arXiv subjects

D. Kirkby

Publications and source records attributed to D. Kirkby.

At least 19 recordsLinked to original sources

arXiv2026

Sizing the Universe with DESI Galaxy Sizes: Plain Fundamentals of Fundamental-Plane Lensing

Weak gravitational lensing provides a powerful way to map cosmic structure, but most current measurements rely on galaxy shape distortions from deep imaging surveys and are affected by systematics such as intrinsic alignments, photometric-redshift uncertainties and shape-measurement biases. Here we present a spectroscopic galaxy-galaxy lensing magnification measurement using Fundamental-Plane (FP) size residuals, $\delta_r\equiv\Delta\log_{10}R_\mathrm{e}$, from 3.26 million DESI luminous red galaxy (LRG) sources behind DESI Bright Galaxy Survey lenses. The FP-like relation predicts the intrinsic sizes of LRGs from lensing-invariant quantities, including velocity dispersion $\sigma_0$ and surface brightness $I_\mathrm{e}$, with a scatter of about 0.06-0.07 dex. Lensing magnifies LRG sizes, giving the direct convergence response $\delta_r(\kappa)=\kappa/\ln 10$, but we show that the full lensing response is modified by magnification bias, because fitting $R_\mathrm{e}$ with $I_\mathrm{e}$ inevitably induces a magnitude dependence in $\bar{\delta}_r(m)$. After calibrating this response, we recover surface-density profiles with uncertainties comparable to those from individual Stage-III shear surveys using 5-20 million higher-redshift sources. The corresponding excess surface-density profiles agree with shear-based measurements. We further show that the estimator is robust to size-measurement uncertainties, with a convergence multiplicative bias only $\simeq -0.2$ times the size bias. FP lensing therefore provides a clean, spectroscopic and complementary probe of cosmic structure.

astro-ph.CO↗

arXiv2026

A Finely-Binned Measurement of the Connected Even-Parity Galaxy 4-Point Correlation Function of DESI Year 1 Luminous Red Galaxies

We present the connected even-parity galaxy 4-Point Correlation Function (4PCF) of DESI Year 1 (Y1) Luminous Red Galaxies (LRGs). This is one of the first measurements of the connected 4PCF on data, and shows a clear detection at $\sim$12 to 17$\sigma$ in the full-sky analysis. We test the robustness of the signal by varying three factors: hemisphere (north vs. south), redshift range (either the full range $0.4 < z < 1.1$ or the higher-number-density interval $0.4 < z < 0.8$), and sample completeness (the full sample vs. only the most complete regions within the number density-motivated cut). Finally, we cross-correlate different patches that are spatially well-separated, and thus should have largely independent noise. This approach, at leading order, is able to remove mismatch between the covariance matrix of the data and that of the mocks; however, in certain cases at the cost of sensitivity. We find $\sim$15$\sigma$ evidence for an even-parity 4PCF in this approach. Overall, the studies of the 4PCF presented here probe sensitively any mismatch between mock catalogs and data, and also open up the prospect of fitting a model, constraining cosmological parameters and galaxy biases, and searching for Baryon Acoustic Oscillation (BAO) features.

astro-ph.CO↗

arXiv2026

A Unified Tracer Analysis of DESI DR2 Baryon Acoustic Oscillations

We improve upon previous efforts to optimally combine overlapping galaxy samples in the DESI baryon acoustic oscillation analysis. By weighting each galaxy by its linear bias, overlapping galaxies are combined into a single, unified catalog, naturally avoiding double counting of cosmic volume and including all auto- and cross- information at the catalog level. Improvements over the previous effort include the addition of QSO out to $z=1.6$ to account for all overlapping DR2 tracers and redshift-dependent bias treatment to improve reconstruction. We report distance measurements using this unified tracer, and find them to be highly consistent with the baseline DR2 BAO analysis. We also test for tracer-dependent systematics within the DESI data, and find no evidence of tracer-dependent systematics within $0.8<z<1.6$. Finally, we take advantage of the unified tracer to rebin the analysis in redshift in order to more finely resolve the redshift-to-distance relation. Dynamical dark energy results on this finer redshift binning indicate that there is no missed feature in the expansion history in the redshifts $0.8<z<1.6$, and reproduces DESI's preference for an evolving dark energy equation of state.

astro-ph.CO↗

arXiv2026

Spatially-Resolved Spectra of Diffuse Galactic Light using 10.8 M DESI Sky Fibers

Using 10.8 million ``blank'' sky spectra from the DESI Year 3 dataset, we measure the diffuse galactic light (DGL) spectrum in the optical at spectral resolution $R \sim 4000$ by correlating with far-infrared emission from IRAS. Subdividing the sky into 54 deg$^2$ pixels (HEALPix, NSIDE = 8), we map the variation of the DGL correlation spectrum and nebular emission lines across the DESI footprint in the high-Galactic-latitude sky. The increased data volume over previous SDSS-based analyses enables several new detections in the DGL, including scattering both onto and out of the line of sight from neutral interstellar sodium and potassium. We further detect direct emission from ro-vibrational transitions of molecular hydrogen in the near-infrared with an absolute radiance of $0.65\substack{+0.13 \\ -0.12}\times10^{-9}~{\rm erg\, cm^{-2}\,s^{-1}\,sr^{-1}}$, roughly consistent with theoretical expectations, but with an apparent ortho-to-para line ratio that is lower by a factor of $0.60\substack{+0.28 \\ -0.27}$. We confirm previous detections of extended red emission (ERE) in the DGL and map its spatial variation. Our spatially resolved DGL maps provide important observational constraints for the radiative transfer efforts that are now possible with recent 3D models of the Milky Way.

astro-ph.GA↗

arXiv2026

Improved constraints on the Milky Way potential using the M68 stream and DESI spectroscopic data

We present a selection of stars belonging to the stellar stream of the M68 (NGC 4590) globular cluster, also known as Fj\"orm. This star selection is an improvement on previous ones that used only Gaia data, as it incorporates spectroscopic measurements from the DESI survey and photometric data from the DESI Legacy Surveys. The selection contains 96 stars, each with five phase-space parameters from Gaia-DR3 and radial velocity from DESI, covering the entire observed section of the stream. This constitutes the largest selection of M68 stream stars with measured radial velocities to date. The observed stream is wider than expected from N-body simulations, and the stars farthest from the centre of the stream appear to be correlated in radial velocity space. This suggests that these stars cannot have been stripped from the cluster in a static axisymmetric potential. By modelling a mock sample of stream stars created using an N-body simulation, we found that we could reliably constrain the disc mass $M_{\rm d}$ and the dark matter halo axis ratio $q_{\rm h}$ of the Milky Way. This is because the stream flows close to and almost parallel to the disc. Using the 44 stars that are consistent with having been stripped from the cluster, combined with measurements of the Milky Way's rotation curve, we constrain the Galactic potential, obtaining $M_{\rm d} = 5.34 \pm 0.57 \times 10^{10}$ M$_{\rm sun}$ and an oblate halo of $q_{\rm h} = 0.83^{+0.06}_{-0.05}$. Additionally, by fitting the stream track, we estimate the Heliocentric distance of M68 to be $r=10.55\pm0.09$ kpc.

astro-ph.GA↗

arXiv2026

Lyman Break Galaxy selection and redshift measurement with supervised contrastive learning

Some of the next steps for high-precision cosmology lie within the high-redshift, high-density universe. Spectroscopic survey experiments such as the Dark Energy Spectroscopic Instrument (DESI)'s second phase DESI Run 2 will shift towards probing Lyman Break Galaxy (LBG) populations from z$\sim$2 to z$\sim$4.5. For this faint sample, spectroscopic redshift measurement and sample decontamination remains a challenge, even after target selection. We propose an approach based on supervised weighted contrastive learning, in order to both learn a redshift representation for spectra and decontaminate the sample from quasars and low redshift emission line galaxies. This strategy generalizes the contrastive learning loss approach with continuous relationship weights, such that the network simultaneously learns redshift and classification tasks. The model shows stronger outlier classification and comparable redshift identification performances when compared to the previous network used for DESI (a modified version of QuasarNET) on the same dataset. In particular, contrastive learning is well suited to the small, visually-inspected sample used for training and testing, especially given the multi-task nature of this work.

astro-ph.CO↗

arXiv2026

DESI DR2 Results IV: Alcock-Paczy\'nski Measurements from the Lyman Alpha Forest and Cosmological Constraints

We present Alcock-Paczy\'nski (AP) measurements from the full shape of Lyman-$\alpha$ (Ly$\alpha$) forest correlation functions measured from the second data release (DR2) of the Dark Energy Spectroscopic Instrument (DESI). Our measurements include information from the Ly$\alpha$ forest auto-correlation and its cross-correlation with quasars. We constrain the AP effect with $1\%$ precision at an effective redshift $z_\mathrm{eff}=2.33$, which is twice as tight as the Baryon Acoustic Oscillation (BAO) constraint from the same data. When using the joint Ly$\alpha$ AP and BAO results, we measure the ratios $D_\text{H}(z_\mathrm{eff})/r_\text{d}=8.600 \pm 0.066$ and $D_\text{M}(z_\mathrm{eff})/r_\text{d}=39.32 \pm 0.33$, where $D_\text{M}$ is the transverse comoving distance, $D_\text{H}$ is the Hubble distance, and $r_\text{d}$ is the sound horizon at the drag epoch. Assuming $\Lambda$CDM, Ly$\alpha$ forest measurements combined with a nucleosynthesis prior produce a constraint on the Hubble constant $H_0=66.5\pm1.3\,\mathrm{km\,s^{-1}\,Mpc^{-1}}$. The Ly$\alpha$ AP result corresponds to a matter fraction constraint $\Omega_\text{m}=0.325\pm0.018$ in $\Lambda$CDM, which is $1.4\sigma$ higher than DESI BAO. This impacts the DESI results relative to the Cosmic Microwave Background (CMB), slightly reducing their discrepancy from $2.4\sigma$ to $2.2\sigma$. We present updated constraints on extended models using the joint DESI DR2 BAO and Ly$\alpha$ forest full shape data, together with external data sets. When considering a time-evolving dark energy equation of state parametrized by $w_0$ and $w_a$, we find it is preferred over $\Lambda$CDM at $2.7\sigma$ for the combination of DESI and CMB data, and at $3.2\sigma$ when also including supernovae. With the new Ly$\alpha$ AP measurement, DESI provides its most precise anchor for the expansion history at $z > 1$ in the matter-dominated Universe.

astro-ph.CO↗

arXiv2026

Validation of the DESI DR2 Ly$\alpha$ forest full-shape analysis

We present the validation of the Dark Energy Spectroscopic Instrument (DESI) Data Release 2 (DR2) Lyman-$\alpha$ (Ly$\alpha$) forest full-shape analysis. This analysis combines three-dimensional Ly$\alpha$ forest auto-correlations and cross-correlations with quasars to extract information from both the baryon acoustic oscillation (BAO) feature and the broadband clustering signal, with primary emphasis on the Alcock-Paczynski (AP) measurement. Compared to the DESI DR1 analysis, the DR2 validation uses substantially larger and more realistic mock datasets, including CoLoRe 2LPT and AbacusSummit Ly$\alpha$ forest simulations. The modeling framework is also improved through analytic marginalization over small scales ($<10$ $h^{-1}$Mpc) and the impact of ultraviolet background fluctuations. The validation program was completed prior to unblinding and defines quantitative requirements for the cosmological parameters of interest, which are evaluated using hundreds of mock realizations. We further test the analysis through independent fits to the auto- and cross-correlations, multiple catalog splits, and a broad suite of analysis and modeling variations applied to both mocks and blinded observational data. We find that the BAO and AP parameters satisfy all validation requirements and remain stable across all tests. In contrast, mock studies reveal a significant bias in the inferred growth-rate parameter $f\sigma_8$, leading us to exclude this measurement from the final analysis. The consistency across mocks, data splits, and robustness tests demonstrates that the DR2 Ly$\alpha$ full-shape analysis provides a reliable and substantially improved broadband AP measurement over previous Ly$\alpha$ forest studies.

astro-ph.CO↗

arXiv2026

CoLoRe-2LPT: Lyman-$\alpha$ mock catalogues for the validation of DESI cosmological analyses

The Lyman-$\alpha$ (Ly$\alpha$) forest has become a crucial probe for studying the large-scale structure of the universe at high redshift ($z > 2$), providing powerful constraints on Baryon Acoustic Oscillations (BAO) and the full-shape (FS) clustering of matter. As a key ingredient for upcoming BAO and FS analyses, we present a new generation of fast cosmological Ly$\alpha$ mocks based on second-order Lagrangian perturbation theory (2LPT). These new mocks significantly improve upon previous log-normal approaches, both at accurately capturing small scale clustering and at recovering the non-linear broadening of the BAO peak. They are able to reproduce Ly$\alpha$ statistics within $10\%$ of the latest DESI measurement; including the Ly$\alpha$ bias and the redshift-space distortion $\beta$ parameter, mean transmitted flux, and 1D power spectrum. The corresponding quasar (QSO) clustering is also improved with respect to previous approaches, calibrated against high-resolution Abacus simulations, recovering the observational QSO linear bias to less than $5\%$ and improving redshift-space distortions via 2LPT velocities and the addition of Fingers-of-God effects. Furthermore, these mocks incorporate high column density systems and metal lines, allowing us to explore the effects and systematics induced by these astrophysical contaminants. This new set of mocks has been key for enhancing the modeling and validation of the DESI DR2 Ly$\alpha$ full shape cosmological analysis. This work provides a physically motivated and computationally efficient tool for simulating current and next-generation Ly$\alpha$ surveys and validating FS and BAO analysis.

astro-ph.CO↗

arXiv2026

Lyman-$\alpha$ forest holography: 3D predictions from 1D measurements

Cosmological analyses of Lyman-$\alpha$ forest clustering rely on either one-dimensional correlations along individual sightlines or three-dimensional correlations between different sightlines. Because these observables probe the matter distribution on very different scales, they have traditionally been analyzed independently. In this work, we bridge this gap using ForestFlow, an emulator trained on a suite of cosmological hydrodynamical simulations that provides a unified description of Lyman-$\alpha$ forest clustering from linear to nonlinear scales. This framework enables us to determine the range of three-dimensional clustering models compatible with the DESI one-dimensional flux power spectrum ($P_{\rm 1D}$). The resulting predictions successfully reproduce the large-scale clustering measured by the DESI BAO analysis and provide physically motivated priors on nonlinear clustering that are used in a companion paper presenting the full-shape analysis of the DESI DR2 Lyman-$\alpha$ forest. We validate our methodology using the large-volume, high-resolution hydrodynamical simulation ACCEL-2, demonstrating excellent agreement across the full range of scales considered. Finally, we combine constraints from the $P_{\rm 1D}$ and BAO analyses on the parameter combinations $b_\delta \sigma_8$ and $b_\eta f \sigma_8$, finding that the two probes provide comparable constraining power while exhibiting complementary parameter degeneracies. Our results establish a direct connection between one- and three-dimensional Lyman-$\alpha$ forest measurements through ForestFlow, an approach we term Lyman-$\alpha$ holography by analogy with the reconstruction of higher-dimensional structure from lower-dimensional information.

astro-ph.CO↗

arXiv2026

Probing the matter-dominated expansion with multi-redshift Lyman-$\alpha$ BAO from DESI DR2

We present a multi-redshift Baryon Acoustic Oscillations (BAO) analysis of the DESI Data Release 2 (DR2) Lyman-$\alpha$ (Ly$\alpha$) forest, splitting the forest auto-correlation and its cross-correlation with quasars into three redshift bins. We obtain BAO measurements at effective redshifts $z_{\rm eff} = 2.13$, $2.40$, and $2.81$ with $\sim2.0$--$2.5\%$ precision per bin in the radial and transverse directions, corresponding to $\sim1.1$--$1.2\%$ precision for the isotropic BAO measurement. Using the same data products and modeling framework as the DESI DR2 Ly$\alpha$ BAO analysis, we validate the pipeline on $400$ synthetic datasets and find unbiased BAO recovery with well-calibrated uncertainties. The measurements show an increase in the isotropic dilation parameter $D_V/r_d$ from $30.26\pm0.39$ to $32.22\pm0.47$ and in the Alcock-Paczy\'nski parameter $D_M/D_H$ from $3.96\pm0.15$ to $5.63^{+0.22}_{-0.24}$. The Hubble distance $D_H/r_d$ decreases from $9.40\pm0.20$ to $7.22\pm0.17$, providing a direct measurement of the expansion history consistent with $\Lambda$CDM and the expected matter-dominated scaling, with $H(z)\propto(1+z)^n$ giving $n=1.34\pm0.16$. The redshift split also provides a self-consistent measurement of clustering evolution: the Ly$\alpha$ forest bias evolves as $(1+z)^\gamma$ with $\gamma_\alpha=3.05\pm0.16$, the RSD parameter has a redshift evolution described by $\gamma_\beta=-0.97\pm0.26$, and the quasar bias evolves with $\gamma_Q=1.56\pm0.23$, consistent with independent quasar clustering measurements. Combining these three-bin BAO measurements with DESI DR2 galaxy and quasar BAO measurements yields cosmological constraints consistent with the single-bin Ly$\alpha$ BAO analysis in flat $\Lambda$CDM and $w_0w_a$CDM and improves curvature constraints by $\sim12\%$ in $\Lambda$CDM$+\Omega_\mathrm{K}$.

astro-ph.CO↗

arXiv2026

Beyond traditional emission-line diagnostics: using autoencoders to uncover active galactic nuclei in DESI spectra

The growing volume of spectroscopic data in modern surveys motivates data-driven approaches that complement traditional emission-line diagnostics for active galactic nuclei (AGN) identification. We present a machine learning framework that exploits the full optical spectrum using unsupervised representation learning within a semi-supervised classification scheme. We use the SPENDER autoencoder to compress DESI galaxy spectra into a low-dimensional latent space and classify sources through a k-d tree nearest-neighbor search. The model is trained on 50,222 DESI Main Survey spectra from the Guadalupe dataset and released as part of Data Release 1 (DR1), restricted to z <= 0.5. We validate the performance using labels derived from FastSpecFit's emission line measurements defining seven galaxy classes: AGN, broad-line (BL), composite, star-forming, passive, retired, and Other. The method achieves high accuracies for AGN (0.952) and broad-line AGN (0.965), reliably identifying these sources even in low signal-to-noise spectra and recovering AGN missed by standard single-diagnostic methods. Our classification metrics are benchmarked against traditional diagnostics, and we show they represent lower limits of the model's true performance. We also find that the learned latent space correlates with key galaxy properties such as stellar mass and star-formation rate, demonstrating that it captures physically meaningful information. These results show that unsupervised spectral representation learning, implemented within a semi-supervised classification framework, provides a scalable and effective approach for constructing more complete AGN catalogues for current and future spectroscopic surveys.

astro-ph.GA↗

arXiv2026

Weak Evolution of Cosmic Atomic Hydrogen over the Past 4.5 Billion Years

The cosmic star formation rate density (CSFRD) has declined sharply toward the present day, but the roles of the atomic and molecular gas reservoirs remain uncertain. We measure the cosmic HI density, $\Omega_{\mathrm{HI}}$, over $0<z<0.41$ by combining HI spectra from the Five-hundred-meter Aperture Spherical Telescope with optical spectroscopy from the Dark Energy Spectroscopic Instrument for $\sim2.5$ million galaxies across $\sim12,000\,{\rm deg}^2$. We measure a raw decrease in $\Omega_{\mathrm{HI}}$ by a factor of $1.35\pm0.10$ over the past 4.5 Gyr. Even after applying the conservative systematic corrections from our forward model, the inferred decline is only $1.12\pm0.10$ -- still far weaker than the CSFRD decline (a factor of 2.46). The molecular gas density, in contrast, is known to evolve more closely with star formation. At fixed stellar mass, the average HI gas fraction evolves by less than 0.2 dex, showing that the weak evolution is present across the galaxy population. These quantitative differences rule out rapid depletion of galaxy HI as the primary driver of the late-time CSFRD decline, and provide a stringent benchmark for models of gas accretion, phase conversion and star-formation regulation.

astro-ph.GA↗

arXiv2026

Clustering of high-redshift quasars with DESI DR2

We present clustering measurements for high-redshift quasars using data from the Dark Energy Spectroscopic Instrument Data Release 2. Our sample consists of quasars with $2.0 < z < 3.5$ in the luminosity range $M_{1450} \leq -19.94$\,mag. We measure the mean quasar bias $b_Q(\bar{z} = 2.48) = 3.61 \pm 0.01$ for the full sample of $\sim 715,000$ quasars and quantify the redshift evolution of quasar bias by dividing the sample into four equal redshift bins. There is strong evolution of the quasar bias with redshift that is well fit by the function $b_Q(z) = a [(1 + z)^2 - 6.565] + b$ with $a=0.230 \pm 0.007$ and $b=2.394 \pm 0.035$, and this fit is also a good match to lower redshift measurements in the literature. This bias evolution is consistent with a characteristic halo mass of $\bar{M}_{\mathrm{h}} \sim 10^{12}\,\mathrm{M_\odot}$ that does not vary significantly with redshift. The inferred duty cycles for quasars in our sample are $f_{\mathrm{duty}} \sim 10^{-2}$, staying mostly constant over redshifts. We investigate the luminosity dependence of quasar clustering by dividing each of our four redshift bins into three luminosity bins. The size of our quasar sample permits the first statistically significant measurement of the luminosity dependence of quasar bias at these redshifts. We measure weak dependence of quasar bias on luminosity at fixed redshift, inconsistent with no dependence, but weaker than predicted by a model in which quasar luminosity is tightly correlated with halo mass. These clustering measurements provide a stringent test for models of active black hole light curves and the black hole-halo connection at high redshift.

astro-ph.CO↗

arXiv2026

Measuring local primordial non-Gaussianity from the clustering of DESI DR1 LRGs and QSOs

We report the first measurement of primordial non-Gaussianity (PNG), parameterized by $f_{\mathrm{NL}}$, in the configuration space two-point correlation function (2pcf). We employ simulation based modeling and a novel approach for the mitigation of imaging systematics. We apply this method to samples of luminous red galaxies (LRG) and quasars (QSO) observed by the Dark Energy Spectroscopic Instrument (DESI) during the first year of its observations (DR1). The observed 68\% CL interval on $f_{\mathrm{NL}}$ is $-3^{+22}_{-21}$ using LRGs, and $ 0^{+17}_{-16}$ using QSOs. The joint measurement yields $f_{\mathrm{NL}} = -3^{+12}_{-12}$ at $\ [68\%]$ CL. Our pipeline imposes a Gaussian prior on the value of $p$ (which defines the PNG bias via the Universality relation), with $p_{\rm LRG} = 1.0 \pm 0.1$ and $p_{\rm QSO} = 1.6\pm 0.1$. The observed constraining power of DESI tracers significantly exceeds that of previous large-scale structure (LSS) surveys, and encouragingly, approaches the sensitivity of CMB probes of PNG.

astro-ph.CO↗

arXiv2026

Cosmological constraints from the DESI DR1 Bispectrum Full-Shape and DR2 BAO

We present cosmological constraints from the combination of DESI DR1 full-shape measurements, including for the LRG bispectrum, and DESI DR2 BAO data. The joint analysis accounts for cross-covariance using mocks, while ShapeFit compression mitigates prior volume effects that hinder beyond-$\Lambda$CDM analyses. In $\Lambda$CDM, the bispectrum (P+B) shifts $\sigma_8$ up by $1.1\sigma$ and $S_8$ by $1.2\sigma$, reducing their uncertainties by $26\%$ and $28\%$, respectively. For $w_0w_a$CDM, DESI-only analyses with the bispectrum shift dark energy parameters toward $\Lambda$CDM, staying consistent with a cosmological constant within $1\sigma$. Adding CMB creates a preference for evolving dark energy: DESI+CMB (P+B) shows a $2.8\sigma$ deviation from $\Lambda$CDM. Including DES-Dovekie supernovae alone reduces this to $1.6\sigma$, while the full combination DESI+CMB+DES-Dovekie gives $3.1\sigma$, driven primarily by the CMB. The bispectrum consistently weakens evidence for time-varying dark energy relative to power-spectrum-only analyses. The bispectrum also enhances sensitivity to massive neutrinos: in DESI-only analysis, the power-spectrum-only posterior for $\sum m_\nu$ is consistent with zero, whereas adding the bispectrum yields a mean of $0.26\pm0.17$~eV and a $95\%$ upper limit of $0.57$~eV, shifting the peak into the positive region and agreeing with oscillation lower bounds. For modified gravity, the bispectrum further constrains $\mu_0 = 0.12\pm0.49$ from DESI-only data, consistent with general relativity. Our analysis shows that accounting for cross-dataset covariances and avoiding prior volume effects yields robust constraints, with the bispectrum raising amplitude parameters and tightening their uncertainties.

astro-ph.CO↗

arXiv2026

Quality Assessment of Spectroscopic Data Reduction Pipelines Using Artificial Intelligence: Scrutinizing Data Release 2 from the DESI Survey

Large spectroscopic surveys now collect data at a scale that makes traditional visual inspection impractical. We present an unsupervised pipeline for spectroscopic quality assessment that requires no labeled training data. The method combines Uniform Manifold Approximation and Projection for dimensionality reduction with Friends-of-Friends clustering to isolate anomalous spectra for targeted review. We apply this pipeline to 58,291,334 spectra across 14,199 tiles from DESI Data Release 2, processing each tile independently to produce a tile-level outlier catalog. In each tile, the pipeline separates a dense core of typical spectra from small, isolated components and singletons, yielding a total of 1,095,816 outlier candidates. The mean tile-level outlier fraction is about 1.96 percent overall, with values of 0.76 percent and 2.36 percent for the dark and bright main-survey programs, respectively. From the visual inspection of 391 outlier candidates from the dark and bright programs of the main survey, we find that 66.8 percent exhibit identifiable spectral anomalies consistent with known reduction and calibration effects. By contrast, only 4.1 percent carry a non-zero quality flag from the standard reduction pipeline. This shows that the method provides a complementary quality-assessment layer to existing pipeline diagnostics and recovers a substantial population of problematic spectra that standard diagnostics miss. Extrapolating to the main-survey catalog, we estimate that approximately 218,000 candidate outliers are free of identifiable reduction artifacts and may correspond to genuine atypical spectra in the context of DESI. The pipeline is scalable, reproducible, and directly comparable across successive data releases, making it a practical quality-assurance monitor for DESI and future multi-object spectroscopic surveys.

astro-ph.IM↗

arXiv2026

Unveiling Hidden Lyman Alpha Emitters in the DESI DR1 Data

We present an automatic method based on machine-learning convolutional neural network (CNN) architecture to detect Lyman alpha emitters (LAE) hidden in the Data Release 1 spectroscopic dataset of the Dark Energy Spectroscopic Instrument (DESI). Those LAEs mostly have incorrect redshift estimations because the current DESI pipeline is not designed to detect and measure the redshifts of galaxies at $z>2$. To uncover those sources, we first visually inspect thousands of DESI spectra and construct a sample, consisting of both LAEs and non-LAEs, for training and testing the CNN-based model to (1) detect LAEs in DESI spectra and (2) determine their Ly$\alpha$ redshifts. The final model yields $95.2\%$ purity and $95.9\%$ completeness for detecting LAEs. We apply this model to approximately $2\times10^{6}$ spectra of sources targeted as emission-line galaxies and detect 19,685 LAEs from $z\sim2$ to $3.5$ within 12 minutes with a single GPU, illustrating the high efficiency of this model for identifying LAEs. The detected LAEs are mostly at the bright end of the luminosity function with Ly$\alpha$ luminosity $L_{\rm Ly\alpha} \gtrsim 10^{43}$ erg/s. The high signal-to-noise composite spectrum of the detected LAEs further shows various spectral features, including P-Cygni profiles of metal lines and MgII emission lines, possible indicators of Lyman continuum escape fraction, revealing the rich astrophysical information in this LAE sample. Finally, this sample can be used to train and validate the pipelines for redshift determination of LAEs for the preparation of the DESI-II survey.

astro-ph.GA↗