PhD Position : Cold atoms for optical reservoir computing

COFUND QuanG; France; Sur place; EUR 25 000 - 30 000; Plein temps

Résumé du poste

COFUND QuanG Grenoble propose un doctorat dans le cadre du programme QuanG2, sur l’étude théorique du reservoir computing avec des atomes froids, dans un modèle tridimensionnel réaliste et avec la polarisation de la lumière. Le candidat idéal possède un Master en physique et des compétences en physique théorique et programmation scientifique.

Ce poste, financé sur trois ans, implique des travaux théoriques et numériques, des échanges avec les partenaires et le respect des règles MSCA.

Qualifications

• Maîtrise en physique et compétences en physique théorique et calculs scientifiques.
• Expérience en programmation pour applications scientifiques est un plus.
• Conformité à la règle MSCA Mobility : ne pas avoir résidé en France plus de 12 mois au cours des 36 derniers mois.
• Les candidats ne doivent pas déjà détenir de doctorat.

Formation

Master’s degree or equivalent

Description du poste

Organisation/Company COFUND QuanG Research Field Physics Researcher Profile First Stage Researcher (R1) Positions PhD Positions Application Deadline 7 Sep 2026 - 12:00 (Europe/Paris) Country France Type of Contract Temporary Job Status Full-time Offer Starting Date 1 Oct 2026 Is the job funded through the EU Research Framework Programme? Horizon Europe – COFUND Reference Number 101261699 Is the Job related to staff position within a Research Infrastructure? No

Offer Description

General Scope:
The concept of reservoir computing arose from the observation that the performance of a large neural network is only weakly degraded when only a part of the network is subject to training instead of all of it. The “passive”, untrained part of the network (termed a “reservoir”) remains useful despite the fact that its response is completely random and not optimized to perform a particular task. Neural networks are often implemented numerically on a computer, but their true power unveils in their physical realizations that often allow for massive parallelization, speed up of operation, and reduction of power consumption. A possible physical realization of a neural network is an ensemble of cold (immobile) two-level atoms playing the role of nodes and interacting via the electromagnetic field (light). Whereas the response of atoms to an external excitation can easily be made nonlinear by increasing the magnitude of excitation—a property
needed to perform nontrivial tasks,—interactions between them cannot be adjusted at will because they are controlled by physical laws (Maxwell equations). When supplemented with an additional layer of “standard” trainable neurons simulated on a computer, such a system represents a hybrid realization of an optical reservoir processor, with atoms constituting its passive part (reservoir). Because atoms are quantum objects and the light by which they are excited and which they emit can be in nonclassical states (Fock, squeezed or entangled states, for example), the cold-atom reservoir processor described above is also suitable for performing quantum tasks and thus qualifies for a “quantum optical reservoir processor”.

PhD Subject:
This thesis project aims at a theoretical study of optical reservoir computing with cold atoms within a realistic three-dimensional model and with proper account for the polarization of light. The idea is to consider an ensemble of N ~ 10 closely located, immobile atoms in the regime of intermediate saturation, in which the richest system dynamics is expected. The input signal is fed to the system by illuminating a small fraction of the atoms (the input “layer” of the network) by a temporally and/or spatially modulated laser. The resulting complex dynamics of the atomic system constitutes the “processing” of the input signal, and the result is read out from a small number of atoms that are chosen to form the output “layer” of the reservoir. A weighted linear combination of signals generated by the reservoir is the output of the reservoir processor, with weights adjusted to optimal values during the training stage to perform a pre-defined task.
Additionally, the system can be controlled by an external strong laser applied to all or some of the atoms. In the simplest, “classical” setting, both the input and the output can be assumed to be the intensities or electric fields of the incident and emitted light, respectively. In the full quantum setting, we will assume to have access to quantum states of the atoms and/or the incident and emitted photons.
Once the theoretical model is built and the numerical model implemented, we will use the latter to explore a number of interesting and nontrivial questions. First, the performance of the system in standard classical (e.g., prediction of chaotic processes) and quantum (e.g., quantifying entanglement of incident light) tasks will be analyzed and the optimal atomic configurations determined. Next, we will study the statistical properties of a random reservoir processor—a processor in which the precise spatial configuration of atoms is not controlled and the atoms are assumed to be distributed randomly in space within a certain volume. This case is conceptually interesting because it should allow to compare the performance of a “typical” random system with the optimal one. It is also of practical importance because precisely controlling positions of several atoms in a small volume may be challenging in an experiment. Finally, we plan to explore the possibility of extending the results obtained for relatively small atomic ensembles (up to N ~ 10 atoms) to larger, “macroscopic” ensembles, where it is not possible to follow the exact dynamics of every atom, but where the statistical properties of the collective atomic response can be estimated in the thermodynamic limit (limit of infinite N).

Required Skills:

• Knowledge and ability to use methods of theoretical physics
• Some experience in scientific programming
• Basics of quantum optics, quantum information, and cold-atom physics
• Communication skills
• Ability to work in a team

Applicants must hold a Master’s degree or an equivalent qualification by the application deadline and must not already hold a doctoral degree. Applicants must also comply with the MSCA mobility rule: they must not have resided or carried out their main activity (work, studies, etc.) in France for more than 12 months during the 36 months immediately preceding the application deadline. Applicants must not be current employees of the host laboratory. There are no nationality or age restrictions.

Selection process

Applications must be submitted through the QuanG2 online application platform by 7 September 2026 at 12:00 PM (Paris time). After the application deadline, all applications will first undergo an eligibility check. Eligible applications will then be reviewed during the pre-selection phase, scheduled for mid-October 2026. Shortlisted candidates will be invited to online interviews at the end of October 2026. Candidates selected following this first interview stage will then be invited to in-person interviews in Grenoble in early December 2026, with the final selection taking place after these interviews. All candidates will be informed of the outcome of the selection process following the final stage.

Additional comments

About the QuanG2 PhD Call

This PhD position is offered as part of the QuanG2 PhD Call for Applications, a doctoral programme coordinated by Université Grenoble Alpes and dedicated to training the next generation of researchers in quantum science and technology. The programme offers fully funded three-year PhD positions within the Grenoble quantum research ecosystem, providing doctoral candidates with a high-level international research environment and dedicated funding for their research and training activities.

[e-mail]