Although scientists have been researching and theorizing black holes for decades, this is the first ever visual image of a black hole. This picture is the result of a massive collaborative effort through a project called the Event Horizon Telescope, or EHT.

The black hole pictured above lies at the center of a galaxy called M-87, which is 55 million light-years away. To understand the effort required to take a picture of it, imagine trying to photograph a grain of sand in Los Angeles—while standing in New York. “Taking the first image of a black hole was a huge endeavor,” said Katie Bouman, one of the lead scientists and developers on the project. “Because it is so small, we needed to build an earth-sized telescope.”

Rather than actually build a massive telescope the size of the earth, over 200 scientists spent decades linking a series of telescopes around the globe into a network, using the precise timing of atomic clocks, to create a massive, virtual telescope capable of taking the picture. “It took many people years of hard work in order to build a computational telescope that made it possible to see the unseeable,” said Dr. Bouman.

The global network of telescopes recorded a huge amount of data, which then took years of processing to finally compose the image. Because it took a whole new “photographic” approach to build and connect the equipment to even take this picture, it similarly meant taking a new approach to processing the data in order to develop it into a single visual image. If building the global network of telescopes was like using a very fancy new camera, then the computational algorithms that are used to process and interpret the collected data to form the image are akin to a very fancy darkroom and photo development process. New equipment requires new development techniques, and in this case, the “development” was in the form of code.

The Speed and Power of Python

Although the HOPS pipeline Dr. Chan helped develop is mainly written in the C programming language, EHT scientists also developed a lot of Python code around it to drive the analysis. Dr. Chan explained that there is a huge advantage in using Python for analysis because it enables scientists to do their work more quickly. Because this was the first time scientists had seen this type of data, they needed to be able to try different things on the fly—and Python is a great language for this explorative approach. “A lot of our work is actually working with and thinking about the data, and Python lets us speed up the process,” he said.

Dr. Chan pointed to NumFOCUS-supported open source tools — Numpy, SciPy, pandas, Astropy, Jupyter, and Matplotlib — as crucial to this iterative scientific thinking process. “It is thought that Python alone would be too slow to process our volumes of data. However, with the open source projects in NumFOCUS, we were able to iterate our algorithms so fast that they enabled us to finish our work in two years, which would be difficult to achieve if we just used C.”