ODIN: Spectroscopic Validation of Ly$\alpha$-Emitting Galaxy Samples with DESI

The ODIN survey successfully validated its narrow-band selection of Lyman-alpha emitting galaxies at redshifts 2.4, 3.1, and 4.5 using DESI spectroscopy, achieving confirmation rates of 92–96% while identifying active galactic nuclei and lower-redshift emission lines as primary contaminants.

The Gist

Imagine the universe as a giant, cosmic ocean. For decades, astronomers have been trying to map this ocean, but it's so vast and dark that they can only see the "islands" (galaxies) that are glowing brightly. One specific type of glowing island, called a Lyman-alpha Emitting Galaxy (LAE), is like a lighthouse. These galaxies are young, forming stars rapidly, and they emit a very specific color of light (Lyman-alpha) that acts as a beacon for scientists studying how the universe grew up.

However, finding these lighthouses is tricky. It's like trying to spot a specific color of firefly in a field full of other bugs, streetlights, and reflections. You might think you've found a firefly, but it could actually be a different bug or a trick of the light.

This paper is about a project called ODIN (One-hundred-deg2 DECam Imaging in Narrowbands). ODIN is a massive survey using a powerful camera on a telescope in Chile to take pictures of the sky through special "sunglasses" (narrow-band filters). These sunglasses are tuned to only let through the specific color of light from those young lighthouse galaxies at three different eras in the universe's history (when the universe was about 3, 4, and 6 billion years old).

The team built a list of thousands of potential lighthouses. But here's the big question: How many of these are real, and how many are imposters?

To answer this, they called in the "police": DESI (Dark Energy Spectroscopic Instrument). DESI is like a super-powered forensic lab. Instead of just taking a picture, it takes a "fingerprint" (a spectrum) of each object. This fingerprint tells them exactly what the object is and how far away it is.

The Investigation: What Did They Find?

The team took 11,599 candidates from the ODIN list and ran them through the DESI lab. Here is the breakdown of their findings, explained simply:

1. The "Firefly" Success Rate (Purity)
The ODIN sunglasses were incredibly good at their job.

• The Result: When they checked the fingerprints, 92% to 96% of the objects they thought were lighthouses actually were lighthouses.
• The Analogy: Imagine you have a bag of 100 marbles, and you think they are all red. You check them one by one, and 93 of them are actually red. That is a very successful bag!
• The "Imposters": The few that weren't real lighthouses were mostly:
  • Active Galactic Nuclei (AGN): These are like super-bright, hungry black holes at the center of galaxies. They glow in the same color, but they are the "villains" of the story, not the young star-forming galaxies ODIN was looking for.
  • Lower-Redshift Galaxies: These are older galaxies that happen to have other glowing lines (like oxygen) that look like the lighthouse signal from a distance. It's like seeing a red stop sign and thinking it's a red firefly because you're squinting.

2. The "Missing" Fireflies (Completeness)
Did ODIN miss any real lighthouses?

• The Result: They missed very few. They estimate that ODIN found about 91% to 98% of the real lighthouses in the areas they looked at.
• Why did they miss some? It wasn't because the sunglasses were bad. It was because some areas were blocked by "fog" (stars in our own Milky Way that were too bright to look past) or because the data wasn't deep enough in those specific spots. It's like trying to find fireflies in a field, but you can't look behind a big oak tree because the leaves are too thick.

3. The "Tricky" Redshifts
The team looked at the three different time periods (redshifts) separately:

• Time 1 (z=2.4): A bit more "noise" here. Some imposters were galaxies with oxygen lines that tricked the computer into thinking they were lighthouses.
• Time 2 & 3 (z=3.1 and 4.5): These were cleaner. The "sunglasses" worked almost perfectly, with very few imposters.

Why Does This Matter?

You might ask, "So what? We found some galaxies."

This is crucial for two reasons:

1. Trust: Before this, scientists had to guess how many "fake" galaxies were in their lists. Now, they know the exact percentage. It's like knowing your scale is accurate to within 1%. This allows them to do precise math on how the universe is expanding and how dark energy works.
2. Future Maps: The ODIN survey is huge. It's mapping out the "skeleton" of the universe. By proving their method works so well, they can confidently use this data to find massive clusters of galaxies (protoclusters) that will eventually become the giant cities of the universe we see today.

The Bottom Line

Think of the ODIN survey as a fisherman casting a net to catch a specific type of rare fish. This paper is the fisherman checking his catch. He pulled up the net, counted 11,599 fish, and brought them to a lab to verify the species.

He found that his net is incredibly efficient. He caught the right fish 93-96% of the time, and he didn't miss many of the right fish hiding in the water. The few "wrong" fish he caught were mostly other types of fish that looked similar from a distance.

This gives astronomers the confidence to say, "We have a very accurate map of the young universe," which helps them solve the biggest mysteries of our cosmos, like what dark energy is and how galaxies are born.

Technical Summary

Here is a detailed technical summary of the paper "ODIN: Spectroscopic Validation of Lyα-Emitting Galaxy Samples with DESI."

1. Problem Statement

The One-hundred-deg² DECam Imaging in Narrowbands (ODIN) survey is the largest photometric survey to date for identifying Lyman-alpha emitting galaxies (LAEs) at high redshifts ($z \approx 2.4, 3.1, 4.5$). While photometric selection using narrow-band (NB) excess is efficient for assembling large samples, it suffers from contamination by "interlopers"—lower-redshift star-forming galaxies (emitting [O II], [O III]) and Active Galactic Nuclei (AGN) that mimic the NB flux excess of LAEs.

To utilize these samples for cosmological studies (e.g., tracing large-scale structure, measuring the Lyman-alpha luminosity function, and identifying protoclusters), it is critical to accurately quantify the sample purity (contamination rate) and completeness (fraction of true LAEs recovered). Prior to this work, the ODIN sample lacked a systematic spectroscopic validation against a large, independent dataset.

2. Methodology

The authors utilized spectroscopic follow-up observations from the Dark Energy Spectroscopic Instrument (DESI) to validate the ODIN photometric selections in the COSMOS and XMM-LSS fields.

• Target Selection:
  • ODIN LAEs: Selected using Source Extractor (SE) photometry with a double-broadband continuum estimate. Criteria required a narrow-band excess corresponding to a rest-frame equivalent width ($EW$) $> 20$ Å.
  • DESI-ODIN (Parent Sample): A broader selection using Tractor model-based photometry, designed to be more inclusive to test the limits of the ODIN cuts.
  • Comparison: The study compared three disjoint categories: Tr-only (DESI selected, not ODIN), Tr+ODIN (selected by both), and ODIN-only (ODIN selected, not DESI).
• Observations:
  • DESI obtained spectra for 11,599 candidates across three observing campaigns (March 2022, Dec 2022–Jan 2023, April 2023).
  • Exposure times ranged from 1.3 to 3.2 hours (median ~2.75 hr), achieving signal-to-noise ratios sufficient for redshift confirmation.
• Visual Classification:
  • A visual inspection campaign by multiple experts assigned a quality flag ($q$) to each spectrum (0–4).
  • Confirmation Criteria: Targets with $q \ge 2.5$ were considered spectroscopically confirmed. This threshold was chosen because LAEs typically show only one strong Ly$\alpha$ line, limiting the maximum possible $q$ to 3.
• Analysis:
  • Calculated Sample Purity: The fraction of confirmed LAEs among all confirmed targets.
  • Calculated Completeness: Estimated by comparing the number of ODIN-selected LAEs to the total number of spectroscopically confirmed LAEs (including those missed by ODIN but found by DESI), accounting for regions excluded by star masks or lack of broad-band data.

3. Key Contributions

• First Large-Scale Validation: This is the first systematic spectroscopic validation of the ODIN survey, providing the first robust measurement of contamination rates for the largest photometric LAE survey to date.
• Quantification of Contaminants: The study identifies and categorizes the specific types of interlopers (AGN, [O II], [O III], Mg II, C IV emitters) and their redshift distributions.
• Completeness Estimation: It provides the first rigorous estimates of the completeness of the ODIN selection, distinguishing between objects missed due to selection criteria versus those missed due to survey geometry (e.g., star masks).
• Methodological Insight: The paper highlights the impact of photometric methods (SE vs. Tractor) and continuum estimation techniques on LAE selection efficiency and contamination.

4. Key Results

A. Sample Purity

The ODIN selection criteria yield highly pure samples across all three redshift bins:

• $z \approx 2.4$ (N419 filter): 93.0% purity (92.2% face value).
• $z \approx 3.1$ (N501 filter): 96.2% purity.
• $z \approx 4.5$ (N673 filter): 92.0% purity.
• Contaminants:
  • The primary contaminants are broad-line AGN at the target redshift (1–2%) and lower-redshift AGN/galaxies.
  • N501 & N673: Contaminants are primarily [O III], Mg II, and C IV emitters. Notably, [O II] contamination is minimal (<1%), validating the $EW > 20$ Å cut.
  • N419: Shows a more complex contamination profile. While most are AGN, there is a small population ($\sim$2%) of star-forming galaxies at $z=0.2-0.8$ and $z=1.0-1.5$. The latter group is hypothesized to possess a strong 2175 Å dust extinction bump, causing the continuum to be underestimated and mimicking an LAE.

B. Completeness

The study estimates the fraction of true LAEs recovered by ODIN:

• $z \approx 2.4$: 91–92%
• $z \approx 3.1$: 95%
• $z \approx 4.5$: 98%
• Reason for Incompleteness: The majority of "missed" LAEs (Tr-only LAEs not selected by ODIN) were located in regions excluded by ODIN's star masks or where HSC broad-band data was unavailable, rather than failing the photometric selection criteria.

C. Demographics and Brightness Dependence

• AGN Dominance: At bright magnitudes ($NB \lesssim 22.5$), AGN dominate the contaminant population, though their absolute numbers are small compared to LAEs.
• LAE Peak: LAEs constitute $>90\%$ of the sample at the peak magnitude ($NB \approx 24$).
• Faint End: At fainter magnitudes ($NB \approx 25.5$), the spectroscopic success rate drops, but the purity remains high.

5. Significance and Future Implications

• Cosmological Utility: The high purity ($>90\%$) and well-characterized contamination rates validate the use of ODIN LAEs for precision cosmology, including Baryon Acoustic Oscillation (BAO) measurements and large-scale structure analysis.
• Protocluster Studies: The validated sample is robust for identifying high-redshift protoclusters, a key goal of the ODIN survey.
• Future Surveys: The results suggest that future surveys (like DESI-2) can rely on similar narrow-band selection techniques. Furthermore, the upcoming Rubin Observatory (LSST) deep imaging will provide the broad-band coverage necessary to eliminate the remaining low-redshift interlopers based on color alone, further enhancing sample purity.
• Methodological Refinement: The study confirms that the double-broadband continuum estimation used by ODIN is effective at removing [O II] contaminants, though it may struggle with rare, highly dust-extincted galaxies at specific redshifts.

In conclusion, this paper establishes the ODIN survey as a premier resource for studying the high-redshift universe, providing the necessary spectroscopic validation to ensure the reliability of its massive LAE catalog for both astrophysical and cosmological research.