The CFHT W3/D3 ancillary program was designed to target the CFHT W3/D3 region in order to construct various samples of quasars that
are useful for testing quasar selection for SDSS-IV/eBOSS. The goals were to find as many quasars as possible to r~22 using a variety
of methods, while also testing typical quasar selection with force-photometered WISE, variability, CFHT-depth fluxes and SDSS_depth fluxes.
We applied 5 main selection methods in the W3/D3 field. The resulting targets are stored in the file ancil-QSO-eBOSS-W3-ADM-dr8.fits,
which also contains a bitmask (W3bitmask) that spcifies the selection method(s):
2^0: W3 color box selection
2^1: SDSS xdqsoz selection
2^2: SDSS-WISE selection
2^3: CFHT variability selection
2^4: PTF variability selection
The bitmask encodes the fact that many of the selection methods overlap. Over the entire file (which is larger than the eventual
footprint of targets that was tiled) there are 6743 total targets that split as follows:
TOTAL TARGETS> 6743 (~4495 over 10 deg<sup>2</sup>)
----TARGETS------
PROGRAM TOTAL UNIQUE
colorbox 3726 1556
xdqsoz 2728 532
wise 3975 1931
CFHTvar 262 150
PTFvar 341 54
OVERLAPPING TARGETS BETWEEN PROGRAMS...
colorbox xdqsoz wise CFHTvar PTFvar
colorbox 3726 1866 1700 90 249
xdqsoz 1866 2728 1735 91 250
wise 1700 1735 3975 102 252
CFHTvar 90 91 102 262 0
PTFvar 249 250 252 0 341
As well as merging targets across the individual programs, any targets that lay in the BOSS bright star mask
allsky_bright_star_mask_pix.ply were excised from the target lists.
Additional useful tags in the target file include KNOWNZ and KNOWNSOURCE, which are high-quality spectroscopic identifications for
some of the targets, mainly from Johan Comparat's CFHT/ELG ancillary program and DEEP2, but also other known sources from ADM's
BOSS known quasar file. The Comparat identifications only include objects with zwarning == 0 and the DEEP2 objects only include
zqualty >= 3, as documented here
The individual selections comprising the program were conducted as outlined below. Whenever it is stated that targets were "matched
to the DR8 sweeps", this means that they were matched to the file produced by DJS that was named "objs-eboss-w3-dr8.fits"...this file
contained all Primary objects in the W3 region from the DR8 sweeps. Matching to the DR8 sweeps both places targets on the DR8 astrometric
system and reduces them to objects that could realistically be targeted by eBOSS.
color box
---------
Objects for the color box were selected from the W3 coadded catalog available at the TeraPix CFHT website
(http://T07.terapix.fr/T07/Wide/W3/Big-Merged/W3_fusion_sm2.cat). The objects were restricted in CFHT magnitudes to g < 22.8. No cut was
made on CFHT morphology.
The following color cuts were conducted in CFHT ugr bands to excise the stellar locus:
x = u - g
y = g - r
w = where( y - 0.5*x lt -0.2 OR y + 0.7*x lt 0.6 )
The targets were then matched to the DR8 sweeps and restricted in SDSS bands to 17 < r < 22 and objc_type==6 (i.e. point source)
Color box targets were assigned a priority of 1, so that they did not collide with variability selected targets.
SDSS xdqsoz
-----------
SDSS xdqsoz classification was directly conducted on the DR8 sweeps, restricted in SDSSbands to 17 < r < 22 and objc_type==6 (i.e.
point source) The xdqsoz algorithm (Bovy, Myers et al. 2012) was then run on the targets and the xdqsoz probability *integrated across
all redshifts* was calculated. The final targets were set to be those sources with PQSO > 0.2. This will suffer from some contamination
at the high redshift end, but is a deliberate choice based on prior discussion on eBOSS targeting phonecons, in which some people
favor a "redshift free" selection with xdqsoz.
SDSS xdqsoz targets were assigned a priority of 1, so that they did not collide with variability selected targets.
SDSS-WISE
---------
The DR8 sweeps file ('objs-eboss-w3-dr8.fits') was force-photometered at the SDSS source positions with the Tractor to extract WISE
fluxes (in the file ' eboss-w3-v4-wise-dr9.fits').
Calling the SDSS objects "a" and the WISE objects "b" the following logic was then applied:
i=where(b.w1 NE 0 AND b.w2 NE 0) & a=a[i] & b=b[i] ; trim to measured objs
flux1 = (a.psfflux[1]+0.8*a.psfflux[2]+0.6*a.psfflux[3])/2.4
flux1model = (a.modelflux[1]+0.8*a.modelflux[2]+0.6*a.modelflux[3])/2.4
flux2 = (b.w1 + 0.5*b.w2)/1.5
mag_opt = 22.5 - 2.5*alog10(flux1>0.1)
mag_model = 22.5 - 2.5*alog10(flux1model>0.1)
mdiff = mag_opt - mag_model ; PSF - model mags
gicolor = -2.5*alog10(a.psfflux[1]) + 2.5*alog10(a.psfflux[3])
mag_wise = 22.5 - 2.5*alog10(flux2>0.1)
iselect = where( (flux2 * a.psfflux[1] GT flux1 * a.psfflux[3] * 10^(3.0/2.5)) $
AND mag_opt GT 17 AND mag_opt LT 22 $
AND (a.objc_type EQ 6 OR mdiff LT 0.10) AND gicolor LT 1.5,cnt)
This logic can be described as:
(1) A stacked flux was created in SDSS gri (weighted as g:0.8r:0.6i), for both MODEL and PSF fluxes
(2) A stacked flux was created in WISE W1 and W2 (weighted as W1:0.5W2)
(3) Only objects with stacked fluxes (in both WISE and SDSS) of more than 0.1 nanomaggies were retained as targets
(4) Only objects with W2 fluxes of more than 0.1 nanomaggies were retained as targets
(5) The targets were cut to those objects with an SDSS stacked flux (mag_opt) of 17 < r < 22
(6) The targets were cut to those objects with SDSS g - i < 1.5
(7) The targets were assigned a slightly relaxed morphological cut such that *either* objc_type == 6 *or*
the difference between PSF and MODEL magnitudes being < 0.1 are targeted
(8) The targets were cut to those objects with (mag_opt - m_wise) > (g-i) + 3.0 where mag_opt is the stacked optical
magnitude and m_wise is the stacked WISE magnitude
SDSS-WISE targets were assigned a priority of 1, so that they did not collide with variability selected targets.
-----------------------------------------------------------------------------------------------------------------------------------------
PRIOR TO THIS POINT, THIS DOCUMENT WAS WRITTEN BY Adam D. Myers, Wyoming. AFTER THIS POINT IT WAS WRITTEN BY Christophe Yeche, CEA/SACLAY
-----------------------------------------------------------------------------------------------------------------------------------------
CFHT variability
----------------
1) Field CFHT deep 3, 1 Sq. deg.
2) 3 years of observation for the light curves
3) 3 variables Chi2, A and Gamma (for the last two variables related to the structure function see in Palanque, 2011 paper the exact definitions)
4) The 3 variables are averaged over the 3 bands gri
5) Color c1 and c3 are defined as:
c1=0.95*(u_mag-g_mag) + 0.31*(g_mag-r_mag) + 0.11*(r_mag-i_mag) ;
c3=-0.39*(u_mag-g_mag) + 0.79*(g_mag-r_mag) + 0.47*(r_mag-i_mag) ;
Than we define two selections:
A) Pure CFHT selection (~1/3 of the targets):
A_gri>0.08&&Chi2_gri>10.0&&gamma_gri>0.3&&(c3<0.6 -.33*c1)&&g_mag<23.0;
B) CFHT + SDSS selection (~2/3 of the targets):
(A_gri>0.08&&Chi2_gri>10.0&&gamma_gri>0.2&&g_mag<22.8&&type==6
CFHT variability targets were set to the highest priority (0)
PTF variability
---------------
1) Field CFHT: 3-4 Sq. deg (Nathalie could you described the 2 PTF fields used).
2) 3-5 years of observation for the light curves
3) 3 variables Chi2, A and Gamma (for the last two variables related to the structure function see in Palanque, 2011 paper the exact definitions)
4) The 3 variables are just for R band linked to the r SDSS observation (with a color correction)
5) Color c1 and c3 are defined as:
c1=0.95*(u_mag-g_mag) + 0.31*(g_mag-r_mag) + 0.11*(r_mag-i_mag) ;
c3=-0.39*(u_mag-g_mag) + 0.79*(g_mag-r_mag) + 0.47*(r_mag-i_mag) ;
Than we define two selections:
A) Pure PTF selection (~1/3 of the targets):
chi2>10.0&&A>0.05&&gamma>0.1&&tmag_g<22.5&&mag_r>18.0&&c3<1.0-0.33*c1
B) CFHT + SDSS selection (~2/3 of the targets):
Chi2>10.0&&A>0.05&&gamma>0.1&&g_mag<22.5&&pQSO>0.1&&type==6)
Note from ADM: PTF variability targets were matched to the DR8 sweeps for astrometry, which discarded some sources (~5%)