Analyses of the 2MASS scan-scan overlap regions has revealed that there can be a constant photometric bias between at least the edges of adjacent scans. The apparent bias can change amplitude both from night-to-night and during a single night. It is often different in the three bands. An example of such a bias is shown in the overlap plot from 970712n which was generated during normal operational processing of that night's data. That plot shows essentially no systematic bias in the J and Ks photometry between scans, and a relatively constant 2-3% bias between scans at H. The scan-overlap plot from 980125n (one of the northern RTB nights) shows approximately 1% of bias in all three bands, although the sign of the bias in H is opposite that in J and Ks.
The importance of understanding and ultimately correcting this effect has now become critical with the analysis of the first 3 southern RTB nights. The cross-scan bias appears significantly greater in the southern data than most anything seen in the north. The plot showing the overlap photometric statistics for 980328s illustrates that the bias between scans can be as large as 8% in J band, 3-4% in H and 4-5% in Ks. These biases are consistent with those seen in the overlap plots from the other southern nights. As a result, the Level 1 Science Requirement for global photometric uniformity may be compromised for these data.
Discussion of the origin of this bias at the July 1998
Science Team Meeting focused on flat-fielding deficiency as the probable
cause of this edge-to-edge bias. Recent analyses of the cross-scan
photometric response that would be sensitive to flat-fielding residuals
proved inconclusive. The analysis presented here examines an alternative
explanation for the photometric bias: variation of the point-spread-function
(psf) across the 2MASS focal planes.
We can examine the difference between the aperture and
profile-fit photometry for sources as a function of cross-scan position
to test whether or not the psf's vary across the focal plane. Since
the photometry of the six in-scan apparitions is effectively averaged in
2MASS, in-scan psf-variations should not contribute to systematic offsets.
The examples below show the photometry comparisons for several nights.
a. Scan-to-scan photometry bias plot from scan overlap regions
c. Median J PSF mag - J aperture mag in 10 pixel bins as a function of cross-scan position
d. Median H PSF mag - H aperture mag in 10 pixel bins as a function of cross-scan position
e. Median Ks PSF mag - Ks aperture mag in 10 pixel bins as a function of cross-scan position
The magnitude difference versus cross-scan position plots
show that there should be <1% bias when comparing the scan overlap
regions (10% of edges) between overlapping J and Ks scans.
There is a systematic slope in the differences across the focal plane at
H-band, however. The amplitude of that slope is 2-3%.
These values are consistent with the biases seen in the scan-overlap plot
in Figure 1a.
2. 980125n - The scan overlap plot for 980125n indicates that there should be at most 1% bias between the photometric differences between the edges of the arrays. Figures 2b-2d show the distribution of the difference between psf-fit and aperture magnitudes as a function of cross-scan position for all high signal-to-noise stars measured during the night.
a. Scan-to-scan photometry bias plot from scan overlap regions
b. Median J PSF mag - J aperture mag in 10 pixel bins as a function of cross-scan position
c. Median H PSF mag - H aperture mag in 10 pixel bins as a function of cross-scan position
d. Median Ks PSF mag - Ks aperture mag in 10 pixel bins as a function of cross-scan position
The magnitude difference plots for this night show essentially no bias between the opposite edges of the arrays in all of the bands. Note that there still appears to be cross-scan structure in the differences in the J-band. H and Ks appear fairly uniform.
3. 980328s - The scan overlap plot for this night indicates a severe edge-edge photometric bias between adjacent scans. Moreover, the amplitude of the bias changed significantly between the first intercalibration period (0h-1h UT) and the later periods (>1h UT). Figures 3b-3d show the distribution of the difference between psf-fit and aperture magnitudes as a function of cross-scan position for the survey scans taken between 0:00 and 1:00 UT, and figures 3e-3g show the magnitude differences for the survey scans taken after 1:00 UT.
a. Scan-to-scan photometry bias plot from scan overlap regions
The difference plots for this night show rather convincingly
that the photometric biases seen in the scan-scan overlap plots correlate
closely with the distributions of differences between psf-fit and aperture
photometry. Most importantly, the changes in the scan-scan overlap
offsets during the night, including a sign flip in the Ks
band, correlate with changes in the difference distributions during the
night.
An important point is the severity of the photometric bias in the scan overlaps for the southern data, and that it is apparently worse than in any yet seen in the north. If the bias is caused by variation of the psf across the focal plane, then it may appear worse in the south primarily because the seeing is so much better, on average, than in the north. Atmospheric seeing degradation will tend to smear and circularize images, thus masking psf-variations due to camera and/or telescope optics. To test this, we should determine if the scan-scan overlap photometric biases and/or the psf-fit/aperture magnitude differences are more pronounced in the north under conditions of very good seeing.
It should be emphasized that the implied variation of the psf across the focal plane is probably very small to produce the <10% effect in the measured photometry for high signal-to-noise sources. Examination of the 3x3 composite psf-images (shown below) from selected scans from the nights discussed do show small differences around the focal plane. It is quite apparent that the psf size is considerably smaller in the southern data, and variations around the focal plane are more obvious.
a. 3x3 composite point source images from scan 049 on 970712n
b. 3x3 composite point source images from scan 095 on 980125n
c. 3x3 composite point source images from scans 012 and 066 on 980328s
Last Update - 28 July 1998
R. Cutri - IPAC