Why I’m still excited about Hyperloop
Last year I spent almost all my time leading OpenLoop, an alliance of six universities competing in SpaceX’s Hyperloop competition. I stepped away from that role in June, but I still find myself pitching the promises of Hyperloop surprisingly often.
This post is a memory-dump of what I’ve learned in the last year, with a focus on the bits I think people get wrong. Corrections are welcome, I don’t claim any real expertise in most of these topics, it’s just what I think.
Note: I’m putting little bracket citations throughout this piece. They aren’t actually citations, they’re just interesting asides I don’t want to interrupt the main point for.
What is Hyperloop:
The term has expanded a bit, so it’s important to be clear. The key details of a Hyperloop system in my head canon are:
- A tube at less than 1/100 atm but more than 1/10000 atm (IE, it’s not really vacuum)
- Small ~60 foot long vehicles without onboard propulsion
- Linear induction motors in the track
- Some form of levitation. Most likely electrodynamic mag-lev in reality, but more on that later.
Important omissions from this definition include:
- A compressor on the front to reduce drag. I’ve heard a lot of credible people claim this isn’t really necessary, and it may be cost effective to instead maintain a lower tube pressure than the alpha paper. This is particularly likely to happen on the first iteration of Hyperloop because the speed/pressure regime involved is wildly different from anything encountered in aviation (transsonic speeds at pressure equivalent to roughly 150,000–210,000 ft above sea level) and developing radically new turbomachinery is not cheap.
- Human passengers. Hyperloop is very exciting for cargo, not just faster human transit. Again, it’s particularly likely the first Hyperloop won’t carry people because that’s much harder to get approved than cargo. Also note that my interest in cargo makes my canonical Hyperloop a lot bigger than the alpha paper’s spec, as I think it should be able to carry 8'x8'x40' intermodal containers.
Hyperloop is exciting because…
- it could be extraordinarily energy efficient
- It amortizes the biggest costs across the network
- it could move enormous amounts of stuff with little infrastructure
- it’s fast
Notice the order of those points. A lot of people oppose the technology because they see it described in the press as ‘faster trains’ and they rightly feel the premium people will pay for quicker transportation can’t possibly fund all this new infrastructure.
Hyperloop isn’t fast, it’s low drag.
The various Hyperloop proposals all involve an operating pressure of about 1/1000 atm.
The energy lost to drag is force * distance, which means the drag losses of a Hyperloop pod are roughly equivalent to those of a semi-truck going 1/10 the speed (about 60 mph).[F1]
In a Hyperloop pod, your dominant source of drag will be the electrodynamic suspension, which has a Lift/Drag coefficient of 100–200, IE 5–10x better than a Boeing 747 at cruise. This low drag force has two important effects.
First, low drag frees you from the need for constant propulsion. A car, a train, a boat, and a 747 all keep their propulsion running whenever they’re moving, because they have to constantly fight drag. Always-on propulsion pretty much means you have to carry your source of propulsion on the vehicle, because building linear induction motors into every inch of track would be prohibitively expensive.
With near-zero drag you can put a LIM boost station every several miles, coast in between them, and carry that much less dead weight around on every vehicle.[F2]
Second, low drag gives you energy efficiency. Linear induction motors can both accelerate and regeneratively brake pods, so ~90% of the energy put into propulsion can be recaptured if it isn’t consumed by drag.
Putting exact numbers to this efficiency win is out of my depth, but we can make an interesting comparison to semi-trucks:
Semi-trucks and Hyperloop pods experience about the same aero drag energy losses. The rolling resistance of semi tires (.004 to .008) converts to a L/D of 125–250, so they’re also roughly equivalent in overall drag. That means they only compete on braking losses. The magnitude of braking losses is hard to figure out, so see [F3] for the full estimate, but all up Hyperloop should be about 4x to 9x more efficient measured in J/kg-km.
The magic of amortized costs
This is about the point where critics should point out that Hyperloop’s low drag isn’t free; Energy hungry vacuum pumps are needed to provide the rarefied air the pods fly through inside the tube. Plus, the tube itself needs to be extremely well built or the leakage rate will make those pumps work harder and use even more energy!
The important thing to understand here is that you pay these costs once per route. If you pump the airlocks down to a small multiple of the tube’s pressure with a roughing pump before launching pods, the amount of pumping that needs to be done is very weakly proportional to the traffic in the tube, and almost entirely a function of the tube welds leaking. The welds leak the same with 1 pod on the track or 1000.
On the actual weld quality, I haven’t got much expertise but I’m convinced it’s not a major obstacle. Fossil fuel pipelines have provided incentives to get really good at welding perfectly sealed tubes over the years, so I would be very surprised if the tech wasn’t there. Plus, Hyperloop One built their big tube last year and we haven’t heard a peep from them about weld quality challenges since, so presumably that test article showed them it’s a nonissue.
Another big amortized cost in Hyperloop is the propulsion system. The LIM rotors attached to each pod are nothing more than inert blades of aluminum, so the per vehicle cost here is extremely low.
Lastly, the energy storage capacity required for propulsion is also amortized, which is a huge win for an electric transportation project. Battery costs and weight are the main obstacle in the electrification of transport — they’re Tesla’s single biggest differentiation, they’re the main obstacle for electric semis, and they’re what categorically rules out electric freight aircraft in the near future. Sharing this resource across all the pods in the network provides enormous savings. Removing battery mass from the vehicles also slims down propulsion power a ton. [F4]
Cargo
The most interesting part of a Hyperloop network is the central control system. [F5] Since the pods only accelerate or decelerate significantly at LIM stations, the whole network lends itself well to centralized control and precise coordination.
For cargo, this means you can launch pods down a tube at very high frequencies and get the ‘utilization’ of your infrastructure very high. Both Hyperloop One and HTT have stated they want to be able to launch pods at a cadence greater than 1 every 20 seconds. Assuming each pod carries a single 40 foot container with a max load of 36,000 kg, this means the max throughput of a single tube is 6500 tons/hour. Measured in shipping containers, this is a Panamax ship every 27 hours.
The precise footprint of a Hyperloop tube + pylons is hotly debated online (turns out lateral force and pylons don’t go together) and I’ll leave the specifics to our friends in civil engineering. However, at the end of the day the footprint for a two-tube Hyperloop route is probably in the same ballpark as that of a 2–3 lane highway.
On a more or less equivalent piece of real estate, you get a near silent zero-emissions piece of infrastructure that can move ~1 Panamax a day instead of a highway with a diesel-burning semitruck passing every 20 seconds for the same cargo.
It’s fast
I don’t like to talk up Hyperloop’s speed because it’s not the system’s biggest strength in my mind. However, it is worth getting excited about.
First let’s talk about speed for cargo: I like to buy things on Aliexpress. Everything’s dirt cheap and a lot of it’s surprisingly well made, but it takes a month to reach me in Upstate NY because they send it across the Pacific on a boat. In a world with a global Hyperloop network, you could get same-day shipping from Shenzhen.
Now, passengers. We’re pretty bad at building passenger transport infrastructure. The highway system was for moving tanks during WW2. Amtrak just piggybacks on the nation’s freight railways. We developed our first airliners by ripping off the B-29 bomber. So I truly think Hyperloop is going to get built for cargo first.
However, once the tracks are laid and the pods aren’t crashing, we’ll eventually approve it for passenger transport. That’s where things get pretty insane: Sacramento to SF in 9 minutes. Boston to NYC in 22.
The advent of new transportation systems is always followed by a change in where people live. With Hyperloop, the radius of acceptable commuting — both travel time and environmental impact wise — around our major hubs grows by a factor of 10. 60 mph → 600 mph. It’ll be like sticking an aggressive Gaussian blur filter on the map of real estate prices.
My sister always says she wants to live on a ranch, unfortunately she’s currently stuck at Berkeley because she really loves studying earthquakes, and that’s not really a job you get on a ranch. If we had a Pacific coast Hyperloop she could commute 200 miles up to Redding (20 mins), take another 20 from the station in an autonomous Tesla and be home at her ranch. That sounds comically wasteful today, but then so did 20 mile commutes before we had highways.
One other important factor in Hyperloop’s speed is its point-to-point nature. The per-vehicle cargo capacity of Hyperloop is far lower than trains and planes, which allows individual vehicles to take far more direct routes.
That whole Mars thing
This section goes off the deep end a little, so I encourage you to stop here if that doesn’t sound fun to you.
I think Hyperloop was designed (or if it wasn’t, should be used) to move ice on Mars, specifically from the poles to the equatorial sites we’re likely to settle.
Please understand I’m not talking about building this anytime soon; I think a realistic timeline to start is when we reach 10,000 people on Mars. Once that happens, it should to be pretty clear we’re on Mars to stay, and a necessary part of staying will eventually be a very large supply of water in the equatorial regions.
Now, why I think that water source will be a Hyperloop takes some explaining. There are a bunch of little arguments here that add up to be reasonably compelling:
First, air bearings don’t make sense on Earth. The long and the short of it is that they don’t float high enough, that forces the track to be really smooth, and that’s expensive. Mag-lev on an inductive surface is just easier.
However, two factors make air bearings a lot more sensible on Mars:
- The reduced gravity gives you 1.5 to 2x the levitation height for the same systems, because your vehicle has 1/3 its Earth weight and levitation height is a funky nonlinear function of load.
- Mag-lev is prohibitively expensive on Mars because metal refining is expensive. An air levitation surface can be made of anything smooth, for example polyethlene from the atmosphere mixed with Martian fines to get more volume out of it.
A whole bunch of common Hyperloop nitpicks also go away on Mars: No seismic activity to cause catastrophic tube failure (or even significant erosion to give you more local disasters), an atmospheric pressure about 6x the spec pressure for Hyperloop means you can probably leave the tube at ambient pressure (no more thermal expansion problems!), long span bridges are easier in 1/3g, and there aren’t any right of ways to give a shit about when you’re laying track across Tharsis (yet).
The second big point is that we really don’t want to melt ice at the poles and try to flow our water to the equator. The grade from the South pole to the equator (~6 km over 3000 miles = 0.12% on average) is enough for water conveyance, but that’s where the advantages end.
To compare apples to apples, let’s go back to our Hyperloop’s throughput of 6500 tons/hour. That’s 1800 liters of water a second, shipping in 8 ft x 8 ft x 20 ft blocks which conveniently weigh about 36000 kg a piece.
A 24" water main carries 18,000 gal/min or about 1100 liter/second. So to get 2/3 a Hyperloop worth of water transport, we’d need to insulate and heat a 3000 mile long 24" dia. pipe to above 0 C on Mars. You also probably want to pressurize the pipe to a decent fraction of Earth atmosphere, because Martian ambient pressure is right below the triple point of water, so if left unpressurized your pipe would be an insane balancing act between a steam pipe and a frozen pipe. To make matters worse, energy at the pole is only available half the year [F6] because the sun sets for several months during Southern Winter / Northern Summer.
Now consider a Hyperloop sliding blocks of ice downhill: 0.12% grade isn’t enough to move the ice for free. Assuming our L/D is 200, we need a grade of 0.5% to get a free lunch, but 0.12% gives us 24% off of our drag losses.
Martian Hyperloops will still need an (unpressurized) tube to keep dust/fines out of their turbomachinery, and for ice transport the tube serves double duty as a sun shade to keep the ice solid until it reaches its destination.
Lastly, an ice Hyperloop simultaneously solves the resource colocation problem for Mars settlement. Choosing a home on Mars involves finding a single region with as many of the necessary raw materials as possible: Water ice, metallic ores, nitrates for crops, and more obscure minerals necessary for industry are very difficult to find in a single site. When you add the requirements for landing (huge flat areas) it’s extremely likely our home will be missing some things. It’s also extremely likely those things will be available somewhere along a 3000 mile Hyperloop route.
Side notes:
F1: A couple people were confused by this… Look at the equation for drag force: The frontal area of a semi truck and Hyperloop is about the same (both carry 8x8 shipping containers). The drag coefficient is roughly the same assuming your Hyperloop has a compressor, and if it doesn’t have a compressor it will further compensate with lower tube pressure. So the difference in drag force comes down to velocity² and fluid density, which is equivalent to air pressure * a constant factor since both systems are in regular air. Hyperloop’s ambient air pressure is roughly 1/1000 the truck’s, so the Hyperloop can go log_2(1000) = 9.965 times faster and experience the same drag.
F2: In practice the distance between LIM motor stations in a Hyperloop track will be dictated by the curvature of the route more than the system’s drag. The pods need to decelerate to make sharp turns, so any section where the radius of curvature is too low needs a braking station ahead of it and a boost station after it. Right of way considerations (on Earth), mountains, and the simple limits of tunnels and bridges will make these boost/brake stations effectively the only ones, unless you’re trying to cross Kansas or an ocean.
F3: What’s missing from this calculation is the relative magnitude of braking losses vs aerodynamic losses. Aerodynamic losses are easy enough. Semi trucks have a Cd of about 0.96 and a frontal area of roughly 9 m². So Fd = 2.7 kN or so. At 50 mph that’s 60 kW of drag power, or about 80 hp. Rolling resistance is about 39 kW or 53 hp. Brake losses are much more handwavy, but as a very conservative estimate we can just look at brake losses due to elevation change. The US highway system has a maximum grade of 6%, so a reasonable guess for the average grade might be 2%. At 50 mph on a 2% grade, the elevation change rate is 1 mph. Half of that change is uphill when the truck is paying energy out from the engine, and the other half is downhill when the truck is burning energy in its brakes. So overall our brake losses due to elevation change are about 79 kW or 106 hp. Finally, if we neglect braking due to traffic, stops, and turns we get that brake energy losses are roughly 44% of the total losses in a semi. Since the Hyperloop wins by 10x on brake losses, a conservative estimate for its energy per kg-km advantage over trucking is 4x. 9x was chosen as an upper bound because 1/9 is about the smallest fraction of braking losses I think it’s reasonable to credit toward elevation change. Note this 4–9x figure also ignores the relative quantities of accelerating and decelerating Hyperloop and semis perform. While Hyperloop may be associated with high acceleration in your head, the metric which relates to the quantity of braking work done is actually average acceleration per distance traveled not average acceleration over the time duration of the trip, and I think that distinction means Hyperloop wins here too.
F4: Another incredibly exciting aspect of the Hyperloop energy problem is the potential for grid balancing. Every LIM station on a Hyperloop route needs an enormous capacitor bank to smooth the short bursts of consumption and generation involved and interface with the power grid. These storage facilities will rarely be near maximum or minimum capacity because you want to keep the network flexible, and as a result they’ll be free to smooth out grid power fluctuations on the timescale of hours. Another big win for Hyperloop’s place in a renewable future.
A sillier version of this idea would be to consider the kinetic energy of the pods as an energy store. 90% round trip efficiency is about what the PowerWall gets, so if you don’t mind speeding up a little here and slowing down a little there it’s not a completely silly thought. It would be immensely tricky to integrate with Hyperloop traffic control though, so I don’t think it’ll happen unless we get Culture style AI Minds to figure that out for us.
F5: OpenLoop studied this problem some. It’s extremely interesting because the cargo in a pod determines how fast it can take a given turn. Heavy or delicate loads need to slow down more to limit centripetal acceleration, so the optimization problem to minimize trip time and the amount of acceleration and braking required (which is proportional to waste energy) is insanely complex.
F6: Solar insolation on Mars is interesting. The peak daily solar power is actually at the poles in Summer. The atmosphere is crazy thin, so you don’t lose nearly as much light to scattering at high latitudes as you do on Earth (ignoring the occasional dust storms). That means that in Summer when the sun’s up all day at the pole, you pretty much get equatorial solar power but for twice as long. Of course, in Winter you get nothing. This paper digs a lot deeper into the general feasibility of Martian solar.
Edit September 2017: I originally wrote and published this article a little less than a year ago, when I was just some student who really liked Hyperloop. Between then and now I spent a few months working on making the tech a reality at Hyperloop One. This piece contains absolutely no proprietary info from Hyperloop One — it was written before I had any of that, and this is the only edit I’ve made since.
