https://www.spintec.fr/wp-content/uploads/2025/10/Plaquette_Spintec_2026.pdf
Context
In recent years, magnetic tunnel junctions (MTJs) have been commercialized as part of nonvolatile memory elements, and are now envisioned as stochastic neurons for unconventional and biosinspired computing schemes. MTJs exhibit two metastable states separated by an energy barrier, depending on the relative orientations of the free and fixed layers. At room temperature, thermal fluctuations can randomly flip the magnetization of the free layer, with mean dwell times which can generally be described by the Arrhenius law: τ= τ0 eΔE/kT, where ΔE is an energy barrier, kT is the thermal energy, and τ0 is a prefactor. While it is common practice in the magnetic community to consider τ0 a characteristic timescale of the dynamics of few nanoseconds, in reality, this prefactor also contains a large entropic contribution related to the number of pathways to the transition state [Desplat & Kim, Phys. Rev. Lett. 125, 107201 (2020)].
These considerations apply at thermal equilibrium. Under negligible currents, we have successfully explained femtosecond mean dwell times measured in 50 nm perpendicularly magnetized MTJs [Soumah, Desplat, et al. Phys. Rev. Appl. 14, L011002 (2025)]. However, to control the magnetization or enable interjunction coupling for computing applications, non-negligible electric currents must typically be applied to these systems.
Work program & Skills acquired during internship
The aim of this internship is to go beyond the current state of the art, and compute mean dwell times in MTJs under applied currents. To achieve this, developments and simulations will be carried out in the magnum.np micromagnetics framework based on the PyTorch library [Bruckner et al., Sci. Rep. 13, 12054 (2023)]. This will be done in collaboration with the magnum.np developers in the group of Dieter Suess at the University of Vienna. During the internship, the student will:
• Familiarize themselves with key micromagnetics concepts, the Hamiltonian, the dynamics including with temperature, and basic concepts of MTJs;
• Learn the basics of computing transition rates: energy barrier computations, equilibrium prefactors, dwell time computation;
• Learn to develop python scripts to run magnum.np;
• Implement a path sampling scheme in magnum.np to compute dwell times under constant electric currents;
• If time allows: explore mean dwell times under time-dependent currents.
- Requested background: Master 2
- Duration: 6 months
- Start period: Feb/ March 2026
- Possibility of PhD thesis : Yes