Why will Starlink satellites use krypton instead of xenon for electric propulsion?

Question

The press kit for the first Falcon 9 Starlink launch and deployment of the first 60 satllites scheduled for May 15, 2019 says:

With a flat-panel design featuring multiple high-throughput antennas and a single solar array, each Starlink satellite weighs approximately 227kg, allowing SpaceX to maximize mass production and take full advantage of Falcon 9’s launch capabilities. To adjust position on orbit, maintain intended altitude, and deorbit, Starlink satellites feature Hall thrusters powered by krypton. (emphasis added)

Most of the electric propulsion systems that I've heard of use xenon. While the lighter krypton would have a higher Isp at a given acceleration voltage, I assume Xenon has a slightly lower ionization potential and so would be easier to ionize.

DC power for electromagnets for confinement and RF power supply for plasma excitation can dominate the weight of an ion propulsion engine (depending on the specific design and principle), so in for these svelte and featherweight spacecraft I would have thought that the'd go with the lower ionization potential of xenon which presumably can be ionized with lower electron energy.

Question: Why will SpaceX's Starlink satellites use krypton instead of xenon for electric propulsion?

Answer 1

It's the same reason SpaceX often does things differently: Krypton is a lot cheaper.

The satellites are designed to control costs. For example, each will maneuver with Hall-effect thrusters—ion thrusters in which propellant is accelerated by an electric field. The conventional fuel for such a thruster is xenon, which offers high performance. The Starlink satellites, however, will use a different noble gas: krypton. It has a lower density, so the satellite fuel tanks need to be larger, and it offers less performance than xenon. But krypton can be bought at just one-tenth the cost of xenon, which matters if a company wants to fuel thousands of satellites.

Price and production rate

I've found wildly different price quotes for the two:

Xenon is listed as \$1200/kg, which would mean SpaceX is getting their Krypton for ~\$120/kg. The source for that Wikipedia quote also lists Krypton, at $300/kg.

This SE answer gives a Xe price in that region too.

On Alibaba I found someone selling Krypton for \$2/kg, but on Alibaba you never know what you get. So I've done some more digging.

The price difference is explained by the production method. One process (air liquefaction) gives you a Kr-Xe mixture:

Using commonly accepted techniques, most of these stations produce a KrXe mixture containing approximately 93% Kr and only 7% Xe.

So ~10 times more Kr than Xe, which makes a price difference of 10x logical.

In 1998, Xenon production was estimated at 5000-7500 m³/year. Going by this answer, 1 kg = 170 l, so 5000 m³ is 29.4 t.

So there's another reason not to use Xe: 10000 satellites carrying 3 kg of propellant each would require the entire world supply of Xe for 1 year, which would spike the price. Better use something more abundant.

Answer 2

I expect they did the math, and found that overall cost was less, even with reduced thrust/watt efficiency, reduced thruster life and increased tankage and solar string mass. The cost of xenon is huge, and supply is very constrained.

When NASA builds a craft, they have to stretch their fuel purchase over several years because there simply isn't enough produced every year to satisfy a long mission like DAWN and the handful of other uses for Xenon.

SpaceX probably can't afford to wait that long or spend that much. They may use xenon for a longer-lived, more slowly deployed second generation satellite.