Nuclear fusion: physics, approaches, engineering

Fusion is the joining of light nuclei into a heavier nucleus, releasing energy. Like fission, the energy comes from the binding-energy curve — but from the light end, where lighter nuclei have less binding energy per nucleon than heavier ones (up to iron). The Sun's core fuses hydrogen to helium through the proton-proton chain.

The terrestrial reaction of greatest engineering interest is deuterium-tritium (D-T) fusion:

$\text{D} + \text{T} \rightarrow \text{He-4} + n + 17.6\, \text{MeV}.$

The products are an alpha particle (He-4, 3.5 MeV) and a fast neutron (14.1 MeV). The He-4 carries 20% of the energy and stays in the plasma (heating it); the neutron carries 80% and escapes the plasma, depositing its energy in surrounding structure as heat for power generation.

Why D-T rather than D-D or other reactions: D-T has the highest reaction cross-section at the lowest temperatures of any fusion reaction. D-D fusion is possible but requires higher temperatures (closer to 500 million K vs ~100 million K for D-T). At any temperature achievable in current systems, D-T is the easiest reaction.

The trade-off: D-T requires tritium, which does not occur naturally in usable quantities (tritium has a 12.3-year half-life). Tritium must be bred from lithium-6 in a blanket surrounding the plasma, using the 14.1 MeV neutron from the fusion reaction itself. The breeding loop is a critical part of any commercial D-T fusion design.

To get net energy out of a fusion reactor, four conditions have to be met simultaneously.

1. High temperature — about $10^8$ K (100 million K) for D-T. At this temperature the fuel is fully ionized (a plasma). The energy comes from collisions: at lower temperatures, electrostatic repulsion between positive nuclei prevents them from getting close enough to fuse.
2. Sufficient density — enough nuclei in a given volume that collisions actually happen. Often quoted as $n$, particles per cubic meter.
3. Sufficient confinement time — the plasma must stay hot and dense long enough for enough fusion reactions to occur. Quoted as $\tau$, the energy confinement time.
4. Energy gain — fusion power produced must exceed the energy input needed to sustain the plasma.

The Lawson criterion (J.D. Lawson, 1955) captures the first three:

$n \cdot \tau \cdot T \gtrsim 3 \times 10^{21}\, \text{keV}\cdot\text{s/m}^3 \quad \text{(D-T, near-ignition)}.$

Approaches that achieve fusion divide into two broad camps based on which factor they push hardest.

• Magnetic confinement holds modest-density plasma ($n \sim 10^{20}/\text{m}^3$) for long times ($\tau \sim$ seconds).
• Inertial confinement compresses fuel to very high density ($n \sim 10^{30}/\text{m}^3$) for a vanishingly short time ($\tau \sim 10^{-10}$ s) before disassembly.

The two strategies face different engineering problems but address the same Lawson-criterion inequality from opposite directions.

Q is the ratio of fusion power produced to external power injected into the plasma:

$Q = \frac{P_{\text{fusion}}}{P_{\text{input}}}.$

Milestones along the Q ladder:

• Q = 0.7 (JET, 1997, transient): the prior world record for tokamak D-T fusion, held for two decades.
• Q = 1: 'scientific breakeven' — the fusion power equals the heating input. The reactor produces as much energy as it took to heat the plasma.
• Q ≈ 5–10: a sustained operating regime where most heating is supplied internally by alpha particles and only a small external input is needed.
• Q → ∞ ('ignition'): the plasma is self-sustaining; the alpha-particle heating is sufficient and no external heating is required.
• Engineering Q ($Q_{\text{eng}}$): a more stringent metric — ratio of net electrical output to total electrical input including all auxiliary systems. $Q_{\text{eng}} > 1$ is the threshold for actually producing net electricity. Plasma physics $Q$ values must reach roughly 10–20 to produce $Q_{\text{eng}} > 1$ after accounting for inefficiencies in heating, magnets, conversion, and auxiliaries.

More granular notation distinguishes:

• $Q_{\text{plasma}}$ — fusion power vs power absorbed by the plasma.
• $Q_{\text{sci}}$ — fusion power vs power into the experiment (more conservative, includes inefficiencies in the heating system).
• $Q_{\text{eng}}$ — net electricity vs gross plant input.

Public headlines often quote the most favorable variant. Reading fusion claims requires asking which Q is being reported. National Ignition Facility (NIF) achieved $Q_{\text{sci}} > 1$ in December 2022 (3.15 MJ fusion output vs 2.05 MJ of laser energy at the target) — a milestone, but with $Q_{\text{eng}}$ much less than 1 because the laser system itself consumed ~300 MJ to deliver the 2.05 MJ of laser energy.

Magnetic confinement fusion (MCF) uses strong magnetic fields to hold the plasma. Charged particles spiral along magnetic field lines instead of crossing them, so a properly-shaped field traps the plasma far from the chamber walls.

Two main MCF architectures.

Tokamak. A torus (donut) with a strong toroidal magnetic field plus a poloidal field induced by a large plasma current. The combination creates twisted helical field lines that confine plasma. Dominant approach since the 1960s.

• Advantages: axisymmetric (rotationally symmetric around the central axis), well-studied, smaller in coil count than stellarators.
• Disadvantages: requires plasma current that limits operation to pulses (or requires non-inductive current drive), plasma instabilities including disruptions and ELMs that can damage the wall.

The largest current tokamak is JET (decommissioned 2023); the international ITER project under construction in France is designed to demonstrate $Q \geq 10$ with 500 MW fusion power output. Private-sector tokamaks (Commonwealth Fusion Systems' SPARC with high-temperature superconducting magnets; Tokamak Energy's ST40 spherical design; others) target smaller, higher-field machines that aim for similar Q values at lower absolute scale.

Stellarator. A torus with a more complex, three-dimensional twisted shape — all the field structure is built into the external coils, with no plasma current required.

• Advantages: steady-state operation, no plasma-current disruption risk, more stable plasma.
• Disadvantages: complex coil shapes that are difficult to manufacture; smaller historical research base.

The largest current stellarator is Wendelstein 7-X in Germany; private programs (Type One Energy, Renaissance Fusion, others) pursue stellarator commercialization arguing the steady-state operation advantage outweighs the manufacturing complexity.

Inertial confinement fusion (ICF) compresses a small target containing D-T fuel to extreme density and temperature using a sudden energy pulse. The fuel reaches fusion conditions before its own inertia lets it disassemble — hence 'inertial' confinement.

The driver supplies the compression energy. Options include:

• Lasers (the largest ICF facilities). NIF (US) uses 192 laser beams converging on a target chamber to deliver ~2 MJ in nanoseconds.
• Heavy-ion accelerators (research programs).
• Z-pinch (the Sandia Z machine uses electromagnetic pulses).

The target itself is typically a millimeter-scale capsule of D-T frozen at cryogenic temperatures. The driver compresses the capsule to roughly 1000× normal density, with central temperature reaching 100+ million K. Fusion ignites a small volume; a 'burn wave' propagates outward through the compressed fuel.

NIF's December 2022 result demonstrated $Q_{\text{sci}} > 1$ for the first time in any fusion approach. The result was scientifically important — confirming that the physical principles of ICF ignition work — and operationally distant from a power plant (the laser system efficiency, pulse rate, and target factory cost would all need to change by orders of magnitude for commercial operation).

ICF for energy faces challenges:

• Pulse rate. Power plants need many shots per second; current ICF systems do a handful per day at best.
• Driver efficiency. NIF lasers are <1% efficient electrically; commercial ICF would need 10–20% efficient drivers.
• Target factory. Each shot consumes one cryogenic D-T capsule that must be manufactured to nanometer-precise specs. Commercial scale requires factory production at very low cost.
• Wall damage and tritium handling. Same as MCF for the latter; ICF-specific issues for the former (the imploding capsule emits neutrons and X-rays that damage the chamber).

Beyond pure MCF and ICF, hybrid approaches occupy the middle ground.

Magnetized target fusion (MTF) / magneto-inertial fusion. Use a moderate magnetic field to confine a moderate-density plasma, then compress mechanically or electromagnetically to fusion conditions. Lower required density than pure ICF, longer required confinement time than pure MCF, with potentially lower engineering demands than either extreme.

Notable programs:

• Helion (US). Uses pulsed-power compression of field-reversed-configuration (FRC) plasmas; targets D-He-3 fusion (different reaction with no neutron output, simpler engineering, but requires He-3 fuel which is rare and would have to be bred separately).
• General Fusion (Canada). Mechanical piston compression of a plasma cavity inside a swirling liquid metal blanket.
• TAE Technologies (US). Beam-driven FRC sustainment aimed at hydrogen-boron fusion (aneutronic, with even simpler engineering if achieved, but vastly harder physics).

Aneutronic fusion. D-He-3 and p-B-11 reactions produce charged particles rather than neutrons. The fusion products can in principle be converted directly to electricity without a thermal cycle, with corresponding efficiency advantages. The trade-off is that these reactions require substantially higher temperatures than D-T, and the engineering and physics demonstrations are at much earlier stages.

The overall structural picture: fusion's path to commercial energy has three main approach families (MCF, ICF, hybrid/MTF), each with one or more public-sector flagship programs (ITER, NIF, etc.) plus a substantial private-sector cohort that has emerged since ~2015. The fundamental physics has been demonstrated in multiple settings; the engineering of materials, tritium handling, magnet systems, target factories, and economics is the remaining work.

Even with the plasma physics solved, several engineering problems remain for any commercial fusion plant.

Tritium handling and breeding. Tritium does not occur naturally in usable quantities. Each fusion neutron leaving the plasma must, on average, produce one tritium nucleus from lithium in a surrounding breeder blanket:

$\text{Li-6} + n \rightarrow \text{He-4} + \text{T} + 4.8\, \text{MeV}.$

The breeding ratio must exceed 1 (to compensate for losses to absorption, leakage, and radioactive decay). Designing a blanket with adequate tritium breeding while also extracting heat is a major engineering task. Tritium itself is radioactive (low-energy beta) and difficult to contain (it diffuses through most metals as hydrogen does).

Neutron damage to structural materials. 14.1 MeV fusion neutrons displace atoms in the structural materials surrounding the plasma. Cumulative damage measured in displacements per atom (DPA) degrades steel's mechanical properties; the inner wall ('first wall') of a commercial fusion reactor may need replacement every few years. Developing radiation-resistant materials (reduced-activation ferritic-martensitic steels, vanadium alloys, silicon carbide composites) is an active research area.

Heat extraction at extreme flux. Power densities at the first wall can reach 10+ MW/m² locally — comparable to a rocket nozzle. Cooling systems (helium, water, molten salt, molten lithium) must remove this heat reliably for decades.

Plasma exhaust and divertor design. Fusion ash (helium) and impurities must be removed continuously from the plasma without quenching the fusion conditions. The divertor handles this exhaust load — currently the most heat-flux-stressed component of any tokamak.

Capital cost and learning rate. Even if the engineering succeeds, the levelized cost of fusion electricity will be high relative to incumbent low-carbon options for many years. The commercial threshold depends on the learning rate as plants are built — whether costs fall with cumulative deployment in the way solar PV and batteries have.

Three structural points for the cursus.

• Fusion physics is well-understood. The Lawson criterion, the D-T cross-section, the energy-gain Q metric are all established. The historical question 'can fusion work in principle' has been answered yes, multiple times, in multiple architectures.
• Engineering is the remaining frontier. Tritium breeding, neutron-resistant materials, sustained operation, capital cost. None of these is a physics question; all are engineering questions whose answers will take a decade-plus per generation of facility to test.
• The Q-value being quoted matters. Q_sci > 1 is a scientific milestone; Q_eng > 1 is what makes a power plant possible; sustained operation at Q ≥ 10 is what makes a power plant economic. Public headlines often quote the easier metric.

The next lesson moves out of nuclear sources into the dominant new electricity additions of the past two decades: solar, wind, and battery storage. These are not constrained by Carnot or by tritium breeding; their constraints come from intermittency and from materials supply chains.