Intrinsic alignment demographics for next-generation lensing: Revealing galaxy property trends with DESI Y1 direct measurements

We present direct measurements of intrinsic alignments (IA) of over 2 million spectroscopic galaxies using DESI Data Release 1 and imaging from four lensing surveys: DES, HSC, KiDS, and SDSS. In this uniquely data-rich regime, we take initial steps towards a more tailored IA modelling approach by building a library of IA measurements across colour, luminosity, stellar mass, and redshift. We map the dependence between galaxy type—in terms of rest-frame colour, strength of the 4000 A˚ break, and specific star formation rate—and IA amplitude; the bluest galaxies have an alignment consistent with zero, across low (0.05 1.15) redshifts. In order to construct cosmic shear samples that are minimally impacted by IA but maintain maximum sample size and statistical power, we map the dependence of alignment with colour purity. Red, quenched galaxies are strongly aligned and the amplitude of the signal increases with luminosity, which is tightly correlated with stellar mass in our catalogues. For DESI galaxies between 0 < z < 1.5, trends in luminosity and colour alone are sufficient to explain the alignments we measure—with no need for an explicit redshift dependence. In a companion paper (Jeffrey et al., in prep), we perform detailed modelling of the IA signals with significant detections, including model comparison. Finally, to direct efforts for future IA measurements, we juxtapose the colour-magnitude-redshift coverage of existing IA measurements against modern and future lensing surveys.

The intrinsic alignment (IA) contaminates cosmic shear studies in two ways: i) physically close galaxies can have correlated shapes and ii) the shapes of galaxies in a foreground cluster are correlated with lensed background galaxies. Cosmic shear studies commonly employ the Non-Linear linear-Alignment model (NLA, Catelan et al. 2001; Hirata & Seljak 2004) to account for IA. The NLA assumes the intrinsic ellipticities of galaxies are linearly related to the density field; the model also includes a non-linear correction to the linear matter power spectrum (Bridle & King 2007), as well as luminosity, mass, and redshift dependence (Joachimi et al. 2011). However, this model is only valid on scales typically larger than those analysed in cosmic shear studies (≳ 6 Mpc h−1). The Tidal Alignment and Tidal Torquing model (TATT, Blazek et al. 2019) adds higher order terms, including tidal torquing, which extend the model validity to smaller scales (≳ 2 Mpc h−1), but the added flexibility comes at the cost of reduced cosmological precision. Additional models have been developed based on the halo model (Fortuna et al. 2021) and effective field theory (Vlah et al. 2020; Bakx et al. 2023; Chen et al. 2024; Maion et al. 2024); these incur additional model parameters that exacerbate the loss of cosmological precision. In cosmic shear studies, models are typically applied without consideration of the properties of the lensing sample. This can result in unnecessary degradation of cosmological precision if a highly flexible model is applied to a galaxy population with weak IA (McCullough et al. 2024). To accurately and precisely account for IA in cosmic shear, we therefore require direct measurements to i) build informative priors on alignment strength as a function of galaxy type and redshift and ii) identify galaxy populations with minimal IA.

Direct IA measurements require precise redshifts and shape measurements across wide areas of sky to correlate shapes and positions on scales between 0.1 to > 100 Mpc h −1. Direct IA measurements are significantly detected for the BGS sample > 10σ, while no alignments are detected for the ELG samples. The LRG measurements are also significantly detected > 10σ and exhibit strong colour and luminosity dependence (see Section 5). The ELG IA measurements are consistent with zero (bottom panel of Figure 2).

The IA shape–density correlation measurements for each DESI tracer are presented in Figure 2. The detection significances of the measurements are reported in Table 2. IA are significantly detected for the BGS sample > 10σ. We divide each DESI BGS cross-matched shape catalogue into blue and red sub-populations by rest-frame colour, Mr − Mz = 0.5, and then further divide by luminosity at Mr = −20.9 and Mr = −21.9 for the blue and red subsamples, respectively; the selection is only applied to the shape catalogues, and the density sample remains the complete BGS.

This scientific paper provides a comprehensive framework for understanding and mitigating Intrinsic Alignment (IA) in galaxy surveys, which is a crucial systematic effect when analyzing cosmic shear data for cosmology.

As an AI researcher, I can suggest several improvements to existing AI systems—specifically those used in weak lensing analysis and cosmological modeling—by integrating the findings from this paper.

Here are the specific improvements and what the resulting improved AI system could achieve:

)

The improved AI system will be a sophisticated, end-to-end pipeline for cosmic shear analysis that incorporates IA-aware galaxy property selection and model-dependent signal extraction.

1. The improved AI system can perform IA mitigation by dynamically selecting galaxy subsamples based on their physical properties (color, luminosity, redshift) to maximize the signal of cosmological shear while minimizing the contamination from intrinsic alignment.

2. It can employ a hybrid modeling approach that switches between linear-alignment models (NLA) and non-linear models (TATT), adapting the complexity of its analysis based on the chosen galaxy population and spatial scale.

Specific improvements derived directly from the paper:

1. The improved AI system can implement a multi-proxy selection strategy for IA mitigation, using combinations of:

2. Rest-frame color (e.g., Mr–Mz),

3. Star Formation Rate (sSFR), and

4. The 4000 Å break strength (DN4000).

This allows the AI to move beyond simple luminosity cuts and target populations that are demonstrably weakly aligned according to the paper's results (e.g., identifying blue galaxies at high redshift as having consistent IA signals near zero).

3. The system can utilize a data-driven approach for optimizing selection, specifically by training a classifier (e.g., using NLA posteriors) to identify galaxy properties that correlate with minimal IA amplitude, thereby informing optimal shear sample selection for future surveys (as suggested in Section 4.3).

4. The AI can employ a model-free intrinsic alignment amplitude calculation using the model-free approach (Equation 11), which scales the measured signal based on the ratio of two-point correlation functions, allowing it to estimate IA amplitudes across vast parameter spaces without relying solely on complex parametric models like NLA or TATT.

5. The system can perform joint fitting of shape–density correlations and density–density clustering signals (as suggested in Section 3.2), enabling the AI to disentangle the effects of intrinsic alignment from galaxy bias and large-scale structure, leading to more robust cosmological parameter constraints (e.g., fitting both NLA and TATT models jointly).

6. The system can incorporate survey-specific calibration adjustments by dynamically calculating shear responsivity matrices (R) based on the input imaging survey (DES, HSC, KiDS, SDSS), accounting for known biases like multiplicative biases and additive biases derived from image simulations.

7. The AI can perform uncertainty quantification using a comprehensive jackknife process (as detailed in Appendix D), providing rigorous error bars on IA measurements to ensure that reported cosmological constraints are not overly optimistic due to underestimation of noise or systematic errors.

We present direct measurements of the intrinsic alignments (IA) of over 2 million spectroscopic galaxies using DESI Data Release 1 and imaging from four lensing surveys: DES, HSC, KiDS, and SDSS. In this uniquely data-rich regime, we take initial steps towards a more tailored IA modelling approach by building a library of IA measurements across colour, luminosity, stellar mass, and redshift. We map the dependence between galaxy type -- in terms of rest-frame colour, strength of the 4000 Angstrom break, and specific star formation rate -- and IA amplitude; the bluest galaxies have an alignment consistent with zero, across low (0.05<z<0.5) and high (0.8<z<1.55) redshifts. In order to construct cosmic shear samples that are minimally impacted by IA but maintain maximum sample size and statistical power, we map the dependence of alignment with colour purity. Red, quenched galaxies are strongly aligned and the amplitude of the signal increases with luminosity, which is tightly correlated with stellar mass in our catalogues. For DESI galaxies between 0<z<1.5, trends in luminosity and colour alone are sufficient to explain the alignments we measure -- with no need for an explicit redshift dependence. In a companion paper (Jeffrey et al., in prep), we perform detailed modelling of the IA signals with significant detections, including model comparison. Finally, to direct efforts for future IA measurements, we juxtapose the colour-magnitude-redshift coverage of existing IA measurements against modern and future lensing surveys.