Manufacturing RF Cavities Quickly and Cheaply
As a graduate student in the Nuclear Science and Engineering Department at MIT with a focus in fusion technology, I worked in the renowned Plasma Science and Fusion Center (PSFC). Having previously come from SLAC Lab, I had an interest in accelerator technology in addition to nuclear fusion, so I decided to focus my research efforts on klystron amplifiers. Klystrons are a type of vacuum electron device that resonantly couples an electron beam to an input drive signal. Through a series of resonant cavities, the electron beam is modulated to increase signal gain. Eventually, the signal leaves the klystron at much higher power. Klystrons are used throughout the world, from radio stations and aircraft, to accelerator facilities and fusion reactors. Despite being invented during World War II, klystrons have not seen much material improvement in their overall electric efficiency as they are notoriously difficult to build. However, if klystrons are to be used as a power amplifier for future fusion power plant RF systems, rapid, iterative design becomes paramount. My research focused on developing methods to quickly and cheaply form klystron bunching circuits, via the use of 3D printing and electroforming.
Here is a 17.14 GHz klystron cut in half, showing the input coupler, resonant cavities, cooling channels, tuning screws and output coupler. These things are quite annoying to manufacture, as cavity blocks are individually machined, polished and brazed together. Considering that additive manufacturing is all the rage, I figured maybe we try to figure out a way to make this more efficient than how they did it in the 1950s.
I could spend a lot of pages describing exactly what I did (90 pages in fact; link to thesis at the top of the page), but here's the gist: instead of making individual cavities and joining them together, what if we used a negative of the klystron? In other words, we create a mandrel of the entirety of the device, electroplate on top of the mandrel and then dissolve away the underlying metal to leave a singular, fully formed klystron bunching circuit? This mandrel method is not unprecedented for accelerator RF cavities, but no one has ever done this for a klystron. So, my project was all about testing out this approach, using both aluminum mandrels (pic to the right) and fused deposition modeling (FDM) 3D printed plastic mandrels.
This project involved dabbling in a lot of different things. First, I did some preliminary static electric field and magnetic field simulations to determine the expected RF qualities per a given cavity geometry. Namely, Poisson Superfish and COMSOL provided adequate resolution for the ~4.6 GHz resonant frequency I was targeting for a single cavity.
A Bambu X1E 3D printer was used to print plastic mandrels. Throughout this project, I had to get pretty handy with SolidWorks, not only for designing and printing mandrels, but also for designing various experimental setups. To achieve sufficient surface smoothness on the plastic mandrels, I designed a rotating vapor smoothing jig. This would allow mandrels to be smoothed uniformly in a safe and repeatable manner. I used either acetone or dichloromethane for the vapor smoothing (yay chemistry!).
Part of determining the potential of 3D printed mandrels was measuring their dimensional accuracy to the underlying CAD geometry. To this end, I learned how to use a Keyence XM-5000 coordinate measuring machine. Key measurements were used to predict the overall frequency shift of a cavity produced via 3D printed mandrels.
Surface roughness measurements were taken via a Mitutoyo SJ201 profilometer. These measurements were critical in determining whether chemical smoothing can achieve "RF quality" smoothness, around 0.4 µ Ra for ~4.6 GHz. (In short, the smoother the surface of a cavity is, the more effective it can be)
More interestingly, I also showed that it is indeed possible to make these same cavities by using chemically smoothed, plastic 3D printed mandrels. These mandrels were then made conductive using air-sprayed conductive paint. Theoretically, a 3D printer would be a LOT easier to make a fully klystron mandrel with than machining completely out of aluminum (though for individual cavities, aluminum is definitely easier). The sample on the right is the best sample I made, can you tell that there's plastic under there?
Fully formed cavities (shown to the left, cut in half using wire EDM), were made by electroplating copper onto the previously shown aluminum mandrel. This mandrel is then dissolved out using sodium hydroxide, leaving the copper undisturbed. This method has been used to form accelerator structures, but never klystron cavities in particular (as far as I know). While similar, klystron cavity geometry is a little bit different than a typical accelerator cavity, so there was no guarentee that this process would work. But it worked beautifully!
The resonant frequency, loaded and unloaded Q of finished cavities (aluminum and 3D printing formed) were measured using a Copper Mountain C4420 vector network analyzer. The two-port insertion loss method gave a semi-consistent measure of the unloaded Q and a pretty accurate determination of the resonant frequency. The mounts were designed in SolidWorks and 3D printed.
Here are the VNA plots for (1) four cavities created from aluminum mandrels, and (2) a single cavity created from an ABS plastic mandrel. Note we were aiming for 4607.07 GHz. The aluminum mandrel cavities landed within ~15 MHz of the design frequency, and the plastic mandrel cavities landed within 50 MHz of the design frequency.
In total, I was able to show that cavities formed by both machined aluminum and 3D printed mandrels exhibit RF properties that are within range of a traditionally manufactured klystron cavity. To my knowledge, no one has ever shown this to be the case for klystron cavities using either mandrel types. Why is this useful? Well, in the realm of kystrons, this could open the door to rapid manufacturing of these devices. More generally, I think it is very surprising that it's possible to successfully build a RF cavity using a structure 3D printed from a consumer grade FDM printer.