A Catalog of Spectral Properties for DESI DR1 Quasars

Introduction

Large, homogeneous spectroscopic surveys have transformed quasar astrophysics by enabling detailed demographic studies and systematic measurements of emission-line and continuum properties across cosmic time. The Sloan Digital Sky Survey (SDSS; [62]) established this paradigm, producing successively larger quasar catalogs through BOSS and eBOSS and culminating in DR16Q ([27]). Value-added measurements from SDSS spectra, such as emission-line properties, quasar luminosities, and single-epoch (SE) black hole (BH) masses, as well as improved systemic redshifts, have become foundational community resources (e.g., [48, 35] (Wu & Shen 2022)).

The Dark Energy Spectroscopic Instrument (DESI; [26]) survey builds on this legacy and represents a major advance in wide-area spectroscopic surveys. Installed on the Mayall 4-meter Telescope at Kitt Peak National Observatory, DESI is the result of an international collaboration effort to deliver efficient, high-throughput, and uniformly processed spectroscopy for tens of millions of targets. DESI Data Release 1 (DR1; [13]) comprises the first 13 months of Main Survey observations, together with uniformly reprocessed Survey Validation data, and delivers high-confidence redshifts for $\sim18.7$ million objects, including $\sim1.7$ million quasars. As the largest spectroscopic compilation to date, DESI DR1 provides an unprecedented data set for studying quasar spectral diversity, improving systemic redshift measurements, and enabling statistical investigations of SMBH growth, outflows, and the quasar luminosity function.

In this work, we present a value-added catalog of spectral properties for DESI DR1 quasars. This paper is ororganized as follows. § 2 describes the DESI DR1 quasar sample and spectroscopy, and § 3 presents the spectral fitting procedure and derived quantities. In § 4, we validate the catalog using systemic redshifts and comparisons with SDSS, and present how emission-line ratios trend with Eddington ratio. We summarize our results in § 5. Throughout this paper, we adopt a flat $\Lambda$CDM cosmology with $\Omega_\Lambda= 0.7$, $\Omega_M = 0.3$, and $H_0 = 70\ \mathrm{km\ s^{-1}\ Mpc^{-1}}$.

Data

Quasar targets in DESI are selected primarily from optical imaging in $g$, $r$, $z$ bands, and WISE W1, W2 using a random forest algorithm to improve quasar selection efficiency [10, 2]. The main-survey selection achieves high quasar purity at magnitude range $16.5 < r < 23$ with more than $\sim200$ quasars per $\mathrm{deg}^2$ (with $60\ \mathrm{deg}^{-2}$ at $z > 2.1$) [10]. Detailed algorithms for the DESI target selections and spectral reduction are presented in Myers et al. [31], Chaussidon et al. [10], Guy et al. [21].

DESI deploys 5,000 fiber-positioning robots with ten corresponding spectrographs with three arms: blue (B; 3600–5800 Å), red (R; 5760–7620 Å), and near-infrared (Z; 7520–9824 Å). We download the HEALPix-coadded one-dimensional spectra produced by the DESI collaboration with sky subtraction and flux calibration from the SPARCL database¹. The combined wavelength coverage is 3600–9824 Å with resolution $R \sim2000$–5200 (arm-dependent) and a fixed pixel size of 0.8 Å [21, 13].

We begin with the DESI DR1 AGN/QSO Value-Added Catalog², which includes all DR1 spectra classified as extragalactic (Juneau et al., in prep). We use SPECTYPE=QSO to select out targets that the Redrock pipeline (Bailey et al., in prep.) assigned the classification, and use AGN_MASKBITS for objects selected by the machine-learning algorithm QuasarNet and the Mg II doublet post-processing pipeline [10, 12, 13]. This yields a total of 1,717,774 unique quasars with 1,774,723 spectra. Figure 1 shows the distribution of the DESI DR1 quasars in the magnitude–redshift plane. Figure 2 compares the $r$-band magnitude distributions of SDSS DR7Q [45, 48], SDSS DR16Q [27] (Wu & Shen 2022), and DESI DR1Q, illustrating the progressively fainter quasar populations reached by successive large spectroscopic surveys over the past two decades. The SDSS DR7Q distribution peaks near $m_r \sim19$ with a relatively sharp faint-end cutoff toward fainter magnitudes due to the relatively bright, color-selected flux limit of the SDSS-I/II quasar survey [40]. The SDSS DR16Q extends roughly two magnitudes fainter, consistent with the deeper eBOSS quasar target selection [30]. DESI DR1Q reaches the faintest magnitudes of the three, with a distribution extending to $m_r \sim24$, which reflects increasingly sophisticated selection algorithms, larger telescope aperture and/or better spectrograph sensitivity, and the consequence of including high-$z$ Ly $\alpha$-forest quasar targets (Chaussidon et al. [10]; Myers et al. [31]).

Methodology

We measure spectral properties for these quasars following the practice in earlier work (Shen et al. [48, 50]; Wu & Shen [60] (2022)) using the publicly available code PyQSOFit (Guo et al. [18]). Each quasar spectrum was first corrected for the Milky Way extinction using the dust maps (Schlegel et al. [44]; Schlafly & Finkbeiner [43]) and then shifted to the rest frame for spectral analysis. For quasars at redshift $z < 1.0$, we perform the host-galaxy decomposition with priors to remove stellar contamination (Ren et al. [38]). Then the continuum is modeled with a combination of a power law, a third-order polynomial, and Fe II templates, fitted to line-free regions of the spectrum. Emission and absorption lines are then modeled with multiple Gaussian components, allowing for measurements of peak wavelength, line fluxes, luminosities, full width at half maximum (FWHM), equivalent widths (EW), and centroid wavelength. More details on the spectral fitting can be found in Wu & Shen (2022). We adopt a Monte Carlo approach with 10 randomly perturbed spectra to estimate the uncertainties. Figure 3 shows an example of our spectral fits.

Bolometric luminosities are estimated from continuum luminosity at 1350 Å, 3000 Å, and 5100 Å using empirical bolometric corrections BC = 3.81, 5.15, and 9.26, respectively (Richards et al. [41]; Shen et al. [48]). BH masses were estimated from the continuum and broad emission lines using SE virial relations (Vestergaard & Peterson [57]). We adopt the same fiducial BH mass recipes on H$\beta$, Mg II and C IV as in previous work (Shen et al. [48]):

where $(a,b) = (0.91, 0.50)$ for H$\beta$, $(0.74, 0.62)$ for Mg II and $(0.66, 0.53)$ for C IV. For C IV we use the FWHM of the broad profile after rejecting Gaussian components that contribute less than 5% of the total broad-line flux. We include the BH mass estimation based on all three lines (H$\beta$, Mg II, and C IV) and the fiducial BH mass based on the redshift (we adopt the Mg II mass for $0.6 \leq z_{\mathrm{pip}} \leq2.2$ when available, otherwise the H$\beta$ mass for $z_{\mathrm{pip}} \leq0.8$ or the C IV mass for $z_{\mathrm{pip}} \geq1.8$). The fiducial $L_{\mathrm{bol}}$ follows the same redshift rule, using $\nu L_{1350}$, $\nu L_{3000}$, $\nu L_{5100}$ respectively. These fiducial BH masses and bolometric luminosities are used to calculate the Eddington ratio with $L_{\mathrm{bol}}/L_{\mathrm{Edd}} \equiv L_{\mathrm{bol}}/(1.26 \times10^{38} M_{\mathrm{BH}})$. The distributions of the bolometric luminosity and mass are demonstrated in Figure 4. A small discontinuity in the virial BH masses is visible near $z \sim2$, where the fiducial mass estimator transitions from Mg II to C IV. This offset is substantially reduced when the sample is restricted to higher-S/N spectra, suggesting that it may arise from increased emission-line measurement uncertainties for the fainter DESI quasars, potentially combined with a mass-dependent systematic difference between the Mg II- and C IV-based virial estimators at low BH masses.

To derive a systemic redshift from the spectral fit, we follow our earlier work (Shen et al. [49]; Wu & Shen [60] (2022)). Each line included in the fit provides an estimate of the systemic redshift, after accounting for the average intrinsic velocity shifts of different lines and the luminosity dependence of the two high-ionization lines C IV and Si IV. We estimate $z_{\mathrm{sys,line}}$ for each line (broad H$\beta$, [O III] 5007, Ca II, [O II] 3727, Mg II, broad C III], C IV, and Si IV) with a $>3\sigma$ flux detection and spectral coverage $N_{\mathrm{pix}} > N_{\mathrm{pix,max}}/2$. For objects with at least three valid lines, we reject $z_{\mathrm{sys,line}}$ values deviating by more than 3 median absolute deviations from the median, and then restore [O III] 5007, Ca II and [O II] 3727 if they are detected at $>3\sigma$. The adopted $z_{\mathrm{sys}}$ is the inverse-variance-weighted mean of the remaining lines. We then decide between $z_{\mathrm{sys}}$ and the DESI pipeline redshift $z_{\mathrm{pip}}$ (Z_FIT, which uses the Ly $\alpha$ BAO redshift where available). We find that $z_{\mathrm{pip}}$ is generally more reliable than $z_{\mathrm{sys}}$ for low-redshift objects ($z \lesssim1$) and for spectra with very low signal-to-noise ($S/N \lesssim1$). In these regimes, host-galaxy features and/or low-SNR template degeneracies can bias line-centroid–based systemic redshift recipes. For objects with large discrepancies, we adopt the weighted $z_{\mathrm{sys}}$ for objects with a SDSS DR16Q counterpart unless $z_{\mathrm{pip}}$ is closer to the DR16Q systemic redshift, and for objects without a counterpart in DR16Q we use $z_{\mathrm{pip}}$ if a Ly $\alpha$ redshift is available and the weighted $z_{\mathrm{sys}}$ otherwise. The spectral measurements and derived quantities are described in Table 2.

Column	Format	Units	Description
SURVEY	STRING		Survey name
PROGRAM	STRING		DESI program type
HEALPIX	LONG32		Healpix number
TARGETID	LONG64		Unique identifier for each object observed by DESI
Z_AGNCAT	DOUBLE		Redshift provided by the DESI pipeline (AGN/QSO VAC)
Z_LYA	DOUBLE		Redshift from the Ly $\alpha$ Y1 BAO analysis ($-1$ if not available)
OBJID	STRING		Object identifier SURVEY-PROGRAM-HEALPIX-TARGETID
RA	DOUBLE	degree	Right ascension (J2000)
DEC	DOUBLE	degree	Declination (J2000)
IN_SDSS_DR16Q	INT16		1 if cross-matched to SDSS DR16Q within $0.5^{\prime\prime}$, 0 otherwise
SDSS_DR16Q_OBJID	STRING		SDSS DR16Q object identifier (NA if no match)
Z_FIT	DOUBLE		Input redshift for PyQSOFit
Z_SYS	DOUBLE		Final systemic redshift (equals Z_FIT where Z_SYS_ERR = $-$
Z_SYS_ERR	DOUBLE		Uncertainty in Z_SYS ($-1$ if Z_FIT is adopted)
Z_SYS_LINES	DOUBLE[8]		Systemic redshift from individual lines (broad H$\beta$, [O III] 5007, C
Z_SYS_LINES_ERR	DOUBLE[8]		Uncertainties in Z_SYS_LINES
SN_RATIO_CONTI	DOUBLE		Median S/N in the continuum windows
SNR_ALL_MEDIAN	DOUBLE		Median S/N per pixel over the full spectrum
EBV	DOUBLE	mag	Milky Way extinction $E(B-V)$ from Schlegel et al. [44]
CONTI_PARA	DOUBLE[5]		Best-fit continuum parameters (PL normalization, slope, and pol
CONTI_PARA_ERR	DOUBLE[5]		Uncertainties in CONTI_PARA
FE_UV_OPT_PARA	DOUBLE[6]		Best-fit UV and optical Fe II template parameters (normalization
FE_UV_OPT_PARA_ERR	DOUBLE[6]		Uncertainties in FE_UV_OPT_PARA
FE_UV_EW	DOUBLE[2]	Å	Rest-frame EW of UV Fe II at 2250–2650 Å, and its uncertainty
FE_OPT_EW	DOUBLE[2]	Å	Rest-frame EW of optical Fe II at 4434–4684 Å, and its uncertainty
LOGL1350	DOUBLE	[erg s$^{-1}$]	Continuum luminosity $\log\lambda L_{\lambda}$ at rest-frame 1350 Å
LOGL1350_ERR	DOUBLE	[erg s$^{-1}$]	Uncertainty in LOGL1350
LOGL1700	DOUBLE	[erg s$^{-1}$]	Continuum luminosity $\log\lambda L_{\lambda}$ at rest-frame 1700 Å
LOGL1700_ERR	DOUBLE	[erg s$^{-1}$]	Uncertainty in LOGL1700
LOGL2500	DOUBLE	[erg s$^{-1}$]	Continuum luminosity $\log\lambda L_{\lambda}$ at rest-frame 2500 Å
LOGL2500_ERR	DOUBLE	[erg s$^{-1}$]	Uncertainty in LOGL2500
LOGL3000	DOUBLE	[erg s$^{-1}$]	Continuum luminosity $\log\lambda L_{\lambda}$ at rest-frame 3000 Å
LOGL3000_ERR	DOUBLE	[erg s$^{-1}$]	Uncertainty in LOGL3000
LOGL4200	DOUBLE	[erg s$^{-1}$]	Continuum luminosity $\log\lambda L_{\lambda}$ at rest-frame 4200 Å
LOGL4200_ERR	DOUBLE	[erg s$^{-1}$]	Uncertainty in LOGL4200
LOGL5100	DOUBLE	[erg s$^{-1}$]	Continuum luminosity $\log\lambda L_{\lambda}$ at rest-frame 5100 Å
LOGL5100_ERR	DOUBLE	[erg s$^{-1}$]	Uncertainty in LOGL5100
CONTI_FIT_STAT	DOUBLE[2]		Number of pixels and reduced $\chi^2$ of the continuum fit
FRAC_HOST_4200	DOUBLE		Host-galaxy fraction at rest-frame 4200 Å
FRAC_HOST_5100	DOUBLE		Host-galaxy fraction at rest-frame 5100 Å
Column	Format	Units	Description
DN4000	DOUBLE		$D_n$4000 index of the host galaxy
HOST_DECOMP_PARA	DOUBLE[5]		Coefficients of the five eigenspectra used in the host decomposition
		Å, $10^{-17}$ erg s$^{-1}$ cm$^{-2}$, [erg s$^{-1}$], km s$^{-1}$, Å, Å	peak wavelength, flux, $\log L_{\mathrm{line}}$, FWHM, EW, 50% flux centroid wavelength
HALPHA	DOUBLE[6]	···	For the entire H$\alpha$ profile (narrow and broad combined)
HALPHA_BR	DOUBLE[6]	···	For the broad H$\alpha$ component
HALPHA_NA	DOUBLE[6]	···	For the narrow H$\alpha$ component
NII6549	DOUBLE[6]	···	For the narrow [N ii] $\lambda6549$ component
NII6585	DOUBLE[6]	···	···
SII6718	DOUBLE[6]	···	···
SII6732	DOUBLE[6]	···	···
HBETA	DOUBLE[6]	···	For the entire H$\beta$ profile (narrow and broad combined)
HBETA_BR	DOUBLE[6]	···	For the broad H$\beta$ component
HBETA_NA	DOUBLE[6]	···	For the narrow H$\beta$ component ($z < 1.02$ only)
HEII4687	DOUBLE[6]	···	For the entire He ii $\lambda4687$ profile (narrow and broad combined)
HEII4687_BR	DOUBLE[6]	···	For the broad He ii $\lambda4687$ component
OIII5007	DOUBLE[6]	···	For the entire [O iii] $\lambda5007$ profile
OIII5007C	DOUBLE[6]	···	For the core [O iii] $\lambda5007$ profile
OIII4960	DOUBLE[6]	···	For the entire [O iii] $\lambda4960$ profile
OIII4960C	DOUBLE[6]	···	For the core [O iii] $\lambda4960$ profile
HGAMMA	DOUBLE[6]	···	···
HDELTA	DOUBLE[6]	···	···
MGII	DOUBLE[6]	···	For the entire Mg II profile (narrow and broad combined)
MGII_BR	DOUBLE[6]	···	For the broad Mg II component
CIII_BR	DOUBLE[6]	···	For the broad C III] component
SIIII1892	DOUBLE[6]	···	···
ALIII1857	DOUBLE[6]	···	···
NIII1750	DOUBLE[6]	···	···
CIV	DOUBLE[6]	···	···
HEII1640	DOUBLE[6]	···	For the entire He ii $\lambda1640$ profile (narrow and broad combined)
HEII1640_BR	DOUBLE[6]	···	For the broad He ii $\lambda1640$ component
SIIV_OIV	DOUBLE[6]	···	For the 1400 Å complex
OI1304	DOUBLE[6]	···	···
LYA	DOUBLE[6]	···	···
NV1240	DOUBLE[6]	···	···
NEV3346	DOUBLE[6]	···	Local fit (linear continuum + one Gaussian) of [Ne v] $\lambda3346$
NEV3426	DOUBLE[6]	···	Local fit of [Ne v] $\lambda3426$
FEVII3586	DOUBLE[6]	···	Local fit of [Fe vii] $\lambda3586$
OII3728	DOUBLE[6]	···	Local fit of [O ii] $\lambda\lambda3727,3729$
FEVII3759	DOUBLE[6]	···	Local fit of [Fe vii] $\lambda3759$
CAII3934	DOUBLE[6]	···	For the Ca II K absorption line (local fit; negative flux and EW)
FEXIV5303	DOUBLE[6]	···	Local fit of [Fe xiv] $\lambda5303$
FEVII5721	DOUBLE[6]	···	Local fit of [Fe vii] $\lambda5721$
FEVII6087	DOUBLE[6]	···	Local fit of [Fe vii] $\lambda6087$
FEX6374	DOUBLE[6]	···	Local fit of [Fe x] $\lambda6374$
Column	Format	Units	Description
SXII7611	DOUBLE[6]	$\cdots$	Local fit of [S xii] $\lambda7611$
FEXI7892	DOUBLE[6]	$\cdots$	Local fit of [Fe xi] $\lambda7892$
		Å, $10^{-17}$ erg s$^{-1}$ cm$^{-2}$, [erg s^{-1}\], km s^{-1}$, Å, Å | Uncertainties in peak wavelength, flux, $\log L_{\mathrm{line}}$, FWHM, EW, 50% flux centroid wavelength
HALPHA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HALPHA_BR_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HALPHA_NA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
NII6549_ERR	DOUBLE[6]	$\cdots$	$\cdots$
NII6585_ERR	DOUBLE[6]	$\cdots$	$\cdots$
SII6718_ERR	DOUBLE[6]	$\cdots$	$\cdots$
SII6732_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HBETA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HBETA_BR_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HBETA_NA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HEII4687_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HEII4687_BR_ERR	DOUBLE[6]	$\cdots$	$\cdots$
OIII5007_ERR	DOUBLE[6]	$\cdots$	$\cdots$
OIII5007C_ERR	DOUBLE[6]	$\cdots$	$\cdots$
OIII4960_ERR	DOUBLE[6]	$\cdots$	$\cdots$
OIII4960C_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HGAMMA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HDELTA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
MGII_ERR	DOUBLE[6]	$\cdots$	$\cdots$
MGII_BR_ERR	DOUBLE[6]	$\cdots$	$\cdots$
CIII_BR_ERR	DOUBLE[6]	$\cdots$	$\cdots$
SIIII1892_ERR	DOUBLE[6]	$\cdots$	$\cdots$
ALIII1857_ERR	DOUBLE[6]	$\cdots$	$\cdots$
NIII1750_ERR	DOUBLE[6]	$\cdots$	$\cdots$
CIV_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HEII1640_ERR	DOUBLE[6]	$\cdots$	$\cdots$
HEII1640_BR_ERR	DOUBLE[6]	$\cdots$	$\cdots$
SIIV_OIV_ERR	DOUBLE[6]	$\cdots$	$\cdots$
OI1304_ERR	DOUBLE[6]	$\cdots$	$\cdots$
LYA_ERR	DOUBLE[6]	$\cdots$	$\cdots$
NV1240_ERR	DOUBLE[6]	$\cdots$	$\cdots$
NEV3346_ERR	DOUBLE[6]	$\cdots$	$\cdots$
NEV3426_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEVII3586_ERR	DOUBLE[6]	$\cdots$	$\cdots$
OII3728_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEVII3759_ERR	DOUBLE[6]	$\cdots$	$\cdots$
CAII3934_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEXIV5303_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEVII5721_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEVII6087_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEX6374_ERR	DOUBLE[6]	$\cdots$	$\cdots$
Column	Format	Units	Description
SXII7611_ERR	DOUBLE[6]	$\cdots$	$\cdots$
FEXI7892_ERR	DOUBLE[6]	$\cdots$	$\cdots$
Ha_STAT	DOUBLE[2]		Number of pixels, reduced $\chi^2$ of the line-complex fit
Hb_STAT	DOUBLE[2]		$\cdots$
HgHd_STAT	DOUBLE[2]		$\cdots$
MgII_STAT	DOUBLE[2]		$\cdots$
CIII_STAT	DOUBLE[2]		$\cdots$
CIV_STAT	DOUBLE[2]		$\cdots$
SiIV_STAT	DOUBLE[2]		$\cdots$
Lya_STAT	DOUBLE[2]		$\cdots$
NEV3346_STAT	DOUBLE[2]		$\cdots$
NeV_STAT	DOUBLE[2]		$\cdots$
FEVII3586_STAT	DOUBLE[2]		$\cdots$
OII_STAT	DOUBLE[2]		$\cdots$
FEVII3759_STAT	DOUBLE[2]		$\cdots$
CaII_STAT	DOUBLE[2]		$\cdots$
FEXIV5303_STAT	DOUBLE[2]		$\cdots$
FEVII5721_STAT	DOUBLE[2]		$\cdots$
FEVII6087_STAT	DOUBLE[2]		$\cdots$
FEX6374_STAT	DOUBLE[2]		$\cdots$
SXII7611_STAT	DOUBLE[2]		$\cdots$
FEXI7892_STAT	DOUBLE[2]		$\cdots$
LOGLBOL	DOUBLE	$[\mathrm{erg}\ \mathrm{s}^{-1}]$	Bolometric luminosity
LOGLBOL_ERR	DOUBLE	$[\mathrm{erg}\ \mathrm{s}^{-1}]$	Uncertainty in LOGLBOL
LOGMBH_HB	DOUBLE	$[M_{\odot}]$	Single-epoch BH mass based on $H\beta$
LOGMBH_HB_ERR	DOUBLE	$\cdots$	Uncertainty in LOGMBH_HB
LOGMBH_MGII	DOUBLE	$\cdots$	Single-epoch BH mass based on Mg II
LOGMBH_MGII_ERR	DOUBLE	$\cdots$	$\cdots$
LOGMBH_CIV	DOUBLE	$\cdots$	Single-epoch BH mass based on C IV
LOGMBH_CIV_ERR	DOUBLE	$\cdots$	$\cdots$
LOGMBH	DOUBLE	$\cdots$	Fiducial single-epoch BH mass
LOGMBH_ERR	DOUBLE	$\cdots$	$\cdots$
LOGLLEDD_RATIO	DOUBLE		Eddington ratio based on the fiducial BH mass
LOGLLEDD_RATIO_ERR	DOUBLE		Uncertainty in LOGLLEDD_RATIO
BAL_FLAG	INT16		BAL flag from C IV BI: 1 = BAL (BI$> 0$), 0 = non-BAL, $-1$ =

Sample properties

Improved systemic redshifts

DESI spectra are processed by the Redrock pipeline, which assigns a spectroscopic classification and redshift to each target (Guy et al. [21]; DESI Collaboration et al. [13]). For quasar candidates, DESI further incorporates the QuasarNet classifier/redshift estimator and an Mg II-based recipe to flag cases where the initial Redrock classification/redshift may be ambiguous or affected by template degeneracies to improve the robustness of the adopted quasar redshifts (Chaussidon (et al. 2023; Alexander et al. 2023). For our spectral fits, we also adopt the updated redshifts from the DESI Year 1 Ly $\alpha$ BAO pipeline whenever available [15, 28]. Overall, the pipeline redshifts perform well for the majority of DESI DR1 quasars. However, redshift inference can be affected by the diversity of quasar spectral properties and by systematic velocity shifts of broad emission lines, especially in regimes where the observed spectrum is dominated by rest-frame UV features. As shown in Figure 6, we compare the composite spectra constructed using the pipeline redshift ($z_{\rm pip}$) and the improved systemic redshift ($z_{\rm sys}$) for a random 10k subset. For objects with small velocity discrepancies, the two stacks are nearly identical, indicating that the pipeline redshifts are highly reliable. As the redshift discrepancy increases, systematic differences become more apparent. The $z_{\rm sys}$ composite spectra exhibit narrower and sharper line profiles compared to those constructed using $z_{\rm pip}$. In particular, high-ionization lines such as C IV are known to exhibit luminosity-dependent blueshifts relative to their rest-frame wavelengths [11, 49], which the $z_{\rm sys}$-based stacks clearly recover. This suggests that the pipeline redshift determination, which relies on the observed line peaks, can be biased by these blueshifted features.

which relies on the observed line peaks, can be biased by these blueshifted features. In addition, objects with large velocity discrepancies display absorption features and asymmetric line profiles, indicating the presence of broad absorption line quasars in this sample. In such cases, the pipeline is more prone to misidentifying the redshift due to partially absorbed emission profiles. Our improved systemic redshifts mitigate such biases by anchoring the redshift to both the emission line features and the quasar luminosity, resulting in better alignment of emission-line centroids across the stacked spectra. We compare $z_{\rm sys}$ to $z_{\rm pip}$ across the full quasar sample, grouped by their velocity difference $v_z$ in Figure 5. The two redshifts agree well across the full redshift range. The majority of objects exhibit small velocity differences: 1,572,196 quasars (88.6%) have $|v_z| \leq500$ km s$^{-1}$, while an additional 149,607 objects (8.4%) fall within $500 < |v_z| \leq1000$ km s$^{-1}$. Only 38,431 quasars (2.2%) show offsets in the range $1000 < |v_z| \leq1500$ km s$^{-1}$, and a small fraction of 14,489 objects (0.8%) exhibit larger discrepancies with $|v_z| > 1500$ km s$^{-1}$. The overall consistency of the two redshift distributions indicates that the DESI pipeline redshifts are generally robust for quasar redshift characterization. The remaining discrepancies likely arise from known challenges in quasar redshift determination, particularly at redshifts where only broad rest-frame UV emission lines are available.

Coronal lines

Coronal lines (CLs) are forbidden transitions from highly ionized species with ionization potentials of $\mathrm{IP}\gtrsim100$ eV. Producing these ions requires an intense extreme-UV or soft X-ray radiation field, making CLs sensitive probes of the hardest portion of the ionizing spectral energy distribution (SED) [14, 39]. Such hard ionizing continua are difficult to produce in normal stellar populations, and strong CL emission is therefore most commonly associated with active galactic nucleus (AGN) activity [39, 17]. High-ionization transitions such as [Ne V] have been successfully used to identify accreting SMBHs in galaxies for which conventional optical AGN signatures are weak or obscured [17, 16, 33, 1]. Alternatively, some studies indicate that fast radiative shocks between high-velocity clouds and the ISM may also contribute to CL production [3, 32].

We search for and fit a set of optical coronal lines (CLs; Table 1) in the DESI quasar spectra over the redshift ranges where each transition is covered. Each line is fitted locally within a $\pm5000\ \mathrm{km\ s^{-1}}$ window centered on its expected wavelength using a linear continuum plus a single Gaussian profile. We classify a line as detected when the fitting window is fully covered, the Gaussian model is preferred over a continuum-only model by $\Delta\mathrm{BIC}>10$, the line has $\mathrm{S/N}\geq3$, and none of the fitted parameters reaches its imposed boundary. Figure 7 presents several CL-detected examples.

Line	$\lambda$ (Å)	$\chi$ (eV)	$N_{\mathrm{fit}}$	$N_{\mathrm{det}}$	$f_{\mathrm{det}}$ (%)
[Ne V]	3346.82	97.1	1,079,952	9,660	0.89
[Ne V]	3426.85	97.1	1,019,083	96,591	9.48
[Fe VII]	3587.34	99.1	897,370	1,606	0.18
[Fe VII]	3759.99	99.1	778,761	3,846	0.49
[Fe XIV]	5304.34	361.0	199,516	261	0.13
[Fe VII]	5722.29	99.1	139,877	4,240	3.03
[Fe VII]	6088.69	99.1	107,900	7,416	6.87
[Fe X]	6376.26	233.6	90,548	332	0.37
[S XII]	7613.10	504.8	28,436	60	0.21
[Fe XI]	7893.97	262.1	18,459	810	4.39

Systematic CL measurements provide several opportunities for quasar studies. Their high ionization potentials make them powerful diagnostics of the EUV/soft-X-ray continuum, which is otherwise poorly constrained observationally. Ratios involving [Ne V] and lower-ionization lines can also improve AGN identification at redshifts where standard optical diagnostics are inaccessible. In addition, CL widths and velocity shifts provide complementary tracers of highly ionized outflows. Combined with broad-line properties, BH mass, and Eddington ratio, these measurements can help connect the BLR, CLR, and NLR to the underlying accretion state.

Line ratios

Emission-line ratios have long served as powerful probes of the ionizing SED in AGNs [22], most notably through the narrow-line BPT diagram [6]. These narrow-line diagnostics become difficult to apply at high redshift, where aperture effects and the evolving ionization state of star-forming galaxies blur the classification boundaries that work well locally [23, 52, 4, 46]. Broad emission lines (BELs) and CLs offer a different route, providing more direct constraints on the incident ionizing continuum than narrow lines alone, since the BLR reprocesses the central ionizing radiation and responds sensitively to continuum shape and intensity [5, 24, 29], and the long history of Eigenvector 1 studies demonstrates that the line strength between the optical Fe II and H$\beta$ already traces accretion rate reasonably well ([8, 54, 47]).

Motivated by this, we test whether a broader set of optical and UV line ratios, spanning a range of ionization potentials, traces $L/L_{\mathrm{Edd}}$ coherently across different line pairs. Figure 8 shows the emission-line ratios as a function of $L/L_{\mathrm{Edd}}$, with the sample divided into $M_{\mathrm{BH}}$ quintiles to help disentangle accretion-rate trends from BH-mass dependence. The three rows present, respectively, the optical diagnostic ratios, UV BEL ratios, and CL-related ratios. The top row contains the classic BPT narrow-line ratios [N II] 6585/H$\alpha$ and [O III] 5007/H$\beta$ ([6, 56, 22]), [O II] 3727/H$\gamma$, and $R_{\mathrm{FeII}}$, the Eigenvector 1 ratio of optical Fe II to broad H$\beta$ ([8, 54, 47]). Compared with the total (broad plus narrow) Balmer flux, [N II] 6585/H$\alpha$ and [O III] 5007/H$\beta$ all decline by roughly $\sim0.5$ dex toward higher $L/L_{\mathrm{Edd}}$, while $R_{\mathrm{FeII}}$ rises modestly, the same Eigenvector 1 behaviour in which [O III] 5007 weakens, and Fe II strengthens with increasing accretion rate ([8, 47]). Replacing the total Balmer flux with the narrow component removes most of the decline: [N II] 6585/H$\alpha_{\mathrm{na}}$ and [O III] 5007/H$\beta_{\mathrm{na}}$ remain near $-0.1$ and $0.55$, respectively, with only a weak decrease at the highest $L/L_{\mathrm{Edd}}$, in the same sense as the $L/L_{\mathrm{Edd}}$ dependence of [N II] 6585/H$\alpha$ reported for X-ray selected AGN ((Oh et al. 2017)) and for low-redshift Type 1 AGN ([53]). The decline of the total-Balmer ratios is therefore driven mainly by the growing broad-line component, with only a modest change in the narrow-line ionization state. The middle row compares BELs that arise in the same region under comparable physical conditions, not split between the NLR and BLR. C IV/C III] and C IV/Si IV increase by $\sim0.3-0.5$ dex and N V/C IV rises by $\sim0.5$ dex with $L/L_{\mathrm{Edd}}$, whereas C IV /He II 1640 is essentially flat. At higher accretion rates, C IV is known to weaken in line strength and show stronger blueshifts ([7, 42, 11]). Temple et al. [55] show that quasars with large C IV blueshifts have on average stronger N V/C IV and (Si IV +O IV])/C IV ratios, and once the C IV blueshift is accounted for, these ratios show little dependence on Eddington ratio, suggesting that the line ratios are more closely tied to the outflow than to the accretion rate itself. These UV line ratios are also sensitive to the BLR photoionization conditions, including the ionization parameter, gas density, and incident SED, so their trends likely reflect a combination of outflow structure and changing BLR physical conditions. The lack of a significant trend in C IV /He II 1640 suggests that the hardness of the ionizing continuum near 50–55 eV changes little over this range, so the trends are more likely governed by the gas conditions than by the SED shape alone. In the bottom row with CLs, [Ne V] /H$\gamma$ and [Ne V] /H$\beta$ decline with $L/L_{\mathrm{Edd}}$ by $\sim0.6-0.7$ dex when compared to the total Balmer lines, and [Fe VII] $\lambda$6087/H$\beta$ shows the same behaviour, matching the optical row, while the [Fe VII] /[O III] 5007 ratio is flat. The ratios between the CLs and the narrow Balmer lines (dotted curves) are also flat, indicating that the CL gas requiring photon energies of $\approx$100 eV or more ([Ne V]: 97 eV; [Fe VII]: 99 eV) tracks the NLR with no measurable change in the soft X-ray/EUV continuum hardness relative to the optical lines across our range. This is consistent with the lack of significant correlations between CL properties and Eddington ratio found in SDSS galaxies ([36, 37]), although that comparison is limited by detection incompleteness. Finally, the dependence on $M_{\mathrm{BH}}$ is consistently weaker than the dependence on $L/L_{\mathrm{Edd}}$ across these panels, and the sample is detection-limited at the lowest and highest accretion rates, and a significant CL detection requires strong lines in high-S/N spectra. Taken together, these ratios are sensitive to the ionizing SED in the EUV and soft X-ray regime, which cannot be observed directly for the vast majority of quasars. These additional line ratios may serve as a proxy for the accretion state of the central SMBH where a reliable single-epoch mass, and hence $L/L_{\mathrm{Edd}}$, is unavailable, including at high redshift, where only rest-frame UV lines fall within typical spectroscopic coverage. Reproducing the simultaneous trends across multi-

ple, independent ratios with physically motivated AGN SEDs and photoionization modeling would place a considerably stronger constraint on the ionizing continuum shape than any single line ratio can provide in isolation [9] (Wu et al. 2025).

Comparison with SDSS Quasar Catalogs

We cross-match our DESI DR1 quasars to the SDSS DR16Q (Wu & Shen 2022) using a positional matching radius of 0.5 arcsec. The matched sample contains 347,201 unique quasars, corresponding to 45.6% of the SDSS DR16Q quasars and 20.2% of the DESI DR1 catalog. Figure 9 presents a comparison of the $z$ and $M_{\mathrm{BH}}$ estimates derived from the two surveys. The left panel shows the comparison of spectroscopic redshifts $z_{\mathrm{sys}}$ from both surveys. The vast majority of sources lie close to the one-to-one relation, demonstrating good overall consistency between the DESI and SDSS measurements. Only a small fraction of objects ($\lesssim0.2\%$) exhibit significant discrepancies with velocity offsets $|v_z| > 5000$ km s$^{-1}$. Most of these outliers occur in the redshift range $0.7 \lesssim z \lesssim2.8$, where redshift determination relies primarily on broad rest-frame UV emission lines such as C IV and Mg II. In this regime, misidentification of emission lines, template degeneracies in automated pipelines, or intrinsic velocity shifts of high-ionization lines can lead to systematic redshift differences. Our visual inspection of these sources suggests that DESI generally provides more reliable redshift estimates, potentially owing to higher spectral quality in these cases and the use of machine-learning-assisted quasar redshift validation [10, 2]. In contrast, such discrepancies are uncommon at $z \lesssim1$, where narrow emission lines and host-galaxy stellar absorption features provide stronger constraints on the systemic redshift. The middle and right panels compare the $M_{\mathrm{BH}}$ measurements. The DESI masses broadly agree with those measured from SDSS spectra.

We test the internal consistency of our single-epoch BH mass estimates by comparing masses derived from different broad emission lines covered in the same spectrum (Figure 10). For reference, we show the corresponding distributions for SDSS DR16Q (Wu & Shen 2022) and SDSS DR7Q [48]. In each catalog, we include all quasars with both relevant mass estimates rather than restricting the comparison to objects observed by multiple surveys. For the DESI sample, host-galaxy contamination is additionally mitigated using the spectral decomposition method of Ren et al. [38], which is particularly important for the optical continuum and $H\beta$-based mass estimates at low redshift. Overall, the Mg II-based masses are consistent with those derived from $H\beta$ and C IV, with the distributions centered close to the one-to-one relation. Some differences among the three catalogs are expected from the spectral-fitting methodology. In particular, SDSS DR7Q adopted local fits around individual broad emission lines. In contrast, DESI DR1 and SDSS DR16Q use global spectral fits, which can introduce small systematic differences in the measured BEL FWHM and fluxes. These offsets are nevertheless modest compared with the intrinsic systematic uncertainty of single-epoch virial BH mass estimators, which is typically of $\sim0.5$ dex, and additional uncertainty arises when recipes calibrated against local reverberation-mapped AGNs are extrapolated to luminous, distant quasars [51]. The DESI DR1 distributions also show broader tails toward large discrepancies between the two mass estimates than those of DR16Q, most notably a tail of objects whose C IV-based masses are much lower than their Mg II-based masses. Restricting the DESI sample to higher-S/N spectra substantially reduces these tails, as the objects in the tail are fainter and have lower S/N. Stacked spectra of these objects show a systematically narrower C IV profile than those of the rest of the sample. We find that they lie at the low-luminosity end of the sample, and their C IV profiles look noticeably narrower, possibly because of a prominent narrow C IV component. The single-epoch mass recipe may therefore be unreliable in this regime.

The substantial overlap between the SDSS and DESI DR1 quasar samples provides a valuable opportunity to investigate quasar spectroscopic variability on a statistical basis. By presenting spectral measurements from both surveys within a consistent analysis framework, our data products enable systematic time-domain studies of quasars across a large, uniformly characterized sample. Spectroscopic variability in quasars offers important insights into the structure and physics of the accretion flow and the BLR. For example, continuum variability can drive the “breathing” effect of the BLR, in which emission-line widths and strengths respond to changes in luminosity [58] (Korista & Goad 2004). Large spectroscopic surveys have also identified changing-look AGNs, where broad emission lines appear or disappear on timescales of years, likely reflecting significant changes in the accretion state or BLR structure [19, 20, 63]. The multi-year baseline between SDSS and DESI is well suited for identifying such long-timescale spectral variability. Moreover, when combined with ongoing photometric surveys such as LSST and Euclid, these datasets provide spectroscopic constraints that serve as a foundation for future studies. dation for large statistical studies of quasar variability in the time domain.

Conclusion

In this work, we present a comprehensive spectroscopic analysis of the DESI DR1 quasar sample and construct a value-added catalog of quasar spectral properties based on uniform measurements of continuum and line emission. By applying a consistent fitting framework, we derive emission-line parameters, BH mass estimates, and improved systemic redshifts across a large and diverse quasar population. For 11.4% of the spectra, our systemic redshifts deviate from the DESI pipeline redshifts by more than 500 $\mathrm{km}\,\mathrm{s}^{-1}$ and mitigate the biases associated with the blueshifts of high-ionization lines at $z > 1$. Comparisons with SDSS quasar catalogs demonstrate overall consistency. We also provide local fits to ten coronal lines, with [Ne V] $\lambda3426$ detected in 9.5% of the spectra with coverage, and we show that some of the ratios of coronal, narrow, and broad lines vary systematically with the Eddington ratio.

The large sample size and homogeneous measurements provided by this catalog open new opportunities for statistical studies of quasar physics, including investigations of BLR structure, accretion properties, and emission-line diversity. Furthermore, the substantial overlap with previous spectroscopic surveys and synergy with ongoing and future time-domain and photometric programs make this dataset a valuable resource for studies of quasar variability and evolution. This VAC is designed to serve as a foundation for a wide range of applications, from precision redshift measurements to physically motivated modeling of AGN emission, and will be further expanded and refined in future DESI data releases. All products from this work, including the catalog, input fitting parameters, individual FITS outputs, Python notebooks, and update notes can be accessed from our data server ³ and GitHub repository ⁴.

Appendix