Years ago, I stayed with the family of a college roommate for a weekend. While browsing their bookshelves, I came across a slim little book that drew me in: To Know a Fly by Vincent Dethier. As I recall, I sunk into an armchair and read it in a single sitting.

Dethier was an entomologist and the book is an irreverent memoir on experimental methods available on a shoestring budget. I got my hands on my own copy recently. This time, I’m taking it a little slower, but it’s still pretty funny and a delightful read (UPDATE: As I got deeper in, it tilted from “delightful” to “cringey”). It’s also more insightful into the reality of working scientists than most pop-science books you’ll find.

When I was physicist, I mostly worked in particle physics, which on the surface is at the complete opposite end of the spectrum of Dethier’s experimental setups: costly, multi-year professional engineering projects, carried out by enormous collaborations. But still, the prototyping work for designing the detectors and targets in a particle physics lab required lots of ramshackle, cobbled together tinkering along the way. That was the best part.

I spent much of the summer of 2005 on the fourth floor of the physics building at the University of Illinois trying to achieve laser-induced nuclear magnetic resonance in a helium sample. The laser and its power supply, were costly high-precision instruments, but nonetheless we were struggling to get the signal we were looking for out of the helium sample.

Laser NMR works when you get the sample in a steady beam of the right wavelength of light. Lasers produce a beam of light with a tight wavelength spectrum (that is, light that is nearly a single, pure color), but the exact wavelength depends on the temperature of the electronics in the laser. For run-of-the-mill applications, that doesn’t matter, but NMR can’t be sustained by a laser with a wandering wavelength.

We couldn’t get the wavelength of our laser to stabilize, mostly due to flickering overhead lights and the HVAC turning on and off. The thing that made it all work was a makeshift cardboard baffle we built up around the tens of thousands of dollars of precision scientific instruments.

This is the real substance of how science is done. Hours of frustration and then saving the day by repurposing mundane objects to hack together a stable system.

Hope Jahrens said it well in Lab Girl:

We need to build a machine. The good news is that we only need to build this machine once, since there’s no chance that anybody besides us will ever want this thing in their bathroom or government office. This frees us to make it as ugly, silly, unwieldy, and inefficient as we want to - we just need to improvise something that works for us. This is how scientific research instruments are built.