For MIT's winter term course 8.S50 (the inaugural term of 8.16 Data Science in Physics) I had the opportunity to tackle 3 projects using open data (LIGO, CERN, and simulated CMB). I thought this one, with the LIGO data, was the coolest: from a horrifying mess of noise in strain data from a hyper-sensitive interferometer, identify the echoes of black holes merging millions of years ago.
Wait, what? If you know nothing about astrophysics, maybe that sounds a bit weird, so let me explain. One of the predictions of general relativity is that accelerating masses should produce waves that can be thought of as distorting spacetime (as gravity can be thought of as spacetime curvature). In theory, this squishing and stretching of space can be measured. In practice this is done using an interferometer. In fact, a geographically separated pair of very big and precise ones, using squeezed light to reduce noise below the typical quantum limit. Over the kilometers-long interferometer arms of LIGO, the distortions lead to path length changes on the atomic scale.
There's all sorts of other clever engineering that went into the LIGO experiment, but that's not for me to explain. The takeaway of this is that these instruments are insanely sensitive to... everything. And so even with exceptional isolation, the output signals are very noisy. My project for the class was to find the gravitational wave.
to be continued...