Stellar Engineering

Stellar engineering means acting on a star itself: accelerating it, removing mass from it, forcing it to mix, or igniting a body that is not one. It is the most extreme entry in this category and it is also the thinnest. The framing under test is that stars themselves are an engineering target.

This page is deliberately written to look weak, because the evidence is. The field consists of a Halley Lecture in 1948, a conference talk in 1983, an astronautical-congress paper in 1987, a note in 1997, a book chapter in 2006, one journal paper in 2019 with a published Comment against it in 2021, an observational constraint in 2020, and a stability theorem in January 2026. The flagship modern paper has seven citations seven years after publication, which is the quantitative statement of the field's size and belongs in the brief rather than in a footnote.

What the page can do, and what the previous version of it did not, is arithmetic. Every performance figure in circulation for a stellar engine can be checked against solar quantities, and doing so changes what the numbers mean: the standard Shkadov thrust turns out to be the Sun's entire radiative momentum flux rather than a design value; the standard star-lifting rate turns out to sit at about 93% of the thermodynamic floor, which describes a lossless machine rather than an engineered one; and the popular-press displacement figures fail an internal consistency check against the accelerations they are supposedly derived from.

The boundary with Dyson Swarms is the star. That brief owns collectors in orbit; this one owns anything that acts on the star. The two share exactly one paper — McInnes (2026), which treats Dyson bubbles and stellar engines in a single stability framework — and each takes its own half. This brief also inherits from it one named, unworked problem: what an enclosing structure does to the star inside it, which nobody has studied.

Speculative The idea is older than the space age, and its founding statement is a lecture rather than a paper. Fritz Zwicky, delivering the Halley Lecture at Oxford on 12 May 1948, proposed that humanity could “accelerate [the Sun] to higher speeds, for instance 1000 km/s directed toward Alpha Centauri,” using “nuclear fusion jets, using the matter constituting the Sun and the planets as nuclear propellants.” It reaches this brief through Badescu and Cathcart's chapter in Macro-Engineering: A Challenge for the Future (Springer, 2006), not through a retrievable original.

Speculative The same chapter carries the other founding claims and they are all conference talks or notes. David Criswell, at the Conference on Interstellar Migration in Los Alamos in May 1983, proposed removing solar plasma annually over about 300 million years to convert the Sun into a stable white dwarf, beginning some time between AD 2170 and 5650. Oliver Knill (1997) proposed deliberately asymmetric fusion — induced solar flares used as rocket motors. Badescu and Cathcart give an optimum stellar-engine radius of about 450 million kilometres and the Class A / B / C taxonomy the field still uses.

Established The dates are themselves the finding: 1948, 1983, 1987, 1997, 2006, 2019, 2020, 2026. Eight substantive contributions across seventy-eight years, from perhaps eight people. For comparison, the Dyson-sphere search literature covered in Dyson Swarms produced four papers in 2024–2026 alone. Stellar engineering is not a small field within astronomy; it is a handful of individuals writing decades apart.

Speculative The canonical design is the Shkadov thruster, and it is a mirror rather than a rocket. Shkadov's Possibility of controlling solar system motion in the Galaxy, presented at the 38th International Astronautical Congress in 1987, places a curved statite reflector on one side of a star, held stationary against gravity by radiation pressure, so that the reflected photon momentum accelerates the whole system. It is a Class A engine: propulsive, using radiation pressure only, adding no mass loss. No accessible copy of Shkadov's paper exists, and everything about it here arrives through Badescu and Cathcart and through secondary tabulations.

Speculative The performance figures in circulation are internally consistent, which is worth checking before repeating them. Thrust for the Sun, 1.28 × 10¹⁸ N; speed after one million years, 20 m/s, for a displacement of 0.03 light-years; speed after one billion years, 20 km/s, for 34,000 light-years. Divide the thrust by the solar mass and the acceleration is 6.4 × 10⁻¹³ m/s²; over a million years that gives exactly 20 m/s and 0.034 light-years. The set reproduces itself.

Speculative But the thrust figure is not a design value, and this is the correction the brief exists to make. The Sun's luminosity divided by the speed of light — 3.828 × 10²⁶ W over 2.998 × 10⁸ m/s — is 1.277 × 10¹⁸ N. The quoted 1.28 × 10¹⁸ N is therefore the entire momentum flux of the Sun's radiation, not the fraction a one-sided mirror could intercept and reflect. A real Shkadov mirror captures some part of that. The tabulated performance is an upper bound presented as a specification, and every account this brief retrieved repeats it without the caveat.

Handwave The popular-press numbers fail the same check, and the failure is worth one line because these are the figures most readers will have seen. ScienceAlert reports a Shkadov thruster that “could probably propel us 100 light years over 230 million years.” Under constant acceleration distance scales as time squared, so 0.034 light-years at one million years implies 0.034 × 230², about 1,800 light-years at 230 million years — eighteen times the reported figure. The published tabulation and the press account are not reconcilable, and averaging them would be worse than either.

Frontier And the design has a known defect that was repaired only this year. Colin McInnes, Stellar engines and Dyson bubbles can be stable (MNRAS 546:1–18, accepted 13 January 2026), finds that ultra-large reflectors in static equilibrium above a star “are always unstable if the reflector comprises a uniform disc”: when the disc's scale approaches its distance from the star the gravitational field deviates from inverse-square, and the deviation destabilises it. His repair is to concentrate the mass at the edge. For a ring-supported reflector the stability condition becomes positive for a scale parameter above 1 over root two — a disc subtending a half-angle of about 55 degrees as seen from the star. The critical areal density for levitation in the solar case is about 1.5 × 10⁻³ kg/m².

Frontier Put that areal density in context, because it is the entire engineering problem in one number, and it is the field's one encouraging result. One and a half grams per square metre is roughly the areal density of a very thin solar sail, and Solar Sail Systems owns the flight record in that regime. A mirror subtending 55 degrees at one astronomical unit covers of order 10²² m², so its mass is of order 10¹⁹ kg — about one ten-thousandth of Mercury. The Shkadov thruster is mass-cheap and time-poor. It is the exact opposite trade to a Dyson swarm, which is mass-expensive and works immediately.

Speculative Buying acceleration with mass loss is the Class D alternative, and it has the field's one modern journal paper. Matthew Caplan, Stellar engines: Design considerations for maximizing acceleration (Acta Astronautica 165:96–104, 2019, Illinois State University): a fusion-powered engine consuming solar material and expelling two jets, one to push the star and one to hold the engine's own position. Reported parameters — mass flow 10¹² kg/s, maximum acceleration 10⁻⁹ m/s², 200 km/s after five million years, 10 parsecs travelled per million years, and a lifetime of 100 million years theoretically, about 10 million practically.

Speculative The mass budget survives and the operating requirement does not. Ten million years at 10¹² kg/s consumes 3.2 × 10²⁶ kg, which is 1.6 × 10⁻⁴ of a solar mass — about fifty Earth masses, and entirely survivable for the star. The acceleration is 1,560 times the Shkadov figure, which is a real improvement. It is bought by running a fusion plant that processes a trillion kilograms per second, continuously, for ten million years. The popular reporting again disagrees with the tabulation — 50 light years in a million years against a tabulated 10 parsecs, about 33 — and the tabulated figure is the one carried here.

Established And here is the measurement of the field's size, which is more informative than any adjective. The OUCI bibliographic record for Caplan (2019) reports seven citations and 34 references. Seven citations for the field's most-covered modern paper, seven years after publication. The same paper drew a formal published Comment — Comments on “Stellar engines: Design considerations for maximizing acceleration”, Acta Astronautica 179:495 (2021) — whose content could not be retrieved for this brief. That the flagship paper attracted a published objection is reportable; what the objection says is not.

Speculative Star lifting is the other branch, and its usual telling escapes the arithmetic. The concept, attributed to Criswell's chapter in Finney and Jones's Interstellar Migration and the Human Experience (UC Press, 1985): build a partial shell of collectors, use the captured energy to drive mass off the star, and thereby reduce its mass, extend its main-sequence life and harvest material. The named methods are thermal-driven outflow, Criswell's huff-n-puff using the star's own magnetic field cycled through ring stations, and a polar or centrifugal method with rings rotating faster than the star, which the literature itself says “suffers from a number of significant complications.”

Speculative The one quantitative anchor is a rate, and checking it against thermodynamics is the most useful thing this brief does. The thermal-driven method at 10% of the Sun's total power output is quoted as giving a mass removal rate of 5.9 × 10²¹ kg/year. Ten per cent of solar luminosity is 3.83 × 10²⁵ W, or 1.21 × 10³³ J per year; divided by 5.9 × 10²¹ kg that is 2.05 × 10¹¹ J per kilogram lifted. The gravitational binding energy per kilogram at the solar surface is GM over R, or 1.91 × 10¹¹ J/kg. The quoted rate therefore corresponds to a lifting efficiency of about 93% — which is to say the figure in circulation describes an essentially lossless machine operating at the theoretical minimum. It is a floor, not an estimate, and it should never be quoted as an engineering performance.

Speculative Now what the rate buys, which is the sentence this brief was written for. Removing 1% of the Sun — 1.99 × 10²⁸ kg — at 5.9 × 10²¹ kg/year takes 3.4 million years. Main-sequence lifetime scales as mass over luminosity, and luminosity goes roughly as mass to the 3.5 power, so lifetime goes as mass to the minus 2.5. A 1% mass reduction therefore buys about 2.5% more main-sequence lifetime — on a ten-billion-year budget, roughly 250 million years. Run a machine consuming a tenth of the Sun's entire output, continuously, for three and a half million years, and you buy about a quarter of a billion years. The return is enormously positive — roughly seventy to one in time — and it is unreachable by anything resembling a human institution.

Speculative The most recent quantitative treatment is by an advocate, and it concedes the case. Gregory Matloff, Star Lifting: An Application for Alien Megastructures (JBIS 70:458–460, 2017), describes lifting by projecting “a collimated electromagnetic beam into the star's photosphere” to increase the solar wind — and then concludes: “To be effective, Star Lifting will be a very-long-duration process. Therefore, an advanced space-faring civilization might elect to preserve its biosphere during the post-main-sequence phase of its host star's existence using interstellar migration as an alternative.” A paper written to make the case for star lifting concludes that migrating is probably easier. That is interest against the finding, from inside the field, and it is the strongest single datum about the technology's standing.

Handwave The adjacent hypotheses reach this brief entirely through one secondary account and are flagged accordingly. Martyn Fogg (1989) proposed stellifying a gas giant — igniting Jupiter using a small black hole in a controlled orbit — to give “100 million years of habitability to the Jovian satellite system.” Martin Beech (1990) proposed extending a star's main-sequence lifetime by forcing mixing, creating a hot spot between core and surface by detonating a series of hydrogen bombs or by directing a concentrated laser at the surface, claiming extension “by factors of ten or more.” Milan Cirkovic proposed the detection side: a stellified Jupiter would sit “ten orders of magnitude or so above the expected luminosity” for its mass, an outlier from the luminosity–mass relation, with spectral anomalies and gamma and X-ray flares during transformation. All of this comes via a single 2016 Centauri Dreams article which itself notes that the originals' full publication details are not given.

Established And then the measurement, which is the established island in an otherwise unmeasured subject. Manasvi Lingam and Abraham Loeb, Constraints on the abundance of 0.01c stellar engines in the Milky Way (arXiv:2009.08874 v2, 31 October 2020; Astrophysical Journal), used Gaia DR2 — about 10⁹ stars with proper motions, about 10⁷ with radial velocities, reducing to roughly 10⁶ solar-mass-or-greater candidates capable of hosting an engine — and set a velocity threshold of 0.01c. The fastest confirmed hypervelocity star, S5-HVS1, moves at about 6 × 10⁻³ c, with another candidate at 5.4 × 10⁻³ c. Nothing reaches the threshold. The published constraint is that the product of the fraction of stars hosting human-level intelligence and the fraction of those deploying a stellar engine is below 10⁻⁶, tightening to 10⁻⁸ with complete Gaia data.

Speculative The second detection route is transit photometry, and its own author's verdict is negative in a useful way. Duncan Forgan, On the possibility of detecting class A stellar engines using exoplanet transit curves (arXiv:1306.1672, 2013; JBIS), notes that a Shkadov mirror is a spherical arc that occults part of the stellar disc, distorting the shape of any planetary transit light curve. His conclusion is that “the a priori probability of detecting a Shkadov thruster during an exoplanet transit remains stubbornly low” even under optimistic assumptions — while remaining “a useful serendipitous SETI technique” able to set local upper limits on Class A engines.

Handwave Finally, the problem this brief inherits and cannot answer. Wright's Dyson spheres review states that “the overall effects of Dyson spheres on the total luminosity of stars have not previously been studied, and might lead to strong observable consequences.” The physics is that stars have negative heat capacity by the virial theorem, so returning energy to a star should make it expand, cool and dim. Nobody knows what harvesting a star does to the star. Planetary Scale Energy Systems named it in one sentence, Dyson Swarms passed it along without pursuing it, and it lands here because it is a thing done to a star. It is the most interesting open question in the slot and there is no literature on it at all.

Handwave Position one, the founding one: the Sun could be accelerated to about 1000 km/s toward Alpha Centauri using fusion jets fed by solar and planetary matter. Zwicky, 1948, in a lecture, with no calculation retrievable. It is carried because it is the origin of the subject and because a reader should be able to see how thin that origin is.

Handwave Position two: solar plasma could be removed annually over about 300 million years to convert the Sun into a stable white dwarf, starting between AD 2170 and 5650. Criswell, 1983 and 1985. The date range is the striking feature: a proposal with a start window and a three-hundred-million-year duration, made at a conference, never followed by a design.

Speculative Position three: a statite mirror can accelerate a star using its own radiation pressure. The Shkadov thruster, and the only Class A design anyone has proposed. Position four, its performance: 1.28 × 10¹⁸ N, 0.03 light-years per million years, 34,000 light-years per billion. Self-consistent, and resting on a thrust equal to the Sun's entire radiative momentum flux — so it is an upper bound rather than a specification, and no retrieved source says so.

Frontier Position five: a uniform-disc stellar engine is always unstable, and an edge-loaded ring subtending more than about 55 degrees is passively stable. McInnes, MNRAS 546 (2026). Position six: levitation requires an areal density near 1.5 × 10⁻³ kg/m² in the solar case, scaling with stellar mass. These two are the only positive engineering results in the field and they are both nine months old.

Speculative Position seven: a fusion-fed Class D engine gives 10⁻⁹ m/s², ten parsecs per million years and 200 km/s in five million years, on a trillion kilograms per second for about ten million years. Caplan, 2019. Position eight: the design has published objections — a formal Comment in Acta Astronautica in 2021, whose content this brief could not retrieve and therefore does not characterise. Position nine: a star tug could reach about 27% of light speed given sufficient stellar mass conversion — Svoronos (Acta Astronautica, 2020, Yale), abstract not retrieved, carried as a claim with a venue and nothing more.

Speculative Position ten: star lifting at 10% of solar luminosity removes 5.9 × 10²¹ kg a year. The field's one quantitative anchor, and it corresponds to about 93% of the thermodynamic floor. Position eleven, derived here: 3.4 million years of that removes 1% of the Sun and buys roughly 250 million years of extra main-sequence life. Position twelve: star lifting works, and interstellar migration is the more likely choice anyway — Matloff (2017), the field's own concession.

Handwave Position thirteen: a gas giant could be stellified with a small black hole, giving about 100 million years of habitability to its moons. Fogg (1989), via a secondary account. Position fourteen: stellar lifetime can be extended tenfold by forced mixing, using staged hydrogen bombs or a concentrated laser. Beech (1990), same route. Position fifteen: stellified objects would be detectable as luminosity outliers about ten orders of magnitude above the mass–luminosity relation. Cirkovic, same route. All three reach this brief through one 2016 article that says the originals' publication details are incomplete.

Established Position sixteen: fewer than one star in 10⁶ hosts a 0.01c stellar engine, tightening to 10⁻⁸ with complete Gaia data. Lingam and Loeb (2020), on real catalogue data with a stated threshold. Position seventeen: Class A engines are detectable as distortions in exoplanet transit curves, at a stubbornly low a priori probability. Forgan (2013), whose own assessment is the negative one.

Handwave Position eighteen, and the most interesting: building a swarm changes the star, because negative heat capacity implies expansion, cooling and dimming — and nobody has studied it. Wright (2020). Position nineteen: the optimum stellar-engine radius is about 450 million kilometres, and the A/B/C taxonomy organises the design space. Badescu and Cathcart (2006). Position twenty: induced asymmetric solar fusion — flares deliberately used as rocket motors. Knill (1997), via the same chapter, with no analysis retrievable.

Speculative The binding constraint in this subject is not a technology and is not a material. It is the ratio of stellar quantities, and no engineering changes it by more than a small factor. A Shkadov thruster's acceleration is bounded by the Sun's radiative momentum flux over its mass, which is 6.4 × 10⁻¹³ m/s² at the absolute limit. A star-lifting machine's energy cost per kilogram is bounded below by GM over R, which is 1.91 × 10¹¹ J/kg. These are not schedules a better technology shortens. They are properties of the star.

Speculative The second bottleneck is duration, and it is the one that makes the subject institutionally unreachable rather than physically impossible. Three and a half million years of continuous operation to move 1% of a star's mass. Ten million years of a trillion kilograms per second for a Class D engine's practical lifetime. A million years to move a star a thirtieth of a light-year with a Class A mirror. There is no human institution with a demonstrated lifetime beyond a few thousand years, and every stellar-engineering proposal requires continuous operation for a thousand times that.

Frontier The third is the one bottleneck that is genuinely engineering-shaped, and 2026 improved it. A Shkadov mirror must hold station without active propulsion at an areal density of about 1.5 g/m², and until this year the design was known to be unstable. McInnes's edge-loading result gives a passive configuration that works. The remaining problem is manufacturing 10²² square metres of it and keeping it intact for geological time, and that is a self-replication problem this brief does not own and no brief on this map does.

Handwave The fourth is that nobody knows what the star does back. Every design assumes the star is a fixed boundary condition — constant luminosity, constant mass distribution, constant photospheric behaviour under a beam or a shell. Wright's observation that this has never been studied means every performance figure in this brief has an unquantified error bar attached to it in an unknown direction. An engineering field whose central object has never been modelled as responding is not yet an engineering field.

Speculative And a fifth that is about the literature rather than the physics: there is almost nothing to read. Seven citations for the flagship paper. One published Comment, unretrievable. A star-lifting rate whose provenance runs through tertiary summaries. Founding proposals in a lecture, a conference talk and a book chapter. The bottleneck for anyone wanting to work on this is that there is no body of work to join.

Speculative The adjudicated dependency is Dyson Swarms, and the line between them is the star. That brief owns anything in orbit around one: swarm architecture, collector mass budgets, orbital mechanics, the shell-buckling result, the waste-heat searches and the Hephaistos candidate list with its resolution. This brief owns anything acting on the star — moving it, lifting mass from it, mixing it, igniting a planet into one. They share exactly one paper by design. McInnes (2026) treats Dyson bubbles and stellar engines in a single stability framework; that brief takes the bubble half and this one takes the Shkadov-mirror half.

Speculative The dependency is real rather than formal, and it runs through the collector. Every Class A design is a statite; every star-lifting design begins with a partial shell of collectors capturing 10% of a star's output. A stellar engine is a Dyson-swarm component pointed at a different purpose, which is why the stability mathematics is shared and why the areal-density number appears in both briefs.

Speculative The second dependency is a scientific result nobody is producing, and it arrives here by handoff. Planetary Scale Energy Systems named the radiative-feedback question in one sentence and Dyson Swarms passed it on. Stars have negative heat capacity; an enclosing structure that returns energy inward should make its star expand, cool and dim; Wright says the effect has never been studied. Until somebody models it, every stellar-engineering design has an unknown-sign correction on its central assumption.

Speculative What this brief does not borrow is as important as what it does. Nothing from FR-I-23's space-based-solar-power economics, beaming efficiencies or Kardashev critique appears here; that inheritance is two removes away and arrives only as the unstudied-feedback question. And every stellar-engineering scenario presumes a manufacturing base far beyond anything Space-Based Manufacturing or Orbital Shipyards documents. That presumption is named here and not supported by borrowing their evidence for it.

Established One experiment in this subject has been performed and it is an astronomical survey. Lingam and Loeb's constraint is a real observational test of the framing at galactic scale: define the signature (a star moving above 0.01c), search a catalogue of 10⁹ objects, and publish the upper limit. It came back negative at 10⁻⁶ and will reach 10⁻⁸ with complete Gaia data. No other proposition in this brief has been tested against anything.

Speculative The second observational test exists on paper and has not been run at scale. Forgan's transit-curve method looks for the distinctive asymmetry a spherical arc mirror imposes on a planetary transit light curve. With millions of stars already photometered, running it is a data-analysis project rather than a facility request. Forgan's own a priori probability is stubbornly low, but a systematic null across the existing transit archive would still be a publishable upper limit on Class A engines, and no retrieved source reports one.

Speculative The third is the stellification test, which is unusually cheap because the signature is enormous. An artificially ignited gas giant would sit about ten orders of magnitude above the mass–luminosity relation. Searching existing catalogues for objects that far off the relation is straightforward, and this brief found no report of anyone having done it. The proposal reaches this pack through a secondary account with incomplete citations, so even the signature specification would need to be reconstructed from primaries nobody has retrieved.

Speculative The fourth is theoretical and is the one that would matter most: model a star with an enclosing structure returning energy to it. This is a stellar-structure calculation with existing codes, not a new instrument, and the physics — negative heat capacity, so heating causes expansion and cooling — is standard. Wright called it an area ripe for investment in 2020 and nothing retrieved for this brief has taken it up. It would put an error bar on every figure on this page.

Speculative And a fifth that has to be stated as impossible rather than merely unfunded. No component of a stellar engine can be tested at scale, because the scale is a star. A statite demonstration would test the levitation regime; nothing tests the thrust, the mass flow, the lifting efficiency or the century-on-century durability, and nothing ever will short of building one. This is a field where the experimental programme is entirely observational by necessity, and where the engineering can only ever be checked by arithmetic.

Speculative Requirements can be listed for the Class A mirror because McInnes specified them, and this is the only design in the brief with a real specification attached. Areal density about 1.5 × 10⁻³ kg/m² for levitation in the solar case. Mass concentrated at the edge rather than uniformly distributed. A half-angle above about 55 degrees as seen from the star, for passive stability. Total area of order 10²² m² at one astronomical unit, for a mass of order 10¹⁹ kg.

Speculative Requirements cannot be listed for anything else in the brief, and the reason is worth stating rather than glossing. The Class D engine has a mass flow rate and an acceleration and no design for the fusion plant that would produce them. Star lifting has an energy rate and no specification of the collector array, the beam, the magnetic hardware or the ring stations. Stellification has a black hole and no account of how one is obtained or steered. A list of performance figures is not a requirements set, and the difference is that a requirements set can be argued with.

Frontier The one genuinely encouraging engineering number in the subject is the mirror mass. Of order 10¹⁹ kg is about one ten-thousandth of Mercury — four orders of magnitude less material than the Dyson-swarm scenarios in Dyson Swarms require. The Shkadov thruster is the cheapest megastructure in this category by mass and the most expensive by time, and that trade is the most useful thing to know about it.

Speculative And the durability requirement, which no proposal addresses at all. A structure of 10²² square metres at gram-per-square-metre thickness must survive micrometeoroids, stellar flares, charged-particle erosion and its own thermal cycling for millions to billions of years, with no retrieved source estimating a lifetime, an erosion rate or a replacement schedule. Continuous operation for geological time is the requirement every design assumes and none specifies.

Speculative The nearest neighbour is Dyson Swarms, and it is a dependency rather than a mere adjacency. A stellar engine is a swarm component with a different purpose: the same areal densities, the same station-keeping physics, the same shared stability paper. The boundary is the star, and it is worth restating because the two subjects are routinely merged in popular accounts.

Frontier The technically closest live field is solar sailing. Solar Sail Systems owns flight-proven areal densities in the grams-per-square-metre regime, which is exactly the regime McInnes's levitation criterion requires, and McInnes is himself a solar-sail dynamicist. A statite — a sail that holds station rather than travelling — would be the first physical hardware bearing on any claim in this brief, and it is within reach of a real mission.

Speculative On the search side the division is purpose rather than dataset. Interstellar Archaeology owns technosignature searching as a research programme — its instruments, methodology and nulls across the whole field. This brief carries Lingam and Loeb's Gaia constraint because it is a measurement about stellar engines specifically, and Forgan's transit method for the same reason. If both briefs cite the same work they should cite it for different claims: abundance of engines here, search methodology there.

Speculative And the far adjacency, which is where the subject's own advocate points. Matloff's conclusion is that a civilisation facing its star's death would probably migrate rather than engineer, which routes the whole subject toward Interstellar Civilization Models and Interstellar Probes. Stellar engineering is the alternative to leaving, and the only person to have costed both in one paper recommended leaving.

Established The institutional statement here is a citation count, and it is the honest measure. Caplan (2019) is the field's most-covered modern paper — the one that produced worldwide press about moving the solar system — and its bibliographic record shows seven citations across seven years. There is no funded programme, no agency line item, no laboratory, no conference series and no institute. The subject exists as a small number of individually motivated papers in general astronomy and astronautics journals.

Speculative The one institutional behaviour worth crediting is that the field publishes in real venues and takes real objections. Caplan's paper is in Acta Astronautica and drew a formal published Comment two years later. Lingam and Loeb published in the Astrophysical Journal. McInnes published in MNRAS. None of this is grey literature, which distinguishes stellar engineering favourably from several better-funded subjects on this map.

Speculative The institutional problem is duration and it has no solution in the retrieved record. Criswell's proposal has a start window between AD 2170 and 5650 and a three-hundred-million-year duration. The star-lifting arithmetic here requires 3.4 million years of continuous operation. No mechanism for maintaining an intention across those timescales appears anywhere in the literature, and the field does not treat it as a question. That silence is more informative than any technical objection.

Speculative And an institutional caution about this page's own sources. NASA's Astrophysics Data System and ScienceDirect were both unreachable throughout the research pass behind this brief. Shkadov's founding paper has no accessible copy anywhere; the published Comment on Caplan could not be read; Svoronos's abstract could not be read; Fogg, Beech and Cirkovic arrive through a single secondary article that flags its own incomplete citations. Roughly half the positions on this page are known only at second hand, and the brief says so at every point rather than presenting them as read.

Speculative Stellar engineering has no near-term ethical content, and pretending otherwise would be the characteristic error of a page like this. Nothing described here is achievable on any horizon a governance system could act on, and no decision available to anyone now is affected by it. The honest statement is that this section is about a hypothetical, and the flag reflects that.

Speculative The one genuinely interesting consideration is irreversibility, and it is unlike anything else in this category. Removing a per cent of a star's mass is permanent on any timescale a civilisation experiences. Igniting a gas giant destroys whatever the Jovian system was. Moving a star relocates every world around it, including any that were not consulted. These are the only proposals on this map whose consequences are literally unrecoverable, and the literature treats every one of them purely as an engineering trade.

Speculative Second, and cutting the other way: the beneficiary is not the builder. Star lifting buys 250 million years of main-sequence life at a cost of 3.4 million years of continuous operation — a project whose payoff arrives after every institution, language and species involved in starting it has ceased to exist in any recognisable form. The field has no account of who would want that, and Matloff's conclusion is effectively that nobody would.

Speculative Third, an epistemic point this page is an instance of. A subject with a dozen papers and enormous press coverage generates a public understanding shaped almost entirely by secondary accounts, several of which fail arithmetic checks against the primaries they describe. Correcting a widely repeated displacement figure by a factor of eighteen is more useful public service than adding another design, and it is the main thing this brief does.

Speculative The civilizational content of stellar engineering is what it says about limits, and the arithmetic says the limits are the star's rather than ours. 0.03 light-years per million years for a Class A mirror. Ten parsecs per million years for the most aggressive design published. 3.4 million years to buy 250 million. None of these numbers improves with technology by more than a small factor, because each is a ratio of stellar quantities — L over c, GM over R. That is a genuinely unusual finding: most speculative technologies are slow because we are early, and this one is slow because of the object.

Speculative The return is nevertheless spectacular in the one case where it can be computed. Seventy to one in time, for star lifting: a machine running for a few million years buys a few hundred million. As an investment ratio that beats almost anything a civilisation could do. The problem is not the return, it is the term of the loan, and no institution in the record has ever maintained a purpose across even a thousandth of it.

Established And the framing gets a direct observational test, which very few speculative subjects do. If stars are an engineering target, somebody should have done it: a galaxy of 10¹¹ stars over 10¹⁰ years is a large sample. Lingam and Loeb searched Gaia DR2 for stars above 0.01c and found nothing, giving fewer than one in a million. Combine that with Dyson Swarms' limits — 2 × 10⁻⁵ within 100 parsecs, and no galaxy in a hundred thousand above 85% reprocessing — and the megastructure hypothesis space has now been probed at three independent scales and returned upper limits at every one.

Speculative The most durable civilizational reading is the one the field's own advocate reached. Matloff set out to present star lifting as an application for alien megastructures and concluded that an advanced civilisation would probably migrate instead. When the argument for a technology, made by someone making it, ends by recommending the alternative, that is the strongest available evidence about the technology's standing — and it points the whole subject toward Interstellar Civilization Models rather than toward engineering.

These horizons track publications and observational limits, because there is no engineering programme in existence to track:

• 10 yr: Frontier The one confident forecast is that the Lingam and Loeb constraint tightens toward 10⁻⁸ as Gaia's final data releases land, which is a real scientific outcome requiring nothing to be built. Speculative Expect a handful more papers of the McInnes kind — specialists from adjacent fields finding a tractable corner — and no funded programme. Speculative A systematic transit-archive search for Class A signatures is the cheapest publishable work available and nobody has announced one.
• 25 yr: Speculative A statite demonstration — a solar sail holding station against solar gravity rather than travelling — is the only physical hardware at this horizon that would bear on anything in this brief, and it belongs to Solar Sail Systems rather than here. Speculative Nothing else in the subject has a plausible milestone at this range, and any date offered for one is an assertion.
• 50 yr: Handwave No basis for a forecast. The nearest thing to a defensible statement is that the unstudied stellar-feedback question could be answered with existing stellar-structure codes at any point, and answering it would change the error bars on every figure on this page. Speculative Whether anyone does is a question about academic attention rather than about capability.
• 100 / 250+ yr: Handwave Beyond forecasting, and the timescales in the subject make that unusually literal: the shortest continuous-operation requirement in the retrieved literature is 3.4 million years. Speculative The defensible structural statement is that stellar engineering is bounded by ratios of stellar quantities that no technology alters, so its horizons are set by physics rather than by progress — and that its own most recent advocate concluded a civilisation would leave instead.

• Depends on One edge, and the line between the two briefs is the star itself. Dyson Swarms owns everything in orbit around a star — collector architecture, mass budgets, orbital mechanics, the shell-buckling result and the whole waste-heat search programme with its resolved candidate list. This brief owns everything that acts on the star: moving it, lifting mass from it, forcing it to mix, igniting a planet into one. The dependency is physical rather than administrative: every Class A engine is a statite and every star-lifting design begins with a partial collector shell, so a stellar engine is a Dyson-swarm component pointed at a different purpose, sharing the same areal-density regime and the same stability mathematics. The two briefs share exactly one paper by design — McInnes (2026), which treats Dyson bubbles and Shkadov mirrors in a single framework — and each cites it for its own half. This brief also receives the handed-down open problem: what an enclosing structure does to the star inside it.
• Requires (not on this map) Two constraints that are not briefs on this map, and the second is the one the field never discusses. The first is a calculation nobody has run: stars have negative heat capacity, so an enclosing structure returning energy inward should make its star expand, cool and dim, and Wright's 2020 review states plainly that the effect has never been studied — which means every performance figure in this brief carries an unquantified correction of unknown sign. The second is institutional rather than technical. The shortest continuous-operation requirement in the retrieved literature is 3.4 million years of star lifting to buy 250 million years of stellar lifetime; Criswell's proposal runs 300 million years from a start date between AD 2170 and 5650. No human institution has maintained a purpose across even a thousandth of that, and no source in the field treats the gap as a problem to solve.
• Enables No enabling edge is claimed and none would be honest. A working stellar engine would move a planetary system, which touches Interstellar Civilization Models and Interstellar Probes — but the field's own most recent quantitative advocate concluded that migration is the likelier choice, the flagship modern paper has seven citations, and nothing here has been demonstrated at any scale. An edge from a capability with no hardware, no programme and a literature of eight items to the briefs that would consume it would record a wish.
• Adjacent Solar Sail Systems is the technically closest live field and the only source of flight-proven hardware in the gram-per-square-metre regime a Shkadov mirror requires — McInnes is himself a solar-sail dynamicist, and a statite would be the first physical test of anything in this brief. Interstellar Archaeology owns technosignature searching as a programme; this brief carries Lingam and Loeb and Forgan only because they are measurements about stellar engines specifically. Planetary Scale Energy Systems is two removes away and contributes only the unstudied-feedback question, which arrives via Dyson Swarms.

Frontier “A Shkadov thruster produces 1.28 × 10¹⁸ newtons.” That figure is the Sun's entire radiative momentum flux — luminosity divided by the speed of light gives 1.277 × 10¹⁸ N, which matches to three figures. Speculative A one-sided mirror intercepts and reflects a fraction of that, so the number in circulation is a hard upper bound presented as a design value, and every account this brief retrieved repeats it without saying so. The derived quantities — 20 m/s and 0.034 light-years per million years — are internally consistent with the bound and therefore also optimistic.

Handwave “A stellar engine could move us a hundred light years in 230 million years.” Under constant acceleration distance goes as time squared, so the tabulated 0.034 light-years at one million years implies about 1,800 light-years at 230 million — eighteen times the widely reported figure. Speculative The published tabulation and the press account are not reconcilable, and the right response is to use the tabulation, note the discrepancy and refuse to average them.

Speculative “Star lifting is a known engineering process with a known rate.” The quoted rate — 5.9 × 10²¹ kg a year at 10% of solar luminosity — works out to 2.05 × 10¹¹ J per kilogram lifted, against a gravitational binding energy at the solar surface of 1.91 × 10¹¹ J/kg. Frontier That is about 93% of the thermodynamic floor, which means the figure in circulation describes a lossless machine. It is a physical limit, not a performance estimate, and quoting it as an engineering rate assumes away every loss in the system.

Speculative “Star lifting would meaningfully extend the Sun's life.” It would, and the arithmetic is worth doing rather than gesturing at. Removing 1% of the Sun at the quoted rate takes 3.4 million years of continuous operation consuming a tenth of the Sun's entire output, and buys about 250 million years of extra main-sequence life on the standard mass–lifetime scaling. Frontier The return is real and about seventy to one. The term is three and a half million years.

Frontier “This is an active research field.” Its flagship modern paper has seven citations seven years after publication. Its founding proposals are a 1948 Halley Lecture, a 1983 conference talk and a 1997 note, none retrievable in original form. Its canonical design paper — Shkadov 1987 — has no accessible copy anywhere. Frontier Eight substantive contributions in seventy-eight years from perhaps eight people is not a field, it is a correspondence.

Handwave “We could ignite Jupiter to warm its moons.” The proposal is real — Fogg (1989), using a small black hole in a controlled orbit, claiming 100 million years of habitability for the Jovian system — and it reaches this brief through a single 2016 secondary article that notes the original's publication details are incomplete. Speculative Nothing about obtaining, moving or controlling a small black hole appears in any retrieved source. The same secondary route carries Beech's tenfold lifetime extension by staged hydrogen bombs and Cirkovic's detection signature, and all three are flagged at the weakest level for that reason.

Established “Nobody has looked for stellar engines.” Somebody has, on real data, with a stated threshold. Lingam and Loeb took Gaia DR2 — about a billion stars with proper motions, ten million with radial velocities — asked how many move above 0.01c, and found that the fastest confirmed hypervelocity star reaches 6 × 10⁻³ c. Frontier The published constraint is fewer than one in 10⁶, tightening to 10⁻⁸ with complete Gaia data. That is a measurement about megastructures and it is the strongest thing on this page.

Speculative “The engineering is the hard part; the physics is fine.” The physics is fine and it is also the constraint. A Shkadov thruster's acceleration cannot exceed L over cM, and a star-lifting machine's energy cost cannot fall below GM over R. Frontier Both bounds are ratios of stellar quantities, so no technology improves them by more than a small factor. This is the rare speculative subject where better engineering genuinely does not help.

Handwave “A Dyson swarm and a stellar engine would leave the star alone.” Nobody knows. Stars have negative heat capacity by the virial theorem, so returning energy to one should make it expand, cool and dim, and Wright's review states that the overall effect of an enclosing structure on stellar luminosity “has not previously been studied.” Speculative Every design in this brief treats the star as a fixed boundary condition, and there is no literature justifying that. It is the field's most interesting open question and it is not a caveat — it is a missing calculation that existing stellar-structure codes could perform.

Frontier And the framing itself. “Stars themselves are an engineering target” fails in three compounding ways. Speculative The field's own most recent quantitative advocate, writing to make the case, concluded that an advanced civilisation would probably migrate instead. Frontier The arithmetic shows the timescales are properties of stars rather than of technology, so they are not schedules that progress shortens. Established And the framing predicts that somebody, somewhere in a galaxy of 10¹¹ stars, would have done it — and the sky says fewer than one in a million have. Three independent tests, all negative, on a subject with a dozen papers.