Cosmic Shear constraints from HSC Year 3 with clustering calibration of the tomographic redshift distributions from DESI

Abstract

We reanalyze cosmological constraints from Hyper Suprime-Cam (HSC) Y3 shear-shear correlation function using new calibration of the tomographic redshift distribution via the clustering redshifts method with DESI spectroscopy presented in Choppin de Janvry et al. (2025a). We present both importance sampling of the original MCMC chains by HSC, applying the weights of our newly calibrated $Δz$ priors, as well as full MCMC analysis with new photometric redshift distributions, finding consistent results between the two. We obtain the growth of structure parameter $S_8\equivσ_8\sqrt{Ω_m/0.3}=0.805\pm{0.018}$, compared to previous HSC Y3 result of $S_8=0.769^{+0.031}_{-0.034}$, which is a 1.8 reduction of error due to the improved clustering redshift calibrations, with the central value shifting considerably higher towards Planck cosmology. With the new photometric redshift calibration, HSC Y3 has comparable constraining power to the recent KIDS Legacy and DES Y3. Combining all three gives $S_8=0.813^{+0.009}_{-0.010}$, which can be compared to $S_8=0.828\pm 0.012$ from CMB lensing. Overall there is no evidence for deviation from Planck on $S_8$ in any of the weak lensing analyses, and combining galaxy lensing with CMB lensing gives a sub-percent constraint $S_8=0.818\pm 0.007$, comparable in both precision and value to the most recent Planck+ACT+DESI constraint $S_8=0.812\pm 0.007$.

Figures (5)

• Figure 1: $n(z)$ distributions for each of the 4 tomographic Bins, spanning $z_p\sim0.3$ to $z_p\sim1.5$ where $z_p$ is the best estimate of the DNNz photo-$z$ algorithm DNNZHSCNishizawa2020 used for tomographic Bin selection, as described in section \ref{['sec:data:hsc']}. The $n(z)$ for each Bin is displayed with the median posterior sample and the $1\sigma$ widths adapted from ChoppinDeJanvry2025a. Here, the $n(z)$ shown includes all corrections to galaxy bias and magnification effects.
• Figure 2: To assess the robustness of the measurement, we exclude Bin 3 and Bin 4 respectively from the chain inference. The measurements are shown to be consistent between the two and consistent with the "All Bins" measurement.
• Figure 3: Comparing the sampled results (section \ref{['sec:method:mcmc']}) from the two $n(z)$ set of distributions at different scale cuts with the importance re-weighting (section \ref{['sec:method:importance']}) approach in real space and to the HSC Y3 fiducial result. The shift posteriors are artificially shifted by the expectation of the inferred shift in ChoppinDeJanvry2025a, as they are by default centered around 0 for the sampling approach, since new $n(z)$ distributions are used.
• Figure 4: $\Omega_m$, $S_8$, $\Delta z_3$, $\Delta z_4$ parameters compared under different measurements. This includes the sampling approach, where the $n(z)$ distributions at two different scale cuts are included, as well as the measurements obtained when excluding either Bin 3 or Bin 4. These are then compared to the importance re-weighting and fiducial results of the Real and Fourier HSC Y3 analyses Li2023_HSCY3_CosmicShearDalal2023CosmoShearHSC. The square marker displays the 1D posterior mode obtained with getDist. The diamond marker is the Maximum A Posteriori (MAP) value. The background gray area represents the projected fiducial measurement obtained with the $0.3-3$h$^{-1}$Mpc scale cut for the $n(z)$. The $n(z)$ distributions are not the same for HSC Y3 samples (derived from HSCClusteringRau2023) and this work (from ChoppinDeJanvry2025a) : the posteriors of this work have been artificially shifted to match the mean of the shifts found in ChoppinDeJanvry2025a.
• Figure 5: Measurements from different surveys measuring the growth of structure, from CMB Measurements and Weak Lensing Measurements. The results are further compared to the results obtained in this work, as well as combined with the $S_8$ constraints obtained. The source of every cited measurement is discussed in text.