Projects

Welcome to my projects page.

My research has focused on:

My career goal is to enable reliable atmospheric characterization of exoplanets, including habitable planets, by pioneering new avenues in atmospheric retrieval modelling.

Here you will find a brief description of the projects I have worked on and are currently published.

Aurora: A Generalized Retrieval Framework for Exoplanetary Transmission Spectra

You can read the full paper open access here. This research was made open access thanks to the Bill and Melinda Gates foundation.

AURORA, logo designed by the great Amanda Smith at the IoA

Authors: Luis Welbanks and Nikku Madhusudhan

Generally, the model complexity considered in atmospheric retrievals has largely been driven by the fidelity of the data – e.g., overly complex models are often not warranted. Due to their observational favourability, most spectro-photometric observations have been of hot Jupiters (e.g., Deming et al. 2013, McCullough et al. 2014, Sing et al. 2016; Nikolov et al. 2018). However, as we look towards the future, observations of low mass planets, theorized to present extreme compositional diversity (e.g., Moses et al. 2013), will be more prevalent. Likewise, the large number of upcoming spectroscopic facilities underscore the need for combining datasets from multiple observatories to infer the atmospheric properties of an exoplanet. An increasingly diverse planet population and higher fidelity data necessarily demand more flexible and complex Bayesian modelling frameworks.

            The development of these new frameworks has been a recent focus of mine. I led the development of a next-generation retrieval framework for the characterization of H-rich and H-poor atmospheres, Aurora (Welbanks & Madhusudhan 2020). Aurora is the only code that relaxes the assumption of a H-rich atmosphere to retrieve the atmospheric composition of any type of exoplanet atmosphere, while considering the presence of inhomogeneous clouds and hazes. This new framework considerably expands the model flexibility required for explaining higher fidelity observations. Focused on transmission spectra, which, due to their geometry, require more complex forward models, Aurora incorporates several optimization and data analysis tools like Gaussian processes to treat correlated noise and next-generation Bayesian samplers.

This framework incorporates new optimization algorithms designed to ease the computational burden when performing parameter estimation in multidimensional atmospheric models. Furthermore, important considerations in the interpretation of transmission spectra such as the impact of stellar heterogeneity, correlated noise, instrumental shifts, and underestimated variances in the data are included within Aurora’s framework. This more advanced modelling framework is essential to reliably extracting atmospheric properties from the data of today and tomorrow.

Mass–Metallicity Trends in Transiting Exoplanets 

You can read the full paper open access here. This research was made open access thanks to the Bill and Melinda Gates foundation.

Authors: Luis Welbanks, Nikku Madhusudhan, Nicole F. Allard, Ivan Hubeny, Fernand Spiegelman, and Thierry Leininger

The key advantage of extra-solar planetary science is the sheer numbers of objects which enable us to address population level hypotheses regarding the origin and evolution of planets. One major objective is to use measured atmospheric observable chemical abundances to test our hypothesis on the primordial formation pathways of exoplanets. A key hypothesis is that the metallicity of a planetary atmosphere should increase with decreasing mass (e.g., Fortney et al. 2013, Mordasini et al. 2016), a hallmark of the core-accretion model of planet formation and a trend that appears within the solar system Jovian planet population. Previous studies, focused on HST WFC3 (1.1-1.6 µm), have aimed to determine H2O abundances relative to solar system planets (e.g. Madhusudhan et al. 2014b, Kreidberg et al. 2014, Kreidberg et al. 2015, Barstow et al. 2017, Pinhas et al. 2019). While many of these studies were able to determine water abundances, there is tension in the underlying interpretation as there exists an intrinsic degeneracy between the atmospheric metal content and the elemental carbon-to-oxygen ratios, which can result in numerous ways to produce a given H2O abundance (e.g. Madhusudhan et al. 2014a).

In order break this intrinsic elemental abundance degeneracy, constraints on more species beyond H2O were desperately needed. Building upon my previously developed frameworks, I set out to determine population level composition trends leveraging data spanning the optical-to-near-infrared, covering not only H2O, but key metallicity tracers, sodium (Na) and potassium (K). This effort (Welbanks et al. 2019) constitutes the largest (i.e., broad wavelength coverage, multiple chemical species, mini-Neptunes to Jupiter sized planets) homogeneous chemical abundance survey for transiting exoplanets to date. While previous studies limited their efforts to hot Jupiters, I extended the sample to mini-Neptunes providing a larger, more diverse sample to provide more leverage in scrutinizing key planet formation hypotheses. In total, I retrieved and analyzed the atmospheric properties of 19 exoplanets ranging from cool mini-Neptunes with temperatures close to 300 K to ultra hot Jupiters with temperatures above 2700 K.

Mass-metallicity relation in transiting exoplanets

The novel contribution of my study was extending the analysis of the so-called mass-metallicity relationship to species beyond H2O. Surprisingly, I found a mass-metallicity trend for H2O abundances significantly below than anticipated from the solar system gas giant planets and from the aforementioned core-accretion predictions. Conversely, I found abundances of alkali species (Na and K) consistent with or higher than solar system derived metallicities. These results are suggestive of superstellar C/O, Na/O, and K/O ratios, meaning stellar or superstellar metallicities but depleted oxygen relative to other species. By acknowledging that different chemical species provide different insights into the atmospheric properties of an exoplanet, this approach was able to break the important degeneracy between C/O ratios and atmospheric metallicity prevalent up to then in the field.

            The consistent depletion of H2O, relative to solar expectations, suggests planet formation pathways that differ from the standard solar system paradigm or longer period planets. Furthermore the differing trends amongst the various species (H2O, Na, K) argue against the use of more constrained retrieval models that rely upon scaled solar elemental abundances – an additional indication that modelling assumptions have a strong influence on atmospheric compositional inferences. Upcoming spectroscopic observations, especially those with JWST at longer wavelengths, will inform whether these observed trends are statistically significant and whether they exist for other chemical species. Deriving credible abundance constrains from these future observations will require not yet developed, novel retrieval modeling frameworks, a focus of my research.

On Degeneracies in Retrievals of Exoplanetary Transmission Spectra

You can read the full paper open access here. This research was made open access thanks to the Bill and Melinda Gates foundation.

Authors: Luis Welbanks and Nikku Madhusudhan

Transmission spectroscopy of transiting exoplanets offers a powerful probe to study their atmospheres. The interpretation of such observations is routinely performed using Bayesian inference tools known as retrievals. At its core, a retrieval framework is composed of an atmospheric forward model that computes a synthetic spectrum and an optimization algorithm that extracts atmospheric constraints from the observables (see e.g. Madhusudhan 2018). While the retrieval approach is a powerful means to derive atmospheric properties of exoplanets, their applicability is inherently limited by the quality (e.g., resolution, wavelength coverage, signal-to-noise) of the data being interpreted and the inherent model assumptions. Indeed, the use of different model assumptions and different datasets have led to contrasting results in the literature (e.g., Kreidberg et al. 2015 – Heng & Kitzmann 2017, Sing et al. 2016 – Barstow et al. 2017 – Pinhas et al. 2019, Sheppard et al. 2017 – Arcangeli et al. 2018).

            To address this problem, I performed a systematic exploration of the degeneracies between different model considerations and the observations that can resolve them (Welbanks & Madhusudhan 2019). This study used a combination of Bayesian atmospheric retrievals and a range of common model assumptions, focusing on H2-rich atmospheres. The models considered increased in complexity and completeness, starting with simple isothermal and isobaric atmospheres (known to be unphysical) to those with full pressure–temperature profiles, inhomogeneous cloud and haze coverage, multiple molecular species (more physically plausible), and data in the optical–infrared wavelengths obtained with HST-STIS and HST-WFC3. This exploration of model considerations demonstrated that inferences derived from observations are strongly influenced by model assumptions.

Inferences derived from observations are strongly influenced by model assumptions

            This work demonstrated that it is possible to overcome the aforementioned limitations by using a combination of physically motivated models with minimal assumptions and broadband transmission spectra with current facilities. I robustly demonstrated that precise estimates of chemical abundances are possible with current transmission spectra when using high-precision optical and infrared spectra, along with models including variable cloud coverage and prominent opacity sources. This work’s key contribution to the field was demonstrating the shortcomings of semi-analytic models employed in previous studies. Furthermore, I demonstrated that the degeneracy between planetary radius and its reference pressure, previously considered a fundamental hindrance for transmission spectra, is well characterized, and has little effect on the derived abundance estimates.

            While my previous work demonstrated that low resolution spectro-photometric observations with facilities such as HST and VLT can provide strong constraints on atmospheric abundances of exoplanets, these considerations will continue to be explored in the context of upcoming facilities such as JWST and ELT. The current modelling paradigm (e.g., 1-dimensional, simplistic clouds, minimal chemical assumptions) will undoubtedly need revision when confronted with observations comprised of higher resolution, longer spectral coverage, and better signal-to-noise anticipated over the next decade. Therefore, in order to maximize the scientific return of tomorrow’s high fidelity observations, we must systematically examine the strengths and weaknesses of the current modelling paradigms in order to develop a foundation for the next-generation of atmospheric retrieval models – the focus of my research.

Our study reveals four key insights:

  • First, we find that a combination of models with minimal assumptions and broadband transmission spectra with current facilities allows precise estimates of chemical abundances. In particular, high-precision optical and infrared spectra, along with models including variable cloud coverage and prominent opacity sources, with Na and K being important in the optical, provide joint constraints on cloud/haze properties and chemical abundances.
  • Second, we show that the degeneracy between planetary radius and its reference pressure is well characterized and has little effect on abundance estimates, contrary to previous claims using semi-analytic models.
  • Third, collision-induced absorption due to H2–H2 and H2–He interactions plays a critical role in correctly estimating atmospheric abundances.
  • Finally, our results highlight the inadequacy of simplified semi-analytic models with isobaric assumptions for reliable retrievals of transmission spectra.

Characterisation of exoplanets

Part of my work has been analysing the spectra of different exoplanets. This has been an exciting opportunity and has been work done in collaboration with many people. Here I highlight some of the work that has been already published.

The atmosphere and interior of K2-18b

You can read the full paper open access here. This research was made open access thanks to the Bill and Melinda Gates foundation.

Authors: Nikku Madhusudhan, Matthew Nixon, Luis Welbanks et. al

This study linked atmospheric observations with mass and radius estimates (bulk properties) to perform an end-to-end analysis of the habitable-zone exoplanet K2-18b. Here, I helped perform the retrievals on the spectra of the mini-Neptune K2-18b. We used different models ranging from clear atmospheres, to inhomogeneous clouds and hazes passing through fully cloudy atmospheres.  We constrained the atmosphere to be H2-rich with a H2O volume mixing ratio of 0.02%-14.8%, consistent with previous studies, and find a depletion of CH4 and NH3, indicating chemical disequilibrium. These results were then used to constrain the internal structure and thermodynamic conditions in the planet. This study finds that K2-18b hosts conditions at its surface that allow for liquid water. These results demonstrate that the potential for habitable conditions is not necessarily restricted to Earth-like rocky exoplanets.

Detection of haze, Na, K, and Li in the super-Neptune WASP-127b

Authors: Guo Chen, Enric Pallé, Luis Welbanks et. al
You can read the paper here.

Observations of the super-Neptune WASP-127b were taken using ground based telescopes (Gran Telescopio Canarias (GTC) and Nordic Optical Telescope (NOT)). I helped perform the analysis of the spectrum using retrievals and we found strong indications of Na, K and Li absorption. This data set was the first to exhibit Li absorption in an exoplanet.

First indication of aluminum oxide in an exoplanet WASP-33 b.

Authors: Carolina von Essen, Matthias Mallon, Luis Welbanks et. al
You can read the paper here.

Using again the ground based observatory Gran Telescopio Canarias, the ultra hot Jupiter WASP-33b was observed. I helped perform a retrieval on these spectral observations. Our analysis found that the feature at ~400 nm was better explained by the presence of AlO, aluminium oxide. This is the first indication of such species in an exoplanet. This indication is an interesting prospect in a field where we have an increasing diversity of ‘exotic’ species found in ultra hot Jupiters.

WASP 127b image credit. Artistic simulation of WASP 127b orbiting a star. Credit: Gabriel Pérez, SMM (IAC).

%d bloggers like this: