ITER Fusion Project: Why Scientists Need a Reactor Hotter Than the Sun

At the center of the Sun, matter is compressed to something like 150 times the density of water, heated to roughly 15 million degrees Celsius, and held there by the sheer weight of everything sitting on top of it  a column of gas some 700,000 kilometers deep. Under those conditions, hydrogen nuclei fuse into helium constantly, releasing the energy that eventually reaches Earth as sunlight. And yet, per cubic meter, the Sun's core is a strikingly weak engine. ITER's own Director-General, Pietro Barabaschi, has pointed out that the power generated in a cubic meter of solar core material is comparable to what an active compost heap produces, and less than the metabolic output of a human body of the same volume. The Sun gets away with this because it is enormous and endlessly patient, converting hydrogen at that lazy rate across a core tens of thousands of times the volume of Earth, for about 4.6 billion years so far, with roughly another five billion still ahead of it.

That fact is the real starting point for understanding ITER, the International Thermonuclear Experimental Reactor now being assembled in Cadarache, in southern France. If a natural fusion reactor can afford to be slow and diffuse, a machine built inside a laboratory cannot. ITER has neither the Sun's gravity nor its patience. Its plasma will occupy about 840 cubic meters  the largest volume any fusion device has confined, though a rounding error next to a star. To get anything useful out of a machine that size, the reaction has to run far harder, hotter, and faster than it does inside the Sun. That single constraint, compressing into a small space and a short pulse what a star spreads across unimaginable space and time, explains almost everything unusual about how ITER is built and what it is actually trying to prove.

Fusion itself is a fairly old idea, and a fairly simple one on paper. Nuclei of light elements combine to form a heavier nucleus, and in doing so shed a small amount of mass as energy, following Einstein's e = mc². This is the mirror image of nuclear fission, in which a heavy nucleus such as uranium-235 splits into lighter fragments, also releasing mass as energy. Both processes draw on the same underlying quantity, binding energy, the energy holding a nucleus together, but from opposite ends of the periodic table. Iron sits near the peak of nuclear stability; elements lighter than iron release energy by joining together, and elements heavier than iron release energy by breaking apart.

For a fusion reactor, the reaction that matters most is between two isotopes of hydrogen: deuterium, whose nucleus holds one proton and one neutron, and tritium, whose nucleus holds one proton and two neutrons. Fuse them and the products are a helium-4 nucleus, a spare neutron, and 17.6 million electron volts of energy, split unevenly between the two, about 3.5 MeV carried by the helium nucleus and 14.1 MeV carried by the neutron. Written out: deuterium plus tritium yields helium-4 plus a neutron plus energy. Of all the fusion reactions physicists have studied, it is the easiest to ignite, which is exactly why every serious fusion reactor built to date, ITER included, is designed around it.

"Easiest" is a relative term here. Deuterium and tritium nuclei both carry positive charge, and like charges repel. The closer two nuclei approach, the harder the electrostatic repulsion, known as the Coulomb barrier, pushes back, right up until they are close enough for the short-range strong nuclear force to take over and bind them together. Getting past that barrier means giving the nuclei enough kinetic energy to force their way through the repulsion, which in a gas translates into temperature, tens of millions of degrees at minimum. At those temperatures atoms don't behave in any ordinary sense. Electrons are stripped away from their nuclei, leaving an electrically charged mixture of free nuclei and free electrons called a plasma, often described as a fourth state of matter distinct from solid, liquid, or gas.

Temperature alone isn't enough. A useful fusion reaction needs three conditions at once: a high enough temperature, a high enough density of nuclei to make collisions likely, and a long enough confinement time for those collisions to actually happen before the plasma's energy leaks away. Push the temperature up without density or confinement time, and reactions still occur too rarely to matter. This is part of why fusion research has taken decades longer than early optimists expected. It isn't one hard problem; it's three problems that all have to be solved together, in the same machine, without any one of them undoing the other two.

The Sun solves the density and confinement problems automatically, through gravity, which is why its core can get away with a comparatively mild 15 million degrees. Gravitational pressure holds the plasma at that extraordinary density indefinitely; there is nowhere for a hydrogen nucleus at the center of the Sun to escape to. ITER has no gravity to lean on. Its plasma density will be almost absurdly low by comparison, closer to an ordinary laboratory vacuum than to the density of water, and it can only be held together, briefly, by magnetic fields rather than by mass. To compensate for that low density and short confinement time, ITER's plasma has to run at roughly 150 million degrees Celsius, about ten times hotter than the Sun's core. That phrase, hotter than the Sun, is often misread as meaning fusion on Earth is somehow more powerful than fusion inside a star. It isn't. The extra heat is compensation, a workaround for everything the Sun has that a laboratory does not.

The machine designed to hold a 150-million-degree plasma without letting it touch a single physical surface is called a tokamak, from a Russian acronym meaning roughly "toroidal chamber with magnetic coils." The plasma sits inside a doughnut-shaped vacuum vessel, confined not by walls but by magnetic fields generated from several directions at once. Coils arranged around the tube produce a toroidal field, running the long way around the doughnut. A current driven through the plasma itself, effectively turning the plasma into the secondary winding of an enormous transformer, generates a second, poloidal field, looping the short way around the tube's cross-section. Together these two fields twist into a helix that spirals around the torus, so that a charged particle drifting outward along one field line is steadily redirected before it can build up enough sideways motion to strike the wall. ITER's field coils are superconducting, wound from niobium-tin cable and cooled to about 4 kelvin, just above absolute zero, so they can carry enormous currents without the resistive losses that would otherwise waste power and generate unmanageable heat. None of this removes the need for extra heating beyond what the plasma current itself provides; ITER will also fire neutral particle beams and radiofrequency waves into the plasma to push it up to fusion-relevant temperatures.

Why does a machine like this have to be so large? Part of the answer is that a bigger plasma volume is, somewhat counterintuitively, easier to keep hot. Fusion heat is generated throughout the plasma's volume, but it leaks out mainly through its surface, where turbulence and instabilities carry energy to the edge. As a device scales up, its volume grows faster than its surface area, so a larger machine has more heat-generating capacity relative to its heat-losing surface. Decades of experiments on smaller tokamaks, including JET in the United Kingdom and JT-60SA in Japan, have produced empirical scaling laws showing that energy confinement improves with machine size in a fairly predictable way. ITER, with a plasma volume around 840 cubic meters, roughly eight times that of JET, is built at close to the scale those scaling laws suggest is needed to reach conditions where fusion self-heating becomes significant compared with the losses that are always present.

None of this means ITER is trying to discover whether fusion works. Fusion has been demonstrated, studied, and measured in laboratories since the 1950s; JET itself produced real fusion power from deuterium-tritium plasmas in the 1990s. What ITER is built to investigate is whether a magnetically confined plasma at reactor-relevant scale can sustain a burning plasma, one dominated by its own internal heating rather than by external heaters, while simultaneously managing the heat exhaust, materials, magnet systems, and fuel handling that a genuine power-producing reactor would need, all integrated into a single machine for the first time. That is a scientific and engineering feasibility question, sharply distinct from whether fusion electricity can ever be produced cheaply enough to compete on a grid. ITER is designed to answer the first question. It says almost nothing, by itself, about the second.

ITER's most quoted figure, Q=10, sits squarely inside that first question. ITER is designed to deliver about 50 megawatts of external heating power into the plasma and, from that, produce roughly 500 megawatts of fusion power within the plasma itself, a tenfold gain, or Q=10. That figure is routinely misread as a claim about electricity, and it isn't one. Q=10 does not mean ITER will generate ten times more electricity than it consumes. ITER has no turbines and no generators; whatever fusion heat is produced will be absorbed by cooling systems and vented, not converted into electrical power. Meanwhile the machine as a whole, its magnets, cooling plants, control systems, and cryogenics, draws far more than 50 megawatts from the grid, at times several hundred megawatts. Q compares fusion power to the heating power delivered specifically into the plasma; it says nothing about the total electrical balance of the facility, which ITER was never designed to close.

Reaching Q=10 depends on the alpha particles produced by the deuterium-tritium reaction doing useful work. Each fusion event produces a helium-4 nucleus, an alpha particle, carrying 3.5 MeV of kinetic energy. Unlike the neutron from the same reaction, the alpha particle carries electric charge, so the magnetic field confining the rest of the plasma traps it too, at least for a while. As it collides repeatedly with the surrounding plasma, the alpha particle gradually surrenders its energy, heating the plasma from within. A plasma is described as burning once this internal heating from alpha particles becomes the dominant source keeping the reaction going, rather than the external heating systems. Burning plasma is one of ITER's central scientific targets, though it is not the same as a self-sustaining reaction that runs forever; ITER's deuterium-tritium pulses are expected to last on the order of several hundred seconds, and external heating and control systems remain very much part of the picture throughout.

Fuel supply looks straightforward for one half of the reaction and genuinely difficult for the other. Deuterium is stable, non-radioactive, and abundant: something like one in every 6,400 hydrogen atoms in ordinary water is deuterium, which makes the oceans, for practical purposes, an inexhaustible supply. Tritium is a different story. It is radioactive, decaying with a half-life of about 12.3 years, and it barely exists in nature; only trace amounts form when cosmic rays strike the upper atmosphere. Every gram of tritium used in fusion research today has been manufactured, largely as a byproduct of certain heavy-water fission reactors, and the global stockpile is small and shrinking as those reactors are retired. A power plant that depended on buying tritium from outside suppliers would run out of fuel almost as soon as it started. The proposed fix is breeding: surrounding the plasma with a blanket containing lithium, so that neutrons produced by fusion strike lithium-6 nuclei and split them into helium-4 and tritium, releasing a small amount of additional energy in the process. Lithium-7 can also react with fast neutrons to produce tritium, though that reaction absorbs energy rather than releasing it. In principle, a well-designed breeding blanket could produce more tritium than a reactor consumes. ITER will not run a complete, self-sufficient breeding system of its own; instead it will host test blanket modules, evaluating breeding concepts under real fusion conditions for the reactors that follow it.

The same neutrons that make breeding possible create a separate, harder problem. A 14.1 MeV neutron carries no electric charge, so the magnetic fields built to trap deuterium, tritium, and alpha particles have no effect on it whatsoever. It travels in a straight line until it strikes something solid, which in a tokamak means the structure surrounding the plasma. There it deposits its energy as heat, but it also does mechanical damage at the atomic scale, knocking atoms out of position in a material's crystal lattice and sometimes transmuting one element into another. Over years of operation this radiation damage embrittles metals, causes swelling, and gradually degrades exactly the properties, strength, ductility, thermal conductivity, that a reactor structure depends on. Fission reactors have decades of operating experience and a mature materials science base behind their neutron environments. Fusion neutrons carry several times the energy of a typical fission neutron, and materials qualified to survive that environment, at the fluence and duration a working reactor would demand, do not yet fully exist. It is one of the least publicized and most consequential gaps between demonstrating fusion and building a fusion power plant.

Nowhere is that gap more physically apparent than at the surfaces closest to the plasma. The first wall, and especially the divertor, a specialized structure at the base of the vessel designed to channel escaping plasma particles and heat away from the rest of the machine, absorb some of the highest heat fluxes engineered materials are ever asked to survive, comparable in places to a spacecraft heat shield during reentry, but sustained repeatedly rather than once. Under its revised 2024 baseline, ITER switched its plasma-facing first wall material from beryllium to tungsten, a metal with an exceptionally high melting point and low erosion rate under plasma bombardment. Tungsten is not a free upgrade: it becomes more brittle after neutron exposure, and can crack or melt locally under sudden heat spikes associated with plasma instabilities. The choice reflects a judgment that tungsten is closer to what future reactors, generally referred to as DEMO-class machines, will actually need, even though managing tungsten's own failure modes remains an active engineering problem rather than a solved one.

ITER's timeline has shifted enough, and often enough, that any single date deserves some skepticism, and the project's own history explains why. The original 2016 baseline aimed for a "First Plasma" milestone as early as 2025, a brief, low-energy test using a deliberately incomplete machine, followed by years of further assembly. By 2024, after delays attributed by ITER's Director-General to the COVID-19 pandemic, component quality problems, and planning that had been too optimistic for a first-of-a-kind, thirty-five-country engineering project, ITER's governing council adopted a substantially revised baseline. It dropped the stripped-down First Plasma concept in favor of a Start of Research Operations, targeted for 2034, using a more complete machine from the outset. From there the plan calls for deuterium-deuterium fusion experiments around 2035, operation at full magnetic energy and 15 megaamps of plasma current by 2036, itself a three-year slip from the prior target, and the beginning of full deuterium-tritium operation, with radioactive tritium fuel, in 2039, four years later than the previous plan. ITER describes the comparison between the old and new schedules as not entirely fair, since the two baselines describe genuinely different machine configurations and research sequences rather than the same plan simply running late.

Despite frequent conflation in press coverage, ITER is not a commercial power station, and it was never meant to be one. It has no connection to any electrical grid delivering power to homes or businesses. Whatever fusion energy its plasma generates will be captured as heat and rejected through cooling systems, not converted into electricity. Its purpose is to demonstrate, at reactor-relevant scale, that a burning deuterium-tritium plasma can be sustained and controlled, and to generate engineering data on magnets, materials, tritium handling, and remote maintenance that future reactors will depend on. Those future reactors, generally grouped under the label DEMO, are being designed separately by several national and regional fusion programs, not as a single unified follow-on project, and their job description is different: producing net electricity while proving out a more complete power-plant system, including continuous or near-continuous operation and full-scale maintenance logistics.

What sits between ITER and a commercial fusion industry is a longer list than most popular accounts acknowledge. Tritium self-sufficiency has to be demonstrated at actual reactor scale, not just in test blanket modules. Structural materials have to survive years, not months, of full-power neutron bombardment. Divertors have to handle sustained heat loads over long, repeated pulses rather than the several-hundred-second bursts ITER is designed for. Reactors have to be maintainable by remote-handling robotics, since no human can enter an activated fusion vessel. And underneath all of that sits the plainest question of all: whether fusion electricity, once technically achievable, can be produced cheaply enough to compete with whatever else the grid looks like by the time it's ready. Scientific feasibility and commercial viability are separate achievements, and ITER is built to address only the first.

Training for the field's next generation continues on its own schedule, regardless of ITER's. This November, the 16th ITER International School convenes at ICTS Bengaluru, India, from November 30 to December 4, jointly organized by India's Institute for Plasma Research and the International Centre for Theoretical Sciences, in partnership with the ITER Organization and Aix-Marseille University. Its theme, "Tokamaks as Fusion Reactors," marks a program mature enough to run twice in a single calendar year for the first time since it began in 2007; a fifteenth school, on heating and current-drive systems, met in Chengdu, China, only months earlier. Neither event changes ITER's physics. Both are a reminder that a research community several thousand strong is actively preparing for decades of work that will outlast the current machine's own construction schedule.

The deepest unresolved problem, though, sits outside ITER's walls entirely, in a facility that does not yet exist in finished form. No neutron source on Earth currently reproduces the full 14.1 MeV fusion neutron spectrum at the volume, intensity, and total accumulated dose that a working reactor's structural materials will experience over years of operation. Existing materials-testing reactors rely on fission-spectrum neutrons, lower in energy and somewhat different in the damage they cause, as an imperfect stand-in. IFMIF-DONES, a dedicated accelerator-driven neutron source under construction near Granada, Spain, is meant to close part of that gap, generating a neutron spectrum close to what a fusion reactor produces so that candidate materials can finally be tested under realistic conditions. Even once complete, it will irradiate small material samples rather than full reactor components, at a scale still short of the total fluence a decades-long reactor lifetime demands. Which means that by the time ITER, or the DEMO reactors meant to follow it, reach sustained deuterium-tritium operation, the materials lining their walls will in some sense still be an open experiment, one that no laboratory on the ground has yet been able to finish running in advance.

YOU MAY ALSO LIKE

CERN’s LHC Shutdown: Inside the High Luminosity LHC Upgrade