The study of Higgs boson pair production is a major focus of Higgs physics at the LHC, particularly in view of the High-Luminosity LHC programme. It provides direct access to the Higgs boson self-coupling, as well as to the VVHH coupling, which is an important parameter for improving our understanding of electroweak symmetry breaking. This coupling can be probed with high sensitivity in di-Higgs events produced through vector-boson fusion (VBF), in particular in the high-mass regions of the HH system.

The exploitation of the full Run 3 dataset, combined with recent developments in the different analyses, will significantly improve the sensitivity to di-Higgs production. The combination of the different final states studied by the ATLAS Collaboration, in particular HH->bbtautau, HH->bbyy and HH->bbbb, could provide the first experimental indications of this process, with a sensitivity potentially reaching the three-standard-deviation level. It will also improve constraints on the properties of the Higgs sector, and in particular on the couplings involved in Higgs boson pair production.

The PhD thesis will make a major contribution to theHH->bbtautau analysis, with a particular focus on boosted topologies. In these configurations, the Higgs boson decay products are highly collimated, requiring the development and optimisation of dedicated object reconstruction and identification techniques. These events are particularly interesting for exploring the high-m(HH) region and probing the VVHH coupling, generally parameterised by the k(2V) coefficient. Particular attention will therefore be paid to VBF topologies and to the kinematic regions providing the highest sensitivity to this coupling. The results of this analysis will contribute to the combination of the different channels studied by ATLAS, exploiting their complementarity to improve the overall sensitivity to HH production and k(2V).

The ATLAS detector is ideally suited to such studies, with a design optimised to reconstruct and identify most of the Standard Model particle decay products produced in rare processes involving Higgs bosons. The thesis will also include a component dedicated to the optimisation of algorithms used to identify jets originating from the hadronisation of b quarks, in preparation for the High-Luminosity LHC phase. These algorithms play a key role in many final states involving b quarks, including Higgs boson and top-quark decays. This work will contribute to preparing the scientific exploitation of the new ATLAS Inner Tracker (ITk), currently under construction, with a significant contribution from CPPM, and scheduled to be commissioned for the start of Run 4 around summer 2030. It will therefore establish a direct link between the exploitation of the Run 3 dataset and the preparation of future Higgs physics analyses at the HL-LHC.

The thesis will be part of a collaborative research effort involving several French laboratories participating in the ATLAS Collaboration at CERN. It will benefit from the strong expertise of the CPPM ATLAS group in b-quark identification, object reconstruction in boosted topologies, and Higgs boson pair production analyses. The PhD student will therefore contribute both to a leading LHC physics analysis and to the preparation of the tools required for its future exploitation, in particular in the context of the commissioning of the new ITk detector.

Applications should include a CV, a motivation letter, academic transcripts from undergraduate to Master's level, and the contact details of two referees who can provide letters of recommendation.

[1] Transforming jet flavour tagging at ATLAS

https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/FTAG-2023-05/

[2] Expected Transformer-Network-based jet flavour tagging performance with the ATLAS Inner Tracker Detector at the High-Luminosity LHC

https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PUBNOTES/ATL-PHYS-PUB-2026-002/

[3] Improved analysis of non-resonant Higgs boson pair production in the final state with fb of data collected at = 13 TeV and 13.6 TeV with the ATLAS detector

https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/HIGP-2024-37/

[4] Combination of ATLAS and CMS searches for Higgs boson pair production at $\sqrt{s}=13$ TeV

https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/HIGP-2025-18/

The discovery of the Higgs boson by the ATLAS and CMS collaborations in 2012 completed the particle content of the Standard Model (SM). A fundamental question nevertheless remains open: the shape of the Higgs potential that drives electroweak symmetry breaking. In the SM, this shape is fixed by the Higgs boson self-coupling, which has never been measured. Its value determines the nature of the electroweak phase transition in the early Universe and the stability of the vacuum, and any deviation from the SM prediction would be a clear sign of new physics. At the LHC, the only direct way to measure the self-coupling is through the production of Higgs boson pairs (HH). This process is about a thousand times rarer than single Higgs production and has not yet been observed, which makes it one of the flagship goals of the LHC Run 3 and High-Luminosity programmes. The ATLAS group at CPPM has played a leading role in Higgs boson studies since the discovery and is now a major contributor to the ATLAS HH search.

This thesis will concentrate on the study of the production of two Higgs bosons decaying to a pair of photons and a pair or b-quarks (HH->bbyy). The detection of such process is a strong proof of the Higgs self coupling and the electroweak symmetry breaking as described by the standard model. The run 3 of the LHC, that ended in 2026, provides enough data to improve the discovery potential of such process. Several analysis techniques are also in development to further improve the signal extraction from the background. However, a combination of several channels of the decay of the produced Higgs bosons will be necessary to reach the evidence of the existence of this process.

The successful candidate will work within this ATLAS collaboration and with a team of four researchers and one PHD student at CPPM. He/She will contribute to all aspects of the analysis including the signal selection and extraction (with machine learning techniques), background estimation with the corresponding uncertainties, and finally the statistical analysis leading to the final results. The student will also work on combination with other channels and the interpretation of the physics results comparing them to the standard model and other new physics models.

The efficient identification of photons in the ATLAS detector is one of the main ingredient for the analysis described above. The candidate is expected to work on improving the photon identification in ATLAS. This work involves understanding the shower shapes that a photon leave in the Liquid Argon calorimeter as well as developing modern methods (based on neural networks) to identify photons and separate them from the background. The improvement of photon identification will target the existing run 3 data as well as the preparation for the run 4 of the High Luminosity LHC that will start in 2030.

Prior knowledge of programming language especially C++/root or python is an advantage but is not mandatory.

This thesis is expected to start in October 2026 (if funding is obtained).

The determination of the CKM matrix element $|V_{ub}|$ is a key objective of flavour physics and provides an important test of the Standard Model description of quark-flavour transitions. In particular, exclusive charmless semileptonic decays such as $B\to\pi\ell\nu$ and $B\to\rho\ell\nu$ , combined with theoretical calculations of the corresponding hadronic form factors, provide a powerful approach to its determination.

The clean kinematic environment of B-factory experiments offers unique opportunities for these measurements. At Belle and Belle II, the reconstruction of the companion B meson provides strong constraints on the signal-side kinematics and enables the reconstruction of the missing neutrino. The performance of the B-tagging algorithm therefore plays a crucial role in the efficiency, background rejection, and kinematic resolution of semileptonic B-decay measurements.

This PhD project will develop and exploit advanced B-tagging techniques for a precision study of charmless semileptonic B decays using Belle and Belle II data. The analysis will focus on channels such as $B\to\pi\ell\nu$ and $B\to\rho\ell\nu$ , with the possibility of extending the programme to complementary exclusive modes. The project will investigate how improved tagging efficiency and kinematic reconstruction can enhance the sensitivity to these decays and reduce the dominant experimental uncertainties.

The measurement will include branching fractions and differential decay distributions, providing access to the $q^2$ dependence of the decay rates and allowing the experimental results to be combined with state-of-the-art theoretical form-factor calculations. The ultimate goal is a precise determination of $|V_{ub}|$ , together with a detailed assessment of the experimental and theoretical uncertainties.

The CPPM Belle II group is actively involved in the development of B-tagging techniques and in the study of semileptonic B-meson decays. The group has expertise in event reconstruction, flavour-physics analyses, and the exploitation of Belle II data, providing a strong environment for the development of this programme. The project will offer the opportunity to contribute to the Belle II physics programme while developing expertise in both advanced reconstruction techniques and precision flavour physics.

By combining advanced event reconstruction with a precision measurement of charmless semileptonic B decays, the project will contribute to a more precise determination of $|V_{ub}|$ , the CKM matrix, and tests of the Standard Model.

In the late 90s, measurements of the distance of Supernovae and the redshift of their host galaxies revealed that the expansion of the Universe was accelerating. More than 20 years after this discovery, the nature of the dark energy at the origin of this phenomenon remains unknown.

The $\Lambda$ CDM concordance model describes a homogeneous, isotropic Universe on large scales, subject to the laws of general relativity (GR). In this model, most of the Universe's energy content comes from cold dark matter and dark energy, introduced as a cosmological constant. The latter behaves like a perfect fluid with negative pressure p, equation of state p = - rho, where rho is the energy density.

Some alternative models (see [1] for a review) introduce scalar fields (quintessence) whose evolution is responsible for the accelerated expansion. These scalar fields can vary in time and space. They can therefore have a time-dependent equation of state and generate anisotropic expansion.

Other models propose to modify the law of gravitation on large scales, mimicking the role of dark energy.

Supernovae remain one of the most accurate probes of the Universe's expansion and homogeneity. In addition, part of the redshift of galaxies is due to a Doppler effect caused by their particular velocities. We can then use supernovae to reconstruct the velocity field on large scales, and measure the growth rate of cosmic structures. This will enable us to test the law of gravitation.

An anisotropy of expansion on large scales, a modification of GR, or an evolution of the equation of state for Dark Energy, would all be revolutionary observations that would challenge our current model.

Until now, supernova surveys have gathered data from multiple telescopes, complicating their statistical analysis. Surveys by the Zwicky Tansient Facility (ZTF: https://www.ztf.caltech.edu/) and the Vera Rubin/LSST Observatory (https://www.lsst.org/) will change all that. They cover the entire sky and accurately measure the distance to tens (hundreds) of thousands of nearby (distant) supernovae.

The CPPM has been working on ZTF data since 2021 and will publish a first cosmological analysis in 2027 with ~3000 SN1a. We have also been involved in the construction and implementation of LSST for years, and are now preparing for the data analysis for the first release which will happen during the PhD.

Within the group, we are working on the photometric calibration of the ZTF survey, essential for the measurement precision we need (see ubercalibration [2,3]). A first PhD student developed a pipeline to simulate ZTF and measure the growth rate of structures ([4], defended in 2023), a second student adapted this exercise to LSST ([5], defended in 2025) and a third one started in 2024 to lead the analysis of real data. In addition, multiple post-docs have joined the group to work on ZTF, and a Chair of Excellence (DARKUNI) is extending this work by combining these data with spectroscopic data from DESI.

The aim of the thesis is to develop and perfect this analysis pipeline for measuring the growth rate of structures. The totality of 30000 SN1a of ZTF will be available to do the final cosmological analysis of this survey. The thesis coincides also with the arrival of the first SN1a catalogs of LSST.

Other aspects may be added to the thesis, such as the study of the homogeneity of the expansion, the photometric calibration of the data, and so on.

This is an observational cosmology thesis, for a candidate interested in cosmology and data analysis.

[1] https://arxiv.org/abs/1601.06133

[2] https://arxiv.org/abs/astro-ph/0703454v2

[3] https://arxiv.org/abs/1201.2208v2

[4] https://arxiv.org/abs/2303.01198 https://snsim.readthedocs.io/

[5] https://arxiv.org/abs/2507.00157

Precision cosmology fundamentally requires millions of accurate galaxy redshifts, obtained either through slitless or large-scale multi-fiber spectroscopic surveys. While slitless spectroscopy has the advantage of delivering spectroscopic observations for all sources present in the field of view with sufficient signal strength, it is severely affected by contamination from overlapping spectral traces of multiple objects. Multi-fiber spectroscopy, by contrast, targets specific objects of interest, virtually eliminating contamination, but requires substantially longer times per pointing to accumulate sufficient spectra and reaches the statistical power needed for precision cosmology. Because statistical power scales with sample size, each generation of instruments has increased its fiber count and fiber density to observe more objects per exposure: SDSS (320 fibers), BOSS (1000), DESI (5000). Next-generation spectrographs (MUST, Spec-S5, WST) push this further still, to 2000 fibers/deg² or more, since denser multiplexing remains the only practical way to reach the required survey depth without proportionally increasing observing time.

This densification, however, runs into a optical limit. The wings of each fiber's PSF spill into neighboring apertures leading to fiber crosstalk: The spectrum measured in each fiber becomes a linear mixture of light from~ neighbors. Furthermore, as spectroscopic surveys probe fainter galaxies to extend to higher redshifts, the target galaxies often lie close to brighter foreground objects with much higher flux, leading to significant contamination from neighboring fibers.

Stage-IV cosmology requires redshifts unbiased while tests from DESI collaboration show that 15% crosstalk already biases redshifts over budget — with 2-5% catastrophic failure rates for contaminated objects. At fiber densities of stage-V spectroscopic survey, crosstalk must be corrected algorithmically, by solving an ill-posed inverse problem in which A is large, ill-conditioned, and variable in wavelength, focal-plane position, and time. The aim of this PhD is to evvaluate the impact of cross-fiber contamination on redshift measurement and cosmological inference, and to develop mitigation strategies for MUST.

RESEARCH PLAN

The thesis will characterize cross-fiber contamination and its impact on redshift and cosmological inference, then evaluate correction strategies, in four steps:

1. Build a predictive optical model of crosstalk from ray-tracing (Zemax) simulations, applicable to MUST ahead of first light.

2. Develop and compare extraction algorithms (multi-fiber optimal extraction, machine-learning extraction pipeline, Bayesian/MCMC joint inference, ...) that recover clean spectra from contaminated data, benchmarking their accuracy and computational cost.

3. Test and validate the correction algorithms on simulated data, and then on real data from the MUST instrument.

4. Propagate residual crosstalk through synthetic galaxy catalogs to quantify its impact on cosmological constraints.

EXPECTED IMPACT

This work will deliver the redshift-accuracy validation and correction pipeline required for MUST to meet its cosmological goals. Beyond MUST, the PhD candidate will acquire deep expertise in spectroscopic data processing and redshift validation that is directly applicable to any multi-fiber survey. As next-generation spectrographs push toward ever-higher fiber densities, these methods will become increasingly critical for maintaining data quality in increasingly multiplexed instruments.

HOST TEAM

The PhD will be supervised by W. Gillard, who leads the RENOIR group at CPPM (Marseille). The group comprises some 34 people covering the full chain from detector characterisation and instrument development to cosmological inference, with contributions to Euclid, DESI, Vera C. Rubin/LSST, ZTF and LISA. This thesis sits exactly at the interface between instrument and cosmology, precisely where the group's expertise lies. On the instrument side, RENOIR has been engaged since 2014 in the construction of the ten DESI spectrographs, with the LAM, the OHP and Bertin/Winlight, and now exploits the DR1 and DR2 data they produced; it also contributed to the construction of the Euclid NISP instrument, taking responsibility for its focal plane, the characterisation of its infrared detectors and its on-ground spectroscopic calibration. On the analysis side, W. Gillard is the Euclid NISP Instrument Scientist, engaged in the coordination of the Euclid Instrument Operation Team and responsible for the quality of the NISP data and calibration underpinning the Euclid releases. He is also PI of the ANR project DlSPERS, dedicated to machine-learning analysis of slitless spectroscopic surveys — the sister problem of the one addressed in this PhD: recovering unbiased redshifts from heavily contaminated spectra. The group is a member of the French consortium forming around MUST, and collaborates with Bertin/Winlight on the characterisation and validation of the first MUST spectrographs.

The Hubble constant $H_0$ tension—a $\sim 5\sigma$ discrepancy between early-universe (CMB) and late-universe (supernovae/Cepheids) measurements—represents one of the most significant anomalies in modern cosmology. Resolving this tension requires independent, orthogonal probes of $H_0$ unaffected by conventional systematic uncertainties. Strong gravitational lensing of time-domain transients provides precisely such a probe: time-delay cosmography measures $H_0$ through the time delays between multiple images of a lensed supernova, tidal disruption event, or $\gamma$ -ray burst, combined with precise lens mass models. This method is theoretically immune to cosmic acceleration assumptions and provides sensitivity to $H_0$ complementary to the cosmic distance ladder.

However, lensed transients are rare—only $\sim20$ systems with time-delay measurements currently exist—and their exploitation remains limited by (1) sparse sampling of light curves, which degrades time-delay precision; (2) degeneracies in lens mass modeling, which propagate into $H_0$ uncertainties; and (3) heterogeneous photometric calibration and spectroscopic characterization across follow-up datasets from diverse facilities, biasing cosmological inference.

The advent of Euclid and LSST transforms this landscape. Euclid will identify millions of strong lens candidates through high-resolution imaging; LSST will discover transients in near-real time. Their synergy—formalized in the July 2026 Euclid-LSST Memorandum of Understanding—enables discovery of 50–100 lensed transients per year, mainly SNIa, sufficient to achieve competitive $H_0$ constraints and break parameter degeneracies through population statistics. Furthermore, type Ia supernovae provide an absolute luminosity scale that helps alleviate the degeneracies in lens mass modelling degeneracy. A PhD candidate joining the Euclid Consortium's Lensed-Transient Working Group will develop the statistical and data-processing infrastructure to convert this discovery stream into precision cosmology.

RESEARCH PLAN

The thesis will develop methods for extracting cosmological constraints—particularly $H_0$ —from heterogeneous lensed-transient samples within the Euclid Consortium framework. The work is structured in four steps:

1 Lens mass modeling and degeneracy mitigation: Build a flexible, Bayesian hierarchical framework for lens mass reconstruction that incorporates high-resolution Euclid imaging and spectroscopic redshifts (from MUST or other surveys) to reduce mass-sheet and density-slope degeneracies. Quantify how Euclid's morphological precision and spectroscopic data constrain the local lensing geometry, and propagate these constraints into $H_0$ uncertainties. Benchmark against simulated Euclid-resolution lens systems.

2 Photometric time-delay measurement and calibration: Develop robust algorithms for extracting precise time delays from LSST's rapid multi-band photometry, accounting for intrinsic source variability, microlensing, and photometric calibration uncertainties across facilities. Implement adaptive follow-up strategies that maximize time-delay signal-to-noise within resource constraints. Establish standardized photometric protocols (coordinated through DESC/SLSC) to minimize systematic biases in multi-epoch datasets.

3 Cosmological inference from lensed-transient populations: Construct a population-level Bayesian inference framework that jointly constrains $H_0$ , the equation-of-state parameters (w?, w?), and spatial curvature from an ensemble of lensed transients with heterogeneous data quality and systematics. Account for selection effects (e.g., preferential discovery of brighter or more separated image pairs) and redshift-dependent biases. Quantify expected $H_0$ precision as a function of sample size and photometric accuracy, and compare forecasts with other late-universe probes.

4 Validation and integration into Euclid Lensed-Transient Working Group: Test methods on existing lensed systems (H1413+117, RXJ1131-1231, etc.), validate on synthetic light curves from lens+source ray-tracing simulations, and apply to real LSST discoveries within Euclid lens catalogs as they emerge. Coordinate $H_0$ measurements across the Euclid Consortium and contribute to joint Euclid-DESC-SLSC publications following the MoU framework.

EXPECTED IMPACT

This work will establish Euclid and LSST as leading contributors to the $H_0$ tension resolution, delivering precision measurements of the Hubble constant that are competitive with or exceed existing methods while remaining independent of cosmic acceleration assumptions. The PhD candidate will become a core member of the Euclid Consortium's Lensed-Transient Working Group, gaining leadership experience in multi-institutional cosmological surveys and contributing directly to Euclid's primary science goals.

Beyond $H_0$ , lensed-transient cosmography constrains dark-energy properties, tests modified gravity theories, and probes the inner mass distributions of elliptical galaxies—opening complementary windows onto fundamental physics. The expertise developed in Bayesian hierarchical inference, photometric systematics, and population-level statistical methods applies broadly to survey cosmology. As LSST and future surveys (Vera Rubin Legacy, next-generation time-domain facilities) generate explosive rates of discoverable transients and lenses, the infrastructure and techniques developed here will become benchmarks for automated cosmological inference across time-domain astrophysics.

The PhD candidate will publish discovery papers and $H_0$ -constraint papers within the Euclid Consortium framework, co-author DESC/SLSC joint publications on lensed transients, and contribute to open-source software for transient vetting and cosmological modeling—positioning them as a leading expert in time-domain gravitational-lensing cosmology.

HOST TEAM

The PhD will be supervised by W. Gillard and D. Fouchez, both members of the Euclid and Vera C. Rubin/LSST collaborations, within the RENOIR group at CPPM (Marseille). The group comprises some 34 people covering the full chain from detector characterisation to cosmological inference, with contributions to Euclid, LSST, DESI, ZTF and LISA. W. Gillard, who leads the group, is the Euclid NISP Instrument Scientist: he takes part in the coordination of Euclid's Instrument Operation Team and is responsible for the quality of the NISP data on which the Euclid releases are built. D. Fouchez leads the group's LSST activities and sits on the DESC Advisory Board. Supernova cosmology is one of RENOIR's core activities: the group built the filter-exchange auto-changer of the LSST camera, finalised the photometric calibration of ZTF — whose 2025 release provides the largest sample of nearby type Ia supernovae ever assembled, and the low-redshift anchor for LSST — and develops machine-learning methods for supernova detection, classification and photometric redshifts, now used within DESC; three PhD theses on type Ia supernovae have been defended here since 2023. The candidate will find both the time-domain and the Euclid-side expertise this project requires.

Twenty years after the discovery of the accelerating expansion of the universe through supernova measurements, the supernova probe remains one of the most accurate means of measuring the cosmological parameters of this recent period in the history of our universe, dominated by the so-called dark energy.

The Rubin Observatory began the Large Survey of Space and Time (Rubin/LSST) in June 2026. Rubin operates an 8.4-metre telescope fitted with a 3.2 billion-pixel camera, the most powerful ever built.

During the survey, half of the sky will be observed every three nights for ten years. This survey will make it possible to measure billions of galaxies with great precision, and to track the variation over time of all transient objects. Together with many other astrophysical studies, it will be a very powerful machine for determining cosmological parameters using many different probes and, in particular, will impose strong constraints on the nature of dark energy. The LSST project aims to discover up to half a million supernovae. This improvement of two to three statistical orders of magnitude over the current data set will enable precise testing of the parameters of dark energy, test general relativity and also impose new constraints on the isotropy of the universe.

As part of this PhD, we are offering the opportunity to contribute to the analysis of the first Rubin/LSST supernovae.

The PhD student will be responsible for further developing deep learning methods for the discovery and identification of supernovae and for applying these methods to the first images.

They will then contribute to the initial analyses using the supernovae they have helped to identify.

The LSST group at CPPM is already involved in precision photometry for LSST, with direct involvement in algorithm validation within DESC/LSST [1][2], and has proposed new deep learning methods to improve detection [5], photometric supernova identification [3] and photometric redshifts [4].

[1] https://www.lsst.org/content/lsst-science-drivers-reference-design-and-anticipated-data-products

[2] https://arxiv.org/abs/1211.0310

[3] https://arxiv.org/abs/1901.01298

[4] https://arxiv.org/abs/1806.06607

[5] https://arxiv.org/abs/2607.05393