Introduction and history of the discipline

What we are made of

The iron in your blood, the oxygen you breathe, the carbon in your bones, the iodine in your thyroid, the calcium in your teeth: almost all of the stable nuclei that make up the human body were not made on Earth. Some nuclei are relics of the first cosmic minutes; others formed in the cores of stars that no longer exist; others were forged in seconds, during explosions so violent that they can be seen from billions of light-years away. The heaviest elements — gold, platinum, uranium, and many neighboring nuclei — carry the signature of extremely neutron-rich environments: neutron-star mergers observed directly, and perhaps also rare collapsing or magnetorotational explosions. We are, almost literally, stardust: a poetic phrase that has the unusual virtue of being physically close to the truth.

Stellar nucleosynthesis is the discipline that tells this story: how the universe, starting from a primordial composition dominated by hydrogen and helium, with traces of deuterium, helium-3 and lithium-7, built the ninety-odd elements we find today in the periodic table. It is neither a single process nor a single event, but a chain of nuclear transformations that has been running since the beginning of time and that, even as you read these words, is still under way in billions of stars.

The synthesis of elements heavier than helium must occur in stars.

Hoyle’s sentence is one of the great syntheses of twentieth-century science, and contains an entire research program. Stated more precisely, the discipline asks which nuclear reactions produce which nuclides, in which astrophysical sites, at which rates. The answers unfold along three planes that today work hand in hand: laboratory nuclear physics, which measures cross sections, decay rates and nuclear masses in increasingly sensitive accelerators; models of stellar and explosive evolution, which describe the structure of the site — temperature, density, duration — where those reactions take place; and observations of abundances — stellar spectroscopy, meteoritic analysis, the interstellar medium — which provide the final court of appeal. Each of the three pillars supports the other two: a newly measured reaction rate can overturn the prediction of a model, a newly discovered metal-poor star can invalidate an astrophysical origin once thought certain, a three-dimensional simulation can reveal a synthesis pathway that no one had thought to consider.

Describing even a single site of nucleosynthesis quantitatively is a major computational undertaking, and this is worth saying up front. Modern codes couple reaction networks — typically ranging from $\sim 10^2$ to $\sim 10^4$ nuclides linked by $\sim 10^3$ to $\sim 10^5$ reactions — to the thermodynamics and hydrodynamics of the stellar plasma, following the time evolution of abundances either in one-dimensional shell-by-shell prescriptions or, increasingly, along Lagrangian tracer particles extracted from three-dimensional magnetohydrodynamic simulations. The sensitivity of the final outcomes to experimental nuclear data, to mass loss, and to the treatment of convection is the subject of a vast literature [Wallerstein et al. 1997] , and any progress on one of the three pillars usually drags the other two along with it. The reference codes that we will encounter throughout this book — MESA for stellar evolution, NuGrid/NuPyCEE for galactic-scale contributions, SkyNet for explosive nucleosynthesis, all built on rate libraries such as JINA REACLIB — are today’s quantitative instrument for interrogating the stars. As for the nuclear inputs, tables of masses, half-lives and decay channels are continuously maintained by the Brookhaven (NNDC) [Brookhaven National Laboratory] and IAEA (IAEA-NDS) [International Atomic Energy Agency] databases, updated each time a new measurement improves on the previous one.

Cosmic kitchens

The universe has, in essence, three great kinds of “kitchens” where elements are manufactured — three environments with very different temperatures, durations and final yields — plus a fourth, smaller one that operates in the void between the stars.

The first kitchen is the oldest of all: the Big Bang itself. In its earliest minutes, when the universe was a hot soup of protons, neutrons, electrons and photons at billions of degrees, nuclear reactions had just enough time to forge hydrogen, helium, and a pinch of lithium. Nothing more: cosmic expansion cooled the plasma too quickly for heavier nuclei to be assembled. The result is a universe that, twenty minutes after the bang, is about 75% hydrogen and about 25% helium-4 by mass, with traces of deuterium, helium-3 and lithium-7 — and that would have remained so, were it not for the stars.

The second kitchen is the stars, which during their long lifetimes — hundreds of millions or billions of years — manufacture carbon, oxygen, nitrogen, silicon and iron-group nuclei through chains of nuclear fusion in their cores and burning shells. The Sun belongs to this category: it has been burning hydrogen into helium for four and a half billion years, and it will go on doing so for about as long again. The most massive stars rush through a sequence of ever faster burnings — helium, carbon, neon, oxygen, silicon — until they build a core dominated by the iron group that brings their existence to an end. A particular sub-class of evolved stars, the AGB giants (asymptotic giant branch), adds a decisive contribution: the s-process of slow neutron capture, which by weaving in neutrons one at a time produces about half of the isotopes heavier than iron in a quiet manner, before ejecting the enriched material in stellar winds.

The third kitchen is that of explosive events: core-collapse supernovae, thermonuclear (Type Ia) supernovae, neutron star mergers, novae, and X-ray bursts on the surfaces of accreting white dwarfs or neutron stars. In these cataclysms matter is compressed and heated to temperatures and densities so extreme that within seconds — sometimes within fractions of a second — nuclei that quiescent burning cannot make in abundance are produced: radioactive iron-peak material, rare proton-rich nuclei, and neutron-rich r-process material. The merger of two neutron stars observed in 2017 as the gravitational-wave event GW170817 [Collaboration & Collaboration 2017] , followed hours later by the optical and infrared kilonova AT2017gfo [Kasen et al. 2017] , provided the first direct evidence that a mass of order $10^{-2}\,M_{\odot}$ of heavy material can be synthesized in events of this kind, turning a sixty-year-old open question into a much more tightly constrained quantitative problem.

There is, finally, a fourth kitchen, by far the smallest in terms of matter transformed but indispensable for explaining a section of the periodic table that the other three cannot produce: cosmic-ray spallation. When a cosmic ray — a high-energy proton or nucleus — collides with a carbon or oxygen atom in the interstellar medium, it can shatter it into the lighter nuclei of lithium, beryllium and boron. These three elements are easily destroyed in stars (they are fragile at temperatures of millions of degrees) and are not efficiently produced either in the Big Bang or in quiescent stellar fusion: without cosmic-ray spallation, the periodic table would have a gap between helium and carbon. It is the cleanest case of nucleosynthesis taking place outside of stars, and chapter 2 of this book treats it alongside primordial nucleosynthesis.

This four-way partition is not just a convenient narrative simplification: it corresponds to radically different physical regimes. Primordial nucleosynthesis (BBN, Big Bang Nucleosynthesis) operates between temperatures of $\sim 1$ MeV and $\sim 30$ keV, lasts about twenty minutes, and unfolds in a homogeneous cosmological setting characterized by a low baryon-to-photon ratio $\eta \sim 6 \times 10^{-10}$. It produces essentially $^{1}\mathrm{H}$, $^{4}\mathrm{He}$, $^{2}\mathrm{H}$, $^{3}\mathrm{He}$, $^{7}\mathrm{Li}$ — and nothing else that survives. Quiescent stellar nucleosynthesis — in main-sequence stars, red giants, AGB stars, advanced phases of massive stars — works at lower temperatures, between $\sim 10^7$ and $\sim 5 \times 10^9$ K, but for incomparably longer times, from $10^3$ to $10^{10}$ years: chains of charged-particle fusion and, in AGB and massive stars, also slow neutron capture. Explosive nucleosynthesis drives matter to extreme conditions — up to $T \sim 10^{10}$ K and neutron densities $n_n \gtrsim 10^{24}\,\mathrm{cm^{-3}}$ — for very short times: rapid neutron capture (r-process), photodisintegration (p- or $\gamma$-process), rapid proton capture (rp-process) on accreting surfaces, freeze-out of nuclear statistical equilibrium. Spallation, finally, operates at single-nucleon energies of hundreds of MeV or more, but on very low target densities — the interstellar medium contains one atom per cubic centimeter, against the $10^{24}$ of a stellar plasma — and yields comparatively tiny amounts of matter.

Every atom heavier than helium that we encounter in everyday life has passed through at least one of these kitchens, and often through more than one. Assigning a single percentage to each element is misleading, however, because the contributions depend on the mass of the progenitor star, its initial metallicity, the type of explosion, and the chemical history of the galaxy in question. In broad terms, oxygen is dominated by hydrostatic and explosive nucleosynthesis in massive stars; carbon and nitrogen receive important contributions from both massive stars and AGB stars; iron comes partly from thermonuclear Type Ia supernovae and partly from core-collapse supernovae; gold, platinum and the actinides require r-process conditions in extremely neutron-rich environments, with neutron-star mergers now demonstrated as a real site, but not necessarily the only one [Kobayashi et al. 2020] [Cowan et al. 2021] . The periodic table, seen in this light, is not a static list: it is a map of trajectories.

The abundance curve

If we take every element in the universe and order them by abundance — how much matter exists as hydrogen, how much as helium, how much as lithium, and so on up to uranium — we obtain a curve with an unmistakable shape. It is one of the most important figures in astrophysics, and it is at once the first piece of evidence and the first challenge for any theory of cosmic nucleosynthesis: every detail of it tells a precise story that the theory must be able to reproduce.

At the far left, hydrogen and helium dominate by a wide margin: together they make up about 98% of all baryonic matter in the universe. Immediately afterwards comes a spectacular collapse: lithium, beryllium and boron are rare, hundreds of thousands of times less abundant than their neighbors. The curve then climbs and wavers, crossing an irregular zone of intermediate-mass elements, until it reaches a prominent peak at iron. Beyond iron the descent is slow and regular, punctuated by two or three small characteristic peaks before fading out at the boundaries of the periodic table. Every bump, every dip, every peak corresponds to a precise physical process that operated in a specific astrophysical site — the curve is, quite literally, a fingerprint of cosmic nucleosynthesis.

nuclide denotes a specific nucleus, distinguished both by $Z$ and $N$, and in nucleosynthesis equations it is the fundamental unit of bookkeeping.

A rigorous measurement of this curve is obtained by combining solar photospheric spectroscopy, analysis of CI chondritic meteorites (which have preserved the composition of the primordial Solar System for all non-volatile elements), and stellar abundances for nearby stars. It is usually displayed as the logarithm of abundance versus atomic number $Z$ — or, more finely, against mass number $A$, since many isotopic details would be invisible in $Z$. The founding paper of modern stellar nucleosynthesis [Burbidge et al. 1957] , universally known as B²FH from the initials of its four authors, introduced eight distinct processes for the synthesis of the heavy elements, each aimed at explaining a particular region of this curve; a review forty years later [Wallerstein et al. 1997] brought its status up to date with the clarity of hindsight, and a more recent review [Thielemann et al. 2017] incorporates the observational revolution of the past decade. The current reference compilation of solar abundances is that of Asplund and collaborators [Asplund et al. 2021] .

The principal features of the curve read like a story:

• The dominance of primordial $\mathrm{H}$ and $\mathrm{He}$ is the direct legacy of the Big Bang: in the primordial composition the mass fractions are roughly $X_{\mathrm{H}} \simeq 0.75$ and $Y_{\mathrm{p}} \simeq 0.247$, with deuterium, $^{3}\mathrm{He}$ and $^{7}\mathrm{Li}$ at trace levels.
• The deep dip at $\mathrm{Li}$, $\mathrm{Be}$, $\mathrm{B}$ reflects the fragility of these nuclei, easily destroyed at stellar temperatures, and the dependence on an “external” source such as cosmic-ray spallation.
• The iron group ($\mathrm{Cr}$, $\mathrm{Mn}$, $\mathrm{Fe}$, $\mathrm{Co}$, $\mathrm{Ni}$) marks the region close to the maximum binding energy per nucleon. The absolute maximum is not exactly at $^{56}\mathrm{Fe}$, but the combination of nuclear stability, statistical equilibrium in explosions and radioactive decays makes the iron group the natural accumulation point of advanced burning.
• The families of peaks around $A \sim 80$-$90$, $130$-$140$ and $195$-$208$ are the signature of the neutron shell closures $N = 50, 82, 126$. Their precise positions differ between the s-process and the r-process because the neutron flux, the path far from stability and the subsequent $\beta$-decay chain are different.
• The local accumulations near lead and bismuth, together with the survival of very long-lived thorium and uranium, mark the practical limit of natural nucleosynthesis before fission, $\alpha$ decays and nuclear instability close the table of durable nuclei.

To talk about these abundances in a way that allows formulas to close, a small notational apparatus is needed. We define the mass fraction $X_i$ of a nuclide $i$, satisfying $\sum_i X_i = 1$, and the number fraction $Y_i = X_i / A_i$, where $A_i$ is the mass number of the nuclide. The fraction $Y_i$ is the natural quantity to use in writing the equations of nuclear reactions, because it literally counts how many nuclei of type $i$ there are per unit mass of plasma, independent of how heavy each one is.

The time evolution of $Y_i$ obeys a coupled reaction network, which in its most general form is written

$$\frac{dY_i}{dt} = \sum_j N^i_j \, \lambda_j Y_j + \sum_{j,k} \frac{N^i_{j,k}}{1 + \delta_{jk}} \, \rho N_A \langle \sigma v \rangle_{jk} \, Y_j Y_k + \sum_{j,k,l} \frac{N^i_{j,k,l}}{1 + \Delta_{jkl}} \, \rho^2 N_A^2 \langle \sigma v \rangle_{jkl} \, Y_j Y_k Y_l$$

where $\lambda_j$ is the decay or photodisintegration rate of nuclide $j$, $\rho$ is the mass density, $N_A$ is Avogadro’s number, $\langle \sigma v \rangle_{jk}$ is the thermal rate of two-body reactions averaged over the Maxwell-Boltzmann distribution, and the coefficients $N^i$ encode the stoichiometry (negative for destruction, positive for production). The three-body terms describe reactions such as the triple-$\alpha$, the crucial bridge that connects helium to carbon and became physically understandable through Salpeter’s treatment of intermediate beryllium-8 and Hoyle’s prediction of the resonant state in carbon-12 [Hoyle 1954] . Solving this coupled equation for thousands of species simultaneously is the task of the nucleosynthesis codes already named — MESA, NuGrid, SkyNet — and which we shall use as a quantitative thread through the central chapters of the book. For a complete formal treatment we refer to the textbook by Iliadis [Iliadis 2015], which remains the standard reference on nuclear reactions in astrophysics.

A science still alive

The picture I have sketched is solid, but it is worth saying right away that it is not complete. Stellar nucleosynthesis is a discipline in full activity, crossed by at least three great open questions that are worth naming from the outset, because they will reappear in the central parts of the book and because they illuminate the way this science actually works.

The first question is the principal site of the r-process, the rapid neutron capture that produces about half of the isotopes heavier than iron, among them many nuclei of gold, platinum and the actinides. For decades the candidates have included neutrino-driven winds from proto-neutron stars in core-collapse supernovae, rare magnetorotational explosions, accretion disks around black holes, and neutron-star mergers. The observation of GW170817 in 2017 and its kilonova provided the first direct measurement of the mass of heavy elements produced in a merger, and showed that mergers of this kind are certainly an r-process site. What remains open is how much they contribute, by themselves, to the chemical history of the Milky Way and to the very early enrichment of galaxies: delay times, event rates and possible alternative channels remain active parts of the debate [Cowan et al. 2021] [Côté et al. 2018] .

The second question is the cosmological lithium problem: the lithium observed in the most ancient stars is roughly three times less abundant than the BBN prediction, even after the baryon-to-photon ratio has been pinned down by the Planck cosmic microwave background data [Collaboration 2020] [Cyburt et al. 2016] . Destruction or dilution of lithium in metal-poor stars through atomic diffusion and mixing, residual nuclear uncertainties, observational systematics, and physics beyond the Standard Model are all discussed possibilities. A stellar-depletion explanation is one of the more plausible lines, but it has not yet turned the problem into a closed case.

The third question, more subtle and less celebrated, is the quantification of the uncertainties on reaction rates for explosive nucleosynthesis processes. Many critical reactions involve nuclei far from the valley of stability — that is, nuclei that live only milliseconds in terrestrial laboratories — and their rates are often known only through theoretical models, with uncertainties that propagate directly to the predicted abundances. Experimental facilities such as FRIB in Michigan, FAIR in Germany, HIE-ISOLDE at CERN, and underground accelerators such as LUNA at the Gran Sasso National Laboratories have transformed the measurement landscape in recent years and will continue to do so throughout the coming decade.

These questions are not defects of the picture: they are precisely the areas where current research is concentrated, and it is there that students and researchers reading this book have the opportunity to contribute. A science is more honest the more transparent it is about what it does not know, and this book tries to be so.

Map of the book

The book follows a thread that is at once historical and cosmological, compressed into eight chapters plus appendices and bibliography. Chapter 1 introduces the framework and reconstructs the history of the discipline from Eddington’s first calculations to the multi-messenger revolution. Chapter 2 covers the foundations of nuclear physics needed to understand the reactions, primordial nucleosynthesis, and the spallation origin of lithium, beryllium and boron. Chapter 3 runs through the general framework of stellar evolution and the quiescent burnings, from hydrogen up to silicon. Chapter 4 describes AGB stars, the s-process and presolar grains as a geochemical archive. Chapter 5 is devoted to explosions: core-collapse supernovae, thermonuclear Type Ia supernovae, the p-process, classical novae and X-ray bursts. Chapter 6 treats neutron-star mergers and the r-process. Chapter 7 turns to the observational side: the measured curve of cosmic abundances [Asplund et al. 2021] and the chemical evolution of the Galaxy over time. Chapter 8 collects the open frontiers. The appendices — a table of relevant isotopes and a glossary — and an annotated bibliography close the book.

The organization follows three parallel cuts that will recur throughout: by cosmological origin (primordial, stellar, explosive, spallation), by physical mechanism (charged-particle fusion, slow and rapid neutron capture, photodisintegration, spallation, rp-process), and by astrophysical site (BBN, main sequence, AGB giants, core-collapse and Type Ia supernovae, novae, X-ray bursts, neutron-star mergers, interstellar medium). The periodic table can be traversed through any combination of these cuts: gold requires the r-process and has neutron-star mergers as an observationally demonstrated site, barium is produced largely by the s-process in AGB stars, iron receives important contributions from both Type Ia and core-collapse supernovae, and beryllium is made mainly by cosmic-ray spallation in the interstellar medium. Every element has an origin, and often an intricate history made of multiple contributions.

The observational sections of the book are not grouped at the end because they are less important, but because they play a specific role in the architecture of the discourse: they constrain and validate the models built in the central chapters. The separation between what we know with reasonable certainty and what remains an open conjecture is a tension that runs through the entire treatment, and is made explicit every time it surfaces — because nucleosynthesis is a young, rich science, still largely to be written.

Before nuclear physics (1859-1925)

The history of stellar nucleosynthesis begins, paradoxically, with a failure. In 1862 William Thomson — the future Lord Kelvin — published a calculation of the age of the Sun based on the only conceivable energy source of the time: gravitational contraction. His estimate, refined by Hermann von Helmholtz, was about twenty million years, a hundred million at the very most. It was a figure that Kelvin defended with the same firmness with which, across the Channel, Charles Darwin maintained that the evolution of species required geological timescales of at least hundreds of millions of years. The debate between the great physicist and the great naturalist was among the sharpest of Victorian science, and by the mid-1880s it looked as if Kelvin would win: physics was an exact science, geology a matter of opinion. In the later editions of On the Origin of Species, Darwin toned down his evolutionary timescales to avoid quarrelling with thermodynamics.

Darwin was right. But to prove it, one had to find an energy source that the physics of 1880 simply did not possess. The beginning of the solution arrived in 1896, when Henri Becquerel accidentally discovered the radioactivity of uranium; Pierre and Marie Curie isolated polonium and radium in the years that followed, and Ernest Rutherford, in the early 1900s, recognized that radioactivity was a nuclear transmutation with an enormous energy budget compared to any chemical process. In 1903 Rutherford speculated, in a vague but prophetic way, that similar processes might power the stars: the energy released by a gram of radium in a year is millions of times that of a gram of coal burned. Radioactivity did not yet provide the modern age of the Sun — precise radiometric dating of Earth and meteorites would mature only many decades later — but it showed that Kelvin’s timescale was not an unbreakable limit imposed by physics.

In the meantime, chemistry and spectroscopy had laid the groundwork with two strokes of luck that deserve to be remembered. The first is the discovery of helium. In 1868 the French astronomer Jules Janssen, observing a solar eclipse in India with a spectroscope, identified an unknown yellow line in the spectrum of the chromosphere; a few months later, the Englishman Norman Lockyer attributed it to a new element and gave it the name from the Greek helios, the sun. Helium was discovered on the Sun before it was discovered on Earth, where it would be isolated only in 1895 by William Ramsay as a decay product of uranium minerals. There is a certain poetic weight to the fact: the second most abundant element in the universe announced itself to astronomy before to chemistry, and it is solar by baptism.

The second stroke of luck was Dmitri Mendeleev’s periodic table of 1869, which arranged the sixty-three known elements in columns of recurring properties and left empty cells for elements not yet isolated — gallium, scandium, germanium — with predictions of their chemical properties. The confirmation of these predictions in the following two decades convinced even the sceptics that the periodic table was not a classification of convenience but reflected an underlying structure. Yet what structure? Only in 1913 did Henry Moseley, firing electrons at atomic targets and measuring the X-ray lines emitted, demonstrate that the atomic number $Z$ is an integer corresponding to the charge of the nucleus. From that moment on the periodic table was no longer a combinatorial mystery: it was a sequence of nuclei with charges $1, 2, 3, \ldots$, and the missing elements were those that nature had not yet built — or had built too little of for us to notice.

The last ingredient of the pre-nuclear landscape was mass spectrometry, developed by Francis William Aston at Cambridge. Between 1919 and 1922, Aston measured with growing precision the masses of the isotopes of the light elements and discovered what we now call mass defect: the mass of a composite nucleus is slightly less than the sum of the masses of its separate constituents. For $^{4}\mathrm{He}$ the defect is about 28 MeV, equivalent to 0.7% of the total mass of the four nucleons. It is a small fraction in percentage terms, but — combined with Einstein’s 1905 relation $E = mc^2$ — it is enormous in energy terms. Aston received the Nobel Prize in chemistry in 1922 for precisely these measurements.

At this point in the story only one cell remains empty, and it is the most important: what are the stars made of? In the 1920s the dominant opinion, vigorously defended by Henry Norris Russell at Princeton, held that stars had a chemical composition broadly similar to that of the Earth: heavy metallic elements in the core, perhaps an envelope of lighter gases on the outside. It is an idea that sounds reasonable — the Earth is made of silicates and iron, why should the Sun be so different? — but it is radically wrong, and the person who proved it was a 25-year-old Englishwoman who in 1925 defended at Harvard the most important doctoral thesis in the history of astrophysics.

Cecilia Payne (later Payne-Gaposchkin, after her marriage to Sergei Gaposchkin) had arrived at Harvard from England in 1923, near certain that at Cambridge — where she had studied — a woman would never obtain a doctorate in physics. Under Harlow Shapley’s supervision, she applied the thermal ionization theory recently published by the Indian astrophysicist Meghnad Saha to stellar spectral lines, and reached a devastating conclusion: stars are made essentially of hydrogen, with a non-negligible fraction of helium and entirely negligible amounts of everything else [Payne 1925]. Russell, sent the thesis as a reader, wrote to Payne that the result was “clearly impossible” and asked her to qualify the conclusion. Payne added a hedging note to her work — “the enormous abundance derived for these elements in the stellar atmosphere is almost certainly not real” — and published the thesis in 1925. It would be Russell himself, in 1929, who publicly acknowledged that Payne had been right, in a paper that cited her thesis only in passing.

The story has two morals. The first is scientific: Eddington in 1920 was calculating the energy of hydrogen-helium fusion without knowing that stars are made of hydrogen; when Payne discovered it, Eddington’s idea became quantitatively consistent with the observed composition, and stellar nucleosynthesis became a research program with the right raw materials in hand. The second is sociological, and worth remembering in a book like this one: one of the greatest discoveries of twentieth-century astrophysics was published with the correct conclusion hidden behind a qualification imposed by an authoritative reviewer. It is a story that recurs in many corners of science, and is worth re-reading every time a counterintuitive result is “softened” to make peace with the consensus.

The question: where do the elements come from? (1920-1932)

The conceptual turning point, in the strict sense, came in 1920. Arthur Eddington, in an address to the British Association delivered in Cardiff on 24 August and published the same year in Nature [Eddington 1920] , observed that the mass of the helium nucleus, according to Aston’s fresh measurements, was less than the sum of the masses of four hydrogen nuclei by about 0.7%. Eddington proposed that the difference, converted to energy according to $E = mc^2$, could be the source of the Sun’s luminosity. The calculation is elementary, but its consequences are enormous: even if only 5% of the Sun’s mass were hydrogen available for fusion, the energy released would suffice for hundreds of billions of years.

If only five per cent of a star’s mass consists initially of hydrogen atoms, which are gradually being combined to form more complex elements, the total heat liberated will more than suffice for our demands.

In the same address Eddington took a further, almost prophetic step: “The store [of subatomic energy] is well-nigh inexhaustible, if only it could be tapped. There is sufficient in the sun to maintain its output of heat for 15 billion years.” The value is not close to the modern solar age of $4.57$ billion years, but it is of the right order to demolish Kelvin’s limit and to show that subatomic energy could sustain a star for geological timescales. Eddington also intuited, in a remarkable passage, that hydrogen fusion might not be the only process: he spoke of “transmutation” of the elements in stars as a general mechanism.

Eddington’s argument is pure energetics, however, and does not specify a mechanism. There is, in fact, a very serious problem to be solved before fusion can be taken as an operational process: the Coulomb barrier. For two protons to come close enough to fuse — at nuclear distances of order one femtometer — they must overcome an electrostatic repulsion of about 1 MeV; the temperatures at the center of the Sun, even for the optimistic estimates of the time, corresponded to thermal energies of roughly 1 keV — a thousand times too low. Eddington himself, when faced with objections of this kind, used a reply that has passed into legend: “We tell the critic to go and find a hotter place.” In other words, fusion in stars must happen because nothing else works. It is a philosophically robust argument but a physically unsatisfying one, and it would have to wait eight years to be resolved.

The decisive tool was quantum mechanics, and it arrived in 1928 from a young Russian-Ukrainian physicist who had just left Leningrad for Göttingen. George Gamow was studying $\alpha$-decay — the spontaneous emission of helium nuclei from heavy nuclei such as uranium — and showed that the phenomenon could be explained as a quantum tunneling through the Coulomb barrier. Independently, and almost simultaneously, Ronald Gurney and Edward Condon reached the same result at Princeton. The astrophysical implication was drawn out the following year by Robert Atkinson and Fritz Houtermans, in a paper published in Zeitschrift für Physik in 1929: applying Gamow’s calculation to proton-proton fusion, they showed that at temperatures of a few tens of millions of degrees the effective cross section — though still tiny — became large enough to account for solar luminosity over billion-year timescales. Atkinson, recalling the episode many years later, said that the conference at which he presented the calculation ended with the audience split between enthusiasm and scepticism: it worked, but it seemed miraculous.

Between 1929 and 1938 the missing pieces of the puzzle fell into place rapidly. James Chadwick discovered the neutron at the Cavendish Laboratory in Cambridge in 1932, completing the picture of nuclear constituents. John Cockcroft and Ernest Walton, also at the Cavendish, inaugurated in the same year the era of artificial accelerators, bombarding lithium with accelerated protons and producing the first artificial nuclear reaction: $^{7}\mathrm{Li}(p,\alpha)^{4}\mathrm{He}$ — a reaction that, not by chance, still figures prominently in modern Big Bang nucleosynthesis concerns. In the same years Ernest Lawrence at Berkeley built the first cyclotron, and Hans Bethe published a long series of articles — the so-called Bethe Bible of 1936-37 — which systematized in encyclopedic form all of known nuclear physics. Around Bethe, and around the universities of Cornell, Caltech, Berkeley, Cambridge, Göttingen, and Bohr’s institute in Copenhagen, there formed the transnational community of nuclear physicists that, before being scattered by the war and by Nazism, laid the foundations of everything that would come after. For the historical reconstruction of the period the prefaces of Clayton’s 1968 monograph and of Iliadis’s [Iliadis 2015] remain essential, weaving together the experimental and theoretical strands with biographical accuracy.

Bethe and the fusion cycles (1938-1939)

In the spring of 1938 George Gamow and Edward Teller organized at George Washington University a closed conference of a week on the sources of stellar energy. It was the fourth meeting in a series begun in 1935, and the attendees were given an explicit goal: close the problem of solar energetics. About twenty physicists took part, among them Hans Bethe, Subrahmanyan Chandrasekhar, Charles Critchfield, Marshak, and Bengt Strömgren. Bethe — who had until then worked on pure nuclear physics and had never seriously turned his attention to stars — accepted the invitation and, as he liked to recall, understood the problem on the train ride home.

The result was the two papers that founded quantitative stellar nucleosynthesis. The first, signed by Bethe and Critchfield in 1938, presented in detail the pp chain (proton-proton): the fusion of two protons into a deuteron with emission of a positron and a neutrino,

$$p + p \to d + e^{+} + \nu_e$$

is a weak reaction — it involves the conversion of a proton into a neutron — and therefore extraordinarily slow, with a timescale of billions of years at solar temperature. Once the deuteron forms, the subsequent steps are fast: $d(p,\gamma)^{3}\mathrm{He}$ and then the termination $^{3}\mathrm{He}+^{3}\mathrm{He}\to ^{4}\mathrm{He}+2p$. The net result is the conversion of four protons into one helium nucleus, with the release of about 26.7 MeV per complete cycle, a small fraction of which is lost as escaping neutrinos.

The following year Bethe published the second paper alone [Bethe 1939] , adding an alternative mechanism that would prove equally important: the CNO cycle. In this case hydrogen is fused into helium using carbon, nitrogen and oxygen as catalysts — nuclei that enter the chain of reactions and emerge unchanged, ready for new cycles. The closed sequence is

$$^{12}\mathrm{C}(p,\gamma)^{13}\mathrm{N}(\beta^{+}\nu)^{13}\mathrm{C}(p,\gamma)^{14}\mathrm{N}(p,\gamma)^{15}\mathrm{O}(\beta^{+}\nu)^{15}\mathrm{N}(p,\alpha)^{12}\mathrm{C}$$

where the $^{12}\mathrm{C}$ at the opening is the same one that re-emerges at the close: four protons have entered, one helium nucleus has come out, and carbon has acted as a nuclear catalyst. Bethe correctly identified the very strong temperature dependence of the CNO cycle — the energy generation per unit mass scales as $\epsilon \propto T^{18}$, because the Coulomb barrier of $^{12}\mathrm{C}$ is six times that of a proton — and the much gentler dependence of the pp chain, $\epsilon \propto T^{4}$. He concluded, however — with an inversion that history would later correct — that the CNO cycle should dominate even in the Sun; in reality the Sun burns mostly through pp, and the crossover between the two regimes sits around $1.3\,M_\odot$, a threshold understood only with the post-war stellar models. The modern picture combines solar models and neutrino data: SNO resolved the solar-neutrino problem by showing flavor oscillations, while Borexino provided the first direct evidence of CNO-cycle neutrinos in the Sun and later an improved measurement of their component [Bahcall et al. 2005] [Collaboration 2020] [Collaboration 2022] .

The paternity of these two mechanisms is in fact shared, and the story is instructive about how the physics of the twentieth century proceeded through parallel discoveries across national borders. The pp chain had been independently proposed by Charles Critchfield, who co-signed the 1938 paper; the CNO cycle had been independently proposed by Carl Friedrich von Weizsäcker in 1937-38 in Germany, in a paper in Physikalische Zeitschrift that Bethe would read only after completing his own. Bethe arrived at von Weizsäcker’s conclusions in the weeks after the Washington conference — the legendary “train ride” is probably a narrative simplification of weeks of intense work. Bethe’s synthesis was, however, the most complete: it includes a discussion of branching ratios, an estimate of timescales, and a critical evaluation of consistency with observed solar luminosity and stellar compositions. For this work Bethe would receive the Nobel Prize in 1967 — nearly thirty years later, the delay owing in part to the fact that in the years after 1939 Bethe devoted himself first to radar at MIT, then to the Manhattan Project at Los Alamos, and only later returned to the reactions of the Sun as an intellectual pastime.

There is one thing Bethe’s 1939 paper does not say, and that something would become the central question of the next decade. Bethe did not consider the synthesis of elements beyond helium. He believed — not unreasonably, given the nuclear rates known at the time — that stars could not reach temperatures sufficient to surmount the next obstacle, namely the $A = 8$ gap: beryllium-8, the would-be product of two helium nuclei fusing, is unstable and decays in $10^{-16}$ seconds. From helium to carbon there seemed to be a chasm that nature did not know how to cross. And without carbon, no life; without any element beyond helium, no planets, no chemistry, no anything. This is the quaestio that would dominate the 1950s.

Big Bang or steady state? (1946-1957)

The outbreak of the Second World War scattered the nuclear-physics community across the Manhattan Project and radar laboratories; for nearly a decade nucleosynthesis stood still. When Hans Bethe published a review on stellar energy in 1946, he had largely to restate what he had said in 1939: nothing of substance had moved in seven years. But in those same months, in an entirely unexpected way, the problem of nucleosynthesis was reopened — not by the astrophysicists but by the cosmologists.

The protagonist again is George Gamow, who after the war moved to George Washington University and became convinced that an expanding, cooling universe must have passed through a phase in which nuclear reactions were possible on a cosmological scale. Gamow called this initial state ylem (from the Greek hyle, primordial matter) and proposed that all the elements of the periodic table had formed there, by successive neutron capture, in the first minutes after the Big Bang. The proposal was worked out in detail in a 1948 paper signed by Ralph Alpher, Gamow’s student, and which Gamow — with academic mischief intent on producing an acronym — got Hans Bethe to co-sign without Bethe having contributed anything to the work [Alpher et al. 1948] . The paper passed into history as αβγ from the phonetic initials of the three surnames (Alpher, Bethe, Gamow), and it proposed that progressive neutron capture on a primordial population of neutrons would yield the full periodic table.

In the same year, in a paper that received much less attention, Alpher and Robert Herman made a prediction worth remembering as one of the most spectacular cases of overlooked prophecy in the history of physics: if the αβγ model is correct, the universe today must still be filled with a thermal background of radiation at about 5 kelvin, the residual heat of the original phase [Alpher & Herman 1948] . This was the first prediction of the cosmic microwave background (CMB), seventeen years before the Penzias-Wilson discovery, and for almost twenty years it remained largely ignored by the observational community.

The αβγ model has a fatal flaw, however, and it is precisely the one missing from Bethe’s 1939 picture: the $A = 5$ and $A = 8$ gaps. Enrico Fermi and Anthony Turkevich, in a calculation not formally published but circulated through seminars and informal contacts around 1950, showed that the chain of neutron capture stops inexorably at $^{4}\mathrm{He}$: $^{5}\mathrm{Li}$ and $^{5}\mathrm{He}$ are too unstable to survive, and crossing the $A = 8$ gap requires densities and timescales unavailable in an expanding universe. The Fermi-Turkevich verdict was decisive: cosmological nucleosynthesis can produce hydrogen, helium, and trace lithium, but no more. Gamow’s dream of explaining the entire periodic table in the first twenty minutes of the Big Bang was dead.

At this point the story takes an ironic turn worth telling carefully, because it contains the narrative engine of everything that follows. While Gamow and Alpher were developing their version of the Big Bang, at Cambridge three physicists — Hermann Bondi, Thomas Gold, and Fred Hoyle — proposed in 1948 an alternative cosmological model: the steady state. Bondi and Gold published together; Hoyle published separately in the same year. In this model the universe expands but remains statistically the same in time, because new matter is continuously created in the vacuum between galaxies, at a rate low enough to be unobservable (one atom per cubic meter per billion years) but enough to compensate for cosmic dilution. There is no initial instant, no dense and hot primordial state; and therefore there is no primordial nucleosynthesis. In a steady-state universe, all elements must be manufactured somewhere else — and the only plausible place, given that the periodic table exists, is the stars.

Fred Hoyle was a complex personality, and he conceded nothing. Convinced that he held the correct cosmology, and convinced that the rival Big Bang theory was “crude” and philosophically unsatisfactory, in a BBC radio broadcast of 28 March 1949 — the series The Nature of the Universe — he christened the rival with a nickname intended as derogatory: “this big bang idea”. The name stuck. Hoyle would later say he had used it to “ridicule” the opposition, but English has appropriated it as neutral, and today no one remembers the original satirical intent. What survives in the historical memory of science, and is the historiographically crucial point, is that Hoyle found himself forced by his own cosmology to develop a complete theory of stellar nucleosynthesis. In the steady state there was no alternative: either stars produced gold, uranium, and everything else, or the steady state was false.

For nearly twenty years, from 1948 to 1965, the two cosmologies coexisted in the community in polemic equilibrium. Big-Bang nucleosynthesis was systematized by Hoyle and R. J. Tayler in 1964, by Peebles in 1966, by Wagoner, Fowler and Hoyle in 1967 — the latter, ironically, with Hoyle himself as a co-author, by then able to run both programs in parallel. But the decision came from experiment: in 1965 Arno Penzias and Robert Wilson, at the Bell Laboratories in Holmdel, New Jersey, discovered by chance a perfectly isotropic background of radiation at 2.7 kelvin [Penzias & Wilson 1965] . It was the Alpher-Herman CMB, seventeen years late. The Big Bang won; the steady state died within a few years. Hoyle, until his death in 2001, would never accept defeat and would propose ever more speculative variants of the steady-state cosmology.

There is an important moral to this whole episode, and it is worth making explicit: stellar nucleosynthesis was born as a research program largely motivated by a cosmology that turned out to be wrong. Without the steady-state theory, it is unlikely that Hoyle would have invested the energy he did in synthesis of the elements in stars. Physics works this way: correct ideas can emerge from the need to save wrong ones, and the results survive even when the motivating context falls away. Stellar nucleosynthesis is today the key to explaining how the primordial hydrogen of the Big Bang gave rise to the iron in our blood: it is the bridge between cosmology and chemistry. And we owe it to a cosmology that lost.

The $A = 8$ obstacle and the Hoyle state (1952-1957)

Back, then, to the beryllium-8 obstacle. There is a gap in the chart of nuclides that held nucleosynthesis in check for more than ten years: no stable nucleus exists with mass number $A = 5$, and none with $A = 8$. The nuclides $^{5}\mathrm{He}$, $^{5}\mathrm{Li}$, and $^{8}\mathrm{Be}$ decay in times shorter than a femtosecond. If the universe is to build carbon (which has $A = 12$) by summing helium nuclei (each with $A = 4$), it runs into a wall: two helium nuclei would make $A = 8$, but beryllium-8 dissociates in roughly $10^{-16}$ s, a timescale unimaginably shorter than any astrophysical process. To Bethe and to Fermi-Turkevich it was a decisive argument against the nucleosynthesis of heavy elements.

The first crack in the wall was opened by Edwin Salpeter at Cornell in 1952 [Salpeter 1952] . Salpeter — a young Viennese physicist, a student of Bethe — observed that, although $^{8}\mathrm{Be}$ is unstable, its mean life is not zero, and at temperatures of $\sim 10^{8}$ K the reaction $\alpha + \alpha \to {}^{8}\mathrm{Be}$ proceeds continually in both directions, establishing a small dynamical equilibrium

$$\alpha + \alpha \rightleftharpoons {}^{8}\mathrm{Be}$$

with an equilibrium concentration of $^{8}\mathrm{Be}$ that is tiny but nonzero — roughly one part in $10^{9}$ relative to helium. On this minuscule transient population, capture of a third $\alpha$ can proceed before the $^{8}\mathrm{Be}$ dissociates, producing net $^{12}\mathrm{C}$. Salpeter calculated the rate of the $3\alpha$ reaction and showed that, in principle, three-alpha fusion was possible.

The problem was quantitative. Even with Salpeter’s equilibrium, the rate of $3\alpha$ was insufficient by many orders of magnitude to account for the observed abundance of carbon in stars. Red giants, which burn helium in their cores at temperatures of about $10^{8}$ K, should produce far less carbon than is actually observed. Fred Hoyle, who had meanwhile moved to California for a sabbatical year at Caltech, made around 1953 a bold prediction of essentially anthropic character: since carbon exists in copious amounts in the universe (and since we, carbon beings, consequently exist to observe it), $^{12}\mathrm{C}$ must possess an excited state at a precise energy that amplifies the cross section of $3\alpha$ by resonance.

The logic of Hoyle’s reasoning deserves to be made explicit. The capture of a third $\alpha$ on a $^{8}\mathrm{Be}$ in dynamical equilibrium is a non-resonant reaction if the resulting $^{12}\mathrm{C}$ has no state at an appropriate energy; in that case the rate is insufficient. If, however, $^{12}\mathrm{C}$ possesses an excited state at an energy near the entrance energy of the $^{8}\mathrm{Be}+\alpha$ system, the effective cross section is amplified by many orders of magnitude — as happens in any resonance, in quantum mechanics as in acoustics. Hoyle estimated the required energy: the state had to lie at roughly 7.65 MeV above the carbon ground state, with spin and parity $J^\pi = 0^+$ to be accessible from $s$-wave collisions between $^{8}\mathrm{Be}$ and $\alpha$.

Hoyle convinced William Fowler to look for it at the Kellogg Radiation Laboratory at Caltech. Fowler was initially sceptical: the prediction rested on an anthropic argument — we exist, therefore carbon exists, therefore the resonance must exist — that to traditional nuclear physicists looked more like metaphysics than science. But he agreed to try. The state was confirmed experimentally by Cook, Fowler, Lauritsen and Lauritsen in 1957, through a measurement of inelastic scattering of alpha particles on carbon-12: the state was where Hoyle had said, with energy $7.65 \pm 0.02$ MeV and properties $0^+$, and with a $3\alpha$ rate amplification exactly as needed to account for observed carbon abundances [Cook et al. 1957] . It is one of the very rare historical examples of a successful anthropic prediction in astrophysics, and it is universally known today as the Hoyle state. Its existence, and the extraordinarily fine adjustment of the nuclear parameters that make it possible, is cited by some authors as a textbook example of fine-tuning of the nuclear Standard Model for the formation of carbon — and therefore of life — in the universe.

Hoyle published an organic treatment of the synthesis of heavier elements in two follow-up papers, in 1954 and 1955. The first [Hoyle 1954] is an almost encyclopedic synthesis of the reactions producing nuclei from carbon to nickel in massive stars, and it introduces for the first time the concept of the onion-shell structure of pre-supernovae — concentric shells of burning, with an iron core at the center, successive shells of silicon, oxygen-neon, carbon-oxygen, helium, and residual hydrogen on the outside. The second paper deals with the subsequent $\alpha$-capture, $^{12}\mathrm{C}(\alpha,\gamma)^{16}\mathrm{O}$, whose cross section remains — even today, after seventy years of experimental effort — the most studied and the least known of stellar nucleosynthesis. It is what William Fowler would call the “holy grail” of the field, and it is taken up in the chapter on helium burning.

B²FH and Cameron: the synthesis (1956-1957)

The year 1957 is the year of the systematic breakthrough. Four authors — Margaret Burbidge, Geoffrey Burbidge, William Fowler, Fred Hoyle — published in Reviews of Modern Physics a work more than a hundred pages long that organized into eight distinct processes everything then known about element synthesis [Burbidge et al. 1957] . This is the “B²FH” paper (from the initials of the four), probably the most cited paper in twentieth-century astrophysics, and for two decades it would serve as the field’s road map. In the same year, and entirely independently, Alastair Cameron published largely overlapping conclusions as an internal report of Atomic Energy of Canada Limited [Cameron 1957]. From that moment forward, stellar nucleosynthesis ceased to be a collection of distinct ideas and became a discipline with a common language.

The ingredient that made B²FH possible was an observational dataset that had appeared one year earlier: the Suess and Urey compilation of 1956 of elemental abundances in the Solar System, published in the same volume of Reviews of Modern Physics where B²FH would appear [Suess & Urey 1956] . Hans Suess (an Austrian physico-chemist) and Harold Urey (Nobel laureate in chemistry for the discovery of deuterium) had combined solar spectroscopy, analysis of CI chondritic meteorites, terrestrial abundances corrected for geochemistry, and laboratory data to produce the first quantitative curve of cosmic abundances worthy of the name. Their curve — reproduced in every astrophysics text from that moment on — showed unambiguously the features that any theory of nucleosynthesis would have to explain: the dominance of hydrogen and helium, the LiBeB dip, the iron peak, the double peaks of neutron capture at $A \approx 90, 138, 208$, the slow decline up to lead and the cutoff beyond. For the first time the nucleosynthesists had a map to decipher, not a fragmentary list.

B²FH read that curve nuclide by nuclide and proposed, for every region, an astrophysical process responsible, each with a specific nuclear mechanism and a candidate stellar site:

The scheme captures the coarse structure of the abundance curve with great accuracy and identifies candidate astrophysical sites for each process: in particular red giants for the s-process — because giants have temperatures and densities suitable for neutron production from $^{13}\mathrm{C}(\alpha,n)^{16}\mathrm{O}$ and $^{22}\mathrm{Ne}(\alpha,n)^{25}\mathrm{Mg}$ — and supernovae for the r-process, a hypothesis that we now know is partially correct but not the whole answer. Cameron’s contribution covers the same ground with greater emphasis on the dependence on stellar models and a more detailed treatment of neutron capture; Cameron, working alone in a Canadian context, received for years fewer citations than he deserved, and would become only later one of the central figures of North American nucleosynthesis.

A historiographic point often overlooked is worth dwelling on, and it concerns the first authorship of the paper. Margaret Burbidge was in 1957 one of very few women active in front-line observational astrophysics. Born Margaret Peachey at Davenport in 1919, she had graduated from University College London and had obtained a position at the Yerkes Observatory in Chicago largely owing to the war (which had emptied U.S. academic posts of male colleagues). When, in the late 1950s, she requested to observe with the major Mount Wilson and Palomar telescopes, including the 200-inch class at Palomar, she encountered rules and customs that excluded women from independent observing runs. She and her husband Geoffrey worked around the rule by submitting applications in Geoffrey’s name, while it was in fact Margaret who used the instrument with Geoffrey acting as company. The spectroscopic observations of S-type giants and barium stars that Margaret collected in the two years before B²FH built on a decisive observation already made in 1952: Merrill’s technetium lines in S stars [Merrill 1952] . Because technetium has no stable isotopes, it must have been synthesized recently by the star itself. That is the observational argument that turned the s-process from a theoretical possibility into a real stellar phenomenon. The paper is “B²FH” — not “FBBH” or anything else — not because of alphabetical order but because Margaret and Geoffrey, then also a married couple, appeared as an indissoluble working unit and the first author of the paper is her. Margaret Burbidge would become in 1976 the first woman president of the American Astronomical Society, and in 1979 director of the Royal Greenwich Observatory.

William “Willy” Fowler, the third author, was the nuclear physicist of the group. Director of the Kellogg Radiation Laboratory at Caltech, he had spent the previous decades measuring stellar reaction cross sections at low energies with ever more refined techniques; his contribution to B²FH is the guarantee that the rates entering the paper were those measured in the laboratory, not those wished for by theorists. Fowler would receive the 1983 Nobel Prize in Physics, shared with Subrahmanyan Chandrasekhar, for his theoretical and experimental studies of the nuclear reactions of importance in the formation of the chemical elements in the universe. He would be the only one of the four B²FH authors to receive a Nobel; that Hoyle was not is considered by many one of the most debated omissions in the history of the prize, attributable in part to his controversial later positions — the steady-state cosmology, the panspermia theory with Wickramasinghe, the critique of Darwin he developed in his later years.

A critical re-reading at forty years’ distance [Wallerstein et al. 1997] shows what B²FH got right and what they missed. Right: the s/r split, the role of the iron group as an accumulation region near the maximum binding energy per nucleon, the identification of red giants as the s-process site, the distinction between neutron-capture and charged-particle fusion processes, the general idea of equilibrium freeze-out for the iron peak. Missed: the main site of the r-process, which B²FH placed generically in core-collapse supernovae and which today remains distributed between neutron-star mergers, rare collapsing sites and still debated chemical-evolution constraints; the nature of cosmic-ray spallation for LiBeB, which B²FH called x-process without specifying a mechanism and which would be clarified by Reeves, Fowler and Hoyle in 1970 [Reeves et al. 1970] ; primordial nucleosynthesis, of which at the time only the general idea of the αβγ paper of 1948 was known but without precise nuclear rates or cosmological parameters that would allow quantitative predictions. The historiography of the paper is well summarized in the proceedings of the “B²FH at 60” symposium held at Edinburgh in 2017 and published in Geochimica et Cosmochimica Acta, with contributions by the surviving authors and the new protagonists of the field.

From 1957 to today: consolidation and revolutions (1957-2026)

The period following B²FH cannot be compressed into a few pages without betraying the complexity of the work done. Let us nonetheless trace the principal turning points and name at least the figures whose careers defined the sub-fields we shall meet in the central chapters of this book.

1960s: from theory to models

In the 1960s stellar nucleosynthesis ceased to be a collection of ideas and became a program of numerical computation. Donald Clayton, a student of Fowler at Caltech, published in 1968 Principles of Stellar Evolution and Nucleosynthesis, the textbook that for decades would be the standard reference of the field. Alastair Cameron, having moved from Atomic Energy of Canada to NASA’s Goddard Institute, developed the first detailed model of neutron capture in AGB stars, with quantitative predictions of s-process abundances that would gradually be confronted with spectroscopic measurements. Bohdan Paczyński, Icko Iben, and others built the first one-dimensional stellar evolution codes capable of following a star from the main sequence to the late phases.

Primordial nucleosynthesis was refounded in 1967 by Robert Wagoner, William Fowler and Fred Hoyle in a paper that is the first detailed calculation of the reaction network in the first minutes after the Big Bang [Wagoner et al. 1967] . Wagoner-Fowler-Hoyle assemble a network of about twenty dominant reactions with laboratory rates, integrate the Friedmann equations coupled to the nuclear network, and produce quantitative predictions of the primordial abundances of $^{1}\mathrm{H}$, $^{2}\mathrm{H}$, $^{3}\mathrm{He}$, $^{4}\mathrm{He}$, $^{7}\mathrm{Li}$ as a function of the baryon-to-photon ratio $\eta$. From that moment BBN becomes a real cosmological discipline: with the 1965 discovery of the CMB and its cosmological constraints, measuring $\eta$ once closes the entire picture, and every new observation of primordial abundances becomes an independent test of standard cosmology.

1970s: neutron capture differentiates

In 1974 James Lattimer and David Schramm, in a short paper in Astrophysical Journal Letters titled Black-Hole-Neutron-Star Collisions [Lattimer & Schramm 1974] , made a proposal worth remembering well because it anticipated by forty-three years the observational confirmation of at least one compact r-process site: mergers involving neutron stars — and more generally catastrophic accretion events onto compact objects — can eject extremely neutron-rich matter. The argument is simple and powerful: a merger can release amounts of order $10^{-2}\,M_\odot$ of decompressed neutron-rich matter, under conditions compatible with the r-process. Lattimer-Schramm is one of those works that the community received with polite incredulity for decades: the prediction seemed too specific, the events too rare, and consensus remained on a core-collapse supernova origin. History would revalue the idea in 2017.

In the same decade Schramm engaged the community on the other cosmological front: together with collaborators he developed the BBN constraint on the number of light neutrino species, showing that the primordial synthesis of $^{4}\mathrm{He}$ is sensitive to the number of relativistic species at BBN epoch, and therefore to the number of neutrino families. The cosmological limit — no more than three families — was later confirmed by the LEP measurements at CERN in the 1990s.

1980s: recognition and a nearby star explodes

In 1983 William Fowler shared the Nobel Prize with Chandrasekhar. It was the first (and until 2017, the only) recognition by the committee for a work on stellar nucleosynthesis. Hoyle was excluded, a decision many — even today — find debatable; the official reasoning is opaque, and community interpretations oscillate between a political judgment on Hoyle’s late positions and the mere happenstance of the prize’s three-laureate-per-year cap.

In the same decade two anomalies that would become permanent open problems of the field began to take shape. The first dates to 1982: Monique and François Spite, at the Meudon Observatory, discovered that extremely metal-poor stars in the galactic halo show an abundance of $^{7}\mathrm{Li}$ that is approximately constant — forming a flat “plateau” as a function of metallicity — at a level about three times less than predicted by modern BBN [Spite & Spite 1982] . The cosmological lithium problem begins here, and has not been definitively closed even today; the leading hypotheses, as we mentioned in the introduction, swing between stellar destruction by atomic diffusion and turbulence, residual nuclear uncertainties, and exotic physics beyond the Standard Model.

The second turning point of the 1980s came on 23 February 1987, when a star in the Large Magellanic Cloud — a blue supergiant catalogued as Sk -69° 202 — exploded as a core-collapse supernova. SN 1987A was the first supernova visible to the naked eye since Kepler’s of 1604, and was also the first astrophysical event in which extrasolar cosmic neutrinos were directly detected: twenty-seven neutrinos in three independent detectors (Kamiokande in Japan, IMB in the United States, Baksan in the Soviet Union) arrived within twenty seconds of each other, a few hours ahead of the optical light. It was the first multi-messenger astronomical observation in history de facto, and demonstrated that the core-collapse mechanism releases about 99% of the event’s gravitational energy in neutrinos — just as theory had predicted. In the months and years that followed, the light curve of SN 1987A showed the radioactive decay of $^{56}\mathrm{Ni}$ into $^{56}\mathrm{Co}$ into $^{56}\mathrm{Fe}$, line by line: one could see iron being synthesized in real time in a real supernova, and there was no longer any doubt that core-collapse supernovae produce the iron group as B²FH had predicted.

The year 1987 is also the year of the discovery of presolar grains. Tom Bernatowicz, Roy Lewis and others, at Washington University in St. Louis, isolated from carbonaceous meteorites tiny grains of silicon carbide (SiC) with isotopic ratios of carbon, nitrogen and silicon so anomalous that they could not have formed in the Solar System. They are fragments of dust produced in individual AGB stars, having survived the collapse of the progenitor molecular cloud and incorporated unaltered into primitive meteorites. For the first time, astrophysicists had laboratory samples of stellar nucleosynthesis: one could measure atom by atom with a mass spectrometer what a single AGB star had produced, and compare it directly with the predictions of the models. Isotopic cosmochemistry was born from this; it has been one of the most productive threads of the field from 1990 on, and is treated in detail in the chapter on AGB stars and presolar grains.

1990s-2000s: the underground, the simulations, the most ancient stars

The 1990s opened the era of underground measurement of astrophysical cross sections. The problem, from Atkinson-Houtermans onward, is that the reactions of astrophysical interest occur at energies in the so-called “Gamow peak” — a few keV for the pp chain, tens of keV for $^{3}\mathrm{He}(\alpha,\gamma)^{7}\mathrm{Be}$, and so on — a hundred times lower than the energies at which laboratory measurements are routinely possible. At those energies the cross sections are so small, and the cosmic-ray background so dominant, that direct measurement at the surface is impossible. In 1992 the LUNA (Laboratory for Underground Nuclear Astrophysics) collaboration installed a 50 kV accelerator at the Gran Sasso National Laboratory, beneath 1,400 meters of rock, where the cosmic-ray background is reduced by a factor of $10^{6}$ compared to the surface. LUNA’s first measurement, of the reaction $^{3}\mathrm{He}(^{3}\mathrm{He},2p)^{4}\mathrm{He}$ at exactly the solar Gamow peak, revolutionized the precision of solar-neutrino predictions and contributed directly to the resolution of the solar neutrino problem, which closed in 2001 with the SNO measurements in Canada and the identification of flavor oscillations. The sequence of reactions measured by LUNA in the years that followed — $^{14}\mathrm{N}(p,\gamma)^{15}\mathrm{O}$ in the CNO cycle, $^{3}\mathrm{He}(\alpha,\gamma)^{7}\mathrm{Be}$ critical for BBN and for $^{8}\mathrm{B}$ solar neutrinos, and many more — is today part of every nucleosynthesis network. JUNA in China (at the Jinping laboratory, 2,400 m of rock) and CASPAR in the United States (at the Sanford Underground Research Facility) follow the same model.

On the astronomical side, the 1990s and 2000s saw the boom of spectroscopy of metal-poor halo stars. The HK survey of Beers and collaborators, the Hamburg/ESO Survey, and later dedicated projects such as HERES and SAGA produced catalogs of thousands of Pop II stars with metallicities below $[\mathrm{Fe/H}] = -2$, down to extremely metal-poor stars at $[\mathrm{Fe/H}] < -4$. Anna Frebel and others discovered stars in which r-process abundances are enormously enhanced — the so-called r-II and r-only stars — that serve as “fossil archives” of nucleosynthesis in the first billion years of the Galaxy. Comparing the r-pattern in these stars with model predictions reveals that the r-pattern is universal: r-II stars all show the same r-process “fingerprint” for the heavy elements (lanthanides and beyond), which strongly constrains the nature of the producing site or sites.

Three-dimensional simulations of supernova explosions became possible on the supercomputers of the mid-2000s. The groups at Garching, Princeton, Oak Ridge and elsewhere built codes that integrate neutrino-driven hydrodynamics, neutrino transport, convection and non-linear instabilities (SASI, Standing Accretion Shock Instability) to follow the hydrodynamic bounce after iron-core collapse. It has been a decade-long, still ongoing endeavor, but it has produced a solid consensus: core-collapse supernovae work via neutrino-driven explosion assisted by convection and SASI, just as Colgate and White had conjectured back in 1966. The quantitative details — which stars explode and which collapse directly into black holes, which yields they produce, how much r-process matter emerges — remain open problems, and are taken up in the chapter on supernovae.

In 2003 the Joint Institute for Nuclear Astrophysics (JINA, today JINA-CEE) was founded at Notre Dame, MSU, and Chicago, institutionalizing the collaboration between experimental nuclear physicists, astronomers and nucleosynthesis theorists. One of JINA’s most tangible legacies is the JINA REACLIB library: an open-access compilation of tens of thousands of nuclear reaction rates in standardized parametrization form, continuously updated and used today by many nucleosynthesis codes [Cyburt et al. 2010] . Together with the National Nuclear Data Center (NNDC) at Brookhaven [Brookhaven National Laboratory] and the IAEA Nuclear Data Section [International Atomic Energy Agency] , REACLIB is part of the nuclear infrastructure that makes the discipline run.

2010-2026: gravitational astronomy and beyond

On 14 September 2015 the LIGO interferometers detected for the first time gravitational waves from a black-hole merger (GW150914): it was the birth of gravitational astronomy. For nucleosynthesis, however, the event that changed everything came on 17 August 2017.

That day LIGO and Virgo detected GW170817 [Collaboration & Collaboration 2017] , a neutron-star merger at about 40 Mpc in the galaxy NGC 4993. Hours later, dozens of optical and infrared telescopes pointed at the region identified by the gravitational waves and located an electromagnetic counterpart: the kilonova AT2017gfo. The kilonova spectrum, gathered in the days after the merger, showed a rapid transition from a blue component — opacity dominated by light and transition elements — to a red component dominated by the opacity of lanthanides, exactly as Daniel Kasen and collaborators had predicted on the basis of r-process calculations in NS-NS mergers [Kasen et al. 2017] . The total mass of heavy elements synthesized was estimated at about $0.05\,M_\odot$, with a significant heavy r-process fraction ($A > 140$) and a r-light component. Lattimer and Schramm, in 1974, had been right.

Since then the situation has become articulated. GW170817 is certainly an r-process site, but the debate about the relative weight of NS-NS mergers compared to alternative candidates — collapsars (collapses of rapidly rotating massive stars into black holes with accretion disks), magneto-rotational supernovae, relativistic jets from peculiar SN II — is still open [Cowan et al. 2021] [Côté et al. 2018] . A recent review [Thielemann et al. 2017] incorporates the multi-messenger revolution into the theoretical framework of heavy-element nucleosynthesis and gives a balanced assessment of what we know and what we don’t. Further kilonova observations with LIGO-Virgo-KAGRA and with optical/infrared surveys should narrow the statistical distribution of r-process mass produced per event, but the relative balance of sites will also remain tied to models of Galactic chemical evolution.

In recent years, finally, experimental nuclear physics has known a further revolution with the commissioning of the Facility for Rare Isotope Beams (FRIB) at Michigan State University in 2022. FRIB is today one of the most powerful sources of short-lived radioactive nuclei in the world, and it is set to measure masses, mean lives and decay channels of the nuclei far from the valley of stability that dominate the r-process — those nuclei that live milliseconds and which we often know only through theoretical models. FAIR in Germany, HIE-ISOLDE at CERN, RIBF at RIKEN in Japan are part of the same global infrastructure. In the next ten to fifteen years many of the nuclear uncertainties that still limit the precision of explosive nucleosynthesis models will be reduced by measurements at these facilities, even if not all can be eliminated: indirect cross sections, fission of superheavy nuclei and astrophysical conditions will remain independent sources of uncertainty.

The story, here, ends — provisionally. Stellar nucleosynthesis remains a young discipline, and the next decade of experimental measurements, multi-messenger observations, and simulations will transform it in ways we cannot yet foresee. It is what makes it a living science.