Recent advances in optimisation techniques have opened a promising path towards computationally exploring the vast design space of new gravitational wave detectors. A key bottleneck in such AI-driven design is the need for repeated numerical gradient evaluations, which makes each simulation computationally demanding.
This project proposes to use our newly developed differentiable interferometer simulator, Differometor, to perform large-scale optimisations that will enable the digital discovery of novel gravitational wave detector designs. Our simulator exploits high-performance computing techniques including hardware acceleration, automatic differentiation, vectorisation, and just-in-time compilation.
By leveraging large-scale HPC resources, this project aims to go beyond designs based on human intuition and discover a new class of tunable gravitational wave detectors with enhanced sensitivity.
Mario Krenn, Eberhard Karls Universität Tübingen, Germany