Deuterium–Tritium Fuel Cycle

The only known way to run a fusion power plant on its own fuel — but the tritium that feeds it must be manufactured inside the reactor wall one atom at a time.

Deuterium–tritium fusion releases 17.6 MeV per reaction: 3.5 MeV stays with the helium-4 ash (which heats the plasma) and 14.1 MeV leaves with a fast neutron. A tritium-breeding blanket surrounds the plasma, intercepts those neutrons, and uses them to transmute lithium into fresh tritium via two reactions: ⁶Li + n → ⁴He + T + 4.8 MeV (exothermic) and ⁷Li + n → ⁴He + T + n′ − 2.5 MeV (endothermic). The tritium is extracted from the blanket, purified in an isotope-separation system, and fed back as fuel — alongside deuterium, which is straightforwardly distilled from seawater. Helium ash is pumped out through the divertor. The entire gas stream is processed in a Tritium Plant, which separates hydrogen isotopes by cryogenic distillation and stores them for reinjection.

The plasma itself holds only a fraction of a gram of fuel at any moment, and the effective burn fraction per pass is roughly 1 %. Unconsumed D and T exhaust from the divertor region via cryopumps, travel to the Tokamak Exhaust Processing subsystem, which strips impurities. Isotope separation then divides deuterium from tritium using cryogenic distillation columns (exploiting the small mass difference). Bred tritium is extracted from the blanket material — either dissolved out of a liquid lithium-lead alloy or thermally released from ceramic pellets (e.g. lithium titanate Li₂TiO₃) by helium purge gas — and fed into the same isotope-separation stream. The entire inventory must be tracked because tritium decays with a 12.32-year half-life and is a regulatory radiological material. A Tritium Breeding Ratio (TBR) greater than 1 — slightly more tritium produced than consumed — is required for fuel self-sufficiency; design targets are typically TBR ≈ 1.05–1.15 to provide a safety margin and cover losses.

In plain terms

Imagine a campfire that, instead of consuming wood you haul in from outside, somehow grows its own logs from the ash and smoke left behind. A D-T fuel cycle does something similar: the energy-releasing reaction spits out a neutron that flies into the reactor wall, hits a lithium atom, and makes a brand new tritium atom — the very fuel the next reaction will need.

Also called: D-T fuel cycle, tritium breeding blanket cycle, closed fusion fuel loop.

Why it matters

Tritium does not exist in useful quantities in nature — the global civilian stockpile is roughly 20 kg as of the mid-2020s, of which ITER alone will consume an estimated 12.3 kg across its D-T operational lifetime.
Deuterium, by contrast, is effectively inexhaustible: every cubic metre of seawater contains 33 g, and it is extracted commercially at industrial scale today.
Closing the fuel cycle is the key technical step separating a fusion experiment (which imports tritium) from a fusion power plant (which is tritium self-sufficient). No device has yet demonstrated TBR > 1 at reactor scale; it remains a central challenge for DEMO.
The lithium blanket simultaneously breeds fuel, absorbs ~80 % of fusion power as heat for electricity generation, and shields the superconducting magnets from fast neutrons — three functions in one component.

Where it lives

JET (Culham, UK, decommissioned 2023) — operated closed D-T loop at engineering scale in DTE1 (1997) and DTE2/DTE3 (2021–2024); tritium purified to >99.5 % purity over eight processing cycles
ITER (Cadarache, France, under construction) — will host Test Blanket Modules to demonstrate tritium breeding for the first time in a real fusion neutron environment
TFTR (Princeton, USA, decommissioned 1997) — first tokamak to use significant tritium quantities, beginning D-T operation in 1993
Darlington Tritium Recovery Facility (Ontario, Canada) — CANDU reactor by-product tritium; primary external supplier for ITER's fuel campaign

Key numbers

D-T reaction energy: 17.6 MeV per fusion event (3.5 MeV to alpha, 14.1 MeV to neutron)
Tritium half-life: 12.32 years (Confirmed by NIST (Lucas & Unterweger 2000); 5.5 % annual decay loss)
Tritium burn fraction (plasma): ~1 % (Per ITER design; rest exhausted and recycled via Tritium Plant)
JET DTE3 record (2024): 69.26 MJ in 6 s (EUROfusion, February 2024; 0.21 mg of D-T fuel)
Global civilian tritium stockpile (mid-2020s): ~20 kg (ITER Organization fuelling page; UKAEA supply studies)
ITER tritium consumption (lifetime): ~12.3 kg (UKAEA/IAEA supply studies; supplied by Ontario CANDU reactors)
Required Tritium Breeding Ratio: >1 (target ~1.05–1.15) (Margin covers decay losses, inventory hold-up, and extraction inefficiency)
⁶Li breeding reaction energy: +4.78 MeV (Exothermic; ⁷Li reaction is −2.47 MeV (endothermic))

Surprising facts

JET's final D-T campaign (DTE3, 2023–2024) released 69.26 MJ of fusion energy in a single 6-second pulse from just 0.21 mg of fuel — that 0.21 mg of D-T is energetically equivalent to burning about 2 kg of coal.
Tritium's half-life is only 12.32 years, meaning any stockpile loses half its inventory in about a dozen years whether you use it or not — making tritium supply and cycle efficiency a genuine economic constraint, not just a physics curiosity.
A commercial 1 GW fusion power plant would need roughly 55 kg of tritium per year. The entire current global civilian stockpile is ~20 kg. Without breeding blankets operating at TBR > 1, fusion power at scale is physically impossible.
The first evidence of alpha-particle heating — fusion products warming the plasma from the inside, a defining feature of a burning plasma — was unambiguously demonstrated at JET in 1997 using its D-T fuel mix, decades before ITER will attempt it.

Open questions

Anomalous tritium retention in plasma-facing walls: During JET's 1997 DTE1 campaign, about 30 % of the injected tritium became trapped in the carbon-fibre-composite tiles and other plasma-facing surfaces — more than three times the predicted level. Even after extended deuterium cleaning, ~6 g (17 % of the total) remained unrecovered. The trapping mechanisms — co-deposition of carbon-tritium layers in shadowed crevices, implantation in sub-surface layers — are qualitatively understood, but predictive models for reactor-relevant tungsten walls under burning-plasma neutron bombardment remain uncertain. Tritium accountancy in ITER, where every gram matters, will depend on closing this modelling gap.

A short history

Deuterium was isolated by Harold Urey in 1932; tritium was first made by Oliphant, Harteck, and Rutherford bombarding deuterium with deuterons in 1934. The D-T fusion cross-section's large peak near 100 keV was characterised in the late 1940s and led to the thermonuclear weapons programme. Controlled D-T experiments in tokamaks began at TFTR (Princeton) in 1993 and at JET (UK) in the Preliminary Tritium Experiment in 1991, with the full DTE1 campaign in 1997 producing 21.7 MJ in a single pulse and 16 MW of peak fusion power. JET's DTE2 (2021–2022) raised the energy record to 59 MJ, and DTE3 (2023–2024) — JET's final campaign before decommissioning in December 2023 — pushed it to 69.26 MJ. The concept of a closed tritium breeding blanket was formalised in the ITER design during the 1980s; ITER's Test Blanket Module programme will be the first attempt to demonstrate TBR > 1 in a real burning-plasma neutron environment.

Sources

ITER Organization — Fuelling (tritium supply and fuel cycle description) (https://www.iter.org/machine/supporting-systems/fuelling)
ITER Organization — Tritium Breeding (Test Blanket Module programme) (https://www.iter.org/machine/supporting-systems/tritium-breeding)
EUROfusion — Breaking New Ground: JET's Latest Fusion Energy Record (DTE3, February 2024) (https://euro-fusion.org/eurofusion-news/dte3record/)
JET DTE1 results — Fusion Engineering and Design 46 (1999), EUROfusion preprint JETP98070 (https://scipub.euro-fusion.org/wp-content/uploads/2014/11/JETP98070.pdf)
Abdou et al. — PPPL/SULI lecture: Introduction to fusion fuel cycles and tritium breeding blankets (2023) (https://suli.pppl.gov/2023/course/SULI%202023%20presentation.pdf)
UKAEA — Tritium resources available for fusion reactors in the long term (CCFE-PR1767) (https://scientific-publications.ukaea.uk/wp-content/uploads/CCFE-PR1767-1.pdf)