Since they were invented in the early 1950s, most money in magnetic confinement fusion research has been spent on two device types: the tokamak and the stellarator. These are also the only two device types that have demonstrated performance scaling by experiment towards that needed for a fusion power plant.
The tokamak is a donut-shaped device in which fusion fuel is contained by a combination of external magnets and a large electric current running inside the fuel itself. In the stellarator, all necessary containment is provided by shaping the external magnets in a particular way, in principle with a negligibly small current running in the fuel.
Of the two, tokamaks have received by far the most funding and investigation. The tokamak is usually described (for example, by the ITER Organization) as the best known prospect for fusion power. It does, however, have disadvantages. Perhaps the most fundamental is that the fuel current drives disruptions – events that end operation or even damage the device – that worsen with tokamak size and performance.
Due to this unfavourable scaling, disruptions will be worse and probably irremediable in any power plant-sized tokamak. While there is hope that this issue can be solved in tokamaks, there do not appear to be specific grounds to believe it will be. The main arguments for the tokamak are, first, that many large tokamaks already exist; second, that the stellarator has its own disadvantages.
The biggest disadvantage of the stellarator is asserted to be the shaping required for its external magnets. Although most tokamak coils are also now shaped, stellarator coils are typically shaped in three dimensions rather than just two. This is believed to be a decisive qualitative difference. For example, the fusion startup Thea Energy asserts that this added complexity is “prohibitive”.
It is not immediately obvious that this should be so. Fusion science grapples with great challenges, not only the engineering challenge of building a power plant based on new generation principles requiring a fuel hotter than the core of the sun, but in fusion plasma theory’s ambition to ‘understand and control’ turbulent plasma heat diffusion it implicitly intends to solve the Millennium Prize problem of fluid turbulence in an even more complex system. It seems to reflect some lack of ambition, or at least lack of proportion, to believe that bending metal into shapes is a comparable or greater challenge and a decisive obstacle to success.
Stellarator coil complexity is a recent interest. A search on Google Scholar for “stellarator coil complexity”, for example, yields only 1,570 results between the invention of the stellarator in 1951 and the year 2000, while the same search yields 5,210 results just between 2000 and 2024. This is not an artefact of a higher general rate of publication in recent years, as searching for “tokamak disruption” shows a much more even pace of publication. Why is this?
One possibility is that interest in stellarator coil complexity is motivated by one of the few notable absolute failures in large-scale experimental magnetic confinement fusion research. This is the cancellation of the US’s National Compact Stellarator eXperiment (NCSX) at Princeton Plasma Physics Laboratory (PPPL) in 2008.
The treatment of this event on Wikipedia, perhaps reflecting the popular understanding, strongly suggests that the project was an absolute technical failure due to coil tolerance requirements, and that the main culprit was the Department of Energy, stating, “the project has been cited as a case study of the hypothetical demon of Bureaucratic Chaos, which “blocks good things from happening” at the United States Department of Energy.”
The primary sources, however, paint a different picture. The Department of Energy’s own “Key Lessons Learned” (Section 8, pg 45) from this project gives only cursory attention to technical factors, instead suggesting that the main cause of the failure was PPPL’s failure to adhere to a clearly stated budget:
“1. Defining the Original Project Budget in an Unconstrained Manner
“In the late 1990s… A conceptual configuration of [NCSX] was developed. During the developmental period leading up to the CD-1 (approval of alternative selection and cost range), the project team was provided budgetary guidance that the NCSX Project should target a TEC of approximately $70M. It is unlikely that the conceptual design was adequately developed to address this budgetary constraint with any degree of cost certainty.”
Six of the seven points raised by the Department of Energy relate to PPPL mis- or underspecifying the design, underestimating costs, failing to estimate costs, and/or disregarding warnings to remain within the original budget. Only one, the fifth out of seven, centres on coil tolerances, stating to PPPL’s credit that, “Even the vendors, who have a history of complex fabrication, underestimated the cost of this requirement.”
Nonetheless, the overwhelming impression is that PPPL simply tried and failed to sell a client a product that was more costly than the client wanted to pay – as the client had clearly stated in advance. Had the project been pitched to the budget the client gave – for example, with a smaller machine, weaker magnets, or (as in early proposals) more re-use of existing components – on the basis of more conservative estimates for cost overruns that are common in R&D, this project would likely have succeeded.
Conversely, even a technically simpler project, if kept on life support indefinitely by the client, could have survived serious mismanagement. It is notable that PPPL is now entering its thirteenth year of attempting to build a medium-sized tokamak with apparently simple coils, while the stellarator was cancelled according to a strict budget criterion in its fourth year of construction. As PPPL stopped producing Annual Reports the year after the NCSX failure, it is not easy or perhaps possible to calculate how much this tokamak effort has cost the US taxpayer based on public information, but it is surely more than the stellarator and perhaps much more. Nonetheless, few would conclude from this that tokamak coils are inherently prohibitively complex.
Ultimately the question and significance of coil complexity should be tested empirically, like everything else in good science. It is not necessary to attempt to draw uncertain inferences from individual events or a mere ‘sense’ that one shape is more complex than another. Enough stellarators have been built successfully with modular coils that the cost per unit mass of magnets of stellarators and tokamaks can be directly compared from real world data, provided that all relevant institutes are willing to release this data in the public interest. This is the most important gap in the existing fusion science literature today.