How To Industrialize Mars: A Strategy For Self-Sufficiency

Casey Handmer

This document is freely available online:

https://docs.google.com/document/d/1pxQg51rGP6JtdD4Eix1xpVek1xX05eQVgc7Jbm6VtDw/edit?usp=sharing

Please feel free to comment and suggest edits using the google docs platform, which automatically tracks changes.

This book is also available electronically on Amazon:

https://www.amazon.com/How-Industrialize-Mars-Strategy-Self-Sufficiency-ebook/dp/B07GN3BJX3/.

I deeply appreciate reviews on Amazon as well as comments here or personal feedback.

The next edition of this book will contain more illustrations^[1]. If you would like to provide an illustration, get in touch!

Praise for “How To Industrialize Mars”:

To Christine

Development Prioritization

Estimating Development Rate

Case Study: Iceland

Key Resource Development

Key Technology Development

Reusable 26 Month Development Plan

Introduction

This book is unabashedly about a dream wrapped up in a fantasy^[2].

For most readers, no further justification is necessary. If you are interested in the “why” question, I can think of no better introduction than the Wait But Why explainer^[3] on the subject. In summary, humans are the only sentient technological life in the universe, as far as we know. This accident of evolution is destined, like all species, for eventual extinction unless we somehow back up the biosphere somewhere remote enough to avoid planet-killing disasters, such as another nearby planet. Today, humans have the technology, the wealth, and the resources to do just that. It won’t be easy. It will never be easy. But there’s every reason to suspect that it may not be possible forever, and thus we should take the opportunity to do it as soon as possible.

That is the “why.” This book is more about the “how.”

In 2016, I wrote “How To Get To Earth From Mars: Solving The Hard Part First,”^[4] a book that described primarily deep space human flight challenges for a general technical audience. My intention was to anticipate SpaceX’s Mars BFR reveal at the IAC^[5] in September that year, and to inform and motivate my fellow nerds to think about and solve these problems.

Since 2016 we’ve seen a rapid evolution of space technology, primarily on the part of SpaceX with their successful demonstrations of first stage reflight and Falcon Heavy. There is still a long way to go! In particular, while flying humans to and from Mars could sustain an outpost similar to the space station or an Antarctic station, it’s not entirely clear how to ramp up from that to a fully self-sustaining, self-sufficient civilization.

In Chapter 22 of “How To Get To Earth From Mars,” I wrote briefly about extended stay and industrialization. Even then, it was obvious to me that “what will we do once we get there?” is a much richer problem than “how will we get there, and back?” But all books must have an end and I just hadn’t thought much about it.

Since then, I’ve written a few blogs on corners of this problem and, perceiving a continuing lack of collated public thinking on Mars industrialization, think it’s time to tackle the matter in expanded form. This book begins with a general explication of the problem of industrialization and self-sufficiency, then digs into the details and logistics required to bring that about. The broader concepts are, I think, also quite applicable to building cities on the Moon, asteroids, or in deep space.

The usual disclaimers apply. I do not regard myself as an authority on space. I identify as a well motivated enthusiast with too much time on my hands. I don’t really know what I’m talking about. But, fortunately for me, there is as yet no proven method for industrializing Mars, so I feel comfortable to abandon certainty and to adopt the scientific method. And there, I do have at least some experience.

My research and development work in optics, gravitational waves, and on the Hyperloop were primarily concerned with problems that were broad, dispersed, and poorly defined. Rather than beginning with design heritage and near certainty, we began with near total systemic ignorance. We were about to do something completely new. And, in such a situation, turning the problem over and over until you know 1% of the solution is a definite improvement over knowing nothing.

In this book, I collate a series of essays examining various aspects of Mars settlement and industrialization. Through these fourteen chapters, I show that despite a lack of absolute certainty, it is possible to build a few bridges into the abyss, to identify the locations and nature of various thorny problems, and to understand the general lie of the land.

Let’s get started!

Literature Review

I’m far from the first person to think about industrializing Mars. In this chapter, I include brief summaries of a selection of other sources on the subject. I know of no other books focused exclusively on Mars industrialization, so I will focus only on the most relevant sections of these related texts. My intent is not to deliver final judgment on other sources, rather I will refer forward to chapters in this text which may offer more detail. This chapter is intended as a jumping-off point for interested readers.

As I developed this book, a number of friends in aerospace recommended Industrial Megaprojects: Concepts, Strategies, and Practices for Success,^[6] by Edward W. Merrow. This book is essential, if extremely dense, reading for anyone contemplating a large scale project. Merrow describes in irresistible detail many of the numerous ways that huge engineering projects can go wrong. His list of risk factors looks like a Mars program specification. Expensive - check. Questionable business case - check. Schedule driven - check. Remote - check. Bad climate - check. New technology - check. Multiple stakeholders - check. Scope confusion - check. Uncertain front end loading - check. Safety issues - check. The single most important takeaway from this book is that time and effort spent planning and getting basic facts right is overwhelmingly valuable. One other salient feature is the difficulty of coordinating across multiple teams and interfaces. The case of Boeing’s 777 and 787 development are great examples respectively of how to, and how not to, manage a development program. Mars industrialization is, of course, in a class of its own, which I expand upon in the chapter on Economic Planning.

Mission to Mars: My Vision for Space Exploration,^[7] by Buzz Aldrin lays out aspects of Mars exploration in a very accessible way, interspersed with charmingly inscrutable diagrams of various cycler concepts. In Chapter 7, “Homesteading the Red Planet,” Aldrin gives a summary of the work of various researchers on the topic of In-Situ Resource Utilization (ISRU), but stops short of the Martian industrial revolution.

Dr Robert Zubrin is a prominent and tireless advocate for crewed space exploration and settlement. He has written numerous articles and is the president of The Mars Society^[8].

In The Case For Mars,^[9] Zubrin vividly describes the technological basis for Mars exploration, focused on a general audience. In Chapter 8, Zubrin explains that Mars has a sufficiently diverse geological history to contain minable ores for many industrial feedstocks. Zubrin recognizes that industrial independence will require millions of people and take many centuries, necessitating some kind of trade with Earth and possibly asteroids to obtain high value, low mass goods. He also describes in great detail possible new technology and economic schemes to help pay for the exercise.

In How To Live On Mars,^[10] Zubrin writes from the perspective of a wiley and sarcastic veteran of the Martian frontier, some decades or centuries in the future. The book is focused mostly on day-to-day survival challenges, such as finding water, food, oxygen, and financial liquidity. It exists within a world of bureaucratic centralized industrial control and hard-scrabble marginal ingenuity. While highly entertaining, it doesn’t address issues of secondary or precision manufacture.

Like Aldrin’s book, Zubrin’s model for Mars development assumes centuries of dependence on regular shipments of advanced technology from Earth within a ruggedly individualistic pioneer or homesteading paradigm. In Chapters 4 and 5 I describe the shortcomings of the homesteading concept, focusing particularly on avoiding scaling problems, indefinite dependence, and labor productivity collapse.

Packing for Mars: The Curious Science of Life in the Void^[11], by Mary Roach is a great introductory text for interested readers trying to get their head around the sometimes unintuitive ways that crewed space programs are run. I was expecting to find a chapter with a checklist of things to bring, like a toothbrush and oxygen, but the book is more focused on space science.

Why is Mars so hard? Much ink has been spilled on this subject, but a review article^[12] by James Oberg describes the unique challenge of not only training the hundreds of thousands of people needed to run a space program, but keeping them in sync and passing their expertise on to the next generation.

The High Frontier: Human Colonies in Space,^[13] published by Gerard O’Neill in 1976 describes how humans could industrialize giant space stations in cis-Lunar space, using raw materials gathered from asteroids or launched from the moon. While thought provoking, the initially bold idea of thousands or millions of people living in rotating space stations is a solution looking for a problem. O’Neill settles on space-based solar power as a fundamental product and then constructs a series of deceptively solid-looking assumptions to justify moving power generation off the Earth, followed eventually by all industry. In the early 1970s population growth had peaked at 2% year-on-year, while energy use was growing at 7% following worldwide recovery from WW2 and exploitation of oil. Today, 45 years later, we can see demographics tending toward population steady state of around 11 billion humans.

Finally, it is worth considering how few industries and processes on Earth are, in any way, constrained by the cost of energy. For example, aluminium production, heating and cooling in extreme climates, and supersonic flight are a few rare examples where energy costs are a substantial fraction of overhead. Another way of looking at this is that in 2018, the cost of a 1000 calorie Big Mac Meal is $5.99, and food doesn’t get much cheaper than this. The same amount of energy in the form of gasoline is only 7c - about 100x cheaper. This is why driving cars is economically possible despite them weighing much more than a human. The cost of one Big Mac Meal worth of electricity is about 14c, but because electric machinery can be much more efficient than gasoline car engines, electricity is usually cheaper for obtaining mechanical work, while gasoline, oil, or gas is cheaper for heating. Gasoline also has an established infrastructure for mobile uses. The point is that energy is already really cheap and readily available on Earth, so giant cis-Lunar space stations will need another business case.

The most comprehensive review of automated space manufacturing from the pre-Shuttle era is probably Advanced Automation for Space Missions,^[14] the proceedings of a NASA-sponsored conference in Santa Clara in 1980. This review sketches in broad brushstrokes a self-replicating robotic factory using local resources on the moon. The authors, nearly 40 years ago, could not have anticipated the revolution in commodity and personal computing. Indeed, the shuttle first flew in 1981! In Chapter 5, I discuss the ways in which human and automated labor mesh together to achieve the greatest degree of efficiency.

Any review of literature on industrializing space would be incomplete without mention of the work done by Paul Spudis, collated on his website^[15]. Spudis has argued consistently that lunar resources, most prominently water, could be mined with a robotic program and used to provide feedstock for hydrogen/oxygen rocket fuel in cis-Lunar space. His papers have not yet contained a side-by-side comparison of the cost/benefit of different program development architectures. In Chapter 5 I describe the necessity of some human labor for manufacturing efficiency. More concerning for the lunar water mining industry, SpaceX’s development of reusable boosters has reduced the cost of launch of cargo from Earth enough to overwhelm the economic advantage of shipping any bulk product off the moon^[16].

That said, when I started work on this book I thought that Mars’ relative availability of raw materials made industrialization there easier than anywhere else in space. While the moon has challenges with ore availability, no atmosphere, and slow rotation, it is definitely much faster to get to than Mars. Furthermore, the magnitude of the total industrialization challenge likely overwhelms any initial advantages or disadvantages between the Moon or Mars. I’m just not sure!

While adventures on Mars are a staple of science fiction, only a few novels have approached the industrialization phase in sufficiently rigorous way to be worth mentioning. They are Red Mars by Kim Stanley Robinson, The Martian by Andy Weir, and Seveneves by Neal Stephenson. While thought-provoking and evocative, I don’t feel any of these novels intended to contribute a lot of insight into the process!

Finally, the complexity and target-focused nature of Mars industrialization lends itself well to computer simulation and gaming. Surviving Mars^[17] is a recently released world-building game but openly elides some of the real world difficulty in favor of familiar game mechanics. SIMOC^[18] is a state-of-the-art minute-by-minute total system simulation engine which may one day power both games and system design efforts.

Defining Autarky

Let’s begin by defining “autarky.”

“Autarky” means, literally, self-sufficiency. Although the word has been around since the early 17th century, it rose to prominence in the 1930s. I think the primary reason is that for the first time warfare became highly technological, so industrial and economic self-sufficiency was required to prevail. Any country in the European theater that was incapable of mass producing munitions, planes, and submarines rapidly lost ground to those that could. And some of the breakthrough technologies of the war, radar, nuclear weapons, and assembly line production of ships, conclusively turned the war in the favor of the allies.

Although settlements on Mars are not expected to have to fight wars, they face a similar need for self-sufficiency. On Earth, there are environments that are sufficiently benign that a single naked human can survive and even thrive indefinitely. Living on Mars is more like surviving in Antarctica, under water, or on the Western Front. Advanced technology is non-negotiable for survival in any adversarial environment.

But surviving on a Mars base created in the image of the Antarctic stations is, at best, postponed annihilation. Any interruption in a tenuous supply chain guarantees eventual death, and so the first goal of any Mars city must be industrial autarky - the ability to survive indefinitely without any external help.

It is difficult to precisely define partial autarky, as even ostensibly self-sufficient countries will conduct trade, even during times of war. It may be possible to define, however, a period of time for which existing supplies and concerted effort can keep an isolated country, Antarctic station, medieval castle, submarine, or Mars city alive. This is not just a matter of having enough food. In any given closed system, there is a finite supply of parts and labor and ever-diminishing process efficiency through exhaustion. An illustrative thought experiment is how one could provision an aircraft carrier to last as long as possible on the open ocean. The salient limiting case is overcoming rust and the structure’s eventual cracking and fragmentation, which would occur easily within the lifetime of the first generation of inhabitants.

For any given city in space, there are different degrees of closure. By “closed,” I am attempting to convey that many, though not all, resources one might take for granted on Earth are limited or much more difficult to make. For an O’Neill cylinder in deep space, access to raw materials could be more difficult than for a Mars city, though both would have access to sunlight. For a city on Mars, “closed” refers to a shortage of labor and advanced technology, rather than poisonous atmosphere and a random assortment of rocks.

It is clear that any closed technologically dependent system must possess excess capacity to rebuild every constituent component from only locally abundant raw materials, at least as quickly as they break down in normal use^[19]. Conversely, after a relatively short period of time, critical processes break down in any closed system smaller than some critical size. For the International Space Station and Antarctic bases, this is about a year. For some tactical nuclear submarines, perhaps five years. A Mars city that has attained total industrial autarky has reached a scale where it is reasonable to believe that total supply interruption, while inconvenient, does not imply certain death in the foreseeable future. Such a settlement would not only be able to make its own food and air, but also its own computers and rockets - a daunting challenge for even large, wealthy countries today.

The primary challenge is scaling up to reach this level. More detailed reasoning in later chapters gives a population requirement in excess of a million people, with commensurate industrial development and total labor productivity, to be confident of indefinite survival on Mars’ extremely hostile surface. If the population starts at 100 and is doubled every 26 month launch window, it would take about 40 years to reach this point. The rest of this book is preoccupied with describing how to plan and achieve such an absurd, sustained rate of growth, and thus minimize the total time window in which the city is vulnerable to supply chain interruption.

The final part of this chapter is concerned with inapplicable historical analogies. Industrialization has occurred, and is currently occurring, in numerous places and different ways around the world. The industrial revolution began in Britain, with the invention of the steam engine. The steam engine was a way to get fungible mechanical (shaft) work at scale that didn’t derive its power from muscle and, ultimately, (recent) solar power captured by photosynthetic plants. Instead, it employed coal (ancient, concentrated solar power) to greatly increase the economic product per capita.

Several other nations or economic areas have also bootstrapped their own industry since then, but most have bought or stolen at least chunks of the technology rather than reinventing the wheel. For a large non-industrialized country today, the path to development and industrialization is relatively clear - borrow money from one of the major world powers, sign deals trading some sovereignty for technical assistance and foreign ownership, then start unpacking shipping containers. Once the foundations are in place, countries with sufficient natural resources can build out their own power stations and factories and develop at their own pace.

Nations that prefer to avoid foreign influence, such as North Korea or Cuba, must attempt to industrialize entirely within their own borders. This is not impossible but, in practice, extremely difficult. Cuba, with a mostly benign environment and a population of 11 million, has taken 20 years to recover from an energy crisis but still must import or recycle practically all advanced industrial products. North Korea, with a harsher climate and larger population, initially led development in the Korean peninsula. Today, an ongoing petroleum shortage has led to the total collapse of previously mechanized agriculture, resulting in persistent starvation.

Several medium-sized countries have actually de-industrialized due to the inherent difficulty of competing with international shipping in a small domestic market. Fortunately, cheap shipping is unlikely to drive Mars industry out of business. In an era of globalization, many economically advanced countries, and their scientific and military outposts, do not have intrinsic industrial self-sufficiency and could not redevelop it if necessary. Rather, they are dependent on foreign manufacture to import advanced industrial products such as cars, planes, computers, medical equipment, and even clothing.

The Mars city may begin like an Antarctic base, depending on external supply for everything except perhaps water, electricity, and carbon dioxide - resources which can readily be piped to the front door. Unlike an Antarctic base, which is hastily constructed to a predetermined design, the Mars city is designed primarily to grow incredibly rapidly, its final form only vaguely conceived at first.

Awkwardly, there is a dangerous decade or three when the city is both much too small to achieve autarky but also much too large to evacuate. Rapid development is intended to shorten this window.

Over time, technology and population both increase to bring a nested hierarchy of industries Mars-side. This technology is developed and tested on Earth and deployed on Mars only when its scaling is sufficiently labor non-intensive.

Ideally, an industrially self-sufficient Mars city would continue to enjoy flights from Earth, but is no longer utterly dependent on the financial and technical health of the Earth-side institutions that built it.

Path To Autarky

In the previous chapter, the concept of “autarky” was introduced and defined. This chapter is intended to explain the path to autarky.

Broadly speaking, there are two complementary ways to achieve autarky over the settlement period, which I hope can be measured in a small number of decades. The first is to scale, or increase the population. The second is through process efficiency enhancement, which decreases the critical population size. Both of these are vitally necessary, and both will have to achieve between three and six orders of magnitude of heavy lifting. If we lived in an era before Moore’s Law, I would give up now. But I think it is possible.

Scale is not merely a matter of dumping a million humans and telling them to get cracking. Aside from the difficulty and expense, each additional person needs to increase the average productivity of the city. This will require meticulous planning and support on Earth. Essentially every new batch of carefully selected Martians each 26 month window is there to instantiate a new industry, and the productivity of the city needs to keep up with the growth. Labor on Mars is the most limited local resource, so it will be primarily occupied with provisioning for the delivery and extraction of more labor. On the Earth side, doubling shipments every two years will require not only steadily increasing numbers of spaceships, but also continual development of larger and faster spaceships.

The other side of the coin is process efficiency enhancement. Fortunately, nearly all the grunt work to improve this can be done on Earth where both labor and breathable air are comparatively free and unlimited. Unfortunately, the challenge is just as daunting as scale. Progressively increasing the productivity per person by a few orders of magnitude is not as simple as 3D printing all the things, although automation and mechanization have already increased personal productivity by a couple of orders of magnitude since the industrial revolution. This process can be observed in real time in China where the children of peasants receive an education and enter the middle class, sustaining 12% GDP growth among other advancements.

Although Mars labor, being nearly entirely immigrants, won't have to include most of child rearing, education, training, and aged care, the required productivity can't be achieved simply by drawing a bigger line around an ever expanding demographic of skilled laborers. Humans and machines on Martian ground will have to get smarter about process efficiency on a weekly basis. Finally, processes that reduce labor demand on Mars may help reduce costs on Earth, ideally providing a mechanism to keep program costs stable even as shipment volumes literally skyrocket.

The surefire sign that process efficiency enhancement has overtaken population scale is the emergence of a Mars diaspora, or city disaggregation. At this point, the Mars settlement topology will shift from a central industrial hub with remote mining facilities to a much larger set of smaller communities that, while engaging in plenty of trade, correctly believe that if necessary, they could survive independently. If the critical population of a million people is achieved in four decades, then on average the population will have increased by a factor of ten every ten years. If the critical population size, which is difficult to estimate, decreases from a hundred million to one million in the same time frame, then it is fair to assume that any technological overshoot could, in only ten more years, decrease the critical population size by another factor of three. If population growth, which is almost entirely emigration from Earth, continues at the same rate, then thirty (!!) independent cities-worth of people could be created on Mars only a decade after achieving autarky.

Post-autarky Mars (and by extension, Earth) is the very opposite of “mission accomplished” boredom. Instead, humans will face a situation of runaway industrial excess capacity for the first time in history. This is the best time to start terraforming in earnest, since any prior attempts that aren’t themselves exponential on similar time frames would be swamped by later almost trivial effort. Other potential applications of post-autarky industrial surplus include settlement of increasingly more hostile environments, of which Mars is but a first step on an almost infinite path, such as various airless moons, asteroids, and deep space. Closer to home, small-scale industrial autarky technology could be applied to comprehensively restore and preserve Earth’s existing environment, and to build the first interstellar spaceships.

The Labor Problem

Labor management is the defining challenge of Mars industrialization. It will be, by far, the longest, largest, and most closely coordinated megaproject ever attempted.

The end condition, autarky, is defined not as the point where anything can be made on Mars, since this could be done with only a thousand people and a lot of time, but as the point where everything can be made on Mars, at least as fast as it breaks down. This requirement for very high productivity is simply a restatement of the labor problem: So much to do, so few people to do it.

This chapter unpacks the labor problem. I begin with subject-specific contextualization, then discuss mechanization and automation, division of labor at different times and scales, and finish up with a prescription for mechanical reliability.

Contextualizing labor

It may not be obvious that there is a latent demand for human labor on Mars. To understand what human labor is for on Mars, let's begin on Earth.

Within living memory in western countries, most human labor was preoccupied with agriculture. Mechanization, irrigation, pest control, fertilizer, variety cultivation and numerous other technologies have greatly increased agricultural productivity, to the point that farming today is done by an ever shrinking fraction of the populace and there is no near-term risk of famine.

Instead, people have specialized in tens of thousands of other jobs to produce all the cool stuff we take for granted. Nevertheless, with total employment in 9 to 5 work there is a labor surplus in the US, and has been since the end of WW2. Some economists worry about finding enough work for everyone and stimulating demand, given that most fundamental needs can now be met rather cheaply. Indeed, it is not too hard to imagine a future where automation has eaten nearly all the jobs and most humans subsist on government-disbursed rations, entering the workforce only to gain access to capital for unusual treats, or because they fundamentally want to work.

On Mars it will be a bit different. An “agriculture first” industrial bootstrapping analogy is not the best place to start, as there is a lot of necessary dumb mass that's easier to make than food. Nevertheless, a relatively tiny number of humans will have an overwhelming amount of work to do.

But what are humans needed for on Mars at all? They are hard to keep alive on frozen planets with poisonous air. I don't think there is anything large numbers of humans can do on Mars which is valuable for commerce on Earth, such as mining or R&D. Indeed, it is hard to find a better, or more motivating, use for humans on Mars than building habitats to keep themselves alive. It's a system with two stable states. The default option is no humans, no city. The interesting option is a city, which needs humans to build it.

Finally, the reason large numbers of humans are necessary is that humans are only efficient enough when specialized on a task. The growing Mars city needs many thousands of people to exploit the efficiencies of division of labor.

Manual-mechanized continuum

Why use humans to build a Mars city, given that today Mars is a planet inhabited solely by some very lonely robots? Why not deploy a “turn key” Mars city with automated robots and have humans show up to cut the ribbon?

The most obvious reason is that we need humans out there learning how to build everything along the way, and a Mars city that achieves autarky does so through demonstrated, ongoing productivity.

The other reason is that a robotically-constructed Mars city capable of a 100% automated industrial stack is science fiction. In the previous chapter, I talk about the Earth-side heavy lift of developing process efficiency technology adequate to reduce the critical population of the Mars city to only one million. An automated city reduces that number to zero, which is essentially a von Neumann machine^[20] in disguise. Von Neumann machines are hypothetical mechanical devices capable of self replication.

Von Neumann machines are not only not forbidden by the laws of physics, they are basically the foundation of biology. Nevertheless, there is a good reason why 100% robotic construction is not an optimal solution in nearly all cases.

For any industrial process, there is a U shaped curve relating labor to cost. The horizontal axis is a continuum from manual to mechanized. Any given process will use a mix of human and mechanical labor to minimize cost. As time and technology have progressed, the bottom of the U has tended towards mechanized methods, shown with the purple arrow.

Mechanization of manufacturing traditionally refers to the use of mechanical power to augment or replace the use of human muscle and manual effort, but more recently automation has also been employed to augment or replace mental effort.

I will supply two examples. The first is product focused, while the second is manufacturing process focused.

First, consider three different assemblies; the pyramids, a house, and a mobile phone.

The pyramids were built about 3500 years ago with perhaps tens of thousands of people over a few decades. The motive force was human muscle, powered by grain. If one built a pyramid today, there would be tractors and cranes involved!

In contrast, a single contractor can build a modern house in weeks because of her access to power tools and prefabricated, standardized components. A handheld power tool is an interesting case study. A human operator provides alignment, troubleshooting, maintenance, and the motor does the work. Such power tools are designed to be able to be handled, albeit often tiring to use. Still, using a grinder is much easier than, say, cutting through a steel beam with a hand file. Imagine the complexity and expense of a cordless drill that could be remotely operated, let alone autonomous. My Ryobi cost $26 including shipping. The Curiosity Mars rover cost $2b, and sometimes its drill doesn’t work for a year at a time.

Finally, consider the motherboard of a mobile phone. Made in enormous volumes, assembly and often even board design is automated. Custom pick and place machines place hundreds of thousands of components an hour, requiring only nominal supervision.

The second example concerns another pinnacle of industry, the car. Although volume production will always use automation, the production ramp represents a real time traversal of the manual-mechanized continuum. Initially, designers, engineers and technicians hand assemble the entire vehicle. During this time they are finding and fixing problems when they are still isolated and relatively cheap to fix. Over time, manual assembly work is pushed back to suppliers or handed over to construction robots, which almost always represent increased rate at the cost of decreased ability to adapt to emerging issues or challenges.

Labor scarcity on Mars will push the manual-mechanized equilibrium further to the right and up against the limits of automation. Nevertheless, responsive and careful use of human labor is the best way to keep the project moving at top speed.

Responsible use of labor

In labor intensive projects on Earth, particularly ones operating in hostile environments, there is an operational preference to cycle staff. Whether executed through organized tours of duty, or through churn and burn hiring practices, this approach doesn't translate well to Mars.

Orbital constraints mean that minimum tours of duty are about 580 days. The cost of transport and onsite training are high enough that there's a good incentive to send staff to Mars permanently. Once on Mars, rotating crews through different styles of work is preferable to match labor demand and availability in a 26 month build cycle^[21], and also to ensure that everyone gets a turn at working in a spacesuit outside the pressurized living areas, or any other area of interest.

Permanent and irreplaceable construction staff will require a somewhat different and rather enlightened approach to labor resource management. Above all, labor utilization must be sustainable to help a person develop and deliver the most value over ten or more years. This philosophy has more in common with labor practices before the onslaught of EPC (Engineering, Procurement, Construction) contracting in the 1990s, which attempted, despite axiomatic impossibility, to commodify skilled labor.

I don't intend to discuss incentive structures here, because at best they cannot overcome issues caused by poor program management. It is imperative that each worker experience the challenges, care, and good sleep they need to gradually cycle through a variety of roles in the city.

Arbitraging relative labor costs

Outsourcing is a standard procedure to arbitrage labor markets. If the cost of offering a service remotely is cheaper than on site, it is hard to argue that local functionality is required. Of course, higher order moments of value than money, or externalities, have additional input to the outsourcing discussion.

In the previous section, I broke down the mix of human vs machine labor. In this section, I add the additional axis of the relative cost of labor on Mars and Earth.

There are few things more expensive than hiring a top software engineer in Silicon Valley, but the cost of Mars labor, including overhead, will put that into perspective. For ballpark numbers, relocation alone is in excess of a million dollars, while the overhead on life support equipment and other infrastructure probably amounts to easily $10m/person/year, at least in the first decade. To first approximation it doesn't even matter what their base salary is, or in what currency it is paid.

With that in mind there is a >100:1 bias to do anything Earth-side possible. This includes design, construction, testing, frustration-free packaging, training, and selection, but also software development, data processing, remote debugging, and remote operation of any process or machine that can cope with the 5-40 minute communication delay between Earth and Mars.

I can even imagine that a Mars-side construction crew would have a dedicated Earth-side team monitoring their data and solving problems on the fly. Why spend time leafing through manuals or IKEA instructions if a head up display can provide the same functionality and less time wastage? Data is by far the easiest thing to move across space. A Mars-side worker could invent a new tool, dispatch a loosely descriptive voice memo, and wake up the following morning with a 3D printed prototype, designed and tested on Earth, waiting for them.

How does the labor lifecycle close? A human being generally consumes labor during childhood and old age, while being a net producer in the middle. Humans mostly will arrive on Mars after extensive training on Earth, at the peak of their productive capability. Even if raising children on Mars wasn't a huge time sink, they simply cannot be produced fast enough to double the population every two years!

Retirement is a different matter. There are several options. Workers could readily fly back to Earth on spaceships making the return journey. Readaptation to Earth gravity, while difficult, is probably not impossible. Alternatively, the unusual demographic structure on Mars means that veterans with more than a decade of experience are always a minority at less than 10% of the total population. I think workers could shift into supervisory roles or semi-retirement without compromising the overall health of the labor pool. Indeed, some latent capacity of seasoned veterans helps insure against unanticipated demand.

Machine labor exploitation

The last important difference between disposition of labor on Earth and Mars is optimal mechanized labor exploitation.

On Earth, it is not unusual for companies and individuals to devote a lot of time and effort to machine maintenance, sometimes in excess of the capital optimal balance. I personally know a few engineers who have lovingly restored antique machine shop tools, partly to have their own when a new one is impossibly expensive, but also to learn that skill.

On Mars, both human and machine labor is much more expensive than on Earth. Further, human labor is relatively much more expensive still. That is, machines on Mars must be commensurately more reliable in order to survive on a leaner maintenance schedule.

More concretely, consider a typical commodity industrial machine with wear parts and a finite lifetime, such as a compressor. On Earth, a new compressor may require one man hour of maintenance for every 10 hours of full duty equivalent use. By contrast, a fighter jet may require hundreds of man hours of maintenance for every hour of use.

On Mars, there is a different calculus for the maintain/replace question. Before local full scale production of a given machine, the cost of a new unit is primarily transportation, or weight, rather than purchase price. And the cost of maintenance is not calling a third party technician once it's out of warranty, but sending a whole extra person plus all the stuff needed to keep them alive, or worse, cannibalizing labor from primary installation teams and breaking everyone's schedule.

This sort of economics is already seen in capital intensive industries such as deep ocean oil drilling. A highly skilled driller knows a troublesome part is not even worth the time to remove it non destructively, let alone attempt a repair. Instead, they throw in a new part that will definitely work and move on to the next task. The only exception to this calculus is in healthcare, since human laborers are not, despite some common misconceptions, intrinsically disposable.

On Mars, broken parts or machines can be readily stockpiled. When the level of local industrialization catches up with that part, it will be easier to make new ones out of recycled materials than undifferentiated raw dirt.

On Mars, imported machines at every level of complexity must be designed and tested to work flawlessly with perhaps 100 times less maintenance than a similar machine on Earth, and to be easily replaceable. On the system level it is easier and cheaper to iterate designs and send essentially disposable new machines every 26 months to top up a stockpile.

A less compelling alternative is to attempt to re-engineer commodity machines to last forever, since the commensurate increase in complexity will outweigh the intended reliability benefit even if only 1% of the parts turn out to unexpectedly require labor-intensive maintenance^[22].

In summary, design that favors simple, reliable, replaceable parts is prefered to complicated, inscrutable, eternal, hard-to-maintain juggernauts.

Development Prioritization

This chapter describes how to answer the question “where does my X on Mars come from?” and how to think about ranking or prioritizing local production capacity.

Lifecycle of individual product

Product procurement will vary as a growing Mars city steadily insources production of things requiring progressively more sophisticated manufacturing. Any given part requiring local manufacture will progress through the following stages.

One off: A specialized workshop with a stock of raw materials can produce nearly anything, with terrible process efficiency. It exists only to insure against catastrophic supply depletion or unanticipated need.
Pilot production: A small scale plant that doesn't meet demand but enables process testing and helps “design for manufacturing” get detail right before large scale investment of resources.
Mass manufacture: Intended to entirely meet local demand, so that shipment bandwidth from Earth can be focused on ever more specialized supplies. Given that the overall scale of the base is increasing exponentially, production rate and labor efficiency must display similar growth to avoid being on the critical path.

The requirement that overall process productivity continue to scale exponentially means that human labor, which is fundamentally limited, must aggressively retreat up the process chain. In one-off manufacture, human hands handle individual components. In pilot production, humans operate machines. In mass production, humans assemble the machines that autonomously produce and dispatch the product. As mass production ramps, machine assembly itself must be gradually automated behind yet another layer of mechanized abstraction. And so on!

Factors that affect development priority ranking of industrial projects

There are three main factors that affect development priority ranking of industrial products. These are the difficulty of local manufacture, the difficulty of shipping, and the criticality of supply.

Complexity or difficulty of manufacture

It is nontrivial to produce an absolute scale of manufacturing difficulty, but broadly speaking there is a continuum between dumb and smart mass. At the low end is oxygen, which is needed mostly to fly spaceships back to Earth, is a single element, and must be produced locally from day one. At the other end are thinking machines, humans and computers, that are difficult and slow to make.

All manufactured items have required sub-components or raw materials whose availability strongly determines manufacturability.

In the first decade of the Mars city, raw materials that can be obtained from the environment in gas or liquid phase are preferable as they can be acquired without multiple parallel mining operations. These include CO2, water, carbon, hydrogen, oxygen, and nitrogen. Soon after, undifferentiated rock and dirt used for masonry, insulation, or shielding can also be procured with minimal effort compared to, say, the processes required to produce 7000 series aerospace grade aluminium. Carbon, hydrogen, and trace additives are adequate to produce structural plastics, by which point the larger fraction of mass and bulk for many products can be produced locally.

As local bulk production grows, steadily more sophisticated secondary and tertiary products are possible. The local capability landscape is essential to understanding a natural ranking for manufacturing.

The third major consideration for local production is the availability of automation and tooling that enables production without overwhelming labor requirements.

Finally, production time scale is much more important in a rapidly growing economy. If a product needs to be built today to match supply after 5 years of production process, that will consume a lot of space with partially processed inventory. More importantly, it will also require outsized production to match demand, since 5 years will see the city grow by a factor of about 4. In other words, products like high quality wine and humans are best produced on Earth, where there is essentially an unlimited fungible supply, then shipped to Mars in the required quantities.

Transportability from Earth

A second factor that affects the local make/buy choice is shipping from Earth. Broadly speaking, this breaks down as mass, bulk, spoilage, and risk.

Mass is the hardest limit. SpaceX wants to build a rocket that can fly 150T of cargo to Mars. This is roughly equivalent to 6 shipping containers. For comparison, the largest containerships can move roughly 2000x as much, and there are hundreds of such ships. Therefore, products that are very heavy, particularly bulk products that are less difficult to make on Mars such as fuel, water, oxygen, and rubble, are good candidates for early Mars manufacture.

Bulk is the second consideration. Objects that are too big to fit through the spaceship door non-destructively are better candidates for local manufacture. I think the door may be vaguely palette sized.

Third, spoilage. Stuff that rots or melts or has handling requirements more stringent than enduring 5gs and lots of vibration are preferred for Mars production. As an example, some fresh lettuce could complement long shelf life imported food, which provides the bulk of the calories.

Finally, supply timing and risk. Orbital mechanics dictates that supplies can only be sent during a launch window that opens for about 2 months every 26 months. Supplies arrive only intermittently, so stockpiling and inventory management are very important. Supplies that have a limited shelf life are difficult to supply consistently from Earth.

More pressingly, expected supplies might not arrive if there are problems with launching or flying. It is wise to plan that some fraction of spaceships might not make it. Accordingly, critical supplies should be split over multiple flights. Further, supplies should be structured such that a slight supply disruption does not affect overall scaling, while all but total catastrophe does not risk project viability at any point. That is, some shipping losses are planned for, and adequate supplies are stockpiled in the event of, say, two consecutive launch windows with no successful flights. As the project proceeds, more accurate actuarial data can be derived and applied.

Predictability of demand and criticality of supply

The final factor that affects production ordering is predictability of demand and criticality of supply.

Given the 26 month gap between supply windows it is important to be able to predict demand at any point during the production ramp. There are all sorts of ways this can go wrong, including consistent under production, feedstock shortages, or labor exhaustion. For small variations, it is reasonable to assume that supply is elastic, but this is not the case in general.

The criticality of supply varies for any given product. It depends on the lifetime or replacement rate of the product, its effect on overall system health, and the degree to which its production depends on other marginal or critical products.

In practice, even the best system design is unlikely to prevent the emergence of a relatively small number of “linchpin” products, which will attract much attention to ensure their smooth supply. It is difficult to predict in advance what they will be!

In general, however, products that neither use or in turn support too many other products reduce bottlenecks, supply web complexity, and industrial interconnectedness. This makes the system topology more diverse and less vulnerable to unforeseen issues. As the city increases in scale, system diversity will also increase, but there are still points of vulnerability.

Tentative ordering of industrial production

At this early stage it is not possible to perform a granular prediction of the prioritization order of local production of stuff. That said, a categorical order can be painted in broad brushstrokes. In particular, some industrial products vary in mass requirement and manufacturing difficulty by at least an order of magnitude. Here is my best guess for production priority, together with a usage estimate.

Initially, each human will require an overhead of cargo measured in the hundreds of tonnes. Over time, local production will reduce that overhead, allowing humans to be an ever larger fraction of the total shipment.

Oxygen. Each spaceship requires 2000T of oxygen to burn with 400T of methane to fly back to Earth for reuse. This is more than can be brought from Earth, and must be produced locally. It dwarfs the requirement for breathing atmosphere, which can be mostly recycled.
Hydrogen, water, methane. 400T methane per spaceship, which is 100T of hydrogen, or about 1000T water, assuming no losses. Ideally, water could be tapped from a liquid aquifer. Water can be partly stored in habitats, where it provides thermal mass and fills swimming pools.
Plastics. A robot arm could be 95% plastic by mass, and 3D printing of plastics, together with plastic pressure vessels, could provide the most easily machined raw materials. Consumption will climb over time but could easily be many tonnes per person per window.
Masonry, concrete, rubble. Digging holes and foundations, or preloading brick vaults uses materials that don't require chemical processing, just big bulldozers. Vaults could be automatically constructed using a bricklaying robot arm, easily to the level of many tonnes per person per year.
Steel. Steel might be the most easily extracted and versatile metal to produce, provided concentrated deposits of iron ore are conveniently located. Will blast furnaces require different geometry in lower gravity?
Food. Food is nearly trivial to grow on Earth, but on Mars will require large areas of insulated green houses or lots of electrical power for stacked palettized hydroponics. One human needs about 200kg/year, depending on caloric quality and physical exercise. Even with intensive farming and good quality soil, which will have to be synthesized, lower intensity sunlight on Mars may require on the order of an acre of cultivated land per person.
Advanced alloys. Materials that use aluminium, copper, titanium, rare earths, magnets, batteries, etc all require metals that are harder to make than steel. Over time, the full range of industrial alloys can be produced in the required quantities.
Electronics. Motors, actuators, electromagnets, batteries, and other bulk electrical components that don't involve silicon are probably next up, as they need a wide range of metals. They form the bulk of work-performing machine components.
Pharmaceuticals, appliances. With generic chemicals and electronics available, most drugs and appliances can be produced on Mars. Per-capita consumption may be a few 10s of kgs per window.
Computers. Computers are hard to make and weigh almost nothing compared to humans, with whom they will be competing for shipping bandwidth by this point. However, given how dependent any Mars city will always be on automation, the ability to produce them locally at scale is vital to achieving autarky.
Humans. Humans are expensive to produce and slow growing, so labor on Mars will be mostly sourced from Earth. I am confident that solving human reproduction issues in space is easier than replicating the entire industrial stack. I think normal human reproduction will be natural in the post-autarky/disaggregation phase, when the number of potentially self-reliant cities on Mars grows rapidly.

Estimating Development Rate

Up until this point I have assumed a growth rate that doubles the population every launch window, which is very roughly equivalent to a factor of 10 every decade. This chapter unpacks the constraints that determine, with the goal of maximizing, growth rate.

Although the rate will vary, and is hard to define on short time scales, it can be measured in terms of either growth per year, or in time required to double the size, which are interconvertible. As an example, China has sustained economic growth of about 12% per year, which implies a doubling time of about 6.12 years. Conversely, Moore’s law implies a doubling of processor performance every 18 months, which is about 59% year over year. I think a Mars city growth rate between these two is probably achievable, and should be as high as possible to build the scale needed for autarky.

If autarky depends on achieving scale and process efficiency, then development rate is determined by the time derivative of these two factors.

Factors that affect scaling rate

The primary limitation on scaling is spaceship production, retirement, and upgrade rate. Provided that production is gradually ramped up, lines duplicated, and steadily more ambitious designs are instantiated, shipping capacity can rapidly increase. As later passengers need less cargo per head, doubling population does not quite require doubling of capacity, but it is pretty close. For example, if the spaceship production rate grows linearly, then cumulative shipping grows quadratically with spaceship reuse. A quadratic approximates an exponential for longer, but maintaining exponential growth requires a steady cadence of capacity increases at every level.

Given a fixed rate of cargo capacity improvement, the next best way to improve scaling is to increase infrastructure efficiency. For example, if each of the first hundred people need 100T of cargo to get started, then each subsequent 10x of population requires 10x less cargo, total cargo shipments remain constant for an ever increasing population. If a 10x increase in population requires instead only 5x less cargo, then the population can scale only half as fast. There are two ways around this: Produce more of the most mass hungry cargo on Mars, and make the cargo sent to Mars lighter. Both of these are engineering challenges that can be mostly solved on Earth, where engineers are the cheapest in the universe.

Given a known shipping capacity and fully optimized cargo manifest, the next most likely cause of scaling headaches is poorly planned technical debt. The most optimal infrastructure deployment strategy will always incur technical debt, meaning a known quantity of work that will have to be redone, or fixed, in the future when it causes problems and resources are available. Ideally, system upgrades and part replacement can be anticipated, planned for, and kept a safe distance from the critical path. But if a lot of technical debt comes due at once, the scaling process may be bottlenecked by both the perpetually short supply of labor and process blockers.

Finally, instantaneous rate limiting steps are day-to-day bottlenecks or gating functions that can't be fixed without major headaches. Any complex, interconnected system will have dozens if not thousands of rate limiting steps. The key to effective system design is to ensure that the subsystem rates are well matched to overall system performance targets. In this case, no subprocess will actually reach its limit and affect the system as a whole. When trying to bring a process up to design rate, various rate limiting gremlins will reveal themselves. Each is, in some sense, a false summit. No matter how quickly each issue is resolved there is always another lurking not far behind. With any luck, Mars city system designers will have the benefit of rapid and effective feedback and learn how to improve efficiency on the fly.

Estimating the required scale

Throughout this book the autarky population of one million has been used. Is there a better way to guess what the critical population might be?

Despite its imperfection as a method of delineating economically distinct units, consider the following list of countries, ranked by population in descending order. Bolded countries have potential industrial self sufficiency, or autarky, as evidenced by the ability to autogenously produce all advanced technology, including planes, turbines, computer components and mobile phones, large ships, advanced weapons systems, and rockets.

China, 1410m
India, 1339m
USA, 325m
Indonesia, 264m
Brazil, 209m
Pakistan, 197m
Nigeria, 191m
Bangladesh, 165m
Russia*, 144m
Mexico, 129m
Japan, 128m
Ethiopia, 105m
Philippines, 105m
Egypt, 98m
Vietnam, 96m
Germany, 82m (as part of larger European bloc)
DR Congo, 81m
Iran, 81m
Turkey, 81m
Thailand, 69m
UK, 66m
France, 65m
Italy, 59m
Tanzania, 57m
South Africa, 57m
Myanmar, 53m
South Korea, 51m (as part of larger east asian bloc)
Colombia, 49m
Kenya, 48m
Spain, 46m

If we were being more rigorous, we might have listed only the population of people in the middle class and above, though it wouldn't change the analysis by much.

Just six countries, and Russia really hasn't been competitive on computer technology since the 1980s. The smallest, South Korea, has 51 million. Politically isolated countries such as Iran (81m), North Korea (25m), and Cuba (11m) have every reason to pursue autarky, and are still highly unlikely to achieve it despite concerted effort.

Could a nation as advanced as South Korea survive if transplanted to Mars? Not even close. The degree of technical pain required to create shirtsleeves environments for much of the economy, and to operate the rest in a frigid vacuum, is a huge effort multiplier.

So the figure of one million people comes with a lot of process efficiency improvement and some hand waving.

Example population scaling simulation

In this section, I present results from one scaling simulation. With a very small set of parameters and simple assumptions, close to exponential scaling is achieved and a population of a million before 2050. This section is adapted from a blog^[23] I wrote on the subject in 2017.

In previous work, I have used the following graph to explain the relationship between population and self-sufficiency under a variety of scenarios, including constant and linearly increasing cargo capacity. It turned out that the final result did not much depend on how many rockets were available when, but the timescale certainly does. In this simulation, I built on the SpaceX exploration architecture, circa 2016, which used slightly larger rockets than the 2017 version. The most fundamental bottleneck is the rate of rocket construction, so we will explore how construction rate affects the population and self-sufficiency timeline.

This graph shows a schematic relationship between population (horizontal axis) and mass self-sufficiency (vertical axis) under a cargo-constrained Mars settlement scenario. The settlement begins at the bottom left and scales towards the top right, where at some population likely exceeding a million people they are sufficiently industrially diverse that they no longer depend on crucial technology to be shipped from Earth.

Before I dig into actual numbers, I’m going to state my assumptions. For better or for worse, a lot of space-exploration themed writing, technical or otherwise, does not hew to the best possible standards for rigor. Here, I’m not going to delve into religious disputes about asteroid mining, lunar fuel stops or any other peripheral concept that’s not related to the core bottlenecks.

Second, there are two primary phases of the settlement timeline. The first, corresponding to the region of the red line below the purple cusp in the diagram above, marks the phase where scaling population within the limits of cargo shipments is a growing challenge. Loosely speaking, this challenge peaks with the successful instantiation of ore mining and refining for every industrially relevant metal and chemical - requiring interaction with the raw, hostile Mars environment and the crushing of many kinds of rocks. This phase is also the phase most directly applicable to current technology and projections.

Assuming the first phase proceeds more or less as planned and everyone doesn’t die, the second phase marks the rush from the cusp to full industrial independence. By this point in the program, at least decades after initial landings, technology at every point of the exercise will have evolved to the point where predictions are difficult to make at this early stage. Specifically, I expect that the forcing function of extreme Mars labor scarcity will result in dramatic improvements in rockets, automation, and manufacturing. It is possible, even likely, that this flowering of technology will reduce the minimum viable population more rapidly than ever-expanding immigration increases it.

That is, at the point of the cusp perhaps 20 years after initial landings, best estimates may still place the minimum viable population at 10 million, at least 30 years away even if the population doubles each launch window. At that cusp, net immigration could be in the tens of thousands per window but would have to increase to 100 times that, something I think rather unlikely with expected technology.

Instead, rapid improvements in extraction and manufacturing technologies will reduce the minimum viable population to less than a million and perhaps less than tens of thousands. As this trend continues, it will be possible to launch entire self-sufficient cities in one go. Perhaps a few decades later, Mars will have thousands of self-sufficient towns, even though the total population may never reach the 10 million this scenario originally thought to be required.

It is important to emphasize that self-sufficiency in reality represents capability rather than practice, since trade and imports will always help increase overall economic efficiency.

I will sketch a picture of phase two, but first I will provide some numbers. Afterall, if the first self-sustaining settlement doesn’t get built, there’ll be no way for the ones that come after.

Phase One - Developing Primary Production

As I explain in my first book^[24], the hard part is getting rockets from Mars to Earth, and to a lesser extent, from Earth to Mars. Here, I’ll explain the constraints on total shipping capacity, then build a model that creates a plausible shipping capacity roadmap.

Important Spaceship Properties

Cargo capacity to the Martian surface. Based on the IAC2016 talk^[25] and subsequent tweets, initial SpaceX Mars ships will have a cargo capacity of around 300T to the Martian surface. Second generation ships may increase this to around 1000T, but further increases are limited by a variety of physical constraints including the thinness of the Martian atmosphere.

Whether it can be reused and rate of reuse. The Mars ship is composed of a spaceship, a tanker, and a booster. Initial boosters and tankers will be flown between 6 and 30 times to refill the spaceship. The first spaceship will fly to Mars, spend nearly 2 years on the surface making propellant, then fly back to Earth. While the first few spaceships will be put near the Smithsonian, later spaceships will be able to fly to Mars every launch window after the emplacement of a fuel/ox production and storage plant near the launch zone, so that the spaceship can immediately fly back to Earth before the launch window closes, as shown in the figure below. Much later, improvements in engines could permit two flights to Mars per launch window. Over this time, the total number of Mars flights a spaceship can perform before retirement will also gradually increase.

Rate of construction. These spaceships are super complicated and difficult to make. Initial spaceships could easily take multiple years to build, and still be overweight. Over time, the construction time will decrease and a single production line can make more of them per launch window, increasing the total number of spaceships. Additional parallel production lines can be built, perhaps by other space agencies using related technology, which also increases the total number of spaceships.

How many spaceships are there per year?

To answer this question I built a Mathematica model^[26] that takes as inputs a function for the construction rate of various types, and outputs all sorts of information about total flights and total mass. Here are the key results.

This table contains a summary of all the different types and versions of spacecraft used in the model.

This is a reconstruction of build rate (per window) from the global manifest data. We see here that as Version 1 reaches rate Version 2 is in the early production phase, on a roughly 8 year design cycle. After 2042, Version 2 production dominates investment and an additional line is added.

This graph shows how spaceship production and reuse increase the payload to Mars year over year. From 2042, Version 2 lifts the total throughput by nearly an order of magnitude.

This graph shows the cumulative cargo transported to Mars, reaching the crucial million tonne mark in about 2052. Given that mass transport begins in 2027, this process takes only 25 years to achieve.

I had a couple of surprises when seeing the results of this model.

First, total payload capacity increases very quickly. The period of time for which an initial settlement is constrained by quasi-constant cargo capacity is basically non-existent. This actually makes sense heuristically, in that it’s easier to build lots of spaceships on Earth than it is to build a complete industry in space. It has a positive consequence too, which is that if the general relation between population and mass independence is maintained, the overall population can be scaled up even more quickly than before.

The second surprise was that there is genuine utility to building a Version 2 spaceship with three times the capacity, as it compresses the timescale to reach a million tonnes of shipped cargo by 15 years.

How quickly does the population scale?

This is another difficult question to answer, but assuming a population-industry trajectory like the red curve given in the first graph above, the total mass each sequential settler has to bring with them can be predicted and a population-mass relation extracted.

This graph shows the total mass payload per person, assuming that the first 10 people, landing in 2027, consume the 900T of payload then available, and that the residual payload is 500kg, enough for a person and the food they have to eat on the journey.

This graph shows how the cumulative mass shipped scales with population. The population reaches a million people as the cumulative mass hits 620,000 tonnes.

This graph shows how population grows as a function of time. Here, the population exceeds a million in the 2050 launch window, 23 years after first landing. It turns out that even though growth is ultimately sub-exponential, it still grows extraordinarily quickly.

This graph shows the window over window fractional population increase. The population grows very rapidly in the first decade to around 10,000 people. This reflects the easy gains of rapidly increasing shipping capacity and gas/water processing for plastics and propellant. 10,000 people is enough to begin mining and processing of metal ores to complete the set of available Martian feedstocks for the development of advanced industry.

Window over window gains drop below 2 from 2045 as all available space in Mars ships is consumed with passengers. If further explosive growth is needed, more ships and more flights are needed to transport people.

What does it cost?

In the previous section we eliminated mass to discover the population-time relations. Here, we reslice the data to discover the mass per passenger on a launch window basis.

This graph shows that by 2033, cargo mass per passenger has fallen to about a tonne, putting a ticket within reach of a middle class family. Someone arriving in this launch window could be the 10,000th person on Mars, and will mark the transition from program-selected specialists to self-selected professionals.

In 2035, a Version 1 spaceship can carry about 300 passengers, each with a metric ton of cargo. By 2044, a Version 2 spaceship can carry about 2000 passengers, each with 500kg of cargo.

Let’s adopt some ballpark numbers. A version 1.5 spaceship+tanker+booster may be constructed for a price comparable to a modern composite passenger jet, say $500m. Each refit costs $100m (of which a tiny fraction is propellant), for a total lifetime cost over 16 reuse cycles of $2b, or $125m per flight. If this is split evenly in time and between 300 passengers in 2035, the per ticket cost is around $420,000. A version 2.5 spaceship+tanker+booster will cost $500m to build, $50m to refit, and fly 30 times. Split evenly, the per-ticket cost in 2044 is $33,000, for a $65m/flight total cost.

Unfortunately it is difficult to be more precise than this, due to multiple cascading uncertainties, or estimation bias. By the onset of “general admission” tickets in 2035, many billions will have already been spent on development and construction of spaceships which may not recover their construction costs in regular service for decades.

That said, I can attempt to estimate development and construction costs. Design rate and cost for both spaceships is $500m and 20/window, which works out to around $4.5b/year. This starts at the beginning of the program, even if the production rate doesn’t reach design rate until 8 years later. Thus construction costs alone reach $4.5b/year in 2022 and $9b/year in 2032.

Reuse costs are initially low due to low numbers of reused spaceships, but eventually dominate overall program costs. By this point, however, ticket revenue will effectively offset this cost, and eventually fund the construction of new ships and the entire program.

The primary financial outlay, then, occurs between 2018 and 2040, and may total $132b at an average of $6b/year, for ship construction alone.

This graph shows how the number of ships built and launched varies over time. If refit costs are 20% building costs, then building costs dominate until about 2040, by which time general admission revenue can begin to cover much of the program’s operating costs.

Phase Two - Leveraging Secondary Production

Earlier, I defined phase one as the era of cargo constraint, and phase two as the era of accelerating returns. As we've seen above, phase two has a different kind of restraint, namely an immigration capacity restraint. By 2045, the critical path for growth is how many people can fit on a Version 2 spaceship, although even without Version 2 a million people are reached only 5 years later, by 2050.

Here, I will wrap up by listing technology concepts that could lift this constraint and permit further high rates of growth into the future.

Higher construction rate of Version 2. Constraints on construction and launch rate are so low that, if adjusted upwards, many thousands of ships could be launched every window. Construction rates could climb into the hundreds per year in a single factory, similar to the 737 today. Ticket revenue could fund this, if a positive margin on launch business was maintained.
Faster ships that can launch multiple times in longer launch windows. This requires better engines and better mass ratios, but eventually there could be cargo and people arriving year round.
Entry of other companies and agencies into the program could achieve 10x, and possibly 100x on rate.

On the flip side, I think it's possible that the minimum viable population requirement will eventually shrink to the point that even small outposts will have the ability to reach full autarky.

Project Timeline

Mars 2020 - Aquifer search probes land, Version 0 ships performing atmospheric tests on Earth.

Mars 2026 - First 10 crew arrive, 3 ships on surface. They scale propellant plant, assemble a lot of base for new arrivals.

Mars 2030 - Middle of explosive growth phase, base population grows to near 1000. Pilots for all primary industries established.

Mars 2035 - First private and 10,000th settler arrives. Mars spaceport hosts dozens of Version 1 ships and the Version 2 prototype, looming over the rest.

Mars 2043 - Ticket prices fall below $100k and the population exceeds 100k. All secondary industries at least in pilot phase. “Mission accomplished.”

Mars 2050 - Population on Mars exceeds a million. Dozens of outposts formed.

Mars 2060 - A web of towns and cities all over Mars, with the first base and by far the largest forming a sort of hub.

Case Study: Iceland

Why Iceland?

Iceland is an apt analogy for a potential Mars city, and a useful way to think about self-sufficiency. Today, Iceland enjoys a high standard of living and is a popular tourist destination famed for its natural beauty. It has a population of about 335,000 and a GDP per capita of about US$60,000. Naturally, it is highly dependent on a range of imports, exporting mainly aluminium and fish^[27].

Further, Iceland is similar to a city on Mars in that it is relatively small, isolated, and cold. It is not a perfect analogy, as it also has breathable air, mainly basaltic geology, and abundant readily accessible geothermal and seafood resources.

It is clear, however, that if Iceland was cut off from international trade, its standard of living would rapidly regress to that of the 19th or 18th century, as first petroleum was exhausted, then parts of the electrical infrastructure suffered failures necessitating parts that could not be made locally.

While many of the goods upon which Iceland depends could be made locally with some forward planning, some parts are truly “magic widgets” that are very difficult to make. The smallest nation that is able to make, for instance, a sufficiently complete set of computer components is probably Japan, with a population of 127 million and a GDP per capita of US$39,000.

In 2018, the world’s largest containership is OOCL Hong Kong^[28], with a cargo capacity of nearly 200,000T. This behemoth can carry, on a single voyage, about as much cargo as I can imagine 50 years of BFRs bringing to Mars.

Here is the thought experiment. Given one large containership of any desired cargo, can one devise a strategy that will forestall Iceland’s post-isolation living standard collapse indefinitely? Can one fit, into a single ship, a complete industrial stack and the tools to run it with a relatively tiny population? It seems like a tough problem, but achieving autarky in Iceland is a much easier task than doing it on Mars.

Population, technology, and environmental hostility

The Icelandic question is a good starting point for a broader discussion of the interplay between population carrying capacity, technology, and environmental hostility.

First, consider the relationship between population and technology level in a given environment, holding hostility constant. For any given level of technology, there is both a maximum carrying capacity (Malthusian limit) and a minimum population required to sustain that technology. In this sense, technology is a disaggregated knowledge base that must be propagated through education even as individual practitioners are born or die. Technologies that increase carrying capacity include germ theory, mechanization of agriculture, and the Haber process for producing nitrate fertilizers, but in practice it is not useful to attempt to be perfectly granular about such a rich constellation of practices.

Provided that the carrying capacity at any level of technology always exceeds the maintenance population, there is a surplus of labor to invent new technology and a stable “band” in which the population can survive without losing or gaining technology. Outside this band, however, the population or technology will fall until stability is achieved. If, for example, Iceland were to suffer a repeat of the 1784 eruption of Laki, environmental fluorine poisoning would increase hostility and reduce carrying capacity, manifesting as a prolonged famine and substantial fall in population.

The asymptotic behaviour of this graph is not well defined, so it has limited utility for predicting the far future of humanity. In particular, it is possible that beyond a certain technology level, the minimum population level stabilizes as computers are doing all the work. On the other hand, it is clear that there is a finite limit on the population capacity of Iceland, the world, or the universe, set by fundamental thermodynamic and volume constraints.

Second, consider the relationship between population and environmental hostility at a constant technology level. This graph shows a markedly different behaviour to the previous figure. As the environmental hostility increases, the carrying capacity decreases, which seems obvious. What is less obvious is that because individual worker efficiency falls in hostile environments, the minimum population required to support a level of technology increases.

On the left part of the graph, the situation is similar to the previous case, with a band of stable population. Moving to the right, the limits converge at the critical hostility level, beyond which there is no stable population size. The early North American colonies Jamestown and Roanoke straddled this point.

As technology level changes, the critical hostility point traces out a trajectory demarcating the fully generalized limits on population and hostile environments.

How do we interpret this in the case of Iceland being artificially isolated from the rest of the world?

Iceland’s imports of advanced technology help to insulate Icelanders from the hostility of their environment. This makes it more livable at a higher standard of living, at the cost of dependency on an extrinsic supply chain. If this technology became unavailable, then Icelandic people would be compelled to operate in a more hostile environment, with the associated labor and health overheads that implies.

This graph shows the full relation between population, hostility, and technology. For any given level of technology there is a family of curves that together define the stable region. As Technology increases, I have suggested that maximum population carrying capacity drops off more gradually with increasing hostility. That is, increased technology lessens the sting of increased hostility. Conversely, the minimum population to support the technology increases more quickly with increased technology, because of the increased complexity of co-dependent industrial processes. In combination, the critical hostility increases with technology, as expected.

This is only a simple, qualitative diagram. A more precise calculation of the system may reveal cusps, stable points, and multiple pockets of relative stability with voids in between. It’s also imprecise to force all of technology onto a single dimension, since environment-specific technologies are more relevant for solving a given problem. Despite the imprecision of this diagram, it does deliver one key insight, which is that management and amelioration of the hostile environment is vitally important for reducing the overall cost and population needed to reach autarky.

It is clear that the vast majority of human labor on Mars should be performed in as controlled an environment as possible. That is, the city is built within a series of enormous, pressurized structures that enable comfortable, safe, efficient shirt-sleeves operations. Even mining operations would be preferentially operated within pressurized tunnels unless automated surface mining machinery reached a sufficiently high level of reliability. Yet establishing a local supply chain for essentially all the chemical elements will require hundreds of labor intensive, hostile mines, and some in remote places. For that reason, the labor requirement with primary resource extraction scales much more aggressively than for secondary manufacturing. Complexifying manufacturing, once large pressure vessels exist, is mostly a matter of direct translation from Earth analogs in a shirt-sleeves environment, and comes with a much less stringent labor requirement.

Is it even possible to build and maintain a habitat on Mars with a sufficiently benign environment for humans to get on with the job? Although the answer seems trivial, there is an overhead for building walls to keep the air in, and this overhead increases as the environment gets more hostile. For instance, it is probably possible to do this on Iceland, or Mars, or the Moon. But a comet? Or isolated city in deep space? What about on the surface of Io, or inside Jupiter, or the bottom of the ocean? While I am sure that a human terrarium on Mars is doable, I am quite sure that no amount of technology would permit humans to build an industrial city inside the sun.

Detailed Icelandic prescription

To return to the Iceland question, what 200,000T of stuff should be obtained to survive a period of indefinite isolation with minimal reduction in standard of living?

Let’s look at trade^[29] to get some ideas about what Iceland will miss the most, and to understand what needs to be duplicated locally. Here’s a treemap of Icelandic imports in 2015^[30], which groups by sector and ranks by total value.

As expected, the major part of Iceland’s imports are connected to its role as one of the world’s largest smelters of aluminium, an energy intensive industrial process that exploits Iceland’s cheap electricity. Iceland exports nearly all its aluminium, so it’s not clear that retaining that industry would be a high priority in the event of total isolation.

The next most salient component is refined petroleum, cars, and other heavy machinery. This component concerns the obtaining and disposal of energy for useful work. Iceland has an abundance of electricity, but gasoline remains the universally preferred way to fuel mobile machines. These machines increase the per-capita work capacity of humans by a factor of 10-100, so it’s clear that maintaining a modern standard of living will require some way to fuel them. This is the single largest challenge for Iceland, and also for a city on Mars. In Iceland’s case, a mixture of high power electrical umbilicals at work sites and battery driven hydraulic power units are probably the best way to retrofit existing machinery and ensure the continuing availability of industrial capacity.

The next largest category are manufactured goods, including parts, electronics, appliances, chemicals, medicines, and so on. A simplified local manufacturing capability for some minimum viable combination of these must be designed. This is a tough problem!

The next largest category are raw structural materials, including metals, plastics, and textiles. This is particularly challenging as Iceland does not have commercial deposits of mineral ores or onshore oil. Limited quantities can be produced from non-standard feedstocks, but it will be labor and cost intensive. This is also a tough problem. While Mars is believed to have ores concentrating all the usual minerals, it does not have a global supply chain so settlers will be limited by what they can access in the immediate vicinity of their base.

The Iceland National Statistics Institute has 4468 categories of traded goods, including live primates, trichloropentafluoropropanes, and nine categories of cranberry juice. A detailed discussion of alternate sources for every last product is beyond the scope of this chapter, but does underscore the utility and diversity of modern trade.

Key Resource Development

A previous chapter outlined heuristics for prioritizing development projects, as well as a rough conceptual ranking. This chapter describes a development plan for a representative example of each product subtype.

Recall that for any given commodity class, the development lifecycle progresses from one-off bespoke through pilot manufacture and eventually, when the preconditions of both local demand and local exponential scaling capacity are met, into mass production. Ideally, each of the following product classes are brought online rapidly and sequentially.

Electricity

In Earth economies, the main energy sources are solar energy through photosynthesis, and combustion, mostly of fossil fuels. On Mars, neither of these energy sources are readily available due to the lack of an oxygen rich atmosphere, coal, or plants. Worse, the Mars city needs much more electricity on a per capita basis, due to widespread automation, heating, artificial light, and electrochemical synthesis.

Taking the South Pole station for comparison, about 130,000 gallons of kerosene supplies 45 people for 300 days over the winter. One substantial use of energy is melting water for drinking! This is about 1GJ/person/day, or 320kWh, which is about 30% more than per capita energy consumption in the USA. Given that the Mars base is preoccupied with building and operating absurdly productive factories that will eventually reproduce the industrial diversity, if not the capacity, of the USA, per capita energy consumption could eventually exceed this baseline by orders of magnitude.

More important, however, is generating fuel and oxidizer to fly rockets back to Earth. If each BFS requires 400T of methane, its energy capacity when fueled is 12TJ. Assuming that fuel production (from water and CO2) and storage is 10% efficient, and operates continuously between launch windows, each BFS has a Mars energy requirement of 1.7MW, or 15GJ/day. This is equivalent to the energy requirement of 15 humans at the south pole.

In the chapter on timeline, I estimated that the first 100 Martians need 100T of cargo each, or most of a BFS. This implies that energy requirements at the Mars base are dominated initially by generating propellant. And as ancillary industries come online they too greatly exceed the energy required merely to operate the life support system.

To pick a number, let's say that Mars base requires 1.5MW per person, initially mostly supplying BFS with propellant and later supporting other industries. In the US, most houses can draw 30A at 110V, which is 3.3kW. At 110V, 1.5MW is 13.6kA, which would require a cable 2” in diameter. Per person! Obviously, this isn't to power their appliances.

There are two clear ways to generate this much power: solar and nuclear. Wind is no good due to low atmospheric density, variable activity, and high mass for imported towers and turbines.

Solar panels on Mars, if angled to the latitude, kept free of dust, and employing the highest efficiency technology, can generate about 1kWh/sqm/sol, or 3.6MJ under clear skies. The per capita area requirement is 36,000sqm, which is about 9 acres. If solar panels and their bus connectors could be made to weigh the same as paper on a per-area basis, this array would weigh only 3T. 3T is an acceptable fraction of the per-capita cargo mass allotment until the city is scaling beyond thousands of people. The batteries to operate at night, however, would weigh maybe 200T per person, so energy storage could instead be performed by running BFS fuel/ox boil off through fuel cells.

A Mars city starting with 100 people would require an area of solar panels equivalent to a square six kilometers on each side. Given Mars’ small size, a human standing in the center of this solar farm would see it stretch to the horizon in every direction.

The best alternative to solar power is nuclear power. Nuclear plants are ideal for providing steady power, and heat, in large quantities. A 150MW reactor for the first 100 people is well within the realms of existing technology, and doesn't require as much land area as solar power. A standardized reactor design could be duplicated in clusters to support the growing base, collocated with propellant generation and storage, and the space port. Opinions differ widely on how mass and labor efficient such a reactor could be made.

Oxygen

Each BFS needs 2000T of oxygen every 26 months, while each human needs about 1kg/day. In other words, the dominant use of unrecoverable oxygen is for propellant, by about a factor of 10,000. In fact, by mass, liquid oxygen probably accounts for >90% of total industrial production on Mars, until warp drive renders chemical propulsion obsolete^[31].

Fortunately, oxygen is super abundant. To first approximation, the rocky planets are a sphere of iron surrounded by a sphere of oxygen, with a few other atoms salted through. Therefore, on Mars, oxygen is available in gas, liquid, and solid form, though almost always chemically bound to other elements.

If the Mars base has been sited near a liquid water aquifer, electrolysis is the easiest way to extract oxygen. Water is over 90% oxygen by weight.

Failing that, Mars’ atmosphere is 95% CO2, which is 73% oxygen by weight. The best method is to take one oxygen and vent CO, which can be done using an electrochemical process and a catalyst^[32]. Compressing ambient gas requires a dust-tolerant first stage pump, since dust filters require more pressure than Mars-ambient to function.

Extracting oxygen from rocks is unnecessarily painful unless as a byproduct of, say, aluminium production.

If each BFS requires 2000T of oxygen, then daily production should be at least 3T/ship. In gaseous form, this is about 60L/s at 500mBar.

If obtained from water, a daily flow rate of 4T, or 3L/minute, is required. An electrolysis system splits water into hydrogen and oxygen which are stored separately.

At Mars’ atmospheric pressure of 7mBar, the system will require one million cubic meters of CO2 per ship per day. If the intake manages a high gas velocity of 20m/s, then the intake must be about 0.5sqm/ship. Gases take up about 1000x more space than liquids at 1 bar.

Once gaseous oxygen has been obtained, it must be compressed, liquefied, and stored in a tank at least as big as the BFS oxygen tank.

Water

Since the start of this century we've come to understand that Mars has lots of water, enough to form a global equivalent ocean depth (GED) of at least 35m. By comparison, the putative paleoshoreline features^[33] and terraforming illustrations on the SpaceX cafeteria wall would require a GED of at least 150m. This ancient ocean might have been blown away by solar wind, or be mixed with dirt and frozen beneath the modern surface. Further, massive aquifers that discharged billions of years ago may have recharged.

In any case, there is enough water on Mars. Water is needed as a source of hydrogen for fuel and oxygen. There are three potential sources for a Mars city.

Water vapor is present in trace quantities in the atmosphere. Extracting enough water through atmospheric processing, however, would be about 100x more volume intensive than the atmospheric oxygen extraction process outlined above. However, like argon and nitrogen, the other main constituents of Mars’ atmosphere, some production as a byproduct is not unwelcome.

Water is present in soil and rocks in varying quantities. Gradual natural processes can even separate ice from dirt, creating (on Earth) seams and diapirs of nearly pure water ice. Such a source could be mined robotically, stockpiled, melted and processed. Propellant production requires about 4T of water per day per BFS, which is fairly trivial if mining a massive drift of pure glacial ice, and extremely tough if it's mixed at, say, 1% purity with rock and sand.

Ideally, however, the Mars city could be located near a liquid geothermally heated aquifer. Scouting probes or even satellites^[34] could locate areas of the surface with unusual thermal signatures and perhaps even signs of water vapor. Defunct hydrothermal vents have been found by rovers, and some volcanic features are very recent on geological time scales, so hydrothermal sources should exist, even if they're hard to find.

Drilling rock to obtain precious liquid is a well understood technology, and the resulting water could be pumped as needed, or stockpiled as ice or liquid water on the surface.

Methane

Spaceships need fuel as well as oxygen. Each spaceship needs about 400T of methane to fly back to Earth. Methane is also used as a fuel and feedstock for Mars industry, but initially the overwhelming demand is for rocket fuel.

On Earth, methane is obtained by drilling, or more rarely as a byproduct of landfill decay. On Mars, there is no similarly and easily obtainable source, so fuel must be synthesized using Sabatier reaction and reverse water-gas shift^[35]. The required inputs are CO2, water, and electricity. In fact, the fuel synthesis process can be combined with the oxygen separation process. Methane is also liquefied and chilled for long term storage.

Brick

In areas where wood is scarce, brick has been a building material for thousands of years. Brick, and masonry more generally, can be used to build large vaults, domes, towers, and walls that can become living space.

If pressurized, arches need to be preloaded with a few meters of rubble to prevent the vault from blowing out. This rubble layer excludes windows but does provide heat insulation and almost total radiation protection.

Bricks are fused using heat, but sand and crushed rocks could also be bound chemically, such as in concrete.

Construction, whether of bricks, concrete, or a combination, will be performed mostly in vacuum and will require mechanization, if not automation. Brick laying robots already exist, and foundation prep work is similarly very automatable. Formless vaults and domes have been employed for thousands of years. The process could initially be supervised locally or remotely by operators in a shirtsleeves environment, but ultimately deploying vaulted tunnels and chambers should be a click, drag, and forget process.

Bulkhead doors, airlocks, and fittings, brought initially from Earth, would be retrofitted into pre-formed spaces in the structure during final fitout.

Plastics

The next product class on the continuum from heavy and easy to make to light and hard to make are petrochemical products, of which plastics are the most important examples.

Plastics are ubiquitous and versatile materials that are derived from the polymerization of repeating subunits. For many plastics, the fundamental unit is derived from ethene, C2H4, which is also a popular welding gas and oil cracking product. On Mars, ethene must be synthesized from methane using heat, pressure, catalysts, and purification, like any other industrial chemical process.

Plastics are useful for generating fibers (textiles, rope), sheets (bags, wraps, windows), and 3D structures. The relative ease of 3D printing plastics in a Mars city that is yet to scale to steel production is attractive as plastics, such as glass fiber reinforced nylon, can be used to produce structural parts for nearly anything.

A robot arm, for instance, could be assembled from Martian plastic “bones” and imported actuators, bearings, sensors, wires, and other specialty components with marginal mass. In this way there is an optimal economic split between locally producing most of the mass while initially avoiding most of the complexity.

While brick and concrete may be readily employed to build robust, windowless living spaces, growth of plants will benefit greatly from natural light, requiring transparent pressure structures. Fluorinated plastics, such as PTFE, are routinely used on Earth to build durable, UV resistant clear structures. On Mars inflatable structures of practically any size could be deployed using plastic sheeting combined with regular tensile anchors to transfer their pressure stress into their foundations. These structures, similar in principle to an inflatable mattress or zorb ball, would need a mechanism to prevent dust accumulation, but are much more mass efficient for enclosing area than masonry. A double walled structure with redundant bulkheads and integral pressure shelters may be necessary to mitigate the consequences of puncture.

Steel

Steel production is analogous to Earth. Concentrated ores are crushed, separated, and processed in a blast furnace. All the required bulk materials (electricity, oxygen, rock crushing robots, high temperature processes) are by this stage already available on Mars. Trace alloying components can be imported where not naturally present in the ore.

Metal 3D printing is a rapidly developing technology, so steel processing will be more versatile than traditional casting, forging, rolling, and welding. It will be necessary to eventually automate as much of the process as possible, which is already becoming the norm on Earth.

Steel is likely to be the first bulk metal available to industry on Mars, and it is extremely versatile. In particular, valves, bulkhead doors, airlocks, bearings, vehicles and other heavy mass can be produced locally, saving cargo. Mars’ lower gravity enables steel structures to be much more spindly, creating the potential for pressurized buildings that stand far above the surface. Steel allows living space to grow vertically, increasing density and areal efficiency.

Food

On Earth, food is easier to make than steel, electricity, and plastics. It literally grows on trees. But Mars is, to quote Mark Watney, a place where nothing grows.

Plants, in general, need many resources which we take for granted on Earth. Water, nutrients, sunlight, oxygen and CO2, a stable temperature, pollination, pest management, and an incredibly complex set of bacteria and fungi in the soil, with which they form a symbiotic ecosystem.

Fortunately on Mars, the natural day length is close enough to Earth’s that artificial lighting is not necessary. If the base is not supply constrained on electricity, artificial lighting over high density automated hydroponics would require less power than spaceship fuel synthesis. But once a terrarium or pressurised greenhouse is artificially created, high intensity farming such as practiced in the Netherlands is business as usual.

Efforts to develop perennial varieties of common crops, planted heterogeneously, would promote healthier soils and reduce the labor required during planting. In general meat production is much less resource efficient than vegetable crops, but some animals can be supported on marginal farming byproducts. Further, if the effort to back up the biosphere doesn't eventually include a really great zoo, it hasn't yet succeeded.

Some food production will occur from the earliest days, in the form of a limited supply of fresh food to increase variety. Over time, more comprehensive farming will be rolled out and the farm development process improved until it is able to keep up with exponential population growth.

Alloys

Loosely speaking, metals that aren't steel include other elements (copper, tin, silver, gold, mercury, tungsten, etc) and thousands of alloys, the most prominent of which is aluminium.

Whereas perhaps four grades of steel are enough to get started, comprehensive industrial use of metals requires the production of dozens of elements. Each of these will require a dedicated robotic mine and processing plant, possibly hundreds of miles from the city. While the electrical, chemical, and metallurgical requirements of this technology are well understood, vertical integration with near total autonomy and exponentially growing scale is a daunting conceptual challenge.

The sheer diversity of materials, processes, and products make the maturation of this industry probably the single most technically challenging aspect of developing autarky. In other words, the required increase in labor and productivity is relatively poorly offset by a reduction in cargo volume from Earth, and is a big source of schedule uncertainty. In my view, this step is the one that requires the most new technology and thus the biggest uncertainty of potential process efficiency improvement.

Electronics

The next industry to stand up is electronic component supply. By this, I mean anything that intermediates electricity and mechanical work or heat, but doesn't contain integral software or photo-etched silicon. This includes magnetic machines such as motors, relays, switches, resistors, superconductors, and all the other stuff that used to be sold in electronics stores bulk bins.

By this point all the precursor materials and automated manufacturing technology are present and maturing in the city. Steadily evolving total system design can drive requirements for standardized parts. Such a set must span the required uses while avoiding both too much near-duplication or any single points of failure. Additionally, part design must be optimized for low labor availability: Autonomous manufacturing, ease of deployment and replacement, and high reliability.

Advanced chemistry

While fuel synthesis, plastics, metallurgy, and other existing industries all exploit chemistry, the likely next step is the capacity for bulk manufacture of arbitrary chemicals, including pharmaceuticals. The great diversity and relatively low mass of most chemicals favors their importation, but autarky requires this capability. Like electronics manufacturing, bootstrapping simpler components or precursors is the natural way to develop arbitrarily sophisticated chemical synthesis.

Semiconductors

The hardest thing to make in bulk on Mars, and the easiest to import, are computers. Today, only a handful of companies can produce the latest technology. While most applications may not require the most powerful GPUs, every industrial process on Mars depends on automation to function, so sustainable self sufficiency requires the ability to make computers.

By the time the Mars city has scaled enough to produce computers, the annual demand could easily be tens of millions of ICs. This is comparable to manufacturing volumes today. The only way the process differs on Mars is that silicon crystals may grow differently in lower gravity!

Humans

Producing humans differs from other sectors covered in this chapter, as the process is well understood and relatively low tech. Despite the relative bulk and spoilability of human cargo, there are two compelling reasons why importation is better than local (re)production.

The first is the externality of education and training. Maximizing per capita labor productivity benefits from the importation of fully trained professionals.

The second is a question of rate. At full tilt, humans can double the population about every 20 years. Even if Mars women were inclined to carry triplets for their entire fertile lives, the population could not be grown quickly enough to reach autarky in less than a few centuries. There is a nearly limitless supply of qualified humans on Earth. Until autarky is achieved and the growth rate can back off, most Martians will be new immigrants.

Key Technology Development

Autarky, or industrial self sufficiency, for any Mars city requires both a large population and large increases in per capita process efficiency. While both these goals are enabled by new technology, it is the latter which demands the most innovation in a wide range of areas.

Forecasting technology five years ahead is a substantial challenge, so fifty years is practically impossible. Nevertheless, Mars autarky will require revolutionary breakthroughs in the following technology areas.

Automation and Remote Operation

One obvious way to improve the per capita productivity is to offload some of the labor burden to machines of various kinds. This is an area of active research on Earth, but operating on Mars will introduce a new set of requirements.

Remote operation implies some degree of spatial and temporal distance between the work and the operator. At one end of the spectrum, a human spins a wrench. Proceeding along this spectrum, an electric drill can drive a socket, intermediating electric mechanical power between the operator and the work. Just this single step greatly increases the productivity of a single human operator.

That drill can be fitted to a robot arm controlled remotely, or, if performing repetitive work, trained once and then operated on playback. On Mars, the operator could be in the next room, in another part of the base, or even on Earth, provided the robots were autonomous enough to not need real time feedback. Even further along the spectrum, humans may not even need to specify the operation - AI can place the robot arm on the manifest, schedule its installation, and infer its operation from context. Or even design the manufacturing process completely, and manage its maintenance.

Reliability

All machines eventually break down. In other chapters I’ve discussed the engineering trade around reliability and replacement. New technology is needed to produce a variety of machines capable of operating with flawless reliability for the full duration of their design lifetime. Ideally, they would be designed on both a unit level and system level to be readily replaceable, and to minimize the labor burden of any required maintenance.

This set of design requirements already exists in some industries, such as fast food kitchens, where cooks must produce a standardized product at consistently high quality with a minimum of hassle. McDonalds, in particular, became famous for endlessly optimizing the design of kitchens to minimize delays between process steps - many workers can do their whole job without having to take a step, which takes time.

Security and Software Assurance

Mars industrial robots will depend on software to achieve the required level of efficiency, versatility, and controllability. It is simply not possible to build a nearly completely automated industrial stack with punch cards and a complex set of cams! This means that nearly every widget will be vulnerable to unintended side effects and possibly malicious hacking.

Not only will industrializing Mars require much more advanced AI algorithms to operate factories, and sophisticated control software to run robots, and tools to develop and evolve this software, it will also require a much more robust approach to software assurance to prevent bugs from having catastrophic consequences.

Scalable hardware processes

Reducing the number of humans needed in industrial processes is not simply a matter of increasing output by increasing the number of machines. If each machine operator doubles the number of machines ever 26 months, then by the time the desired scale is reached, each operator will be operating a thousands times more machines than they began with.

Exponential improvement in productivity has to be designed in. Hardware processes need to be developed that prioritize the ability to scale faster in future. Any change in the near future needs to be assessed on the longest possible time scale.

For processes that have a minimum footprint, there’s only so much headroom on increasing the speed of machines. Future productivity increases will require the development of new factories in newer, larger habs. These factories will have to be built and operated with steadily decreasing levels of human involvement.

The usual paradigm is that as processes mature and continue to scale, human involvement has to steadily retreat up the process chain to ever higher levels of abstraction. As an example, consider a factory that assembles a basic part such as bearings. In the first 26 months of operation, it produces 1000 bearings a day with humans operating rolling machines, ball mills, and the assembly process. In the second 26 months, output doubles to 2000 bearings a day with the same human involvement. The operators, in concert with engineering help on Earth, automate the most labor-intensive parts of the process, eventually assembling stock that arrives on automated trucks with robotic labor. Humans merely assemble the machines that produce the bearings. Two windows later, the rate has increased to 8000 bearings a day. At this point, the assembly and commissioning of bearing producing and assembly tools is also managed by a larger, slower, and more versatile machine - which is assembled and commissioned by humans. In this way, all humans involved in secondary manufacturing on Mars eventually converge to a common job description - assembler of the single most versatile, complex robotic assembly machine in operation. Division and specialization of labor is eventually entirely mechanized.

This approach to metafactory development that has no target rate, merely a target growth rate, is very unusual in conventional terrestrial projects.

Recycling

Some resources developed on Mars are intended to be used destructively and irrecoverably. Most prominent among these is rocket fuel, which is literally exhausted back into the atmosphere or space. For this reason, primary production of methane and oxygen is an enormous consumer of electricity and industrial effort.

Other resources, while energetically expensive to obtain, are not consumed in use and are readily recovered through recycling. Lead and aluminium are routinely recycled on Earth, reducing the demand for primary production. In the same way, recycling wherever possible is important on Mars.

In addition to raw materials, particularly reduced rare metals, the rapid scaling of industrial capacity provides an opportunity for time-delayed recycling of broken and discarded parts.

As an example, consider a heat exchanger. In the early days of the city, there may not be adequate labor resources to do more than throw it in a pile with other broken things after routine replacement. Later, a broken heat exchanger could be cannibalized for spare parts for other machines, or melted down as a source of raw materials. Still later, when heat exchangers are being produced locally, sufficient commonality of design may incentivize remanufacturing of the original unit, particularly if it serves obsolete legacy systems that are not readily retrofitted with more modern components.

On the other hand, exponential growth of capabilities and demand mean that the relative handful of early discards may not be worth the effort to restore original functionality. Why restore four different obsolete heat exchangers if you normally operate a line that produces a thousand better ones every year?

Piloting on Earth

New technology can’t be deployed on Mars directly from a napkin sketch. Given the difficulty of transportation and criticality for the continuing success of the project, new technology will have to be piloted on Earth.

There is a standard procedure for the development of new technology and progressing through Technology Readiness Levels, or TRLs. The challenge will be to get everything ready and in time, while leaving the door open to future improved parts being adopted late in the design cycle.

Ideally, numerous new companies will materialize which specialize in providing a particular process or service and which simplify the initial centralized planning. Care is required, however, to ensure that all the relevant incentives are appropriately aligned.

Human Resource Management

The final consideration, on a technology level, for effective and rapid industrialization, is the continued deployment of technology that increases productivity.

Organization

According to Edward W. Merrow^[36], the probability of success for any industrial megaproject, defined as having a capitalization of over a billion dollars, is about 50%. Attempting to integrate new technology, adhere to a tight schedule at all costs, or operate in a remote/hostile environment reduces the probability of success to about 10%.

This isn’t good enough. There needs to be a reasonable chance of total success, which demands next level organizational management. I think it is fair to say that developing and improving an organizational system which can manage the staggered remote deployment over several decades of insanely advanced technology supporting tens of thousands of carefully selected specialists in an absurdly hostile environment and a race against time building a machine that is widely considered impossible even on Earth, is a daunting technological challenge.

Crew Selection

Traditional projects have selected the best people for the job, wanting to minimize the risk that when push comes to shove, the team is unable to deliver a product on a fixed, near term timeline. The Mars project is a bit different. Given that the productivity per capita is continually increasing, the majority of a given person’s contribution will occur in the last two years of their deployment. The important thing, therefore, is to select crew who will grow into the roles that emerge around them and that work effectively in rapidly scaling teams, as well as being able to follow IKEA instructions and airlock checklists.

Labor Assist

One additional aspect of crew support is the adaptation of an emerging technology, the helmet-mounted head up display^[37]. Such a unit can record and transmit the process for later reference and analysis on Earth. It can query a voice-activated database. It can show instructions and call other nearby experts on the fly. In addition to reducing time spent on training and confusion, it can also be used to record process memos for other workers on similar projects, which are tagged with semantic data. Eg “When operating valve 93, it can stick at the 5 o’clock position, and be unstuck by operating valve 92 in sequence.” More powerfully, asynchronous requests can be handled offline by a support team on Earth. For example, a worker could invent and loosely describe a tool to help with an ongoing operation. A team on Earth would rapidly design and test the tool, then transmit its design to 3D printers on Mars, where it would be ready for use on the following shift.

Reusable 26 Month Development Plan

This book presents a model of Mars development that enables exponential growth in population, industrial capacity, and self sufficiency. This permits a reusable 26 month development plan, because it’s a discrete scaling time translation symmetry.

For example, although the population doubles every window, the ratio of the required living space construction to the number of construction workers remains the same. Each reusable plan is a substantially similar schedule lasting about 670 Martian sols. This schedule guides resource balancing between different operations that include unloading, storage, deployment, construction, installation, testing, and qualification.

There are substantial historical parallels for nations undergoing periods of centrally planned, rapid industrialization, and with varying levels of success. In a later chapter I will break down aspects of successful central planning, in this chapter I describe one potential plan.

Schedule construction approach

The schedule developed in this chapter is not intended to be the final word on Mars industrialization. Like the rest of this book, my goal is to list and rank the most important questions and to sketch a constructive method for finding answers.

I take the standard approach for determining schedule, beginning with milestones, sketching the critical path, and finally labor balancing.

Milestones

Milestones are boundary conditions. Here, one sets the beginning and end state of any given subsystem, and ranks the relative importance of partial completeness. For example, over the course of a single window, the enclosed and pressurized living space needs to double in volume. Any given industry needs to reach the next step on its path to full production.

Critical Path

The critical path traces the most optimal route from initial state to final state, ranked by the most difficult process. This stage is ideal for identifying, disentangling, and protecting various gating interdependences. For example, orbital mechanics imposes a stiff penalty for failing to turn around the Mars ship in less than 28 days. Similarly, failing to complete the foundation of a new pressurized vault prevents its construction, completion, and fitting out.

Labor Balancing

At any job site, and in particular at Mars, there is a fundamental limit on available labor. The efficacy of any particular person on a job varies greatly depending on the availability of fundamental resources such as time, tools, self care, management, parts, team. Mechanical labor, such as a brick laying robot, is similarly limited and gated by numerous ancillary resources. Project success depends on closely matching labor demand, supply, and wellbeing in every subsection. For instance, maximizing bulk infrastructure will require time investment in the building and maintenance of automated construction robotics, which will then need to be continuously operated for maximum utility.

Major Project Subsections

This section is not intended as a prototype org chart. Instead, I am attempting to break down the various subsections of the industrialization project in the most intuitive way. For that reason, I start with labor utilization considerations.

Wellbeing Operations

Wellbeing operations include every process designed to maximize the labor output. This includes life support, food storage and prep, crew shifts, crew task rotation and recreation, and scheduling critical equipment maintenance cycles. Like nearly every other process, wellbeing operations are continuous.

As outlined in Chapter 5, maximizing labor efficiency in this context is not the same as maximal human exploitation. Rather, considering the inherent difficulty in moving humans to and from Mars, and the extremely long term nature of the project, labor efficiency is a reflection of investment in worker health and happiness.

One critical consideration for determining labor efficiency is the danger and comfort of the environment. For the purposes of this chapter, I break available work spaces on Mars down into four different categories:

Surface. Exposure to vacuum, cold, radiation, dust.
Pressurized area, but not fault tolerant. For example, the interior of a single wall pressurized inflatable volume enables human operations without a pressure suit, but not without precautions.
Pressurized area, industrial zone. Negligible risk of depressurization, but high voltages, temperatures, noise, heavy machinery, and risk of fire are present. Could be subdivided further by level and nature of risk.
Pressurized area, residential area. Designed and intended to be an entirely benign environment.

Roughly speaking, each subsequent category is an order of magnitude easier to work in than the last. For this reason, the first priority of pressurized infrastructure is to ensure that there is always adequate pressurized volume to get downstream processes done.

Structures

The purpose of structures is to create large volumes for industry to be built in a shirtsleeves environment. In this chapter, I am agnostic about structure design and construction method, except that I assume that some fraction will utilize masonry or metal for primary structure (dark sky) and some will use layers of pressurized fluorinated plastic (light sky).

Structures are composed of several interdependent parts, each of which is deployed sequentially by specialized machines and workers in a rolling, campaign style designed to maximize productivity increases. These parts include:

Foundation preparation and construction.
Pressure structure construction, including bulkheads and pass throughs.
Pre-placement of critical utilities infrastructure, including pipes, electricity, and monitoring systems.
Structures for habitation or industrial installation. Because everything is walled, many components or bulk dunnage are best prepositioned inside before being walled in, particularly if they incorporate components bigger than the available doors or passageways.

Just-in-time manufacturing is an approach that minimizes the volume, and thus expense of parts and unsold inventory. On Mars’ rapidly growing and changing industrial base, inventory in transit is devalued exponentially, further favoring just-in-time. That said, adequate pressurizable space is so vitally important for leveraging labor productivity that structures, like other fundamental infrastructure, must be “built ahead” to provide room to grow. In practice, just-in-time manufacturing will produce the most value when employed for the final integration of various products just prior to deployment.

Other Fundamental Infrastructure

Structures are the first and probably most difficult primary industry to get right, since they literally form the medium by which humans interact with, and are protected from, the deadly Martian environment. As other industries are developed and come online, they too will eventually join the stack of critical, fundamental infrastructure upon which future advances are based. Deploying and maintaining this infrastructure will take several windows, but will eventually include:

Electricity production, fuel production. Electricity and fuel are more fundamental than structures, but have a smaller interaction cross section with the outside world.
Logistics, cargo warehousing and transport, port operations. Uniquely among Mars projects, port operations will be extremely busy for one month out of 26. Labor balancing port operations represents a particular challenge.
Mining, metal production.
Farming and other operations within large scale pressurized plastic greenhouses.

Progressive Industrial Rollout

In a previous chapter, I described the sequence of key resource development, as each of many dozens of industries is developed on Mars. Although different industries will develop differently, all begin with a pilot industry test and mature with the successful deployment of steadily increasing levels of automation abstraction and exponential productivity. In any given window, up to 10 industries could be at various points along this path.

Table of Interdependencies

Subsystem	Milestones after 26 months	Critical dependencies	Labor balance
Cargo ops	Cargo handling capacity doubles	Short turnaround time for BFS, fuel, warehousing	Cargo arrival coincides with more people
Logistics	Capacity meets demand	Warehousing, airlocks, passthroughs	Warehouse ops and fulfillment can be automated
Well being	Capacity to serve new arrivals	Decouple services from living space, whose growth will lag	Human-human operations are hard to scale
Maintenance	Technical debt stable	Simplifying process dependencies	Human actions limited to long term reliability assurance
Structures	Available volume at least doubles	Construction rate and scale must avoid blockers	Automation of vacuum work must scale smoothly, final assembly is more fungible
Pilot industries	Successful demonstration or proof of concept	Adequate space, power, trained personnel	Labor intensive, will require cross training
Scaling industries	Attainment of design quality	Sufficient tolerances and material pipeline	Full production labor intensive, automating slowest parts first
Ramping industries	Substantially closer to ultimate production rate and growth rate target	Adequate provenance of technology and supporting parts	Automation of material handling and management
Mature industries	Output doubles while labor steady, through combination of duplication and machine velocity increases	Requires sophisticated machines, spare parts, feedstock, careful planning	Continued scaling requires layers of automated abstraction with exponentially suppressed human involvement at the lowest levels

26 Month Gantt Chart

A Gantt chart is a traditional method of depicting project timelines with interdependencies and resource loading, and to keep track of progress. This Gantt chart is intended to be reusable in any given 668 sol period beginning with the landing of the next contingent of settlers on Mars, and ending just before the next landing.

There are three broad types of work that repeat every 668 days. The first is maintenance-type operations, that must freely scale to keep up with demand. The second are infrastructural, which employ a rolling production system to keep large scale machinery continuously employed and a steady supply of finished product, multiple times per window. The third is industrial scaling, which occurs over perhaps four consecutive windows and thus has multiple industries at different levels of maturity developing in parallel.

The Gantt chart can be accessed here: https://docs.google.com/spreadsheets/d/18EGmAp_a80Mb6U-lbLJ2I-zfZSavY42dO66hAlBoO70/edit?usp=sharing

Economic Planning

By this part of the book, I think most aspects of the industrialization problem have been mentioned if not fully scoped out. One important question remains, which is designing, maintaining and evolving an organizational structure that gives the project the best chances for success.

This chapter is not particularly prescriptive, as I am definitely not an expert on management of megaprojects^[38]. I will begin with some observations about project risk profiles, then discuss broader methodology.

For projects on Earth, a good measure of scale is total cost. Cost is a measure of labor, materials, risk, schedule, and the local regulatory framework. In “Industrial Megaprojects: Concepts, Strategies, and Practices for Success”^[39] the author Edward Merrow describes how scale cuts across all other traditional indicators of project health in determining initial odds of success. In particular, as a project scales beyond a billion dollars, project coordination, schedule adherence, and success all drop dramatically.

One key reason for low success rates is that such projects are so big, and so rare, that teams usually reach retirement age before getting a second chance, which breaks traditional feedback mechanisms for successful and failure-prone management methods. A particularly salient example is Boeing’s development of the 777, which successfully integrated 280-odd teams and finished on budget and schedule, and Boeing’s development of the 787, which suffered nearly every project problem in the book, and some new ones. Boeing’s inability to rewind the tape and maintain its mature project development strategies, despite every incentive to do so, is a cautionary tale about the difficulty of maintaining this rather intangible yet vital organizational skill.

In addition to overwhelming scale, cost, and schedule, the Mars industrialization project also incorporates many other problematic elements, including limited labor, hostile environment, remote operation, cross-cultural/language coordination, deployment of new technologies, questionable infrastructure, and rapidly shifting scale. These factors, which would doom any conventional project, require a next-level approach to retain any hope of eventual success.

There are many different approaches to organizing groups of people to most effectively execute a project, and different approaches work better at different scales. In particular, as projects scale up from a few people to hundreds of thousands or more, project complexity and interface dimensionality consume more and more bandwidth. This is axiomatically true, so optimal and evolving partitioning of the problem is necessary to place interfaces in the best places, and to operate them effectively.

Consider the following list of teams of various scales and naturally effective approaches for organization.

Rock climbing team. Task oriented, multifunctional, autonomous, self-organizing.
Restaurant. Clear task and goals. Single leader. Different specializations.
Medium sized company. Executive team, department leaders form two levels of management. Department leaders primarily responsible for interfacing, but may require intervention from executive team because this structure is very prone to damage.
Larger company. Multiple teams developing a spectrum of different products, and each team has a dedicated interfacing professional who manages interactions with other teams and external partners. In this system, interfacing responsibility has been pushed out from a central authority (which eventually dies of congestion) to satellite orgs. Has the ability to rapidly and asynchronously make decisions.
Consortium. Once multiple companies with different structures, cultures, and org structures begin to coordinate, interfacing responsibility often moves from technical communication professionals to sales, marketing and legal. Ironically, this step is designed to reduce legal risk and trade secret exposure but often ends up injecting a lot of technical risk and destroying value.
Economy as a whole. Economies with centralized control have not proven to be as efficient, and thus competitive, as decentralized capitalism. This model moves away from legally bound consortia to a producer-consumer model, largely mediated by market mechanisms. These mechanisms include advertising, marketing, and sales industries, as well as a strong common regulatory framework which attempts to reduce volatility and avoid known market failures. The attractive feature of this approach is that the cost of organizational overhead is never born entirely by one central entity, even though, in practice, many companies duplicate a lot of interfacing work.

At different scales and different times, aspects of the Mars industrialization project will include these different styles of organization.

The free market is a powerful organizational method, because it pays for efficiency via the profit motive. The problem with the free market is that, unless extremely lucky or careful, the profit motive can also incentivize counterproductive behaviours, including greed, corruption, predatory dumping, slavery, rent seeking, monopolization, and anti-competitive behaviour. The natural mechanism for punishing failure within the free market, or some naturally occuring ecology, is starvation and death. Yet poor organizational health incurs other, hidden costs that hurt the overall project. One of the reasons that Silicon Valley has become such a powerful economic ecosystem is that it reduced the cost of failure and provided a mechanism to preserve and propagate hard-earned wisdom.

How should different industries, teams, and organizations on Mars interface to maximize the health of the project as a whole? How can the power of market economics be exploited while greatly reducing the risk of total economic collapse or catastrophic failure of critical infrastructure?

I don’t think it’s particularly useful to speculate in great detail about hypothetical economic systems or to fantasize about the implementation of some imaginary pet scheme, such as crypto-anarcho-techno-blockchain-libertarianism. There are more productive things to worry about!

In broadest terms, a “winner takes all” or “winner takes most” system incentivizes greed and zero-sum deal making. Alternatively, a group performance-based incentive doesn’t adequately punish individual laziness or free-riding. And any sufficiently sophisticated performance measurement system incentivizes the expenditure of precious resources to game that system rather than just do the right thing. Therefore, I think it’s probably ideal to use a simple monetary system with a simple tax structure to allow both measured individual reward and provide for the common good. The ideal implementation will evolve toward greater complexity with the growth of the overall industrial ecosystem on Mars.

Urban Planning

This chapter builds on a blog^[40] on the topic inspired by some slides presented by SpaceX at IAC2017, which showed a SimCity base growing on Mars. If Mars cities won't look like this, what will they look like? This chapter is preoccupied with urban planning principles, or the practice of systems integration at the city level.

The first city on Mars is oriented toward rapidly developing autarky, to minimize the number of people and period of time when the effort is dependent on shipments from Earth. This fact differs a lot from a mostly static Antarctic station outpost, and this determines how it must be planned. No one knows for sure how many people are needed for autarky, but is likely at least a million and will thus require decades of blistering growth. The primary role of fixed infrastructure on Mars then, besides keeping death out, is enabling growth.

In the chapter on pathways to autarky, I attempt to estimate growth rates using ship construction and utilization estimates, combined with a population/self sufficiency relation. The Earth-Mars launch window occurs every 26 months, and initial population growth targets are a factor of 4 per window, later dropping to 2 depending on ship production, capacity, and reuse.

No city in history has needed to or managed to sustain growth this fast. On Mars, the primary task is building more city for impending arrivals, and the primary constraint is labor availability. To maximize production efficiency, construction will need to use mechanization, automation, and wherever possible, a shirtsleeves environment.

All this is fairly obvious. Can we now draw a map? Not really. I don't know what a self-sustaining Mars city looks like and I probably will not live long enough to find out. Indeed, attempting to learn from experience that doesn’t yet exist is a pointless endeavor. But I can wave my hands a bit about the first decade.

In addition to enabling its own maximal growth, the Mars city will perform every other kind of function from life support, transport, recycling, and entertainment to privacy, education, mining, manufacturing, communication, and emergency management. Some of these functions will be distributed, others more centralized. To maximize the utility of limited living space, for instance, compact apartment geometries can be imported from Earth, while landing and launch operations, and other especially hazardous activities, will have to be separated from more vulnerable or less defensible areas. Practically speaking, all functions span a continuum from local to centralized. Somewhere in the middle of this continuum is a point of mandatory separation, and it is here around which individual pressure vessels, habs, vaults, tunnels, domes, and vehicles will be divided from each other.

All of the more local functions (health, education, libraries, sport, food, recreation, spirituality, music, common space, food distribution, non-transformative recycling, life support, temperature control, atmospheric processing, grid stabilization, communications, data storage, residential, and non-hazardous industrial live-work spaces) are ideally collocated. Since these all take place in a climate controlled pressure vessel, each pod is self-contained and resilient enough to withstand substantial extrinsic challenges, while nominally sharing capacity with adjacent systems. Vacuum operations are required only for initial construction and exterior maintenance. Everything else is done in shirtsleeves at minimal marginal labor cost.

Opinions vary on ideal structure design and material, and methods will no doubt continue to evolve drastically during deployment. My personal preference is for hangar-like structures. A cylindrical roof spreads pressure, requires no internal support, can be shielded with dirt, and unlike spherical domes, has simple curvature and decouples ideal volume from geotechnical concerns in the foundation. With few or no windows, the interior could be somewhere between a modern submarine or a Vegas casino. Both structures are quite comfortable despite uninhabitable exterior environments. I envision structures ranging in size from Quonset huts to Hangar One at Moffett field and larger.

Arched structures of various scales (left to right, top to bottom). Quonset huts, Project Iceworm, South Pole logistics archways, some building in Hawthorne, Hangar One, Atlantic City Convention Hall.

They may be connected by sealable bulkhead doors, while the roof can support solar panels or farms, especially the equatorward face of east-west oriented pods. They also have good volume to material/labor ratios. Building materials can range from prefab panels to locally produced concrete or brick. Brick vaults can be assembled robotically without formwork using a variety of techniques. Brick and concrete structures are compressional so need preloading before pressurization with several meters of dirt. Numerous other materials and methods are possible, including inflatables.

A growing Mars base, then, could be a densely packed crosscutting network of arched pods, with outskirts being built out at ever more ambitious scale. Manufacturing, chemical work and other non-residential activities can be confined to dedicated pods, which can be repurposed (loft conversion!) over time as demand shifts.

Primary demand for water, (nuclear or solar) power, fuel synthesis and storage is associated with launch, so it makes sense to collocate much of this capacity at the spaceport outside the city. Pads with retractable hoses and robot arms can handle ship surface operations. If the city outgrows the spaceport, construction of new pads and pipes is much less labor intensive than demolishing and/or rebuilding pressurized pods. Spaceports will ideally be located 5-10km north and/or south of the city to keep east-west approach/departure paths clear. On Mars this is well over the horizon!

Manhattan is one of the highest density cities on Earth, with about 25,000 people per square kilometer. The island as a whole, with an area of about 60 sqkm, houses 1.6 million, and is thus a reasonably good geographical analogue for a future industrially self-sufficient Mars city. This area would include industrial and commercial spaces, with purely residential parts occupying perhaps 15% by surface area assuming 500 sqft of living space per person and 5 levels of buildings. Of the 500 sqft, most would be individual space, while some would count towards the more generous common living spaces favored in all high density living situations.

The industrial spaces use the bulk of enclosed volume because, despite a relatively low density of humans, the total volume of productivity required through automation is comparable to a medium sized country.

A larger use still of enclosed volume is farmland, though farming isn’t an immediate priority. Farming on Earth requires between 1000 and 10,000 square meters per person. Farming on Mars will have more in common with high-intensity agriculture as practiced in the Netherlands than open range cattle grazing, but would still require an area of between 1000 and 10,000 square kilometers to feed everyone. For comparison, while the city occupied an area comparable to Manhattan, the farms would consume an area comparable to Long Island.

It makes sense to operate at the same pressure as the base, rovers, and suits (perhaps 340 mbar) but with enriched CO2 for plant growth, and every other trick worked out by decades of dedicated research that hasn’t happened yet! The primary difference between farming and habitation structures is that farms need transparent roofs, though ideally still with a layer of water, ice, or glass to mitigate radiation. I like the idea of transparent ETFE^[41] inflatables sealed to the ground at the periphery and anchored with cables at regular intervals to spread the pressure load into the ground. This structure is the tensile analogue of traditional compressional masonry vaults, and would look a bit like this.

There's no reason why habitation pods couldn't be interspersed between greenhouses. Adequate single fault tolerance, as well as thermal and radiation protection, would be required for humans to normally live under an inflatable sky. Intermittent work is possible through use of backup air shelters or sealable vacuum escape pods.

One final consideration is scaling and congestion. Pod bulkhead doors are natural choke points. While clever neighbourhood designs will keep the base walkable for most activities (food, hygiene, schooling, training, recreation), movement of large equipment or lots of people may require progressively larger thoroughfares. This is a great problem to have, since that many people on Mars implies many other problems have already been solved! I think careful use of tunnel boring machines for subgrade roads and repurposing of legacy structures will prevent problematic congestion.

I have, no doubt, overlooked many details and obvious issues in this first pass at Mars urban planning principles. I’m interested in developing sensible system design axioms from which any given city plan can be readily derived. I don’t know of much other work done on urban planning with such an aggressive focus on growth, so I hope this chapter has helped to facilitate a better understanding of the issue.

Earth

This chapter is intended to capture aspects of the Mars industrialization project which are more Earth-focused, such as policy, investment, and risk management. Rather than conduct a futile search for inapplicable historical analogies, I will instead discuss the known issues as they present themselves today, and how they may reasonably evolve in the next century.

Law

I am not an expert on law or space law, but there are a few particularly interesting legal aspects of Mars autarky. Broadly speaking, they come down to jurisdiction. In “The Martian,” author Andy Weir speculates that the law of the sea will apply on Mars^[42], at least before a local governance structure is set up.

This is an interesting problem because sooner or later there will be critical infrastructure and other sources of embodied wealth, which on Earth requires legal machinery to provision for ownership. Other writers have speculated about legal paradigms which avoid aspects of property law, and anything is possible particularly in an environment where enforcement could exist anywhere between anarchy and totalitarianism.

This discussion is more than academic. Conway’s Law states that “organizations which design systems … are constrained to produce designs which are copies of the communication structures of these organizations.”^[43] A nascent life support production team that opts for centralized control and distribution may be inadvertently laying foundations for absolutist totalitarian control in the future city. Any vacuum-embedded city is so vulnerable to destruction that legal structures favoring the health of the whole at the expense of the dissident individual are quite likely to emerge.

Multi-party cooperation

Although SpaceX is the only major organization actively working on Mars industrialization, it is quite clear that its scope and scale is so huge that success may require participation from many companies and ultimately nations. Whether through tax incentives or fear of irrelevance, it is my hope that all possible contributors jump in to help with this grand project.

Smooth coordination will probably require a governing agency or standards organization, together with intelligent subdivision of the overall program into minimally intersecting parts. Technology and space policy would need to evolve to enable the requisite tech transfer and promote mutually beneficial cooperation.

For example, Caterpillar and Komatsu could cooperate on surface mobility and mining tech, while AECOM could build hab management systems. SpaceX would focus on transportation with help from Boeing and Roscosmos, and Amazon could help coordinate Earth-side supply chain logistics.

Money

I am not currently aware of any human space industrialization schemes that could make money. Indeed, it is clear that human space exploration is very expensive and the time frame over which this project must be executed spans longer than any non-government investor could be interested. At best, many many decades are needed for a modest financial ROI, and requiring only a few dozen miracles.

That said, governments and organizations routinely spend huge sums of money employing people with a range of technical skills to work on projects of dubious economic merit, because cost and profit are not the only important considerations. In particular, all advanced nations^[44] have space programs for prestige and strategic investment in domestic aerospace expertise.

Today, the US government spends about a trillion dollars every year on defense, much of which is spent on weapons development. A Mars industrialization program would employ the same people under the same contractors (Lockheed, Boeing, and friends) in the same congressional districts and building much the same sorts of things. At perhaps $10b a year, a Mars industrialization program wouldn’t have to cut substantially into the regular defense programs either. In comparison to the F-35 development program, it would barely dent the bottom line.

There is one other good reason for technical agencies and companies to invest in moonshot programs. Organizations that routinely tackle big challenges gain an experienced workforce and new technology. For example, despite there being no market requirement, SpaceX’s ambitious reusable rocket program has delivered a product markedly better than earlier versions and has cemented its dominance in the sector.

Risk

While the Mars industrialization program will plan for cargo losses, total cessation of resupply flights for multiple windows will be disastrous. And while cargo losses are mostly an engineering problem, any number of problems on Earth could also disrupt regular shipments needed to continue the industrialization project. I don’t think any of these challenges are insurmountable, but they may require a certain degree of focus to overcome.

I don’t think this is the place to dwell here on broader political or environmental challenges, though they continue to threaten Earth’s advanced industrial civilization. It is, however, unwise to expect that there will be no problems at all over the next century, so plans should be made to safeguard the most vulnerable and dependent outposts, including half-built space cities, against foreseeable and preventable supply chain interruptions.

What Don’t We Know?

In the introduction, I argued that it is impossible to learn the lessons of experience that hasn’t yet happened. However, despite our overwhelming ignorance about the challenges of Mars industrialization, it was still worthwhile to identify the best questions and to describe potential solution finding algorithms.

I hope you, the reader, are inclined to agree that this is a valuable exercise. I started by defining autarky, then creating a logical chain of actions which lead from some initial landing to gradual insourcing of all industrial processes. I discussed why labor availability was the primary constraint, and how it relates to other variables. I explained how maturing industries and capabilities projected to double every 26 months with constant labor cost would have to create steadily growing layers of abstraction to intermediate human hands and finished products. I elaborated flight rate constraints on growth, listed technologies and resources in order of local exploitation, then finished with an exploration of schedule and organizational issues.

In my view, the single most important unsolved problem is transportation between Earth and Mars. For operations on Mars, the problems are many and difficult, but have closer analogs to familiar problems on Earth.

Today, the scale of the challenge is daunting. The process of writing this book has served to reinforce my perception of the sheer difficulty of the project. An encouraging sign of progress would be the reindustrialization of smaller countries due to development of compactified industrial processes. Any sign that the critical population necessary to attain autarky had reversed its centuries of growth would indicate that manufacturing technology has matured to the point where a Mars city of perhaps only a million people could credibly maintain adequate labor specialization and achieve industrial closure.

As this project matures and gets underway, further challenges will doubtless occur. If this book has any impact or enduring relevance, it will be in its approach to difficult, poorly defined problems. Begin by asking the next question.

[1] Here are some conceptual sketches: https://photos.app.goo.gl/TUKzHkV9zC9xvqaX8

[2] But let’s not call it “colonization.” “Settlement,” “industrialization,” or “base-building” are much better.

[3] https://waitbutwhy.com/2015/08/how-and-why-spacex-will-colonize-mars.html

[4] Available for comments and suggestions at https://docs.google.com/document/d/1qfztXXRWr1km6U4H44dSpyG7I-Xspd4GkBQmKVjKmbM/edit?usp=sharing and on Amazon https://www.amazon.com/dp/B01MR6BOP0/.

[5] http://www.spacex.com/mars

[6] https://www.amazon.com/Industrial-Megaprojects-Concepts-Strategies-Practices/dp/047093882X

[7] https://www.amazon.com/Mission-Mars-Vision-Space-Exploration/dp/1426210175

[8] http://www.marssociety.org/

[9] https://en.wikipedia.org/wiki/The_Case_for_Mars

[10] https://www.amazon.com/How-Live-Mars-Guidebook-Surviving/dp/0307407187

[11] https://en.wikipedia.org/wiki/Packing_for_Mars

[12] http://www.thespacereview.com/article/1448/1 James Oberg’s website (http://www.jamesoberg.com/policy.html) is a wealth of information on every space topic.

[13] https://en.wikipedia.org/wiki/The_High_Frontier:_Human_Colonies_in_Space

[14] https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19830007077.pdf

[15] http://www.spudislunarresources.com/

[16] While beyond the scope of this book, I outline the marginal nature of lunar resource economics in this blog: http://caseyexaustralia.blogspot.com/2017/04/does-lunar-resource-exploitation-make.html.

[17] https://en.wikipedia.org/wiki/Surviving_Mars

[18] https://simoc.space/

[19] Frailty is the analogous consequence of repair capacity shortfall in biological systems.

[20] https://en.wikipedia.org/wiki/Self-replicating_machine

[21] Orbital dynamics restrict launches to Mars to a 2 month window that occurs every 26 months.

[22] This design requirement is borne out by experience in the ISS, where astronauts must spend a disproportionate fraction of their time cleaning air filters and maintaining other systems that were designed to never break, and thus never be disassembled, on orbit.

[23] http://caseyexaustralia.blogspot.com/2017/09/estimating-mars-settlement-rates.html

[24] Available for comments and suggestions at https://docs.google.com/document/d/1qfztXXRWr1km6U4H44dSpyG7I-Xspd4GkBQmKVjKmbM/edit?usp=sharing and on Amazon https://www.amazon.com/dp/B01MR6BOP0/.

[25] https://www.youtube.com/watch?v=H7Uyfqi_TE8

[26] Available on Github: https://github.com/CHandmer/mars-cargo-model

[27] https://en.wikipedia.org/wiki/Economy_of_Iceland

[28] https://en.wikipedia.org/wiki/OOCL_Hong_Kong

[29] https://statice.is/, https://comtrade.un.org/

[30] https://atlas.media.mit.edu/en/visualize/tree_map/hs92/import/isl/all/show/2015/

[31] Not going to happen.

[32] https://en.wikipedia.org/wiki/Mars_Oxygen_ISRU_Experiment

[33] https://en.wikipedia.org/wiki/Mars_ocean_hypothesis

[34] https://en.wikipedia.org/wiki/InSight and https://en.wikipedia.org/wiki/Thermal_Emission_Imaging_System being the most prominent examples.