Inverse design of an optical power splitter using a global optimization method

Contexte et atouts du posteThis internship project is part of a collaborative project between the Atlantis project-team from Inria Research Center at Université Côte d’Azur, Sophia Antipolis and Ansys Lumerical R&D, La Farlède.Atlantis is a team from Inria Research Center at Université Côte d’Azur located in Sophia Antipolis. It gathers researchers in numerical mathematics and computational physics, with an interdisciplinariy focus. The team has developed a specific expertise in the efficient numerical modeling of propagation of electromagnetic wave in complex media with a strong emphasis on nanoscale light-matter interactions. Through the years, the Atlantis team has developed a strong expertise in the design, analysis and development of dedicated efficient numerical methods (based on high order accurate Discontinuous Galerkin finite elements methods). More recently, the team has also acquired a know-how of numerical optimization using various techniques, and a solid experience on high performance computing practices (parallel numerical algorithms and parallelization strategies for large-scale problems). This materializes concretely through the DIOGENeS software suite [1] that has already proven its crucial efficiency in nanophotonics. DIOGENeS will be the corner stone to numerically address the various complex scenarios in this internship project.Mission confiéePhotonic Integrated Circuits (PICs) are considered among the most promising approaches to ensure progress in computer technology. Their possible applications range from improving the performance and efficiency of modern AI accelerators to realizing solid state LIDAR detectors for autonomous driving. However, the design of large-scale PICs still lags the design process of electronic circuits and presents unique challenges.The goal of our research is to improve photonic components by reducing their footprint, enhancing their performance, and improving their robustness against manufacturing variations. To do so, we develop and employ advanced simulation and optimization methods, such as inverse design and topology optimization.Principales activitésThis internship project aims at realizing a preliminary (and preparatory) study prior to a more ambitious project that will be undertaken in the context of a Cifre PhD thesis between Ansys R&D and Inria. The internship will take place in the Atlantis project-team from Inria Center at Université Côte d’Azur in Sophia Antipolis. The overall objective will be to develop and assess an inverse design strategy leveraging a modern global optimization algorithm. Specifically, we want to implement the global optimization of a PIC device by combining the following approaches:A DGTD (Discontinuous Galerkin Time-Domain) Maxwell solver [1] to simulate a device and compute a specified figure-of-merit (FoM) for a given design;The EGO (Efficient Global Optimization) method, which is a statistical learning-based global optimization algorithm [2] belonging to the family of Bayesian optimization methods [3];A shape parameterization technique, which is compliant with the fact that the DGTD method makes use of an unstructured tetrahedral mesh for the simulations.As an initial use case, we will optimize a symmetric Y-branch (50:50 optical splitter). This basic and well-studied device [4] requires a relatively low number of parameters (< 20) and is therefore well-suited for global optimization.The individual software components mentioned above are already available in the DIOGENeS software suite developed by the Atlantis project-team. The internship focusses on the geometrical modeling and parametrization of the PIC device as well as the integration of the individual components. The results of the internship will be integrated in DIOGENeS.[1] S. Lanteri, C. Scheid and J. Viquerat. Analysis of a generalized dispersive model coupled to a DGTD method with application to nanophotonics. SIAM Journal on Scientific Computing, Vol. 39, No. 3, pp. A831–A859[2] D. Jones. Efficient global optimization of expensive black-box functions. Journal of Global Optimization, Vol. 13, No. 4, pp. 455-492[3] R. Garnett. Bayesian Optimization. Cambridge University Press[4]CompétencesTechnical skills and level required :Master or engineering degree in numerical mathematics or scientific computingSound knowledge of numerical analysis for PDEsBasic knowledge of physiscs of electromagnetic wave propagationSoftware development skills : Python and Fortran 2003Relational skills : team worker (verbal communication, active listening, motivation and commitment)Other valued appreciated : good level of spoken and written englishAvantagesSubsidized mealsPartial reimbursement of public transport costsLeave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)Possibility of teleworking and flexible organization of working hoursProfessional equipment available (videoconferencing, loan of computer equipment, etc.)Social, cultural and sports events and activitiesAccess to vocational trainingContribution to mutual insurance (subject to conditions)RémunérationTraineeship grant depending on attendance hours #J-18808-Ljbffr