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 recently posted our next paper on arXiv: HOLISMOKES XI. It follows up on Paper VI and presents new neural networks to identify strongly lensed galaxies. While we applied the residual neural network to new real HSC images and presented the newly identified lenses in Paper VI, Paper XI focuses now on detailed performance tests of network architecture, hyper parameter, and training data variations. It is trained on mock images as previously, and tested on mock images as well as confirmed real lenses.
  • We publicly released now the neural network to model strongly lensed galaxies observed with the Hyper-Suprime Cam. For details see Paper IX.

News & Press

Science Magazine reports on the Target and Observation Manager TOM

We contribute to the TOM for gravitational lensing events. Our TOM is planned to take events classified as supernovae by event brokers, evaluate whether this new supernova is lensed, and, if it is lensed, schedule follow-up observations.

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


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.

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 a catalogue of HSC lens candidates identified with a residual neural network (HOLISMOKES VI and HOLISMOKES VIII). We further release 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


D. Sluse (STAR Institute)


Y. Shu (PMO)


S. Blondin (LAM)


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


S. Schuldt (UniMi)


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


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

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