Sunday, September 16, 2018

Unpopular opinions in space

One of the fun things about speculative technology is encountering people with different views, and then trying to understand why and how, essentially, the same set of axioms leads to a different conclusion. In some ways, the community can be divided up into camps, representing the acolytes of various well-known thinkers in the area. As an example, concerning where humans should live in space, some people suggest the Moon, others Mars, others asteroids, Venus, or giant stations in deep space. For a second example, opinions differ about where the money may come from. I previously dealt with the funding question in another blog: 

In this post, I'm going to air three unconventional if not unpopular opinions, then spend a few thousand words explaining my views on them. I will endeavor to use accessible math and be quite clear about which axioms can be altered. My intention is to continue the conversation and to have, in one place, a concise summary of a point of view which can be referred back to as necessary.

Rule 1: There is no problem in space that can't be most effectively solved by building a bigger rocket on Earth.

Rule 2: There is no commodity resource in space that could be sold profitably on Earth.

Rule 3: Self-replicating robots and matter compilers do not exist.

The implications of these rules are not as dire as they may seem. In particular, my recent book on Mars industrialization ( shares these axioms and retains an optimistic tone.

Why are these rules true? Read on!

Rule 1: There is no problem in space that can't be most effectively solved by building a bigger rocket on Earth.

In short, big rockets are expensive, but managing interfaces in a vacuum is much, much more expensive. Case in point: The ISS.

Unfortunately there are very few large rocket development programs to use as baselines, but there are a few. The Saturn V, which could launch 110T to LEO, cost about $1.2b to develop in 2016 dollars. The Soviet Energia was considerably cheaper. The SLS has already consumed tens of billions, but it is well understood in the industry that it is not exactly a lean program. SpaceX is developing the BFR using only internal funds, and will probably spend a similar amount to the Saturn V, though with a lower per-flight cost and much higher overall performance. 

In contrast, the ISS, which tested the idea of assembling a space station from modular parts launched using a partially-reusable shuttle, has cost $150b, and has taken the better part of 30 years to build, including on-Earth fabrication. And for that cost, a disproportionately high fraction of the overall station mass is consumed by the interfaces: heavy airlocks, narrow connecting passages, and architectural constraints. Further, the station will need to be retired in the next decade or so, as the "sausage link" interfaces are subject to bending fatigue that is gradually weakening them.

In a recent Quora question, I pointed out that a single expendable launch of the BFS would deliver much more volume to LEO than the entire space station ( That is to say, a Skylab-style station based on BFR could be launched in one go, at a cost equivalent to the marginal cost of a single spaceship at retirement, which is close to nothing. 

Why is this true? The reason is that all large pressurized volumes require assembly from various sub-components. For Boeing jets, this is done in Renton, near Seattle. Even there, in a climate controlled factory, it is a serious headache. It requires a small army of engineers and technicians. But it is still cheaper to do it there, in the factory, than in flight or on the side of a small runway somewhere in the middle of Wyoming. 

In short, as much construction and integration as possible should occur on Earth, where labor costs are about a million times lower than in space. Payloads should be launched in the largest possible units, in plug-and-play configuration. And that is why, even though building an enormous rocket is extremely expensive, it is the cheapest way to do business in space.

Are there limits to this? Yes, of course. The largest rockets ever flown delivered about 100T to orbit. It is not clear to me that rockets could efficiently deliver more than about 1000T to orbit in a single shot, with chemical rockets. This is due to fundamental limitations on the strength of materials in pressurized combustion chambers, fuel and material density, and Earth's gravitational field. But 1000T to LEO is a very, very large chunk of stuff compared to the current model for doing business.

Note: While I personally think that there are all sorts of good reasons to pursue reusable rockets, and that larger rockets in the correct configuration are easier to make reusable due to improved margins, this rule doesn't imply that reusable rockets are necessary. In particular, the space station would have been much cheaper and faster to build if it were launched on a Shuttle-derived expendable heavy lift stack. 

Rule 2: There is no commodity resource in space that could be sold profitably on Earth.

One possible exception: The elixir of life, if it could only be obtained on the Moon. 

The usual examples range from water or Helium-3 mined on the Moon, to platinum-rich asteroids, to space-based solar power.

This rule comes down to a discussion of intrinsic value. Clearly, to be worthwhile, a space-based resource has to command an exceptionally high value-per-mass. There are a handful of commodities on Earth with values as high as $100,000/g, such as listed here: 

It is important to note that none of these are intrinsically valuable. The illegal drugs are expensive because there's a high cost to being caught making them. The diamonds are expensive because their market is manipulated. And rare metals are expensive because they're very hard to chemically extract from rocks, but also because they're basically never used in industry. That is to say, there is no demand for them. 

In particular, of all rare, valuable commodities there isn't a single one with a high level of baseline usage. This means that if the supply suddenly increased, because of an additional discovery, the price would collapse. Even if there was an asteroid of pure platinum orbiting the Moon, and there most certainly is not, increasing the global supply beyond baseline of about 40T/year would simply reduce the market price. 

As for water on the Moon, water isn't even intrinsically valuable or rare on Earth. In fact, as I showed in a previous blog (, it is probably cheaper to import water from Earth to the Moon than to extract it there. And even if it were cheaper to obtain water on the Moon, there is zero case to export water from the Moon to anywhere else. 

Finally, let's consider Helium-3. Helium-3 is a nice example, because it is relatively much more abundant on the Moon, and it is currently very expensive on Earth due to rarity. It is even used in some industrial and scientific processes as a refrigerant. But in order to make a business mining it on the Moon, adequate demand to both keep costs and revenues high must exist. For this, we are told, Helium-3 is a natural fuel for nuclear fusion. There may come a time when lunar Helium-3 fuels fusion-powered interstellar voyages, but I am unable to not put that in the science fiction bucket. 

So what would a space resource have to look like, quantitatively, to make a business selling it on Earth? My interest here is to be inclusive, so I will underestimate fixed costs as much as possible on the first pass. Let's say that although currently it costs about $3000/kg to launch something to LEO on a reusable F9 flight, SpaceX's BFR reduces that further to $100/kg. Let's suppose that further the cost of delivering cargo to the Moon using hardware based on the BFR is $1000/kg and the cost of returning cargo from the Moon $10,000/kg, which assumes at least local oxygen propellant production. By comparison, the cost of shipping a container half way around the world is about $0.10/kg. 

The question, then, is what commodity is relatively much more available on the Moon than the Earth to make up for the fact that shipping it is 100,000 times as expensive. I am not aware of any physical matter, short of the elixir of life, that would make this worthwhile. Yes, a tiny number of high net worth individuals may want to travel there for tourism, but that doesn't approach a billion dollar industry.

But what about space-based solar power, popularized by Gerard O'Neill in The High Frontier? While shipping matter to and from space is enormously expensive, it is much cheaper to beam microwaves as they have no intrinsic mass. 

Gerard O'Neill's book was written in the early 1970s, when it seemed as though the world was headed for a Malthusian crisis of population and energy consumption. This is not the case anymore. Indeed, the fundamental challenge with space solar power is that although the solar resource in space is about 3 times as good as the best places on Earth, the transmission losses from space are comparable in magnitude. Economically, it is much cheaper to deploy solar photovoltaic panels on Earth than in space, where at best the delivery costs are 1000 times higher, the maintenance costs a million times higher, and the environment much more difficult to deal with. 

As Elon Musk has concisely pointed out, the fundamental problem with space solar power is that it's obtaining a commodity, power, somewhere where it's expensive and selling it somewhere where it's cheap. This is not a good business. Indeed, it would make more sense to beam power from Earth to space stations, if they needed it. And, more generally, the same goes for supply chains for any other product.

That's not to say that microwaves have intrinsically low value. The trick is to use them for something other than carrying power, namely, information. And indeed, the majority of the space industry, and almost all of the non-military space industry, is dominated by microwave communications. The dispatching of information, through space, from specialized satellites made in factories on Earth. And SpaceX has a play in this market too, with their StarLink internet constellation. The right kinds of information, at the right place and time, are very valuable indeed.

Note: This rule doesn't necessarily apply to Earth-based manufacturing with a step performed in zero-G LEO. There are a number of companies pursuing niche products that exploit zero-G processes to make stuff, and at a potential level of revenue adequate to cover the cost of launch and recovery. These products however, do not necessarily lead to a generic space industrial capacity, or generalize in any particular way. There is no good reason, for instance, to extract material precursors from a passing asteroid instead of launching them from Earth.

Rule 3: Self-replicating robots and matter compilers do not exist.

In answer to the previous two rules, some proponents of space settlement argue that it's not necessary to launch giant payloads from Earth. All that needs to occur is the launch of a small, robotic egg to a convenient asteroid. Once there, it will process the raw materials to produce whatever it needs, build thrusters, antennas, copies of itself, habitats, food, televisions, whatever. This asteroid will then be a glorified robot and can be steered back to the Earth in preprocessed, highly valuable form to be used as a mine or space station or interstellar spaceship or whatever.

This idea sounds great, and variations of it have been kicking around since at least the time of the ancient Greeks. The fundamental problem is that such a compact universal factory, or egg, simply does not exist.

That's not to say it could not exist. Indeed, E. coli is a very capable self-replication machine, given the appropriate environment. However, no known life form prospers in vacuum, and that's not for a lack of searching on the part of astrobiologists. So any matter-compiling asteroid-munching probe would have be mechanical in nature, not biological. And there are no robotic self-replicating factories in existence, not even close. 

Indeed, the reason that there are so few countries capable of heavy industry, and all of them are very large, wealthy countries, is that industry is big. Why must industry be so big? The self-replicating machine that is modern industry requires about a million different kinds of specialists. Specialists train for years to be sufficiently efficient at their given tasks, without which the final product, such as an iPhone or fighter jet, may well take infinite time to complete.

There are numerous theoretical approaches to matter compilers, or rapid, atomic-level 3D printers, but I am not aware of any that pose a credible threat to the current industrial status quo. It would be cool, but as far as I'm concerned, we're more likely to have a vibrant lunar Helium-3 mining industry in 50 years than access to universal matter compilers. 

What does this mean? There is no way to do advanced industry in space without thousands to millions of humans. There are no miracle shortcuts. We just have to find a way to support thousands to millions of humans in space, probably on a planet with a diverse array of natural resources.

Tuesday, September 4, 2018

How to Industrialize Mars

How to Industrialize Mars
Casey Handmer

Remarks from talk given at the Mars Society Convention, August 23 2018, now in blog form.

I always prefer to start talks with a recognizable image. In this case, this is a Mars Global Surveyor image of Gale Crater, looking south. The crater is about 150km wide, and the Curiosity Rover is currently driving around near the right hand edge of the central mound.

Although I’ve probably seen this photo a thousand times, I only just noticed the Pancake Delta feature, just right of center at the top of the image.

I’ll start with just a bit about me. I have many interests, though I’m formally trained in physics. I earned my PhD at Caltech in 2015, studying gravity in an effort to understand warp drive. Since it turned out to be impossible, I switched to working on the Hyperloop, where I was the levitation engineer, and more recently I’ve moved to JPL, where I work on GPS instruments. These pictures show me exploring the remnants of Siberian industry, a Tesla coil I helped build for Burning Man, and pointing at the Great Unconformity, a billion year gap in the geological record visible at the bottom of the Grand Canyon.

This is also the place for the great disclaimer, beginning with the obvious stuff. It is a great honor to be here speaking to you all, but I must stress here I represent my own views and research, not that of my employer!
Second, I don’t have a crystal ball. I don’t know for sure how to build a city on Mars. But this applies to most of us. We are all mostly not doing Mars settlement full time. Instead, we come together at events like these to share our ideas and enthusiasm and then spread it like the conference flu.
Finally, this talk/blog is derived from a book on the subject available gratis at:, also available on Amazon at for the princely sum of 99c. This work is a hobby of mine, it’s not my main gig, and my principal goal is to develop and workshop these ideas and disseminate them as widely as I can.
The goal here is to think systematically about visions of the future. It’s to inspire a more vibrant, rigorous level of discourse. It’s to help us find ways to find the right questions, questions that help us find useful answers.

Let’s begin by defining autarky. It means economic independence or self-sufficiency.

Google trends shows that this word became very important in the 1930s when impending total war in Europe demanded industrialization for survival. Even though the word had been around since the early 17th century, it was the second world war, or the war of machines, that drove home the importance of industrial self-sufficiency.

The Martian, with Matt Damon, is one of my all time favorite films. When we think of self-sufficiency on Mars, we think of growing potatoes. But what would have happened to Watney without a ride home? The book is quite explicit about this. Even if he had plenty of supplies? Even with a thousand qualified friends and greenhouses more like the incredible farms in Holland, which have the highest productivity on a per-area basis on Earth? Death is inevitable.
In the book of The Martian, Andy Weir writes that Watney depends on life support machines, including an oxygen generator, water reclaimer, and pressure vessel, to survive. These machines were made on Earth and beyond a handful of very basic repairs, cannot be built or maintained in space, and will not last forever.
Unlike the popular but flawed traditional colonial picture, humans cannot survive on Mars with 40 acres and a donkey. Like European geopolitics in the 1930s, survival in such an inherently hostile environment is simply not possible by analogy with the US history of rugged agricultural pioneers on the frontier. In fact food isn’t even the first thing to be produced locally, being rather hard to grow without lots of related infrastructure.
It’s worth stating that while self-sufficiency for a Mars city is a worthy goal, in all but the most catastrophic scenarios, Mars and Earth would continue to trade essential parts. In the early days of the Mars city, actuaries will be able to calculate the consequences of supply interruption, just as they do for remote outposts and bases on Earth. Although manifests will be designed to minimize the disruption caused by a low level of supply interruption, total isolation will inevitably result in death after a few months or years. As the city develops greater industrial capacity, this grace period will gradually extend until the point where isolated survival is possible indefinitely, even though it would be far from optimal! The capacity for indefinite isolated survival is autarky.
What is the alternative to agricultural analogies? How can we think systematically about industry in space?
I would hope that my readers become very familiar with Cody Reeder! He has one of the most incredible YouTube channels: Cody’sLab. On this channel, he demonstrates the basis of a lot of primary production, including mining, farming, prospecting, and chemical purification. But even a thousand Codys would not survive very long on Mars.

Unlike Mark Watney, Cody has chemical and technical expertise adequate to build the necessary equipment from scratch. But while a thousand Codys could probably make *anything,* they could not make *everything* faster than the rate at which it breaks down in normal use. Indeed, if you watch the videos about his ranch, just keeping all the tractors and trucks running, in Utah with breathable air and access to McMaster-Carr, is sometimes pretty tough.

It is simply not enough to grow some food. Living on Mars is more like living indefinitely in a submarine. It will require automated manufacturing and, among other things, lots of metal. So let’s think big! Really big!
This is Australia’s main steel works, BlueScope Steel, which employs 16,000 people and produces about three million tonnes a year. It takes 30 minutes to walk across, but probably wouldn’t be big enough for a self-sustaining city on Mars.
This is the Tesla Gigafactory, which is pioneering the next generation of industrial automation. The Gigafactory makes cars and batteries, but Mars will eventually need at least basic chip fabs, an advanced composites supply chain, and active industrial research.
Again, the aim here isn’t to be completely descriptive. My goal is that this helps you develop a new thought, an incisive question, or a new strategy.
This is a good point to deal with a common diversionary tactic. Wouldn’t it be nice if self-replicating machines existed? 3D printers are very exciting precisely because they offer a technological shortcut for certain kinds of manufacturing. But they are not self-replicating machines, not by a long shot. They require very carefully curated input material and can produce only a limited range of parts.
Are self-replicating machines possible? Yes! Given the right resources, biological organisms can reproduce themselves, including my favourite here, Mr Platypus, and everyone else’s favorite, e coli.

But the platypus makes eggs containing baby platypuses, and e coli produces e coli. Convincing e coli to print a CPU or the platypus to lay an air filter would be something else!
So when we think of a self-replicating machine to solve all the industrial problems, we’re really describing a self-replicating factory or process that actually can produce anything. Which is nothing less than our modern globalized industrial society, in total. At least until someone builds a matter compiler, a regular staple of science fiction and asteroid mining concepts.
So we return to the original question, how to compactify the entire industrial stack and ship it to Mars?
What sort of scale are we talking here anyway? Is this a big problem? Oh yes!
China, 1410m
Mexico, 129m
Turkey, 81m
India, 1339m
Japan, 128m
Thailand, 69m
USA, 325m
Ethiopia, 105m
UK, 66m
Indonesia, 264m
Philippines, 105m
France, 65m
Brazil, 209m
Egypt, 98m
Italy, 59m
Pakistan, 197m
Vietnam, 96m
Tanzania, 57m
Nigeria, 191m
Germany, 82m
South Africa, 57m
Bangladesh, 165m
DR Congo, 81m
Myanmar, 53m
Russia*, 144m
Iran, 81m
South Korea, 51m

Here’s a list of the 27 most populous countries. The bolded ones contain essentially a complete industrial stack, by which I mean the ability to produce, within its own borders, all or nearly all technology necessary to produce the most advanced machines, including container ships, fighter jets, rockets, computers, mobile phones, and nuclear weapons. Russia doesn’t anymore, but it did until quite recently. South Korea does, sandwiched between Japan and China. Germany does, as a hub of sorts for the rest of Europe.
These data strongly suggest that a scale in the hundreds of millions is necessary to have enough labor specialization to support a complete industrial stack, and that’s on a planet on which we have evolved to survive essentially naked. Launching one hundred million people to Mars would be a major headache.
Let’s look at some counterexamples. Economically isolated countries like Albania, Cuba, North Korea, and Iran have every reason to attempt industrial autarky, and in many cases have tried really really hard. Yet even Cuba, with 11 million people, a very benign climate, and ample natural resources, has not succeeded. Australia, with 22 million, is not even close.
I like to think that with further advances, it might be possible to achieve autarky on Mars with *only* a million people, after 50-100 years of transport and building. But this would not be easy.
For comparison, imagine taking Iceland in 2018, a country with 350,000 people. Without imports, Iceland would regress to 18th century standards of living within a few years.
The OOCL Hong Kong, the world’s largest containership, can carry about 22,000 containers, which is roughly equivalent to 2000 flights of SpaceX’s BFR. I estimate is at least several decade’s worth of flights. So, given only unlimited money and one containership of gear, one has to reproduce the industrial versatility if not the might of Japan in Iceland by 2050, without substantial population growth. It’s almost unimaginable. But not quite!

Let’s talk about how to increase per capita human productivity in Iceland, or on Mars, by a factor of a hundred or so. The trick is the mechanization of labor, which is related to why whales have big mouths.
Consider a pre-industrial agricultural society, such as the fields pictured above. All available energy in the form of work is derived from solar power, from photosynthesis, and all available physical labor is from human muscle. Therefore, the total output of the system is limited, fundamentally, by how much energy all the humans can consume and digest.
Yet the gap in GDP between industrialized and pre-industrialized societies is a factor of 30-60. By freeing themselves from the requirement of work of sweat off the brow, a single human can control a gigantic, usually gasoline-powered, machine, or even remotely program one, to perform labor on its behalf.
We’ve already seen that while on Earth, a sufficiently motivated, knowledgeable individual can survive in many places with no resources, an industrial Mars will require the production efficiency of a hundred million humans. The fundamental problem for industrial human societies on Mars is a terrible shortage of labor, so the solution is to automate, mechanize, and outsource non-local tasks. By how much?
Consider the manual-mechanized continuum graph. For any level of technology, there is an optimal blend of human and machine labor. Compare, for example, the manual construction of the pyramids and the rapid modern construction of a house by a skilled contractor.
As technology improves, the optimal point moves to the right, but it is *never* most efficient to completely automate something, even the construction of a moon or Mars base. Complete automation is a subset of the self-replicating machine problem. A complimentary mixture of humans and machines is the way to go.
Take for example these two drilling systems. One is eight orders of magnitude more expensive, but doesn’t need a human to “line it up”. And sometimes it breaks down for a year at a time. Perhaps NASA paid too much for the rover, and it’s only 7 orders of magnitude more expensive than a human-controlled tool?
from future import industry
How can we achieve a Japanese level of industrial versatility and power with Iceland-style population? We must grow the fraction of labor that’s automated as the base scales. For instance, if the base doubles in size every launch window, the productivity of the supporting industries must also double, without doubling their labor requirement. Instead, individual labor productivity must also double at close to the same rate.
What does increasing individual productivity by three orders of magnitude look like? For inspiration, let’s consider the development of programming languages. The answer is the sequential interposition of layers of automated abstraction between the human and the physical stuff. This approach is only cost effective in situations of profound labor shortage, such as keeping up with exploding capability and complexity of modern computers.
Computer languages have evolved some hierarchy like machine code, Assembly, C, C++, Python. Each step encapsulates another layer of abstraction between the human writer and the fundamental logical operations, allowing much more powerful things to be done with a given labor pool. But let’s not stretch the analogy too much.

The final question I want to cover is order of industrial roll out. What resources get made first? As a case study, consider this incredible open source robot arm, the BCN3D-Moveo ( Mars will need a lot of robot arms, so this isn’t a bad place to start.

Here’s (the interesting half of) the BOM, or bill of materials. This is a shopping list from which the arm can be made. Broadly speaking, components fall into five groups: Structure, fasteners, bearings, motor, and power.

  • Structure, which are the bones, base, and carry the weight of the arm.
  • Fasteners, such as bolts, rivets, screws, nuts, clips, and so on.
  • Bearings, which enable two adjacent hard parts to swivel past each other with low friction and wear.
  • Motors, which provide the forces. Motors are deceptively cheap due to mass manufacture, but often require weird magnets and low sulfur high conductivity copper and other things that are hard to come by on any randomly selected part of Mars.
  • Power, which includes cables that move electricity around but also printed circuit boards, control logic, microcontroller chips, and other components which are relatively cheap on Earth, light, and very difficult to make from scratch.

I tabulated the cost as a proxy for manufacturing difficulty and the mass as a proxy for transportation difficulty from Earth. While a workshop on Mars could make any of these parts, mass local manufacture will proceed in order of mass divided by difficulty of manufacture, which is the order tabulated above.

This is what you’ve been waiting for, the roadmap for industrialization. There’s a lot going on in this figure, so let’s unpack it!
On the left, I’ve ranked successive orders of magnitude of industrial “closure”, or local production capacity. Starting with oxygen, then water and fuel, plastics and some food, then masonry, structural metals, then alloys, electronics, advanced chemistry and computer processors.
On the bottom, we have population. Today, we are in the bottom left, with only robots. With local production of oxygen and fuel, humans can explore and even operate outposts like the Antarctic stations. But at some point vast quantities of cargo and humans will have to be shipped to traverse this dangerous area of potential collapse. This area is dangerous because the population is too large to be evacuated and too small to be self-sufficient. The city traverses the graph toward the top right, such as the trajectory marked in red. Ultimately, the city has a large population and a diverse, self-sufficient industrial base.
The major primary industries deal with mining of any desired element. Because each mine will have to operate in the hostile Mars environment, emplacement of primary industry incurs a much steeper labor penalty than increasing complexification of secondary manufacturing, which can be conducted entirely inside large, pressurized, climate controlled habs.
For this reason, from the exploration phase until the cusp, marked with a purple dot, each order of magnitude of mass self-sufficiency requires more than that of people. Beyond this point, marking the completion of a local basic material supply chain, relatively small additions of population have an outsized effect on industrial closure. I estimated the critical stage on this graph is from about 1000 people to 100,000.
The fundamental limit here is Earth-Mars cargo capacity, as illustrated by the grey lines. Cargo capacity is determined by how fast we can build gigantic rockets here on Earth. Today, SpaceX can build about 20 cores a year, but Boeing can build 560 737s a year, a machine of comparable complexity. So I think this is a tractable problem, within the capabilities of our current civilization.
In summary, autarky is possible, but requires a really bold vision for scale, lots of giant rockets, lots of people, lots of ongoing, though non-infinite, investment of money and effort on Earth.
What questions do you have?