Saturday, May 13, 2017

A roadmap to an industrially self-sufficient Mars base in the minimum time

Edit: After this post was initially published (May 12 2017), it generated a flurry of wonderfully constructive comments, particularly on Hacker News ( I have used them all to improve the text, flag some limitations, and better understand the problem. Let's keep the conversation going!

Dear reader(s), let’s talk about how to get to a self-sustaining Mars base as quickly as possible. This is a challenging question to approach, because we just don’t know enough about huge slabs of the problem. Nevertheless, it is possible to approach this problem in a rigorous way and paint, at least in broad brushstrokes, much of the solution. Some of this material is introduced in Chapter 22 of my book “How to get to Earth from Mars: Solving the hard part first” published in 2016 (, but this blog post will take a slightly different approach.

The problem of a self-sustaining Mars base requires the development of much technology that does not exist. Copious and reliable electrical power will be required on Mars, provided most likely by a nuclear fission plant(s) or solar, but is beyond the scope of this discussion. Similarly, a transportation system capable of flying to and from the planet is a substantial problem, but not one I will deal with here. I will be assuming something like SpaceX’s baseline ITA system is available, capable of delivering payloads exceeding at least hundreds of tonnes every 2.2 years, coinciding with the launch window. Details can be found at This blog attempts to answer the question of “What will we do once we’re there?”

Let’s illustrate a picture of how emplacement of industry on Mars may occur, bearing in mind that this will be a rather ambitious timeline, then fill in some of the detail. Technology and ability to fly cargo and humans to and from Mars may not exist forever. Therefore it is wise to try to achieve self-sustainability within a fixed timeline, of perhaps 50 years.

Today on Earth, which is better adapted for life than Mars, between 10 and 100 million people are needed for a sufficiently diverse economy to support the “full industrial stack”, which includes primary resource production, secondary manufacturing of basically everything, and other tertiary services. The number of economic blocs capable of “making anything” number perhaps 5: China, Japan, USA, Europe, India, and perhaps South Korea. Several larger countries are not sufficiently economically advanced. Cuba, North Korea, Australia and Russia (once part of the former club but now enduring industrial decline) all have populations well over 10 million but are entirely dependent on trade to obtain some advanced technology such as computers, aircraft, container ships, engines, cars, and so on. It is impossible to predict with certainty the minimum number of specialists needed to create industry efficient enough to support itself on Mars with the technology of 2060, but one million is probably within an order of magnitude of the true number. To elaborate slightly, I can imagine a machine fab shop with 1000 very clever engineers who can make basically anything from ore given enough time, but sooner or later  (sooner with fewer people) they would be unable to make parts rapidly enough to replace them faster than they break in real world use. Sufficient manufacturing efficiency demands a higher production rate with fewer resources, most prominently human labor!

Scaling to one million people in 50 years, or around 20 launch windows, implies a doubling of population every launch window, which is about a factor of 10 every decade. One decade per decade. Ambitious, indeed, and a great place to start crystallizing an approach.

It seems wise to assume, at least initially, that cargo capacity is closer to constant than exponentially increasing. Therefore, each increase in population mandates a commensurate increase in self-sufficiency, so that the same total cargo capacity can bring enough machinery and supplies to keep everything working.

In the following figure, I plot a hypothetical trajectory of a population from exploration/outpost phase to full self-sufficiency assuming limited cargo transfer capability. On the vertical axis I have constructed a rough order of goods by some metric of inverse manufacturability, and on the horizontal axis I have population. The red line marks a hypothetical population- independence trajectory, and the purple dot the inflection point at which demand for cargo reaches its maximum. Beyond this point, industrializing becomes easier.

The “cusp of settlement viability” is an important concept. It is possible to imagine the dropping of cargo and humans on Mars with instructions to “get cracking”. But every machine and human on Mars represents a future liability for the replacement of that machine and life support of that human, a liability which has to be fully priced into the future. Scaling more quickly than technology and shipping capacity can support guarantees a point in the near future when those liabilities come due, machinery and local industrial capacity undergoes dramatic collapse, and everyone dies of suffocation. There is a serious side to this speculation.


Let’s dig a little deeper into the list of goods or capabilities on the vertical axis. For this figure, I ranked commodities according to specific cost, that is, their cost on Earth normalized by their mass. The reason for this is that the major cost of importation to Mars is driven by the mass of the item, while the cost is a very blunt proxy for manufacturing difficulty. For reference, the cost of flying a tonne of cargo to Mars will not be less than $1m, and could easily be 10x or 100x this, at least initially.

In terms of mass, the greatest requirement on the surface, by far, is oxygen. Oxygen is an underrated element, but accounts for something like 89% of the mass of water, the majority of the mass of rocks, and we also need it to breathe. More importantly, each SpaceX Mars ship needs thousands of tonnes of it for propellant to fly back to Earth. In fact, any non-trivial Mars return flight requires oxygen to be made on Mars, so that’s the first thing on the list.

Fortunately, oxygen is readily available on Mars as the atmosphere is mostly CO2, which is 73% oxygen by weight. It is also worth pointing out that not all in-situ resources are created equal. Atmosphere-derived materials (oxygen, carbon) are easier to obtain than liquid water (via an aquifer or well), which in turn are MUCH easier to obtain than metals from various ores on the surface, or anything that requires digging (although: Boring Company!). The next most important thing to obtain is fuel, of which the SpaceX Mars ship also requires hundreds of tonnes to return to Earth. Potentially the vehicles could bring enough hydrogen from Earth to make methane on Mars, but doing so would consume much of their cargo payload. Therefore, the capability to make enough fuel on the surface of Mars entirely from local resources marks the “efficient cargo utilization threshold”.

The list of items in the figure are based on Earth-costs of production, which do not always map perfectly to cost of production on Mars. In particular, human labor is vastly less available on Mars, and arable land is non-existent. The cost of producing food (carbs) is therefore higher and perhaps should be promoted at least above masonry. One other salient point is that beginning the process of a masonry-producing industry does not mean that the oxygen production plant no longer requires shipment of any parts or humans from Earth. Making a product locally implies an improvement in overall mass efficiency, but not the complete elimination of supporting cargo shipments, something which is not well illustrated in the diagram.

Human labor is so expensive, in fact, that it is worth considering the trade between maintaining and replacing machinery. Obviously machinery sent to Mars must be designed with a high level of reliability, but labor is so constrained on Mars that machines must be capable even of self-maintenance or problem diagnosis. This is a completely different paradigm to the “rugged individual trying to survive” such as Mark Watney in The Martian. I estimate that a machine must have at least 99.9% no-worry uptime reliability to be worthwhile, because the marginal cost of sending and supporting another human solely to maintain the machines is so high. Human labor is so expensive on Mars that it will have to be employed almost exclusively on the deployment on new equipment, rather than constant maintenance of existing machines. For Earth-supplied machinery, it will be more cost effective to provide machines that operate with very little to no intervention and replace them frequently, than to have a labor-intensive machine shop and humans working in it. For Mars-manufactured machines, the calculus is a little different, since it is easier to make a new machine from an old machine than from raw materials. As we will see below, however, there is likely to be little direct human involvement in the (re)manufacturing of machines on Mars.

The situation is even more dire than that, in terms of the scarcity of labor. Not only is there not enough labor available to maintain a constant level of productivity given inevitably decreasing machine health, productivity has to scale with the scale of the base. In the above diagram, the population increases by a factor of 10 every 10 years. Each decade, a new industry is brought online. Therefore, there are four launch windows to deploy, pilot, test, and scale that industry. The next decade will bring 10x as many people, but those people will be primarily devoted to that new, more labor intensive industry. The first 10 people who operate the oxygen plant are mostly “locked in”, while the productivity of that plant has to scale aggressively to meet the needs of the growing base. There are few industries on Earth that can point to doubling the productivity of a human every 2.2 years, but to maintain the schedule that will have to be achieved during the early phases of base construction.

During the latter phase, per-capita productivity will have grown enough that it will not be necessary to send a million people in the final decade to do state-of-the-art computer microchip fabrication, but it is difficult to predict how many will be needed, or even exactly what computers will look like by then. The rate at which individual productivity grows or tapers largely determines the shape and progress of the red trajectory, with the win and peak population demand occurring very soon after the purple cusp. At present, all we can say for certain is that any progress in the initial decades depends on rapid exponential growth of the capabilities of the first generation of settlers.

It seems clear that no matter how fast the Mars base astronauts can swing wrenches, growing demands for productivity will mandate the deployment and exploitation of automated labor. Humans will, in some sense, nurse into existence a base populated mostly by and for robots. An individual robot is probably even less capable of scaling its productivity beyond its design than a human, therefore the number of robots will also have to double every launch window, which provokes another interesting modification to the technology priority acquisition timeline.

The Mars base will need to make robots, or at least parts of robots, as soon as possible. Fortunately, the capability to make copious methane fuel creates the foundation for ethene and polymers: plastics. That is, a Mars base that has not yet scaled to the mining of solid ores is able to make plastics accounting for the majority of the mass and bulk of a robot arm (or leg), and scaling this ability will be of paramount importance. Other essential robot components like actuators and processors are extremely labor intensive to produce, but relatively light and can be flown from Earth while local manufacturing scales according to its needs. Pumps, valves, filters, bearings, latches, brushes, robots, and regulators are all wear parts, some of which can be made locally of printable or machineable plastics where doing so is cheaper than importation from Earth.

The thought of huge facilities full of brightly colored 3D printed plastic robots building each other at a fabulous pace is not what I had in mind when I started thinking about Mars industrialization, but it is a compelling vision. Large scale integrated robotic factories are currently being developed around the world, such as the Tesla gigafactories. Tech development that's good for Mars also makes a lot of economic sense on Earth. In fact, some hobbyists on Earth have gotten dangerously close to building microgigafactories in their garages. The reprap project ( represents a microcosm of the overall problem - a 3D printer which can print (most of) itself. A more complete vision for hobbyists might be the creation of a “robotic garden” with commercial off the shelf components generating plastics from CO2 and water, and gradually 3D printing replacement parts until the entire manufacturing chain has evolved into a self-maintaining software-defined plastic ecosystem requiring minimal hands-on human involvement.

At the beginning of this post, I mentioned that all but five or six countries on Earth were incapable of making enough stuff to be self-sufficient. I am a big fan of trade and economic efficiency provided by trade, but it has left smaller nations vulnerable to industrial dependency, economic weakness, and potentially global trade disruption. In fact, any disruption of the global economy in its current hyperinterdependent phase may not be recoverable, seeing as we’ve already depleted all the easy-to-obtain surface resources. It is much easier to emplace industrial self-sufficiency even in some bone-dry valley in central Nevada than on Mars, so the development of technology which permits that is an essential safeguard for civilization on Earth, as defined by the ability to make or obtain “anything” with a trade-competitive level of overhead.

Although we have had to remain agnostic about huge facets of Mars industrialization, including precise numbers on who, when, where, how much $, how big rockets, and so on, we have made some progress. We have seen that a labor-cost focused approach, normalized by the requirement of “self sufficiency, ASAP”, has illuminated the importance of understanding the relative value of transportation, human labor, maintenance, and robotic labor.

Monday, April 17, 2017

Does Lunar resource exploitation make sense?

Hello loyal reader(s). Although I haven't blogged now for a few months, that doesn't mean I've been doing nothing. On the contrary, I have been advancing several super cool projects and today I'm going to write about an aspect of one of them.

Every few years (roughly coinciding with congressional budgeting schedules) NASA gets antsy and proposes some new ideas. Recently, they have included the asteroid capture mission, the Europa lander mission, and all sorts of other cool concepts. On the crewed side, however, NASA is (and has been) stuck in an organizational quandary, wherein it is allocated just enough $$ to do what it has been doing, and not quite enough to make a solid start on any of its mandated new programs, such as the Mars mission.

I have written extensively about crewed Mars exploration in the past, and a distillation of much of that is kept at . The main problem with Mars exploration is that there is no way of doing it with existing rockets. Developing new rockets is expensive, large rockets particularly so, and so the hunt has always been on for finding smarter ways of getting more mass to (and from) Mars using rockets that fit, somehow, within the current budget. This is a conceptual mistake, in that huge new rockets are certainly expensive, but they are cheap compared to the programmatic costs incurred by having a rocket that while undeniably huge, is just not quite huge enough. I am reliably informed that similar cost inefficiencies can occur in other areas too!

This blog post deals with one particularly baroque proposal, namely the installation of a robotic fuel mining base and "gas station" on the Moon, to refuel spaceships on their way to other places. This proposal has been floating around for a while but has recently gotten a lot more attention than is, perhaps, warranted, hence this blog. The topic is quite arcane so I will do my best to keep the writing both concise and precise. First, I will summarize the results, then delve into entirely inappropriate levels of detail.

Much of space exploration advocacy is performed by way of analogies. Unfortunately there is no good analogy for this particular proposal, so instead I have used math to compute some best case cost estimates for Lunar resource exploitation, and compared them to the alternate method (Earth-launched resources) computed using median case cost estimates. This biases the comparison toward Lunar fuel, but will it be enough?

This table shows the per year cost for a program designed to deliver 100 metric tonnes of cargo (such as water) per year to various locations in cis-Lunar space. It also estimates the development and deployment time to reach rate after program start.

Earth originEarth originEarth originLunar originLunar origin
Δv (km/s)
Cost ($m/year) expendableCost ($m/year) reusableTime to reach rate (years)Cost ($m/year) reusableTime to reach rate (years)
Earth surfaceKSC0NA0.150.05NA>15
Low Earth orbitLEO9.43001202>1000x5>15
Geosynchronous transfer orbitGTO2.446002402>1000x5>15
Trans-Lunar injectionTLI0.687503002>1000x5>15
High lunar orbitHLO0.147503002>1000x5>15
Low lunar orbitLLO0.689003602>1000x4>15
Lunar surfaceLS1.7318007205>1000>10

The most optimistic cost estimate for the robotic Lunar port suggests costs of $1b/year for 15 years to reach rate, and that's what I've used in this graph. I think all reasonable experts would agree it's highly unlikely to cost less than that, or to reach rate (100T/year delivery to some location) faster than that. The xN quantities encode the fact that moving fuel from the Moon to other locations uses >75% of that fuel in delivery. So if $1b/year for 10 years is enough to produce 100T of water a year on the Moon, additional time and tech and fuel and money is required to move that fuel to, say, Low Earth Orbit.

In contrast, we see that using today's technology at today's prices, the same quantity of water (or any cargo) can be delivered from the Earth to all the same locations at a fraction of the cost and a fraction of the time. Employing reusable rockets, such as those currently being pioneered by SpaceX, may reduce costs even further, to the point that the cheapest, fastest way to get even raw materials on the Moon is to launch it from Earth instead of mining it locally.

Before I dive into the nitty gritty, it is worth stating that a similar analysis focused on the use of Mars' atmosphere (rather than the deep frozen heavy metal-laced dirty snow of the moon) for propellant production shows a clear advantage over launching all the fuel required for the Mars-Earth trip from the Earth. 

Now I can dive into the nitty gritty. First I'm going to write about the why, then I'm going to write about the how.

A really good rocket can launch about 4% of its initial mass into low Earth orbit (LEO). For the Saturn V (the most powerful rocket ever built), the orbital payload was about 140T. To get from LEO to the moon, Mars, or elsewhere, yet more fuel has to be burned. For LEO-Mars, around 25% of the LEO mass can be payload, the rest has to be fuel and oxidizer. At this point even the Saturn V can launch only 35T to Mars and that's not really enough to keep four brave astronauts alive for a three year mission and then bring them back.

Instead, the 140T in LEO can be the payload and spaceship with empty tanks. 3 more launches of the Saturn V can increase its mass to 560T, at which point it has enough fuel to fly to Mars with 140T of payload, which is much better.

Unfortunately, four launches of the Saturn V is much more expensive than one, and building a rocket 4x bigger than the Saturn V, while exciting, is not part of the solution space NASA is presently looking at, possibly because the manufacturing facility at Marshall Space Center in Alabama couldn't fit it through the door. 

If ~400T of propellant is needed in LEO, however, perhaps it could be obtained from the Moon? But how? Remember that the baseline expense case is three more launches of an already existing launch vehicle, so any alternate scheme should be some combination of cheaper, safer, faster, or more scale-able.

The best Lunar resource extraction architecture I've come across so far looks something like this.

The following new robotic vehicles are developed on Earth:
- A solar electric propelled orbital tug.
- A hydrogen/oxygen powered lunar orbital shuttle and lander, based on the Centaur upper stage.
- A solar powered fuel processing plant with some capacity for remachining or replacing worn out components.
- A Lunar orbital nanosat platform containing numerous guidable lead or steel rods.
- A battery powered combine harvester robot that ingests lunar regolith.
- A battery powered generic transfer truck with robot arms and useful tools.
- A solar powered deep space electrolysis cryogenic fuel depot. 

The lunar components (in sufficient numbers) are deployed near one of the permanently shadowed regions at the lunar pole, landing on the landing vehicle. The orbital nanosats deorbit cavalcades of dense metal rods to precisely impact the mine site, performing a kinetic drill and blast procedure. The combine harvesters scoop up the fractured regolith, physically process it for water and other volatiles, and transfer the ore to shuttle trucks while dumping the depleted material, which can also be used (eg sintered) to make roads or landing pads. The trucks shuttle the physically separated ore back to the fuel processing plant, which performs chemical separation and packages water ice in aluminized mylar coated pallets for transportation. It also performs limited electrolysis to make fuel for the lunar orbital shuttle's ascent flight.

The shuttle flies the water ice to low Lunar orbit, depositing it at one of the deep space fuel depots, refuels with electrolysed fuel from that depot, and returns to the lunar surface. That part of the operation has a mass efficiency of just 20%. That is, 80% of the extracted water is used propelling the shuttle to and from the Lunar orbital depot. Hydrogen boiloff may be mitigated by (eg) platinum catalysis and conversion back to water.

Non-hydrolyzed water ice is collected at the lunar orbital fuel depot and transported by solar electric tug back to low Earth orbit, consuming a relatively trivial fuel fraction but taking at least several weeks. Water ice is stockpiled at the low Earth orbiting depot(s), which must hydrolyse it all in time for the required launch to Mars, or wherever, and requiring huge solar arrays to do so. 

There are numerous other proposed systems which are less mass or time efficient, or have less overall benefit. As an example, it may be possible to fly a Mars vehicle to land itself on the Moon, refuel there, and then fly on to Mars. However, it would take less fuel to fly from LEO to Mars directly. Similarly, the mass benefit of any post-launch refueling drops off extremely quickly for any depot beyond LEO. Although the Moon has relatively low gravity, its lack of an atmosphere extracts a toll in both directions; launch and landing.

If the above scheme for mining propellant from the moon sounds complicated, that's because it is! In fact, of millions of potential failure modes, the net outcome is the same - not enough water delivered to LEO, or even none at all. To mitigate the programmatic risk for the crewed flight to Mars, a mechanism for the delivery of water from Earth to top up the LEO-based solar powered fuel depot must be provisioned for. At which point, of course, it is (by the table above) far cheaper and quicker to cancel the lunar program entirely and refuel the depot, or the Mars vehicle itself, using that same Earth-launched mechanism. 

I really do not believe there is much more to say about the Lunar-derived fueling concept. Here are some links to other resources if, for some reason, your curiosity is not entirely sated.