next previous
Up: The 1997 reference

12. Extragalactic background light

  For the extragalactic background radiation no generally-accepted measured values exist in the UV, optical or infrared wavebands. However, upper limits from surface photometry and lower limits from galaxy counts are available. We present a critical evaluation and tabulation of the available results.

Extragalactic background light (EBL) in UV, optical and near-IR (tex2html_wrap_inline15825) is thought to consist mainly of redshifted starlight from unresolved galaxies; more hypothetical contributions would be, e.g., from stars or gas in the intergalactic space, and from decaying elementary particles (e.g. neutrinos). In the mid- and far-infrared the main contribution is thought to be redshifted emission from dust particles, heated by starlight in galaxies.

Observations of the EBL are hampered by the much stronger foreground components of the night sky brightness described in the other sections. Unlike the other components the EBL is isotropic which, in combination with its weakness, complicates its separation. Recent reviews of the observational and theoretical status of the EBL have been given by Mattila (1990), Tyson (1990, 1995), Mattila et al. (1991) for the optical; by Bowyer (1991), Henry (1991), Henry & Murthy (1995) and Jakobsen (1995) for the ultraviolet; by Matsumoto (1990), Franceschini et al. (1991a), Hauser (1995a,b, 1996) and Lonsdale (1995) for the infrared; Longair (1995) has given a general review covering all wavelengths.

The observational results presented here are summarised for each wavelength range in a separate table in the corresponding subsection. They are also put together in overview in Fig. 79 (click here) at the end of this section, where in the visual and near-infrared region some model predictions are added for comparison with the data, which stretch over a wide range of brightnesses at these wavelengths. Otherwise, model prediction of EBL brightness are not the topic of this reference. For this matter see, e.g., the conference proceedings by Bowyer & Leinert (1990) and Rocca-Volmerange et al. (1991) or the work of Franceschini et al. (1991b).

Figure 79: Summary of present observational limits and model predictions for the EBL. The photometric upper limits by Dube et al. (DWW), Toller (T) and Mattila & Schnur (MS) in optical and the UV upper limit of 300 photon units at 170 nm (see text) are shown as downward pointing arrows. The COBE/DIRBE and COBE/FIRAS dark sky (total) brightnesses between 1.25 tex2html_wrap_inline10901m and 500 tex2html_wrap_inline10901m are shown as crosses and open triangles, respectively. The ranges of isotropic residuals after foreground subtraction are shown by vertical bars for the DIRBE 1.25 tex2html_wrap_inline10901m - 240 tex2html_wrap_inline10901m bands (Hauser 1996). The Mather et al. (1994) estimate for an upper limit of possible sub-mm excess above the CMB spectrum is shown as a dashed line between 500 tex2html_wrap_inline10901m and 1000 tex2html_wrap_inline10901m. The claimed tentative detection of CIBR by Puget et al. (1996) is shown as a solid line between 400 tex2html_wrap_inline10901m and 1000 tex2html_wrap_inline10901m. Solid lines at 10 tex2html_wrap_inline10901m - 40 tex2html_wrap_inline10901m are the possible detections from Dwek & Slavin (1994); the upper line is for H0 = 100 km s-1 Mpc-1 and the lower one for 50 km s-1 Mpc-1. The results from galaxy counts are are shown with different symbols: Cowie et al.: black squares; Tyson: solid circles; Morgan and Driver: open circles (two values at each wavelength band, see Table 46 (click here) and text). In the visual range, some model calculation results are shown as well for comparison: solid lines are after Yoshii & Takahara (1988) for evolving galaxy models, labeled with q0 and tex2html_wrap_inline16374, where tex2html_wrap_inline16374 means the epoch (measured by redshift) of galaxy formation; the dashed line is for a non-evolving galaxy model with q0 = 0.02. The two dash-dotted lines are after Väisänen (1996) for models which include the estimated effect of low-surface-brightness and faint blue galaxy populations: the upper line is with Ferguson & McGaugh (1995) luminosity function and with luminosity evolution (model labeled FMB-LE in Väisänen 1996); the lower line is with a luminosity function evolution in accordance with the findings of Lilly et al. (1995), i.e. extra brightening of the blue galaxies over the passive luminosity evolution, and an excess of a non-evolving blue population of faint galaxies (model labeled BBG in Väisänen 1996)

12.1. Ultraviolet

An extragalactic component is certainly present in the UV/FUV since the summed flux of galaxies is present at some level. Early in the Space Age it was realized that searches in the FUV had substantial advantages over searches in the UV, because the zodiacal light component is not present at a measurable level and contributions from stellar sources were expected to be small. In particular, it was hoped that emission from a very hot (10 6 K) or hot (105 K) intergalactic medium might be detected. These measurements were far more difficult to carry out than was originally anticipated, and a wide range of conflicting results were reported.

At this point, the most cited argument that some of the Far UV diffuse background is extragalactic in origin is that most measurements of this flux show a correlation with galactic neutral hydrogen column density, and the extrapolation to zero hydrogen columns yields fluxes that are in the range of 50 to 300 photon units. These results are only upper limits to an extragalactic background, however, since there is no guarantee that galactic components are not producing this flux.

While the total picture is far from clear, some aspects of a possible extragalactic flux have been established. Quasar absorption line studies definitely constrain emission from a diffuse intergalactic medium to a marginal role (Jakobsen 1991). Paresce & Jacobsen (1980) had shown before that integrated light from QSOs and AGNs will not produce a significant contribution to the diffuse FUV background. However, Armand et al. (1994) have used data on galaxy counts obtained at 2000 Å with a limiting magnitude of 18.5, to calculate the ultraviolet flux due to the integrated FUV light of all galaxies. The extrapolation is small and leads to an expected flux of 40 to 130 photon cm-2 s-1 sr-1Å-1. Hence it seems certain that there is at least some extragalactic flux present in the diffuse FUV background. It is interesting to note that the flux predicted by Armand et al. is consistent with the (uncertain and controversial) observational results for a possible extragalactic diffuse FUV background.


Summed from all galaxies 40 to 200
QSOs/AGNs <10
Intergalactic medium <10
observed upper limit 50 to 300
Table 45: Possible components of a diffuse extragalactic far ultraviolet background with their estimated intensitiesa

a Intensities are given in units of photons cm-2 s-1 sr-1 Å-1.

12.2. Visual



Author(s) tex2html_wrap_inline10929 tex2html_wrap_inline10957 tex2html_wrap_inline10957 tex2html_wrap_inline14031(revised) Method
(Å\) revised, 1tex2html_wrap_inline14579 limits erg s-1cm-2sr-1
Dube et al. (1979) 5115 1.0 tex2html_wrap_inline11177 1.2 S10 4.0 tex2html_wrap_inline11177 1.2 S10 photometry
tex2html_wrap_inline15880 3.4 S10 tex2html_wrap_inline15880 5.2 S10 tex2html_wrap_inline15880 4.0 10-5
Toller (1983) 4400 1.3 tex2html_wrap_inline11177 1.3 tex2html_wrap_inline11131 1.9 tex2html_wrap_inline11177 4.8 tex2html_wrap_inline11131 photometry
tex2html_wrap_inline15880 3.9 tex2html_wrap_inline11131 tex2html_wrap_inline15880 6.7 tex2html_wrap_inline11131 tex2html_wrap_inline15880 3.2 10-5
Mattila & Schnur 4000 6.5 tex2html_wrap_inline11177 2.5 cgs* tex2html_wrap_inline15880 9.0 cgs* tex2html_wrap_inline15880 3.6 10-5 photometry
Cowie et al. (1994) 3400 (U') 1.3 10-6 galaxy counts
4470 (B) 1.8 10-6 (tex2html_wrap_inline15928)
5425 (V) 3.1 10-6
8340 (I) 4.7 10-6
22000 (K) 5.2 10-6
Tyson (1995) 3600 (U) 2.5(+.07 -.04) 10-6 galaxy counts
4500 (B) 2.9(+.09 -.05) 10-6 (Bjtex2html_wrap_inline1588029tex2html_wrap_inline15948/tex2html_wrap_inline15950)
6500 (R) 2.9(+.09 -.05) 10-6
9000 (I) 2.6(+.3 -.2 ) 10-6
22000 (K) 7.2(+1 -1 ) 10-6
Morgan & Driver 4500 (B) 1.9 10-6 galaxy counts
(1995) 5500 (V) 1.3 10-6 (tex2html_wrap_inline15968)
6500 (R) 3.2 10-6
9000 (I) 3.5 10-6
4500 (B) 4.7 10-6 galaxy counts
5500 (V) 6.4 10-6 (tex2html_wrap_inline15978 tex2html_wrap_inline1588038tex2html_wrap_inline15948)
6500 (R) 8.2 10-6
9000 (I) 10.0 10-6
Table 46: Observational upper and lower limits to the EBL intensity as determined from surface photometry or galaxy counts

* cgs = 10-9 ergs cm-2 s-1 sterad-1 Å-1 .

A selection of upper limits from photometric measurements as well as lower limits obtained from galaxy counts are summarised in Table 46 (click here). In the table, the author(s) and date of publication are given in Col. (1). The wavelength of observation and the tex2html_wrap_inline15998 value (or its upper limit) as given in the original publication are listed in Cols. (2) and (3). In Col. (4) we give our critical revision (upper limit) of each tex2html_wrap_inline15998 value; in deriving these "revised values" we have tried to consider the effects of some additional uncertainties or corrections which in our opinion were not sufficiently discussed in the original paper. In Col. (5) we give tex2html_wrap_inline16002 for the revised EBL values. The last Col. (6) gives the method used.

12.2.1. Photometric upper limits

Three surface photometric measurements are included in Table 46 (click here):

(1) Dube et al. (1979) observed the total night sky brightness from the ground in eleven high-latitude fields. As a mean value of the 11 fields Dube et al. gave an EBL+DGL value of tex2html_wrap_inline16010. Because it was not possible to estimate the DGL contribution the result was interpreted as a 2tex2html_wrap_inline14579 upper limit to the EBL of 3.4 S10 or 5.1 10-9 ergs cm-2 s-1 sterad-1 Å-1. A basic problem with this method is that it starts with the total night-sky brightness which is a factor of tex2html_wrap_inline10939 100 brighter than the EBL. Thus, very accurate measurements of the absolute intensities of ZL and airglow are required. The most critical point in the data analysis of Dube et al. was the way they corrected for the airglow. They assumed that airglow is a linear function of sec z and used linear extrapolation to sec z = 0 to eliminate airglow. This method is doubtful since the sec z - dependence of the airglow is not strictly linear but follows the so-called van Rhijn's (1921) law. Mattila et al. (1991) have reanalysed the airglow problem using, as far as possible, the observational values given in Dube et al. (1979) and in Dube (1976). They have thus found that Dube et al. probably have overestimated the airglow intensity by tex2html_wrap_inline10939 3 S10. Thus the residual value for EBL + DGL should be increased by this amount, resulting in tex2html_wrap_inline16038 = 4.0 tex2html_wrap_inline11177 1.2 S10 or an tex2html_wrap_inline16044 upper limit of 5.2 S10.

(2) Toller (1983) utilized measurements of a photometer aboard Pioneer 10 as it moved out of the zodiacal dust cloud (tex2html_wrap_inline16048). From these he subtracted integrated starlight and gave a value for the average brightness of the diffuse background light of tex2html_wrap_inline16050 = 3.3 tex2html_wrap_inline11177 1.2 tex2html_wrap_inline11131 He estimated tex2html_wrap_inline16056 to be 2.0 tex2html_wrap_inline11177 0.4 tex2html_wrap_inline11131. As a final result Toller thus obtained an EBL intensity of 1.3 tex2html_wrap_inline11177 1.3 tex2html_wrap_inline11131 which he expressed as a tex2html_wrap_inline16066 upper limit of tex2html_wrap_inline15998 tex2html_wrap_inline15880 3.9 tex2html_wrap_inline11131.

Since Toller's EBL value has been frequently cited as the EBL reference value, it deserves a detailed discussion of errors. The basic problem for his EBL determination is the large field of view (tex2html_wrap_inline16074 tex2html_wrap_inline16076) of the photometer. Thus, the starlight entered with full weight into the measured sky brightness, and in order to derive the small residual EBL one must know the ISL very accurately in the Pioneer 10 photometric system. This was not fully the case. The ISL values of Roach & Megill (RM, 1961) and Sharov & Lipaeva (SL, 1973) are based on the Harvard-Groningen (Pickering et al. 1918, 1923, 1924; van Rhijn 1929) and Mount Wilson starcounts (Seares et al. 1930) the magnitudes of which were calibrated by using photographic techniques. Sharov & Polyakova (1972) have shown that the Harvard-Groningen photographic magnitude scales are in need of positive corrections of as much as 0.4 mag to 0.5 mag in order to reduce stars of 7 to 16 mag from tex2html_wrap_inline16078 to the B system. In their ISL summation SL tried to take these photometric corrections into account and thus their ISL values should be given the preference over the RM values. Then an average tex2html_wrap_inline16050 of 3.9 tex2html_wrap_inline11131, instead of 3.3 tex2html_wrap_inline11131, is obtained. There is a the remaining systematic error of the Sharov and Lipaeva ISL values due to the scale errors which is at least 15%. With an average ISL value of 25 tex2html_wrap_inline11131 this amounts to 3.8 tex2html_wrap_inline11131. The systematic error of the Pioneer 10 photometry itself (e.g. due to calibration) has been given as 8% (Weinberg & Schuerman 1981), which for 25 tex2html_wrap_inline11131 corresponds to 2.0 tex2html_wrap_inline11131. A further uncertainty of 1.6 tex2html_wrap_inline11131 results from variations in the cutoff for bright stars. The total error resulting from quadratically adding the systematic and statistical errors then is 4.8 tex2html_wrap_inline11131.

Thus we end up with a revised EBL value of tex2html_wrap_inline15998 = 1.9 tex2html_wrap_inline11177 4.8 tex2html_wrap_inline11131, which corresponds to 2.3 tex2html_wrap_inline11177 5.7 10-9 ergs cm-2 s-1 sterad-1 Å-1 or to a one tex2html_wrap_inline14579 upper limit of 8.0 10-9 ergs cm-2 s-1 sterad-1 Å-1.

(3) Mattila & Schnur (1990), on the basis of their observations in the dark cloud area L1642, have presented a preliminary estimate for the EBL of 6.5 tex2html_wrap_inline11177 2.5 10-9 ergs cm-2 s-1 sterad-1 Å-1. In this method an opaque dark cloud is used as a zero point where the EBL is negligible or at least much smaller than in the transparent surroundings of the cloud. Starlight, zodiacal light and the atmospheric components are eliminated in the differential surface brightness measurements used in this method. The main problem is the elimination of the scattered light (DGL) which is present both in the opaque parts of the cloud as well as in its (semi)transparent surroundings. In view of the preliminary character of the above-mentioned value we prefer to interpret it as an upper limit, tex2html_wrap_inline15998 tex2html_wrap_inline15880 9.0 10-9 ergs cm-2 s-1 sterad-1 Å-1.

12.2.2. Galaxy counts

Deep galaxy counts by Cowie et al. (1994), Tyson (1995) and Morgan & Driver (1995) have provided estimates for the contribution of galaxies to the EBL. These lower limits to the EBL are given in Table 46 (click here) for several wavelength bands between 3400 Å and 2.2 tex2html_wrap_inline10901m.

(1) The EBL values of Cowie et al. are for a K-band-limited sample with tex2html_wrap_inline16160 22 mag. The K-band counts have a slope tex2html_wrap_inline16164 = 0.26 at K = 22 mag which implies that the surface brightness contribution per magnitude interval is converging. Contrary to this the B-band counts have a divergent slope tex2html_wrap_inline16164 = 0.45 (Gardner et al. 1993). Thus it is suggested that the total EBL at U to visual wavelengths may be substantially higher than the values given in the Table.

(2) The EBL values of Tyson are for a sample to an isophotal limiting magnitude of 29 Bj magnitude arcsec-2. The limiting magnitudes of the counts are tex2html_wrap_inline10939 27 mag at Bj, 26 mag at R, and 24 mag at I. Galaxies fainter than 20 mag at Bj contribute about 75% of the EBL at 4500 Å. Most of the EBL flux originates from galaxies around B = 24 mag.

(3) The EBL values of Morgan & Driver (1995) are from counts with limiting magnitudes of B = 26 mag, V = 24.5 mag, R = 26 mag and I = 22.5 mag. Morgan & Driver adopted two approaches in estimating the EBL: firstly they used direct observations of the number counts to sum up the EBL to the limiting magnitude; secondly they used a dwarf-dominated luminosity function to extrapolate the number counts down to a limiting magnitude of tex2html_wrap_inline15978 = 38 mag. The EBL values for the second case are seen to be a factor of 2 to 5 higher than for the first case. This gives an impression of the possible importance of the contribution by very faint galaxies, tex2html_wrap_inline16200 30 mag, to the EBL.

12.3. Infrared



tex2html_wrap_inline10929 tex2html_wrap_inline13261 Reference tex2html_wrap_inline13261 Reference Ref.
tex2html_wrap_inline10901m nW m-2 sr-1 nW m-2 sr-1
1.25 393 tex2html_wrap_inline1117713 DIRBE dark sky 50 - 104 DIRBE residual 1
2.2 150 tex2html_wrap_inline111775 '' 15 - 26 '' 1
3.5 63 tex2html_wrap_inline111773 '' 15 - 24 '' 1
4.9 192 tex2html_wrap_inline111777 '' 9 - 22 '' 1
12 2660 tex2html_wrap_inline11177310 '' 102 - 164 '' 1
25 2160 tex2html_wrap_inline11177330 '' 136 - 210 '' 1
60 261 tex2html_wrap_inline1117722 '' 31 - 42 '' 1
100 74 tex2html_wrap_inline1117710 '' 20 - 35 '' 1
140 57 tex2html_wrap_inline111776 '' 12 - 63 '' 1
240 22 tex2html_wrap_inline111772 '' 8 - 33 '' 1
111 108 FIRAS dark sky 1
143 63 '' 1
250 30 '' 1
500 6 '' 1
500-5000 680/tex2html_wrap_inline10929(tex2html_wrap_inline10901m) FIRAS residual 2
400-1000 3.4 (tex2html_wrap_inline10929/400 tex2html_wrap_inline10901m)-3 FIRAS residual 3
10 - 40 6 h (tex2html_wrap_inline10929/tex2html_wrap_inline10901m)0.55 tex2html_wrap_inline12391-ray method 4
Table 47: Upper limits and claims of tentative detections of the cosmic infrared background radiation

References: 1 Hauser (1996),2 Mather et al. (1994), 3 Puget et al. (1996), 4 Dwek & Slavin (1994).
h = H0/100 km s-1 Mpc-1.
For conversion of the units to MJy/sr see Table 6.

The Diffuse Infrared Background Experiment (DIRBE) aboard the Cosmic Background Explorer (COBE) was designed to perform a sensitive search for the Cosmic Infrared Background Radiation (CIBR). Special care was paid in the design to supression of stray radiation. During the 10-month cryogenic operation of COBE in 1989 - 1990 DIRBE mapped the whole sky with high redundancy in ten wavelength bands between 1.25 and 240 tex2html_wrap_inline10901m. DIRBE is completely free from any residual atmospheric effects or contamination by rocket exhaust which have made many of the previous balloon or rocket borne experiment results problematic. The main obstacles for a determination of the CIBR are the remaining strong foreground components which contribute to the infrared sky brightness with varying importance over the whole wavelength region. As detailed in previous section, these are the zodiacal light, the light of unresolved stars and the thermal emission by interstellar dust (cirrus). Since there is no distinct spectral signature known in the CIBR, the separation of the foreground components has to be based on modelling of their different spatial or broad band spectral distributions. In the case of the zodiacal component also its temporal variation during a year can be utilized. This modelling process is still being worked on by the DIRBE team. The most recent, still preliminary results have been presented by Hauser (1996). They are reproduced in Table 47 (click here) for the ten DIRBE bands as well as for selected wavelengths based on FIRAS data. The first column of results gives the upper limits on the CIBR based on the darkest spots observed on the sky. Because no foreground components were subtracted, these values are stringent upper limits to any isotropic component of the infrared sky brightness. In the second column of results the range of DIRBE sky brightness residuals at high galactic latitude after subtraction of "best models" currently available for zodiacal light and emission, starlight and interstellar cirrus are given. These numbers, as emphasized by Hauser (1996), are still preliminary. The uncompleteness of the foreground modelling is reflected in the fact that the spectrum of the residual brightness at mid-IR wavelengths shows a resemblance with the zodiacal light spectrum. For ease of comparison, part of Table 6 on total infrared sky brightness in dark spots is repeated here in the left part of the table.

Kashlinsky et al. (1996) have tried to convert the smoothness of the spatial distribution of DIRBE light into upper limits to a CIBR radiation component coming from clustered matter evolving according to standard scenarios. They find that the upper limits to such a component between 1.25 tex2html_wrap_inline10901m - 100 tex2html_wrap_inline10901m are by a factor of 4 to 100 lower than the Hauser et al. (1996) residual isotropic brightnesses given in Table 47 (click here). These values have to be taken with caution, however, since their derivation is strongly model-dependent.

Using COBE/FIRAS data Mather et al. (1994) have estimated that the CMB spectrum between 0.5 mm and 5 mm deviates from a 2.726 K blackbody shape by less than 0.03% of the peak intensity. Taking twice this amount as an upper limit to an additional CIBR implies tex2html_wrap_inline16388 W m-2 sr-1.

Puget et al. (1996) have claimed a tentative detection of far-IR CIBR using COBE/FIRAS data. They have modelled and removed the foreground components above 140 tex2html_wrap_inline10901m. For estimating the interstellar cirrus emission they used its correlation with HI 21-cm data, and for zodiacal emission its spectral and spatial distribution as determined at shorter wavelengths, tex2html_wrap_inline10907 100 tex2html_wrap_inline10901m. The residual isotropic component claimed for the 400 tex2html_wrap_inline10901m - 1000 tex2html_wrap_inline10901m range can be represented by tex2html_wrap_inline16406 W m-2 sr-1.

An indirect method for measurement of the mid-IR CIBR is based on the spectra of tex2html_wrap_inline12391-ray sources, since tex2html_wrap_inline12391-rays interact with intergalactic IR-photons by pair production, giving rise to energy-dependent extinction. A recent application gives, for tex2html_wrap_inline16416, the result tex2html_wrap_inline16418 W m-2 sr-1 (Dwek & Slavin 1994). The result depends on the Hubble constant h = H0/100 km s-1 Mpc-1. This estimate is by a factor of tex2html_wrap_inline10939 10 lower that the DIRBE isotropic residuals at 10 and 25 tex2html_wrap_inline10901m, but fits nicely to the DIRBE isotropic residuals at shorter and longer wavelengths (see Fig. 79 (click here)). Again, there are uncertainties in this method, since the intrinsic high energy gamma ray spectra before attenuation by interaction with the cosmic infrared radiation field are not really known.

12.4. Overview on EBL observations

Figure 79 (click here) summarises the current observational limits to the extragalactic background light in the wavelength range from 0.1 tex2html_wrap_inline10901m to 1000 tex2html_wrap_inline10901m. In the visual and near-infrared range, where discrepancies between different methods of determination are particularly large, we also plot a few selected model predictions for comparison, without the intent to discuss them here. Compared to the situation ten years ago, the gap between upper limits from direct measurements, lower limits from galaxy counts, and model predictions is getting smaller, being mostly less than a factor of ten by now. A comparison with the total sky brightness values shown in Fig. 1, which are typically brighter by two orders of magnitude, is informative. In this comparison please note that tex2html_wrap_inline13261 and tex2html_wrap_inline14031 are identical units of brightness.


We are grateful to M. Cohen for providing the estimates of integrated starlight shown in Figs. 61 and 62 and in Tables 24 and 29. We thank P. Feldman, R. MacQueen, H. Kimura and H. Lauche for helpful advice and discussions and M. Rowan-Robinson for supplying his infrared sky maps in digital form. We are grateful to P. Väisänen, who contributed the first and the last figure of this paper, and we very much thank D. Anders, K. Meissner-Dorn, and in particular M. Weckauf for the careful and patient work in producing or making computer-readable most of the numerous figures of this paper.

next previous
Up: The 1997 reference

Copyright by the European Southern Observatory (ESO)