The DESI Transients Survey: Legacy Classifications and Methodology

TL;DR

The paper demonstrates the DESI Transients Survey as a scalable framework to discover and spectroscopically classify extragalactic transients using DESI, extending reach to faint and intermediate-redshift events. By combining serendipitous DESI spectra with a spare-fiber ToO program and a robust classification pipeline (NGSF with two redshift modes and human vetting), the study classifies hundreds of transients, including multiple tidal disruption events and SN types, down to $m \lesssim 22$ and $0.05 < z < 0.25$ (with potential to reach higher redshifts for certain subclasses). Key results include the legacy and newly classified transients, infant TDE observations (e.g., 2022emf and 2023ant), and host-subtracted spectra that improve classification accuracy. The work demonstrates DESI’s capacity to supplement Rubin LSST by providing timely spectroscopic classifications at intermediate redshifts, with data releases to TNS and Zenodo and a roadmap toward ~$O(1000)$ spectra per year in the LSST era.

Abstract

We present the first systematic spectroscopic observations of extragalactic transients from the Dark Energy Spectroscopic Instrument (DESI), as part of the DESI Transients Survey program. With 5,000 fibers and an ${\sim} 8$ deg$^2$ field of view, we exploit DESI as a machine for the discovery and classification of transients. We present transient classifications from archival DESI data in Data Releases 1 and 2, relying on a combination of a secondary target program and serendipitous observations. We also present observations from the first 6 months of the DESI spare fiber program dedicated to transients. The program is run in coordination with a dedicated DECam time-domain survey, serving as a pathfinder for what we will be able to achieve in conjunction with the Rubin Observatory Legacy Survey of Space and Time (LSST). We classify over 250 transients, of which the majority were previously unclassified. The sample comprises thermonuclear and core-collapse supernovae and tidal disruption events (TDEs), including a TDE observed before its discovery in imaging. We demonstrate DESI's ability to classify a population of faint transients down to $r\sim 22.5$ mag during main survey operations, with negligible impacts on DESI's main observations. can you make this plain text for arxiv abstarct

Figures (12)

• Figure 1: TNS transients that have serendipitous observations by DESI per reporting group. Transients come from ZTF (ALERCE sanchez-saez_alert_2021, ZTF Partnerships bellm_zwicky_2018, and AMPEL nordin_transient_2019), Pan-STARRS chambers_pan-starrs1_2016, ATLAS tonry_atlas_2018, DESIRT palmese_desirt_2022hall_decam_2025, The Young Supernova Experiment jones_young_2021,
• Figure 2: The number of targets of opportunity that fall into a DESI tile on a given survey night (Potential ToOs) versus the number actually observed (Observed ToOs). On nights where DARK or BRIGHT tiles are observed about $10\%$ of all possible targets are observed. We find that DESI is able to average about one transient per night from TNS alone and around three per night including TNS, ZTF, and DESIRT. This large jump in observed and potential ToOs can be seen with the restarting of the DESIRT program in February 2025.
• Figure 3: A histogram of the number of transients that have been observed a certain number of days after being initially triggered. The line shows the cumulative distribution of targets observed within that time span.
• Figure 4: Treemap showing the count of each broad type of extragalactic transients discovered in DESI DR1, DR2 and the spare fiber survey desi_collaboration_data_2025desi_collaboration_desi_2025
• Figure 5: Top: lightcurves of TDEs 2022emf (left) and 2023ant (right), two TDEs newly classified by this work; in both note the long slow decline and lack of clear color evolution in the lightcurve. Bottom: DESI spectra of the same TDEs with the "best fit" template from Superfit is plotted in gray. The broad He II and H$\alpha$ emission in each spectrum are consistent with TDEs. The vertical line shows the time of the DESI spectrum. The spectrum for 2022emf was taken even before the transient discovery in imaging. DESI serendipitous and spare fiber spectroscopy has immense potential in its ability to do infant TDE and SNe studies. The lightcurves are provided by the ZTF forced-photometry service masci_new_2023.