---
title: "How to read the DES Year 3 lensing and clustering analysis"
description: "Read the Dark Energy Survey Year 3 joint lensing and clustering paper through its shape catalog, redshift calibration, alignment model and key figures."
canonical_url: "https://sandboxphysics.com/stories/weak-lensing/reading-des-y3-lensing-and-clustering/"
language: "en"
page_type: "story"
alternate_languages:
  zh-CN: "https://sandboxphysics.com/zh/stories/weak-lensing/reading-des-y3-lensing-and-clustering/"
last_modified: "2026-11-14"
---

[Weak gravitational lensing and mass maps](https://sandboxphysics.com/stories/weak-lensing/)

# How to read the DES Year 3 lensing and clustering analysis

Sandbox Physics · 2026-11-14

The Dark Energy Survey Year 3 paper combines three two-point functions: cosmic shear, galaxy clustering, and the shear of background galaxies around foreground galaxies. Read it in this order. Section II says how 100 million shapes were calibrated and binned in redshift. Section IV and Table I say which nuisance parameters absorb what remains uncertain. Figures 5 and 8 show the shear data and the intrinsic-alignment amplitudes. Section V and Figures 25 and 26 show how much the answer moves when those choices change. The headline is a clustering amplitude $S_8=0.776\pm0.017$, consistent with the Planck prediction.

This guide reads arXiv:2105.13549v3, dated 21 March 2022, and the abstract of its cosmic-shear companion, arXiv:2105.13543v2, dated 29 September 2022. We did not compare these with the journal versions. We did not reproduce any DES number. Our [tutorial](https://sandboxphysics.com/stories/weak-lensing/recompute-shear-correlations/) tests estimators and failure modes on Gaussian simulations, nothing more.

## What the three functions are and why combine them

Cosmic shear, $\xi_\pm(\theta)$, correlates the shapes of source galaxies in four redshift bins. Galaxy clustering, $w(\theta)$, correlates the positions of foreground lens galaxies. Galaxy-galaxy lensing, $\gamma_{\rm t}(\theta)$, measures the tangential shear of sources around lenses. The paper calls the combination $3\times2$pt. Each function depends differently on the matter density, the clumpiness of matter and the bias that relates galaxies to matter. Combining them lets the data pin down nuisance parameters that would be degenerate in any one function. The paper calls this self-calibration, in Section V. [Cosmological Constraints from Galaxy Clustering and Weak Lensing, Abstract and Sec. V](https://arxiv.org/abs/2105.13549v3)

## Section II: the shape catalog

The shapes come from the Metacalibration pipeline, which calibrates the shear estimate from the measured galaxy images themselves, including biases from how galaxies are selected. It does not correct a detection bias tied to blended galaxies, which the paper puts at 2 to 3 percent in the multiplicative shear. Image simulations calibrate that remainder. The final catalog has 100 million galaxies over 4143 square degrees, an effective density of 5.9 per square arcminute and shape noise $\sigma_e=0.26$.

Two sentences in Section II answer the question of point-spread-function errors. Errors from modelling the telescope's blurring were shown to be negligible for Year 3, thanks to improved PSF modelling. The B-mode signal is consistent with zero. For the evidence behind both statements, Section II points to the shape-catalog paper.

## Section II: redshifts

Source redshifts come from a self-organizing map. Wide-field galaxies are grouped by colour into cells; deep-field galaxies, with more bands and less noise, are grouped into finer cells. Spectroscopic and many-band redshifts in the deep fields give each deep cell a redshift distribution. A simulation that injects synthetic galaxies into real images, called Balrog, then maps wide cells to deep cells. The four source bins have edges $z=0$, $0.36$, $0.63$, $0.87$ and $2.0$. Blending correlates the shear calibration with the shape of each bin's $n(z)$, and the analysis models the two together for the first time.

Our [tomography chapter](https://sandboxphysics.com/stories/weak-lensing/shear-correlations-and-tomography/) shows why this matters: moving 20 percent of far galaxies into a near bin raises its signal by 42 percent in our toy setup.

## Section IV and Table I: what is marginalized

The model has 6 cosmological parameters and 25 nuisance parameters in $\Lambda$CDM. For the sources, Table I lists Gaussian priors on a shift of each bin's mean redshift and on its multiplicative shear bias $m$.

**Table 6.1** Source-side priors from Table I of arXiv:2105.13549v3, mean and width, both times 100.

| Bin | Redshift shift, $10^2\,\Delta z$ | Shear bias, $10^2\,m$ |
| --- | --- | --- |
| 1 | $0.0\pm1.8$ | $-0.6\pm0.9$ |
| 2 | $0.0\pm1.5$ | $-2.0\pm0.8$ |
| 3 | $0.0\pm1.1$ | $-2.4\pm0.8$ |
| 4 | $0.0\pm1.7$ | $-3.7\pm0.8$ |

Intrinsic alignments use the tidal alignment and tidal torquing model. It has an amplitude $a_i$ and a redshift power law $\eta_i$ for each of two terms, with flat priors from $-5$ to $5$, plus a source bias $b_{\rm TA}$ between $0$ and $2$ and a pivot redshift $z_0=0.62$. The commonly used nonlinear alignment model is the subspace $a_2=b_{\rm TA}=0$. The 2004 analysis by Hirata and Seljak explains why alignment terms can bias or even reduce the lensing signal, including in cross-correlations between bins. [Intrinsic alignment-lensing interference as a contaminant of cosmic shear](https://arxiv.org/abs/astro-ph/0406275v3)

## Figures to read closely

Figure 5 shows $\xi_\pm(\theta)$ for every pair of source bins with the best-fit model. Read the bottom strip of each panel, the residual in units of the error, and the shaded regions. Dark and light shading mark scales excluded from the fiducial analysis, mainly because baryons reshape the matter power spectrum on small scales. The lighter shading marks scales that a $\Lambda$CDM-optimized variant puts back in. After all cuts, 462 of 1300 data points remain, with a total signal-to-noise of 87.

Figure 8 shows the alignment amplitudes $a_1$ and $a_2$ per source bin, fitted with two different lens samples. The two lens samples agree. The caption warns that the redshift evolution is forced to be a smooth power law.

Figure 25 replaces the mean-shift treatment of $n(z)$ with sampling over many full $n(z)$ realizations, and also drops the shear-ratio data. Figure 26 swaps in a nonlinear galaxy bias, free lens magnification, or the simpler alignment model. Appendix E reports that the simpler alignment model tightens the $\sigma_8$-$\Omega_{\rm m}$ constraint by only 13 percent. Because the extra amplitudes come out small, the paper judges the simpler model a fair approximation.

## Results and their status

**Table 6.2** Headline numbers as printed, mean with 68 percent limits.

| Analysis | $S_8$ | Status |
| --- | --- | --- |
| $3\times2$pt, $\Lambda$CDM | $0.776^{+0.017}_{-0.017}$ | arXiv v3 abstract |
| $3\times2$pt, $\Lambda$CDM, $\Omega_{\rm m}$ | $0.339^{+0.032}_{-0.031}$ | arXiv v3 abstract |
| Cosmic shear alone, fiducial | $0.759^{+0.025}_{-0.023}$ | companion v2 abstract |
| Cosmic shear alone, $\Lambda$CDM-optimized | $0.772^{+0.018}_{-0.017}$ | companion v2 abstract |

[Cosmology from Cosmic Shear and Robustness to Data Calibration, Abstract](https://arxiv.org/abs/2105.13543v2)

The two papers describe the Planck comparison with different statistics. The $3\times2$pt paper quotes a probability to exceed of 0.13 from the parameter difference and 0.48 from a second metric, the suspiciousness, and calls the results consistent; its Table VIII adds two more metrics, one of which gives 0.10. The cosmic-shear paper finds its $S_8$ values lower than the Planck prediction by $2.3\sigma$ and $2.1\sigma$, with $p=0.02$ and $0.05$, and still calls them statistically consistent. Both are published statements from the collaboration. The difference is in the data combination and the metric, not a contradiction between the papers.

The cosmic-shear companion names baryonic effects and intrinsic alignments as the limiting systematics, ahead of shape and redshift calibration. That is the open question this guide leaves you with. The fiducial analysis handles baryons by cutting the small scales they affect, as the shading in Figure 5 shows, so the information on those scales is still unused.

## What you can check yourself

- In [the weak-lensing mass map Lab](https://sandboxphysics.com/labs/weak-lensing-mass-map/), move the sources in front of a halo and watch the lensing efficiency vanish. That is the reason a redshift shift changes the predicted shear.
- In the tutorial, rotate shapes by $45^\circ$ and watch $\xi_-$ change sign. That is the B-mode logic behind the null test in Section II.
- The DES numbers themselves require the DES data release, covariance and likelihood code. We have not run them, and nothing on this site depends on unreleased survey products, including those of the Euclid mission.

## Try it in the Lab

[Weak-lensing mass map](https://sandboxphysics.com/labs/weak-lensing-mass-map/): Move the sources in front of a lens and see why redshift errors change what shapes measure.

## Sources

1. [Cosmological Constraints from Galaxy Clustering and Weak Lensing, Abstract and Sec. V](https://arxiv.org/abs/2105.13549v3)
2. [Intrinsic alignment-lensing interference as a contaminant of cosmic shear](https://arxiv.org/abs/astro-ph/0406275v3)
3. [Cosmology from Cosmic Shear and Robustness to Data Calibration, Abstract](https://arxiv.org/abs/2105.13543v2)

## Keep reading

Weak gravitational lensing and mass maps · Reading path

### From galaxy shapes to a map of mass

1. [Article **How can galaxies that are barely stretched map dark matter?**](https://sandboxphysics.com/stories/weak-lensing/galaxy-shapes-weigh-dark-matter/)
2. [Article **Convergence, shear and what a galaxy shape measures**](https://sandboxphysics.com/stories/weak-lensing/convergence-shear-and-galaxy-shapes/)
3. [Article **Turning a shear field into a mass map**](https://sandboxphysics.com/stories/weak-lensing/from-shear-to-a-mass-map/)
4. [Article **Shear correlation functions and redshift tomography**](https://sandboxphysics.com/stories/weak-lensing/shear-correlations-and-tomography/)
5. [Article **Recompute shear correlations and tomography from a simulation**](https://sandboxphysics.com/stories/weak-lensing/recompute-shear-correlations/)
6. [You are here **How to read the DES Year 3 lensing and clustering analysis**](https://sandboxphysics.com/stories/weak-lensing/reading-des-y3-lensing-and-clustering/)
