Skip to main content
SearchLoginLogin or Signup

A machine learning approach for black hole reverberation modeling

Presentation #107.68 in the session Stellar/Compact Objects - Poster Session.

Published onMay 03, 2024
A machine learning approach for black hole reverberation modeling

X-ray Reverberation Mapping, where X-rays produced close to the black hole reverberate off inspiraling gas, allows us to map out scales close to the event horizon – orders of magnitude better than the resolution of our telescopes. Reverberation lags result from the light travel time difference between the direct coronal emission and the reflected disk emission, and therefore the lag properties probe the disk-corona geometry. We adopt a novel, machine learning approach to black hole reverberation modeling that treats the cross spectra (a spectral timing product involving the time lags) as “images” with photon energy and Fourier frequency as axes. With simulated images using a reverberation mapping model called reltrans, we train a basic convolutional neural network and find the two key parameters, coronal height and black hole mass, can be perfectly recovered. The predictions of the trained neural network for real data of MAXI J1820+070 are also close to the results found through conventional modeling method.

No comments here