I. Introduction

---

6. Cautionary Notes

b. Point Source Catalog (PSC)

1. Composite Nature of the PSC
   
2. Depth of the PSC
   
3. The Default Magnitude
   
4. Origin and Quality of Photometry - PSC Quality Flags
   
5. Photometric Uncertainties
   
6. Sources of Unreliability
   
7. Single-Band Sources
   
8. Photometric Biases
   
9. General Effects of Confusion
   
10. Very Bright Stars
    
11. Extremely Faint Sources
    
12. Extended Sources and Galaxy Contamination
13. Position Reconstruction
    
14. Tile Overlap Regions
    
15. Solar System Objects
    
16. Optical Associations
    

i. Composite Nature of the PSC

The 2MASS PSC is a composite release that is comprised of two subsets contained in a single database. The first part is the signal to noise ratio (SNR) >10, high reliability Catalog that meets or surpasses all of the Level 1 Requirements for sensitivity, uniformity and reliability in unconfused regions of the sky. Very bright stars that saturate the shortest 2MASS exposures are included in this set even though they do not satisfy the photometric accuracy requirements. The second component of the PSC is a faint extension that contains sources that reach 0.5-1.0 mag beyond J<15.8, H<15.1 and K_s<14.3 mag levels, where the PSC is >99% complete. The completeness, reliability and uniformity of the faint extension sources are not as good as the high-reliability Catalog subset of the PSC, and may not satisfy all of the specifications described in the 2MASS Level 1 Requirements. They are included in the PSC because, when used with some care, they can be a valuable resource for certain classes of investigations in which homogeneous, statistical datasets are not necessary.

Sources are included in the PSC if they have a SNR>7 detection in at least one band, or have SNR>5 and are detected in all three 2MASS bands. Sources in the high reliability Catalog meet the following band-by-band criteria:

1. ph_qual="A" in a band OR (rd_flg="1" OR rd_flg="3") in that band
2. use_src="1"

where ph_qual, rd_flg and use_src are quality parameters that accompany each source record in the PSC and are described in subsections iii and xiv below. Criterion 1 specifies that a source must have a flux measurement with signal to noise ratio (SNR) >10 or it must be brighter than the detector saturation limits in either the 1.3 s Read_2-Read_1 or 51 ms Read_1 exposures. Criterion 2 above refers to sources that fall within the overlap region between adjacent Survey Tiles, and which could have been detected more than once. If the source is multiply-detected, only the apparition that that falls closest to the center of its respective Tile is included in the PSC (see V.4). If a source is not detected in all observations, then it is still included in the PSC, but is flagged with use_src="1" only if it lies closer to the center of its Tile than any of the other Tiles in which it could have been detected. This requirement ensures the strict detection uniformity of the Survey.

Table 1 provides a breakdown of the total number of detections, and the number of sources which satisfy the high-reliability Catalog criteria by band (b). For the column labeled "All", the count of high-reliability Catalog sources includes the total number that satisfy the above criteria in at least one band.

Table 1 - Band-by-Band Breakdown of Full PSC and High-SNR Catalog Detections

Name	Criteria	N(J)	N(H)	N(Ks)	All
Full PSC	rd_flg[b]!="0"	455371014	436797138	375248502	470992970
High Reliability Catalog	(ph_qual[b]="A" OR rd_flg[b]="1" OR rd_flg[b]="3") AND use_src="1"	321117915	269533860	210729664	341317908

The high-reliability Catalog and faint extension of the PSC are illustrated graphically in Figures 1a, 1b, and 1c which show the differential J, H and K_s source counts for ~1.9 million sources in the north and south Galactic Caps (|b|>75°). Counts for the high SNR, high reliability Catalog are shown by the blue, green and red lines in the J, H and K_s plots, respectively. The faint extension components are shown by the cyan, yellow and magenta lines in the three plots, respectively. Black lines show the total source counts in all plots.

The Catalog component is divided into two regions, distinguished by the heavy and thin lines in the Figures, that correspond to the two parts of Criterion 1. The majority of Catalog sources, denoted by the heavy lines, satisfy all of the Level 1 Requirements for the mission, in the absence of confusion. The bright part of the Catalog, represented by thin lines, consist primarily of detections of bright stars that saturate the 51 ms Read_1 exposures (rd_flg="3"), with a very small contribution of fainter rd_flg="1" sources which are usually associated with confused regions (see I.6.b.vi below). Positions and brightness estimates for the very bright, saturated stars are now included in the All-Sky PSC (they were not in the Incremental Release PSC versions). These detections are believed to be highly complete and reliable, and to have excellent positions, but the quality of the photometry is relatively poor because of the saturation. The Level 1 Requirements place no specifications on these sources, and they do not satisfy the photometric accuracy specification that applies to unsaturated sources.

The faint extension sources, denoted by the cyan, yellow and magenta lines in Figures 1a, 1b, and 1c, comprise 29.5%, 38.3% and 43.8% of all J, H and K_s detections in the PSC, respectively. Most of these sources, which make up the peaks in the distributions near J~16.5, H~16 and K_s~15.5 mag, have lower photometric accuracy (ph_qual="B","C" or "D") simply because they are faint. There is also a population that extends brightward of these peaks up to the approximate levels at which 1.3 s "Read_2-Read_1" exposures saturate (J~9, H~8.5 and K_s~8 mag). These objects, which constitute about 1% of all PSC sources over this brightness range, are cases where the measurement accuracy is degraded because of confusion with other sources or other artifacts.

Section II.2.b describes further the composite nature of the PSC, and contains tips and guidelines on how to select samples appropriate to address a variety of scientific investigations.

PSC |b|>75^oSource Counts

Figure 1a	Figure 1b	Figure 1c

ii. Depth of the PSC

The flux levels reached by the PSC are not uniform across the sky. At the magnitude limits corresponding to the SNR=10 levels specified by the Level 1 Requirements (J=15.8, H=15.1 and K_s=14.3 mag), the completeness of the Survey is ultimately limited by confusion noise due to astronomical source backgrounds. At fainter flux levels, the depth of the Survey measurements is governed by the atmospheric conditions at the time of the observations.

The PSC is >99% complete for J<15.8, H<15.1 and K_s<14.3 mag across the sky, except in regions of high source density. This is illustrated in Figures 2, 3, and 4 which show the differential surface density of PSC sources in 0.2 mag wide bins at these brightness levels in Galactic aitoff projections. Away from the Galactic Plane and Bulge, the surface density varies smoothly across the sky. Within +90^o of longitude of the Galactic Center the source counts drop dramatically due to the influence of confusion noise. The threshold for source detection used in pipeline processing is scaled to the point source noise levels estimated from the zero-sum-filtered Atlas Images (cf. IV.4.a.ii). This threshold is sensitive to confusion noise, so the detection depth decreases with increasing source density. The decrease in sensitivity can also occur in more isolated regions of high source density, such as in Tiles that contain large globular clusters (e.g., 47 Tucanae).

Figures 5, 6, and 7, show the differential surface density of PSC sources one magnitude fainter than the nominal SNR=10 brightness levels. These will be characteristic of the faint extension of the PSC. At these flux levels, the density of extracted PSC sources varies on spatial scales corresponding to the Survey Tiles (III.2b). These variations are caused by the influence on survey depth by atmospheric transparency, seeing and emissive backgrounds, which could change from scan-to-scan during the Survey. The zone of incompleteness in the Galactic Plane and Bulge is also correspondingly larger at these fainter magnitude limits.

The completeness and depth of the PSC as a function of brightness is discussed in more detail in VI.7a.

Differential PSC Surface Density at J=15.9, H=15.1 and K_s=14.3 mag

Figure 2	Figure 3	Figure 4

Differential PSC Surface Density at J=16.9, H=16.1 and K_s=15 .3 mag

Figure 5	Figure 6	Figure 7

iii. The Default Magnitude

Point source photometry is measured differently for three brightness regimes in the PSC. The J, H and K_s default magnitudes (j_m, h_m, k_m) listed for each object represent what are the best available measurements in each band, as determined by the automated processing. The rd_flg specifies the origin of the default magnitude in each band.

Because different algorithms are used for each regime, and various normalizations are used to tie the photometry in each brightness interval together, photometric biases can occur between them. These biases are discussed in subsection vii below.

Non-Saturated "Read 2-Read 1" (J>9, H>8.5, K_s>8 mag)

The majority of sources in the PSC have brightnesses below the saturation threshold of the 1.3 s "Read 2-Read 1" (R_2-R_1) exposures. R_2-R_1 saturation occurs near J~9, H~8.5 and K_s~8 mag, but can vary by up to ~0.5 mag depending on the atmospheric transparency and seeing, and the location of a source relative to pixel centers. Most default magnitudes for non-saturated R_2-R_1 sources in the PSC are measured using profile-fit photometry (see IV.4b) performed simultaneously on all six individual 1.3 s R_2-R_1 exposures covering the source. These sources have rd_flg=2 in the appropriate bands. The profile-fit photometry is normalized to curve-of-growth-corrected aperture photometry (see IV.4c), which is also measured for non-saturated R_2-R_1 sources. The aperture photometry is supplied in the j_m_stdap, h_m_stdap and k_m_stdap fields of the PSC for non-saturated R_2-R_1 sources.

Occasionally, the profile-fitting photometry routines fail for sources in crowded environments or that lie in regions with complex backgrounds. If an aperture magnitude is available, it will be listed in the default magnitude field in the appropriate band, and the rd_flg=4 in the affected bands. However, these magnitudes are highly uncertain due to the confused environment.

In a small number of cases, both profile-fitting and aperture photometry fail to return a valid measurement for detections made on the R_2-R_1 exposures. In such cases, the default magnitude is null and rd_flg=9 in the affected band, indicating there is technically a detection at that position, but no useful measurement can be made. In practice, this occurs in confused environments.

Non-Saturated "Read 1" (4.5<J<9, 4<H<8.5, 3.5<K_s8 mag)

Sources brighter than J~9, H~8.5 and K_s~8 mag will saturate in the 1.3 s R_2-R_1 exposures. Up to brightnesses of J~4.5, H~4 and K_s~3.5 mag, these sources are non-saturated on the short 51 ms "Read 1" (R_1) exposures. The default magnitudes for such sources are measured using aperture photometry (see IV.4c) on the R_1 exposures, and rd_flg=1 in the appropriate band.

If a source is detected on the R_1 exposures, but the aperture photometry is unable to extract a valid measurement, then the default magnitude is null and rd_flg=9 in the affected band. As for faint sources, this often occurs in confused environments. The northern observatory K_s detector exhibited very high noise levels in its central column in data taken prior to 1997 December 25 UT (an update to the northern camera controller electronics solved this problem). This noisy column was masked off, so detections that fell on or near the column often had failed measurements (see III.1b). Such sources will have good photometry in J and H, but a detection with no valid measurement in K_s.

Saturated "Read 1" (J<4.5, H<4, K_s<3.5 mag)

Stars brighter than approximately J~4.5, H~4 and K_s~3.5 mag saturate even the 51 ms exposures. A new feature of the 2MASS All-Sky Release PSC is that brightness and accurate position estimates derived directly from 2MASS data are now provided for very bright stars. Photometry for saturated R_1 stars is performed using a 1-d radial profile fit to the azimuthally-averaged image profile on the R_1 exposures, and the rd_flg value for the appropriate band is "3".

If a R_1-saturated source is detected, but the 1-d radial profile fit is unable to extract a valid brightness estimate, the default magnitude is null and rd_flg=9 in the affected band. As with non-saturated R_1 sources, this occurs when sources fall in confused environments, or when they fall on the masked central column of the northern K_s array (see III.1b).

iv. Origin and Quality of Photometry - PSC Quality Flags

The origin and general quality of the default magnitude photometry listed in the PSC are summarized by a number of informative flags that accompany each source. It is essential that users refer to these flags when interpreting photometry for any source in the Catalog. The primary quality indicator flags include rd_flg, ph_qual, cc_flg and bl_flg. Each of these is comprised of three characters, each corresponding to one band; the first character is the J-band value, the second is the H value, and the third is the K_s value.

rd_flg - Origin of Point Source Photometry

The rd_flg denotes the origin of and algorithm used to derive the quoted default magnitudes for each 2MASS Point Source. Table 2 contains the possible values of rd_flg, their interpretation and the number of instances of each value in each band in the PSC.

Table 2 - Definition of rd_flg and Occurrences in the PSC

rd_flg Value (1 per band)	Source of Photometry	N(J)	N(H)	N(Ks)
0	Not detected in this band. The default magnitude is a 95% confidence upper limit derived from a 4´´ radius aperture measurement taken at the position of the source on the Atlas Image. The sky background is estimated in an annular region with inner radius of 14´´ and outer radius of 20´´. The uncertainty in the default magnitude, [jhk]_cmsig, is null in these cases.	15621956	34195832	95744468
1	R_1 aperture photometry. The default magnitude is derived from aperture photometry (IV.4a.i) measurements on the 51 ms R_1 exposures. The aperture radius is 4´´, with the sky background measured in an annulus with an inner radius of 14´´ and an outer radius of 20´´. The photometric uncertainty in these cases is the RMS variation of the measured flux from all frames on which a source falls. If the source was measured on only one frame, the [jhk]_cmsig value becomes a flag, and has a value of >8.0 mag in the appropriate band.	1256974	2080493	1840686
2	R_2-R_1 profile-fit photometry. The default magnitude is derived from a profile-fitting measurement (IV.4b) made on the 1.3 s R_2 exposures. The profile-fit magnitudes are normalized to curve-of-growth-corrected aperture magnitudes. The uncertainty for the profile-fit measurements, [jhk]_cmsig, are derived from a goodness-of-fit metric that includes terms for uncertainties in the photon noise in the sky background reference, the noise in the source brightness, and the how well the fitted profile matches the actual point-spread-function of the source.	447910694	426256434	360945591
3	Saturated in R_1. The default mag derived from 1-d radial profile fit to star image made on the 51 ms R_1 frames. The photometric uncertainty in these cases is the RMS variation of the measured flux from all frames on which a source falls. If the source was measured on only one frame, the [jhk]_cmsig value becomes a flag, and has a value of >8.0 mag in the appropriate band.	18554	26950	24062
4	R_2-R_1 aperture photometry. The default magnitude is derived from curve-of-growth-corrected 4´´ radius aperture photometry measurements (IV.4c) on the 1.3 s R_2 exposures. The sky background is measured in an annulus with an inner radius of 14´´ and an outer radius of 20´´. The photometric uncertainty in these cases is the RMS variation of the measured flux from all frames on which a source falls.	2323	1878	2301
6	Inconsistent deblend. The default magnitude is the 95% confidence upper limit derived from a 4´´ radius aperture measurement taken at the position of the source on the Atlas Image. The sky background is estimated in an annular region with inner radius of 14´´ and outer radius of 20´´. This is used for pairs of sources which are detected and resolved in another band, but are detected and not resolved this band. This differs from a rd_flg="0", because in this case there is a detection of the source in this band, but it is not consistently resolved across all bands. The [jhk]_cmsig value is null for these cases.	6182450	8430977	12434963
9	Detection at this location, but no useful brightness measurement possible. The default magnitude and uncertainty are null in these cases. This occurs for sources that were nominally detected in this band, but for which either profile-fitting or aperture photometry could not return a useful measurement, usually because of confusion or proximity to image frame edges where a significant fraction of the measurement aperture or sky annulus falls off the focal plane.	19	406	899

ph_qual - Photometric Quality

The photometric quality flag, ph_qual, provides a concise guide to the quality of the default point source photometry that is based on signal to noise ratio, measurement quality, detection statistics, etc. Sources with ph_qual="A" in a band, have <10% measurement uncertainties, and generally meet or surpass all of the Level 1 Requirements for the Survey in unconfused areas of the sky, one of the distinguishing features of sources that make up the high SNR, high-reliability Catalog subset of the PSC (see Section i). Measurement quality decreases with alphabetically increasing values of ph_qual. The values of "U" and "X" are reserved for non-detections/inconsistent deblends or otherwise unmeasurable detections in a band. The ph_qual value is set in each band according to the precedence listed in Table 3 below. For example, a measurement in a band that satisfies the conditions for "X" is not tested for other conditions. Table 3 also gives the number of sources in each band that have the corresponding quality value.

Table 3 - Definition of ph_qual and Occurrences in the PSC

ph_qual Value (1 per band)	Meaning	N(J)	N(H)	N(Ks)
X	There is a detection at this location, but no useful brightness estimate can be extracted using any algorithm. rd_flg=9 and default magnitude is null.	19	406	899
U	Upper limit on magnitude. Either source is not detected in this band (rd_flg=0), or it is detected, but not resolved in a consistent fashion with other bands (rd_flg=6). A value of ph_qual="U" does not necessarily mean that there is no flux detected in this band at the location. Whether or not flux has been detected can be determined from the value of rd_flg. When rd_flg="0", no flux has been detected. When rd_flg="6", flux has been detected at the location where the images in all three bands (JHKs) were not deblended consistently.	21804406	42626809	108179431
F	This category includes rd_flg=1 or rd_flg=3 sources where a reliable estimate of the photometric error, [jhk]_cmsig, could not be determined. The uncertainties reported for these sources in [jhk]_cmsig and [jhk]_msigcom are flags and have numeric values >8.0.	1533	3916	7085
E	This category includes detections of any brightness or SNR where: 1) the goodness-of-fit quality of the profile-fit photometry was very poor (rd_flg=2 and [jhk]_psf_chi10.0); or 2) where profile-fit photometry did not converge, and an aperture magnitude is reported (rd_flg=4); or 3) where the number of frames on which a source was detected was too small in relation the number of frames in which a detection was geometrically possible (rd_flg=1 or rd_flg=2). The 2 limit results in many relatively bright (SNR>>10) unresolved double stars being found in this category.	2376875	3067220	2100224
A	Detections in any brightness regime where valid measurements were made (rd_flg=1,2 or 3) with [jhk]_snr10 AND [jhk]_cmsig0.10857.	321736060	269905059	211009193
B	Detections in any brightness regime where valid measurements were made (rd_flg=1,2 or 3) with [jhk]_snr7 AND [jhk]_cmsig0.15510.	93014634	82796316	60899695
C	Detections in any brightness regime where valid measurements were made (rd_flg=1,2 or 3) with [jhk]_snr5 AND [jhk]_cmsig0.21714.	28268193	58181272	59859408
D	Detections in any brightness regime where valid measurements were made (rd_flg=1,2 or 3) with no [jhk]_snr or [jhk]_cmsig requirement.	3791250	14411972	28937035

cc_flg - Contamination and Confusion

The cc_flg values identify sources that are believed to be real astrophysical sources, but whose detection and/or measurement are contaminated by proximity to either image artifacts caused by bright stars or nearby sources of equal or greater brightness. Photometry in a band for any source that contains a cc_flg value other than "0" may be corrupted and/or biased by the nearby artifacts or stars. Users should select sources with cc_flg="000" if they need samples of sources that have the lowest probability of contamination.

The algorithms used to identify confused and/or contaminated sources and to set the values of the cc_flg are discussed in IV.7. The sources of contamination encoded in the cc_flg are summarized below.

• Latent Images

  NICMOS3 detector material produces a latent image after exposure to a bright source which fades over a timescale of order 10 sec. This persistence effect produces a trail of spurious sources in the wake of a bright star at exactly the 2MASS frame spacing. Since the position and relative brightness of these artifacts are well known, detections associated with them are identified during data processing. Because a real source can lie under one of these after-images, candidate latent images are flagged with a "probability of persistence" that is derived using a chi-squared probability distribution defined by the predicted position and brightness of the latent image (cf. IV.7.ii). The probability distribution was normalized such that sources with persistence probabilities >50% are believed to be spurious latent image detections, and those with probabilities between 10% and 50% are believed to be real sources that may be contaminated by proximity to the latent images. Detections with persistence probability >50% were excluded from the PSC. Sources with persistence probability between 10% and 50% are indicated with cc_flg="p" in the appropriate band.

• Scattered Light and/or Photometric Confusion

  Bright stars effectively mask out regions of the sky to fainter source detection because of the intensity in the wings of their point spread function. These bright wings often cause spurious detections, so sources lying within a magnitude-dependent circular region around bright stars are flagged as confusion artifacts, and are not passed to the PSC if they are below a specified brightness relative to the "parent" bright star. Objects brighter than this threshold remain in the PSC, but are flagged with cc_flg="c" in the appropriate band(s) to indicate possible measurement contamination. Section IV.7.v contains a description of the magnitude-dependent terms used for this process.

  Point sources that lie close to another source of equal or greater brightness may suffer some bias in its photometry because of gradients in the underlying background produced by the nearby source. The degree of bias is a function of the magnitude difference between the sources and their separation. Sources for which the bias is believed to exceed 5% are also marked with cc_flg="c" value in the appropriate band. The magnitude-separation relationships used to identify sources with this potential photometric confusion are presented in Section IV.7.vi.

  A related parameter in the PSC that is useful for identifying sources that may be affected by confusion is the prox value which gives the separation (in arcsec) between each source and its nearest neighbor in the catalog.

• Diffraction Spikes

  Diffraction spikes from bright stars can trigger false detections or contaminate the measured flux for a real source that falls on or near them. Sources believed to be false detections triggered by diffraction spikes were not passed to the PSC. Sources believed to be real, but which may have photometry contaminated by diffraction spikes, have cc_flg="d". The algorithms used to identify sources confused and contaminated by diffraction spikes are described in Section IV.7.i.




• Electronic "Stripes"

Low frequency cross-talk in the 2MASS camera read-out electronics resulted in bright stars producing an "echo" that was manifested as horizontal "stripes" in the image frames offset from the star by +128 camera pixels (256 arcsec). These stripe artifacts are usually of quite low surface brightness and so did not trigger many spurious detections. However, they can contaminate the measurements of sources falling on or near them. Therefore, such sources are flagged with cc_flg="s" in the PSC. The algorithm used to identify sources affected by the stripes is described in Section IV.7.iv.



• Confusion During Bandmerging

Source detection and characterization is conducted separately in the J, H and K_s bands during the processing of each Survey scan. The detections from each band are then positionally "merged" into single source records containing the cross-band information (cf. IV.4.e). Because the band-to-band instrumental position offsets for the 2MASS cameras are stable and well-characterized, the bandmerging process is usually unambiguous. However, in regions of very high source density, or in instances were a source is split in one band and not others (i.e., for close double stars), then the source is flagged with cc_flg="b" indicating possible confusion in the bandmerging process. Sources so indicated are usually in confused environments, and can have aberrant colors because of an incorrect matching of flux levels between bands.

The possible values of the cc_flg, the conditions they indicate, and the number of sources in each band having those cc_flg values in the PSC are summarized in Table 4 below. Sources that are affected by more than one of the contamination conditions have cc_flg set in the priority shown by the order of the values in the table. For example, a source that is identified as both contaminated by proximity to a latent image and a diffraction spike from a nearby bright star will have cc_flg="p" in the affected bands.

Table 4 - Definition of cc_flg Values and Occurrences in the PSC

cc_flg Value (1 per band)	Nature of Contamination or Confusion	N(J)	N(H)	N(Ks)
0	Source is unaffected by artifacts, or source not detected in that band	375059297	389911797	410775947
p	Persistence (Latent) Image from Bright Star	2438794	3919008	1550140
c	Confusion with nearby brighter source	90679500	73584699	54196294
d	Nearby Diffraction Spike	133283	563982	707105
s	Horizontal "Stripe" due to Bright Star	2667409	2995595	3740717
b	Confusion in Bandmerging	14687	17889	22767

bl_flg - Blended Sources

The blend flag (bl_flg) included with each point source record indicates the number of detections or source components measured simultaneously during the photometric measurement process. For aperture photometry (rd_flg="1","4") and 1-d radial profile fitting of saturated stars (rd_flg="3"), the bl_flg value is always "1", since these measurements did not attempt to deal with blended sources in any way. For the profile-fitting photometry of non-saturated Read_2-Read_1 sources (rd_flg="2"), if two or more detections fell within approximately 5´´--6´´, multiple PSFs were fit simultaneously to the ensemble of detections to compensate for the overlapping source profiles.

Sources that have bl_flg values >1 are in confused environments, and have poorer photometric accuracy than do isolated sources of similar brightness (bl_flg="1"). Most deblended sources will also be flagged with cc_flg="c" in the appropriate band, indicating likely photometric degradation due to confusion. To select sources with the highest photometric accuracy, users should select those with bl_flg1 in all three bands.

Source deblending during the profile-fitting photometry was triggered only if multiple detections with small separations were passed to the photometry routines. Close multiple sources that triggered only one detection were not split into multiple sources, even if the ² goodness-of-fit to a single profile was large (IV.4.b). Thus, ² >> 1.0 for rd_flg="2", sources may indicate that a source is an unresolved multiple (N.B.: Large ² values are also characteristics of single, extended sources and unreliable detections of image artifacts, hot pixels and meteor trails).

Table 5 contains the possible values bl_flg for sources in the PSC, and the number of sources in each band that have each value. The vast majority of PSC sources are fit with single components (bl_flg="1"). The maximum number of components fit to any source in the PSC is 7.

Table 5 - Definition of bl_flg Values and Occurrences in the PSC

bl_flg Value (1 per band)	Interpretation	N(J)	N(H)	N(Ks)
0	Source not detected in that band (rd_flg="0")	21804425	42627215	108180330
1	Single profile fit to isolated source (rd_flg="2"), or detection measured with aperture photometry (rd_flg="1","4") or 1-d radial profile fitting (rd_flg="3")	423830534	406502420	344169247
2	Two sources detected in small region and fit simultaneously during profile-fit photometry (rd_flg="2").	24537688	21024130	17963320
3	Three sources detected in small region and fit simultaneously during profile-fit photometry (rd_flg="2").	795238	806744	656531
4	Four sources detected in small region and fit simultaneously during profile-fit photometry (rd_flg="2").	24404	31493	22931
5	Five sources detected in small region and fit simultaneously during profile-fit photometry (rd_flg="2").	658	945	603
6	Six sources detected in small region and fit simultaneously during profile-fit photometry (rd_flg="2").	18	23	8
7	Seven sources detected in small region and fit simultaneously during profile-fit photometry (rd_flg="2").	5	0	0

Figure 8 shows an example of a deblended close triple source in the PSC. Some of the attributes of each source are listed in Table 6. The central, brightest component of the triple, 2MASS J22061452+5345303, is blended with both other components of the triple, so was fit with three PSFs simultaneously (bl_flg="333"). The other two components, 2MASS J22061426+5345345 and 2MASS J22061487+5345253, were each blended with 2MASS J22061452+5345303, so they were each fit simultaneously with two PSFs (bl_flg="222").

Table 6 - Source attributes for the resolved triple shown in Figure 8

|      designation|   j_m|j_cmsig|   h_m|h_cmsig|   k_m|k_cmsig|ph_qual|rd_flg|bl_flg| prox|
|             char|double| double|double| double|double| double|   char|  char|  char|doubl|
|                 |   mag|    mag|   mag|    mag|   mag|    mag|       |      |      |arcse|
 22061426+5345345  13.836   0.027 13.595   0.046 13.545   0.052 AAA     222    222      4.7
 22061452+5345303  13.320   0.043 12.591   0.051 12.426   0.041 AAA     222    333      4.7
 22061487+5345253  14.456   0.042 13.807   0.051 13.721   0.065 AAA     222    222      6.0

Multiple sources with separations close to the resolution limit of 2MASS may not be resolved consistently across all detected bands. In such cases, the deblended magnitudes will be listed for each component in the resolved band(s). In the unresolved bands, the default magnitude will be the 95% confidence brightness upper limits measured in a 4´´ aperture on the Atlas Image at the location of the source in the resolved band(s), and rd_flg="6". An example of this is shown in Figure 9, which shows the close source pair 2MASS J14352630-2615587 and 2MASS J14352601-2615582 (sep=4.0´´). A subset of the source information for these stars is shown in Table 7. Both components of this double were detected in the K_s band, and fit simultaneously during the profile-fitting procedure. The double was not resolved in the J and H bands, even though the image is clearly extended in those bands. Thus, the source listings for these two stars show deblended profile-fit measurements (rd_flg="2" and bl_flg="2") in the K_s band, and inconsistent deblend upper limits (rd_flg="6" and bl_flg="0") in the J and H bands. The upper limits given in the J and H bands capture much of the flux from two components, though, so contain useful photometric information. Inconsistent deblend upper limits (rd_flg="6") should not be confused with non-detection upper limits (rd_flg="0").

Table 7 - Source attributes for the partially resolved double shown in Figure 9

|      designation|   j_m|j_cmsig|   h_m|h_cmsig|   k_m|k_cmsig|ph_qual|rd_flg|bl_flg| prox|
|             char|double| double|double| double|double| double|   char|  char|  char|doubl|
|                 |   mag|    mag|   mag|    mag|   mag|    mag|       |      |      |arcse|
 14352630-2615587  12.346    null 11.984    null 13.131   0.041 UUA     662    002      4.0
 14352601-2615582  11.717    null 11.406    null 11.328   0.035 UUA     662    002      4.0

Figure 8	Figure 9

v. Photometric Uncertainties

Three measures of photometric uncertainty are provided for each 2MASS PSC "default magnitude". The first, [jhk]_cmsig, are the measurement errors returned by the default photometry algorithm in J, H and K_s. For profile-fit photometry (rd_flg="2"), the measurement uncertainties have been corrected (see IV.4b) to be consistent with the statistical variation of repeated measurements. For aperture photometry and 1-d radial profile-fitting of the saturated R_1 sources, the uncertainties are the RMS variation of the measured flux from all frames on which a source falls (typically six). In the latter case, if there are less than two frames available for the aperture measurements, [jhk]_cmsig becomes a flag, and has a value >8.0 mag in the appropriate band. The [jhk]_cmsig values are appropriate when comparing photometry of sources over small spatial scales (i.e., within a 2MASS Tile).

The second photometric uncertainty quoted in the PSC is [jhk]_msigcom, which is the root-sum-square (RSS) combination of [jhk]_cmsig with the uncertainty in the nightly photometric calibration solution, the estimated flat-fielding residual (0.005 mag), and, for bright rd_flg="1" sources, the R_1 photometric normalization uncertainty (0.012 mag). The [jhk]_msigcom values provide a better estimate of total photometric errors, useful when comparing photometry of sources across large spatial regions.

Since error estimation is a statistical process, faint sources will occasionally have unphysically low measurement errors. Bright sources can also have unusually large uncertainties if the photometry process is contaminated by confusion with a nearby source, masked pixels, or cosmic rays. For these reasons, measurements in each band are also supplied with the scan signal to noise ratio ([jhk]_snr) which are the ratios of the default flux in each band to the characteristic point source measurement noise within each Survey Tile.

vi. Sources of Unreliability

Data processing algorithms, quality assurance, and criteria for generating the 2MASS Catalogs are all designed to produce highly reliable lists of detections of solar system and astronomical objects. By reliable, here, we mean that if an observer conducts a follow-up observation at the position listed for a 2MASS source, then a source of near-infrared radiation will be found at that location. This is modulated by proper motion of the source (which can be significant for solar system objects), and that the accuracy of the reported flux will be limited by the brightness and potential confusion of that source. Even at the strict <0.05% unreliability limits stipulated by the Level 1 Requirements, though, the PSC is so large that there may be >170,000 unreliable sources extant in the high-reliability Catalog subset, and considerably more in the full PSC.

In this section, we describe the known causes of unreliable sources in the PSC, and provide tips on how to identify such "detections" when using the catalog. Confirmed unreliable sources in the PSC will be maintained in an Anomaly List (see II.6.g) that will be updated periodically.

In general, unreliable sources exhibit one or more of the characteristics described in Table 8. Users should exercise some caution when encountering such sources. However, possession of any one or more of these attributes does not necessarily imply that a detection is unreliable, only that it merits examination.

Table 8 - Common Characteristics of Unreliable PSC Sources

Condition	Source Attribute
Single band detection	rd_flg="0" in two bands
Contamination or confusion flagging	cc_flg != "000"
Large value of profile-fit photometry 2 value	rd_flg="2" and [jhk]_psfchi >> 1
Low frame-detection count for SNR>7 sources	ndet[1,3,5] 1 and [jhk]_snr >7
Read_1 detection that is fainter than the Read_2 saturation limits	(rd_flg[1]="1" and j_m>9) or (rd_flg[2]="1" and h_m>8.5) or (rd_flg[3]="1" and k_m>8)
Source is fragment of an extended source	gal_contam="2"
Source is associated with a solar system object	mp_flg="1"

Unflagged Artifacts from Bright Stars

Scattered-light halos, diffraction spikes, latent images, and other optical ghosts associated with bright stars can trigger spurious source detections and contaminate the photometry of real sources that lie close to them. These spurious detections and contaminated sources are identified and flagged (see IV.7), using geometric algorithms based on the position and brightness of the "parent" bright star. The spurious detections are filtered out of the PSC during the Catalog Generation process (see V.3). The artifact identification process is not perfect, however, so a number of bright star artifacts are present in the PSC. Some of these will be flagged as real sources contaminated by artifacts (cc_flg="p","c","d","s") in one or more bands, and others may not.

The residual artifacts occur for several reasons:

• The geometric artifact search algorithms are based on statistical averages of large amounts of survey data and may not be optimum for each individual star.
• In the vicinity of very bright stars there can be large numbers of spurious detections triggered on the wings of the star image on both the R_1 and R_2-R_1 frames. The resulting confusion in trying to link the R_1 and R_2-R_1 detections and in discriminating the artifact detections from the true source detections occasionally allowed a R_1 artifact detection to pass through.
• The brightest stars can cast diffraction spikes and scattered-light halos across scan boundaries, into adjacent scans that may have been observed at very different times. Since many of the brightest near-infrared stars are variables, the artifact search parameters used may be based on the incorrect brightness of the parent at the time of the adjacent scan observation.
• A small number of very bright stars were not cleanly detected, so were not used in the list of "parent" stars during the artifact detection process (cf. subsection x below).

An analysis of residual bright star artifacts in the PSC has been made, and most are found to be single-band and occasionally two-band sources with SNR < 10 (i.e., ph_qual !="A"), and profile-fit photometry values ² > 2 (i.e., rd_flg="2" and [jhk]_psfchi >2).

A special class of bright star confusion "artifact" that appears in the PSC is the "faint" rd_flg="1" detection. "Faint", here, means that these sources are fainter than the nominal R_2-R_1 saturation level of J~9, H~8.5 and K_s~8 mag. These sources usually found in the wings of very bright stars, and can sometimes be detections of real objects. The real sources are usually detected in both the R_1 and R_2-R_1 exposures, but the R_2-R_1 detections are usually flagged as confusion artifacts and filtered out of the processing. However, because R_1 and R_2-R_1 artifact identification was handled slightly differently, the corresponding R_1 detections of the sources were not always filtered out. These "faint" R_1 unreliable sources are usually single band (rd_flg="1" in one band and rd_flg="0" in two bands) and about 80% of the occurrences are flagged as being affected by artifacts, having cc_flg="c","d" or "p" in the detected band.

Meteor Trails and Other Image Artifacts

Most meteor trails were identified and masked during construction of the 2MASS Atlas Images (see IV.3), so relatively few spurious detections on the trails are present in the PSC. However, masking of the trails could not be made 100% complete without compromising survey coverage, and a few trails have persisted in the images (see I.6.a). It is estimated that at most a few hundred such detections may have survived into the PSC. Such sources are usually distinguished by having a profile-fit photometry value ² > 2 in one or more bands (rd_flg="2" and [jhk]_psfchi >2), by being detected in only one of the 6 or 7 2MASS frames that cover each point on the sky (ndet[1,3,5]="1"), and by having no optical association within 5´´ (i.e., a is null).

Insects crawling (see Anomalies Gallery) across the camera windows during 2MASS scanning also produce image artifacts that trigger spurious point source detections. Most scans that were contaminated by insects were rejected during the quality assurance (QA) process and reobserved, but a small number were not caught during QA. Insect artifacts are usually diffuse blobs that are impossible to mistake for point sources, but that could be mistaken for extended sources on the Atlas Images. Spurious point source detections, due to insects, are believed to be quite rare in the PSC (i.e., < 100 occurrences).

Hot Pixels

A small number of pixels on each of the 2MASS NICMOS 3 detector arrays exhibited bi-stable behavior, which would toggle between relatively normal noise and responsivity and an intermittent high-signal, or "hot" state. Pixels read out when in the hot state often triggered spurious detections on a succession of frames in a scan. These detections were fixed on the array, but not on the sky. Thus, they produced what could be relatively bright single-frame, or "solo" detections. Most incidents of "hot pixel" events were eliminated during scan processing using two filters. First, an automated procedure built into the sky-bias correction step of the instrumental frame correction (see IV.2) masked pixels that exhibited large variances in the stack of 42 frames used to generate the bias frame. Second, a solo-blanking procedure (see IV.4) identified bright single-frame detections and filtered them out of the detection stream that was passed on to later stages of data processing.

Despite these protections, a number of "hot pixel" detections are known to have passed through these filters and into the extracted source database. A detailed report on the spurious hot pixel detections shows that their characteristics are:

• They are single-band detections (rd_flg="0" in two bands)
• They have profile-fit photometry values ² >2 (rd_flg="2" and [jhk]_psfchi>2)
• They can have SNR10--15
• They are detected on 1 individual frames, despite having SNR7 (ndet[1,3,5]1).
• They fall in narrow bands of cross-scan position (x_scan) corresponding to the positions of the hot pixels.

The most egregious hot pixel was one on the southern H-band array detector that produced ~84,000 spurious detections, with cross-scan positions in the range 153+x_scan+155, that would have otherwise been passed onto the PSC. Because ~70,000 of the artifacts from this one pixel would have satisfied the criteria of the high SNR, high reliability Catalog, more than 14 times as many produced by all of the other hot pixels combined, these particular hot pixel detections were removed from the PSC before the Release.

There may still be as many as 100,000 hot pixel detections remaining in the full PSC, but only ~2500 are in the high reliability subset of the PSC with ph_qual="A". Users should be cautious when dealing with single-band PSC sources with [jhk]_psfchi>2 and ndet[1,3,5]1. The Atlas or Quicklook Images of spurious detections caused by hot pixels do not show an source apparent at the source position.

Solar System Objects

PSC Sources associated with the predicted position of known asteroids, comets, planets and planetary satellites have the parameter mp_flg="1". While these are for the most part good detections and measurements of these objects, they do not meet the strict definition of source reliability because they will not be at the reported position in follow-up observations.

vii. Single-Band Sources

As summarized in II.2.c, the PSC contains approximately 9.2 million, 1.3 million and 960 thousand single-band J, H or K_s sources, respectively. While these sources are not rare, they should be regarded with some caution because single-band detections are the least reliable class of source in the catalog. Single-band sources have rd_flg="0" in two bands.

At high Galactic latitudes, real, faint J-only sources are not uncommon, because the 2MASS system is most sensitive in the J-band, and the intrinsic colors of most stars and galaxies are blue enough to make them preferentially detectable in J. Legitimate K_s-only sources are generally associated with large amounts of intrinsic or foreground dust extinction, and are usually found close to the Galactic Plane, in molecular clouds, and in the Magellanic Clouds. H-only detections are extremely difficult to produce with known astrophysical objects, and are probably the least reliable sources in the PSC. In principle, a bona fide H-only detection could be produced by a source with strong emission lines in the H-bandpass, e.g., a QSO or planetary nebula, but these are expected to be rare.

Very often single-band detection is a symptom of unreliability, due either to confusion between sources, spurious detections of image artifacts, or transient events, such as cosmic ray strikes or hot pixels. Users are strongly urged to scrutinize single-band sources; examining their images using the 2MASS All-Sky Image Server is an excellent way to discriminate between clean sources and spurious, or confused, detections.

viii. Photometric Biases

The photometry of high SNR/high reliability Catalog subset of the PSC is demonstrated to be uniform to the ~4% level in each band over the unconfused sky (cf. VI.1b). However, there are known to be systematic photometric biases at levels below this that affect the photometry of sources in the Catalog over a range of spatial scales. In addition, a much larger bias is known to exist in the photometry of stars that saturated the 51 ms "R_1" exposures (rd_flg="3") that produces a prominent discontinuity in the global PSC source counts near the R_1 saturation level. The known sources of bias in the PSC are summarized here.

Observatory Bias

2MASS photometric uniformity is enforced on the largest scales by the Survey's photometric calibration strategy (see III.2.d). Nightly photometric transformations are derived from hourly measurements of one of 35 calibration fields, 12 of which are located on the celestial equator and are observed from both observatories, often on the same night. The photometric system is tied between the 2MASS observatories by referencing all photometry, in part, to measurements of the same standard stars. The measured mean residual bias between stars observed from the two observatories on the same night is -0.003, -0.009 and -0.004 mag in J, H and K_s, respectively, with measurements from Mt. Hopkins being brighter than those from Cerro Tololo. Brighter stars show a bias up to ~0.005 mag larger than fainter stars. The bias also shows a weak variation around the sky, with the differences being largest in the range 0°<ra<120°.

Uncorrected Seeing Effects for rd_flg="3" Photometry

Brightness estimation for stars that saturated the 51 ms R_1 exposures is performed using a 1-d radial profile fit to the non-saturated portion of the azimuthally-averaged image profile on the R_1 exposures (IV.4.a.iii). In general, the statistical accuracy of these brightness estimates is ~10--20%, poorer than for non-saturated point sources. In addition, the photometry of these stars can exhibit a substantial bias relative to non-saturated sources because of uncorrected effects of seeing.

The 1-d radial profile fits were made to single analytical templates, one for each band and observatory, that were derived from the observations of a relatively small number of non-variable, bright stars with previously measured brightness. Photometry derived from fits made to stars observed under seeing conditions substantially different from the mean condition of the profile-calibration stars is systematically biased relative to the calibrators. The amplitude and sign of the bias is a systematic function of the seeing, and can be as large as ~0.3 mag. However, the bias is on average closer to ~0.1 mag in the sense that the quoted magnitudes are too faint. The seeing has been characterized to some degree. However, seeing corrections were not applied to the rd_flg="3" photometry in the PSC because they always produced stellar colors that showed larger post-correction dispersions.

The seeing calibration bias produces a substantial discontinuity in the PSC source counts in the vicinity of the 51 ms R_1 exposure saturation level because sources are redistributed into fainter flux bins. Figure 10 shows a section of the source count curve for all PSC J-band detections, along with the Tycho 2 V_T source counts for reference. The discontinuity seen near J~5 mag is not due to incompleteness of the PSC. Similar discontinuities can be seen in the H and K_s source count curves, as can be seen in Figure 11.

Figure 10	Figure 11

R_1 Exposure Photometry Normalization Uncertainty

The aperture photometry of sources made on the 51 ms R_1 exposures (rd_flg="1") was normalized to the curve-of-growth-corrected aperture photometry from the 1.3 s R_2-R_1 exposures. The normalization was made by adding a seeing-dependent constant to the R_1 exposure photometry. The seeing-indexed normalization constants were derived for each band at each observatory by computing the mean offsets between the R_1 and R_2-R_1 exposure aperture photometry using stars in the brightness range to be detected on the R_1 exposures and non-saturated on the R_2-R_1 exposures (IV.4.c.i). The standard deviation of the mean offsets in each seeing bin was typically 1--2%. This component of the uncertainty has been factored into the combined photometric uncertainties ([jhk]_msigcom) for rd_flg="1" sources.

The J-band PSC source count curve shown in Figure 10 shows a second, smaller discontinuity at J~9 mag, which is near the R_2-R_1 saturation level, and thus the transition between rd_flg="1" and rd_flg="2" sources. This discontinuity is better seen in Figure 12, which shows the J-band PSC source counts normalized to a power-law. This offset is consistent with a ~1% systematic bias between the R_1 and R_2-R_1 photometric scales. However, we cannot rule out that the PSF-fit normalization table error, described in the next subsection, does not also contribute to this discontinuity by biasing the rd_flg="2" source photometry calibration by ~1-2%.

Figure 12

PSF-Normalization Table Correction

Outdated corrections were used to normalize the profile-fit photometry to the curve-of-growth-corrected aperture photometry (IV.4.c) in high source density regions, where the empirical correction could not be reliably determined. This resulted in a systematic overestimate of source brightnesses, the amplitude of which is a function of seeing FWHM and the photometric band. The overestimation ranges from <0.01 mag in good seeing, to a maximum of 0.02--0.03 mag for the worst seeing in the Survey. Because the amplitude of the bias can be different in each band, this can produce a residual color-bias on ~15´ scales along a Tile, and ~8.5´ across Tile boundaries. This bias is believed to cause at least part of the discrete color jumps correlated with Atlas Image boundaries, seen in spatially-averaged color maps of high source density regions, such as those illustrated in Figure 13. However, these jumps can be as large as 5%, so they cannot be fully accounted for by the PSF-normalization errors. We do not have a complete understanding of the origin of these color jumps, but they are known to be correlated to changes in the PSF used in profile-fit photometry. The biases are not seen in colors derived from aperture photometry.

Figure 13

PSF-Seeing Mismatch

The assumed form of the PSF used for profile-fit photometry (rd_flg="2") was obtained from a table lookup that is indexed by the seeing shape parameter. The seeing shape parameter is related to image full-width at half-maximum (FWHM) via the relation:

FWHM(arcsec) = 3.135*shape-0.464.

The mean seeing shape was estimated at intervals of ~20 sec or longer, corresponding to the time for the telescope to scan over the length of approximately one Atlas Image (15´). Variations between the instantaneous seeing and the mean value can lead to mismatches between the adopted PSF and the true image shape. Such mismatches result in systematic errors in the profile-fit photometry that are coherent over scales equal to the seeing variation time. The maximum bias expected from this effect is <0.02 mag, and it is largest for high SNR sources where the photometric uncertainties are dominated by PSF-mismatch rather than photon statistics. The impact of PSF errors is discussed in IV.4b.

Cross-scan Photometric Bias

The profile-fit photometric algorithm (see IV.4b) is sensitive to small variations in the shape of the point-spread-function (PSF) across the focal plane. Because the 2MASS optical systems are slightly astigmatic, point source images become elongated towards the focal plane edges, and there can be a resulting bias of of the profile-fit photometry of sources close to scan edges relative to the centers of the scans. This cross-scan bias is corrected ex post facto during data processing by normalizing to the aperture photometric measurements of sources as a function of cross-scan position. Residual cross-scan biases of PSC sources are measured to be << 1%.

Proximity to Sources of Equal or Greater Brightness

Photometry of sources in close proximity to another source of equal or greater brightness is biased, due to contamination of the sky reference annulus by the light from the nearby star. The sense of the bias is that the source brightness is overestimated, and the degree of overestimation increases the closer the source falls to the nearby source. Sources that are believed to have fluxes overestimated by 5% are flagged with cc_flg="c" in the appropriate band(s) in the PSC (IV.7). This photometric confusion affects nearly 50% of sources in the highest source-density regions of the sky.

Stars Near the R_1 Saturation Level

Bright stars very near the R_1 saturation levels of J~4.5, H~4 and K_s~3.5 mag may exhibit a larger scatter in their photometry because they were measured on only one or two frames, and therefore suffer from the effects of intrapixel response variations. The point source response of the 2MASS detectors has been measured as a function of sub-pixel position. Response variations are small for the northern and southern J and K_s and original northern H arrays, but can be as large as 0.2 mag for the southern H-band and the new northern H-band arrays. Normally, the quoted brightness for stars in the unsaturated R_1 regime is the average of the flux measured on six or seven unsaturated frames. For stars just at the R_1 saturation level, though, the PSC magnitude is derived from measurements only on those frames that are unsaturated, and can be taken from as few as one unsaturated frame (stars saturated on all available frames have brightnesses estimated differently). The number of frames used to derive the default magnitudes for non-saturated PSC sources (rd_flg="1", "2" or "4") is given in the first, third or fifth fields of the ndet parameter for each source. Photometry for sources with rd_flg="1" in J, H or K_s and that have ndet[1,3,5]<3 is of poorer quality. Such sources have degraded values of the photometric quality flag, ph_qual="E".

Compromised J-band Photometry for Sources with dist_edge_ew=251E and hemis='s'

Approximately one million stars in the All-Sky Relase PSC have been found to have erroneous J-band photometry. The affected stars fell on the central column of the J-band array of the Southern hemisphere 2MASS system. Stars that fell on this column have reported J-band fluxes that are approximately 10% lower than their actual fluxes and thus 2MASS J-H and J-K colors that are too red. Queries of the 2MASS All-Sky Release PSC, Survey Point Source Reject Table, and the 6x and Calibration Point Source Working Databases can avoid/identify these stars by including the following condition

hemis='s' and dist_edge_ew=251 and (dist_edge_flg='se' or dist_edge_flg='ne').

Examples of this bias and a discussion of its potential impact are given in VI.9.

ix. General Effects of Confusion

Source Detection Thresholds in High Source Density Regions

Faint source detection is done by identifying intensity maxima in the 2MASS Atlas Images that are greater than 3 times the point-source-filtered noise levels (see IV.4a for details). The noise estimator is sensitive to confusion noise, and thus increases in areas of increasing source density, such as the Galactic Plane. Thus, the effective depth of the PSC will decrease in regions of high source density, as is illustrated in Figures 2--4 and 5--7. The variation of completeness of the PSC as a function of source density is discussed in Section VI.7.

Artifact Identification Efficiency

In areas of high source density, it is very likely that an image artifact from a bright source will fall on top of one or more real sources. The process of artifact identification (see IV.7) results in many real sources being filtered out of the Catalogs, or at least flagged as being affected by artifacts. This filtering produces outlines of missing sources that trace the "mask" around bright sources defined by the geometric parameters used for artifact identification. This is especially apparent in the cores of globular clusters and near the Galactic Center.

Blended Sources

Close double or multiple sources are not reliably resolved by 2MASS if their angular separation is <6´´. The effective resolution of the 2MASS system is approximately 5-6´´, and is governed by the camera pixel sizes, image dithering efficiency, atmospheric seeing and source extraction algorithms. Figure 14 shows the distribution of separations for all source pairs in the PSC out to 16´´ separation. The resolution limit is reflected in the rapid roll-over for separations <5´´. There are source pairs with apparent separations well below this limit. These are invariably associated with blended sources or sources in complex environments, where the detections have been split into different Catalog entries in different bands. Users should treat any source with a prox value <5--6´´ with some caution.

Figure 14

Bandmerging Confusion

Close multiple sources may trigger numerous detections in a small area that differ from band-to-band, leading to potential confusion in the bandmerging process (IV.4.e). This can lead to the incorrect matching of detections between bands, resulting in unusual source colors. Sources affected by such confusion have cc_flg="b" in the appropriate bands. Photometry in such cases is very likely to be contaminated due to confusion.

Multiple sources with separations near the resolution limit of 2MASS may be resolved in one or two bands, but not necessarily in all detected bands. In such cases of inconsistent resolution or deblending, the sources involved have the measured point source magnitude reported in bands that are resolved, but have 95% confidence upper limits quoted in the bands that are not resolved. The upper limits are derived from aperture measurements made on the Atlas Images, so generally capture most of the flux of the blended sources in the unresolved band(s). Such sources have rd_flg="6" in the unresolved band(s), and usually have bl_flg > 1 in the resolved band(s).

Photometric Confusion

As discussed in above, sources close to objects of equal or greater brightness will have quoted fluxes that are systematically too bright, because of contamination in the sky reference annulus used for photometry. This photometric confusion is produced by stars of all brightnesses, so in regions of high source density, a significant fraction (~50%) of all sources can have biased photometry. Sources with photometry that is believed to be overestimated by >5% are flagged with cc_flg="c" in the affected band(s).

x. Very Bright Stars

Stars brighter than approximately J~4.5, H~4 and K_s~3.5 mag saturate the 51 ms exposures. A new feature of the 2MASS All-Sky Release PSC is that brightness estimates and accurate positions derived directly from 2MASS data are now provided for all bright stars, up to the brightest near-IR source on the sky, alpha Orionis (Betelgeuse = 2MASS J05551028+0724255), which has J=-2.99, H=-4.01 and K_s=-4.38 mag. Thus, the 2MASS PSC has an effective dynamic range in brightness of nearly 20 magnitudes.

Brightness Estimation Limitations

Photometry for saturated R_1 stars is performed using a 1-d radial profile fit to the non-saturated portion of the azimuthally-averaged image profile on the R_1 exposures (IV.4a). The accuracy of this photometry is not as good as for the non-saturated stars, and the photometric uncertainties listed for these objects (rd_flg="3") are generally >0.2 mag. Users should respect these uncertainties.

In addition to having large random uncertainties, the photometry of saturated R_1 stars can also exhibit biases of up to 0.1--0.2 mag, because of uncorrected seeing-dependencies in the 1-d profile-fitting (IV.4a). The maximum effect of this bias can be ~0.3 mag, but is much smaller for most stars. The current best estimate of the seeing correction is discussed in IV.4a.

The uncorrected seeing bias in the saturated R_1 photometry produces small discontinuities in the PSC source counts, as can be seen in Figure 10, near J=4.5 mag. The extracted photometry for saturated stars is on average too faint, resulting in a redistribution of counts into slightly fainter bins. This produces an excess of sources on the faint side, and a deficit on the bright side.

Confusion and Missing Detections

Because of the large extent of their source profiles, the effective confusion radii of very bright stars are larger than for non-saturated stars. Thus, very bright double and multiple stars become increasing difficult to resolve and measure. Although many saturated doubles are separated down to ~7´´, a few very bright doubles of larger separation are known to not have been resolved. The most prominent example is Centauri AB, with components separated by 17.7´´. A small number of other very bright stars are missing measurements in one or two bands, or are missing entirely, due to processing problems.

xi. Extremely faint sources

Sources were required to have a SNR7 in at least one detected band, or to have SNR5 and be detected in three bands to be included in the PSC. In the former case, fluxes in the remaining two bands may lie well below the SNR7 threshold, may be reported as detections (rd_flg="2" or"4") entirely dominated by noise, or may be reported as upper limits (rd_flg="0" or "6"). The magnitudes in these remaining bands can be unphysically faint ([jhk]_snr < 3), or will have associated uncertainties that imply a <3 detection ([jhk]_cmsig>0.36).

xii. Extended Sources and Galaxy Contamination

The PSC contains entries and point source-processed flux measurements for virtually all extended sources listed in the XSC. Because these objects are not point-like, profile-fit photometry and curve-of-growth-corrected aperture photometry will not provide accurate measurements of their brightness.

Extended sources in the PSC have non-null values of the ext_key (e.g., ext_key is not null), which is the cross-reference to the source in the XSC. PSC sources that correspond identically to XSC sources with semi-major axes >10´´ in size are also flagged in the gal_contam column, with gal_contam="1". The ext_key field is a more reliable indicator of source extension, since it cross-references all extended sources and not just larger galaxies.

Point sources that are superimposed on background galaxies or nebulae may have contaminated photometry, because of the complex background. The 341,724 PSC entries that fall within the elliptical area defined by the r_k20fe semi-major axis + 10% and ellipticity for all extended sources with r_k20fe > 10´´ are denoted by gal_contam="2".

Thus, to avoid sources in the PSC that are resolved by 2MASS, or that may have photometry contaminated by an extended sources, users should select sources where ext_key is null and gal_contam=0.

xiii. Position Reconstruction

Position Uncertainties

The accuracy of position reconstruction and the position uncertainties quoted for PSC sources has been assessed by comparing 2MASS star positions with those in the Tycho 2 and UCACr10 astrometric reference Catalogs. A detailed summary of this analysis is presented here. The list below summarizes the accuracy of quoted position uncertainties in the PSC in the three different brightness regimes.

• Over most of the unsaturated R_2-R_1 exposure brightness range (rd_flg="2"), the positional accuracy is 70-80 mas with respect to these comparison catalogs, and the quoted positional uncertainties overestimate the errors by 10-20%.



• In the non-saturated R_1 exposure range (rd_flg="1"), the 2MASS astrometric accuracy is approximately 100 mas, and the quoted 2MASS uncertainties underestimate the observed mean offsets by ~30% on average. The position error models for the R_1 frame detections were calibrated using the dispersion of positions measured on the group of six frames which sampled each source. The measurements from each band were assumed to be independent when positions were combined during the bandmerging procedure. However, it is now known that the R_1 detection position errors are correlated between bands because the short 51 ms R_1 exposures effectively freeze the seeing. Thus, the position errors were typically underestimated by between a factor of 1.0 and sqrt(3).



• The positional accuracy of saturated R_1 sources (rd_flg="3") with respect to the comparison catalogs is 100--130 mas. The 2MASS position uncertainties typically overestimate the true uncertainties by approximately a factor of two.

"Read_1" vs. "Read_2" Detection Position Bias

Frame distortion maps were used in the position reconstruction process for detections on the 1.3 s R_2-R_1 exposures (rd_flg="2"), but not for the detections on the 51 ms R_1 exposures (rd_flg="1","3"). This discrepancy results in a small position bias between PSC sources in the two broad brightness regimes that varies with cross-scan position and between observatories. This bias is illustrated in Figure 15 which shows the the mean cross-scan (approximately RA -- upper panels) and in-scan (approximately DEC -- lower panels) position differences between 2MASS and UCACr10 stars, plotted as a function of cross-scan position (x_scan). Northern observatory results are shown in the left panels and southern in the right. The maximum bias occurs in cross-scan positions (RA), and ranges between ±50 mas in the data from both observatories. The residual in-scan bias are <10--20 mas. Because the sign of the bias cross-scan bias between rd_flg="2" and rd_flg="1","3" sources changes across the focal planes, the net bias (i.e., averaged over all cross-scan positions) is 15 mas.

Figure 15

Galactic Coordinates

The approximate galactic coordinates of each 2MASS source are supplied in the PSC (glon, glat). These coordinates are provided to a precision of only three decimal places (3.6´´), though. This is done because the galactic coordinate system is not the astrometric reference system of 2MASS, and there is currently an ambiguity in the appropriate way to convert J2000 ICRS coordinates to the galactic system. The PSC galactic coordinates were derived by precessing J2000.0 coordinates to B1950, then using the rotation transformations into the l_II,b_II system. This transformation method produces galactic coordinates that can differ by up to 0.4´´ from those, e.g., produced using the direct J2000-to-galactic transformations, proposed by Murray (1989, AsAp, 218, 325).

Because there is not currently an IAU-standard for the ICRS-to-galactic transformation, we provide approximate galactic coordinates to facilitate coarse searches only.

xiv. Tile Overlap Regions

Sources that lie in the overlapping regions between 2MASS Survey Tiles (see III.2b) may be detected more than once during Survey observations. To insure the uniformity of depth, the PSC was constructed by selecting only one apparition of multiply-detected (or "duplicate") sources, rather than averaging measurements. Because faint sources near the detection level of the measurements have more than one chance to be detected in the overlap regions, the effective sensitivity is slightly enhanced in those regions. The full PSC contains sources that were detected on only a subset of all possible Tiles regardless of their location in the overlap strip. However, the high-reliability Catalog subset of the PSC contains only those non-multiply-detected sources that satisfy a purely geometric selection that removes the sensitivity bias. The use_src and dup_src values in each source record provide a guide for understanding the process that selected the apparitions for the PSC and high-reliability Catalog (cf. V.4).

use_src and dup_src

The duplicate source resolution process for the PSC began by positionally correlating sources from all overlapping Tiles, using a match radius of 2´´. All sources that had matching counterparts in overlapping Tiles had the dup_src value set to 1. Sources without duplicate apparitions have dup_src=0. For each group of matched sources, the apparition that fell farthest from its respective Tile edge was selected for the PSC. The selected source was assigned use_src=1. Duplicate sources that were not closest to their Tile centers were assigned use_src=0. Sources not in Tile overlap regions have use_src=1 and dup_src=0.

Sources in the overlapping regions that were not detected on all possible Tiles were treated specially to insure uniformity of sensitivity in the high-reliability Catalog subset of the PSC. A source that was missing one or more possible apparitions was assigned use_src="1" only if it was closer to its respective Tile center than the apparitions in the overlapping Tiles would have been to their Tile centers, had they been detected. This selection is based solely on geometry and not on source brightness, so it minimizes the non-uniformity of sensitivity in the Tile overlaps. Non-matching sources in the overlap regions that are farther from their Tile centers than their "virtual-duplicate" counterparts are included in the PSC, but have use_src="0".

Sources were included in the PSC if they had use_src="1" OR (use_src="0" AND dup_src="0"). The requirements for a source to be part of the highly-reliable, uniform Catalog subset of the PSC is use_src="1".

Other Reasons That Sources May Not Be Detected in All Possible Tiles

In addition to faint sources toggling above and below the 2MASS detection thresholds, there are several other reasons a source may fail to be detected on all possible Tiles in which it was observed. Some, but not all of these are reason to treat sources with use_src=0 and dup_src=0 with some caution.

• Motion: If a source moved >2´´ between subsequent observations, it would not be matched in the duplicate source processing. Asteroids and comets can move this much even between Tiles observed consecutively on a night. However, because some adjacent Tiles on the sky were observed up to ~3 years apart, stars with moderate proper motions could move out of the matching window.



• Confusion: Multiple sources with small component separations may not be consistently resolved in adjacent Tiles if, for example, the seeing differed between the time each was observed. This could lead to situation were a close double is matched with a single, unresolved source in the overlapping Tile. Depending on the relative separation of the resolved and unresolved sources, one component of the resolved double may be matched to the unresolved source in an adjacent scan, and the other component is treated as an unmatched source that can have use_src set to 0 depending on its location. If the unresolved source is matched to both components in the other Tile, the duplicate processing is in a confused situation. In such cases, the value of dup_src is set to >1, the exact value depending on the degree of confusion.



• Spurious Detections Sources that are spurious detections of either image artifacts, meteor trails, or hot pixel events will usually not repeat between Tiles.



• Scan Direction/Confusion with Artifacts: A persistence artifact may have been superimposed on a source in one Tile but not in another because the direction of scanning (North or South) was different. The unaffected source will have dup_src=0 and use_src=0 but will otherwise be perfectly valid detection.

xv. Solar System Objects

No attempt was made to remove non-inertially fixed sources from the PSC. However, sources that are at the positions of known asteroids, comets, planets or planetary satellites at the time of the observations are flagged by having mp_flg="1". These are positional associations and are not necessarily identifications. Therefore, they remain in the PSC. Some fraction of the putative detections are not valid, but are chance superpositions of the predicted positions with background sources. This is particularly likely at low Galactic latitudes, where the density of background stars is large. See IV.9 for more details.

xvi. Optical Associations

All PSC sources were positionally correlated with the Tycho 2 and USNO-A2.0 optical catalogs. 2MASS sources that have optical catalog counterparts within ~5´´ of the PSC position have listed for convenience the optical catalog identifier (a="T" or "U"), blue and visual-or-red magnitudes, the 2MASS/optical position separation (dist_opt) and position angle (phi_opt) information listed in the PSC source record, for convenience. Matches with the Tycho 2 catalog take precedence over the USNO-A2.0 matches in the PSC listing.

It is stressed that these are positional associations, and not necessarily identifications. The positional accuracy of 2MASS and the optical catalogs is sufficiently high that inertial sources should have positions that match to within ~1´´. Optical associations with separations dist_opt >1´´ are either proper motion candidates, or possible cases of chance alignments with the IR sources.

The optical magnitudes listed for the optical associations in the PSC are derived from the Tycho 2 and USNO-A2.0 values. In the case of Tycho 2 associations, the optical magnitudes listed are Johnson B and V magnitudes and are derived from the Tycho blue (B_T) and visual (V_T) magnitudes using the transformations given by (Høg et al. 2000):

V = V_T - 0.090*(B_T - V_T)
B-V = 0.850*(B_T - V_T)

For USNO-A2.0 associations, the optical magnitudes are the photographic blue and red magnitudes taken explicitly from the USNO-A2.0 Catalog.

Incorrect "a" Values for Optical Associations with dist_opt5´´

A software error in setting the optical catalog identifier flag, a, resulted in an incorrect value of a="0" being assigned to optical associations that have dist_opt5´´. There are 361,663 sources in the full PSC with this error. These objects can be identified as having a="0" and non-null values in the other optical association fields: dist_opt, phi_opt, b_m_opt, and vr_m_opt. The optical catalog identifier, a, should not be used as the only test for optical counterparts in PSC sources.

Missed Optical Associations Near the Equatorial Poles

An error in the positional correlation procedure between 2MASS and USNO-A2.0 sources resulted in associations being systematically missed in the vicinity of the north and south equatorial poles. In Figure 16 is shown the fraction of bright, three-band 2MASS PSC sources without optical counterparts (nopt_mchs=0) plotted as a function of declination within 20° of the north equatorial pole. The fraction of PSC sources without reported counterparts increases rapidly and monotonically with proximity to the pole for declinations > 86°. Optical associations for PSC sources near the south equatorial pole exhibit identical behavior.

An independent correlation between 2MASS and USNO-A2.0 sources in the vicinity of the poles shows that most PSC sources do have USNO-A2.0 counterparts within 5´´. Furthermore, good agreement between 2MASS and USNO-A2.0 positions near the poles, and the analysis of 2MASS astrometric accuracy (e.g., VI.6a.ii) indicate that there is no evidence for systematic problem with 2MASS or USNO-A2.0 positions near the poles. Therefore, the apparent increase in fraction of sources without optical counterparts shown in Figure 16 is an artifact of the correlation processing.

For sources with |dec|>86°, the optical association information is not reliable. The absence of optical association data in the PSC record does not necessarily mean that there is not an optical source within 5´´ of the 2MASS position.

Figure 16

[Last Updated: 2009 September 7; by R. Cutri and M. Skrutskie]

---

Previous page. Next page.
Return to Explanatory Supplement TOC Page.