Hybrid Very Long Baseline Interferometry Imaging and Modeling with themis

Avery E. Broderick*, Dominic W. Pesce, Paul Tiede, Hung Yi Pu, Roman Gold

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

39 Citations (Scopus)


Generating images from very long baseline interferometric observations poses a difficult, and generally not unique, inversion problem. This problem is simplified by the introduction of constraints, some generic (e.g., positivity of the intensity) and others motivated by physical considerations (e.g., smoothness, instrument resolution). It is further complicated by the need to simultaneously address instrumental systematic uncertainties and sparse coverage in the u-v plane. We report a new Bayesian image reconstruction technique in the parameter estimation framework Themis that has been developed for the Event Horizon Telescope. This has two key features: First, the full Bayesian treatment of the image reconstruction makes it possible to generate a full posterior for the images, permitting a rigorous and quantitative investigation into the statistical significance of image features. Second, it is possible to seamlessly incorporate directly modeled features simultaneously with image reconstruction. We demonstrate this second capability by incorporating a narrow, slashed ring in reconstructions of simulated M87 data in an attempt to detect and characterize the photon ring. We show that it is possible to obtain high-fidelity photon ring sizes, enabling mass measurements with accuracies of 2%-5% that are essentially insensitive to astrophysical uncertainties, and creating opportunities for precision tests of general relativity.

Original languageEnglish
Article number9
JournalAstrophysical Journal
Issue number1
Publication statusPublished - 2020 Jul 20
Externally publishedYes

ASJC Scopus subject areas

  • Astronomy and Astrophysics
  • Space and Planetary Science


Dive into the research topics of 'Hybrid Very Long Baseline Interferometry Imaging and Modeling with themis'. Together they form a unique fingerprint.

Cite this