II. User's Guide to the 2MASS All-Sky Data Release

2. Point Source Release

b. How to Use the 2MASS Point Source Release

  1. Achieved Performance
  2. Catalog Access
  3. Image Access
  4. Sample Catalog Queries
  5. Potential Hazards
  6. Obtaining Uniform Samples
  7. Graphical Query Results

The 2MASS All-Sky Point Source Release contains 470 million source extractions with a broad range of photometric quality. Various flags (rd_flg, cc_flg, bl_flg, ph_qual, prox, ext_key) permit users to distinguish between sources with reliable photometry and sources with suspect and/or poor photometry. The subsections below describe the overall nature and quality of the Point Source data and provide guidance for selecting reliable samples.

N.B.: The 2MASS All-Sky Point Source Release is deceptively named. "Point Source" refers to the fact that the fluxes and other parameters were extracted by 2MASS point source processing software. In the vast majority of cases, sources were indistinguishable from the system point-spread function (PSF) and the fluxes are valid point source flux estimates. In a small fraction of cases (about 1.6 million of the 470 million sources in the Release), the source profile was larger than the telescope PSF, indicating that the source was extended. Point source profiles fit to an extended profile produced an underestimate of the source flux. These same sources were detected and processed by the 2MASS extended source processor and form the basis of the 2MASS Extended Source Catalog. In the Point Source Release these sources have the ext_key flag set to the corresponding counter number in the Extended Source Catalog. Correct fluxes for extended sources can be found there. Queries seeking reliable point source photometry should exclude sources with non-null values of ext_key: (SQL: ext_key is null).

The admonition above is but one of several characteristics of the Release dataset which could produce misleading results for an uninformed user. The table below (as well as the discussion of the Point Source "caveats" (I.6.b) highlights several similar "traps" which could result in an embarrassing referee's report.

i. Achieved Performance

Dynamic Range >20 magnitudes (-4 through 16 mag at Ks)
10 photometric sensitivity
at J, H, Ks in unconfused regions
100% of coverage better than 15.9, 15.0, 14.3 mag
50% of coverage better than 16.4, 15.5, 14.8 mag
Most sensitive coverage at 16.8, 16.0, 15.3 mag
Photometric Precision [jhk]_cmsig < 0.03 mag for bright ([jhk]_m < 13.0 mag) stars
Photometric Uniformity < 2% maximum photometric bias
for any point on the celestial sphere
Astrometric Accuracy<100 mas absolute relative to the
Hipparcos reference frame for [jhk]_m < 13.5 mag
Completeness > 99% for ph_qual="A" sources (i.e. >10)
Reliability > 99.95% for ph_qual="A" sources

Dynamic Range

2MASS photometry has a dynamic range of >20 magnitudes, owing to extraction software that address three different exposure regimes. The Survey's nominal frame exposure time was 1.3 s, obtained by differencing two readouts of the NICMOS3 array separated by that time interval. Both readouts were independently recorded and saved for future processing. The first readout (often referred to as "Read_1" in subsequent sections) occurred 51 ms after array reset and represents an effective 51 ms exposure image. The second read ("Read_2") occurred 1.351 s after reset. The point source database includes a rd_flg flag, which distinguishes between different techniques for populating the "default magnitude" column for sources, depending on their level of saturation (all XSC photometry derives from the 1.3 s "Read_2-Read_1" difference). Since different techniques can be used in different wavelength bands for a given source, the rd_flg is a three-character flag with an independent value for each band, revealing the source of the default magnitude (i.e. j_m, h_m, or k_m) column in the Point Source Release database.

Photometric Sensitivity

Rigorous evaluation of repeated photometry, and thus limiting sensitivity, as a function of observing conditions (primarily seeing and airglow background), permitted assignment of limiting sensitivity for each survey Tile, in the absence of confusion (and re-observation of Tiles which fell below the Survey's requirements for photometric sensitivity). The histograms shown in Figures 1 and 2 document the achieved 10 sensitivity for every Tile in the entire sky in each band. Details of the sensitivity estimation and distribution of sensitivity on the sky are given in Section VI.2.

Figure 1Figure 2

Photometric Precision

Repeated observations of sources provided empirical uncertainties, as a function of magnitude, as a means of validating the statistical uncertainties assigned to individual source by the extraction algorithms. The quoted default uncertainties are consistent with the repeatability statistics. Figure 3 illustrates the locus of quoted default uncertainty ([jhk]_cmsig) for Ks-band sources across the entire dynamic range of 2MASS. At faint magnitudes (Ks>13 mag) uncertainties rise, due to the dominance of background noise. For 8.5 < Ks (mag) < 13 default uncertainties are consistently in the range 0.02-0.03 mag. This limiting profile-fit uncertainty is attributable to undersampling, due to the coarse pixel size of 2MASS (2.0´´), relative to the seeing-limited image size (typically 1.0´´). Stars with 4 < Ks (mag) < 8.5 saturated in the 1.3 s duration 2MASS primary exposure. They were measured in the 51 ms "Read_1" exposure and extracted using aperture photometry receiving rd_flg=1. As a result, their mean uncertainties (light blue points) are slightly smaller than the profile-fit extractions. Stars with Ks < 4 mag saturated even in the 51 ms exposure. Their fluxes were extracted with a procedure that fit to the scattered light wings of the stellar profile. Figure 3 shows the substantially larger uncertainty associated with this procedure. The actual magnitude boundaries between these regimes vary with seeing and background conditions. Users should consult the rd_flg to track the origin of photometry for a given source. Details of the photometric extraction process and assignment of uncertainties are found in Sections IV.4.b and IV.4.c.

Figure 3

Photometric Uniformity

Nightly/hourly calibration errors in 2MASS were estimated to be of order 0.01 mag (see IV.8). In addition, large scale drifts in the 2MASS photometric system were minimized (and evaluated) using a global photometric procedure, which calculated a simultaneous optimization of all available calibration measurements. Comparison of these solutions, obtained independently between hemispheres, shows <1% bias in the photometric calibrators around the sky (see the Photometric Uniformity analysis in VI.1).

Astrometric Accuracy

Internal comparison of repeated 2MASS observations, and external comparison with more precise astrometric catalogs have demonstrated that the limiting astrometric accuracy of 2MASS is 100 milliarcseconds (mas) relative to the Hipparcos/Tycho coordinate system. The Astrometric Accuracy analysis in VI.1 addresses these comparisons in detail.


Some of the 1° × 8.5´ 2MASS calibration fields were observed thousands of times during the course of the Survey. The recoverability of stars in these fields, as a function of magnitude, provides a definitive characterization of the Survey's completeness statistics. There are sufficient trials that these statistics can binned as a function of 10 limiting magnitude, providing a means of selecting uniform samples at specified completeness levels, as discussed in subsection II.2b.6 below.


Reliability has been assessed through

In both cases, reliability for sources which contain an "A" for the ph_qual flag (i.e., Catalog-level sources) exceeds 99.97% in all bands. A number of known sources of unreliability have been identified and excised from the Catalog (for example, bright star artifacts, hot pixels, and meteor trails). Less egregious examples of these phenomena survive, but contribute fewer than 0.03% of sources at the Catalog level. Despite this stringent reliability, a few hundred thousand unreliable sources lurk in the Catalog of nearly 0.5 billion objects. The Reliability analysis of VI.1 details the evaluation of reliability and known sources of unreliability.

ii. Accessing the 2MASS Point Source Release

The Infrared Processing and Analysis Center (IPAC) processed the 2MASS images and produced Catalogs of Point and Extended sources. These Catalogs, which include photometry, astrometry, and an extensive set of quality flags, are or will soon be publicly available from the following sources:

iii. Accessing the 2MASS J, H, and Ks Images

Discussion of 2MASS Point Source data would not be complete without highlighting the available Images which underlie the Catalogs. The Images are resampled to a common pixel scale and registered to a common center, to aid in the production of color images and mosaics. The images were generated by summing the six-deep, overlapping 1.3 s Survey exposures. Most rd_flg=1 and all rd_flg=3 sources will be saturated in these images, despite having useful flux extractions in the Catalog.

iv. Basic Database Queries

Before attempting work with 2MASS Point Source data, users should read the Point Source Cautionary Notes (I.6b) in detail and familiarize themselves with the explicit contents of the columns provided in the Release (II.2.a).

The GATOR website provides its own "user's guide" and tutorial, which provide specific examples of how to query the 2MASS Point Source database using the IRSA GATOR interface. In addition to a general user interface, the GATOR server permits queries based on Structured Query Language (SQL). Below are a few introductory examples. The SQL given in these examples can be inserted into the "User Defined Constraints" section on the GATOR query page (after removing the word "where," which is already assumed).

v. Potential Hazards

H-Ks colors toward the Galactic Center
PSF Aperture PSF-Aperture Delta(H-Ks) Histogram Empirical PSF correction
Figure 7 Figure 8 Figure 9 Figure 10 Figure 11

vi. Obtaining Uniform Samples over Large Areas

Uniformity in Unconfused Regions

The two figures below compare stellar density in two narrow J-band magnitude slices. Figure 12 is from a magnitude range (J = 15.8--16.0 mag), where the Survey is 99% complete for virtually all of the sky. Little, if any, imprint of the Survey scan pattern is evident. Figure 13 shows J-band source counts from a J-band magnitude range that is 0.8 mag fainter (J = 16.6--16.8 mag). In this faint magnitude range the stellar density is biased by the survey Tile pattern, because of the varying sensitivity of the Survey, driven primarily by airglow and seeing. Conversely, this figure indicates that large contiguous regions of the sky are complete to flux levels as much as a magnitude fainter than the nominal 2MASS completeness limits.

Figure 12Figure 13
15.8<j_m<16.0 16.6<j_m<16.8

The 10 sensitivity limit, computed for every individual Tile, based on seeing, background, and calibration zeropoint, provides a guide to selection of flux limits which provide uniform catalogs over steradian scales and selection of modest contiguous areas of the sky, which are complete at fainter magnitudes.

2MASS tracked sensitivity empirically as a function of observing conditions, by monitoring the statistics of repeated observations of sources under varying conditions of seeing and background (see VI.2). The repeated observations, in this case, were Survey calibration observations which occurred several times each night. Each calibration observation consisted of six independent observations of every star in a 1° × 8.5´ field. Assessing the flux uncertainty empirically from the standard deviation of the repeated measurements permitted an estimate of uncertainty vs. flux for the instantaneous seeing, airglow, and system zeropoint which prevailed at the time of observation. Collecting these observations from the range of conditions encountered during the three-year history of observations yielded a means of characterizing the sensitivity of any independently-observed Survey Tile using these key variables. In fact, many Survey Tiles were re-observed when the conditions indicated that the Tile's sensitivity fell below the Survey's sensitivity specification. Virtually the entire sky was observed at least once with sensitivity at or better than Survey specifications as a result. The 10 limiting sensitivity of each Survey scan is quantified in a column of the ancillary Scan Information Table.

The Survey's thousands of repeated calibration observations also provided a means of assessing the Survey's completeness as a function of magnitude, simply by tracking the number of times a given source was detected, compared with the number of opportunities for detection. Since calibration observations were obtained over a broad range of 10 limiting sensitivity, these empirical completeness statistics can be binned by sensitivity. Figure 14 shows that the empirical completeness percentage declines steeply faintward of the 10 sensitivity threshold and scales faintward as expected with limiting 10 magnitude. The table below provides guidance on setting a completeness limit vs. the 10 magnitude for any given Tile, based on the data in Figure 14. 99% or better completeness can be obtained for the entire unconfused sky by selecting sources with a limit 0.1 magnitude brighter than the poorest 10-limiting sensitivity in the histogram in Figure 1 -- or 15.9, 15.0, and 14.2 mag at J, H, and Ks, respectively.

Completeness threshold relative to 10 limiting magnitude
Achieved Completeness> 50%> 80%> 90%> 95%> 99%
Completeness threshold for
10 magnitude=M

Figure 14

In addition to appropriate flux limits, use of the "use_src" flag is important in obtaining truly uniform sampling of the sky. 2MASS Tiles overlap by a small amount. About 20% of the sky was observed two times or more as a result of this overlap. A selection algorithm has chosen a single apparition of each source with the potential for multiple observations, to remove any bias associated with multiple opportunities to observe a source (see V.4). If a source was observed only once in two opportunities, the algorithm would decide whether or not to include the source in the Catalog in an unbiased manner. Occasionally, bright sources would be missed on one of two apparitions, due to effects, such as one apparition falling on the persistence image of a nearby bright star. These bright sources are included in the Catalog for completeness, but should not be used when selecting uniform samples. The use_src flag permits this distinction. Sources selected with use_src=1 will be representative of a uniform 2MASS selection.

Uniformity in confused regions

The complete thresholds described below pertain to unconfused regions of the sky. In high-source density regions, such as globular clusters or the Galactic Plane, confusion noise from unresolved point sources will limit the point source completeness and photometric accuracy. This subsection presents an empirical determination of the uniformity of the 2MASS Point Source Catalog in high density regions.

A simple, yet powerful, visual representation of the 2MASS completeness limit over the entire sky can be obtained by viewing the "movies" in Figures 15, 16, and 17. These movies cycle through images of the differential star counts, in 0.2 mag bins, presented in the Galactic Aitoff projection. The Galactic Plane is clearly visible in these movies as the bright horizontal band. The star count maps show a patchy distribution in the Galactic Plane, especially at J, due to extinction from dust within the Galaxy. Beginning at approximately J=14, H=13.5, and Ks=13 mag, the star count maps start developing holes (as represented by the black regions), first near the Galactic center, and expanding to larger galactic latitudes for fainter magnitudes. These holes are a result of confusion noise.

GIF Movies of the 2MASS Sky
Figure 15Figure 16Figure 17

The above impressions where quantified by computing, for each 10 × 10 arcmin2 pixel in the star count maps, the magnitude bin which contains the peak star counts. The magnitude at the peak star counts was taken as a measure of the completeness. Note that this definition of completeness differs from that used to characterize the 2MASS Point Source Catalog, which was defined to be the faintest magnitude bin which recovers 99% of the expected source counts and was measured based on repeated observations of 2MASS calibration fields and overlap regions in the 2MASS survey. The definition of completeness used here is not as stringent.

Figures 18, 19, and 20 show images of the derived completeness limits in the Galactic Aitoff projection for the J, H, and Ks bands. Further, Figures 21, 22, and 23 show the average completeness limit, as a function of galactic longitude, for several latitude bins. Relative to the average completeness at high galactic latitude (|b|> 30°), incompleteness in the Galactic Plane occurs within approximately (a) 80° of the Galactic Center for |b| < 3°, and (b) 6° of the Galactic Center.

In addition to the Galactic Plane, the completeness limit will occur at bright magnitudes in high density regions, such as the Large Magellanic Cloud and globular clusters. For these regions, it is instructive to view the completeness limit as a function of source density. Figures 24, 25, and 26 show magnitude of the peak star count bin as a function of source density. The source density has been computed for sources brighter than magnitude=12, since Figures 18--23 indicate that nearly the entire sky is complete to that magnitude limit. The color scales in Figures 24, 25, and 26 are normalized such that peak value is unity. These figures show that at low stellar densities (i.e., high galactic latitudes), the magnitude histograms peak at approximately J=16.5, H=15.7, and Ks=15.2 mag. As the stellar density increases, the magnitude histograms systematically peak at brighter magnitudes.

Figure 18 Figure 19 Figure 20

Figure 21 Figure 22 Figure 23

Figure 24 Figure 25 Figure 26

vii. Example Query Results

The subsections that follow present the results of increasingly refined queries on the Point Source Release database, aimed to produce source fluxes of the greatest accuracy and astrophysical relevance.

Photometric Characteristics of the Complete Dataset

The color-color and color-magnitude diagram in Figure 27 result from a random sampling of 0.1% of the 2MASS Release database. Evident is a substantial smearing of the color-magnitude diagram, due to Galactic reddening. This reddening appears as the narrow spike extending toward the upper right in the color-color diagram (demonstrating the relative uniformity of the Galactic extinction law). Distinct features in the color-magnitude diagram (seen in the figures in the subsection below) are similarly smeared beyond recognition by reddening.

Figure 27

Photometric Characteristics of the Complete High-Latitude Dataset

SQL: where abs(glat)>35.

The effects of source density and reddening in the Galactic plane are discussed in detail in Section VI.7. Since the present subsection focuses on producing the highest quality photometry, subsequent examples address only high galactic latitudes, where reddening does not significantly bias stellar colors. The diagrams in Figure 28 show the results of selecting 2MASS sources at galactic latitude of 35° or more from the Galactic plane. The effects of reddening are no longer evident, and the distribution of stars in the diagram reflects true astrophysical colors of various stellar classes. This diagram, however, derives from all sources above the galactic latitude restriction, without regard for noisy detections (ph_qual), confusion (cc_flg), and upper limits (rd_flg). The colors of some of the sources are thus either contaminated or simply nonsense.

Figure 28

High-Latitude Sources with Valid Three-Band Detections

SQL: where abs(glat)>35. and ph_qual not matches '[EFUX]'

The diagonal features at faint magnitudes in the previous example arise from the use of upper limits calculated at the positions of non-detections. Sources must be detected in all three bands (and thus have a rd_flg value of 1, 2 or 3, in each rd_flg position) for valid color-color analysis. In addition, they must have valid photometric quality. All positions in the ph_qual flag must contain an "A", "B", "C", or "D". This condition is obtained by selecting against the other possible values of ph_qual that indicate non-existant or corrupted photometry ("E", "F", "U", and "X") The ph_qual restriction naturally delivers sources with valid rd_flg values. Figure 29, which includes only sources with valid three-band detections, is dominated entirely by astrophysical features convolved with the Survey's noise characteristics. The vertical spikes in the color-magnitude diagram are common dwarfs (left) and giants (right) smeared in apparent magnitude by their varying distance. The spikes are "sharp" at the bright end and broaden toward faint magnitudes, due to increasing flux uncertainty. In the color-color diagram the sources largely cluster along the coincident giant and main-sequence track for stars of earlier spectral type than K5. The "hook" at the red end of this distribution arises from K7/M0 main-sequence stars and cooler, which now deviate from the nearly linear trend (giants continue along the linear sequence as shown in Figure 30). The sources redder than J-Ks=1.0 mag at faint magnitudes are largely unresolved galaxies.

Figure 29Figure 30

Exclusion of Photometrically-Contaminated Sources and Known Galaxies

SQL: where abs(glat)>35. and ph_qual not matches '[EFUX]' and cc_flg='000' and gal_contam=0 and ext_key is null

Contaminated photometry and extended sources must be eliminated in order to obtain the most photometrically pristine sample of stars. Point sources contaminated by nearby extended sources have gal_contam set to a non-zero value. The cc_flg is set to a value other than "0" in a given band when diffraction spikes ("d"), artifacts ("p"), or simply scattered light ("c") from an adjacent star potentially contaminate a source's flux. The value cc_flg="c" is set to flag sources whose flux is likely to be in error by 5% or more, due to flux contamination from a nearby star. (Alternatively, additionally requiring prox > 8.0, in arcsec, provides more conservative protection from contamination from close faint stars at the expense of eliminating significant portions of the sources in high density regions.) Figure 31 contains sources which have cc_flg='000' and gal_contam=0. Sources included in this plot also have "null" values in their ext_key column, indicating that they were not identified as an extended object by the extended source processing. Plenty of galaxies remain in the Point Source Release after this exclusion -- all unresolved. These unresolved galaxies form the faint red haze beyond the red end of the main sequence in the color-color diagram and the extended red pedestal at faint magnitudes in the color-magnitude diagram. Comparison with Figure 29 shows that the removed extended sources populated the brighter portion of the red extragalactic distribution (see Figure 32).

Figure 33 shows the result of an "inverse" query intended to select potentially contaminated objects. This diagram contains sources with prox < 8 and with non-zero values in the cc_flg. By and large, this "contaminated" sample resembles the "clean" sample in Figure 31, in part reflecting the conservative nature of the confusion flagging and in part indicating that flux contamination may have lesser influence on stellar colors. Given that a source is flagged as photometrically contaminated when the flux is likely biased by 0.05 mag, and given that the vertical scale in the color-magnitude diagrams spans more than 10 magnitudes, the fact that there is little obvious bias in the "contaminated" sample arises mainly from the small contamination needed to trigger the flag.

Figure 34 similarly shows sources that received gal_contam=2 indicating they were contaminated by a nearby extended source. Surprisingly, the diagram seems to contain obvious red giant branches and red clumps. This result occurs because globular clusters are identified as extended by the extended source processor, and many of their stars lie within the 20 mag arcsec-2 isophote used to set the gal_contam flag. In addition to these features, there is a significant haze of points in the red regions populated by galaxies, indicative of true extragalactic contamination.

Figure 31
cc_flg='000' and
Figure 32
"blink" vs.
Figure 4
Figure 33
cc_flg!='000' or
Figure 34

Restriction to High Photometric Quality

SQL: where abs(glat)>35. and ph_qual not matches '[EFUX]' and cc_flg='000' and gal_contam=0 and ext_key is null and ph_qual="AAA"

Since detection above SNR>7 in any one band or SNR>5 in two bands is sufficient to include all three bands of source extractions in the Release database, many sources may have exceptionally large flux uncertainties in some bands. The previous figures included sources with relatively low signal to noise ratio and, thus, large photometric uncertainty. A variety of means exists to select samples of sources with small photometric uncertainty. A very simple approach involves using the "photometric quality" flag -- ph_qual. This three-character (one for each band) flag is set to "A" in a given band when the photometric uncertainty for the source extraction is <0.109 mag and the source flux to sky noise ratio exceeds 10. Selecting sources with ph_qual="AAA" ensures modest uncertainty in all three bands (Figure 35). Most apparent in these figures is the absence of the wide pedestals in the color-magnitude diagram. Since SNR=10 is typically achieved at Ks~14.5 mag, fainter sources do not survive the ph_qual="AAA" restriction.

Figure 35

Obtaining the Most Precise Profile-Fit Colors

SQL: where abs(glat)>35. and ph_qual not matches '[EFUX]' and cc_flg='000' and gal_contam=0 and ext_key is null and j_msigcom<0.05 and h_msigcom<0.05 and k_msigcom<0.05


SQL: where abs(glat)>35. and ph_qual not matches '[EFUX]' and cc_flg='000' and gal_contam=0 and ext_key is null and j_msigcom<0.03 and h_msigcom<0.03 and k_msigcom<0.03

The large (2´´) 2MASS camera pixels limit the Survey's ultimate precision to about 0.02 to 0.03 mag. This level of precision is regularly achieved for sources not limited by other intrinsic noise sources. Figures 36 and 37 illustrate the result when the final combined uncertainty ([jhk]_msigcom) constrains the sample selection. The samples do not extend to as faint a magnitude in the color-magnitude diagrams as in previous figures, due to the stricter effective requirement on signal to noise ratio. The color-color distribution is considerably sharper than for previous figures, with Figure 36 more distinct than Figure 37, due to the more stringent uncertainty selection.

Figure 36
Figure 37

Aperture Photometry

SQL: select j_m_stdap,h_m_stdap,k_m_stdap where abs(glat)>35. and ph_qual not matches '[EFUX]' and cc_flg='000' and gal_contam=0 and ext_key is null

Aperture photometry provides the ultimate photometric precision for bright sources ([jhk]_m_aper < 13.0), since it does not depend on the perfect matching of a source's spatial profile with a pre-determined PSF. At faint magnitudes ([jhk]_m > 13.0), however, the reduced spatial footprint of a PSF profile fit yields a superior flux relative to an aperture measurement (which necessarily incorporates a large background area). Figure 38 presents aperture photometry for bright stars. The color-color relationship is more sharply defined, compared with Figure 37. Figures 39 and 40 make direct comparisons between aperture and profile-fit results in two narrow magnitude ranges. For 11.0<Ks(mag)<12.0, aperture photometry is superior to profile-fit photometry. For 13.0<Ks(mag)<13.5, the reverse is true.

Bright stars are absent in the color-magnitude diagram of stars with [jhk]_m_stdap aperture magnitudes in the PSC database (Figure 38). This column was populated with an aperture magnitude only if the source had a valid profile-fit detection (and thus only if the star did not saturate the 1.3 s exposure). Aperture photometry was used exclusively to produce rd_flg=1 photometry extracted from the 51 ms "Read_1" exposure; however, this value becomes the default magnitude, [jhk]_m, for the source and does not appear in the ( [jhk]_m_stdap) column. The [jhk]_m_stdap column is null for rd_flg=1 detections.

Figure 38
Aperture Photometry
Figure 39
Blink Comparison with
profile-fit 11.0<Ks<12.0
Figure 39
Blink Comparison with
profile-fit 13.0<Ks<13.5

[Last Updated 2003 Oct 3; M. Skrutskie & J. Carpenter]

Return to Section II.2.