of over 1000. To keep the driver length within practical limits, it must use high acceleration (several hundred Earth-gravities).

The mass-driver employs superconductors only in the buckets; the stationary drive windings are of ordinary aluminum. For commercially available superconductors, field-strength limits are not reached even at accelerations of more than 1000 g. (Mainly for historical reasons, the earlier calculations on mass-drivers were based on rectangular geometries similar to those of an MIT Magneplane transportation concept." Calculations indicate tight-coupled axial structures, advocated by H. Kolm of MIT, to be lighter by more than a factor of 2 for the same performance.")

At a nominal 60 flights per year, the Shuttle can bring to low orbit approximately 1700 tons of payload and 2100 tons of reaction mass (i.e., external tanks). For spiral orbits, missions to lunar distance require an engine with specific impulse in the general range of 1000 sec, if the engine is to be returned and a substantial payload carried. A massdriver of that sort would have these characteristics based on the latest 1977 Study results:

Round trip time to lunar orbit, days.. .....200 Reaction mass use rate, tons/yr.... . 2,100 Payload from low-Earth orbit to lunar orbit, tons/yr ....... 1,300 The 1977 Study identified as a major necessary development for reaching this performance reducing the weight of a silicon-controlled rectifier to only a few times the weight of the silicon wafer alone. Present silicon rectifiers, designed for entirely different service with high duty cycles, weigh far more.

Until the 1977 Study, the mechanical stability of the long, thin mass-driver structure as buckets accelerated and decelerated remained unclear. After a great deal of work the Study concluded that the machine would just remain stable. Therefore, the system mass includes a yardarm and guywire structure with slow-acting servomechanisms to maintain straightness. (The Stanford Two-Mile Electron Accelerator, located almost on top of the San Andreas earthquake fault line, is given occasional adjustments based on a laser-beam sight line. Its aperture is about 1 cm.)

The nominal design just outlined would give more than twice the specific impulse of the Space Shuttle Main Engine, which is the state of the art for large chemical rockets.

To the best of my knowledge, no physical upper limit on mass-driver output velocity exists (other than c); but the optimum I, for missions in the

Earth-Moon system is 800-1000.

So we should be able to lift about 1300 tons/yr of Shuttle payloads from low-Earth to lunar orbit, without having to devote additional Shuttle flights to carrying propellant. What minimal facility would then be necessary on the Moon to convey lunar materials into space?

My original estimates centered on 1050 tons. A more detailed evaluation carried out during the 1977 study concluded with a figure of just under 1000 tons. The axial mass-driver for the lunar surface contributes less than a quarter of the total, excluding its foundations. That machine would be sized for an eventual throughput of 600,000 tons/yr, but the photovoltaic array first sent to the Moon could handle only a twentieth as much throughput. The machine parameters are:

F-2

In mission scenario for mass-driver reaction engine (MDRE) operating as an interorbital transfer tug, total mass in low-Earth orbit (LEO) would be 2075 tons. The engine would expend 1000 tons of powdered Shuttle tankage in the spiral-orbit climb to the Moon. After dropping a 730-ton payload in lunar orbit, the unmanned tug would spend the remaining reaction mass in a relatively quick return to low-Earth orbit for maintenance, refurbishment of solar cell arrays, and reloading with payload and reaction mass.

......93

Photovoltaic power supply (for 30,000 tons/
yr)
Downrange course-correction stations . . . . 50
Habitat (temporary for 50, long-tour for 10)..

300 Life-support resupply (50 people, 1 year, at 10 kg/d)... .. 180 Silicon and glass-fiber plant for materials encapsulation.... ........ 30

In addition, resupply for a 10-person crew using 10 kg per person per day would total 37 tons per year. To soft-land that equipment, an equal quantity of propellant must be brought to lunar orbit. For operations both on the lunar surface and in space, we must ante up about 2800 tons of Shuttle payloads-about 100 Shuttle flights-to commence transporting materials from the lunar surface to a precise point in space. The 2800 tons breaks down as follows:

Total lunar surface equipment and supplies..

Interorbital tug (LOX-H, engine, tanks, controls). 8 Supplies for personnel in space (for 1 yr at 10 kg per person) 110

The 1977 Summer Study obtained a similar total from a slightly different mix. Lunar landers and personnel transfer tugs had more mass, but converting spent Shuttle external tanks to 21-person habitats for use on the Moon and in space saved substantial mass. Those habitats could be spun on cables to create normal gravity, and would be divided into private apartments, one for each person in the original workforce.

At the first plateau, a number of milestones would be reached: proof of the concept of lunar materials transport, provision of large quantities of reaction mass for all later operations, and opening of a large reserve of mass for shielding manned operations at geosynchronous or higher orbit.

In reaching this plateau it will not have been necessary to learn how to chemically process or fabricate from lunar materials, nor to learn how to set up agriculture in space.

Once at the plateau, with the availability of lunar soil as reaction mass, the throughput of equipment from the Earth to geosynchronous and higher orbits will no longer depend on powdered Shuttle tankage. (That might be a good time to deploy a Shuttlederived lift-vehicle, so that subsequent launch operations can be of lower cost.)

The next step requires chemical processing, making oxygen in lunar materials (40% by weight) available as propellant for lunar landers, as

Throughput/power, tons/year per Mw... 650 Process and fabrication plant throughput (includes radiators but not power supply), tons/year-ton of plant 200 Photovoltaic power in space, tons/Mw... 5.3 Labor utilization (Si plus metals, fabricated), tons/year per person

Resupply needs (after oxygen extraction becomes practical), kg/person per day...... 2 LEO to high orbit mass penalty (personnel transfer). Factor of 4

Totals for process and fabrication:
Throughput, ton/yr...

30,000

The 1977 Study updated these numbers, but the major conceptual breakthrough on chemical processing during 1977 was laying out a new basic process-flow chart depending on a carbo-chlorination reaction rather than on carbothermic reduction of aluminum-bearing minerals. The change reduced the peak temperature in the system by a thousand degrees to bring it in line with normal industrial practice here on Earth.

Once chemical processing is tested and working, the next step is fabricating a limited range of products from chemically separated lunar materials, such as photovoltaic arrays (lunar material, 20% silicon by weight) and thermal radiators to upgrade both the lunar launcher and the processing/fabrication plants in space. Products might later include structural components. We estimate an output of about 9000 tons per year of fabricated components-of these, about 20% would be photovoltaic arrays and 80% would be habitat components.

Together with 1800 tons/year from the Earth (21 flights/year of a Shuttle-derived freight rocket) and 750 people per year (13 Shuttle flights), this system would grow at a rate of 150,000 tons of additional throughput per year. When that growth reaches 630,000 tons throughput per year, or 190,000 tons/year of fabricated output products, the system output would equal in mass two 10,000Mw satellite power stations per year."

Minimum transport investment up to that point comes to just under 300 flights from Earth to low orbit, or about 42 flights per year averaged over seven years; of these, over half would be Shuttle flights (STS) and the remainder Shuttle-derived lift vehicles (SDLVs).

F-3 summarizes recommended transport investment. It includes additional safety factors-mainly 96 additional SDLV flights to provide the 3000-person workforce in space with 8400 more tons of supplies. These may be in the form of 1 ton of organics per person to stock later space farming, of 2 kg per person per day of emergency resupplies adequate for two years, and 4300 tons of Earth

built components for habitats or for products exportable to geosynchronous orbit.

If (as assumed) space agriculture has not yet been developed, 25 SDLV flights per year will be required to transport food to the 3000-person work force, at 2 kg per person per day. At $20 million per flight that will add about a dollar per pound to the cost of fabricated products. (The rate of developing agriculture in space will likely be affected less by direct lift-cost economics than by the need to

maintain workforce morale by providing enjoyable fresh foods.)

The range of output products and the rate of diversification of products will depend on the economics of construction in space vs. lift from Earth. To sustain rapid exponential growth, an increasing share of the total must be built in space. For example, if 5% of the output products' mass must still be obtained from Earth when 600,000 tons per year is being processed, about 100 flights per year of a Shuttle-derived freight rocket will be

decade, to a point where the entire market for new generator capacity worldwide apparently might be provided by space-built satellite power. That Study found the market 25 years from now $200-400 billion per year. It found exponential growth essential for beating interest costs, and ran into the biggest uncertainty over what fraction of the products assembled in space must be hauled from Earth. As both studies in the past two years have concluded, effort must go into designing fabricated products optimized for construction from lunar

F-4 R&D COST ESTIMATES

export labor. The latter two have been ignored in this simplified treatment. The time-line appears rather insensitive to those numbers until the first plateau; after that, the actual growth rate depends on them more.

The takeoff point in space manufacturing will reflect our cultural values. It might be defined in one way as the point at which the rate of valuegeneration exceeds the rate of investment, the latter including development, construction, transport, and interest charges.

R&D cost estimation is a notoriously uncertain business. Given the transport economies just outlined, R&D costs may dominate the total in

vestment required to reach the takeoff point. R&D costs have ranged from as little as $6 million per ton in 1960 dollars for a throwaway rocket whose engines had already been developed, the Saturn V, to over $200 million per ton for the Apollo Command Module.

To estimate R&D costs for space manufacturing, it seems best to use a modern example: Shuttle R&D totals $6.3 billion. Of that total, I take two-thirds as devoted to the Orbiter, a complex but reusable 68-ton vehicle. Orbiter R&D costs approximately $62 million per ton. The relatively simple external tank and one SRB total 76 tons, and cost $28 million per ton for R&D.

The components of the space-manufacturing program just described have been analyzed for cost on that basis, as summarized in F-4. Mass-driver structures were estimated on the basis of individual non-equivalent sections every 5% in bucket

velocity. The crew-transport module for 60 people was taken as a Shuttle payload of 11 tons dry mass, and the workforce habitat module was assigned a mass of 12 tons. To compare this primitive approach with the more detailed estimates made during the 1977 study, all mass-driver R&D in F-4 costs $2.4 billion. Several months later different persons obtained a figure of $3.3 billion, which includes construction of several mass-driver tugs and an extensive program of unmanned observations by lunar polar orbiter and exploration by roving lander. No sensible person would conclude from that a figure of $2.7 billion 20%. Other estimates, equally defensible, were higher and lower by factors of two.

F-4 shows a range of cost estimates for some items for which there are data from previous studies of orbital transfer vehicles, space stations, a lunar base, and a Shuttle-derived lift vehicle. All estimates are necessarily rough and do not consider overall integration, government infrastructure, and commonality with other applications, but they do serve to show that the investments fall in a reasonable range.

Both studies reveal clearly that reaching the economic takeoff point does not require large monolithic habitats ("space colonies"). Furthermore, colonies could not be developed like aerospace hardware, for R&D cost reasons. When the time comes to build such structures, they will have to be designed and constructed like ships or like buildings rather than like jet aircraft. Spacefarms, like monolithic habitats, do not appear to be essential in reaching the takeoff point.

The 1976 Ames Study concluded that over the range of 1 to 30 kg/sec in throughput, chemical extraction and fabrication plant mass and power should be proportional to throughput. For minimizing R&D costs, therefore, it seems best to design such plants for moderate size (1 kg/sec or 30,000 ton/yr) and then to parallel identical units.

It is reassuring to find that most of the necessary early R&D steps have potential benefit to all largescale high-orbital activity, whether or not using lunar materials.

To indicate the uncertainty of R&D cost estimation, in a recent NASA Johnson Space Center study a large, fully reusable Class IV launch vehicle was assigned an R&D cost of $6 million per ton-a factor of ten below the figure used in this article, and one-seventh of the average for the Shuttle. " Until far more detailed work is done, it seems premature to argue which, if any, of these estimates is near the truth. I use the higher figures mainly to be on the safe side.

The required R&D investment falls naturally into three blocks: