THE COBE DIFFUSE INFRARED BACKGROUND EXPERIMENT SEARCH FOR THE COSMIC INFRARED BACKGROUND. I. LIMITS AND DETECTIONS

Transcription

1 THE ASTROPHYSICAL JOURNAL, 508:25È43, 1998 November 20 ( The American Astronomica Society. A rights reserved. Printed in U.S.A. THE COBE DIFFUSE INFRARED BACKGROUND EXPERIMENT SEARCH FOR THE COSMIC INFRARED BACKGROUND. I. LIMITS AND DETECTIONS M. G. HAUSER,1 R. G. ARENDT,2 T. KELSALL,3 E. DWEK,3 N. ODEGARD,2 J. L. WEILAND,2 H. T. FREUDENREICH,2 W. T. REACH,4 R. F. SILVERBERG,3 S. H. MOSELEY,3 Y. C. PEI,1 P. LUBIN,5 J. C. MATHER,3 R. A. SHAFER,3 G. F. SMOOT,6 R. WEISS,7 D. T. WILKINSON,8 AND E. L. WRIGHT9 Received 1998 January 7; accepted 1998 June 22 ABSTRACT The Di use Infrared Background Experiment (DIRBE) on the Cosmic Background Exporer (COBE) spacecraft was designed primariy to conduct a systematic search for an isotropic cosmic infrared background (CIB) in 10 photometric bands from 1.25 to 240 km. The resuts of that search are presented here. Conservative imits on the CIB are obtained from the minimum observed brightness in a-sky maps at each waveength, with the faintest imits in the DIRBE spectra range being at 3.5 km (I \64 nw m~2 sr~1, 95% conðdence eve) and at 240 km (I \ 28 nw m~2 sr~1, 95% conðdence eve). The bright foregrounds from interpanetary dust scattering and emission, stars, and interstear dust emission are the principa impediments to the DIRBE measurements of the CIB. These foregrounds have been modeed and removed from the sky maps. Assessment of the random and systematic uncertainties in the residuas and tests for isotropy show that ony the 140 and 240 km data provide candidate detections of the CIB. The residuas and their uncertainties provide CIB upper imits more restrictive than the dark sky imits at waveengths from 1.25 to 100 km. No pausibe soar system or Gaactic source of the observed 140 and 240 km residuas can be identiðed, eading to the concusion that the CIB has been detected at eves of I \ 25 ^ 7 and 14 ^ 3nWm~2 sr~1 at 140 and 240 km, respectivey. The inte- grated energy from 140 to 240 km, 10.3 nw m~2 sr~1, is about twice the integrated optica ight from the gaaxies in the Hubbe Deep Fied, suggesting that star formation might have been heaviy enshrouded by dust at high redshift. The detections and upper imits reported here provide new constraints on modes of the history of energy-reeasing processes and dust production since the decouping of the cosmic microwave background from matter. Subject headings: cosmoogy: observations È di use radiation È infrared: genera 1. INTRODUCTION The search for the cosmic infrared background (CIB) radiation is a reativey new Ðed of observationa cosmoogy. The term CIB itsef has been used with various meanings in the iterature; we deðne it here to mean a di use infrared radiation arising externa to the Miky Way. Measurement of this distinct radiative background, expected to arise from the cumuative emissions of pregaactic, protogaactic, and evoved gaactic systems, woud provide new insight into the cosmic dark ages ÏÏ foowing the decouping of matter from the cosmic microwave background (CMB) radiation (Partridge & Peebes 1967; Harwit 1970; Bond, Carr, & Hogan 1986, 1991; Franceschini et a. 1991, 1994; Fa, Charot, & Pei 1996). The search for the CIB is impeded by two fundamenta chaenges: there is no unique spectra signature of such a 1 Space Teescope Science Institute, 3700 San Martin Drive, Batimore, MD 21218; hauser=stsci.edu. 2 Raytheon STX, Code 685, NASA Goddard Space Fight Center, Greenbet, MD Code 685, NASA Goddard Space Fight Center, Greenbet, MD Caifornia Institute of Technoogy, IPAC/JPL, MS , Pasadena, CA Physics Department, University of Caifornia at Santa Barbara, Santa Barbara, CA Lawrence Berkeey Laboratory, Space Sciences Laboratory, Department of Physics, UC Berkeey, CA Massachusetts Institute of Technoogy, Room 20F-001, Department of Physics, Cambridge, MA Princeton University, Department of Physics, Jadwin Ha, Box 708, Princeton, NJ UCLA, Astronomy Department, Los Angees, CA background, and there are many oca contributors to the infrared sky brightness at a waveengths, severa of them quite bright. The ack of a distinct spectra signature arises in part because so many di erent sources of primordia uminosity are possibe (e.g., Bond et a. 1986), in part because the radiant characteristics of evoving gaaxies are imperfecty known, and in part because the primary emissions at any epoch are then shifted into the infrared by the cosmic redshift and possiby by dust absorption and reemission. Hence, the present spectrum depends in a compex way on the characteristics of the uminosity sources, on their cosmic history, and on the dust-formation history of the universe. Setting aside the difficut possibiity of recognizing the CIB by its anguar Ñuctuation spectrum (Bond et a. 1991; Kashinsky et a. 1996b; Kashinsky, Mather, & Odenwad 1996a), the ony identifying CIB characteristic for which one can search is an isotropic signa. Possibe evidence for an isotropic infrared background, or at east imits on emission in excess of oca foregrounds, has been reported on the basis of very imited data from rocket experiments (Matsumoto, Akiba, & Murakami 1988; Matsumoto 1990; Noda et a. 1992; Kawada et a. 1994). Puget et a. (1996) have used data from the COBE Far Infrared Absoute Spectrophotometer (FIRAS) to concude that there is tentative evidence for a CIB at submiimeter waveengths. Indirect upper imits, and even possibe ower imits, on the extragaactic infrared background have been inferred from the apparent attenuation of TeV c-rays in propagation from distant sources (de Jager, Stecker, & Saamon 1994; Dwek & Savin 1994; Bier et a. 1995; Stecker 1996; Stecker & de

2 26 HAUSER ET AL. Vo. 508 Jager 1997). However, the detection of TeV c-rays from Mrk 421 recenty reported by Krennrich et a. (1997) casts some doubt on the infrared background inferred in this manner. The integrated energy density of the CIB in units of the critica density might, on the basis of pre-cobe observations (Resse & Turner 1990), exceed that of the CMB, ) \ 1 ] 10~4 h~2, and preiminary DIRBE resuts (Hauser CMB 1995, 1996a, b) ony set imits on the integrated CIB that are comparabe to the energy density in the CMB. Direct detection of the CIB requires a number of steps. One must sove the formidabe observationa probem of making absoute brightness measurements in the infrared. One must then discriminate and remove the strong signas from foregrounds arising from oneïs instrument or observing environment, the terrestria atmosphere, the soar system, and the Gaaxy. Particuar attention must be given to possibe isotropic contributions from any of these foreground sources. This paper summarizes the resuts of the DIRBE investigation, in which a direct measurement of the CIB has been made by measuring the absoute sky brightness at 10 infrared waveengths and searching for isotropic radiation arising outside of the soar system and Gaaxy. We report upper imits on the CIB from 1.25 to 100 km and detection of the CIB at 140 and 240 km. Section 2 brieñy describes the DIRBE instrument and the character of its data. Section 2 aso summarizes the procedures used to mode foreground radiations and for estimating the random and systematic uncertainties in the measurements and the modes. Because the foreground modes are critica to our concusions, they are aso described more extensivey in separate papers. Detais of the interpanetary dust (IPD) mode used to discriminate the sky brightness contributed by dust in the soar system are provided by Kesa et a. (1998, hereafter Paper II). Arendt et a. (1998, hereafter Paper III) describe the Gaactic foreground discrimination procedures and summarize systematic errors in the foreground determination process. Section 3 of this paper summarizes the observationa resuts, presented in compact form in Tabe 2. Dwek et a. (1998, hereafter Paper IV) show in detai that the isotropic residuas detected at 140 and 240 km are not ikey to arise from unmodeed soar system or Gaactic sources. Section 4 of this paper summarizes that anaysis, provides a comparison of the DIRBE resuts with other di use brightness and integrated discrete source measurements, presents imits on the integrated energy in the cosmic infrared background impied by the DIRBE measurements, and brieñy discusses the impications of these resuts for modes of cosmic evoution. A more extensive discussion of the impications is provided in Paper IV. Independent con- Ðrmation of the DIRBE observationa resuts and extension of the CIB detection to onger waveengths is provided by Fixsen et a. (1998), as discussed in The remainder of this section provides an overview of the rather extensive arguments presented in this paper as a guide to the reader. From absoute brightness maps of the entire sky over 10 months of observation, the faintest measured vaue at each waveength is determined ( 3.1 and Tabe 2). These dark sky ÏÏ vaues are either direct measurements of the CIB (if we were fortuitousy ocated in the universe), or yied conservative upper imits on it. Since the measured infrared sky brightness is not isotropic at any waveength in the DIRBE range, it cannot be concuded that these dark sky vaues are direct detections of the CIB. As expected, the dark sky vaues are east near 3.5 km, in the reative minimum between scattering of sunight by interpanetary dust and reemission of absorbed sunight by the same dust, and at the ongest DIRBE waveength, 240 km, where emission from interstear dust is decreasing from its peak at shorter waveengths and the cosmic microwave background has not yet become signiðcant. To proceed further, the contributions from the soar system and Gaaxy to the DIRBE maps are determined. The contribution of interpanetary dust is recognizabe because motion of the Earth in its orbit through this coud causes annua variation of the sky brightness in a directions. An empirica, parametric mode of the IPD coud ( 2.3 and Paper II) is used to extract the IPD contribution. Athough this mode is not unique, Paper II demonstrates that the impications for the CIB are reasonaby robust, that is, rather insensitive to variations in the mode. The Gaactic contribution from discrete sources bright enough to be detected individuay is simpy deeted from further anaysis by banking a sma surrounding region in the maps. The integrated contributions of faint discrete Gaactic sources are cacuated at each waveength from 1.25 to 25 km from a detaied statistica mode of Gaactic sources and their spatia distribution ( 2.3 and Paper III). The contribution from the di use interstear medium (ISM) at each waveength is obtained by scaing a tempate map of ISM emission to that waveength. At a waveengths except 100 km, the tempate is the residua 100 km map after remova of the IPD contribution, a map where ISM emission is prominent. To remove the ISM contribution without removing some fraction of the CIB at other waveengths, the 100 km extragaactic ight is Ðrst estimated by extrapoating the H IÈ100 km correation to zero H I coumn density for two Ðeds, the Lockman Hoe (Lockman, Jahoda, & McCammon 1986) and north eciptic poe, where there is known to be itte other interstear gas (moecuar or ionized) in the ine of sight ( 3.4). This estimate is subtracted from the 100 km map before scaing it to other waveengths. The ISM tempate at 100 km was chosen to be the map of H I emission, scaed by the sope of the H I to 100 km correation ( 2.3 and Paper III). Ceary, drawing the proper concusions from the DIRBE measurements and foreground modes is criticay dependent upon assessment of the uncertainties in both the measurements and the modes. These uncertainties are discussed at ength in Papers II and III and are summarized here in 2.4 and Tabe 2. Because the foreground emissions are so bright, the deðnitive search for evidence of the CIB is carried out on the residua maps after remova of the soar system and Gaactic foregrounds in a restricted region of the sky at high gaactic and eciptic atitudes (designated the highquaity ÏÏ region B, HQB, discussed in 3.3 and deðned in Tabe 3). The HQB region is the argest area in which the residua maps do not ceary contain artifacts from the foreground remova and covers about 2% of the sky. It incudes regions in both the northern and southern hemispheres and aows isotropy testing on 8140 map pixes over anguar scaes up to 43 degrees within each hemisphere and from 137 to 180 degrees between hemispheres. In this region, the mean residuas are determined and their uncertainties are estimated. More precise estimates of the mean residuas at 100, 140, and 240 km are obtained from a weighted average

3 No. 1, 1998 DIRBE DETECTIONS AND LIMITS 27 of vaues determined in the HQB region and in we-studied faint regions toward the Lockman Hoe, and the north eciptic poe. The residuas are tested for signiðcance by requiring that they exceed 3 times the estimated uncertainty incuding both random and systematic e ects. The Ðna step toward recognition of the CIB is to test for isotropy of the residuas. Athough a number of approaches are discussed ( 3.5), the concusions are Ðnay based upon the absence of signiðcant spatia correations of the residuas with any of the foreground modes or with gaactic or eciptic atitude and the absence of signiðcant structure in the two-point correation function in the HQB region. Ony at 140 and 240 km do the resuts meet our two necessary criteria for CIB detection: signiðcant residua in excess of 3 p and isotropy in the HQB region ( 3.6). These isotropic residuas are unikey to arise from unmodeed soar system or Gaactic sources ( 4.1 and Paper IV), eading to the concusion that the CIB has been detected at 140 and 240 km. At each waveength shorter than 100 km, an upper imit to the CIB is set at 2 p above the mean HQB residua, which in a cases is a more restrictive imit than the dark sky imit. At 100 km, the most restrictive imit is found from the weighted average of the residuas in the HQB region, the Lockman Hoe and the north eciptic poe. The ast ine of Tabe 2 shows the Ðna CIB imits and detected vaues. 2. DIRBE, DATA, AND PROCEDURES This section provides a brief review of the important features of the DIRBE instrument, the data it provides, and our reduction of the data with the goa of extracting the CIB. These topics are more thoroughy described in the COBE/DIRBE Expanatory Suppement (1997) and Papers II and III DIRBE Instrument Description The COBE Di use Infrared Background Experiment was the Ðrst sateite instrument designed speciðcay to carry out a systematic search for the CIB in the 1.25È240 km range. A detaied description of the COBE mission has been given by Boggess et a. (1992), and the DIRBE instrument has been described by Siverberg et a. (1993). The DIRBE observationa approach was to obtain absoute brightness maps of the fu sky in 10 broad photometric bands at 1.25, 2.2, 3.5, 4.9, 12, 25, 60, 100, 140, and 240 km. Tabe 1 summarizes the instrumenta parameters, incuding the e ective bandwidth, beam soid ange, detector type, and Ðter construction. Athough inear poarization was aso measured at 1.25, 2.2, and 3.5 km, the poarization information has not been used in this anaysis. DIRBE characteristics of particuar reevance to the CIB search incude the foowing: 1. Highy redundant sky coverage over a range of eongation anges.èbecause the di use infrared brightness of the entire sky varies as a resut of our motion within the IPD coud (and possibe variations of the coud itsef), the DIRBE was designed to scan haf the sky every day, providing detaied ight curves ÏÏ with hundreds of sampes over the mission for every pixe. This samping provides a strong means of discriminating soar system emission. The scanning was produced by o setting the DIRBE ine of sight by 30 from the COBE spin axis, which was normay Ðxed at a soar eongation ange of 94, providing samping at eongation anges ranging from 64 to 124. Such samping aso moduates the signa from any nearby sphericay symmetric Sun- or Earth-centered IPD component, which woud otherwise appear as a constant (i.e., isotropic ÏÏ) signa. 2. Sensitivity.ÈTabe 2 ists 1 p instrumenta sensitivities, p(i ), for each 0.7]0.7 Ðed of view over the compete 10 months of cryogenic operation. These singe Ðed-of-view vaues are generay beow the actua sky brightness and beow many of the predictions for the CIB. Averaging over substantia sky areas, once foregrounds are removed, increases the sensitivity for an isotropic signa. 3. Stray ight rejection.èthe DIRBE optica conðguration (Magner 1987) was carefuy designed for strong rejection of stray ight from the Sun, Earth imb, Moon, or other o -axis ceestia radiation, as we as radiation from other parts of the COBE payoad (Evans 1983). Extrapoations of the o -axis response to the Moon indicate that stray ight contamination for a singe Ðed of view in faint regions of the sky does not exceed 1 nw m~2 sr~1 at any waveength (COBE/DIRBE Expanatory Suppement 1997). 4. Instrumenta o sets.èthe instrument, which was maintained at a temperature beow 2 K within the COBE superñuid heium Dewar, measured absoute brightness by chopping between the sky signa and a zero-ñux interna TABLE 1 DIRBE INSTRUMENT CHARACTERISTICS ja * e b Beam Soid Ange Absoute Caibration Band (km) (Hz) (10~4 sr) Detector Type Fiter Constructionc Reference Source d 5.95 ] InSbe Coated gass Sirius d 2.24 ] InSbe Coated gass Sirius d 2.20 ] InSbe Coated germanium Sirius ] InSbe MLIF/germanium Sirius ] Si:Ga BIB MLIF/germanium/ZnSe Sirius ] Si:Ga BIB MLIF/siicon NGC ] Ge:Ga MLIF/sapphire/KRS5/crysta quartz Uranus ] Ge:Ga MLIF/KC/CaF /sapphire Uranus ] Si/diamond boometer Sapphire/mesh grids/baf 2 /KBr Jupiter ] Si/diamond boometer Sapphire/grids/BaF /CsI/AgC 2 Jupiter 2 a Nomina waveength of DIRBE band. b E ective bandwidth assuming source spectrum I \ constant. c MLIF \ mutiayer interference Ðter. d Linear poarization and tota intensity measured. e AntireÑection coated for the band center waveength.

4 TABLE 2 RESULTS OF THE DIRBE SEARCH FOR THE CIB WAVELENGTH (km) MEASURED QUANTITY (ISM2) 1. p(i ) S(gain) (%) S(o set) I (dark) ^ ^ 4 60^2 183 ^ ^ ^ ^ ^ ^ 8 22^3 22^3 5. I (NEP)... 40(3.0)(18) 6(1.3)(11) 6(0.7)(6) 16(0.3)(7) 125(0.3)(89) 145(0.2)(98) 30(0.2)(17) 21(0.3)(8) 16(1.4)(12) 7.2(0.5)(5) 22(1.3)(12) 6. I 0 (SEP)... [9.2(6.4)(18) [5.7(3.0)(11) [2.4(1.1)(6) 16(0.6)(7) 112(0.4)(88) 122(0.2)(98) 21(0.1)(17)... 37(2.6)(12) 20(0.8)(5) 16(2.4)(12) 7. I 0 (NGP)... 42(2.1)(21) 21(0.9)(12) 15(0.5)(6) 32(0.2)(8) 261(0.4)(138) 247(0.2)(156) 24(0.1)(27) 29(0.2)(10) 11(1.8)(12) 5.5(0.6)(5) 17(1.7)(12) 8. I 0 (SGP)... 36(2.3)(20) 19(0.9)(11) 16(0.5)(6) 36(0.3)(8) 255(0.5)(128) 232(0.3)(145) 22(0.1)(25) 24(0.2)(9) 7.0(1.9)(12) 4.5(0.6)(5) 20(1.8)(12) 9. I 0 (LH)... 69(5.9)(19) 26(2.4)(11) 16(1.1)(6) 30(0.6)(7) 208(0.3)(102) 215(0.3)(112) 24(0.1)(19) 23(0.2)(9) 18(3.4)(12) 11(1.0)(5) 14(3.2)(12) 10. I 0 (HQA) (0.5)(21) 13(0.2)(12) 11(0.1)(6) 26(0.1)(8) 195(0.2)(138) 192(0.2)(156) 22(0.1)(27) 21(0.1)(10) 19(0.3)(12) 9.4(0.1)(5) 16(0.3)(12) 11. I (HQB) (1.6)(21) 14.9(0.6)(12) 11.4(0.4)(6) 24.8(0.2)(8) 192(0.2)(138) 192(0.2)(156) 20.6(0.1)(27) 19.0(0.1)(9.5) 18.5(0.5)(12.0) 9.7(0.2)(5.2) 13.8(0.5)(12) 12. I 0 (LH@, HI) (0.2)(6.1) 26.6(0.9)(7.3) 13.9(0.3)(2.6) I (NEP@, HI) (0.4)(6.3) 23.1(1.9)(13.2) 14.3(0.6)(4.5) L I (HQB) ^ ^ 10.9 [19.5 ^ 5.3 [33.4 ^ 2.7 [159 ^ 3.6 [115 ^ ^ 0.9 [26.0 ^ ^ ^ ^ 14.6 b L I (HQB)... b 0 [9.3 ^ ^ ^ ^ ^ ^ 1.9 [0.8 ^ ^ 0.9 [9.5 ^ 7.2 [3.5 ^ ^ I (95% CL)... \75 \39 \23 \41 \468 \504 \75 \38 \43 \20 \ SI 0 T ^ ^ ^ Isotropy 0... No No No No No No No No Yes Yes Yes 19. CIB... \75 \39 \23 \41 \468 \504 \75 \ ^ ^ NOTE.ÈSubscript 0 means residua. Resuts are in units of nw m~2 sr~1. Heading row: nomina waveength of the DIRBE photometric bands. The ast coumn presents resuts at 240 km using the two-component ISM mode. Row 1: mission-averaged 1 p instrument sensitivity per 0.7]0.7 Ðed of view. Row 2: percent uncertainty in the absoute gain caibration. Row 3: uncertainty in the instrumenta o set (zero point) caibration. Row 4: minimum observed sky brightness. Errors are estimated from the quadrature sum of the gain and o set uncertainties. Rows 5È9: residua intensity derived using the 100 km map as a tempate for the ISM at sma patches toward the north (NEP) and south (SEP) eciptic poes, the north (NGP) and south (SGP) gaactic poes, and the Lockman Hoe (LH). The 100 km coumn shows residuas using the Be Labs H I map as the ISM tempate. Fina coumn shows resuts using the two-component (100 and 140 km) ISM tempate. Random (systematic) 1 p errors are shown in the Ðrst (second) parentheses. Rows 10È11: residua intensity derived using the 100 km map as a tempate for the ISM (Be Labs H I tempate in the 100 km coumn and two-component tempate in the ast coumn) at the high-quaity ÏÏ HQA and HQB regions deðned in Tabe 3 in units of nw m~2 sr~1. Random (systematic) 1 p errors are shown in the Ðrst (second) parentheses. Rows 12È13: residua intensity derived from the H I correation at the Lockman Hoe (LH) and the north eciptic poe (NEP). Random errors (systematic errors) are shown in the Ðrst (second) parentheses. Rows 14È15: spatia gradients of the residua intensities in the HQB region with respect to csc o b o and csc o b o. Row 16: 2 p upper imit on the CIB, derived from the HQB region using random and systematic uncertainties combined in quadrature. The 240 (ISM2) entry is derived from the dark sky vaue, since it yieds a more restrictive imit. Row 17: weighted average of the residuas at the HQB region, the Lockman Hoe, and the north eciptic poe (rows 11, 12, and 13), weighted by the quadrature sum of the random and nonècommon-mode systematic errors in each region. Errors in the mean incude the random and a systematic errors. Row 18: resut of the two-point correation function isotropy test of the residua intensities in the HQB region: no \ not consistent with isotropy; yes \ consistent with isotropy. Row 19: DIRBE detections of the CIB or 95% conðdence upper imits.

5 DIRBE DETECTIONS AND LIMITS 29 reference at 32 Hz. Instrumenta o sets were measured about 5 times per orbit by cosing a cod shutter ocated at the prime focus. A radiative o set signa in the ongwaveength detectors arising from junction Ðed e ect transistors (JFETs, operating at about 70 K) used to ampify the detector signas was identiðed and measured in this fashion and removed from the DIRBE data. Because the o set signa was stabe over the course of the mission, it woud appear as an isotropic signa if eft uncorrected. To estabish the origin of the radiative o set signa and determine if its vaue was the same whether the instrument shutter was cosed (when the o set was monitored) or open (when the sky brightness pus o set was measured), specia tests were conducted during two one-week periods of the mission. In these tests, power to individua JFETs was turned o sequentiay whie measuring the o set (shutter cosed) and sky brightness (shutter open) with a remaining operating detectors. The sky brightness measurements at each waveength with JFETs o and on at other waveengths were carefuy compared. The o sets measured in this fashion were consistent with those measured by cosing the shutter in norma operations, demonstrating that changing the position of the shutter did not signiðcanty modify the o set. The Ðna uncertainties in the o set corrections, shown as S(o set) in Tabe 2, are dominated by the uncertainties in the resuts of these specia tests due to the imited amount of time devoted to them. The uncertainties are quite negigibe at waveengths ess than 140 km. The accuracy of the DIRBE measurement zero point at 140 and 240 km, where the o set uncertainty exceeds 1 nw m~2 sr~1, has been independenty conðrmed by comparison with COBE/ FIRAS data, as discussed beow. 5. Gain stabiity.èshort-term stabiity and inearity of the instrument response were monitored using interna radiative reference sources that were used to stimuate a detectors each time the shutter was cosed. The highy redundant sky samping aowed the use of stabe ceestia sources to provide precise photometric cosure over the sky and reproducibe photometry to D1% or better for the duration of the mission. 6. Absoute gain caibration.ècaibration of the DIRBE photometric scae was obtained from observations of a few isoated bright ceestia sources (COBE/DIRBE Expanatory Suppement 1997). Tabe 1 ists the DIRBE gain reference sources, and Tabe 2 ists the uncertainties in the absoute gain, S(gain), for each DIRBE spectra band. An independent check of the DIRBE o set and absoute gain caibrations at 100, 140, and 240 km has been performed by Fixsen et a. (1997) using data taken concurrenty by the FIRAS instrument on board COBE. The FIRAS caibration is intrinsicay more accurate than that of the DIRBE, but the FIRAS sensitivity drops rapidy at waveengths shorter than 200 km, e ectivey ony partiay covering the DIRBE 100 km bandpass. In genera, the two independent caibrations are consistent within the estimated DIRBE uncertainties. Quantitativey, Fixsen et a. (1997) evauated the gain and o set corrections needed to bring the two sets of measurements into agreement. Taking account of the absoute FIRAS caibration uncertainty and the uncertainty arising from the comparison process itsef (owing in part to the need to integrate the FIRAS data over the broad DIRBE spectra response in each band and to integrate the DIRBE data over the arge FIRAS beam shape to obtain comparabe maps), Fixsen et a. (1997) found statisticay signiðcant, but sma, corrections (3 p or greater) to the DIRBE caibration ony at 240 km. A resuts in this paper are based upon the DIRBE caibration and its uncertainties. The sma e ect of adopting the FIRAS caibration at 140 and 240 km, which has no quaitative e ect on the concusions presented here, is discussed in T he DIRBE Data The caibrated DIRBE photometric observations are made into maps of the sky by binning each sampe into a pixe on the COBE sky-cube projection in geocentric eciptic coordinates (COBE/DIRBE Expanatory Suppement 1997). The projection is neary equa area and avoids geometrica distortions at the poes. Pixes are roughy 20@ on a side. Forty-one weeky maps have been produced by forming a robust average of a observations of each pixe taken during a week. About one-haf of the sky is covered each week; compete sky coverage is achieved within 4 months. Data used in this anaysis originate from the weeky sky maps produced by the 1996 Pass 3b DIRBE pipeine software, as documented in the COBE/DIRBE Expanatory Suppement (1997). A anaysis is performed on maps in the origina skycube coordinate system. For iustrationa purposes, the maps shown in Figure 1 of this paper are reprojected into an azimutha equa-area projection. The DIRBE surface brightness maps are stored as I in units of MJy sr~1. Many of the resuts in this paper are presented as I, where I (nw m~2 sr~1) \ (3000 km/j)i (MJy sr~1) Foreground-Remova Procedures Conservative upper imits on the CIB are easiy determined from the minimum sky signa observed at each waveength; these resuts are quoted in 3.1. In order to derive more interesting imits or detections, one must address the probem of discriminating the various contributions to the measured sky brightness. The procedures used to discriminate and remove foreground emissions from the soar system and Gaaxy are carefuy based on distinguishing observationa characteristics of these sources. Isotropy of the residuas was not assumed or imposed, but was rigorousy tested ( 3.5). The approach adopted here is to derive, for each DIRBE waveength, j, an a-sky map of the residua intensity I remaining after the remova of soar system and Gaactic res foregrounds from the observed sky brightness I : obs I (, b, j) \ I (, b, j, t) [ Z(, b, j, t) [ G(, b, j), (1) res obs where and b are gaactic ongitude and atitude, t is time, Z(, b, j, t) is the contribution from the interpanetary dust coud, and G(, b, j) is the contribution from both stear and interstear dust components within the Gaaxy. Both Z and G are derived from modes. The choice of modes is motivated by the primary goa of ensuring that no part of the CIB is inadvertenty incuded in the interpanetary dust coud or Gaactic emission components. Figure 1 presents maps of I as derived from the foreground-remova process. res The DIRBE IPD mode (Paper II) is a semiphysica, parametric mode of the sky brightness simiar, but not identica, to that used to create the IRAS Sky Survey Atas (Wheeock et a. 1994). The mode represents the sky bright-

6 30 FIG. 1.ÈDIRBE residua intensity maps after remova of foreground emission at 1.25È240 km, shown in gaactic coordinates with an azimutha equa-area projection. The eft (right) circe represents the projected north (south) Gaactic hemisphere with b \]90 (b \[90 ) in the center and b \ 0 at the edge. Contours of Ðxed atitude are concentric circes with r P [(1 [ sin o b o )/2]. Longitude ines run radiay from the poe to the edge and increase cockwise (countercockwise) on the eft (right) hemisphere. The ongitude \ 0 runs from the center to the bottom edge of each projected hemisphere. A maps are potted on a inear scae with coor-coded brightness ranges, from shortest to ongest waveength, of ([0.05, 0.3), ([0.05, 0.3), ([0.01, 0.2), (0, 0.2), (0, 2), (0.5, 3), (0, 3), (0, 15), (0, 20), and (0, 20) in units of MJy sr~1. Vaues beow (above) the pot range are shown in back (white). The ast pane indicates the sky ocations of the Ðve sma patches ( 3.2) and the two seected high-quaity regions ( 3.3): the tiny white square in the eft hemisphere is the Lockman Hoe, the centered white square in the eft (right) hemisphere is the north (south) Gaactic poe, and the remaining white square in the eft (right) hemisphere is the north (south) eciptic poe; the arge bue area in the eft (right) sphere is the north (south) high-quaity region A, and the sma yeow area in the eft (right) hemisphere is the north (south) high-quaity region B.

7 31 FIG. 1.ÈContinued

8 FIG. 1.ÈContinued

9 DIRBE DETECTIONS AND LIMITS 33 ness as the integra aong the ine of sight of the product of an emissivity function and a three-dimensiona dust density distribution function. The emissivity function incudes both therma emission and scattering. The therma emission at each ocation assumes a singe dust temperature for a coud components. The temperature is a function ony of distance from the Sun and varies inversey as a power aw with distance. The density distribution incudes a smooth coud, three pairs of asteroida dust bands, and a circumsoar dust ring. The mode is intrinsicay static, except that structure within the circumsoar ring near 1 AU is assumed to co-orbit the Sun with the Earth. The apparent seasona brightness variation arises from the motion of the Earth on an eccentric orbit within the coud, which is not required to be symmetric with respect to the eciptic pane. Anaytica forms are assumed for the density distributions, scattering phase function, and therma emission characteristics of the dust. Parameters for the anaytica functions are determined by optimizing the mode to match the observed tempora variations in brightness toward a grid of directions over the sky. By Ðtting ony the observed time variation to determine the mode parameters, Gaactic and extragaactic components of the measured brightness are totay excuded. However, it must be emphasized that this method cannot uniquey determine the true IPD signa; in particuar, an arbitrary isotropic component coud be added to the mode without a ecting the parameter vaues determined in our Ðtting to the seasona variation of the signa. No such arbitrary constants are added to the brightnesses obtained directy from our mode, and imits on unmodeed isotropic components of the IPD coud emission are set based upon independent knowedge of the nature of the coud ( 4.1). Once the optima mode parameters are determined, the IPD mode is integrated aong the ine of sight to evauate Z at the mean time of observation of each DIRBE pixe for each week of the mission. The cacuated IPD map is then subtracted from each DIRBE weeky map, and an average mission residua is computed. This simpe mode represents the IPD signa fairy we, but there are ceary systematic artifacts in the residuas at the eve of a few percent of the IPD mode brightness (Paper II). Because the zodiaca emission is so bright, uncertainties in the residua sky maps at 12È60 km are dominated by the uncertainties in the IPD signa. The Gaactic mode G is removed from the missionaveraged residuas formed after remova of the IPD contribution (Paper III). The Gaactic mode actuay consists of three separate components: bright discrete sources, faint discrete sources, and the interstear medium. Both stear and extended discrete sources whose intensity above the oca background exceeded a waveength-dependent threshod are excuded by banking a sma surrounding region from each of the 10 maps. The banked regions appear back in Figure 1 and are most evident in the 1.25È4.9 km maps and at ow gaactic atitude. The contribution from faint discrete sources beow the bright-source banking threshod at 1.25È25 km is then removed by subtracting the integrated ight from a statistica source-count mode based on that of Wainscoat et a. (1992), with eaborations by Cohen (1993, 1994, 1995). We ca this the faint source mode (FSM). The use of a source-count based mode ensures that the reated intensity represents ony Gaactic sources. The stear contribution is negected at waveengths ongward of 25 km. The basic mode of emission from interstear dust, G (, b, j), consists of a standard spatia (waveength independent) I tempate of the brightness of the interstear medium (ISM), scaed by a singe factor R(j) at each waveength. The factor R(j) is determined by the sope of a inear correation of the standard spatia tempate with the intermediate residua map at waveength j obtained from the measured map, I (, b, j, t), by subtraction of the IPD mode, banking of bright obs sources, and subtraction of the FSM. The ISM spatia tempate is constructed so that it does not contain di use extragaactic emission. To the extent that this is successfu, when the scaed ISM tempate at any waveength, G (, b, j), is subtracted from the intermediate residua map at I that waveength, any CIB signa in the resuting Ðna residua map I (, b, j) is not modiðed. This inear ISM mode works we res in that it removes the evident cirrus couds, especiay in the high gaactic atitude regions where the search for the CIB is conducted. Severa approaches have been used to create the ISM spatia tempate. In one approach, the 100 km ISM map, G (, b, 100 km), obtained by subtracting the contributions I from the IPD and bright and faint discrete Gaactic sources from the observed map at 100 km, was used as the spatia tempate for a other waveengths from 12 to 240 km. No signiðcant ISM emission coud be identiðed at 1.25 and 2.2 km, and a modiðed form of this procedure was required to detect the weak ISM emission at 3.5 and 4.9 km (Paper III). The use of the 100 km ISM emission as the tempate at other waveengths has the advantages of good signa-tonoise ratio and an idea match of anguar resoution with the other DIRBE data. Furthermore, the use of an infrared map as the tempate automaticay incudes contributions from dust in a gas phases of the ISM. The procedure used to estimate the 100 km CIB signa so as to remove it from the 100 km ISM map is described brieñy beow and in 3.4. For additiona anaysis of the 240 km map, a twocomponent mode of the ISM emission ( ISM2 ÏÏ) was aso generated. This mode used a inear combination of the DIRBE 100 and 140 km ISM maps as a tempate. Athough the one-component (100 km) mode ( ISM1 ÏÏ) appears to work adequatey at high atitudes, where we coud best test for isotropic residuas, the ISM2 mode can account for spatia variations in dust temperature throughout the ISM (Paper III). This eads to a more accurate mode of the ISM emission, particuary at ow gaactic atitudes, and a residua map I (, b, 240 km) that is more weaky correated with the ISM res tempate than in the case of the ISM1 mode. Figure 1 shows maps of I (, b, j) at240km for both the ISM1 and ISM2 modes. res To search for evidence of an isotropic CIB residua at 100 km, an ISM spatia tempate independent of the measured 100 km map was needed. For this purpose a veocityintegrated map of H I coumn density was used as the spatia tempate of the ISM emission. The range of veocities in the H I map was restricted so that it contained ony Gaactic H I emission. The success of this procedure of course depends on the accuracy with which the H I traces the dust distribution, at east at the high gaactic atitudes of interest here. Paper III provides extensive discussion of the uncertainty in the correation of infrared brightness with H I coumn density. The H I spatia tempate used to remove ISM emission from the map at 100 km was the Be Labs H I survey (Stark et a. 1992). This survey has the advantages of a we-

10 34 HAUSER ET AL. Vo. 508 estabished baseine and arge area coverage, but the disadvantage of ower anguar resoution than the DIRBE data. Higher resoution H I data (Evis, Lockman, & Fassnacht 1994; Snowden et a. 1994) obtained in sma regions where there are observationa constraints on the amount of moecuar and ionized materia (Paper III) and caibrated with the Be Labs H I survey were used to estabish the scaing factor between the H I and 100 km ISM emission. These same high-resoution data were used to estimate the 100 km brightness at zero H I coumn density so as to remove di use extragaactic emission from the ISM spatia tempate, G (, b, 100 km), used at a other waveengths as discussed above I (see 3.4). The 100 kmèh I correation was aso evauated using the new Leiden/Dwingeoo H I survey (Hartmann & Burton 1997), but this made itte di erence in the scaing factor or the residua intensity I (, b, 100 km). Use of the Leiden/ Dwingeoo H I survey as res the spatia tempate of the ISM at 100 km produces a ceaner map of residua emission I (, b, 100 km) than does use of the Be Labs data because res of a better match to the DIRBE anguar resoution, but the differences are not very apparent in maps made in the projection and scae of those in Figure 1. Resuts quoted in this paper are based on the Be Labs H I survey and other observations that are directy caibrated to that data set (Evis et a. 1994; Snowden et a. 1994) Uncertainties For this anaysis it is usefu to make distinctions between three forms of uncertainties. First are the random uncertainties, which incude instrumenta noise, uncorrected instrument gain variations, random Ñuctuations of the stear distribution, and certain deðciencies in the foreground modeing procedures. The key property of random uncertainties is that they are reduced as one averages over onger time intervas or arger regions of the sky. Tabe 2 ists typica vaues for the detector noise per pixe averaged over the entire mission, p(i ), assuming 400 observations per pixe. The boometer detectors used at 140 and 240 km are distincty ess sensitive than the other detectors. The second form of uncertainty is the gain uncertainty. This is the uncertainty in the gain factor used in the absoute caibration of the DIRBE data. Athough the gain uncertainty does a ect the quoted intensities, incuding the residua intensities, in a systematic way, it does not ater the signa-to-noise ratio of the resuts or the detectabiity of an isotropic residua signa using our methods. We therefore distinguish the gain uncertainty, shown as S(gain) for each waveength band in Tabe 2, from other systematic errors. Finay there are the systematic uncertainties, which are the uncertainties in the data and the foreground modes that tend to be isotropic or very arge scae. The systematic uncertainties cannot be reduced by averaging and therefore are the utimate imitations in the detection of the CIB. Tabe 2 ists the detector o set uncertainties, p(o set). The o set uncertainties are important contributors to the tota uncertainty ony at 140 and 240 km. The systematic uncertainties of the IPD mode, the stear emission mode, and the ISM mode are important, respectivey, at 1.25È100, 1.25È4.9, and 100È240 km. Papers II and III discuss in detai the estimation of the systematic uncertainties in the foreground modes; Tabe 6 of Paper III ists the systematic uncertainty associated with each foreground. The systematic uncertainty in each residua shown in Tabe 2 of this paper is the quadrature sum of the individua contributions identiðed in Paper III. The tota uncertainties used to state our most restrictive upper imits on the CIB and the uncertainty in the CIB detections at 140 and 240 km are estimated as the quadrature sum of the random and systematic uncertainties. Tabe 2 of this paper and Tabe 6 of Paper III ceary show that the tota uncertainties are dominated by the systematic uncertainties in removing the foreground contributions to the infrared sky brightness. 3. OBSERVATIONAL RESULTS 3.1. Dark Sky L imits The most conservative direct observationa imits on the CIB are derived from the minimum observed sky brightnesses. In each DIRBE weeky sky map, the faintest direction has been determined for each waveength. At waveengths where interpanetary dust scattering or emission is strong, the sky is darkest near the eciptic poes. At waveengths where the IPD signa is rather weak (i.e., ongward of 100 km), the sky is darkest near the gaactic poes or in minima of H I coumn density. The smaest of these vaues at each waveength over the duration of the mission is the dark sky ÏÏ vaue, isted in Tabe 2 as I (dark). The uncertainty shown for each vaue is the quadrature sum of the contributions from the gain and o set 1 p uncertainties. We deðne dark sky ÏÏ upper imits to the CIB at the 95% conðdence eve (CL) as 2 p above the measured dark sky vaues Residuas in Sma Dark Patches After removing the contributions of interpanetary dust (Paper II), bright and faint discrete gaactic sources, and the interstear medium (Paper III) from the measured sky brightness, the residua signa at high gaactic and eciptic atitudes is positive and generay rather featureess, athough ow-eve artifacts from systematic errors in the modes are ceary present. To iustrate the magnitude of the foreground signas, Figure 2 shows the DIRBE spectrum of the tota observed sky brightness averaged over a 5 ] 5 region at the Lockman Hoe, the region of minimum H I coumn density at (, b) D (150, ]53 ) [geocentric eciptic coordinates (j, b) D (137, ]45 )] (Lockman et a. 1986; Jahoda, Lockman, & McCammon FIG. 2.ÈContributions of foreground emission to the DIRBE data at 1.25È240 km in the Lockman Hoe area: observed sky brightness (open circes), interpanetary dust (trianges), bright gaactic sources (crosses), faint gaactic sources (stars), and the interstear medium (squares). Fied circes show the residua brightness after removing a foregrounds from the measurements.

11 No. 1, 1998 DIRBE DETECTIONS AND LIMITS ). Figure 2 aso shows the individua contributions from the foreground sources and the residuas after removing the foreground contributions. Scattering and emission from the interpanetary dust dominates a other signas from 1.25 to 100 km. This is true even at 3.5 km, the spectra window ÏÏ between the maxima of the scattered and emitted IPD signa. Ony at 140 and 240 km does some other foreground signa, that from the interstear medium (infrared cirrus), become dominant. Some insight into the residuas is provided by ooking at severa high-atitude regions (Hauser 1996a, 1996b). For this purpose, we have examined the residuas in 10 ] 10 Ðeds at the north and south Gaactic and eciptic poes (designated NGP, SGP, NEP, and SEP, respectivey) and a 5 ] 5 Ðed in the Lockman Hoe (LH). Tabe 2 ists the mean residua brightnesses for these Ðve patches after a of the foreground-remova steps isted above. As discussed in 2.3, the 100 km map was used as the ISM tempate in producing the residua maps at a waveengths except 100 km. At 100 km, the Be Labs H I map was used as the ISM tempate. Athough the range of residua vaues at each waveength is substantia, typicay a factor of 2 or more, comparison with the dark sky vaues shows that these residuas are sma fractions, approaching 10% at waveengths shortward of 100 km, of the dark sky vaues. However, the fact that the residuas are brightest in the region of peak IPD therma emission, 12 to 25 km, strongy suggests that signiðcant foreground emission sti remains, at east in the midde of the DIRBE spectra range. This is not surprising in view of the very apparent residua IPD modeing errors at these waveengths (e.g., Fig. 1, especiay 4.9 to 100 km; and Paper II) Residuas in High-quaity Regions Athough each of the sma dark patches ( 3.2) is situated where one of the IPD, stear, or ISM foregrounds is minimized, each patch is aso ocated in a region where the other foregrounds may be strong. Therefore, we deðned highquaity ÏÏ (HQ) regions where a foregrounds are expected to be reativey weak. The range of eciptic atitude, b, was restricted to excude bright scattering and emission from the IPD, and the range of gaactic atitude, b, was restricted to excude regions with bright stear emission. To avoid regions with bright ISM emission, ocations where the 100 km brightness, after the IPD contribution was removed, was more than 0.2 MJy sr~1 above the oca mean eve were aso excuded. The argest region that can reasonaby be considered as high quaity covers D20% of the sky between the Gaactic and eciptic poes and is designated HQA. A much more restrictive region, designated HQB, ies in the center of the HQA region and incudes D2% of the sky. Tabe 3 ists the constraints for the HQ regions, and the ast pane of Figure 1 shows the areas covered by the HQ regions. Each HQ region is composed of corresponding northern and southern segments. Tabe 2 ists the mean residua intensities, I (HQA) and I (HQB), for the HQ regions after a foregrounds 0 have been 0 removed. As in the anaysis of the sma dark patches ( 3.2), the 100 km map was used as the ISM tempate in producing the residua maps at a waveengths except 100 km. At 100 km, the Be Labs H I map was used as the ISM tempate. The statistica uncertainty of the mean, which is cacuated from the observed rms variation of the residua emission over the region, is aso shown. For HQB, the tota systematic uncertainty estimated for each band is aso isted in Tabe 2. Athough some portions of the systematic uncertainty needed to be evauated at regions other than the HQ regions (see Papers II and III), the numbers isted here shoud be appropriate for HQB. The systematic uncertainties are arger when deaing with other areas where the foreground emission removed was stronger Residuas at the L ockman Hoe and the North Eciptic Poe The intercept of a inear Ðt to the correation between the infrared emission and the H I coumn density yieds an estimate of the isotropic residua component of infrared emission. This technique was used ( 2.3 and Paper III) to estabish the amount of emission that needed to be removed to create the 100 km tempate of the ISM. The H I data were from Snowden et a. (1994) for a 250 deg2 region covering the Lockman Hoe (LH@) and from Evis et a. (1994) for a 70 deg2 region around the north eciptic poe (NEP@). These regions are denoted with primes to distinguish them from the dark patches ÏÏ LH (5 ] 5 patch) and NEP (10 ] 10 patch) at simiar ocations but of somewhat di erent size, which are discussed in 3.2. Figures 7 and 8 of Paper III show that the 100 km brightness and H I coumn density are ineary reated at ow coumn density in these regions. Within these regions, inear Ðts to the correations between the 140 and 240 km emission and the H I coumn density were aso cacuated. Tabe 2 ists the intercepts of these Ðts as I (LH@, H I) and I (NEP@, H I). 0 0 The advantage of this technique for estimating the CIB at 140 and 240 km, over our standard method using the 100 km data for the ISM tempate, is that the systematic uncertainties of the 100 km data, incuding those caused by uncertainties in the 100 km IPD mode and in the extrapoation of the 100 kmèh I correation to zero H I coumn density, are not propagated into the 140 and 240 km resuts. Thus, the systematic uncertainties for I (LH@, H I) are smaer than those for I (HQB). For the NEP@ 0 region, the intercept of the correation 0 must be extrapoated over a onger interva of H I coumn density and from fewer data, so the systematic uncertainties for I (NEP@, H I) are ony smaer than those of I (HQB) at and 240 km. A disadvantage of this 0 technique is that the H I does not trace other phases of the ISM (ionized and moecuar gas) that may aso contribute to the observed infrared emission. Any part of the emission from other phases that is not TABLE 3 HIGH-QUALITY REGION DEFINITIONS o b o imit o b o imit 100 km ISM imit Areaa Areaa Areaa Region (deg) (deg) (MJy/sr) (pixes) (deg2) (sr) HQA... [30 [25 \ HQB... [60 [45 \ a Bright-source remova reduces these areas by up to 35% at near-ir waveengths.