SpaceX's Starship SN8 explosion - what you should know
On 9 December 2020 5:45PM local time, the latest Starship prototype, SN8 -- short for Serial Number 8 -- took off from SpaceX’s facility near Boca Chica in South Texas. It soared 12.5 kilometers into the sky (higher than most commercial jets cruise), executed a number of complex aerial maneuvers, and 6 ½ minutes later crash-landed near the launch stand in a blazing ball of fire:
(Source)
You can watch the whole thing here.
At first blush this seems like a devastating failure. This is why SpaceX CEO Elon Musk’s reaction is noteworthy:
Musk elaborated on what happened at the end in a follow-up tweet:
You’re probably perplexed at this point. What’s all this about low fuel header tank pressure? What’s so exciting about getting some data that you’d overlook such a catastrophic and costly ending? What’s the context for all of this? And what on earth is RUD anyway?
To make sense of all this, it helps to take a couple of steps back.
Elon Musk founded SpaceX in 2002 with the goal of radically lowering space transportation costs to enable the colonization of Mars, as a first step in making humanity a multi-planetary species. His reasoning was straightforward, albeit Spock-like in logic and staggering in scope:
if we stay on Earth forever, we’ll get wiped out by some eventual mass extinction event, either natural (of which there have been several) or, more likely, manmade [1]
Becoming a multi-planetary species is therefore “life insurance” against any single event taking out humanity (Musk considers this a “strong humanitarian argument”, just as fundamental a human concern as eradicating disease and global poverty)
The first step is to colonize one planet, because this creates a “a strong economic forcing function to improve space travel, in turn facilitating the colonization of the whole solar system”
The logical first choice is Mars because all other options are far less suited to scale up to become a self-sustaining civilization [2]. Mars has decent sunlight and is not too cold, has a primarily CO2 atmosphere that lets us grow plants, has a day length close to Earth’s at 24.5 hours, is resource-rich, and is ultimately more terraformable
These factors matter because the idea isn’t just to establish an outpost, but to build a self-sustaining civilization, which requires “recreating the entire industrial base in a much more difficult environment than Earth -- no trees, no oxygen, no oil”. In Musk’s estimate, this requires at least a million people [3]
For Musk the bottleneck isn’t technology but cost: “there is no intersection of the sets of people who want to go and people who can afford to go”. Using “traditional methods” (e.g. Apollo program-style approach) you can optimistically get a $10 billion per person price tag. To get a million people, Musk wants to bring this down to under $500,000 [4]
This radical cost reduction, along with the challenge of ferrying enough people and cargo to create an entire self-sustaining civilization (his estimate: “100,000 trips of a giant spaceship in a century”), requires fundamental breakthroughs in rocketry, so Musk founded SpaceX with that goal (keep the ‘giant spaceship’ in mind)
To fund the R&D to develop these fundamental breakthroughs, SpaceX’s transport system would be architected so it “can pay for itself through economic spaceflight activities in the near-Earth space zone” like delivering payloads for governments and corporate clients, and (more profitably) the Starlink satellite mega-constellation [5]
Tim Urban at Wait But Why has a cute infographic summarizing SpaceX’s plan:
Eventually, Musk’s self-sustaining Martian civilization may look something like this artist’s impression:
(Source)
Two remarks:
Going to Mars sounds like typical tech billionaire outlandishness, but Musk isn’t alone. In 2015, NASA published a 35-page report outlining its plan to “establish a sustainable human presence on Mars, not to visit but to stay, within the next few decades”. In 2016, Boeing announced that it would beat SpaceX to Mars. In 2017, the UAE revealed its goal of establishing “the first inhabitable human settlement on the Red Planet by 2117”. And the idea of colonizing Mars has been a staple of science fiction for decades. What separates Musk from the rest is his contextualization above: while “humans on Mars” seems to be the end goal for the latter, it’s the first step for him in his vision of making humanity multi-planetary as life insurance against existential risks
A common impression is that the option of going to Mars is “an escape hatch for the rich”. Musk notes that this doesn’t make sense: “Your probability of dying on Mars is much higher than Earth. Really the ad for going to Mars would be like Shackleton’s ad for going to the Antarctic [in 1914]. It’s gonna be hard. There’s a good chance of death, going in a little can through deep space. You might land successfully. Once you land successfully, ... there's a good chance you'll die there. We think you can come back; but we're not sure.”
But why is going to Mars, or just space in general, so expensive in the first place?
Fundamentally, your spacecraft needs enough velocity to escape Earth’s gravity well: over 40,000 km/h. This requires a lot of energy. The only way we know how is via chemical rockets, so this entails a lot of fuel: ~90% of a rocket’s mass is propellant. This means very thin error margins for very high-performance engineering carrying very sensitive payloads. Getting this right costs a lot of money
Traditionally, rockets weren’t reusable. (Including the Space Shuttle [6].) Musk calls this “the fundamental problem with space exploration”. Throwing away all that expensive hardware after every launch costs a lot of money
Current systems can only bring 3-8 astronauts per trip, so sending each person costs a lot of money
The only companies in aerospace are huge, therefore risk-averse. This often entails using proven legacy components half a century old, like Russian rocket engines built in the 60s (not designed in the 60s, built)
Big aerospace companies tend to outsource everything to subcontractors, who in turn outsource to sub-subcontractors, 4-5 layers down. 4-5 layers of profit overhead costs a lot of money [7]
So how is SpaceX’s Mars transportation system supposed to lower cost with respect to the factors above? Recall that Musk needs to reduce per person cost from ~$10 billion to <$500,000, or around 4.5 orders of magnitude. Respectively:
SpaceX is pioneering the use of methane as fuel, which is harder to work with than industry-standard kerosene & hydrogen, but gives more thrust, and exists aplenty on Mars. That said, there isn’t much cost reduction here
SpaceX first achieved (partial) reusability in 2017 with the Falcon 9 rocket, so named because it uses 9 Merlin engines. Musk eventually eventually expects full reusability to drive costs down by 2-2.5 orders of mag, making it the biggest cost reduction driver
SpaceX plans to carry 100+ people per Mars trip in their Starship, still in development, driving down cost by 1-1.5 orders of mag (more on this below)
SpaceX’s “move fast and break things” approach to rocketry R&D is the exact opposite of aerospace incumbents’ risk aversion, and has paid off very early -- SpaceX held the record for the most cost-efficient rocket ever to launch as far back as the Falcon 1 [8]. It’s also led to them using methane as rocket fuel, vs kerosene like everyone else [13]
SpaceX is heavily vertically integrated, manufacturing 80-90% of its rockets, engines, electronics and so on. This keeps costs down
This brings us to the SpaceX Starship system, the “giant spaceship” referred to in the Mars plan above. (It’s gone through a number of name changes over the years. At first it was the Mars Colonial Transporter or MCT. Later Musk renamed it to Interplanetary Transport System or ITS, to emphasize that the system would be able to go “well beyond Mars”. That name only lasted a year before it was rechristened the SpaceX Starship system.)
Here it is in the middle; to the left is the Apollo program’s Saturn V rocket, the largest and most powerful ever built, and to the right is SpaceX’s Falcon 9 [12].
(Source)
There are two stages to this gargantuan launch vehicle: a reusable booster stage called Super Heavy powered by 28 Raptor engines (with 3 times the thrust-to-weight ratio of the Merlin engine, already the most efficient rocket engine ever), which will make it the most powerful rocket ever built, and a reusable spacecraft doubling as a second stage called Starship with 6 Raptor engines.
(Confusingly, both the entire system and the second stage/reusable spacecraft are called Starship. It shouldn’t be too hard to disambiguate in context though.)
Starship is a magnificent piece of engineering, straight out of the Golden Age of science fiction. The picture below is of a test flight version, not the final design (which is why it’s missing its body flaps -- more below), but it’s still glorious:
(Source - real picture by the way)
No, seriously -- here’s Ray Bradbury’s The Martian Chronicles, serialized in 1946; tell me those aren’t Starships:
(Source)
Starship will execute a “belly flop” maneuver in the process of landing on Mars. Or if you prefer a diagram:
(Source - the details differ for Earth and Mars because Earth’s atmosphere is over 100x denser)
It’s worth asking why do such a complex maneuver, without precedent in the history of spaceflight, since Falcon 9’s first stage does propulsive landing more straightforwardly:
(Source)
The reason is higher reentry velocity. Propulsive landing is easier for the the first stage booster because it’s ejected only 2 minutes after launch and so returns to Earth from a lower altitude, hence “only” reaching low hypersonic speeds of about Mach 6, or 6x the speed of sound at sea level i.e. ~7,200 km/h.
Starship can’t do this because it’ll be returning from orbit at ~Mach 25 (NASA calls this regime “high hypersonic speed” [9], where the dominant design consideration is thermal control), so it needs to dissipate 99% of its energy via aerobraking before reigniting its engines to bring it to a full stop. But aerobraking converts all that kinetic energy to heat, so it needs a heat shield to protect the payload. Traditionally heat shields are something like this artist’s rendition of the Apollo command module returning to Earth:
(Source)
But this design isn’t reusable. The Space Shuttle orbiter is reusable, but it used huge wings to glide onto a runway; heavy wings reduce payload capacity, and anyway there’s no runway on Mars.
SpaceX’s radical solution is to have Starship do a belly flop, i.e. enter the atmosphere at a 90 degree angle, meaning it’s fully stalled. So it uses one side of entire body as a heat shield [10], as in this artist’s rendering:
(Source)
This makes Starship unique, because normally aircraft avoid stalling -- it’s a condition where an airfoil’s angle of attack goes beyond a certain critical angle, like in this illustration of a plane wing cross-section:
(Source)
This flow separation leads to a decrease in lift that makes the aircraft fall, as in this example diagram:
(Source)
Which is precisely what Starship intends to do to aerobrake. The issue is that a full stall is inherently unstable, with aerodynamics impossible to predict, which is why it’s never been tried before. (Think of a leaf fluttering to the ground, instead of dropping in a predictable trajectory.) This is where active control comes in. Starship has 4 electrically-driven flaps used to control its belly flop freefall, a little like a skydiver uses her limbs.
These are the rear body flaps:
(Source)
Again, since these complex aerial maneuvers have never been done before in the history of spaceflight, SpaceX wants to collect as much data as it can; consistent with their “move fast break things” approach, they’re willing to sacrifice a lot of testbeds to do so. This is precisely what the Serial Number X prototypes are: testbeds for collecting data to refine aerial maneuvers.
Now we’re in a position to understand SN8. It was the most ambitious such test yet, flying ~100x higher than the highest previous test, testing how well the electrically-driven flaps could stabilize Starship in full stall belly-flop position as it descended, and checking SpaceX’s own aerodynamic modeling of the descent.
SN8 tested the subsonic portion of Starship’s landing:
(Source)
Musk himself gave it a 1 in 3 chance of success:
(Source)
Given these expectations, it’s no wonder Musk & Co. were elated by what happened: SN8 comfortably surpassed expectations. That SN8 was completely destroyed wasn’t a costly setback, because it wasn’t that expensive, perhaps $15-20 million [11], and at this point they can churn out a new prototype every month or so.
This tweet summarizes SN8’s “scorecard” concisely:
(Source)
But why did SN8 fail to stick the landing? Let’s go back to Musk’s tweet above:
In a bit more detail: Starship has two sets of liquid fuel tanks, the big ones used for ascent and small fuel header ones for propulsive landing. Each set comprises one oxygen tank and one methane tank. On the way up, gravity and thrust provide the pressure that ensures the engines are constantly getting propellant, but on the way down Starship is in freefall so the main tanks lose pressure. The header tanks are kept under pressure with small cylinders of helium gas. The helium gas cylinder for the methane header tank failed, so the engines got too much oxygen and not enough methane. This meant that not all the oxygen was burned, so thrust plummeted and SN8 hence failed to slow down enough, leading to what SpaceX employees call “Rapid Unscheduled Disassembly”, or RUD, which is another way to say “rocket exploded”.
The thing about superhot pure oxygen is that anything that can burn pretty much will burn with it. This includes the main part of the Raptor engine’s “bell housing”:
(Source)
This bell is made primarily of a bronze alloy, which contains copper, and copper burns bright green in oxygen. That’s why the exhaust flames turned green towards the end:
(Source)
So what’s next?
SpaceX engineers will fix the low pressure header tank issue to ensure adequate thrust throughout propulsive landing, then SN9 will be a high-altitude “hop” attempt, and later prototypes will test more extreme aerodynamic regimes -- transonic, supersonic, hypersonic -- sticking to suborbital tests in the near term. Then they’ll test the Super Heavy first booster stage, both launching & landing. They’ll refine Starship’s heat shield further to ensure it survives multiple reentries from orbital velocity, as will be needed to shuttle cost-effectively between Earth and Mars. After that, they should be testing the next critical piece of the Starship system’s cost-lowering infrastructure -- in-orbit refueling:
(Source)
Given that Mars launches are limited to a window of every 26 months due to Earth-Mars orbital alignment, the schedule above means they’ll likely miss the 2022 window, so the earliest SpaceX will do unmanned Starship trips will be in 2024, and the earliest manned trips in 2026. Here’s what the schedule looked like back in 2018:
My own vague impression, from SpaceX’s progress over the years, is that Musk tends to overpromise timelines by 2-3x, so I don’t think we’ll see manned trips to Mars this decade. But I’d love to be proven wrong!
Footnotes
[1] Toby Ord, senior fellow at Oxford’s Future of Humanity Institute, summarizes the existential risk research community’s best current estimates of the probability that humanity will go extinct due to natural or manmade catastrophes in the following table, from his book The Precipice. This underscores how much greater the existential risks of our own making are:
[2] Other options considered and rejected as first step, albeit still potential options much further down the road:
Venus is a high-pressure high-temperature acid bath (although its upper atmosphere is more habitable)
Mercury is too close to the Sun, and has no protective atmosphere
The larger moons of Jupiter and Saturn are much further out than Mars
The Moon, although much closer, doesn’t have an atmosphere, is much smaller and less resource-rich, and has a much longer day, so while it makes a great outpost it’s less suited to ultimately scale up to be a self-sustaining civilization
[3] The minimum viable population required to be robust against extinction from natural disasters and genetic/environmental stochasticity is much less than a million, but it isn’t the real minimum bound here. The minimum bound is what’s needed to recreate an entire industrial base necessary to support a self-sustaining civilization. Musk noted that “even at a million, you’re really assuming an incredible amount of productivity per person”.
[4] Musk has a cute infographic illustrating this:
[5] Musk said that the goal for Starlink is to “rebuild the Internet in space” by having the majority of long-distance Internet traffic go over this network and about 10 percent of local consumer and business traffic. Revenue-wise, Musk said that Starlink makes sense for SpaceX because there’s more money in satellite manufacturing than in space-launch services: “Looking at the long term, what’s needed to create a city on Mars? Well, one thing’s for sure – a lot of money. So we need things that will generate a lot of money”, but doesn’t give any estimates.
This hasn’t stopped others from trying. For instance, Morgan Stanley analyst Adam Jonas estimated that by 2040, Starlink will bring in $21 per customer per month from 364 million subscribers, or ~5% of the world population, for an annual gross revenue of $92 billion. (This is a lot, but not a lot; Walmart for instance made $524 billion in revenue in 2019.) I’m not sure however how Jonas arrives at those figures, and personally think 5% penetration is a bit on the low end.
[6] Only the Space Shuttle orbiter was really reusable. The rockets needed 9 months to refurbish, which nullified the cost benefits; the large orange gas tank wasn’t reusable at all. The original vision was $20 million per flight, as often as once a week. In reality, the fleet never flew more than 9 missions in a single year, and the $209 billion cost of the entire program for 134 missions works out to ~$1.6 billion per flight, or 80x more expensive than envisioned.
[7] Musk criticized the aerospace industry’s outsourcing practices: “You have to go four or five layers down to find somebody actually doing something useful—actually cutting metal, shaping atoms. Every level above that tacks on profit—it’s overhead to the fifth power. … [I asked], What is a rocket made of? Aerospace-grade aluminum alloys, plus some titanium, copper, and carbon fiber. And then I asked, what is the value of those materials on the commodity market? It turned out that the materials cost of a rocket was around 2 percent of the typical price—which is a crazy ratio for a large mechanical product…So, I thought, we should be able to make a much cheaper rocket given those materials costs.”
[8] SpaceX’s total development cost for both the Falcon 1 and Falcon 9 rockets, according to NASA’s own independently verified numbers, was estimated at ~$390 million. NASA estimated in 2011 that based upon its own traditional contracting processes, it would have cost $4 billion to develop a rocket like Falcon 9 (10x more).
[9] Wikipedia has an informative classification of Mach regimes that explains why you’d want to distinguish between “low” and “high” hypersonic speeds, the former relevant for the first-stage Super Heavy booster and the latter for the second-stage Starship:
[10] It’s worth noting that Starship’s heat shield is one of their biggest unknowns at the moment. Every test so far has had failed heat shield tiles from short hops (to say nothing of hypersonic reentries), and each test tried a different configuration. Musk has tweeted that shielding the areas around the moving wing flaps is a big problem:
[11] The usual quote for a complete Starship is $20 million, so SN8 (being a prototype) should cost less. Once upfront costs are paid for, $20M is believable given the following assumptions:
Starship weighs 100 tons, mostly 304 stainless steel at ~$3,400/ton, for ~$350k
The Raptor engines are only ~$1M each because they’re 3D-printed, and SN8 only had 3 instead of 6 for the final configuration
The motors & batteries are Tesla car parts
The computers, pumps, valves etc are made in-house (because SpaceX is vertically integrated)
Assuming 1,000 employees at Boca Chica paid $100k/yr (high side, given most are construction workers) and the capacity to turn out a Starship prototype every month gives ~$8M labor cost
[12] And this is just Starship V1, which is 9m across. Musk already has plans for Starship V2, which will be double the diameter and 4x the payload capacity (ignore the LEO figures in the illustration, they don’t make sense -- ultimately what limits the rocket’s height is the Raptor’s thrust-to-weight ratio; since the number of Raptor engines only scales as the square of rocket size, so does payload capacity, height being constant):
[13] There are many advantages to using methane over kerosene or hydrogen:
Methane engines produce higher specific impulse (a measure of how effectively rockets use fuel: how many seconds can this propellant accelerate its own mass at 1g?)
Methane is far easier to store and handle than hydrogen
Methane takes up far less volume, so fuel tanks can be smaller and lighter
Methane burns cleaner than kerosene (its only byproducts are CO2 and water), eliminating problems with carbon deposits causing “hot spots” that can endanger the rocket, complicating overhaul before reuse
Methane can be made anywhere in space given power and raw materials, including on Mars
Methane is cheap and plentiful on Earth (as natural gas)
Methane’s boiling point is close to oxygen’s, so you can easily prevent one from freezing the other with a little insulation. In traditional rockets, liquid hydrogen can freeze oxygen, and liquid kerosene can be frozen by liquid oxygen, so extra weight is needed to insulate them, reducing payload, driving up cost