THE RESOLVED NEAR-INFRARED EXTRAGALACTIC BACKGROUND

The near-infrared (NIR) extragalactic background light (EBL) is the total light from resolved and unresolved extragalactic sources in the NIR. This represents the integrated light from all star and galaxy formation processes over the history of the universe that has been redshifted into the NIR. Some fraction of the NIR EBL can be resolved as the light from individual galaxies (integrated galaxy light, IGL), but the fractional contribution from unresolved (and perhaps unresolvable) sources is not well constrained. A measurement of the total NIR EBL minus the IGL will provide insight into the energy budget of the early universe.

While a direct unresolved measurement of the NIR EBL is technically difficult due to complex foregrounds, several authors have reported a measurement of the total (resolved plus unresolved) NIR EBL (Hauser et al. 1998; Dwek & Arendt 1998; Gorjian et al. 2000; Wright 2001; Matsumoto 2001; Matsumoto et al. 2005; Cambrésy et al. 2001; Levenson et al. 2007) and found it to be a factor of 2 or more above the IGL obtained through source counts (Madau & Pozzetti 2000; Totani et al. 2001; Thompson 2003). This is known as the NIR background excess (NIRBE). The spectral energy distribution (SED) of this measured excess, in some cases, appears very similar to that of zodiacal light, which suggests there may be a foreground contamination issue. Another possible solution to this excess is a large population of undetected faint galaxies and/or Population III (PopIII) stars contributing a large fraction of the NIR background (see Kashlinsky 2005; Hauser & Dwek 2001 for reviews). For example, Matsumoto et al. (2005) found a large NIRBE with a distinct spectral break at ∼1 μm, which they interpreted as a possible signature of a Lyman limit break in PopIII stellar spectra at a redshift of ∼10.

Since the NIR EBL presents a source of opacity for TeV γ-rays via pair production, the density of the background light can, in principle, be measured via direct observation of TeV blazars. This method relies on assumptions of the intrinsic blazar spectrum and the SED of the EBL from ultraviolet to the NIR, both of which are poorly observationally constrained. Nevertheless, TeV γ-rays provide an independent estimate of the NIR EBL that can be used to help determine what fraction of the background light can be attributed to resolved sources and how much may arise from faint and possibly exotic sources in the early universe.

Dwek et al. (2005) explored the possibility that the NIRBE is of extragalactic origin by looking for the absorption imprint of the ∼1 μm spectral break on the γ-ray SEDs of several TeV blazars. They concluded that the apparent NIR excess is likely not of extragalactic origin because its signature is not detected in the blazar SEDs. Recent observations of TeV blazars with the High Energy Stereoscopic System (HESS) have placed limits on the NIR EBL that allow for the possibility of little or no excess above integrated galaxy counts (Aharonian et al. 2006).

Mazin & Raue (2007) used 13 TeV blazars and a grid of NIR background SEDs to further constrain the background light. They derived limits on the NIR EBL that are roughly in agreement with those of Aharonian et al. (2006). However, systematic uncertainties in these γ-ray opacity measurements due to uncertainties in the blazar SED and that of the EBL result in large error bars that could still allow for a significant NIRBE (of the same order as the IGL) to be present, even if their assumptions about the intrinsic blazar SED and that of the EBL are assumed to be reasonable.

As such, the goal of this paper is to derive a firm lower limit to the NIR IGL so as to facilitate a robust comparison with future new results from direct and indirect measurements of the NIR EBL. We present the best current estimate of the NIR IGL by combining our deep wide-field NIR data set with a compilation of galaxy counts from the literature and ultra-deep K_s-band galaxy counts from a subfield of the Subaru Multi Object Infrared Camera and Spectrograph (MOIRCS) image presented in Barger et al. (2008) and Wang et al. (2009).

The structure of this paper is as follows. We describe our NIR data set briefly in Section 2. We calculate the contribution of resolved galaxies to the NIR EBL in Section 3. We compare with previous measurements of the NIR EBL and IGL and discuss the implications of our results in Section 4. We summarize in Section 5. Unless otherwise noted, all magnitudes given in this paper are in the AB magnitude system (m_AB = −2.5log f_ν − 48.60 with f_ν in units of erg cm⁻² s⁻¹ Hz⁻¹).

We presented the observations, data reduction, star–galaxy separation, and bright galaxy counts for our deep wide-field NIR data set in Keenan et al. (2010, K10 hereafter). The data consist of five widely separated fields at high galactic latitudes, including one centered on an Abell cluster (Abell 370). The survey covers approximately 3 deg² reaching a 5σ limiting magnitude of JHK_s ∼ 22–23 over ∼2.75 deg² with another ∼0.25 deg² to JHK_s ∼ 24, making it one of the deepest wide-field surveys to date in the NIR. The fields covered in this survey are described in Section 2.1.

In this paper, we supplement the above data with those from an ultra-deep K_s-band mosaic composed of imaging done by our group and Japanese groups led by various investigators using the MOIRCS instrument on the 8.2 m Subaru Telescope. The MOIRCS imaging is described in Section 2.2.

2.1. Wide-field Imaging and Galaxy Counts

Our first two fields are centered on the Chandra Large Area Synoptic X-ray Survey (CLASXS; Yang et al. 2004; Steffen et al. 2004) and the Chandra Lockman Area North Survey (CLANS; Trouille et al. 2008, 2009). Each of these fields covers ∼1 deg² in JHK_s. These fields are located in the Lockman Hole region of extremely low Galactic H i column density (Lockman et al. 1986). Our third field covers a 0.25 deg² area centered on the Chandra Deep Field North (CDF-N; Brandt et al. 2001; Alexander et al. 2003). The CDF-N contains the Great Observatories Origins Deep Survey North (GOODS-N; 145 arcmin² Hubble Space Telescope Advanced Camera for Surveys observation; Giavalisco et al. 2004). Our fourth field is the Abell 370 (A370) cluster and surrounding area (∼0.5 deg²). A370 is a cluster of richness 0 at a redshift of z = 0.37. Our fifth field is a ∼0.2 deg² area centered on the "Small-Survey-Area 13"(SSA13) from the Hawaii Deep Fields described in Lilly et al. (1991).

In K10, we used the bright NIR galaxy counts from our survey in combination with counts we generated from the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), the UKIRT Infrared Deep Sky Survey (UKIDSS; Lawrence et al. 2007), and other selections from the literature to explore local large-scale structure via the slope of the galaxy counts curve as a function of position on the sky. Here, we extend the counts from K10 to fainter magnitudes using the methods described below.

We generated average galaxy counts for our ensemble of NIR images by the methods described in K10. We corrected the counts for completeness through a simulation in which relatively bright (JHK_s = 19) galaxies were extracted from each image and then scattered at random to other parts of the image to test how well they were recovered as a function of apparent magnitude. We faded the test galaxies in 0.25 mag steps until the recovered fraction fell below 25%. We only included counts in the final average where completeness was better than 50%.

In our completeness simulations, given that we know the total magnitude of objects we are scattering throughout the image, we are able to compare the recovered magnitudes in various apertures to total magnitudes and thereby derive a magnitude correction factor for faint objects. Figure 1 shows example results from this exercise. In Figure 1(a), we show the fractional completeness (N_recovered/N_scattered) as a function of the apparent magnitude for the CDF-N K_s band. In all fields we reliably recover, on average, relatively bright objects (JHK_s < 20) at the correct magnitude using the SExtractor MAG_AUTO, which fits a Kron-like elliptical aperture to each object (Kron 1980). We find that the 3'' aperture shows a roughly constant offset (∼−0.2) from total magnitude. Based on these results, we measure total magnitudes in the MAG_AUTO aperture for objects brighter than 20th magnitude and switch to corrected 3 arcsec aperture magnitudes for objects fainter than 20th magnitude using the measured difference at 20th magnitude. This magnitude correction only affects our calculated IGL at the 0.1 nW m⁻² sr⁻¹ level, which is roughly a factor of 10 below the uncertainty in the IGL due to the uncertainty in the counts themselves.

Zoom In Zoom Out Reset image size

2.2. MOIRCS Imaging and Galaxy Counts

In Figure 2, we show our completeness-corrected and averaged galaxy counts with those drawn from the literature. However, before combining all of our data in this way, we first investigated field-to-field variations for the IGL for our five fields over the magnitude range 14.5 < JHK_s < 22.5, where all five are highly complete and have good counting statistics. We found the IGL over this magnitude range to be consistent across the four non-cluster fields (CLASXS, CLANS, CDF-N, and SSA13 ∼7–8 nW m⁻² sr⁻¹) with a 1σ dispersion of ±0.5 nW m⁻² sr⁻¹. As such, we find that the IGL is consistent with large-scale isotropy, an expected signature of the EBL (see Kashlinsky 2005; Hauser & Dwek 2001 for reviews).

Zoom In Zoom Out Reset image size

In the A370 cluster field, our IGL measurements were ∼2–3 nW m⁻² sr⁻¹ higher in all bands with the peak contribution to the excess light arising from galaxies at JHK_s ∼ 17, consistent with an excess of L_* galaxies at a redshift of ∼0.4 as in A370. We include the cluster field in our average counts for this study because in a survey of a few square degrees such as this, roughly one massive cluster will be present.

In Figure 2, we use black diamonds to denote the counts for our five wide fields and black squares to denote counts for the MOIRCS image. The counts determined by K10 from the 2MASS field with Galactic latitudes of |b|>30 are denoted by red asterisks. The counts determined by K10 from the 2MASS-6x Lockman Hole survey are denoted by green triangles. The counts determined by K10 from three subfields of the UKIDSS LAS are denoted by blue plus symbols. The three subfields that make up this section of the LAS are described in detail in K10. Other data points are taken from the publications listed in the plots. The dashed curves represent the error-weighted least-squares running average from which we calculate the IGL. We describe the methods we used to calculate this average in Section 3.

Figures 2((d)–(f)) display the same data as in panels (a)–(c) after dividing by an arbitrarily normalized Euclidean model (of slope α = 0.4 in the form N(m) = A× 10^αm and A is a constant) to expand the ordinate and demonstrate where resolved galaxies contribute the most to the IGL. A flat line in panels (d)–(f) would imply an equal contribution to the IGL at all magnitudes. The areas of positive slope show where galaxies contribute a larger fraction to the IGL as one moves toward fainter magnitude. The steep negative slope beyond JHK_s > 23–24 demonstrates the diminishing contribution of resolved galaxies to the IGL at the faintest magnitudes.

To calculate the IGL from all the counts data displayed in Figure 2, we used an error-weighted least-squares fit of the slope and normalization parameters "A" and "α" in the function N(m) = A × 10^αm. We fit the data over a range of ±2 mag above and below each magnitude bin to get two estimates for the counts in that bin. In other words, at a given apparent magnitude "m," we fit the range m − 2 to m and m to m + 2 separately. Each fit yielded an estimate of the counts at apparent magnitude m. We then recorded the average of these two values as the best estimate for counts in that bin. Each of the two fits also yielded a 1σ confidence interval for the counts in each bin. Given these fitted values and confidence intervals we found the maximum and minimum allowable counts over the 1σ range and recorded $\frac{\rm {maximum}-\rm {minimum}}{2}$ as the final 1σ error associated with the average value for the counts in that bin. We then integrated these fitted values for the galaxy counts to calculate the IGL and its associated 1σ confidence interval for the magnitude range 11 < JHK_s < 28. Our fitted values for the galaxy counts with the confidence intervals as described above are given in a tabular form in Table 1.

The results of this calculation are shown in Figure 3. In Figure 3(a), the solid curves represent the cumulative contribution of galaxies to the IGL as a function of apparent magnitude, and the dashed curves show the percent contribution in each 1 mag bin. Figure 3(b) shows the calculated slope of the galaxy counts curve as a function of apparent magnitude. Due to simple geometrical effects, when the slope drops below α = 0.4 the total light from galaxies begins to converge.

Zoom In Zoom Out Reset image size

Figure 3(b) shows that the counts slope for all three bands begins near the expected Euclidean value of α = 0.6 at bright magnitudes and then drops steadily beyond JHK_s ∼ 17. Due to these trends in slope, the vast majority (∼90%) of the resolved IGL arises from galaxies in the range ∼15 < JHK_s < 24. Over this entire range our study is highly complete and has good counting statistics.

Furthermore, from Figure 3 it can be seen that for sources fainter than JHK_s = 28 to make any appreciable contribution to the IGL the counts curve would have to rise dramatically over several magnitudes beyond the limits of current surveys. Such a rise would quickly result in several (or more) galaxies per square arcsecond, rendering any counting exercise impossible due to confusion. Faint galaxies certainly exist beyond the limits of the deepest NIR surveys, because at high redshifts the faintest apparent magnitudes observed are only probing a few magnitudes fainter than L_* down the luminosity function. It is unknown whether there is a steep upturn in the luminosity function toward faint magnitudes, but if such an upturn exists and faint galaxies contribute significantly to the NIR EBL, they would need to be so numerous as to be unresolvable. As such, it may be safe to say that the resolvable portion of the NIR EBL (the IGL) has, for the most part, been measured and that the most important contribution to the resolved portion comes from galaxies in the magnitude range ∼15 < JHK_s < 24, for which the deep wide-field data presented here are optimized.

3.1. K_s-band Faint Counts Slope and Missing Flux

We use our ultra-deep MOIRCS data to test for a possible steepening of the counts slope at faint magnitudes in the K_s band, as appeared possible from the SDF (Maihara et al. 2001) and SSDF (Minowa et al. 2005) K_s-band counts. We see from Figure 2(c) that the slope of the counts in the MOIRCS field continues to become flatter out to the faintest magnitudes. We detect no upturn in the slope for very faint galaxies. As mentioned above, the MOIRCS counts have not had a completeness correction applied. Such a correction would steepen the curve slightly at the faintest magnitudes, but this effect would not be strong enough to alter significantly the trend in the slope.

Thompson et al. (2007) estimated the missing flux component in the J and H bands from the faint outer parts of galaxies and from galaxies below their detection limits using a histogram of flux in all pixels associated with detected objects. We use a similar method to estimate the flux missed in the K_s band. In Figure 4, we show a histogram of the number of pixels versus flux for all pixels associated with detected galaxies (object pixels) in our MOIRCS K_s-band mosaic. Noting the linear trend for fluxes ∼0.005–0.4 μJy, we fit a line to the data over this range (blue dashed line). The portion of the histogram used in the fit represents 60% of all object pixels in the image and 99% of object pixels with fluxes greater than the turnover in the histogram at ∼0.005 μJy. The slope of the linear fit is −0.86. Thompson et al. (2007) find a slightly steeper value of −1 by simply estimating the slope by eye. Assuming that the true flux distribution for faint pixels (<0.005 μJy) continues along the same trend, we extrapolate the linear fit to approximate the shape of the histogram when all pixels in the image are accounted for. Using this method, we calculate an estimate for flux missed in the faint outer parts of galaxies and in galaxies that are below our detection limits. We find the missing flux component to be ∼22% (1.9 nW m⁻² sr⁻¹) of the total K_s-band light from resolved galaxies.

Zoom In Zoom Out Reset image size

3.2. The Stellar Component of NIR Light

We present our IGL results in tabular form in Table 2. In Figure 5, we show our IGL results alongside a summary of measurements of the NIR IGL (black symbols) and EBL (red symbols) over the past decade, including the results obtained via integrated galaxy counts (Madau & Pozzetti 2000; Totani et al. 2001; Thompson 2003), via total light minus stars and zodiacal light (Dwek & Arendt 1998; Hauser et al. 1998; Gorjian et al. 2000; Wright 2001; Cambrésy et al. 2001; Matsumoto et al. 2005; Levenson et al. 2007), via modeling of the NIR IGL (Dominguez et al. 2010), and via TeV γ-ray opacity (blue hashed area; Aharonian et al. 2006).

Zoom In Zoom Out Reset image size

The blue hashed area of Figure 5 shows the allowed (1σ) NIR EBL intensity (14 ± 4 nW m⁻² sr⁻¹) derived from HESS observations of TeV blazars (Aharonian et al. 2006). Mazin & Raue (2007) use 13 TeV blazars and a grid of NIR background intensities to further constrain the NIR EBL and find approximate agreement with the results of Aharonian et al. (2006). As noted earlier, however, estimates of the NIR EBL from TeV γ-ray opacity measurements rely on assumptions about the intrinsic SEDs of blazars and that of the EBL, both of which are poorly constrained observationally. Our results (black diamonds) are some 10%–20% higher than previous estimates of the IGL, which puts them closer to EBL estimates from TeV blazar observations and the most recent total NIR EBL measurements. We also find excellent agreement with the recent efforts to model the IGL directly from galaxy SEDs over a wide range in redshift by Dominguez et al. (2010; purple solid and dashed line in Figure 5).

A large NIRBE has been found by several groups, with perhaps the most striking result being that of Matsumoto et al. (2005). This excess, when combined with other NIR and optical background measurements, showed an apparent spectral break around 1 μm. The excess was originally attributed to PopIII stars, with the break corresponding to the redshifted Lyman limit for these stars at z ∼ 10. However, a search for the possible absorption imprint of this break on the γ-ray SED of blazars did not find evidence for such a feature (Dwek et al. 2005).

Madau & Silk (2005) discuss the implications of a NIRBE due to massive PopIII stars in the early universe and demonstrate that an excess of even a few nW m⁻² sr⁻¹ in the J band would imply energetic requirements that are "uncomfortably high." They note that ∼5% of baryons would be formed into such stars by a redshift of 9 (roughly twice the fraction converted into stars since then) and that the metals created in these PopIII stars would have to be hidden in intermediate-mass black holes (IMBHs) for the universe to avoid exceeding solar metallicity by a redshift of 9. They point out that the IMBHs themselves could produce the NIR excess as miniquasars, as suggested by Santos et al. (2002), but that accretion onto these IMBHs would dominate the soft X-ray background and their mass density in the present-day universe would exceed that of black holes found in galactic nuclei by 3 orders of magnitude. Finally, they show that the ionizing flux produced by these PopIII stars would exceed by 3 orders of magnitude that required to produce the Wilkinson Microwave Anisotropy Probe observed electron scattering depth at z = 17.

We have presented measurements of the IGL in J, H, and K_s using deep wide-field NIR photometry in combination with ultra-deep MOIRCS K_s-band data and selections from the literature. These results place the best current constraints on the total NIR light from resolved galaxies and serve as a new lower limit to the total NIR EBL. While these results are in relative agreement with previous measurements, our numbers are 10%–20% higher, bringing them into better agreement with those derived from γ-ray experiments and the most recent measurements of the total NIR EBL. We also find excellent agreement with the recent efforts to model the IGL directly from galaxy SEDs.

We find the IGL to be roughly isotropic and consistent with the expectation of large-scale isotropy in the EBL. We confirm that the starlight subtraction for the most recent total NIR EBL measurements is correct, so if there still exists a foreground subtraction issue in these measurements, it most likely is associated with the zodiacal light.

While our measurements cannot rule out the existence of a NIRBE due to PopIII stars or other exotic early universe objects, our new lower limits on the IGL and the upper limits found from TeV γ-ray experiments (Aharonian et al. 2006) could now be considered in rough agreement with the most recent total NIR EBL measurements in the J band, and in near agreement in the K band.

We thank the anonymous referee for a careful review of this paper and the comments and suggestions that helped to improve the content and presentation of this work.

We gratefully acknowledge support from the NSF grants AST 0708793 (A.J.B.) and AST 0709356 (L.L.C.), the University of Wisconsin Research Committee with funds granted by the Wisconsin Alumni Research Foundation, and the David and Lucile Packard Foundation (A.J.B.). R.C.K. was supported by a Wisconsin Space Grant Consortium Graduate Fellowship, a Sigma Xi Grant in Aid of Research, and an NSF East Asia and Pacific Summer Institutes Fellowship during portions of this work.

This publication makes use of data products from the Two Micron All Sky Survey (2MASS), which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.

This work is based in part on data products from the UKIRT Infrared Deep Sky Survey (UKIDSS).

This work is based in part on observations obtained with WIRCam, a joint project of CFHT, Taiwan, Korea, Canada, France, and the Canada–France–Hawaii Telescope (CFHT) which is operated by the National Research Council (NRC) of Canada, the Institute National des Sciences de l'Univers of the Centre National de la Recherche Scientifique of France, and the University of Hawaii.

This work is based in part on data collected at the Subaru Telescope, which is operated by the National Astronomical Observatory of Japan.

This work has made use of NASA's Astrophysics Data System.