Advanced Nuclear Propulsion

Advanced nuclear propulsion means using fission to move a spacecraft, and it is two systems that share a fuel, a regulator and a political constituency and almost nothing else. Nuclear thermal propulsion (NTP) runs a reactor as a heat exchanger: hydrogen is pumped through the core, heated to a few thousand kelvin, and expanded through a nozzle. Thrust is high, specific impulse is roughly double a hydrogen–oxygen stage, and the reactor is on for minutes at a time. Nuclear electric propulsion (NEP) runs a reactor as a power plant, converts heat to electricity and feeds ion or Hall thrusters. Specific impulse is several times higher again, thrust is three to four orders of magnitude lower, and the reactor runs for years.

This brief sits in a part of Category I that works differently from the rest of it, and the difference is the organising claim of the page. Warp Drives waits on negative energy density that general relativity does not obviously permit. Reactionless Propulsion waits on momentum conservation being wrong. Gravity Modification waits on a coupling nobody has measured. Those briefs are gated by nature. This one is not. The physics here is closed and the record is experimental: reactors were built, fired, instrumented and photographed, and the engineering questions that remain are questions about materials, facilities, supply chains and appropriations.

That contrast is true and it is also the framing this page is built to test, because the unqualified version of it is wrong. The framing under test is nuclear thermal and nuclear electric propulsion are ready and waiting on funding. The single most important correction on the page is that the physics is settled and the system is ready are different sentences, and the second is false. Two of the four things this brief records as prerequisites are capital facilities and supply chains that do not exist at the required scale, and two are institutional conditions that sixty years of record suggest are the hardest of the four to obtain. Money is necessary here in a way it is not in the exotic-physics briefs. It is nowhere near sufficient.

The page is therefore organised around four named requirements rather than around a technology roadmap: a HALEU fuel supply qualified for flight; a ground test stand that contains and scrubs reactor exhaust; a launch-safety regime that approves flying a reactor; and appropriations that outlast a single programme cancellation. Section 13 records those as typed constraints. Sections 2 to 5 supply the evidence for them one at a time, in that order.

Established The two systems have different hard problems and the brief treats them separately, because conflating them is how a reader ends up believing that a solved problem in one is progress in the other. NTP's difficulties are materials and hydrogen — surviving roughly 2,700 K in flowing hydrogen, and storing liquid hydrogen at 20 K across a multi-year mission without boil-off. NEP's difficulties are mass and heat rejection — the whole system has to come in near 20 kg/kWe, and the radiators alone can dominate the mass budget. Established They share fuel, regulatory approval and a constituency, which is exactly why they were terminated together in one budget line in 2025. That is an institutional fact rather than a technical one, and it is the first evidence on the page that the binding constraint is not physics.

Frontier The modern design trade space is published and it is modest. NASA Marshall's Generation-1 NTP concepts (Duchek, Nikitaeva, Harnack, Burns and Polzin) span 10–25 klbf thrust; specific impulse of 750–900 s on hydrogen, 350–494 s on methane and 375–450 s on ammonia; propellant loads from 1.25 to 90 t; a ΔV range of 1 to 11.8 km/s; and only 3 to 120 minutes of full-power operation. Established The authors state plainly that “none of the Gen 1 concepts in the trade space exceed the aggregate operating time required for Generation 2 mission concepts or for a crewed Mars mission.” A NASA team writing that its own concept family does not reach the mission it is famous for is the kind of statement that should be read carefully rather than skipped.

Frontier The NEP side has a point design, and the point design is mostly radiator. NASA Glenn's Mars-opposition piloted concept (Oleson and colleagues) uses a 1.6 MWe class system at 68.5 We/kg — about 14.6 kg/kWe — driving sixteen 100 kWe Hall thrusters on xenon at 2,600 s specific impulse, with roughly 2,500 square metres of radiator, about 300 t assembled in a near-rectilinear halo orbit including some 250 t of propellant, delivered 45 t at a time. Speculative None of that has been built. The 14.6 kg/kWe figure is a paper value against a 20 kg/kWe target that no megawatt-class integrated system has ever been assembled to test. Established The image worth carrying away is the 2,500 square metres. A megawatt-class nuclear electric vehicle is, by area and increasingly by mass, a radiator with a reactor attached.

Established The historical test record is real, large, and better documented than most treatments admit — and it is the strongest evidence for the framing under test. Projects Rover and NERVA ran from 1955 to 1973 and spent about $1.4 billion in then-year dollars without ever flying. Robbins and Finger's NASA history counts 20 reactor tests and 2 engine-system tests, with 17 hours of accumulated operating time, six of them above 2000 K. The National Academies count 22 reactors built and tested. Frontier The discrepancy is not an error in either source: it turns on whether the two engine-system tests and the Nuclear Furnace are counted as reactors. Where two good sources disagree this brief prints both numbers and says what the disagreement is about, rather than averaging them.

Established The individual results are the substance, and they are specific. Kiwi A3 ran at 112.5 MW in October 1960, in an early gaseous-hydrogen series in which unclad graphite was badly attacked by the propellant. NRX-A2 ran about six minutes 40 seconds at full power on 24 September 1964, at 811 s ideal specific impulse, and twenty minutes on 15 October; the NRX-A series overall reached 1,100 MW thermal, 55,000 lbf and 810–870 s. Phoebus 1B averaged 1,290 MW in February 1967 and released 240,000 curies — the dirtiest test of the programme. Phoebus 2A reached 4,082 MW peak in June 1968 at 200,000 lbf and 820–850 s, and remains the highest-power nuclear reactor ever operated by anyone. Established Pewee 1 set the record that still matters most: 503 MW in November 1968, fuel temperature 2,750 K, propellant exit 2,550 K, about 875 s specific impulse. XE-Prime, the closest thing to a flight configuration, ran nozzle-down between December 1968 and September 1969 at 1,140 MW, 2,272 K exhaust, 247 kN and 710 s, through 28 starts and shutdowns and 115 minutes of total run time. NRX/EST accumulated nearly two hours of operation with 28 minutes at full power in 1966.

Established The failure modes are on the record too, and a brief that prints only the successes is selling the programme rather than describing it. Transverse fuel-element cracking disrupted coolant flow; peripheral elements suffered mid-band corrosion; clad flaked off under differential thermal expansion. Full niobium-carbide cladding on peripheral elements resolved most of it by the programme's end. Frontier That detail comes from an enthusiast compilation of the primary test reports rather than from a primary itself, and this brief flags it as such and uses it only where it agrees with the National Academies and the NASA history. Established The baseline flight engine was specified and never built: 75,000 lbf, 825 s, reactor power “somewhat in excess of 1000 MW”, about 2,300 K. Established Rover and NERVA did not end on a technical verdict. The Nixon administration terminated the project over Congressional objection, under Vietnam-era cost pressure and a post-Apollo collapse in deep-space appetite.

Established Every one of those reactors ran highly enriched uranium graphite fuel, and that single fact is the first of the brief's four requirements. The National Academies are blunt about the consequence: “The only data available in the United States … are from HEU reactor-engine subsystems in the 1960s and 1970s; there are no experimental data on … HALEU NTP subsystems.” Their formal Recommendation 2 is that NASA and the Department of Energy “conduct a comprehensive assessment of the relative merits and challenges of highly enriched uranium and high-assay, low-enriched uranium fuels”. Established A committee does not recommend that a comparison be performed if the answer is known. Frontier Modern reference designs are uranium-nitride cermet at 40–70% packing density, with three architectures in contention: cercer coated particles, cercer solid solutions, and cermet coated particles. None has been irradiated in an NTP configuration.

Established The material supply behind that fuel is thin, single-sourced, allocated, and already oversubscribed. Centrus Energy's American Centrifuge Operating plant at Piketon, Ohio is described by the Department of Energy as “the only source of HALEU enrichment in the Western world” and the only licensed US HALEU production facility — a cascade of 16 advanced centrifuges licensed to 19.75% U-235. First production was 20 kg, announced 7 November 2023, the first domestic HALEU in roughly seventy years. Cumulative output reached 900 kg by the end of June 2025, with a contract extension for a further 900 kg over the following year. Against that, Congress directed the Department to make 21 metric tons of HALEU available by 30 June 2026. Established And the allocation is explicitly rationed in the Department's own words: “If there is insufficient supply to satisfy all HALEU requests, preference will be given to those activities that are most likely … to best further its missions,” with the Department reserving “the right to rescind allocations”. Priority goes to the X-energy demonstration reactor and TerraPower's Natrium. Frontier Do the arithmetic out loud: production is running at roughly 900 kg a year against a 21-tonne statutory target — about 4% of its own mandate — and the queue ahead of any propulsion programme is two commercial power reactors with named allocations. A nuclear rocket does not merely need HALEU. It needs HALEU in a fuel form nobody has irradiated, from a chain that is short of its own target and already spoken for.

Established The launch-approval regime exists, is documented, and is far more tractable than the popular version of it — with one boundary that matters enormously here. NSPM-20, signed 20 August 2019, replaced the single presidential-approval pathway that had stood since 1977 with three tiers. Tier I covers radioactive material up to 100,000 times the IAEA A2 value with no credible accident above 5 rem, and not a reactor; the sponsoring agency head authorises. Tier II covers material above that threshold, credible exposure between 5 and 25 rem, or a low-enriched-uranium fission system; the agency head authorises but the Interagency Nuclear Safety Review Board must independently review the Safety Analysis Report. Tier III covers credible exposure above 25 rem at a probability of one in a million or greater, or a fission system on non-low-enriched fuel, and requires presidential authorisation. Frontier The tier boundary therefore runs exactly along the HEU/HALEU line. A HALEU nuclear rocket is a Tier II decision. An HEU one — which is what every historical Rover and NERVA reactor was — goes to the President. The fuel-supply requirement and the launch-approval requirement are the same decision seen from two directions, and a programme that solves one by choosing HEU has made the other harder.

Established The base rate cuts for the framing and should be stated as clearly as the constraints. Since 1961 the United States has flown 32 space launches carrying nuclear power sources, with one unplanned reentry involving dispersal, Transit 5-BN-3 in 1964. The Interagency Nuclear Safety Review Board replaced an ad hoc panel that had operated for roughly fifty years, and was to stand up by February 2020 with the Nuclear Regulatory Commission designating its own equal member. Frontier A think-tank reading of the memorandum describes it as easing approval for routine missions by removing presidential bottlenecks at Tiers I and II while potentially complicating higher-risk projects through mandatory board review and stricter documentation. Established But every one of those 32 launches carried a radioisotope source, not an operating reactor. The only US reactor ever operated in space is SNAP-10A, in 1965, and it ran 43 days.

Established Which brings the page to DRACO, the live institutional fact, and to a point about sourcing this brief will not skate over. The Demonstration Rocket for Agile Cislunar Operations began with Phase 1 contracts in April 2021 — General Atomics at $22M for the reactor concept, Blue Origin at $2.5M, Lockheed Martin at $2.9M. A non-reimbursable interagency agreement was signed on 11 January 2023; the NASA–DARPA partnership was announced on 24 January 2023 at a joint value of $499M; Lockheed Martin with BWXT took Phase 2/3 on 26 July 2023; launch was targeted for 2027. In January 2025 the launch went to indefinite hold, with cited “technical and regulatory challenges such as the complex safety and testing requirements for ground-based nuclear reactor validation”. Cancellation was confirmed in the FY2026 President's Budget Request on 30 May 2025, and DARPA's stated rationale in late June was that costs no longer matched the benefits.

Established There is no standalone DARPA cancellation press release, and this brief says so rather than implying a cleaner record exists. What exists is three primary artefacts, and each should be named as what it is. First, DARPA's own programme page, which now reads “This program is now complete” and “This content is available for reference purposes. This page is no longer maintained.” That is the agency's institutional record of termination, on a page that also states the programme's goals — thrust-to-weight roughly 10,000 times that of electric propulsion, specific impulse two to five times in-space chemical — and that DARPA “concentrated on the highest technical risk first”. Second, NASA's FY2026 Budget Technical Supplement, in which the Administrator's message states: “Given that our near-term human exploration and science needs do not require nuclear propulsion, current demonstration projects will end.” The same document shows Space Technology falling from $1,100.0M enacted in FY2024 to a $568.9M request in FY2026 — a 48% cut of $531.1M — and speaks of diverting resources from “underperforming space propulsion projects”. Established That is the strongest single citation on this page: NASA reporting the termination of its own programme, in its own budget request, in the Administrator's own message. Interest runs hard against the finding, which is the strongest form a claim of this kind can take. Third, the interagency agreement itself.

Established The agreement is the underreported structural fact, and it is the evidence for the fourth requirement. It divides the work precisely: NASA led nuclear thermal rocket engine development and fabrication, reactor development and fabrication, non-nuclear engine development and engine-to-vehicle integration; DARPA led vehicle-to-launch-vehicle integration, launch procurement and approvals, scheduling and contracting, the ground segment and operations, and programme security. Milestones ran a system requirements review in FY23, preliminary design review in FY24, nuclear-engine and vehicle critical design reviews in FY25, engine acceptance in FY26, and launch readiness and launch in FY27. And then the clause: “There will be no transfer of funds between the Parties under this Agreement and each Party will fund its own participation.” Frontier DRACO had no shared purse. Either agency could withdraw its own money unilaterally and the programme would collapse without anyone cancelling a joint appropriation. It was built, by construction, as two single points of failure.

Frontier The stated reason for cancellation matters more than the cancellation, and it is the thing that most damages the framing. DARPA's deputy director's rationale was that the original justification — high launch costs making propellant efficiency decisive — had been undermined by falling launch prices and projected heavy-lift capability. Frontier That quotation reaches this brief second-hand: the trade report carrying it could not be retrieved directly, and it is carried at the weaker flag for that reason. Established But the shape of the argument survives the sourcing question. DRACO was not defunded because nobody had the money. It was defunded because a competing technology got cheaper and the cost–benefit case for a nuclear upper stage weakened. That is a different failure mode from underfunding and it does not respond to advocacy: if chemical launch keeps getting cheaper, the case for NTP keeps getting weaker no matter how well the reactor works.

Established The sharpest institutional observation available to this page is a comparison of two documents ten weeks apart. On 4 August 2025 — roughly two months after the propulsion line was zeroed — NASA's acting administrator issued a Directive on Fission Surface Power: a minimum of 100 kWe, “readiness to launch by the first quarter of FY30”, a programme executive designated within 30 days, a request for proposals within 60 days, and authority to award two providers within six months of release. Established The directive contains no mention of nuclear thermal propulsion, nuclear electric propulsion, or nuclear propulsion of any kind. Frontier Trade reporting adds that surface power moved from the Space Technology Mission Directorate to Exploration Systems Development, runs on roughly fifteen full-time engineer equivalents in a deliberately minimum structure, and sits inside a new $350M Mars Technology programme growing to $500M in FY27; a NASA–DOE memorandum of understanding on 14 January 2026 specified 100 kWe, a closed Brayton cycle, a heavy lander carrying up to 15 t and launch readiness in Q1 FY2030. Frontier Within six months the same agency killed space nuclear propulsion and accelerated space nuclear power, using much of the same fissile material, the same launch-approval regime, the same contractor base and much of the same safety analysis. What differed was mission pull. Space nuclear is not underfunded in aggregate. It is redirected on a cycle shorter than the development time of any of its systems.

Established And the DRACO cancellation is not an event; it is an instance. Bhavya Lal, a former NASA acting chief technologist and associate administrator, and Roger Myers of the National Academy of Engineering count the record: SNAP-10A in 1965 remains the first and only US reactor ever to operate in space and ran 43 days; since then there have been “about a dozen start-stop-restart efforts to launch a nuclear fission system in space”; JIMO/Prometheus consumed “over $400 million” before abandonment when estimates “spiraled above $20 billion”; KRUSTY achieved a successful ground experiment and never became an operational system; DRACO never advanced past demonstration. Their total is “tens of billions of dollars (there is no official tally)” with no operational space nuclear system to show for it. Frontier Their four diagnosed failure modes are the best available articulation of this brief's fourth requirement: no anchoring mission, producing open-ended technology development with no deployment pathway; technological overreach, with Prometheus attempting to leap from paper to a flagship Jupiter mission before validating systems; political cycles of one, two and four years against nuclear development horizons; and no central owner, with responsibility fragmented across NASA, DOE, Defense, the FAA and the NRC — contrasted explicitly with the Navy's nuclear programme. Frontier Both authors have an interest in more space-nuclear funding, which cuts for the framing under test. Their specific diagnosis is that money alone has repeatedly failed, which cuts against it. That tension is worth naming rather than resolving.

Established The first open question is not open at all, and the page should say so before opening the rest. That NTP roughly doubles chemical specific impulse at comparable thrust is demonstrated: Pewee at 875 s, the NRX series at 810–870 s, XE-Prime at 710 s in a flight-like configuration. Frontier What is contested is everything downstream of that demonstration, and the contest is mostly about fuel, facilities and money.

Frontier Position one: an aggressive programme could field NTP for a crewed Mars mission around 2039. This is the National Academies' own judgement, resting on the Apollo precedent of eight years from decision to landing, with cargo missions from 2033. Frontier It is a serious position from an independent committee. It is also conditioned in the same report on a three-phase ground campaign requiring new facility construction, which no one has begun.

Frontier Position two: HALEU can substitute for HEU in an NTP core at an acceptable performance cost. This is the design choice DRACO and BWXT made, and it is the assumption on which every modern programme rests. Frontier It is not demonstrated. The National Academies record zero experimental data on HALEU NTP subsystems and formally recommend that the comparison be performed. A lower-enrichment core needs more uranium, a larger core, or a different moderator arrangement, and the penalty is a design output nobody has measured in a hot-fire.

Frontier Position three: NEP at one to two megawatts electric is the better crewed-Mars architecture, not NTP. Held by the National Academies as a baseline and by NASA Glenn's point-design team. Speculative Position four, downstream of it: 20 kg/kWe is achievable. The Glenn design claims 14.6 kg/kWe on paper. No megawatt-class integrated space power system has ever been built or tested, ground electric-propulsion testing is limited to under 50 kWe, and the largest flight electric thruster is 6.9 kWe. The paper number is three orders of magnitude beyond anything demonstrated in an integrated test.

Speculative Position five: bimodal NTP/NEP — one reactor doing both jobs — is the right architecture. Various NASA and contractor studies advocate it. It is attractive on paper because it amortises the hardest single component. It also inherits both sets of hard problems, and no bimodal system has been tested at any scale.

Frontier Position six is a forecast about launch prices rather than about reactors, and it is currently the winning position inside the US government. If falling launch cost permanently undercuts the value of propellant efficiency, then NTP's core economic argument does not recover. Frontier The counter is that this is a claim about a market, made at a moment, by an agency that had already decided to stop. Launch prices could stop falling; a mission requiring high ΔV with a mass constraint that cheap launch does not relieve could appear. Neither has happened yet, and the brief should not pretend the argument has been refuted merely because it is convenient.

Frontier Position seven: the binding constraint is institutional continuity rather than technology. Lal and Myers's diagnosis, with a strong evidentiary base and a contested reading. Speculative Position eight, which the National Academies explicitly reject: ground testing can be replaced by in-space demonstration. The committee's words are that subscale in-space flight testing “cannot address many of the risks and potential failure modes … and therefore cannot replace full-scale ground testing”. This position has advocates in the flight-first community and no committee support.

Speculative Position nine: commercial actors can carry space nuclear through political discontinuity where agencies cannot. The two-provider structure of the surface-power award is the closest thing to an institutional bet on this. Speculative No commercial space reactor has flown, no commercial entity has ever held a Tier III authorisation, and the fissile supply is government-allocated, so the mechanism by which commercial actors would achieve independence from the political cycle is unclear. Handwave The strongest version of the position — that a private company could simply buy its way past the four requirements on this page — runs into the fact that two of them are federal facilities and one is a presidential signature.

Established Bottleneck one is a fuel form, not a fuel. HALEU exists; HALEU in an NTP-qualified fuel form does not. The gap is not enrichment but irradiation data: how a uranium-nitride cermet at 40–70% packing density behaves at 2,700 K in flowing hydrogen across dozens of thermal cycles, and nobody has run that test on a low-enriched core. Frontier Every performance projection for a modern engine crosses this gap by assumption.

Established Bottleneck two is a building that has never existed, and it may be the hardest of the four. A nuclear thermal rocket on a test stand exhausts fission products into the open air unless something catches them. In the entire Rover and NERVA programme exactly one test did that: the Nuclear Furnace NF-1 in June and July 1972, the first to incorporate an effluent scrubber, which ran four times at full power — 44 MW — for 108.8 minutes total. The scrubber worked, although some krypton-85 leaked. Established 44 MW is about 4% of XE-Prime's 1,140 MW and roughly 1% of Phoebus 2A's 4,082 MW. No one has ever scrubbed the exhaust of a full-scale nuclear rocket engine. The historic infrastructure at Jackass Flats — Test Cells A and C, the engine and reactor maintenance buildings, engine test stand ETS-1, the radioactive-material storage facility, all in Area 25 of the Nevada Test Site from 1962 — was built for open-air exhaust and no longer exists in usable form.

Established The National Academies specify what would be required and close the obvious escape hatch. Three phases: zero-power critical testing for neutronic verification; operational tests with full fission heating and liquid-hydrogen flow through startup, operation, shutdown and restart; and integrated reactor-engine-propellant testing at full scale, which requires new facility construction. And then: subscale in-space flight testing “cannot replace full-scale ground testing”. They add that US NTP testing “took place nearly 50 years ago, and did not fully address flight system requirements”, and that recapturing the capability will be “costly and time-consuming”. Frontier This is where “ready and waiting on funding” first breaks. The funding in question is not a line item for engineering hours. It is capital construction of a nuclear test facility that has never been built at the required scale, in a regulatory environment that did not exist in 1972.

Established Bottleneck three is hydrogen, and it is peculiar to NTP. A reactor that doubles specific impulse only does so on hydrogen, and hydrogen at 20 K boils off. A multi-year mission needs zero-boil-off or near-zero-boil-off cryogenic storage at scale for a propellant with the worst density-to-performance trade in the business. Frontier Methane and ammonia are storable alternatives and cost roughly half the specific impulse — 350–494 s and 375–450 s against 750–900 s — which is most of the reason to have built the reactor.

Established Bottleneck four is peculiar to NEP and is thermal. Radiators dissipating at around 500 K need roughly 1,500–3,000 square metres single-sided for one to two megawatts electric; for Prometheus at 200 kWe the heat-rejection subsystem alone was projected at 10.1 kg/kWe, about half the entire mass budget. Established Alongside it sits an electronics problem stated flatly by the National Academies: radiation-hardened power management and distribution or processing units at megawatt electric levels “have never been developed”. Frontier Heat rejection is the most under-narrated constraint across this whole cluster, and it recurs in every one of the neighbouring briefs at a larger scale.

Frontier Bottleneck five is not technical and is the one the record says binds hardest. The observed institutional half-life of a US space nuclear programme is shorter than the development time of any system in it. That is not editorialising: it is a counted record of about a dozen start–stop–restart efforts since 1965, one of which spent over $400M and another of which was structured with no shared funding between its two sponsoring agencies.

Established This brief records no dependency on another brief, and that absence is the finding rather than a gap. Nothing on this map has to be discovered before a nuclear thermal engine can be built. There is no physics result outstanding, no material that needs inventing, no theoretical question whose answer would change the design. That is precisely what distinguishes this slot from the rest of Category I, where the honest adjudication routes through Quantum Gravity or waits on a coupling nobody has measured.

Established What it records instead are four constraints that are not briefs, because they are facts about the world rather than results anyone could produce by research. A fuel supply, a test facility, a regulatory decision and a budget line. Each has a named institution behind it, a published number, and a documented shortfall. Frontier A reader who takes only one thing from this page should take that shape: a technology can be complete and unavailable at the same time, and when it is, the dependency graph stops looking like a science tree and starts looking like a procurement schedule.

Frontier Two of the four are physical and two are institutional, and the record says the institutional pair is harder. HALEU production is rising and a test facility is a construction project with a known cost shape; both are the kind of problem money solves. The launch-approval regime already functions — 32 launches, one dispersal, a tiered process with a standing review board. The appropriations condition is the one with a sixty-year record of failure, and it is the only one of the four on which more money has repeatedly been spent to no effect.

Established Downstream, three briefs on this map wait on this one. Fusion Spacecraft inherits every problem here — radiators, power conversion, launch approval, ground-test infrastructure, multi-decade appropriations — and adds a reaction nobody has run at net power in a flyable package; the best-developed fusion propulsion concept in the NASA portfolio works by driving fission in a uranium liner. Antimatter Propulsion sits above that again. Interstellar Probes names this brief as one of three drives it has no independent path around.

Established The required experiments are specified by an independent committee and this brief simply reports them, because a list assembled here would be an opinion and the committee's list is a finding. Phase one: zero-power critical testing to verify the neutronic model of a HALEU core. Phase two: operational testing with full fission heating and liquid-hydrogen flow through the complete cycle — startup, steady operation, shutdown, restart. Phase three: integrated reactor, engine and propellant testing at full scale, which requires a facility that does not exist.

Established The single most consequential missing experiment is the fuel comparison the National Academies formally recommended. A comprehensive assessment of HEU against HALEU for NTP, backed by irradiation data on candidate low-enriched fuel forms. Frontier Until that exists, every performance number quoted for a modern engine is an extrapolation from a different fuel in a different regulatory tier, and the page should be read accordingly.

Established The second is the scrubbed full-scale hot-fire. NF-1 demonstrated effluent scrubbing at 44 MW in 1972 with some krypton-85 leakage. Nothing since has scrubbed a nuclear rocket exhaust at all, and the step from 44 MW to a flight-class engine above 1,000 MW is a factor of twenty-five in fission-product inventory, in a modern permitting environment. Frontier This is an experiment whose principal difficulty is a construction permit and an environmental impact statement rather than a measurement technique.

Established On the electric side, the gap is measurable and specific. Ground testing of electric propulsion is limited today to under 50 kWe; the largest thruster flown is 6.9 kWe; the point designs call for sixteen 100 kWe units. Frontier A megawatt-class electric propulsion ground facility, and a radiation-hardened megawatt power management system to feed it, are both prerequisites to testing an NEP vehicle and neither exists. Speculative A 2,500 square metre deployable radiator has never been flown, deployed, or thermally cycled in vacuum at scale.

Frontier And one experiment the record already ran, which is worth knowing about because it answers a question the public asks. Kiwi-TNT, 12 January 1965: control drums accelerated from 45 degrees per second to 4,000 with shims removed and excess reactivity set at six dollars prompt critical. The result was 3.1 × 10²⁰ fissions in 12.4 milliseconds, core temperatures above 17,500 degrees Celsius, five to fifteen per cent vaporisation, and roughly 100 pounds of TNT equivalent of kinetic energy. Frontier A nuclear rocket core deliberately driven to destruction produced a hundred pounds of TNT, not a nuclear explosion. The figure comes from an enthusiast compilation of primary reports rather than from a primary, and is flagged accordingly.

Established The NTP engineering requirement set is short, specific and mostly about one material problem. A core that survives about 2,700 K in flowing hydrogen — hydrogen being the most chemically aggressive coolant available at that temperature — through repeated thermal cycles, with fuel elements that neither crack transversely nor lose their cladding. Pewee reached 2,750 K fuel temperature in 1968 with coated particles and niobium-carbide cladding, so this is a demonstrated capability on HEU graphite fuel. Frontier It is an undemonstrated capability on any fuel form a modern engine would fly.

Established Around the core sit turbopumps, a nozzle, and a hydrogen tank that has to survive years. The flight engine that was specified in 1972 and never built was 75,000 lbf at 825 s from a reactor somewhat over 1000 MW. Modern Gen-1 concepts are smaller — 10–25 klbf, 3 to 120 minutes of full-power operation. Frontier Nothing about the mechanical engineering is exotic. The exotic part is the qualification programme, and the qualification programme needs a facility.

Established The NEP requirement set is dominated by two numbers. System specific mass near 20 kg/kWe, and radiator area of roughly 1,500–3,000 square metres for one to two megawatts electric at around 500 K rejection temperature. Speculative The 68.5 We/kg — 14.6 kg/kWe — in the Glenn point design is what a well-optimised paper vehicle achieves; the honest reading is that it is a target with a plausible derivation, not a measured value.

Established The power-conversion and distribution chain is the requirement most likely to be underestimated, because it sounds like a solved industrial problem and is not. Megawatt-class radiation-hardened power management and distribution has never been developed. Frontier Terrestrial megawatt power electronics exist in quantity and none of it is qualified for a decade in a reactor's neutron and gamma environment at spacecraft mass fractions.

Established Assembly is a requirement in its own right and it is usually omitted. The Glenn concept masses about 300 t assembled in a near-rectilinear halo orbit, delivered 45 t at a time. That is a multi-launch on-orbit assembly campaign for a nuclear vehicle, with the reactor started only after assembly. Frontier Deep Space Infrastructure owns the assembly capability; this brief owns the fact that the architecture assumes it.

Established The nearest neighbour on this map is not a propulsion brief at all. Advanced Fission owns reactor physics, coolants, fuel cycles and the qualification of nuclear materials, and the HALEU supply chain described here is the same chain its commercial demonstration reactors are drawing on — which is exactly why a propulsion programme finds itself third in a queue. Lunar Energy Infrastructure owns surface power, and the 100 kWe fission surface power directive belongs there; this brief's claim on it is only the contrast, that propulsion was cancelled and power accelerated within ten weeks using much of the same technology base.

Established Within this cluster the split is by reaction and by mission. This brief owns fission: reactors that have been built, fuel that is enriched and fabricated today, and a regulatory tier structure written for fission. Fusion Spacecraft owns fusion and depends on this brief. Antimatter Propulsion sits above that. Interstellar Probes owns the mission rather than the drive — lifetime, power, communications, arrival — and names this brief as one of its three drives. Solar Sail Systems and Beam-Powered Propulsion are the propellantless alternatives and share none of this brief's constraints except heat rejection.

Frontier Two adjacencies are boundaries this brief deliberately does not cross. The first is launch vehicles: DRACO's cancellation rationale is a claim about launch cost, not about nuclear thermal propulsion, and letting this page become an argument about heavy-lift economics would misplace the evidence. The second is Electrodynamic Propulsion Concepts and the electric-thruster literature generally: the Hall and ion thrusters an NEP vehicle would fly are mature technology with flight heritage, and the unsolved part of NEP is the power plant, not the thruster.

Established Off-map, the relevant expertise sits in four communities that mostly do not talk to each other: naval reactor engineering, which is the one US institution with an unbroken multi-decade nuclear development record and is the explicit comparison Lal and Myers draw; uranium enrichment and fuel fabrication; cryogenic hydrogen storage, shared with launch vehicles and with hydrogen energy systems; and space thermal management, which is where the radiator problem actually lives.

Established Two of this brief's four requirements are institutional, and they are the ones with the worst record. The launch-safety regime is the tractable one: NSPM-20 gives a three-tier structure with agency-head authorisation for low-enriched fission systems, an independent review board, and a documented base rate of 32 launches with one dispersal since 1961. That is a functioning process, not a barrier, and the popular claim that a reactor cannot be launched is wrong.

Frontier The appropriations requirement is the one that has never been met. The counted record is a dozen start–stop–restart efforts since SNAP-10A, a $400M+ write-off at Prometheus, a successful ground experiment at KRUSTY that never became a system, and DRACO terminated four years in. Established And DRACO's interagency agreement makes the failure mode structural rather than accidental: no transfer of funds between the parties, each funding its own participation. A programme built that way has two independent cancellation points and no shared sunk cost to defend.

Frontier The diagnosis with the most explanatory power is the absence of a central owner. Responsibility for space nuclear is distributed across NASA, the Department of Energy, the Department of Defense, the Federal Aviation Administration and the Nuclear Regulatory Commission, none of which owns the outcome. The comparison Lal and Myers draw is with the Navy's nuclear programme, which has one owner, one authority and a continuous sixty-year development line. Frontier That comparison is the most useful institutional exhibit on the page, and it is also the one that most resists implementation, because creating a single owner means a statutory reorganisation rather than an appropriation.

Established The supply chain is an institutional fact as much as an industrial one. HALEU allocation is a Department of Energy discretionary process with an explicit preference rule and an explicit power to rescind, running at roughly 900 kg a year against a 21-tonne statutory target. A propulsion programme does not buy HALEU on a market; it applies for an allocation and is ranked against commercial demonstration reactors with named priority. Frontier That is a governance decision with a technical consequence, and it is invisible in every popular treatment of nuclear rockets.

Frontier The most instructive institutional evidence is the pair of documents ten weeks apart. A budget request stating that near-term needs do not require nuclear propulsion, and a directive requiring a 100 kWe surface reactor ready to launch in Q1 FY30 with two providers under contract within six months. Established The money exists. The regulatory pathway exists. The contractor base exists. What moves is the mission, and it moves faster than any of these systems can be developed.

Established The launch-accident question has a measured base rate and it should be stated before anything else, because the discourse around it is largely uncalibrated. Thirty-two US launches carrying nuclear power sources since 1961; one unplanned reentry with dispersal, Transit 5-BN-3 in 1964. Established The graded response to that record is NSPM-20's tier structure, in which the depth of independent review scales with credible dose and with the nature of the source, and a reactor is categorically separated from a radioisotope generator.

Established The ground-test question is different and is the one with a genuine unresolved externality. Every full-scale nuclear rocket test in history except one vented fission products to open air at the Nevada Test Site; Phoebus 1B alone released 240,000 curies. Frontier A modern programme cannot do that, which is why the scrubbed test stand is a requirement rather than a preference — and why the environmental permitting for such a facility is plausibly a longer pole than its construction.

Frontier The fuel choice carries a nonproliferation dimension that maps exactly onto the regulatory tiers. Highly enriched uranium gives the best NTP performance and the historical data, and it is weapons-usable material requiring presidential authorisation to launch. Low-enriched fuel is the nonproliferation-compatible choice, sits in Tier II, and has no experimental basis. Frontier The performance argument for HEU and the nonproliferation argument against it are both sound, and the honest position is that this is a real trade rather than a solved question.

Established One public fear can be answered with data. A nuclear rocket core cannot detonate as a nuclear weapon, and this was tested deliberately: Kiwi-TNT in 1965 drove a core to prompt criticality on purpose and produced roughly a hundred pounds of TNT equivalent along with a substantial radiological release. Frontier The radiological hazard is real and the explosive hazard is not, and conflating them makes the actual risk harder to discuss.

Frontier And there is an equity question that the aggregate spending figure raises. Tens of billions of dollars, on Lal and Myers's reckoning with no official tally, spent across sixty years on a capability that has never operated. Speculative Whether that is a scandal or the normal cost of a hard capability depends entirely on whether the next attempt is structured differently from the last dozen, and nothing in the current record suggests it is.

Frontier The payoff, stated at the size the evidence supports, is a solar system that is a few times more accessible rather than one that is transformed. NTP roughly doubles specific impulse at high thrust; NEP multiplies it several times more at very low thrust. That shortens crewed Mars transits, widens launch windows, and makes outer-system orbiters rather than flybys affordable. It does not change the character of interplanetary travel and it does not touch interstellar distances at all.

Established The most civilizationally interesting thing about this brief is not the technology but the demonstration it provides. Here is a capability whose physics was closed sixty years ago, whose hardware was built and fired twenty-two times, whose regulatory pathway exists and functions, and which is nonetheless unavailable. Frontier That is a general lesson about how technological capability actually decays: not through forgotten physics but through demolished facilities, lapsed qualification data, retired fuel forms and discontinued appropriations. The Rover programme's test stands are gone; its fuel is the wrong enrichment; its personnel are dead; and the $1.4 billion in then-year dollars bought knowledge that has to be substantially re-earned.

Frontier The second lesson is about mission pull, and it generalises past space. The same agency that found no need for nuclear propulsion issued an accelerated directive for nuclear surface power ten weeks later. What moved was not capability, cost or safety but a decision about where people are going. Frontier Technologies in this category are not funded because they are ready; they are funded because something else has been decided.

Speculative If the four requirements were met, the realistic long-run picture is a small fleet of reusable in-space nuclear stages, refuelled and maintained in cislunar space, doing the jobs chemical propulsion does badly. That is a substantial industrial capability and it is not a revolution. Handwave Treatments that present nuclear propulsion as the thing that opens the solar system are running the argument at the wrong magnitude: a factor of two to five in specific impulse is a valuable engineering improvement, not a change of regime.

These horizons track facilities, fuel qualification and appropriations rather than discoveries — which is the whole distinction this brief draws against the rest of Category I. Each is conditioned on the four requirements in section 13:

• 10 yr: Established No US nuclear propulsion flight demonstration is currently funded; the demonstration line was zeroed in the FY2026 request and the programme page reads “This program is now complete”. Frontier Expect the space nuclear centre of gravity to be surface power rather than propulsion: a 100 kWe lunar reactor with a stated Q1 FY30 launch-readiness target, two providers, and a NASA–DOE memorandum already signed. Frontier Expect HALEU production to grow but to remain allocated ahead of propulsion, and expect the HEU-versus-HALEU assessment the National Academies recommended to be the pivotal document if a programme restarts. Speculative A restarted flight demonstration within a decade is possible and would require a mission that needs it, which currently does not exist.
• 25 yr: Frontier If a crewed Mars programme is committed to, this is the horizon on which either NTP or megawatt-class NEP could plausibly fly, and the National Academies' 2039 judgement sits inside it. Speculative The pacing item is the ground-test facility, not the engine: a scrubbed full-scale stand is a multi-year capital project with an environmental permitting process that has no precedent at this scale. Speculative A plausible alternative trajectory, and on the base rate the more likely one, is another start–stop cycle: a programme announced, a preliminary design review passed, and a cancellation on a cost–benefit reassessment before hardware. Frontier Surface fission power operating on the Moon is a genuinely likely outcome at this horizon and would supply much of the fuel, safety and contractor base a propulsion programme would need.
• 50 yr: Speculative A mature in-space nuclear stage, reusable and refuelled in cislunar space, is a coherent picture at this horizon and depends on an industrial demand that does not yet exist. Speculative Megawatt-class NEP is the harder and more transformative of the two and requires the radiator, power-conversion and assembly capabilities to arrive together. Handwave Any specific date here is a convention. The honest statement is that the technical path is short and the institutional path has never been walked.
• 100 / 250+ yr: Speculative At this horizon fission propulsion is either routine infrastructure or a historical curiosity displaced by something better, and the evidence does not distinguish between those. Handwave What can be said is that no physics discovered between now and then is required for the first outcome, which is not something the other briefs in this category can say.

• Depends on Nothing on this map. This is the unusual case in Category I of a brief that waits on no scientific result whatsoever: fission propulsion needs no discovery, no new material class and no theoretical advance. The reactors were built and fired, the specific impulse was measured, and the flight engine was specified. What is missing is not knowledge. The four rows below record what is missing instead, and the reason this brief carries no depends_on edge is a finding about the subject rather than an omission in the adjudication.
• Requires (not on this map) Four constraints, none of which is a research result. Fuel: every one of the 22 reactors ever tested ran highly enriched uranium, the National Academies record no experimental data on high-assay low-enriched NTP subsystems, and the only Western HALEU plant produced 900 kg by mid-2025 against a 21-tonne statutory target, on an allocation the Department of Energy reserves the right to rescind. Test stand: exactly one nuclear rocket test in history scrubbed its exhaust — the Nuclear Furnace at 44 MW in 1972, about 1% of Phoebus 2A — and full-scale integrated testing requires new facility construction that subscale flight testing explicitly cannot replace. Launch approval: the regime exists and works, with 32 nuclear-source launches since 1961 and one dispersal, but its tier boundary runs along the HEU/HALEU line, so a high-enrichment engine needs the President's signature and a low-enrichment one needs fuel data nobody has. Appropriations: SNAP-10A remains the only US reactor ever operated in space, about a dozen start–stop–restart efforts have followed, Prometheus consumed over $400M, and DRACO was signed under an agreement stating that “there will be no transfer of funds between the Parties … each Party will fund its own participation” — two single points of failure, by construction.
• Enables Fusion Spacecraft depends on this brief, and the dependency is substantive rather than thematic: a fusion vehicle inherits the radiators, the power conversion, the launch-approval regime, the ground-test problem and the appropriations problem recorded here, and the best-developed fusion propulsion concept in the NASA portfolio drives fission in a uranium liner. Interstellar Probes names this brief as one of three drives it has no independent path around. Through those two, Antimatter Propulsion sits downstream as well.
• Adjacent Advanced Fission supplies the reactor physics, the fuel qualification methods and the HALEU chain this brief competes for; Lunar Energy Infrastructure owns the surface-power programme that was accelerated as this one was cancelled; Electrodynamic Propulsion Concepts owns the mature thrusters an electric vehicle would fly; Deep Space Infrastructure owns the on-orbit assembly a 300-tonne vehicle assumes; and Solar Sail Systems and Beam-Powered Propulsion are the propellantless alternatives that share only the heat-rejection problem. Off-map: naval reactor engineering, uranium enrichment and fuel fabrication, cryogenic hydrogen storage, and space thermal management.

Handwave “The physics is settled, so the system is ready.” This is the framing under test and it fails on a distinction the page exists to draw. The physics is settled; the record is a genuine experimental record; and none of that makes a system available. Established Twenty-two reactors were tested on a fuel a modern engine will not fly, in facilities that no longer exist, under a regulatory regime that has since been rewritten, funded by a programme that was cancelled over Congressional objection in 1973. Every element of the demonstration has been retired except the knowledge that it worked.

Handwave “Nuclear propulsion is just waiting on funding.” The most recent test of this claim is DRACO, and it failed in an informative direction. DRACO was not defunded because nobody had the money — it was a $499M joint programme past preliminary design. Established It was ended because the case weakened: NASA's own budget message says near-term needs “do not require nuclear propulsion”, and the stated rationale was that falling launch costs had undercut the value of propellant efficiency. Frontier That is a mission-pull failure, and it is immune to the argument “just fund it”. The same agency issued an accelerated 100 kWe surface-power directive ten weeks later, which demonstrates that the money exists and the mission is what moves.

Established “DRACO was cancelled for safety reasons, or under regulatory pressure, or because of protest.” The stated reasons were cost–benefit and mission need. Frontier The January 2025 hold did cite “complex safety and testing requirements for ground-based nuclear reactor validation”, which is a real constraint and is the second of this brief's four requirements — but the termination five months later was recorded in a budget document as a decision about need, not about risk. Attributing it to obstruction gets the lesson backwards and makes the actual failure mode invisible.

Established “There is a DARPA press release announcing the cancellation.” There is not, and this brief says so rather than citing a news summary as though it were the primary. Established The primary record is three artefacts: a programme page that now reads “This program is now complete”, a NASA budget supplement in which the Administrator states the demonstration projects will end, and the signed interagency agreement showing what was cancelled and how the work was split. Naming what the record actually is, is more useful than implying a tidier one.

Handwave “HALEU is a drop-in substitute for the fuel the old engines used.” The National Academies' words are that there are “no experimental data on … HALEU NTP subsystems”, and their formal recommendation is that the HEU-versus-HALEU comparison be carried out. Established The impressive test record is a record of a different fuel, in a different regulatory tier, in a core geometry that lower enrichment changes. Every modern performance projection crosses that gap by assumption.

Established “You cannot get permission to launch a reactor.” Wrong, and this is the misconception the page is happiest to correct. The United States has flown 32 launches with nuclear power sources since 1961 with a single dispersal event; NSPM-20 gives a graded three-tier process; a low-enriched fission system is a Tier II decision authorised by an agency head with independent board review. Frontier The regulatory half of the story is the most tractable of this brief's four requirements. What is true is narrower: no operating reactor has flown since SNAP-10A in 1965, and a high-enriched core would need presidential authorisation.

Handwave “A nuclear rocket could blow up like a bomb.” Tested, on purpose, in January 1965. Kiwi-TNT drove a core to six dollars prompt critical with the control drums accelerated nearly a hundredfold; the yield was roughly a hundred pounds of TNT equivalent from 3.1 × 10²⁰ fissions in 12.4 milliseconds. Frontier The hazard from a nuclear rocket accident is radiological dispersal, not nuclear yield — and the radiological hazard is real, which is why the tier structure exists and why Phoebus 1B's 240,000-curie release is in this brief.

Established “NTP and NEP are two flavours of the same thing.” They share fuel, a regulator and a political constituency, and almost nothing else. One is a heat exchanger that runs for minutes at high thrust and lives or dies on materials in hot hydrogen; the other is a power plant that runs for years at very low thrust and lives or dies on kilograms per kilowatt and square metres of radiator. Frontier The one thing they genuinely share is a budget line, which is why they were cancelled together — and treating that shared fate as evidence of shared technology is exactly backwards.

Speculative “We can skip the ground campaign and demonstrate in space.” The National Academies address this directly and reject it: subscale in-space flight testing “cannot address many of the risks and potential failure modes … and therefore cannot replace full-scale ground testing”. Frontier The position has advocates and no committee support, and the practical objection is that the failure modes that ended fuel elements in the 1960s — transverse cracking, mid-band corrosion, clad flaking — are found by instrumented destructive testing, which is what a ground stand is for.

Established “The 1960s engines are on a shelf somewhere.” XE-Prime was the closest article to flight configuration and was still a ground test article; the 75,000 lbf flight engine was specified and never built. Frontier The test stands at Jackass Flats were built for open-air exhaust, and the one scrubbed test in the whole programme ran at 44 MW. What survives is documentation, and documentation of a fuel form nobody intends to fly.

Frontier And the correction that matters most for reading the rest of this category. It is genuinely true that this cluster is gated by money and institutions rather than by nature, and that is a real and important difference from Warp Drives or Gravity Modification. Established But “gated by money” is not the same as “a cheque away”. Two of the four things this brief requires are a supply chain running at 4% of its own mandate and a facility that has never been built at the needed scale. The other two are a presidential-tier regulatory decision and a sixty-year record of appropriations that do not survive a change of administration. Nature is not the obstacle here. That does not make the obstacle small.