aa r X i v : . [ a s t r o - ph ] J a n Massive stars as cosmic enginesProceedings IAU Symposium No. 250, 2008Fabio Bresolin, Paul Crowther & Joachim Puls, eds. c (cid:13) Imprint of first stars era in the cosmicinfrared backround fluctuations
A. Kashlinsky
SSAI and Observational Cosmology Lab, Code 665, Goddard Space Flight Center, Greenbelt,MD 20771, U.S.A.email:
Abstract.
We present the latest results on CIB fluctuations from early epochs from deep Spitzerdata. The results show the existence of significant CIB fluctuations at the IRAC wavelengths(3.6 to 8 µ m) which remain after removing galaxies down to very faint levels. These fluctuationsmust arise from populations with a significant clustering component, but only low levels ofthe shot noise. There are no correlations between the source-subtracted IRAC maps and thecorresponding fields observed with the HST
ACS at optical wavelengths. Taken together, thesedata imply that 1) the sources producing the CIB fluctuations are individually faint with S ν < a few nJy at 3.6 and 4.5 µ m; 2) have different clustering pattern than the more recent galaxypopulations; 3) are located within the first 0.7 Gyr (unless these fluctuations can somehowbe produced by - so far unobserved - local galaxies of extremely low luminosity and with theunusual for local populations clustering pattern), 4) produce contribution to the net CIB flux ofat least 1-2 nW/m /sr at 3.6 and 4.5 µ m and must have mass-to-light ratio significantly belowthe present-day populations, and 5) they have angular density of ∼ a few per arcsec and are inthe confusion of the present day instruments, but can be individually observable with JWST . Keywords. (cosmology:) diffuse radiation, early universe, large-scale structure of universe
1. Introduction
The cosmic infrared background (CIB) is a repository of emissions throughout theentire history of the Universe. The recent years have seen significant progress in CIBstudies, both in identifying and/or constraining its mean level (isotropic component) andfluctuations (see Kashlinsky 2005 for a recent review). The CIB contains emissions alsofrom objects inaccessible to current (or even future) telescopic studies and can, therefore,provide unique information on the history of the Universe at very early times. Oneparticularly important example of such objects, of particular reference to this conference,concerns Population III stars (hereafter Pop III), the still elusive zero-metallicity starsexpected to have been much more massive than the present stellar populations (seeBromm & Larson 2004 for a recent review). Herebelow I will use the term ”era of thefirst stars”, or ”Pop III era”, with the understanding that the actual era may be composedof objects of various nature from purely zero-metallicity stars, to low- metallicity starsto even possibly mini-quasars whose contribution to the CIB is driven by energy releasedby gravitational accretion, as opposed to stellar nucleosynthesis.Extensive numerical investigations of collapse and fragmentation of the first objectsforming out of density fluctuations specified by the standard ΛCDM model suggest thatPop III stars were quite massive and lived at z >
10, well within the first Gyr of theUniverse’s evolution. If predominantly massive, they are expected to have left a significantlevel of diffuse radiation redshifted today into the IR, and it has been suggested that theCIB contains a detectable contribution from Pop III in the near-IR, manifest in both its1 A. Kashlinskymean level and its anisotropies (e.g. Bond et al 1986, Santos et al 2002, Salvaterra &Ferrara 2003, Cooray et al 2004, Kashlinsky et al 2004).In the past several years a group of us (Kashlinsky, Arendt, Mather & Moseley 2005,2007a,b,c - hereafter KAMM1, KAMM2, KAMM3, KAMM4) have used deep-integration
Spitzer data to measure the CIB fluctuations component arising from early populations.These provide first observational insights into the global evolution of the Universe at earlycosmic epochs. Our measurements revealed significant CIB fluctuations at the IRACwavelengths (3.6 to 8 µ m) which remain after removing galaxies down to very faintlevels (KAMM1, KAMM2). These fluctuations must arise from populations that havea significant clustering component, but only low levels of the shot noise (KAMM3).Furthermore, there are no correlations between the source-subtracted IRAC maps andthe corresponding fields observed with the HST
ACS at optical wavelengths (KAMM4).Taken together, these data imply that 1) the sources producing the CIB fluctuationshave a very different clustering pattern than galaxies at intermediate redshifts and areindividually faint with S ν < a few nJy at 3.6 and 4.5 µ m; 2) are located within the first ≃ . /sr at 3.6 and 4.5 µ m and must have mass-to-light ratio significantly below thepresent-day populations, and 4) their angular density is ∼ (a few) arcsec − , so they are inthe confusion of the current instruments, but can be individually observable with JWST .Below, I will discuss the latest measurements of the fluctuations in the CIB by ourteam (Kashlinsky, Arendt, Mather & Moseley - KAMM) and explore their implicationsfor the nature of the sources contributing to these anisotrpies, specifically in the era ofthe first stars. Following the Introduction, Sec. 2 reviews the current measurements atboth near-IR (IRAC) and optical (ACS) wavelengths and Sec. 3 discusses the nature ofthe cosmological populations producing these CIB anisotropies.
2. Source subtracted CIB fluctuations vs optical galaxies
Before we discuss the interpretation of the KAMM measurements, it is important toreview the steps done in the analysis leading to the measured CIB fluctuations. Thedata have been assembled from the individual AORs using the self-calibration methodfrom Fixsen et al (2000). The exposure times ranged from ∼ − ′ × ′ QSO 1700 field (KAMM1) to ∼ −
25 hours/pixel in each of the fourGOODS fields of 10 ′ × ′ (KAMM2) The latter have been observed at two differentepochs separated by ∼ µ m adopted fromKAMM2. The detected signal is significantly higher than the instrument noise and the D 11. First stars and cosmic infrared background µ m where Galactic cirrusmay contribute to the measured signal. There was a statistically significant correlationbetween the channels for the regions of overlap meaning that the same population isresponsible for the fluctuations. The correlation function at deeper clipping cuts, whentoo few pixels were left for Fourier analysis, remains the same and is consistent with thepower spectrum numbers (KAMM1). The signal is to a good accuracy isotropic on thesky, as required by its extragalactic origin, and must thus contain contributions from the“ordinary” galaxies and from unresolved populations at high z .The extragalactic signal is made of two components: 1) shot noise from the remainingfaint galaxies (shown with dotted lines in Fig. 1), and 2) on arcminute scales the fluc-tuations are produced by clustering of the emitters. It is important to emphasize thatas fainter foreground galaxies are removed so that the remaining shot noise is reducedthe details of the fluctuations change. This is due to the varying contribution from theremaining foreground galaxies. The large-scale part of the fluctuations remains as theforeground sources are removed down to the lowest shot-noise levels. Left panels of Fig.2 show the decrease in the shot-noise power, P SN , as progressively higher iterations inthe source removal are reached. The final shot noise reached by us with the GOODSdata is shown with horizontal lines and is a factor of ∼ m AB > −
22 for IRAC beam). The figure shows that galaxy removal is efficient to m AB > −
27 and the signal in Fig. 1 comes from very faint sources.
Figure 1.
Left : Source-subtracted CIB fluctuations from KAMM2 at 3.6 and 4.5 µ m. Four setsof symbols correspond to the four GOODS fields. Dotted lines show the shot-noise contribution.Solid line shows the slope of sources at high- z with the ΛCDM model spectrum of the sameamplitude at 3.6 and 4.5 µ m. Right : CIB fluctuations due to ACS galaxies for the 972 ×
972 0.6 ′′ pixel field at HDFN-Epoch2 region for the ACS B and z-bands. Filled circles correspond to ACSgalaxies fainter than m = 21 with the mask defined by the clipping. Filled diamonds, trianglesand squares correspond to fluctuations produced by sources fainter than m + 2 , m + 4 , m + 6. GOODS fields have also been observed at optical wavelengths with the
Hubble
ACSinstruments reaching source detection levels fainter than 28 AB mag. This allowed us tofurther test the origin of the source-subtracted CIB fluctuations. If the latter come fromlocal populations there should be a strong correlation between the source-subtractedIRAC maps and the ACS sources. Conversely, there should be no such correlations if theCIB signal arises at at epochs where the Lyman break (at rest ∼ . µ m) gets shiftedpassed the longest ACS z-band at ≃ . µ m. To test for this in KAMM4 we have con-structed synthetic maps, overlapping with the GOODS fields, using sources in the ACSB, V, i, z bands from the ACS sources catalog of Giavilsco et al (2004). These maps were A. Kashlinsky Figure 2.
Left : Decrease of the shot noise power vs the iteration number of the source cleaning.Horizontal lines show the levels reached in Kashlinsky et al (2007), which are a factor of ∼ Right : Shot noise power P SN estimated by integrating the counts. Symbols identical to those on the left denote the fourGOODS fields; open circles correspond to counts for the QS1700 field. Solid line shows P SN according to the fit to IRAC counts of Fazio et al (2004) used in KAMM1. then convolved with the IRAC 3.6 and 4.5 µ m beams. Finally, we applied the clippingmask from the IRAC maps and computed the fluctuations spectrum produced by theACS sources and their correlations with the IRAC-based maps.The fluctuations in the diffuse light produced by the ACS galaxies are shown in theright panels of Fig. 1. The contrast between the spectrum of the source-subtracted CIBfluctuations and those produced by the optical galaxies is obvious. The former has thepower spectrum such that the fluctuations are flat to slowly rising with increasing angularscale, whereas the latter have power spectrum with fluctuation amplitude decreasing withincreasing scale in agreement with CIB measurements from deep 2MASS data arisingfrom galaxies at z ∼ removed by KAMM prior to computing the remaining CIB fluctuations.The excess source-subtracted CIB fluctuation on arcminute scales in the 3.6 µ m channelis ∼ . /sr; KAMM measure a similar amplitude in the longer IRAC bandsindicating that the energy spectrum of the arcminute scale fluctuations is flat to slowlyrising with increasing wavelength at least over the IRAC range of wavelengths. This isillustrated with the solid line in the left panels of Fig. 1.
3. Cosmological implications
Any interpretation of the KAMM results must reproduce three major aspects: • The sources in the KAMM data were removed to a certain (faint) flux limit, so theCIB fluctuations arise in populations with magnitudes fainter than the correspondingmagnitude limit, m lim . Furthermore, these sources are not present among the opticalACS galaxies as demonstrated by the absence of correlations between these galaxies andthe IRAC source-subtracted CIB maps. • These sources must reproduce the excess CIB fluctuations by KAMM on scales
D 11. First stars and cosmic infrared background Figure 3.
Left : Correlation coefficient between clipped/masked ACS and KAMM data. Largeand small symbols correspond to the IRAC Ch 1 and Ch 2; the four sets of symbols correspond tothe four GOODS fields. Open symbols correspond to correlations with the maps of the removedsources and filled symbols with the residual KAMM maps which contain the fluctuations shownin Fig. 1.
Right : Solid lines show the dimensionless correlation function between the diffuse lightin the ACS and KAMM maps for
B, V, i, z -bands in order of increasing thickness. Dotted lineshows the dimensionless correlation function of the KAMM maps, C KAMM ( θ ) /σ , whichremains positive out to ∼ ′′ and is better viewed in log-log plots as in Fig. SI-4 of KAMM1. > . ′ . They must also reproduce the measured spectrum of the CIB fluctuations, whichis different from the observed clustering pattern of ordinary galaxies at intermediate z . • Lastly, the populations fainter than the above magnitude limit must account not onlyfor the correlated part of the CIB, but - equally important - they must reproduce the(low) shot-noise component of the KAMM signal, which dominates the power at < ′ .The discussion below by-and-large follows KAMM3:1) Magnitude limits and epochs . Since the ACS galaxies do not contribute to thesource-subtracted CIB fluctuations, the latter must arise at z > ∼ . µ m getting redshifted past the ACS z -band of peak wavelength ≃ . µ m. This would place the sources producing the KAMM signal within the first 0.7Gyr. If the KAMM signal were to originate in lower z galaxies which escaped the ACSGOODS source catalog because they are below the catalog flux threshold, they wouldhave to be extremely low-luminosity systems ( < × h − L ⊙ at z =1) and these galaxieswould also have to cluster very differently from their ACS counterparts.2) Clustering component . Solid lines in Fig. 1 show the expected CIB fluctuationsfrom sources with the (biased) concordance ΛCDM power spectrum at z >
5. The fit isreasonably good making such sources a plausible candidate for producing the observedsignal. At the same time, the observed galaxy populations out to z ∼ − observed difference in the clustering patterns.3) Net CIB levels from the new sources : The angle of 1 ′ in the concordancecosmology subtends comoving scales of 2.2-3 Mpc at 5 z
20. For ΛCDM densityfields with reasonable biasing one can reach relative arcminute-scale fluctuations of ∼ µ m is at least 1-2 nW/m /sr, which is well within the uncertainties of the recent CIBmeasurements of Thompson et al (2007).4) Shot noise constraints . The amplitude of the shot-noise power gives a particularlystrong indication of the epochs of the sources contributing to the KAMM signal. Thiscan be seen from the expressions for the shot noise (Kashlinsky 2005a): P SN = Z m AB >m lim f ( m ) dF ( m ) ≡ f ( ¯ m ) F tot ( m AB > m lim ) (3.1) A. Kashlinskywhere f ( m ) is the flux in Jy of a source of magnitude m and F tot ( m AB > m lim ) is thenet CIB flux produced by the remaining sources. Above it was shown that the sourcescontributing to the fluctuations must have CIB flux greater than a few nW/m /sr andcombining this with the values for P SN ∼ − nW /m /sr, reached in the KAMM2analysis, leads via eq. 3.1 to these sources having typical magnitudes m AB < −
30 orindividual fluxes < Such faint sources are expected to lie at very high z . Mass/light ratio of the new populations : This information on the nature ofthe populations responsible for these CIB fluctuations, can be obtained from the factthat the significant amount of flux ( > /sr) required to explain the amplitudeof the fluctuations must be produced within the short time available at these high z (cosmic times < ≡ M/L . The smaller the value of Γ, the fewer baryonsare required to explain the CIB fluctuations detected in the KAMM studies. It turnsout that in order not to exceed the baryon fraction observed in stars, the populationsproducing these CIB fluctuations had to have Γ much less than the solar value, typical ofthe present-day populations (KAMM3). This is consistent with the general expectationsof the first stars being very massive.6)
Resolving the new sources : In order to directly detect the faint sources re-sponsible for the CIB fluctuations with fluxes below a few nJy, their individual fluxmust exceed the confusion limit. If such sources were to contribute to the CIB requiredby KAMM data, at 3.6 and 4.5 µ m they had to have the average surface density of¯ n ∼ F /P SN ∼ − . In order to avoid the confusion limit and resolve thesesources individually at, say, 5-sigma level ( α = 5) one would need a beam of the area ω beam α − / ¯ n ∼ × − arcsec or of circular radius below ∼ JWST could be able toresolve these objects given its sensitivity and resolution.
Acknowledgements
I thank my collaborators, Rick Arendt, John Mather and HarveyMoseley for many contributions to the KAMM results, and the NSF AST-0406587 grantfor support.
References
Bond, J.R., Carr, B.J., & Hogan, C.J. 1986,
Ap.J. , 306, 428Bromm, V. & Larson, R. 2004,
ARA&A , 42, 79Cooray, A. et al 2004,
Ap.J. , 606, 611Fazio, G. et al 2004,
Ap.J.Suppl. , 154, 39Fixsen, D., Moseley, S.H. & Arendt, R.G. 2000,
Ap.J.Suppl. , 128, 651Giavilsco, M. et al 2004,
Ap.J. , 600, L93Hogbom, J. 1974,
A&AS , 15, 417Kashlinsky, A. 2005,
Phys. Rep. , 409, 361Kashlinsky, A. 2005,
Ap.J. , 633 L5Kashlinsky, A., Odenwald, S., Mather, J., Skrutskie, M. & Cutri, R. 2002,
Ap.J. , 579, L53Kashlinsky, A., Arendt, R.G., Gardner, J.P., Mather, J. & Moseley, S.H. 2004,
Ap.J. , 608, 1Kashlinsky, A., Arendt, R.G., Mather, J., & Moseley, S.H. 2005,
Nature , 438, 45 (KAMM1)Kashlinsky, A., Arendt, R.G., Mather, J., & Moseley, S.H. 2007,
Ap.J. , 654, L5 (KAMM2)Kashlinsky, A., Arendt, R.G., Mather, J., & Moseley, S.H. 2007,
Ap.J. , 654, L1 (KAMM3)Kashlinsky, A., Arendt, R.G., Mather, J., & Moseley, S.H. 2007,
Ap.J. , 666, L1 (KAMM4)Santos, M., Bromm, V. & Kamionkowski, M. 2002,
MNRAS , 336, 1082Thompson, R., Eisenstein, D., Fan, X., Rieke, M. & Kennicutt, R. 2007,
Ap.J. , 659, 667Thompson, R., Eisenstein, D., Fan, X., Rieke, M. & Kennicutt, R. 2007,