Weak Lensing Convergence Maps Research Articles

C3NN: Cosmological Correlator Convolutional Neural Network an Interpretable Machine-learning Framework for Cosmological Analyses

Modern cosmological research in large-scale structure has witnessed an increasing number of machine-learning applications. Among them, convolutional neural networks (CNNs) have received substantial attention due to their outstanding performance in image classification, cosmological parameter inference, and various other tasks. However, many models based on CNNs are criticized as “black boxes” due to the difficulties in relating their outputs intuitively and quantitatively to the cosmological fields under investigation. To overcome this challenge, we present the Cosmological Correlator Convolutional Neural Network (C3NN)—a fusion of CNN architecture and cosmological N-point correlation functions (NPCFs). We demonstrate that its output can be expressed explicitly in terms of the analytically tractable NPCFs. Together with other auxiliary algorithms, we can open the “black box” by quantitatively ranking different orders of the interpretable outputs based on their contribution to classification tasks. As a proof of concept, we demonstrate this by applying our framework to a series of binary classification tasks using Gaussian and log-normal random fields and relating its outputs to the NPCFs describing the two fields. Furthermore, we exhibit the model’s ability to distinguish different dark energy scenarios (w 0 = −0.95 and −1.05) using N-body simulated weak-lensing convergence maps and discuss the physical implications coming from their interpretability. With these tests, we show that C3NN combines advanced aspects of machine learning architectures with the framework of cosmological NPCFs, thereby making it an exciting tool to extract physical insights in a robust and explainable way from observational data.

The FLAMINGO project: baryonic impact on weak gravitational lensing convergence peak counts

ABSTRACT Weak gravitational lensing convergence peaks, the local maxima in weak lensing convergence maps, have been shown to contain valuable cosmological information complementary to commonly used two-point statistics. To exploit the full power of weak lensing for cosmology, we must model baryonic feedback processes because these reshape the matter distribution on non-linear and mildly non-linear scales. We study the impact of baryonic physics on the number density of weak lensing peaks using the FLAMINGO cosmological hydrodynamical simulation suite. We generate ray-traced full-sky convergence maps mimicking the characteristics of a Stage IV weak lensing survey. We compare the number densities of peaks in simulations that have been calibrated to reproduce the observed galaxy mass function and cluster gas fraction or to match a shifted version of these, and that use either thermally driven or jet active galactic nucleus feedback. We show that the differences induced by realistic baryonic feedback prescriptions (typically 5–30 per cent for κ = 0.1–0.4) are smaller than those induced by reasonable variations in cosmological parameters (20–60 per cent for κ = 0.1–0.4) but must be modelled carefully to obtain unbiased results. The reasons behind these differences can be understood by considering the impact of feedback on halo masses, or by considering the impact of different cosmological parameters on the halo mass function. Our analysis demonstrates that, for the range of models we investigated, the baryonic suppression is insensitive to changes in cosmology up to κ ≈ 0.4 and that the higher κ regime is dominated by Poisson noise and cosmic variance.

The MillenniumTNG Project: the impact of baryons and massive neutrinos on high-resolution weak gravitational lensing convergence maps

ABSTRACT We study weak gravitational lensing convergence maps produced from the MillenniumTNG simulations by direct projection of the mass distribution on the past backwards lightcone of a fiducial observer. We explore the lensing maps over a large dynamic range in simulation mass and angular resolution, allowing us to establish a clear assessment of numerical convergence. By comparing full physics hydrodynamical simulations with corresponding dark-matter-only runs, we quantify the impact of baryonic physics on the most important weak lensing statistics. Likewise, we predict the impact of massive neutrinos reliably far into the non-linear regime. We also demonstrate that the ‘fixed & paired’ variance suppression technique increases the statistical robustness of the simulation predictions on large scales not only for time slices but also for continuously output lightcone data. We find that both baryonic and neutrino effects substantially impact weak lensing shear measurements, with the latter dominating over the former on large angular scales. Thus, both effects must explicitly be included to obtain sufficiently accurate predictions for stage IV lensing surveys. Reassuringly, our results agree accurately with other simulation results where available, supporting the promise of simulation modelling for precision cosmology far into the non-linear regime.

Starlet higher order statistics for galaxy clustering and weak lensing

We present a first application to photometric galaxy clustering and weak lensing of wavelet-based multi-scale (beyond two points) summary statistics: starlet peak counts and starlet ℓ1-norm. Peak counts are the local maxima in the map, and ℓ1-norm is computed via the sum of the absolute values of the starlet (wavelet) decomposition coefficients of a map, providing a fast multi-scale calculation of the pixel distribution, encoding the information of all pixels in the map. We employ the cosmo-SLICS simulations sources and lens catalogues, and we compute wavelet-based non-Gaussian statistics in the context of combined probes and their potential when applied to the weak-lensing convergence maps and galaxy maps. We obtain forecasts on the matter density parameter Ωm, the reduced Hubble constant h, the matter fluctuation amplitude σ8, and the dark energy equation of state parameter w0. In our setting for this first application, we consider the two probes to be independent. We find that the starlet peaks and the ℓ1-norm represent interesting summary statistics that can improve the constraints with respect to the power spectrum, even in the case of photometric galaxy clustering and when the two probes are combined.

Position-dependent correlation function of weak-lensing convergence

We provide a systematic study of the position-dependent correlation function in weak-lensing convergence maps and its relation to the squeezed limit of the three-point correlation function (3PCF) using state-of-the-art numerical simulations. We relate the position-dependent correlation function to its harmonic counterpart, i.e., the position-dependent power spectrum or equivalently the integrated bispectrum (IB). We use a recently proposed improved fitting function, BiHalofit, for the bispectrum to compute the theoretical predictions as a function of source redshifts. In addition to low redshift results (${z}_{s}=1.0--2.0$) we also provide results for maps inferred from lensing of the cosmic microwave background (CMB), i.e., ${z}_{s}=1100$. We include a Euclid-type realistic survey mask and noise. In agreement with the recent studies on the position-dependent power spectrum, we find that the results from simulations are consistent with the theoretical expectations when appropriate corrections are included. Performing a rough estimate, we find that the signal-to-noise (S/N) for the detection of position-dependent correlation function from Euclid-type mask with ${f}_{\mathrm{sky}}=0.35$, can range between 6--12 depending on the value of the intrinsic ellipticity distribution parameter ${\ensuremath{\sigma}}_{\ensuremath{\epsilon}}=0.3--1.0$. For reconstructed $\ensuremath{\kappa}$ maps using an ideal CMB survey the $\mathrm{S}/\mathrm{N}\ensuremath{\approx}1.8$. We also found that a 10% deviation in ${\ensuremath{\sigma}}_{8}$ can be detected using IB for the optimistic case of ${\ensuremath{\sigma}}_{\ensuremath{\epsilon}}=0.3$ with a $\mathrm{S}/\mathrm{N}\ensuremath{\approx}5$. The S/N for such detection in case of ${\mathrm{\ensuremath{\Omega}}}_{M}$ is lower.

Almanac: Weak Lensing power spectra and map inference on the masked sphere

We present a field-based signal extraction of weak lensing from noisy observations on the curved and masked sky. We test the analysis on a simulated Euclid-like survey, using a Euclid-like mask and noise level. To make optimal use of the information available in such a galaxy survey, we present a Bayesian method for inferring the angular power spectra of the weak lensing fields, together with an inference of the noise-cleaned tomographic weak lensing shear and convergence (projected mass) maps. The latter can be used for field-level inference with the aim of extracting cosmological parameter information including non-gaussianity of cosmic fields. We jointly infer all-sky $E$-mode and $B$-mode tomographic auto- and cross-power spectra from the masked sky, and potentially parity-violating $EB$-mode power spectra, up to a maximum multipole of $\ell_{\rm max}=2048$. We use Hamiltonian Monte Carlo sampling, inferring simultaneously the power spectra and denoised maps with a total of $\sim 16.8$ million free parameters. The main output and natural outcome is the set of samples of the posterior, which does not suffer from leakage of power from $E$ to $B$ unless reduced to point estimates. However, such point estimates of the power spectra, the mean and most likely maps, and their variances and covariances, can be computed if desired.

Weak lensing trispectrum and Kurt-spectra

We introduce two kurt-spectra to probe fourth-order statistics of weak lensing convergence maps. Using state-of-the-art numerical simulations, we study the shapes of these kurt-spectra as a function of source redshifts and smoothing angular scales. We employ a pseudo-Cℓ approach to estimate the spectra from realistic convergence maps in the presence of an observational mask and noise for stage-IV large-scale structure surveys. We compare these results against theoretical predictions calculated using the FFTLog formalism, and find that a simple nonlinear clustering model — the hierarchical ansatz — can reproduce the numerical trends for the kurt-spectra in the nonlinear regime. In addition, we provide estimators for beyond fourth-order spectra where no definitive analytical results are available, and present corresponding results from numerical simulations.

Investigating cosmological GAN emulators using latent space interpolation

ABSTRACT Generative adversarial networks (GANs) have been recently applied as a novel emulation technique for large-scale structure simulations. Recent results show that GANs can be used as a fast and efficient emulator for producing novel weak lensing convergence maps as well as cosmic web data in 2D and 3D. However, like any algorithm, the GAN approach comes with a set of limitations, such as an unstable training procedure, inherent randomness of the produced outputs, and difficulties when training the algorithm on multiple data sets. In this work, we employ a number of techniques commonly used in the machine learning literature to address the mentioned limitations. Specifically, we train a GAN to produce weak lensing convergence maps and dark matter overdensity field data for multiple redshifts, cosmological parameters, and modified gravity models. In addition, we train a GAN using the newest Illustris data to emulate dark matter, gas, and internal energy distribution data simultaneously. Finally, we apply the technique of latent space interpolation as a tool for understanding the feature space of the GAN algorithm. We show that the latent space interpolation procedure allows the generation of outputs with intermediate cosmological parameters that were not included in the training data. Our results indicate a 1–20 per cent difference between the power spectra of the GAN-produced and the test data samples depending on the data set used and whether Gaussian smoothing was applied. Similarly, the Minkowski functional analysis indicates a good agreement between the emulated and the real images for most of the studied data sets.

Higher order spectra of weak lensing convergence maps in parametrized theories of modified gravity

ABSTRACT We compute the low-ℓ limit of the family of higher order spectra for projected (2D) weak lensing convergence maps. In this limit these spectra are computed to an arbitrary order using tree-level perturbative calculations. We use the flat-sky approximation and Eulerian perturbative results based on a generating function approach. We test these results for the lower order members of this family, i.e. the skew- and kurt-spectra against state-of-the-art simulated all-sky weak lensing convergence maps and find our results to be in very good agreement. We also show how these spectra can be computed in the presence of a realistic sky-mask and Gaussian noise. We generalize these results to 3D and compute the equal-time higher order spectra. These results will be valuable in analysing higher order statistics from future all-sky weak lensing surveys such as the Euclid survey at low-ℓ modes. As illustrative examples, we compute these statistics in the context of the Horndeski and beyond Horndeski theories of modified gravity. They will be especially useful in constraining theories such as the Gleyzes–Langlois–Piazza–Vernizzi (GLPV) theories and degenerate higher order scalar-tensor theories as well as the commonly used normal-branch of Dvali–Gabadadze–Porrati model, clustering quintessence models and scenarios with massive neutrinos.

CosmoGAN: creating high-fidelity weak lensing convergence maps using Generative Adversarial Networks

Inferring model parameters from experimental data is a grand challenge in many sciences, including cosmology. This often relies critically on high fidelity numerical simulations, which are prohibitively computationally expensive. The application of deep learning techniques to generative modeling is renewing interest in using high dimensional density estimators as computationally inexpensive emulators of fully-fledged simulations. These generative models have the potential to make a dramatic shift in the field of scientific simulations, but for that shift to happen we need to study the performance of such generators in the precision regime needed for science applications. To this end, in this work we apply Generative Adversarial Networks to the problem of generating weak lensing convergence maps. We show that our generator network produces maps that are described by, with high statistical confidence, the same summary statistics as the fully simulated maps.

Weak lensing by voids in weak lensing maps

ABSTRACT Cosmic voids are an important probe of large-scale structure that can constrain cosmological parameters and test cosmological models. We present a new paradigm for void studies: void detection in weak lensing convergence maps. This approach identifies objects that relate directly to our theoretical understanding of voids as underdensities in the total matter field and presents several advantages compared to the customary method of finding voids in the galaxy distribution. We exemplify this approach by identifying voids using the weak lensing peaks as tracers of the large-scale structure. We find self-similarity in the void abundance across a range of peak signal-to-noise selection thresholds. The voids obtained via this approach give a tangential shear signal up to ∼40 times larger than voids identified in the galaxy distribution.

On the effect of projections on convergence peak counts and Minkowski functionals

The act of projecting data sampled on the surface of the celestial sphere onto a regular grid on the plane can introduce error and a loss of information. This paper evaluates the effects of different planar projections on non-Gaussian statistics of weak lensing convergence maps. In particular we investigate the effect of projection on peak counts and Minkowski Functionals (MFs) derived from convergence maps and the suitability of a number of projections at matching the peak counts and MFs obtained from a sphere. We find that the peak counts derived from planar projections consistently overestimate the counts at low SNR thresholds and underestimate at high SNR thresholds across the projections evaluated, although the difference is reduced when smoothing of the maps is increased. In the case of the Minkowski Functionals, V0 is minimally affected by projection used, while projected V1 and V2 are consistently overestimated with respect to the spherical case.

Distinguishing general relativity andf(R)gravity with the gravitational lensing Minkowski functionals

We explore the Minkowski functionals of weak lensing convergence map to distinguish between $f(R)$ gravity and the general relativity (GR). The mock weak lensing convergence maps are constructed with a set of high-resolution simulations assuming different gravity models. It is shown that the lensing MFs of $f(R)$ gravity can be considerably different from that of GR because of the environmentally dependent enhancement of structure formation. We also investigate the effect of lensing noise on our results, and find that it is likely to distinguish F5, F6 and GR gravity models with a galaxy survey of $\sim3000$ degree$^2$ and with a background source number density of $n_g=30~{\rm arcmin}^{-2}$, comparable to an upcoming survey dark energy survey (DES). We also find that the $f(R)$ signal can be partially degenerate with the effect of changing cosmology, but combined use of other observations, such as the cosmic microwave background (CMB) data, can help break this degeneracy.

Mass–concentration relation and weak lensing peak counts

The statistics of peaks in weak lensing convergence maps is a promising tool to investigate both the properties of dark matter haloes and constrain the cosmological parameters. We study how the number of detectable peaks and its scaling with redshift depend upon the cluster dark matter halo profiles and use peak statistics to constrain the parameters of the mass - concentration (MC) relation. We investigate which constraints the Euclid mission can set on the MC coefficients also taking into account degeneracies with the cosmological parameters. To this end, we first estimate the number of peaks and its redshift distribution for different MC relations. We find that the steeper the mass dependence and the larger the normalisation, the higher is the number of detectable clusters, with the total number of peaks changing up to $40\%$ depending on the MC relation. We then perform a Fisher matrix forecast of the errors on the MC relation parameters as well as cosmological parameters. We find that peak number counts detected by Euclid can determine the normalization $A_v$, the mass $B_v$ and redshift $C_v$ slopes and intrinsic scatter $\sigma_v$ of the MC relation to an unprecedented accuracy being $\sigma(A_v)/A_v = 1\%$, $\sigma(B_v)/B_v = 4\%$, $\sigma(C_v)/C_v = 9\%$, $\sigma(\sigma_v)/\sigma_v = 1\%$ if all cosmological parameters are assumed to be known. Should we relax this severe assumption, constraints are degraded, but remarkably good results can be restored setting only some of the parameters or combining peak counts with Planck data. This precision can give insight on competing scenarios of structure formation and evolution and on the role of baryons in cluster assembling. Alternatively, for a fixed MC relation, future peaks counts can perform as well as current BAO and SNeIa when combined with Planck.

Reconstructing the thermal Sunyaev–Zel'dovich effect in 3D

The thermal Sunyaev-Zel'dovich (tSZ) effect measures the line-of-sight projection of the thermal pressure of free electrons and lacks any redshift information. By cross correlating the tSZ effect with an external cosmological tracer we can recover a good fraction of this lost information. Weak lensing (WL) is thought to provide an unbiased probe of the dark Universe, with many WL surveys having sky coverage that overlaps with tSZ surveys. Generalising the tomographic approach, we advocate the use of the spherical Fourier-Bessel (sFB) expansion to perform an analysis of the cross-correlation between the projected (2D) tSZ Compton $y$-parameter maps and 3D weak lensing convergence maps. We use redshift dependent linear biasing and the halo model as a tool to investigate the tSZ-WL cross-correlations in 3D. We use the Press-Schechter (PS) and the Sheth-Tormen (ST) mass-functions in our calculations, finding that the results are quite sensitive to detailed modelling. We provide detailed analysis of surveys with photometric and spectroscopic redshifts. The signal-to-noise (S/N) of the cross-spectra $\mathcal{C}_{\ell} (k)$ for individual 3D modes, defined by the radial and tangential wave numbers $(k;\ell)$, remains comparable to, but below, unity though optimal binning is expected to improve this. The results presented can be generalised to analyse other CMB secondaries, such as the kinetic Sunyaev-Zel'dovich (kSZ) effect.

Cross-correlating Sunyaev–Zel'dovich and weak lensing maps

We present novel statistical tools to cross-correlate frequency cleaned thermal Sunyaev–Zel'dovich (tSZ) maps and tomographic weak lensing (wl) convergence maps. Moving beyond the lowest order cross-correlation, we introduce a hierarchy of mixed higher order statistics, the cumulants and cumulant correlators, to analyse non-Gaussianity in real space, as well as corresponding polyspectra in the harmonic domain. Using these moments, we derive analytical expressions for the joint two-point probability distribution function for smoothed tSZ (y) and convergence (κ) maps. The presence of tomographic information allows us to study the evolution of higher order mixed tSZ–wl statistics with redshift. We express the joint PDFs pκy(κ, y) in terms of individual one-point PDFs [pκ(κ), py(y)] and the relevant bias functions [bκ(κ), by(y)]. Analytical results for two different regimes are presented that correspond to the small and large angular smoothing scales. Results are also obtained for corresponding hotspots in the tSZ and convergence maps. In addition to results based on hierarchical techniques and perturbative methods, we present results of calculations based on the lognormal approximation. The analytical expressions derived here are generic and applicable to cross-correlation studies of arbitrary tracers of large-scale structure including, e.g., that of tSZ and soft X-ray background. We provide detailed comparison of our analytical results against state of the art Millennium Gas Simulations with and without non-gravitational effects such as pre-heating and cooling. Comparison of these results with gravity only simulations, shows reasonable agreement and can be used to isolate effect of non-gravitational physics from observational data.

Cosmology with Minkowski functionals and moments of the weak lensing convergence field

We compare the efficiency of moments and Minkowski functionals (MFs) in constraining the subset of cosmological parameters $({\ensuremath{\Omega}}_{m},w,{\ensuremath{\sigma}}_{8})$ using simulated weak lensing convergence maps. We study an analytic perturbative expansion of the MFs [T. Matsubara, Phys. Rev. D 81, 083505 (2010); D. Munshi et al., Mon. Not. R. Astron. Soc. 419, 536 (2012)] in terms of the moments of the convergence field and of its spatial derivatives. We show that this perturbation series breaks down on smoothing scales below ${5}^{\ensuremath{'}}$, while it shows a good degree of convergence on larger scales ($\ensuremath{\sim}{15}^{\ensuremath{'}}$). Most of the cosmological distinguishing power is lost when the maps are smoothed on these larger scales. We also show that, on scales comparable to ${1}^{\ensuremath{'}}$, where the perturbation series does not converge, cosmological constraints obtained from the MFs are approximately 1.5--2 times better than the ones obtained from the first few moments of the convergence distribution---provided that the latter include spatial information, either from moments of gradients or by combining multiple smoothing scales. Including a set of either these moments or the MFs can significantly tighten constraints on cosmological parameters, compared to the conventional method of using the power spectrum alone.

PROBING PRIMORDIAL NON-GAUSSIANITY WITH WEAK-LENSING MINKOWSKI FUNCTIONALS

We study the cosmological information contained in the Minkowski Functionals (MFs) of weak gravitational lensing convergence maps. We show that the MFs provide strong constraints on the local type primordial non-Gaussianity parameter f_NL. We run a set of cosmological N-body simulations and perform ray-tracing simulations of weak lensing, to generate 100 independent convergence maps of 25 deg^2 field-of-view for f_NL = -100, 0 and 100. We perform a Fisher analysis to study the degeneracy among other cosmological parameters such as the dark energy equation of state parameter w and the fluctuation amplitude sigma_8. We use fully nonlinear covariance matrices evaluated from 1000 ray-tracing simulations. For the upcoming wide-field observations such as Subaru Hyper Suprime-Cam survey with the proposed survey area of 1500 deg^2, the primordial non-Gaussianity can be constrained with a level of f_NL ~ 80 and w ~ 0.036 by weak lensing MFs. If simply scaled by the effective survey area, a 20000 deg^2 lensing survey using Large Synoptic Survey Telescope will give constraints of f_NL ~ 25 and w ~ 0.013. We show that these constraints can be further improved by a tomographic method using source galaxies in multiple redshift bins.

From weak lensing to non-Gaussianity via Minkowski functionals

We present a new harmonic-domain approach for extracting morphological information, in the form of Minkowski Functionals (MFs), from weak lensing (WL) convergence maps. Using a perturbative expansion of the MFs, which is expected to be valid for the range of angular scales probed by most current weak-lensing surveys, we show that the study of three generalized skewness parameters is equivalent to the study of the three MFs defined in two dimensions. We then extend these skewness parameters to three associated skew-spectra which carry more information about the convergence bispectrum than their one-point counterparts. We discuss various issues such as noise and incomplete sky coverage in the context of estimation of these skew-spectra from realistic data. Our technique provides an alternative to the pixel-space approaches typically used in the estimation of MFs, and it can be particularly useful in the presence of masks with non-trivial topology. Analytical modeling of weak lensing statistics relies on an accurate modeling of the statistics of underlying density distribution. We apply three different formalisms to model the underlying dark-matter bispectrum: the hierarchical ansatz, halo model and a fitting function based on numerical simulations; MFs resulting from each of these formalisms are computed and compared. We investigate the extent to witch late-time gravity-induced non-Gaussianity (to which weak lensing is primarily sensitive) can be separated from primordial non-Gaussianity and how this separation depends on source redshift and angular scale.

Comparison of halo detection from noisy weak lensing convergence maps with Gaussian smoothing and MRLens treatment

Taking into account the noise from intrinsic ellipticities of source galaxies, we study the efficiency and completeness of halo detections from weak lensing convergence maps. Particularly, with numerical simulations, we compare the Gaussian filter with the so called MRLens treatment based on the modification of the Maximum Entropy Method. For a pure noise field without lensing signals, a Gaussian smoothing results in a residual noise field that is approximately Gaussian in terms of statistics if a large enough number of galaxies are included in the smoothing window. On the other hand, the noise field after the MRLens treatment is significantly non-Gaussian, resulting in complications in characterizing the noise effects. Considering weak-lensing cluster detections, although the MRLens treatment effectively deletes false peaks arising from noise, it removes the real peaks heavily due to its inability to distinguish real signals with relatively low amplitudes from noise in its restoration process. The higher the noise level is, the larger the removal effects are for the real peaks. For a survey with a source density ng ∼ 30 arcmin−2, the number of peaks found in an area of 3 × 3 deg2 after MRLens filtering is only ∼ 50 for the detection threshold κ = 0.02, while the number of halos with M > 5 × 1013 M⊙ and with redshift z ⩽ 2 in the same area is expected to be ∼ 530. For the Gaussian smoothing treatment, the number of detections is ∼ 260, much larger than that of the MRLens. The Gaussianity of the noise statistics in the Gaussian smoothing case adds further advantages for this method to circumvent the problem of the relatively low efficiency in weak-lensing cluster detections. Therefore, in studies aiming to construct large cluster samples from weak-lensing surveys, the Gaussian smoothing method performs significantly better than the MRLens treatment.