Comparison of Polarized Radiative Transfer Codes used by the EHT Collaboration

Ben Prather, Jason Dexter, Monika Mościbrodzka, Hung-Yi Pu, Thomas Bronzwaer, Jordy Davelaar, Ziri Younsi, Charles F. Gammie, Roman Gold, George N. Wong, Kazunori Akiyama, A. Alberdi, W. Alef, Juan Carlos Algaba, Richard Anantua, Keiichi Asada, Rebecca Azulay, U. Bach, Anne-Kathrin Baczko, David W. Ball, Mislav Baloković, John Barrett, Michi Bauböck, B. A. Benson, Dan Bintley, Lindy Blackburn, R. Blundell, Katherine L. Bouman, Geoffrey C. Bower, Hope Boyce, Michael Bremer, Christiaan D. Brinkerink, Roger Brissenden, S. Britzen, Avery E. Broderick, Dominique Broguière, Sandra Bustamante, Do‐Young Byun, J. E. Carlstrom, Chiara Ceccobello, Andrew Chael, Chi‐kwan Chan, Dominic O. Chang, Koushik Chatterjee, Shami Chatterjee, Ming‐Tang Chen, Xiaopeng Cheng, Ilje Cho, Pierre Christian, Nicholas S. Conroy, J. E. Conway, J. M. Cordes, T. M. Crawford, G. Crew, Alejandro Cruz-Osorio, Yuzhu Cui, Mariafelicia De Laurentis, Roger Deane, J. T. Dempsey, G. Desvignes, Vedant Dhruv, Sheperd S. Doeleman, Sean Dougal, Sergio A. Dzib, Ralph P. Eatough, Razieh Emami, H. Falcke, Joseph Farah, Vincent L. Fish, Ed Fomalont, H. Alyson Ford, Raquel Fraga-Encinas, William T. Freeman, Per Friberg, Christian M. Fromm, Antonio F. Fuentes, Peter Galison, Roberto García, Olivier Gentaz, Boris Georgiev, C. Goddi, Arturo I. Gómez-Ruiz, José L. Gómez, Minfeng Gu, Mark Gurwell, Kazuhiro Hada, Daryl Haggard, Kari Haworth, M. H. Hecht, Ronald Hesper, Dirk Heumann, Luis C. Ho, Paul T. P. Ho, Mareki Honma, Wu Jiang, Lei Huang, D. H. Hughes, Shiro Ikeda, C. M. Violette Impellizzeri

Abstract: Interpretation of resolved polarized images of black holes by the Event Horizon Telescope (EHT) requires predictions of the polarized emission observable by an Earth-based instrument for a particular model of the black hole accretion system. Such predictions are generated by general relativistic radiative transfer (GRRT) codes, which integrate the equations of polarized radiative transfer in curved spacetime. A selection of ray-tracing GRRT codes used within the EHT collaboration is evaluated for accuracy and consistency in producing a selection of test images, demonstrating that the various methods and implementations of radiative transfer calculations are highly consistent. When imaging an analytic accretion model, we find that all codes produce images similar within a pixel-wise normalized mean squared error (NMSE) of 0.012 in the worst case. When imaging a snapshot from a cell-based magnetohydrodynamic simulation, we find all test images to be similar within NMSEs of 0.02, 0.04, 0.04, and 0.12 in Stokes I, Q, U, and V respectively. We additionally find the values of several image metrics relevant to published EHT results to be in agreement to much better precision than measurement uncertainties.

Introduction

When the Event Horizon Telescope (EHT) produces an image of a black hole, scientists compare the observed data against millions of simulated images to determine the black hole's mass, spin, and magnetic field structure. These simulations rely on General Relativistic Radiative Transfer (GRRT) codes.

While previous studies validated how these codes handle total light intensity (Stokes I), polarized light (Stokes Q, U, V) is significantly more complex. Polarization carries vital information about magnetic fields but is sensitive to "numerical weather"—small errors in how a code handles curved spacetime, the parallel transport of light vectors, and plasma effects like Faraday rotation. This paper presents a rigorous cross-validation of six different simulation codes to ensure that theoretical predictions for polarized light are robust and consistent, backing the scientific interpretation of images like M87 and Sgr A.

Methodology and Datasets

The authors evaluated six independent ray-tracing GRRT codes: BHOSS, ipole, ipole-IL, grtrans, Odyssey, and RAPTOR. These codes solve the same fundamental equations but utilize different numerical schemes, such as varying integrators or methods for handling plasma physics.

The evaluation followed a three-tiered "stress test" of increasing complexity:

  1. Analytic Integration: Integrating the transfer equations with constant coefficients to check basic mathematical accuracy. This baseline ensures the fundamental solvers are functioning correctly.
  2. Thin-Disk Model: Simulating an opaque Novikov-Thorne disk around a spinning Kerr black hole. This tests the code's ability to handle light-bending (geodesics) and the parallel transport of polarization vectors in curved spacetime.
  3. GRMHD Snapshot: The most realistic test, using a "snapshot" from a 3D General Relativistic Magnetohydrodynamic (GRMHD) simulation of magnetized plasma (SANE model). This tests how codes interpolate fluid data and calculate complex plasma emission and absorption.
Table 1: Coefficients for the analytic integration tests, defining emission, absorption, and rotation parameters for the Stokes vector.

Key Findings

Stochastic vs. Systematic Errors

Most differences between the codes appear to be stochastic (random noise) rather than systematic biases. This is a positive result for the collaboration, as random errors tend to average out over the large libraries of simulations used for parameter estimation.

The Stokes V Challenge

Circular polarization (Stokes V) showed the highest disagreement, with a Normalized Mean Squared Error (NMSE) of 0.12. This is because V is often a tiny fraction of the total light and is highly sensitive to "Faraday conversion," where linear polarization is twisted into circular polarization by the plasma.

Precision vs. Accuracy

The codes agree with each other to a much higher precision than the actual EHT telescopes can currently measure. For example, code-to-code uncertainty in linear polarization is approximately 0.21%, while the detector uncertainty is roughly 1.35%. Furthermore, code variations would only affect mass estimates by about 0.4% and observer angle estimates by 0.8°—values far smaller than other astrophysical uncertainties.

Limitations and Future Work

While the codes are highly consistent, several areas for refinement remain. The current tests assume "fast light," where light travels across the black hole environment instantaneously; future work will need to compare "slow light" versions where the plasma moves during light transit. Additionally, the study noted that "cold" electrons in the disk can cause massive Faraday rotation, necessitating more sophisticated models for electron temperature and potentially accounting for anisotropic plasma emission.

In conclusion, this cross-validation demonstrates that the choice of GRRT simulation software does not bias the scientific interpretation of black hole images, ensuring the robustness of EHT results.

Summary updated

Abstract

Interpretation of resolved polarized images of black holes by the Event Horizon Telescope (EHT) requires predictions of the polarized emission observable by an Earth-based instrument for a particular model of the black hole accretion system. Such predictions are generated by general relativistic radiative transfer (GRRT) codes, which integrate the equations of polarized radiative transfer in curved spacetime. A selection of ray-tracing GRRT codes used within the EHT collaboration is evaluated for accuracy and consistency in producing a selection of test images, demonstrating that the various methods and implementations of radiative transfer calculations are highly consistent. When imaging an analytic accretion model, we find that all codes produce images similar within a pixel-wise normalized mean squared error (NMSE) of 0.012 in the worst case. When imaging a snapshot from a cell-based magnetohydrodynamic simulation, we find all test images to be similar within NMSEs of 0.02, 0.04, 0.04, and 0.12 in Stokes I, Q, U , and V respectively. We additionally find the values of several image metrics relevant to published EHT results to be in agreement to much better precision than measurement uncertainties.
BibTeX
@article{prather2023comparison,
  author = {Ben Prather and Jason Dexter and Monika Mościbrodzka and Hung-Yi Pu and Thomas Bronzwaer and Jordy Davelaar and Ziri Younsi and Charles F. Gammie and Roman Gold and George N. Wong and Kazunori Akiyama and A. Alberdi and W. Alef and Juan Carlos Algaba and Richard Anantua and Keiichi Asada and Rebecca Azulay and U. Bach and Anne-Kathrin Baczko and David W. Ball and Mislav Baloković and John Barrett and Michi Bauböck and B. A. Benson and Dan Bintley and Lindy Blackburn and R. Blundell and Katherine L. Bouman and Geoffrey C. Bower and Hope Boyce and Michael Bremer and Christiaan D. Brinkerink and Roger Brissenden and S. Britzen and Avery E. Broderick and Dominique Broguière and Sandra Bustamante and Do‐Young Byun and J. E. Carlstrom and Chiara Ceccobello and Andrew Chael and Chi‐kwan Chan and Dominic O. Chang and Koushik Chatterjee and Shami Chatterjee and Ming‐Tang Chen and Xiaopeng Cheng and Xiaopeng Cheng and Ilje Cho and Pierre Christian and Nicholas S. Conroy and J. E. Conway and J. M. Cordes and T. M. Crawford and G. Crew and Alejandro Cruz-Osorio and Yuzhu Cui and Mariafelicia De Laurentis and Roger Deane and J. T. Dempsey and G. Desvignes and Vedant Dhruv and Sheperd S. Doeleman and Sean Dougal and Sergio A. Dzib and Ralph P. Eatough and Razieh Emami and H. Falcke and Joseph Farah and Vincent L. Fish and Ed Fomalont and H. Alyson Ford and Raquel Fraga-Encinas and William T. Freeman and Per Friberg and Christian M. Fromm and Antonio F. Fuentes and Peter Galison and Roberto García and Olivier Gentaz and Boris Georgiev and C. Goddi and Arturo I. Gómez-Ruiz and José L. Gómez and Minfeng Gu and Mark Gurwell and Kazuhiro Hada and Daryl Haggard and Kari Haworth and M. H. Hecht and Ronald Hesper and Dirk Heumann and Luis C. Ho and Paul T. P. Ho and Mareki Honma and Wu Jiang and Lei Huang and D. H. Hughes and Shiro Ikeda and C. M. Violette Impellizzeri},
  title = {{Comparison of Polarized Radiative Transfer Codes used by the EHT Collaboration}},
  journal = {The Astrophysical Journal},
  year = {2023},
  doi = {10.3847/1538-4357/acc586},
}
Figure