Wednesday, March 2, 2016

Interorbital Exchange - part 7, electric vehicles

All previous posts described all-chemical systems that could be built and operated profitably in the near term. This one focuses on electrical propulsion systems.

The defining features of most electric propulsion:
 - High efficiency (high Isp)
 - Low thrust
 - High power requirements
 - Long trip times
 - Long operating life

 I chose a specific paper (Frisbee, Mikellides) to examine since the authors thoughtfully included most of the interesting parameters for a reusable NEP Mars cargo tug. I don't really dive into how to calculate this for yourself because the problem is quite difficult without modeling software.

It all comes down to the details; the question of NEP vs. SEP vs. Chemical depends on the specific mission goals and technologies used.

The summary:
23 tons dry mass for a nuclear-electric tug of ~6 MW thermal / 1.2 MW electric
64 tons cargo capacity from low Earth orbit to Phobos-Mars orbit
Just under 40 tons of water propellant for the outbound trip and another 7.2 tons acquired at Phobos for the return
2.2 years outbound, slightly less inbound
Two round trips between thruster refits, five round trips between reactor refits

More after the break.

(skip to the next section if you are already familiar with electric propulsion)

The general idea is to use electrical power to dump energy into a propellant and then release it at very high speed.

The simplest of these is electrothermal.
 An electric current produces heat and the propellant is passed through it. First up is the resistojet, where a resistor somewhat like an incandescent lightbulb filament is heated and then the propellant is pumped past it. These are common devices in the RCS systems of satellites. Second is arcjet, which passes an electric arc directly through the propellant instead of through a resistor. These can reach higher Isp because they can heat the propellant beyond material limits for resistor elements.
 Efficiency is moderate (Isp of 500 to 1000, well above any realistic chemical system but easily an order of magnitude lower than the most efficient electrics). Preferred fuels are low atomic weight with no particulates (hydrogen, water, ammonia, hydrazine). Thermal power is most efficient when there are few degrees of freedom for the molecules so more of the energy can be applied to macroscopic motion, so hydrogen or hydrogen-rich propellants are ideal.

Next up is electrostatic.
 These are the 'traditional' ion thrusters, NSTAR, Hall effect, etc. The propellant is ionized by a strong electrostatic field (with some variations) and then the ions are accelerated by a negatively-charged grid or electrode. An electron gun is used to neutralize the electric charge of the ion beam and keep the spacecraft electrically neutral.
 Efficiency is high. with Isp ranging from 1000 to 5000 for most designs and a few reaching as high as 10,000. Preferred fuels are high atomic weight gases or elements with a very low first ionization energy; argon and xenon are frequently used. Sometimes iodine or metals like tin, magnesium or sodium are used, while caesium and rubidium can be used in a FEEP thruster.

Lastly, electromagnetic.
 These are the 'new generation' plasma thrusters like VASIMR, MPD, PIT, etc. Note the idea is not new, it's just that these technologies have been getting a lot of press lately. In fact, a pulsed plasma thruster (from this family) was the first electric thruster flown in space. In these devices the propellant is ionized by arc discharge, microwave heating or other means and accelerated by a magnetic field rather than an electrostatic grid.
 Efficiency is typically variable but can be very high, with an Isp of 1000 to 30,000 (most commonly about 1500 to 6000). Preferred fuels are the same as electrostatic thrusters for the most part, with lithium making an appearance in MPD thrusters.

 All three families have been in use for decades, while each family has relatively recent members pushing the limits. All share common fundamental physics with regard to their efficiencies. Sources of loss are in the power processing units, ionization energy of the propellant, dissociation energy of the propellant if it is a molecule rather than a pure element, ion impacts with the grid or body and in the charge density and geometry of the exhaust.


 Of key concern for a reusable vehicle is the propellant should be available in space. Xenon, argon and nitrogen are available in the atmosphere of Mars. Small amounts of nitrogen may be available in lunar cold traps or bound in soils of Ceres and many C-type asteroids. Water appears to be widely available. Alkali metals would be available on the Moon and in most asteroids.

 Another critical factor is that the thruster system should be reliable over the long term.  Electrostatic systems have already demonstrated very long operational lifespans in the range of 5 to 10 years of active thrust. Electromagnetic systems don't yet measure up in demonstrated lifespan, but that is mainly because most systems of this type are low Isp / high thrust RCS components designed for relatively short operating times rather than main drive units designed to last a decade. As a practical engineering concern, there are difficult challenges in improving lifespan of systems with exposed electrodes or grids. Designs without grids, whether electrostatic or electromagnetic, have the potential for decades-long operation.

 A practical system will be able to service a cargo route in a reasonable time. To illustrate this, let's dive into a paper on a pulsed inductive thruster proposal that includes payload transits to Mars, Saturn and solar escape. A reusable cargo tug with about 64 tons of payload and a 2.2-year Mars transit would require about 2 megawatts of electricity and about 275 tons fueled mass at departure. Propellant would be plain water, though ammonia works as well. If propellant supplies are available at Mars then we require only 1.2 megawatts and 165 tons fueled mass. Of that, the tug itself is just under 23 tons. Enough spare components would be included to make two round trips between thruster refits. A refit would mean swapping out the entire thruster pod with a fresh assembly, quick and easy.
 2.2 years is a long trip, but it is right at the synodic period for Earth and Mars. A cargo tug launched at one opportunity would arrive just in time to go / no go the crew launch at the next opportunity. Part of the problem is that a low-thrust vehicle has to produce about 16 km/s of dV for this trip, far more than a high-thrust chemical system requires thanks to the Oberth effect. Assuming an Isp of 6000 seconds the fuel mass ratio is 23.81% or 39.285 tons of water. An empty return trip would require 7.2 tons of propellant. Put another way, each ton of fuel delivers 1.6 tons of payload. Contrast that with my chemical tug's ratio of 0.8 tons of payload per ton of fuel and you can see the advantage; twice the mass delivered for the same quantity of fuel. Of course, the two approaches trade off costs between dry structure and fuel; electric propulsion is not automatically better but it certainly lets you do a lot more with the same starting mass.

 Power can come from two sources, solar or nuclear. The tug described above is nuclear; its reactor should be good for five or possibly six round-trips out of the box but could be designed for a much longer lifespan of 40 to 50 years with a bit more mass. This is partly because the same propulsion system is intended for exploration missions to the outer planets, where solar power is minimal.

 A Mars cargo tug could certainly use solar power instead, with a lifespan in the 20 to 30 year range and less complicated refitting / disposal. The ~19 tons of reactor, radiators and conversion hardware would be replaced by very large solar panel arrays of about 30 tons, 3.0 MW at Earth beginning-of-life yielding 1.2 MW at Mars end-of-life, with a whole-system specific power of 100 W/kg, 20-year useful life, 20% degradation and 50% of Earth-normal power available at Mars. If PV refits are available every two or three trips then the allowance for degradation can be reduced to perhaps 10%, saving about 3.5 tons.

 Reactors also degrade over time as their fuel decays; some designs such as traveling wave, drum reflector and pebble bed can level the power output over time by only burning part of the nuclear fuel load at any one time. These designs can be life-extended by including more fuel during construction and / or by replacing fuel elements. Since the nuclear fuel is only a tiny fraction of a reactor's mass, this life extension adds very little mass to the overall system. This is the same reason why increasing a reactor's power takes less mass than increasing a solar panel array's power; for low power outputs the solar panels are nearly always lighter due to the reactor's heavy power conversion equipment and shielding, but as the power output grows the reactor eventually beats PV. As you can see from the above example, 1.2 megawatts at Mars after 20 years is firmly in nuclear territory given current state of the art solar performance. Even so, I would bet there is room in the design space for solar PV to be competitive at this power level, design life and solar distance.

Future work:

I think the next step is to work up some EML2 to Phobos tether-capture cargo runs and see how they compare to the baseline LEO to LMO mission. Keeping used nuclear reactors out of LEO is a good idea, as is keeping large solar arrays out of the Van Allen belts. I'll try for an electric interplanetary tug with similar payload to my chemical tug. These would have the added bonus of providing abundant power while parked; there may be a case for a set of tugs such that one is always at Phobos providing megawatt-scale electrical power.


  1. A very interesting write up. I wish I had as much technical knowledge for use in my own sf blog.

    A couple of questions:

    How scalable is the design?

    Why are electrothermal drives limited to such low Isp?

    If you have 6MW of onboard thermal power, why not use it directly as a nuclear thermal rocket, with possible simpler design?

    At what power output do you think nuclear designs would become more efficient than solar designs, in your example?

    1. Greetings, and thanks.
      For a full investigation I strongly recommend Atomic Rockets as a reference. For example:

      The reactor design presented in the paper I referenced is scaled up to 5 MW in their graphs. I think we can take that as a high-confidence estimate. Terrestrial nuclear reactors range up to about 1.3 GW (many examples at this power level and a small number higher), so the potential to go bigger is certainly there. I think it would be fair to scale up the Promethius design to perhaps 50 MW with an alpha of 18 kg/kW including the reactor, shadow shield, power conversion equipment, radiators and thrust structure. That would be 900 tons, so we're talking about a very large craft.
      Their thruster design is also scalable, but it looks like they chose to use an optimized thruster design and simply used enough of them to provide sufficient thrust, spares and propellant throughput. A 50 MW JIMO-like spacecraft should have enough area at the back to mount enough thrusters to maintain desired acceleration and lifespan.

      Electrothermal drives are basically rocket engines that get their heat from electricity instead of combustion. A cryogenic engine burning liquid hydrogen and liquid oxygen is able to cool its combustion chamber and nozzle by passing the liquid propellant through cooling channels, so these engines can have extremely high combustion temperatures. By contrast, the electrothermal engine is typically using hydrazine or a pressurized gas at ambient temperature. Cooling becomes a serious problem at high temperatures.

      State of the art electrothermal devices do have slightly higher Isp than direct nuclear-thermal engines because the filaments or electrodes can be made of tungsten, which can handle temperatures over 3200 °C. Nuclear-thermal engines have to be kept below the melting point of the fuel elements, cladding, etc., so zirconium oxide and uranium oxide are the typical limiters to around 2400 °C.

      You are correct that using the reactor for NTR would be simpler in some ways, but it would be less efficient. This would be offset somewhat by the much higher levels of thrust, gaining some Oberth benefit and much shorter trip times. An NTR cargo tug would be viable and is a core option in NASA's current Mars mission plan. Still, using the same power output to drive thrusters with an Isp of 6000 or so yields far more delta-V for the same mass.

      The crossover point varies by mission and by vehicle size. At Earth's orbit, tens to hundreds of tons of cargo and for missions of 10-20 years solar power is going to win probably every time. One possible exception would be a dedicated heavy-cargo LEO to GEO tug that crosses the Van Allen belts multiple times per year; the added mass needed for radiation shielding might be beaten out by a very efficient reactor. If you want to build a 50k ton GEO power satellite then you will probably want a few heavy nuclear cargo tugs to move components out of LEO and EML1/2.
      Mars missions for reusable craft that need to make 5 or more round trips with tens to low hundreds of tons of cargo are in the right ballpark for crossover. Disposable ships should probably just use solar. Heavier cargo favors nuclear.
      Main Belt bodies like Ceres make nuclear even more competitive even for single-shot missions, while the gas giants and outer bodies require nuclear for almost any useful mission.

    2. Then I see you've previously linked that same project rho page on your blog. Excellent.

  2. Its always fascinating to have estimates based on real world numbers, not only idealized equations, so thank you for that. I'm surprised that you've taken a look at ToughSF too.

    Three questions:

    -Could there be a contactless version of an electrothermal engine, such as one where the propellant is heated by radio or by magnetic induction?
    -You mentioned crossing the Van Allen belts. Are modern solar panels for space use specially vulnerable to that sort of radiation? Can they be made to resist damage?
    -How or where should I contact you to discuss a near future setting that would deal with the initial steps to colonize the solar system?

    1. The VASIMR engine heats its propellant with microwaves, for exxample, so yes it is possible. You could say that the distinction between electrothermal and electromagnetic is a bit fuzzy; I don't think there is a particularly sharp boundary between the two. In common use the term electrothermal usually refers to the simpler low-Isp devices.

      The silicon cells of solar panels are covered in a 'front glass' layer for protection. Panels intended for space usually have this layer deposited directly on the silicon, so the thickness is tuned for the specific mission. Thicker glass (or alumina or sometimes silica gel) absorbs more radiation and allows for a slower decay of efficiency at the cost of more mass.
      Silicon is tougher than human flesh with regard to radiation. Even for fine details like in memory or processors it typically takes several damage events to disable a single feature (though high Z cosmic rays can definitely do real damage with just one impact). For bulk PV silicon the slow accumulation of lattice defects and implanted nuclei leads to a slow and even degradation of efficiency. A typical solar panel meant for GEO could survive a few months in the lower Van Allen belt and only lose perhaps 10% of its output, then lose another 10% over the following 10 years in GEO. Something built as a LEO to GEO tug would need much thicker protection to withstand the radiation for a few years.

      You can reach me at christopher (dot) wolfe {at} gmail com. As you've no doubt noticed I can be a bit erratic with communication, but I would be glad to be of help.

    2. I refresh this page very often, don't worry :)

      Thank you for the reply, once again.

      Please do create a Google Plus account.

    3. Done, and following a few people now as well.
      I managed to mangle my own email address, the accurate one is christopher(dot)wolfe(dot)public(at)gmail(dot)com.

    4. Ah, that might be why. I'll send again :)