HOLISMOKES

Highly Optimised Lensing Investigations of Supernovae, Microlensing Objects, and Kinematics of Ellipticals and Spirals

We measure the current expansion rate of the Universe using time-delay cosmography with strongly lensed supernovae. We also employ the lensing effect to study the very early phases in the evolution of supernovae.

Supernova Refsdal, the first supernova that is strongly lensed by a foreground galaxy cluster into multiple resolvable images (indicated by the arrows in the inset). This event was discovered serendipitously by Kelly et al. (2015). Image credit: NASA, ESA, and S. Rodney (JHU) and the FrontierSN team; T. Treu (UCLA), P. Kelly (UC Berkeley), and the GLASS team; J. Lotz (STScI) and the Frontier Fields team; M. Postman (STScI) and the CLASH team; and Z. Levay.



HOLISMOKES is funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (LENSNOVA: grant agreement No 771776).

Latest Results

  • We are happy to announce that our HSC lens candidates from the HOLISMOKES collaboration, presented in Paper VI , Paper VIII , and Paper XIII, are now also available through the SuGOHI data base.
  • We recently posted another paper on arXiv:HOLISMOKES XIV. In this work, We present a method, to retrieve time delays and the amount of differential dust extinction between multiple images of lensed type IIP supernovae through their color curves, which display a kink in the time evolution. With multiple realistic mock color curves based on an observed unlensed supernova from the Carnegie Supernova Project, we demonstrate that we can retrieve the time delay with uncertainties of ±1.0 days for light curves with 2-day cadence and 35% missing data due to weather losses. With the upcoming Rubin Observatory Legacy Survey of Space and Time, hundreds of strongly lensed supernovae will be detected and our new method for lensed SN IIP is readily applicable to provide delays.
  • We recently posted our next paper on arXiv: HOLISMOKES XIII. It follows up on Paper VI and presents several new strong lensing candidates using HSC imaging data. In addition, we study the lens environment from our visually identified lens candidates from Paper VI and this work, resulting in a total of 546 identified grade A or B lens candidates, and present the first lensing clusters identified by CNNs. For the environment, we consider three different techniques: (1) matching to cluster catalogs, (2) visual inspection, (3) photometric redshift distribution of objects within 200". The novel criteria defined for the lens cluster selection (see figure) are easy to compute and usable for large samples from e.g., LSST or Euclid. For more details see Paper XIII .

News & Press


Sherry Suyu receives the 2021 Berkeley Prize


“Thank you very much for this fantastic news! It is a great honor, not just for me, but for the entire H0LiCOW team! Our results took many years of hard work, and we are excited to establish a completely independent and competitive probe of cosmology. None of this would be possible without my wonderful H0LiCOW collaborators, and I am truly grateful to them.” -- Sherry Suyu, PI of H0LiCOW and HOLISMOKES

Berkeley Prize for Sherry Suyu announced by the ERC


Publications

HOLISMOKES. I. Highly Optimised Lensing Investigations of Supernovae, Microlensing Objects, and Kinematics of Ellipticals and Spirals

We present the HOLISMOKES programme on strong gravitational lensing of supernovae (SNe) as a probe of SN physics and cosmology. We investigate the effects of microlensing on early-phase SN Ia spectra using four different SN explosion models. We find that distortions of SN Ia spectra due to microlensing are typically negligible within ten rest-frame days after a SN explosion (<1% distortion within the 1σ spread and ≲10% distortion within the 2σ spread). This shows the great prospects of using lensed SNe Ia to obtain intrinsic early-phase SN spectra for deciphering SN Ia progenitors. ...

HOLISMOKES. II. Identifying galaxy-scale strong gravitational lenses in Pan-STARRS using convolutional neural networks

We present a systematic search for wide-separation (with Einstein radius θE ≳ 1.5″), galaxy-scale strong lenses in the 30 000 deg2 of the Pan-STARRS 3π survey on the Northern sky. With long time delays of a few days to weeks, these types of systems are particularly well-suited for catching strongly lensed supernovae with spatially-resolved multiple images and offer new insights on early-phase supernova spectroscopy and cosmography. We produced a set of realistic simulations by painting lensed COSMOS sources on Pan-STARRS image cutouts of lens luminous red galaxies (LRGs) with redshift and velocity dispersion known from the sloan digital sky survey (SDSS). ...

HOLISMOKES. III. Achromatic phase of strongly lensed Type Ia supernovae

To use strongly lensed Type Ia supernovae (LSNe Ia) for cosmology, a time-delay measurement between the multiple supernova (SN) images is necessary. The sharp rise and decline of SN Ia light curves make them promising for measuring time delays, but microlensing can distort these light curves and therefore add large uncertainties to the measurements. An alternative approach is to use color curves where uncertainties due to microlensing are significantly reduced for a certain period of time known as the achromatic phase. In this work, we investigate in detail the achromatic phase, testing four different SN Ia models with various microlensing configurations. ...

HOLISMOKES. IV. Efficient mass modeling of strong lenses through deep learning

Modelling the mass distributions of strong gravitational lenses is often necessary to use them as astrophysical and cosmological probes. With the high number of lens systems (>105) expected from upcoming surveys, it is timely to explore efficient modeling approaches beyond traditional MCMC techniques that are time consuming. We train a CNN on images of galaxy-scale lenses to predict the parameters of the SIE mass model (x, y, ex, ey, and θE). To train the network, we simulate images based on real observations from the HSC Survey for the lens galaxies and from the HUDF as lensed galaxies. ...

HOLISMOKES. V. Microlensing of type II supernovae and time-delay inference through spectroscopic phase retrieval

We investigate strongly gravitationally lensed type II supernovae (LSNe II) for time-delay cosmography incorporating microlensing effects, which expands on previous microlensing studies of type Ia supernovae (SNe Ia). We use the radiative-transfer code TARDIS to recreate five spectra of the prototypical SN 1999em at different times within the plateau phase of the light curve. The microlensing-induced deformations of the spectra and light curves are calculated by placing the SN into magnification maps generated with the code gerlumph. We study the impact of microlensing on the color curves and find that there is no strong influence on them ...

HOLISMOKES. VI. New galaxy-scale strong lens candidates from the HSC-SSP imaging survey

We have carried out a systematic search for galaxy-scale strong lenses in multiband imaging from the Hyper Suprime-Cam (HSC) p survey. Our automated pipeline, based on realistic strong-lens simulations, deep neural network classification, and visual inspection, - is aimed at efficiently selecting systems with wide image separations (Einstein radii θE ∼ 1.0–3.000 ), intermediate redshift lenses o(z ∼ 0.4–0.7), and bright arcs for galaxy evolution and cosmology. We classified gri images of all 62.5 million galaxies in HSC Wide rt with ...

HOLISMOKES. VII. Time-delay measurement of strongly lensed SNe Ia using machine learning

The Hubble constant (H0) is one of the fundamental parameters in cosmology, but there is a heated debate on the >4σ tension between the local Cepheid distance ladder and the early Universe measurements. Strongly lensed Type Ia supernovae (LSNe Ia) are an independent and direct way to measure H0, where a time-delay measurement between the multiple supernova (SN) images is required. In this work, we present two machine learning approaches to measure time delays in LSNe Ia, namely, a fully connected neural network (FCNN) and a Random Forest (RF). For the training of the FCNN and the RF, we simulate mock LSNe Ia from theoretical SN Ia models including observational noise and microlensing. We test the transfer learning capability of both machine learning models, by using a final test set based on empirical LSN Ia light curves ...

HOLISMOKES. VIII. High-redshift, strong-lens search in the Hyper Suprime-Cam Subaru Strategic Program

We carry out a search for strong-lens systems containing high-redshift lens galaxies with the goal of extending strong-lensing-assisted galaxy evolutionary studies to earlier cosmic time. Two strong-lens classifiers are constructed from a deep residual network. Applying the two classifiers to the second public data release of the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP), we identify 735 grade-A or B strong-lens candidates in total, of which 277 are discovered for the first time. This is the single largest set of galaxy-scale strong-lens candidates discovered with HSC data to date, and nearly half of it contains lens galaxies with photometric redshifts above 0.6.

HOLISMOKES. IX. Neural network inference of strong-lens parameters and uncertainties from ground-based images

While we presented in HOLISMOKES IV a convolutional neural network to predict point estimates of the SIE mass parameters, we present here a residual neural network that predicts the SIE mass parameters together with the external shear component. It further predicts the 1σ uncertainty of each parameter. The network is trained on realistic HSC lens mock images generated using observed images of galaxies, together with measured redshift and velocity dispersion of the lens galaxy.

HOLISMOKES. X. Comparison between neural network and semi-automated traditional modeling of strong lenses

To test the performance of our residual neural network presented in HOLISMOKES IX on real data, we apply the network to 31 real HSC lenses. We further model them with our newly developed code glee_auto.py, a fully automated modeling code for ground-based galaxy-scale lenses relying on Monte-Carlo Markov-Chain optimization. We further present glee_tools.py, a flexiple code to automate individual steps for manual refinment of mass models. We find a good agreement espectially for the Einstein radius, while stronger differences remain on the external shear as expected from the simulated test set.

HOLISMOKES. XI. Evaluation of supervised neural networks for strong-lens searches in ground-based imaging surveys

While supervised neural networks have become state of the art for identifying the rare strong gravitational lenses from large imaging data sets, their selection remains significantly affected by the large number and diversity of nonlens contaminants. This work evaluates and compares systematically the performance of neural networks in order to move towards a rapid selection of galaxy-scale strong lenses with minimal human input in the era of deep, wide-scale surveys. We used multiband images from PDR2 of the Hyper-Suprime Cam (HSC) Wide survey to build test sets mimicking an actual classification experiment, with 189 strong lenses previously found over the HSC footprint and 70,910 nonlens galaxies in COSMOS covering representative lens-like morphologies. Multiple networks were trained on different sets of realistic strong-lens simulations and nonlens galaxies, with various architectures and data pre-processing, mainly using the deepest gri bands. Most networks reached excellent area under the Receiver Operating Characteristic (ROC) curves on the test set of 71099 objects.

HOLISMOKES. XII. Time-delay Measurements of Strongly Lensed Type Ia Supernovae using a Long Short-Term Memory Network

Strongly lensed Type Ia supernovae (LSNe Ia) are a promising probe to measure the Hubble constant (H0) directly. To use LSNe Ia for cosmography, a time-delay measurement between the multiple images, a lens-mass model, and a mass reconstruction along the line of sight are required. In this work, we present the machine learning network LSTM-FCNN which is a combination of a Long Short-Term Memory Network (LSTM) and a fully-connected neural network (FCNN). The LSTM-FCNN is designed to measure time delays on a sample of LSNe Ia spanning a broad range of properties, which we expect to find with the upcoming Rubin Observatory Legacy Survey of Space and Time (LSST) and for which follow-up observations are planned.

HOLISMOKES. XIII. Strong-lens candidates at all mass scales and their environments from the Hyper-Suprime Cam and deep learning

We have performed a systematic search for galaxy-scale strong gravitational lenses using Hyper Suprime-Cam (HSC) imaging data, focusing on lenses in overdense environments. To identify these lens candidates, we exploit our residual neural network from HOLISMOKES VI, which is trained on realistic gri mock-images as positive examples, and real HSC images as negative examples. Compared to our previous work, where we have successfully applied the classifier to around 62.5 million galaxies having i-Kron radius ≥ 0.8", we now lower the i-Kron radius limit to ≥ 0.5". This results in an increase by around 73 million sources to more than 135 million images. During our visual multi-stage grading of the network candidates, we now also inspect simultaneously larger stamps (80" × 80") to identify large, extended arcs cropped in the 10" × 10" cutouts and classify additionally their overall environment.

HOLISMOKES - XIV. Time-delay and differential dust extinction determination with lensed type II supernova color curves

We have presented a method, to retrieve time delays and the amount of differential dust extinction between multiple images of lensed type IIP supernovae through their color curves, which display a kink in the time evolution. With multiple realistic mock color curves based on an observed unlensed supernova from the Carnegie Supernova Project, we have demonstrated that we can retrieve the time delay with uncertainties of ±1.0 days for light curves with 2-day cadence and 35% missing data due to weather losses.

Data Products

Below is a link to data products from the HOLISMOKES project. This includes a catalogue of candidate strong-lensing systems identified in PS1 data with the help of a convolutional neural network (HOLISMOKES II) and catalogues of HSC lens candidates identified with a residual neural network (HOLISMOKES VI, HOLISMOKES VIII and HOLISMOKES XIII). We further publicly released the produced microlensed spectra and light curves of HOLISMOKES VII and the residual neural network to model strongly lensed galaxies from HOLISMOKES IX.

Who we are

HOLISMOKES is an international collaboration with people from around the whole world

Belgium

D. Sluse (STAR Institute)

China

Y. Shu (PMO)

France

S. Blondin (LAM)
R. Cañameras (LAM)

Germany

S.H. Suyu (PI, MPA / TUM / ASIAA)
J. Bayer (MPA / TUM)
S. Huber (MPA / TUM)
M. Kromer (HITS)
A. Melo Melo (MPA)
U. Nöbauer (MPA)
S. Taubenberger (MPA)
C. Vogl (MPA)

Italy

S. Schuldt (UniMi)

Switzerland

F. Courbin (EPFL)

United Kingdom

S. Sim (QUB)

United States of America

J.H.H. Chan (CUNY / FutureLens)

... & Friends!

D. Ghoshdastidar (TUM)
L. Leal-Taixé (TUM)
T. Meinhardt (TUM)
A. Yıldırım

Contact

Is there anything you want to ask us? Do not hesitate to contact us!
Sherry is the PI of HOLISMOKES, and the lead author of Paper I.

  • Address

    Sherry Suyu
    Technical University of Munich
    TUM School of Natural Sciences
    Department of Physics
    James-Franck-Str. 1
    85748 Garching
    Germany

    Max Planck Institute for Astrophysics
    Karl-Schwarzschild-Str. 1
    D-85741 Garching
    Germany