C/O Ratios of Transiting Exoplanet Host Stars

The second half of my Ph.D. thesis moved from observing exoplanets to their host stars; I focused on measuring carbon and oxygen in transiting exoplanet (TEP) host stars in order chemically connect exoplanets and their host stars. One of the most robust findings in this field is that hot Jupiter host stars have higher metallicities (measured as [Fe/H], the amount of iron in a star relative to how much iron is in the Sun), which can be explained by the core-accretion model of exoplanet formation. However, the metallicity trend is weaker for Neptune-sized planets and is even weaker (or disappears entirely) for terrestrial-sized planets. In the current study of exoplanets, the atmospheric compositions of exoplanets that transit their host star (pass in front of the star in our line of sight) can (begin to) be constrained. Measurements of both star and exoplanet combined, in a variety of systems, can be compared for differences or trends that give insight into planet formation processes in these systems beyond the giant planet-metallicity trend.

The relative abundances of C and O have long been recognized as diagnostics of Galactic chemical evolution, and are also the most promising elements for measurement in both stellar and exoplanetary atmospheres. Carbon and oxygen are of particular interest in exoplanet atmospheres because the C/O ratio significantly affects the molecular composition, and hence observed spectral signatures, through thermochemical equilibrium partitioning of carbon in CO, CH4, and CO2. Hot Jupiters are ideal probes of different chemistry induced by different C/O ratios because their IR spectra are dominated by these molecules. Also, the C/O ratio of an exoplanet can give clues as to where in the protoplanetary disk it formed. Since the main molecular reservoirs of C and O have different condensation temperatures, their relative amounts vary at different temperatures/disk radii, as do the fractions of gas versus grains and their movement in the disk. These factors can change the relative ratios of C and O in the planet compared to those in the parent star.

Figure from Öberg et al. (2011), showing how in a simple equilibrium chemistry model of a protoplanetary disk, the C/O ratio of the gas and dust can vary with distance from the star.

Therefore, the C/O ratio of a planet does not necessarily reflect the disk-averaged C/O ratio, and is affected by where formation occurs, how much of the atmosphere is accreted from gas versus solids, and how isolated the atmosphere is from the core . Previous work hinted at a new class of carbon-rich planets, unlike anything in our Solar System, that JWST and other upcoming space-based missions will be able to better characterize.

Figure from Madhusudhan et al. (2012), showing how the atmospheric C/O ratio might be used as a primary characteristic to explain the varied observations of transiting hot Jupiter atmospheres.

The more precise abundance analysis that is possible right now for host stars can help constrain their exoplanets’ formation histories. Already, host star C/O ratios have been used to motivate compositional modeling of their exoplanets.

To explore the possibility of C-rich planet atmospheres from a new perspective, I  measured C/O in 16 host stars of well-studied transiting hot Jupiter exoplanets. I proposed for and conducted Subaru/HDS and Keck/HIRES high-resolution optical echelle observations of the host stars, and reduced the data “by hand” using IRAF echelle routines (and then verified with the HDS pipeline). I derived the stellar parameters, as well as several elemental abundances including C and O, of these host stars using SPECTRE to measure line equivalent widths (by hand/individually) and MOOG to determine the best stellar models (using an interpolated suite of Kurucz ATLAS models) and resulting abundances (using either the collectively measured lines for an element or the spectral synthesis technique for particular elemental lines that were blended). A non-automated analysis is necessary to derive the most accurate C and O abundances -- determinations of stellar C/O ratios can be particularly challenging because indicators for both elements are weak, blended with other lines, and/or subject to non-LTE e ects (Teske et al. 2013b,c).

Below: Top: From Teske et al. (2014). Example of spectral synthesis fit to forbidden [Oi] line (λ6300.3) for HD189733. The data are shown as black open circles. The full synthesis fit is represented by a solid red line, with components shown with blue dash-dotted ([O i]), green dashed (Ni i), and pink dotted (CN) lines. Bottom: From Line et al. (2015). Plotted are the the lines measured in this work to determine the C/O ratio of GL 570A. Dots represent the spectrum of GL 570A, while triangles represent the solar standard spectrum. In the top two plots, synthesis fits to the lines are shown in red dot-dashed lines.

I found that previously-measured stellar C/O ratios (e.g., Delgado Mena et al. 2010; Petigura & Marcy 2011) were likely overestimated (Teske et al. 2014), as suggested by Fortney (2012), Nissen (2013), and later Gaidos (2015). What I did not nd were trends between C/O_{host star} and measurements of the HJ atmospheres (e.g., Spitzer/IRAC NIR secondary eclipse fluxes) or their bulk physical properties. This supports the idea that how a planet's atmospheric C/O ratio diff ers from its host star may help determine where/when in the disk it formed (Madhusudhan et al. 2014; Oberg et al. 2011; Ciesla & Cuzzi 2006; Stevenson & Lunine 1988). There is likely a C/O ratio range in individual protoplanetary disks, related to other planet formation parameters (mass of the disk, grain growth and composition, distance away from the star, etc.) that influence the final C/O_{planet}. I also measured C/O ratios in two additional types of host stars -- 55 Cnc, a cool and metal-rich star hosting a purported "diamond-planet" (Teske et al. 2013b) and two cool K dwarfs, each hosting a late T dwarf (Line et al. 2015). These studies further demonstrated the caution required when determining stellar carbon and oxygen abundances.

Above: From Teske et al. (2014). C/O vs. [Fe/H] from Delgado Mena et al. (2010) and Nissen (2013) (all Nissen (2013) hosts are in the Delgado Mena et al. (2010) host sample). Non-host stars from Delgado Mena et al. (2010) are plotted with gray open squares, while host stars from Delgado Mena et al. (2010) and Nissen (2013) are plotted with gray and blue asterisks, respectively. Quoted typical error bars are in the upper left. Measurements from this work are plotted as red filled circles, with error bars included. All C/O ratios normalized have been to the same C/O adopted in my paper, C/O = 0.54 (logN(C) = 8.39, log N(O) = 8.66; Asplund et al. 2005).