Observing---and Imaging---Active Galactic Nuclei with the Event Horizon\n Telescope

Vincent L. Fish, Kazunori Akiyama, Katherine L. Bouman, Andrew Chael, Michael D. Johnson, Sheperd S. Doeleman, Lindy Blackburn, J. F. C. Wardle, William T. Freeman, The Event Horizon Telescope CollaborationarXiv 2016 · DOI · arXiv

Abstract: Originally developed to image the shadow region of the central black hole in Sagittarius A* and in the nearby galaxy M87, the Event Horizon Telescope (EHT) provides deep, very high angular resolution data on other AGN sources too. The challenges of working with EHT data have spurred the development of new image reconstruction algorithms. This work briefly reviews the status of the EHT and its utility for observing AGN sources, with emphasis on novel imaging techniques that offer the promise of better reconstructions at 1.3 mm and other wavelengths.

Introduction: The Challenge of Sparse Interferometry

While the Event Horizon Telescope (EHT) is famous for capturing the first images of black hole shadows, it is also a powerful tool for studying the high-energy jets of Active Galactic Nuclei (AGN). The EHT operates using Very Long Baseline Interferometry (VLBI) at a wavelength of 1.3 mm, linking telescopes across the globe to create a "virtual telescope" the size of the Earth.

However, this telescope is extremely "sparse." Because there are only a handful of observatory sites, the EHT only samples a tiny fraction of the spatial frequencies (the Fourier transform) of the actual image. This sparse $(u,v)$ coverage leads to significant artifacts (sidelobes) when using traditional imaging methods.

The sparse $(u,v)$ coverage of the EHT array, illustrating the limited sampling of spatial frequencies which necessitates advanced reconstruction algorithms.

Furthermore, at 1.3 mm, atmospheric phase noise is chaotic, making it nearly impossible to calibrate the absolute phase of the signal. To resolve features like a black hole shadow, researchers require "super-resolution"—the ability to see details smaller than the nominal diffraction limit of the telescope.

Beyond CLEAN: Regularized Maximum Likelihood

Traditionally, radio astronomers use the CLEAN algorithm, a "greedy" deconvolution method that assumes the sky is composed of point sources. While effective for many arrays, CLEAN struggles with the EHT’s specific challenges. Consequently, the EHT collaboration is moving toward Regularized Maximum Likelihood (RML) methods. Instead of "cleaning" a dirty map, these "forward-imaging" methods search for an image that best fits the observed data while satisfying specific "priors" or assumptions about the image structure.

Key RML methodologies include:

  • Polarimetric Maximum Entropy Method (MEM): Uses "entropy" as a regularizer to find the simplest image consistent with the data, extended to handle polarization (Stokes Q and U) for mapping magnetic fields.
  • Bi-Spectrum Sparse Modeling (BSSpM): Employs "compressed sensing" (specifically $\ell_1$ norm and Total Variation) to favor sparse images with sharp edges. It fits to closure phases, a robust observable that cancels out atmospheric noise.
  • CHIRP (Continuous High-resolution Image Reconstruction using Patch priors): Developed by a team including Katie Bouman, CHIRP uses a library of "natural" and astronomical image patches to learn a probability distribution of likely features, helping to fill in data gaps.

Comparative Performance and Validation

The authors evaluated these new algorithms using simulated EHT data of black hole shadows and real-world VLBA data of the quasar 3C 279. These tests demonstrated that the EHT’s imaging challenges are more akin to those of optical interferometry than traditional low-frequency radio astronomy.

Comparison of reconstruction performance. The plot (left) shows Normalized Root Mean Square Error (NRMSE) vs. restoring beam size, indicating that MEM and other RML methods can achieve lower error than CLEAN. The image grid (right) compares model-based reconstructions with various algorithmic outputs.

Key results from these comparisons include:

  • Super-resolution is Attainable: BSSpM can recover features at roughly 40% of the nominal fringe spacing, surpassing the theoretical limit.
  • Closure Phases are Essential: Fitting to the closure phase is the only reliable way to reconstruct images at 1.3 mm due to atmospheric corruption of absolute phase.
  • CHIRP Performance: In unsupervised comparisons, CHIRP consistently outperformed CLEAN and MEM, especially for extended source structures and low signal-to-noise ratios.

Conclusion and Future Directions

While RML methods offer significant advantages, they are computationally expensive and the choice of "regularizers" still involves human judgment, which can introduce bias. The inclusion of the Atacama Large Millimeter/submillimeter Array (ALMA) into the EHT has been the most critical factor for improving sensitivity. Looking forward, observations at 0.87 mm will provide even higher angular resolution, further pushing the boundaries of our ability to image the most extreme environments in the universe.

Summary updated

Abstract

Originally developed to image the shadow region of the central black hole in\nSagittarius A* and in the nearby galaxy M87, the Event Horizon Telescope (EHT)\nprovides deep, very high angular resolution data on other AGN sources too. The\nchallenges of working with EHT data have spurred the development of new image\nreconstruction algorithms. This work briefly reviews the status of the EHT and\nits utility for observing AGN sources, with emphasis on novel imaging\ntechniques that offer the promise of better reconstructions at 1.3 mm and other\nwavelengths.\n
BibTeX
@misc{fish2016observing,
  author = {Vincent L. Fish and Kazunori Akiyama and Katherine L. Bouman and Andrew Chael and Michael D. Johnson and Sheperd S. Doeleman and Lindy Blackburn and J. F. C. Wardle and William T. Freeman and The Event Horizon Telescope Collaboration},
  title = {{Observing---and Imaging---Active Galactic Nuclei with the Event Horizon\textbackslash\{\}n Telescope}},
  year = {2016},
  eprint = {1607.03034},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1607.03034},
}
Figure