Validation of the DESI DR2 Ly alpha BAO analysis using synthetic datasets

arXiv:2503.14741 · astro-ph.IM, astro-ph.CO · Submitted 2025-03-18 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics. The day in astrophysics: observations, models and what the data forces.

Vera: I'm Vera.

Jocelyn: And I'm Jocelyn, here with Subrahmanyan.

Vera: We're looking at the paper: "Validation of the DESI DR2 Ly alpha BAO analysis using synthetic datasets".

Jocelyn: Here's the gist of the paper. The validation of the DESI DR2 Lyα BAO analysis using synthetic datasets is important because it rigorously tests the cosmological inference pipeline against a significantly larger dataset than previously used The second data release (DR2) of the Dark Energy Spectroscopic Instrument (DESI), containing data from the first three years of observations, doubles the number of Lyman-α (Lyα) forest spectra in DR1 and it provides the largest dataset of its kind Mock Dataset Improvements The authors made significant updates compared to DR1 to both the mocks and the analysis framework used in the validation In particular, they present CoLoRe-QL, a new set of Lyα mocks that use a quasi-linear input power spectrum to incorporate the non-linear broadening of the BAO peak They also increased the number of realisations used in the validation to 400, compared to the 150 realisations used in DR1 Synthetic Quasar Spectra Generation

Title and authors: Vera: So let's talk about the title and who wrote this paper, "Validation of the DESI DR2 Ly alpha BAO analysis using synthetic datasets." It’s clear that this work is focused on testing the accuracy of their method when they use the second data release from DESI.

Jocelyn: And it highlights that they aren't just doing a simple check; they are validating an entire analysis pipeline against a significantly larger dataset than what was used in previous studies.

Subrahmanyan: From my side, what this means is that if their method works well on these synthetic mocks, then the cosmological inferences they get from the actual DESI DR2 data should be more reliable for constraining dark energy.

Vera: That’s right. They are using these mocks to see how much noise or systematic error creeps in when you scale up the input data size and complexity.

Jocelyn: And they also mention that they increased the number of realisations used in their validation, moving from one hundred fifty realisations in DR1 to four hundred realisations for this work.

Subrahmanyan: A larger number of simulations gives them a more statistically robust picture of the error bars associated with their final measurements.

Vera: And they also introduced some new mock sets called CoLoRe-QL, which use a quasi-linear input power spectrum to better include that non-linear broadening we talked about earlier when dealing with the BAO peak.

Jocelyn: That sounds like a targeted fix for a known issue in modeling the signal shape itself, moving beyond just using Gaussian mocks.

The paper's summary: Vera: The summary of this paper boils down to how they improved their mock generation and the analysis framework itself to better match what we expect from DESI DR2.

Jocelyn: They’re focusing on several key improvements that make the synthetic data more realistic, like including astrophysical contaminants such as high column density systems and broad absorption lines in their simulations.

Subrahmanyan: That's important because real observations aren't just clean Ly alpha forest; they have these other things mixed in that can mess up the signal if you don't model them correctly.

Vera: They also added a step to simulate redshift errors by adding random shifts to the quasar redshifts, following a Gaussian distribution with a dispersion of four hundred kilometers per second <ref:2503.14741#pg2>.

Jocelyn: That directly tests how much those estimation errors from things like Redrock or QuasarNET impact the final BAO measurements.

Subrahmanyan: And they developed an alternative analysis method called the "No close pairs" analysis, which discards problematic quasar pairs based on certain angular and velocity separations.

Vera: They also integrated metal absorption line modeling into their correlation function analysis to account for misestimated pixel redshifts when assuming everything comes from Ly alpha.

Jocelyn: Plus, they use a comprehensive set of free nuisance parameters, including biases for Ly alpha systems, HCDs, and linear biases for each metal line in the model fitting process.

The paper's improvements: Vera: So when we look at the specific improvements they made to their methodology and mocks, it's quite detailed. They focused on better incorporating non-linear broadening of the BAO peak using an anisotropic Gaussian damping term.

Jocelyn: That term is PQL parallel, k parallel, k perpendicular, z, which accounts for how the shape of the BAO peak changes based on direction and redshift in a non-linear way.

Subrahmanyan: Modeling that non-linear evolution is crucial because it’s what gets you away from simpler linear approximations when looking at structure formation across different scales.

Vera: They also developed a specific strategy to mitigate BAO scale bias in their real data analysis by using the "No close pairs" method, which they found was less biased than baseline configurations when analyzing real data stacks.

Jocelyn: It’s interesting that they also used this same "No close pairs" technique as a strategy to mitigate bias when analyzing the actual DR2 data, suggesting it's a good tool here.

Subrahmanyan: So, in terms of what this means for cosmology, these improvements show they are systematically addressing known sources of systematic error in the analysis pipeline.

Vera: They also used metal absorption line modeling to account for the contamination from pixel redshift misestimation when assuming all absorption is just Ly alpha.

Jocelyn: That level of detail suggests they’re trying to be as thorough as possible, accounting for various sources of systematic error present in the mock data.

Conclusion: Vera: So wrapping up this discussion on "Validation of the DESI DR2 Ly alpha BAO analysis using synthetic datasets," it seems like the main implication is that they’ve built a much more rigorous and thoroughly tested framework for extracting cosmological information from the DESI DR2 data.

Jocelyn: They've used these detailed mock improvements—like incorporating non-linear broadening and modeling redshift errors—to give us confidence in the resulting BAO measurements.

Subrahmanyan: For the bigger picture, this work strengthens our ability to use Lyman alpha forest BAO to constrain dynamical dark energy because they’ve shown how well their methodology handles real-world complexities.

Vera: They showed that by testing against these extensive synthetic datasets, they can quantify the impact of various systematic errors on their final measurements.

Jocelyn: It points toward a more reliable path for using this specific probe to measure the expansion history of the universe.

Subrahmanyan: I think we need more work on how this analysis scales up to even larger volumes or different redshift ranges, but this validation is a solid step forward in making sure the fundamental physics modeling is sound.

Vera: Absolutely. It sets a high bar for future analyses that want to use these kinds of spectroscopic surveys for cosmological constraints.

Jocelyn: So that's what we have with this paper on the DESI DR2 Ly alpha BAO analysis using synthetic datasets, and it leaves us with a much more robust tool for the job.

Institut de Fíica d’Altes Energies (IFAE) · Institut d…Astrophysique de Paris · IRFU CEA Université Paris-Saclay · Lawrence Berkeley National Laboratory · Boston University · Leinweber Center for Theoretical Physics University of Michigan · University of Michigan · ▲ Instituto Avanzado de CosmologΡ A. C. · Instituto de Ciencias Fíicas Universidad Nacional Autonoma de México · Instituto de Fíica Universidad Nacional Autonoma de M⃡xico · Department of Physics & Astronomy University of Rochester · INAF-Osservatorio Astronomico di Brera · Department of Physics & Astronomy University College London · Institute of Cosmology and Gravitation University of Portsmouth · ▲ Department of Physics and Astronomy The Ohio State University · University of Queensland School of Mathematics and Physics, Department of Physics The Ohio State University, Department of Physics Southern Methodist University, Department of Physics and Astronomy University College Irvine, Sorbonne Universite CNRS/IN2P3 Laboratoire de Physique Nuclére et de Hautes Energies · Department of Astronomy and Astrophysics UCO/Lick Observatory University of California Santa Cruz · Excellence Cluster ORIGINS Boltzmannstrasse Garching Germany, Ludwig-Maximilians-Universität Munich

astro-ph.IM, astro-ph.CO

Submitted: 2025-03-18

Updated: 2026-11-10

Comments: This DESI Collaboration Publication is part of the Data Release 2 publication series (see https://data.desi.lbl.gov/doc/papers)

Journal ref: Phys. Rev. D 113, 023520 (2026)

DOI: 10.1103/fvgh-kswf

Code: https://github.com/damonge/CoLoRe

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 87/100

The gist: The validation of the DESI DR2 Lyα BAO analysis using synthetic datasets is important because it rigorously tests the cosmological inference pipeline against a significantly larger dataset than…

Key concepts

CoLoRe-QL mocks
These are new synthetic Lyman-alpha (Ly$\alpha$) mock datasets designed to better represent real observations. They use a quasi-linear input power spectrum, which helps accurately model the non-linear broadening that occurs in the BAO peak, leading to a more realistic test of the analysis pipeline.
Redshift errors in mocks
The simulations now include random shifts in quasar redshifts following a Gaussian distribution (400 km/s dispersion). This directly tests how much uncertainty in measuring quasar distances affects the final BAO measurements, mimicking real-world estimation errors.
"No close pairs" analysis
This is an alternative method used to clean up data by discarding quasar pairs that are too close angularly or have small velocity separations. This strategy minimizes contamination from redshift errors during continuum fitting while retaining most of the useful data.

Terminology

Summary

The validation of the DESI DR2 Lyα BAO analysis using synthetic datasets is important because it rigorously tests the cosmological inference pipeline against a significantly larger dataset than previously used The second data release (DR2) of the Dark Energy Spectroscopic Instrument (DESI), containing data from the first three years of observations, doubles the number of Lyman-α (Lyα) forest spectra in DR1 and it provides the largest dataset of its kind

Mock Dataset Improvements

The authors made significant updates compared to DR1 to both the mocks and the analysis framework used in the validation In particular, they present CoLoRe-QL, a new set of Lyα mocks that use a quasi-linear input power spectrum to incorporate the non-linear broadening of the BAO peak They also increased the number of realisations used in the validation to 400, compared to the 150 realisations used in DR1

Synthetic Quasar Spectra Generation

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved mock generation for CoLoRe-QL mocks by using a quasi-linear input power spectrum to better incorporate non-linear broadening of the BAO peak, as this leads to better agreement with DESI DR2 measurements. This allows for more accurate modeling of the BAO signal compared to previous Gaussian mocks.

  2. Enhanced spectral simulation for DESI DR2 mocks by including various astrophysical contaminants such as High Column Density systems (HCDs), Broad Absorption Lines (BALs), and other transition lines, which are then multiplied by random unabsorbed quasar spectra. This provides a more realistic representation of the observed Ly alpha forest spectra.

  3. Implementation of redshift errors in mock catalogs by adding random shifts to the redshifts of simulated quasars following a Gaussian distribution with a dispersion of 400 km/s, emulating statistical uncertainties from quasar redshift estimation algorithms like Redrock or QuasarNET. This directly tests and quantifies the impact of redshift errors on BAO measurements.

  4. Development and application of an alternative analysis method, such as the No close pairs analysis, which discards problematic pairs of quasars based on angular separations smaller than 20 arcmin and velocity separations smaller than 4000 km/s. This minimizes contamination from redshift errors in continuum fitting without losing a significant fraction of the data.

  5. Integration of metal absorption line modeling into correlation function analysis by accounting for the misestimation of pixel redshifts due to the assumption that all absorption originates from Ly alpha, modeling this contamination with the same method used in DR1 analysis but modified for DR2. This ensures more accurate measurement of BAO scale across different spectral regions (A and B).

  6. Use of a comprehensive set of free nuisance parameters, including three for Ly alpha and quasar biases, three for HCD systems, four linear biases for each metal line, and two for redshift errors in the model fitting process. This allows the analysis to simultaneously account for various sources of systematic error present in the mock data.

  7. Incorporation of a model to describe non-linear broadening of the BAO peak using an anisotropic Gaussian damping term PQL(k parallel, k perpendicular, z) = PBAO(k, z)e−(k squared parallel sigma squared parallel + k squared perpendicular sigma squared perpendicular)/2+Psmooth(k, z). This accounts for nonlinear evolution effects on the BAO peak shape.

  8. A strategy to mitigate BAO scale bias by removing close pairs of quasars in real data analysis using the No close pairs method, which has been shown to be significantly less biased than baseline configurations when analyzing real data stacks.

Abstract

The second data release (DR2) of the Dark Energy Spectroscopic Instrument (DESI), containing data from the first three years of observations, doubles the number of Lyman- α (Ly α) forest spectra in DR1 and it provides the largest dataset of its kind. To ensure a robust validation of the Baryonic Acoustic Oscillation (BAO) analysis using Ly α forests, we have made significant updates compared to DR1 to both the mocks and the analysis framework used in the validation. In particular, we present CoLoRe-QL, a new set of Ly α mocks that use a quasi-linear input power spectrum to incorporate the non-linear broadening of the BAO peak. We have also increased the number of realisations used in the validation to 400, compared to the 150 realisations used in DR1. Finally, we present a detailed study of the impact of quasar redshift errors on the BAO measurement, and we compare different strategies to mask Damped Lyman- α Absorbers (DLAs) in our spectra. The BAO measurement from the Ly α dataset of DESI DR2 is presented in a companion publication.

Sources

Related papers