Building Toolbox to Characterize the NEID Earth Twin Survey Detection Efficiency
PI: Suvrath Mahadevan (Astronomy & Astrophysics)
Team: PI: Suvrath Mahadevan (Astronomy & Astrophysics) Co-I (methodology advisor): Eric Ford (Astronomy & Astrophysics, ECoS, ICDS)
Level of effort: 3-6 months of effort from a postdoc or assistant research professor. Remainder of postdoctoral researcher’s salary will be supported by the Center for Exoplanets and Habitable Worlds if one of the center postdocs is selected for this.
Summary Sentence: This project aims to develop performant, parallelized statistical tools to characterize the detection efficiency of the NEID Earth Twin Survey as a function of the mass, orbital period and eccentricity of exoplanets orbiting the target stars.
Description: Detecting and characterizing potentially Earth-like planets is a priority of the recent Astronomy & Astrophysics Decadal Survey. Ground-based observations such as the NEID survey play a critical role in identifying which stars host planetary systems worth detailed characterization with current and future space-based observatories (Crass et al. 2021). Penn State built the NEID spectrograph currently searching for rocky planets orbiting nearby sun-like stars (Gupta et al. 2021). In addition to discovering exciting exoplanets (Gupta et al. 2025), the survey was designed to support statistical studies of the population of exoplanets as a function of their mass, orbital period and eccentricity. Realizing that goal will require combining modern statistical tools (e.g., approximate Bayesian computing, simulation-based inference) with an accurate model for the survey detection efficiency as a function of stellar, planetary, and orbital properties (e.g., Hsu et al. 2019). Existing packages (e.g., radvel; Fulton et al. 2018) lack the ability to model multiple outputs (e.g., estimated radial velocity and stellar variability indicators) which are crucial for achieving the highest sensitivity to low-mass planets. Recent progress in applying multi-output Gaussian processes to analyze Doppler exoplanet survey time series appears to have improved the sensitivity of surveys, but so far the extent of these improvements in survey sensitivity are not well quantified (Hara & Ford 2023).
In this project, an ICDS junior researcher would develop an open-source, performant, parallelized pipeline that combines both a forward model for generating simulated survey data with a fast method for estimating the significance of planet detections. We can provide code snippets to generate simulated data (Gupta et al. 2024), to perform uncertainty quantification using Octofitter (Gilbertson et al. 2020; Thompson et al. 2023), and to perform Bayesian model comparison (Nelson et al. 2020). The Junior Researcher would integrate these into a single Julia pipeline for calculating NETS’s detection efficiency as a function of stellar, planetary, and survey design properties, while accounting for the impact of offsets in the measured velocity and other observables (e.g., due to changes in instrument and calibration process). They will also implement unit/integration tests using simulated data, and document and release their code so that other research projects and teams can apply it to other surveys and/or extend it to incorporate additional exoplanet detection methods.
Expertise/skills of interest:
– Statistics: Periodograms; Bayesian Statistics: Uncertainty Quantification, Approximate Bayesian Computing and/or Simulation-based Inference
-Astronomy: Exoplanets, Radial velocity method.
– Programming experience using Julia and Python or substantial experience with one and willingness to learn the other.
-Familiarity with the following would be helpful, but is not required: MCMC, Gaussian Processes regression, Bayesian model comparison.
Expectations:
-Postdoc or assistant research professor with experience and/or training in Astronomy & Astrophysics, Physics, or a related field.
-Write code organized into small functions with appropriate documentation and unit/integration tests that will be released as open-source software.
-Produce Jupyter notebooks, scripts, and project environment files that support conclusions and make results readily reproducible.
-Participate in weekly group meetings of Mahadevan and Ford groups and provide project updates (~2 hour/week). Participare and listen in to NEID Science Team telecons (~1 hour/2 weeks).
Goal: Characterize the detection sensitivity of the NEID Earth Twin Survey as a function of planet mass and orbital. Develop a toolbox that can support a future funding proposal to NASA ROSES and/or the NSF AAG call.
Specific Objectives:
1. Develop a module providing a forward model for generating simulated NETS datasets, including effects of both planetary systems and stellar variability.
2. Develop a module to compute Bayesian periodograms for global search of simulated NETS datasets that makes use of both estimated RV and one to a few stellar variability indicators.
3. Develop a module to estimate the Bayesian evidence for models with different numbers of planets, while including the possibly of correlated “noise” due to intrinsic stellar variability using a Gaussian Process noise model.
4. Optimize and parallelize above code as necessary to allow for Monte Carlo simulations of surveys.
5. Develop a module to compute detection efficiency for each star in the survey as a function of planetary and orbital parameters.
6. Begin incorporating above into an Approximate Bayesian Computing or Simulation-base Inference framework for characterizing the exoplanet population. [if time permits]
Engagement: Mahadevan has existing collaborations with ICDS co-hire Ford and is interested in exploring future collaborations with other ICDS co-hires (Tak, incoming Hu & Patil). His research group would benefit from the expertise of an ICDS junior researcher with expertise in statistical characterization of exoplanet populations.
Connection to ICDS: This project combines modern statistical and computational methods for Bayesian uncertainty quantification and model comparison to quantify the sensitivity and robustness of surveys to discover and characterize low-mass planets; thus it synthesizes expertise in exoplanets and some of the most precise planet search instrumentation – developed at Penn State – to make these measurements, with advanced statistical approaches to address questions of scientific importance.