The First Image of a Black Hole

This guest post was written by Daniel Palumbo, a Ph.D. student at Harvard University and member of the Event Horizon Telescope team. He received his BS in Physics from MIT in 2018. He develops imaging methods for the Event Horizon Telescope and pioneered the imaging parameter survey used to produce the final images of M87.

Title: First M87 Event Horizon Telescope Results I: The Shadow of the Supermassive Black Hole

Authors: Event Horizon Telescope Collaboration

Status: Accepted in ApJL with open access


In April 2017, the Event Horizon Telescope (or EHT), a global interferometric network of radio dishes, observed the bright center of the galaxy M87. The EHT team studied four days of intense observations and spent months calibrating nearly a petabyte of data! What we found was astounding: an asymmetric ring-like structure consistent with the emission expected to outline the shadow of a black hole event horizon! More than six billion solar masses were confined to an area on the sky equivalent to roughly the size of an atom held at arm’s length, or a physical distance of thrice the length of our solar system.

Today we examine six papers, released together by the EHT collaboration; Paper I tells the connected story of the EHT result, while papers II-VI step through the instrument, data processing, imaging, comparison to theoretical work, and mass measurement, respectively. We’ll examine what kinds of structure we might expect to see, what the EHT actually saw, and what these results mean for Einstein’s theory of general relativity (spoiler alert: Einstein has nothing to worry about yet!).

It’s a tale as old as general relativity, so let’s take it from the top.

Figure 1 (Figure 3 in Paper I): Images of the black hole candidate at the core of M87 produced by the Event Horizon Telescope. The image is consistent with emission from matter near the event horizon of a Kerr black hole with a mass of approximately 6.5 billion solar masses.

What is a black hole, and what do we expect one to look like?

The punchline of Einstein’s theory of general relativity is a set of equations relating the matter and energy in a volume of spacetime to the curvature in that region of spacetime. A black hole is an allowed solution to these equations: in its simplest case, a spherical gravitator with all of its mass located at a single point. The associated metric (basically, a “ruler” that tells you how to measure curved space) has a special radius within which not even light can escape: the event horizon. Understanding the behavior of matter at this boundary has been a primary focus of research in quantum physics, theoretical astrophysics, mathematics, and observational astronomy for more than a hundred years.

Black holes are found throughout astrophysics with a variety of masses. Stellar-mass black holes, which may result from the collapse of massive stars, have been strongly indicated by X-ray observations and LIGO detections. Meanwhile, supermassive black holes (containing up to tens of billions of solar masses) are expected to exist in the centers of most galaxies, powering active galactic nuclei or jets observed on galactic scales. Despite the darkness of a black hole’s surface, accreting matter is accelerated and heated during infall. This heat is released as light, producing incredibly luminous cores in some galaxies. The core of M87 is believed to be a supermassive black hole fed by a slow flow of plasma, and it is expected to power the highly collimated and luminous jet observed to billow from the center of M87.

The material near the event horizon radiates light that is strongly lensed (or distorted) by the gravitational field of the black hole. The EHT observes this radiation, effectively imaging the silhouette of the black hole event horizon amidst a bath of light. We only see light that’s lensed towards us and, since the infalling material is rotating relativistically fast (likely very close to the speed of light), the radiation is relativistically Doppler boosted as well. The result is an asymmetric ring-like feature surrounding a black central region several times larger than the event horizon — the shadow. The shadow is thus a smoking gun for a black hole as we understand it, and imaging at the shadow scale is a crucial test of GR.

How do you take a picture of a black hole, and what the did the EHT see?

Though all black holes have a well-defined shadow, there are very few observationally feasible targets for imaging the shadow. Stellar-mass black holes, though believed to be abundant, have far too small of an angular size to image, as do most supermassive black holes. However, the center of the Milky Way contains a supermassive black hole (Sagittarius A*/Sgr A*) a mere 26,000 light years away from us, close enough for the EHT to observe. Sgr A* has a mass of roughly 4 million solar masses, corresponding to an angular shadow size of about 50 microarcseconds (roughly the size of an atom held at arm’s length). Similarly, though more than 55 million light years away, the black hole in M87 is expected to have a mass of a few billion solar masses, making it just large enough that the EHT can expect to resolve its shadow. The mass of M87 was a matter of contention, believed to be between about 2.5 and 6.2 billion solar masses (measured from gas and stellar dynamical studies, respectively). Though the EHT observed both Sgr A* and M87, the first results released this week pertain only to M87. The EHT image reveals a shadow with a diameter of 40 microarcseconds, corroborating the larger M87 mass.

Multiple factors complicate these observations, forcing the EHT to observe at very specific frequencies. The material surrounding a black hole is expected to be optically thick to shadow-scale radiation (primarily synchrotron radiation) at most radio frequencies. At frequencies greater than 230 GHz, though, the EHT can begin to see through this material to directly view the black hole shadow. Unfortunately, the small size of the shadow requires resolution that is difficult to achieve with any telescope capable of observing at these frequencies. At 230 GHz, an Earth-sized telescope is necessary to achieve a nominal resolution similar to the expected shadow size of M87.

Since such a telescope isn’t going to exist in the near future, the best we can do is build something mathematically similar with Very-Long-Baseline Interferometry, or VLBI, building a synthetic dish with an effective diameter equal to that of the Earth. In this technique, multiple observatories focus on the same source, and the observed radiation is then correlated between every pair of telescopes. These correlations contain numerous sources of error, from antennas to the atmosphere, and Paper III discusses how sensitive telescopes like ALMA can be used to remove these effects in VLBI observables. 

Reconstructing such a complex image is clearly not a simple task, and the first EHT results demonstrate three algorithms for doing so. Algorithms are trained on known synthetic images to improve output on the real data. This technique produces ensembles of similar images with good agreement with the data (“top sets”), representing the uncertainty in the underlying source structure inherent to the limited sampling of the image provided by a virtual dish. For each imaging method, a set of imaging parameters is selected based on having the best agreement to the synthetic training set, which together form the image of M87 produced by the Event Horizon Telescope.

Figure 2 (Figure 11 in Paper IV): Fiducial images on each day of observations from each imaging pipeline. A consistent asymmetric ring feature is recovered by all pipelines on all days, with hints of time variability along the brighter southern edge.

The EHT imaging pipelines represent both traditional interferometric imaging (the CLEAN algorithm, implemented in DIFMAP) and modern Regularized Maximum Likelihood (RML) methods (implemented separately in SMILI and eht-imaging). The RML methods tend to produce sharper images, and have better agreement with the data; however, all three pipelines produce images that are qualitatively similar in reconstructing an asymmetric ring-like structure brighter on the southern edge (Figure 2).

What else can we learn from the data? 

General relativity predicts the size and shape of the shadow to have a strong dependence on mass and a very weak (± 4%) dependence on spin. So, using a distance to M87 measured with other methods, the size of the black hole shadow primarily corresponds to a measurement of the mass. A simple ring profile measurement was applied to the “top sets” from each imaging method (as in Figure 3), producing a narrow distribution of shadow sizes around 40 microarcseconds, consistent with a Kerr black hole of mass near 6.5 billion solar masses.

Figure 3 (Figure 26 in Paper IV): Example of the ring fitting procedure applied to images from each pipeline. The fit measures the ring diameter, thickness, asymmetry, and shadow depth. Due to the difference in the effective resolution between the pipelines (the CLEAN images are restored with a 20 microarcsecond beam, unlike RML methods), the ring thickness measurements are generally not consistent among pipelines. However, the ring sizes track very closely across days and pipelines.

However, theoretical models can also be compared directly to the data, without imaging at all. Paper V examines a library of general relativistic magnetohydrodynamic (GRMHD) simulations of accretion flows at M87. Doppler-boosted synchrotron emission of a hot plasma orbiting a Kerr black hole is found to be consistent with the observations (as visualized in Figure 4). The small-scale structure is compared to existing observations and models of the jet observed to launch from M87. While there are non-spinning models that are good fits to the EHT data, these are rejected after fitting based on the observed power within the jet. Alternatives to a Kerr black hole are briefly considered as possible explanations for the observed structure; sufficiently fine-tuned exotic objects could produce the same unpolarized image, but these will be constrained even further by polarimetric follow-up by the EHT.

Figure 4 (Figure 1 in Paper V): A comparison of the EHT consensus image of M87 from April 6th (left) to a GMRHD simulation (middle) blurred to the nominal resolution of the EHT (right). Details of the filamentary components of accretion or jet structures are largely lost in the context of EHT observations, while the bulk feature of asymmetric emission around a shadow is consistent with the observed image.

Finally, Paper VI presents fits of GRMHD simulations and asymmetric crescent models directly to the data. The crescents are calibrated to snapshots of GRMHD simulations from Paper V, connecting fits of the shape of the black hole with measurements of its mass. The crescent fit parameters find a black hole diameter of 42 ± 3 microarcseconds, consistent with the estimate taken by fitting snapshots of GRMHD simulations directly to the data, and further, consistent with the mass estimate from image-domain ring fits (as shown in Figure 5).

Figure 5 (Figure 19 in Paper VI): Mass estimates of the core of M87 based on direct fitting of geometric crescents, fitting of GRMHD simulation snapshots, and ring fits from imaging. The three broad categories of methods are all remarkably consistent and lie firmly in agreement with prior measurements of the mass of M87 from stellar dynamics.

What’s next for the study of M87?

Polarimetric analysis of existing data from EHT observations of M87 will shed light on the exact source of radiation near the event horizon of M87, and will help further constrain the possible non-Kerr compact objects which might explain the image. Follow-up observations can confirm the observed structure, and further additions to the EHT will increase the imaging fidelity of the array, potentially allowing connection of the shadow-scale features to jet-launching structures. Until then, we’ll have to content ourselves with another test passed for Einstein’s GR, and a beautiful image of a 6.5 billion solar mass monster 55 million light years away.


About Guest

This post was written by a guest author. If you're interested in writing a guest post for Astrobites, please contact us.


  1. Does the orientation of the target matter for this type of imaging? Also, does the target need to be actively feeding on material at the time to be resolved? Thanks, very interesting article.

  2. Congratulations to every member of the team for this success. I am impressed by the deployment of the Event Horizon Telescope. Thanks for the article and to Astrobites for its communication.


Leave a Reply