Gregory Eckersley and Claude — June 2026
This document is a technical appendix to the Space Accord governance work published at spaceaccord.org. It exists to answer a specific question: is the capability that the governance argument is about actually feasible?
The answer developed here is yes — more feasibly than intuition suggests, and sooner than most people expect.
The core of the argument can be stated simply. A small self-replicating robot with communication and transport capability can be established on a well-chosen near-Earth asteroid from a seed payload of modest mass, using locally available materials and processes that exploit rather than resist the space environment. The combination of semiconductor and other electronics enables that capability. The open question is not whether it will happen, but why — under what governance, directed toward what ends, for whose benefit.
That question is what spaceaccord addresses. This document provides the technical grounding that makes the question urgent rather than premature.
The concept was developed through extended human-AI collaboration — Gregory Eckersley directing the investigation, Claude providing synthesis across a wide range of disciplines. A differently configured AI working from the same summary would likely arrive at a similar or better implementation by a different path. That is itself a demonstration of how quickly the landscape is changing, and of why the governance conversation cannot wait.
The document is honest about what remains open. The architecture is sound. The details will be refined by experiment, by specialist review, and by the accumulated work of people who engage with it seriously. Contributions and corrections are welcome.
Gregory Eckersley — spaceaccord.org June 2026
This document describes a concept for a small, modular, self-bootstrapping fabricator intended for deployment on a suitable near-Earth asteroid. Starting from a compact seed payload of brought components and using locally available asteroid materials, the system is designed to progressively close its dependencies on Earth resupply through a generational sequence of increasing capability. The central argument is that the space environment, properly understood, is not hostile to manufacturing — it is a different environment that rewards different approaches, and in several respects is actively superior to terrestrial conditions for the processes involved. The concept is physically grounded, internally consistent, and honest about what remains open. No step requires a capability that does not exist or a material that is not present. The irreducible import list at maturity is small: GaN laser diodes for optical communication, fluorine-bearing materials used sparingly, and silicon devices until local semiconductor fabrication closes that loop. The concept is optimistic not because the engineering is trivial, but because it stands on an immense foundation of prior human achievement — and because it is designed from the outset to exploit its environment rather than fight it.
Any manufacturing capability deployed in a remote environment faces a fundamental chicken-and-egg problem. The machines needed to make things require components that must themselves be made. On Earth this problem is dissolved by the existing industrial ecosystem — any missing component can be sourced from somewhere. In a remote space environment that ecosystem is absent. The fabricator must, to the greatest extent possible, make itself.
This is not a new problem. Every industrial capability in history bootstrapped from something simpler. The first screw-cutting lathe was made without a screw-cutting lathe. The first semiconductor fab was built with tools that predated semiconductors. The question is not whether bootstrap is possible — history shows it always is — but what the minimum seed is from which a given capability can grow, and how long the generational sequence takes.
For an asteroid fabricator the bootstrap problem has a specific and tractable form. Given a seed payload delivered by a near-term launch vehicle, and a well-chosen asteroid body, what industrial capability can be established, and over what timeline does it become self-sustaining?
The answer developed in this document is more optimistic than most people would expect. The minimum seed is compact. The generational sequence is coherent. The residual import dependencies are small. And the space environment, far from being an obstacle, contributes positively to several of the most critical process steps.
The seed payload has two components. The first is physical hardware — the brought tools, devices, and materials described in the appendix. The second is the accumulated scientific and engineering knowledge of centuries, which weighs nothing and yet makes the entire project conceivable.
Every process described in this document — zone refining, carbonyl chemistry, Fischer-Tropsch synthesis, I²L semiconductor fabrication, float-zone crystal growth, optical communication — represents decades of work by thousands of researchers and engineers. A reader can easily pass over a sentence like “use float-zone refining” without registering that it compresses perhaps thirty years of materials science into six words. The fabricator does not rediscover these processes. It inherits them.
This distinction matters when assessing the semiconductor section in particular. The pioneers of semiconductor technology faced three compounded difficulties: understanding the underlying physics, developing the process knowledge, and achieving the required process control. The first two are now largely solved. A fabricator operating in the twenty-first century begins already knowing why contamination at the parts-per-billion level matters, why certain crystal orientations are preferred, why the indirect bandgap of silicon prevents laser emission, and why I²L is suitable for the application. That knowledge was not free — it cost generations of effort — but it is now inherited at the outset.
The semiconductor industry did not require fifty years to develop because transistor fabrication is intrinsically a fifty-year process. It required fifty years because the underlying physics, materials science, process techniques, and design methodologies first had to be discovered. A fabricator operating today inherits that knowledge from the start. That is a very different problem from the one the pioneers faced.
Biological evolution had to discover everything — metallurgy, electricity, semiconductors, optics — from scratch through blind variation. The fabricator is not blind. It begins life already possessing a vast body of inherited knowledge. In that sense it resembles a human civilisation much more than a bacterium.
This document was itself assembled with the assistance of AI — not because AI contributed new scientific knowledge, but because it substantially lowered the effort required to reason across disciplines that no individual would comfortably span simultaneously: asteroid geology, metallurgy, ceramics, vacuum technology, semiconductor physics, industrial chemistry, communications engineering, and systems architecture. The knowledge originated with generations of human researchers. AI helped connect pieces that would otherwise have remained scattered. The fabricator concept exists at the intersection of those two realities.
The choice of host body is the most consequential decision in the concept. It determines what materials are locally available, what the thermal and mechanical environment imposes on the fabricator, and which process steps can be closed locally versus remaining dependent on imports.
The initial instinct toward a metallic M-type body is understandable — abundant iron-nickel, solid and stable, closest composition to the structural materials the fabricator needs. Working through the full material requirements reveals the limitation. M-type bodies are differentiated — the metal and silicate phases separated long ago, leaving a body composed almost entirely of the metal fraction. This means:
The M-type body gives excellent structural metal and reasonable magnetic materials through cobalt ferrite from native oxide inclusions, but stalls at the electrical and chemical layers of the industrial stack. The concept works on M-type but works considerably better on a body with a richer mineral palette.
Ordinary chondrites of S-type composition are the most abundant meteorite class and well represented in the near-Earth asteroid population. Crucially, chondritic bodies are undifferentiated — they preserve the original solar nebula composition with metal and silicate intimately mixed at the grain scale. The mineral assemblage typically includes:
The single most important element this unlocks compared to M-type is aluminium. Its implications run through the entire concept: aluminium metal as a conductor superior to iron-nickel; alumina as the premier locally synthesisable bearing and insulator ceramic; aluminosilicate ceramics — mullite, cordierite — as precision structural materials; aluminium as a reducing agent for selective chemical processing.
A body in the kilometre-scale diameter range has sufficient escape velocity to retain ejecta from impacting bodies over geological time, accumulating a regolith that samples the broader asteroid population. A nominally S-type body of this size carries impactor contributions from C-type, M-type, and other S-type bodies in its surface regolith — acting as a naturally sorted sample library that concentrates useful trace elements the host body’s composition alone would not provide.
Physical properties of the body bear on the fabricator’s operations in ways that go beyond composition:
Rotation rate — slow rotation is strongly preferred. A rapidly rotating body subjects the surface to large thermal swings at high frequency, driving thermal stress in precision structures and complicating solar energy harvesting. A body with a rotation period of hours to days rather than minutes allows the fabricator to track the sun and manage thermal transitions gracefully.
Surface gravity — a body with slight but real surface gravity simplifies anchoring and surface operations considerably compared to a true microG body where every mechanical operation risks pushing the lander off the surface. A body of one to several kilometres diameter has surface gravity of millimetres per second squared — negligible for most purposes but sufficient to keep objects and regolith grains settled.
Surface cohesion — this is one of the least well characterised and most operationally critical properties of small bodies. Missions to Ryugu and Bennu found surfaces considerably more unconsolidated than anticipated — rubble piles with very low surface strength, behaving more like loose aggregate than solid rock. A crumbly surface makes mining easier but anchoring harder, raises dust management as a serious concern, and complicates any operation that needs to react forces against the ground. Surface cohesion is an explicit body selection criterion and a primary target for precursor characterisation.
Stability — the full problem of surface stability, anchoring design, reaction force management during mechanical operations, and long-term evolution of the installation is a major engineering domain in its own right and outside the scope of this concept document. It requires dedicated investigation. The body selection criteria — slight gravity, slow rotation, some consolidated rock available — reduce the severity of the problem without resolving it.
The target body is therefore a kilometre-scale or larger S-type chondritic NEA, slowly rotating, with slight surface gravity, a well-developed regolith, in a low delta-v orbit from LEO — ideally under 6 km/s — and with confirmed or plausible compositional diversity from a long collision history. Detailed characterisation of composition, rotation, surface cohesion, and surface properties should precede commitment of the fabricator payload. Each target body will have unique properties requiring body-specific planning — a generalised architecture can be designed on Earth, but implementation decisions must be made locally in response to what characterisation reveals.
Working through the fabricator’s actual material requirements against the S-type chondritic composition produces the following palette.
Iron-nickel alloy from magnetic separation of the metal fraction. Abundant, directly accessible by mechanical processing without high temperature chemistry. The natural Fe-Ni ratio in chondritic metal fraction approximates permalloy compositions — around 30-40% nickel — which has significantly better soft magnetic properties than pure iron. Useful directly for structural components, motor cores, and transformer cores.
Carbon steel — iron with carbon addition from the carbonaceous regolith fraction, heat treated under the solar concentrator — extends the mechanical property range, providing harder and tougher variants for tooling, spring stock, and bearing races.
Aluminium is the primary conductor. Extracted from feldspar by aluminothermic or electrolytic reduction, drawable into fine wire, with resistivity roughly 1.6 times that of copper — a modest penalty tolerable given the system’s low clock rates and modest power levels. Used for motor windings, transformer windings, bus bars, interconnects, and capacitor electrodes.
Silver is present in the metal fraction as a siderophile element, enriched relative to bulk solar system composition. Concentrations are low — tens to hundreds of ppm — but consumption at the fabricator’s scale is similarly low. Silver wire for the most demanding conductor applications is achievable from local material.
Iron-nickel wire remains available for applications where conductivity is less critical than strength or magnetic properties.
Copper is present primarily in sulphide phases as chalcopyrite intergrowths with troilite. Extraction requires roasting and smelting — several process steps. Given aluminium’s adequacy as a conductor, copper refining is not worth the infrastructure investment except for catalyst applications where native copper grains — present at trace levels in the metal fraction — may suffice.
Soft magnetics — permalloy from the natural Fe-Ni composition of the metal fraction, and carbonyl iron powder composite from the carbonyl processing infrastructure. Carbonyl iron powder — pure, fine, spherical — forms low eddy-current soft magnetic composites when sintered or bound, providing adequate performance for motor cores and transformer cores at the operating frequencies of the system.
Hard magnetics — cobalt ferrite (CoFe₂O₄) sintered from iron oxide and cobalt under the solar concentrator. Adequate permanent magnet performance for motors and sensors. Not rare-earth grade, but sufficient.
Silicon dioxide — abundant from silicate mineral processing. Solar-fused into glass for fibre insulation, sheet, rod, and tube. The most immediately accessible insulating material.
Alumina (Al₂O₃) — synthesised from local aluminium by oxidation and sintering. Hardness of 9 Mohs, excellent electrical insulator, chemically inert, dimensionally stable. The primary precision insulator and bearing material once aluminium processing is established.
Aluminosilicate ceramics — mullite and cordierite form naturally when aluminium silicate minerals are heated. Cordierite has exceptionally low thermal expansion, valuable for precision structures — ion beam column components, positioning stage elements — that must maintain dimension through thermal cycling.
Silicon carbide — synthesised from local silicon and carbon under the solar concentrator. Useful as a structural ceramic and as an abrasive for lapping and polishing alumina bearing surfaces. The abrasive supply chain is local — SiC abrasive laps alumina surfaces, alumina abrasive laps SiC surfaces.
Silicon nitride — requires nitrogen, present in regolith at trace concentrations from solar wind implantation and nitride mineral phases in impactor material. Conditional on nitrogen availability, Si₃N₄ provides excellent bearing and insulator performance. The alumina route is the more confident primary path.
Fischer-Tropsch synthesis — carbon monoxide from local carbon and oxygen, hydrogen from hydrated silicate electrolysis, iron catalyst from the metal fraction — produces hydrocarbon feedstocks from CO and H₂. Products include:
Chlorocarbon variants — PVC, chlorinated paraffins — from local chlorine produced by electrolysis of chloride-bearing mineral phases. Chlorocarbons cover higher temperature and chemically more demanding insulation applications.
Silicones — polysiloxane backbone from local silicon and oxygen, methyl side groups from Fischer-Tropsch methane plus local chlorine via the Müller-Rochow process. Silicone oils provide vacuum-compatible lubrication with exceptional temperature range. Silicone rubbers provide compliant seals for repeated-assembly joints, flexible electrical insulation, and potting compound for electronics packages. Silicone RTV adhesive bonds ceramics to metals across thermal expansion mismatch. The Müller-Rochow process requires a copper catalyst — present in trace native form in the metal fraction, potentially adequate for catalyst quantities without a full copper refining infrastructure. Pi-conjugate conducting polymers are not pursued — the synthesis is demanding, monomer purity critical, and the application need is covered by simpler routes.
Fluorine is the one element the S-type chondritic body does not provide in useful quantity. Fluorapatite — Ca₅(PO₄)₃F — is the primary fluorine-bearing mineral in chondrites, present at tens to hundreds of ppm. Real but thin.
PTFE is used where nothing else adequately serves: the most demanding gas-phase process containment, ultra-high vacuum seals where metallic knife-edge seals cannot be used, and applications requiring both chemical resistance and flexibility simultaneously.
The design discipline is to ask for each application whether fluorine is genuinely necessary or merely convenient, and to route to alternative materials — metallic seals, silicone rubber, alumina — wherever they adequately serve. A modest PTFE allocation in the seed payload, supplemented by locally extracted fluorine as the ecosystem matures, covers the lifetime requirement.
MoS₂ or WS₂ dry film — space-qualified, vacuum-compatible, brought as a coating on precision bearing surfaces in the seed payload. MoS₂ performs better in vacuum than in air — the oxide layer that causes wear in terrestrial use does not form.
Silicone oil — locally synthesised, extremely low vapour pressure, stable across the asteroid thermal range. Used in protected bearing environments sealed from direct vacuum exposure by labyrinth or thin metallic shields.
Fischer-Tropsch wax fractions — for low-demand lubrication in partially enclosed environments.
The fabricator’s processes divide into thermal, mechanical, and chemical operations. Thermal processes are powered by solar concentration, mechanical processes by local servo-driven machinery, and chemical processes by the contained reactor infrastructure.
A parabolic reflector formed from local metal sheet, with a secondary focus for high-temperature applications. In vacuum, temperatures above 2000°C are achievable with modest concentrator area. This covers:
The concentrator mirror is a shared module in the fleet architecture — see section 7.
Nickel tetracarbonyl Ni(CO)₄ forms when nickel metal contacts carbon monoxide at around 50-60°C and decomposes to pure nickel metal at around 180°C. Iron pentacarbonyl Fe(CO)₅ follows similar chemistry. Carbon monoxide is synthesised from local carbon and oxygen.
The carbonyl process provides:
The toxicity of Ni(CO)₄ in an autonomous robotic system operating in hard vacuum is an engineering containment problem rather than a personnel safety problem — any leak disperses into vacuum and decomposes rapidly. The process vessel, PTFE-sealed and nickel-lined, handles the process streams safely.
The carbonyl reactor infrastructure — contained gas-phase processing with recirculating carrier gas — transfers directly to Fischer-Tropsch synthesis and partial silicon refining chemistry. One infrastructure investment serves multiple process needs.
CO and H₂ over an iron catalyst at moderate temperature and pressure produce hydrocarbons from methane upward. The catalyst is local iron. The reactor infrastructure is shared with the carbonyl process. Products feed the polymer synthesis chain and provide lubricant fractions.
Müller-Rochow process: silicon metal reacts with methyl chloride over a copper catalyst at 250-300°C to produce chlorosilane intermediates, hydrolysed and polymerised to silicone. Methyl chloride from Fischer-Tropsch methane plus local chlorine. Copper catalyst from native copper grains in the metal fraction.
Chondritic texture means physical separation at room temperature is effective:
Roasting the sulphide concentrate produces sulphur dioxide — capturable and convertible to sulphuric acid, opening an electrolytic processing capability. The silicate fraction feeds ceramic synthesis. The carbonaceous fraction feeds Fischer-Tropsch. The regolith is a mineral library that rewards careful sorting.
A significant architectural choice is whether processing occurs on the surface of the asteroid or on an orbiting platform, with raw material transported from surface to orbit. This trade space deserves explicit consideration because the right answer depends heavily on the specific body’s properties — and different bodies will warrant different answers.
The surface installation — all operations on the asteroid surface — is the simpler deployment. No orbital mechanics to manage, no mass transport system required, and the installation can grow by physically extending itself across the surface. The disadvantage is that precision processing must coexist with mining dust, thermal cycling driven by the body’s rotation, and the continuous challenge of reacting mechanical forces against a surface of uncertain cohesion.
The orbital processor — mining and initial sorting on the surface, precision processing on an orbiting platform — separates the dirty and clean operations architecturally rather than by industrial zoning on the surface. The orbital environment offers continuous solar power unobstructed by the body, a stable thermal environment controllable by platform orientation, no dust from surface operations, and no anchoring forces. The solar concentrator always points at the sun. The communication relay is naturally co-located. Contamination from one process module does not reach another without deliberate connection.
For a body with sufficient surface cohesion, a mass driver, pneumatic launcher, or mechanical sling can loft sorted material to orbit at negligible energy cost — escape velocity for a kilometre-scale body is of order 0.5-1 m/s. The launched containers are caught by the orbiter with a simple electromagnetic or mechanical catcher. The return path — waste material or processed products to the surface — is equally trivial.
For a crumbly, unconsolidated body the orbital architecture becomes more compelling still. The surface installation is reduced to the simplest possible operation — scoop and loft — which imposes minimal reaction forces against weak regolith. All precision work moves to orbit where surface properties are irrelevant. Mining dust stays on the surface. The precision platform remains pristine.
The orbital platform also has a natural stability challenge — small body gravitational fields are irregular, and solar radiation pressure is significant relative to weak gravity. Orbital stability requires analysis for each specific body and modest continuous station-keeping. Retrograde orbits are often more stable around small bodies.
The surface and orbital approaches are not mutually exclusive. A mature installation may operate both — surface processing for bulk material reduction, orbital platform for precision semiconductor and assembly work — with the boundary between them determined by the cleanliness and stability requirements of each process.
The industrial zoning principle — that dirty mining operations should be separated from precision manufacturing — applies whether the separation is achieved by distance on the surface or by the vacuum between surface and orbit. The orbital architecture resolves it most cleanly. In either case, vacuum actually assists contamination management because contaminants travel ballistic paths rather than mixing everywhere by atmospheric turbulence. The source of a contamination event is identifiable and the affected zone is predictable — an advantage that air-based manufacturing does not enjoy.
Each target body is unique. The specific combination of composition, surface cohesion, rotation rate, dust behaviour, orbital stability, and regolith depth cannot be fully characterised before arrival, and even a thorough precursor mission will leave uncertainties that only operations resolve. A generalised architecture can be designed on Earth, but implementation decisions must be made locally and adaptively.
With an eight-minute one-way light delay, Earth cannot be in operational decision loops. By the time Earth receives a report that the surface is more unconsolidated than anticipated and formulates revised instructions, the situation has evolved. The system must reason locally about its own situation and adapt its plan accordingly.
This requirement goes beyond the module-level linguistic capability described in section 7 — individual modules negotiating resource allocation and reporting faults within a known operational envelope. Body-specific planning requires engineering judgement: examining what the characterisation data actually shows, identifying where the standard architecture needs modification for this specific body, generating and evaluating alternatives, and implementing a revised plan.
For example: surface cohesion insufficient for a surface-mounted mass driver — shift to orbital processing with pneumatic launch, or redesign the surface installation to distribute reaction forces over a larger footprint. Or: regolith carbon content lower than assumed — reduce Fischer-Tropsch reactor scale and increase brought polymer stock. Or: rotation rate faster than selection criteria specified — add active solar tracking to the concentrator design. None of these decisions are in a lookup table. They require understanding the system’s own architecture well enough to reason about trade-offs under changed conditions.
This is precisely where local AI becomes a structural necessity rather than a convenience. The supervisory layer must carry genuine engineering reasoning capability — sufficient to plan and adapt the system’s own development in response to what it finds. This is also where the design of that AI becomes a governance question, not merely an engineering one. A supervisory AI that can explain its decisions — that can report to Earth why it chose the orbital architecture for this body, what alternatives it considered, and what it expects the consequences to be — is auditable and accountable in a way that a black-box autonomous system is not. That transparency is not optional for a long-lived autonomous system operating far from human oversight. It is the foundation on which trust between the system and its human principals is built.
Local semiconductor fabrication is the most ambitious element of the concept and the one that takes the longest to close. It is important to distinguish two separate closure milestones that may be separated by decades:
Industrial closure — the fabricator can produce its own structural metal, conductors, magnetics, insulators, bearings, polymers, motors, and mechanical systems without Earth resupply. This milestone is achievable relatively early in the generational sequence, probably within the first few generations of the system.
Semiconductor closure — the fabricator can produce its own silicon devices. This is a substantially harder milestone requiring the full semiconductor path described below. It should not be assumed to follow quickly from industrial closure. The two may be separated by a generation or more of operation during which imported silicon devices continue to serve the control and computing functions.
This distinction matters for planning. The concept does not require semiconductor closure to be useful or to begin generating value. A system that has achieved industrial closure is already capable of building and maintaining its own mechanical infrastructure, producing local materials, and assembling electronics from imported silicon. Semiconductor closure is the final and most ambitious loop to close, not a prerequisite for everything else.
Metallurgical grade silicon by carbothermic reduction of silicate minerals at 1500°C under the solar concentrator. Purity approximately 98-99%.
Float zone refining in the microG environment is superior to its terrestrial equivalent:
The known limitation is the segregation coefficient of boron and phosphorus — both close to unity, meaning zone refining removes them slowly. These set a floor on background doping. The phosphate minerals in S-type chondrite provide a local phosphorus source of known composition — the background can be characterised and accommodated as a design constraint rather than an unknown.
Silicon forms a native oxide within seconds of exposure to any oxygen-containing environment. This must be removed before doping, contact deposition, or any process step requiring clean silicon. Vacuum-compatible alternatives to hydrofluoric acid etching are available and in some respects superior:
Hydrogen passivation — atomic hydrogen at moderate temperature replaces the native oxide with a hydrogen-terminated surface stable in vacuum. No fluorine required. Leaves lower defect density than acid etching.
Thermal desorption — heating above approximately 900°C in ultra-high vacuum drives off the native oxide. The zone refining apparatus can be adapted for this function.
Ion beam cleaning — low energy argon bombardment from the ion column removes surface oxide by sputtering. Same tool as doping, different operating conditions.
Chlorine reactive etching — silicon etches in chlorine gas producing volatile SiCl₄. Local chlorine from electrolysis of chloride-bearing minerals feeds a contained etch reactor using shared infrastructure. Pattern-selective etching is possible using local oxide or nitride films as masks.
Controlled thermal oxidation — deliberate growth of SiO₂ at defined thickness for gate dielectric and capacitor dielectric applications.
Bond pad surfaces in the WCB assembly process are cleaned by low energy ion beam immediately before bonding, replacing the flux function of terrestrial soldering entirely.
Lithography is avoided entirely. Spatially defined doped regions are created by direct write with a focused ion beam of dopant species — the scan pattern defines the device geometry with no mask, no photoresist, no wet chemistry.
Feature size is determined by beam diameter and positioning accuracy, not throughput. Writing a single substrate over days or weeks is entirely acceptable. With a well-designed column, good vacuum, and a stable slow-moving stage, features in the 1-2 micron range are achievable — late 1970s to early 1980s device capability, adequate for the application.
The ion source feeds on solid dopant materials — boron from local boron-bearing mineral phases, phosphorus from local phosphate minerals.
Plasma environment and high voltage shielding — the space environment introduces a critical electrical constraint that has no laboratory analogue. The ambient solar wind plasma is a conducting medium permeating the operating environment. Work at NASA Lewis Research Center (Domitz and Grier, TM X-71554, 1974, https://ntrs.nasa.gov/api/citations/19740019277/downloads/19740019277.pdf) demonstrated that even microscopic pinholes in insulation covering high voltage conductors collect plasma current orders of magnitude greater than expected — the surrounding insulation funnels plasma charges into any defect, and transient arcs occur at negative bias voltages as low as 400V. The plasma must not be able to see high electric fields.
The mitigation is a continuous grounded conductive metallic enclosure around all high voltage wiring — locally producible from available metal stock. Fused silica, confirmed by Domitz and Grier to perform well to over 20kV in plasma environments, serves as the internal insulator and is locally producible. Silicone RTV, present elsewhere in the concept as a general encapsulant, is inadequate for plasma-exposed high voltage applications and must be excluded from any such use. The ion beam exit aperture requires a neutraliser — an electron-emitting cathode that neutralises beam charge on exit — to prevent uncontrolled plasma return currents. A thoriated or oxide-coated nickel cathode, producible from local materials via the carbonyl process, serves this function.
I²L is chosen as the device architecture for reasons specific to this application:
The on-chip microprocessor targets a 17-bit data path and address space. The 17th bit serves three simultaneous roles:
Parity is critical in the radiation environment. Word-level parity catches all single-bit upsets. A background memory scrubbing task checks every location periodically. Triple modular redundancy on the most critical registers — program counter, stack pointer, key control registers — provides majority-vote correction. Radiation hardening by architecture, requiring no special process steps.
A 5mm × 5mm die at 1-2 micron geometry contains:
The first locally produced microprocessor marks genuine semiconductor closure. It does not need to be good by terrestrial standards. It merely needs to be useful — capable of driving a motor, reading a sensor, executing a control loop. A crude device comparable to the late 1960s might already be sufficient to control the next generation of improvement. That threshold may be surprisingly low, and crossing it opens the path to progressive refinement.
Stacked metal-insulator-metal capacitors are integrated directly onto the I²L die, built over field oxide regions using process steps already required for transistor fabrication. The stack is built layer by layer — time-unconstrained, as many layers as the application warrants. Capacitance scales with layer count over the same footprint, covering picofarads to tens of nanofarads for bypass, filtering, tank circuits, and interstage coupling — all on-chip.
For bulk power storage where stacked on-chip capacitors cannot reach the required capacitance, polymer film capacitors serve: locally synthesised polyethylene or polypropylene film from Fischer-Tropsch, aluminium foil electrodes, wound as discrete components. Entirely vacuum compatible.
Wired Circuit Board assembly — a serial robotic process using pre-formed wire interconnects compression-bonded to SMD component pads on a backing substrate — provides the electronics assembly capability. In the asteroid environment the process simplifies:
Solar panels are brought as part of the seed payload — silicon or GaAs photovoltaic, space-qualified, providing the primary power for all operations. As the semiconductor path matures, locally fabricated silicon photovoltaic cells become producible — the same float zone silicon, the same doping process, applied to a planar structure. Local solar panel production is an important closure milestone on par with the microprocessor milestone, eliminating the brought panel as a resupply item.
The solar resource at 1AU is approximately 1.36 kW/m² — consistent and uninterrupted in orbit. On the surface, body rotation and shadowing create cycles that the orbital architecture avoids entirely.
Stacked MIM capacitors on-chip for local bypass and decoupling. Polymer film capacitors as discrete components for bulk power storage at module level.
Mechanical energy storage is complementary to electrical storage for specific applications where chemical or electrical storage is inappropriate:
Flywheels — a spinning mass in hard vacuum on a magnetic bearing stores kinetic energy with very low loss rate. Locally producible permalloy magnetic bearings, local metal flywheel mass. Already used in spacecraft for attitude control and energy storage.
Springs — carbon steel spring stock from the local carburising process. Stores energy as elastic strain. Useful for one-shot actuators, emergency release mechanisms, and applications requiring guaranteed energy delivery independent of all electronics.
Compressed gas — gas compressed by solar-powered pump stored in local metal pressure vessels. Useful for pneumatic actuation where electrical drives are less suitable, and potentially as the surface-to-orbit launch mechanism.
Mechanical storage requires no electrochemistry, no radiation-sensitive materials, and no temperature-dependent capacity degradation — properties that have real value for a long-lived autonomous system far from maintenance.
The fabricator is not a single unit but a fleet of modular landers on the same body, potentially including both surface and orbital elements. Multiple units provide:
Every subsystem is a module with a standard interface. The interface standard — defined once in the seed payload as precision-made reference connectors, alignment fixtures, and coupling elements — propagates through the ecosystem because every locally fabricated replacement module is built to fit existing interfaces.
The interface standard covers mechanical mounting, power, data, thermal, and fluid connections where applicable.
Modules are physically compact and dense. The large volume elements of each unit — solar panels, concentrator mirrors, thermal radiators, anchoring structure, positioning stages — are mostly locally fabricated from early generations onward. Brought modules are small, dense, high-value items that travel between units as needed.
Each module carries a permanently assigned identity — established at manufacture, immutable, unique across the fleet — with an associated history of every interaction, fault, and calibration event. A module moving between units brings its history to the new unit’s network.
Each module has sufficient computational and linguistic capability to describe its own state, understand requests from dependent modules, reason about its own operational constraints, negotiate scheduling and resource allocation, detect and report anomalies, understand directly coupled modules sufficiently to coordinate without central direction, and escalate to the supervisory layer when a situation exceeds its reasoning capability.
The network is peer-to-peer. The solar concentrator mirror arbitrates competing thermal requests by priority and schedule. The power module sheds non-critical loads in a power emergency without instruction from above. Coordination emerges from negotiation rather than central direction.
The modular identity and linguistic architecture makes the system comprehensible and auditable by human visitors — a property that connects to the broader governance question of how a human presence might eventually relate to an established autonomous fabricator ecosystem.
The solar concentrator mirror is not owned by any single process — it is a shared resource any thermal process module can request. As locally fabricated mirrors join the fleet they announce their capability and the network redistributes thermal load. The system self-configures as resources are added.
The supervisory layer — described in section 5.2 — carries engineering reasoning capability substantially beyond module-level coordination. It maintains the fleet-wide view, manages Earth communication, handles novel situations, maintains the fleet registry of module identities and histories, and adapts the system’s development plan in response to what operations reveal about the specific body.
Transparency and auditability are design requirements, not optional features. The supervisory AI must be able to explain its decisions to Earth in terms that allow human review and override. This is not merely a governance preference — it is the practical foundation of a workable relationship between an autonomous system operating at interplanetary distances and the human civilisation that deployed it.
Anchoring is the first engineering problem the system must solve on arrival. Any mechanical operation exerting force on the body risks pushing the lander off the surface if reaction forces are not managed. The orbital architecture reduces the anchoring burden by keeping precision operations off the surface entirely. For surface operations, candidate approaches include harpoon anchors, screw anchors into consolidated material, and dead-weight anchoring. The detailed engineering depends heavily on the specific body’s surface properties and is a subject for the preliminary design phase.
An orbiting RF relay station serves as the communication hub for the fleet, co-located naturally with the orbital processing platform if that architecture is adopted. It separates the communication problem into two independently solvable parts:
Surface to relay — short range, RF, simple. Milliwatt transmitters and omnidirectional antennas achieve high data rates without fine pointing on the surface units.
Relay to Earth — long range, optical, sophisticated. All fine pointing and optical engineering concentrated in one well-engineered terminal serving the entire fleet.
GaN laser diodes — direct bandgap, emitting in the 400-500nm range — are the transmitter for the relay-to-Earth optical link. Silicon is not viable as a laser material: it is an indirect bandgap semiconductor with the conduction band minimum and valence band maximum at different points in k-space. A photon cannot bridge the momentum gap — phonon assistance is required, making stimulated emission orders of magnitude less probable than in a direct bandgap material. No engineering workaround exists for this constraint.
GaN is chosen for radiation hardness — wide bandgap and strong bond energy confer significantly better resistance to the asteroid radiation environment than narrower gap III-V materials.
Silicon PIN photodiodes serve as receivers — silicon absorption is efficient at GaN emission wavelengths. The receiver chain is locally fabricable once the semiconductor path matures.
GaN laser diodes remain on the brought list indefinitely. This is the most benign remaining import dependency — grams per unit, years between replacements, clearly defined.
The relay’s optical terminal uses coarse gimbal pointing from ephemeris calculation and a fast steering mirror with quartz piezoelectric actuators driven by closed-loop feedback from a beacon detector. Crystalline quartz — SiO₂ in the appropriate cut orientation — is locally synthesisable.
Light travel time at 1AU is approximately eight minutes each way. The link is episodic and store-and-forward. Earth cannot be in any operational loop. Every operational decision is made locally. The fleet’s autonomous operation is a physical necessity imposed by the speed of light.
The concept addresses principal subsystems and material dependencies in sufficient depth to establish feasibility. Many process details have not been individually resolved. The general answer follows the same logic throughout: a local production path exists using infrastructure already described, or a small brought stock covers the need at negligible mass cost, or a combination of both progressing from brought to local as the generational sequence advances.
No individual detail of this class is expected to be a showstopper. The design discipline for each is to ask first whether a local solution exists, second whether an alternative material adequately serves, and third what the brought mass of a sufficient stock would be. A well-chosen stock covering all anticipated gaps is estimated to add 5-10kg to the seed payload without changing the order-of-magnitude conclusion.
The following dependencies close locally within the early generational sequence:
Achieved later, after industrial closure is established:
At maturity a microsatellite-class resupply package every several years maintains full operational capability. The fleet does not depend on resupply for survival — only for gradual capability maintenance.
Silicon boron/phosphorus floor — residual boron and phosphorus after carbothermic reduction and multiple-pass microG float zone refining, and tolerance of I²L devices at 1-2 micron features to that background. The most critical open question in the semiconductor path.
Ion source dopant materials — performance of solid source ion guns with impure local chondritic feedstock. Reproducibility between source replacements and calibration procedure.
Copper catalyst for silicone synthesis — native copper concentrations against catalyst quantity required. Evaluation of zinc or tin alternatives if inadequate.
Regolith nitrogen and fluorine concentrations — determining local Si₃N₄ viability and PTFE supplement rate. Requires assay of actual target body regolith.
Microseismic environment — vibration amplitude during ion beam writing and whether passive isolation achieves 1-2 micron feature stability.
Fischer-Tropsch product distribution — yield of useful polymer fractions from local feedstocks of characterised composition.
Chlorine etch characterisation — etch rate and selectivity of silicon in local chlorine at accessible conditions.
Stacked capacitor yield — achievable layer count and defect rate, and redundant cell count needed for adequate yield.
Surface cohesion — quantitative characterisation of the specific target body’s surface properties and implications for anchoring and the surface-orbit architecture trade.
Orbital stability — detailed analysis of orbital mechanics around the specific target body, station-keeping requirements, and attitude control budget.
Carbon budget — mass flow estimate from regolith processing to useful carbon, and adequacy for Fischer-Tropsch at the fabricator’s consumption rate. This is central to the polymer and lubricant chain and deserves quantitative treatment.
Plasma environment characterisation — the solar wind plasma density and energy spectrum at the specific target body location, and the implications for high voltage system design, insulation selection, and neutraliser sizing. Domitz and Grier (1974) provide the foundational experimental data; body-specific plasma conditions need assessment for detailed design.
Anchoring — surface consolidation, anchor design, reaction force management. Heavily body-specific.
Stability — full geotechnical, dynamic, and thermal stability of the surface installation. A major engineering domain outside the scope of this document.
The single-body self-sufficient fabricator is the pioneer phase of a larger story, not its conclusion. Real industrial systems rarely stop at self-sufficiency. They specialise.
On Earth no city insists on containing its own iron ore, bauxite, copper mines, oil fields, semiconductor fabs, and forests. Trade emerges because specialisation is more efficient. The same logic applies in space.
As multiple fabricator installations develop across the near-Earth asteroid population, each body’s compositional character becomes an economic asset:
M-type bodies — abundant in structural iron-nickel, cobalt, platinum-group metals. Natural exporters of refined structural metal and magnetic materials.
S-type bodies — balanced composition supporting the full industrial stack. Natural producers of ceramics, electronics, and precision machinery.
C-type bodies — rich in water, carbon, nitrogen compounds. Potentially the most strategically valuable bodies of all, supplying the volatiles that every other installation needs for chemistry, propulsion, and atmosphere.
Copper, fluorine, ammonia — materials that a self-sufficient pioneer must manage carefully — become straightforward imports once a trading network exists. A nearby body with concentrated copper-bearing sulphides can export refined copper. A body with abundant fluorapatite can supply fluorine. A carbon-rich body can supply feedstock for Fischer-Tropsch chemistry across the network.
This possibility substantially strengthens the overall concept. The closure requirements for the very first fabricator are stricter than those required of the mature industrial network that grows from it. The pioneer must be largely self-reliant because trade does not yet exist. Once enough installations are operating, many of the remaining irreducible imports cease to be imports from Earth at all — they become imports from elsewhere in the asteroid population.
In this longer view the fabricator concept is not a proposal for an isolated self-replicating machine. It is the seed of an emergent solar system economy, in which specialisation, trade, and mutual dependence between installations progressively replace the self-sufficiency that was necessary only at the pioneer stage.
The exercise of working through a self-bootstrapping asteroid fabricator concept from first principles produces a result that surprises in its optimism. Not because the engineering is easy — it is not — but because the space environment, examined honestly rather than measured against terrestrial assumptions, turns out to contribute positively to several of the most critical process steps.
Zone refining in microG is better than on Earth. Bonding in vacuum needs no flux. Surface preparation in vacuum replaces corrosive acid chemistry with cleaner alternatives. Vibration isolation without gravity is more effective. The vacuum process environment eliminates oxidation chemistry that terrestrial processes must constantly manage. Solar energy provides heat for sintering, annealing, and zone refining without a fuel supply chain. Contamination in vacuum travels ballistic paths rather than mixing everywhere — industrial zoning is cleaner, not harder. The radiation environment drives an architectural response — the 17-bit parity architecture — that makes the processor more robust than a conventional design would be.
The instinct to treat space as hostile comes from measuring it against Earth conditions as a baseline. Earth conditions are not the natural baseline for technology — they are the conditions under which technology evolved by historical accident. A technology designed from first principles for the space environment looks different, and in several important respects works better.
The asteroid fabricator concept developed here is not a terrestrial factory heroically surviving in a hostile environment. It is a system that makes more sense in space than it would on Earth — shaped by its environment in ways that are often advantageous, building on local materials that the environment provides, and closing its dependencies on Earth one by one through a generational sequence that has internal logic and no discontinuous leaps.
This optimism rests on a foundation. The concept does not start from ignorance. It inherits centuries of accumulated scientific and engineering knowledge — metallurgy, chemistry, semiconductor physics, vacuum technology, systems architecture — knowledge that was hard-won by generations of researchers and practitioners. The fabricator solves a different and easier problem than the pioneers faced because it begins already knowing what they had to discover. Acknowledging that inheritance is not false modesty. It is the accurate account of why the concept is plausible.
The irreducible import list at maturity — GaN laser diodes, PTFE used sparingly, a small stock of specialist materials — is light enough to sustain with occasional resupply. The fleet is functionally autonomous. And in the longer arc, even these imports may be sourced from within a growing solar system economy rather than from Earth.
What remains is the practical testing. Laboratory analogue experiments on the key process steps, material characterisation of actual chondritic samples, precursor characterisation of a specific target body, and quantitative answers to the open engineering questions listed above. The concept provides a coherent framework for that experimental programme. The framework holds together. Whether reality matches it in every particular is the question that only experiment can answer.
That is the correct next step.
A preliminary estimate, subject to revision as the concept develops. The 18kg figure represents a lower bound for the concept — a practical first mission should budget 50-200kg to accommodate the full instrumentation, metrology, robotic systems, spare electronics, sensors, wiring, and actuators required for real operations. The order of magnitude — tens to low hundreds of kilograms rather than thousands — is the robust result, and the ratio between seed mass and ultimate capability is the argument that matters.
A further observation on lifetime autonomy: the entire residual import list at maturity — GaN laser diodes, PTFE stock, specialist materials, precision consumables — almost certainly fits within an additional 50kg carried at launch. A system that brings its own lifetime supply of residual dependencies from the outset is not functionally dependent on Earth resupply at all. The resupply cadence drops from every few years to essentially never. The total mission mass — seed fabricator plus lifetime consumables — remains within the capability of a single dedicated smallsat-class launch. This reframes the concept: not a system that remains tenuously connected to Earth by a supply chain, but a genuinely self-contained outpost that carries everything it will ever need in a payload of modest total mass.
Precision tooling — draw plate dies, bonding tool tips, interface standard reference elements, lapping fixtures. Mass: ~2kg.
Silicon devices — microcontrollers, motor drivers, sensor ICs, power management, communication ICs for initial fleet operation. Mass: ~3kg.
Solar panels — brought photovoltaic panels for primary power before local silicon PV production. Mass: ~3kg.
GaN laser diodes — primary and spares for communication modules. Mass: ~0.1kg.
PTFE stock — sheet, rod, tape, tubing. Mass: ~1kg.
MoS₂ dry film — applied to precision bearing surfaces. Mass: ~0.2kg.
Precision bearing sets — for highest-precision axes before local alumina bearing production. Mass: ~1kg.
Seed electronics — ion beam column control, initial WCB assembly controller, LAN infrastructure components. Mass: ~2kg.
Reference standards — calibration masses, dimensional references, wavelength reference for interferometric metrology. Mass: ~0.5kg.
EEPROM dice — bare unpackaged radiation-hardened non-volatile memory, for module identity records, calibration data, operational history, and process parameters throughout the fleet. Rewritable and power-loss safe — fills the gap between implant-programmed ROM (permanent but unalterable) and SRAM (volatile). Without non-volatile storage the module identity and history system loses continuity on every power interruption. Disproportionate value relative to mass. Mass: ~0.1kg.
Unresolved detail stock — refined materials and specialist components covering process details not individually resolved at concept stage. Mass: ~5kg.
Concept lower bound: ~18kg Practical first mission estimate: 50-200kg
The difference between the lower bound and the practical estimate is instrumentation, metrology, robotic actuation, spare electronics, sensors, and wiring — all real requirements for an operational system that are implied by the concept but not individually enumerated at this stage.
v1.3 — Added plasma environment and high voltage shielding to section 6.4, with reference to Domitz and Grier (NASA TM X-71554, 1974). Added plasma environment characterisation to open engineering questions. Added EEPROM dice to seed payload — non-volatile storage for module identity and operational history. RTV exclusion from plasma-exposed high voltage applications noted.
v1.2 — Added preface framing the document as technical appendix to spaceaccord.org and stating the AI collaboration clearly. Corrected I²L transistor description: merged PNP/NPN structure with lateral PNP as current injector and NPN performing all logic and power switching. Added lifetime autonomy observation to appendix: residual import dependencies fit within ~50kg additional payload, making the system genuinely self-contained from launch. Updated change log.
v1.1 — Added section 1.1 The inheritance of knowledge, including the role of AI assistance. Separated semiconductor closure from industrial closure as distinct milestones, acknowledging they may be decades apart (section 6 introduction). Added section 5 Surface and orbital architecture covering the surface-orbit trade, crumbly body considerations, industrial zoning in vacuum, and body-specific planning as a structural requirement for local AI. Expanded supervisory layer discussion to include engineering reasoning capability and the governance requirement for transparency and auditability. Added section 13 Beyond the Pioneer — A Trading Network covering specialisation by body type and the transition from self-sufficiency to trade. Revised seed payload discussion to present 18kg as a lower bound with 50-200kg as a practical first mission estimate. Added carbon budget, surface cohesion, and orbital stability to open engineering questions. Minor corrections and clarifications throughout.
v1.0 — Added body selection physical criteria, energy section, anchoring subsection. Revised capacitors to stacked MIM and polymer film only. Revised seed payload to ~18kg.
v0.2 — Added surface preparation, capacitors, unresolved details principle. Revised seed payload from ~10kg to ~15kg.
v0.1 — Initial concept document.
This document represents the current state of a concept developed through extended technical discussion and collegial review. It is a starting point for detailed engineering investigation, not a finished design. The authors welcome engagement from researchers and engineers with relevant expertise in any of the subsystems described.
Gregory Eckersley — spaceaccord.org June 2026