Deep learning algorithms are growing in popularity in the field of exoplanetary science due to their ability to model highly non-linear relations and solve interesting problems in a data-driven manner. Several works have attempted to perform fast retrievals of atmospheric parameters with the use of machine learning algorithms like deep neural networks (DNNs). Yet, despite their high predictive power, DNNs are also infamous for being 'black boxes'. It is their apparent lack of explainability that makes the astrophysics community reluctant to adopt them. What are their predictions based on? How confident should we be in them? When are they wrong and how wrong can they be? In this work, we present a number of general evaluation methodologies that can be applied to any trained model and answer questions like these. In particular, we train three different popular DNN architectures to retrieve atmospheric parameters from exoplanet spectra and show that all three achieve good predictive performance. We then present an extensive analysis of the predictions of DNNs, which can inform us - among other things - of the credibility limits for atmospheric parameters for a given instrument and model. Finally, we perform a perturbation-based sensitivity analysis to identify to which features of the spectrum the outcome of the retrieval is most sensitive. We conclude that for different molecules, the wavelength ranges to which the DNN's predictions are most sensitive, indeed coincide with their characteristic absorption regions. The methodologies presented in this work help to improve the evaluation of DNNs and to grant interpretability to their predictions.
We present a theory for interpreting the sodium lines detected in transmission spectra of exoplanetary atmospheres. Previous analyses employed the isothermal approximation and dealt only with the transit radius. By recognising the absorption depth and the transit radius as being independent observables, we develop a theory for jointly interpreting both quantities, which allows us to infer the temperatures and number densities associated with the sodium lines. We are able to treat a non-isothermal situation with a constant temperature gradient. Our novel diagnostics take the form of simple-to-u…
Constraining planet formation based on the atmospheric composition of exoplanets is a fundamental goal of the exoplanet community. Existing studies commonly try to constrain atmospheric abundances, or to analyze what abundance patterns a given description of planet formation predicts. However, there is also a pressing need to develop methodologies that investigate how to transform atmospheric compositions into planetary formation inferences. In this study we summarize the complexities and uncertainties of state-of-the-art planet formation models and how they influence planetary atmospheric com…
One of the desired outcomes of studying exoplanetary atmospheres is to set decisive constraints on exoplanet formation theories. Astronomers often speak of the "metallicity" in broad terms. When modeling the bulk metallicity, workers refer to the elemental iron abundance (Fe/H). By contrast, when exo-atmospheric astronomers speak of the "metallicity" derived from analysing low-resolution Hubble and Spitzer spectrophotometry, they are referring to the elemental abundances of oxygen (O/H), carbon (C/H) and nitrogen (N/H): in decreasing order of importance, since spectra from the Hubble Wide Fiel…
We present a new method for performing atmospheric retrieval on ground-based, high-resolution data of exoplanets. Our method combines cross-correlation functions with a random forest, a supervised machine learning technique, to overcome challenges associated with high-resolution data. A series of cross-correlation functions are concatenated to give a "CCF-sequence" for each model atmosphere, which reduces the dimensionality by a factor of ~100. The random forest, trained on our grid of ~65,000 models, provides a likelihood-free method of retrieval. The pre-computed grid spans 31 values of both…
Constraining the lives and times of exoplanets through evolutionary Bayesian retrievals
Institute of Astronomy, University of Cambridge（剑桥大学天文学研究所）；Kapteyn Astronomical Institute, University of Groningen（格罗宁根大学卡普坦天文研究所）；Helmholtz AI, Technical University of Munich（慕尼黑工业大学赫尔姆霍兹人工智能）；University of Technology Nuremberg（纽伦堡技术大学）
The Effects of M Star Age Dependent Ultraviolet Emission on Detecting and Interpreting Exoplanet Biosignatures
University of California Santa Cruz（加州大学圣克鲁兹分校）；SETI Institute（SETI研究所）；University of Washington（华盛顿大学）；NASA NExSS Virtual Planetary Laboratory（NASA NExSS虚拟行星实验室）；Arizona State University（亚利桑那州立大学）；University of Maryland, Baltimore County（马里兰大学巴尔的摩分校）；NASA Goddard Space Flight Center（美国宇航局戈达德太空飞行中心）；United States Naval Observatory（美国海军天文台）
Polarimetry is one of the keys to enhanced direct imaging of exoplanets. Not only does it deliver a differential observable providing extra contrast, but when coupled with spectroscopy, it also reveals valuable information on the exoplanetary atmospheric composition. Nevertheless, angular separation and contrast ratio to the host-star make for extremely challenging observation. Producing detailed predictions for exactly how the expected signals should appear is of critical importance for the designs and observational strategies of tomorrow's telescopes. We aim at accurately determining the mag…
Spectral retrieval has long been a powerful tool for interpreting planetary remote sensing observations. Flexible, parameterised, agnostic models are coupled with inversion algorithms in order to infer atmospheric properties directly from observations, with minimal reliance on physical assumptions. This approach, originally developed for application to Earth satellite data and subsequently observations of other Solar System planets, has been recently and successfully applied to transit, eclipse and phase curve spectra of transiting exoplanets. In this review, we present the current state-of-th…
The Ariel mission, due to launch in 2029, will obtain spectroscopic information for 1000 exoplanets, providing an unprecedented opportunity for comparative exoplanetology. Retrieval codes - parameteric atmospheric models coupled with an inversion algorithm - represent the tool of choice for interpreting Ariel data. Ensuring that reliable and consistent results can be produced by these tools is a critical preparatory step for the mission. Here, we present the results of a retrieval challenge. We use five different exoplanet retrieval codes to analyse the same synthetic datasets, and test a) the…
FlopPITy: Enabling self-consistent exoplanet atmospheric retrievals with machine learning
Kapteyn Astronomical Institute, University of Groningen（格罗宁根大学卡普坦天文研究所）；Netherlands Space Research Institute (SRON)（荷兰空间研究所（SRON））；Centre for Exoplanet Science, University of Edinburgh（爱丁堡大学系外行星科学中心）；School of GeoSciences, University of Edinburgh（爱丁堡大学地球科学学院）
TelescopeML -- I. An End-to-End Python Package for Interpreting Telescope Datasets through Training Machine Learning Models, Generating Statistical Reports, and Visualizing Results
NASA Ames Research Center（美国国家航空航天局艾姆斯研究中心）；Bay Area Environmental Research Institute（湾区环境研究所）；Universities Space Research Association（大学空间研究协会）；Helmholtz-Zentrum Dresden-Rossendorf（亥姆霍兹德累斯顿罗森多夫研究中心）；University of California, Riverside（加州大学河滨分校）
The MUSCLES Extension for Atmospheric Transmission Spectroscopy: Spectral energy distributions for 20 exoplanet host stars that JWST observed in Cycle 1
The use of machine learning is becoming ubiquitous in astronomy, but remains rare in the study of the atmospheres of exoplanets. Given the spectrum of an exoplanetary atmosphere, a multi-parameter space is swept through in real time to find the best-fit model. Known as atmospheric retrieval, it is a technique that originates from the Earth and planetary sciences. Such methods are very time-consuming and by necessity there is a compromise between physical and chemical realism versus computational feasibility. Machine learning has previously been used to determine which molecules to include in t…
Determining the habitability and interpreting atmospheric spectra of exoplanets requires understanding their atmospheric physics and chemistry. We use a 3-D Coupled Climate-Chemistry Model, the Met Office Unified Model with the UK Chemistry and Aerosols framework, to study the emergence of lightning and its chemical impact on tidally-locked Earth-like exoplanets. We simulate the atmosphere of Proxima Centauri b orbiting in the Habitable Zone of its M-dwarf star, but the results apply to similar M-dwarf orbiting planets. Our chemical network includes the Chapman ozone reactions and hydrogen oxi…
Kapteyn Astronomical Institute, University of Groningen（格罗宁根大学卡普坦天文研究所）；Netherlands Space Research Institute (SRON)（荷兰空间研究所（SRON））；Centre for Exoplanet Science, University of Edinburgh（爱丁堡大学系外行星科学中心）；School of GeoSciences, University of Edinburgh（爱丁堡大学地球科学学院）
Exoplanet observations are currently analysed with Bayesian retrieval techniques. Due to the computational load of the models used, a compromise is needed between model complexity and computing time. Analysis of data from future facilities, will need more complex models which will increase the computational load of retrievals, prompting the search for a faster approach for interpreting exoplanet observations. Our goal is to compare machine learning retrievals of exoplanet transmission spectra with nested sampling, and understand if machine learning can be as reliable as Bayesian retrievals for…
Using a 3D general circulation model (GCM), we investigate the sensitivity of the climate of tidally-locked Earth-like exoplanets, Trappist-1e and Proxima Centauri b, to the choice of a convection parameterization. Compared to a mass-flux convection parameterization, a simplified convection adjustment parameterization leads to a $>$60% decrease of the cloud albedo, increasing the mean day-side temperature by $\approx$10 K. The representation of convection also affects the atmospheric conditions of the night side, via a change in planetary-scale wave patterns. As a result, using the convection …
We simulate the space environment around AU Microscopii b and the interaction between the magnetized stellar wind with a planetary atmospheric outflow for ambient stellar wind conditions and Coronal Mass Ejection (CME) conditions. We also calculate synthetic Ly$α$ absorption due to neutral hydrogen in the ambient and the escaping planetary atmosphere affected by this interaction. We find that the Ly$α$ absorption is highly variable due to the highly-varying stellar wind conditions. A strong Doppler blue-shift component is observed in the Ly$α$ profile, in contradiction to the actual escape vel…
We are now on a clear trajectory for improvements in exoplanet observations that will revolutionize our ability to characterize their atmospheric structure, composition, and circulation, from gas giants to rocky planets. However, exoplanet atmospheric models capable of interpreting the upcoming observations are often limited by insufficiencies in the laboratory and theoretical data that serve as critical inputs to atmospheric physical and chemical tools. Here we provide an up-to-date and condensed description of areas where laboratory and/or ab initio investigations could fill critical gaps in…
Beyond the mass-radius plane: Integrated radiative-convective and interior structure simulations of the exoplanet continuum
Institute of Astronomy, University of Cambridge（剑桥大学天文研究所）；Atmospheric Oceanic and Planetary Physics, University of Oxford（牛津大学大气海洋与行星物理系）；Kapteyn Astronomical Institute, University of Groningen（格罗宁根大学卡普坦天文研究所）
We aim to interpret future photometric and spectral measurements from these instruments, in terms of physical parameters of the planets, with an atmospheric model using a minimal number of assumptions and parameters. We developed Exoplanet Radiative-convective Equilibrium Model (Exo-REM) to analyze the photometric and spectro- scopic data of directly imaged planets. The input parameters are a planet's surface gravity (g), effective temperature (Teff ), and elemental composition. The model predicts the equilibrium temperature profile and mixing ratio profiles of the most important gases. Opacit…
The complexity of atmospheric retrieval models is largely data-driven and one-dimensional models have generally been considered adequate with current data quality. However, recent studies have suggested that using 1D models in retrievals can result in anomalously cool terminator temperatures and biased abundance estimates even with existing transmission spectra of hot Jupiters. Motivated by these claims and upcoming high-quality transmission spectra we systematically explore the limitations of 1D models using synthetic and current observations. We use 1D models of varying complexity, both anal…
While the occurrence rate of wide giant planets appears to increase with stellar mass at least up through the A-type regime, B-type stars have not been systematically studied in large-scale surveys so far. It therefore remains unclear up to what stellar mass this occurrence trend continues. The B-star Exoplanet Abundance Study (BEAST) is a direct imaging survey with the extreme adaptive optics instrument SPHERE, targeting 85 B-type stars in the young Scorpius-Centaurus (Sco-Cen) region with the aim to detect giant planets at wide separations and constrain their occurrence rate and physical pro…
Dozens of habitable zone, approximately earth-sized exoplanets are known today. An emerging frontier of exoplanet studies is identifying which of these habitable zone, small planets are actually habitable (have all necessary conditions for life) and, of those, which are earth-like. Many parameters and processes influence habitability, ranging from the orbit through detailed composition including volatiles and organics, to the presence of geological activity and plate tectonics. While some properties will soon be directly observable, others cannot be probed by remote sensing for the foreseeable…
Quantifying the differences in transmission and emission spectra for hot irradiated gaseous exoplanet atmospheres: A comparison of 1D and 3D modeling using JWST
Transmission spectroscopy provides us with information on the atmospheric properties at the limb, which is often intuitively assumed to be a narrow annulus around the planet. Consequently, studies have focused on the effect of atmospheric horizontal heterogeneities along the limb. Here we demonstrate that the region probed in transmission -- the limb -- extends significantly toward the day and night sides of the planet. We show that the strong day-night thermal and compositional gradients expected on synchronous exoplanets create sufficient heterogeneities across the limb to result in importan…
Ultraviolet Radiation Effects on the Optical Properties of Water-Dominated Exoplanet Hazes
University of Arizona（亚利桑那大学）；Space Telescope Science Institute（太空望远镜科学研究所）；Jet Propulsion Laboratory（喷气推进实验室）；University of Science and Technology of China（中国科学技术大学）；Johns Hopkins University（约翰斯·霍普金斯大学）；Planetary Science Institute（行星科学研究所）；Cornell University（康奈尔大学）
Interpreting the spectra of brown dwarfs is key to determining the fundamental physical and chemical processes occurring in their atmospheres. Powerful Bayesian atmospheric retrieval tools have recently been applied to both exoplanet and brown dwarf spectra to tease out the thermal structures and molecular abundances to understand those processes. In this manuscript we develop a significantly upgraded retrieval method and apply it to the SpeX spectral library data of two benchmark late T-dwarfs, Gl570D and HD3651B, to establish the validity of our upgraded forward model parameterization and Ba…
