PhD student in Multimodal Machine Learning for Energy Storage Materials

Descrizione

Testo originale dell'annuncio (in inglese)

The Department of Chemistry - Ångström conducts research and education in the chemistry field. The department has more than 270 employees and has a turnover of 300 million SEK. At the department's six programs, we conduct very successful research of a high international standard. We have a large number of externally funded research projects, often with international cooperation and we see continued good growth in our subject area. The department has education assignments in engineering programs and master's programs. More information is available on our website . Read more about our benefits and what it is like to work at Uppsala University Project description Progress in the energy storage materials of the future rests on a fundamental understanding of how their chemical composition and structure evolve during operation in the applications. No single approach can capture the whole picture. Instead, researchers rely on several complementary techniques, each providing an incomplete and indirect view of the same underlying process, traditionally interpreted manually and in isolation. Multimodal machine learning offers a principled alternative: integrating heterogeneous experimental and simulation data streams into a single, physically grounded model. In focus of the project is the development of such a framework and its specific application to electrode/electrolyte interphases in batteries, where surface-sensitive X-ray scattering techniques (X-ray reflectivity, GISAXS and GIWAXS) and electrochemical quartz crystal microbalance with dissipation monitoring (EQCM-D) are combined in operando experiments. At the core of the methodology is data assimilation, in which the underlying properties of the interphase (mass, volume, density, porosity and morphology) are treated as a latent state evolving in time, with each technique contributing a noisy, partial observation through its own forward model. The work spans sequential Bayesian inference, deep state-space models, and uncertainty-aware representations trained on both experimental data and simulations, such as phase-field simulations of metal nucleation and growth coupled to lattice-Boltzmann modelling of the acoustic quartz crystal response. Machine-learning-accelerated surrogate models make these simulations fast enough to serve as transition and observation operators within the inference engine. Finally, the learned latent space is analysed to extract physical meaning, and the generative capability of the models is exploited to translate data between techniques. The goal is an interpretable, uncertainty-quantified model of interphase formation and growth that deepens our understanding of metal plating and surface passivation in Li, Zn and Cu based systems, and that is transferable to other materials and analytical techniques. The doctoral candidate will develop and implement this framework in close collaboration between chemistry and computer science at Uppsala University and together with our German partners. The work includes building and validating inference and simulation pipelines, actively participating in the planning and execution of operando experiments – including beamtimes at the PETRA III synchrotron (DESY, Hamburg) – and gaining in-depth expertise in probabilistic machine learning, data assimilation, scientific computing and the analysis of electrochemical interphases. The core focus of the PhD program is research. Teaching responsibilities may be assigned but will not exceed 20% of the total working hours. The candidate will also complete third-cycle coursework closely aligned with the research themes. We are looking for a highly motivated individual who will contribute to high-quality scientific work within a team setting, aiming to publish findings in leading peer-reviewed journals and present them at both national and international scientific meetings. Requirements The candidate must have: been awarded a second-cycle qualification, or have completed at least 240