In situ particle measurements are fundamental to characterizing and understanding plasma processes and associated risks in heliophysics, planetology, and space weather. The necessary measurements, performed by satellite-borne instruments, concern ions, electrons, and neutral atoms over a wide energy range, from eV to MeV. These measurements often require instruments to be equipped with wide fields of view, for example through the use of multiple identical detectors or measuring heads.These instruments are calibrated on the ground before launch and then throughout the mission. 

The increasing complexity of instruments and missions, the need for greater precision for reliable measurements over long periods, and the limitations of traditional methods— time-consuming manual calibration that is difficult to reproduce and subject to human bias or in-flight degradation—are leading to the development of predictive models based on ground and in-flight calibration data.As part of the ESA M-MATISSE mission, IRAP is developing the SP@M (Solar Particles at Mars) instrument, a set of identical, multiple medium/high-energy particle detectors.

The SP@M instrument comprises 16 detection heads based on silicon detectors. The large number of detectors will probably not allow for a complete particle accelerator calibration of all detectors. Furthermore, in-flight calibration will be very limited. A complete prototype of the SP@M instrument was developed at IRAP as part of the M-MATISSE Phase A, which was completed in early 2026.

The objective of this thesis is to prepare the calibration of multiple measurement heads for the detection of charged particles using the example of the SP@M instrument by developing a machine learning model that can predict the response of the detector in flight based on ground calibration data and generalize the calibration to conditions that have not been tested experimentally.This model will enable systematic instrumental effects (temperature, electronic noise, radiation degradation, etc.) to be taken into account, as well as the effects of background noise caused by cosmic particles. It will also enable intercalibration in terms of energy and efficiency between the various detectors and can be coupled, if necessary, with physical models for simulating the detectors. 

This model will be used throughout the instrument's lifetime to adjust instrumental parameters (detector bias, etc.) by integrating in-flight calibration dataThese machine learning methods can be applied to the calibration of various space instruments, such as low-energy plasma analyzers, mass spectrometers, and gamma/X-ray detectors, as well as to intercalibration between low-energy and high-energy detectors. 

They could also be applied to multi-sensor observatories, such as constellations of small satellites.The tasks to be performed are:

• Developing benchmarks to compare the performance of different machine learning algorithms

• Participating in the calibration of prototype detectors and qualification models with radioactive sources and particle accelerators

• Developing a machine learning model and training it with calibration data.

• Integration of explainability mechanisms to ensure confidence in predictions.

=================

For more Information about the topics and the co-financial partner (found by the lab!);contact Directeur de thèse - pierre.devoto@irap.omp.eu or direction@irap.omp.eu

Then, prepare a resume, a recent transcript and a reference letter from your M2 supervisor/ engineering school director and you will be ready to apply online  before March 13th, 2026 Midnight Paris time!