100 Buckingham Drive, No. 253, Santa Clara, California 95051, USA.
Journal of the British Interplanetary Society, Vol. 33, pp.
251264 1980.
Note: This web version is derived from an earlier draft of the paper and may possibly differ in some substantial aspects from the final published paper.
In the classical "Bracewell probe" contact scenario [2, 7], the automated device enters our Solar System, detects radio emissions of an unnatural character emanating from Earth, and subsequently positions itself in some convenient parking orbit around our planet. Upon receiving some arbitrary human transmission, the intelligent probe beams an identical copy of the message back to the transmitter site in hopes of gaining our attention. Once accomplished, language lessons soon follow; hopefully, meaningful discourse and cultural exchange between humanity and the automated alien ambassador ultimately take place.
Project Daedalus, a preliminary design study of a flyby probe mission to Barnard's star recently completed by members of the British Interplanetary Society [8], has demonstrated the feasibility of this approach to interstellar exploration and communication using foreseeable human technology sometime in the next century. The automated probe strategy has received further support from Freitas [3], who shows that transmission of information across interstellar distances using energymarkers (photons) or mattermarkers (probes) may be energetically and alternatives for highly developed technological civilizations Only such advanced societies realistically can afford either radio beacons or starprobes, and secondary distinguishability criteria suggest the possible superiority of intel ligent automata for contact and communication missions between extraterrestrial civilizations. The search for alien space artifacts in our own Solar System has already begun [9].
A major alternative to both the Daedalus flyby and "Bracewell probe" orbiter is the concept of the self reproducing starprobe. Replicating spacefaring machines recently have received a cursory examination by Calder [4] and Boyce [5], I but the basic feasibility of this approach has never been seriously considered despite its tremendous potential. In theory, each self reproducing device dispatched by the launching society would become an independent agent, slowly scouting the Galaxy for evidence of life, intelligence and civilization. While such machines might be costlier to design and construct, given sufficient time a relatively few replicating starprobes could search the entire Milky Way.
The present paper addresses the plausibility of selfreproducing starprobes and the basic parameters of feasibility. A subsequent paper [10] compares reproductive and nonreproductive probe search strategies for missions of interstellar and galactic exploration.
2. THEORY OF SELFREPRODUCING AUTOMATA
There is little disagreement among cyberneticians that full machine self reproduction is possible in principle. Quite a number of techniques have been explored by modern theoreticians; however, it was von Neumann [11][12] who first investigated several different models of automata replication.
The simplest of these schemes is usually called the "tessellation model." Space is divided into cubical cells, and each part of the machine and each piece of raw material from which it constructs itself occupies exactly one cell. All processes are quantized in time as well as space  activity occurs in step with regular clock cycles. In the tessellation model examined by von Neumann, space becomes twodimensional for simplicity, with the machine occupying a connected set of squares. Each square is in one of a specified set of states. The machine is surrounded by inert squares which it must organize into an exact duplicate of itself.
Von Neumann's simplest example calls for a box consisting of 80x400 contiguous cells plus a long linear "tail" of coded instructions 150,000 squares in length. The box consists of a specific arrangement of three kinds of parts  neurons, transmission cells, and muscle cells  and performs two primary functions. First, it follows the instructions encoded in the tail, in sequence. Second, it copies the tail, which contains a coded description of a tailless box. Reproduction is complete when the box organizes the inert units around it into a copy of the tailless box, then copies its own tail and attaches this to the second box.
The second von Neumann scheme, commonly known as the "kinematic model," is somewhat more abstract and generalized. It involves the activity of a Universal Turing Machine, a device designed conceptually by Turing [13] which is able to process any specific algorithm. Since its output is pure information, von Neumann conceived of a Universal Automaton capable of building material objects using any specific construction algorithm.
Imagine three Universal Automatons A, B, and C. If the description of A is fed into B, the output of B is the behaviour of A. If the behaviour of A is the construction of C then the output of B is the machine C. But if C is identical to B, then B is producing copies of itself  in essence, it is reproducing. Of course, for B to function properly it must have access to a stockroom well supplied with all component parts necessary to build each Universal Automaton.
It may be argued that the above is merely a design for a general purpose assembler robot, whose output happens to be copies of itself. This may be true, but still the process is properly termed "self reproduction." Given access to the proper environment, B can indeed replicate itself. This is not cheating, for human beings also must have access to a very specialized environment in order to reproduce  a chemical "stockroom" full of air, water, and food containing assorted proteins, fats, carbohydrates, vitamins, minerals, and so forth. Lifeforms, like Universal 'Automatons, cannot produce order out of complete chaos. Rather, each can only transform more simply organized matter into more complexly organized matter.
An example of a modern approach is given by Laing [14] who has investigated two machine reproduction schemes involving complete self inspection. In the first version the original machine is visualized as a onedimensional string of components consisting of a constructor sequence (initially active) and a sequence containing the instructions to build an analyzer (initially passive). The constructor operates on the passive set and builds a working analyzer which, acting independently, produces a complete description of the original machine. Using this description, the original constructor builds an exact duplicate of the original machine, and reproduction is complete. In the second version the original machine consists of two parallel strings, each containing a working constructor and analyzer. The first string to be activated analyzes its passive partner and reconstructs a copy of it; the second string is next activated, resulting in a copy of the first. Thus the system has reproduced itself without recourse to an external self description.
Much has appeared in the literature affirming the possibility of automata replication and development [1527]. Computer programs and numerical patterns that reproduce themselves have been created [2829] and several simple but ingenious physical machines capable of selfreproduction in specialized favourable environments have already been designed, constructed, and successfully operated by Penrose [3032], Jacobson [33] and Morowitz [34]. Other machines have been built over the years which demonstrate the ability to feed, metabolize, learn, respond to stimuli, recognize the self, and move about in physical space with goaloriented behaviour [3538]. Surprisingly, many of these devices are quite modest in complexity, sometimes requiring as few as 30 bits [39] for complete physical description.
A sophisticated self reproducing starprobe must be able to function in highly generalized environments. It will not be able to pick up its parts (or bits of structural information) "free" from the environment, hence it must carry with it much more descriptive data than any replicating machine built to date. But there is little doubt that such a machine can, in theory, be designed.
REPRO is designed from the payload up. To contain the necessary reproductive equipment the payload of the original Daedalus probe was approximately doubled, causing the vehicle mass to increase from 5.45x10^{7} kg to 1.07x10^{8} kg. While Daedalus is a flyby mission, REPRO must slow to a halt in the target solar system so that reproductive activities may commence. The doubleDaedalus configuration must be used to decelerate the 10^{6} kg payload from interstellar cruise velocity down to parking orbit velocity. Hence, to accelerate doubleDaedalus up to interstellar cruise velocity will require a "0th Stage" much larger than Daedalus itself. Using the same vehicle launch mass to payload ratio as the BIS starship, about 100:1 [40], the total fueled mass of REPRO must be 1.07x10^{10} kg.
Assuming humanity is the launching civilization, REPRO is constructed and fueled in Jovian orbit much like Daedalus. After inspection and certification for flight readiness the Oth Stage is ignited, propelling REPRO to about 12%c after a burn time of 4 years. Following a coast phase of 43 years to Barnard's star, the empty Oth Stage hulk is jettisoned and the remaining vehicle structure is rotated 180° so that it points backwards along the direction of flight. Stage I and Stage 2 are ignited in turn, dropping the payload down to interplanetary velocities (<10 km/sec) in about 4 years.
REPRO has a number of subprobes much like Daedalus but these are not constrained to perform simple flyby explorator~ missions since REPRO has fully decelerated. Orbiter, balloon/floater, rocket plane, and even surface lander missions on interesting planets or moons in the target solar system are possible. Detailed planetological data may be accumulated and processed, and a variety of xenobiological investigations under taken in the search for alien life and intelligence [41]. Sophisticated messenger probes of the kind envisioned by Bracewell could eventually be dispatched to parking orbits around selected planetary bodies in the target system.
Theory suggests that most single star systems should be accompanied by at least one jovian planet, possibly more [42]. In order to reproduce itself, REPRO needs 1.01x10^{10} kg of fusion fuel mined from a jovian atmosphere and 5.60x10^{8} kg of nonfuel mass. Upon arrival in the target system, the vehicle uses its remaining 2nd Stage fuel to guide itself into orbit around a small moon of a jovian gas giant. About half the payload, 4.43x10^{5} kg, is designated SEED. SEED is deorbited to the surface of the jovian moon where, over the next 500 years, it builds and launches a number of interplanetary probes. Its primary function, however, is to produce FACTORY, an automated manufacturing complex whose output (following rearrangement of its modular buildingblocks [43]) is exactly one new REPRO every 500 years.
In order to plan the replication of REPRO the material requirements must be determined with some specificity. This requires a detailed knowledge of the structure of the vessel to an accuracy of definition at least as good as the Daedalus vehicle. Unfortunately, no such design study exists for a star ship the size of REPRO, so its ultimate feasibility remains unknown. However, since a fleet of ~200 Daedalus vehicles could place 10^{6} kg of payload in orbit around Barnard's star or one of its planets, then assuming we accept as valid the techniques and conclusions of Project Daedalus the basic plausibility of REPRO cannot seriously be questioned. The technical details are not yet available, but apparently the mass/energy/time requirements are satisfied adequately by the gross physical parameters suggested above for the REPRO vehicle.
To determine an element inventory for REPRO the various Oth Stage component masses were calculated from the basic Daedalus starship model on the reasonable basis of comparative propellant mass ratio. This procedure was deemed acceptable in lieu of a sophisticated 10,000 manhour design study since about 2/3 of the nonfuel Daedalus mass is directly related to the handling of fuel or the distribution of propulsive force. If M_{0}, M_{1} and M_{2} are the REPRO propellant masses for the 0th, 1st and 2nd Stages and C_{0}, C_{1} and C_{2} are specific component masses in the three Stages, respectively, then the scaling for C_{0} given C_{1} and C_{2} is:

(1) 
Taking M_{0} = 10^{10} kg from the previous calculation involving mass ratios, and M_{1} = 9.20x10^{7} kg and M_{2} = 8.16x10^{6} kg as estimates from the figures given by White and Parfitt [44] the above expression reduces to:

(2) 
The estimator given in Eqn. (2) was applied to every major component system in the Daedalus design. The results are in Table 1. No detailed analysis of the technical feasibility of these derived masses was attempted, although a quick check of two particular subsystems (engine reaction chamber, fuel tank mass and refrigeration hardware) indicated substantial agreement with the engineering constraints assumed for Daedalus.
The material requirements for a new REPRO are determined by breaking down the vehicle mass distribution into its elemental constituents. In many cases the precise alloys and materials to be used for a Daedalus component are specified in the summary report [4448], and these values are adopted for REPRO. However, about 55% of nonfuel REPRO mass is unspecified in this fashion. Unspecified mass is treated in two categories: (1) Unidentified nonfuel "Vehicle Other" (labeled V.O. in Table 1), and (2) unidentified SEED or "FACTORY Other" (labeled F.O. in Table 1).
The elemental distribution of V.O. is determined in a similar manner. Each element was carefully considered on the basis of its most common use and classified either as aerospace/electronics/optical Or as nonaerospace/electronics/optical. Elements in the latter category were arbitrarily demoted to 0.1% of their annual US consumption value. Elements in the former category retain their original consumption values, as these substances are more likely to comprise a larger relative fraction of the starprobe vehicle mass. The revised consumption figures were again reduced to fractional abundances by weight for each element.
The final element inventory estimate for REPRO appears in the last columns of Table 2. These totals were obtained by summing the following three quantities: (1) Mass of specific elements specified in the Daedalus report, plus (2) Mass of unspecified "Vehicle Other" (3.08x10^{8}kg) multiplied by each of the V.O. element abundances, plus (3) Mass of unspecified "FACTORY Other" (the 4.43x10^{5} kg SEED) multiplied by each of the F.O. element abundances.
An interesting alternative to jovian propellant acquisitionis the cometary capture technique. In this scheme, REPRO must seek out and rendezvous with a large longperiod (> 100 years) cometary body in a loweccentricity solar orbit. While cometary nuclei are believed to average 110 km in diameter, a few exceptionally large objects (P/SchwassmannWachmann I, Haley's Comet) are known to have diameters from 50  100 km and masses from 10^{17}10^{18} kg. The exact chemical composition is unknown but is thought to represent original. condensate from the protostellar nebula at roughly "cosmic" abundances, perhaps with a slight depletion of volatiles. The main body of the nucleus may consist of perhaps 74% ices (water, ammonia, methane, and solid clathrate hydrates) and 26% stonyiron dust, for a net density of about 1100 kg/m^{3} [52]. The cosmic abundance by mass of He^{3} may be computed from Cameron [53] as 2.67x10^{5} and as 2.50x10^{5} for deuterium. Even assuming the helium isotope is depleted by as much as three orders of magnitude, a large comet might still contain 10^{10} kg of He^{3} and 10^{13} kg of D, enough to permit reproduction of at least one REPRO vehicle. Further comparison of nonfuel materials requirements with cometary "cosmic" abundances indicates that all needs may be satisfied by completely cannibalizing one large body of the kind discussed above.
It seems likely that there are two main cometary belts, an Inner Cloud within 40 AU and the Oort Cloud out to about 100,000 AU. Recent estimates place the total population Of the Inner Cloud at about 3x10^{6} bodies [54]. These remain in relatively circular orbits until deflected into the inner Solar System by planetary gravitational perturbations. Orbital inclinations of Inner Cloud members are random [55], so starprobes arriving along arbitrary trajectories should be able to find suitable candidates reasonably close to the plane of entry. Assuming one comet of every thousand is sufficiently large for utilization by a reproductive probe, then about 3000 suitable objects exist within a heliocentric search volume of radius 40 AU. If REPRO must locate at least one large comet within 5 years of target system entry, starprobe sensors must be capable of scanning 10^{6} AU^{3} sec^{1}, necessitating a field diameter of at least 1 arcsec scanning to a depth of 40 AU with an integration time of 100 seconds. Within this range a 100kmwide comet with albedo 0.1 has an apparent magnitude of +23 or brighter, observable using a 2metre search telescope somewhat smaller than NASÁs 2.4metre Space Telescope [56] or Daedalus's 5metre optical telescopes [41].
Once it has been decided that the bulk of the mass of REPRO must be drawn from a jovian planet in the target star system, efficiency dictates that the base site be located somewhere in jovian orbit. Heavy elements comprising the nonfuel mass can most easily be drawn from one of four sources  atmosphere, jovian moon, jovian trojan asteroids, or the asteroid belt [52, 5758].
Element abundances for "average asteroidal" material were assumed to include 74.7% silicate, 19.6% nickeliron, and 5.7% troilite after Mason [59], and appear in Table 3. Data for the rare gases AT, He, Kr, Ne, and Xe are estimated from Heymann [60], and the value for H is averaged from data in Heide [61] and Brown [62]. A recent study of four large Trojan asteroids of Jupiter and two of the outer Jovian moons Himalia (J6) and Elara (J7) indicates an abnormally low surface albedo a result which calls into question the validity of the current "ice model" for these bodies [63]. Although the blackening may be only a surface effect, the highly irregular shape of Hektor, the largest Trojan body, implies that it cannot be composed of solid ice because ice has insufficient structural strength to support the irregular shape against collapse. Hartmann [64] calculates that if Hektor is composed of chondritic material it should be stable against internal crushing. If this conclusion is applicable to Himalia and Elara, smaller bodies but similarly low in albedo, then asteroidal capture becomes a more likely explanation for the origin of the outer Jovian moons.
Jovian atmospheric heavy element abundances generally fall about two or three orders below "average asteroidal, " assuming "cosmic" abundances [53] for the gas giant. (See Table 3). There are only eight elements for which jovian abundance may be higher than the asteroidal; Ar, C, H, He, Kr, N, Ne Xe. Extraction of these elements thus should be delegated to the aerostat balloon factory system.
Transportation out to the asteroidal belt or to the trojan clusters from the target system jovian is far more expensive if time and fuel than transportation of an equal mass to an outer jovian moon. In our Solar System, there are four known such moons orbiting at a distance of about 1.1x10^{7} km from Jupiter with diameters 170 km (J6), 80 km (J7), 40 km (J10), and 20 km (J13); another five moons orbit about 2.2x10^{7} km from Jupiter, with diameters 50 km (J8, J11), 40 km (J9), 30 km (J12), and 20 km (J14).
In this paper it is assumed that a jovian moon about 100 km in diameter having "average asteroidal" composition can be located and colonized by REPRO.
Upon arrival in jovian lunar orbit, the first step in the reproductive process is to deorbit SEED, all remaining functioning wardens, most of the 412 MW Stage 2 power supply, and other miscellaneous equipment totalling ~5x10^{5} kg to the surface of the jovian moon. In the weak 7 milligee gravity field this manoeuvre should require 290 MJ from an orbit 100 km above the surface. an eight hour trip using a 10 kW propulsion system. This is a deceleration of 4 milligees, well below the 13113 milligee design limit [44] for all powered flight phases of the Daedalus mission.
Following moonfall, SEED and its wardens unpack and proceed slowly to erect FACTORY during the next 500 years. SEED operates under the direction of its own autonomous computer system which is activated upon landing, in coordination with the flight computer which remains in orbit aboard the 2nd Stage hulk. The 4.43x10^{5} kg SEED grows into a 3.46x10^{8} kg FACTORY, which is reorganized, then produces a 1.07x10^{10} kg REPRO starprobe during the second 500year period. The mass growth rates are 1.3%, and 0.690%, respectively, fairly modest for modern automated industrial systems and an order lower than typical rates for biological organisms.
The actual reproductive apparatus consists of 13 distinct robot species, including Chemists, Aerostats, Miners, Metallurgists, Computers, Fabricators, Assemblers, Warehousers, Crawlers, Tankers, Wardens, Verifiers and Power Plants. These collectively perform all of the functions of a living system [65]. SEED carries representatives of only the first nine species mentioned. Several aging wardens are provided "free" from the derelict Stage 2, together with its 412 MW power supply. All Tankers and Verifiers must be constructed in situ from scratch.
A subsidiary function of Chemists is to combine various elements and simple molecules into perhaps 100 basic reagents needed for SEED and FACTORY extraction, manufacturing and testing processes. Johnson and Holbrow [69] describe an aluminum processing plant for which 13% of the total refinery mass consists of five nonelemental process chemicals (H_{2}O, H_{2}SO_{4}, Na_{2}SO_{4}, NaCl, and LiCl. Preparation of industrial process chemicals generally is quite easy in comparison with element extractions from ores. Most of these substances, both inorganic and organic, are one or two step preparations, so the additional mass of SEED and FACTORY Chemists to perform this function should again represent only a few per cent of the total Chemist mass.
FACTORY must produce essentially all nonfuel REPRO mass in 500 years. The mass flow requirements for each element, based on the REPRO inventory in Table 2, are divided by the jovian moon elemental abundances in Table 3 and the total production time of 500 years to obtain the lunar ore throughput rates tabulated in Table 3. The total mass flow rate is 3860 kg/sec, so in 500 years FACTORY Chemists must process 6.06x10^{13} kg of jovian moon ore.
Fusion breeding [7071] of the most troublesome elements from more abundant precursors was considered in an attempt to reduce Chemist mass. flow rates. Plausible breeding pathways were found for the production of Mo from Zr (4.24x10^{2} mole/sec), Rh from Ru (3.68x10^{4} mole/sec), W from Hf (8.14x10^{5} mole/sec), Au from Pt (3.88x10^{5} mole/sec), Ta from Hf (9.26x10^{6} mole/sec), In from Cd (1.88x10^{7} mole/sec), Cu from Fe (5.88x10^{3} mole/sec), Ti from Ca (5.56x10^{2} mole/sec), and Al from Mg (8.57x10^{1} mole/sec), for a total of 9.61x10^{1} mole/sec breeder throughput. Utilization of all nine processes reduces the mass flow rate from 3860 kg/sec to 1110 kg/sec of jovian moon ore.
Setting aside technological difficulty, the main problem appears to be energy. Each of the proposed breeding pathways requires each precursor atom to absorb 27 neutrons to achieve majority conversion, coupled with continuous extraction of desired isotopic species to prevent degradation. If on average 100 2.45MeV neutrons must be liberated by a DD fusion reactor to transmute one precursor atom into one rare atom (27% transmutation efficiency), then to breed all nine elements requires a power supply of 2.27x10^{13} watts. Even assuming 10,000 watts/kg for the higher temperature DD reactor the mass of this supply is 2.27x10^{9} kg, nearly seven times the mass of FACTORY. It is concluded that fusion breeding is not feasible within REPRO mission constraints.
Current ore processing technology for aluminum, REPRO's mostneeded element, permits a 4750ton refinery to produce 1.74 kg/sec Al metal using soil from Earth's Moon [69]. This material is only 5.6%, A] by weight so the factory must process 31 kg/sec soil to achieve the stated output. This is a net processing rate of 6.53x10^{6} (kg ore/sec)/kg refinery. Assuming one order improvement from technological advances, then if the values given for the aluminum plant are typical for element extractions FACTORY may be able to process jovian moon ore at the rate of 6.53x10^{5} (kg/sec)/kg. This gives a total FACTORY Chemist mass of (3860)/(6.53x10^{5}) = 5.91x10^{7} kg, equivalent to about 124 individual 475ton refineries. The power requirement for the original aluminum plant is 40 watts/kg; adopting this value, the total for FACTORY Chemists is 2360 MW.
SEED Chemists must process enough jovian moon ore to produce the entire FACTORY mass of 3.46x10^{8} kg. However, replacement mass must be added to take account of break downs and repairs. Daedalus carries 20 tons of repair shops and spare parts to service a 450ton payload for 50 years, or 8.9x10^{4} (kg spares/kg serviced)/year. The REPRObuilding phase lasts 500 years, so the extra mass of shops and spares needed to maintain FACTORY is estimated as (8.9x10^{4})(500)(3.46x10^{8})=1.54x10^{8} kg. The FACTORYbuilding phase also lasts 500 years; the logarithmic average mass of the growing SEED during this time is 1.24x10^{7} kg, so the extra mass of shops and spares needed to maintain SEED is estimated as 5.52x10^{6} kg by the same method as above. Including main tenance support, SEED Chemists must produce a total FACTORY mass M_{F} = 5.06x108.
It will be noted that the function of FACTORY is to produce REPRO, hence production of REPRO must be linear; on the other hand, since the function of SEED is to produce FACTORY, and since FACTORY consists of "universal" modular components of the same types used in SEED, then in effect SEED is building more SEED, hence production of FACTORY must be exponential. If M(t) is the mass of the growing SEED at time t, then the rate of growth dM(t)/dt = XM(t), where X is the mass of new SEED generated per unit time interval, per unit mass of existing SEED at time t. The equation of growth may be written as:

(3) 
where M_{s} is the mass of the Original SEED at t = 0.
The value of X and the FACTORY mass M_{f} are determined as follows. From Eqn. (3) we have M(t) =M_{s} · exp(Xt) = M_{t}, where t is the 500year production time of FACTORY. Also, X may be fixed by assuming that the SEED mass at t = t, it is not reorganized into FACTORY hardware capable of pro ducing REPRO, but rather retains its structure as fully mature SEED. This nonreconfigured SEED can produce more SEED components instead of new REPRO mass at a rate of M_{t}/t kg/sec, where M_{t} is nonfuel REPRO mass. However, this fraction must be corrected for the difference in element abun dances between the output of FACTORY (new REPRO, with REPRO abundances) and the output of the mature SEED (more SEED, with F.O. abundances). To obtain a relation for X, the above production rate must be multiplied by the ratio of the appropriate output abundances relative to pvian moon soil, and divided by the mass of existing SEED at t = t. Hence, X = (M_{t}/t)(R_{t}/R_{f})/M_{f}, where R_{t} = S{(REPRO abundances)/(jovian moon abundances)} and R_{f} = S {((SEED or F.O. abundances)/(jovian moon abundances)}. Since M_{t} (5.60x10^{8} kg), M_{s} (4.43x10^{5} kg), and t (1.57x10^{10} sec) are flyby mission parameters, and Rg (Rg(1.10x10^{5}) an R_{f}(2.68x10^{4})2.68x10^{4}) are readily calculated from the data in Tables 1, 2, and 3, this leaves two equations and two unknowns. Solving simultaneously, M_{f} = 3.46x10^{8} kg and X = 6.66/t.
The same value of the rate constant X may be used to calculate the individual SEED mass of each exponentially growing component robot species. The most general definitionof the component rate constant X. is mass flow output 01 component per unit mass of existing, component, so X_{c} = (M_{t}t) (R_{t}/R_{f})(M_{fc}/M_{f})/M_{fc} where M_{fc} is the mass of the particular component in FACTORY. The dependence on M_{fc} drops out so X_{c} = X for all components. Since SEED components must produce M_{F}rather than Mf, SEED component mass Ms is given by:

(4) 
Plugging in M_{fc} = 5.91x10^{7} kg, from Eqn. 4, M_{sc} =1.11x10^{5} kg for SEED Chemists. The initial power requirement is 4.4 MW.
Aerostats designed for operation on Jupiter for Project Daedalus have an empty mass of 180.2 tons, including ascent ferry mass, and consume 10 MW of power, all of it generated by the onboard reactor. Each factory can process 680 kg/sec of gas giant atmosphere, and is designed for fully automatic operation in the highradiation, turbulent jovian environment.
From the jovian throughput rates in Table 3 it is clear that the mass flow for He^{3} is controlling. A total Aerostat fleet of (14,500)/(680) = 21.3 units must on average operate continuously for 500 years to provide all the He^{3} and D propellants for a single REPRO starprobe. Although Parkinson suggests a lifespan minimum of 8 years, industrial factories generally are assigned a maximum working life of 30 years. If the average of these is taken as typical for Aerostats, 19 years, then in 500 years of FACTORY operation the entire Aerostat fleet must be replaced 26.3 times. A total mission fleet of 560 units must therefore be constructed in order to maintain an average working fleet of 21.3 units in continuous operational status. The total mass of FACTORY Aerostats is 1.01x10^{8} kg.
Parkinson has suggested that workable devices might be designed on a somewhat smaller scale. If SEED carries two small Aerostats (each one 15% normal mass) to gather fuel and other isotopes during the FACTORY building phase, then the total mass of SEED Aerostats is 54.1 tons.
One minor difference between Daedalus and REPRO Aerostats is the incorporation in the latter of additional facilities for the extraction of the elements Ar, C, Kr, NN, Ne, and Xe because of their expected greater abundance in the jovian atmosphere than on the jovian moon. The extra production requires the continuous service of only about 10^{4} Aerostat, a negligible additional production load for the FACTORY fleet. All six elements are readily obtained using simple cold traps, since the extraction processes for recovering them need an efficiency of only ~10^{6} compared to He^{3} and D processors. Each full load of fusion fuel, 16.8 tons, thus is accompanied by ~260 gm of the six specified elements.
The Miners are semiintelligent general purpose excavation robots, charged with the responsibility of locating and digging up jovian moon material for processing by Chemists. Physical functions include drilling, blasting, dozing, shoveling, stripping, and lifting large quantities of ore. Search functions are highly specialized, confined to locating surface ore veins and local pockets of mineralogical enrichment (survey satellites do most of the finding  see Section 6.5).
FACTORY Chemists must process 3860 kg/sec of jovian moon ore, so the Miners must dig it up. Typical performance figures for power shovels, bulldozers and shovel dozers are (3x10^{3} (kg excavated/sec)/kg machine and 4 watts/kg machine during continuous operation [72]. These values, derived from experience working in the 1 gee terrestrial environment. may be somewhat pessimistic when directly applied to the 7 milligee jovian moon environment, but hopefully not unduly so. If they are taken as appropriate, with a duty cycle of 50% to account for delays due to the finding function, then the total mass of FACTORY Miners is 2.57x10^{6} kg with a power consumption of 10 MW.
SEED Miners must locate and excavate enough jovian lunar ore to provide the entire mass of FACTORY plus shops and spares. Using Eqn. (4), for M_{fc} = 2.57x10^{6} kg M_{sc} = 4820 kg with an initial power requirement of 19.3 kW.
FACTORY Metallurgists must produce M_{t}/t = 3.57x10^{2} kg/sec in order to complete REPRO on schedule. Taking the glass refinery figures as typical, the total mass of FACTORY Metallurgists would be 29.7 tons. However, a subspecies of Metallurgist robots handles electronicgade materials of the highest purity, performing specialized functions such as ultrapure crystal growth, bulk wafer fabrication, and selected doping and drifting prelithographic operations. Although the mass flow is low, machine complexity is high because of the extreme standards of purity. If electronics Metallurgists also have a mass of 29.7 tons, then the total mass of FACTORY Metallurgists is 59.4 tons with a power consumption ~130 kW.
SEED Metallurgists must process the entire mass of FACTORY plus shops and spares. From Eqn. (4), M.= I I I kg; this is probably below the minimum feasible mass for a working system of Metallurgist robots using presentday technology, so Msc is arbitrarily taken to equal 1000kg with an initial power requirement of 2.2 kW. With more than an order of magnitude extra starting mass, SEED Metallurgists should be able to complete their tasks in only 335 years, well ahead of schedule.
The computer system designed for Daedalus is extremely complex. It must be able to engage in autonomous activity during a 50100 year mission, varying the operational goals embedded in its own software and designed in black box modular form to permit a "repairbyrepair" strategy [74]. It must also be capable of heuristic functions, including lateral thinking and intuition.
FACTORY and REPRO Computers are at least an order of magnitude more complicated. They must guide a vehicle equal in mass to ~200 Daedalus vessels to another star. Then, rather than a comparatively simple flyby manoeuvre, they must select an appropriate jovian moon by remote sensing and enter orbit around it. Once SEED deorbits and its Computers are activated, it must erect virtually from scratch an automated factory system over the course of 500 years, and then operate it successfully for at least another 500 years. All aspects of production, scheduling, operations, repairs and inspections must be coordinated by the Computers, as well as various peripheral functions including target system reconnaissance, transmission of updated status reports back to the home planet .or sending civilization, and various responses to emergencies and unexpected events. Computers must supervise the launch and control of a network of remote sensing survey satellites placed in low orbit around the jovian moon to help map the terrain and to scout ore deposits to evaluate mineral reserves.
The Daedalus computers have a mass of 10 tons and service a payload of 450 tons, with a power consumption of 1 kW. If FACTORY and REPRO Computers must be an order more complex, if we assume one order technological improvement this gives 2.2x10^{2} (kg Computer)/kg serviced) and 0.1 watts/kg Computer. Hence, to service the entire FACTORY complex requires a total FACTORY Computer mass of 7.61x10^{1} kg, using 761 kW.
Applying the same ratio to the mass of SEED gives a Computer mass of 9.75 tons. However, this ignores the Caches, a feature unique to REPRO. SEED carries a total of six Memory Caches, each permanently impressed with a readonly description of the entire starprobe together with the appropriate instructions and algorithms for building it, SEED, and the FACTORY complex. A typical value for the information content of automobiles, spacecraft, and other machines is ~10^{4} bits/kg as compared to ~10^{6}' bits/kg for encyclopedias and ~10^{9} bits/kg for moving surface magnetic data storage devices (disks, tapes, drums). The sum of nonfuel REPRO mass and FACTORY mass is 9.06x10^{8} using the ~10^{4} bits/kg estimate, each Cache must hold 9.06x10^{12} bits. Six moving surface magnetic storage Memory Caches thus have a mass of 54.4 tons, so the total mass of SEED Computers is 64.1 tons, with power consumption 6.41 kW.
Three Caches remain near the FACTORY site, physically separated but datalinked to SEED and FACTORY Computers. These Caches are activated only for very brief periods to transfer large blocks of information into the data banks of the main Computers. Two Caches are selected at random for information transmission in parallel; in case of disagreement the third Cache is polled to break the tie vote. The three remaining Memory Caches are carried by Crawlers to the opposite side of the jovian moon and buried in small individual rock vaults so the data cannot be lost even if a major catastrophe befalls the FACTORY. Before a new REPRO is launched, six clean Caches stored in its SEED are impressed with Cache data polled from at least three of the original memory stores. The probability that the same memory position is damaged in at least two original Caches (via cosmic rays, internal radioactivity, thermal fluctuations, etc.) is exceedingly small. Much like von Neumann, tessellation model, no Cache contains a full description of itself but each contains instructions which permit a Computer to copy the contents of a Cache onto a clean unit. Data movement is rapid  moving surface memories can transfer up to 10' bits/sec, so the contents of an entire Cache may be transcribed in ~250 hours.
Fabricator robots convert raw material stocks produced by the Metallurgists into working parts of all description, from hull sidings and doublewalled reaction chambers to threaded bolts and semiconductor chips. These parts are then passed to the Assemblers, who assemble the parts into the working black box modules which comprise most of FACTORY, and later into the completed sections of the REPRO starprobe itself. The functions of Fabricators and Assemblers are closely related, hence. will be discussed together. More significant is the type of fabrication and assembly labour involved, whether bulk material or electronics material.
Johnson and Holbrow [69] give 50 tons/personyear as the typical productivity in bulk processing and heavy industries on Earth. If each person can be replaced by a semiintelligent 1000 kg robot, the mass flow for bulk Fabricators and Assemblers may be about 1.6x10^{6} (kg/sec)/kg robot. Human beings have a specific power consumption of about 2 watts/kg. If robots can be designed which are 10% as energy efficient, bulk processor robots may consume 20 watts/kg.
Estimates from Oldham [75] suggest that a typical value for the production of microelectronic chips in average size semiautomated factories is ~10^{6} (kg/sec)/person. Again assuming each person is replaced by a 1000 kg robot, the mass flow for electronics Fabricators is ~10^{9} (kg/sec)/kg robot. Assembly of chips into functional circuitry should proceed somewhat faster, but testing and debugging time may offset this gain since a circuit of chips has more states than any individual chip and thus more ways to go wrong. The value of ~10^{9} (kg/sec)/kg is adopted for electronics Assemblers as well. A power requirement of 20 watts/kg is assumed for both.
FACTORY bulk Fabricators and Assemblers must process M_{t}/t = 3.57x10^{2} kg/sec, so total FACTORY mass of each robot species is 22.3 tons. To produce the REPRO Computer in 500 years requires FACTORY electronic Fabricators and Assemblers to produce the REPRO Computer, its included SEED Computer, and siz Caches in a time t, a rate of 4.72x10^{6} kg/sec This requires 4.72 tons of electronics robots. so the FACTORY total mass of Fabricators or Assemblers is 27.0 tons, consuming 540 kW of power each.
SEED bulk Fabricators and Assemblers must produce M_{cf} kg of FACTORY mass exponentially in 500 years; from Eqn. (4), M_{sc} = 50.6 kg. If a minimum feasible working mass of M_{sc} = 1000 kg is arbitrarily adopted these robots can finish their jobs in only 276 years. However, a FACTORY electronics mass equal to the FACTORY Computer plus shops and spares must also be produced by the electronics Fabricators and Assemblers of SEED. This presents a problem, since it represents a rate of 1.11x10^{7}/t = 7.09x10^{4} kg/sec which implies a SEED mass of Fabricators or Assemblers of 7.09x10^{5} kg. nearly twice the mass of SEED.
The strategy used to overcome this difficulty relies upon the modular character of FACTORY robot species. Bulk Fabricators and electronics Fabricators, in other words, are made of much the same basic components. To solve the problem posed above, the allotted electronics Fabricator/ Assembler mass is divided into two parts. The first part consists of extra bulk processing robots which proceed to build only more of their own components for a time T, after which this initial mass plus the new exponentially created mass is reorganized as electronics Fabricator/Assembler robots to begin contributing to FACTORY Computer mass. The second part produces Computers continuously from t = 0. Lin ear optimization calculations indicate that minimum SEED mass is achieved when the two parts are equal in mass. In this case, T = 267 years and each part has a minimum mass of 40.7 tons. Adding the 1 ton for bulk processors, SEED Fabricato, or Assembler mass is 82.5 tons, requiring 1.65 MW initially.
Warehousers are designed to contain ~10^{8} kg nonfuel material and ~10^{10} kg fusion fuel for REPRO. The nonfuel mass may be contained within a FACTORY Warehouse' mass of 10^{6} kg, if the ratio of structure volume to contained volume is 10^{2} and the densities of the structural material and the stored matter are approximately equal. From Eqn. (2) the mass of the refrigeration equipment required to maintain III,: He^{3} and D fuel pellets at the proper temperature is 3.68 x10^{4} kg. the value of 0.5 kg/Kw for refrigeration hardware give" by White and Parfitt [4], the power required by Warehouse" to perform the cooling function is 73.5 MW. Hence, the maximum mass of FACTORY Warehousers is 1.04x10^{6} kg with a consumption of at most 2080 MW.
SEED Warehousers initially must store ~5x10^{5} kg deorbited with SEED, which should require a mass of no more than 5 tons since little of the contained mass is fusion fuel requiring careful refrigeration. The maximum power draw is 10 MW.
To dig out the necessary ore the Miners must carve an annular pit 20 metres deep and at least 20 km in radius on the jovian lunar surface around the base site. If the mean oneway trip distance for a Crawler is 15 km at a speed of 10 km. /hr, the mean travel time is 5400 see so the fleet load carried each trip is 4.17x10^{7} kg. A properly designed carrier vehicle easily can carry its own weight in a 1gee gravity field. In a 7 milligee field, Crawlers should be able to haul (7x10^{3})^{1} = 143 kg/kg, so the fleet mass of FACTORY Crawlers is 292 tons. A typical power consumption figure for surface vehicles on Earth is 20 (joules/metre)/kg, direct application of which gives a FACTORY Crawler fleet requirement of 16.2 MW.
To carry the entire deorbited SEED mass of ~5x10^{5} kg requires a mass of SEED Crawlers of 3.5 tons, with an initial power requirement of 194 kW.
One Tanker arrives at the jovian every 6.44 days. The ascent ferries from all Aerostats arrive closely spaced in time to facilitate rapid cargo transfer. A total loading and unloading time of 10^{2} year is allowed each Tanker per circuit. If the mass of the fueled but unloaded Tanker is twice the mass of the load it may carry, then the total fleet mass of FACTORY Tankers is 7.16x10^{7} kg. On a single round trip lasting 1.77 years, each Tanker consumes a total of 2.5GM_{j}m_{l}(R_{1}^{1}R_{2}^{1}) = 1.61x10^{15} joules, about 33 kg of fusion fuel. The mean power output is 29 MW continuous and the mean acceleration 0.14 milligees, a power/thrust ratio within the limits of current technology (76) although fusion propulsion engines have not yet been constructed.
SEED carries no Tankers.
As described in the Project Daedalus final report, the Wardens are a manipulator subsystem of the Computers [4], 77]. Their duties include maintenance and repair as well as rearrangement and restructuring of instruments and ship components. Wardens have a high degree of operational autonomy and mobility. Each 5ton unit deployed by REPRO on the jovian lunar surface can transport ~ 10^{8} kg from surface to orbit on a single charge of 7500 kg propellants (assumed to deliver 1.34x10^{7} joules/kg, the value for LH_{2}/LO_{2} combustion).
The original Daedalus design calls for two Wardens to service a 901ton payload plus2ndStage over a nominal 50year mission. The equivalent REPRO mass (doubleDaedalus) is about 119 times greater over the same flight time, which would seem to demand 238 Wardens to achieve the specified Daedalus reliability figure of 99.99% [77]. However, if a reliability of 99.9% is acceptable during the 50year flight portion of the mission (a mere 5% of each reproductive cycle) then REPRO can get by with as few as 10 operational Wardens. The 99.99% reliability is retained for the remaining 95% of the reproductive cycle.
If two Wardens are required to service a 901ton mass for 50 years to achieve 99.99% reliability, then 7700 Wardens are required to service FACTORY for 500 years with 99.99% reliability, a total mass of FACTORY Wardens of 3.85x10^{7} kg. Most of these, perhaps 99%, are ground units travelling on tracks or wheels, drawing power from onboard fuel cells energized by electrical energy provided by FACTORY Power Plants. Each Daedalus Warden has a total propellant energy resource of 1.01x10^{11} joules expended during 50 years of service, or 64 watts/Warden. (This is comparable to the estimated 350500 watts required for human EVA during the Apollo missions [78].) A FACTORY Warden fleet of (99%) (7700) units thus requires 4.88 kW of power. The remaining 1% of the Warden fleet must have space manoeuvring capabilities, which requires an adequate supply of propellants. Each Daedalus Warden carries 7500 kg fuel for a 50 year service lifetime, so a fleet of (1%) (7700) units needs 578 tons of Warden propellants. This subfleet is capable of hauling ~10^{10} kg into jovian lunar orbit (100 km altitude), which is approximately the mass of REPRO.
Since their primary function is the gathering of information rather than bulk processing of matterenergy, the FACTORY Verifier fleet should have a fairly low mass. A firstcut estimate is 1%, of the Warden fleet, about 3.91x10^{5} kg. Total power requirement is about 4.88 kW.
Table 4 gives total mass and power requirements for each major robot system comprising SEED and FACTORY. A power supply of 4470 MW is necessary for minimum FACTORY operation; a 5000 MW Power Plant is assumed, with 11% of the energy uncommitted as a generous allowance for losses, accidents, special uses, and unforeseen circumstances.
The fission power plants of Daedalus are replaced by compact, hightechnology fusion reactors in REPRO, SEED and FACTORY. The original Stage 1 and 2 thermonuclear propulsion systems designed for Daedalus require reaction chambers with power densities of 31,000 watts/kg and 50,800 watts/kg, respectively [45]; the Daedalus Ist Stage power supply, using lessefficient fission fuel, achieves a power density of 14,300 watts/kg less shielding and about 1000 watts/kg for the total shielded system using current technology [46]. Allowing for one order of magnitude improvement it is not unreasonable to expect that fusion power supplies eventually may be built with a power density of ~10,000 watts/kg. Using this figure, the mass of FACTORY Power Plant is 500 tons. Allowing an additional 10% for transmission lines and other peripherals raises the total to 5.5x10^{5} kg.
The large uncommitted share (95%) of the SEED Power Plant permits maximum deferral of FACTORY Power Plant construction so that early priority may be given to other robot component systems.
A simple extension of the basic REPRO mission is the possibility that FACTORY might continue to produce copies of REPRO at the steady rate of one every 500 years. This could continue until irreversible errors and defects reached a lethal level. It is conceivable that with suitable planning, error correction and verification procedures the lifetime of a FACTORY might be extended from 1000 years to perhaps 5000 years, permitting a total of 9 REPRO starships to be born and launched to neighbouring solar systems.
It is also possible to imagine that the data stored in the ultrasecure Cache memories could be used to entirely rebuild a FACTORY once its defects have reached the lethality stage. This approach is conceptually equivalent to a scenario in which the last viable REPRO built by the aging FACTORY is retargeted to land back at the site of its parent, using its SEED to build a new FACTORY either by overhauling or by cannibalizing the remains of the old (as a supplement to normal reproductive activities). Such a mission would be limited in the number of REPRO starcraft it might ultimately produce by only two major factors: (1) The total accessib4e mass on the jovian moon, enough for ~40,000 REPROs assuming a 100 kmdiameter body; and (2) the maximum number of viable generations permitted the lineal descendants of the original REPRO, depending on the stability of the Cache data stores (i.e., each FACTORY could regenerate itself for as many generations as remained to its offspring). Viability may be enhanced by multiplying the number of Caches.
Another interesting possibility is that SEED could simply be allowed to grow for, say, 2t rather than only t to take even better advantage of exponential growth. Maturation for 2t would result in a FACTORY of mass 2.70x10^{11} kg, capable in theory of churning out REPROs at the astonishing rate of three every two years. Maturation for 3t would produce a FACTORY so large that its production of REPROs would consume the resources of the entire jovian moon in only 33 years.
REPRO could also be tested in our own Solar System before we launch one to the stars. The advantages are threefold, First. we get a free test, a full dress rehearsal of the most important part of the mission, and an opportunity to debug the design. Second, we get a free source of REPRO starprobes, assuming multiple reproduction is feasible as suggested above. Third, our total initial monetary and material investment in the project is just one SEED, far less even than the price of a single complete starship. The principal disadvantages of this approach are that in testing the probe our Solar System gets cannibalized (why not let some uninhabited star system bear this burden?), and also that we must wait ~10^{3} years for testing to be completed and the first "free" offspring to leave the Solar System. Humanity today may balk at these difficulties, but the costs and delays may prove acceptable to an extraterrestrial civilization with a serious longterm commitment to the exploration of their galaxy.
To augment economic viability, smaller self reproducing probes may be conceived. For example, one simple alternative is to redesign REPRO with an element distribution more closely matched to that expected to be found on the jovian moon. The abundance ratios calculated in Section 6.1 are a measure of the "exotic" character of the structural materials as compared with their source ore. Thus REPRO, with R_{t} = 1.10x10^{5}, is about four times "more exotic" an extraction from jovian lunar soil than FACTORY, with R_{f} = 2.68x10^{4}. Engineers frequently optimize for low cost, light weight, reliability, maintainability, or a hundred other factors, so there is no reason why elemental distribution could not be treated similarly. If the extraction ratios R_{t} and R_{f} could be designed down as low as 10^{2}10^{3}, the mass or replication time (and hence the cost) of REPRO would drop an order of magnitude or more.
Another remote possibility for downscaling involves using the Daedalus vehicle itself without any modifications. In this scheme, a maximum of 200 tons is set apart from the subprobe payload mass, equipped with a fusion engine with a 100:1 vehicle launch mass to payload ratio similar to Daedalus, and fired backwards at exactly 12.2%c as Daedalus sweeps through the target solar system on its normal flyby mission. Correctly executed, this manoeuvre would suffice to place a payload of 2000 kg in orbit around any planet in the star system. Forward (79) has suggested that a 100 kg payload at relative rest might be sufficient for nearinterstellar missions of exploration, so it is certainly plausible to view Daedalus as a prototype delivery vehicle for Bracewell probes. However, a successful SEED design in under 2000 kg would require micromechanical engineering techniques more than an order of magnitude beyond current technology. For instance, preliminary estimates indicate that the total mass of SEED Chemists would be ~400 kg; if 124 individual refineries are required this gives 3.2 kg/refinery. We cannot say this is an impossible figure to reach; however, it lies. many orders beyond the technology otherwise assumed for Daedalus and for REPRO.
The situation improves markedly if provision is made for sexual, rather than asexual, reproduction. This alteration of the basic REPRO replication strategy permits the acquisition of variation in parallel, not serially as in the original starprobe design. Assuming a sufficiently dense population, significant evolution could occur in as few as 10^{3}10^{4} generations (80), or ~10^{6}10^{7} years. This is of the order of the exploration time of the Galaxy, enough time for machine speciation to occur. Niche specialization is plausible and there is a remote possibility that a simple machine ecology might have time to arise, complete with predators and prey. Sexual starprobe designs may be imagined as REPRO vehicles preprogrammed for target star overlap every few generations. The usual reproduction scenario might then include two starprobes landing on opposite sides of the same jovian moon and jointly engaging in the construction of two FACTORY complexes and two REPRO offspring. Memory Caches could be compared, evaluated and edited "consciously," offering the exciting possibility of yet faster development by means of intelligent participative evolution.
On a strictly material basis, the presence of a single REPRO in a star system represents a de minimis mass loss to the native inhabitants thereof. A typical jovian atmosphere will contain enough fusion fuel to fill ~10^{13} self reproducing starships, and a single large jovian moon (100 km in diameter) may contain sufficient molybdenum to construct ~10^{5} REPRO machines. Even if 10100 offspring are generated by each FACTORY the mass loss is negligible. Nevertheless, it is highly unlikely that humanity would take kindly to an alien starcraft landing on one of the Jovian moons Himalia or Elara and reproducing itself there without at least first asking our permission. Probably we would regard it as one of Dyson's "technological cancers loose in the Galaxy" and attempt to destroy or disable it.
The theory of interstellar colonization ethics is discussed elsewhere in greater detail [8185].
The author wishes to thank Dr. Francisco Valdes, for helpful discussions and criticism of the manuscript, and Dr. Ronald Bracewell, whose scepticism in discussions with the author provided the initial impetus for the present paper.