Leak Detection with Vis/IR Spectroscopy

Leaks are notoriously difficult to track down in high vacuum systems. When research labs build a vacuum system, they normally also use a Helium leak detector or a Residual Gas Analyzer (RGA). These both work by detecting the amount of Helium in the system at normal operating conditions, then you blow Helium gas into one joint at a time and look for a sharp rise in Helium concentration. If a sharp increase is found, there is likely a leak at that joint. These systems are incredibly useful for large vacuum systems with hundreds of connections, but they are also extremely expensive.

With my background in Astronomy and Astrophysics, I became very comfortable reading spectra and the principles in how spectra are created. I already had a high voltage system that creates a consistent plasma, so I could generate a emission spectrum to figure out what the major components of the atmosphere inside the chamber are. This could theoretically be used for Helium lead detection.

Other Ways of Checking for Leaks

Before setting out to make something new, I tried suggestions from other amateur fusor builders. One of these suggestions was to use air duster compressed gas. Because of the working principle of my vacuum gauges, different gasses have an effect on their readings. This is noted extensively in the manuals where a chart is given that contains corrections for different types of gasses. This means that I can blow this heavy hydrocarbon gas into the joints and look for drastic changes on the vacuum gauge readings.

I did attempt this, but I never saw fluctuations in the pressure of the system. This either means that the leak is too small to let enough of the gas in at once, or that I have good seals and the rise in pressure is due solely to adsorbed volatiles on the surface of my chamber. I was producing a good enough vacuum for my needs, but I believed that I could put my knowledge of plasma physics to the test by making this spectrometer.

The Spectrometer Build

There are several videos online for the construction of a similar type of spectrometer. It uses a cheap web camera as the light sensor and a plastic diffraction grating. The IR filter was removed from the camera, allowing it to see into the Near-IR band. I purchased a general purpose electrical enclosure for this build, which needed to be light tight.

The light enters the slit at the front to approximate a point source. This beam travels to the end of the box where the diffraction grating is glued to the front of the camera. I chose to angle this to get the first diffraction spike from the beam, which is the brightest and allows for a higher signal to noise ratio.

One of the other important features is the painted egg crate foam. The biggest source of noise for these spectrometers besides light leaks is the light reflected inside of the box from off-angles. This is especially a problem for the glossy plastic box. I chose to use egg crate foam, as the bumps help to limit direct and specular reflections. This is similar to the foam triangles used in anechoic chambers. I noticed that the foam was fairly matte, but each of the cells in the foam reflected at certain angles, making the whole surface appear to shimmer when moved. These are direct reflections on the flat surfaces of each of the cells that make up the foam. I purchased a popular extreme matte black paint to cover the surface of the foam and camera. This paint leaves a surface that reminds me of kerosene soot.

The software used is called Theremino Spectrometer and is a free way to generate a spectrum from a camera. This software allows you to select the region where the diffracted light appears, and uses its left and right position to add up the light in each coordinate, generating a spectrum.

Theory of Operation

My 35kV power supply is more than enough to ionize any gasses in the vacuum chamber and create plasma, which is seen in my fusor pictures. The regular air plasma is purple/pink, which comes from the Nitrogen and Oxygen. If I recorded the spectrum of the air plasma and checked it for changes when introducing Helium, I could possibly detect leak locations. The main issue is the speed of the system. Commercial Helium lead detectors work in real-time, so you can check the entire system very quickly.

My spectrometer is not sensitive enough to detect small changes in Helium concentration, so a workaround is needed. I devised a way of wrapping each system joint in an air-tight rubber membrane, which I then filled with Helium every couple of hours. Most of this Helium escaped to atmosphere, but the composition of the atmosphere around the joint was mainly Helium. This meant that if there was a leak on that joint, it would be sucking in Helium gas. I could come back to the system after a day or so and reignite the plasma to see if its spectrum changed at all. This needed to be tested one joint at a time, which significantly slowed down the process. This is feasible for my system which only has about 20 joints, but it would be impractical for a much larger system, justifying the cost a commercial system.

It may have been possible to simply wrap the joints in the rubber membrane and wait to see if it becomes concave due to pressure loss, but I believed that the leak rate was so low that it would either take days to check one joint, or that the diffusion of air into the rubber would match or exceed the leak rate. The spectrometer setup is surprisingly sensitive to changes in the composition of the chamber atmosphere.

With this method, I was able to detect two minor leaks from the Conflat flanges. These thankfully just required extra tightening and did not need disassembly. My leak rate was significantly reduced after these were tightened, but I still had considerable pressure increase in the chamber which is due to adsorbed volatiles on the chamber walls. This was solved using the bakeout process described in my vacuum system.