The science of Commonwealth Fusion Systems sorts into three kinds of claim, and telling them apart is the most important thing to understand about the corpus. The magnets have been built and measured. The plasma performance has been modeled and published in advance of any machine that could test it. And a third body of work, easy to mistake for either of the other two, consists of real measurements taken on tokamaks the company does not own — 📝JET, 📝ASDEX Upgrade, 📝MAST Upgrade, 📝TCV, 📝DIII-D, 📝Alcator C-Mod — and used to anchor the predictions. All three are real work. They are not the same kind of claim. This hub is organized by that distinction first and by 📝Commonwealth Fusion Systems's own topic taxonomy second — magnets, physics, ecosystem — so that a reader can always tell which of the three a given number came from.
Magnets — what has been measured
This is the demonstrated half. 📝VIPER: An Industrially Scalable HTS Cable qualified the conductor in 2020 across thousands of mechanical cycles and dozens of 📝quench transients. 📝The SPARC Toroidal Field Model Coil Program (Paper) records what that conductor then did: 20.1 📝tesla peak field on conductor at 40.5 kiloamps, on 5 September 2021, in the 📝Toroidal Field Model Coil (TFMC) led by 📝Zachary Hartwig. 📝The TFMC Engineering Papers are the seven companion publications on how it was built, cooled, powered and deliberately quenched. 📝Charlie Sanabria's 📝PIT VIPER: A Low-AC-Loss Cable for Pulsed Magnets extended the cable to pulsed magnets and the fast-ramping 📝central solenoid.
What follows a demonstration is production, and that is the half of the magnet record the model coil papers do not cover. 📝The Production Magnet Papers — fifteen papers in IEEE Transactions on Applied Superconductivity between 2023 and 2026 on conductor acceptance, cable manufacture, test facilities and quench detection — record a one-off coil becoming a manufacturing line. 📝HTS Coils for a Magnetic Mirror shows the same technology built and measured for a machine that is not a tokamak: two sub-two-tonne 20-tesla 📝REBCO coils, designed in 2023 and tested by 2025. Underneath sits a supply chain — 📝YBCO Wire Production for Fusion — and, further underneath, 📝The Antecedent Papers from 2001 and 2012 that CFS did not write. 📝Superconductors for Fusion: A Roadmap supplies the field-wide view these papers cannot supply about themselves.
Measurement does not stop at the magnets. 📝ARC Engineering Beyond the Physics Basis reports cryogenic low-cycle fatigue data for 316 stainless steel and IN718 at 4 K and 77 K — laboratory results, not projections, for structures belonging to a power plant that does not yet exist. The distinction the rest of this hub turns on is not between old work and new work; it is between what an instrument recorded and what a code computed.
Physics — what has been predicted
This is the projected half, and it is much larger. 📝The 2015 ARC Design Paper is the conceptual study the company was founded on. 📝The SPARC Physics Basis (2020) is the seven-paper collection projecting what 📝SPARC will do — 📝Q ≈ 11 nominal against a Q > 2 mission, per 📝Overview of the SPARC Tokamak and framed by 📝Status of the SPARC Physics Basis. 📝The 2022 SPARC Physics Basis Overview is the ten-page successor that revised parts of it two years later, recording the 📝tungsten wall decision and a changed 📝divertor heat flux.
The clearest thing this half of the corpus has to show is not a number but a number moving. 📝Revising the SPARC Core Performance Prediction follows the projected gain of SPARC's reference discharge across four papers: Q ≈ 9 in 2020 from a reduced turbulence model, no gain figure at all in 2021 while nonlinear 📝gyrokinetics established what that model had assumed, Q ≈ 8 in 2022 once surrogate modeling made a converged gyrokinetic solution affordable, and in 2024 a different question entirely — what breakeven takes in a first 📝L-mode campaign. No measurement existed at any point in that sequence. The prediction was revised twice anyway, in public, because the modeling got better. That chain is the organizing principle of this hub demonstrated rather than asserted: a projection published in advance is the only kind that can be seen to move.
The rest of the SPARC physics record extends the 2020 collection paper by paper. 📝The Pedestal and Boundary Follow-Ups carry the 📝pedestal and edge case forward from 2024 to 2026; 📝The SPARC Disruption Modeling Papers finish the self-consistent 📝disruption modeling the 2020 chapter left open, including the unmitigated damage case; 📝The Runaway Mitigation Coil Papers and 📝Thin-Wall Eddy Current Modeling develop the passive coil meant to handle 📝runaway electrons, a device no disruption on any machine has yet tested. 📝The SPARC ICRF Conference Papers cover the machine's only external heating, and 📝The SPARC Neutronics and Shielding Papers a branch of the record CFS's own publication index does not list.
📝The ARC Physics Basis (2026) does for the power plant what the 2020 collection did for SPARC, projecting 1.13 gigawatts and 400 megawatts to the grid in 📝Overview of the Physics Basis for the ARC Fusion Power Plant. Its longest-standing commitment predates it by eight years: 📝The ARC Divertor Papers of 2018 and 2019 are where ARC's long-legged, tightly baffled 📝X-point target divertor came from.
Not one of those plasma figures has been measured. 📝ARC has not been built; SPARC has not yet produced a 📝plasma. 📝SPARC Early Campaign Diagnostics and 📝The SPARC Diagnostics Papers — roughly fifty instruments described across eighteen papers — are the bridge between the halves, and the reason the bridge matters is that their job is to convert every projection above into a result that can be checked or refuted.
Measured elsewhere, applied here — the third category
Between the demonstrated half and the projected half sits a body of work that belongs to neither, and mistaking it for either one is the most common way to misread this library. 📝CFS Science on Other People's Machines is its anchor: ten peer-reviewed papers published between 2023 and 2026 in which CFS-affiliated authors report measurements taken on tokamaks the company neither owns nor operates. An infrared 📝bolometer watching the radiation front move along the 📝separatrix on MAST Upgrade. Argon injection on JET radiating more than three times as much near the injector as toroidally distant from it. The HEAT code checked against ASDEX Upgrade infrared thermography. Blob velocities from Langmuir probes on six machines at once. These are results, not projections — and not one of them is a SPARC result.
The category is necessary rather than merely tidy because these measurements are load-bearing for the predictions above. 📝The Separatrix Operational Space Paper builds its framework on 6,688 measured ASDEX Upgrade points and then uses it to project an ELM-free operating point for SPARC. The ThinCurr eddy-current code was benchmarked against DIII-D conducting structures before it was pointed at SPARC's mitigation coil. The pedestal follow-ups reach back into Alcator C-Mod density data taken years before SPARC existed on paper. Every projected number in the section above rests, somewhere below it, on a number an instrument recorded on a machine in Oxfordshire, Garching, Lausanne, San Diego or Cambridge.
The rule this library follows is to name the machine, every time. A reader who takes a JET measurement for a SPARC result has been misled exactly as badly as one who takes a projection for an achievement — and the third category is where that mistake is easiest to make, because everything in it is genuinely measured.
Ecosystem — why any of this is public
A private company has no funder requiring it to publish, which makes 📝The Role of Peer Review in Commercial Fusion the paper that explains the rest of this hub. 📝Matt Reinke, 📝Brandon Sorbom and 📝Martin Greenwald argue that peer review is how a claim about an unbuilt machine acquires standing, and say plainly where proprietary work limits that. Read against the corpus, it is also the strongest defense of the prediction-heavy half: publishing projections in advance, with their codes and conservative cases attached, is a harder claim to make than publishing results afterward.
Publication is one form of openness and access is another. 📝SPARC as a Platform to Advance Tokamak Science makes the second case — that a privately built machine should carry outside experiments and release its data — and it is the most-cited non-magnet paper in this corpus, which is itself a result of some kind. 📝The Academic Research Ecosystem Paper supplies the constraint underneath both: a faculty headcount showing that the United States academic base for 📝fusion needs to grow in absolute terms and to spread across more institutions, because a field that runs on graduates cannot expand faster than the people teaching it.
📝The Public-Private Funding Record is the part of the record the publication list does not show at all — eleven Department of Energy and OSTI technical reports plus a recorded seminar, from 2022 to 2026, documenting collaborations with national laboratories and universities. The engineering roadmap is more visible in those reports than in the journal literature, which is a fair description of where the peer-reviewed corpus stops.
Reading the line
The mechanical test for any figure in this library is which of the three categories it came from, and the test is answerable in one question: what recorded it, and where. Twenty tesla, 40.5 kiloamps, four kilometers of cable fabricated, a quench at 31.5 kiloamps, low-cycle fatigue at 4 kelvin — measured here, on hardware CFS built. Q ≈ 11, Q ≈ 8, 140 megawatts, 1.13 gigawatts, one disruption per day, argon at 0.9 percent — predicted here, by codes, for machines that have not run. And 6,688 separatrix measurements, a hundredfold drop in thermal electron density in five milliseconds, a radiated thermal energy fraction of 0.84 — measured elsewhere, on ASDEX Upgrade, DIII-D and JET, and applied here. A figure that cannot be sorted into one of those three has not been read carefully enough yet.
The 📝Fusion Glossary is the descent path for any term in any of the three, and every paper memo here carries a DOI for readers who want the source rather than the summary. When SPARC produces its first plasma, the second category starts converting into the first, roughly half this hub becomes checkable, and the memos that need rewriting are the ones that say "projected."