Frank Tipler jolted the astrophysics community in 1980 when he introduced self-replicating interstellar probes into discussion of the Fermi Paradox. The mathematical model of self-replication came from John von Neumann, and was codified in 1966 (after von Neumann’s death) by Arthur Burks in Theory of Self-Reproducing Automata (1966). SF fans will also know of Fred Saberhagen’s berserker novels and short stories (the first appeared in 1963). I’ve found an even earlier SF reference but will leave that for a future post. Right now I want to introduce Peter Marinko, who today weighs in on self-replication and the problems therein. Based in Uppsala, Sweden Peter holds an M.Sc. in metallurgy and has a career background in industrial process engineering. He has studied SETI under Erik Zackrisson at Uppsala University, and his current work explores the thermodynamics of technological civilizations — including a manuscript on high-exergy technospheres and the longevity of detectable civilizations, currently under peer review at the International Journal of Astrobiology. A preprint is available on Zenodo.
by Peter Marinko
Discussions of von Neumann probes — here and elsewhere — tend to treat replication as a systems problem: the probe arrives, mines local material, and builds a copy of itself. The hard part is usually assumed to be propulsion, navigation, or AI. As someone who has spent a career in metallurgy and industrial process engineering, I would like to suggest that the hardest part is the one that gets a single sentence: “mines local material and builds a copy.”
Let me raise four concrete problem areas, in increasing order of difficulty.
1. Beneficiation without gravity, water, or atmosphere
“Asteroid mining” is a misleading phrase. Mining is the easy part; the problem is beneficiation — concentrating useful elements out of undifferentiated regolith. Every terrestrial concentration process relies on things an asteroid lacks: gravity-driven sedimentation, water-based flotation, density separation in fluids, atmospheric combustion. Electrostatic and magnetic separation in microgravity are conceivable in principle, but neither has been demonstrated at industrial scale, and both work poorly on the fine, cohesive, electrostatically charged dust that dominates regolith.
2. Reduction metallurgy without an industrial hinterland
All terrestrial metal production rests on an invisible foundation: carbon or hydrogen as reducing agents, fluxes, and — critically — refractory materials for the furnaces. Refractories are the forgotten enabling technology of civilization. A furnace lining must itself be manufactured, at high temperature, in a furnace. Bootstrapping this loop from raw regolith, with fully closed chemical cycles (no atmosphere to vent to, no water to waste), is a chicken-and-egg problem that no study I am aware of has worked through at the level of actual process flowsheets.
3. The closure problem, honestly accounted
The classic NASA study (Freitas et al., 1980) assumed ~90–96% “closure” — the fraction of its own components a system can reproduce — with the remainder supplied as “vitamins” from home. But the missing few percent are not marginal; they are precisely the hardest items: semiconductors, precision bearings, sensors, and insulation. Consider something as unglamorous as wire insulation. Virtually all electrical insulation on Earth is organic polymer, resting on a petrochemical industry, resting in turn on a biosphere that spent hundreds of millions of years concentrating carbon. Inorganic alternatives (glass fiber, ceramics, mica) exist but are brittle, heavy, and require entirely different process chains to apply to fine conductors. A modern semiconductor fab is arguably the most complex artifact humanity has built, drawing on tens of thousands of specialized inputs. Shrinking that into a 500 kg seed — or even Freitas’ original 100-ton seed — is not an engineering detail. It may be the entire problem.
4. Aging over interstellar timescales
Even a probe that could replicate must first arrive functional after a voyage of tens of thousands of years. We have essentially no empirical data on machine longevity beyond ~50 years (Voyager, surviving on redundancy and switched-off instruments). Over interstellar timescales, materials face cumulative radiation damage and lattice defects, embrittlement and transmutation; creep and solid-state diffusion (solder joints, thin films and interfaces are only kinetically frozen, not thermodynamically stable); tin and zinc whisker growth; outgassing and cold welding in vacuum. The repair systems age too. Replication must outrun degradation — and degradation never sleeps.
A thermodynamic framing
These four problems share a common structure. A self-replicating probe is, in effect, a miniaturized high-exergy technosphere that must rebuild its entire exergy cascade — from raw, unconcentrated feedstock to precision components — at every node, before its own irreversible degradation catches up. The feasibility question is then not “does physics forbid it? (it does not) but “can accessible exergy per node sustain full process closure faster than irreversible losses accumulate?
This is the same ratio, I would argue, that governs the longevity of detectable civilizations generally — a question I explore in a recent preprint on the thermodynamics of technological civilizations. But the probe case is a cleaner test, because the system boundary is sharp and the accounting is (in principle) tractable.
Questions for discussion
1. Has anyone attempted an actual process flowsheet — not a block diagram — for closing even a simple metallurgical loop (say, iron from chondritic material to finished machine parts) without terrestrial inputs?
2. Is there a credible inorganic-only pathway for electrical insulation and semiconductor packaging?
3. What is the realistic closure fraction if “vitamins” are disallowed — and does the seed mass then grow beyond anything launchable?
4. Are there materials strategies (amorphous metals? self-annealing designs?) that could plausibly survive 10,000+ years of transit?
My suspicion, as a practitioner, is that von Neumann probes are constrained not by the laws of physics but by process-chain closure and materials aging — both, at root, thermodynamic limits. If that is right, it bears directly on the Fermi paradox: the galaxy may be quiet not because nobody tried, but because replication is harder than arithmetic suggests.
I would be glad to be proven wrong on any specific point above — ideally with a flowsheet.
Thanks for this. These are the sorts of clearheaded questions that must be asked if discussions of advanced and/or extraterrestrial technologies are to move beyond pleasant speculation and “get real.”
Putting all this into a text to speech generator yields quite an appealing way of absorbing and thinking about the article.
Q1 There is plenty of atomic free iron and nickel in asteroids seeded by violent interstellar explosions which can be removed by magnetic fields, you may need a charge negator maybe an electron gun to control it. Electron guns can be very small every nano scale.
Q2 Due to smaller voltages insulation is not a massive issue, silicon dioxide or even glassyified regolith would be sufficient, distance is also your friend in electronics.
Maybe we need a octopus like device that takes what it needs from moving through its surroundings rejecting what it does not need and it creates another octopi and so on.
This is a wild guess, but I wonder if you could solve the problem by emitting a highly focused beam of electrons, capable of striking a specific part of each individual atom targeted, with the electrons spaced in such a way as to do “vibrational ladder climbing”, pushing a specific chemical bond to its breaking point. I had a long chat with an AI and got to this paper: https://news.mit.edu/2019/manipulate-atoms-graphene-quantum-0517 There’s also this talk: https://www.youtube.com/watch?v=HSS4QVeQPSY I haven’t really watched it all but there is a cute demonstration at 29 minutes.
I’m thinking the most important feedstock ought to be the carbon and hydrogen atoms harvested from the stellar wind. The goal would be to come up with a network of probes in space that look approximately (or exactly?) like cosmic dust.
Space is full of organic compounds. Indeed polycyclic organic hydrocarbons are found in asteroids and are of obvious abiogenic origin.
Refractories is an issue. At least space has plenty of cheap solar power to generate high temperatures. I had a friend who got his PhD in chemistry who was involved in L5-Society. He was doing research on developing metallurgical processes in space, including the refractory problem. He said it was difficult, but could be developed within a decade or two. The author of this piece is correct that too few people are working on this problem. I consider it the key technology necessary to enable space colonization in our own solar system.
Semiconductor manufacturing is by far the biggest hurtle for self-replication manufacturing. A better process than depo/etch/patterning has to be developed to enable cost effective manufacturing in space. Indeed, this hurtle has to be overcome here on Earth to realize cost-effective AI. Molecular electronics or some kind of bio-fabrication? Or at minimum printable semiconductors in a web process. The one thing that can be said about space is that it is UHV environment. Since the vacuum is there, you don’t need vacuum chambers for the current processes.
In the industry we say that building chips rather than carving them (current processes) is the holy grail of computing technology.
Both of these hurtles can be overcome, but will take time and $$$ to do so.
All true.
But robust and successful self-replicating systems DO exist. They cover the earth, and have done so continuously for billions of years, without any “intelligent” design or supervision. I refer to Life itself. Not only have living organisms occupied the planet, they have also managed to modify it profoundly, as well as survived all sorts of cosmic and planetary catastrophes . Life has not only done this, it has simultaneously flourished by increasing in extent and complexity in the variety of environments it has occupied and transformed. And it has accomplished all this without any evidence of planning, design or intelligence. Counter-entropic systems not only exist, it appears that they MUST exist, they seem to arise spontaneously and evolve inevitably without any supervision or control. Its almost as if they are an intrinsic property of space-time itself. Life is, and acts, like a self-replicating technology, and we can conceive of other possibilities as well. This leads me to believe that the concept is not just reasonable, it may be inevitable.
Of course, the earth’s ecosphere, or any analogues to it we may fabricate, may not be able to carry out the functions we would like to assign to our self-replicating systems (such as galactic exploration, conquest and settlement). But nothing in nature, natural or artificial, does that, does it? In fact, as far we can tell, it appears there is a reason for everything, but a purpose to nothing.
All we know for certain is “shit happens”. That which is not forbidden is mandatory. Matter and energy, interacting in space and time, will always generate complexity, structure and pattern–as long as energy is distributed anisotropically and available to drive the process. The “laws of physics” only provide constraints.