DESI z ≳ 5 Quasar Survey. I. A First Sample of 400 New Quasars at z ∼ 4.7–6.6

We report the first results of a high-redshift (z ≳ 5) quasar survey using the Dark Energy Spectroscopic Instrument (DESI). As a DESI secondary target program, this survey is designed to carry out a systematic search and investigation of quasars at 4.8 < z < 6.8. The target selection is based on the DESI Legacy Imaging Surveys (the Legacy Surveys) DR9 photometry, combined with the Pan-STARRS1 data and J-band photometry from public surveys. A first quasar sample has been constructed from the DESI Survey Validation 3 (SV3) and first-year observations until 2022 May. This sample includes more than 400 new quasars at redshift 4.7 ≤ z < 6.6, down to 21.5 magnitude (AB) in the z band, discovered from 35% of the entire target sample. Remarkably, there are 220 new quasars identified at z ≥ 5, more than one-third of existing quasars previously published at this redshift. The observations so far result in an average success rate of 23% at z > 4.7. The current spectral data set has already allowed analysis of interesting individual objects (e.g., quasars with damped Lyα absorbers and broad absorption line features), and statistical analysis will follow the survey’s completion. A set of science projects will be carried out leveraging this program, including quasar luminosity function, quasar clustering, intergalactic medium, quasar spectral properties, intervening absorbers, and properties of early supermassive black holes. Additionally, a sample of 38 new quasars at z ∼ 3.8–5.7 discovered from a pilot survey in the DESI SV1 is also published in this paper.


INTRODUCTION
Quasars at z ≳ 5, residing in the cosmic times from the reionization era to the post-reionization epoch, are unique probes of the intergalactic medium (IGM) evolution during the last stage of the neutral hydrogen phase transition in the IGM (e.g., Eilers et al. 2018;Yang et al. 2020b;Bosman et al. 2022) and the growth of early supermassive black holes (SMBH) (e.g., Shen et al. 2019;Yang et al. 2021;Farina et al. 2022).The study of quasar luminosity function and black hole mass function over cosmic epochs tracks SMBH accretions (e.g., Shen & Kelly 2012;Kelly & Shen 2013).In addition, luminous high-redshift quasars are thought to trace the most massive dark matter halos in the young Universe (e.g., Costa et al. 2014).These luminous background sources also provide valuable sightlines for the investigations of intervening high-redshift H i in the circumgalactic medium or interstellar medium, e.g., the damped Lyα (DLA) absorbers, and metal absorbers (e.g., Chen et al. 2017;D'Odorico et al. 2018) .
Recent successful high-redshift quasar surveys have increased the number of known z ≥ 5 quasars to ∼ 600 (e.g., Fan et al. 2006;Willott et al. 2010;Mortlock et al. 2011;Venemans et al. 2015;Wu et al. 2015;Bañados et al. 2016;Jiang et al. 2016;Mazzucchelli et al. 2017;McGreer et al. 2018;Reed et al. 2019;Matsuoka et al. 2019a,b;Wang et al. 2016Wang et al. , 2018Wang et al. , 2019;;Yang et al. 2017Yang et al. , 2019a,b;,b;Ross & Cross 2020;Yang et al. 2021) with the three most distant quasars at z > 7.5 (Bañados et al. 2018;Yang et al. 2020a;Wang et al. 2021).After the end of high-redshift quasar searches using the Sloan Digital Sky Survey I-IV (e.g., Schneider et al. 2010;Pâris et al. 2017;Lyke et al. 2020), recent discoveries of highredshift quasars are mainly based on single-object observations, which has a limited efficiency of spectroscopy.As a result, most existing high-redshift quasar surveys either focus on luminous quasars in a wide-field (e.g., Bañados et al. 2016;Wang et al. 2016;Yang et al. 2017) or target faint objects in a smaller area (e.g., McGreer et al. 2018;Matsuoka et al. 2019a).The Dark Energy Spectroscopic Instrument (DESI; DESI Collaboration et al. 2022) employs a set of 10 multi-fiber spectrographs with wide spectral coverage, offering a unique opportunity to search for high-redshift quasars with high spectroscopic efficiency.The combination of deep wide-field imaging from the DESI Legacy Imaging Surveys (Dey et al. 2019, hereafter the Legacy Surveys) and DESI multiobject spectroscopy will allow a new wide-field survey of high-redshift quasars, extending through the entire DESI footprint and to a fainter luminosity range.
In this paper, we report a DESI secondary program designed to search for quasars at z ∼ 4.8 − 6.8 using DESI.This is our program's first publication to introduce the survey and publish discoveries from the DESI observations by 2022 May.We will describe the photometric datasets and target selection in Section 2. We then report the observations and the first sample of new quasars in Section 3. In Section 3, we will also present examples of scientific analyses using the current quasar spectra.In Section 4, in addition to a summary, we will discuss the future science projects using quasars from this survey.Additionally, we also publish a sam-ple of new quasars from a pilot program of searching for z ∼ 4 − 5.3 quasars during the DESI Survey Validation (SV) observations, which will be briefly described in Section 2. In this paper, we adopt a ΛCDM cosmology with parameters Ω Λ = 0.7, Ω m = 0.3, and H 0 = 70 km s −1 Mpc −1 .Photometric data in the optical are reported in the AB system after applying the Galactic extinction correction (Schlegel et al. 1998;Schlafly & Finkbeiner 2011); photometric data from infrared surveys (e.g., in J, W 1, and W 2 bands) are on the Vega system.

A HIGH-REDSHIFT QUASAR SURVEY IN DESI
The DESI survey is designed to determine the nature of dark energy through the most precise measurement of the expansion history of the universe (Levi et al. 2013).With 5,000 fibers mounted in the focal plate and a 3degree-diameter field of view (DESI Collaboration et al. 2016b;Silber et al. 2023), DESI will measure the spectra of about 40 million galaxies and quasars covering 14,000 square degrees over its 5-year observing campaign (DESI Collaboration et al. 2016a, 2022).DESI target selection is based on the imaging data from the Legacy Surveys (Zou et al. 2017;Dey et al. 2019, Schlegel et al. 2023 in preparation), including targets of the Milky Way Survey (MWS; Allende Prieto et al. 2020;Cooper et al. 2023), the Bright Galaxy Survey (BGS) (Ruiz-Macias et al. 2020;Hahn et al. 2023), the Luminous Red Galaxy (LRG) sample (Zhou et al. 2020(Zhou et al. , 2023)), the Emission Line Galaxy (ELG) sample (Raichoor et al. 2020(Raichoor et al. , 2023)), and the quasar (QSO) sample (Yèche et al. 2020;Chaussidon et al. 2023).The details of the pipelines designed for DESI target selections, fiber assignments, survey operations, spectral reduction, and spectral classifications can be found in Myers et al. (2023), Raichoor et al. (2023 in preparation), Schlafly et al. (2023), Guy et al. (2023), andBailey et al. (2023 in preparation).The DESI survey will not only use quasars as direct tracers of the matter distribution mostly at z < 2.1, but also use quasars at z > 2.1 as backlights for the intervening matter distribution via the Lyα forest.The quasar selection for the DESI main survey is based on a Random Forest algorithm and is mainly focused on quasars at z ≲ 4.5 (Chaussidon et al. 2023).
We present an additional quasar survey program designed to systematically search for quasars at z ≳ 5 using DESI, which is part of the DESI secondary target program of the DESI survey.More details of DESI secondary target programs and the DESI target pipeline can be found in Myers et al. (2023).The wide spectral coverage (3600 -9800 Å) and high efficiency of the DESI spectrographs allow sensitive identification of quasar Lyα lines up to redshift ∼ 6.8 (covering wavelength range blueward of 1250 Å in the rest-frame).In addition, DESI's spectral resolution (R ∼ 3000 − 5000 at λ > 5500 Å) is higher than that of typical discovery spectra (R ≲ 1000 ) used in most other highredshift quasar surveys.Thus, the DESI spectra can be used to directly construct a dataset for a wide range of scientific analyses with no need of optical follow-up spectroscopy.For example, a complete DESI spectral dataset of z ≳ 5 quasars will allow us to investigate quasar luminosity function, quasar clustering, quasar rest-frame UV spectral properties, IGM evolution, and intervening absorbers, as well as the early supermassive black holes growth, together with multi-wavelength follow-up observations.

Photometric Datasets
The candidate selection of our survey is mainly based on the photometric data from the Legacy Surveys (Dey et al. 2019;Schlegel et al. 2023 in preparation), which images more than 19,700 square degrees of the extragalactic sky visible from the Northern hemisphere in g, r, and z bands.The Legacy Surveys consists of three programs: the Beijing-Arizona Sky Survey (BASS; Zou et al. 2017), covering an area in the North Galactic Cap with Dec > 32.375 deg in g and r bands, the Mayall z-band Legacy Survey (MzLS; Dey et al. 2019), imaging the same area as BASS in z band, and the Dark Energy Camera Legacy Survey (DECaLS; Dey et al. 2019), mapping the entire South Galactic Cap and the regions in North Galactic Cap with Dec < 34 deg in g, r, and z bands.There are also other public DECam grz data within the DESI footprint.For example, the data from the Dark Energy Survey (DES; The Dark Energy Survey Collaboration 2005), which overlaps with the Legacy Surveys in an era of 1,130 deg 2 .The Legacy Surveys program makes use of the DES raw data instead of reobserving that era.
The Legacy Surveys DR9 1 includes images in all the three bands with ≥ 3 observational passes over 14,750 square degrees, with median PSF depth of 24.7 (24.2 in BASS area), 24.2 (23.7 in BASS), and 23.3 in the g, r, and z bands, respectively.The optical data are complemented by photometry in infrared bands (W 1 and W 2) from the all-sky data of the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010).The W 1 and W 2 data in the Legacy Surveys catalog are based on forced photometry in the unWISE images at the locations of the Legacy Surveys optical sources.The un-WISE images are the unblurred coadds of WISE (Meisner et al. 2017(Meisner et al. , 2018) ) including all imaging through year 6 of NEOWISE-Reactivation (Mainzer et al. 2014).
We have also included photometric data from the Pan-STARRS1 survey (PS1; Chambers et al. 2016), which covers the 3π sky at Dec > −30 deg in the optical grizy bands.In our survey, the PS1 i-band photometry is required for target selection, and data in PS1 z and y bands are included if available (see details in the next section).The PS1 survey has 5σ depth of 23.1, 22.3, and 21.4 magnitude in the i, z, and y bands, respectively.In the following discussions of color selection, we use z to represent the z-band photometric data from the Legacy Surveys and use z P1 for the PS1 photometry.
In addition, in order to further reduce contamination rate of quasar selection, photometric data in the nearinfrared (NIR) J-band are used for objects with available J-band photometry.We collect J-band data from the three public large-area NIR imaging surveys: the UKIRT Hemisphere Survey (UHS) (Dye et al. 2018), the UKIRT InfraRed Deep Sky Surveys-Large Area Survey (ULAS, Lawrence et al. 2007), and the VISTA Hemisphere Survey (VHS, McMahon et al. 2013).The UHS and ULAS surveys cover an area of 17,900 square degrees in the northern sky with a depth of 19.6 mag (Vega) in J.The VHS survey aims to map the entire Southern sky and has a depth of 20.2 mag (Vega) in the J band.The quasar selection of our survey is based on opticalinfrared color cuts, which have been previously applied to successful z ∼ 5 − 6.5 quasar surveys (e.g., Bañados et al. 2016;Wang et al. 2016Wang et al. , 2019;;Yang et al. 2017Yang et al. , 2019a,b),b).We start our selection with the Legacy Surveys DR9 catalog and first apply a set of standard cuts: 1) BRICK PRIMARY = T to select objects within the brick boundary; 2) MASKBITS not in [1,10,12,13], which requires that the targets do not touch pixels in the vicinity of bright stars, "bailout" blob (areas leading to fitting issue due to a very high source density), large galaxies, or globular clusters; 3) TYPE = PSF OR (dchisq[1]dchisq[0])/dchisq[0] < 0.01 to select objects with stellar morphology.We next require that all targets should be observed in all bands and limit the signal-to-noise (S/N) in the z (> 5σ), W 1 (> 3σ), and W 2 (> 2σ) bands.Then we apply very relaxed cuts in the gr bands (i.e., S/N g < 5 or g > 24.5 or g − r > 1.8; S/N r < 5 or r − z > 1.0) to reduce the sample size.These generate a pre-selected sample with ∼ 13 million sources within the DESI footprint for following color-color selection.
One basic technique used for the color selection of high-redshift quasars is based on color 'drop-out', which .The red points on the track represent quasars at z = 5.0, 5.5, 6.0, and 6.5.The blue crosses denote all existing quasars known at 4.8 ≤ z ≤ 6.8 (e.g., Fan et al. 2006;Willott et al. 2010;Mortlock et al. 2011;Venemans et al. 2015;Wu et al. 2015;Bañados et al. 2016Bañados et al. , 2018;;Jiang et al. 2016;Mazzucchelli et al. 2017;McGreer et al. 2018;Matsuoka et al. 2019a,b;Pâris et al. 2017;Reed et al. 2019;Wang et al. 2016Wang et al. , 2018Wang et al. , 2019Wang et al. , 2021;;Yang et al. 2017Yang et al. , 2019aYang et al. ,b, 2020aYang et al. , 2021)), while the grey points and the black squares show the loci of M and L/T dwarfs (Kirkpatrick et al. 2011;Mace 2014;Best et al. 2015), respectively.The orange dashed lines bound the selected regions for z ∼ 5 and z ∼ 6 quasar candidates. is caused by the Lyα break in quasar spectra due to significant IGM absorption at the wavelength blueward of the Lyα emission line at these redshifts.The Lyα line of a z ∼ 5 quasar is located in the i band (7294.02Å at z = 5).Thus, the i band is typically used as a detection band for the selection of z ∼ 5 quasars, and the g and r are the dropout bands.At z ≳ 5.7, Lyα line moves into the z band and the i band becomes the dropout band, so the i band also plays a critical role in the selection of z ∼ 6 quasars.We therefore include the i-band photometric data from the PS1 survey.
We cross match the pre-selected sample selected above with the PS1 DR1 and DR2 catalogs.We only use DR1 data when an object is not in the DR2 catalog but is in the DR1.Then we build up different selection criteria for three redshift ranges, z ∼ 4.8 − 5.4, 5.7 -6.4, and 6.4-6.8, according to the quasar color tracks in different color-color spaces.The main color-color selections are the r − i/i − z for z ∼ 5 and 6, and the z − y/y − W 1 for z ∼ 6 and 6.5.The z − W 1/W 1 − W 2 color cuts are used for all redshifts.The redshift range of z ∼ 5.4 − 5.6 is the redshift at which quasar locations fully overlap with those of middle-type M dwarfs in the riz colorcolor space.Multiple NIR colors are necessary to select quasars in this redshift range, as demonstrated in our previous successful z ∼ 5.5 quasar survey (Yang et al. 2017(Yang et al. , 2019a)).Thus we exclude this redshift range in this survey.Below we briefly describe our selections of quasars in the three redshift ranges, and show the main color-color diagrams in Figures 1 and 2. The selection criteria are listed in Appendix A.
z ∼ 4 .8− 5 .4quasar candidates -1) We first select objects with g-band dropout by applying the cuts of S/N g < 3 or g − r > 2.5.2) We next require objects to have > 5σ detection in the i P1 band and apply the r − i/i − z color cuts as shown in Figure 1. 3) Then we apply the z − W 1/W 1 − W 2 cuts (Fig. 1, bottom).4) Additionally, for objects that have > 3σ detections in the y P1 band, we use a y P1 − W 1 color (Fig. 2) to improve the purity.All objects without y P1 data are kept.
z ∼ 5 .7 − 6 .4quasar candidates -At z ≳ 6, the Lyα emission line moves into the z band. 1) Thus we first require the objects to be dropouts in the g and r bands (see criteria in Appendix A). 2) Then we select objects without i-band detection (S/N iP1 < 3) or meeting the cut of i P1 − z > 2.0.3) We also apply the z − W 1/W 1 − W 2 cuts, which are the same to the z ∼ 5 selection.4) In addition, if one object has > 3σ detections in the PS1 z and y bands, z P1 and y P1 photometry are employed to further reduce the contamination rate.The z P1 band covers a wavelength range bluer than the Legacy Sur- veys' z band, so it can provide additional constraints on colors.We limit y P1 − W 1 > 2.399 and z P1 − y P1 < 1.0, as shown in Figure 2. Objects without z P1 and y P1 detections are kept.
z > 6 .4quasar candidates -1) We first select objects that are dropouts in the gri bands (see criteria in Appendix A). 2) At the same time, we apply a more stringent z − W 1 cut as shown in Figure 1 (bottom).
3) At such high redshift, most quasars only have the z-band detection, except for very luminous ones, which could be detected in the i band due to a bright Lyβ line.Thus, we require additional photometric data in the y P1 band for all candidates.We select objects with > 3σ detection in the y P1 and require y P1 − W 1 > 2.399 (Fig. 2).4) We reject objects that have > 3σ detections in z P1 band but have colors of z P1 − y P1 ≤ 1.0 (Fig. 2) or z P1 − W 1 < 4.0.
NIR photometry can play an important role in further rejecting M/L/T dwarf contaminants in high-redshift quasar selections by leveraging the power-law continuum of quasars.Among the NIR broad bands (i.e., Y JHK), the J band has been mapped in the widest sky coverage.We therefore include J-band data into our selection, using photometry from the UHS, ULAS, and VHS surveys.We reject all objects that are covered by J-band photometry but have bad J-based colors (e.g., J −W 1 < 1.5 or y P1 −J > 2.0 if y P1 is available; J AB = J Vega +0.938).Objects without J-band data are kept.
In this first year of DESI observations, our highredshift quasar selection focuses on a relatively bright candidate sample, with a z-band magnitude limit of 21.4 Yang et al.
for z ∼ 5 quasar candidates and 21.5 for z ∼ 6 as well as z > 6.4 candidates.This is already one magnitude deeper than the previous wide-area z ∼ 5 quasar surveys (Wang et al. 2016;Yang et al. 2017Yang et al. , 2019a)), while this depth only covers the luminous end of z > 6 quasar population.We will extend the quasar selections towards a fainter range in subsequent years of the DESI survey.
In total, we obtain ∼ 3,500 targets for z ∼ 5 quasar selection, ∼ 3,000 targets from z ∼ 6 selection, and ∼ 450 targets for z > 6.4 quasars within the DESI footprint, with an average target density of 0.5 per square degree.These targets have a bit-name of Z5 QSO in the DESI target catalog.We are not excluding known quasars that pass our selection criteria in order to construct a uniformly observed spectral dataset, which is important for future spectral analysis.Multi-epoch spectra will also allow the study of quasar variability.At the survey depth, about 60-90% of known quasars can be selected by our selections in three different redshift ranges, as shown in Figure 3, which can be treated as a rough estimate of the selection completeness.A careful measurement of the selection function using a simulated quasar sample will be presented with our final quasar sample.

Selection of Quasars at
In this paper, we also report results from a selection of z ∼ 4 − 5.3 quasars using photometric data from the Legacy Surveys only, and this selection is used as a pilot selection during the DESI SV1 (Myers et al. 2023;DESI Collaboration et al. 2023).DESI conducted SV observations, prior to the main, five-year mission, to validate the systems, to refine the purity and completeness of the targeting algorithms, and to stress-test the procedures that would be needed (DESI Collaboration et al. 2023).SV1 is the first iteration of SV ran.The pilot survey of z ∼ 4 − 5.3 quasars was designed to test the quasar selection without the i-band photometry.The selection was based on grz and zW 1W 2 colors.The selection first adopted the g-band and r-band dropout techniques.The r−z color was applied instead of the riz colors, and the zW 1W 2 cuts were used with a relaxed This selection yielded ∼ 60 quasars during the SV1 observations between 2020 December and 2021 May.Among them, there were 38 new quasars at 3.8 ≤ z ≤ 5.7, as listed in Table 2 and shown in Figure 8 in the Appendix B. The selection of z ∼ 5 quasars without i-band photometry could help to construct a quasar sample that is independent of the Lyα line luminosity.Indeed, we were able to discover weak line (WL) quasars and broad absorption line (BAL) quasars at z ∼ 5 (Fig. 8), which are missed by the selection mentioned above and also by the quasar surveys in previous works (e.g., Wang et al. 2016;Yang et al. 2017Yang et al. , 2019a)).However, the lack of i-band photometry led to a high contamination rate; the success rate of this selection was only 2-3%, which is quite low compared to the efficiency of our main selection described above (see details of efficiency in Section 3).Therefore, this selection was not retained after the SV observations.We publish the 38 new quasars discovered in SV1 in this paper.

Observations and Redshift Measurements
The selected z ∼ 4.8 − 6.8 quasar candidates are mainly observed as dark-time targets in the DESI Main Survey, which started on May 14, 2021.A minor part (1.6%) of candidates were observed during the DESI 1% survey (SV3; in 2021 April) and the DESI SV1 (before April 4, 2021; for 27 targets overlapped with the SV1 selection).These targets have the same fiber assignment priority to the primary QSO targets (Schlafly et al. 2023).All our targets are designed to first have a single-exposure with an effective exposure time2 (Guy et al. 2023) of 1000s for the purpose of identification.A S/N > 3 per pixel is expected on the Lyα line to identify a target.Then all targets identified as quasars will be observed with three repeat exposures by DESI for high quality spectra.As of May 14, 2022, there are ∼ 2370 targets that have been observed from our program, which is about 35% of the entire candidate sample.The spectra of our targets after a first-pass observation have sufficient S/N for quasar identification, as shown in Figure 4 and Figure 7.
We identify quasars using the daily coadd spectra in the database, which are products of the DESI spectroscopic pipeline (Guy et al. 2023).The spectroscopic pipeline is designed to extract spectra from the raw data, subtract sky model, flux-calibrate spectra based on standard star exposures, and then measure their classifications as well as redshifts.We identify quasars via visual inspection and estimate quasar redshifts from visual spectral fitting, which is not included in the general DESI spectra visual inspection (Alexander et al. 2023;Lan et al. 2023).The visual spectral fitting is based on a semi-automated toolkit, A Spectrum Eye Recognition Assistant (ASERA; Yuan et al. 2013), which has been used to measure redshifts for a number of highredshift quasars (e.g., Wang et al. 2016Wang et al. , 2019;;Yang et al. 2017Yang et al. , 2019a)).All spectra of our program have been visually inspected by at least two inspectors.When fitting spectra, we match the observed spectrum with the SDSS quasar template based on quasars' broad emission lines, Lyβ, Lyα, O i, Si iv, and C iv if visible, as well as the continuum emission.Figure 4 shows three examples of the DESI spectra with emission lines identified.
They represent the quality of single-epoch observation spectra used for our identification and visual spectral fitting.Such redshift measurements have a typical uncertainty of ±0.03 (e.g., Wang et al. 2016;Yang et al. 2017).For weak line quasars and strong broad absorption line quasars, the uncertainty could be as large as ∼ 0.05 -0.1.We have one primary inspector and at least one secondary inspector.If redshifts from primary and secondary inspectors have a difference ≤ 0.03, we use redshift from the primary inspector.For a few cases that we get a discrepancy > 0.03, the inspectors repeat the VI together until they obtain a consistent result.
3.2.The Discovery of More than 400 New Quasars Among the observed targets, there are 556 quasars identified in the redshift range of z ∼ 4.4−6.8,including 144 known quasars and 412 new quasars.These new quasars span a redshift range from 4.44 to 6.53, with 220 quasars at z ≥ 5 and 25 quasars at z ≥ 6.The redshift distribution of these quasars is presented in Figure 3.So far, the observations of these targets result in an average success rate of 23% of quasars at z > 4.7.If we count the success rate of z ∼ 4.8 − 5.4, z ∼ 5.7 − 6.4, and z > 6.4 quasar selections separately, we obtain 39%, 8%, and 5%, respectively.The main contaminants are M/L/T dwarfs (> 90%) and red galaxies, as shown in the Appendix C (Fig. 9).In Figure 4, we show examples of z ∼ 5, 6, and 6.5 quasars newly identified from our program.Quasar DESI J144355+350055 at z = 6.53 is the highest-redshift new quasar identified from our survey to date.
Although our survey has not been completed, it has already significantly increased the number of quasars known at z ≥ 4.8 (377 new quasars).The total number of known quasars at z = 4.8 − 6.8 from previous works is about 810, and thus the new quasars from this program already expand the existing known quasar sample by 46%.At z ≥ 5, our survey so far has increased the known quasar sample by more than one third, with 220 new discoveries and 628 existing known quasars.Within the DESI footprint, our new discoveries have more than doubled the quasar sample size in the redshift range of 4.8 − 5.4.
We list all 412 new quasars in Table 1 and plot all spectra of these new quasars in Fig 7 .The spectra plot- ted here are based on the coadd data of the DESI daily spectra.All spectral data will be made public in digital form in future DESI data releases (e.g., DESI Collaboration et al. 2023).

Examples of Scientific Analysis
So far, the DESI high-redshift survey and the repeat observations of newly identified quasars have not been finished, so this sample is not complete enough for statistical analysis.However, there are already a number of spectra that have sufficient quality to allow us to start science analysis for individual objects.
As an example, Figure 5 shows the spectrum of quasar DESI J112903.54+023706.9, which has a z-band magnitude of 21.3, among the faintest targets in our candidate sample.The spectrum has ∼ 4,400s exposure time, and the average S/N over the Lyα+N v line region is 9 per 0.8 Å pixel and 3 per pixel on continuum.All quasars from this survey will be observed repeatedly by DESI to have comparable or longer exposure time.In addition, as shown in Figure 5, the Si iv and C iv BAL features can be seen clearly from this spectrum.The balnicity index BI (Weymann et al. 1991) is derived as ∼ 2000 km s −1 and 2500 km s −1 for the Si iv and C iv absorptions, respectively, using a 3-pixel binned spectrum.The BI (Weymann et al. 1991) Cdv, where f (v) is the normalized spectrum, and C is set to 1 only when f (v) is continuously smaller than 0.9 for more than 2000 km s −1 , otherwise it is set to 0.0.The value of v min is set to 0. The final spectral dataset will allow us to identify BAL quasars and measure the BAL fractions at different redshift bins.
These quasar spectra also enable us to search for highredshift DLA absorbers.Figure 6 shows an example of a z = 4.018 DLA identified in the spectrum of z = 5.03 quasar DESI J100828.30−021229.8.The spectrum has a coadd exposure time of ∼ 5,000s.The redshift of this DLA is derived using a set of metal lines.Metal lines Si II λ1526, C IV λ1548, C IV λ1550, Fe II λ1608, and Al II λ1670 have been identified.The column density of this DLA is determined as log N HI = 21.2 ± 0.2 by fitting a Voigt profile.We place the centroid of a Voigt profile to the redshift of the low-ion metal-line transitions (z = 4.018) and manually select values of N HI to fit the profile, assuming a single component of DLA plus possible blended Lyα forest in the wings.The uncertainties of N HI estimations for high-redshift DLAs are dominated by continuum uncertainty and Lyα forest line blending (Rafelski et al. 2012).The column density of metal lines is derived from the apparent optical depth method using the routines publicly available in the LINETOOLS package (Prochaska et al. 2017).We then obtain logN = 14.64 ± 0.03 for Si using the Si II λ1526 line, which implies a metallicity of −2.06 ± 0.21.This follows the DLA metallicity-redshift evolution trend (Rafelski et al. 2012).A systematical search will be carried out using the complete DESI spectral sample.
In addition, we study radio detections among these new quasars by cross-matching (2 ′′ ) new quasars with the catalogs from the Faint Images of the Radio Sky at Twenty-cm (FIRST;Becker et al. 1995), the first epoch of the Very Large Array Sky Survey (VLASS; Lacy et al. 2020;Gordon et al. 2020), and the second data release from the ongoing LOw-Frequency ARray (LOFAR) Two-metre Sky Survey (LoTSS) (Shimwell et al. 2022).The cross-match radius is chosen following recent LOFAR studies (e.g., Retana-Montenegro & Röttgering 2018;Gloudemans et al. 2022), although it might be small for matching FIRST catalog.As a first test, we choose this radius for a reliable detection.We find six quasars detected in the FIRST catalog with peak flux densities from 1.2 to 54.7 mJy/beam at 1.4 GHz.The same six objects are in the VLASS catalog.Two quasars have significant differences in the peak flux density between the measurements from FIRST and VLASS, which might be useful to study quasar radio variability.We also obtain eleven additional radio detections (z ∼ 4.7 − 6.0) in the LOTSS DR2 catalog, with peak flux densities from 0.3 to 1.1 mJy/beam at 144 MHz.In the future, using the final sample, we will be able to measure the radio-loud fractions in different redshift bins.

SUMMARY AND FUTURE STUDIES
In this paper, we describe a high-redshift (z ∼ 4.8-6.8)quasar survey with DESI, as a secondary program of the DESI survey.Using the photometric data from the Legacy Surveys DR9 (g, r, z, W 1, and W 2), the PS1 survey (i P1 , z P1 , and y P1 ), and several NIR surveys (J), we carried out selections of quasars in three redshift ranges: z ∼ 4.8 − 5.4, 5.7 − 6.4, and 6.4 − 6.8.The observations during the DESI SV1 and the main survey before May 14, 2022 have covered 35% of candidates and identified more than 550 quasars, yielding 412 new quasars at 4.44 ≤ z ≤ 6.53, with 377 quasars at z ≥ 4.8 and 220 at z ≥ 5.0.The observations to date result in an average success rate of 23%.Assuming this success rate, we would expect ∼ 1,000 new quasars from this survey.The new discoveries from this survey have increased the sample size of known quasars by ∼ 46% in the redshift range that we are targeting.The high quality spectra allow us to start the study of individual quasar spectra, although the current dataset is not yet complete enough for statistical analysis.
Based on this survey, we expect to construct a large z ≳ 5 quasar sample for scientific analyses.We will present details of the full quasar sample and the selection function when the identification of the entire sample is complete.The quasar luminosity functions will be also calculated at that time.In subsequent papers, we will present spectral fitting and basic quasar spectral properties (e.g., continuum slope, luminosity, and line width).When all repeat observations of quasars have been completed, the DESI quasar spectral dataset will allow a set of science projects, including but not limited to z ∼ 5 − 6 quasar clustering, quasar proximity zone, IGM evolution, radio-loud fraction, WL and BAL quasar fractions, as well as intervening DLA and metal absorbers.Follow-up observations in the NIR will help to build a sample for the study of early SMBHs growth and the BH mass function.We will also push the high-redshift quasar survey to a fainter magnitude in the future DESI survey.Bañados, E., Venemans, B. P., Decarli, R., et al. 2016, ApJS, 227, 11 Bañados, E., Venemans, B. P., Mazzucchelli, C., et al. 2018, Nature, 553, 473 Bañados, E., Schindler, J.-T., Venemans, B. P., et al. 2023, ApJS, 265, 29 Figure 9.The example spectra of contaminants from our quasar survey, rebinned with seven pixels.The first two are M dwarfs and the other two are galaxies.They all have red colors and thus contaminate the quasar selection.

Figure 1 .
Figure1.Top: The riz color-color diagram used for the quasar selection.The blue solid line shows the color track of quasars from z = 4.8 to z = 6.8 with a step of 0.1 (blue filled circles).The red points on the track represent quasars at z = 5.0, 5.5, 6.0, and 6.5.The blue crosses denote all existing quasars known at 4.8 ≤ z ≤ 6.8 (e.g.,Fan et al. 2006;Willott et al. 2010;Mortlock et al. 2011;Venemans et al. 2015;Wu et al. 2015; Bañados et al. 2016 Bañados et al.  , 2018;;Jiang et al. 2016;Mazzucchelli et al. 2017;McGreer et al. 2018; Matsuoka et al.  2019a,b;Pâris et al. 2017;Reed et al. 2019;Wang et al. 2016Wang et al. , 2018Wang et al. , 2019Wang et al. , 2021;;Yang et al. 2017Yang et al. , 2019a Yang et al.  ,b, 2020aYang et al. , 2021)), while the grey points and the black squares show the loci of M and L/T dwarfs(Kirkpatrick et al. 2011;Mace 2014;Best et al. 2015), respectively.The orange dashed lines bound the selected regions for z ∼ 5 and z ∼ 6 quasar candidates.Bottom: The zW 1W 2 color-color diagram.All symbols are the same as in the top panel.The orange dashed lines represent the cuts used for z ∼ 5 and z ∼ 6 quasar candidates.A more stringent z − W 1 cut (orange dotted line) is used for the selection of z > 6.4 quasars.Note that the photometry in W 1 and W 2 are in the Vega system (W 1Vega = W 1AB − 2.699, W 2Vega = W 2AB − 3.339).
Figure1.Top: The riz color-color diagram used for the quasar selection.The blue solid line shows the color track of quasars from z = 4.8 to z = 6.8 with a step of 0.1 (blue filled circles).The red points on the track represent quasars at z = 5.0, 5.5, 6.0, and 6.5.The blue crosses denote all existing quasars known at 4.8 ≤ z ≤ 6.8 (e.g.,Fan et al. 2006;Willott et al. 2010;Mortlock et al. 2011;Venemans et al. 2015;Wu et al. 2015; Bañados et al. 2016 Bañados et al.  , 2018;;Jiang et al. 2016;Mazzucchelli et al. 2017;McGreer et al. 2018; Matsuoka et al.  2019a,b;Pâris et al. 2017;Reed et al. 2019;Wang et al. 2016Wang et al. , 2018Wang et al. , 2019Wang et al. , 2021;;Yang et al. 2017Yang et al. , 2019a Yang et al.  ,b, 2020aYang et al. , 2021)), while the grey points and the black squares show the loci of M and L/T dwarfs(Kirkpatrick et al. 2011;Mace 2014;Best et al. 2015), respectively.The orange dashed lines bound the selected regions for z ∼ 5 and z ∼ 6 quasar candidates.Bottom: The zW 1W 2 color-color diagram.All symbols are the same as in the top panel.The orange dashed lines represent the cuts used for z ∼ 5 and z ∼ 6 quasar candidates.A more stringent z − W 1 cut (orange dotted line) is used for the selection of z > 6.4 quasars.Note that the photometry in W 1 and W 2 are in the Vega system (W 1Vega = W 1AB − 2.699, W 2Vega = W 2AB − 3.339).

Figure 2 .
Figure 2. The zP1yP1W 1 color-color diagram, used for the selection of z ∼ 6 and 6.5 quasars.All symbols are the same as Figure 1.A yP1 − W 1 cut is applied for all objects, and the two regions separated by the zP1 − yP1 color (vertical orange dashed line) are the selections for z ∼ 6 and z > 6.4 quasar candidates, respectively.
8 quasars due to the bluer W 1 − W 2 color of z ∼ 4.5 quasars.The selection criteria are listed in the Appendix B.

Figure 3 .
Figure3.A stacked histogram presenting the redshift distribution of new quasars (red, "New QSOs") from this survey, compared with all existing known quasars that are in the DESI footprint and within our survey depth (i.e., < 21.5 in the LS z band).Known quasars that pass our selections are shown in orange ("Selected known QSOs"), while the quasars that do not meet our selections are in the color of blue ("Non-sel Known QSOs").The redshift range that our survey focuses on has been marked in red (z = 4.8 − 6.8).

Figure 4 .
Figure 4. Examples of the DESI spectra for quasars at z = 5, 6, and 6.5, binned with seven pixels.The three spectra are from observations with ∼ 800 − 1000s actual exposure time.These three quasars are 20.55,20.72, and 21.20 magnitude in the Legacy Survey z band.The quasar broad emission lines (i.e., Lyβ, Lyα, O i, Si iv, and C iv) used for redshift measurements are marked with blue dotted lines.The spectra of all 412 new quasars are shown in Figure 7.

Figure 5 .
Figure5.The DESI spectrum of a z = 5.04 quasar J112903.54+023706.9, coadded from four exposures with a total exposure time of ∼ 4,400s.The spectrum plotted here is binned with 3 pixels for the BAL absorption measurements.The grey line is the spectral uncertainty.The target has a z-band magnitude of 21.3.This spectrum has sufficient S/N for quasar classification and study of strong absorption features.The Si iv and C iv BAL absorption troughs are denoted by the blue shaded regions.

Figure 6 .
Figure 6.The DESI spectrum of a z = 5.03 quasar J100828.30−021229.8 without binning.The grey line represents the spectral uncertainty.A DLA system at z = 4.018 (red dashed line) is clearly present in the absorption spectrum.Its redshift is estimated using metal lines.The Voigt profile fitting (inset plot) yields a column density of log NHI = 21.2 ± 0.2.

Figure 7 .
Figure 7. Continued.We are presenting the first two pages of the spectral sample, and the complete figure set (10 pages, 412 spectra) is available in the online journal.

Figure 8 .
Figure 8.The spectra of new quasars from the SV1 selection ordered as in Table2.The y-axis shows the flux density (f λ ) in units of 10 −17 erg s −1 cm −2 Å−1 .The spectra have been binned with 11 pixels for the purpose of plotting.The blue vertical lines denote the observed wavelengths of the emission lines, including (from left to right) Lyβ, Lyα, O i, Si iv, and C iv. Quasar J0548-2233 is an example of WL quasar, and J0719+4249 is an example of BAL quasar.

Table 1 .
The 412 New Quasars from Our Main Selection.

Table 1 continued
Table1(continued) Redshift from visual fitting using quasar template with a typical uncertainty of 0.03.For strong BAL quasars and WL quasars, the uncertainty could be ∼ 0.05 − 0.1.