Zoe Ko

Giant planets orbiting low-mass M dwarfs challenge current planet formation theories, which predict that such planets are unlikely to form. Characterizing their atmospheres can provide key insight into their origins, but analysis is complicated by stellar contamination in transmission spectra. We present a JWST/NIRSpec PRISM transmission spectrum of TOI-3235 b, a 604 K, 0.665 M _Jup giant planet orbiting a 0.39 M _⊙ M dwarf. We apply a hierarchical retrieval framework incorporating an atmospheric model, a parametric stellar contamination model, and a Gaussian process (GP) to capture residual structure not explained by these deterministic components. We find that the inferred atmospheric properties are model dependent: retrievals that treat stellar contamination deterministically yield a constrained CH ₄ abundance corresponding to a subsolar metallicity that would challenge standard expectations from both core accretion and gravitational instability, and potentially elevated CO and CO ₂ abundances that would require chemical disequilibrium. However, residual wavelength-correlated structure suggests the deterministic model is incomplete. Including a GP to marginalize over this structure broadens the atmospheric posterior distributions, preventing robust constraints on the atmospheric composition, and highlighting the challenge of characterizing giant planet atmospheres around M dwarfs when physical models are incomplete. We demonstrate that an eclipse observation could clarify these model-dependent ambiguities: detected emission features would constrain the planet’s metallicity and test formation scenarios, while a featureless spectrum would confirm stellar contamination dominates the transmission spectrum, providing an empirical M dwarf contamination spectrum.

Gravitational wave observations of binary black hole mergers probe their astrophysical origins via the binary spin, namely the spin magnitudes and directions of each component black hole, together described by six degrees of freedom. However, the emitted signals primarily depend on two effective spin parameters that condense the spin degrees of freedom to those parallel and those perpendicular to the orbital plane. Given this reduction in dimensionality between the physically relevant problem and what is typically measurable, we revisit the question of whether information about the component spin magnitudes and directions can successfully be recovered via gravitational-wave observations, or if we simply extrapolate information about the distributions of effective spin parameters. To this end, we simulate three astrophysical populations with the same underlying effective-spin distribution but different spin magnitude and tilt distributions, on which we conduct full individual-event and population-level parameter estimation. We find that parametrized population models can indeed qualitatively distinguish between populations with different spin magnitude and tilt distributions at current sensitivity. However, it remains challenging to either accurately recover the true distribution or to diagnose biases due to model misspecification. We attribute the former to practical challenges of dealing with high-dimensional posterior distributions, and the latter to the fact that each individual event carries very little information about the full six spin degrees of freedom.

The Search for Extraterrestrial Intelligence has traditionally been conducted at radio wavelengths, but optical searches are well-motivated and increasingly feasible due to the growing availability of high-resolution spectroscopy. We present a data analysis pipeline to search Automated Planet Finder (APF) spectroscopic observations from the Levy Spectrometer for intense, persistent, narrow-bandwidth optical lasers. We describe the processing of the spectra, the laser search algorithm, and the results of our laser search on 1983 spectra of 388 stars as part of the Breakthrough Listen search for technosignatures. We utilize an empirical spectra-matching algorithm called SpecMatch-Emp to produce residuals between each target spectrum and a set of best-matching catalog spectra, which provides the basis for a more sensitive search than previously possible. We verify that SpecMatch-Emp performs well on APF-Levy spectra by calibrating the stellar properties derived by the algorithm against the SpecMatch-Emp library and against Gaia catalog values. We leverage our unique observing strategy, which produces multiple spectra of each target per night of observing, to increase our detection sensitivity by programmatically rejecting events that do not persist between observations. With our laser search algorithm, we achieve a sensitivity equivalent to the ability to detect an 84 kW laser at the median distance of a star in our data set (78.5 ly). We present the methodology and vetting of our laser search, finding no convincing candidates consistent with potential laser emission in our target sample.