Article by: Andacs Robert Eugen, on 02 June 2023, at 10:22 am PDT

The groundbreaking observations of the Event Horizon Telescope (EHT) in 2017 allowed scientists to capture the first-ever horizon-scale images of black holes. The images of the black hole in the M87 galaxy and the Galactic Center black hole, Sagittarius A*, revealed a bright ring of emission surrounding a central brightness depression. These images were reconstructed using model-agnostic imaging algorithms, which were successful in capturing the ring-like morphology of the black holes. However, the sparse interferometric coverage of the EHT array posed challenges in accurately reconstructing the images. To overcome this limitation, a novel image reconstruction algorithm called Principal-Component Interferometric Modeling (PRIMO) has been developed.

PRIMO leverages machine learning and Principal Component Analysis (PCA) to reconstruct high-resolution images of black holes from sparse interferometric data. The algorithm utilizes a large library of high-fidelity, high-resolution general relativistic magnetohydrodynamic (GRMHD) simulations to train the model. PCA is applied to the simulated images, resulting in an orthogonal basis of image components that capture the correlations in Fourier space. PRIMO then uses a Markov Chain Monte Carlo (MCMC) approach to sample the space of linear combinations of these PCA components, producing reconstructions that match the EHT data.

Compared to previous image reconstructions using general-purpose imaging algorithms, PRIMO offers superior resolution and reveals finer details of the black hole's structure. By employing machine learning to fill in Fourier space and using a Butterworth filter instead of a Gaussian filter to blur the image, PRIMO achieves better utilization of the EHT array's resolution and dynamical range. The resulting images exhibit a substantially narrower ring and enhanced ability to discern finer features. This improved resolution is attributed to PRIMO's ability to learn correlations between low-frequency and high-frequency structures from the GRMHD simulations.

The reconstructed image obtained using PRIMO for the M87 black hole based on the 2017 EHT data closely resembles the image published by the EHT collaboration in 2019. Both images exhibit a bright ring of emission, a central brightness depression, and a north-south brightness asymmetry. However, the PRIMO image demonstrates a significantly narrower ring width and reveals more intricate details. Comparisons of the PRIMO reconstruction with the EHT data in terms of Fourier amplitudes and closure phases show excellent agreement, with minimal residual structures.

A parameter study using different total fluxes and varying numbers of PCA components confirms the robustness of the main image features, such as the ring presence, size, and brightness depression. Changes in total flux and the number of PCA components do not significantly alter the morphology of the image, highlighting the stability and reliability of the PRIMO technique. The analysis consistently favors a black hole spin axis pointing away from the observer at 17° away from the line of sight, consistent with previous findings.