← carlofanara.meDataphysics ↗
FUSION, PART I — RATES AND BARRIERS

Inside the Gamow Window

D-T fuses easily and dirtily; p-¹¹B fuses cleanly and more reluctantly. The gap between them ultimately gets to the number of fusion events you can get. Here's where that number comes from, and what it costs in real power.

Follows on from "Two Ways to Split the Atom's Ledger" Part 1 of the fusion series

Two nuclei on a collision course are trying to do something the electric force forbids: get inside each other's Coulomb repulsion, close enough (a few femtometres) for the strong force to take over. Classically, unless they're aimed at each other with more energy than the barrier height, they never touch. Quantum mechanically, they don't need to clear the barrier — they can tunnel through it, at a probability that falls off sharply with distance and charge. Nearly everything that distinguishes "easy" fusion fuels from "hard" ones lives in that single tunneling probability.

This piece starts from the two curves that make D-T the default fuel, the cross-section and the reactivity, and then works underneath them, through the chain in the order it actually happens: how likely a pair of nuclei is to have enough energy in the first place (Maxwell-Boltzmann); how likely they are, given that energy, to tunnel through the barrier (the Gamow factor); where those two probabilities multiply out to a peak — the Gamow window — and how wide it is; and then what comes out the other side: a reaction rate, a power density, and a blunt numerical answer to "how much harder is p-¹¹B, really?"

1. Cross-sections and reactivity: why D-T gets picked first

Two plots do the real talking here. The first is the fusion cross-section, σ, as a function of the relative collision energy: how likely two nuclei are to fuse when they collide at a given energy. This is measured in units of area (imagine it as the surface area of a target as seen from an impinging particle: the larger it is, the higher the impact probability). The second is the reactivity, ⟨σv⟩, which averages that cross-section over a Maxwellian plasma at temperature T (how many of the particles have a given velocity v – or energy E). This is the number that actually goes into a reactor's power-density calculation.

Fusion cross-section versus center-of-mass energy for D-T, D-D, D-3He, p-11B, T-T, T-3He, and 3He-3He reactions.
Fig. 1 Fusion cross-section σ vs. collision energy, ECM, for the candidate fuel pairs.
Fusion reactivity, sigma-v, versus plasma temperature for the same set of candidate fuel pairs.
Fig. 2 Fusion reactivity ⟨σv⟩ vs. plasma temperature T (bottom axis in keV, top axis in millions of kelvin). D-T's curve sits an order of magnitude above everything else across most of the practical range — the reason it's the reference reaction for every near-term reactor concept, despite being the "dirtiest" in neutron terms.

The reactivity panel is a rough ranking of how readily each fuel reaches its maximum reactivity: D-T's reactivity climbs earliest and peaks around 60–70 keV (roughly 700–800 million kelvin), which is why it dominates every current MFE and ICF program.

D-D needs noticeably higher temperatures for the same reactivity, and its energy yield per reaction is about a quarter of D-T's. D-³He needs higher temperatures still, but rewards you with a largely aneutronic reaction and a higher energy yield. p-¹¹B sits at the hard end — its reactivity only becomes competitive well above 100 keV, and even then it needs boron-11 mixed at high concentration with protons, which is its own plasma-physics challenge. That curve is, in a real sense, a map of how much harder each "cleaner" fuel choice is to actually burn.

The whole story of fusion fuel choice is a trade: D-T is the easiest reaction to ignite and the hardest exhaust to live with; p-¹¹B is nearly the reverse.

It's worth observing how compressed that reactivity plot is — three, sometimes four orders of magnitude separate D-T from p-¹¹B at the same temperature. That gap is a large part of why most funded near-term reactor concepts, whatever their confinement scheme, still plan to start on D-T, and why the aneutronic-fuel programs (several of the "other approaches" in the introductory piece are explicitly chasing D-³He or p-¹¹B) are making a harder bet on plasma performance in exchange for a much simpler power-conversion and materials story. The rest of this piece goes underneath those curves, to ask where the gap comes from.


2. The barrier, and why it depends so hard on charge

The height of the Coulomb barrier between two nuclei of charge Z₁e and Z₂e is irrelevant on its own — what matters is the probability of tunneling through it at a given collision energy E. That probability is the Gamow factor:

Ptunnel(E)=exp(−2πη),2πη=EGE
η is the Sommerfeld parameter; EG is the Gamow energy — the single number that encodes how hard a given pair of nuclei is to fuse.

The Gamow energy itself depends on nothing but the charges and the reduced mass μ of the pair:

EG=2μc2(παZ1Z2)2
α ≈ 1/137 is the fine-structure constant. EG grows with the square of the charge product Z₁Z₂ — this is the whole story in one exponent.

That Z₁Z₂ dependence is why boron is such a hard sell as a fusion partner: swapping tritium's Z=1 for boron's Z=5 doesn't make the barrier five times higher, it makes the tunneling suppression exponent scale with (Z₁Z₂)² inside a square root — and boron-11 also happens to be roughly five times heavier than tritium, which raises the reduced mass and compounds it further.

Table 1. Gamow energy for the candidate fuel pairs, computed from charge and reduced mass alone. For D-T it reproduces, to four significant figures, the Gamow constant BG = √EG ≈ 34.38 √keV that Bosch & Hale tabulate for their reactivity formula[1].
reactionZ₁Z₂reduced mass (amu)EG (keV)
p + p (for reference — the Sun's bottleneck)10.504493
D + D11.007986
D + T11.2071,182
T + T11.5081,476
D + ³He21.2074,728
T + ³He21.5085,905
p + ¹¹B50.92322,590
³He + ³He41.50823,616

p-¹¹B's Gamow energy is nineteen times D-T's. That single ratio explains most of the claims that p-¹¹B is "harder" to ignite.

These Gamow energies hold for every pair in the table, p-¹¹B included: they are set by charge and reduced mass, not by nuclear structure. What the barrier calculation cannot see is what happens once the nuclei are through it. For D-T that hardly matters; for p-¹¹B it matters a great deal, as discussed in Section 4.


3. Two probabilities multiplied: the Gamow window

A hot plasma doesn't hand you nuclei at one fixed energy — it hands you a Maxwell-Boltzmann spread, with the population falling off as exp(−E/kT) (weighted by a phase-space factor of √E) as you go to higher energy. Raising E makes tunneling more likely, but it also makes that energy rarer. The product of a falling population and a rising tunneling probability is not monotonic: it has a maximum, the Gamow peak. Its width, the Gamow window, is the narrow band of collision energies where fusion actually happens, sitting well above the thermal average kT, on the joint tail of both curves.

Two stacked panels, D-T at 20 keV above and p-11B at 300 keV below, showing the Maxwell-Boltzmann tail, the tunneling factor and their product, the Gamow peak. The caption gives the details.
Fig. 3 Each curve normalized to its own peak, so both can share one axis. The Gamow peak (filled) is where the falling Maxwell-Boltzmann population meets the rising tunneling probability. Note p-¹¹B's window sits at ~800 keV even though its plasma is "only" at 300 keV — sixteen times further out in energy than D-T's window, although, measured against its own thermal energy (E₀/kT ≈ 2.7 against 2.5), it sits at almost the same place on the tail. For p-¹¹B this is the barrier-only window: it assumes a smooth S-factor, while the real reaction energies are pulled towards the ¹²C resonances near 150 keV and 620 keV (see "Where the resonances actually sit").

The peak location and width both fall out of finding where the exponent −E/kT − √(EG/E) is stationary:

E0=(EG(kT)24)1/3Δ=43E0kT
E₀ is the Gamow peak energy; Δ is its 1/e width. Both depend only on EG and the plasma temperature — no nuclear physics yet.

Note: the curves in Fig. 3 correspond to the second (D-T at 20 keV) and fifth (p-¹¹B at 300 keV) rows of Table 2 below.

Table 2. Where the action happens, and how rare it is. The last column, exp(−3E₀/kT), is the exponential penalty relative to a hypothetical reaction with no barrier at all — it's the price of admission. The p-¹¹B rows are barrier-only estimates for a smooth S-factor; Section 4 shows what the p-¹¹B resonances change.
reactionkT (keV)E₀ (keV)E₀ / kTΔ (keV)exp(−3E₀/kT)
D-T10313.1419.4 × 10⁻⁵
D-T20492.5726.4 × 10⁻⁴
D-T50901.81554.4 × 10⁻³
p-¹¹B1003843.84521.0 × 10⁻⁵
p-¹¹B3007982.71,1303.4 × 10⁻⁴
p-¹¹B6001,2672.12,0131.6 × 10⁻³

Read the last column as "how rare is a successful collision, compared to just having thermal energy kT at all." At D-T's working point (kT ≈ 20 keV), only about 6 in 10,000 of the joint energy-and-tunneling weight survives — everything a reactor does is try to make that rare event happen often enough, often enough times a second, in enough particles. p-¹¹B at kT = 300 keV is actually paying a similar exponential penalty to D-T at 20 keV — the point isn't that its window is astronomically rarer per collision: it is that reaching a window that shallow requires fifteen times the plasma temperature, which is where the real engineering cost sits (radiation losses, confinement, materials).


4. From tunneling to an actual reaction: the two-step rate

Tunneling through the barrier isn't the same as reacting. Getting close is necessary, not sufficient — what happens once the nuclei are within range of the strong force is nuclear physics: resonances, spin and parity matching, the structure of the compound nucleus. The standard way to keep these two contributions visible is to split the cross-section in two:

σ(E)=S(E)E×exp(−2πη)
The astrophysical S-factor S(E) carries everything except barrier penetration — it's the "given that they touched, did they react" term, and it varies slowly with energy except where nuclear resonances sit.

So the three factors that count are:

How many particles even have enough energy to try

Set by the Maxwell-Boltzmann distribution at temperature T — mostly a bookkeeping question, no nuclear physics involved.

Of those, how many actually tunnel through

The Gamow factor exp(−2πη), fixed entirely by charge and mass (the Sommerfeld factor 2πη is defined in Section 2). This is what makes the Gamow window narrow and pushes it out onto the thermal tail.

Of those that tunnel, how many react

The nuclear S-factor S(E) — resonances, angular momentum matching, whether the compound nucleus likes to exist. This is why D-T (with a broad resonance sitting almost exactly at its Gamow peak) so dramatically outperforms a naive charge-and-mass estimate, and why p-¹¹B has resonances of its own near 150 keV and 600 keV that a smooth Gamow-window picture alone won't capture.

Folding the Maxwellian average through σ(E) and evaluating the integral by expanding around the Gamow peak (steps 1 and 2, with S(E) treated as locally constant — step 3, held fixed at its value near E₀) gives a closed-form reactivity:

⟨σv⟩≈2μS(E0)Δ(kT)−3/2exp(−3E0kT)
Self-derived from the steepest-descent (saddle-point) approximation to the Maxwellian-averaged cross-section integral. Validated below against the exact D-T reference formula.

Checking this against the exact, universally-used D-T reference fit (Bosch & Hale, Nuclear Fusion 32, 1992)[1] — by backing out the effective S(E₀) the formula needs to match the real curve at kT = 20 keV — recovers S(E₀) ≈ 21 MeV·barn, inside the literature range quoted for D-T near its resonance. The approximation tracks the exact curve to within ~ 10% between 5 and 20 keV (the range that matters for any real reactor concept) and drifts to roughly 2–4× too high above 50 keV, where the resonance's energy dependence stops looking "locally constant." That's how the smooth-S(E) picture breaks down: not because the Gamow-window location is wrong, but because the underlying nuclear physics isn't flat. (The saddle-point derivation itself follows the standard treatment in Iliadis's Nuclear Physics of Stars[3].)

Log-log plot of reactivity against temperature: solid reference curves for D-T and p-11B, dashed closed-form approximations. A dotted curve shows the 2026 p-11B evaluation. The D-T pair agree below about 20 keV; the p-11B approximation sits well below its reference curve.
Fig. 4 The closed form above (dashed) against reference fits (solid), each fit drawn only over its quoted range of validity. For D-T, with S₀ fitted at 20 keV, the closed form tracks the Bosch & Hale curve[1] to about 10% from 5 to 20 keV and overshoots above 50 keV. For p-¹¹B, a single constant S₀ (pinned near 600 keV) falls short of the Nevins & Swain fit[2] by a factor of 3.7 at 300 keV and about 11 at 100 keV, because its resonances make S(E) anything but constant. The dotted curve is the newest cross-section evaluation (Wang et al. 2026[7]) integrated numerically here: it matches Nevins & Swain to within 2% up to 150 keV and runs 7% higher at 300 keV and 26% higher at 500 keV. The other fuels are compared in Fig. 2.
What's exact here and what's illustrative. The Gamow energies in Section 2 are exact for every pair: they follow from charge and reduced mass alone. The peak locations, widths and suppression factors in Section 3 are exact for a smooth S-factor, which describes D-T well and p-¹¹B poorly. The solid curves in Fig. 4 are published reference fits and the dotted one the newest p-¹¹B cross-section integrated here; the dashed ones are the constant-S₀ closed form, shown to make its limits visible. The power figures in Section 5 use the reference fits.

Where the resonances actually sit

The constant-S₀ picture above is a convenient fiction — both channels have real nuclear structure riding on top of the smooth Gamow-window curve. For D-T, the reaction proceeds through a single, broad resonance in the compound nucleus ⁵He: a 3/2⁺ state around Ecm ≈ 64 keV above threshold (a deuteron lab energy near 107 keV), with a width on the order of a few hundred keV to ~1 MeV. That width is comparable to or broader than the Gamow window itself at reactor-relevant temperatures — which is exactly why D-T's reactivity curve looks smooth rather than sharply peaked, and why the resonance sits so conveniently close to temperatures a plasma can actually reach.

For p-¹¹B the landscape is very different. The compound nucleus ¹²C, formed at high excitation, breaks up into three alpha particles, mostly by way of ⁸Be. Over the energies that matter, there is a narrow resonance near Ecm ≈ 148 keV, a few keV wide, then a strong, broad one around Ecm ≈ 600–650 keV[2], and further broad states exist above 1 MeV. At p-¹¹B temperatures the Gamow window is hundreds of keV wide (Table 2), so it spans this structure instead of sitting on one smooth peak, and S(E) changes several-fold across it. That is why the constant-S₀ estimate misses by a factor of 3.7 at 300 keV in Fig. 4, and why a trustworthy p-¹¹B reactivity has to come from the measured, resonance-resolved cross-section integrated numerically (as in the Nevins & Swain fit[2]). That cross-section has since been re-measured and re-evaluated[4–7]. Integrating two of the newer evaluations here, the reactivity at 300 keV comes out 7% (Wang et al. 2026[7]) to 30% (Sikora & Weller 2016[4]) above the Nevins & Swain fit, and 26–50% above it at 500 keV, in line with Tentori & Belloni[6]. Most of that spread comes from how counted alpha particles are turned into a cross-section: Sikora & Weller completed partly measured alpha spectra with a model of the three-alpha break-up, while the newer measurements behind Wang et al. record the whole spectrum directly. The resonance newly observed near 4.7 MeV[7] changes the reactivity by less than 1% below 600 keV. These corrections are well within a factor of two and do not change the picture, so they are only quoted (Table 3); the full treatment would belong to a dedicated p-¹¹B piece.


5. From rate to power — no reactors yet

Nothing above has mentioned magnets, lasers or containment. Given only a fuel density and a temperature, the reactivity alone fixes a volumetric reaction rate and a power density:

R=n1n2⟨σv⟩P=RQ
R: reactions per cubic metre per second. Q: energy released per reaction (17.6 MeV for D-T, 8.7 MeV for p-¹¹B). P: fusion power density.

Taking the same hypothetical fuel density for both — n₁ = n₂ = 5 × 10¹⁹ m⁻³ (total ion density 10²⁰ m⁻³, typical in the range plasma physicists work with) — and comparing D-T at a temperature where it's already well past ignition-relevant (20 keV) against p-¹¹B at the two temperatures people actually propose for it, we get the reactivities and power densities of Table 3.

Table 3. Purely from cross-section physics — same assumed density, no confinement scheme. ⟨σv⟩ from the Bosch & Hale (D-T) and Nevins & Swain (p-¹¹B) fits; 600 keV is just beyond the latter's quoted range. The "newer data" column integrates the Wang et al. 2026[7] and Sikora & Weller 2016[4] cross-sections numerically (lower and upper value).
fuelT (keV)⟨σv⟩ (m³/s)power density (MW/m³)power density, newer data (MW/m³)time to find a partner
D-T204.3 × 10⁻²²3.05—~46 s
p-¹¹B3003.4 × 10⁻²²1.181.26–1.54~59 s
p-¹¹B6003.7 × 10⁻²²1.281.78–2.07~54 s

At the same density, D-T at 20 keV beats p-¹¹B at 300 keV on power density by a factor of about 2.6, and going up to 600 keV, thirty times hotter than D-T needs to be, barely helps: the gap is still a factor of 2.4, because the Nevins & Swain fit levels off near 500 keV. The newer cross-section evaluations[4, 7] narrow the gap to about 2.0–2.4 at 300 keV and 1.5–1.7 at 600 keV, where their reactivity is still rising. This is the "purely physics" efficiency that the above reaction-rate machinery was built to answer: given a certain number of reactants, how many actually fuse, and how much power comes out — before a single word about magnetic fields, lasers, or wall materials. (It also understates p-¹¹B's real handicap: at 300–600 keV, bremsstrahlung radiation losses — driven by boron's higher charge — start competing with the fusion power itself, a further purely-physics complication for a future part.)

A related, dimensionally cleaner way to express the same "efficiency": the product n⟨σv⟩τ (density × reactivity × how long the fuel is held together) is the expected number of times a given particle finds a fusion partner during that time — a probability, not tied to any specific confinement scheme, of order n⟨σv⟩τ for small values. To burn 1% of the fuel at n = 10²⁰ m⁻³, D-T at 20 keV needs about 0.23 seconds of confinement; p-¹¹B at 300 keV needs about 0.30 seconds, and at 600 keV about 0.27 seconds. Same order of magnitude in time — the real difference is entirely in the temperature you have to sustain that time at.


6. Should we be using Maxwell-Boltzmann at all?

Every formula above assumed the fusing nuclei follow a classical Maxwell-Boltzmann energy distribution — an assumption worth checking. Quantum statistics (Bose-Einstein for integer-spin particles, Fermi-Dirac for half-integer spin) only diverge from the classical Maxwell-Boltzmann limit when a gas is degenerate — when quantum wavefunctions of neighboring particles start to overlap. The standard yardstick is the degeneracy parameter nλ³, where λ is the thermal de Broglie wavelength:

λ=h2πmkTdegenerate whennλ3≳1
Table 4. Degeneracy parameter for realistic fusion conditions versus a genuinely degenerate astrophysical comparison.
species / conditionn (m⁻³)Tn λ³
D ions, D-T-relevant plasma10²⁰20 keV5.3 × 10⁻²⁰
p ions, p-¹¹B-relevant plasma10²⁰300 keV2.6 × 10⁻²¹
electrons, same D-T plasma10²⁰20 keV1.2 × 10⁻¹⁴
electrons, white-dwarf-core-like10³³5 keV0.95 ← degenerate

For every fusion-relevant ion population — D-T, p-¹¹B, anything at reactor-plausible densities — nλ³ sits fifteen to twenty orders of magnitude below 1: the gas is purely classical, so Maxwell-Boltzmann is essentially exact, and Bose-Einstein or Fermi-Dirac corrections to the velocity distribution would be undetectable. Even the much lighter electrons stay non-degenerate by fourteen orders of magnitude at these temperatures. The one place where this genuinely changes is in dense, cold, electron-degenerate matter — white dwarf interiors, neutron star crusts — where degenerate electron screening changes the effective barrier that nuclei see, giving rise to pycnonuclear fusion. That's a different regime than anything a lab reactor operates in.

One more nuance worth separating from the occupation-number question: reactions between identical particles — D-D, T-T, ³He-³He — carry a distinct quantum-statistics effect that has nothing to do with degeneracy. Deuterons are bosons (integer spin), while tritons and ³He nuclei are fermions (half-integer spin), and the symmetry of the two-particle wavefunction restricts which relative angular momenta are allowed to contribute. This shapes the fine structure of σ(E) for same-species reactions, but it's a statement about indistinguishability in the scattering channel, not about whether the surrounding plasma has a classical or quantum velocity distribution.

The answer to "does quantum statistics matter here" isn't about which particles are involved — it's about how close together they are relative to their own thermal wavelength. For any fusion plasma anyone has built or seriously proposed, that answer is no, by a very wide margin.

Put together: most of the D-T/p-¹¹B gap traces back to one exponent, (Z₁Z₂)², buried inside a square root inside another exponential. Everything downstream — the Gamow window sitting at hundreds of keV instead of tens, the reactivity needing a much hotter plasma to reach a comparable rate, the power density that still trails even when you pay that temperature cost — is arithmetic on top of that one number, with p-¹¹B's resonances clawing a little of it back. And yet p-¹¹B is exactly what several serious efforts — companies among them — are betting on, precisely because the payoff (direct charged-particle energy conversion, no tritium breeding, no 14 MeV neutron wall damage) is large enough to be worth the higher energy cost, if the plasma physics can be made to cooperate. The newest nuclear data already tilt that balance a little: measured with current techniques, p-¹¹B reacts 7–30% more readily at 300 keV, and up to 50% more at 500 keV, than the long-standing reference assumes (Table 3). Analyses built on those data, which follow the fast alpha particles as they heat the fuel and let the electrons run cooler than the ions, find temperature windows where fusion power outruns the bremsstrahlung losses[5–7]. On current data, then, p-¹¹B is likely to fare better than the classic estimates given here suggest. Which raises the question for what comes next: if the smooth, thermal, Maxwell-Boltzmann picture in this piece makes p-¹¹B look this expensive, what happens once you stop assuming thermal equilibrium — non-thermal, beam-driven ion populations, or the extreme transient densities of a compressed pellet or pinch? That's Fusion, Part II.

Dataphysics Limited works on energy-systems modelling and physics-informed AI. → dataphysics.ae

§