A Global Effort to Capture the Most Elusive Culprit
For something presumed so essential to the structure of galaxies and, by extension, the universe, we know surprisingly little about black holes. A major reason for knowing so little is due to the fact we can’t even see them; we see light, and not even light can escape a black hole’s immense gravitational pull. Nearly a century’s theoretical physics indicates that black holes were pervasive throughout the cosmos, a widely accepted view among physicists, despite the absence of any tangible evidence.
Albert Einstein had a few doubts as to whether his theory -that matter can curve space and time- was correct. His theory of general relativity hinged on the idea of areas of space with massive amounts of gravity. We’ve witnessed hints of the existence of black holes since 1990, typically when a star ventured too close to one, but a direct observation of an event horizon remained elusive. Until now.
With the publication of the first ever picture of a black hole this month, any residual doubt that they exist is gone. The image is a groundbreaking achievement for the Event Horizon Telescope (EHT), an international observatory spanning the globe. Two years ago, an international collective of scientists joined forces to take pictures of two black holes located at the centers of galaxies: one in the Milky Way, known as Sagittarius A*, and one in a nearby galaxy called M87, known as M87. Scientists linked eight radio telescope observatories in Chile, Mexico, Antarctica, Spain, Arizona and Hawaii and combined their images taken over the course of a week in April 2017. Linking radio dishes across the Earth created a virtual planet-sized telescope with a magnifying power capable of imaging black hole event horizons.
“Rather than looking at the black hole itself—which does not permit light to escape—researchers looked at gas surrounding it in the event horizon. The gas in this area heats up to billions of degrees, creating a silhouette which can be measured,” said Dr. Jorge Dias, Professor of Computer Imaging at Khalifa University. “Just taking pictures of the black hole from various points around the Earth isn’t enough, as the light waves rolling in from M87 were never collected at a single focal point. Instead, the data was received by each telescope and physically carried to a single location to be processed”.
Dr. Jorge Dias is part of the team at Khalifa University that works on Image and Signal Analysis. His work impacts the viability of a vast swath of future tech, from computer vision, to robotics, and machine learning. Unlike our own vision, which we primarily experience seamlessly and simultaneously to cognition, machines take in data through sensors and transmit numbers, which are interpreted through predefined algorithms. Machine learning is tightly related to sensory input and remains a focus for researchers.
“The Global mm-VLBI Array had to cancel out the background static created by taking images from across the world and sharpen them. As there are no direct connections between the radio dishes, the recordings at each site needed to be stable enough to be compared without ‘jitters’, with VLBI using atomic clocks to time-stamp the recorded data. To ensure recordings were made simultaneously, VLBI required synchronization at the level of a millionth of a second, achieved through using Global Positioning Service clocks at each geographical location,” explained Dr. Dias.
Once all the data was measured and aggregated, the picture still needed to be created. The light collected gives an indication of the structure of the black hole, but since there are only eight telescope locations, there still wasn’t enough data to reconstruct a picture. To make an image possible, imaging algorithms were developed to fill in the gaps.>
“Emerging computational methods push the boundaries of interdisciplinary imaging to fantastic results. The Continuous High-resolution Image Reconstruction using Patch priors (CHIRP) algorithm used machine learning to fill in the gaps, much like completing a jigsaw puzzle with missing pieces,” Dr. Dias.
Astronomers had so far only observed black holes by the behavior of the objects around them. The visible light, x-rays and radio waves emitted by stars can be seen by advanced telescopes to measure a black hole’s effect on its surroundings—meaning scientists had an idea of what a black hole would look like. Computer-powered observatories scan for and record bright points of light that are emitted as a black hole affects a nearby star; the CHIRP algorithm takes this data and identifies common patterns among black holes. It then learns these patterns and uses them to predict what would appear in the areas we can’t get data for using the EHT. And it’s not like we didn’t have a lot of data: in one night, the EHT generated enough data to fill half a ton of hard drives. Getting access to the data from the South Pole Telescope required waiting for the end of the Antarctic winter, so the hard copies could be shipped out.
The CHIRP algorithm is an example of the concept of neural networks. Their mechanism is closely related to how the human biological neural network functions—learning from examples. Neural networks have a set of inputs and one output, which they are taught to give based on some fixed input patterns. If a neuron encounters an input pattern it has not been taught, it outputs something as closely associated with its taught input pattern as possible.
Continuing the puzzle analogy, if you know the puzzle is supposed to show a face, you can assemble the outline and then use the computer algorithm to create a recognizable image. The problem? There are billions of different faces and it’s impossible to know which face would be the right one. We could have an infinite number of possibilities for the image of a black hole, so how do we know which is correct when we don’t know what it looks like in the first place?
To combat this, the Event Horizon Telescope Collaboration split into four separate teams to analyze the data independently and ensure no bias affected the resulting image. Different features were imposed on the input to the algorithms, and the output images were compared. If a lot of different features give the same kind of final image, the algorithm can be trusted. An elaborate series of tests was conducted to ensure the image was not the result of a technical glitch or fluke, especially since creating the image required filtering out the noise caused by atmospheric humidity warping radio waves and precisely synchronizing the signals captured by the telescopes, among other factors adding to the difficulty.
After months of the teams working independently, they reconvened in Cambridge, Massachusetts, and ran their algorithms in the same room, at the same time. The result? The now famous image of the supermassive black hole at the center of the M87 galaxy.
The image shows a bright ring formed as light bends in the intense gravity around a black hole that is 6.5 billion times more massive than the sun. The accretion disk—the ring of light—is on its side with regards to Earth, with the hole facing us and spinning clockwise. The image is brighter where gas flows around towards us. M87* is massive even by supermassive standards but located 54 million light-years away. Despite Sagittarius A* sitting a mere 26,000 light-years away, M87* was easier to image—and what we’re seeing in the black hole image.
Photo evidence of a black hole has been postulated for years, but finally accomplishing it is a stunning accomplishment of machine learning and Image and Signal Analysis. Modern physics starts with basic assumptions, builds verifiable theories, and then verifies them: that’s what’s happened here. A theory has to be given every possible new opportunity to fail, and the theory of General Relativity has withstood this one. This image represents the first steps into a profound new kind of astronomy and paves the way for an array of space telescopes throughout cislunar space (the volume inside the Moon’s orbit) as we seek ever sharper and clearer images of M87*, Sagittarius A* and the supermassive black holes to be found in the center of every nearby galaxy.
News and Features Writer
1 May 2019