So, until about a year ago, I was victim of a misunderstanding that made me very, very confused about how plasmas were made. I had read, in many places, about the designs for ion thrusters and plasma thrusters and VASIMR and even plasma enhanced chemical vapor deposition systems and sputter coating systems, that plasmas were generated by “radio frequency sources” or “RF” for short.
I had heard of radios, and I knew they communicated by sending out radio waves, which are waves of light that oscillate in a spectrum range between a few kilohertz and a few gigahertz that we have dubbed the “radio spectrum” or the radio frequency range. So, I made what I think is a very natural assumption, which is that these RF sources were shooting radio frequency photons at gases to ionize them and make a plasma.
[I thought it was this. It is not this.]
I was wrong, but it didn’t matter, because I also didn’t really need to understand in detail how this stuff worked. That is, until about a year and a half ago, when I decided to become a full-time ion engine builder, and suddenly it became much more important for me to understand these things.
I started looking at the pictures and design drawings people published about thrusters that used RF to ionize their fuel or generate their plasmas, and I started getting more and more confused, because where I was expecting to see something that looked like a radio dish or maybe a laser, what I actually saw was usually a copper coil wrapped around the plasma chamber. Sometimes the shape of the coil was sort of interesting, like in VASIMR, where it was supposed to guide some kind of helicon plasma flow, and then I would think “So, maybe it is some sort of waveguide?” but most times it just looked like a coil, like you would use to make an electromagnet.
[Many many diagrams look like this. A coil. A label somewhere saying “RF.” No explanation.]
Then, as I started learning more about how ionization works, I became even more confused. See, any given atom or molecule has an energy threshold that would need to be met to make an ion. As an example, the threshold for xenon is 12.13 electron volts (eV) – hit the atom with that much energy at once, and it will usually knock off an electron, and make the atom into a positive ion (Xe+). But, a photon from the radio part of the spectrum (even the most energetic part of the RF spectrum, so, call it 5 mm wavelength) will contain only 0.000248 eV of energy. This topic is on my mind now because it came back up with my recent look into lasers, and there are actually a bunch of other complicated dynamics with gas-phase photoionization like excitation lifetimes and interaction cross sections at different wavelengths, but you can get the sense of the situation just from the dummy math: 12.13 / 0.000248 = 48,912 radio wavelength photons that would need to hit a xenon atom simultaneously to ionize it.
Simultaneously, here, means “within the excitation lifetime of the initial excitation from the first photon that hits” which for most materials and most photon energies is on the order of a few nanoseconds to a few femtoseconds. Atoms, as you may know, are quite small. We are lucky if we can get two photons to hit the same atom at the same time. 50,000 ain’t gonna happen.
So, clearly, my photon-based thinking was completely wrong, and now I needed it to be right. So I dove into the research, and eventually (after an embarrassingly large amount of reading and feeling like I must be crazy), I figured out my error.
[It actually is this.]
Turns out there are other things that can oscillate at the same frequencies as radio waves, namely electric fields. This is what people are actually talking about when they refer to RF plasmas or RF power sources: they are using an AC electric current, except that instead of switching polarities at 60 hz or 50 hz like your wall electrical socket, they switch polarities at khz to Ghz, like radio waves. Probably the most common frequency for plasma generation is 13.56 Mhz – if this was a photon frequency, that photon would be a radio wave, hence they are called “radio frequency” oscillations. But, it’s not a photon.
The way this actually works to generate a plasma is to make electrons dance. As the polarity of the RF source switches, an electron will feel an electric (or magnetic – either will work, just different configurations) field that pulls it first one way, then another. The amount of acceleration that the electron can pick up during each oscillation is determined by the voltage at any given frequency – higher voltage, faster electron, more energy added to the gas. Free electrons in the gas will hit neutral molecules, and if you make the oscillation voltage high enough, eventually those collisions start to happen with enough energy to knock electrons off the neutrals, which become free electrons, and fly off to create more ions; it starts a chain reaction, called an avalanche, of ionization. There are some factors that impact the rate of ionization, such as recombination, the efficiency of energy absorption into the plasma, and energy loss through de-excitation (light emission) and thermal collisions, but fundamentally if you keep pumping in power from these RF oscillations, and the voltage is sufficient, you will end up with a lot of free electrons, which will maintain a very highly ionized plasma.
These systems have a few dynamics that should be noted:
First, you need free electrons to seed the system – the voltages used are way below the voltages required for an electric field to just yank electrons off of molecules, so the only electrons that really interact with the oscillations in field are those already out in free space. This means that a perfectly neutral gas cannot be turned to plasma by an RF source. In practice, there are almost always a few free electrons in a gas because of cosmic rays that are always passing through, and DO have enough energy in single photons to ionize an atom in free space (some have enough energy in one photon to flip a bit in a computer!). But, if you get the gas thin enough (reduce its pressure quite close to vacuum) there are not enough atoms in a given volume to get hit by these rays, and you might need to supply free electrons independently to spark and maintain the plasma, using a cathode of some kind. And, if the gas gets really thin, you start having electrons make it all the way to a grounding surface before they hit anything (i.e. the mean free path becomes long compared to the oscillation length or the physical dimensions of the plasma chamber) and electrons are just oscillating their way through free space until they hit the walls – no way to start the chain reaction of ionizing and freeing more electrons to go make more ions, so you can’t spark a plasma at all. So there is a lower bound on operating pressure for these systems.
Second, the ions (once you spark the plasma and have ions) in the system are barely affected at all by the oscillating field. Ion masses are thousands of times as much as the electron mass, so they are thousands of times slower to accelerate in an electric field. Thus, if the oscillations are fast enough, the positively charged ions barely get moving at all before the polarity has switched, and starts pushing them back the opposite direction. They accelerate through so little of the potential field during each oscillation that almost none of the energy ends up in the ions and nearly all ends up in the electrons. This is useful for two reasons:
- It means that the electrons will always have very high relative velocity to the ions, which is why the impacts are energetic – if they danced at the same speed in the same field, then they would always be moving in parallel and the electrons could not ionize the atoms.
- It means that an applied DC field across the plasma can extract the ions over time – because the oscillation has almost no net effect, the ions will drift almost entirely under control of the DC bias, which is used in processes like sputtering and plasma etching for semiconductor fabrication to direct ion bombardment at from a plasma at a target.
Now, of course, I started thinking about how this applied to ion thrusters (because, you know, that’s sort of my thing). RF is a very efficient way to make a plasma. 50% to 90% of the wattage you spend on the RF oscillating signal is deposited directly into your fuel gas. And, because of the avalanche chain reaction effect when you get to ionizing energies, you get a very high percentage of the gas that is ionized at any given time. It’s a simple, robust, easy-to-use system.
However, resulting plasma is net-neutral – all the electrons are still mixed in there with all the ions, which means that the electrons are constantly recombining with the ions to try to get back to neutral again. Every time they do that, they will emit energy as a photon, which is typically lost energy from the system. And, if you want to do anything with the plasma other than bombard the inside of the plasma chamber (slowly wearing out your system), you need a way to extract the ions.
The gridded thruster approach to extraction is a positively biased grid on one end of the plasma chamber – if an ion gets into enough collisions in the plasma to gain enough thermal energy to overcome the positive bias, it can enter one of the grid apertures, and then be accelerated through the gap between the positive grid (called the screen grid, because it screens most of the plasma from the accelerator system) and a second grid, usually at a negative bias (called the accelerator grid, because, you know, it accelerates the ions). This is much less efficient than the process of pulling energy into the plasma itself. Typically, gridded thrusters will spend 80-100 eV to generate each extracted ion. So, despite the 50-90% efficient energy use to make the plasma, you are still spending ~8x the fuel ionization energy for each ion you get; you could call it an ionization-to-extraction efficiency of 12.5% or so. Gridded thrusters accelerate ions very efficiently after extraction, so the overall efficiency from power to thrust is usually 40-60%, which tells you ionization-extraction is the main power loss in the process.
And that’s why Orbital Arc doesn’t use RF to make our ions. It would be the easy path to ionization (field effect ionization is much more complex from an up-front fabrication standpoint). But, you lose too much energy making each ion because extraction is so inefficient. When we make our ions, we will not have a quasi-neutral hot plasma that we then have to manage, we will have a cold, positively charged gas, which we can accelerate directly with a DC electric field. This simplifies our electronics (saves mass) and also should improve our efficiency significantly. (We’ll report exactly how much efficiency improvement we get when we have experimental data to back our modeling, but the modeling looks… promising ;-).)