Damped Lyman-alpha (DLA) object catalog

This document describes the content and construction of DLA catalogs from CNN and GP finder for DESI Data Release 1 (release date March 2025).

Overview

Our VAC file contains a DLA catalog derived from the Lyman-alpha forest of DESI first year. This catalog combine damped Lyman-alpha (DLA) catalogs from two machine learning DESI DLA finders: a convolutional neural network (CNN) model and a Gaussian Process (GP) model. For more information about these DESI DLA finders see Wang et al. (2022).

Data Access

Data URL: https://data.desi.lbl.gov/public/dr1/vac/dr1/dla-cnn-gp/

NERSC access:

/global/cfs/cdirs/desi/public/dr1/vac/dr1/dla-cnn-gp/

Documentation

The combination of CNN and GP catalog was performed in the following way:

• Match each absorber from two models (\(\Delta v\) < 800 km/s).
• For absorbers detected by both models: use GP_NHI and GP_Z_DLA from GP.
• For absorbers only detected by the CNN model: use CNN_NHI and CNN_Z_DLA from CNN.
• For absorbers only detected by the GP model: use GP_NHI and GP_Z_DLA from GP.

Files

• dla_catalog_cnngp_combine_dr1.fits - DLA catalog

Data Model

The detail for each column of the catalog is:

Name	Type	Units	Description
TARGET_RA	float64	deg	Target Right Ascension (J2000)
TARGET_DEC	float64	deg	Target Declination (J2000)
Z_QSO	float64	-	Best-fit redshift after masking
TARGETID	int64	-	Unique 64-bit identifier for each object observed by DESI
S2N	float64	-	Uses slices in 1420–1480 Å in QSO rest-frame to calculate the continuum signal-to-noise.
DLAID	str	-	Unique DLA ID, TARGETID+ 3-bit code. The 3-bit code is sorted by redshift from low to high, ‘000’ to ‘001’, etc.
CNN_NHI	float64	-	CNN model logarithmic HI column density. The physical HI column density can be computed as \(10^{CNN_{NHI}} \mathrm{cm}^{-2}\). If this DLA was not detected by one DLA finder, then this finder’s prediction will set to 0.
GP_NHI	float64	-	CNN model logarithmic HI column density. The physical HI column density can be computed as \(10^{GP_{NHI}} \mathrm{cm}^{-2}\). If this DLA was not detected by one DLA finder, then this finder’s prediction will set to 0.
CNN_Z_DLA	float64	-	CNN model redshift. If this DLA was not detected by one DLA finder, then this finder’s prediction will set to 0.
GP_Z_DLA	float64	-	GP model redshift. If this DLA was not detected by one DLA finder, then this finder’s prediction will set to 0.
CNN_DLA_CONFIDENCE	float64	-	The confidence level predicted by CNN DLA finder. We suggest to select absorbers with CNN_DLA_CONFIDENCE>0.2 as valid detections for S2N>3, DLA_CONFIDENCE>0.3 for S2N<3.
GP_DLA_CONFIDENCE	float64	-	The confidence level predicted by GP DLA finder. We suggest to select absorbers with GP_DLA_CONFIDENCE>0.9 as valid detections. See descriptions in Ho et al. (2021).
ABSORBER_TYPE	str	-	One of DLA (log NHI > 20.3), SUBDLA (log NHI < 20.3), or LYB (Lyman-beta absorbers corresponding to DLAs in the same sightline).
NHI	float64	-	For absorbers detected by both models use NHI=GP_NHI. For absorbers only detected by the CNN model use NHI=CNN_NHI. For absorbers only detected by the GP model use NHI=GP_NHI. The physical HI column density can be computed as \(10^{NHI} \mathrm{cm}^{-2}\).
Z_DLA	float64	-	For absorbers detected by both models use Z_DLA=GP_Z_DLA. For absorbers only detected by the CNN model use Z_DLA=CNN_Z_DLA. For absorbers only detected by the GP model use Z_DLA=GP_Z_DLA.

Contact

These catalogs were generated by DLA Finder Group, contact Jiaqi Zou if you have any questions.