r-Process Nucleosynthesis

The problem the r-process solves

Stars are nuclear furnaces, but they hit a wall at iron. Fusing two iron-56 nuclei costs energy rather than releasing it, because iron sits at the peak of the binding-energy-per-nucleon curve. Charged-particle reactions — proton fusion, helium burning, carbon burning, and so on up the chain — can build everything from hydrogen to iron. They cannot build copper, silver, gold, or uranium. The Coulomb barrier between two highly charged nuclei climbs faster than stellar temperatures can keep up, and the binding-energy curve is going the wrong way regardless. To make elements heavier than iron, you have to give up on charged-particle fusion and start dropping in neutral particles instead.

Neutrons fit the bill perfectly. They are uncharged so no Coulomb barrier resists them, and they are abundant in the interiors of evolved stars. A nucleus (Z, A) that captures a free neutron becomes (Z, A+1). The new isotope is heavier but has the same number of protons — it is still the same element, just a different isotope. If the new isotope is stable, you stop there until you capture another neutron. If it is unstable to β⁻ decay (a neutron inside the nucleus turning into a proton + electron + antineutrino), then sooner or later it decays, the proton number Z increases by one, and you are now on a heavier element.

The whole architecture of heavy-element nucleosynthesis comes down to a competition between two timescales: how fast neutrons are captured, and how fast the resulting unstable isotopes β-decay. If captures are slow relative to decays, every capture has time to decay back to stability before the next neutron arrives — you crawl up the chart of nuclides one element at a time along the valley of stability. If captures are fast relative to decays, you stack neutrons onto the same element, pushing it far out into neutron-rich territory before any decay intervenes. These two limits are the s-process and the r-process. They build almost all of the elements past iron, in roughly equal share, by sharply different routes through the same chart of nuclides.

r-process vs s-process

The two pathways were first identified by Burbidge, Burbidge, Fowler and Hoyle in their 1957 review "Synthesis of the Elements in Stars" — the famous B²FH paper that organised stellar nucleosynthesis into eight processes still used today. The distinction is clean.

	s-process	r-process
Speed of capture vs β-decay	slower	much faster
Path through chart of nuclides	along valley of stability	far on neutron-rich side
Neutron density needed	10⁷–10¹¹ cm⁻³	~10²⁴ cm⁻³
Timescale of full nucleosynthesis	10³–10⁵ yr	~1 s
Heaviest element built	²⁰⁹Bi	U, Th, transuranic
Abundance peaks (A)	88, 138, 208	80, 130, 195
Site	AGB and massive star interiors	NS mergers (+ rare supernovae)
Confirmed since	1950s (stellar spectra)	2017 (GW170817)

The two processes leave distinct fingerprints on the chart of nuclides because each magic-number bottleneck happens at a different point. In the s-process, the path runs through stable isotopes, so the abundance bump appears at the stable-isotope mass numbers for closed neutron shells (A = 88 for N = 50, A = 138 for N = 82, A = 208 for N = 126). In the r-process the path is twenty or thirty mass units off to the neutron-rich side, so the magic-number bump is at lower A by exactly that offset (A = 80 for N = 50 after β-decay back, A = 130 for N = 82, A = 195 for N = 126). The solar abundance curve shows both sets of peaks side by side — the smoking-gun observation that two distinct heavy-element production channels are at work.

The mechanism: a path through the chart of nuclides

Picture the chart of nuclides as a grid with neutron number N along the horizontal axis and proton number Z along the vertical. Each grid cell is one nuclide. Stable nuclei form a narrow valley winding upward to the right with a roughly Z ≈ 0.45 N + offset slope — light elements have N ≈ Z, heavier elements run more neutron-rich because the extra strong-force binding of neutrons partly compensates for proton-proton Coulomb repulsion.

The r-process starts with seed nuclei in the iron-peak region (say ⁵⁶Fe or ⁷⁸Ni). Free neutrons rain down at densities of ~10²⁴ cm⁻³, and each seed captures one after another:

(Z, A) + n  →  (Z, A+1)
(Z, A+1) + n  →  (Z, A+2)
(Z, A+2) + n  →  (Z, A+3)
…

Each step pushes the nucleus one cell to the right on the chart — the proton number stays the same, the neutron number climbs. After many captures the nucleus is wildly neutron-rich. At some point it reaches an isotope so weakly bound that the next neutron is essentially unbound — the so-called (n, γ) ⇌ (γ, n) equilibrium. There, the path stops capturing and waits: the nucleus β-decays, increases Z by one, and starts capturing again on the new element row. This "waiting point" alternation is what makes the r-process path look like a horizontal sweep across each row, hitting a wall, jumping diagonally up-left, and sweeping rightward again on the next row.

The defining condition is

τ_n  ≪  τ_β

where τ_n = 1 / (n_n <σv>) is the typical timescale to capture another neutron given free neutron density n_n, and τ_β is the β-decay half-life of the current isotope. β-decay half-lives along the r-process path are typically milliseconds to seconds; to be much faster than that, captures need n_n ≈ 10²⁴ cm⁻³ at temperatures ~10⁹ K, sustained for of order a second. Anywhere those conditions hold, the r-process runs.

Why magic neutron numbers make abundance peaks

Nuclei with closed neutron shells — N = 50, 82 and 126 — are anomalously tightly bound, like noble gases in atomic physics. A nucleus that has just filled a shell is much less inclined to capture another neutron, because the next one would have to go into a higher, weaker shell. The neutron separation energy S_n drops sharply on crossing a magic number.

What this means dynamically is that the r-process path, sweeping right along a row, hits a wall when N reaches a magic number. The (n, γ) ⇌ (γ, n) equilibrium turns around — photodisintegration becomes competitive with capture — and the path stops growing in N. To make further progress, the nucleus has to β-decay first, increasing Z. Each magic-N column is therefore a "waiting point" where r-process material accumulates while it waits for β-decay.

When the neutron exposure shuts off (typically because the merger ejecta has expanded enough that n_n falls), the accumulated material at each waiting point β-decays back along its row to the corresponding stable isotope at the same mass number A. Because waiting points are at magic N, and the back-decay preserves A, the stable end-products pile up at three specific values of A determined by the magic numbers. With Z(N) ≈ stable-valley Z minus β-decay path offset, the three magic-N points become the three r-process abundance peaks:

N = 50  →  A ≈ 80   (Se, Br, Kr peak)
N = 82  →  A ≈ 130  (Te, I, Xe peak)
N = 126 →  A ≈ 195  (Os, Ir, Pt peak)

These three bumps are visible in the solar-system abundance curve as a function of mass number A. They are the most direct evidence that the r-process really runs along closed-shell waiting points and is not, for instance, a smoothed-out s-process. The slight kink at A ≈ 165 (rare-earth peak) is a fine-structure feature of the path before the third major peak.

Where does the r-process actually happen?

The required neutron density n_n ~ 10²⁴ cm⁻³ is more than 10 orders of magnitude above anything in an ordinary stellar interior. For decades the candidate astrophysical sites fell into two main families.

• Core-collapse supernovae (neutrino-driven winds). When a massive star's iron core collapses to a proto-neutron star, an intense flux of neutrinos drives a wind off the proto-NS surface. In the right conditions — sufficient entropy, low electron fraction Y_e — this wind can synthesise heavy nuclei. Early hopes were that this was the r-process site, but more careful modelling shows the neutron-richness is marginal: a "weak r-process" up to A ≈ 130 is plausible, but reaching the third peak (A ≈ 195) and the actinides is difficult. Special supernovae — magnetorotational, collapsar — may do better.
• Neutron-star mergers. When two neutron stars in a binary spiral together and merge, several percent of a solar mass of cold, extremely neutron-rich material (electron fraction Y_e < 0.2) is ejected through three channels: tidal tails, shocked polar dynamical ejecta, and post-merger accretion-disk wind. This ejecta runs the r-process essentially to completion — Y_e is so low and the neutron-to-seed ratio so high that the third peak is naturally reached. The decaying r-process isotopes power the visible "kilonova" transient that follows the merger.

The debate was settled empirically on 17 August 2017.

GW170817 and AT2017gfo — direct detection

At 12:41 UTC on 17 August 2017, the LIGO Hanford detector registered a 100-second-long inspiral chirp consistent with a binary of two compact objects in the neutron-star mass range merging at a distance of about 40 megaparsecs. Two seconds later the Fermi Gamma-ray Burst Monitor detected a short γ-ray burst from the same sky region. Within hours, optical telescopes in Chile (Swope, DLT40) localised an emerging transient — designated AT2017gfo — in NGC 4993, an early-type galaxy at exactly the LIGO-Virgo distance estimate.

Over the next two weeks, ground- and space-based observatories tracked AT2017gfo through ultraviolet, optical and near-infrared bands. Two distinct components emerged. An early blue component, peaking at about t = 1 day, was consistent with a low-opacity outer ejecta dominated by lighter r-process elements (A < 140). A longer-lived red/IR component, peaking at t = 3–10 days, showed broad spectral features that were modelled successfully as the absorption blanket of lanthanide-rich, heavy-r-process ejecta. The total ejecta mass inferred from light-curve modelling was about 0.05 solar masses; the lanthanide fraction was several per cent. Extrapolating to the cosmological merger rate inferred from the LIGO-Virgo detection rate, NS mergers can produce essentially all of the heavy r-process material in the universe.

This was the first direct, multimessenger detection of heavy-element nucleosynthesis in action. Gravitational waves told us what was happening dynamically — two neutron stars merging — and the electromagnetic counterpart told us what was being produced — lanthanides and heavier r-process material radioactively powering a kilonova. The combination eliminated supernovae as the dominant heavy-r-process site and established mergers as the workhorse.

Kilonova: how the freshly made r-process atoms light up

The ejecta from a neutron-star merger is initially opaque, neutron-rich and expanding at v ≈ 0.1c–0.3c. Free neutrons decay (τ ≈ 15 min) only a fraction of the way before they have already been captured onto seed nuclei, so the r-process runs to completion in about one second after the merger. The ejecta is now composed of unstable, neutron-rich isotopes — and these isotopes β-decay back toward stability over hours, days and years.

The radioactive heating from those β-decays is the kilonova power source. Roughly

L_kilonova(t)  ≈  M_ej × ε_th × ε̇_radioactive(t)

where ε̇_radioactive ≈ 10¹⁰ erg/g/s × (t/day)^(-1.3) is the radioactive heating rate per gram of r-process material (a near-universal power-law from the convolved decays of thousands of unstable isotopes) and ε_th is the thermalisation efficiency. For M_ej ≈ 0.05 M☉ this gives peak luminosities of 10⁴¹–10⁴² erg/s — about a thousand times brighter than a typical classical nova, hence the name kilonova (Greek "kilo" = 1000), and about a hundred times dimmer than a supernova.

The colour evolution traces the opacity structure of the ejecta. Lanthanides have huge bound-bound opacity in the optical because of their many low-lying f-shell transitions — the same physics that makes rare-earth ions strongly coloured in solution. A lanthanide-rich layer therefore traps optical photons until it expands enough to thin out, releasing them on a timescale of weeks. A lanthanide-poor outer layer, in contrast, is roughly an order of magnitude lower opacity and shines blue for the first day or two. AT2017gfo showed both — outer blue, inner red — confirming that mergers eject heterogeneous r-process material.

What r-process elements are common in everyday life

• Gold (Au, Z = 79). The solar-system gold abundance is reproduced almost entirely by r-process production, with a few percent contribution from the s-process. Every wedding ring, every conductor in a high-end PCB connector, every leaf of gold in art — colliding-neutron-star material.
• Platinum-group metals (Pt, Pd, Rh, Ru, Os, Ir). All sit near the third r-process peak. The catalytic converters in petrol cars and the catalysts in industrial chemistry rely on these. Earth's mantle inherited them from r-process-enriched ISM.
• Europium (Eu, Z = 63). A lanthanide. Used in the red phosphors of older CRT monitors and modern white LEDs; its absorption band in lanthanide-rich ejecta is one of the kilonova fingerprints. Sneden's Star and other ultra-metal-poor halo stars enriched by single r-process events show enhanced Eu/Fe by factors of 30+.
• Uranium (U, Z = 92) and Thorium (Th, Z = 90). The two long-lived natural actinides. Heat from their radioactive decay drives plate tectonics in Earth's interior and fuels nuclear reactors. Their primordial ratio is set by the r-process; comparing present-day ratios in old halo stars to the r-process prediction gives ages via cosmochronology.
• Tellurium, Iodine, Xenon (Te, I, Xe). The second r-process peak. Iodine in your thyroid, xenon in arc lamps — second-peak r-process products.

Cosmochronology with Th and U

Because thorium-232 and uranium-238 are produced in the r-process in calculable ratios but decay at very different rates (half-lives 14.05 Gyr and 4.47 Gyr respectively), their present-day ratio in an old, metal-poor star is a clock. If a halo star was enriched by one or a few r-process events early in its life and has been quiescent since, then

[Th/U]_now  =  [Th/U]_initial × exp[(λ_U - λ_Th) × t]

where λ = ln(2) / T_½. Inverting for t gives the time since enrichment. Applied to CS 22892-052 (Sneden's Star, [Fe/H] ≈ -3.1) the method gives t ≈ 13 Gyr, in agreement with the age of the universe from CMB and Cepheid calibrations to within ~1 Gyr. The same technique applied to CS 31082-001 ("Cayrel's star") and HE 1523-0901 gives consistent ages 12–14 Gyr. The fact that independent astrophysical clocks all point at the same number is one of cosmology's quiet triumphs.

Open questions

• What is the merger rate? GW170817 is one event. The inferred local rate is 320 (+490, -240) Gpc⁻³ yr⁻¹ from LIGO O3, with very large uncertainty. Bigger merger samples will sharpen the mass yield estimates and tell us whether mergers alone account for all r-process material or whether supernovae must contribute too.
• Do all NS mergers eject r-process material? The mass of the merger remnant — whether it forms a hypermassive NS that survives for some time, or collapses promptly to a black hole — strongly affects the ejecta amount and composition. Different progenitor mass ratios may produce qualitatively different kilonovae.
• What fraction of the lightest r-process comes from supernovae? The light r-process (Sr, Y, Zr, ~A = 90) shows scatter in metal-poor stars that hints at multiple contributing sources. Disentangling supernova and merger contributions to the weak r-process is an active observational and theoretical programme.
• Are there exotic r-process sites? Collapsars (long-GRB progenitors), magnetorotational supernovae, and accretion-induced collapse of white dwarfs have all been proposed. Some may matter for specific element ratios. The early-universe r-process (z > 6) is poorly constrained.
• Nuclear physics inputs. Many isotopes along the r-process path are too short-lived to study in terrestrial laboratories. Their β-decay rates, neutron-capture cross-sections, and fission probabilities are estimated from theoretical mass models, and uncertainties propagate into predicted abundance patterns. The FRIB facility (Michigan) is the current frontier for measuring nuclear properties of the most exotic neutron-rich isotopes.

Common pitfalls

• Confusing r-process and s-process peaks. Both processes produce three abundance bumps tracking the magic neutron numbers, but at different mass numbers because the r-process path is offset to neutron-rich side. The s-process peaks at A = 88, 138, 208 (closed-shell stable isotopes); the r-process peaks at A = 80, 130, 195 (β-decay products of the closed-shell waiting points).
• "Gold comes from supernovae." The popular-science slogan from the 1960s–2000s. After GW170817 it is more accurate to say "gold comes from neutron-star mergers." Supernovae make the lighter r-process elements (perhaps up to A ≈ 130) at most.
• Mistaking kilonova for nova or supernova. A nova is a thermonuclear surface eruption on an accreting white dwarf (10³⁹ erg total). A supernova is the death of a massive star or thermonuclear runaway of a white dwarf (10⁴⁴ erg). A kilonova is the radioactively powered transient from a neutron-star merger (10⁴¹–10⁴² erg). Three different physics, three different energy scales.
• Treating "neutron capture" as if it acts only on Fe seeds. The r-process operates on whatever seed-nucleus distribution the merger ejecta provides — typically a mix dominated by iron-peak nuclei but with a tail extending to A ≈ 90. Heterogeneity in the seed distribution affects the final abundance pattern.
• Forgetting fission cycling. If the r-process pushes nuclei into the actinide region and beyond, the very heaviest fissile nuclei spontaneously split, returning fragments to lower A and adding to the abundance at A ≈ 130. This "fission cycling" can flatten predicted abundance patterns and is an active area of investigation for the heaviest part of the spectrum.