O'Neill Cylinders

An O'Neill cylinder is a rotating pressure vessel large enough to hold an ecology: kilometres of radius, spun for apparent weight, shielded against galactic cosmic rays, and built from material that does not have to climb out of Earth's gravity well. The framing under test is that free-space habitats beat planetary surfaces. That is a comparative claim, and the way to make it a subject rather than a restatement is to insist on the quantity that decides it and is almost never quoted: shielding mass per square metre.

There are exactly three quantities on which a free-space habitat can beat or lose to a planetary surface, and each has published values. First, shielding mass per unit area for a given annual dose — a surface gives you half the sky blocked for free plus unlimited local regolith, and free space gives you neither. Second, gravity and the cost of producing it — a surface gives a fixed value you cannot change; a rotating habitat gives any value you want at a structural and physiological price. Third, where the mass comes from — a surface habitat is built on top of its own feedstock and a free-space habitat must import every kilogram from somewhere with a gravity well. Everything else in the classical argument — sunlight, no dust, no night, controllable climate — is real and second-order against those three, and this brief is structured around them.

Two seams are strict. Historical Space Colonization Concepts owns the 1970s in full: O'Neill's papers and books, the 1975 NASA Ames and Stanford summer study, the Stanford torus, Mass Driver 1, the L5 Society, and the solar-power-satellite reviews that removed the economic rationale. This brief narrates none of it, and uses a classical design parameter only as a numerical input to a present-day calculation. And Artificial Gravity owns all rotation physiology — Coriolis and vestibular tolerance, adaptation rates, the AGBRESA trial, the partial-gravity animal data, the head-to-foot gradient. What this brief owns of rotation is one equation and a table of design points. The near neighbour Space Habitats owns habitats that exist or are funded; this one owns habitats designed to be self-contained at settlement scale.

Established Start with the number that dominates every design and almost never appears on a page about this subject. Globus and Strout, using NASA's OLTARIS radiation transport tool, assume limits deliberately stricter than astronaut limits because a settlement contains children and pregnancies: 20 mSv/yr for the general population and 6.6 mGy/yr for pregnant women. Their computed deep-space requirements are 6–7 tonnes per square metre of polyethylene, a similar figure for water, and 10–11 tonnes per square metre of lunar regolith — regolith being explicitly called “ineffective due to low hydrogen content”, needing about 1.6 times the areal mass of polyethylene for the same result. Those three numbers are the whole economics of free-space settlement, and no version of this page has carried them.

Established An agency source confirms the same order by a different route and adds the scaling law. Jones at NASA Ames gives a solar-particle-event safe haven at 20 g/cm² in 2.5 tonnes; galactic cosmic radiation reduced to about half of baseline at 8 g/cm² and to a quarter at 50 g/cm²; and per-crew shielding mass at 100 m³ of living space of 13 tonnes per person at 8 g/cm² and 78 tonnes per person at 50 g/cm². Established Then the geometry that is the strongest single argument in the entire O'Neill tradition, and which the popular treatment states as aesthetics rather than arithmetic: shielding mass per person falls as N^−1/3. Shielding is a surface, population is a volume, so at 1,000 crew the same 8 g/cm² gives 1.0 tonne per person and the same 50 g/cm² gives 6.5 tonnes per person — a thirteenfold improvement bought purely by building bigger. O'Neill designs are enormous because of that exponent, not because of taste.

Established The reality check comes from an actual design, and the number it produces is startling. Globus and Arora's revised Kalpana One is a cylinder of 250 m radius and 325 m length, rotating at 2 rpm for about 1 g, housing 3,000 people at 170 m² each, shielded with 10 tonnes per square metre of lunar or near-Earth-asteroid regolith placed on the interior of the hull, at a total mass of about 7 million tonnes; the cylinder geometry is chosen because it minimises the ratio of hull area to living area, at about 1.77. Frontier Derived, from that design's own figures: 7 million tonnes for 3,000 people is about 2,300 tonnes per resident. That is roughly 180 times the Jones figure at 8 g/cm² and 30 times his figure at 50 g/cm², because Kalpana One shields at 10 t/m² — about 1,000 g/cm² — and gives each resident 170 square metres rather than a bunk. Established The reframing this makes possible is the most useful thing in the section: the range from 1 tonne per person to 2,300 tonnes per person is a design choice about dose target and living standard, not a disagreement about physics.

Established And then the finding that makes the whole calculation contingent on an orbit rather than on a technology. Globus and Strout's actual argument is not about deep space at all. They claim that “space settlements in Equatorial Low Earth Orbit (ELEO) below about 500 km are likely to meet this standard with little or no dedicated radiation shielding”: at 500–600 km and minimal inclination, radiation stays below 20 mSv/yr with essentially nothing, and only 0.01 tonnes per square metre of polyethylene is needed to meet the pregnancy threshold. Citing Johnson (1975), they argue that eliminating shielding reduces total settlement mass by a factor of 19 to 155 depending on geometry. Frontier Mark the interest and the caveat, because both are load-bearing. Globus was affiliated with San Jose State University under a NASA arrangement and the paper explicitly advocates ELEO over the classical L5 location, so it has a position. And the authors themselves say the findings “should be considered preliminary” because of gaps in understanding chronic low-level high-LET exposure, “particularly regarding pregnancy, children, and testicular effects”. A 19- to 155-fold mass reduction resting on a choice of orbit is an extraordinary claim, it is the strongest modern argument for free-space settlement, and it is made by an advocate and self-labelled preliminary.

Frontier The classical design contains an apparent contradiction and the arithmetic resolves it in the design's favour. O'Neill's cylinder is 6.4–8.0 km in diameter and 32 km long, rotating about 28 times per hour — roughly 0.47 rpm — for 1 g, with six lengthwise stripes of which three are windows, agricultural rings outside, material from the Moon and later asteroids by mass driver, and a half-pressure atmosphere at 20% oxygen and 30% nitrogen of sea-level partial pressures, chosen specifically to reduce wall thickness. Its shielding claim is that “at this scale, the air within the cylinder and the shell of the cylinder provide adequate shielding against cosmic rays.” Frontier Derived, and the derivation is elementary: atmospheric column mass per unit area is pressure divided by gravitational acceleration, so at about 50 kPa and 9.81 m/s² the air column alone is roughly 5,100 kg/m² — 5.1 tonnes per square metre, or about 510 g/cm². Set against Globus and Strout's 6–7 t/m² of polyethylene and Jones's 50 g/cm² for a fourfold GCR reduction, O'Neill's air is doing genuine shielding work — within a factor of about 1.3 of the polyethylene requirement before the hull is counted. Frontier So the classical design does not conflict with the modern numbers; it anticipates them, and it pays for them the same way. The shielding is not free: it is the atmosphere, and the atmosphere is mass that has to be lifted. An O'Neill cylinder shields itself because it is enormous and full of air, and the mass of that air is comparable to the mass of purpose-built shielding. That is a stronger statement than “the physics is textbook”, and no fetched source states it in this form.

Established What this brief owns of rotation is one equation and a table. Acceleration equals radius times angular velocity squared — Jones states it as acceleration in m/s² equal to radius in metres times the square of 2π times rpm over 60 — and gives the canonical pairing: at a maximum tolerable 4 rpm, one g requires a radius of 56 metres, a 112-metre diameter. The published designs then read as points on that curve: Jones's low-Earth-orbit baseline at 56 m and 4 rpm for 1 g; Kalpana One at 250 m and 2 rpm for about 1 g; Habitat Bennu at 3,000 m for 0.3 g; and the classical O'Neill cylinder at 3,200–4,000 m and about 0.47 rpm for 1 g. Frontier And the design-relevant inversion, which the popular treatment has backwards: rotation rate has never been the binding constraint on habitat design. Every one of those designs sits below the classical comfort limit by construction — nobody proposes a habitat at 10 rpm — because designers satisfy the rate constraint trivially by making the thing large, and making it large is what makes it expensive. The binding constraint is mass, and rotation enters mass only through radius. The artificial-gravity literature worries about rpm; the habitat literature should worry about tonnes. All physiology routes to Artificial Gravity.

Frontier Structure turns out not to be the problem, and there is now a peer-reviewed number rather than an assertion. Miklavčič, Siu, Wright, Debrecht, Askari, Quillen and Frank, in Frontiers in Astronomy and Space Sciences, analyse spinning habitats built from rubble-pile near-Earth asteroids and report a containment tensile requirement of about 200 MPa for the baseline design, with the explicit finding that hoop stress values “are well within the range of many currently existing construction materials”. Their configuration is a 300-metre asteroid expanded to a 3 km radius habitat at 0.3 g, with a 2-metre rubble layer giving protection equivalent to 100 mSv/yr; their honest caveat is that “while our study clearly relies on engineering capacities that do not exist at present”, the physics is sound. Established State the structural conclusion flatly: a kilometre-scale rotating habitat does not require exotic materials, and 200 MPa is ordinary structural steel. The usual claim that the concept needs “no new physics or exotic materials” becomes a finding rather than a slogan at that point.

Frontier Which relocates the entire problem, and explains the shape of the adjudication. If the containment vessel is undemanding, the mass is almost entirely shielding and atmosphere, and the question becomes purely logistical: where do millions of tonnes of regolith come from and how are they moved. The four dependencies on this brief are therefore not a list of related topics but the four terms of the problem — physiology sets the radius (FR-I-02), fabrication and assembly set whether you can build it (FR-I-24), and the Moon and the asteroids are the only two candidate sources of the mass (FR-II-01, FR-II-08). Frontier One apparent discrepancy between designs is worth resolving rather than presenting as a dispute. Habitat Bennu takes 2 m of rubble as giving 100 mSv/yr; Globus and Strout require 10–11 t/m² of regolith for 20 mSv/yr. At a bulk density near 1.5 t/m³, 2 m of rubble is about 3 t/m². The designs are consistent — different dose targets, different areal densities, roughly the right ratio. Dose target, not physics, separates a 3 t/m² design from an 11 t/m² one, and a dose target is a policy choice.

Frontier Now the ledger against launch cost, with every input published. The design points give 13 t/person (Jones, 100 m³, 8 g/cm²), 78 t/person (same, 50 g/cm²), 1.0 and 6.5 t/person at 1,000 crew under N^−1/3 scaling, about 2,300 t/person for Kalpana One, and about 0.01 t/m² of shielding only for an ELEO settlement below 500 km. Against a reused Falcon 9 at about $2,700 per kilogram to low Earth orbit — and noting that the compiler of that figure explicitly labels Starship's quoted $100–200/kg “a design target, not an achieved price” — the derived per-resident cost of launched shielding alone runs $2.7 million at 1.0 t/person, about $35 million at 13 t/person, and about $6.2 billion at 2,300 t/person. Frontier What that spread says, stated carefully. The lowest figure is not absurd — $2.7 million per person is comparable to a large piece of terrestrial infrastructure. The highest is prohibitive by three orders of magnitude. The difference is entirely dose target, living area and population scale: three design choices, not three technologies. That is the most useful economic observation available to this subject and no version of this page has made it. Frontier And the reason O'Neill's own answer was “don't launch it”: the classical architecture imports mass from the Moon by mass driver precisely to avoid this multiplication. There is no operating mass driver, no lunar launch infrastructure and no demonstrated regolith beneficiation, and the audited $1.2 million per kilogram in Lunar Industry is a cost of delivering to the lunar surface, not of exporting from it.

Established Closure is the same wall as Mars, and a free-space habitat has to clear it by more. There is no atmosphere, no soil and no water outside the hull, so life support must be more closed than a planetary settlement's, not less. The record, which Mars Colonization analyses and this brief cites: 98% water recovery achieved on the ISS; six months as the longest human closure ever run, one to three people at 95.4% in BIOS-3; 60% food closure as the best ever achieved, Lunar Palace 1, four crew over a year, with waste recycling not closed; the 2025 statement that “no nation has demonstrated a completely closed BLiSS system” including waste processing; and four to eight years of operational experience estimated to reach full readiness. Speculative The scaling question is this brief's own and it has no evidence base in either direction. Closure for 3,000 people in a cylinder is an ecology, not a machine, and its failure modes are ecological — trophic collapse, pathogen sweep, atmospheric drift. It may well be easier than closure for four people, since more buffer and more inertia is a real argument. But that is a hypothesis, not a result, and this brief flags it speculative rather than assuming that big closure is easy. The one historical attempt at that scale belongs to Historical Space Colonization Concepts.

Established And the brief should not be one-sided, because four free-space advantages are real and quantified. Gravity is a design variable rather than a given: a surface offers 1/6 g or 0.38 g and no alternative, while a rotating habitat offers 1 g, the only value for which human physiology has long-run data — and given that the one quantitative modelled health floor may sit at 0.4 g, above Mars, this is not a small point. Continuous solar power: no fourteen-day night, no dust storms, no seasonal variation, and none of the 40 kWe fission unit that a lunar base needs because, in NASA's words, “the sun is only available roughly two weeks out of every month”. No dust: the best-evidenced operational hazard of surface work — every environmental and gas seal failed on Apollo, suits worn out in eight hours, dials scratched unreadable, radiators degraded — simply does not exist. Location choice: ELEO puts a settlement inside the geomagnetic shield, and no planetary surface offers a choice of radiation environment. Frontier Each has an honest counterweight. Gravity is bought with radius and structural mass. Continuous sun is offset by having to radiate all waste heat with no ground to conduct into. Dust is replaced by micrometeoroid and debris exposure — ESA counts about 54,000 objects larger than 10 cm, 1.2 million between 1 and 10 cm, and 130 million between 1 mm and 1 cm, with 9.8 non-deliberate fragmentations per year on a two-decade average and business-as-usual risk projected at four times the acceptable sustainability threshold. And location choice is available only in LEO, which is exactly where the debris environment is worst.

Speculative Position one is the framing: free-space habitats beat planetary surfaces. Held by the O'Neill tradition, and explicitly by Jones at NASA Ames, whose paper concludes that surface bases “cannot adequately protect human health”. For it: gravity as a design variable, continuous solar power, no dust, N^−1/3 shielding scaling, and a choice of radiation environment. Against it: every kilogram must be imported, nothing above 420 tonnes has ever been assembled in orbit, and micrometeoroid and debris exposure replaces dust. frontier on the physiological argument, speculative on the whole comparative claim.

Frontier Position two is the most consequential live claim in the subject: ELEO settlements need essentially no shielding, which changes the economics by two orders of magnitude. Globus and Strout state it explicitly, with OLTARIS results below 500 km at minimal inclination meeting 20 mSv/yr with little or nothing, 0.01 t/m² for the pregnancy threshold, and a mass reduction of 19 to 155 times. Against it: the authors are advocates, they label the result preliminary on chronic high-LET exposure and on paediatric and pregnancy effects, and low Earth orbit is where the debris environment is worst. If it holds, free-space settlement is a different and far cheaper subject. If it does not, the deep-space numbers apply and the mass is enormous.

Established Position three: structure is not the constraint and ordinary materials suffice. Miklavčič and colleagues put the containment requirement at about 200 MPa with hoop stresses “well within the range of many currently existing construction materials”. Frontier The structural analysis is established; the inference that structure is therefore not a problem is frontier, because assembly at that scale is untested and the analysis itself concedes it “relies on engineering capacities that do not exist at present”. Established Position four: shielding mass per person falls as the cube root of population, so only very large habitats are economic. The geometry is established — 13 t/person at roughly 100 crew becomes 1.0 t/person at 1,000 crew at the same 8 g/cm². The economic argument built on it is frontier, because it assumes the entire shell is built at once, which is precisely the assembly problem nobody has solved.

Frontier Position five: a rotating habitat's atmosphere is most of its shielding. Held by O'Neill's original design, by assertion. The derived air column at half an Earth atmosphere is about 5.1 t/m², against a 6–7 t/m² polyethylene requirement — so the assertion is arithmetically defensible, and this brief carries it as derived because no fetched source states it in this form. Speculative Position six: asteroid material, not lunar material, is the right feedstock. Habitat Bennu's premise is that a rubble-pile near-Earth asteroid is simultaneously the structure and the shielding — a 300 m body expanded to a 3 km habitat needing only a containment envelope. That is a genuinely different logistics answer from lunar export and it is the main live alternative; against it, no asteroid material has ever been returned in quantity. Route to Asteroid Mining.

Speculative Position seven: closure at settlement scale is easier than closure at module scale. Implied whenever a closed ecology is assumed for a large habitat, and there is no evidence in this pack in either direction. The largest demonstrated human closure is six months with three people at 95.4%. The brief's job is to mark that as an untested and rarely stated assumption rather than to adopt it. Frontier Position eight: rotation rate is a binding constraint on habitat design. The popular treatment holds it; every published design sits well below the classical comfort limit by construction, and the binding term is radius, which binds through mass. The inversion is worth stating explicitly because the intuition runs the other way.

Speculative Position nine: free-space habitats will be built from Earth-launched material once launch is cheap enough. The modern cheap-launch argument. Against it: at the best achieved price of about $2,700/kg, per-resident launched shielding runs from $2.7 million to $6.2 billion depending on design point, and the price that would change the arithmetic — $100–200/kg — is labelled by the compiler who published it as a design target rather than a price. Speculative Position ten: O'Neill cylinders could house millions. Kalpana One houses 3,000 for 7 million tonnes. Scaling that to millions is roughly the mass of a small mountain, assembled in vacuum. Give the per-person mass and let the reader multiply; the claim runs from speculative to handwave depending on how it is stated.

Established Position eleven: nothing about the concept requires new physics. Correct, and now supported by a specific structural number rather than an assertion. Frontier And the brief should immediately add that “no new physics” is a very weak claim, since it is equally true of a bridge to the Moon. What decides this subject is not physics but four numbers that are all logistics: tonnes per square metre, tonnes per person, tonnes assembled in orbit to date, and dollars per kilogram to get there.

Established The binding bottleneck is mass, and it has a number for every design point. 6–7 tonnes per square metre of polyethylene or 10–11 of regolith for a 20 mSv/yr deep-space target; 13 tonnes per person at 8 g/cm² and 78 at 50 g/cm²; about 2,300 tonnes per person for a design that gives its residents 170 square metres each. Nothing else in this brief is close to being a constraint of that magnitude, and the constraint is not technological — it is arithmetic about areal density and dose.

Frontier The second is assembly, and the gap is four orders of magnitude. The largest structure ever assembled in orbit is the ISS, at about 420 tonnes across dozens of flights over more than a decade, at a cost of roughly $150 billion — about $0.36 billion per tonne, the only empirical figure anyone has for assembled orbital mass. Kalpana One is 7 million tonnes, which is roughly 17,000 ISSs. Space-Based Manufacturing owns how one would close that gap; this brief owns the ratio, and the ratio is the most concrete statement of the assembly problem available.

Frontier The third is that no export capability exists from either candidate mass source. The Moon has the material and a gravity well twenty-two times shallower than Earth's, and nothing has ever been launched off it except Apollo and Chang’e samples. Asteroid material has no gravity well worth speaking of and has never been returned in quantity. Both candidate supply chains are at the concept stage, and the classical answer — the mass driver — belongs to programme history rather than to any current capability.

Speculative The fourth is closure at a scale nobody has attempted. Six months, three people, 95.4% is the record; a settlement is thousands of people for decades with no resupply and no atmosphere outside the hull. Whether ecological closure gets easier or harder with scale is untested and untheorised in this pack, and treating it as easier because a bigger system has more buffer is a hypothesis dressed as a reassurance. Frontier The fifth is the debris environment, and it is worsening measurably in exactly the orbit that makes the shielding argument work. ESA's 2026 report counts 1.2 million objects between 1 and 10 cm — too small to track reliably, too large to stop with a Whipple bumper — and projects business-as-usual risk at four times the acceptable sustainability threshold. Space Habitats carries the operational side of that.

Frontier And a sixth that is not usually listed: pressure-boundary ageing over time. The only long-run data anyone has is the ISS, which has been losing air through cracks in the Zvezda transfer tunnel since 2019, with no identified root cause, a risk rating at the top of NASA's scale, a rate that rose to two pounds per day in June 2026, a structural repair paused as too risky, and a crew placed in a docked Dragon as a safe haven. A settlement hull must hold for centuries, and the one hull humanity has operated for decades has an unrepaired structural leak.

Frontier The adjudication carries four edges and they are the four terms of the problem rather than a list of related subjects. Physiology sets the radius; fabrication and assembly set whether it can be built; and the Moon and the asteroids are the only two candidate sources of the mass. Stating it that way makes the tech tree do work rather than decorate the page.

Established Artificial Gravity sets the radius. The design equation is a = rω² and the free parameter is the tolerable rotation rate; every published habitat design is a point on that curve. FR-I-02 owns what rate is tolerable, for whom, after what adaptation, and whether partial gravity protects — and this brief re-derives none of it. What follows here is that the radius sets the hull area, the hull area times the areal density sets the shielding mass, and the shielding mass is the whole cost.

Frontier Space-Based Manufacturing sets whether it can be built at all. Nothing habitat-scale has ever been assembled in space. The demand this brief states is 7 million tonnes for a 3,000-person cylinder against 420 tonnes assembled to date; FR-I-24 owns in-space fabrication, robotic assembly and whether large structures can be built rather than launched whole.

Frontier Lunar Industry and Asteroid Mining are the two candidate mass sources and this brief owns only the demand side. How many tonnes, at what areal density, for what dose target — that is here. Where the tonnes come from is theirs. The one fact this brief must carry from the lunar side, because it is routinely misread, is that the audited $1.2 million per kilogram is a cost of delivering to the lunar surface, not of exporting from it, and no lunar export capability of any kind exists. On the asteroid side, Habitat Bennu's premise — a rubble pile that is simultaneously structure and shielding — is the main live alternative and is genuinely different in kind.

Frontier What this brief does not depend on is materials science. A containment requirement of about 200 MPa is ordinary structural steel. That is a real result and it should be read as relocating the problem rather than solving it: the binding constraints are all logistics, and a stronger material would not move any of the four numbers that decide the subject.

Frontier The experiment that would decide the most is biological rather than structural: chronic low-dose-rate high-LET exposure in the populations a settlement actually contains. Globus and Strout label their own ELEO result preliminary for exactly this reason, naming gaps in understanding of chronic exposure “particularly regarding pregnancy, children, and testicular effects”, and their assumed limits — 20 mSv/yr general population, 6.6 mGy/yr in pregnancy — are the terms of the whole mass calculation. Move the dose target and the mass moves with it, by a factor of three or more between the design points in this brief.

Frontier Second: measure the ELEO radiation environment against a settlement dose target rather than an astronaut one. The claim is that below about 500 km at minimal inclination the annual dose stays under 20 mSv with essentially no shielding. That is a modelled result from OLTARIS, published by an advocate; it is the single highest-leverage claim in the subject and it is testable with instruments that already exist on vehicles that already fly. Nothing in this brief would be cheaper to check or would change more if it failed.

Frontier Third: assemble something an order of magnitude larger than has been assembled, and record what it costs per tonne. The only empirical figure is the ISS at $0.36 billion per tonne, from a programme with dozens of flights over more than a decade. Whether that number falls with automation, in-space fabrication and standardised structure is the question Space-Based Manufacturing owns, and it is the difference between a habitat costing what a national programme costs and one costing what a civilisation costs.

Speculative Fourth: run a closed ecology at a scale between a module and a settlement. The record is six months with three people. There is no result at thirty people, or three hundred, and no theory in this pack that predicts whether the failure modes get better or worse. The interesting experiment is not a bigger closure but a closure large enough for ecological rather than mechanical failure modes to appear — trophic instability, pathogen dynamics, atmospheric drift — because those are the modes that would matter at settlement scale.

Frontier And fifth, the one experiment this brief explicitly hands off: rotate something with people in it. No habitat has ever rotated and nothing funded rotates before the 2030s. The physiology, the tolerable rate and the adaptation question all belong to Artificial Gravity; what this brief needs from that experiment is a single number — the maximum tolerable rpm — because it sets the minimum radius and therefore the hull area and therefore the mass.

Established The engineering requirements for this subject are unusually easy to state and unusually hard to meet, because they are all areal densities and masses. Containment: about 200 MPa tensile in the hoop direction for a kilometre-scale spinning habitat, which ordinary structural steel meets. Shielding: 6–7 t/m² polyethylene or 10–11 t/m² regolith for 20 mSv/yr in deep space, or about 0.01 t/m² if the settlement sits in equatorial low Earth orbit and the magnetosphere does the work. Rotation: 56 m radius at 4 rpm for 1 g, scaling as the inverse square of angular velocity.

Established The one requirement that is genuinely counter-intuitive is that shielding is not monotonic in the material that habitats are usually made of. NASA's thick-shielding project finds a dose-equivalent minimum at 20–30 g/cm² of aluminium, above which secondary neutrons and light particles raise the dose again. That is why the design literature specifies hydrogen-rich materials and why lunar regolith is explicitly called ineffective for its hydrogen content, needing about 1.6 times the areal mass of polyethylene. A habitat's shield is a materials choice before it is a thickness.

Frontier Thermal is the requirement free space quietly imposes in exchange for continuous sunlight. A surface habitat conducts into the ground and shades half its sky; a free-space habitat radiates everything, and radiator area scales with the heat rejected rather than with the population's convenience. The lunar comparison is instructive in reverse: the same 40 kWe reactor needs 133.4 m² of radiator at the lunar pole and 216.2 m² at the equator, a 62% penalty, purely because the thermal environment is worse. Free space avoids the fourteen-day night and pays for it in radiators.

Frontier And the requirement nobody specifies: a pressure boundary with a design life measured in centuries. The ISS has one leak of unidentified root cause after twenty-five years, rated at the top of its operator's risk scale, with a structural repair paused as too dangerous to attempt. A settlement hull has no ground crew, no resupply of makeup gas from a planet, and no possibility of evacuation to a nearby capsule. Nothing in the design literature states a leak-rate requirement, an inspection method or a repair concept at habitat scale, and this brief names that as an absence rather than filling it.

Established The nearest neighbour is Space Habitats and the cut between them is regime, not size. FR-II-07 owns habitats that exist or are funded — the ISS, Tiangong, BEAM, Sierra Space's LIFE module, the Commercial LEO Destinations programme and its six companies — and the near-term engineering of pressure vessels, leaks, micrometeoroid protection and life support in practice. This brief owns habitats designed to be self-contained at settlement scale and the three comparative quantities in section 1. Rotation appears in both and divides cleanly: FR-II-07 covers whether any funded programme will actually spin anything; this brief covers what radius a settlement needs and why. Neither carries the other's source table.

Established Historical Space Colonization Concepts owns the 1970s and this brief narrates none of it. O'Neill's 1974 paper and The High Frontier, the 1975 Ames and Stanford summer study and NASA SP-413, the Stanford torus, Mass Driver 1, the L5 Society, and the 1981 NRC and OTA reviews of solar power satellites that removed the economic rationale — all FR-X-08's. Where this brief needs a classical parameter — the 6.4–8.0 km diameter, the half-atmosphere, the mass-driver supply chain — it states the parameter as an input to a present-day calculation and names FR-X-08. In the other direction, FR-X-08 should not carry the shielding areal densities, the 200 MPa structural result, the per-person masses or the ELEO argument; those are present-day feasibility and belong here.

Established Artificial Gravity owns all rotation physiology. This brief owns a = rω² and the design table, in one paragraph, and does not restate AGBRESA, the animal data, the historical comfort criteria or the vestibular literature. Frontier Space-Based Manufacturing owns assembly; this brief owns the mass requirement and the ISS ratio. Lunar Industry and Asteroid Mining own supply; this brief owns demand. Moon-Based Manufacturing is where the shielding-as-mass-per-area finding does its most surprising work, since it makes regolith bagging the top-ranked construction technique. And Space Resource Economies owns any claim about who pays or what a habitat is worth.

Established The institutional situation is stark and the numbers are current. GAO reported in June 2026 that NASA's projected $1–1.5 billion for FY2026–31 on commercial low-Earth-orbit destinations is, in NASA's own March 2026 clarification, “only sufficient to support one commercial space station”, against six companies in development and ISS operations running at about $3 billion a year. That is the funding environment against which a 7-million-tonne settlement has to be read, and the gap between them is not rhetorical.

Established The only empirical cost of assembled orbital mass comes from one programme. The ISS is about 420 tonnes for roughly $150 billion — $0.36 billion per tonne — and it is the denominator for every serious estimate in this subject, including the Mars cost re-derivations in Mars Colonization. Frontier A brief that wants to argue the number will fall has to argue it from in-space fabrication and automation rather than from launch price, because launch is a minority of the ISS's life-cycle cost in the same analyst's decomposition.

Frontier The advocacy institutions in this subject deserve naming rather than dismissing. The National Space Society published both the shielding paper and the Kalpana One design; Globus was affiliated with San Jose State University under a NASA arrangement; the papers explicitly advocate a position. They are also the only sources that compute the numbers this brief needs, because the agencies do not fund settlement design. Established The correct treatment is the one this page performs: take the numbers, mark the interest, print the authors' own caveats, and check the results against agency sources where any exist — which is why the Jones per-person masses and NASA's thick-shielding minimum appear alongside them.

Speculative And the institutional gap that would have to be filled before any of this is a programme rather than a literature: nobody owns the question. No agency has a settlement-scale habitat programme; no regulator has a standard for a permanent civilian population in orbit; the dose limits used in the design literature are assumed by the designers rather than set by anyone with authority; and the 6.6 mGy/yr pregnancy figure — which drives a mass requirement of hundreds of thousands of tonnes — is a designer's choice, not a regulation. The most consequential number in the subject has no institutional owner at all.

Established The ethical content of this subject sits inside a number that reads like an engineering parameter. Globus and Strout set 20 mSv/yr for the general population and 6.6 mGy/yr for pregnant women precisely because a settlement is not a crew: it contains children who did not consent, and pregnancies whose exposure is involuntary in a stronger sense than an astronaut's. Adopting an astronaut limit for a settlement would cut the shielding mass by a large factor and would be the wrong analogy, and this is one of the rare cases where the ethical judgment and the mass budget are the same decision.

Frontier The authors' own caveat is the honest version of the problem. They label the findings preliminary because of gaps in understanding chronic low-level high-LET exposure, “particularly regarding pregnancy, children, and testicular effects”. The population whose protection sets the design is the population about which least is known, and there is no ethical route to the missing data other than modelling, animal work and eventual exposure.

Speculative Second, the ELEO argument has an ethical shape that its economics conceals. A settlement that is habitable because Earth's magnetic field shields it is a settlement dependent on Earth, and residents of an orbital suburb below 500 km are within hours of rescue and within a resupply line. That is a better place to live and a worse second home, and the two evaluations point in opposite directions on the same design.

Frontier Third, the externality. Placing a permanently inhabited structure in the orbital regime with the highest debris flux — 1.2 million untrackable objects between 1 and 10 cm, 9.8 unintentional fragmentations a year, business-as-usual risk at four times the sustainability threshold — changes the calculus for everyone else who uses that shell, and no framework allocates that. Space Habitats carries the operational hazard; the governance question has no owner.

Speculative And fourth, the one this brief can only name. A settlement is a jurisdiction. Who may leave, who may be born there, who repairs the hull, and what happens when the atmosphere that is also the shielding starts to leak are questions of politics rather than engineering, and Space Resource Economies and Multi-Planetary Civilization own them. What this page contributes is the observation that the hull, the shield and the atmosphere are the same object, so life support failure and structural failure are not separable in the way they are on a planet.

Frontier The comparative verdict this brief can actually support is a conditional, and it is a good one. On the numbers in this pack: free space wins on gravity, decisively if the modelled 0.4 g health floor holds, since both the Moon and Mars sit below it and a rotating habitat can sit at 1 g. Free space wins on power and dust — no fourteen-day night, no abrasive electrostatic regolith, no seals failing in eight hours. Free space loses on mass, because a planetary surface habitat sits on top of unlimited free shielding and gets half its sky blocked for nothing. And free space loses on demonstrated capability by about four orders of magnitude: 420 tonnes assembled in orbit, at $150 billion, against 7 million tonnes for one 3,000-person cylinder.

Frontier The finding that cuts hardest against the framing is that the strongest modern argument for free-space settlement is not an argument for free space. Globus and Strout's result is that an equatorial low-Earth-orbit settlement below about 500 km meets a general-population dose target with little or no dedicated shielding, cutting total mass by a factor of 19 to 155. That is the finding that would make free-space settlement plausible, and it does so by abandoning the classical case entirely: it works because Earth's magnetosphere is doing the shielding. A habitat that depends on Earth's magnetic field for its habitability is not an independent second home; it is an orbital suburb. Deep space — L5, the classical location — needs 6–7 t/m² of polyethylene or 10–11 of regolith, and Kalpana One's own design carries 10 t/m² and about 2,300 tonnes per resident.

Established What the subject has genuinely gained since the 1970s is that its central quantities are now computed rather than asserted. The shielding requirement has an areal density from a NASA transport code. The structural requirement has a peer-reviewed tensile figure. The per-person mass has a scaling law with an exponent. The assembly gap has an empirical dollars-per-tonne from a completed programme. None of that existed when the concept was formulated, and all of it points the same way: this is a logistics problem of a size nobody has ever attempted, wrapped around a physics problem nobody needs to solve.

Speculative The terminal claim, stated as a ratio rather than a date. A million-person free-space civilisation at Kalpana One's standard of living is of order a billion tonnes of assembled mass. At the ISS's empirical $0.36 billion per tonne it is not financeable by any entity that exists; at a hypothetical order-of-magnitude improvement it is still not; and the only architecture that escapes the multiplication is the one that never launches the mass from Earth at all. That is why the four dependencies on this brief are the mining and manufacturing ones, and why a page about cylinders is really a page about where several million tonnes of regolith come from.

These horizons track mass, assembly capability and one radiation result, because those are the three things that decide the subject — not materials, and not rotation:

• 10 yr: Established Expect no settlement-scale habitat and no rotating habitat of any kind: GAO reports NASA funding sufficient for one commercial station against six competitors, and nothing funded rotates before the 2030s. Frontier Expect the ELEO shielding claim to be either checked against flight dosimetry at settlement dose targets or to remain a modelled result from an advocacy paper; it is cheap to test and it is the highest-leverage claim in the subject. Frontier Expect the assembly cost per tonne to be the number worth watching, since it is the only one that could move by an order of magnitude on this horizon.
• 25 yr: Speculative A rotating habitat at the low end of the design table — tens of metres of radius, a handful of people, inside the magnetosphere — is the plausible article at this horizon, and it is a spacecraft rather than a settlement. Speculative The settlement question turns on whether any lunar or asteroid export capability exists, and today there is none: no mass driver, no lunar launch infrastructure, no beneficiation, and no asteroid material returned in quantity. Route to Lunar Industry and Asteroid Mining. Frontier And the closure question would need a facility running for years, which nobody is building.
• 50 yr: Speculative A habitat housing hundreds is coherent at this horizon if in-space fabrication and a non-terrestrial mass source both arrive, because the structural requirement is ordinary steel and the physiology is bought with radius. Speculative A habitat housing thousands at Kalpana One's standard means moving millions of tonnes, which is a different civilisational activity rather than a bigger project, and the honest thing a brief can do is give the per-person mass and let the reader multiply.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The defensible statement is which constraints are permanent and which are not: the areal density needed for a given dose is physics and does not improve; the cost of putting a tonne in orbit and the cost of assembling it are engineering and can. Speculative A subject whose entire difficulty is logistics rather than physics has no natural ceiling and no base rate either — which is a more honest statement than a date, and more useful than either enthusiasm or dismissal.

• Depends on Four edges, and they are the four terms of the problem rather than a list of neighbours. Artificial Gravity sets the radius: the design equation is a = rω², the free parameter is the tolerable rotation rate, and at a maximum tolerable 4 rpm one g requires a 56-metre radius. Space-Based Manufacturing sets whether it can be built: the largest structure ever assembled in orbit is about 420 tonnes at roughly $150 billion, and Kalpana One is 7 million tonnes — about 17,000 ISSs. And Lunar Industry and Asteroid Mining are the only two candidate sources of the mass, since the shielding requirement is 6–7 t/m² of polyethylene or 10–11 t/m² of regolith for a 20 mSv/yr deep-space target and every kilogram must come from somewhere with a gravity well. This brief owns the demand side — how many tonnes, at what areal density, for what dose target — and hands off supply, noting that the audited $1.2M/kg is a cost of delivering to the lunar surface and that no lunar export capability exists.
• Requires (not on this map) The first sets the dose target and therefore the entire mass budget: the design literature assumes 20 mSv/yr for a general population and 6.6 mGy/yr in pregnancy, and its own authors label those results preliminary because chronic low-level high-LET effects on pregnancy, children and testicular tissue are not understood. Move that target and the shielding mass moves by a factor of three or more. The second is not a brief's result but an industry that does not exist: there is no operating mass driver, no lunar launch infrastructure, no demonstrated regolith beneficiation and no asteroid material returned in quantity, and every kilogram of a deep-space habitat has to come from one of those two places or be launched from Earth at about $2,700 per kilogram.
• Enables No typed enabling edge is claimed. A settlement-scale free-space habitat would enable a great deal — permanent off-Earth population, industry with no gravity well, a genuine second location for the species — but nothing at that scale has been built, the largest orbital structure ever assembled is four orders of magnitude short, and an edge from a capability that does not exist to capabilities that would depend on it records an aspiration rather than a dependency. Multi-Planetary Civilization owns the argument about what such a habitat would mean, and takes this brief's numbers as inputs from its own side.
• Adjacent Space Habitats across a regime boundary rather than a size one: that brief owns habitats that exist or are funded and the practical engineering of leaks, debris and life support, and this one owns settlement-scale design. Historical Space Colonization Concepts owns the 1970s programme history in full. Moon-Based Manufacturing is where this brief's shielding-as-areal-density finding does its most surprising work, since it makes regolith bagging the top-ranked lunar construction technique; Mars Colonization supplies the closure ladder and the measured deep-space dose; and Space Resource Economies owns who pays.

Established “It needs materials we don't have — carbon nanotubes, unobtainium, something.” It does not. The peer-reviewed structural analysis of a kilometre-scale spinning habitat puts the containment tensile requirement at about 200 MPa and states that hoop stresses “are well within the range of many currently existing construction materials”. That is ordinary structural steel. Frontier The finding overturns the usual framing rather than confirming it: if structure is undemanding, then the mass is almost entirely shielding and atmosphere, and the subject is a logistics problem rather than a materials one.

Established “No new physics is required, so it is basically ready.” The first half is true and now well supported. The second does not follow, and “no new physics” is a very weak claim — it is equally true of a bridge to the Moon. Established What decides this subject is four logistics numbers: tonnes per square metre of shielding, tonnes per person, tonnes ever assembled in orbit, and dollars per kilogram to get there. The largest structure ever assembled in orbit is 420 tonnes at about $150 billion; one 3,000-person cylinder is 7 million tonnes.

Frontier “The rotation rate is the hard part — people will get sick.” Rotation rate has never been the binding constraint on habitat design, because every published design sits well below the classical comfort limit by construction: 4 rpm at 56 m, 2 rpm at 250 m, about 0.47 rpm at 3,200–4,000 m. Frontier Designers satisfy the rate constraint trivially by making the habitat large, and making it large is what makes it expensive. The binding term is radius, and radius binds through mass. Whether a given rate is tolerable, for whom and after what adaptation, belongs entirely to Artificial Gravity.

Established “Shielding is a detail.” Shielding is the design. 6–7 tonnes per square metre of polyethylene, or 10–11 tonnes per square metre of lunar regolith, for a 20 mSv/yr general-population target in deep space; 13 tonnes per person at 8 g/cm² and 78 tonnes per person at 50 g/cm² at 100 m³ of living space. Established And regolith is explicitly the worse material — “ineffective due to low hydrogen content”, needing about 1.6 times the areal mass of polyethylene — which is inconvenient, because regolith is the material that is actually available.

Established “More shielding is always safer.” NASA's own thick-shielding project finds a dose-equivalent minimum at 20–30 g/cm² of aluminium, beyond which secondary neutrons and light particles raise the dose again — reported explicitly against the conventional assumption. Interest running against the finding, from the programme whose primary lever it constrains.

Frontier “Deep space at L5 is the natural location.” The only modern analysis that makes the mass arithmetic work does so by moving the settlement into Earth's magnetosphere: equatorial low Earth orbit below about 500 km, meeting a 20 mSv/yr general-population limit with little or no dedicated shielding and cutting total settlement mass by 19 to 155 times. Frontier That result is the strongest modern argument for free-space settlement and it dissolves the classical case in the same breath — a habitat whose habitability depends on Earth's magnetic field is an orbital suburb, not a second home. It is also made by an advocate and labelled preliminary by its own authors.

Frontier “Equatorial LEO is the safe orbit, then.” It is the orbit with the worst debris environment. ESA counts about 54,000 objects larger than 10 cm, 1.2 million between 1 and 10 cm, and 130 million between 1 mm and 1 cm, with 9.8 unintentional fragmentations per year on a two-decade average and business-as-usual risk projected at four times the acceptable sustainability threshold. Frontier The 1–10 cm population is the dangerous one — too small to track reliably, too large to stop with a bumper — and the two hazards, radiation and debris, trade against each other with no source in this pack analysing them jointly. Name that as a gap.

Speculative “Cheap launch will build them.” At the best achieved price — a reused Falcon 9 at about $2,700 per kilogram — per-resident launched shielding alone runs from $2.7 million to $6.2 billion depending on design point. Established The figure that would change the arithmetic, $100–200/kg, is labelled by the compiler who published it as “a design target, not an achieved price”, on a vehicle with twelve flights, seven successes, and an unattempted ship-to-ship propellant transfer.

Speculative “A big habitat's ecology will be more stable than a small one's.” There is no evidence in either direction. The largest human closure ever run is six months, one to three people, at 95.4%; the best food closure is 60% over a year with waste uncontained; and no nation has demonstrated a completely closed system including waste. Speculative More buffer and more inertia is a real argument, and it is a hypothesis rather than a result, and the failure modes at settlement scale are ecological rather than mechanical, which is a category the record does not cover at all.

Established And the framing itself. “Free-space habitats beat planetary surfaces” is not true or false; it is underspecified, and the useful thing this page can do is specify it. Established Free space wins on gravity, power and dust. It loses on mass, by a margin that depends entirely on the dose target chosen — a policy decision, not a physical one — and it loses on demonstrated capability by four orders of magnitude. Frontier And the one design regime where the mass margin disappears is low equatorial Earth orbit, where the shielding is Earth's. That is the honest terminal position, and it is a conditional rather than a verdict.