Your favourite hot Jupiter in detail: HD189733b from the NUV to the MIR

JWST is revolutionizing our understanding of exoplanet atmospheres. In particular, MIRI gives us access to the 10μm feature from silicate clouds that have long been theorized (e.g. Helling+16) and recently observed with JWST (e.g. Dyrèk+23). Clouds can drastically alter the gas composition of the atmospheres (e.g. Helling+19). It is crucial to understand them to derive reliable elemental ratios and metallicities for the atmospheres. In this work, we present a detailed characterisation of the hot Jupiter HD 189733 b from the NUV to the MIR combining observations from HST, Spitzer and JWST. We present the detection of silicate clouds in its atmosphere with JWST/MIRI. The simultaneous presence of strong water absorption bands in the HST/WFC3 spectrum (naively indicative of a mostly cloud free atmosphere) makes the interpretation of this set of observations a puzzle. Moreover, the spectrum shows a pronounced slope in the optical hinting at the presence of hazes in the atmosphere. We explore whether a mixing regime exists which reproduces the strong cloud and molecular features observed in the spectrum, or whether these features are produced in different terminators (e.g. having clear evenings and cloudy mornings). We run retrievals with complex models including microphysical clouds. The temperature structure, the chemistry, and the clouds are self-consistent with each other (Ormel+19). We assume equilibrium chemistry and consider disequilibrium chemistry due to vertical mixing. We also include a simple prescription for photochemistry to produce high altitude hazes that reproduce the slope observed in the optical. This is the first time such a retrieval has been run, owing to it being computationally unfeasible with traditional methods (e.g. nested sampling or MCMC). We use a novel machine learning retrieval tool (FlopPITy, Ardévol Martínez+23) that can speed up retrievals up to ~20x. In this talk we also briefly introduce this tool. This work showcases the possibilities that machine learning opens and sets the stage for the detailed characterisation of exoplanet atmospheres using complex self-consistent models.