US Edition
Your source for latest news
ScienceFusion Energy

New Theory Predicts the Power Fusion Reactors Need to Reach Peak Efficiency

A physics-based theory from EPFL researchers, rather than one fitted to past experiments, finds that SPARC, the compact reactor under construction in Massachusetts, may have little margin to spare.

PS
By PressTemps Science DeskPublished Yesterday, 17:24 ET · 6 min read
What to know
EPFL physicists derived, from first principles, a theory predicting the heating power needed to trigger H-mode, the high-confinement plasma state essential to fusion reactor performance.
The theory, published Nov. 13 in Physical Review Letters and announced by EPFL on Nov. 20, 2026, also explains why the threshold depends on which way a reactor's magnetic field points — a 44-year-old open question.
It predicts ITER needs about 43 megawatts of heating power against a 73-megawatt capacity, while SPARC needs about 26 megawatts against a planned 25-megawatt capacity — a far tighter margin.
The prediction has not yet been tested on an operating reactor; SPARC's first plasma is expected in 2027, ITER's research operations around 2034.

Physicists at the Swiss Federal Institute of Technology in Lausanne (EPFL) say they have derived, from first principles, the amount of heating power a fusion reactor needs to flip its plasma into a more efficient, energy-retaining state — a question that has shadowed tokamak design for 44 years. The theory, published Nov. 13 in Physical Review Letters and detailed in an EPFL announcement posted Nov. 20, gives engineers on the world's two most closely watched fusion projects a physics-based way to check whether their machines will have enough muscle to reach that state at all.

The projects are ITER, the 35-nation reactor under construction in southern France, and SPARC, the compact tokamak Commonwealth Fusion Systems is building in Devens, Massachusetts. Both depend on reaching what plasma physicists call H-mode, or high-confinement mode, a condition in which turbulence at the edge of the superheated gas suddenly collapses and heat stops leaking out as fast. Without it, a reactor can burn plasma for a long time without ever generating useful amounts of energy.

A 44-year-old gap in fusion theory

Researchers first saw the transition in 1982, at the German ASDEX tokamak: raise the heating power past some threshold, and the plasma's edge turbulence drops sharply, letting the core hold onto its heat. The effect has since been reproduced on dozens of machines, but no one had worked out, from the underlying plasma physics, exactly what that threshold power should be. Designers instead relied on empirical formulas fitted to results from existing, smaller tokamaks and then stretched to predict the behavior of much larger ones — an extrapolation that leaves real uncertainty when applied to reactors that do not yet exist.

The EPFL team, led by doctoral researcher Brenno De Lucca of the university's Swiss Plasma Center with colleagues Paolo Ricci, Benoit Labit, Davide Mancini, Louis Stenger and Zeno Tecchiolli, built three-dimensional simulations of plasma turbulence in a realistic reactor geometry. The simulations showed the turbulence itself generating sheared flows near the plasma's edge, and those flows suppressing the very turbulence that created them — the mechanism behind the H-mode transition. The model also reproduced a quirk long seen in experiments but never fully explained: the threshold power is noticeably lower when the reactor's magnetic field points in one direction around the machine than the other. The team traced that asymmetry to collisions between particles breaking a time-reversal symmetry in the turbulent flow. Using the simulations, the researchers then derived mathematical scaling laws for the threshold power, including the plasma density at which the least power is needed, and found the formulas matched existing experimental databases at least as well as the empirical scalings now in use.

Tight margins for two flagship reactors

Applied to real hardware, the theory produces numbers with immediate design implications, according to the paper and the EPFL summary:

  • For ITER, the predicted minimum heating power to reach H-mode is about 43 megawatts, against a planned auxiliary heating capacity of 73 megawatts — a comfortable margin.
  • For SPARC, the predicted minimum is about 26 megawatts, versus a planned heating capacity of 25 megawatts, leaving, in the researchers' description, considerably less room for error.
  • The underlying 1982 discovery, at the ASDEX tokamak in Germany, is the origin point for a search that has run for more than four decades.
  • The paper appears in Physical Review Letters, volume 137, article 205101, with an earlier version posted to the arXiv preprint server in May and revised through October 2026.

SPARC is the demonstration machine for Commonwealth Fusion Systems, the MIT spinout working to show it can get more energy out of a fusing plasma than it takes to heat it, a milestone the company has said it is targeting for 2027. Commonwealth Fusion Systems has reported the reactor's vacuum vessel and magnet assembly are well advanced, with first-plasma operations following completion. If the new threshold prediction holds, SPARC's design will have little spare heating capacity to compensate for unexpected losses or diagnostic uncertainty once operators try to trigger H-mode. ITER, by contrast, appears to have a wider cushion, which matters given the project's own history: its schedule has already slipped years behind the original plan, and current estimates point to first operations around 2034, according to reporting on the project's repeated delays. ITER's own facts and figures page lays out the scale of the machine the theory is being tested against: a tokamak designed to produce ten times more fusion power than the power used to heat the plasma.

Who is watching the result

The audience for the work extends beyond the two headline projects. Every tokamak-based design under study for a future power plant, including the ARC reactor Commonwealth Fusion Systems plans to build near Richmond, Virginia, in the next decade, and various national and private efforts in the United States, Europe and Asia, needs some assurance that its heating systems can push the plasma into H-mode before it is built, not after. A theory that predicts the threshold from physical principles, rather than from curve-fitting past experiments, gives designers an independent check on those plans and a way to flag machines, like SPARC, where the margin for error is thin.

The result has not gone through the kind of experimental confirmation that would settle the matter outright. The scaling laws were checked against existing multi-device databases rather than against a newly built reactor, and researchers in the field continue to refine competing empirical fits separately, including a 2026 reanalysis of threshold data from metal-walled tokamaks published in the journal Nuclear Fusion. The EPFL group's theory is also built on simulations of a specific, simplified diverted-tokamak geometry; translating it cleanly to the particular engineering of ITER or SPARC, each with its own wall materials, divertor design and heating hardware, is itself a nontrivial step that outside researchers are likely to scrutinize.

What comes next

The immediate test will come from the machines themselves. ITER's construction has been proceeding through a sequence of sector installations this year, and the project's current plan calls for a hydrogen-and-deuterium research campaign to begin around 2034 rather than the low-power "first plasma" milestone once envisioned for 2025. SPARC's builders expect to reach first plasma in 2027, which would make it the sooner of the two opportunities to compare the new prediction against an actual machine, albeit one operating in early, lower-power campaigns rather than at full design heating power right away. Confirmation or contradiction on either device would carry weight well beyond the two projects, since both are treated across the field as reference points for the scaling laws that feed into every subsequent reactor design. For now, the EPFL group's contribution is a prediction with a number attached to it, derived from the physics of the plasma edge rather than from a fit to past machines, and a modest but specific warning that one of the world's most advanced fusion prototypes may be cutting its margin closer than its builders would like.

More on this story

All Science