Bernhard-1: An Eccentric Binary Periodically Obscured by its Misaligned Circumbinary Disk

Bernhard-1 is a proposed KH 15D-like circumbinary disk occultation (CBO) system, but its binary nature and disk geometry have not previously been confirmed. We present new optical and near-infrared spectroscopy together with multi-band photometric monitoring of the system. The radial velocities confirm that Bernhard-1 hosts a highly eccentric binary with e = 0.80 ± 0.09, confirming that the periodic photometric variability arises from occultation by a misaligned circumbinary disk. Joint modeling of the spectra and phase-dependent spectral energy distributions yields pre-main-sequence components with masses of ∼ 1.1 M⊙ and ∼ 0.8 M⊙. Combining stellar isochrones with the measured lithium abundance yields a system age of ∼ 10 Myr. Together with the spatial, astrometric, and metallicity properties of Bernhard-1, this suggests that Bernhard-1 is probably a member of the open cluster Dolidze 42. By combining the RV orbit with a semi-transparent occultation-screen model, we infer a disk–binary mutual inclination of roughly 50◦ or 130◦, with the degeneracy arising from the unknown disk rotation direction. This geometric method can be applied to any CBO system once radial velocity monitoring yields an orbital solution. The new light curves deviate from earlier model predictions, consistent with ongoing disk precession, while the phase-dependent Hα profiles indicate pulsed accretion near periastron. Bernhard-1 therefore joins KH 15D and Bernhard-2 as a rare spectroscopically confirmed CBO system.

Improvements to AI Systems:

1. Orbital-Disk Geometry Inference Engine

• Improvement: Train a neural network to directly invert radial velocity (RV) curves and photometric light curves into binary parameters (eccentricity, masses, inclination) and disk properties (semi-transparency, precession rate, mutual inclination) without requiring manual joint modeling.

• What it can do: Automatically classify candidate CBO systems from time-series data, resolve the disk rotation direction degeneracy (50° vs 130°) by incorporating asymmetric light-curve features, and predict future occultation timings for precessing disks.

2. Precession-Aware Light Curve Predictor

• Improvement: Develop a generative model (e.g., a variational autoencoder or diffusion model) trained on synthetic CBO light curves that includes disk precession as a latent variable. Use the observed deviation from earlier model predictions (as seen in Bernhard-1) to update the precession phase and rate in real time.

• What it can do: Provide continuously updated ephemerides for monitoring campaigns, flag anomalous deviations that indicate disk warping or fragmentation, and simulate long-term photometric behavior for target selection in surveys like LSST.

3. Pulsed Accretion Detector from Spectral Line Profiles

• Improvement: Build a transformer-based classifier that ingests phase-resolved Hα (and other Balmer lines) profiles to detect periastron-driven accretion bursts. Train on Bernhard-1, KH 15D, and Bernhard-2 to learn the spectral signatures of pulsed accretion vs. steady accretion.

• What it can do: Automatically identify new CBO candidates with eccentric binaries where accretion is modulated by disk occultation, and estimate mass accretion rates from line equivalent widths and velocity shifts without manual fitting.

4. Stellar Population Age-Dating via Multi-Wavelength SED + Lithium

• Improvement: Create a multi-task learning model that combines isochrone fitting (from SEDs) with lithium abundance measurements to jointly infer age, mass, and membership probability in open clusters. Use Bernhard-1’s successful Dolidze 42 membership as a training example.

• What it can do: Rapidly assign ages and cluster memberships to thousands of pre-main-sequence stars in Gaia and 2MASS catalogs, improving star formation history mapping and identifying coeval populations.

5. Semi-Transparent Occultation Screen Simulator

• Improvement: Implement a differentiable physics-based simulator (e.g., using JAX or PyTorch) that models a semi-transparent, misaligned circumbinary disk occulting a binary star. Train the simulator to match Bernhard-1’s observed light curves and RV data, then use it for inverse modeling via gradient descent.

• What it can do: Enable fast Bayesian inference of disk geometry (inclination, node angle, scale height) and binary orbit from sparse photometry, and generate synthetic training data for other AI models to detect CBOs in large surveys.

6. Automatic CBO Candidate Finder from Time-Series Photometry

• Improvement: Use a recurrent or temporal convolutional network trained on Bernhard-1, KH 15D, and Bernhard-2 light curves (with known periodic dips and precession) to distinguish CBOs from eclipsing binaries, exoplanet transits, and stellar spots.

• What it can do: Scan light curves from TESS, Kepler, and future missions to flag high-probability CBO systems, prioritizing them for RV follow-up and reducing human inspection time by orders of magnitude.

Bernhard-1 is a proposed KH 15D-like circumbinary disk occultation (CBO) system, but its binary nature and disk geometry have not previously been confirmed. We present new optical and near-infrared spectroscopy together with multi-band photometric monitoring of the system. The radial velocities confirm that Bernhard-1 hosts a highly eccentric binary with e = 0.80 plus or minus 0.09, confirming that the periodic photometric variability arises from occultation by a misaligned circumbinary disk. Joint modeling of the spectra and phase-dependent spectral energy distributions yields pre-main-sequence components with masses of about 1.1,M and about 0.8,M. Combining stellar isochrones with the measured lithium abundance yields a system age of about 10 Myr. Together with the spatial, astrometric, and metallicity properties of Bernhard-1, this suggests that Bernhard-1 is probably a member of the open cluster Dolidze 42. By combining the RV orbit with a semi-transparent occultation-screen model, we infer a disk--binary mutual inclination of roughly 50 or 130, with the degeneracy arising from the unknown disk rotation direction. This geometric method can be applied to any CBO system once radial velocity monitoring yields an orbital solution. The new light curves deviate from earlier model predictions, consistent with ongoing disk precession, while the phase-dependent H alpha profiles indicate pulsed accretion near periastron. Bernhard-1 therefore joins KH 15D and Bernhard-2 as a rare spectroscopically confirmed CBO system.