TL;DR
This paper presents the first results from the visual inspection (VI) of quasar spectra from the Dark Energy Spectroscopic Instrument (DESI) survey. By comparing visual classifications to automated pipeline outputs, the authors demonstrate that a modified pipeline significantly improves quasar identification and redshift accuracy, crucial for DESI's cosmological goals.
What Problem Does This Paper Address?
Large spectroscopic surveys like DESI aim to map the universe by collecting millions of spectra. Automatically classifying these spectra and determining accurate redshifts is a monumental task. While automated pipelines are essential for efficiency, they can misclassify objects or fail to find reliable redshifts, especially for complex or faint targets. This paper addresses the need to validate and improve these automated pipelines by using human visual inspection of a subset of DESI's quasar spectra. This validation is critical for ensuring the scientific integrity of the survey's cosmological measurements.
Methods
The core of this work is a detailed visual inspection (VI) of DESI quasar spectra. Experienced inspectors classified spectra and determined redshifts, creating a "truth table" to evaluate the performance of the standard DESI spectroscopic pipeline (Redrock). The authors then developed and tested a "modified pipeline" that incorporates "afterburner" algorithms (QuasarNet, Mg II afterburner) to identify quasars missed by Redrock. This modified pipeline aims to improve both the identification of quasars and the accuracy of their redshifts. The study also involved constructing composite spectra to understand the characteristics of quasars that are missed by the standard pipeline.
Datasets
The primary dataset used is the optical spectroscopic data from the Dark Energy Spectroscopic Instrument (DESI) Survey Validation (SV) phase. This includes deep coadded spectra from three specific SV tiles (XMM-LSS, Lynx, and COSMOS) and shallower spectra from 27 additional SV tiles. For redshift accuracy assessment, DESI quasar spectra were matched to the SDSS DR16 quasar catalog.
Key Insights and Ideas
- Visual inspection is crucial for validation: Even with advanced pipelines, human eyes are essential for catching subtle errors and understanding the nuances of spectral classification.
- "Afterburners" are key to improving quasar yield: Post-pipeline algorithms can significantly recover quasars missed by the standard Redrock pipeline, especially those with host-galaxy contamination or dust reddening.
- Modified pipeline offers a substantial boost: Combining Redrock with "afterburner" techniques leads to a significant increase in the number of reliably identified quasars and improved redshift accuracy.
- Missed quasars are often host-galaxy dominated or dust-reddened: These objects present a challenge for standard pipelines due to their spectral characteristics, resembling galaxies or being obscured by dust.
- DESI's depth reveals a diverse quasar population: The survey is capable of identifying fainter and more typical quasars than previous surveys, including those with prominent host galaxies and broad absorption lines.
Key Results
- The majority (~70%) of main-survey targets are spectroscopically confirmed as quasars, with ~16% galaxies, ~6% stars, and ~8% low-quality spectra.
- The standard Redrock pipeline reliably identifies high-quality quasars (~97% for main selection), but misses a significant fraction (~11-14%) of visually confirmed quasars.
- The modified pipeline recovers ~94% of visually inspected quasars, a substantial improvement over the standard pipeline (~86%).
- Both pipelines achieve good redshift purity (~99% for main DESI survey depth) and median redshift precision (~100 km/s).
- The modified pipeline shows improved redshift accuracy, particularly for challenging targets.
- DESI's capabilities allow for the identification of a diverse range of quasar types, including low-luminosity analogs of Seyfert galaxies and dust-reddened quasars.
Limitations and Future Work
- Exposure time dependence: Results from the deeper SV spectra are more optimistic than those achievable with the shallower main DESI survey exposures. However, the study mitigates this by constructing simulated shallower spectra.
- Catastrophic redshift failures: While the modified pipeline significantly reduces these, some still occur, particularly for Ly-alpha QSOs, likely due to complex spectral features and limitations in template libraries.
- Future work: The authors plan to re-visually inspect spectra after each data-assembly release to continuously improve the VI dataset and re-compute survey metrics. Further validation of "afterburner" approaches is also ongoing to enhance the DESI quasar survey's performance.
Summary updated