ASTRO-F, super-iras, the All-Sky Infrared Survey

Transcription

1 Mon. Not. R. Astron. Soc. 347, 3 9 (4) ASTRO-F, super-iras, the All-Sky Infrared Survey Chris P. Pearson,,3,5 H. Shibai, T. Matsumoto, 3 H. Murakami, 3 T. Nakagawa, 3 M. Kawada, T. Onaka, 4 H. Matsuhara, 3 T. Kii, 3 I. Yamamura 3 and T. Takagi Astrophysics Group, Imperial College of Science Technology and Medicine, Blackett Laboratory, Prince Consort Rd, London SW7 BW Infrared Astrophysics Laboratory, Nagoya University, Furo-cho, Chikusa-ku, Nagoya , Japan 3 Institute of Space and Astronautical Science, Yoshinodai 3--, Sagamihara, Kanagawa 9 85, Japan 4 Department of Astronomy, School of Science, University of Tokyo, Tokyo 3-33, Japan 5 Centre for Astrophysics and Planetary Science, University of Kent at Canterbury, Canterbury, Kent Accepted 3 October. Received 3 October ; in original form October 6 INTRODUCTION In order to understand such fundamental cosmological questions as type dependent galaxy evolution, star formation history and the large-scale structure of the Universe, large data sets of homogeneous data are required. Most recently, these large data sets have been mainly accumulated at near-infrared (NIR)/optical wavelengths (e.g. The Sloan Digital Sky Survey (SDSS) 6 galaxies over one-quarter of the sky, Stoughton et al., the two degree field galaxy redshift survey (dfgrs) 5 galaxies over deg of the sky (Colless et al. ). To investigate galaxy evolution and star formation in the high-redshift z > Universe, due to finite observation times, a trade off must be made between the area and the depth of a given survey, that is to say, the greater the sensitivity or depth required, the narrower the survey has to be, e.g. the Hubble and SUBARU Deep Fields (Williams et al. 996; Maihara et al. ). Hence, although the two HDF fields have produced excellent images and some spectroscopic redshifts with >3 galaxies at z > (Cohen et al. ; Dennefeld ), cpp@ic.ac.uk Homepage: cpp/astrof/ ABSTRACT We review the next-generation Japanese infrared space mission, ASTRO-F. ASTRO-F will be the first survey of the entire sky at infrared wavelengths since the IRAS mission almost years ago. ASTRO-F will survey the entire sky in four far-infrared bands from 5 to µm and two mid-infrared bands at 9 and µm to sensitivities times deeper than the IRAS satellite at angular resolutions of 5 45 arcsec (cf. 5 arcmin for IRAS). ASTRO-F can be considered as a super-iras. Using the galaxy evolution model of Pearson, we produce expected numbers of sources under three different cosmological world models. We predict that ASTRO- F will detect of the order of tens of millions of sources in the far-infrared wavelength bands, most of which will be dusty luminous infrared/ultraluminous infrared galaxies, of which as many as half will lie at redshifts greater than unity. We produce number redshift distributions, flux redshift and colour colour diagrams for the survey and discuss various segregation and photometric redshift techniques. Furthermore, we investigate the large-scale structure scales that will be accessed by ASTRO-F, discovering that ASTRO-F and SIRTF-SWIRE probe both different scales and redshift domains and concluding that the two missions will supplement rather than supplant one another. Key words: surveys galaxies: evolution cosmology: observations infrared: general. they are effectively limited to two lines of sight (HDF-N, 5.3 arcmin, Williams et al. 996 and HDF-S,.7 arcmin, Gardner et al. ; Williams et al. ) and are therefore somewhat susceptible to statistical uncertainty (e.g. from large-scale structures). For example, almost one-quarter of the total z <.,.65-µm luminosity comes from just four galaxies. In the past few years considerable progress has been made in the understanding of the star formation rate of high-redshift galaxies using the Lyman drop out technique to isolate and identify galaxies at redshifts of 4 (Madau et al. 996; Steidel et al. 996). Well over 5 Lyman-break systems have now been observed at z > 3. These are ultraviolet (UV)-bright, blue spectra, actively starforming galaxies. Estimations of star formation rates of a few to 5 M yr led to the popular Madau plot (Madau et al. 996) and the somewhat premature implication that the star formation rate of the Universe peaked at z. However, the early analysis of the optical/uv star formation rate neglected the effects due to dust. Subsequent analysis correcting for extinction did indeed imply corrections to the star formation rates of Lyman-break galaxies of factors of 7 (Dickenson 998; Pettini et al. 998; Calzetti & Heckman 999; Meurer, Heckman & Calzetti 999; Steidel et al. 999; Goldader et al. 3). Downloaded from by guest on October 8 C 4 RAS

2 4 C. P. Pearson et al. Indeed, studies with the Infrared Space Observatory (ISO) (Kessler et al. 996), of the Hubble Deep Fields have revealed star formation rates at least comparable to or higher than those of the optical/uv studies (Rowan-Robinson et al. 997; Mann et al. ). At submillimetre wavelengths, surveys with SCUBA on the JCMT (Holland et al. 999) revealed a large (>3 deg at S 85 > mjy) strongly evolving population of sources with bolometric luminosities > L and star formation rates of 3 M yr lying at redshifts > (Smail, Ivison & Blain 997; Barger et al. 998; Hughes et al. 998; Barger, Cowie & Sanders 999; Blain et al. 999; Dunne et al. ; Fox et al. ). The straightforward interpretation of these results is that most massive star formation occurs in dense molecular clouds that obscure the view at optical wavelengths. What is seen at optical/uv wavelengths represents star formation at the periphery of the clouds and is but the tip of the iceberg. Of course this revelation is not new. The first significant evidence came from the IRAS satellite, launched in 983, which surveyed almost the entire sky in four wavebands,, 6, 5 and µm, detecting more than 5 galaxies and providing a huge legacy of data. IRAS revealed a new population of galaxies in the IR indicating that most of the star formation history had in fact been missed by the optical observations. The local ratio of L IR /L opt was found to vary from 3 per cent for normal quiescent galaxies (Rowan-Robinson, Helou & Walker 987; Corbelli, Salpeter & Dicky 99) to as much as 9 per cent for luminous L IR L and ultraluminous L IR L galaxies (e.g. Soifer et al. 986). Further evidence of a darker side to star formation has come from measurements, detections and upper limits of the integrated extragalactic background light at., 3.5, 4 and 4 to µm by the COBE, ISO and IRTS missions (Puget et al. 996; Fixsen et al. 998; Hauser et al. 998; Finkbeiner et al. ; Gorjian et al. ; Lagache et al. ; Lagache & Puget ; Matsumoto et al. ; Wright & Rees ; Hauser ). These data have indicated that the far-infrared (FIR) background is comparable to the optical nearinfrared background, i.e. that 5 per cent of the integrated rest frame optical/uv emission from hot, young stars is reprocessed by dust and reradiated at mid-infrared (MIR) and far-infrared wavelengths. Therefore, the rest frame optical/uv luminosities of galaxies only contribute a lower limit to the star formation history of the Universe (Chary & Elbaz ). In order to tightly constrain general mass evolution of galaxies and the star formation history of the Universe, large area, wide field infrared surveys to redshifts are required. For example, although the optical/uv studies of Lyman-break galaxies sample large volumes at high redshift (the volume encompassed at < z < is times the volume at z <, =.3, =.7) most cosmological evolution takes place at lower redshifts. The ISO and SIRTF (Rieke ) missions are observatories making surveys of the order of tens of deg. The IRAS mission produced a pioneering survey of the infrared sky but could only reach out to relatively low redshifts (e.g. z med.3 for the PSCz, Saunders et al. ) and Table. Far infrared surveyor all-sky survey parameters. is now more than yr old. To extend unbiased measurements of the star formation rate out to higher redshifts a new all-sky infrared survey is required. ASTRO-F, also known as the Infrared Imaging Surveyor (IRIS, Murakami 998) will be the second infrared astronomy mission of the Japanese Institute of Space and Astronautical Science (ISAS). ASTRO-F is a 67-cm cooled telescope and will be dedicated for all-sky and large area surveys in the infrared, one of the primary objectives of the survey being to map the entire sky in four FIR bands ranging from 5 to µm. With the timely launch of ASTRO-F (almost on the th anniversary of the IRAS mission) it can indeed be regarded as a super-iras. In this paper, we review the ASTRO-F mission its objectives and predicted results (see Takeuchi et al. 999; Jeong et al. a,b for further simulations and discussions of the ASTRO-F mission). Note that although ASTRO-F will also provide equally valuable information in the areas of galactic star formation, planetary systems and innumerable other fields of galactic astronomy and astrophysics, here we will concentrate solely on its extragalactic all-sky survey, as this is one of the primary motivations behind the mission. In Section we introduce the ASTRO-F mission and the infrared allsky survey. Section 3 explains the model parameters used to predict the survey results. In Section 4 we present the predictions of the All-Sky Survey and in Section 5 we discuss the simple methods of colour discrimination of sources in the All-Sky Survey. A summary of results and the conclusions are given in Section 6. THE ASTRO-F ALL-SKY SURVEY The ASTRO-F infrared space telescope (Murakami 998) incorporates a 67-cm diameter silicon carbide Ritchey Chretien design telescope (-kg primary mirror) cooled to 5.8 K (the detectors to.8.5 K) by 7 l of liquid helium contained within a light weight cryostat (Kaneda et al. 3). The temperature of the outer wall of the cryostat is expected to be maintained below K via radiation cooling and a pair of -stage Stirling-cycle coolers ensure minimum heat flow from the outer wall of the cryostat, thus almost doubling the lifetime of the helium providing a mission lifetime of more than 55 d. Furthermore, the use of mechanical coolers means that the near-infrared detectors will still be usable even after helium exhaustion. ASTRO-F is scheduled to be launched in mid 4 into a sun-synchronous polar orbit with an altitude of 75 km with the M-V launch vehicle of ISAS. ASTRO-F incorporates two focal plane instruments covering the entire IR region from the K-band to µm. The first focal plane instrument is the far-infrared surveyor (FIS, see Table, Kawada 998; Takahashi et al. ), which will survey the entire sky simultaneously in four FIR bands from 5 to µm with angular resolutions of 5 45 arcsec. The FIS has two kinds of detector arrays; a short-wavelength array (5 µm) comprised of unstressed Ge:Ga detectors and a long-wavelength array ( µm) comprised of stressed Ge:Ga detectors. These two arrays Downloaded from by guest on October 8 Filter Wavelength band Array size Pixel size (diffraction beam) 5σ sensitivity N µm 6.79 arcsec (.6 arcsec) 44 mjy N7 5 µm arcsec (6 arcsec) 55 mjy WIDE S 5 µm arcsec (3 arcsec) mjy WIDE L µm arcsec (5 arcsec) 3 mjy C 4 RAS, MNRAS 347, 3 9

3 ASTRO-F, the All-Sky Infrared Survey 5 Table. Infrared camera all-sky survey parameters. Filter Wavelength band FOV Spatial (in pixel) resolution 5σ sensitivity IRC-MIR-S9W 6.5 µm arcmin arcsec 5 mjy IRC-MIR-LW 4 6 µm arcmin arcsec mjy are divided into two further narrow and wide bands known as N6, WIDE S, N7, WIDE L, respectively (see Table ). The two arrays are tilted against the scan direction by 6.5. such that the interval between the scan paths traced by the detector pixels is half of the physical pitch of the pixels. The long- and short-wavelength arrays observe almost the same area of sky, the only difference being due to the difference in the size of the arrays. As well as the all-sky survey mode the FIS may also be used for pointing observations and includes a Fourier-transform spectrometer mode that can be operated via the wideband arrays with a resolution of. cm (λ/ λ = 5, λ = 5 µm allowing imaging spectroscopy of selected sources. In survey mode, ASTRO-F will perform a continuous scan of the sky. ASTRO-F spins around the Sun pointed axis once every orbit of min, keeping the telescope away from the Earth. The result is to trace out a great circle with a solar elongation of 9 scanning the sky at a rate of 3.6 arcmin s. The second focal-plane instrument is the infrared camera (IRC, Matsuhara 998; Watarai et al. ; Wada et al. 3). The IRC employs large-format detector arrays and will be used to take deep images of selected sky regions in the near-infrared and mid-infrared range after the All-Sky Survey has been performed. The IRC has three cameras each with three filters. The IRC-NIR has bands in K, L, M plus a grism and prism from.5 to 5 and. to 5 µm, respectively. The IRC-MIR-S has narrow bands at 7 and µm and a wide band at 9 µm, plus two grisms at 5 8 and 7 µm. The IRC-MIR-L has narrow bands at 5 and 4 µm, a wide band at µm, plus two grisms covering the range 9 and 8 6 µm. Pearson et al. () made detailed predictions for small, deep and large, shallow area survey strategies using the IRC instrument in pointing mode. The possibility of adding the two wide band channels of the IRC-MIR (S9W, SW) in scan mode is currently under investigation (Ishihara et al. 3). In all-sky survey mode the IRC would utilize just one line (perpendicular to the scan direction of the telescope beam) of the MIR detector arrays of the MIR-S and MIR-L channels. Data from these channels would be sampled and downloaded to the ground simultaneously with the FIS All-Sky Survey data. Typical angular resolutions of arcsec are expected (since integrated signal will be read out every arcsec). A significant portion (but not all) of the sky could be surveyed. In this paper we offer a preliminary analysis of an additional simultaneous all-sky survey using these two detector channels. The specifications of the IRC all-sky survey mode are shown in Table. The ASTRO-F All-Sky Survey will be far superior to the IRAS survey having almost two orders of magnitude better sensitivity at µm. Optimally, ASTRO-F will achieve sensitivities of mjy in the far-infrared (cf..5.5 Jy for IRAS at, 5, 6 and µm, respectively, (Soifer, Houck & Neugebauer 987). Spatial resolution with IRAS was limited to 5 arcmin. ASTRO-F will attain spatial resolutions of 5 7 arcsec in the short (5 µm) and long ( µm) wavelength bands, respectively (see Table ). Fig. plots the sensitivities expected for the FIS All-Sky Survey in Table against the spectral energy distributions (SEDs) of two sources from the Canada UK Deep SCUBA Survey (CUDSS, Lilly et al. Figure. ASTRO-F survey sensitivities (5σ ) for IRC and FIS instruments assuming all-sky survey sensitivities for the FIS (FIR wavelengths) and IRC (MIR wavelengths). IRAS flux limits and non-survey pointing sensitivities from Pearson et al. () for the IRC are also shown for comparison. Sensitivity limits are compared with two sources from the Canada UK Deep SCUBA Survey (Lilly et al. 999). CUDSSA has a spectroscopic redshift of.55 and a flux at 85 µm ofs 85 = 4.8 mjy. For CUDSS3A (S 85 = 3.4 mjy), the adopted redshift is.6, which gives a minimum chi-square fit to the SED (Takagi et al. ). Both of these sources would be detected by the IRC. CUDSSA would also be detected by the FIS in all four bands and even CUDSS3A would be detected marginally in the two most sensitive FIS bands. 999, SED fits from Takagi, Arimoto & Hanami ): CUDSSA and CUDSS3A, which have redshifts of.55 and.6 and 85- µm fluxes of 4.8 and 3.4 mjy, respectively. Both of these sources would be detected in the All-Sky Survey (CUDSS3A in only the two most sensitive FIS bands) and both sources would be easily detectable in all nine bands of the IRC instrument in pointing mode where detection limits in the near mid-infrared are expected to be 5 µjy (see Pearson et al. for IRC sensitivities in pointing mode). The NASA SIRTF mission (Fazio et al. 998), due for launch in early 3, offers better sensitivity at FIR MIR wavelengths [e.g. SIRTF-SWIRE covering 7 deg at 6 µm(s = 7.5 mjy), 7 µm (S =.75 mjy), 4 µm(s =.45 mjy), Lonsdale ]. However, these two missions should be seen as complementary rather than rivalling each other since SIRTF is an observatory and ASTRO-F a surveyor (a similar comparison may be made between the IRAS and ISO missions). ASTRO-F will cover the entire sky at far-infrared wavelengths, providing the first all-sky survey at 7 µm. In pointing mode, in the near mid-infrared ASTRO-F will be able to reach similar detection limits to SIRTF in nine bands, over areas of deg due to the larger field of view (FOV). Furthermore, in terms of large-scale structure, SIRTF will be sensitive to small spatial scales at high redshifts, while conversely ASTRO-F will be sensitive to large spatial scales at low redshifts (see Section 4.3). At.5 5 yr, Downloaded from by guest on October 8 C 4 RAS, MNRAS 347, 3 9

4 6 C. P. Pearson et al. SIRTF has a longer projected lifetime than ASTRO-F, although it has no capability at near-infrared wavelengths (<5 µm). ASTRO- F has the added capability of being an observatory and we will in fact be able to continue to use this instrument even after liquid helium exhaustion, since the NIR instrument it is sustained by the Stirling-cycle coolers. At present, the ASTRO-F mission can be divided into four distinct phases. (i) Phase, performance verification, 6 d. Initial check out and calibration of telescope and instruments. A mini-survey may also be attempted. (ii) Phase, all-sky survey, 8 d. ASTRO-F will scan the entire ecliptic longitude in survey mode. The all-sky survey will have first priority although there may be as many as 5 independent IRC pointings and 3 FIS pointings planned. (iii) Phase, pointing observations, 3 d. Any supplemental survey observations will be performed. The rest of this phase will be dedicated to pointing observations with the IRC and FIS instruments. This phase will last until helium exhaustion during which 6 pointings are expected. (iv) Phase 3, helium boil off, >365 d? Even after the liquid helium has been exhausted the mechanical coolers will keep the temperature low enough to enable the continuation of observations using the IRC-NIR detector. In this phase there is hope to be as many as 5 pointings. 3 MODEL PARAMETERS The galaxy evolution models of Pearson (, hereafter CPP) are used to estimate the source counts, the number of sources detected and other observables such as number redshift distributions and expected galaxy colours in the ASTRO-F All-Sky Survey. The CPP model is an improved extension of the original model of the 996 Pearson and Rowan Robinson model that provided an extremely good fit toiras data. The model incorporates a four component parametrization segregated by IRAS colours of galaxies (Rowan- Robinson & Crawford 989) consisting of normal, starburst and ultraluminous galaxies (Sanders & Mirabel 996), defined at 6 µm and an active galactic nuclei (AGN) population defined at µm. The cool [high S ( µm)/s (6 µm)] and warm component [low S ( µm)/s (6 µm)] 6-µm luminosity functions of Saunders et al. (99) are used to represent the normal, starburst and ultraluminous infrared galaxy (ULIG) galaxies, respectively, the ULIGs comprising the high luminosity tail of the warm luminosity function (L 6µm > L ). The AGN population utilizes the -µm luminosity function of Lawrence et al. (986) modelled on the sample of Rush, Malkan & Spinoglio (993) (this IRAS selected sample is consistent with the ISO subsample of Spinoglio, Andreani & Malkan ). Spectral energy distributions are required both for the K- corrections of sources and to transform the luminosity functions from their rest wavelength to the observation wavelength [i.e. by a factor L(λ obs )/L(λ LF ) where λ LF is the wavelength at which the luminosity function is defined and λ obs is the wavelength of observation]. These model spectra templates are taken from the new radiative transfer models of Efstathiou, Rowan-Robinson & Siebenmorgen (a) and Efstathiou & Rowan-Robinson (3) for the starburst and ultraluminous galaxy components and the normal galaxy component, respectively. The AGN spectral template is represented by the dust torus model of Rowan-Robinson (995). The CPP model assumes that the starburst, ULIG and AGN components evolve with cosmic time. The normal galaxy population does not evolve. The starburst and AGN components undergo positive luminosity evolution of the form L(z) = L()( + z) 3. to a redshift of z =.5 (Pearson & Rowan-Robinson 996), which provides a good fit to the source counts, at least at brighter fluxes from the radio to the X- ray regime (e.g. Boyle, Shanks & Peterson 988; Benn et al. 993; Pearson & Rowan-Robinson 996; Pearson et al. 997; Serjeant et al. ; Oliver et al. ). However, the subsequent strong evolution evident in the galaxy population at FIR wavelengths as detected by ISOPHOT on ISO (Dole et al. ; Efstathiou et al. b), at MIR wavelengths with ISOCAM (especially at 5 µm, where the source counts span two orders of magnitude in flux (Oliver et al. 997; Altieri et al. 999; Aussel et al. 999; Flores et al. 999a,b; Gruppioni et al. 999; Biviano et al. ; Elbaz ; Serjeant et al. ) and in the submillimetre with SCUBA on the JCMT (Smail et al. 997; Hughes et al. 998; Scott et al. ) has introduced the need for some extreme evolutionary scenarios in galaxy evolution models. Therefore, the ULIG population is assumed to undergo extreme evolution in both density and luminosity (see Pearson for details of these evolutionary models). The density evolution is of the form D(z) = + g exp [ (z z p ) /σ ] with g = 5, z p =.8, σ =. to z p and decaying quickly towards higher redshifts. The luminosity evolution is of the form L(z) = + k exp [ (z z p ) /σ ] with k = 4, z p =.5, σ =.58 to z p and decaying quickly to higher redshifts. In this scenario the luminosity evolution corresponds to the accretion of matter on to a core initiating star formation at high redshifts. The density evolution corresponds to major mergers at a later epoch. Locally, ULIGs have a space density comparable to optically selected QSOs (. deg ), although analysis of the IRAS Faint Source Catalogue leads to the conclusion that they should be much more numerous at higher redshift (Kim & Sanders 998). Moreover, both the ISO 5-µm differential source counts (Elbaz et al. 999) and the dust enshrouded star formation rate (Chary & Elbaz ; Elbaz et al. ) exhibit a peak due to sources at z.8. The evolutionary models of CPP provide an excellent fit to the source counts, number redshift distributions and the integrated background radiation from submillimetre to NIR wavelengths (Pearson and references therein). Although the FIR SED of galaxies is relatively featureless, the spectral templates are convolved with each individual FIS filter bandwidth to achieve smoothed model spectral templates for all components in all ASTRO-F FIS wavebands. The SEDs are also convolved with the two MIR wide bands. This convolution is more significant as unidentified infrared bands [UIBs probably due to polycyclic aromatic hydrocarbon (PAH) features in the galaxy SEDs, Puget & Leger 989, Boulade 996; Lu et al. 996; Vigroux et al. 996] may significantly affect any MIR observations (Xu et al. 998, ; Pearson et al. ). Indeed, the equivalent widths of these unidentified bands can be as much as µm, comparable to the ASTRO-F IRC-MIR wide filter bandpasses of S9W (6.5 µm) and LW (4 6 µm) (see Table ). Following growing consensus between observation and theory (e.g. Kravtsov, Klypin, Khokhlov 998; Balbi et al. ; Freedman et al. ), the model assumes H o = 7 km s Mpc, =.3, =.7 unless otherwise explicitly stated. 4 SURVEY PREDICTIONS The IRAS All-Sky Survey probed out to redshifts of.3., ASTRO-F is hoped to extend this range out to redshift. In order to assess the visibility of sources in the ASTRO-F All-Sky Survey, we plot in Fig. the luminosity-redshift distributions for each model C 4 RAS, MNRAS 347, 3 9 Downloaded from by guest on October 8

5 lg(l 6 /L o ) lg(l 6 /L o ) lg(l 6 /L o ) lg(l 6 /L o ) NORMAL GALAXIES N6 (44mJy) N7 (55mJy) WIDE-S (mjy) WIDE-L (3mJy) WIDE-9 (5mJy) WIDE- (mjy) STARBURST GALAXIES ULTRALUMINOUS GALAXIES SEYFERT GALAXIES (AGN) Figure. Visibility of ASTRO-F All-Sky Survey sources. A source must have 6-µm luminosity to be included in the all-sky survey of sensitivity, S(mJy) at redshift, z. Results are plotted for the four FIR bands (N6, N7, WIDE S, WIDE L) and proposed WIDE 9 and WIDE µm MIR bands for the four components in the galaxy evolution model. ASTRO-F, the All-Sky Infrared Survey 7 galaxy component at each all-sky survey wavelength. In essence Fig. plots the rest frame 6-µm luminosity that a source must have if it is to be detected in the survey in a given wavelength band with limiting flux S. We see that the normal galaxy population is quickly shifted beyond log(l 6 /L ) =.5 (the arbitrary limit for the normal galaxy population in CPP) at redshifts of.3 and.8 in the N6 and N7 bands, respectively, i.e. there will be no high-redshift (z > ) normal galaxies in the sample. However, the starburst population [starbursts, luminous infrared galaxies (LIGs), ULIGs] enjoy a stronger negative K-correction in the longer-wavelength bands as the bandpass climbs the Rayleigh Jeans slope towards the dust peak around 7 6 µm in the observed frame. Thus an L 6 = L ULIG (corresponding to a total FIR luminosity of.5 L, Helou, Soifer & Rowan-Robinson 984) should be detectable out to redshifts of z in the long-wavelength bands. Similarly the AGN have an almost flat luminosity curve out to redshifts of.5 3 as the wavelength bandpass slowly move up the long-wavelength SED towards the peak at around 3 µm. Fig. also emphasizes the constraints placed on the MIR all-sky survey by their relatively low sensitivity. At MIR wavelengths almost all the sources would be expected at z except for the AGN that have high luminosities and are also enjoying a sampling wavelength closer to the peak in their SED. In summary, from Fig. we may broadly expect four regimes of increasing redshift encompassing the normal, starburst, ULIG and AGN populations, respectively. 4. Source counts The ASTRO-F All-Sky Survey will provide a huge data base for the statistical study of galaxies. The zeroth-order moment of this statistical data base (for a complete sample of sources) is simply the number counts of detected sources. Integral galaxy number counts as a function of flux will indeed form the front line of the results from the ASTRO-F mission. Fig. 3 shows the predicted source counts calculated from the model explained in Section 3. Numbers for the all-sky survey are calculated using three different cosmologies assuming a Hubble constant of 7 km s Mpc,aflat =.3, =.7 universe, an open =.3, = universe and a flat =, = universe (note, changing the value of the Hubble constant will have no effect on the source counts). The source counts in the N6 and N7 narrow bands can be compared directly with observations by IRAS and ISO, respectively, whilst counts in the wide FIR bands will probe down to and below the confusion limit due to point sources. In general, the ultimate sensitivity for any FIR telescope at high galactic latitude will be determined by one of three contesting noise values. The instrument noise, the noise due to diffuse IR cirrus emission (Low et al. 984) which is spatially dependent or the confusion limit due to point sources (Helou & Beichman 99). In Table 3 we calculate the limiting flux sensitivity corresponding to the confusion limit due to point sources for each of the ASTRO-F All-Sky Survey bands for three different cosmological world models. We assume the classical confusion criteria of a source density of one source per 4 beams (Condon 974; Hogg ) of the observing instrument, where the beam diameter is given by d =. λ /D, where D is the telescope diameter (67 cm). Here we calculate the source confusion assuming a beam diameter equivalent to the full-with half maximum of the Airy disc, d =. λ /D such that the beam diameter at 7 µmis arcmin corresponding to a source confusion density of sources (where the subsequently calculated limiting flux will of course be model dependent). From Table 3 we see that for the shorter FIS bands and the MIR bands the confusion limit due to point sources is Downloaded from by guest on October 8 C 4 RAS, MNRAS 347, 3 9

6 8 C. P. Pearson et al. Figure 3. Predicted integral (top row) and normalized differential (middle row) source counts for the ASTRO-F All-Sky Survey in the four FIR bands (N6, N7, WIDE S, WIDE L) compared with expected sensitivities in mjy. Observed counts shown for N6 band are from IRAS data at 6 µm, Lonsdale et al. (99), Hacking & Houck (987), Rowan-Robinson et al. (99a), Saunders (99) and Gregorich et al. (995). Observed counts shown for N7 band are from ISO data at 7 µm, Kawara et al. (998), Puget et al. (999) and Matsuhara et al. () (fluctuation analysis, dotted box), Dole et al. (). The observed data at 6 and 7 µm are also overplotted as unfilled symbols for comparison, on the counts corresponding to the WIDE-S and WIDE-L bands. The bottom row shows the predicted source counts at MIR wavelengths in the proposed WIDE 9 and WIDE µm all-sky MIR survey bands. The sensitivities shown correspond to the nominal survey mode (highest values), enhanced survey sensitivity obtained by a lower data sampling rate in mjy (solid arrows) and the 5σ pointing sensitivity of the IRC instrument in µjy (dotted arrow). Table 3. Estimated confusion limits due to point sources for ASTRO-F All-Sky Survey bands. Band Flux Confusion source density Confusion limit (mjy) 5σ mjy (no of sources deg ) =.3, =.7 =.3, = =, = N N WIDE S 54 WIDE L WIDE WIDE Downloaded from by guest on October 8 not really dependent on the world model assumed and is on or well below the expected sensitivity capabilities of the instruments. For the longer-wavelength bands (N7, WIDE L) we would expect that the galaxy counts will be quite severely restricted by the source confusion due to the extremely steep slope of the source counts expected ( 3.3 compared with.5 expected for a non-evolving population in a Euclidean universe, Kawara et al. 998; Dole et al. ). Indeed, observations at 7 5 µm from the ISO FIRBACK survey (Lagache & Puget ) and Japanese Lockman Hole survey (Matsuhara et al. ) place even more severe limits on the source confusion level C 4 RAS, MNRAS 347, 3 9

7 suggesting that the confusion due to point sources would become severe at flux levels of 45 mjy. Although SIRTF boasts a larger diameter mirror (85 cm) than ASTRO-F and also has much better sensitivity at 7 µm (.75 mjy) and 6 µm (7.5 mjy), it too will feel the pinch of the source confusion. Estimated limits for SIRTF are mjy at 7 µm and 6 mjy at 6 µm. It should be noted that various innovative fluctuation analysis methods can be used to push below these limits (e.g. Matsuhara et al. ). The source counts in Fig. 3 are plotted by galaxy component and tabulated in Table 4. The source counts are rapidly increasing down to fluxes of mjy where they begin to flatten off. Approximately,.6 6 sources will be detected in the N6 band compared with more than 6 in the N7 band. The N7 band samples higher up the source count slope (closer to the turn over) and is more sensitive to higher-redshift galaxies due to the negative K-corrections (see Fig. ). The dust hump at the peak of the emission spectrum at and 6 µm for normal and starburst galaxies passes through the N7 band at redshifts of.7 and.8, respectively, where as the emission in the N6 band will already be decreasing. In the wide bands, approximately 6 and 47 6 galaxies may be detected in the WIDE-S and WIDE-L bands, respectively. The vast majority of sources detected in all bands will be luminous and ultraluminous galaxies. Expected numbers range fromto4 6 from N6 to WIDE L, respectively, although the exact numbers will be dependent on the assumed evolutionary scenario. The longer-wavelength FIR bands (N7/WIDE L) will be particularly sensitive to these galaxies. AGN will be preferentially detected in the short-wavelength FIR bands (N6/WIDE S) as the AGN emission peaks at 5 µm meaning that the longer wavelengths are still climbing the Rayleigh Jeans slope of the SED out to redshifts of 6 (see Fig. ). Approximately 7 AGN will be detected in the N6 band, with as many as 4 in the superior WIDE-S band. The N7 band will detect between 7, while the WIDE-L band should see as many as 45 (these numbers being dependent on the assumed cosmology). Comparison with the IRAS integral counts at 6 µm (N6 and WIDE-S bands) shows that the ASTRO-F survey will probe between 3 times deeper than the various IRAS surveys (.5 Jy for IRAS-PSC and mjy for the VFSS, Gregorich et al. 995; Bertin, Dennefield & Moshir 997). Moreover, a comparison with the IRAS 6-µm differential source counts (Hacking & Houck 987; Rowan- Robinson et al. 99a; Saunders 99; Gregorich et al. 995) emphasizes that ASTRO-F will be reaching very interesting flux levels where some discrimination between competing evolutionary scenarios will be possible. At the limit of the IRAS-PSC the counts are Euclidean (i.e. flat in the normalized differential counts, the middle panel of Fig. 3), the IRAS-VFSS survey extended this flux limit down to mjy and suggested an upturn in the differential source counts at 3 mjy similar to that observed in the differential counts at radio wavelengths due to the emergence of a radio submjy population of starburst galaxies (Condon 984). Kim & Sanders (998) re-analysed the IRAS-FSS data concluding that there was evidence of much stronger evolution at higher redshifts to z. The nature of this evolution cannot be constrained by the IRAS source counts at 6 µm with both luminosity evolution or density evolution being equally acceptable (Oliver, Rowan-Robinson & Saunders 99; Pearson ). ASTRO-F will be able to detect and quantify the upturn in the differential counts at fluxes < mjy. ASTRO-F will be the first ever all-sky survey at 7 µm. Since the peak of the energy output of dust enshrouded star formation lies in the FIR at 6 µm the N7 and WIDE-L bands will be particularly sensitive to high-redshift (z > ) galaxies as the peak C 4 RAS, MNRAS 347, 3 9 ASTRO-F, the All-Sky Infrared Survey 9 of the SED emission is redshifted to 7 µm atz.8 producing the phenomenon where galaxies are brighter at higher z than lower z as made apparent in Fig., (Franceschini et al. 99; Pearson & Rowan-Robinson 996). To date the largest survey at 7 µm was carried out by the FIRBACK team covering 4 deg down to 8 mjy (Dole et al. ). The ASTRO-F All-Sky Survey will be aptly placed below the limit of the FIRBACK survey in both the N7 and WIDE-L bands. Although the WIDE-L band will almost certainly be source confused, valuable work can still be carried out on and below the confusion limit (cf. the fluctuation analysis of ISO 7-µm source counts (Matsuhara ). The sensitivity of the FIR bands will also be good enough to provide discrimination between different evolutionary scenarios in the differential counts. The FIRBACK differential counts have already shown a steep non- Euclidean slope of 3.3 ±.6 at fluxes less than 5 mjy (cf. a Euclidean slope of.5). ASTRO-F should detect any turnover in the differential source counts. The lower panels of Fig. 3 shows the predictions for the possible additional survey in the two wide MIR wavebands at 9 and µm. In this case, the survey sensitivities will be somewhat lower than compared with the FIR bands. This is due to constraints on the data transmission rate of the ASTRO-F satellite. Nominal sensitivities are currently 5 and mjy at 9 and µm, respectively, with some possibility of increasing this to 3 and 7 mjy, respectively, by using a lower sampling rate. Predicted numbers in the WIDE 9 µm band are approximately 7 sources in total, mostly at local redshifts with approximately starburst galaxies and AGN. Results from the longer WIDE µm band are similar with approximately sources in total of which around half would be star-forming galaxies. Note that in the case of IRAS only 893 galaxies were obtained from the IRAS-FSS ( b > 5, >. Jy at µm, Moshir 99; Rush et al. 993). Furthermore, around sources detected by ISO were used to construct the ISO number counts at 5 µm Elbaz et al. 999; Gruppioni et al. 999; Serjeant et al. ). ASTRO-F will obtain an order of magnitude larger sample than the IRAS sample and the ISO 5-µm samples. Thus the local luminosity function of IR galaxies can be accurately derived since all galaxies at 9 µm must be detected by the FIS survey. ASTRO-F will also be an ideal tool for the discovery of rare objects such as hyperluminous (L 6µm > 3 L, Rowan-Robinson ) and primeval galaxies, cf. IRAS F4+474 discovered at the limit of the IRAS-FSS with S 6 =. Jy, at z =.86 and FIR luminosity 4 L (Rowan-Robinson et al. 99b). Totani & Takeuchi () have suggested that starbursting primordial ellipticals may have higher dust temperatures (4 8 K) than previously assumed. Such starbursting primordial ellipticals may also be observed in the ASTRO-F All-Sky Survey. The best strategy for such detections are (i) wide area shallow as opposed to narrow, deep surveys, since these objects are rare but bright with source densities of between.. deg depending on the evolutionary scenario and cosmological world model. (ii) Far-IR wavelengths since the bulk of any emission (9 99 per cent) should emerge at these wavelengths due to the copious amounts of dust expected due to star formation. (iii) Multiwavelength survey for IR colours for photometric redshifts, SED fitting and discrimination from other less luminous sources. An F4 type hyperluminous galaxy with luminosity of the order of a few 4 L would be detectable out to redshifts 4 in the FIS bands and at z > in the IRC-MIR bands. Depending on the cosmology, we could expect of the order of 3 such sources over the entire sky detected in the WIDE-S and WIDE-L bands and between 4 in the WIDE 9 and WIDE MIR bands, respectively. Corresponding follow-up and cross-correlations with future Downloaded from by guest on October 8

8 C. P. Pearson et al. Table 4. ASTRO-F All Sky Survey model predictions at expected 5σ sensitivities for three cosmological models: a flat =.3, =.7 universe, an open =.3, = universe and a flat =, = universe. The predicted number of sources is segregated by galaxy type for total numbers, number higher than a redshift of, number higher than a redshift of and percentage fractions. Numbers on the lower columns for the MIR bands are the expectations assuming a higher sensitivity due to a lower sampling rate. ALL COMPONENTS NORMAL GALAXIES STARBURST GALAXIES LIG/ULIGS AGN N N> %> N> %> N N> % > N> % > N N> %> N> %> N N> %> N> %> N N> %> N> %> N6 (44 mjy) = = = N7 (55 mjy) = = = WIDE-S ( mjy) = = = WIDE-L (3 mjy) = = = WIDE- ( mjy) (7 mjy) = = = WIDE-9 (5 mjy) (3 mjy) = = = Downloaded from by guest on October 8 C 4 RAS, MNRAS 347, 3 9

9 ASTRO-F, the All-Sky Infrared Survey lg(number of Sources) lg(number of Sources) lg(number of Sources) N6 (44mJy) All Components Normal Galaxies Starburst Galaxies LIG/ULIG Galaxies AGN WIDE-S (mjy) mJy WIDE-9 µm (5mJy) mJy lg(number of Sources) lg(number of Sources) lg(number of Sources) N7 (55mJy) WIDE-L (3mJy) WIDE- µm (mjy) mjy 7mJy Figure 4. Predicted number redshift distributions for the ASTRO-F All- Sky Survey in the four FIR bands (N6, N7, WIDE S, WIDE L) and proposed WIDE 9 and WIDE µm MIR bands assuming 5σ survey sensitivities in mjy. Also shown for the two MIR bands are the total contribution assuming an enhanced survey sensitivity obtained by a lower data sampling rate (long dashed lines). Vertical axis gives the total number of sources contained within a redshift bin of width δ z =.. A =.3, =.7 cosmology is assumed. millimetre wave instruments such as LMT ( e.g. a large area survey of a few deg ) would be particularly sensitive and will further constrain the detections of such objects, since a detection at millimetre wavelengths with no corresponding ASTRO- F counterpart will automatically imply a high redshift (and therefore high luminosity) for any given source. 4. Number-redshift distributions Fig. 4 shows the expected number redshift distributions for the ASTRO-F All-Sky Survey in the four FIS bands along with predictions for the MIR all-sky survey currently under consideration. Beyond a redshift of.3 the number redshift distributions in all the FIS bands are dominated by luminous/ultraluminous infrared galaxies consistent with the results of deep surveys with ISO (e.g. Dole et al. ; Elbaz et al. ). The narrow FIS bands (N6, N7) are only sensitive to normal and starburst galaxies out to redshifts of.5, the shortest band only detecting a few per cent of these galaxies above a redshift of unity. The longer-wavelength bands, due to the large K-corrections encountered (see Fig. ), do better at detecting sources at z > with approximately half the total number of detections expected to be at redshifts higher than unity. The WIDE-L band (and to a lesser extent the WIDE-S band), due to their better sensitivity are also sensitive to starburst galaxies at z with between / /8 and /3 /3 depending on the cosmology being found at z > in the WIDE-S and WIDE-L bands, respectively. As a result of the nature of the assumed AGN SED almost all AGN detected in the long-wavelength bands will be at z > (3 5 per cent for the short-wavelength bands), thus providing a unique sample of far-infrared, high-redshift AGN especially at 7 µm where the low detection rate means that only a large area (all-sky) survey such as the ASTRO-F survey can provide us with a statistically viable sample of such objects. In summary, the ASTRO- F survey will provide the data base for a comprehensive redshift survey out to redshift for normal and starburst galaxies and to redshift and higher for ULIG and AGN sources thus bridging the dark gap between the IRAS observations in the local Universe and the optical and submillimetre studies in the high-redshift regime. To investigate any correlation between the number redshift distributions further, we plot in Fig. 5, the N z residual. That is, the difference between the N z distributions at two wavelengths. Residual numbers are plotted as N(λ ) N(λ ). A positive value indicates that N(λ ) > N(λ ), a negative value that N(λ ) < N(λ ) and a value of zero indicates that all galaxies are seen in both wavelength bands, i.e. that the N z distributions are identical. The apparent periodicities in the distributions arise due to two contributing factors,the form of the evolution assumed and the galaxy K-correction as a function of redshift. These two factors conspire to produce peak source densities at redshifts that are dependent on the wavelength (see Table 5). From Fig. 5 we conclude that normal galaxies are preferentially detected in the long-wavelength bands due to the fact that the peak of their SED is around µm as are the LIG/ULIG sources due to the strong K-corrections incurred at redshifts up to.8. For starburst galaxies, at lower redshifts we are sampling the dust emission hump at 6 µm, thus the sources are preferentially detected in the short-wavelength bands, moving to higher redshift pushes the residuals in favour of the longer-wavelength bands. For AGN, the shorter-wavelength bands are preferred due to the flat nature of the SED from 3 to 3 µm. The 6-µm band climbs out of the Rayleigh Jeans region of the SED at z whereas the 7 µm remains on the Rayleigh Jeans slope until z > 4. Note, the residuals corresponding to the wide-narrow bandpasses simply show the increase in detections due to the enhanced sensitivity of the wide band filters over the narrow band filters. We find average redshifts, median(mean), of.9(.34),.99(.96),.63(.67),.97(.98) for the N6, N7, WIDE-S, WIDE-L bands, respectively. By comparison the IRAS QDOT sample had a median redshift of.3 (Rowan-Robinson et al. 99a) and even the IRAS Very Faint Source Survey analysis was limited to z <.3 (Gregorich et al. 995). In Table 5 a detailed analysis of the expected redshift values (mean, mode, median) is tabulated for all bands and all components for the ASTRO-F All-Sky Survey. Distinct populations sample distinct redshift regimes at factors of deeper than the IRAS surveys. The number-redshift distribution of the proposed MIR all-sky survey at 9 and µm shown in the bottom pair of panels in Fig. 4 is somewhat limited to redshifts <.5 (z < in the sparse sampling rate scenario, see the figures in italics in Table 4). However, even this relatively shallow sample will enable a more accurate determination of the local 9- and -µm luminosity functions. Previous recent studies of the and 5-µm luminosity functions with IRAS have been limited to z <.3 at S 5 > 5 mjy (Fang et al. 998; Shupe, Fang & Hacking 998). Even with the limitations of the proposed all-sky MIR survey (median redshift.5), the MIR luminosity function of galaxies will be accurately determined out to at least a Downloaded from by guest on October 8 C 4 RAS, MNRAS 347, 3 9

10 C. P. Pearson et al. factor of higher in redshift and a factor of 5 better sensitivity in flux. Figure 5. The residual number redshift distributions calculated from the difference in the N z distributions at two wavelengths for combinations given in the figure legend. The vertical axis gives residual number as N(λ ) N(λ ), a positive value indicating that N(λ ) > N(λ ) and a negative value indicating that N(λ ) < N(λ ). A value of zero indicates that all galaxies are seen in both wavelength bands. The apparent periodicities observed in the distributions are caused by the source K-corrections and evolution as a function of redshift (see the text). 4.3 The visibility function The ASTRO-F All-Sky Survey will be an invaluable tool in the study of large-scale structure. The IRAS survey mapped the local largescale structure of the Universe out to redshifts.. in a series of studies including 35 deg down to 6 mjy at 6 µm and deg down to 5 mjy at, 5, 6 µm (e.g. Lonsdale et al. 99; Rush et al. 993; Oliver et al. 996; Shupe et al. 998; Saunders et al. ) using the IRAS PSC (Beichman et al. 988) and FSS (Moshir 99) data sets. Wide sky coverage is a necessity for large-scale structure studies, as studies on smaller scales can be effected by localized fluctuations. ASTRO-F will probe to times deeper than IRAS over 5 times the area of the SIRTF-SWIRE program (Lonsdale ). A detailed analysis of the usefulness of ASTRO-F in the study of large-scale structure is beyond the scope of this present work and will be addressed in a later paper; in this work, we briefly outline the potential of such studies using the all-sky survey data. We approach our assessment via the density function and derived visibility function of galaxies in the ASTRO-F All-Sky Survey (see Von Hoerner 973; Condon 984 for a more detailed derivation and explanation of the visibility function). In general, the total number of sources observable down to some limit S is given by N(S) = φ(l) d log L z dv (z), () where φ is the local differential luminosity function per decade in luminosity L integrated over the volume element dv out to some limiting redshift z. Discarding the second half of the integral leaves φ(l)d log L, the number density of galaxies in a given redshift bin at redshift z. This is effectively the galaxy luminosity function in a bin of width dz centred on a redshift z, we call this the density function ρ (z). The density function for the ASTRO-F All-Sky Survey is shown in Fig. 6(a) for all the FIS and MIR bands. The density function is essentially a measure of the survey sensitivity and plotting the corresponding results for the SIRTF-SWIRE survey the superiority of the sensitivity of the latter can be clearly seen. By taking logarithmic steps in the density, a threshold redshift z o (ρ o ) can be read from the density function in Fig. 6(a). Using this redshift, z o, a corresponding cosmological volume is calculated which will be the sampled volume above a given galaxy density for the proposed survey, where V(ρ o ) = V(z o ). Convolving this volume with the total coverage of the survey gives us the quantity we refer to as the visibility function of sources in the all-sky survey and is a function of both the depth (sensitivity) and the coverage of any given survey. The visibility function is plotted in Fig. 6(b) and clearly shows the power and potential of the ASTRO-F All-Sky Survey for large-scale structure studies with two to three orders of magnitude in the volume being sampled at densities of log (ρ) = 7to 4 Mpc 3 compared with the equivalent SIRTF- SWIRE bands. In truth, the ASTRO-F and SIRTF-SWIRE surveys supplement rather than supplant one another and are extremely complementary with ASTRO-F covering scales from to h Mpc from z. to and the SIRTF-SWIRE survey covering the range from scales of to h Mpc from redshifts >. to 5. Downloaded from by guest on October 8 C 4 RAS, MNRAS 347, 3 9