Martin Kilbinger

Me in front of the Li river, near Yangshuo in GuangXi province of China.

I am a faculty researcher (ingénieur-chercheur) at the SAp which is part of IRFU at CEA Saclay, near Paris/France. At SAp I am a member of the CosmoStat laboratory, and the AIM mixed research unit together with the University Paris-Didérot (Paris 7). I am also associated to the IAP. Our institute is part of the new Université Paris-Saclay.

Download my CV.
My publications: ( ads | arXiv )
My Erdős-Number is at most 4 (via Starck - Donoho - Chui - Erdős).


Contact Information

CEA Saclay/Irfu/SAp, AIM
Orme des Merisiers
Bât 709
F-91191 Gif-sur-Yvette
Phone: +33 1 69 08 17 53
Fax:      +33 1 69 08 65 77
Office: 280
E-mail: martin.kilbinger (at)

github orcid_logo

Research Interests

My main research field is cosmology, in particular weak gravitational lensing. Using observational data from large galaxy surveys, I constrain cosmological models and infer information about dark matter and dark energy. To learn more about the topic, have a look at my (technical) review here, or check the video abstract on the companion web page.

Weak cosmological lensing

I am interested in using cosmic shear, the distortion of galaxies by the large-scale structure in the Universe, to measure cosmological parameters. From CFHTLenS, we have obtained constraints on dark matter, dark energy and modified gravity parameters, using second- and higher-order statistics, tomography, 2D and 3D weak lensing techniques. See those papers for more details. The ESA space mission Euclid is expected to improve those constraints by orders of magnitude.

One particularly interesting weak-lensing observable are peak counts. This higher-order statistic is sensitive to the non-Gaussian aspects of the large-scale structure. Together with Chieh-An Lin, we have developed a new model prediction approach for peak counts. We have compared this model to N-body simulations, explored its stochastic nature for strategies to constrain cosmological parameters, and looked at different filtering techniques.

Dark matter, galaxies, and galaxy clusters

I am interested in the relation between galaxies and dark matter, and how galaxies form and evolve in dark-matter structures. This relation can be studied by combining weak gravitational lensing and galaxy clustering statistics. This can help us to understand when and in which galaxies residing in dark-matter halos stars are formed efficiently.

We used weak lensing to measure masses of galaxy clusters in XXL and SDSS, being an important step for cosmological analyses using the number of clusters in mass and redshift.

Sampling the likelihood

I have been developing and implementing the sampling method Population MonteCarlo (PMC) which is an efficient and massively parallelizable method to sample from an arbitrary posterior distribution. PMC readily provides an estimate of the Bayesian evidence. Recently, I have looked into the likelihood-free technique Approximate Bayesian Computation (ABC) and other likelihood estimation methods, together with my PhD student Chieh-An Lin.


PhD for 2017, 50% co-funding from CNES (French space agency) already obtained: The Euclid space mission: Redshift measurements uncertainties and their effect on cosmological constraints using weak gravitational lensing and galaxy clustering, co-supervised by Yannick Mellier (IAP). Deadline 31/03/2017.

See also the CosmoStat job page.

Collaborations and projects

 cfhtls-logo CFHTLS (Canada-France Hawaii Telescope Legacy Survey) is a deep and wide imaging survey of 170 square degrees in five optical bands, observed between 2003 and 2009 with the wide-field camera MegaCam on the 3.6m telescope CFHT.
 cfhtls_logo CFHTLenS (Canada-France Hawaii Telescope Lensing Survey) is the weak-lensing collaboration using data from the CFHT legacy survey. 
 xxl-vertical XXL (The ultimate XMM extragalactic survey) is an extragalactic survey of 2 x 25 square degrees, with very deep X-ray coverage from XXM. Deep multi-wavelength data is available over the XXL fields (e.g. CFHTLS, BCS), and follow-up observations are underway and planned.
 Euclid Consortium Logo Euclid is a large upcoming ESA space mission with the goal to map galaxies and dark matter in the Universe, and to constrain the nature of dark energy. I am deputy lead of the weak-lensing science working group (WLSWG), and further work in shear and science-level (Level 3) data processing and validation. 
 desi_logo  The Dark-Energy Spectroscopic Instrument is a next-generation galaxy clustering (BAO, RSD) observatory, with the aim to measure dark energy properties.
CFIS Logo The Canada-France Imaging Survey, a 5,000 deg2 survey in the Northern hemisphere. Observations are starting in early 2017. See here for more details.


  • Camelus, a model for weak-lensing peak counts. The codes comes with statistical analysis tools for cosmological parameter constraints. In particular, it exploits the stochastical nature of the model predictions using Approximate Bayesian Computation (ABC). [Main author is Chieh-An Lin.]
  • CosmoPMC, Cosmology sampling with Population Monte Carlo, an adaptive importance-sampling method to explore high-dimensional posteriors to estimate parameter constraints and the Bayesian evidence. CosmoPMC comes with various cosmology modules such as weak lensing, galaxy clustering, SNIa, BAO distance priors and CMB (via CAMB/WMAP software). The cosmology module is based on nicaea.
    The most recent version is 1.2 (December 2012). The user's guide is on the archive (v3).
  • athena, a tree code to calculate second-order correlation functions, including shear-shear ("cosmic shear"), shear-position ("galaxy-galaxy lensing") and position-position (spatial angular correlation).
    The most recent version is 1.7 (March 2014).
  • nicaea: NumerIcal Cosmology And lEnsing cAlculations. It involves basic cosmology functions (distances, growth factor), dark-matter power spectrum and lensing observables (power spectrum, second-order shear statistics). The non-linear power spectrum are the Peacock&Dodds (1996) and Smith, Peacock et al (2003) fitting formulae. The code is easy to use and fast enough to be used for Monte-Carlo sampling.
    The most recent version is 2.5 (October 2014).

Lecture Notes and Talks

External Links

Class, the Cosmic Linear Anisotropy Solving System, by Julien Lesgourgues et al.

What might happen at The end of the Universe?

Non-Science Stuff

Check out my piano music: videos and music scores available