Trained on 111,000 simulated images, the model extracts the n=1 ring hidden beneath the main emission with a correlation of approximately 0.99, with a view to future missions such as BHEX.
Very long-baseline interferometry could one day photograph the ring of photons of a black hole. Its substructure, known as n=1, is valuable: its radial profile and its variations in brightness around the ring provide information on the rotation of the black hole. But in observations, it is superimposed on the main ring n=0.
Researchers trained a convolutional lattice to separate the two components, based on 111,000 synthetic images produced with eht-imaging. The lattice retrieves the radial and angular profiles of the n=1 ring with an average normalized cross-correlation of about 0.99, and reports no false positives when the image does not contain any.
More discrepancies on realistic simulations
Tested on images from KerrBAM and relativistic magnetohydrodynamic simulations, it reproduces the radial shape of the ring well, but less faithfully its intensity. The authors see it as a basis for analysing the high-resolution images expected from missions such as the Black Hole Explorer (BHEX).
Read the full article on the original source
Go to the source →