Diameters and Temperatures VII: High-angular resolution measurements of Solar-type stars with the CHARA Array

We present interferometric measurements of angular diameters for 27 nearby solar-type stars obtained with the Precision Astronomical Visible Observations (PAVO) beam combiner at the CHARA Array. The sample spans a broad range of metallicities, includes several known exoplanet hosts, and covers evolutionary stages from the zero-age main sequence to mildly evolved subgiants. Uniform-disk and limb-darkened angular diameters were measured for each target and combined with bolometric fluxes and Gaia parallaxes to determine precise, model-independent stellar radii, effective temperatures, and luminosities. We achieve typical uncertainties of ∼ 1% in radius and ∼ 1.5% in effective temperature. Comparisons with multiple stellar evolutionary model grids yield mass and age estimates and enable assessment of grid-to-grid systematics, highlighting sensitivities to abundances, evolutionary stage, and proximity to grid boundaries. Our results provide empirical benchmarks for testing stellar evolutionary theory, refining surface brightness-color relations, and improving the characterization of exoplanet host stars.

Improvements to AI Systems:

1. Stellar Parameter Inference with Uncertainty Quantification

• Train a Bayesian neural network or Gaussian process regressor on the measured angular diameters, bolometric fluxes, and Gaia parallaxes to predict stellar radii, effective temperatures, and luminosities with calibrated 1% (radius) and 1.5% (temperature) uncertainties.

• The improved AI can directly output posterior distributions for these parameters from photometric/spectroscopic inputs, replacing traditional model-dependent fitting.

2. Evolutionary Model Grid Emulation and Systematics Detection

• Use the multi-grid comparisons (mass, age, metallicity) to train an ensemble model that learns grid-to-grid discrepancies (e.g., systematic offsets near grid boundaries or for subgiants).

• The improved AI can flag when a star’s inferred mass/age is unreliable due to model grid limitations, and automatically interpolate between grids with uncertainty penalties.

3. Surface Brightness–Color Relation Refinement

• Apply the new empirical benchmarks to train a deep learning model that maps multi-band photometry (e.g., Gaia, 2MASS, Tycho) to limb-darkened angular diameters, replacing linear/parametric fits.

• The improved AI can predict angular diameters for any unresolved star with sub-1% accuracy, enabling direct radius estimation for millions of stars without interferometry.

4. Exoplanet Host Star Characterization

• Fine-tune a transformer-based model on the exoplanet hosts in this sample to incorporate stellar radius and temperature priors into planetary radius and insolation flux calculations.

• The improved AI can automatically propagate stellar parameter uncertainties into exoplanet density and equilibrium temperature estimates, reducing systematic biases in habitability assessments.

5. Synthetic Stellar Population Generation

• Use the empirical radii, temperatures, and luminosities to train a generative adversarial network (GAN) that produces realistic stellar populations across metallicity and evolutionary stage, conditioned on the observed grid-to-grid systematics.

• The improved AI can generate synthetic catalogs for testing survey pipelines (e.g., PLATO, TESS) and for training other models on rare subgiant or metal-poor stars.

6. Automated Model Selection and Grid Boundary Warning

• Develop a reinforcement learning agent that, given a star’s measured parameters, selects the optimal evolutionary model grid and assigns confidence scores based on proximity to grid boundaries and abundance sensitivity.

• The improved AI can autonomously flag stars requiring custom stellar modeling (e.g., alpha-enhanced or low-metallicity) and recommend additional observations.

What the Improved AI System Can Do:

• Provide model-independent stellar radii, temperatures, and luminosities for any star from photometry alone, with known uncertainties.

• Detect and correct for systematic biases in stellar evolution codes, improving mass and age estimates for exoplanet host stars.

• Generate realistic synthetic stellar samples for testing exoplanet detection algorithms and stellar population synthesis.

• Automatically warn users when a star’s inferred properties are unreliable due to model grid limitations, reducing false positives in exoplanet characterization.

We present interferometric measurements of angular diameters for 27 nearby solar-type stars obtained with the Precision Astronomical Visible Observations (PAVO) beam combiner at the CHARA Array. The sample spans a broad range of metallicities, includes several known exoplanet hosts, and covers evolutionary stages from the zero-age main sequence to mildly evolved subgiants. Uniform-disk and limb-darkened angular diameters were measured for each target and combined with bolometric fluxes and Gaia parallaxes to determine precise, model-independent stellar radii, effective temperatures, and luminosities. We achieve typical uncertainties of about1 % in radius and about1.5 % in effective temperature. Comparisons with multiple stellar evolutionary model grids yield mass and age estimates and enable assessment of grid-to-grid systematics, highlighting sensitivities to abundances, evolutionary stage, and proximity to grid boundaries. Our results provide empirical benchmarks for testing stellar evolutionary theory, refining surface brightness-color relations, and improving the characterization of exoplanet host stars.