J/ApJS/275/19 High-z QSOs from DESI DR9 and machine-learning (Ye+, 2024)
Machine learning-based search of high-redshift quasars.
Ye G., Zhang H., Wu Q.
<Astrophys. J. Suppl. Ser., 275, 19 (2024)>
=2024ApJS..275...19Y 2024ApJS..275...19Y
ADC_Keywords: QSOs; Surveys; Photometry, ugriz; Redshifts; Spectra, optical
Keywords: Supermassive black holes ; Quasars
Abstract:
We present a machine learning search for high-redshift (5.0<z<6.5)
quasars using the combined photometric data from the Dark Energy
Spectroscopic Instrument (DESI) Imaging Legacy Surveys and the
Wide-field Infrared Survey Explorer survey. We explore the imputation
of missing values for high-redshift quasars, discuss the feature
selections, compare different machine learning algorithms, and
investigate the selections of class ensemble for the training sample,
then we find that the random forest model is very effective in
separating the high-redshift quasars from various contaminators. The
11 class random forest model can achieve a precision of 96.43% and a
recall of 91.53% for high-redshift quasars for the test set. We
demonstrate that the completeness of the high-redshift quasars can
reach as high as 82.20%. The final catalog consists of 216,949
high-redshift quasar candidates with 476 high probable ones in the
entire Legacy Surveys DR9 footprint, and we make the catalog publicly
available. Using Multi Unit Spectroscopic Explorer (MUSE) and DESI
early data release (EDR) public spectra, we find that 14 true high-
redshift quasars (11 in the training sample) out of 21 candidates are
correctly identified for MUSE, and 20 true high-redshift quasars (11
in the training sample) out of 21 candidates are correctly identified
for DESI-EDR. Additionally, we estimate photometric redshift for the
high-redshift quasar candidates using a random forest regression model
with a high precision.
Description:
We use the photometric data from the Data Release 9 (DR9) of the
Legacy Survey, which includes observations obtained by the DECam at
the CTIO 4m (DECaLS), an upgraded MOSAIC camera at the KPNO 4m
telescope (Mayall z-band Legacy Survey), and the 90Prime camera at the
Steward Observatory 2.3m telescope (Beijing-Arizona Sky Survey).
In addition to the Legacy Survey data, we also take advantage of the
WISE data release, providing infrared photometry of four bands at
central wavelengths of 3.4, 4.6, 12, and 22um (W1, W2, W3, W4) over
the entire sky.
See Section 2.
Verification using MUSE -- see Section 5.3. The Multi Unit
Spectroscopic Explorer (MUSE), an integral field unit mounted on the
Very Large Telescope (Yepun, UT4), has a field of view (FOV) of
1arcmin2 (R∼3000 at ∼7000Å). Among the entire high-redshift
candidates, 21 of them are located within the MUSE field.
Verification using DESI -- see Section 5.4. Currently, there are 306
objects at z>5 in the DESI-EDR catalog (DESI Coll.+ 2023). After
spectra inspection, we find that only 22 of them are true
high-redshift quasars at z>5, 12 of which are in our training sample,
and 10 of which are new high-redshift quasars.
The magnitudes used in this paper are generated based on the AB system
after applying Galactic extinction corrections. We use a ΛCDM
cosmology with ΩΛ=0.7, Ωm=0.3, and H0=70km/s/Mpc.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table11a.dat 144 568188 High-z quasar candidate catalog; flux model
table11b.dat 137 272424 High-z quasar candidate catalog; mag model
table11c.dat 136 216949 High-z quasar candidate catalog;
overlapping candidates
--------------------------------------------------------------------------------
See also:
II/294 : The SDSS Photometric Catalog, Release 7 (Adelman-McCarthy+, 2009)
II/363 : The band-merged unWISE Catalog (Schlafly+, 2019)
V/154 : Sloan Digital Sky Surveys (SDSS), Release 16 (DR16) (Ahumada+, 2020)
VII/289 : SDSS quasar catalog, sixteenth data release (DR16Q) (Lyke+, 2020)
III/286 : APOGEE-2 DR17 final allStar catalog (Abdurro'uf+, 2022)
VII/292 : DESI Legacy Imaging Surveys DR8 photometric redshifts (Duncan, 2022)
II/389 : The Pan-STARRS release 1 (PS1) Survey - DR2 (Magnier+, 2025)
J/ApJ/662/72 : Lyman alpha forest of 55 QSOs (Becker+, 2007)
J/ApJS/180/67 : Phot. selection of quasars from SDSS. II. (Richards+, 2009)
J/ApJS/219/39 : QSOs selection from SDSS and WISE (Richards+, 2015)
J/ApJS/227/11 : PS1 z>5.6 quasars follow-up (Banados+, 2016)
J/ApJ/819/24 : z>4.5 QSOs with SDSS and WISE. I. Opt. spectra (Wang+, 2016)
J/AJ/155/131 : z∼5 QSO luminosity function from the CFHTLS (McGreer+, 2018)
J/A+A/632/A56 : KiDS-BEXGO catalog (Khramtsov+, 2019)
J/A+A/624/A13 : KiDS DR3 QSO catalog (Nakoneczny+, 2019)
J/MNRAS/494/789 : NIR and MIR photometry of known z≥5 quasars (Ross+, 2020)
J/A+A/649/A81 : KiDSDR4 QSOs phot. redshifts catalog (Nakoneczny+, 2021)
J/AJ/162/72 : Random forests method to discover high-z QSOs (Wenzl+, 2021)
J/ApJ/923/262 : NIR spectroscopic obs. of z>6.5 quasars (Yang+, 2021)
J/ApJS/269/27 : DESI z≳5 quasar survey. I. New z∼4.7-6.6 QSOs (Yang+, 2023)
Byte-by-byte Description of file: table11a.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 F10.6 deg RAdeg Right Ascension (J2000) (G1)
12- 21 F10.6 deg DEdeg [-68.2/84.8] Declination (J2000) (G1)
23- 27 F5.2 mag gmag [22/38]? Apparent g-band AB magnitude (G1)
29- 33 F5.2 mag rmag [18.9/37.2]? Apparent r-band AB magnitude
(G1)
35- 39 F5.2 mag zmag [15.4/21.5] Apparent z-band AB magnitude
(G1)
41- 45 F5.2 mag W1mag [10.3/22.7] Apparent W1-band AB magnitude
(G1) (G2)
47- 51 F5.2 mag W2mag [9.7/23] Apparent W2-band AB magnitude (G1)
(G2)
53- 57 F5.2 mag grzmag [17/32]? Constructed grz-band magnitude
59- 63 F5.2 mag Wmag [10.1/22.6] Constructed W-band magnitude
65- 73 E9.2 mW/m2/Hz gFlux [-39/1.6] g-band flux (G1)
75- 83 E9.2 mW/m2/Hz rFlux [-41.1/26.4] r-band flux (G1)
85- 90 F6.2 mW/m2/Hz zFlux [2.5/645] z-band flux (G1)
92- 99 F8.2 mW/m2/Hz W1Flux [0.8/69819] W1-band flux (G1)
101- 109 F9.2 mW/m2/Hz W2Flux [0.6/129743] W2-band flux (G1)
111- 116 F6.2 mW/m2/Hz grzFlux [-0.9/141] Constructed grz-band flux
118- 125 F8.2 mW/m2/Hz WFlux [0.9/84800] Constructed W-band flux
127- 132 F6.4 --- HizProb [0.16/0.98] Probability for high-z class
134- 138 F5.3 --- zphot [4.76/6.31] Photometric redshift from
regression model
140- 144 F5.3 --- zspec [5/6.5]? Spectroscopic redshift determined
from spectrum
--------------------------------------------------------------------------------
Byte-by-byte Description of file: table11b.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 F10.6 deg RAdeg Right Ascension (J2000) (G1)
12- 21 F10.6 deg DEdeg [-68.2/84.8] Declination (J2000) (G1)
23- 27 F5.2 mag gmag [23.5/38]? Apparent g-band AB magnitude (G1)
29- 33 F5.2 mag rmag [19.8/35.4]? Apparent r-band AB magnitude
(G1)
35- 39 F5.2 mag zmag [16.2/21.5] Apparent z-band AB magnitude
(G1)
41- 45 F5.2 mag W1mag [13.9/22.7] Apparent W1-band AB magnitude
(G1) (G2)
47- 51 F5.2 mag W2mag [13.9/23] Apparent W2-band AB magnitude (G1)
(G2)
53- 57 F5.2 mag grzmag [17.9/23.2]? Constructed grz-band magnitude
59- 63 F5.2 mag Wmag [13.9/22.6] Constructed W-band magnitude
65- 72 E8.2 mW/m2/Hz gFlux [1e-06/0.4] g-band flux; erg/s/cm2/Hz (G1)
74- 81 E8.2 mW/m2/Hz rFlux [7e-06/12] r-band flux; erg/s/cm2/Hz (G1)
83- 88 F6.2 mW/m2/Hz zFlux [2.5/315] z-band flux; erg/s/cm2/Hz (G1)
90- 96 F7.2 mW/m2/Hz W1Flux [0.8/2649] W1-band flux; erg/s/cm2/Hz (G1)
98- 104 F7.2 mW/m2/Hz W2Flux [0.7/2604] W2-band flux; erg/s/cm2/Hz (G1)
106- 110 F5.2 mW/m2/Hz grzFlux [0.5/69] Constructed grz-band flux;
erg/s/cm2/Hz
112- 118 F7.2 mW/m2/Hz WFlux [0.9/2638] Constructed W-band flux;
erg/s/cm2/Hz
120- 125 F6.4 --- HizProb [0.17/0.97] Probability for high-z class
127- 131 F5.3 --- zphot [4.78/6.23] Photometric redshift from
regression model
133- 137 F5.3 --- zspec [5/6.5]? Spectroscopic redshift determined
from spectrum
--------------------------------------------------------------------------------
Byte-by-byte Description of file: table11c.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 F10.6 deg RAdeg Right Ascension (J2000) (G1)
12- 21 F10.6 deg DEdeg [-68.13/84.8] Declination (J2000) (G1)
23- 27 F5.2 mag gmag [23.57/38]? Apparent g-band AB magnitude
(G1)
29- 33 F5.2 mag rmag [19.8/35.4]? Apparent r-band AB magnitude
(G1)
35- 39 F5.2 mag zmag [16.2/21.5] Apparent z-band AB magnitude
(G1)
41- 45 F5.2 mag W1mag [15/22.6] Apparent W1-band AB magnitude (G1)
(G2)
47- 51 F5.2 mag W2mag [14.7/22.8] Apparent W2-band AB magnitude
(G1) (G2)
53- 57 F5.2 mag grzmag [17.9/23.2]? Constructed grz-band magnitude
59- 63 F5.2 mag Wmag [14.9/22.6] Constructed W-band magnitude
65- 72 E8.2 mW/m2/Hz gFlux [1e-06/0.4] g-band flux; erg/s/cm2/Hz (G1)
74- 81 E8.2 mW/m2/Hz rFlux [7e-06/12] r-band flux; erg/s/cm2/Hz (G1)
83- 88 F6.2 mW/m2/Hz zFlux [2.5/315] z-band flux; erg/s/cm2/Hz (G1)
90- 95 F6.2 mW/m2/Hz W1Flux [0.9/995] W1-band flux; erg/s/cm2/Hz (G1)
97- 103 F7.2 mW/m2/Hz W2Flux [0.8/1307] W2-band flux; erg/s/cm2/Hz (G1)
105- 109 F5.2 mW/m2/Hz grzFlux [0.5/69] Constructed grz-band flux ;
erg/s/cm2/Hz
111- 117 F7.2 mW/m2/Hz WFlux [0.9/1073] Constructed W-band flux;
erg/s/cm2/Hz
119- 124 F6.4 --- HizProb [0.18/0.97] Probability for high-z class
126- 130 F5.3 --- zphot [4.78/6.23] Photometric redshift from
regression model
132- 136 F5.3 --- zspec [5/6.5]? Spectroscopic redshift determined
from spectrum
--------------------------------------------------------------------------------
Global notes:
Note (G1): In the DESI Imaging Legacy Survey DR9 catalog.
Note (G2): The conversion between AB and Vega magnitude for W1 and W2 is
W1AB=W1Vega+2.699 and W2AB=W2Vega+3.339, respectively.
--------------------------------------------------------------------------------
History:
From electronic version of the journal
(End) Prepared by [AAS], Emmanuelle Perret [CDS] 05-Sep-2025