Thursday, September 29, 2016

SpaceX Mars plan analysis

Update: Try your own simulations of Mars EDL with my code:

On September 27, 2016, SpaceX finally revealed their Mars transportation architecture ( It was a very exciting moment. Regular readers will know that I have engaged in idle speculation on the topic, and I was gratified to see I got the details mostly right, though their system is a lot larger (up to 450T cargo!) than what I initially had in mind. If you're interested, see my best guess from 2015:
That said, I would be very surprised if the final product looks much like what we saw today, especially for the ship. The architecture presented is fairly conceptual, especially on some of the less mass-affected issues.
With that in mind, what SpaceX presented today can be broken down into a few different parts: a conceptual concept of operations (conops), a CAD draft and pretty animations, and most enticingly, early demos of core tech. Let's look at each in turn, and I'm going to assume familiarity with the presentation: .
The architecture is designed around the principle of "cheaper is better" which almost always drives "simpler is better". Yes, it is possible to get more mass (maybe) with less fuel if there is an intermediate stage or multiple cores, but the most overlooked handle is the size of the rocket. Mars requires a developing a new super heavy lift rocket anyway, so it may as well be BIG! The SpaceX booster, with a nominal 550T to LEO capacity, fits the bill. 
Having total reusability drives a big Mars vehicle that can fly from Mars back to Earth with a single stage, requiring about 8km/s of delta-V. Indeed both ship, tanker, and booster can fly single stage to orbit on Earth, albeit with no payload. The same Mars vehicle also has to perform entry descent and landing on Mars, and has enough fuel to fly from Earth LEO to Mars, and from a suborbital boost to LEO. This means it has to be refueled along the way: In orbit by 3-5 tanker flights, depending on how the masses wash out down the line, and on the surface of Mars. The rest of the presented masses and thrusts all check out. The engine clustering on the ship is an interesting approach, with 6 vacuum engines and 3 sea level engines (smaller bells). Thrust wise, sea level engines are only needed to land on Earth or under high dynamic pressure on Mars, and one is plenty. Three provides some redundancy, and may figure in some launch abort scenario. Mars is so close to vacuum that the vacuum raptors will work there too. Given that landing on Earth happens at the very end, it may even be possible to detach part of the expansion bell so that the vacuum raptor engines can function in the Earth's atmosphere.
Areas that were light on detail include the transition to powered flight during descent on Mars (or Earth). The video showed it nosing up indefinitely, though that would require terrific pitch authority and amazing anti-slosh fuel tank baffles. Downmass capability on Mars is driven by aerodynamic constraints, so I ran the SpaceX sizes and masses through my Mars EDL simulator:

(Click to expand) Left panel: Historical data from robotic missions, showing Mars entry profiles. Parachute descent typically commences in the bottom left at around 500m/s. Central panel: Results from my ballistic motion simulation reproducing behaviour of previous landings, validating the code. Right panel: Entry profiles of several hypothetical future Mars vehicles, with Curiosity for reference. LDSD levels out a little higher (depending on total loading), while Red Dragon needs a significant mass offset to achieve enough lift to not hit the ground. The three curves marked ITA (Interplanetary Transportation Architecture) represent different lift parameters for the SpaceX ship. Horizontal flight represents banked turns to prevent multiple skips out of the atmosphere. Their high lift and high entry speed compensate for their high mass, and they don't get too close to the ground. Mars' highest mountains are >20,000m tall.

I was pleased to see that despite the high mass (up to 800T) the high entry speed, generous cross section, and lifting body concept results in an entry profile that doesn't involve a compulsory crater. Thermally speaking, SpaceX claimed a maximum temp on entry of 1700C, which seems a little low. If PICA can endure 1.2kW/cm^2 heat load, that implies a peak heat shield temperature of about 3800K, given a sensible surface emissivity. A fully loaded ship decelerating at 6gs is dissipating more like 67kW/cm^2, but most of that turns into a very hot, shiny, pretty wake like a shooting star.
Similarly, the propellant farm was presented as a series of chemical reactions, without specifics on mass, efficiency or output rate. About a megawatt of electrical power, continuous, is required to refuel the ship on Mars is a single year (365 days). Most of this power is spent on electrolysis. A solar array capable of producing this (without tracking) would be around 10,000m^2, which is not impossibly large. Solar panels are virtuous, in the mass sense, since they can be made practically two dimensional. 

CAD Models
The CAD models look great, but clearly represent an early draft. The interior space of the crewed module is a bit spartan (needs bulkheads), while the oxygen feed lines to the 42 raptor engine cluster look a lot like a brain angiogram scan. Getting prop feed to 42 engines that are throttling and pogoing, across a giant thrust structure trampoline, while damping every instability and cavitation, sounds like a nightmare/worthy engineering challenge to me.
Similarly, I'm not convinced about the giant window or the downward facing aero strakes, but these parts are less important at this time. The long lead stuff is engines and tanks, and those parts in the CAD are nicely specced out. 

Core Tech Demos
This was the most exciting part by far. The reusable architecture calls for single stage return from Mars. It's all very well to draw spaceships (spaceship!) all day long, but when the rubber hits the road, the system requires a monster engine, as well as fuel tanks with practically imaginary mass. That's a good place to start, and that's what SpaceX has been working on.
I don't know enough to comment on their carbon fiber fuel tank prototype (though I liked the chandelier), so I'll focus on the engine. The Raptor engine has haunted my dreams for years. Unlike the rest of the architecture, here cheaper does not drive simpler, at least at the combustion cycle level. When it comes to high efficiency, the Raptor uses every trick in the book and probably a few that aren't written down yet.
These include full flow staged combustion, multi stage pumps, very high chamber pressure, and the latest in materials and manufacturing tech. Big moving parts in this engine have to withstand high pressure high temperature preburned (ionized) oxygen, which makes a lava-proof submarine look easy by comparison.
And to their credit, SpaceX designed and built the hardware, and showed a video of a short test firing, probably at around the 20% thrust level. Obviously, the engine is far from qualified. But a working demo is a long way from a paper study, it convincingly demonstrates that SpaceX has world leading vision and core competence in rocket engine design.
Final Thoughts
The SpaceX Mars plan is a compelling vision for moving lots of humans to Mars. A complete system will be much more detailed and probably a bit different, but importantly this lays a technical foundation and is a great starting point for future system discussions.
To read more about these and other technical challenges facing crewed Mars exploration, check out my book "How to get to Earth from Mars" at .


  1. I wonder if they plan to rotate during aerobrake to spread the energy around the entire rocket.

    1. Heat shield all around! I would hope it has some dihedral for stability, but anything is possible.

    2. Dr. Handmer.
      Hello. I would be interested in conferring with you more on EDL. Please find me at your leisure FB/Linkedin: Donald C Barker (Houston).

  2. I find myself stunned, even now, that this venture is being forecast using chemical rockets. It's utter madness, and the description of the Raptor and aerobraking in this article only reinforced my predudice. A fission thruster is required for the transfer stage, that can also detach as a lander, leaving the habitat in LMO. It can then collect reaction mass on the surface quickly and easily.Such a machine could make multiple journeys between the surface and orbiting habitat.

    1. I think if you have a fission engine, or warp drive, there's not much you can't do in space. But the specific thrust of a fission engine probably isn't good enough for launch, and there's also the niggling issue that they have a very low TRL.

  3. I don't see your EOMs explicitly listed anywhere in your code, so I can't be sure, but it looks to me like you are running a 2D planar simulation well outside the range of its validity. For example, you are showing ITS engaging in level flight at 20 km and 6000 m/s. That's impossible for this vehicle, even by flying inverted. Can you confirm whether or not these charts were generated from planar or spherical EOMs?

    1. There's no aero going on under the hood, and no aerodynamic inertia either. What do you mean by impossible?
      The EOMs are on line 5ish, called "eqn" or something like that.

    2. Sustained level flight at higher than circular orbital velocity doesn't just require bank modulation, it requires bank modulation while flying inverted. And you show the vehicle capable of doing so at altitudes as high as 30 km. We may be assuming different maximum lift values, but as far as I can tell, that is impossible. At that altitude and speed, the vehicle can't generate enough lift to avoid skipping, even by flying with the lift vector fully down.

      As such, I was wondering if you might be using simplified EOMs that omitted centrifugal terms.

    3. Yep, the world is flat. I'll throw in centrifugal terms and see what happens.

    4. I will update the code shortly. Centrifugal terms added. Regarding inverted flight at 7000m/s, given the IMS numbers, a Cd of 1.5 and a Ld of 0.88, which is reasonable for a lifting body like this, the atmosphere is thick enough to maintain horizontal flight below about 40km.

    5. I thought so. Don't worry, forgetting centrifugal terms when extrapolating from atmospheric to orbital flight is a common mistake. And by common, I mean I've made it 3 times in a professional setting. A good rule of thumb is that if your velocities are being expressed in km/s, then you should include centrifugal terms.

      If you got those aerodynamic numbers from shuttle, then I would say they are optimistic. For a wingless blunted vehicle of this size, I would recommend using values closer to the DRM lifting body:

      C_d = 2.5
      C_l = 1.4
      L/D = .56

      I think you will find that this also changes your integrated thermal load to unrealistically high levels. It is for this (and other) reasons that I strongly suspect the vehicle will be making at least one braking orbit before entering.

    6. I think this code started life as an orange cannon trajectory modelling system!

      That said, there's more to the physics than that. Both drag (and hence lift) and centrifugal force scale with v^2. So the most important handle is ambient density. But in the regimes we're talking about, the entering vehicle is pulling 4-8gs, Mars contributes 0.38gs, and the centrifugal term may be as high as 1.5gs, so I think there's still a good deal of headroom when it comes to the heating constraint.

      The heating problem was the one part of the SpaceX presentation I thought looked a bit funny, as a 2000K max heat shield temp seems a bit low.

      The aerodynamic numbers are somewhat worse than shuttle. Closer to AMaRV, and bracketed with 100% variation in the parameter. I also turned the mass up as high as it would go.

      If you're creative enough with the lift function, you can get my code to do a skip entry. Depending on the heat shield material, you may want to avoid more than one swing in temperature. Aerobraking sometimes confers a delayed-action heat pulse that starts making trouble. I think the ablators, like PICA-X, partially side step this issue.


If you want to comment without a google or open-ID account, sign enough of your name that at least I know who you are, or leave an email address or bank account details or something.