SDSS Bright Red Galaxy Selection |
This page documents the results of an initial exploration of the simulated data using the bright red galaxy algorithm as written down in the document we submitted to Steve Kent in 1995-1996, found here.
Discussion on sdss-clusters has made it clear that instead of Petroisan magnitudes or 3" aperture magnitudes we will use some form a best-fit model magnitudes to form colors. This is in effect a matched filter way of optimally finding the colors of faint objects.
The simulation test data have several problems. The galaxies were emplaced with magnitudes too bright by 2 mags, colors that are too red by 0.25 mags. The k-correction is too steep by a factor of 2. (See The Test Data.)
An exploration with a real field galaxy redshift survey suggests that our preliminary estimate of 10% of the main galaxy survey remains reasonable, and that 2.5 sigma cuts will provide the target 7.5 BRG/sq-degree. (See Real Data.)
Our document to Steve Kent has several parameter that are incorrect. A search of the literature has produced better numbers. (See The Algorithm Parameters.)
The current testing group includes: Neta Bahcall, Francisco Castander, Andy Connolly, Bob Nichols, Marc Postman, Michael Strauss, and Jim Annis. (See The Testing Group for what we are looking into.)
A coded version of the algorithm may be found in
/data/dp2.8/yanny/targets2/target_algs/redGals.tcl
It may be run by loging in to sdssdp2 and
unix> cd /data/dp2.8/yanny/targets2/ unix> setup astrotools unix> astrotools astls> source target_algs/redGal.tcl astls> set chain [fits2Schema tsCOBJC-000581-1.fit COBJC] astls> select_bigRed $chainIt returns a chain containing the selected bright red galaxies.
In the entire run of 100 sq-degrees we find 1150 BRG, for 11.5 per sq-degree. The population has the following characteristics:
median clipped st. dev. phot z 0.49 0.14 real z 0.35 0.05 M_r -26.13 0.87 Pet r 16.69 0.26
BCE
M_r = -22.3 Schneider et al., inside 26 kpc diameter aperture
= -22.9 corrected to "total" magnitude
sigma = 0.33 Postman and Lauer
g-r = 0.47 Schneider et al.
sigma = 0.055 Schneider et al.
The M_r=-22.3 refers to a 26 kpc diameter aperture. The conversion
to total or our Petrosian mag is roughly 0.6 mags.
The measured standard deviation of BCE mags is 0.33, admittedly inside
the 26 kpc aperture.
The aim is for 7.5 BRG per sq-degree. This translates to 10% of the total fibers allocated to galaxies.
The data are redshifts and restframe b-R colors (R CCD and b APM blue magnitudes). The data cover 248 square degrees down to a magnitude limit of R=17.7, with a sampling factor of 0.7, and totals 7471 galaxies.
We use the cut parameters corresponding to that described above: in the R band, BCE M_R=-22.5, the total mag is M_R = -23.1, and use Postman and Lauer's observed B-R dispersion of 0.055 mags. Further, we assume that the mean BCE color is that of the E/S0 ridgeline, observed here to be b-R=1.35. We set the BCE cuts at 3 sigma, giving M_r = -22.13, and b-R=1.25.
Using that cut on the Los Companos survey data gives 320 objects. This is 1.8/sq-degree at R=17.7, 4.6 at R=18.7, and 12 at R=19.7. (The extrapolation to fainter magnitudes assumes a number count slope of 0.4 log#/mag.) The R=19.7 number is close to our actual limits. The total number, 12. objects per sq-degree is fine: we expect to have 5-10 objects per sq-degree.
The cuts are shown on the following graph, where the x-axis is absolute R mag, and the y-axis is b-R in the restframe.
If we ask how sensitive we are to changing the cuts we find the following:
Culmulative
#/sq-degree
Bin total red %red R=17.7 R=18.7 R=19.7
< -22 286 179 62.59% 1.03 2.59 6.50
< -21.5 1265 665 52.57% 3.83 9.61 24.13
< -20 3134 1380 44.03% 7.95 19.95 50.08
< -20.5 5522 2232 40.42% 12.86 32.27 81.00
3-sigma cuts:
< -21.8 568 320 56.34% 1.84 4.63 11.61
2-sigma cuts
< -22.13 181 104 57.46% 0.60 1.50 3.77
Differential
#/sq-degree
Bin total red %red R=17.7 R=18.7 R=19.7
< -22 286 179 62.59% 1.03 2.59 6.50
-22.0->-21.5 979 486 49.64% 2.80 7.03 17.64
-21.5->-21. 0 1869 715 38.26% 4.12 10.34 25.95
-21.0->-20.5 2388 852 35.68% 4.91 12.32 30.92
We proposed a number of 100,000 BRGs, 10% of the main galaxy survey.
This works out to 7.5 per sq-degree. The Las Campanas data suggests we
will achieve those numbers with 2.5 sigma cut, in line with what we
expected.
The simulated data has 3 times more ellipticals than spirals!
The simulated data ellipticals have a mean g-r=1.00. We expect closer to g-r=0.75.
The simulated data has a large population of hyperluminous red galaxies at z=0.5. This is due to a magnitude conversion error between David Weinberg's catalog and Brian Yanny's simulation.
Of the 5810 unsaturated galaxies in a 2.0 sq-degree patch, 4114 are luminous enough, and 1883 are red enough. This is a factor of 125 more than our desired 7.5 per sq-degree.
A feature can be seen on the color magnitude diagram already shown above: there are not that many blue galaxies, and galaxies at r > 18 get red quite fast. I believe this is due to a k-correction that is much steeper than we expect.
The following plot shows the restframe color-absolute magnitude diagram based on the real redshift of each object. The objects meeting the selection critera (in the upper left) have a median redshift of z=0.47. This spray of objects are the simulated data's "hyperluminous galaxy" population.
A plot showing the photometric redshift vs. real redshift (both scaled by x10) shows the effectiveness of this scheme. The objects at 0 are stars. The line shows phot-z= real z. Excluding the stars, the mean difference photZ-realZ = 0.12, with standard deviation 0.08. The dispersion is not much more than that found in the calibration step, 0.03->0.06. The offset shows that the polynomial equation is sensitive: calibrated with Petrosian magnitudes, it fails systematically for fiber-mags scaled to Petrosian r.
We will need to calibrate the photometric redshift relation in the test year.
Bob Nichol is going to use spectro to test how faint we can go in an hour on brightest cluster ellipticals and still get redshifts.
Daniel Reichart and Bob Nichol are going to explore the Edinburgh-Durham galaxy catalog in conjunction with the Las Campanas survey to learn more about what cuts will give us 7.5 BRG/sq-degree. Further, they will use the Edinburgh-Durham cluster catalog and Rosat pointings to learn about what BRG selection means in terms of clusters found.
Francisco Castander is going to explore what synthetic spectra can tell us about photometric redshifts and k-corrections. One idea is to develop a photometric redshift equation designed for brightest cluster ellipticals.
I am pursuing the application of these ideas and code on the test data.