Saturday, May 23, 2020

The High Frontier, space based solar power, and space manufacturing

The High Frontier, published in 1976 by Gerard K O'Neill, lays out a vision of economically profitable space colonization in artificial orbital habitats, and guessed (while disclaiming it as prediction) that it was "unlikely" that a space community would not be established in 30 years. I was interested in why those forecasts were made, and why they turned out wrong, as data points for thinking about forecasting future technological developments. The book lays out a case that in the long run space habitats can support immensely larger populations and wealth than the planets in the Solar System. In the medium term it argued that a government program to invest hundreds of billions of dollars to build space factories and Lunar mining facilities would eventually let them produce solar power a few times more efficiently than terrestrial solar power production, and that this would drive space colonization. This seems to have been doomed for multiple reasons, radically underestimating launch costs and likely fatally underestimating the increased costs of space production (to be paid for out of a 2-3x improvement in solar radiation), as well as requiring immense government funding. As a means to improve solar power cost-effectiveness, it would have been far inferior to solar cell R&D. Subsequent orders-of-magnitude improvement in launch costs per kW of solar cells make space-based solar more plausible than at the time, but the challenge of competing with terrestrial solar and especially terrestrial scale economies of industry remains high.

Space colonization as a means to ease ultimate terrestrial resource limits

The book was written when concerns about population growth were a much larger topic of discussion than today, and during the 1970s energy crisis when oil prices had spiked enormously. It discusses historical growth of energy use by 7% per year, and the strong linkage between economic growth (which has been attenuated since the 1970s as economies reduced energy intensity improved their production of economic output per unit energy consumption).

O'Neill mentions but does not rely on the historical superexponential growth of population, the cancelled singularity:

"Viewed on a time scale of many centuries, though, the population-growth rate has itself increased continuously. This has led to such papers as that of Von Hoerner, which shows that up to 1970 the best mathematical fit to the population-growth curve would lead to a true "explosion": an infinite number of people about fifty years from now...Thhis sort of study is of great value in calling attention to the growth problem, but it is best understood as a statement that within the next few decades the growth rate must reduce, and radically. For purposes of this book I will use the much more conservative growth-rate figures of the U.N.: the situation is already serious enough without the need to overstate it...The U.N. hardly dares to predict what will happen [in the 21st century]...but if we project their graphs we find that the 10 billion mark will be reached by 2035"

Space colonization is supposed to provide centuries of headroom in low energy, land-equivalents, and materials, in line with four guiding principles:

  1. A proposal to improve the human condition makes sense only if, in the long term, it has the potential to give all people, whatever their place of birth, access to the energy and materials needed for their progress.
  2. A technical "improvement" is more likely to beneficial if it reduces rather than increases the concentration of power and control.
  3. Improvements are of value if they tend to reduce the scale of cities, industries, and economic systems to small size, so that bureaucracies become less important and direct human contact becomes more easy and effective.
  4. A worthwhile line of technical development must have a useful lifetime "without running into absurdities" of at least several hundred years.
O'Neill argues against 'planetary chauvinism,' noting that the land area of the Moon and Mars is only about the dry land area of Earth. He describes the O'Neill cylinder, habitats that rotate to produce pseudogravity via centrifugal force. While solar power reaching the earth is about 2e17 W, the luminosity of the Sun is ~ 4e26 W, a difference of more than a billionfold. Asteroid materials also dwarf terrestrial resources near the surface.

So space habitats permit vastly larger eventual populations and economic output than the Earth alone can sustain. There is some talk of the long-run possibilities, e.g. that with population growth of 1/6th per generation, 20,000x growth would be attained in 5,000 years, and with doubling times of the day (~35years), growth would max out the solar system's capacity long before that. But even if most of the long-run potential of the solar system lies in space habitats, that doesn't mean that at the current margin it makes sense to build or live in space habitats.

The Earth provides its own atmosphere, gravity, radiation shielding, ecosystem services, and proximity to other humans. It is much cheaper to build homes and factories on Earth, even setting aside the dominant (and enormous) launch costs. While agricultural land is limited, expanding cultivation of marginal land on Earth is cheaper than building greenhouses and hydroponics facilities, let alone space agriculture. Continued population and economic growth would eventually yield much higher land and resource prices, reducing or reversing this price differential, but in the meantime it would only make economic sense to settle space to take advantage of some large special advantage.

The existing satellite industry exploits the high altitude of Earth orbit to increase the portion of the Earth within line of sight of a transmitter, but this does not call for the scale of construction that O'Neill is interested in. The space advantage that he focuses on is the greater availability of solar energy in space without (1) attenuation by the atmosphere, (2) blockage by the Earth, with reduced insolation at night and seasonally. In space at a distance from the sun of 1 astronomical unit the solar irradiation is ~1.36 kW/m^2. Averaging over the surface of the Earth across weather, day, and season irradiation is about 0.18 kW/m^2, In sunnier areas such as deserts close to the equator the numbers can be above 0.3 kW/m^2:

So space habitats could enjoy a 4x advantage in solar flux for solar power or agriculture over terrestrial deserts (which have large areas of unused land that is much less hostile than space and vastly cheaper to transport to and from). In the vision of the High Frontier, this is the difference that is supposed to motivate the construction of space habitats and industries.

Space based solar power as the industry driving space settlement

O'Neill is clear that large scale space colonization requires that it be advantageous for those funding initial construction and those eventually migrating to space, remarking that "even our first space colonies must pay their way, and they can only do so if they do not price themselves out of their markets." So space solar power production must be able to deliver large profits, even considering the costs of launch and industry in space.

A 4x space solar advantage in photovoltaic productions vs terrestrial solar sites (O'Neill likes to compare to the average across the United States, but that doesn't seem to be the right marginal case, although solar power satellites can be directed to different locations) is attenuated by losses in transmitting the power to Earth, by beaming microwaves at fields of receivers on Earth. The book argues for losses of about 2x in this process. There could be improvements on this, but also the technologies have not been tested to reveal other in practice costs.

Combined with increased radiation on the space-based solar panels damaging them, the expense of transmitters and receivers, and the increased costs of space labor (remote controlled robots, or very expensive humans), this seems to leave only a modest or negative potential profit to the solar panels from radiation in space, less than a doubling, even with ~free space transportation at scale.

For comparison, terrestrial solar cell cost-efficiency has doubled about every 5-7 years through technological advance and economies of scale, with the faster progress in recent years:

Recently, as solar becomes economically viable at larger scales, solar costs have dropped 5x in 10 years. These advances came with cumulative R&D spending that is quite small compared to the sorts of investments discussed by O'Neill. E.g. from 1948-2018 the U.S. federal government provided $29.35 BB for all renewable energy  research, of which solar only captured a portion.

Private manufacturer R&D has been increasing recently, but is still small compared to the costs discussed in High Frontier (hundreds of billions of 2020 dollars), e.g. this dataset of public company announced  solar R&D:

20 PV Manufacturers (Publically Listed) Annual R&D Expenditure (US$) Millions 2007 to 2016.

It still appears that government funding for solar R&D would have been a far more helpful than space launch and construction subsidies for the goal of cheaper solar power, even if the plan could have proceeded as claimed.

But in fact space transportation is not free, but enormously prohibitively expensive. The Apollo rockets costs thousands of 1976 dollars per kg (4.5x as much in 2020 dollars) to lift material to orbit, several times that for moving to the LaGrange points and geosynchronous orbit and ~$20,000 (1976) per kg sent to the moon. Such prices render shipping solar power satellites from Earth completely economical: a 2x improvement in solar efficiency cannot pay for orders of magnitude increase in costs.

Bootstrapping space manufacturing to reduce launch costs
The original edition of the book had two stages to its model of how transport costs could fall enough to permit space power satellites to pay for themselves. The first step was accepting NASA estimates of its launch costs using the space shuttle, which turned out to be optimistic by more than an order of magnitude.

This was despite the following in the appendix:

Each of the specialists expressed strongly the opinion that the critical numbers assumed for the work so far, and quoted in this book (mass-driver acceleration and efficiency, HLV lift costs, lunar power-plant mass, etc.) were too conservative and could be improved substantially without great technical risk.

The second step was to acquire materials and construct the power satellites using materials acquired from the Moon and eventually asteroids. Terrestrial launch capacity would send workers and industrial equipment to a space manufacturing facility, with mining equipment and electromagnetic mass drivers to extract materials and send them to the manufacturing facilities at manyfold reduced costs, as a result of lower Lunar gravity, and the advantages of mass drivers over rockets.

In a 1975 Science article, O'Neill lays out lift costs for constructing a first manufacturing facility, from $70BB to $380BB in today's dollars (plus somewhat smaller wage and construction/development costs for a total of 142-834BB). With actual shuttle costs, this would be in the trillions of dollars, and completely dominated by terrestrial solar R&D and investment as a means of acquiring cheaper solar energy.

The article assumes this facility would then produce further facilities, equipment, and solar power satellites with minimal further input from Earth (except high value products such as chips and new workers) autarkically, with exponential growth of colony industrial base. Industrial production rates seem to be based off of terrestrial rates (in generic 'tons of production'), without inflation for the high costs of work in space, and especially the lack of economies of scale. We know from terrestrial data that there are quite drastic economies of scale (for both solar and the industries feeding into it, which would compound), and a facility of several thousand people replacing the entire supply chains of the solar, mining, and space habitat construction industries without a 2x loss of productivity from scale effects seems quite unlikely to me, pushing out the minimum scale of space community needed to profitably produce solar power satellite.

The structure of the plan also requires large investments developing all the necessary methods, and astronomical investments in building a space community, in hopes of eventually reaching profitability (which looks unlikely to me given the technology of the time). O'Neill recognized such a thing could only be undertaken by government:

By now our planning group benefits from the advice of senior executives in the electric utilities and investment communities. From them we have learned a good many realities that help us in guiding our research. For one thing, it seems almost certain that we cannot expect private capital to invest in space manufacturing until the risks have been reduced almost to zero. Government funding...will have to carry the program at least until a pilot SSPS, not necessarily made from lunar materials, has supplied energy to the Earth...Above all, the economic studies made at that time will have to show that SSPS power can undersell all competition.
But governments would do vastly better financing solar energy research on Earth, and >Apollo program commitments are not reliable for questionable unproven technologies without returns along the way in any case.

Change since 1976
There were a number of further NASA reports on this subject, but ultimately the US government rejected O'Neill's plans as infeasible. In an update to the book in 1988 O'Neill addressed the shuttle launch costs, citing work on telerobotics and self-replicating machinery. Remote controlled robots could be much cheaper than humans in space, although still far more expensive than human workers (terrestrial solar and mining supply chains weren't completely automated for good economic reason), likely overwhelming the benefits of space solar.

A push towards lowering launch-costs, and progress in making solar cells lighter per unit power production (which accordingly reduces launch costs) could potentially drastically improve the cost-benefit today. SpaceX has cut launch costs by a factor of 20-50x vs the space shuttle to close to $1,000 per kg, with room to improve at least an order o f magnitude if reusable spacecraft can be efficiently turned around (without refurbishment costs close to construction costs). Fuel costs are still substantially less than 1% of SpaceX rocket costs.

Progress in solar cells and thin film cells has increased power per kg several fold, and potentially tremendously for thin film cells. But at the same time the scale of solar production facilities has increased enormously, as has the sophistication of their production: to compete with terrestrial solar panel production, space-based manufacturing facilities would need more sophisticated facilities than O'Neill proposed (and which I had difficulty believing previously), increasing minimum scale. Going from a modern supply chain of millions of people across many factories, mines, etc to a robotic omni-factory looks like a lot more than a 2x or 3x cost difference to justify space production with solar flux.

Earth-launched power satellites may become feasible as an extension of broader solar and launch trends, before robotic space factories that can compete with terrestrial supply chains (even with the advantage of high orbit).

No comments: