. Our results are compared with Spitzer counts at 70 and 160 µ m, respectively,in GTO fields (Dole et al. 2004) and xFLS field (Frayer et al. 2006a). The Spitzer data wereconverted at 65 µ m and 140 µ m with a flat spectrum of ν × F ν = const as assumed in theirpaper. The Spitzer GTO data in the intermediate flux range show a field variance betweenthose taken in two different fields (Marano and CDFS; Chandra Deep Field South). At bothwavelengths, our counts are slightly higher than the Spitzer counts but agree with them withinerror bars and the field variance.In Fig. 4, partial number counts for the 90 µ m sources that are constituents of the 65 µ m counts are plotted with an open circle. About half of the 90 µ m counts in a range from100 to 160 mJy consists of the 65 µ m sources. The average color temperature of these sourceswithout color correction is T ∼
40 K for β =1.5, corresponding to the color ratio S /S ∼ T ∼
30 K for β =1.5 (Frayer et al. 2006a). The main reason that our 65 µ m counts are slightly higher thanthe Spitzer counts is the field variance; most of the 65 µ m sources lie in the lower DEC fieldwhere more bright sources exist, as described in section 5.1. Note that our 140 µ m counts arealso consistently higher than the Spitzer 160 µ m counts. The remained half sources in the 90 µ m counts would have lower color temperatures as T <
30 K to be consistent with the Spitzerresults. Such redder color populations may contribute to the 65 µ m counts at fluxes below thedetection limit.
6. Discussions
As described in the last section, our source counts at 65 and 140 µ m are fairly consistentwith Spitzer observations. In this sense, our count data at 90 µ m, which lie in between 65and 140 µ m, are simply expected to agree with recent evolutionary models that show excellentagreement with the Spitzer counts. In the following, we describe the comparison of our countswith the models.In Fig. 4, 5 and 6, source counts predicted from models at each wavelength are shown;dashed: Pearson (2007) and dash-dotted: Lagache et al. (2003, 2004). In Fig. 4, Lagache et al.(2003) model taken from Heraudeau et al. (2004) was used, because the model counts at thewavelength near 90 µ m is not presented in Lagache et al. (2004), but the difference betweentheir models of 2003 and 2004 are negligible in the AKARI wavelength bands (Lagache et al.2004). For Fig. 5 and 6, the predicted counts at 70 and 160 µ m in Lagache et al. (2004)were converted to 65 and 140 µ m, respectively, assuming the flat spectrum. A non-evolutionmodel by Pearson (2007) is also shown as a reference. Lagache et al. (2003, 2004) usesphenomenological approach aiming to build the simplest model. They assume that infraredgalaxies are mostly powered by starformation and use typical SEDs of normal and starburstgalaxies. The luminosity function is represented by these two activity types. The modelparameters are adjusted to fit the number counts in multi-wavelength bands from mid-infraredto sub-mm, the redshift distributions, the local luminosity functions at 60 and 850 µ m, andCIB. The contemporary galaxy evolution model of Pearson (2007) uses a backward evolutionprocess based on the IRAS all-sky PSCz multi-component luminosity function defined at 60 µ m comprising of cool normal quiescent galaxies, and a warmer component defined by infraredluminosity as L IR < L ⊙ starburst galaxies, L IR > L ⊙ LIRG sources, L IR > L ⊙ ULIRG sources and AGN. Both luminosity evolution and density evolution is included in themodels. This model fits the observed source counts from the 2–1200 µ m from the IRAS, ISO,Spitzer missions and the SCUBA/JCMT, MAMBO/IRAM instruments. Differences betweenthese models are small at brighter fluxes, but at flux levels below 100 mJy strong evolution11ffects set in and significant differences between models appear.A significant result of the model comparison is that the observed counts at 90 µ m ina flux range of 26–160 mJy are apparently lower than the predicted counts using a model byLagache et al. (2003, 2004), which show excellent agreement with Spitzer counts in all MIPSbands of 24, 70 and 160 µ m and also with our data at 65 and 140 µ m, as shown in Fig. 5 and6. Descrepancy between the model and data is clear at fainter levels; the model continues toincrease while the observed counts keep constant. The model by Pearson et al. (2007) showsbetter fit with our 90 µ m counts, but it still predicts slightly too high counts to fit the data.Other models, e.g. Chary and Elbaz (2001), King and Rowan-Robinson (2003), and Balland,Devriendt and Silk (2003), also reproduce various observables including the source counts forthe infrared luminous galaxies within uncertainties of the observation data, and all of thempredict similar counts to above two models. Consequently, no model to explain the low 90 µ mcounts is currently available.The incompleteness of our data might lower the 90 µ m counts at fainter flux levels, butit is unlikely, because the 65 and 140 µ m counts derived from the 90 µ m selected catalogsare consistent with or even higher than the Spitzer data and the models. Separation betweenmodels and data starts already at ∼
80 mJy, where the signal-to-noise is high (
S/N ∼
10) enoughnot to lose many sources. If the discrepancy is due to the flux calibration, our flux calibrationhas to be different by a factor of ∼
2, which is unreasonably larger than our estimate of thecalibration error. In order to explain our results, the evolutionary models may require somemodifications.It should be emphasized that our results provide strong constraints on the evolutionaryscenarios. A new model to explain our results may be based on more complicated SEDs thanpreviously used; faint galaxies may have exceptionally low emissivity at 90 µ m due to dustproperties. There also exists a room to tune model parameters to represent high redshiftcomponents (e.g. ULIRG), because their redshifts and SEDs are not well defined. In any cases,the AKARI observations of much larger number of SED samples are essential for constructingthe new model. Combination of source counts and redshift distributions are much strongerdiscriminators for various evolutionary models (e.g. Le Floc’h et al. 2005). Although theredshift measurements by optical spectroscopy for some of the ISO 90 µ m sources have beendone (Oyabu et al. 2005, Rodighiero et al. 2005), only 9 spectroscopic samples in their catalogsmatch with the AKARI 90 µ m sources. In the spectroscopic samples, 8 sources have redshiftsof z < .
4, and the remaining 1 source is z = 1 .
1. Such statistically poor samples with lowredshifts cannot constrain any evolutionary models. Further measurements and analysis of theredshift distribution of the AKARI 90 µ m sources are required.Unresolved sources below the detection limit form the CIB. The integrated flux of our90 µ m counts down to 26 mJy corresponds to the surface brightness of 0.021 MJy/sr. This isonly ∼
7% of the CIB estimated by Lagache et al. (2003) and Dole et al. (2006). Dole et al.122004) reported that the Spitzer counts in the GTO field at 70 and 160 µ m down to 15 and50 mJy account for about 23% and 7% of the CIB, respectively. For the xFLS survey downto 8 and 50 mJy, approximately 35% and 15% of the CIB were resolved at 70 and 160 µ m,respectively (Frayer et al. 2006a). Frayer et al. (2006b) also claimed that ultradeep Spitzercounts down to 1.2 mJy account for about 60% of the CIB at 70 µ m and that a turn-over ofthe differential counts, as seen in the Spitzer 24 µ m counts (Papovich et al. 2004), appears at ∼
10 mJy. If we simply assume the flat spectra for all galaxies, the turn-over point of the 90 µ m counts is expected to be ∼
13 mJy. These results suggest that number of the bulk galaxiesof the CIB lying below the point-source detection limit at 90 µ m may steeply increase towardsthe turn-over point. In order to search for such missing sources, much deeper survey at 90 µ mis demanded. The flux limit of the present survey is still above the confusion limit, and sourcecounts to deeper levels can be obtained by spending much exposure time.
7. Conclusions
We presented far-infrared observations of the Lockman hole with AKARI satellite. Inthe performance verification phase, we performed slow scan observations in a 0.6 deg × σ ) at 65, 90 and 140 µ m, respectively. The counts at 90 µ mare ∼ µ m counts were several times lower than the predicted counts by recent evolu-tionary models that show good agreement with the Spitzer/MIPS data. On the contrary, theobserved counts at 65 and 140 µ m are consistent with previous measurements with Spitzer at70 and 160 µ m. Our 90 µ m counts above 26 mJy accounts for ∼
6% of the CIB, and the bulkgalaxies of the CIB may cause a steep rise of the counts below the detection limit. These resultsprovide strong constraints on the evolutionary models.The source counts presented in this paper were not color-corrected, because limitedsamples were available to examime the infrared colors of galaxies. Further study of the sourcecounts could be done by proper color correction based on the measured SED of each source.A far-infrared deep survey in the lowest cirrus region with an area of ∼
10 square degrees nearthe south ecliptic pole, as a part of AKARI deep survey programs (Matsuhara et al. 2006), isexpected to provide us a large number of samples over 1000 sources down to ∼
20 mJy at 90 µ m. Moreover, the AKARI all-sky survey will provide us extremely high statistics of galaxysamples on precedented sensitivity levels over the whole sky. The data sets obtained by suchlarge-area surveys are promising not only for producing the SED templates for various galaxypopulations but also for extending the flux range of the source counts to much brighter levelsthan we obtained in this work. 13he AKARI mission is operated by Japan Aerospace Exploration Agency (JAXA),Nagoya University, the University of Tokyo and National Astronomical Observatory Japan,European Space Agency (ESA), Imperial College London, University of Sussex, Open University(UK), University of Groningen / SRON (Netherlands), and Seoul National University (Korea).The far-infrared detectors were developed under collaboration with The National Institute ofInformation and Communications Technology (Japan). The authors would like thank all theAKARI project members for their intensive effort. Appendix 1. Comparison with the ISO source catalogs in the Lockman hole
To confirm the flux calibration at lower flux levels, we compared the measured 90 µ mfluxes of relatively bright sources with the ISO data for the same sources in Fig. 7. The dot-dashed and dashed lines are the best-fit lines with linear coefficients (scaling factor) of 0.88 ± ± µ m (1.22 Jy) is perhaps overestimated. In fact,the flux of this source measured with AKARI is 0.56 ± µ m and lower than theIRAS flux by a factor of ∼
2. This factor is in good agreement with the scaling factor betweenKawara et al. (2004) and our measurements for the sources as described above. A tendencyfor the IRAS FSC to overestimate the flux is also pointed out by Heraudeau et al. (2004) inthe data analysis for the ELAIS survey. Our recent result of the flux comparison between theAKARI all-sky survey data and the IRAS point source catalog (PSC) shows large scattering ofthe AKARI/IRAS flux ratio up to ∼ µ m mapas background. Many of the sources in their catalog are not identified by AKARI and viceversa, though only bright sources are identified. Some of their sources were misidentified asfaint sources detected by AKARI at S/N < eferences
Abraham, P., Leinert, C., Acosta-Pulido, J., Schmidtobreick, L., & Lemke, D. 1998, in The Universeas seen by ISO, ed. P. Cox, & M. Kessler, ESA SP-427, 145Balland, C., Devriendt, J. E. G., & Silk, J. 2003, MNRAS, 343, 107Blain, A. W., Smail, I., Ivison, R. J., Kneib, J.-P., & Frayer, D. T. 2002, Phys. Rep., 369, 111Blain, A. W., Barnard, V. E., & Chapman, S. C. 2003, MNRAS, 338, 733Caputi, K. I., et al. 2006, ApJ, 637, 727Chary, R. & Elbaz, D. 2001, ApJ, 556, 562Devriendt, J. E. G. & Guiderdoni, B. 2000, A&A, 363, 851Dole, H., et al. 2001, A&A, 372, 364Dole, H., Lagache, G., & Puget, J.-L. 2003, ApJ, 585, 617Dole, H., et al. 2004, ApJS, 154, 87Dole, H., et al. 2006, A&A, 451, 417Efstathiou, A., et al. 2000, MNRAS, 319, 1169Finkbeiner, D. P., Davis, M., & Schlegel, D. J. 2000, ApJ, 544, 81Frayer, D. T., et al. 2006a, AJ, 131, 250Frayer, D. T., et al. 2006b, ApJ, 647, L9Franceschini, A. 2000, in Galaxies at High Redshifts, Proceedings of the XI Canary Islands WinterSchool of Astrophysics, ed. I. Perez-Fournon, M. Balcells, F. Moreno-Insertis and F. Sanchez(Cambridge University Press)Franceschini, A., Aussel, H., Cesarsky, C. J., Elbaz, D., & Fadda, D. 2001, A&A, 378, 1Hacking, P. B., & Soifer, B. T. 1991, ApJ, 367, L49Hauser, M. G., et al. 1998, ApJ, 508, 25Hauser, M. G., & Dwek, E. 2001, ARA&A, 37, 249Helou, G., & Beichman, C. A. 1990, Proc. 29th Liege International Astrophysical Colloquium FromGround-Based to Space-Borne Sub-mm Astronomy, ESA SP-314, 117Heraudeau, Ph., et al. 2004, MNRAS, 354, 924Jeong, W.-S., et al. 2007, PASJ, this volumeKawada, M., et al. 2007, PASJ, this volumeKawara, K., et al. 1998, A&A, 336, L9Kawara, K., et al. 2004, A&A, 413, 843King, A. J. & Rowan-Robinson, M. 2003, MNRAS, 339, 260Lagache, G., Abergel, A., Boulanger, Desert, F. X., & Puget, J.-L. 1999, A&A, 344, 322Lagache, G., Dole, H., & Puget, J.-L. 2003, MNRAS, 338, L555Lagache, G., et al. 2004, ApJS, 154, 112Landsman, W.B. 1993, in Astronomical Data Analysis Software and Systems II, A.S.P. ConferenceSeries, Vol. 52, ed. R. J. Hanisch, R. J. V. Brissenden, and Jeannette Barnes, p. 246.Le Floc’h, E., et al. 2005, ApJ, 632, 169Linden-Vornle, M. J. D., et al. 2000, A&A, 359, 51Matsuhara, H., et al. 2000, A&A, 361, 407Matsuhara, H., et al. 2006, PASJ, 58, 673 urakami, H., et al. 2007, PASJ, this volumeNeugebauer, G., et al. 1984, ApJ, 278, L1Oyabu, S., et al. 2005, AJ, 130, 2019Papovich, C., et al. 2004, ApJS, 154, 70Pearson, C. 2007, Adv. Space Res., in pressPerez-Gonzalez, P. G., et al. 2005, ApJ, 630, 82Puget, J.-L., Abergel, A., Berbard, J. P., Boulanger, F., Burton, W. B., Desert, F.-X., & Hartmann,D. 1996, A&A, 308, L5Puget et, J.-L., et al. 1999, A&A, 345, 29Rieke, G. H., et al. 2004, ApJS, 154, 25Rodighiero, G., Lari, C., Franceschini, A., Gregnanin, A., & Fadda, D. 2003, MNRAS, 343, 1155Rodighiero, G., & Franceschini, A. 2004, A&A, 419, L55Rodighiero, G., Fadda, D., Franceschini, A., & Lari, C. 2005, MNRAS, 357, 449Werner, M. W., et al. 2004, ApJS, 154, 1 able 1. Parameters for aperture photometry
Wavelength Pixel size Aperture radius Net integration time Point-source noise Pixel noise[ µ m] [arcsec] [arcsec] per pix [s] (1 σ ) [mJy] (1 σ ) [MJy / sr]65 30 37 12 26 0.2090 30 39 12 9 0.06140 50 58 20 76 0.39160 50 61 20 288 0.60 Table 2.
Integral and differential counts at 90 µ m Integral counts Differential counts S Number N ( > S ) Average S bin width dN/dS × S . [Jy] [sr − ] [Jy] [Jy] [sr − Jy . ]0.119 4 1.92 e4 0.140 0.042 3.39 e3 ± ± ± ± ± ± µ m 65 µ m140 µ m 160 µ m Fig. 1.
The final co-added map of the Lockman hole field in all photometric bands in equatorial coordi-nates with the pixel size of 30 arcsec. Upper-left: 90 µ m, Upper-right: 65 µ m, Lower-left: 140 µ m, andLower-right: 160 µ m. The image size is approximately 1.2 deg in RA × F l u x [ Jy ] Observed wavelength [ m m] 200 Fig. 2.
Examples of the measured SEDs of bright sources in Lockman hole (ID76: circles, ID87: trianglesand ID104: diamonds) as functions of the observed wavelength are compared with modified graybodyspectra; T =27 K, β =1.5 and α =2.4 for ID76 (thin line), T =23 K, β =1.5 and α =2.4 for ID87 (dashedline), and T =25 K, β =1 with a redshift of z = 0.362 (effectively T =18 K) for ID104 (dotted line). Filledand open symbols denote the data without and with the color correction, respectively. The model fluxesare scaled to fit the measured fluxes. ISO LH (Kawara 2004)ISO LH (Rodighiero 2003)ISO ELAIS (Heraudeau 2004)ISO LH (Linden-Vornle 2000)This workNon-evolution (Pearson 2007)DEC < 57.28 degDEC > 57.28 deg N u m be r c oun t [ s r - ] Flux [Jy] m m Fig. 3.
Integral counts at 90 µ m with no correction for completeness (filled circles). Previous results ofthe ISO surveys at 90 µ m in the Lockman hole are shown; Linden-Vornle et al. 2000 (open squares),Kawara et al. 2004 (crosses), Rodighiero et al. 2004 (open triangles). The ISO counts in the ELAIS field(Heraudeau et al. 2004) are plotted with diamonds. The solid line is a non-evolution model by Pearson(2007). The dashed and dotted lines are the source counts for sub-samples in the fields of DEC < > ISO LH (Kawara 2004)ISO LH (Rodighiero 2003)ISO ELAIS (Heraudeau 2004)This work65 m m selected d N / d S * S . [ s r - Jy . ] Flux: S [Jy] m m Fig. 4.
Differential counts at 90 µ m with no correction for incompleteness. The ISO results are plottedwith the same symbols as Fig. 3. A cross symbol at lower left shows total uncertainty including fluxcalibration error and color correction uncertainty (see text). Recent evolutionary models are also shown;solid and dashed lines: non-evolution and evolution model by Pearson (2007), and dash-dotted line:Lagache et al. (2003) taken from Heraudeau et al. (2004). Spitzer xFLS (Frayer 2006)Spitzer GTO (Dole 2004)This work d N / d S * S . [ s r - Jy . ] Flux: S [Jy] m m Fig. 5.
Differential counts at 65 µ m for the 90 µ m selected sources (filled circles) compared with theSpitzer counts at 70 µ m scaled to 65 µ m (triangles: Dole et al. 2004, crosses: Frayer et al. 2006a). Across symbol at upper left shows total uncertainty of the measured flux (see text). The volutionary modelssame as Fig. 4 are also shown. Spitzer xFLS (Frayer 2006)Spitzer GTO (Dole 2004)This work d N / d S * S . [ s r - Jy . ] Flux: S [Jy] m m Fig. 6.
Differential counts at 140 µ m plotted with the same symbols as Fig. 5. .010.10.01 0.1 Kawara 2.09 +/- 0.17Rodighiero 0.88 +/- 0.12 I S O m m f l u x [ Jy ] AKARI 90 m m flux [Jy] Fig. 7.
Comparison of the AKARI measured flux for bright sources in the Lockman hole with ISO results(Kawara et al. 2004, Rodighiero et al. 2005). Fig. 8.
Comparison of the source positions between Rodighiero et al. (2005) and this work (backgroundcontour). In this linear contour map, darker pixels have higher surface brightness as gray scale. Largecircles denote brightest sources listed in the catalog of Rodighiero et al. (2005) and also in Oyabu etal. (2005). Small circles denote the remained faint sources in their catalog. Crosses are AKARI sourcesdetected at
S/N >
3. Many of the faint ISO sources are not identified by AKARI and vice versa. In Fig. 7,all the ISO sources identified by AKARI are plotted.3. Many of the faint ISO sources are not identified by AKARI and vice versa. In Fig. 7,all the ISO sources identified by AKARI are plotted.