ATLAS is one of the four main detectors at the Large Hadron Collider (LHC) and one of the two general-purpose detectors. It has a cylindrical symmetry that covers almost the entire solid angle and the onion-like layers of sub-detectors allow for unambiguous reconstruction of the final state objects from the high energy collisions.

Electrons are fundamental objects that are reconstructed and identified by the ATLAS detector. They are reconstructed from tracks in the inner detector and matching energy deposits in the electromagnetic calorimeter. Ensuring high electron reconstruction efficiency is crucial for the ATLAS physics programs. Therefore, assessing the performance of the electron reconstruction is very important. Additionally, measuring the electron reconstruction efficiency in both data and Monte Carlo (MC) simulations allows deriving correction factors (or Scale Factors) that are used throughout the ATLAS physics analyses to correct the electron reconstruction efficiency in MC simulations to the efficiency observed in data.

The candidate will work closely with the e/gamma team on performing the measurement of electron reconstruction efficiency with the freshly collected data in 2022 and early 2023. Besides the performance assessment, the selected candidate will work on the optimization of the existing code that performs the measurement, as well as work on improving the systematic uncertainties of the measurement.

Good command of Python and C++ programming languages would be an advantage but their prior knowledge is not mandatory.

The predictivity of the Standard Model (SM) of particle physics remains unchallenged by experimental results. After the tantalizing discovery of the Higgs boson at LHC, the measurements of properties such as its mass, spin, parity and its couplings with other SM particles have confirmed its SM-like nature. This goes hand in hand with the absence of direct signs of TeV physics beyond the SM from current direct searches.

Indeed, the excellent performance of the LHC in terms of delivered luminosity allowed the ATLAS and CMS experiments to set stringent limits on new particle masses well beyond the EW scale, thus worsening the naturalness problem. If the new physics scale lies well above the present experimentally probed energies, one would be left with the only experimental perspective of searching for deviations within the LHC precision measurements, and with no solid theoretical explanation of why the new physics should be so unnaturally heavy. There is, however, another logical possibility: new physics may be hidden at lower energies although weakly coupled to the SM known particles, so that its signals could be swamped in the SM background.

To process the enormous amount of data provided by the LHC, ATLAS uses an advanced “trigger” system to tell the detector which events to record and which to ignore. The ATLAS trigger is a two-level system composed of the first level, Level 1 (Level-1) trigger implemented in custom hardware, and High Level Trigger (HLT) which relies on selections made by algorithms implemented in software. The trigger is designed in such a way that an initial rate of collisions of 40 MHz is decreased to about 100 kHz after L1 and further decreased to 3 kHz at the HLT. This is a harsh limit on the possibility of recording low energetic events, swamped by high rate background, where signal of new physics could be hidden.

The Phase-I ATLAS Level-1 calorimeter trigger consists of a series of upgrades in order to face the challenges posed by the Run 3 LHC luminosity. The trigger upgrade benefits from new front-end electronics for parts of the calorimeter that provide the trigger system with digital data with a tenfold increase in granularity. This makes possible the implementation of more efficient algorithms to maintain the low trigger thresholds at much harsher LHC collision conditions.

The candidate will work on the analysis of the LHC data recorded in 2022 and 2023 to assess the quality of the data recorded by the upgraded ATLAS calorimeter system.

The data acquisition and trigger electronics of the ATLAS liquid argon calorimeter will be fully replaced as part of the second phase of upgrade of the ATLAS detector. The new backend electronics will be based on high-end FPGAs that will compute on-the-fly the energy deposited in the calorimeter before sending it to the trigger and data acquisition systems. New state-of-the-art algorithms, based on neural networks, are being developed to compute the energy and improve its resolution in the harsh conditions of the HL-LHC.

The candidate is expected to take a role in the development of data processing algorithms allowing to efficiently compute the energies deposited in the LAr calorimeters in the high pileup conditions expected at the HL-LHC. These algorithms will be based on AI techniques such as recurrent neural networks will be adapted to fit on hardware processing units based on high-end FPGAs. The successful candidate will be responsible of designing the AI algorithms, using python and keras, and assessing their performance. The candidate will also assess the effect of employing such algorithms for electromagnetic object reconstruction (especially at trigger level). She/he will work closely with the engineers designing the electronic cards at CPPM in order to adapt the AI algorithm to the specifics of FPGAs. Candidates with a strong interest for hardware will be encouraged to take part in the design of the firmware to program the FPGAs.

Prior knowledge of keras, python and C++ is desirable but not mandatory.

During the LHC Run 2, the ATLAS experiment was equipped with a fourth cylindrical layer of pixel sensors, which became the innermost layer, at an average radius of 3.5 cm. This layer makes use not only of planar pixel sensors but also of 3D sensors.

3D pixel sensors are a recent development of silicon pixel sensors.

Classical planar sensors have their p-n junction across their thickness, while 3D sensors have it between p and n columns drilled through the thickness. The inter-column distance being a fraction of thickness, desertion voltage and drift distance of charge careers are reduced, making 3D sensors more radiation hard than planar sensors.

The candidate will analyse real data from the ATLAS experiment collected during 2015 to 2018, and will focuse on the response of 3D sensors.

The candidate will measure collected charge, cluster size, check the expected zero-value of the Lorentz angle and measure the performance degradation du to radiations, and compare it with planar sensor's.

The ATLAS (http://atlas.cern) and CMS collaborations at the LHC have celebrated this year the 10th anniversary of the Higgs boson discovery, which led to a Nobel Prize for F. Englert and P. Higgs in 2013. LHC has started this year its new Run 3 data-taking period, which will allow to collect a lot of new data, in order to better characterize the Higgs boson and to possibly find evidences of new physics beyond the Standard Model. However, in order to increase by a factor 100 the amount of useful data we already have, the LHC and its detectors will be upgraded for the High-Luminosity phase of LHC (HL-LHC, 2029-2040). The ATLAS group at CPPM, building on its previous expertise, is developing a new pixel detector and the corresponding reconstruction algorithms to this end.

This high-tech detector plays a fundamental role to measure the trajectories of charged particles and to identify jets of particles stemming from the hadronization of bottom quarks. This ability, also known as b-tagging, is instrumental to the success of the ATLAS and LHC physics program and has played a major role in the past observation of the associated production mode of a Higgs boson with top quarks and in the search for the production of a pair of Higgs bosons. Recent b-tagging algorithms based on Deep-Learning techniques have already demonstrated sizeable improvements with the current Run 3 ATLAS detector and are therefore investigated also for HL-LHC.

The student will use detailed Monte-Carlo simulations to assess the b-tagging performance associated with the most recent detector simulation and investigate potential improvements for those. The project provides an opportunity for the student to get an exposure to a broad spectrum of topics: LHC physics notably the Higgs sector; basics of silicon detectors, track-finding and pattern-recognition; b-tagging algorithms based on Deep-Learning techniques. The project requires the use of existing analysis frameworks and plotting scripts, mostly based on Python. The prior knowledge of this language is desirable but not mandatory.

Being forbidden in the Standard Model (SM) of particle physics,lepton flavor violating decays are among the most powerful

probes to search for physics beyond the SM. In view of the recent anomalies seen by LHCb on tests of lepton flavor

universality in $b \to s \ell \ell$ and $b \to c\ ell \nu$ processes, the interest of tau lepton flavor violating

decays has been greatly reinforced. In particular, several new physics models predict branching

fractions of $\tau \to \ell V^{0}$ and $\tau \to \ell \ell \ell$ just below the current experimental limits.

The Belle II experiment located at KEK, Japan, has started the physics data taking in 2019, and is aiming at

50 times more data than its predecessor. Thanks to its clean environment and high $\tau^+ \tau^-$ cross section,

it provides an ideal environment to study tau decays. The CPPM group searches for lepton flavour violating $\tau$ decays,

such as $\tau^{\pm} \to \ell^{\pm} V^{0}$ or $\tau^{\pm} \to \ell^{\pm} \ell^{\mp} \ell^{\pm}$ , with V0 being a neutral vector meson and

$\ell^{\pm}$ an electron or muon.

The goal of this internship is to develop and use a Graph Neural Network (GNN) to reject background events.

Other architectures and implementations could be studied as well. The candidate will prepare data for training samples,

get familiar with the GNN, assess its performances and explore various formulation of the problem and different architectures.

The final objective is to use the GNN in the analyses of the channels studied in the group.

This internship can be continued with a PhD thesis.

Application including a CV, grade records and a motivation statement should be sent to giampi@cppm.in2p3.fr and serrano@cppm.in2p3.fr.

References:

https://arxiv.org/abs/1808.10567

https://hflav-eos.web.cern.ch/hflav-eos/tau/spring-2017/lfv-limits-plot.html

https://arxiv.org/abs/1903.11517

https://arxiv.org/pdf/1806.05689.pdf

https://arxiv.org/abs/2208.14924

The CTA (Cherenkov Telescope Array) is a worldwide project to construct the next generation ground based very high energy gamma ray instrument [1]-[2]. CTA will use tens of Imaging Air Cherenkov Telescopes (IACT) of three different sizes (mirror diameter of 4 m, 12 m and 23 m) deployed on two sites, one on each hemisphere (La Palma on the Canary Islands and Paranal in Chile). CTA will detect gamma-rays with energy ranging from 20 GeV up to 300 TeV by imaging the Cherenkov light emitted from the charge particle shower produced by the interaction of the primary gamma ray in the upper atmosphere.

The CTA unconventional capabilities will address some of the most intriguing questions of astroparticle physics such as the origin of very high energy galactic cosmic rays. The observatory completion is foreseen in 2025 but the first large size telescope (LST1) is already installed and taking data in La Palma. This telescope has a key role in the definition and validation of the methods and software tools for the future observatory.

This internship concerns the reconstruction of LST1 data by the use of innovative convolution neural network methods (CNN). Standard IACT event reconstruction is based on parametrization of the shower images and on machine learning algorithms trained with these parameters, for the estimation of the energy, the direction and the gammaness (the probability to be a gamma) of the primary particles [3]. Recently, several innovative reconstructions based on CNN have been developed in the context of CTA and of LST1 in particular [3]. Till now, these methods have been tested only on Monte Carlo simulated data. The goal of the internship is to verify their performance on data coming from real observations. The student will first make use of the newly proposed CNN methods to access their performance on simulated observations of the Crab Nebula (the standard candle of gamma-ray astronomy). Then, she/he will apply the same methods to Crab real data taken during the present LST1 commissioning phase.

The candidate needs a medium knowledge of the python programming language.

The internship will be in co-supervision with Dr. Rubén López Coto from the University of Padova (Italy). Candidates should send their CV and motivation letter as well as grades (Licence, M1 as well as their M2 if available) to cassol@cppm.in2p3.fr

A PhD contract can eventually follow the internship.

References:

[1] Science with the Cherenkov Telescope Array: https://arxiv.org/abs/1709.07997

[2] https://www.cta-observatory.org/

[3] Lopéz-Coto, R. et al. Physics Performance of the Large-Sized Telescope prototype of the Cherenkov Telescope Array, Proceeding of 37th International Cosmic Ray Conference (ICRC 2021)

[4] Grespan, P. et al., Deep-learning-driven event reconstruction applied to simulated data from a single Large-Sized Telescope of CTA, Proceeding of 37th International Cosmic Ray Conference (ICRC 2021)

KM3NeT/ORCA (Oscillation Research with Cosmics in the Abyss) is a deep sea

neutrino telescope currently under construction at a depth of 2500m in the

Mediterranean Sea off the coast of Toulon. ORCA is optimised for the detection of

low energy (3-100 GeV) atmospheric neutrinos and will allow precision studies

of neutrino oscillation properties. ORCA is part of the multi-site KM3NeT research infrastructure, which also incorporates a second telescope array (in Sicily) optimised for the detection of high-energy cosmic neutrinos.

The first ORCA detection strings have been operating for more than a year and are providing high quality data. During this stage the student will apply machine learning techniques to the data analysis with the aim to improve the angular and energy resolutions of the current event reconstruction algorithms. It is expected the candidate will follow this stage with a PhD on measuring the neutrino oscillation parameters.

Links:

http://antares.in2p3.fr

http://www.km3net.org

http://www.cppm.in2p3.fr/rubrique.php3?id_rubrique=259

Doing neutrino astronomy is a long quest in astroparticle physics. IceCube and ANTARES have found the first evidences of a few neutrino sources, mainly related to blazars (active galactic nuclei with their jets posting toward the Earth) and tidal disruption events. Most of those explosive events can release enormous amounts of energy both in electromagnetic radiation and in non-electromagnetic forms such as neutrinos and gravitational waves. This is at the basis of multi-messenger astronomy.

KM3NeT, the second generation neutrino detectors in the Mediterranean Sea, is in construction. It is taking data with a sensitivity much larger than ANTARES in the whole energy range, from GeV to PeV thanks to the complementarity of the 2 detectors: ORCA and ARCA. Already with the 30-40 detection units in operation, KM3NeT has significant better performances, either in term of effective area or in term of angular resolution.

In CPPM, we are mainly working on the implementation of multi-messenger analyses with high-energy neutrinos detected with ANTARES and KM3NeT neutrino telescopes. In this context, we are developing an analysis framework that is able to receive and process a time and spatial correlation analysis with high-energy neutrinos in coincidence with selected potential external triggers. Those analyses can be performed in real-time or offline including the most refined knowledge of the detector. In the last years, IceCube has provided alerts from selected high-energy neutrinos, and for some of them, a bright blazar has been located in the error box of the neutrino and found in active state with concomitant multi-wavelength observations.

During this internship, the student will perform an optimised analysis of the KM3NeT data for those interesting associations. It consists of the development of a neutrino selection based on the outputs of the event reconstructions and the event topology classifiers. This selection can be made with machine learning tools. After the event selection, the student will implement the correlation analysis.

This internship can be continued with a PhD in our group on the multi-messenger analyses with KM3NeT.

The analyses will be performed using C++ or python.

KM3NeT/ORCA (Oscillation Research with Cosmics in the Abyss) is a deep sea neutrino telescope currently under construction at a depth of 2500m in the Mediterranean Sea off the coast of Toulon. ORCA is optimised for the detection of atmospheric neutrinos in the energy range 3-100 GeV and will allow precision studies of neutrino properties. Currently the detector takes data with 11 detection strings hosting more than 6000 photomultiplier tubes.

During this internship at the Centre de Physique des Particules de Marseille, the student will analyse data taking with the ORCA detector in the period 2020 to 2022. The goal is to obtain detector response functions for neutrinos in the relevant energy range and to derive sensitivities for the observation of sterile neutrinos. Software tools which have been developed at CPPM will be used.

Links:

http://www.km3net.org

http://www.cppm.in2p3.f/rubrique.php3?id_rubrique=259

The Neutrino Team at CPPM is strongly involved in the KM3NeT/ORCA neutrino telescope, under construction in the abyss (-2500m) of the Mediterranean sea, 40km offshore Toulon. The first detection units that have been deployed are successfully collecting data. The detector is now large enough to access yet unexplored physics territories. A very exciting topic is the search for tau neutrinos appearing in the neutrino flux created in the collisions of cosmic rays in the atmosphere. The appearance probability is poorly known and KM3NeT/ORCA has a unique potential to measure it. Such measurements could lead to a major discovery regarding the existence of sterile neutrinos.

One of the keystones for these studies is the tag of the neutrino flavours (electron, muon, or tau); hence, in this project, the student will develop Machine Learning algorithms to perform this kind of identifications. The expected skills are to master the basics of neutrino oscillation and to program in python, c++, ROOT.

Doing neutrino astronomy is a long dream in astroparticle physics. IceCube and ANTARES have found the first evidences of a few neutrino sources, mainly related to blazars (active galactic nuclei with their jets posting toward the Earth) and tidal disruption events. For most of those explosive events can release enormous amounts of energy both in electromagnetic radiation and in non-electromagnetic forms such as neutrinos and gravitational waves. This is at the basis of multi-messenger astronomy. KM3NeT, the second generation neutrino detectors in the Mediterranean Sea, will have significant better performances, either in term of effective area or in term of angular resolution.

In CPPM, we are mainly working on the development of multi-messenger analyses with high-energy neutrinos detected with ANTARES and KM3NeT neutrino telescopes. In this context, we are developing a real-time analysis framework that is able to send neutrino alerts and to receive and process a cross-match analysis with high-energy neutrinos in coincidence with selected potential external triggers.

During this intern ship, the student will implement a neutrino selection module that takes in inputs the reconstructed and classified neutrino streams. To reach a sustainable false alert rate (1-2 per month), It will be necessary to filter on the topology of the events, the multiplicity, the energy and the estimate of the reconstruction error. The student will have to implement such module based on machine learning tool.

The analyses will be performed using C++ or python.

Description

Deadline to apply: December 30th

Scientific Context

According to modern physics, matter should not have emerged from the Big Bang and its origin remains one of the most profound riddles in fundamental physics. The heart of this mystery is the CP symmetry, i.e. the fact that the laws of physics are the same for matter and anti-matter. During the Big Bang, this symmetry should have maintained particles and anti-particles in equal quantities while they were gradually annihilating each other leaving, at the end, nothing but pure energy. The existence of matter thus requires the CP symmetry to be violated, which is one of the so-called Sakharov conditions (Sakharov, 1967).

The works awarded by the 2015 Nobel Prize imply that neutrino physics can break this symmetry via a quantum phenomenon called neutrino oscillations. Neutrinos can be produced in three types – or flavors – the electron neutrino ( $\nu_{e}$ ), the muon neutrino ( $\nu_{\mu}$ ) and the tau neutrino ( $\nu_{\tau}$ ). Experimental evidence showed that the flavor of neutrinos oscillates when they propagate. In theory, this oscillation can be different for neutrinos and anti-neutrinos which would break the CP symmetry. Discovering such an effect would be a major breakthrough in fundamental physics. Intense experimental efforts (Acciarri at al., 2015; Abe et al., 2018) are thus ongoing worldwide to study the neutrino oscillations. However, the experimental techniques used so far are reaching their limits (Branca, et al., 2021). New techniques are thus urgently needed. The goal of this Master 2 project is to design such a new technique: the neutrino tagging (Perrin-Terrin, 2022).

Neutrino Tagging

At accelerator-based experiments, neutrinos are produced by colliding protons on a target. These collisions produce secondary particles, in particular, $\pi^+$ and $\pi^-$ , which decay as $\pi^+ \to \mu^+ \nu_\mu$ and $\pi^- \to \mu^- \bar\nu_\mu$ and so produce a $\nu_\mu$ and anti-$\nu_\mu$ beam. The optimal propagation distance to observe the neutrino oscillations depends on the neutrino energy and ranges between 100~km to 1000~km. A neutrino detector consisting of an instrumented target is installed at this distance to measure the neutrino flavor and energy. Conventionally, the neutrino characteristics are measured from their interaction in a densely instrumented detector, as shown in Fig 1-(left). The complexity of the interaction mechanisms induces strong limitations on the precisions of the energy measurements.


The tagging technique (Perrin-Terrin, 2022) proposes to determine the neutrino characteristics using the production mechanisms, as shown in Fig 1-(right). These mechanisms, the $\pi \to \mu \nu$ decays, are extremely simple processes. Hence, once the $\pi$ and $\mu$ characteristics (time, momentum, charge: t,p,±) are measured, a simple kinematical relation allows to derive precisely the neutrino characteristics. For example, while the precision of the neutrino energy measurement based on the interaction plateaus at about 15%, the one of the tagging can easily reach 1%. In these conditions, the only task left to the neutrino detector is to identify the flavor of the neutrino after propagation. These relaxed requirements allow to use seawater neutrino detectors such as KM3NeT/ORCA (Adrian-Martinez et al., 2016) under construction at a depth of 2450~m offshore Toulon. These detectors are extremely large (several Mton) which increases the probability for neutrinos to interact in them. The key element of the technique is the tagger, i.e. the detector based on which the neutrino properties are estimated. The tagger will be composed of several planes of cutting-edge high time precision silicon pixel detectors (Lai, 2018) able to sustain the extremely high rate of particles in the beam line: 1012 per second! A proof-of-concept of this technique is being performed using the NA62 experiment (Cortina Gil at al., 2017) at CERN as a miniature neutrino experiment (Martino, 2022).

Objectives of the Project

The project aims to co-design the detector layout (time resolution, number of planes and location etc.) and the algorithms to estimate the neutrino characteristics. Achieving this objective will require to:

• design a statistical model describing the physical setup,
• develop a simulation of the detector,
• calculate the optimal performance bounds,
• optimize the detector setup based on the calculated bounds,
• test/apply the model and bounds to NA62.

Working Environment

The student will be based at IM2NP, located at the La Garde campus in Toulon, or/and at CPPM located in the Luminy campus in Marseille, at the entrance of the Parc National des Calanques. Both places are located near remarkable natural sites and offer pleasant working and living conditions.

The student will integrate the Signal And Tracking (STr) Team of IM2NP and the Neutrino team of CPPM. The STr team is composed of ~10 people who are specialized in applied statistics and signal processing for different domains (astrophysics, optics, RADAR, SONAR, LIFI, etc. (Roueff, Arnaubec, Dubois-Fernandez, Refregier, 2011; Roueff, et al., 2020; Roueff, Roux, Réfrégier, 2009)). The team at CPPM is composed of ~30 people, researchers, postdocs, PhD students, engineers and technicians with a large panel of skills and working on the construction and exploitation of the KM3NeT detectors. A PhD student is also working on the proof-of-principle of the neutrino tagging technique at NA62.

The student will be involved in a dynamic international team. The project will be developed in close collaboration with CERN. Indeed, the tagger studied in this project is meant to be installed in the neutrino beam line that CERN has started to study to for the tagging. Regular video-conference meetings which CERN collaborators will be held during the project, and short trips to CERN for in-person meetings can also be envisaged.

The last step of the project (application and test at NA62) will be done in collaboration with several members of the NA62 experiment at University of Birmingham (UK), Ecole Fédérale Polytechnique de Lausanne (Switzerland) and in Université Catholique de Louvain (Belgium) who are all actively contributing to the proof-of-principle of the neutrino tagging technique.

Student Profile

We are seeking for a highly motivated student who could consider continuing this work for a PhD thesis. The student should ideally have:

• basic knowledge of experimental particle physics,
• skills in applied statistics,
• previous experience in machine learning techniques using c/c++, matlab or python,
• oral and written proficiency in English and French.

Application Procedure

The student interested in applying for the internship must provide:

• CV
• Motivation letter
• Grades from M1
• Available grades from M2
• Desired internship duration
• Desired internship starting date (optional)
• Reference letter or reference contact (optional)

This information should be sent to antoine.roueff@univ-tln.fr and mathieu.perrin-terrin@cern.ch before December 30th. Interviews will be conducted beginning of January.

PhD Perspectives

Beyond this Master Project, the student could then enroll in a PhD thesis in the context of the KM3NeT experiment, the NA62 experiment and the neutrino tagging studies at CERN.

Bibliography

Abe, K., at al. (2018, May). Hyper-Kamiokande Design Report. Hyper-Kamiokande Design Report.

Acciarri, R., at al. (2015). Long-Baseline Neutrino Facility (LBNF) and Deep Underground Neutrino Experiment (DUNE).

Adrian-Martinez, S., at al. (2016). Letter of intent for KM3NeT 2.0. J. Phys., G43, 084001. doi:10.1088/0954-3899/43/8/084001

Branca, A., Brunetti, G., Longhin, A., Martini, M., Pupilli, F., Terranova, F. (2021). A New Generation of Neutrino Cross Section Experiments: Challenges and Opportunities. Symmetry, 13, 1625. doi:10.3390/sym13091625

Cortina Gil, E., at al. (2017). The Beam and detector of the NA62 experiment at CERN. JINST, 12, P05025. doi:10.1088/1748-0221/12/05/P05025

Lai, A. (2018). A System Approach towards Future Trackers at High Luminosity Colliders: the TIMESPOT Project. (pp. 1–3). Sydney: IEEE. doi:10.1109/NSSMIC.2018.8824310

Martino, B. D. (2022, July). Tagged Neutrino Beams. Tagged Neutrino Beams. Zenodo. doi:10.5281/zenodo.6785370

Perrin-Terrin, M. (2022). Neutrino tagging: a new tool for accelerator based neutrino experiments. Eur. Phys. J. C, 82, 465. doi:10.1140/epjc/s10052-022-10397-8

Roueff, A., Arnaubec, A., Dubois-Fernandez, P. C., Refregier, P. (2011). Cramer–Rao Lower Bound Analysis of Vegetation Height Estimation With Random Volume Over Ground Model and Polarimetric SAR Interferometry. IEEE Geoscience and Remote Sensing Letters, 8(6), 1115–1119. doi:10.1109/LGRS.2011.2157891

Roueff, A., Gerin, M., Gratier, P., Levrier, F., Pety, J., Gaudel, M., . . . Sievers, A. (2020, May). C18O, 13CO, and 12CO abundances and excitation temperatures in the Orion B molecular cloud: An analysis of the precision achievable when modeling spectral line within the Local Thermodynamic Equilibrium approximation. AA 645, A26 (2021). doi:10.1051/0004-6361/202037776

Roueff, A., Roux, P., Réfrégier, P. (2009, April). Wave separation in ambient seismic noise using intrinsic coherence and polarization filtering. Signal Processing, 89, 410–421. doi:10.1016/j.sigpro.2008.09.008

Sakharov, A. D. (1967). Violation of CP Invariance, C asymmetry, and baryon asymmetry of the universe. Pisma Zh. Eksp. Teor. Fiz., 5, 32–35. doi:10.1070/PU1991v034n05ABEH002497

Twenty years after the discovery of the current acceleration of the expansion of the universe by supernova measurements, the supernova probe remains the most accurate way to measure the parameters of this recent period in the history of our universe dominated by the so-called dark energy.

The precision measurements that can be performed by the supenova probe will be a crucial element that, in combination with other probes (LSS, weak lenses, CMB, etc.), will put strong constraints on the nature of dark energy. This will be made possible by the exceptional Supernova data set to be provided by LSST, with a combination of huge statistics and extreme calibration accuracy.

The Rubin observatory with the Large Survey of Space and Time (Rubin/LSST) project will be commissioned in 2022 and will run at full speed by the end of 2023. It is an 8.4-metre telescope with a 3.2 billion pixel camera, the most powerful ever built.

This telescope will take a picture of half the sky every three nights for ten years. This survey will make it possible to measure billions of galaxies with great accuracy and to track the variation over time of all transient objects. With many other astrophysical studies, it will be a very powerful machine for determining cosmological parameters using many different probes and, in particular, it will impose strong constraints on the nature of dark energy. The LSST project aims to discover up to half a million supernovae. This two to three orders of magnitude improvement in statistics over the current data set will allow accurate testing of dark energy parameters and will also impose new constraints on the universe's isotropy.

In this Master 2 internship we propose to prepare the first analysis of LSST supernova data by performing an analysis using LSST software and our deep learning method for identifying supernova on existing HSC/Subsaru data. Indeed, the HSC data has characteristics that are very close to what we expect with Rubin/LSST. The CPPM LSST group is already engaged in precision photometry work for LSST with direct involvement in algorithm validation within DESC/LSST [1][2][3] and has proposed a new deep learning method to improve the photometric identification of supernovae [4] and photometric redshifts [5].

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

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

[3] https://www.lsst.org/about/dm

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

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

[6] https://arxiv.org/abs/1401.4064

More than 20 years after the discovery of the accelerating expansion of the Universe, the mystery as to the nature of dark energy remains. The current data are in agreement with the ?CDM concordance model, consistent with a cosmological constant. However there are tensions between the different measurements that could herald fundamental discoveries in cosmology. For example, the current H0 expansion, measured with observations of the nearby universe is higher than that predicted by ?CDM and the cosmological diffuse background data with an incompatibility of the order of 5?.

Type 1a supernovae (SN1a) are standardizable candles, i.e. objects whose luminosity can be predicted. By comparing these observed luminosities with the emitted one, their distance (D) can be inferred accurately. By combining with measurements of the redshifts (z) of the spectra of the host galaxies, redshifts due to the expansion of the universe, we can construct a Hubble diagram (D vs z), which allows us to measure the acceleration of the expansion. By calibrating these distances locally with cepheids for example, we can measure the current expansion H0. Moreover, if we assume to know our cosmological model, we can predict what is the redshift for a given distance, which allows us to measure other shifts, such as those due to the particular velocities of SN1a. We can then reconstruct the cosmic velocity fields, which are due to the collapse of matter during the formation of structures, dictated by the law of gravitation.

The constraints on the concordance model, as well as on modified gravitation theories, will be improved by the new generation of sky surveys: Zwicky Transient Facility (ZTF) and the Rubin observatory Legacy Survey of Space and Time (LSST) which have the particularity to observe SN1a on more than half of the sky. ZTF has been taking data since 2017, and LSST will begin in the fall of 2024.

A pipeline has been set up at CPPM to study the growth rate of structures (fs8) [1,2] by reconstructing the particular velocity of supernovae from realistic simulations of the ZTF survey.

During the M2 Master internship, the student will develop tools either to adapt this pipeline to LSST [3], or to study other aspects, such as the anisotropy of the expansion of the Universe [4,5,6], also made possible by these surveys over a very large part of the sky.

The LSST group of the CPPM is already involved in precision photometry for LSST with a direct involvement in the validation of algorithms within LSST and has proposed a new deep learning method to improve the photometric identification of supernovae [7] and photometric redshifts [8]. A thesis is already in progress to develop an analysis pipeline for the measurement of structure growth rate for ZTF. The student will work in this framework by improving the pipeline and adapting it to LSST, and may also prepare an analysis of the anisotropy measurement of the expansion of the Universe.

[1] https://arxiv.org/abs/1609.08247

[2] https://arxiv.org/abs/2001.09095

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

[4] https://arxiv.org/abs/2205.12692

[5] https://arxiv.org/abs/1808.04597

[6] https://arxiv.org/abs/1912.02191

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

[8] https://arxiv.org/abs/1401.4064

Although the universe is well described by the concordance model ?CDM, the nature of its components, dark matter and dark energy, remains a major puzzle of modern cosmology. While historically most attention has been paid to the overdense regions, the underdense regions account for about 80 per cent of the total volume of the observable Universe and strongly influence the growth of large-scale structure. As voids are nearly devoid of matter, they have proved to be very promising objects for exploring the imprint of possible modifications of General Relativity (GR) such as f(R) gravity or extended gravity theories.

The RENOIR cosmology team at CPPM focuses on the understanding of the history and composition of our Universe, particularly on its dark components. The team is particularly involved in large spectroscopic surveys Dark Energy Spectroscopic Instrument at Mayall, US and the European space mission Euclid, that will provide the observation of 40 million of galaxies, the largest 3D map of the Universe ever made.

A promising way to probe modified gravity models is to constrain the growth of structure of the Universe using information from Redshift Space Distortions around cosmic voids. The aim of the internship is to test and quantify the importance of reconstruction methods which aims to separate peculiar velocity components from the Hubble flow in redshift space, and see the impact on the construction of void catalogs.

This subject can be pursued by a thesis on the extraction of cosmological constraints using Alcock-Paczynski deformation information and RSD information around voids, with DESI data which started its observations in June 2021 for 5 years, and the Euclid mission that will be launched in July 2023.

Contexte :

La mission Euclid (http://www.euclid-ec.org) est un projet majeur de l'ESA qui lancera en 2023 un télescope spatial dédié à la compréhension de l'Univers et réalisera une cartographie de tout le ciel. D'une précision jamais atteinte auparavant, ces mesures des grandes structures de l'Univers lointain permettront de tester le modèle cosmologique et en particulier de questionner la nature de l'énergie noire. La cartographie sera obtenue grâce au spectrophotomètre NISP et les 16 détecteurs infrarouges de son plan focal dont le CPPM a réalisé la calibration au sol, étape fondamentale pour valider les performances de l'instrument.

Activité principale :

Les détecteurs infrarouges du NISP ont été développés expressément pour la mission Euclid. À la pointe de la technologie, chacun est constitué d'une matrice de 2048 x 2048 pixels. Leur calibration fine a été réalisée au CPPM et a donné lieu à l'enregistrement de 500 To de données à analyser. Ces données montrent clairement la présence de persistance qui vient polluer les données pendant plusieurs heures d'acquisition. Afin d'acquérir une meilleure compréhension du phénomène, l'ingénieur/l'ingénieure-stagiaire cherchera à caractériser la persistance et à comprendre l'influence des paramètres environnementaux sur celle-ci.

Pour ce faire, l'ingénieur/l'ingénieure-stagiaire devra~appliquer les méthodes d'analyse classiques (fit par des fonctions plus ou moins simples) suivant les étapes :

- Implémenter les méthodes choisies en python

- Extraire des grandeurs comme constante de temps et amplitude à partir des données de calibration existantes

- Analyser les corrélations entre la persistance et les données environnementales

Connaissances requises :

- Base solide en programmation en langage python

- Bonnes connaissances en traitement du signal

- Connaissances en physique du semi-conducteur

The imXgam research team conducts interdisciplinary research activities for imaging applications of ionizing radiation in the health and energy fields. It participates in the TIARA (Time-of-flight Imaging Array) project, whose objective is to reduce the uncertainties related to the path of protons during proton therapy treatments through the development of a detector for time-of-flight imaging of prompt gamma rays (PGs) created during irradiation.

The accuracy of proton therapy is currently limited by the uncertainties related to the proton path, which result from the composition of the patient's tissues, physiological movements or transient changes in the anatomy and which lead to the use of large safety margins (up to 1 cm) to avoid irradiating healthy tissues. The purpose of PG imaging is to allow real-time control of tumor treatment [1]. To fully exploit its potential, an innovative real-time treatment control detector based on time-of-flight imaging of PGs with a temporal resolution of 100 ps is proposed [2]. This detector consists of a set of lead fluoride Cherenkov converters of about 1 cm³ each surrounding the irradiated volume and read in coincidence with a beam monitor. The principle consists in measuring precisely (better than 100 ps) the time difference between the time of passage of the protons in the beam monitor based on a diamond detector and the time of arrival of the PGs in the Cherenkov converters, which corresponds to the time-of-flight of the proton between its passage in the beam monitor until its interaction in the tissues followed by the time-of-flight of the PG emitted during this interaction until its detection by TIARA. This time difference, knowing the position of the TIARA detectors, constrains the coordinates of the point of emission of the PGs, which allows a 3D reconstruction of the proton path in real time with a millimeter precision [3].

A 3D reconstruction algorithm of the proton path in real time specific to the TIARA detector and its physics has been developed and validated on analytical data. The objective of this internship is to evaluate the performance of this algorithm on more realistic data obtained by Monte-Carlo simulation. In a first part, the TIARA experiment will be modelled on the Monte-Carlo simulation platform GATE and data corresponding to a realistic benchmark will be generated. In a second part, the robustness and accuracy of the 3D reconstruction algorithm already developed will be evaluated on these Monte-Carlo data.

These developments will mainly use Python and GATE/Geant4 languages.

Candidates are invited to contact the person in charge of the thesis subject by attaching a CV with a letter of motivation and the last transcript (the one of the previous year and the one of the current semester, if available).

[1] J Krimmer et al., Prompt-gamma monitoring in hadrontherapy: A review, Nucl. Instrum. Methods A 878 (2018) 58-73

[2] S. Marcatili et al., Ultra-fast prompt gamma detection in single proton counting regime for range monitoring in particle therapy, Phys. Med. Biol. 65 (2020) 45033

[3] M. Jacquet et al., A time-of-flight-based reconstruction for real-time prompt-gamma imaging in protontherapy, Phys. Med. Biol. 66 (2021) 135003

Context

The imXgam research team conducts interdisciplinary research activities for imaging applications of ionising radiation in the field of health and energy. The internship topic proposed here aims at improving the temporal performance of gamma ray detectors in the context of time-of-flight positron emission tomography (PET).

The coincidence time resolution (CTR) of state-of-the-art clinical TOF-PET cameras is around 210 ps FWHM. A CTR of 10 ps FWHM would allow the position of an electron-positron annihilation to be located at better than 1.5 mm FWHM, making it possible to obtain a PET image virtually without tomographic inversion [1]. One possible way to improve the temporal performance of detectors is to exploit the Cherenkov radiation generated by the motion of photoelectrons in a transparent interaction medium [2]. If this transparent medium is also scintillating, then two types of visible light photons are emitted with temporal distributions different from each other, the first almost simultaneously by the Cherenkov effect and the second slightly delayed by de-excitation of a radiative centre causing the scintillation phenomenon [3]. The photons are then likely to undergo reflections from the faces of the transparent medium before being collected by a photodetector(s) in order to accurately label the photoelectric interaction with a detection time. The existence of different temporal distributions makes the measurement of CTR complex [4].

Aim of the internship

The aim of the internship is to determine whether deep learning techniques can be used to label these two populations of visible light photons given their time and location of detection and, if so, to accurately date the photoelectric interaction in order to improve the temporal resolution of the coincidence. The trainee will build Monte Carlo datasets using GATE [5] to simulate the interaction of 511 keV gamma rays in a scintillating medium and exploit Monte Carlo truth to learn the position of the photoelectric interaction, the type of emission (Cherenkov or scintillation) and the time of the interaction, and then seek to reduce the dimension of the phase space as much as possible while preserving the accuracy of the observables retrieved by deep learning.

Required knowledge: Python programming, knowledge of radiation-matter interactions, notions of deep learning (DL)

[1] P. Lecoq, C. Morel et al. Roadmap towards the 10 ps time-of-flight PET challenge, Phys. Med. Biol. 65 (2020) 21RM01

[2] S.K. Kwon et al, Ultrafast timing enables reconstruction-free positron emission imaging, Nat. Photon. (2021) https://doi.org/10.1038/s41566-021-00871-2

[3] D. Yvon et al, Design study of a scintronic crystal targeting tens of picoseconds time resolution for gamma ray imaging: the ClearMind detector, J. Instrum. 15 (2020) P07029

[4] J. Nuyts et al. Estimating the relative SNR of individual TOF-PET events for Gaussian and non-Gaussian TOF-kernels, in Proc. Fully-3D'2021, G. Schramm, A. Rezaei, K. Thielemans and J. Nuyts eds, pp. 19-23.

The imXgam research team conducts interdisciplinary research activities for imaging applications of ionising radiation in the field of health and energy. The internship topic proposed here aims at improving the performance of an automatic segmentation process of liver tumours in the context of small animal CT.

Context

This internship is part of the DePIcT project financed by the Mission pour les Initiatives Transverses et Interdisciplinaires of the CNRS (https://miti.cnrs.fr/projet-multi-quipe/depict/ , https://www.in2p3.cnrs.fr/fr/cnrsinfo/palmares-des-80prime-2020-4-projets-pilotes-par-lin2p3-decrochent-un-financement). As part of a preclinical study on hepatocellular carcinoma, longitudinal in vivo follow-ups are carried out using the PIXSCAN-FLI, a photon counting micro-CT developed at the CPPM. It has been demonstrated that photon counting guarantees very high contrast on the 3D images.

In addition, the ultra-fast acquisition (100 images per second) allows the capture of the respiratory movements of the mouse. This study is based on an imaging protocol established at the CPPM (Cassol et al. 2019), which consists in labelling the liver with Barium nanoparticles, a contrast agent absorbed by liver macrophages. Tumours appearing in negative can then be observed and characterized due to the radiopacity of the contrast agent surrounding the tumours. This technique allows the differentiation of liver from tumours and the estimation of a series of important tumour parameters over time (size, shape, etc.)

Objectives

The aim of the internship is to implement and evaluate the performance of a state-of-the-art Deep Learning method in micro-CT for automatic segmentation (Léger et al. 2018, Brion et al. 2020). The aim here is to automatically segment the liver as well as liver tumours. The trainee will be able to rely on a large real-world database for which tumour segmentation by an expert is already available. If these methods prove satisfactory, they will be incorporated into the PIXSCAN-FLI automatic data processing pipeline for routine use.

This study may include analysis and correction of mouse breathing movements to improve the sharpness of 3D images.

Skills required: Python programming, deep learning. Knowledge of the context and physics of CT imaging will be appreciated.

Bibliography

E. Brion et al, Domain adversarial networks and intensity-based data augmentation for male pelvic organ segmentation in cone beam CT in Computers in Biology and Medicine https://dial.uclouvain.be/pr/boreal/object/boreal:245104

J. Léger et al, Contour Propagation in CT Scans with Convolutional Neural Networks in Advanced Concepts for Intelligent Vision Systems https://dial.uclouvain.be/pr/boreal/object/boreal:203221

F. Cassol et al, Tracking dynamics of spontaneous tumours in mice using Photon Counting Computed Tomography, iScience 21 (2019) 68-83 https://www.sciencedirect.com/science/article/pii/S2589004219303943

The data acquisition and trigger electronics of the ATLAS liquid argon calorimeter will be fully replaced as part of the second phase of upgrade of the ATLAS detector. The new backend electronics will be based on high-end FPGAs that will compute on-the-fly the energy deposited in the calorimeter before sending it to the trigger and data acquisition systems. New state-of-the-art algorithms, based on neural networks, are being developed to compute the energy and improve its resolution in the harsh conditions of the HL-LHC.

The candidate is expected to take a role in the development of data processing algorithms allowing to efficiently compute the energies deposited in the LAr calorimeters in the high pileup conditions expected at the HL-LHC. These algorithms will be based on AI techniques such as recurrent neural networks will be adapted to fit on hardware processing units based on high-end FPGAs. The successful candidate will be responsible of designing the AI algorithms, using python and keras, and assessing their performance. The candidate will also assess the effect of employing such algorithms for electromagnetic object reconstruction (especially at trigger level). She/he will work closely with the engineers designing the electronic cards at CPPM in order to adapt the AI algorithm to the specifics of FPGAs. Candidates with a strong interest for hardware will be encouraged to take part in the design of the firmware to program the FPGAs.

Prior knowledge of keras, python and C++ is desirable but not mandatory.

Our group is developing machine learning for embedded trigger systems for the ATLAS detector, and as such the developed neural networks need to be optimized for speed and resource consumption on Field Programmable Gate Arrays (FPGAs). The networks are developed using Tensorflow/Keras which is a Python library, but historically C++ has been the language used in particle physics for its speed. While Python has a reputation as a slow language, that is not anymore the case with new methods such as Numba that uses LLVM and Just In Time (JIT) compilation to optimize the performance. The purpose of this internship is to develop methods for fast data processing with Python that would replace and complement the existing parts written in C++ or (slow) Python. This work consists of writing and evaluating Numba as a method to speed up Python used in the data processing and analysis as well as possibility to merge Python with C++, such as pybind11. This internship would also consist of developing neural networks with tools suitable to create more optimized networks that would retain their original accuracy, but be compressed to run fast and consume low resources when running on FPGAs.

The project is done in English. The candidate is expected to have knowledge in Python programming and good communication skills in English. The project length is expected to be 2 to 3 months and can start in spring 2022.

The CPPM is developing an application to present the laboratory's activities to a high school audience. The digital communication assistant trainee will be part of the project team and will contribute to the production of multimedia contents, with several components: an editorial section on fundamental topics, video interviews presenting the jobs and skills, knowledge learning tests. The duration of the internship is 3 months.

Time-domain astronomy has received a considerable boost in recent years due to its ability to study extreme physics, to track cataclysmic phenomena like the birth of stellar mass black holes or the mergers of neutron stars, to probe distant

regions of the Universe, and to identify candidate sources for multi-messenger astrophysics. These explosive events can release enormous amounts of energy both in electromagnetic radiation and in non-electromagnetic forms such as neutrinos

and gravitational waves. They lie at the frontier of our understanding of the laws of physics under the most extreme conditions. Multi-messenger astronomy – the observation of astrophysical objects and processes using combinations of different messengers such as electromagnetic radiation, neutrinos, cosmic rays and gravitational waves – has emerged as a major new field in astronomy during the last years.

In CPPM, we are mainly working on the development of multi-messenger analyses with high-energy neutrinos detected with ANTARES and KM3NeT neutrino telescopes. In this context, we are developing a real-time analysis framework that is able to send neutrino alerts and to receive and process a cross-match analysis with high-energy neutrinos. In the next years, the LSST telescope in Chile will be one of the major discover of optical transients. Around a million triggers are expected each night. To account for these large numbers, LSST is developing some brokers to filter the alerts. In France, some colleagues are implementing the FINK broker (https://arxiv.org/abs/2009.10185). Some actual data are available with the ZTF telescope in US.

During this intern ship, the student will implement a filter chain in the broker to identify the most interesting candidates for the neutrino searches. It will filter on the nature of the transient, the number of detections, the light-curve and some cross-matches with astrophysical catalogues.

The analyses will be performed using C++ or python.

The KM3NeT experiment is a next-generation neutrino telescope, currently under construction in the two sites: ORCA (Oscillation Research with Cosmics in the Abyss) and ARCA (Astroparticle Research with Cosmics in the Abyss).

The ORCA site foresees the installation of 115 detection units (DUs), at 2500m depth in the Mediterranean Sea off Toulon and it is optimized for the atmospheric neutrino detection in the 3-100 GeV energy range, allowing for precision studies of neutrino oscillation parameters. Currently, only partially instrumented, KM3NeT/ORCA has already been operating for several months and it is collecting high-quality data. During this internship, the student will have the opportunity to become familiar with neutrino physics and analyze the KM3NeT/ORCA data collected so far, for studying neutrino oscillation properties. It is expected that the candidate will pursue similar studies in an M2 internship and eventually a Ph.D. in our group.

The accelerated expansion of the Universe and the mass of the neutrino species are fundamental problems in physics that can be tackled with cosmological observations. The Dark Energy Spectroscopic Instrument is an ongoing spectroscopic survey mapping the three-dimensional distribution of matter in the Universe through precise redshift measurements. At high-redshifts (2

The imXgam research team conducts interdisciplinary research activities for imaging applications of ionizing radiation in the health and energy fields. It participates in the ClearMind project whose objective is to develop an optimized detector for highly time-resolved applications, in particular for time-of-flight positron emission tomography (PET).

The measurement of the time of flight of a pair of annihilation photons, i.e. the time between the detection of the two 511 keV photons, allows to constrain the tomographic inversion within a back-projection range determined by the accuracy of the time-of-flight measurement, which is given by the coincidence time resolution (CTR). Knowing that the speed of light in vacuum is 30 cm/ns, a CTR of 10 ps FWHM would allow to localize the electron-positron annihilation with an accuracy of 1.5 mm FWHM, which would be sufficient to obtain an image of the distribution of annihilation points virtually without reconstruction and thus limit the dose required to obtain an image quality equivalent to that of the clinical PET cameras. Currently, state-of-the-art cameras achieve a CTR of 215 ps FWHM. The objective of the ClearMind project is to improve the temporal resolution of the detectors by using a scintillating lead tungstate (PWO) crystal as the input window of a microchannel plate photomultiplier tube (MCP-PMT) and to deposit a photocathode directly on the inner face of the PWO crystal in order to avoid total reflections of scintillation and Cherenkov photons on the PWO/photocathode interface to improve the collection of Cherenkov photons whose emission is practically instantaneous when a photoelectric electron is emitted at a speed higher than the speed of light in the PWO [1].

A mechanical bench for tomographic experimentation called tomXgam has been built at CPPM on which the first prototypes obtained in the framework of the ClearMind project are mounted. The objective of the internship is to participate to the first measurement campaign on tomXgam in order to confirm its readiness and to characterize the performances of detectors embedded on this device.

Candidates are invited to contact the person in charge of the thesis subject by attaching a CV with a letter of motivation and the last transcript (the one of the previous year and the one of the current semester, if available).

[1] D. Yvon et al., Design study of a scintronic crystal targeting tens of picoseconds time resolution for gamma ray imaging: the ClearMind detector, J. Instrum. 15 (2020) P07029

Le CPPM (CNRS et Aix-Marseille Université) a pour mission d'explorer et d'accroitre ses connaissances dans le domaine de la physique de la matière et de l'Univers.

Le Correspond Formation du CPPM est chargé de recueillir les besoins en formation du personnel afin de les retranscrire dans un document appelé Plan de Formation du L'Unité.

Missions

Le/la stagiaire rejoindra le service administratif du CPPM sous la responsabilité de Madame Isabelle Richer-Gonzalez, Correspondante Formation afin de l'assister dans la réalisation du plan de formation.

Activités principales

Le ou la stagiaire pourra réaliser les missions suivantes :

• Recueil des demandes de formation

• Établissement d'un bilan et de tableaux de bords des actions de formation passées

• Établissement du bilan des actions demandées / actions réalisées

• Aide à la rédaction du plan de formation de l'unité

• Diffusion de l'information et des annonces de formation

• Suivi des dossiers de demandes individuelles

Connaissances appréciées

• Droit de la formation

• Dispositifs de la formation continue

• Bonne maîtrise des outils bureautiques (Excel, Word, Powerpoint…)

• Techniques de présentation écrite et orale

Compétences opérationnelles

• Aptitudes relationnelles

• Capacités de synthèse et d'analyse

• Conception de tableaux de bord

Formations possibles

Ce stage conviendrait par exemple à un étudiant (Bac+1 ou Bac+2) préparant un diplôme en ressources humaines, un BTS ou un DUT Gestion des Entreprises et des Administrations (GEA) option « Ressources humaines ».

Stage pouvant aller de 4 et 7 semaines entre le 15 mai et 15 juillet 2023.

Contact : CV + lettre de motivation à Isabelle Richer-Gonzalez - Correspondant formation du CPPM

Tél : +33 4 91 82 72 26- Mél : richer@cppm.in2p3.fr

Le Centre de Physique des Particules de Marseille, unité mixte CNRS/Aix-Marseille Université, est un des laboratoires de l'Institut National de Physique Nucléaire et de Physique des Particules (IN2P3), institut du CNRS qui regroupe les moyens de la physique des particules. Le CPPM travaille notamment pour l'expérience LHCb (http://lhcb-public.web.cern.ch/lhcb-public/), installée sur le LHC, l'accélérateur de particules et collisionneur proton-proton le plus puissant du monde, au CERN à Genève (http://www.cern.ch). Cette expérience s'intéresse à la différence entre matière et anti-matière ainsi qu'à l'extensions du modèle standard de la physique.

Fort de l'expérience acquise lors de la conception et fabrication du système d'acquisition capable de traiter 30Tb/s via 10 000 liens optiques à 5Gb/s en temps réel, le CPPM s'intéresse à la future génération qui devra avoir une puissance de calcul 10 fois supérieure. Pour parvenir à cela, Intel nous confie en avant-première la dernière génération de FPGA Agilex doté de 4 millions d'éléments logiques et de transceiver capables de transmettre jusqu'à 112Gb/s. De tels liens sériels à très haut débit nécessitent notamment la maîtrise d'une nouvelle technique de modulation en amplitude du signal.

Activité principale :

Il s'agit d'œuvrer aux développements firmware en vue de tester l'implémentation de liens sériels à haute vitesse entre 10Gb/s NRZ et 112Gb/s PAM4 sur des kits de développement Intel. Ces développements de firmware et de software de contrôle permettront de préparer les tests sur la carte prototype du futur système d'acquisition actuellement en développement au CPPM. La caractérisation des liens sériels ainsi que l'étude des paramètres d'optimisations par technique de pré-accentuation et égalisation permettront de tester les limites des systèmes en jeu. Le travail proposé pourra être adapté selon les intérêts et les connaissances du ou de la candidat(e).

Le ou la stagiaire sera accueilli(e) au sein du service électronique du CPPM qui possède un savoir-faire étendu dans la programmation des FPGA et en conception de cartes à très haute densité.

Le travail s'effectuera dans un environnement de recherche international. Quelques déplacements au CERN (Genève) seront possibles en vue d'assister à des réunions de collaboration.

Connaissances appréciées :

Les compétences suivantes seront appréciées et une formation sur les outils utilisés sera donnée

• Utilisation d'appareils de mesure~: Oscilloscope, Analyseur de spectre

• Simulations électriques pour l'analyse d'intégrité de signal

• Transmission de signaux rapides

• Conception FPGA en langage VHDL

• Langage Python et éventuellement PyQt

Contact :

CV + lettre de motivation avec la référence « Ingenieur-2223-EL-05 » à

Frédéric HACHON

Ingénieur de Recherche - Correspondant stages techniques du CPPM

Tél : +33 4 91 82 76 71- Mél : hachon@cppm.in2p3.fr

Le stage de 6 mois sera conventionné et rémunéré.

Le stage se déroule au Centre de Physique des Particules de Marseille (CPPM). C'est est une unité mixte de recherche (UMR 7346) qui relève de l'IN2P3, institut regroupant les activités de physique des particules et de physique nucléaire au sein du CNRS et d'Aix-Marseille Université.

Le CPPM participe depuis plusieurs années au projet à l'expérience ATLAS du CERN à Genève, au sein d'une collaboration internationale de plus 3 000 scientifiques issus de 174 instituts, représentant pas moins de 38 pays.

L'un des grands challenges techniques au niveau de l'expérience ATLAS réside dans l'augmentation du nombre de données à acheminer depuis le détecteur vers les centres de calculateurs et les circuits intégrés spécifiques très haut débit et durcis contre les irradiations sont des éléments essentiels pour la transmission de ces données.

Le CPPM fait partie d'une collaboration R&D du CERN (RD53) pour le développement de la puce électronique de lecture des détecteurs à pixels pour les futures mises à niveau concernant l'ensemble des expériences du LHC. Cette puce a été développée avec le process CMOS 65 nm et comporte 160000 pixels de 50×50 µm2.

La haute luminosité anticipée dans le cadre des futurs upgrades du LHC ou pour les prochaines générations de collisionneurs se manifeste par un très fort taux de hits par unité de surface au niveau des différents détecteurs et particulièrement dans le cas des détecteurs à pixels proches du point d'interaction. Dans ce contexte, une bonne précision spatiale qui passe par la réduction de la taille des pixels s'avère indispensable. L'ajout d'informations temporelles de haute précision à la précision spatiale existante (4D tracking), est une caractéristique intéressante qui pourrait ouvrir la voie à de nouvelles améliorations de la reconstruction des traces.

Activité principale :

Le but du stage est de proposer une architecture de l'étage frontal analogique de détection de charge composé d'un l'amplificateur de charge et d'un discriminateur. L'objectif est d'atteindre une résolution temporelle de 50 ps tout en maintenant une consommation réduite et un niveau de bruit électronique compatible avec les exigences du seuil d'énergie requis dans le détecteur de traces.

Le stage de 6 mois sera organisé en plusieurs étapes :

• Etude du système de détection de traces actuel utilisé pour l'expérience ATLAS

• Etude, conception et optimisation du circuit amplificateur de charge actuel avec le process CMOS 28 nm

• Simulation et optimisation du cricuit sous Cadence Virtuoso

• Dessin des masques sous Cadence

Connaissances requises :

• Bonnes connaissances en conception de circuits analogiques CMOS

• Le développement de bancs de test basés sur des composants programmables de type FPGA est considéré comme un avantage

Contact : CV + lettre de motivation avec la référence « Ingenieur-2223-EL-02 » à

Frédéric HACHON

Ingénieur de Recherche - Correspondant stages techniques du CPPM

Tél : +33 4 91 82 76 71- Mél : hachon@cppm.in2p3.fr

Le stage de 6 mois sera conventionné et rémunéré.

BelleII est un détecteur polyvalent du collisionneur SuperKEKB au Japon. Il a été conçu et construit afin de tester de nouveaux modèles de physique et rechercher les signatures de nouvelles particules. Le détecteur BelleII est un détecteur de particules qui mesure 7,5 m de long, 7 m de haut. Il est composé principalement d'un détecteur de vertex et d'un calorimètre. En augmentant la luminosité de l'accélérateur de particule SuperKEKB, le sous-ensemble Vertex de BelleII devra aussi évoluer et être mis à jour d'ici 2026.

Une collaboration internationale s'est ainsi structurée afin de réfléchir et concevoir cette jouvence du détecteur.

Au CPPM, un groupe d'une dizaine de physiciens, ingénieurs et techniciens est impliqué dans le projet BelleII et s'intéresse en particulier à l'évolution du détecteur de vertex (VXD), détecteur interne le plus proche du point d'interaction. Ce détecteur de traces (trajectographe) est destiné à suivre le passage des particules dès leur formation.

La brique élémentaire du trajectographe est un circuit intégré spécifique (ASIC) matriciel de plusieurs millions de transistors. Ce circuit opère comme un appareil photo à pixels, qui doit prendre une image de la détection des particules. Plusieurs contraintes de conception sont imposées sur l'électronique, comme la surface, la rapidité, la consommation et la précision. De plus, afin de fonctionner en toute autonomie, le circuit a besoin de fonctions générales, comme un «~bandgap reference~», un capteur de température, un buffer analogique et son ADC, des circuits numériques de décisions et mémoires, ou encore un système de distribution des alimentations ou polarisations des étages. Des étages d'entrée/ sortie à hautes vitesses comme les standards LVDS ou CML seront aussi intégrés.

Activité principale :

Dans un premier temps, le/la stagiaire doit mener une recherche bibliographique détaillée sur le circuit servant de référence au projet (TJ-MONOPIX2) ainsi que sur les détecteurs à pixels monolithiques et sur les fonctions générales. Ensuite, il lui sera proposé d'étudier et concevoir une des fonctions qui soit le mieux adaptée à l'application selon le cahier des charges fourni.

En fonction de l'avancement du projet, le/la stagiaire aidera l'équipe de conception à finaliser le circuit prototype OBELIX, pour une fabrication courant 2023.

• Etude bibliographique sur les architectures de la fonction.

• Conception, simulation sous Cadence

• Dessin des masques (Layout)

• Simulation post-layout

• Des tests sur d'anciens circuits sont à prévoir.

Connaissances requises / apréciées :

• Bonnes connaissances en conception de circuits intégrés en technologie CMOS

• Connaissance dans la manipulation d'instruments de mesure

Contact : CV + lettre de motivation avec la référence « Ingenieur-2223-EL-01 » à

Frédéric HACHON

Ingénieur de Recherche - Correspondant stages techniques du CPPM

Tél : +33 4 91 82 76 71- Mél : hachon@cppm.in2p3.fr

Le stage de 6 mois sera conventionné et rémunéré.

Le Centre de Physique des Particules de Marseille, unité mixte CNRS/Aix-MarseilleUniversité, (http://marwww.in2p3.fr) est un des laboratoires de l'Institut National de Physique Nucléaire et de Physique des Particules (IN2P3), institut du CNRS qui regroupe les moyens de la physique des particules.

Le CPPM travaille notamment sur des systèmes d'acquisition sur le LHC, l'accélérateur de particules et collisionneur proton-proton le plus puissant du monde, au CERN à Genève (http://www.cern.ch).

Le nombre de collisions va dans un avenir proche être multiplié par 10, rendant difficile l'identification des particules générées tant elles seront nombreuses.Une piste possible est d'utiliser des algorithmes neuronaux au plus près du détecteur pour trier et identifier les particules générées ainsi que les phénomènes recherchés.

Le challenge est que le LHC génère 40 millions de collisions par seconde, chacune d'entre elles «~illuminant~» des dizaines, voire des centaines de milliers de capteurs. Il est donc nécessaire d'implémenter ces algorithmes «~au vol~» sur des FPGAs très puissants.L'objet du projet THINK (Testing Hardware Instantiations of Neural Kernels) est d'évaluer la capacités de FPGAs ou de circuits spécialisés tels que des chips neuromorphiques à traiter ce type de données en temps réel.

Activité principale :

L'évaluation consistera à hiérarchiser les performances de différents types de FPGAs ou chips neuromorphiques en implémentant plusieurs benchmarks communs. L'évaluation portera non seulement sur les performances mais aussi sur la qualité des outils de mise en œuvre, notamment leur facilité d'emploi ou leur versatilité.

Deux types de FPGAs relativement différents seront étudiés~:

• Le Stratix NX d'Intel doté d'AI Tensor Blocks répartis dans le FPGA

• Le Versal AI de Xilinx doté d'un processeur scalaire avec accélérateur de fonctions AISi le temps le permet un chip neuromorphique ou un GPU sera également étudié.

Connaissances appréciées :

• Conception FPGA en langage VHDL, HLS

• Langage Python

• TensorFlow, Queras, Pytorch

Contact : CV + lettre de motivation avec la référence « Ingenieur-2223-EL-04 » à

Frédéric HACHON

Ingénieur de Recherche - Correspondant stages techniques du CPPM

Tél : +33 4 91 82 76 71- Mél : hachon@cppm.in2p3.fr

Le stage de 6 mois sera conventionné et rémunéré.

ATLAS («~A Toroidal LHC ApparatuS») est une expérience de physique des particules installée sur le LHC («~Large Hadron Collider~») au CERN («~Centre Européen pour la Recherche Nucléaire~») situé à Genève. Elle a été conçue pour tester de nouveaux modèles de physique et rechercher les signatures de nouvelles particules, telles que le boson de Higgs découverte expérimentalement en 2012.

En prévision d'une jouvence complète du détecteur à pixel de l'expérience, une collaboration internationale, RD53 a été mise en place pour développer le prochain circuit de lecture associé au détecteur en technologie CMOS~65~nm. Le CPPM fait partie de cette collaboration et a en charge plusieurs cellules implantées dans le circuit global comme un ADC de type SAR permettant la numérisation des informations provenant de références de tension, dosimètres, et capteurs de température. Il a également la responsabilité de la conception de mémoires tolérantes au SEU («~Single Event Upset~») et exerce une activité dans le groupe «~Radiation Tolerance~» puisque l'ASIC de lecture devra fonctionner dans un environnement très radioactif, supportant une dose totale de 500~Mrad (5~MGray) pendant 5 ans d'exploitation.

Activité principale :

Plusieurs prototypes de circuits intégrés (CI) ont été conçus en différentes technologies, 65~nm, 28~nm, etc… et testés sur table ainsi qu'en irradiation au CERN. Les tests de ces CI permettent la validation de leur architecture auprès de la collaboration où les résultats sont présentés. Plusieurs bancs de tests de CI prototypes a été développé au CPPM. Cela peut représenter le pilotage d'appareillages de précision (type Keithley) ou encore la gestion d'une électronique embarquée à base d'une carte du commerce nanoPC de type BeagleBone, communicant avec un FPGA (Altera-Cyclone III) via un bus parallèle de type GPMC (General-Purpose Memory Controller). Les séquences de tests sont préalablement implantées dans le FPGA (programmation VHDL), Le contrôle-commande s'effectue au niveau de la carte BeagleBone en C++. D'autres paramètres tels que, la consommation, la température, les niveaux d'alimentation, sont enregistrés via un bus I2C.

Ces éléments sont resuiq pour s'assurer du bon fonctionnement du circuit.

Le stage de 6 mois devra comporter plusieurs étapes~:

• Prise en main du banc de test

• Maitrise, débogage des différentes fonctions du banc de test

• Amélioration et finalisation de l'ensemble, paramétrage intuitif et convivial via une interface utilisateur de type Qt Python.

Connaissances requises :

• Bonnes bases en électronique

• Solides connaissances en programmation LabView, VHDL, C++, Qt Python

Contact : CV + lettre de motivation avec la référence « Ingenieur-2223-EL-03 » à

Frédéric HACHON

Ingénieur de Recherche - Correspondant stages techniques du CPPM

Tél : +33 4 91 82 76 71- Mél : hachon@cppm.in2p3.fr

Le stage de 6 mois sera conventionné et rémunéré.

Cagire est une caméra sensible dans le proche infrarouge, qui sera placée au foyer du télescope GFT Colibrí installé au Mexique pour faire l'imagerie et la photométrie des régions du ciel où apparaissent des sources transitoires dignes d'intérêt qui seront détectées par l'imageur spatial Eclairs à bord du satellite franco- chinois SVOM. Lorsqu'il reçoit une alerte, l'objectif principal du télescope Colibrí est la recherche rapide d'une contrepartie visible ou proche infrarouge, qui se manifeste comme l'apparition d'une nouvelle étoile, qui peut être identifiée tant par son absence des catalogues d'objets connus que par ses variations rapides de luminosité.

\bf{Activité principale :}

Projet soutenu et financé par le CNES, la caméra Cagire sera équipée du détecteur Alfa, le premier détecteur scientifique bas bruit grand format européen développé par la société Lynred, et premier qui sera mis sur le ciel. Au sein du projet, le CPPM est responsable de la caractérisation des performances scientifiques du détecteur. Cette caractérisation fine sera réalisée dans la salle propre du laboratoire en collaboration avec le laboratoire IRAP, maître d'oeuvre de la caméra et le CEA qui réalise la sélection du meilleur détecteur pour le projet.

Afin de fournir au projet les produits de calibration qui seront utilisés par le pipeline de traitement des données de la caméra CAGIRE, l'ingénieure/l'ingénieur-stagiaire devra :

1) Participer aux tests qui auront lieu au CPPM au premier semestre 2023 incluant

? la prise en main du banc et la participation aux shifts d'acquisition,

? la mise en place de codes de vérification de qualité des données,

? la définition et la mise en œuvre d'un plan de test dédié au phénomène de persistance

2) Adapter des codes d'analyse existants dans l'équipe afin d'extraire les paramètres principaux du détecteur (courant d'obscurité, bruit de lecture, linéarité, gain de conversion etc.)

3) Participer aux réunions de projet pour exposer son travail

Connaissances requises :

• Base technique solide en instrumentation

• Base solide en programmation (langage python)

• Bonnes connaissances en traitement du signal

Contact : CV + lettre de motivation avec la référence « CAGIRE » à Aurélia Secroun, Ingénieure Chercheure

Le stage de 6 mois sera conventionné et rémunéré.

L'astronomie des neutrinos et activités pluridisciplinaires dans le cadre du projet KM3-NeT / Numerenv. Les télescopes à neutrino ont pour but d'étudier les neutrinos cosmiques de haute énergie avec un réseau de photo-détecteurs installé au fond de la mer.

Pour alimenter ces lignes de détections, des connecteurs hybrides sont utilisées et la connexion est faite par des petits robots sous-marins.

Nous nous inspirons d un outillage de démultiplication utilisé au laboratoire en augmentant la démultiplication.

L'élève travaillera en étroite collaboration avec un ingénieur. Des études en CAO seront confiées ainsi que du suivi de sous traitance .

The Marseille Particle Physics Center is a mixed research unit (UMR 7346) dependent on the CNRS and Aix-Marseille University, which deploys its research activities both in the field of fundamental physics and also for applications based on ionizing radiation.

Satiety or addiction circuits are driven in the brain by negative or positive feedback loops using neurotransmitters. These circuits can be imaged by positron emission tomography (PET) thanks to the labeling of neurotransmitters by positron-emitting radioactive ions, such as, for example, cocaine labeled with 11C.

However, PET scans require the subject to be anesthetized, which does not capture the actual behavior of the brain under waking conditions.

The CPPM is participating in the MAPSSIC project, which consists of developing an intracranial CMOS pixel probe for positron imaging in vigilant and free-moving rats. The IMIC probe, which forms a needle of several hundred active CMOS pixels, was developed by the IPHC in Strasbourg to be permanently implanted in the brain of a rat which, equipped with a backpack including a battery and a wireless transmitter linked to the CMOS pixels, will make it possible to directly image the positrons emitted during the disintegration of the nuclei of a radioactive tracer attached to the molecules of the neurotransmitter studied.

Main activity :

The trainee will be integrated into the MAPSSIC project will participate in the design study and the implementation of a wireless solution to ensure the monitoring and control and transmission of data collected simultaneously by 4 probes IMIC to an acquisition PC.

This wireless solution must be carried in a backpack adapted to the build of a rat and be able to achieve an autonomy of several hours corresponding to several periods of decay of the radioactive tracer used to label the neurotransmitter.

Required profile :

• Practice of C / C ++ language

• Embedded systems programming (µC) in C / C ++

• Knowledge of python is a plus

The 6-month internship will be remunerated.