arXiv subjects

Kristen Menou

Publications and source records attributed to Kristen Menou.

At least 19 recordsLinked to original sources

arXiv2026

Stargazer: A Scalable Model-Fitting Benchmark Environment for AI Agents under Astrophysical Constraints

The rise of autonomous AI agents suggests that dynamic benchmark environments with built-in feedback on scientifically grounded tasks are needed to evaluate the capabilities of these agents in research work. We introduce Stargazer, a scalable environment for evaluating AI agents on dynamic, iterative physics-grounded model-fitting tasks using inference on radial-velocity (RV) time series data. Stargazer comprises 120 tasks across three difficulty tiers, including 20 real archival cases, covering diverse scenarios ranging from high-SNR single-planet systems to complex multi-planetary configurations requiring involved low-SNR analysis. Our evaluation of eight frontier agents reveals a gap between numerical optimization and adherence to physical constraints: although agents often achieve a good statistical fit, they frequently fail to recover correct physical system parameters, a limitation that persists even when agents are equipped with vanilla skills. Furthermore, increasing test-time compute yields only marginal gains, with excessive token usage often reflecting recursive failure loops rather than meaningful exploration. Stargazer presents an opportunity to train, evaluate, scaffold, and scale strategies on a model-fitting problem of practical research relevance today. Our methodology to design a simulation-driven environment for AI agents presumably generalizes to many other model-fitting problems across scientific domains. Source code and the project website are available at https://github.com/AIPS-UofT/Stargazer and https://aips-uoft.github.io/Stargazer/, respectively.

cs.LG↗

arXiv2025

Meridional circulation molecular-weighted

Meridional circulation in stratified stellar/planetary interiors in the presence of stable molecular weight gradients remains poorly understood, thereby affecting angular momentum transport in evolutionary models. We extend the downward control principle of atmospheric sciences to include compositional stratification. Using a linearized analysis we show that stable compositional gradients slow down the penetration of circulation into the depths, emphasizing the importance of time-dependent solutions. However, additional effects such as horizontal turbulence or magnetic fields are needed to halt it completely. We also find limits demarcating linear and nonlinear regimes in terms of Schmidt and Rossby numbers. Nonlinear simulations exhibit compositional mixing due to meridional currents, enabling deeper penetration than otherwise. We propose slowly evolving and steady-state scenarios for the solar tachocline, and helioseismically observed heavy element abundances, while acknowledging the absence of constraints on the radial variation of the Schmidt number. In the context of stellar evolution, differential rotation profiles of solar-type main-sequence stars may follow analytical solutions extending from Banik & Menou (2024), thereby aiding further probes into magneto/hydrodynamic instabilities and outward angular momentum transport.

astro-ph.SR↗

arXiv2025

Gravity-Bench-v1: A Benchmark on Gravitational Physics Discovery for Agents

Modern science emerged from reasoning over repeatedly-observed planetary motions. We present Gravity-Bench-v1, an environment-based benchmark that challenges AI agents on tasks that parallel this historical development. Gravity-Bench-v1 evaluates agents on the discovery of physics concealed within a dynamic environment, using rigorous gravitational dynamics simulations. Gravity-Bench includes out-of-distribution cases, i.e. with physics that deviates from the real world, to evaluate true scientific generalization capabilities. Agents must plan to collect data within an experimental budget and must perform a dynamic form of data analysis and reasoning to solve tasks efficiently. Our benchmark admits an open-ended space of solutions. Reference solutions for each task are provided to calibrate AI performance against human expertise. Technically at an upper-undergraduate level, our benchmark proves challenging to baseline AI agents. Gravity-Bench-v1 and planned extensions should help map out AI progress towards scientific discovery capabilities.

cs.AI↗

arXiv2024

Hot Jupiters are asynchronous rotators

Hot Jupiters are typically assumed to be synchronously rotating, from tidal locking. Their thermally-driven atmospheric winds experience Lorentz drag on the planetary magnetic field anchored at depth. We find that the magnetic torque does not integrate to zero over the entire atmosphere. The resulting angular momentum feedback on the bulk interior can thus drive the planet away from synchronous rotation. Using a toy tidal-ohmic model and atmospheric GCM outputs for HD189733b, HD209458b and Kepler7b, we establish that off-synchronous rotation can be substantial at tidal-ohmic equilibrium for sufficiently hot and/or magnetized hot Jupiters. Potential consequences of asynchronous rotation for hot Jupiter phenomenology motivate follow-up work on the tidal-ohmic scenario with approaches that go beyond our toy model.

astro-ph.EP↗

arXiv2024

Climate Transition to Temperate Nightside at High Atmosphere Mass

Our recent work shows how M-Earth climates and transmission spectra depend on the amount of ice-free ocean on the planet's dayside and the mass of N2 in its atmosphere. M-Earths with more ice-free ocean and thicker atmospheres are hotter and more humid, and have larger water vapour features in their transmission spectra. In this paper, we describe a climate transition in high-pN2 simulations from the traditional ``eyeball" M-Earth climate, in which only the substellar region is temperate, to a ``temperate nightside" regime in which both the dayside and the nightside are entirely ice-free. Between these two states, there is a ``transition" regime with partial nightside ice cover. We use 3D climate simulations to describe the climate transition from frozen to deglaciated nightsides. We attribute this transition to increased advection and heat transport by water vapour in thicker atmospheres. We find that the nightside transitions smoothly back and forth between frozen and ice-free when the instellation or pCO2 is perturbed, with no hysteresis. We also find an analogous transition in colder planets: those with thin atmospheres can have a dayside hotspot when the instellation is low, whereas those with more massive atmospheres are more likely to be in the ``snowball" regime, featuring a completely frozen dayside, due to the increased advection of heat away from the substellar point. We show how both of these climate transitions are sensitive to instellation, land cover, and atmosphere mass. We generate synthetic transmission spectra and phase curves for the range of climates in our simulations.

astro-ph.EP↗

arXiv2024

Meridional Circulation Streamlined

Time-dependent meridional circulation and differential rotation in radiative zones are central open issues in stellar evolution theory. We streamline this challenging problem using the downward control principle of atmospheric science, under a geostrophic f-plane approximation. We recover the known stellar physics result that the steady-state meridional circulation decays on the length scale proportional to N/f sqrt(Pr), assuming molecular viscosity is the dominant drag mechanism. Prior to steady-state, the meridional circulation and the zonal wind (= differential rotation) spread together via radiative diffusion, under thermal wind balance. The corresponding (4th-order) hyperdiffusion process is reasonably well approximated by regular (2nd-order) diffusion on scales of order a pressure scale-height. We derive an inhomogeneous diffusion (equiv. advection-diffusion) equation for the zonal flow which admits closed-form time-dependent solutions in a finite depth domain, allowing for rapid prototyping of differential rotation profiles. In the weak drag limit, we find that the time to rotational steady-state can be longer than the Eddington-Sweet time and be instead determined by the longer drag time. Unless strong enough drag operates, the internal rotation of main-sequence stars may thus never reach steady-state. Streamlined meridional circulation solutions provide leading-order internal rotation profiles for studying the role of fluid/MHD instabilities (or waves) in redistributing angular momentum in the radiative zones of stars. Despite clear geometrical limitations and simplifying assumptions, one might expect our thin-layer geostrophic approach to offer qualitatively useful results to understand deep meridional circulation in stars.

astro-ph.SR↗

arXiv2024

Water Vapour Transit Ambiguities for Habitable M-Earths

We have shown in a recent study, using 3D climate simulations, that dayside land cover has a substantial impact on the climate of a synchronously rotating temperate rocky planet such as Proxima Centauri b. Building on that result, we generate synthetic transit spectra from our simulations to assess the impact of these land-induced climate uncertainties on water vapour transit signals. We find that distinct climate regimes will likely be very difficult to differentiate in transit spectra, even under the more favourable conditions of smaller planets orbiting ultracool dwarfs. Further, we show that additional climate ambiguities arise when both land cover and atmosphere mass are unknown, as is likely to be the case for transiting planets. While water vapour may be detectable under favourable conditions, it may be nearly impossible to infer a rocky planet's surface conditions or climate state from its transit spectrum due to the interdependent effects of land cover and atmosphere mass on surface temperature, humidity, and terminator cloud cover.

astro-ph.EP↗

arXiv2023

Physics simulation capabilities of LLMs

[Abridged abstract] Large Language Models (LLMs) can solve some undergraduate-level to graduate-level physics textbook problems and are proficient at coding. Combining these two capabilities could one day enable AI systems to simulate and predict the physical world. We present an evaluation of state-of-the-art (SOTA) LLMs on PhD-level to research-level computational physics problems. We condition LLM generation on the use of well-documented and widely-used packages to elicit coding capabilities in the physics and astrophysics domains. We contribute $\sim 50$ original and challenging problems in celestial mechanics (with REBOUND), stellar physics (with MESA), 1D fluid dynamics (with Dedalus) and non-linear dynamics (with SciPy). Since our problems do not admit unique solutions, we evaluate LLM performance on several soft metrics: counts of lines that contain different types of errors (coding, physics, necessity and sufficiency) as well as a more "educational" Pass-Fail metric focused on capturing the salient physical ingredients of the problem at hand. As expected, today's SOTA LLM (GPT4) zero-shot fails most of our problems, although about 40\% of the solutions could plausibly get a passing grade. About $70-90 \%$ of the code lines produced are necessary, sufficient and correct (coding \& physics). Physics and coding errors are the most common, with some unnecessary or insufficient lines. We observe significant variations across problem class and difficulty. We identify several failure modes of GPT4 in the computational physics domain. Our reconnaissance work provides a snapshot of current computational capabilities in classical physics and points to obvious improvement targets if AI systems are ever to reach a basic level of autonomy in physics simulation capabilities.

cs.AI↗

arXiv2022

Compositional Turbulence and Layering in the Gaseous Envelopes of Forming Planets

Differential settling and growth of dust grains impact the structure of the radiative envelopes of gaseous planets during formation. Sufficiently rapid dust growth can result in envelopes with substantially reduced opacities for radiation transport, thereby facilitating planet formation. We revisit the problem and establish that dust settling and grain growth also lead to outer planetary envelopes that are prone to compositional instabilities, by virtue of their inverted mean-molecular weight gradients. Under a variety of conditions, we find that the radiative envelopes of forming planets experience compositional turbulence driven by a semi-transparent version of the thermohaline instability ('fingering convection'). The compositional turbulence seems efficient at mixing dust in the radiative envelopes of planets forming at super-AU distances (say $5$ AU) from a Sun-like star, but not so at sub-AU distances (say $0.2$ AU). We also address the possibility of compositional layering in this context. Distinct turbulent regimes for planetary envelopes growing at sub-AU vs. super-AU distances could leave an imprint on the final planets formed.

astro-ph.EP↗

arXiv2022

Semi-Transparent Shear Turbulence in Hot Jupiter Atmospheres

Turbulent transport driven by secular shear instabilities can lead to enhanced vertical mixing in hot Jupiter atmospheres, impacting their cloudiness, chemistry and overall vertical structure. We discuss the turbulent regime expected and evaluate theoretical uncertainties on the strength of the vertical mixing (i.e, $K_{\rm zz}$ values). We focus our work on three well-studied hot Jupiters with a hierarchy of atmospheric temperatures: HD189733b ($T_{\rm eq} \simeq 1200$K), HD209458b ($T_{\rm eq} \simeq 1450$K) and Kepler7b ($T_{\rm eq} \simeq 1630$K). $K_{\rm zz}$ uncertainties are large. They are dominated by i) the poorly understood magnitude of turbulent transport and ii) the semi-transparent nature of shear turbulence near the planetary photosphere. Using a specific Moore-Spiegel instability threshold, we infer that the cooler HD189733b is not subject to enhanced mixing from semi-transparent shear turbulence while the daysides of the hotter Kepler7b and (marginally so) HD209458b are. Enhanced vertical mixing is generally expected to manifest on hot enough exoplanets, with $T_{\rm eq} > 1500-1600$K. On a given planet, day and night $K_{\rm zz}$ profiles can differ by an order of magnitude or more. Vertical mixing is slightly favoured in equatorial regions, where the atmospheric zonal shear is strongest. In all three planetary cases studied, momentum feedback on the atmospheric mean flow is minor to negligible.

astro-ph.EP↗

arXiv2022

TOI-1452 b: SPIRou and TESS reveal a super-Earth in a temperate orbit transiting an M4 dwarf

Exploring the properties of exoplanets near or inside the radius valley provides insights on the transition from the rocky super-Earths to the larger, hydrogen-rich atmosphere mini-Neptunes. Here, we report the discovery of TOI-1452 b, a transiting super-Earth ($R_{\rm p} = 1.67 \pm 0.07$ R$_{\oplus}$) in an 11.1--day temperate orbit ($T_{\rm eq} = 326 \pm 7$ K) around the primary member ($H = 10.0$, $T_{\rm eff} = 3185 \pm 50$ K) of a nearby visual binary M dwarf. The transits were first detected by TESS, then successfully isolated between the two $3.2^{\prime\prime}$ companions with ground-based photometry from OMM and MuSCAT3. The planetary nature of TOI-1452 b was established through high-precision velocimetry with the near-infrared SPIRou spectropolarimeter as part of the ongoing SPIRou Legacy Survey. The measured planetary mass ($4.8 \pm 1.3$ M$_{\oplus}$) and inferred bulk density ($5.6^{+1.8}_{-1.6}$ g/cm$^3$) is suggestive of a rocky core surrounded by a volatile-rich envelope. More quantitatively, the mass and radius of TOI-1452 b, combined with the stellar abundance of refractory elements (Fe, Mg and Si) measured by SPIRou, is consistent with a core mass fraction of $18\pm6$ % and a water mass fraction of $22^{+21}_{-13}$%. The water world candidate TOI-1452 b is a prime target for future atmospheric characterization with JWST, featuring a Transmission Spectroscopy Metric similar to other well-known temperate small planets such as LHS 1140 b and K2-18 b. The system is located near Webb's northern Continuous Viewing Zone, implying that is can be followed at almost any moment of the year.

astro-ph.EP↗

arXiv2022

Fundamental Challenges to Remote Sensing of Exo-Earths

Inferring the climate and surface conditions of terrestrial exoplanets in the habitable zone is a major goal for the field of exoplanet science. This pursuit will require both statistical analyses of the population of habitable planets as well as in-depth analyses of the climates of individual planets. Given the close relationship between habitability and surface liquid water, it is important to ask whether the fraction of a planet's surface where water can be a liquid, $χ_\text{hab}$, can be inferred from observations. We have produced a diverse bank of 1,874 3D climate models and computed the full-phase reflectance and emission spectrum for each model to investigate whether surface climate inference is feasible with high-quality direct imaging or secondary eclipse spectroscopy. These models represent the outcome of approximately 200,000 total simulated years of climate and over 50,000 CPU-hours, and the roughly-100 GB model bank and its associated spectra are being made publicly-available for community use. We find that there are correlations between spectra and $χ_\text{hab}$ that will permit statistical approaches. However, spectral degeneracies in the climate observables produced by our model bank indicate that inference of individual climates is likely to be model-dependent, and inference will likely be impossible without exhaustive explorations of the climate parameter space. The diversity of potential climates on habitable planets therefore poses fundamental challenges to remote sensing efforts targeting exo-Earths.

astro-ph.EP↗

arXiv2022

ExoPlaSim: Extending the Planet Simulator for Exoplanets

The discovery of a large number of terrestrial exoplanets in the habitable zones of their stars, many of which are qualitatively different from Earth, has led to a growing need for fast and flexible 3D climate models, which could model such planets and explore multiple possible climate states and surface conditions. We respond to that need by creating ExoPlaSim, a modified version of the Planet Simulator (PlaSim) that is designed to be applicable to synchronously rotating terrestrial planets, planets orbiting stars with non-solar spectra, and planets with non-Earth-like surface pressures. In this paper we describe our modifications, present validation tests of ExoPlaSim's performance against other GCMs, and demonstrate its utility by performing two simple experiments involving hundreds of models. We find that ExoPlaSim agrees qualitatively with more-sophisticated GCMs such as ExoCAM, LMDG, and ROCKE-3D, falling within the ensemble distribution on multiple measures. The model is fast enough that it enables large parameter surveys with hundreds to thousands of models, potentially enabling the efficient use of a 3D climate model in retrievals of future exoplanet observations. We describe our efforts to make ExoPlaSim accessible to non-modellers, including observers, non-computational theorists, students, and educators through a new Python API and streamlined installation through pip, along with online documentation.

astro-ph.EP↗

arXiv2021

Climate uncertainties caused by unknown land distribution on habitable M-Earths

A planet's surface conditions can significantly impact its climate and habitability. In this study, we use the 3D general circulation model ExoPlaSim to systematically vary dayside land cover on a synchronously rotating, temperate rocky planet under two extreme and opposite continent configurations, in which either all of the land or all of the ocean is centred at the substellar point. We identify water vapour and sea ice as competing drivers of climate, and we isolate land-dependent regimes under which one or the other dominates. We find that the amount and configuration of land can change the planet's globally averaged surface temperature by up to 20K, and its atmospheric water vapour content by several orders of magnitude. The most discrepant models have partial dayside land cover with opposite continent configurations. Since transit spectroscopy may permit observations of M-dwarf planets' atmospheres, but not their surfaces, these land-related climate differences likely represent a limiting uncertainty in a given planet's climate, even if its atmospheric composition is known. Our results are robust to variations in atmospheric CO2 concentration, stellar temperature, and instellation.

astro-ph.EP↗

arXiv2020

Climate Diversity in the Solar-Like Habitable Zone due to Varying Background Gas Pressure

A large number of studies have responded to the growing body of confirmed terrestrial habitable zone exoplanets by presenting models of various possible climates. However, the impact of the partial pressure of background gases such as N$_2$ has not yet been well-explored, despite the abundance of N$_2$ in Earth's atmosphere and the lack of constraints on its typical abundance in terrestrial planet atmospheres. We use PlaSim, a fast 3D climate model, to simulate many hundreds of climates around Sun-like stars with varying N$_2$ partial pressures, instellations, and surface characteristics to identify the impact of the background gas partial pressure on the climate. We find that the climate's response is nonlinear and highly sensitive to the background gas partial pressure. We identify pressure broadening of greenhouse gas (such as CO$_2$ and H$_2$O) absorption lines, amplification of warming or cooling by the water vapor greenhouse positive feedback, heat transport efficiency, and cooling through Rayleigh scattering as the dominant competing mechanisms that determine the equilibrium climate for a given N$_2$ partial pressure. Finally, we show that different amounts of N$_2$ should have a significant effect on broadband reflected light observations of terrestrial exoplanets around Sun-like stars.

astro-ph.EP↗

arXiv2020

A Large Repository of 3D Climate Model Outputs for Community Analysis and Postprocessing

As the number of known exoplanets has climbed into the thousands, efforts by theorists to understand the diversity of climates that may exist on terrestrial planets in the habitable zone have also accelerated. These efforts have ranged from analytical, to simple 0-D, 1-D, and 2-D models, to highly-sophisticated 3D global climate models (GCMs) adapted from Earth climate and weather models. The advantage of the latter is that fewer physical processes are reduced to simple parameterizations and empirical fits, and may instead be represented by physically-motivated algorithms. However, many such models are difficult to use, and take a long time to reach a converged state relative to simpler models, thereby limiting the amount of parameter space that can be explored. We use PlaSim, a 3D climate model of intermediate complexity, to bridge this gap, allowing us to produce hundreds to thousands of model outputs that have reached energy balance equilibrium at the surface and top of the atmosphere. We are making our model outputs available to the community in a permanent Dataverse repository (https://dataverse.scholarsportal.info/dataverse/kmenou). A subset of our model outputs can be used directly with external spectral postprocessing tools, and we have used them with petitRADTRANS and SBDART in order to create synthetic observables representative of fully-3D climates. Another natural use of this repository will be to use more-sophisticated GCMs to cross-check and verify PlaSim's results, and to explore in more detail those regions of the exoplanet parameter space identified in our PlaSim results as being of particular interest. We will continue to add models to this repository in the future, including more than 1000 models in the short- to medium-term future, expanding the diversity of climates represented therein.

astro-ph.EP↗

arXiv2020

Keys of a Mission to Uranus or Neptune, the Closest Ice Giants

Uranus and Neptune are the archetypes of "ice giants", a class of planets that may be among the most common in the Galaxy. They hold the keys to understand the atmospheric dynamics and structure of planets with hydrogen atmospheres inside and outside the solar system; however, they are also the last unexplored planets of the Solar System. Their atmospheres are active and storms are believed to be fueled by methane condensation which is both extremely abundant and occurs at low optical depth. This means that mapping temperature and methane abundance as a function of position and depth will inform us on how convection organizes in an atmosphere with no surface and condensates that are heavier than the surrounding air, a general feature of giant planets. Owing to the spatial and temporal variability of these atmospheres, an orbiter is required. A probe would provide a reference atmospheric profile to lift ambiguities inherent to remote observations. It would also measure the abundances of noble gases which can be used to reconstruct the history of planet formation in the Solar System. Finally, mapping the planets' gravity and magnetic fields will be essential to constrain their global composition, atmospheric dynamics, structure and evolution. An exploration of Uranus or Neptune will be essential to understand these planets and will also be key to constrain and analyze data obtained at Jupiter, Saturn, and for numerous exoplanets with hydrogen atmospheres.

astro-ph.IM↗

arXiv2020

Predicting the long-term stability of compact multiplanet systems

We combine analytical understanding of resonant dynamics in two-planet systems with machine learning techniques to train a model capable of robustly classifying stability in compact multi-planet systems over long timescales of $10^9$ orbits. Our Stability of Planetary Orbital Configurations Klassifier (SPOCK) predicts stability using physically motivated summary statistics measured in integrations of the first $10^4$ orbits, thus achieving speed-ups of up to $10^5$ over full simulations. This computationally opens up the stability constrained characterization of multi-planet systems. Our model, trained on $\approx 100,000$ three-planet systems sampled at discrete resonances, generalizes both to a sample spanning a continuous period-ratio range, as well as to a large five-planet sample with qualitatively different configurations to our training dataset. Our approach significantly outperforms previous methods based on systems' angular momentum deficit, chaos indicators, and parametrized fits to numerical integrations. We use SPOCK to constrain the free eccentricities between the inner and outer pairs of planets in the Kepler-431 system of three approximately Earth-sized planets to both be below 0.05. Our stability analysis provides significantly stronger eccentricity constraints than currently achievable through either radial velocity or transit duration measurements for small planets, and within a factor of a few of systems that exhibit transit timing variations (TTVs). Given that current exoplanet detection strategies now rarely allow for strong TTV constraints (Hadden et al., 2019), SPOCK enables a powerful complementary method for precisely characterizing compact multi-planet systems. We publicly release SPOCK for community use.

astro-ph.EP↗