BRENNEN

Industries / Energy, water and waste / Power generation / Small modular reactors

Reactor designs

This is the work of choosing a small reactor's coolant, fuel form and layout, building it and running it, from Army and Navy plants of the 1950s to helium-, salt-, sodium- and lead-cooled machines.

6

Milestones achieved, from 1961 to 2023.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1961

    One rod, three deaths

    SL-1, in Idaho, was a 3 MW prototype boiling water reactor designed for the US Army, the first of several meant to power remote bases. On 3 January 1961 a night crew finishing maintenance after the Christmas break pulled its central control rod out by hand, and the reactor underwent a nuclear excursion, a sudden runaway burst of power, that killed three men. A 1962 Atomic Energy Commission review found that this one rod, withdrawn alone, could make the reactor critical, that is, start a self-sustaining chain reaction. Afterwards, by INL's account, procedures that had been two pages long "were expanded into thick books."

    Lessona design in which one control rod can restart the chain reaction depends on every person who handles that rod.

    osti.govinl.gov

    Credit

    US Atomic Energy Commission (investigation; review by A. Nelson Tardiff); Idaho National Laboratory (record)

  2. 1965

    Fuel that flows

    Oak Ridge National Laboratory's Molten-Salt Reactor Experiment (MSRE) was an 8 MW (thermal) test reactor whose uranium fuel was dissolved in a molten fluoride salt at about 650 °C (1,200 °F), pumped through a core of graphite bars. It first went critical on 1 June 1965, and by March 1968 it had run the equivalent of 9,000 hours at full power, critical 80% of the time over its last 15 months on uranium-235, by its operators' count. They then changed the fuel to uranium-233 and restarted in October 1968, and call it the world's first reactor fuelled with that material. Its records are still in use: a 2025 benchmark by M. Elhareef and colleagues tests reactor computer codes against its pump tests.

    media.githubusercontent.comosti.gov

    Credit

    Oak Ridge National Laboratory (Paul Haubenreich and J. R. Engel, account)

  3. 1986

    A reactor shuts itself down

    Experimental Breeder Reactor II was a 20 MW (electric) sodium-cooled fast reactor in Idaho; a fast reactor keeps its neutrons at high speed instead of slowing them with water or graphite. On 3 April 1986, with the reactor at full power and its automatic shutdown disabled, Argonne's team cut power to the pumps that drive coolant through the core; in a second test they stopped heat removal from the primary loop. Both times the metal fuel, sodium and steel expanded as they heated, which slowed the chain reaction, and the reactor shut itself down without damage to the fuel. About 60 visitors, 13 of them foreign specialists, watched the tests.

    osti.govne.anl.gov

    Credit

    Argonne National Laboratory; Pete Planchon (test lead)

  4. 2008

    A pebble bed ran hot

    Germany's AVR at Jülich, a 46 MW (thermal) test reactor, ran from 1967 to 1988 on fuel held in graphite balls, called pebbles, that move slowly down through the core, cooled by helium leaving at up to 990 °C. A pebble bed cannot carry instruments, so its temperatures were calculated. In a 2008 re-evaluation, Jülich researcher Rainer Moormann concluded that the core had run more than 200 K hotter than calculated, which was found only about a year before shutdown, and that this, more than poor fuel, explained why the cooling circuit was contaminated with strontium-90 and caesium-137 "some orders of magnitude" beyond prediction.

    Lessonwhen a core cannot be measured directly, its calculated temperatures need checking against whatever can be measured.

    api.core.ac.uk

    Credit

    Forschungszentrum Jülich (Rainer Moormann, re-evaluation)

  5. 2020

    Icebreaker reactors in series

    Russia's nuclear icebreaker Arktika raised its flag in Murmansk on 21 October 2020 and sailed on its first working voyage on 14 November. It carries two RITM-200 reactors of 175 MW (thermal) each. The RITM-200 is an integral design: its steam generators sit inside the reactor vessel, which Rosatom says makes it almost half the weight and size of earlier icebreaker reactors. Sister ships Sibir and Ural entered service in 2022 and Yakutia in 2024, making it one of the few small reactor designs built as a series; by early October 2025 Arktika had escorted 677 vessels, by Rosatom's count.

    rosatomnewsletter.comwnn.world-nuclear.org

    Credit

    OKBM Afrikantov (reactor design); Baltic Shipyard (ships)

  6. 2023

    Pebble-bed modules go commercial

    China's HTR-PM at Shidao Bay, Shandong, entered commercial operation on 6 December 2023, as announced by China's National Energy Administration. It pairs two small helium-cooled reactors of 250 MW (thermal) each with one 210 MW (electric) turbine, and each reactor holds more than 400,000 fuel pebbles 60 mm across. World Nuclear News calls it the world's first modular high-temperature gas-cooled reactor plant. In August and September 2023 Tsinghua University's team stopped normal cooling on each module at 200 MW (thermal) and reported that both cooled down without operators or emergency systems, a result it published in the journal Joule in 2024.

    world-nuclear-news.orgpowermag.com

    Credit

    Tsinghua University's Institute of Nuclear and New Energy Technology (design; chief engineer Zhang Zuoyi); China Huaneng (owner and operator); China National Nuclear Corporation (construction and fuel)

Watch for

  1. Linglong One awaits the grid

    China National Nuclear Corporation (CNNC) says its ACP100, also called Linglong One, will be the world's first commercial land-based small modular reactor. It is a 125 MW (electric) pressurised water reactor of integrated design, with the main cooling parts inside the reactor vessel, at Changjiang on Hainan island. Construction began on 13 July 2021 with a planned 58 months; CNNC finished a turbine test without nuclear heat on 23 December 2025, and a CNNC research official gave the first half of 2026 for commercial operation. In August 2026 World Nuclear News reported the unit was still undergoing pre-commissioning tests.

    Achieved whenthe unit is connected to the grid and China's regulator or the IAEA's reactor database records it in commercial operation.

    world-nuclear-news.orgworld-nuclear-news.org

    Claimed by

    CNNC, December 2025 to January 2026

  2. A lead-cooled fast reactor

    Rosatom is building BREST-OD-300, a 300 MW (electric) fast reactor cooled by molten lead, at Seversk in Siberia, beside plants that make its uranium-plutonium nitride fuel and will reprocess used fuel on the same site. Construction began in 2021 with operation first targeted for 2026. By the end of 2025 the reactor vessel was about 70% assembled, and in October 2026 Rosatom said the turbine foundation would be finished by the end of the year.

    Achieved whenthe reactor reaches criticality and supplies electricity to the grid, as recorded by Russia's regulator, Rostechnadzor, or the IAEA.

    world-nuclear-news.org

    Claimed by

    Rosatom, October 2026

Open

  1. 01

    Rerun the HTR-PM cooling test

    An independent computer model of the HTR-PM, built only from published design data, that predicts what Tsinghua measured in its two 2023 loss-of-cooling tests: how fast power fell and how hot the core and reactor vessel became, compared against the curves Tsinghua published.

    The brief
    Credit · unclaimed
  2. 02

    The Army's reactors as data

    One sourced, machine-readable operating record for the US Army's small reactor program of the 1950s to 1970s, plant by plant (SM-1, SM-1A, PM-1, PM-2A, PM-3A, SL-1, ML-1 and MH-1A): hours run, share of time available, each long outage and its cause, crew size where reported, and why each plant closed. Most of these plants were built off site and shipped in pieces to places such as Greenland, Alaska and Antarctica.

    The brief
    Credit · unclaimed

Open 01 · the brief

Rerun the HTR-PM cooling test

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An independent computer model of the HTR-PM, built only from published design data, that predicts what Tsinghua measured in its two 2023 loss-of-cooling tests: how fast power fell and how hot the core and reactor vessel became, compared against the curves Tsinghua published.

Why it is still open

The only measured results are Tsinghua's own, published in Joule in 2024. Idaho National Laboratory's 2026 simulation of HTR-PM loss-of-cooling accidents was checked against "similar solutions" from other studies, not against the tests, and found its results for the pressurised case "highly dependent" on how finely the model was divided. The Nuclear Energy Agency runs an international code benchmark on loss-of-cooling tests, but at Japan's HTTR test reactor, not the HTR-PM. Searches of OSTI and Crossref found no outside comparison with the HTR-PM measurements.

powermag.comosti.govoecd-nea.org

Built with what exists

OpenMC, an open-source code that simulates neutrons in a reactor core; the open-source MOOSE framework from Idaho National Laboratory for heat flow; WebPlotDigitizer, a free tool that turns published graphs into numbers; one workstation.

  1. First step

    Digitize the power and temperature curves in the Joule paper and list every model input, marking which ones are published and which must be estimated.

  2. Done looks like

    For both tests, a published model that reports its error against the digitized measurements, in °C for peak temperature and in hours for the time of the peak, at every measured location over the first 48 hours, with every input traced to a public source, all files on GitHub, and a second group rerunning the model and matching the first group's numbers.

Open 02 · the brief

The Army's reactors as data

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

One sourced, machine-readable operating record for the US Army's small reactor program of the 1950s to 1970s, plant by plant (SM-1, SM-1A, PM-1, PM-2A, PM-3A, SL-1, ML-1 and MH-1A): hours run, share of time available, each long outage and its cause, crew size where reported, and why each plant closed. Most of these plants were built off site and shipped in pieces to places such as Greenland, Alaska and Antarctica.

Why it is still open

The program's reports are scanned and free on OSTI, from a 1959 review of the first plant's operating experience to PM-3A's design reports. The summaries found cover one plant or stop early. The Atomic Energy Commission's 1968 booklet "Power Reactors in Small Packages" lists designs and start dates, and Oak Ridge's 2025 paper on microreactors for Antarctica gives PM-3A's ten years in one paragraph, saying it closed because the inspections and repairs it needed "were not cost effective." Searches of OSTI found no compiled record.

osti.govosti.govosti.gov

Built with what exists

OSTI's free search API and scanned PDFs; optical character recognition (Tesseract); a vision-language model that reads scanned tables and returns each number with its page; two people to check.

  1. First step

    Pull every OSTI record for the eight plants through the API and list which reports give operating hours and outages.

  2. Done looks like

    All eight plants in one open CSV file, every figure linked to a page of a public report, and a second person re-checking a random 20% of figures with the error rate published.