Rockets of Today


Of all the articles on this page, this is the most updated and revised, due to the frequent and public changes to the design of the proposed rocket. And also due to the fact that of all the rockets listed here, this one is the most truly revolutionary. If it works, it will expand human access to the solar system in a fundamental way that no other rocket can match. There may be other big rockets which on paper could launch a hundred ton spaceship, but no others have a way of then loading that ship with a thousand tons of fuel. And once you’re up in space with that much fuel... you can go anywhere.

Elon Musk’s SpaceX may not have announced anything to compete directly with the New Glenn, but what they’re working on now is a leap right past it: the most ambitious rocket design ever budgeted. Its goal is no less than to enable colonization of other planets! This rocket starts with an enormously powerful methane-burning reusable booster, far larger than the New Glenn, and on top of it, they plan to put a huge passenger-carrying spacecraft. And this is the scaled-down version: originally they wanted to make it twelve meters wide, but in 2017 they reduced that to nine meters — not quite as big around as a Saturn V. Since then that diameter has remained unchanged, though there have been constant changes to the height, the expected weight, the number of engines, the aerodynamic surfaces, and even the construction material.

This spacecraft would have plenty of engine power, and would act as the second stage. But unlike any previous upper stage, it would be fully reusable, thereby eliminating the last part of a traditional rocket stack that has to be thrown away, and getting them much closer to the dream of a spacecraft which can be reflown for only the cost of the fuel.

orbital refueling

To go to Mars or wherever, this spacecraft would put itself in orbit with passengers aboard, and then dock with a tanker for refueling. The tanker would just be another Starship, but with the partitions between the fuel tanks moved closer to the nose to increase its capacity. The passenger ship’s tanks can hold about twelve hundred tons of propellant (the lox being most of the weight), so it will take many tanker flights to fill up the passenger vessel. They are now talking about storing the fuel in an orbiting depot, which would just be a tanker with thermal insulation instead of a heat shield, and maybe solar powered refrigeration, so that the passenger ship can do all its refueling at once. Such a depot might be able to store fuel for many months with very little evaporative loss.

At first they claimed the payload would be 150 tons, then they said 100, and now they only say it will end up somewhere between 100 and 200. Fewer tanker loads will suffice if you don’t need the maximum delta-V. The tanker could be launched by another flight of the same booster, which is designed to be reused rapidly with minimal maintenance. They’ve talked about eventually doing tanker launches as rapidly as three times a day with one booster, which certainly won’t happen in the short term. One part of this hope is the idea of landing the booster right back onto the launchpad, or close enough that it can be swung back into launching position with a simple crane built into the tower. Obviously this is a risky strategy, as a slip could destroy the launch complex. The way they ended up pursuing this was by building a tower with a pair of arms near the top which the booster’s grid fins would come to rest on, so it could then set the booster down onto the launch platform. The same arms would lift or catch the upper stage by its forward wings, and then lower it onto the booster. It would, of course, only support them with mostly empty tanks.

And this is where we start to see what a gamble the Starship plan is. Any launch beyond low orbit, whether to the Moon or the Gateway or Mars or Pluto, would require the ability to launch multiple tankers within a span of a few months — or without a good depot, a few weeks. If they follow a more traditional launch schedule, the fuel might be lost from the orbiting ship as fast as they can bring up more. And this rapid cadence also depends on being able to land the boosters and tankers right next to the pad, which means landing them where there’s a maximum risk that a crash could stop the whole operation for a year or more. Rapid reusability has to be completely nailed down as a safe, commonplace, routine operation before it will become possible to get good use out of the Starship for distant destinations, especially if passengers are on it.

Once fully fueled, the passenger ship would then cruise to Mars, possibly with some extra speed to shorten the trip, but I doubt that. It would use its heat shield to slow down in the Martian atmosphere, then slow the rest of the way with its engines, which would take more fuel than a landing on Earth despite the reduced gravity, because the air is too thin to slow anything to a comfortable subsonic speed.

To come back from Mars... well, they haven’t fully worked that out yet. As presently designed, it has enough fuel to land there but not to take off again. They are planning to devise a means for people on the surface of Mars to manufacture their own methane and lox out of carbon dioxide and water (2 H2O + 1 CO2 → 1 CH4 + 2 O2) via the Sabatier process, which involves splitting the water and then reacting the hydrogen with the CO2 via a metallic catalyst such as nickel. They hope to land machines for this purpose on an unmanned flight, and it’s conceivable they might be able to start filling up tanks of return fuel before the human passengers set off... but I’d bet that human beings would have to get the machine running after they arrive, as it will be hard to automate a process to dig up the ice it needs, which does exist on Mars but is not abundant or easily accessible, except at the poles.

And not only will they be faced with the daunting prospect of mining huge amounts of ice, but they will also need about a megawatt of electric power to run the process, which will not be easy to generate. Solar cells for that much power would have to be made very thin and lightweight, and therefore very fragile, unless an additional one-way rocket is allocated to bring them. Plus they would be subject to dust storms. The fuel production would have to be sustained for months without any mishaps or setbacks. All in all, the plan to get people back to Earth when or if necessary definitely needs work... the ideas they’ve got would work fine in a mature well-established colony where most people are staying permanently, but won’t work at all for the sort of mission where people explore and then return home. For that we’d probably have to make something much smaller to come home in.

Anyway, it is this ability to refuel in orbit, 100 or more tons at a time, which turns a rocket into a spaceship. Without that, it’s just a second stage with a nonremovable fairing, but with it, it gains a level of delta-V never previously available to heavy deep-space missions — up to about eight km/s depending on payload — making any body in the solar system reachable with a bit of gravity assist. Of course, if you put people in it, voyages beyond Mars are still implausible due to the time it would take, but the ship itself can get there. Give it a smaller rocket as a payload, and it should be no problem to perform previously impossible feats of delta-V such as soft-landing a probe on the surface of Io or Mercury, or flinging something past Voyager 1.

Raptor engine

Though on paper they had what looked like a complete design as early as 2016, first under the name “Interplanetary Transport System” and then the placeholder name “BFR”, very little of this was built or tested until 2019 except for the “Raptor” engines. Even the early versions were not all that much heavier than their existing Merlins but produced twice the thrust — about three fourths as much as Blue Origin’s far larger BE-4. These new motors get a much better specific impulse from methane than the Merlin can get from kerosene, thanks in part to an exceptionally high combustion chamber pressure of 300 bar — a goal which sounded unlikely but was achieved by the Raptor 2 version in 2020. (A Raptor 3 has reached 350 in tests. The BE-4 only uses 134 bar, a safer value which might lend itself to a longer reuse lifespan, but still outdoes the Merlin. The Space Shuttle’s main engines, which were the most advanced of the 20th century, reached 206 bar.)

Unlike the BE-4, or any previous engine, the Raptor uses “full flow staged combustion”. That enormous 300 bar chamber pressure is applied to all of the fuel and all of the oxidizer, as each is preburned with a minimal amount of the other to power the pump on its side. (On the methane side, the mostly unburned result is then used as coolant on the outside of the bell and combustion chamber.) This means it doesn’t lose specific impulse by having a separate low-pressure exhaust pipe coming out of the turbopump, as the Merlin does. The BE-4, by contrast, uses oxygen-rich partial staged combustion, which is the same approach preferred by Energomash. This puts it ahead of the Merlin in efficiency. The oxygen-rich and full-flow fuel cycles both face the challenge of handling hot oxygen, which can eat rapidly into most metals, but the temperatures are not all that high. SpaceX claimed “long life through... more benign turbine environments”, and said they have a new oxygen-resistant alloy to make the turbine from. But in typical SpaceX fashion, they take what could be a safety margin and instead push this new alloy even harder, and raise the pressure in the engine to even more unheard-of levels. To have 300 bar in the chamber, the pressure in the turbines is double that, and the pressure inside the pumps is even higher — estimated at 700 bar for the oxygen and near 900 for the methane. Both Energomash and Aerojet Rocketdyne have experimented with full flow, but no such engine has left the ground before. Staged combustion of any kind is rare in American engines, the RS-25 Space Shuttle Main Engine (which used a fuel-rich cycle) being the outstanding example. Most have not seen full-flow as worth the extra complexity, as ordinary one-sided staged combustion is almost as good.

It is probably tricky and complicated to start the Raptor up. That’s an issue with traditional staged combustion too; it’s why the Space Shuttle’s hydrogen engines had to be started seven seconds before lighting up the solid boosters. The Raptor has multiple turbine stages to produce all that pressure, making the issue even trickier. Even the simpler Merlin has to be started by first running helium through the turbine to get it turning. (Historical note: in the Apollo, the initial flow to get the pumps turning was powered by gravity. The rocket was so tall that even standing still, the lox entering the engine was under quite a bit of pressure.) With two separate sets of turbines on independent axles, both have to be spun up with gas, which in the current version is apparently high-pressure helium. Test flights have demonstrated that the Raptor can, as claimed, start up quickly. This would need to happen if, for instance, the Starship suddenly has to detach from a failing booster. The escape plan in this case is mostly to trust the ship’s steel hull to protect it from the booster’s fireball for one second, rather than to try to outrun it completely as you would have to with something like a solid-fuel booster.

The first full-scale Raptor was completed and fired early in 2019, apparently with substantial changes from the smaller prototypes, and later that year they started flying it in a low-altitude test vehicle. It did have issues with durability, with some tests producing what is euphemistically called “engine-rich exhaust” as its insides started to burn out, but apparently it fully met its ambitious goals for power and thrust. A year later they had the engine flying with no such issues... except when fuel-pressure problems made it run too lean. By 2022 they had revised the engine enough to call it Raptor 2, with the thrust increasing from 1.8 to 2.3 meganewtons, which might reach 2.5 later. If it does reach 2.5, it would be enough to beat the BE-4 while still being significantly smaller. (It would still take three of either one to equal the F-1 engine from the Saturn V. Three Raptors would be a lot smaller than an F-1, as it had a bell the size of a Dragon spacecraft. Three BE-4s would not be smaller.) The Raptor 2 is also very noticeably simplified from the original, having far fewer external parts and costing half as much to make. But they may still be struggling with reliability; tests still sometimes produce engine-rich exhaust. They are still pushing this simplification further — a version with the working name of LEET 1337 is now in the works, where they are trying to push the parts count as far down as it can go.

design and construction

Even as they were building test articles, fundamentals of the rocket’s overall construction kept undergoing deep changes. For instance, the intent had originally been to build the ship out of carbon composites, until in late 2018 Musk suddenly said they were moving toward metal construction. They decided that the most weight-efficient material that could survive re-entry might be, of all things, stainless steel! It can be polished to reflect away most radiant heat (which, astonishingly, scales with the eighth power of speed during reentry), and it can be cooled from the back with piped fluids — or in the tank sections, perhaps just by splashing the cryogenic propellants against the interior walls. At one point they said they would also put tiny pores in it so the coolant fluids (i.e. extra propellant) boil out to produce a gaseous protective layer... but this idea was fraught with risks, so they wisely backed away from it. They then developed a thin ceramic tile to put over the steel, getting it to re-emit as much radiant heat as possible, which means it’s near black in color. Because steel can tolerate far more heat than aluminum or carbon fiber, the tiles don’t have to deflect the entire heat load, which means they can be thinner and lighter than previously used materials such as the Space Shuttle’s tiles, or a traditional ablative heat shield. They can even have little gaps between them, which act as expansion joints. Any hot gases leaking through will first encounter a blanket of insulating fibers, and then the shiny steel surface. I’d had the impression that these tiles would be fairly thin ceramic plates, but it turns out they’re essentially made from the same formula as shuttle tiles, consisting of a low density block of fine silica fibers topped with a black ceramic glaze doped with boron. They are somewhat thinner, but not nearly as much so as I had thought. They get fastened to the ship with welded-on steel pegs that act as mounting points, which allow them to be easily removed and replaced... though getting them to stay securely attached to these pegs during launch has been a struggle.

To be more specific, the material they use is a special new chromium steel which is treated at cryogenic temperatures. Apparently it was the breakthrough which made this new stainless steel available which caused them to change their minds about using carbon fiber. They call it “30X” for now. Besides tolerating heat, this new steel also likes cold: it doesn’t turn all glassy and brittle when chilled to the temperature of liquid oxygen, as many materials do — in fact, it’s actually stronger cold than at room temperature, which for steel is kind of bizarre. Insulation for the tanks will essentially be nonexistent: the idea is to make the tank wall and the outer hull a single piece to avoid redundancy, but since steel is much more thermally conductive than carbon composite is, keeping things cool will be a problem. It looks like the current plan is to just accept the extra boiloff. Whether this might affect the capacity to keep fuel available on months-long missions may still be an open question, as liquid oxygen can produce a vapor pressure far beyond the capacity of the fuel tanks to contain, long before getting near room temperature.

In current prototypes the steel walls in the upper stage are about four millimeters thick — I don’t know if the booster uses a heavier gauge. With this thickness, the ship’s walls would weigh in the ballpark of forty tons, with the total dry weight for a tanker maybe being somewhere around eighty tons. They say they may be able to make sections of it thinner in the future. Areas that have to support a lot of weight, such as the skirt at the bottom end of each stage, are reinforced with additional corrugated layers.

Even after sending prototypes on high altitude hop tests they kept making design changes. Musk’s philosophy is to have no attachment to sunk costs, to have no reluctance to ask other engineers to change their part if it’s making your part difficult, and that if a design turns out to be difficult to implement, that means the design is flawed and should change. The result that this philosophy arrives at may turn out to look amazingly simple and crude, belying the amount of sophistication that went into pursuing complex alternatives. He intends to keep pursuing rapid iteration and revision as they build and launch numerous Starships, so that in a year or two they arrive at something much better than what they’ve got now. They are free to do this because so far, almost nobody is actually paying them to launch on one, so the only thing these test flights will be used for is putting up their own Starlink satellites.

Originally, SpaceX also announced plans to develop reaction-control thrusters that burn gaseous methane and oxygen, replacing the hypergolic “Draco” thrusters that the Dragon capsules are equipped with, as well as the inefficient compressed-nitrogen jets that they now use to steer Falcon boosters during re-entry. Musk said they would have quite a lot of thrust. I presume that these would be turbineless, and use electric pumps or heat to pre-pressurize their fuel so they can start instantly... but reliably igniting a gas mixture when surrounded by vacuum is a tricky challenge in itself, so it’s understandable why most makers of deep space thrusters would rather use hypergolics, or even monopropellant. But eventually they realized that they have no real need for high-impulse thrusters, that cold gas is adequate for the job... and they’ve got a plentiful supply of cold gas in the form of propellant boiloff — the “ullage gas” which pressurizes the tanks. Once in orbit, they were already planning to bleed that pressure down from around five bar to less than two, so there’s your thruster propellant for free. And in the booster, this boiloff gas is more plentiful and will handle the steering for reentry until the steel grid fins take over at lower altitudes. But I question whether dumping a lot of boiloff gas is a good idea on a lengthy cruise to Mars.

They also won’t need the large helium tank that the Falcon uses to pressurize its kerosene, since during launch the methane and oxygen can be pressurized enough to feed the thirsty turbopumps just by backfeeding some heat from the engines, once you reach a point where ambient warmth is no longer giving you all the pressure you can stand. Since they also plan to use spark plugs for ignition (like the Space Shuttle’s main engines), instead of the pyrophoric lighter fluid that kerosene-burning engines usually require, this could someday eliminate the need for any other consumable chemical to be supplied — an important consideration if your journey is starting from someplace other than Earth. But this is an ideal, and even approaching it will have to wait for a future version. The first high-altitude flight of a Starship prototype (SN-8) crashed because the autogenous pressurization didn’t work for the small “header” tank that stores fuel for the landing burn. This was positioned between the two main tanks, with the lox header up in the nose of the ship to balance the weight during reentry, so getting hot gas into them is not so simple. (For some reason, the main lox tank is at the bottom with the much lighter methane above it; I would think that during reentry it would balance better the other way around, so the weight is more centered. The booster also has the lox on the bottom, which is rare in first stages because you usually want the heavier part forward, as this makes the steering easier and more stable.) For a short while, they went back to using helium pressurant, but this caused a new failure mode so they switched back. And time will tell whether a spark igniter can be reliable enough, or whether they will be able to do engine starts without helium. I think the booster still uses helium for starting, plus a substantial amout of carbon dioxide for fire suppression in the engine bay.

The number of engines per stage has changed repeatedly. By 2020 they had settled on six engines for the ship, and dropped the number on the booster to 31, saying they might go as low as 28 if they can get the thrust high enough. In mid-2021 the plan was for 29, and by year’s end they said it would soon be 33. The upturn apparently arose from increases in the size of the ship on top, making the whole rocket even bigger. This high degree of numeric redundancy means that even if several booster engines have issues at once and shut down, they could still reach orbit just fine, as long as it didn’t happen too close to the start of the flight. It also means that propulsive landings would be safer, as they could also survive an engine failure. They now plan to burn three engines when landing the ship, and be prepared for any one of the three to fail. (Falcon booster landings traditionally depend on the central engine alone, even if two additional ones were sometimes used for part of the landing burn.) They feel that keeping the engine relatively small and (by rocket standards) mass-produced will be the secret to cheapness.

The booster engines will be nongimballed except for a central cluster of seven — no, nine — no, thirteen. The fixed outer engines will have higher thrust than the central ones, and have no ability to restart themselves — they are started with gases supplied by the launch tower. Currently there are twenty of these outer engines jammed closely together (which still don’t fit beneath the tank, so the bottom of the rocket has to flare out a bit). Inside those is a looser ring of ten steerable engines, and within this ring are three more at the center. They may yet cram more into the middle.

On the spaceship, the outer engines have large bells optimized for vacuum, about 2.4 meters in diameter, and these are likewise nongimballed. These are the engines they’d use for orbital maneuvers, such as sending the vehicle from Earth orbit toward Mars, and hopefully for the lion’s share of second stage boosting in the upper atmosphere. For most of Starship’s history they’ve only planned on three of these vacuum engines, but now they’re talking about increasing that to six for the second version, so that the tanker variant of it can be full of fuel all the way to the nose, and the primary variant can lift more payload. They also said they might put both header tanks in the nose for balance, which means more hot gas has to be fed forward for pressurization, but they may have dropped that.

The ship’s inner three engines have small bells, about the same as the inner engines on the booster, and can gimbal widely and quickly. These would be used for landing, and I think for takeoff from Mars. Those large-belled outer engines (called “R-Vac” for short) would have a claimed specific impulse of 3.7 km/s, possibly even 3.8, which is over mach ten.

At any diameter, they had hoped to give these engines the unusual ability to throttle down to just 20% of max thrust. The only current engine I’m aware of with such a range is the BE-3 hydrogen burner in the New Shepard, which needs to go that low in order to hover before landing. One issue is that low throttling at sea level is supposed to be incompatible with even a moderately large bell size. But this low throttling probably only needs to be used in low gravity and little or no atmosphere, not on Earth. The current Raptor 2 did not meet this goal; its minimum throttle is about 40%.

Throughout most of 2017 and 2018, I kept hoping that at some point they’d budget some time and attention to giving this “BFR” project a real name, like Condor or Pterodactyl or something, instead of Big, uh, Falcon Rocket. I took to just calling it “the Beef” for short. This allowed me to ask regularly where it is. Then late in 2018 they decided to name the upper stage “Starship”, and the lower stage “Super Heavy Booster”. These names are disappointing, since “Starship” manages to be both blandly generic and preposterously overblown, and “Super Heavy Booster” is just a category name that could be used by anyone.


Of course, this Mars colonization scheme may go nowhere, but if so, the Raptor should still be a very competitive engine or other uses. They could use it to make a cheaper Falcon with fewer motors, or make a new rocket in the fifty ton payload class, or make a compact model with one Raptor. Elon said that he hoped that the Raptor can make a reusable second stage possible for the Falcon; the Air Force was helping to fund some development for second stage use, but this was dropped. Lots of uses are possible; after all, once you’ve got the engines working, you’ve done two thirds of the development work — the rest of the booster is simpler stuff and takes less time, so all kinds of body designs are doable.

But while such uses might all be profitable, Musk is threatening to dedicate his life and fortune to promoting a Mars colonization effort involving thousands of ships, at a hoped-for cost of around $200,000 per colonist. Personally, I don’t see the need to rush. But here’s how seriously he takes the idea: he now says the plan is to replace their entire fleet of Falcons and Dragons with these new rockets, so that astronauts will ride a Beef — er, Starship to the space station, and commercial satellites will be launched from a Starship with a huge cargo bay that opens like an alligator jaw. This makes me envision the thing making stops at a dozen different orbits to drop off payloads, like a mailman doing his rounds... but just as likely, this means that a lot of flights are going to leave the Starship’s payload area mostly empty.

(Hey, maybe they could fit the interior with some kind of catapult, to help fling the satellites into different orbits. It might be a good use of all that available space, and would save fuel. Telescoping rails, perhaps? If it can unfold to forty meters long, then with 5G of acceleration it could impart 44.7 m/s of orbital change velocity; with 10G, you’d get 63.2 m/s — enough to save at least one ton of fuel per change of orbit.)

Anyway, this easing of space constraints would be a big relief to satellite builders, who grumble about the “tyranny of the fairing”. (One area where the Falcon 9 does lag behind its competitors is in fairing size.) This roomy cargo bay would particularly enable the deployment of big telescopes.

And of course with refueling, a Starship could also take a substantial load to the moon, which is a destination that has been getting a lot of renewed interest lately, and is definitely on Elon’s list. He says they are officially budgeting all post-Falcon R&D to building the Starship and its booster. The idea is that using it for their regular business of orbital launches will allow the entire Mars scheme to pay for itself. Their bet is that using an absurdly oversized rocket to launch ordinary satellites will make economic sense because the 100% reusability will make each flight so cheap... but that might not work unless the market for heavy cargo expands considerably as prices drop. If they can’t do large volume business, these savings may remain elusive... but mass-producing ships out of sheet steel does look like a good start on keeping them inexpensive. But if they pull it off, then maybe once the competition also has high reusability, at that point a smaller launcher would save money again. A Mini-FR or Starboat that scales this design back down to, say, a thirty ton capacity might someday end up being the even cheaper way to carry satellites and astronauts. And going still smaller, like say a five ton capacity, would probably have economic value as well.

(Some builders have started to target this idea: now that the plans for the Starship are well known, there are some outfits — Relativity Space for one, and also a Chinese government program — which are seeking to do a similar sort of build on a smaller scale, so that a midsized sat launcher can have full reusability of all stages. I don’t think any of them are going for in-orbit refueling, though.)

Because of these priorities, Musk’s plan is now to devote most of the development effort to the ship rather than the booster, since it’s a much more complex task and requires far more innovation. They were sending prototype ships on suborbital hops before the booster had even started being built. He says the heat shield is particularly challenging, due to the high velocities it’ll be asked to handle. But I bet that’ll be nothing compared to the million details it’ll take to go from a usable unmanned Starship to one that can carry passengers safely for months. I would not be surprised to see more than five years go by between the first uncrewed commercial launch, and the first crewed one. (Their proposed schedule is to cover this gap in just two years.)

Whereas most super-heavy rockets are exotic expensive beasts that launch only a few times, Musk wants the Starship to be mass-produced. That’s the only way to have a fleet big enough to colonize Mars, after all, and I suppose he hopes that if the flights are cheap enough, it will catalyze a boom in demand for other uses. Even as they built the earliest prototypes (and quite often destroy them in tests, because they have fully embraced the principle of “fail fast”), they were looking at every step for ways to make the process scalable to high volume assembly line production. Musk is talking about increasing the human race’s orbital lift capacity by about two orders of magnitude within a decade or so, if they really do start making these in large numbers. That could be enough to build whole cities in orbit, and large bases on the Moon. Longer term, a full-sized Mars fleet might make that three orders of magnitude. And he’s considering making future Starships of a size far larger yet.

One factor which Musk has basically ignored — and which NASA seems to be turning a blind eye toward as well — is the possibility that there is already life on Mars. The Viking tests back in the seventies are difficult to explain by other means, and strange color blooms have been seen in certain patches of soil in moist areas near the equator... if such life exists, colonizing on top of it could be a bad look, and possibly even mortally dangerous. The fact that NASA has sent multiple probes after Viking but didn’t follow up on biological experiments until recently is a strange omission.

The silliest proposal for the system is to use Starships for rapid transportation here on Earth: build spaceports around major cities (most likely out in the ocean), and fly passengers to the far side of the world in 45 minutes. Can you imagine the regulatory nightmare that would be, and what a tempting target it would be to use one for a terror attack? And as someone once said of this sort of ballistic passenger transportation scheme when it was first dreamed of fifty years ago, “For half of the flight you can’t get to the toilet, and for the other half it’s out of order.” Well, it’s not as unrealistic as Musk’s “Hyperloop” plan.

Since Musk is always thinking about the project after next as well as the next project... what would come after Starship? What he’s talked about as a successor is the same thing only bigger — a fat Starship. They’re considering increasing the diameter, perhaps as big as eighteen meters. Even if there is little increase in height (and I don’t think raising the height would be easy), that would give us a rocket weighing around twenty thousand tons, with a payload capacity of several thousand cubic meters.

getting it operational

They finally tried the first full launch in 2023, and it went rather poorly, though the rocket did manage to get a good distance away before blowing up, which was supposed to mean that the ground facilities would be undamaged. But it didn’t mean that at all — the area took substantial damage, due to a dumb decision by Elon that apparently nobody could say no to. (Rumor has it that there are people at each of his companies whose job is to talk him down when he gets enthusiastic about doing something stupid, and of course they can’t always succeed at this. The job may be getting more difficult over time.)

See, normally the launch pad of a large rocket has heavy systems in place to protect the pad area from the engine blast. These usually involve either curved plates to turn the exhaust sideways, large volumes of water, or both. The Starship pad in Texas had none of these things. Why not? Because of a whim of Elon’s — he wanted to see if they could get away without any of this stuff, just to launch sooner. The pad was just flat reinforced concrete... and when the booster fired up, with enough power to shake the ground ten miles away, that concrete pad turned into a crater. The legs of the support structure, also concrete, came apart as well. This resulted in everything around the area being pelted with jagged debris, including tanks full of fuel.

Three of the booster’s engines got knocked out immediately, which caused it to have a sluggish liftoff... which compounded over the next few minutes as three more engines failed, sometimes in rather explosive fashion. As the rocket struggled to reach sufficient altitude, it tilted upward so that the supersonic wind was hitting it at an angle. At this point a camera showed that one of the upper stage’s fins was badly bent — whether from the wind or the concrete I don’t know.

Finally, as it neared the time for stage separation, the rocket could struggle no more, and ended up completely sideways, out of control. (Apparently by this point, the guidance avionics were on fire.) They had to push the boom button. But the prototype booster had one more failure left to give: after the flight termination system detonated its charges, it kept flying under power for another forty seconds! The tanks were punctured but the holes weren’t big enough — fuel sprayed out the side, but not fast enough to collapse the tanks, or to ignite. That eventually happened once the inside pressure was low enough for the booster to crumple under the assault of the supersonic crosswind. It certainly showed some impressive durability.

Normally they’d be ready to try another rocket within weeks, but with the pad wrecked, it took a lot longer to get ready. Elon tried to say they’d go again in two months, but industry watchers immediately called bullshit. It ended up being seven months. But they did build something pretty quickly: a big steel plate on the launch pad with a lot of holes for water to squirt upward. It’s known as the “booster bidet”, or more formally as the deluge system, and should help reduce noise as well as ground damage. (They eventually decided that for their second launch pad, they would put in an actual flame trench.) Fixing the rocket itself took longer, as did satisfying all the government agencies they’d pissed off by spraying debris all over an ecologically sensitive area.

In that period they announced another design change: an open vented interstage like an old Soviet rocket, so they could ignite the upper stage while still attached to the booster. Somehow this is supposed to increase payload capacity, though I don’t see how that adds up. Also, Elon started talking about pushing the chamber pressure to 350 bar and the thrust to 2640 kilonewtons per engine at sea level, which is more than a BE-4 and might allow the Starship to lift a 200 ton payload... and maybe taking it even further in later versions.

On the second test launch we saw why he was talking about thrust increases: with all engines working it was still rather slow to lift off. But the pad and tower survived mostly intact, and the booster flew exactly as planned right up through the second stage ignition, and we saw the booster turn around as the ship continued. But we also saw the booster venting a lot of gases, and over the next several seconds a bunch of engines started failing, often producing bright flashes, and a little bit later the entire booster disappeared in a huge detonation. It looks like the steel hat they added to the top was sufficient for surviving the “hot staging”, but the sudden reverse push on that hat, at a time when the booster had shut off most if its engines, may have caused fuel to slosh forward, letting gas into the fuel lines, which may have been compounded as the booster flipped over. That’s one theory.

But the upper stage looked fine, unaffected by the problems in the now-detached booster. Until a few minutes later, that is, when it too started producing big gas clouds as it vented excess propellant, though all six engines were still running. It was lost about half a minute before the expected engine shutdown, with the venting having led to an explosion. It reentered in pieces near Puerto Rico, though one might have thought it was moving fast enough to have gone a lot further. I guess the altitude of 150 km was low enough for the atmosphere to pull it back down pretty quickly.

Close examination of the launch footage showed some heat shield tiles had come loose in the first minute, so it would have had little chance of surviving reentry if it had finished its ascent burn in one piece. But it was still a lot more successful than the first test, and it did reach space.

For the third launch they added more slosh baffles in the booster tanks, and it got through hot staging without incident, though the booster did later fail at its landing burn. By some reports it may have blown up a few hundred meters above the water after some engines did not relight; if not, it hit the water at high speed, producing a similar outcome. But the ship got into the desired orbit, though a lot of gas venting continued for the whole time, releasing so much that it started building up on the hull as ice. They pumped fuel around between the main tanks and header tanks, and they opened a slot-shaped payload door, but it looked like it was a struggle to get it closed again. They planned to try relighting one raptor engine in space, but the automated system in charge of that decided to skip the burn, as they’d half expected. The problem was that the ship was rolling too much — they concluded later that valves for some of the thrusters had somehow gotten clogged, though I can’t imagine what with. A few missing tiles could be seen.

It reentered over the Indian Ocean near Australia, with the flaps flexing around, and they actually got video of the plasma glow for a while without losing the signal. It looked like the ship was still rolling, which meant different sides were exposed to the wind, which is not what you want if the goal is to get down intact. Contact was not regained after reentry; it presumably burnt up due to plasma hitting the un-tiled side.

’tis but a scratch

The fourth test was spectacular at the end. The booster lost one engine on the way up, and another on the way back down, but managed to successfully do a soft splashdown off the Texas coast. The ship kept its orientation during reentry, and we got to see the hot plasma most of the way down, as the Starlink connection did amazingly well at getting data through the interference of hot plasma. One camera stayed alive the whole way, though by the end it had big cracks in the lens and was spattered with slag. And in the harshest part of reentry... we could actually see molten metal parts flying off one of the flaps as its hinge and one edge were torched away. Yet somehow it still worked! It still pivoted as if it were intact, and still helped guide the ship down in the lower atmosphere. I can’t imagine how that was possible with the amount of destruction we’d seen it undergo. But the scarred ship apparently still managed to do a landing burn, flip upright, and approximate a soft splashdown of its own. I had hoped that something might be left afloat to be recovered, but apparently not.

So the heat shield, as expected, is the main area that needs work. That’s a very difficult area to test in the lab. (NASA has devices that combine a wind tunnel, indoor lightning, and big lasers, and they do a good job of blasting the crap out of any substandard material, but they still fall short of what real reentry will do to you.) The new plan is to add a backup layer of ablative heat shield material underneath the reusable tiles... which might well save the ship if one of the tiles breaks, but will also worsen the difficulties that the ship is already having with making weight. If they can’t get it light enough, the whole orbital refueling idea isn’t going to work.

Overall, mishaps are ongoing, and they’re a long way yet from making something safe to fly in, but each test continues to get a lot further than the one before.

Super Heavy and Starship: mass ~5000 t, diam 9 m, thrust 74400 kN, imp 3.55 km/s, full-flow staged combustion (methane), payload 100‒150 t (2.0‒3.0%) with reuse, cost maybe $2M/t at first (much less later), record 0/2/1. Yipe!

(...How does 5500 tons compare to the biggest historical rockets? Apollo was just under 3000 metric tons, N-1 (the failed Soviet moon rocket) was 2750 tons, the SLS is about 2500 tons in its initial version, the Energia was 2400 tons without payload, and the Space Shuttle was 2000 tons. As far as I am aware, those are the only rockets that have ever been bigger than a New Glenn or Falcon Heavy.)

(The original 2016 BFR design was expected to be 10,500 tons. If you think that was absurdly gigantic, well, it still isn’t as big as the legendary “Sea Dragon” which was designed in the early sixties but never built — a very basic but very large rocket designed to make heavy lifting as cheap as possible for the time. It would be built out of sheet steel in a shipyard, and use a simple pressure-fed motor on each of its two stages. That thing would have weighed eighteen thousand tons, and the two engine bells were well over twenty meters wide — large enough to fit comfortably over my house. I doubt the idea would have worked, with combustion instability tending to worsen as engine size goes up. And Boeing once proposed a far bigger rocket called the Large Multipurpose Launch Vehicle, with a core stage so huge that it could use ten strap-on solid boosters that were each the size of an Apollo. The heaviest configurations might weigh thirty thousand tons, and lift almost two thousand tons to orbit.)