Deficiencies With "Dark-Sky" Flats

Analysis of 2MASS Protocamera data indicate that there are significant inhomogeneities in the photometric response across the array (see References 1--3). Figure 1 shows a graphical representation of these variations. This image was constructed from the ROUND analysis that computes the average ratio of the brightnesses of stars in each of their six apparitions on the detector during a 2MASS scan to that in the 3rd apparition. The array is subdivided into 4 super-columns, so the resulting ''map'' of the photometric response is a 4x6 grid. Note that the values in each grid column have been normalized to the third row, so there is no quantitative cross-scan information other than the fact that there is structure in the cross-scan direction. The peak-to-peak variations in the response map are 23%.

  
Figure: The relative photometric response for different regions across the 2MASS Protocam focal plane, as measured from June 1994 data. The peak-to-peak variation in response is 23%. This image was constructed from ROUND analysis (References 1 and 2).

There are a number of factors that contribute to the photometric response variations:

1. Bias fluctuations: The in-scan pattern visible in the photometric response map shown in Figure 4 is similar to the "reset decay" pattern seen in Nicmos 3 arrays. The influence of this pattern and possible fluctuations are discussed in Reference 4. As discussed in Reference 4, the effect of this pattern may be minimized or eliminated entirely by operating the array at a slightly warmer temperature, or by using a slightly longer delay after the detector is reset.

2. Significant background from telescope/optics: At infrared wavelengths, the thermal emission from the telescope can contribute significantly to the illumination of the array. Therefore, there is a "local" component to the background that is being used to construct the flat-field images. If the local component varies because of temperature fluctuations or flexure in the camera then the background illumination of the array will change, possibly leading to errors in the flat-field construction ( i.e. see Reference 5).

3. Focused and diffuse light paths different within telescope/camera: The optical paths taken by focused and diffuse light sources that illuminate a given pixel can be quite different. Therefore, using a diffuse illumination source, such as the sky background, to determine the flat-field that will be used primarily to calibrate images of point and small extended sources can lead to significant photometric errors ( i.e. see Reference 6). This effect has been seen in many different near-infrared camera systems using both HgCdTe and InSb arrays.

4. Sky Color: Since the quantum efficiency of the NICMOS 3 detectors is wavelength dependent, the flat-field correction should be constructed using illumination that has a similar color to the typical astronomical sources to be measured. In the near infrared the night sky emission arises primarily from OH airglow lines. This is a relatively poor color match for virtually all of the sources expected to be observed by 2MASS. The differences in the colors between flat-field and astronomical sources may introduce subtle color terms.

The large observed variations may at first seem inconsistent with the photometric repeatibility tests reported in Reference 7. However, the normal 2MASS scanning technique ensures that each source is observed six times at different in-scan positions along the array. Therefore, any in-scan response variations are effectively averaged out. Cross-scan variations are not, however. The repeatibility tests described in Reference 7 were run for multiple scans that differ very little in cross-scan position on the sky. Therefore, they will yield no information about residual scatter in the photometry due to the cross-scan photometric response variations.