ORIGINS Lensing Day

Thursday Nov 14, 2024, 9:00 AM → 6:20 PM Europe/Berlin

MIAPbP Seminar Room (MIAPbP@ORIGINS)

The ORIGINS Lensing Day is a one day meeting for researchers in the Munich/Garching area with an interest in gravitational lensing. 

This meeting is supported by ORIGINS Research Unit D: From the Large-Scale Structure to Galaxies, Stars and Planets and organized by Simona Vegetti, Conor O'Riordan, Aymeric Galan (MPA) and Leon Roman Ecker (LMU).

9:00 AM → 10:30 AM Morning session 1

Clusters and weak lensing

Convener: Aymeric Galan (TUM/MPA)

9:00 AM Introduction

9:05 AM Strong lensing by cluster galaxies

Speaker: Massimo Meneghetti (INAF)

9:50 AM Strong lens modelling of galaxy clusters with JWST data

The James Webb Space Telescope is revolutionizing the field of strong gravitational lensing in galaxy clusters, allowing us to go further and fainter in our search for multiply imaged galaxies. I will present the recent results of the Canadian NIRISS Unbiased Cluster Survey (CANUCS), targeting 5 massive clusters with 3 instruments onboard JWST. With the example of MACS 0416 - the cluster with the largest dataset of multiple images with spectroscopic redshifts - I will demonstrate the effectiveness of JWST (in particular its NIRISS slitless spectrograph) compared to HST and MUSE, which have until now been the cornerstone facilities for strong lensing studies. I will also go through some of the lessons we learned when modelling the mass distributions with the extended datasets.

Speaker: Gregor Rihtaršič (University of Ljubljana)

Rihtarsic_origins_lensing.pdf

Rihtarsic_origins_lensing.pptx

10:10 AM Weak lensing mass calibration of clusters (SPTxDES data)

In this talk, I will discuss how gravitational weak lensing data can be leveraged to calibrate the masses of galaxy clusters. The South Pole Telescope (SPT) survey detects clusters via the thermal Sunyaev-Zel'dovich (SZ) effect, but this method does not yield direct mass measurements. Gravitational weak lensing provides a complementary approach, enabling accurate mass estimations that are essential for deriving robust cosmological constraints based on galaxy cluster abundance. I will outline the methodology used in the latest SPTxDES analysis, conducted by our team at LMU, to achieve precise mass calibration through weak lensing.

Speaker: Asmaa Mazoun (Technische Universität München (TUM))

10:30 AM → 11:00 AM Coffee break

11:00 AM → 12:40 PM Morning session 2

Weak lensing and radio lens finding

Convener: Leon Roman Ecker (LMU Munich (USM)/ Max-Planck Institute for extraterrestrial physics (MPE))

11:00 AM Cosmological Constraints with Weak Lensing Scattering Transform

CMB observations show that early density fluctuations were nearly Gaussian. The gravitational evolution of matter formed large structures, and nonlinear evolution introduced non-Gaussianity, making it challenging to analyze. In this project, we employ the scattering transform, a powerful statistical tool that shares ideas with convolutional neural networks (CNNs) but requires no training or tuning. The scattering transform generates a compact set of coefficients—scattering coefficients—that capture non-Gaussian information through hierarchical, interpretable summary statistics. This method is particularly suited for fields with localized structures and hierarchical clustering, like cosmological density fields. We apply the scattering transform to weak lensing convergence maps from the CosmoGridV1 simulations, using Gaussian-enveloped, harmonic kernels. These compact, stable descriptors serve as robust statistics, allowing us to make tight cosmological parameter forecasts with the Fisher information matrix, which is 14 times tighter than that of the power spectrum (Cheng et al. 2020). Additionally, we develop a deep-learning emulator trained across cosmologies, incorporating noise, to predict scattering coefficients and employ MCMC to constrain cosmological parameters. Further, a tomographic analysis using auto and cross correlations further tightens these constraints.

Speaker: Sijin Chen (LMU Munich)

11:20 AM Higher-order WL Statistics and Intrinsic Alignment

Higher-order WL statistics can capture the non-Gaussian nature of the Cosmic Web and exploit that to offer stricter constraints on Cosmological/MG parameters. In this talk I will start by elaborating on the significance and advantages of looking into these higher-order statistics.

There are certain known systematic errors (like galaxy Intrinsic Alignment) that exist in WL data, which need to be accounted for to help the future Stage-IV surveys infer cosmological statistics and constraints effectively. IA is anticipated to significantly impact lensing statistics, and hence, is important to be accounted for.

My ongoing Mater’s thesis investigates the effect of infusing IA as a systematic to WL Maps generated with FORGE-BRIDGE, MGLenS using two-point and beyond two-point WL statistics for the f(R) and the nDGP MG models. I will proceed by presenting some results that I already have obtained and end by giving a glimpse into what is the aim of my project: The statistical measures are to be used to train a Gaussian Process Regression Emulator to reliably interpolate between IA, MG and Cosmological Parameters to provide physical insight wherever analytical models are not available, followed by MCMC analysis of the trained Emulator.

Speaker: Mehar Chawla

WL Day Talk.key

11:40 AM Cosmology with weak lensing peaks and voids

Maximising the information that can be extracted from weak lensing measurements is a key goal for LSST and Euclid. This is typically achieved through statistics that are complementary to the cosmic shear two-point correlation function.

In this talk I will present the development of two such complementary statistics.

The first is weak lensing peaks. Typically, only the peak abundance is used to test cosmology. I will show how the clustering of peaks can be used to significantly improve the sensitivity of weak lensing peaks to cosmological parameters.

Secondly, I will present weak lensing voids, a new weak lensing statistic, that corresponds to a new void definition, which also captures a wealth of cosmological information.

I use the cosmoSLICS, FORGE, and BRIDGE simulations to measure the weak lensing peak and void statistics for a range of cosmological and gravitational parameters. The simulation data is used to train a machine learning emulator, which is used to generate parameter constraint forecasts from mock observations.

I will show that both peaks and voids can double our constraining power on cosmological models, relative to the standard shear two-point correlation function.

Finally, I will present ongoing work applying these methods to the DES Y3 data.

Speaker: Chris Davies (USM LMU)

12:00 PM Watershed on the curved sky

Cosmic voids are vast underdense regions in the universe that offer unique insights into dark energy and the large-scale structure of the cosmos. This talk presents a novel method for identifying voids in weak lensing convergence maps using a 2D watershed algorithm. By analyzing data from the CosmoGridV1 simulation and the Dark Energy Survey (DES), we demonstrate the algorithm's capability to detect voids and extract their statistical properties efficiently. Our results highlight the relationship between void characteristics and cosmological parameters, emphasizing the utility of void statistics as powerful cosmological probes. This approach provides a robust framework for future studies leveraging data from upcoming surveys like Euclid and LSST, with implications for advancing our understanding of cosmic evolution.

Speaker: Xu Han (LMU Munich)

Watershed Void Finder.pdf

12:20 PM Finding lenses with LOFAR: A case study

The International LOFAR Telescope (ILT) is currently observing the entire northern sky at a resolution of 0.3 arcsec as part of the LOFAR Two-metre Sky Survey (LoTSS). Its large survey area and high resolution combined with steep number counts towards low frequencies make the ILT well suited for finding galaxy-scale lenses, but the inherent complexity of radio source morphologies make the automated identification of these lenses a challenging task.
In this talk I will present a case-study where I apply convolutional neural networks, trained on realistic ILT visibility simulations, to real ILT data, and will discuss some of the challenges of finding lenses at radio wavelengths.

Speaker: Willem de Roo

12:40 PM → 2:10 PM Lunch

2:10 PM → 3:50 PM Afternoon session 1

Galaxy scale modelling and substructure

Convener: Conor O'Riordan (Max Planck Institute for Astrophysics)

2:10 PM Low redshift Strong Lensing

The Euclid mission is expected to discover approximately 150,000 strong gravitational lenses, greatly expanding the known lens population (Collett et al. 2015). Low-redshift early-type galaxies (ETGs) offer a unique opportunity to probe the stellar initial mass function (IMF) as their Einstein radii, typically smaller than the effective radius, enable mass measurements more sensitive to baryonic matter since at these scales it dominates dark matter by far.

Prior studies of massive ETGs found Milky Way-like IMFs (Smith & Lucey 2013; Smith et al. 2015), conflicting with stellar population models that suggest heavier IMFs (Conroy 2014). Dynamical modeling with MUSE and SINFONI data also reveals substantial mass-to-light ratio gradients (Mehrgan et al. 2024; Neureiter et al. 2023), highlighting the potential of combining dynamical and lensing techniques for IMF constraints (Thomas et al. 2011; Newman et al. 2017).

The project focuses on low-redshift strong lens ETGs to facilitate combined analyses. VIS imaging will enable initial lensing mass models, providing a basis for follow-up with LBT-LUCY and VLT-MUSE, where we will examine velocity dispersion fields and stellar population metrics, advancing our understanding of the IMF and mass distribution in massive galaxies.

Speaker: Leon Roman Ecker (LMU Munich (USM)/ Max-Planck Institute for extraterrestrial physics (MPE))

2:30 PM Strong-lensing and kinematic analysis of CASSOWARY 31: can strong lensing constrain the masses of multi-plane lenses?

We present a mass measurement for the secondary lens along the line of sight (LoS) in the multi-plane strong lens modeling of the group-scale lens CASSOWARY 31 (CSWA 31). The secondary lens at redshift z = 1.49 is a spiral galaxy well aligned along the LoS with the main lens at z = 0.683. Using the MUSE integral-field spectroscopy of this spiral galaxy, we measure its rotation velocities and determine the mass from the gas kinematics. We compare the mass estimation of the secondary lens from the lensing models to the mass measurement from kinematics, finding that the predictions from strong lensing tend to be higher. By introducing an additional lens plane at z = 1.36 for an overdensity known to be present, we find a mass of ≃ 1010 M⊙ enclosed within 3.3 kpc from the centroid of the spiral galaxy, approaching the estimate from kinematics. This shows that secondary-lens mass measurements from multiple-plane modeling are affected by systematic uncertainties from the degeneracies between lens planes and the complex LoS structure. Conducting a detailed analysis of the LoS structures is therefore essential to improve the mass measurement of the secondary lens.

Speaker: Han Wang (MPA/TUM)

2:50 PM Modelling of the Quadruply Lensed Quasar WG J0214-2105 for Time-delay Cosmography

Strong gravitational lensing is an astrophysical effect that happens when the gravitational field of a massive object bends light from a background source, creating multiple, often distorted, images of the same distant object. Accurate modelling of the mass distribution of strong gravitational lenses is crucial in order to use them as astrophysical and cosmological probes, such as for time-delay cosmography. We present the status of the robust modelling of WGJ0214-2105, a quadruply lensed quasar in the southern celestial hemisphere. Using the Gravitational Lens Efficient Explorer (GLEE) software and Hubble WFC3 imaging data in three photometric bands (F160W, F475X, and F814W). We construct a composite model of the system for the dark and baryonic mass distribution of the foreground lens galaxy and the light distributions of the galaxies, while also considering effects from nearby perturbers and the large-scale structure of the universe. This model will be instrumental for future time-delay cosmography, offering a robust mass distribution that enhances the system’s utility as a cosmological probe

Speaker: Allan Schweinfurth (TUM / MPA)

3:10 PM Subhaloes vs. line-of-sight haloes

Strong gravitational lensing has proven to be a promising tool for detecting low-mass haloes through observing perturbations in the lensing potential, therefore allowing the testing of different dark matter (DM) models by comparing the theoretical predictions for the abundance of such haloes in different DM models with observations. However, to robustly constrain the mass function of low-mass haloes, it is important to distinguish between various sources of perturbations to the smooth potential. In this talk, I present a systematic study of the degeneracy in the lensing effect between subhaloes and haloes that lie along the line of sight and the robustness of distinguishing between them, given the observational limits and modelling challenges. Finally, I will discuss the results of modelling two lens systems with detected perturbers using models that include field haloes.

Speaker: Maryam Tajalli (Max Planck Institute for Astrophysics)

3:30 PM Sequential simulation-based inference for strong gravitational lensing images

Galaxy-galaxy strong gravitational lenses provide a unique laboratory for probing small-scale structures and testing the $\Lambda$CDM paradigm. However, performing precise statistical analysis of such observations is extremely challenging since it requires disentangling the source galaxy’s light from the lens’ mass distribution and marginalizing over different dark matter substructure configurations. Research in this field can be broadly separated into works that aim to directly detect individual perturbers and works that aim to statistically constrain the matter distribution by looking at collective perturbations caused by an unresolved population of perturbers. In this talk, I will present recent advances in both of these approaches using a multi-stage method that combines parametric lensing models and an implicit-likelihood neural simulation-based inference technique. I will explain how this technique enables measuring both the properties of individual subhalos and directly the parameters of the subhalo mass function, overcoming some computational limitations of likelihood-based analyses.

Speaker: Noemi Anau Montel (MPA)

3:50 PM → 4:10 PM Coffee break

4:10 PM → 6:10 PM Afternoon session 2

SBI, lens finding, lensed SN and TDE

Convener: Aymeric Galan (TUM/MPA)

4:10 PM Weak Lensing Cosmology using Simulation Based Inference

We present a simulation-based inference analysis framework for a higher-order weak lensing observable called the integrated 3-point correlation function. For this we have created a forward model based on N-body simulations. This forward model can create realistic shear maps including survey masks, realistic shape noise, and relevant systematic effects. Furthermore, I present a Python package for the efficient evaluation of the integrated 3-point correlation function.
Combining this I have created a set of measurements of correlation functions that have been validated against theoretical predictions.
Based on these, a simulation-based inference pipeline has been established. For this, generative models are trained to learn the likelihood of the chosen summary statistics given the set of parameters. This ensemble is shown to be able to recover the correct cosmological parameters at the fiducial cosmology, as well as to pass coverage tests.
Applying this to the Dark Energy Survey year 3 galaxy shape catalogue, in our preliminary analysis we find $S_8 = 0.76 \pm 0.03$, agreeing with other late-time cosmological probes.

Speaker: David Gebauer (LMU Munich)

4:30 PM A Transformer-based Foundation Model for Strong Lensing

We present a foundation model for strong lensing based on diffusion transformers, which can solve a wide range of different simulation-based inference problems via representation learning. Image-like data such as observations are split into smaller patches, which together with their positions are embedded as tokens. Similarly, parametric models are also encoded via tokens. A strong lens system is therefore fully represented by a sequence of tokens. Each token in this sequence can be masked and the diffusion transformer is trained to infer masked tokens. Different types of inference problems such as source reconstruction, forward modeling or lens parameter estimation correspond to specific token masks. The representation as a sequence of tokens offers great flexibility. Our diffusion transformer can be trained on observations coming from different measurement instruments with different resolutions, pixel sizes or point spread functions. At the same time, both pixelated and parametric models can be used interchangeably within the same framework.

Speaker: Benjamin Holzschuh

4:50 PM Search for strong gravitational lenses combining ground-based and space-based imaging

In the past, researchers have relied on single-resolution images from individual telescopes to detect gravitational lenses. We propose a search for galaxy-galaxy lenses that, for the first time, combines high-resolution single-band images (in our case Hubble Space Telescope, HST) with low-resolution multi-band images (in our case Legacy survey, LS) using machine learning. To compensate for the scarcity of lensed galaxy images for network training, we generated simulated lenses by superimposing arc features onto HST images, saved the lens parameters, and replicated the lens system in the LS images. We test four architectures based on ResNet-18: (1) using single-band HST images, (2) three-band LS images, (3) stacking these images after interpolation the LS images to HST resolution for simultaneous process, and (4) merging before the fully connected layer, a ResNet branch of HST with a ResNet branch of LS. Our results demonstrate that
models integrating images from both the HST and LS significantly enhance the detection of galaxy-galaxy lenses compared to models relying on data from a single instrument, and could be use in the future for LSST and Euclid.

Speaker: Alejandra Melo Melo (Max Planck Institute for Astrophysics / Technical University of Munich)

5:10 PM Lens confirmation and machine-learning-assisted lens modelling as milestones for an effective lensed-SN follow-up

Strongly lensed supernovae (SNe) hold great promise for measuring the expansion rate of the Universe and studying SN progenitors and environments. However, finding lensed SNe resembles searching for the needle in a haystack, and following them up is very demanding and costly due to their faintness. For the best possible use of available resources, it is thus essential that we can securely identify lensed systems to beat down the large number of false alerts expected in the era of LSST, and predict the number and time delays of trailing images promptly to optimise the follow-up. In this talk, I will present some of the steps taken by the HOLISMOKES team in this direction, including a fast machine-learning-assisted extraction of lens parameters and a comparison of spectroscopic and lens-modelling-based techniques for confirming candidate systems as genuine strong lenses.

Speaker: Stefan Taubenberger (TUM / MPA)

HOLISMOKES_modelling_confirmation_no_animation.pdf

5:30 PM Spectroscopic and photometric time-delay inference of strongly lensed type II supernovae

With the upcoming start of Rubin Observatory Legacy Survey of Space and Time
(LSST), we expect to detect hundreds to thousands of strongly lensed supernovae,
which can be used for time-delay cosmography.
As part of the HOLISMOKES program, we focus on developing methods to measure time delays of strongly lensed type II supernovae (SNe II) to determine the Hubble Constant.
In my talk, I will present two methods of retrieving time delays from lensed SN II.
The first one uses spectra with which we can measure the time delay with an uncertainty of <1 day using multiple spectral lines. For this, we also investigated microlensing effects, which can be neglected in the early phases of the type IIP spectra as they are mostly achromatic at that phase.
I will also present our most recent work, using photometric data to determine the
time delays. The SN IIP color curves show a kink structure around 35 days
after the explosion, which is used as an anchor to retrieve a change in color and the temporal shift of color curves between two lensing images. We can retrieve time delays and differential extinction coefficients using this feature.

Speaker: Jana Grupa (Max Planck Institute for Astrophysics)

5:50 PM Gravitational Lensing of Tidal Disruption Events

In the coming years, surveys such as Rubin Observatory Legacy Survey of Space and Time (LSST) are expected to increase the number of observed tidal disruption events (TDEs) substantially. This analysis is a followup to Szekerczes, et al. (2024). We similarly employ Monte Carlo integration to calculate the TDE rate as a function of limiting magnitude. We investigate multiple black hole mass functions (BHMFs) and TDE luminosity models. The predictions of the unlensed TDE rate range from $1.6$ to $5 \; 440 \; \mathrm{yr}^{-1} \; 20 \, 000 \; \mathrm{deg}^{-2}$ for the Zwicky Transient Facility, only matching the observed rate at the low end. An overestimation is expected, as we do not consider observational effects such as dust extinction or host galaxy light. However, our analysis still reveals that the redshift evolution of the BHMF is not a significant factor, even for LSST. The predicted lensed rates display a similarly large scatter, ranging from $0.008$ to $15 \; \mathrm{yr}^{-1} \; 20 \, 000 \; \mathrm{deg}^{-2}$ for LSST. This scatter does reduce, when we consider the fraction of lensed TDEs, allowing us to estimate that approximately one in ten thousand TDEs will be lensed.

Speaker: Elias Mamuzic (Technical University of Munich / Max Planck Institute for Astrophysics)

Lensed TDEs_ Lensing Day.pptx