The deuterium-tritium (DT) reaction is a type of nuclear fusion in which one deuterium nucleus (the abundant and non-radioactive hydrogen 2 isotope) fuses with one tritium nucleus (the rare radioactive hydrogen 3 isotope), giving one stable helium 4 nucleus and liberating one free neutron as well as a vast amount of energy. The DT reaction is the most favored of fusion reactor designers because its reactivity is two orders of magnitude larger than any other reaction at accessible operating temperatures. However, this reaction has at least five major drawbacks that revolve around tritium or the neutron reaction product (Jassby 2017; 2018). The present article is concerned with only one of these drawbacks, namely, shortcomings in tritium fuel supply.

Tritium is radioactive with a half-life of 12.3 years. Because of this short half-life, there are no natural resources of this nuclide, and it must be manufactured in nuclear reactors or accelerators. By contrast, its reaction partner (deuterium) does not decay and is easily extracted from ordinary water, being one part in 6,500 of natural hydrogen.

Almost all tritium for civilian use is currently sourced from CANDU (Canada Deuterium Uranium) fission reactors in Canada, where it is unwelcomely formed by unavoidable neutron capture in the heavy-water (deuterium-oxide) moderator, then extracted and stored. Other CANDU reactors are found in Argentina, China, Romania, and South Korea, but tritium is being extracted currently only in South Korea. Table 1 shows assured and potential quantities of tritium that can be extracted from all heavy-water reactors (HWRs) worldwide, including many units in India. Current world stockpile is no more than 30 kilograms (kg), and this quantity will decrease drastically beginning around 2040 as source reactors complete their useful life and the tritium decays. Also, 12 kg may be allocated to ITER (initially, the International Thermonuclear Experimental Reactor), the world’s largest experimental fusion tokamak currently being built in France by a consortium of several dozen countries (Claessens 2023).

Required tritium inventories

There are two broad approaches toward developing terrestrial fusion reactors: magnetic confinement or inertial confinement. In magnetic confinement fusion (MCF), such as in tokamaks and stellarators, magnetic fields are used to confine the hot fusion fuel in the form of a large plasma that is maintained for seconds or longer. In inertial confinement fusion (ICF), laser or particle beams are used to compress and heat a small capsule of fusion fuel to generate a micro-explosion of nanosecond duration.

In the 75-year history of fusion research and development, only two MCF devices have ever used tritium, namely the Tokamak Fusion Test Reactor (TFTR) in Princeton, New Jersey (Strachan et al. 1997) and the Joint European Torus (JET) in the United Kingdom (Keilhacker et al. 1999). Many ICF devices have used tritium, including the National Ignition Facility (NIF) near Livermore, California, and the OMEGA facility in Rochester, New York, both based on powerful laser beams.

One of the most critical reactor parameters is the percentage of tritium that is burned (fused) with deuterium as it passes through the reacting plasma. The reaction vessel is continually pumped, and the output must be processed to recover and re-inject the unburned tritium. If the processing cycle takes 24 hours, then the required reactor tritium inventory is simply the amount of tritium burned daily divided by the percentage burned in each pass.

The burnup fraction is at least an order of magnitude (10 times) higher in inertial confinement systems than in magnetic confinement systems. In ICF fuel capsules, the tritium is installed exactly where the temperature will become the highest, as the tritium itself defines the “hot spot,” and it retains its relative position during implosion and while the capsule disassembles. In magnetic confinement devices, however, only a quarter of the injected fuel reaches the hot reacting core plasma and then immediately begins to escape, so that continuous refueling is necessary.

Table 2 compares the tritium inventory required for both ICF and MCF systems for two processing cycles of 12 hours and 24 hours. The numbers shown are for reactors producing 2,000 megawatts (MW) of fusion power, which requires burning 300 grams of tritium per day. (That fusion rate could be converted to about 800 MW of gross electric power.) The fact that ICF systems require 20 to 40 times less tritium inventory than MCF systems reflects the much higher tritium burnup fraction of the inertial confinement approach.

Table 3 summarizes the cost per gram of tritium from various sources and the total tritium cost needed to start up a 2,000 MW fusion reactor. The least expensive source of tritium is from CANDU reactors, costing about $30,000 per gram. For scale, fusing one gram of tritium produces 550 gigajoules of thermal energy, which can theoretically be converted to $7,000 worth of electricity at 10 cents per kilowatt-hour. That is a small fraction of the cost of one gram of tritium from any source.

Possible scenarios for tritium use

Relevant nuclear reactions vary depending on the dominant fuel of the fusion reactor (Figure 1). Every DT reaction that consumes one triton (the nucleus of the tritium atom) produces one neutron, which can be absorbed in lithium to regenerate the triton. There are four conceivable scenarios for tritium use and production in civilian fusion devices over the next half century: the conventional, the back-up, the chimerical, and the realistic.

The conventional scenario

Present plans for ITER call for the acquisition and consumption of one or two kilograms of tritium annually beginning in 2039, and a total consumption of 12 kg, which would amount to one-third of the world’s assured tritium stockpile and one-quarter of the maximum potential stockpile (see Table 1).

Then, two DEMO (demonstration) reactors or pilot plants are started up with the remaining tritium in the 2040s or later (Gianfranco 2022). These two DEMO reactors have lithium blankets that “breed” (that is, produce) adequate tritium for self-sufficiency, as well as excess tritium to start up new reactors, ad infinitum. This last objective, however, is unlikely to be viable.

The back-up scenario

If no tritium is available for starting up any DEMO reactor, the same reactors using only deuterium fuel can gradually build up tritium from deuterium-deuterium (DD) reactions (see Figure 1). However, the rate of the DD reaction is only one percent of that of the DT reaction under similar plasma conditions. During the three to five years that would be needed to build up the kilograms of required tritium inventory, and with zero revenue, the facility must service its debt for construction as well as pay for internal power consumption (at least 200 MW) and other operational costs. Analysts estimate that the effective cost of tritium would be of order $1 million per gram (e.g., see Kovari et al. 2018).

The scramble for tritium could enlist commercial fission reactors, as the US nuclear weapons program does, by replacing standard control rods with lithium-containing neutron absorber rods (NNSA 2016). Commercial light-water reactors (LWRs) can be fitted to supply limited quantities at a price estimated at $100,000 per gram, giving a startup cost of $500 million to $1 billion in US dollars (see Table 3).

However, modification of fission reactors to produce significant tritium can lead to disasters, such as the infamous 1957 Windscale fire—the United Kingdom’s worst nuclear accident to date. When a large amount of tritium was needed for immediate military use, one fission reactor of the Windscale plant was modified to breed tritium from a lithium compound, but with inadequate attention paid to the possible consequences of this change. The modifications led to overheating of some reactor components with an ensuing fire. Today, stringent safety precautions are responsible for some of the huge cost of tritium produced in modified lightwater reactors.

The chimerical scenario

In a recent phenomenon of so-called “strategic planning,” national governments have apparently begun to believe the preposterous claims of several dozen “fusion start-ups” that they will put fusion-based electricity on the grid during the next decade. Consequently, the governments of the United States, United Kingdom, South Korea, and Japan, among others, have decided to put aside their own plans for relatively conservative DEMO reactors planned for the 2040s or later in favor of far more ambitious electricity-producing “fusion pilot plants” on a shorter timescale, by 2040 (NASEM 2021).

This date of 2040 happens to be when, according to its plans, ITER will first require tritium. This means that ITER and just two pilot plants would essentially gobble up all the assured tritium stockpile. Moreover, there is no obvious source of tritium for the host of fusion power reactors that private companies claim they will build in the 2030s.

This deployment scenario, embraced by the Fusion Industry Association and by national departments of energy, is entirely fantasy, as there are three fundamental show-stoppers: First, to date, no fusion concept, other than radiatively driven implosion as practiced by the NIF, has come anywhere close to demonstrating scientific feasibility, defined as the ability to reach thermonuclear ignition (Jassby 2022). Second, every fusion device consumes megawatts to hundreds of megawatts of electricity, but no device has ever produced even a token amount of electricity (kilowatts) while gorging on megawatts. Third, 80 percent of DT fusion energy emerges as streams of hugely energetic neutrons, but no one in any line of endeavor—reactors or accelerators—has ever converted neutron barrages into electricity.

The realistic scenario

First, ITER will probably never operate. Delays in the ITER schedule have been announced periodically for nearly 30 years, and the project has recently postponed the startup date to 2034 and the first use of DT plasmas to 2039 (Cleary 2024). But if precedent means anything, new problems will arise leading to further schedule delays. ITER has turned out to be “a bridge too far.” The experimental reactor was supposed to lay to rest the uncertainties for MCF devices of achieving the conditions for a power reactor. It’s unlikely, however, that ITER will ever become operational, so that the fundamental assumptions underlying DEMO operation for MCF-based facilities may not be tested.

Second, if ITER operation is a daydream, then DEMOs and pilot plants are hallucinations. Aside from the fundamental show-stoppers listed above, none of the DEMO reactors or pilot plants can possibly be built because of non-existent technologies, insuperable manufacturing issues, and price tags in the tens of billions of dollars. Until these three show-stoppers are resolved, it is likely that successful DEMO reactors and pilot plants of any type, not just ITER-based, will remain indefinitely far away.

In any given rational scenario, there will be no tritium shortage for non-military use in the foreseeable future, not because the supply will suddenly become generous but because demand will remain limited. The global non-military consumption of tritium has never been more than 400 grams per year and is unlikely to exceed one kilogram for decades. The numerous research and development facilities expected to be implemented over the next decade in many countries for research on tritium diffusion, recovery, and processing, as well as the SPARC tokamak being built by Commonwealth Fusion Systems in the United States (Creely et al. 2020), will each require about 100 grams of tritium every five or 10 years. This amount can readily be accommodated by the continued extraction of tritium from currently operating heavy-water reactors. But most tritium is likely to continue to simply decay in place at its point of origin.

The critical matter of tritium replenishment

While startup of any reactor would require typically five to 10 kg of tritium, a reactor generating 2,000 MW of fusion power (converted to 800 MW of electricity) would consume 100 kg per year at 90 percent capacity factor. Given the exorbitant cost of producing tritium (see Table 3) and the severely limited supply in any event, it’s obvious that every DT-burning reactor must replenish all of its own fuel. A 10-percent shortfall in a single reactor (10 kg) would consume the world’s entire stockpile in just a few years (see Table 1).

Designers of conceptual fusion reactors commonly assume that all the burned tritium will be replaced by absorbing the fusion neutrons in a lithium-containing “blanket” completely surrounding the reacting plasma (Figure 2). However, the lithium blanket can only partly surround the reactor because of the gaps required for vacuum pumping, heating and fuel injection in magnetic confinement systems, and for driver beams and removal of target debris in inertial confinement systems. Heat transfer structure within the lithium blanket will also absorb neutrons. Nevertheless, the most comprehensive analyses indicate that, by the judicious use of neutron multipliers, there can be a five to 15 percent surplus in regenerating tritium under ideal conditions (Abdou et al. 2021). In practice, any surplus will be needed to accommodate the incomplete extraction and processing of the tritium bred in the blanket and to compensate for tritium decay during downtimes.

Replacing the burned-up tritium in a fusion reactor, however, addresses only the first major issue of tritium replenishment. The second major problem is compensating for the irretrievable loss of tritium in reactor components and processing systems. In magnetic confinement devices, only a quarter of the injected fuel reaches the hot reacting core and then immediately tries to escape. While there will be some recycling from the vessel wall, less than five percent of the injected fuel will actually be burned before it is pumped out or lost. Most of it simply goes out with the plasma exhaust, but a significant fraction of the injected tritium must be scavenged from the surfaces and interiors of the reactor’s myriad sub-systems. All the tritium, whether readily pumped or painstakingly scavenged, must be re-injected some 20 times before it is completely burned. During their two dozen traverses of the plasma, vacuum, reprocessing, and fueling systems, some tritium atoms will be permanently trapped in the vessel wall and in-vessel components, and in plasma diagnostic and heating systems. In magnetic confinement systems, if only one percent of the unburned tritium is not recovered and re-injected, even the largest theoretically conceivable surplus in the lithium-blanket regeneration process cannot make up for the lost tritium.

In ICF reactors, the tritium burnup per pass will be 35 percent or more, so that tritium losses of even one percent will be insignificant compared to the amount burned. ICF systems, therefore, not only have much smaller inventories than MCF systems (see Table 2), they have a much better chance of maintaining their fuel supply.

No MCF project other than TFTR and JET has ever dared to use tritium because of its cost and safety issues. In each of the two cases, the on-site inventory did not exceed 50 grams, about one percent of that characteristic of putative power reactors (see Table 2). In the TFTR and JET campaigns of the 1990s, approximately 10 percent of the injected tritium was never recovered (Tanabe et al. 2003). Then, in JET’s 2021 and 2023 campaigns, using a new beryllium wall that absorbed less tritium in the plasma vessel, the loss was reduced to about one percent, but only after extensive baking and other treatments of the vessel components requiring downtimes that would not be practical in a power reactor (Matveev et al. 2023).

In the TFTR and JET experiments, the actual fractional burnups of tritium were an astoundingly low 0.01 percent. The only new MCF facilities funded and with plans to use tritium are SPARC around 2030 and ITER around 2040. Both reactors are supposed to have a tritium burnup of 0.5 percent or less. Even the conceptual DEMOs are expected to have burnup of no more than one percent, so that a one-percent tritium loss would require an impossible tritium breeding ratio of two. By contrast, in recent years ICF experiments at the NIF have demonstrated burnups of two to five percent, levels that may never be reached in MCF devices.

Fusion engineers love to calculate doubling times, the number of years necessary to double the reactor’s tritium inventory so that an additional reactor can be started up (Abdou et al 2021). The calculations always assume the theoretical maximum tritium breeding ratio and ignore all tritium losses as well as extraction difficulties. But since self-sufficiency is likely impossible, the concept of doubling time is meaningless. In fact, it would be more constructive to speak of “tritium halving times” instead, because the initial tritium inventory will gradually disappear.

To make up for the inevitable shortfalls in recovering unburned tritium fuel, it would be tempting to call on fission reactors to provide makeup tritium. But an appropriately modified fission reactor could supply less than two percent of the tritium required by a fusion reactor with the same power output—and at exorbitant cost. In any event, it seems preposterous that fusion reactor operation must depend on fission reactors, given that fusion promoters constantly boast of the alleged superiority of envisioned fusion power plants over actual fission power plants. From the viewpoint of fuel supply, since support from external tritium production is impractical, it is likely that only fusion reactors fueled solely with deuterium can ever be viable.

Paucity of tritium breeding experiments

While six decades of textbooks on fusion energy have assumed that burned tritium would be replenished in a lithium-containing blanket, no one has ever done an experiment in an actual fusion device. To date, tritium-breeding experiments have used only point neutron sources installed within blocks of lithium compounds. Essentially, they provide confirmations or corrections of nuclear cross sections, but no experiment has ever been done with an external volume neutron source (see Figure 2), which is characteristic of all fusion reactors.

Around 1980, my colleagues and I at the Princeton Plasma Physics Laboratory realized that the TFTR tokamak, then under construction at our institute, could be fitted with one or more blanket modules to gain quantitative insight into the validity of the neutronics calculations of blanket breeding performance with a volume neutron source. The Electric Power Research Institute funded the design, development, and actual fabrication of one blanket module called the LBM, for Lithium Blanket Module (Jassby et al. 1985). The LBM was completed by 1984. However, the TFTR administration later refused to install it, for unclear reasons, and the LBM was sent to other laboratories for old-fashioned irradiation with point-neutron sources. The JET project, too, had no interest in tritium-breeding blanket modules.

Perhaps it’s unremarkable that both the TFTR and JET administrations were reluctant to host tritium-breeding experiments. After all, an unfavorable result would challenge the fusion energy myth of “unlimited fuel supply.”

The LBM was similar to some of the blanket modules that have been designed for installation on ITER, which is supposed to have six different tritium-breeding modules comprising less than two percent of the vessel wall. But these experiments would shed little light on the vital global breeding ratio, the value that takes into account all the blanket penetrations that lead to neutron loss (see Figure 2). Even as it’s unlikely that ITER will ever become operational, Commonwealth Fusion Systems claims that their ARC tokamak reactor, a follow-on facility that will be much larger than SPARC and is supposed to be operating by the mid-2030s, will contain a tritium-breeding blanket enclosing most of the plasma vessel.

Gearing up for deuterium-based reactors

No fusion power reactor design appears to be viable from the viewpoint of fuel supply unless it is based on the DD cycle, in which all the tritium is generated and burned within the fusioning plasma, so that no tritium need be purchased or bred in an enveloping blanket. The drawback of DD reactions, however, is that their reactivity is 100 times smaller than that of DT reactions at the same temperature. The ratio is less forbidding in practice, as the triton and helium 3 reaction products (see Figure 1) can also be burned up, so that the total energy yield is 2.5 times that of the DT reaction, and the non-neutron yield which is used to maintain the plasma temperature is eight times that of the DT case. Nevertheless, the operating temperature must be at least a factor of three higher, and the product of density and confinement time must be 10 times larger than for DT.

The fuel issue is reminiscent of the early stage of the hydrogen bomb development—from 1947 to 1951—when weaponeers faced a similar quandary of how to provide the large amounts of tritium that seemed necessary to start and sustain a thermonuclear burn in deuterium alone. The impasse was resolved by the Teller-Ulam configuration invented in 1951, in which extreme compression of the fuel resulted in a much higher reaction rate so that deuterium could be heated to ignition before the package disassembled. This technique enabled the thermonuclear stage of nuclear weapons to be fueled either with deuterium alone as in the 1952 Ivy Mike nuclear test, or with deuterium and lithium as in the 1954 Operation Castle test series. As illustrated in Figure 1, DD reactions produce both tritium and neutrons within a superdense plasma that holds lithium and thermalizes neutrons. The tritium reacts with the deuterium and the neutrons react with lithium in the dense plasma to produce even more tritium. All the tritium is generated and burned in situ.

In the Teller-Ulam configuration, the required extreme fuel compression is attained by ablation using the X-ray output of a fission explosive, which is basically the same technique used in the successful NIF experiments to ignite DT (Zylstra et al 2022). In the latter case, the X-rays are produced by laser beams heating a container closely surrounding the fuel capsule. To ignite a capsule containing predominantly deuterium fuel with a small DT core “hot spot,” the highest possible density can be attained by effecting much greater pressure during the implosion. This condition demands a driver beam energy of 50 to 100 megajoules, at least a factor of 10 higher than for DT alone. As noted, much higher fuel density results in greater reaction rates for both DT and DD that propel the temperature to the three times higher value needed to ignite and maintain DD burn before disassembly of the capsule becomes significant. The reaction catalyst, tritium, is concurrently generated and burned within the fuel capsule, according to the formulas shown in Figure 1.

Turning to magnetic confinement systems, such as the tokamak or stellarator, radiation losses and severe limits on plasma pressure will likely preclude MCF devices from reaching the high temperature and density conditions for deuterium-based burning. In fact, MCF systems are a long way from showing scientific feasibility even in DT, defined as the ability to reach thermonuclear ignition, or at least to produce 10 times as much fusion energy as the energy needed to heat the plasma— ITER’s principal operational objective. There are plausible solutions to the remaining issues that challenge scientific feasibility in DT, but success is still far away and highly uncertain. In contrast, the scientific feasibility of ICF was first demonstrated when the NIF reached ignition conditions two years ago (Kritcher et al 2024).

Deuterium-based reactors will afford truly infinite fuel supply. To make such reactors viable, fusion programs worldwide must embark on many decades of research and experimentation. For example, the required 100-megajoule driver for ICF is achievable with ion beams, but not with today’s nanosecond pulse lasers. A lengthy research and development period would not delay the advent of practical fusion power. As noted, only one fusion concept has demonstrated scientific feasibility in DT, namely ICF with X-ray-induced implosion, and the host of technologies required to transform this technique into a viable power generator even with DT will require a half-century to invent and develop.

In a sense, the ICF approach is unpalatable, because the implosion technique is closely related to that employed in nuclear weapons. But if one wants a fusion power reactor with an assured fuel supply, there may be no choice for the indefinite future!

Conclusions

Deuterium-tritium reactors require substantial tritium fuel for startup, and the global stockpile for civilian use will be only a few tens of kilograms for the indefinite future. Nevertheless, concern about adequate tritium resources is misplaced as the perpetually delayed ITER is unlikely to operate, while proposed demonstration reactors and pilot plants will be unbuildable for many decades because of unresolved issues of plasma performance, non-existent technologies, manufacturing hurdles, and insupportable cost.

A genuine supply crunch may occur half a century from now. If eventually feasible, DT reactors must be self-sustaining in tritium after startup, because providing adequate makeup fuel from fission reactors will be impossible. However, tritium fuel self-sufficiency may be unattainable because of shortcomings in the internal breeding system and irreplaceable loss of unburned tritium in reactor subsystems. Loss of tritium may be crippling for magnetic confinement devices where minuscule tritium burnup requires at least two dozen cycles of fuel reprocessing and re-injection. The old concept of doubling time that would permit startup of additional reactors is unworkable in practice.

In the long run, the only viable fuel solution for both starting and sustaining operation are deuterium-based reactors where the tritium is concurrently generated and burned in the fusioning plasma, and no external supply or internal breeding is required. Deuterium-based reactors should be feasible with inertial confinement fusion systems by going to supersized fuel capsules and laser or particle beams delivering 50 to 100 megajoules.