Following the work of Hauser et al. (1998), there has been significant progress in direct determination of the CIB in the near infrared window using DIRBE data. Since the uncertainty in the statistical model of the faint stellar contribution used by Arendt et al. (1998) was a major source of uncertainty at these wavelengths, several approaches have been used to reduce that uncertainty.
Dwek & Arendt (1998) assumed that the CIB at 2.2 µm is close to the integrated light from galaxies at this wavelength. Subtracting the integrated galaxy light and the zodiacal light (Kelsall et al. 1998) from the DIRBE 2.2 µm maps yielded a map of starlight at 2.2 µm (the ISM contribution at this wavelength is negligible). Using this 2.2 µm starlight map as a spatial template for starlight at 3.5 µm, they obtained a significantly positive 3.5 µm residual, I = 9.9 + 0.312[ I(2.2) - 7.4]±2.9 nW m-2 sr-1, where I(2.2) is the actual CIB at 2.2 µm. They tentatively identified this 3.5 µm residual as the CIB, though they did not demonstrate that it was isotropic. They also obtained somewhat improved upper limits on the CIB at 1.25 and 4.9 µm using the same stellar template (Table 1).
Gorjian, Wright, & Chary (2000) reduced the stellar foreground uncertainty more directly by measuring all of the stars brighter than 9th magnitude at 2.2 and 3.5 µm in a dark 2° × 2° field near the north Galactic pole. They calculated the contribution of fainter stars using the statistical model of Wainscoat et al. (1992). The uncertainty in the calculated light from objects below this faint limit is negligibly small, even at high galactic latitude. They also used a model for the zodiacal light contribution which differed from that of Kelsall et al. (1998). They argued that the Kelsall et al. model left a high galactic latitude residual at 25 µm which is dominated by IPD emission. The IPD model used by Gorjian et al. was similar to that of Kelsall et al. in that it required that the apparent annual time variation be reproduced, but it further required that the residual brightness at 25 µm after zodiacal light removal be constant at a value of zero at high galactic latitude (the "very strong no zodi principle" of Wright 1997). After removing the IPD and stellar contributions, Gorjian et al. found significant positive residuals at 2.2 and 3.5 µm which they identified as probable detections of the CIB (Table 1). Had they used the IPD model of Kelsall et al, their results would have been ~ 40% higher than those shown in Table 1, a clear illustration of the uniqueness problem in modeling the zodiacal light. With a field covering only ~ 8 DIRBE beams, they did not demonstrate the isotropy of these signals.
Wright & Reese (2000) compared the histograms of the pixel intensity distributions in the DIRBE 2.2 and 3.5 µm maps in five fields at high galactic and high ecliptic latitudes with the histograms predicted from the star count model of Wainscoat et al. (1992). The IPD contribution had first been removed from the observations using the model of Gorjian et al. (2000). They found that the predicted histograms had the same shape as those observed, but had to be displaced by a constant intensity to agree with the observed histograms. The necessary shift was consistent in the five fields analyzed. They interpreted this shift as the CIB, and obtained average values for the five fields consistent with the values found by Gorjian et al. (2000) (Table 1). They noted that the histogram method is statistically more powerful for finding a real residual and less subject to systematic errors in the star count model than the subtraction approach used by Arendt et al. (1998).
Wright (2000) considerably strengthened the case for detection of the CIB at 2.2 µm. He used the newly released 2MASS catalog to remove the contribution of Galactic stars brighter than 14th magnitude from the DIRBE maps at 1.25 and 2.2 µm in 4 dark regions in the North and South galactic polar caps. Each region was about 2° in radius. Using the same IPD model as Gorjian et al. (2000) Wright obtained 2.2 µm residuals in his 4 fields consistent with each other, and consistent with those of Gorjian et al. and Wright & Reese (2000). Hence, there is a strong case for detection of an isotropic CIB at 2.2 µm. The average value of the 2.2 µm CIB determinations by these three methods is I = 21.8±5.5 nW m-2 sr-1. The scatter in the 1.25 µm residuals was too large to claim a detection. Wright's analysis does provide the strongest current upper limit on the CIB at that wavelength. The dominant uncertainty in all of these analyses of the near infrared CIB remains the uncertainty in the zodiacal light contribution.
Kashlinsky & Odenwald (2000) found that the amplitude of the fluctuations in the DIRBE 1.25-4.9 µm maps varied with csc(b). Extrapolating to csc(b) = 0, they found positive intercepts which they attributed to fluctuations in the CIB. They reported rms fluctuations of 15.5+3.7-7.0, 5.9+1.6-3.7, 2.4+0.5-0.9, and 2.0+0.25-0.5 at 1.25, 2.2, 3.5, and 4.9 µm respectively, where the errors are 92% confidence limits. Adopting their argument that the fluctuations are expected to be 5-10% of the CIB, these results are generally consistent with the direct background determinations listed in Table 1.