I’ve had a lot of friends ping me today about my thoughts on Elon’s Mars talk today. I was in a meeting when it happened, and literally was pinged by half a dozen people during the meeting… Now that I’ve had a time to chew and digest things a bit, here’s a bit of a stream of consciousness take on the plan.
Overall my feelings are mixed. I think the plan has a lot of good points, could probably work given enough money, and would likely be a better use of money than whatever NASA does for its so-called Journey to Mars, but also am really skeptical about a lot of the technical choices made, and the likelihood of hitting the price points SpaceX is predicting.
First, what I liked (some minor, some major):
- In-Space Refueling: Elon’s plan is built around the idea of refueling his ITS (Interplanetary Transport Ship) in LEO prior to Mars departure. The BFR would have to be a lot bigger, and/or 3STO, to loft a fully-fueled ITS without LEO tanking. This is a point that ULA has been making for over a year now as well–“Distributed Lift” allows you to do a lot more with a given sized booster. As someone running a company trying to develop the rendezvous/capture and propellant transfer technologies needed for Distributed Lift, this was a welcome choice.
- Mars ISRU: Bob Zubrin has been beating on this drum since I was a teenager, and while I think Elon’s handwaving the challenge of generating 1000s of tonnes of propellant on the Martian surface, not having to haul that all the way from Earth saves a lot of launch mass to LEO, and makes reuse of the lander a ton easier.
- Lifting, High-Alpha Mars Entry and Supersonic Retropropulsion: By coming in sideways, and doing a somewhat lifting reentry, they’re able to bleed off enough energy early on so that the supersonic retropropulsion delta-V can stay modest even though ballistic coefficients suffer from the square cube law. If they had tried a base-first, low-lift entry, they probably would need a lot more landing propellant.
- Reusability in General: I’ve always been a fan of both in-space and earth-to-orbit reuse. I think it’s a key to get costs anywhere close to what Elon wants.
- Depots for Extension Beyond Mars: I also really liked Elon’s point about how his architecture can eventually include depots in various orbits, and that once you do that, you could theoretically use these vehicles for traveling almost anywhere else in the solar system you want. Many of those flights would require more radiation shielding, and possibly some level of artificial gravity to work out, but he’s totally right that once you have depots, the solar system is your oyster. Regular readers of this blog won’t find that too surprising.
Now for what I didn’t like as much (some minor some major):
- Too Big: Something 3.5x the size of a Saturn V seems like overkill. Doing a BDB-sized launch vehicle but with near-SSTO mass ratios and the highest chamber pressure large propulsion systems to ever fly is a challenge, and I think SpaceX is likely underestimating the challenges with doing an RLV that size, and with that many engines. One thing I wonder about is if they’ve done the acoustics analysis on launching something that big. We’re talking about something several times bigger than Saturn V or Shuttle, and an order of magnitude bigger than the Falcon 9 powered landings they’ve already done. May be a non-issue, but I think both Bezos and Musk are making a mistake with how big of vehicles they’re going after.
- Swiss-Army Knife ITS: Trying to make the ITS do so much doesn’t bode well to me for keeping it affordable. We’re talking a near-SSTO performance level reusable upper stage, who’s ascent and landing propulsion has to double as a rapid-response launch abort system, with a what amounts to a 100 person space station on top of it. Many SpaceX fans seem to think that slapping more requirements onto the most challenging piece of the overall architecture will somehow save costs compared to developing two or three more optimized system elements, but I’m really, really, really skeptical. This seems like repeating one of the dumbest mistakes from the Space Shuttle.
- Vertical Mars Landing: I know SpaceX is sticking with what they know, but the crew and payload section on ITS looks like it’s ~30 meters off the ground level. That’s some really long ladders and/or elevators or cranes, which aren’t going to be light. Heck, ITS makes the LSAM look reasonable when it comes to payload accessibility. I know a horizontal powered landing approach ala DTAL/XEUS would require additional engines and potentially landing complexity (depending on what changes you made to the rest of the architecture), but they would probably require less mass than hauling everything down from what amounts to an 8-9 story tall building.
- Not Refueling In Mars Orbit: I don’t have the exact EDL delta-V requirements for ITS, but it’ll probably take close to 1/3 of the Mars Orbit mass in propellant. Aerocapture/braking first into Mars orbit doesn’t really change the vehicle requirements much, it allows you to cut down on the TMI mass by 1/3. Launching ITS from the surface with only enough prop to get to Mars orbit, and then refueling in Mars orbit, can also dramatically cut down on the overall ITS size. I’m not sure which mission phase is the driver for the ITS propellant tanks and landing engines, but I wouldn’t be surprised if in-space refueling in Mars orbit didn’t cut down both on the overall size of ITS per passenger complement, or if it cut down on the IMLEO of the system. And really, you already have to develop ITS tankers and ISRU on the Mars surface. Just ship a tanker or two along with your original landing group. Use the prop it shipped for refueling the other landers, making sure to leave enough for it to land empty. Even without an actual depot in Mars orbit you can take advantage of that. No new tech, no new elements, but likely a decent savings right from the start. Refuel Early, Refuel Often.
- Methane Uber Alles: I think Elon oversells his case on how awful Hydrogen is compared to Methane. Sure, for his specific architecture, Methane might make more sense, but I can think of many other architecture where LOX/LH2 could probably be quite competitive for all but maybe Mars Ascent and Landing. Sure, if you insist on having one vehicle do it all, sticking with one propellant makes sense, and sticking with one like Methane probably makes your life easier. But there are so many assumptions baked into that logic chain.
- Expensive Work In Process Inventory: The major cost driver on the mission is the ITS, because it can only do a Mars trip once per synodic period at best. This is somewhat the nature of the beast for Mars travel–whatever you send to/from Mars is going to take a long time to get there and back, which means you’ll have to amortize its cost over a lot fewer missions. Which is why it would seem like you would want to minimize the cost of the assets that get tied up like that. Having your Swiss Army Knife vehicle be the one that can only fly 12 times in a half century seems like a poor way to optimize for the problem.
- No Landing Gear on the BFR Stage: I know that my colleagues at Masten also really like the landing-cradle approach, but I’ve never been a fan. Is it doable? Sure. Does it save a lot of cost and time when it works? Sure. But how reliable is it really? I can’t honestly answer this question, but my gut suggests that if your vehicle has a decent chance of surviving landing on an unprepared surface, there are going to be many situations where an abort, a large last second disturbance, or some other error could be survived when a gear-less vehicle is toast. Missing the landing cradle by 10m due to a last-second engine-out scenario probably means loss of vehicle and major pad repairs, where with landing gear it’s a non-event. We could totally build jetliners today without landing gear, using landing trolleys or other things. Or fighter jets landing on sleds on carriers. But we don’t because there’s no way aircraft would be as reliable as they are without having things like landing gear that give them options when something goes off-nominal. Let me put it this way–I don’t think you’re likely to ever see an RLV design that can survive long enough to average 1000 reuses (like SpaceX has baked into their BFR economics) without including landing gear.
- Crazy Raptor Performance: 4500psi chamber pressure with a LOX-rich preburner sounds like a recipe for fun engine development. This is probably doable, and the Russians eventually tamed RD-180 class engines which have almost as high of chamber pressure, but how reliable will they really be, and how long lived will they really be? I just worry that SpaceX is trying to have its cake and eat it to, by pushing the bleeding edge of performance always, while also trying to push down manufacturing costs and up reusability at the same time. My guess is something is going to give–either the engines will end up being more expensive and finicky, or less reliable than they’ll really want for such a system, or nowhere near as reusable. I just don’t think 4500psi staged combustion seems like a good recipe for a 1000 flight engine. And the failure modes of a 4500psi staged combustion engine when you have 42 of them on your first stage also doesn’t sound likely to be graceful. I could, and genuinely hope I’m wrong. Once again repeating some of the mistakes of the Shuttle by trying to push crazy performance out of their first attempt at a staged-combustion rocket engine.
Ticket Price Economics Thoughts:
- Per-Vehicle Costs: BFR costs seem to be assuming a per kg cost less than half that of F9 FT. Which seems optimistic to me given the higher performance, more complex engines, and the use of composites for the propellant tanks, and the general scale of the thing. Once again, trying to do a Big Dumb Booster with bleeding-edge performance. But it’s the ITS that I’m really skeptical about. You’re really going to make a spacecraft that has to have all the life support capabilities of ISS, but for 16x the crew, and cram it into a high performance upper stage, a reentry/landing vehicle, and all of that for less than half the cost of a 747-8? Especially given that you’re likely only making a tiny fraction of the number per year. Dragon currently costs probably $30-40M each to produce, and we’re saying that a Dragon designed for 14x the number of people, and 30-50x the duration, with a nearly Saturn V first stage class propulsion system built in is going to only cost 5x as much? Color me extremely skeptical. I think that they’d be lucky to have the production cost of an ITS with all of its subsystems necessary to get 100 people safely to Mars and return reliably down below $1B each anytime in the foreseeable future.
- BFR Reuse Numbers: I’m also really skeptical they’ll get BFR reliability or engine life high enough to get anywhere near the 1000 reuses they’re claiming. I think they’d be lucky to average 100 flights each once again through the foreseeable future, based on the technology choices they’ve made.
The upshot is that if I’m right on those three items, you’re still talking less than $500M per Mars mission, and a ticket price in the ~$5M per person price range. That’s still three orders of magnitude better than what NASA could realistically do with its architecture. So while I think Elon doesn’t have an architecture that really gets down into the “cost of a median US house” range, he is getting into a range that a lot of people could afford. Having a Mars architecture this affordable would still be absolutely amazing, even if I think it could be done better.
How would I do things differently? Honestly I haven’t put as much thought into it as Elon has, but I have a few high-level thoughts:
- Split Things Up More: I’d separate TMI propulsion, the actual transfer habitat, and the Mars to surface and back into three separate elements. If you combine with the next point, the TMI stage can literally just be a normal upper stage like the Falcon 9 upper stage or ACES. The transfer hab would want to aerocapture at Mars, but could do so with a much more modest propulsion, and the Mars landing system can be smaller and higher flight rate. Honestly I think developing three systems that are more optimally split like this will not only cost less to develop than the swiss army knife approach, but will also be lower cost to operate, and open things up more for technology advancements over time.
- Go a Bit Smaller: Unless Induced Torpor works out, 100 people in a single vehicle seems really big for the transport stage. Breaking things up into convoys of smaller say 10-20 person vehicles might make more sense. This would mean transfer habitats would be small enough that you could use a TMI stage that is more reasonably sized for use in Cislunar space, so you don’t need a dedicated TMI stage. The transfer habs could be small enough that you have a range of options for aerocapture (inflatable, deployable, in-space assembled, or if it pans out magnetoshell aerocapture). These transfer habs will also have a lot more in common with LEO and cislunar orbital habitats, and possibly early mars surface habitats.
- Post TMI or TEI Boostback: I think Dave Masten and I have discussed this in the past, and Robert Zubrin hit on it today in his comments–It probably makes sense to have a separate TMI propulsion system that does the equivalence of a boostback maneuver after the TMI burn is done, to decelerate back into a highly-elliptical Earth orbit, where it can then aerobrake back to LEO. By not sending that along with the transfer hab, you enable it to be reused a lot more, since it’s not tied up for four years now. While not doing transfers to Mars, it can be sending payloads to/from Cislunar space, or to/from GEO. On a similar note, you could send one or two TMI stages along with the transfer habs to serve as TEI stages on the Mars side of things, using a similar post-TEI boostback maneuver and aerobraking to return back to Mars orbit for reuse. The more of your architecture is in the “can get 100+ flights in its lifetime” category vs “synodics mean I can only fly 12x in my lifetime” category, the cheaper things will be overall.
- Reusable Horizontal Mars Landers: Having separate landers, possibly smaller than the transit habs makes a lot of sense. The same landers can be used both for hauling people/cargo to/from Mars during arrival season, but also can be used for prepositioning propellant in Mars orbit, and Martian suborbital point-to-point transportation when not being used for Mars arrival landing. Horizontal landing on Mars is trickier than on the Moon, and I’m not totally wedded to the concept, but it seems like a much better way of getting people and heavy cargo onto/off of the surface.
- Using LH2 for More of the In-Space Elements: Once you’ve split-up Mars landing/ascent from the TMI/TEI burns, it makes sense to start looking again at LOX/LH2 for those segments. Those are two of the highest delta-V portions of your mission, so the higher performance could help. And LOX/LH2 can be made from Martian, Lunar, NEO, and possibly Phobos/Deimos ISRU sources. I know SpaceX is allergic to LH2, but most of the people I know who’ve worked with it have said “sure it’s a pain, but it’s not as evil as people make it sound”.
- Find Ways to Use Transfer Habs for Other Destinations: Say you can only realistically send a reusable Mars transfer hab on every other Mars window. That leaves a decent amount of down-time in-between. If the hab could be used for say taking tourists to/form the Moon or Venus when waiting for the planets to re-align from Mars, you can get much higher utilization out of the transfer hab elements. If you can take the one element in your system that currently can only be amortized over 12 flights (~50 year lifetime), and add in cislunar trips say during the “off-season”, you’re now amortizing it over 100+ flights instead of just 12. If you look at Elon’s architecture, 2/3 of the cost of a Mars ticket is due to the transfer hab’s low number of flights (the same if you make my more pessimistic hardware costs). If you could divide that over 100 flights instead of 12, that would make more of a difference for Mars ticket prices than almost anything else. Could you theoretically do this with the ITS as is? Sure, but without a source of CH4 on the Moon, you’d need to fly a lot more lunar tankers to make that work. Not impossible, but the economics aren’t as good as it would be if ITS could run on LOX/LH2.
- Leverage the Moon and NEOs for ISRU More: I still think there are ways that lunar ISRU can eventually beat earth-launched RLV prices for propellants in orbit. Especially if you stage out of a highly-elliptical earth orbit or EML-1 or 2. Investigating if there’s a way to tap into that wouldn’t be a bad idea, and would be a lot easier with the other suggestions I’ve given above.
Anyhow, that’s kind of off-the-cuff, but those are some of how I’d do things differently. As I said above, Elon has a lot of great architecture ideas, but I really don’t think he’s found the “One True Way” to get people to Mars as inexpensively as possible. Worlds better than NASA’s Journey to Mars? Definitely. Technically feasible? Probably. Cheap enough to be interesting? Sure. The best path forward from where we are today? That’s what I’d quibble with.

Jonathan Goff

Latest posts by Jonathan Goff (see all)
- NASA’s Selection of the Blue Moon Lander for Artemis V - May 25, 2023
- Fill ‘er Up: New AIAA Aerospace America Article on Propellant Depots - September 2, 2022
- Independent Perspectives on Cislunar Depotization - August 26, 2022
“” A couple extra decades of technology demonstration and robotic precursor missions are a real problem for that “”
But such missions conform more to reality than one man’s hubris-driven dream.
There’s a whiff of desperation among many space enthusiasts. They are understandably despondent that great things have not been done ion space with any regularity, and with big chunks of downtime. Such people want to see these long-cherished dreams be fulfilled in their lifetimes. But crash programs that neglect the still-existing need to do the preliminaries are a bad idea — especially when they involve wanting to great a very, very large colony virtually right off the bat. It is a fantasy in a vacuum — no pun intended.
Pardon the typos. I saw no edit functions, and dashed this (above) out in a hurry. 🙂
Bob, Chris, Roderick,
A friend once suggested that a good definition for a shortcut was “the longest possible path between any two points.” Overestimating how fast you can really move on a project can cause you to lock things down before you ought to, which could drive up the cost of your overall architecture to the point where it fails (either due to not being able to raise the money, or running into some other problem with how overambitious the scale is). With how much potential Torpor and MAC have, I’d take Elon more seriously if he were working with those companies to pay to accelerate when we can do a proof-of-concept test. He doesn’t have to do that in-house at SpaceX, but even 1% of the money he’s probably spent to-date on Raptor and that big ITS tank would likely pay for a cubesat MAC flight demo that could be flown before Raptor was even ready for flight. Or could pay for some ground tests of the induced Torpor. There’s no reason those technologies have to be drug out over a decade other than scarcity of funds.
~Jon
~Jon
@Jon, Bob, Chris, etc:
Regarding timelines and technologies that have yet to be tested, I think the key thing to remember is that SpaceX isn’t just proposing this as a possible way to do it, they are telling us about what they are building right now.
To fly people in eight years, you need to fly unmanned full-scale lander to Mars in six years, and to do that you need to have it working nicely by then. Given how long Falcon Heavy is taking, I’d guess they are not taking any chances and hope to have part of it flying (at least suborbital) in under two years. That doesn’t give much time to develop new technologies of possible inclusion in the system, that time has past.
They have a system they think they can build, and that they think will work. What more do they need?
And again on size, the large scale of the system should be considered a feature, not a bug. In the Q&A, Elon mentioned building parts of the system in some of the same states that built the Saturns. If SpaceX hopes to get NASA/govt. finance in a “public private partnership” or whatever, this can only help. Those cities might think, it was awesome when we built the rockets to go to the moon, and this one is twice as big! Interpreted, twice as awesome.
in situ says:
” To fly people in eight years, you need to fly unmanned full-scale lander to Mars in six years, and to do that you need to have it working nicely by then. Given how long Falcon Heavy is taking, I’d guess they are not taking any chances and hope to have part of it flying (at least suborbital) in under two years ”
Essentially, whether you intended to or not, this paragraph “argues” that the SpaceX timeline is woefully unrealistic.
Any smart plan would take longer, and would be scaled back in size and scope at first for a considerable time. Ironically, unlike the SpaceX operational model that exists now, this Mars proposal makes huge assumptions about success rates. It is “success-oriented” in the way that the Shuttle program was. Not a good model, all things considered.
I believe that Musk used deltaV of his intended earth return burn to size the ITS vehicle. The repeated claim that he will reuse the vehicle every 26 month synod, the roughly 9 km/s total deltaV and the specified earth entry velocity of 12.5 km/s all match a trajectory which would involve about a 100 day stay on mars followed by a 1.3 YEAR long trajectory with apohelion beyond mars orbit which would bring the craft back to Earth just in time for the next departure window. It’s the only trajectory that fits all of the data as unbelievable as it seems.
Personally I consider this trajectory a complete non-starter on human health factors as it is about a month longer then even the world record held by Valeri Polyakov and would expose passengers to a huge radiation dosage and a crippling duration of zero-g. This seems to be why Musk never talks about the return leg of the trip other then to say it is ‘free’ but the dirty little secret seems to be it’s a free deathtrap. So I think he is so really expecting an almost entirely one-way colonization process.
Impaler,
If that really is what’s sizing the ITS, then it seems like refueling in Mars orbit before the return trip would be as much of a no-brainer as refueling in Earth Orbit before flying to Mars…
~Jon
Yes that’s, the assessment I came too as well, I read your description of a refueling choreography but think I have a simpler one. Send one full tanker to mars aero-braking into orbit and then down into LMO, I calculate 720 mt of propellant with a slow hohmann transfer and no propellant consumed at arrival. Send the crew carrying ship directly to the surface as planned and fill with just enough propellant to reach orbit which should be only around 300 mt on the surface rather then 1950 mt, a huge reduction in the ISPP need. Then in orbit take on all the propellant needed for the rest of the return transfer which is only around 500 mt which leaves the tanker will a good reserve. The tanker can then either make it’s own slow return to Earth or land on mars if it’s needed their, if the tanker is at it’s end of lifespan then this makes an excellent swan-song mission as it won’t hurt amortization. This makes for a much safer startup and then transitions into using more local propellant by just altering the choreography but leaving the vehicles as they are.
Now in reality I would prefer to use this large propellant and available DeltaV to make a return to Earth at conjunction with inbound time of flight on par with the 3-5 month outbound flight. Only when both flight times are at that length can I believe that health issues have a chance of just being ‘outrun’ which is the strategy Musk embraces. Unfortunately a conjunction return mean a 2 synod cycle and half the amortization chances for the vehicle which leads directly into Zubrin’s spitting of the TMI propulsion system from the lander which I agree with as a significant improvement over the ‘Swiss army knife’ vehicle as it should both lower development and operating costs.
A smaller lander with about 5-6 km/s total DeltaV could be lighter and shorter, improving launch abort speed at Earth and stability on landing and ease of egress at Mars, ideally having the cargo bay be at the base so large monolithic equipment or vehicles can be simply rolled out via ramps. The 5-6 km/s range is I think the sweet-spot for a lander as it would allow the vehicle to cycle between mars surface and orbit taking cargo down and small amounts of propellant up each time to be stored in a propellant depot in preparation for the synod window, and if completely empty of any cargo or passengers the 6 km/s gets the vehicle back to Earth directly from the surface on a slow hohmann. This gets you good amortization at Mars and a vehicle that can be as high as 20 percent dry weight rather then in single digits.
TMI from highly eccentric earth orbit requires less delta-V than from LEO. A vehicle to Mars could be assembled/fueled in HEEO gradually, with ferry vehicles of modest size, even if the systems that perform the final TMI (at perigee) accompany the vehicle on to Mars.
Pingback: The case for space – Part 3: Martian delusions | daryanenergyblog