Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Validation and Calibration of Semi-Analytical Models for the Event Horizon Telescope Observations of Sagittarius A*

Ali SaraerToosi, Avery Broderick

TL;DR

The study tackles the challenge of expensive forward-modeling for Event Horizon Telescope data by introducing ALINet, a fast generative model that produces radiatively inefficient accretion flow images as functions of physical parameters such as the spin $a$, inclination cosine $\mu$, and disk geometry. Integrated with Themis, ALINet enables Bayesian parameter estimation while incorporating unmodeled physics like interstellar scattering and intrinsic variability through systematic error budgets, validated on mock Sgr A* data and GRMHD-based tests. Key contributions include demonstrating self-consistency and interpolation validity of ALINet within Themis for RIAF images, quantifying scattering-induced posterior broadening, and assessing robustness to model misspecification via GRMHD scenarios that reveal multimodal posteriors and the relative resilience of MAD-like configurations. The findings advance fast, uncertainty-aware inferences of black hole properties from EHT observations and provide practical guidance for incorporating scattering and variability into parameter constraints under model misspecification. Overall, the work bridges fast surrogate image generation with rigorous Bayesian calibration, improving reliability of Sgr A* inferences in the presence of observational and physical uncertainties.

Abstract

The Event Horizon Telescope (EHT) enables the exploration of black hole accretion flows at event-horizon scales. Fitting ray-traced physical models to EHT observations requires the generation of synthetic images, a task that is computationally demanding. This study leverages \alinet, a generative machine learning model, to efficiently produce radiatively inefficient accretion flow (RIAF) images as a function of the specified physical parameters. \alinet has previously been shown to be able to interpolate black hole images and their associated physical parameters after training on a computationally tractable set of library images. We utilize this model to estimate the uncertainty introduced by a number of anticipated unmodeled physical effects, including interstellar scattering and intrinsic source variability. We then use this to calibrate physical parameter estimates and their associated uncertainties from RIAF model fits to mock EHT data via a library of general relativistic magnetohydrodynamics models.

Validation and Calibration of Semi-Analytical Models for the Event Horizon Telescope Observations of Sagittarius A*

TL;DR

The study tackles the challenge of expensive forward-modeling for Event Horizon Telescope data by introducing ALINet, a fast generative model that produces radiatively inefficient accretion flow images as functions of physical parameters such as the spin , inclination cosine , and disk geometry. Integrated with Themis, ALINet enables Bayesian parameter estimation while incorporating unmodeled physics like interstellar scattering and intrinsic variability through systematic error budgets, validated on mock Sgr A* data and GRMHD-based tests. Key contributions include demonstrating self-consistency and interpolation validity of ALINet within Themis for RIAF images, quantifying scattering-induced posterior broadening, and assessing robustness to model misspecification via GRMHD scenarios that reveal multimodal posteriors and the relative resilience of MAD-like configurations. The findings advance fast, uncertainty-aware inferences of black hole properties from EHT observations and provide practical guidance for incorporating scattering and variability into parameter constraints under model misspecification. Overall, the work bridges fast surrogate image generation with rigorous Bayesian calibration, improving reliability of Sgr A* inferences in the presence of observational and physical uncertainties.

Abstract

The Event Horizon Telescope (EHT) enables the exploration of black hole accretion flows at event-horizon scales. Fitting ray-traced physical models to EHT observations requires the generation of synthetic images, a task that is computationally demanding. This study leverages \alinet, a generative machine learning model, to efficiently produce radiatively inefficient accretion flow (RIAF) images as a function of the specified physical parameters. \alinet has previously been shown to be able to interpolate black hole images and their associated physical parameters after training on a computationally tractable set of library images. We utilize this model to estimate the uncertainty introduced by a number of anticipated unmodeled physical effects, including interstellar scattering and intrinsic source variability. We then use this to calibrate physical parameter estimates and their associated uncertainties from RIAF model fits to mock EHT data via a library of general relativistic magnetohydrodynamics models.

Paper Structure

This paper contains 15 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Astrophysical Models
  3. RIAF Models
  4. Image Models
  5. Data Generation and Analysis
  6. Data Generation
  7. Analysis
  8. Self Consistency Tests
  9. ALINet Self Test
  10. RIAF Test
  11. Scattered RIAF Test
  12. Robustness to Model Misspecification Tests
  13. GRMHD Simulations
  14. Scattered Mean GRMHD
  15. Scattered Variable GRMHD

Figures (5)

  • Figure 1: ALINet Self test truth image vs. best fit image from Themis, and joint posteriors for a representative ALINet self test. Contours show $68.3\%$, $95.4\%$, and $99.7\%$ cumulative probability regions, indicating 1, 2, and 3$\sigma$ confidence levels, respectively. Gold points indicate the truth values.
  • Figure 2: RIAF test Truth image vs. best fit image from Themis, and parameter fits triangle plots. The gold circles in the triangle plot denote the truth values.
  • Figure 3: Scattered RIAF test Truth image, scattered image to which fitting is done, best fit image from Themis, and parameter fits triangle plots. The gold circles in the triangle denote the truth values. The threshold noise used is $7$ mJy.
  • Figure 4: Violin plots for three test cases for scattered RIAF test (on the left), mean GRMHD (middle), and variable GRMHD (right). The X shows the truth value. The violins show the resulting fitted distributions from Themis. The bars in each distribution show minimum, maximum, and the median values of the distribution. In each column, the first two from the left are MADs, and the third is a SANE.
  • Figure 5: Scattered GRMHD test including truth image, scattered image to which fitting is done, best fit to mean GRMHD data from Themis (top), truth image and best fit modes (two modes) to variable GRMHD data from Themis (bottom). We used a threshold noise of $7$mJy and $20$mJy for fitting to the mean GRMHD and variable GRMHD case, respectively.