讨论班

Three dates on one book

The revised edition of the first Séminaire de Géométrie Algébrique immediately separates three kinds of time. The title says 1960–61. The series number, Lecture Notes in Mathematics 224, belongs to a publishing system that did not yet exist when the seminar began. The copyright and imprint say 1971. Between the work named by the first date and the object delivered by the last lay oral exposition, typed fascicles, corrections, reorganizations, and the labor of more than one mathematician. The book is therefore evidence of a seminar, but it is not a transcript; it is a later construction made from a research season.11Grothendieck, Revêtements étales et groupe fondamental (SGA 1).

That distinction matters especially in Grothendieck’s case because the later books became so authoritative that they can make their own prehistory disappear. A reader encountering SGA 1 in its 1971 form may imagine a finished theory placed before an audience eleven years earlier. A reader citing SGA 4 from the three volumes published in 1972 and 1973 may silently transport every revision back to 1963–64. The result is a false picture of mathematical creation: immaculate treatises appearing at the instant of conception. The documents instead disclose a slower, layered process. Problems were selected for a season; arguments were divided into exposés; notes circulated; gaps or dependencies became visible; and a publishable architecture was assembled only later.22Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Michael Artin et al., eds., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2, Lecture Notes in Mathematics 270 (Springer, 1972), https://doi.org/10.1007/BFb0061319; Michael Artin et al., eds., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3, Lecture Notes in Mathematics 305 (Springer, 1973), https://doi.org/10.1007/BFb0070714.

The word seminar can sound modest beside the size of the resulting corpus. Here it named neither an occasional lecture series nor a passive survey course. It was a device for making a field. The successive programs addressed étale coverings and the fundamental group, local cohomology and Lefschetz theorems, group schemes, topoi and étale cohomology, ℓ-adic cohomology and L-functions, then intersection theory and Riemann–Roch. Each subject demanded definitions that could bear transport from one problem to another. Each depended upon pieces of the scheme-theoretic language being written simultaneously in EGA. The seminar was where the language met resistance. EGA was where much of the foundational grammar was stabilized. Neither description is complete without the other.33Grothendieck and Dieudonné, “Éléments de géométrie algébrique II: Étude globale élémentaire de quelques classes de morphismes”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, première partie”; Alexander Grothendieck and Jean Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, seconde partie,” Publications Mathématiques de l’IHÉS 17 (1963): 5–91, https://www.numdam.org/item/PMIHES_1963__17__5_0/; Alexander Grothendieck and Jean Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, première partie,” Publications Mathématiques de l’IHÉS 20 (1964), https://www.numdam.org/item/PMIHES_1964__20__5_0/; Alexander Grothendieck and Jean Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, seconde partie,” Publications Mathématiques de l’IHÉS 24 (1965), https://www.numdam.org/item/PMIHES_1965__24__5_0/; Alexander Grothendieck and Jean Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, troisième partie,” Publications Mathématiques de l’IHÉS 28 (1966), https://www.numdam.org/item/PMIHES_1966__28__5_0/; Alexander Grothendieck and Jean Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, quatrième partie,” Publications Mathématiques de l’IHÉS 32 (1967), https://www.numdam.org/item/PMIHES_1967__32__5_0/; Grothendieck, Revêtements étales et groupe fondamental (SGA 1); Alexander Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2), Advanced Studies in Pure Mathematics 2 (North-Holland, 1968), https://webusers.imj-prg.fr/~leila.schneps/grothendieckcircle/SGA2.pdf; Michel Demazure and Alexandre Grothendieck, eds., Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3), Lecture Notes in Mathematics 151 (Springer, 1970), https://doi.org/10.1007/BFb0058993; Michel Demazure and Alexandre Grothendieck, eds., Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3), Lecture Notes in Mathematics 152 (Springer, 1970), https://doi.org/10.1007/BFb0059005; Michel Demazure and Alexandre Grothendieck, eds., Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3), Lecture Notes in Mathematics 153 (Springer, 1970), https://doi.org/10.1007/BFb0059027; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3; Luc Illusie, ed., Cohomologie ℓ-adique et fonctions L (SGA 5), Lecture Notes in Mathematics 589 (Springer, 1977), https://doi.org/10.1007/BFb0096802; Pierre Berthelot et al., eds., Théorie des intersections et théorème de Riemann–Roch (SGA 6), Lecture Notes in Mathematics 225 (Springer, 1971), https://doi.org/10.1007/BFb0066283.

In 1961 the institutional setting was still young. Léon Motchane had founded the Institut des Hautes Études Scientifiques in 1958. The institute’s own history describes its purpose as providing time for theoretical research outside an ordinary university timetable and lists Grothendieck as a permanent professor from 1958 to 1970. Its retrospective profile adds that Jean Dieudonné, recruited as the first permanent mathematician, made Grothendieck’s recruitment a condition of his acceptance. That story comes from the institution’s later account, not from a stenographic record of the negotiation, and it indicates the mathematical alliance on which the early program depended.44Institut des Hautes Études Scientifiques, “Alexander Grothendieck, Permanent Professor from 1958 to 1970”; Institut des Hautes Études Scientifiques, “The Foundation”.

The surviving bibliography does not support a cinematic reconstruction with a particular blackboard, a fixed row of chairs, or a winter view of the Bois Marie estate. “Bois Marie” belongs to the published name of the series; it does not by itself establish where every early exposé was delivered or how many people heard it. Later witnesses describe Grothendieck lecturing with speed and clarity, and institutional photographs can establish buildings, but a biography must not turn such fragments into a composite afternoon. The documentary anchor is more austere: a sequence of named programs, exposé bylines, seminar-year labels, fascicles, and revised volumes.55Grothendieck, Revêtements étales et groupe fondamental (SGA 1); Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2); Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3; Illusie, Cohomologie ℓ-adique et fonctions L (SGA 5); Berthelot et al., Théorie des intersections et théorème de Riemann–Roch (SGA 6); O’Connor and Robertson, “Alexander Grothendieck”; Michael Artin et al., eds., “Alexandre Grothendieck 1928–2014, Part 1,” Notices of the American Mathematical Society 63, no. 3 (2016): 242–55, https://doi.org/10.1090/noti1336; Michael Artin et al., eds., “Alexandre Grothendieck 1928–2014, Part 2,” Notices of the American Mathematical Society 63, no. 4 (2016): 401–13, https://www.ams.org/notices/201604/.

That evidentiary limit directs attention away from a theatrical individual-creator account and toward a system of work. Grothendieck’s power in the seminar did not consist only in answering isolated questions. It consisted in deciding that a family of questions should be reorganized, naming the structures that made them comparable, and asking other mathematicians to inhabit the resulting language. An expositor had to fit a theorem and its proof into a chain of definitions, base-change statements, functorial constructions, and references. When it failed to fit, the friction might reveal that the language itself needed enlargement. The seminar made abstraction accountable to an audience.

The account must also resist the opposite romance: that a collective seminar was a perfect republic in which authorship dissolved. The title pages do not support it. They distinguish direction, collaboration, exposition, and editing. Grothendieck’s authority was exceptional; younger participants often worked inside programs he had set. Yet the published record also refuses the shorthand by which every theorem in SGA becomes “a theorem of Grothendieck.” Some exposés have named authors; whole volumes name co-directors; later editors performed indispensable reconstruction. The seminar was collective without being equal, and Grothendieck was central without being solitary.66Grothendieck, Revêtements étales et groupe fondamental (SGA 1); Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2); Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3; Illusie, Cohomologie ℓ-adique et fonctions L (SGA 5); Berthelot et al., Théorie des intersections et théorème de Riemann–Roch (SGA 6).

The first program: coverings without paths

The first numbered program, SGA 1, bears the seminar years 1960–61. Its theme, étale coverings and the fundamental group, addressed an old topological desire in a new algebraic setting. On a connected topological space, covering spaces and the behavior of loops can be organized by a fundamental group. An algebraic variety over an arbitrary field need not supply usable paths, and over a finite field there is no complex manifold whose loops can simply be borrowed. The strategy was to reverse the emphasis. Instead of beginning with paths and deriving covers, begin with the category of finite étale covers and ask what group is encoded by all of them.77Grothendieck, Revêtements étales et groupe fondamental (SGA 1).

Let $\bar x$ be a geometric point of a connected scheme $X$. Every finite étale cover of $X$ has a finite fiber over $\bar x$. Morphisms of covers give maps between those fibers. The automorphisms of this fiber functor form a profinite group, written $\pi_1^{\acute et}(X,\bar x)$, and finite étale covers correspond to finite sets with a continuous action of that group. This is a Galois theory of coverings. Its force lies in the categorical reversal: the group is reconstructed from the objects it acts upon.88Grothendieck, Revêtements étales et groupe fondamental (SGA 1).

The theory records every finite unramified cover at once and recovers the symmetry common to their fibers. The analogy with ordinary topology illuminates why “fundamental group” is an apt name. It stops being reliable if it suggests that the algebraic group records all possible homotopy. It is profinite, tuned to finite étale covers; the choice of geometric base point matters up to the familiar inner ambiguity. Its arithmetic content can be present even when there is no path to draw. This precision was characteristic of the new architecture: intuition was welcomed, provided its boundary was stated.

The program also forced the seminar to deal with descent. A cover is useful only if objects and morphisms constructed after passing to it can sometimes be recognized as coming from the base. Faithfully flat descent, already developed in Grothendieck’s Bourbaki sequence, supplied a systematic answer for important categories. Thus the title of SGA 1 is narrower than the infrastructure required to make it work. The fundamental group, étale morphisms, fiber functors, and descent were not detachable topics placed side by side. They formed a chain: cover the space, compare what appears over overlaps, glue when effectivity permits, and read symmetry from the entire category.99Grothendieck, Revêtements étales et groupe fondamental (SGA 1); Grothendieck, “Technique de descente et théorèmes d’existence en géométrie algébrique. I. Généralités. Descente par morphismes fidèlement plats”.

The later book credits Grothendieck together with Michèle Raynaud. That name needs care because histories frequently collapse Michèle Raynaud and Michel Raynaud into one person or silently masculinize the initial. The documentary rule is simple: retain the name and role printed in the volume. It is equally important not to infer from the 1971 credit that every stage of the 1960–61 work had the final form visible in the book. The credit belongs to the extant volume; the seminar date belongs to the research season it records.1010Grothendieck, Revêtements étales et groupe fondamental (SGA 1).

Already the seminar’s characteristic scale was visible. A familiar word—covering— was generalized only after its surrounding structure had been rebuilt. The program had to establish what counted as unramified, how covers behaved under base change, which categories possessed enough limits or exactness, what notion of point was appropriate, and which local objects descended. A theorem about one curve could not answer these questions. A language capable of asking them uniformly could. This is why the apparent preliminaries grew. They were not ornamental foundations laid before “real” results. They were the machinery by which results from different regions could become instances of one method.

A second year, a different kind of locality

The title of SGA 2 records the seminar year 1961–62: Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux. Its revised book appeared in 1968, three years before the revised SGA 1, a reminder that numbered programs and publication order do not coincide. SGA 2 was not the second installment of a prewritten encyclopedia advancing toward an inevitable endpoint. It was a research program with its own problems, and its production history followed its own pace.1111Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2).

Local cohomology asks what part of a sheaf’s cohomology is supported on a specified closed subset. If ordinary sections answer what exists globally, sections with support insist that the answer live near a chosen locus; their derived functors measure the higher obstructions concentrated there. In the geometry of hyperplane sections and singularities, such information feeds Lefschetz-type theorems, which compare a space with a sufficiently well-behaved section. SGA 2 investigated local and global forms of those comparisons for coherent sheaves.1212Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2).

The vocabulary “local” could therefore mean more than one thing in the seminar years. In descent it meant passage to a cover and compatibility on overlaps. In local cohomology it meant support near a closed subset. In the emerging étale theory it would mean locality for a new Grothendieck topology. These meanings are related by sheaf-theoretic habits but should not be collapsed. Grothendieck’s architecture did not discover a single mystical principle called locality. It supplied formal settings in which several exact local-to-global questions could be separated and then compared.

The published SGA 2 added a separate exposé credited to Michèle Raynaud. Again, the bibliographic fact is small but consequential. It shows that the volume’s finished form was a later construction rather than the frozen speech of its director. The same work could pass through a seminar, receive written elaboration, and acquire an additional exposé before publication. In the mythology of a lone creator, such additions become footnotes. In the actual architecture, they are part of how a research program becomes transmissible.1313Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2).

During the same period, EGA was appearing in Publications Mathématiques de l’IHÉS. EGA II and the first part of EGA III both carry the date 1961. The second part of EGA III followed in 1963. These are publication dates, not labels for later editions: the fascicles entered the literature while the early SGA programs were under way. EGA II treated global elementary properties of classes of morphisms. EGA III treated the cohomology of coherent sheaves. The proximity of its subject to SGA 2 did not make the two texts duplicates. EGA established a systematic reference line; the seminar pursued a concentrated research front.1414Grothendieck and Dieudonné, “Éléments de géométrie algébrique II: Étude globale élémentaire de quelques classes de morphismes”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, seconde partie”.

The title pages named Grothendieck and added, in French, “with the collaboration of Jean Dieudonné.” That formula is more exact than either of two common caricatures. It does not make the men equal coauthors in the printed designation. It also does not make Dieudonné a typist who merely regularized another man’s thoughts. His role in the redaction, organization, and proof culture of EGA was integral to the work’s form. The title-page hierarchy is evidence; so is the sustained appearance of the credit across fascicles. A responsible biography keeps both facts in view.1515Grothendieck and Dieudonné, “Éléments de géométrie algébrique II: Étude globale élémentaire de quelques classes de morphismes”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, troisième partie”.

The parallel workload is difficult to grasp if one counts titles alone. Each EGA fascicle depended upon a web of definitions and prior propositions. Each SGA program could expose missing foundations that had to be supplied elsewhere. Publication did not wait for the whole projected EGA to be complete; seminar research did not wait for every reference to reach its final edition. Participants worked in a landscape whose roads were being paved while traffic already moved over them. Cross-references, recalled hypotheses, and long foundational detours were the cost of that simultaneity.

Group objects over a moving base

SGA 3 occupied the seminar years 1962–64 and was later published in three volumes, LNM 151, 152, and 153, all in 1970. Its direction was credited to Michel Demazure and Grothendieck. The subject was group schemes. Here the relative viewpoint acquired a particularly concrete test. A group need not be a fixed set of symmetries standing outside geometry. It could itself be a scheme over a base $S$, with multiplication, identity, and inverse given by morphisms over $S$. Its fibers could change as the point of the base changed.1616Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3).

A group scheme is a group object in the category of schemes. That concise sentence is not an attempt to hide the familiar group axioms. It translates them into commutative diagrams. Associativity says that two composites from $G\times_S G\times_S G$ to $G$ agree. Identity and inverse are likewise morphisms satisfying diagrammatic equalities. The gain is functorial: after any base change $S'\to S$, the pulled-back object $G_{S'}$ remains a group scheme.1717Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3).

The elementary intuition is a family of groups parameterized by the base. It must be used carefully. Knowing each fiber as an abstract group does not determine the total scheme. Nilpotent structure, specialization, flatness, smoothness, and the way fibers fit together carry information invisible in a list. The seminar’s insistence on the relative object was precisely an insistence that variation is part of the thing being studied. A bad fiber was not necessarily an exception to remove; it could reveal the geometry of degeneration.

Group schemes joined algebraic geometry to arithmetic in ways ordinary algebraic groups over a field could not exhaust. Finite group schemes behaved differently when their order met the characteristic. Tori, groups of multiplicative type, abelian schemes, and reductive groups demanded distinct structure theorems. Questions of descent returned because a group that acquired a simple form after a cover had to be recognized over the original base. Classification and construction were thus again inseparable from local-to-global technique.

The three-volume scale of the eventual publication conveys the size of the program but obscures the seminar’s risk. In 1962 there was no guarantee that every structural path would lead to a finished chapter. The later reader sees a taxonomy. The participants faced questions whose proper statements were being chosen. The act of placing group laws inside schemes multiplied the cases that a classical statement could ignore. Generality did not shorten the work at first. It enlarged the duty of proof.

Demazure’s printed role as co-director is therefore not decorative. The title disciplines retrospective attribution. One can say that Grothendieck’s relative vision set a powerful agenda and that he directed a school around it. One cannot therefore append his name alone to every argument in three volumes. Where an exposé has a byline, the byline governs. Where a theorem has a traceable prior source, that source governs. Where the history of an idea moved through discussion and revision, the biography should describe the movement rather than invent a moment of solitary discovery.

This rule is not pedantry applied from outside mathematics. It corresponds to the very architecture the seminar taught. A claim is not secure because it sounds plausible or because it fits a powerful person’s style. It is secure when its maps commute, its hypotheses are visible, and its dependencies can be followed. Credit deserves the same discipline.

The foundation arrives in fascicles

While SGA 3 unfolded, EGA III’s second part appeared in 1963. EGA IV then began its serial publication: the first part in 1964, the second in 1965, the third in 1966. The fourth did not appear until 1967. This sequence is one of the clearest protections against retrospective compression. By the Moscow congress in August 1966, three parts of EGA IV were in print; the fourth was not. The seminar could use concepts under active development, but the biographer may not cite the eventual four-part whole as though it had been physically available from the start.1818Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, troisième partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, quatrième partie”.

EGA IV’s subject, the local study of schemes and morphisms, sounds narrower than the cohomological projects surrounding it. In practice it supplied much of their load-bearing vocabulary. Properties of morphisms—flat, smooth, unramified, étale, proper, of finite presentation—had to be defined so that they behaved under composition and base change. Fiberwise statements had to be related to statements in neighborhoods. Dimension, specialization, constructibility, and openness had to be controlled. Without that discipline, the phrase “vary in a family” remained a promise rather than a theorem.

This foundation was not neutral bookkeeping. Choosing a definition changed which theorems could be stated economically. If a property was stable under arbitrary base change, it could survive the relative maneuvers on which the program depended. If it was local for a chosen topology, it could be checked after a cover. If it was compatible with limits, finite-presentation arguments could connect immense general objects to manageable ones. The seminar’s ambitions continually pressed on these design decisions.

The recurring word “machine” should not imply automatic proof. A mathematical machine is an arrangement of lemmas, functors, and compatibility theorems that turns a class of inputs into a class of outputs. Constructing it can be harder than solving the first problem that motivated it. Its value appears when new problems enter. Grothendieck repeatedly accepted the initial cost. The seminar distributed that cost across definitions, examples, special cases, and exposés. EGA attempted to make the investment permanent.

There was a danger in this method, perceptible even to sympathetic readers. The foundational chain could become so long that motivation disappeared behind references. A participant might master the formal route while forgetting why a particular cohomology theory was needed. The best seminar exposition had to perform two opposite tasks: insist on exact dependencies and keep the motivating problem visible. Later recollections that praise Grothendieck’s clarity are valuable partly because they indicate that, in person, he could sometimes restore the movement that the printed architecture concealed. They remain recollections, however, and cannot authorize invented lecture scenes.1919O’Connor and Robertson, “Alexander Grothendieck”; Leila Schneps, ed., Alexandre Grothendieck: A Mathematical Portrait (International Press, 2014); Artin et al., “Alexandre Grothendieck 1928–2014, Part 1”; Artin et al., “Alexandre Grothendieck 1928–2014, Part 2”.

The physical fascicle also had a social meaning. Before a revised SGA volume could stand on a library shelf, typed or duplicated notes could circulate among people equipped to correct and use them. Distribution was limited, versions could differ, and a result known in the network might remain awkward for an outsider to cite. Later revision increased durability at the price of delay. The modern reader, with a stable PDF or reprint, encounters a single surface where the historical actors encountered layers.

That layering helps explain why the seminar should be treated as an institution inside the institution. IHES supplied time, visitors, and an intellectual address. The SGA program supplied a rhythm of collective obligation. A proposed language had to be teachable; a proof had to be writable; a body of notes had to survive its season. Grothendieck’s personal intensity mattered, but intensity alone does not produce twelve published volumes. The work required a chain of custody for ideas.

A site large enough for arithmetic

SGA 4, bearing the seminar year 1963–64, made the chain more ambitious. The printed title names the theory of topoi and the étale cohomology of schemes. The revised publication credits Michael Artin, Grothendieck, and Jean-Louis Verdier as directors; its volumes appeared only in 1972, 1972, and 1973. The delay is not a side note. Concepts developed during the seminar continued to be reworked, and the published object embodies a later stage of the theory.2020Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

The underlying problem had been visible for years. Coherent sheaves on the Zariski topology, used with great success by Serre and then in EGA, did not by themselves provide the finite constant-coefficient cohomology imagined in Weil’s program. The Zariski topology was too coarse for that purpose. One needed more local neighborhoods—but algebraic varieties over arbitrary fields did not come with an ordinary finer topology that would do the job. The response was to generalize the notion of topology itself.2121Grothendieck, “The Cohomology Theory of Abstract Algebraic Varieties”; Serre, “Faisceaux algébriques cohérents”; Serre, “Géométrie algébrique et géométrie analytique”.

A Grothendieck topology declares which families of maps in a category count as covers. A site is a category supplied with that declaration; sheaves are functors whose local sections glue with respect to it. The category of sheaves forms a topos. In the étale site, the local objects over $X$ are étale maps $U\to X$, and jointly surjective families provide covers in the usual small-site formulation. This chapter need not anticipate the full explanation of Chapter 33. What matters here is the seminar mechanism: a targeted arithmetic problem forced a redefinition of the arena in which “local” and “sheaf” were understood.

Artin’s Harvard seminar notes on Grothendieck topologies date from spring 1962, before the SGA 4 seminar year. They demonstrate that the transmission network did not run only from a blackboard at IHES to a later Springer volume. Ideas crossed the Atlantic, were taught in another institutional setting, and returned in a larger collaborative formalism. It would be wrong to treat Artin’s mimeograph as an early printing of SGA 4. It is a distinct documentary object and evidence of the theory’s movement.2222Michael Artin, Grothendieck Topologies: Notes on a Seminar (Harvard University, Department of Mathematics, 1962).

Verdier’s presence was equally structural. Derived categories and the organization of derived functors were becoming indispensable to duality and cohomology. Their mature published form should not be projected wholesale into the first EGA fascicles, but the developing language shaped the work that SGA 4 sought to assemble. Here again collaboration did more than divide pages. Different mathematical strengths altered the architecture of the shared object.2323Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3; Robin Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64, Lecture Notes in Mathematics 20 (Springer, 1966), https://doi.org/10.1007/BFb0080482.

The topos was a particularly revealing product of the method. A narrow solution might have defined just enough étale sheaves to compute a desired group. SGA 4 instead studied the general environment in which many notions of covering and sheaf could live. The wider theory gave comparison maps, morphisms of topoi, cohomological tools, and set-theoretic conventions a common home. This generality was not proof that every possible generalization was useful. It was a wager that the correct ambient category would make essential operations natural.

Set-theoretic size entered because categories of all relevant objects can be too large for casual manipulation. Grothendieck universes supplied a convention for choosing a horizon of sets and then speaking of small and large categories relative to it. Later logicians and category theorists would find other uses for topoi, but those later developments must not be smuggled into the motives of 1963. Within this story, the immediate pressure came from geometry, sheaves, and cohomology.2424Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1.

An annual numbering that was not annual

The familiar list of SGA titles creates an apparent march from 1 to 7. The calendar is less tidy. SGA 3 and SGA 4 overlap in their printed seminar ranges. There is no numbered SGA volume corresponding simply to 1964–65. SGA 5 bears 1965–66. SGA 6 begins in 1966–67. The numbering identifies programs, not a perfectly regular annual course catalog. The gap and overlaps should remain visible because they show several fronts being developed and redacted at once.

SGA 5, Cohomologie ℓ-adique et fonctions L, records the 1965–66 seminar. Its revised volume did not appear until 1977, edited by Luc Illusie. A twelve-year distance separates the seminar label from publication. No responsible narrative can treat the 1977 table of contents as a minute-by-minute account of what stood proved in 1965. The book is primary evidence for the program and for its eventual formal statement; questions of exact interim knowledge require earlier fascicles, correspondence, and contemporaneous talks.2525Illusie, Cohomologie ℓ-adique et fonctions L (SGA 5).

The subject brought the architecture into direct contact with arithmetic point counting. Finite étale coefficients, inverse systems, Frobenius actions, compact supports, trace formulas, and L-functions had to be made compatible. Beyond attaching groups to a variety, the program had to obtain operations and finiteness properties strong enough to translate the number of rational points into traces of an endomorphism. The seminar therefore tested the entire preceding sequence: schemes and morphisms from EGA, coverings and descent from SGA 1, sheaf and derived methods, and the étale topos of SGA 4.

That test did not end with a proof of everything. Dwork had already proved rationality of zeta functions in 1960 by another method. The general weight statement was still unproved in 1966; Deligne established it in 1974. The seminar’s importance lies partly in the fact that it built a machine whose powers and missing parts could both be stated. An unfinished program can be a major mathematical achievement when it changes the form of the remaining problem.2626Bernard Dwork, “On the Rationality of the Zeta Function of an Algebraic Variety,” American Journal of Mathematics 82, no. 3 (1960): 631–48, https://doi.org/10.2307/2372974; Pierre Deligne, “La conjecture de Weil I,” Publications Mathématiques de l’IHÉS 43 (1974): 273–307, https://www.numdam.org/item/PMIHES_1974__43__273_0/.

SGA 6, on intersection theory and Riemann–Roch, bears the years 1966–67 and appeared in 1971 under the direction of Pierre Berthelot, Grothendieck, and Illusie. It belongs chiefly to the next movement of the biography, but its beginning within 1966 prevents the Fields Medal from serving as a neat full stop. While an international congress prepared to honor work already accomplished, another seminar program was opening.2727Berthelot et al., Théorie des intersections et théorème de Riemann–Roch (SGA 6).

SGA 4½ requires an additional warning. Published in 1977 under Deligne’s editorship, it was a compilation and simplification within the orbit of SGA 4 and SGA 5; it did not record a seminar held in some fractional academic year. The whimsical number is useful shorthand only if its documentary status is remembered. Without that care, bibliography turns into fiction.2828Pierre Deligne, Cohomologie étale (SGA 4½), Lecture Notes in Mathematics 569 (Springer, 1977), https://doi.org/10.1007/BFb0091516.

What an exposé did

An exposé is more than a chapter and less than an unmediated voice. Its byline can identify the person responsible for the written or delivered treatment, but the underlying theorem may have a longer genealogy. A director can set the problem and supply arguments; an expositor can repair or reorganize them; another participant can contribute a lemma; a later editor can reconcile the text with developments that occurred after the seminar. The finished SGA corpora contain all these layers. The safe unit of attribution is therefore the particular claim in the particular document, not the prestige of the series as a whole.

This layered authorship resembles the mathematics of descent more than a heroic biography might expect. Local pieces are not erased when they glue. Their compatibilities are what permit a global object to exist. The analogy has a strict limit: human collaborators are not interchangeable restrictions of one mind, and social agreement is not a cocycle theorem. Yet the comparison helps identify what the seminar valued. An idea became part of a common structure when it could survive translation into another person’s exposition and still retain exact relations.

Grothendieck’s centrality appears in the scale and continuity of the programs. The same preoccupations—relative objects, base change, representability, local-to-global methods, cohomology—recurred across them. Their design indicates his ability to treat apparently separate obstacles as symptoms of a missing concept. His institutional authority then allowed collective work to be reorganized around that diagnosis. This concentration accelerated several linked research programs.

Authority carried costs. A program framed at such a level could determine which problems seemed central and which forms of proof counted as satisfactory. Younger mathematicians entered a language already marked by its director’s taste. The labor of exposition could become invisible once a theorem was absorbed into “Grothendieck geometry.” Later disputes about credit cannot be settled by imagining the seminar as either benign family or oppressive factory. The contemporary record is granular: titles, bylines, letters, editorial credits, and dates. A biography should remain equally granular.

The published Grothendieck–Serre correspondence illustrates a different relation. Serre was not an apprentice assigned an exposé. He was a sustained interlocutor, able to propose examples, question general claims, communicate arithmetic pressures, and answer with results of his own. Letters could move faster and more privately than seminar volumes. They reveal mathematics before the impersonal voice of a treatise has removed hesitation and address. Yet even correspondence must be handled exactly: a dated letter establishes what that letter says, not everything either man privately knew.2929Alexander Grothendieck and Jean-Pierre Serre, Grothendieck–Serre Correspondence: Bilingual Edition, ed. Pierre Colmez and Jean-Pierre Serre, trans. Catriona Maclean (American Mathematical Society; Société Mathématique de France, 2004), https://bookstore.ams.org/view?ProductCode=CGS.

Harvard supplied another mode. Artin’s 1962 notes transmitted Grothendieck topologies; the 1963–64 seminar on Grothendieck duality was later written by Robin Hartshorne as Residues and Duality and published in 1966. These records make the geographical phrase “the Grothendieck seminar” insufficient. The center at IHES generated programs, but the network extended through visitors, universities, letters, and redactions elsewhere.3030Artin, Grothendieck Topologies: Notes on a Seminar; Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64.

Many problems, many entries

Identified witnesses attributed the seminar’s attraction partly to Grothendieck’s charisma and intellectual force. The documentary structure suggests an additional, bounded explanation: the program offered problems large enough to contain many entries. One person might develop a comparison theorem, another a finiteness result, another the language of derived categories, another a difficult class of group schemes. A common formalism allowed those efforts to meet without becoming identical.

Generality in this setting widened the range of possible contributions, though not effortlessly. A statement relative to an arbitrary base could admit arithmetic and geometric specializations. A functorial construction could be carried by another mathematician into a case the originator had not calculated. A categorical definition could make the same diagram meaningful for several kinds of object. Entry still required training, and long reference chains remained difficult. The structure’s potential reach did not make it easy to read.

Grothendieck’s seminar practice also challenged a narrow idea of communication. Popular exposition often seeks to remove technical apparatus so that a result can be seen at once. SGA communication added apparatus until the exact route could be repeated. It treated definitions, conventions, and compatibility proofs as shared equipment. This can produce pages that appear impersonal, yet the impersonality had a social aim: no reader should have to trust a private intuition that could not be transported.

This chapter does not use retrospective natural or building metaphors attributed to Grothendieck without an exact original locator. The contemporary books instead show logical dependence directly: remove a base-change theorem or a descent result and a distant cohomological argument may fail. Their structure matters because later claims carry the weight of earlier ones.

That transfer was temporal as well as logical. Work from the 1960–61 seminar could receive its durable edition in 1971. Work from 1963–64 could wait until 1973. Participants acted before a stable book existed; later generations learned from the stable book and easily forgot the action. The seminar was the bridge between those conditions. It made provisional knowledge public enough to be tested and eventually formal enough to outlive its first public.

Versions, corrections, dependencies

The distance between seminar and book changes what “primary source” means. A revised SGA volume is primary evidence for the mathematical text it contains and for the credits printed on it. It is not automatically primary evidence that every sentence was spoken in the seminar year. The distinction resembles the difference between a dated letter and a collected edition. One records an utterance at a time; the other may give a more accurate final text while incorporating knowledge obtained later. For intellectual history, greater mathematical polish can mean less chronological transparency.

This does not make the revised volumes deceptive. Their purpose was mathematical use, not the preservation of every false start. A research book properly removes errors, strengthens hypotheses, and coordinates notation. The danger arises when a later historian mistakes editorial success for instantaneous origin. SGA’s authority depends partly on the revisions that separate it from the live event. To honor the work is to preserve that process, not to deny it.

The exposé structure made dependencies visible in a distinctive way. A theorem might refer backward to an earlier exposé for a general lemma, sideways to EGA for a property of morphisms, and forward in practice to a result not yet available in its final form. Long chains of reference can frustrate readers, but they map the architecture. They show which parts of the machine carry which load. A shorter paper may hide the same dependencies under phrases such as “standard arguments.” SGA often preferred to build the standard.

That preference affected mathematical risk. Once a proposition was installed as a foundation for later exposés, an unnoticed defect could travel. Correction therefore required more than changing one line. Definitions, numbering, and downstream proofs might need adjustment. A seminar audience could test a local argument; redaction had to test the network. Publication delay included the time required to turn interdependent research into a reference system.

There was no single threshold at which provisional notes became “the theory.” Mathematicians could use a result from an early fascicle, cite a seminar announcement, receive an argument in correspondence, or wait for a revised book. Different communities therefore possessed different levels of access. Proximity to the network was an epistemic advantage. A visitor who had heard an exposé or received notes could work with ideas unavailable to a distant graduate student. The eventual books widened access, though their density imposed another barrier.

The history of SGA is accordingly also a history of media. Blackboard exposition allowed rapid adjustment but depended on presence. Duplicated fascicles traveled but could be scarce and unstable. Journal fascicles such as EGA carried formal publication dates and wide bibliographic identities. Lecture Notes volumes gave seminar work a durable commercial form years later. Today’s digital scans collapse the effort of finding these objects, but they do not collapse the dates without distorting them.

A line of biography should therefore identify which medium supports it. “The seminar treated group schemes from 1962 to 1964” rests on the SGA 3 title record. “The three revised volumes appeared in 1970” rests on their imprints. “A participant knew a particular corrected theorem in 1963” would require a dated fascicle, letter, or testimony. Without such evidence, the first two statements do not entail the third. This modest logic prevents a large class of confident inventions.

The same care governs terminology. Later editions may replace an early word, expand a convention, or adopt notation stabilized only after the seminar. EGA itself moved from the language of préschémas toward the modern use of schémas. A clean modern explanation can use current terms, but a historical sentence must not imply that every participant used them in precisely that sense at every date. Translation is necessary; anachronism is not.

From problem to program

Grothendieck’s distinctive move was often to convert a difficult theorem into a program of prerequisites. This can look like avoidance if one sees only the distance from the original question. In fact it changed the distribution of difficulty. Suppose a point-counting conjecture appears to demand an unknown cohomology theory. One response is to search directly for a clever formula. Another is to ask what properties any satisfactory theory must possess, what category of local data could support it, what finiteness and comparison theorems would make it computable, and what functorial operations a trace formula would require. The second route replaces one wall with a landscape of roads.

The seminar was suited to that replacement because roads can be built in parallel. One exposé could establish a base-change theorem while another clarified a class of coefficients. A sequence could begin with categorical foundations and end with an arithmetic application. Examples could reveal that a proposed general statement needed a constructibility hypothesis. The audience participated in determining what would count as a sufficient system.

The program form also preserved incompletion. A direct attack either proves a theorem or does not. A large architecture can succeed unevenly. It may establish rationality, duality, or comparison while leaving a weight estimate out of reach. The remaining gap becomes sharper because the rest of the machine isolates it. This was the situation of the Weil program by 1966: the cohomological language had transformed the problem, yet the decisive general estimate on Frobenius eigenvalues was not possessed.

To say that the seminar “aimed at the Weil conjectures” is thus true but incomplete. Not every SGA program was a subroutine designed solely for that destination. Group schemes had their own arithmetic and geometric questions. Local cohomology and duality answered needs far beyond point counting. Descent and representability reorganized moduli. The Weil conjectures provided a powerful horizon, not the only reason each road mattered.

This distinction protects collaborators from another form of erasure. If every technical result is valued only as a step toward the most famous theorem, the expositor’s actual problem disappears. The historian then knows the destination and forces each actor to face it. A seminar program is better read at two scales: through the local theorem as its author understood it, and through the larger architecture that made unforeseen later uses possible.

Grothendieck’s generality was productive partly because it declined to decide every future use in advance. A representable functor, a stable base-change property, or a topos could travel. This portability differs from vagueness. The more widely a concept was meant to travel, the more exact its compatibility conditions had to become. Universality did not mean saying less. It meant specifying the relation to every test object.

The functor-of-points habit captures that ethic. It replaces a question about the points an object contains over one field with a question about the families of points it assigns to every test scheme and how those assignments change under maps. The answer may characterize the object. Socially, the seminar made an analogous demand: an idea had to remain coherent when another mathematician pulled it into a new context. The analogy again has a boundary, but it identifies why exposition was part of discovery. A construction that could not survive transport was not yet in its strongest form.

The programs also trained taste. Participants learned to ask whether a statement was relative to a base, whether it was local for an appropriate topology, whether it commuted with base change, whether the object was representable, and whether the proof was functorial. These questions were not neutral checkboxes. They expressed a view of geometry in which relations and variation were primary. The seminar made that view habitual across a generation.

Work at different speeds

Grothendieck’s mathematical production during these years cannot be measured by a list of papers. Conventional journal articles were only one channel. EGA fascicles, SGA programs, Bourbaki exposés, correspondence, and lectures formed a coupled system. A theorem announced in one place might receive foundations in another and its durable proof in a third. Counting only publication dates would underestimate the live work; treating every later volume as contemporary would overstate what was settled.

EGA’s serial rhythm offers the cleanest spine. Two fascicles appeared in 1961, one in 1963, then one in each of 1964, 1965, and 1966. That is already a formidable printed record. Alongside it ran the seminar sequence. The two tracks demanded different kinds of attention. EGA sought continuity and exhaustive foundations. SGA could concentrate on a frontier, assume some common background, and recruit a group around a theme. Their interaction allowed the program to move faster than a single completed treatise would have allowed, while still pursuing the ideal of a durable foundation.

Dieudonné’s collaboration was crucial to this division of labor. The printed formula does not establish the exact origin of every sentence, and a responsible account should not pretend it does. It establishes that EGA was not produced by Grothendieck alone. Later testimony about Dieudonné’s redaction helps explain the work’s consistency, but the strongest evidence remains in the sustained mathematical object: thousands of linked propositions carrying a common notation across years.

SGA solved the labor problem differently. Direction was joined to assignment and exposition. A participant might be responsible for mastering a region sufficiently to present and write it. That process distributed work, but it also created dependence on the director’s program and standards. The resulting education was intense. To write an exposé was to learn not only a theorem but the network of definitions in which it belonged.

Later recollections can make this system sound effortless because they select its most dazzling moments. The books show the opposite. Their length, revisions, and publication gaps are material traces of effort. Even unusually rapid mathematical insight does not typeset itself, reconcile notation, check references, or prepare an index. Collective intelligence is not a cloud of inspiration. It is organized work.

The phrase also should not romanticize unpaid or uncredited labor. This biography follows explicit credits because personal contribution can be lost when a school is named after its leader. The available documentation is itself uneven: some conversations left no record, and some editorial interventions are more visible than others. Where evidence stops, the text must stop too. A visible gap is preferable to a compensating legend.

At the same time, refusing unsupported claims should not flatten the evident asymmetry of creative direction. Grothendieck selected extraordinarily fertile problems and supplied concepts that repeatedly reorganized them. The recurrence is documented across programs. His leadership was not a later brand attached to unrelated volumes. It was a mathematical force. Accurate collective history can say this without making every proof his property.

What was visible in 1966

Imagine not a room but a shelf at the end of 1966. On it could stand EGA II, both parts of EGA III, and the first three parts of EGA IV. Nearby one might find early SGA fascicles or seminar notes, depending on access, but not the revised Springer books by which later readers know SGA 1, 3, 4, 5, and 6. The North-Holland edition of SGA 2 followed in 1968. Hartshorne’s Residues and Duality, however, was published in 1966 from the Harvard seminar of 1963–64, and Grothendieck’s algebraic de Rham paper also appeared that year.3131Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64; Alexander Grothendieck, “On the de Rham Cohomology of Algebraic Varieties,” Publications Mathématiques de l’IHÉS 29 (1966): 95–103, https://www.numdam.org/item/PMIHES_1966__29__95_0/.

This documentary shelf prevents two errors. The first is to underestimate the period by saying that the major SGA books were “not yet published,” as though no working theory existed. Seminar participants and correspondents had notes, talks, and arguments. The second is to overestimate it by placing the revised books on the shelf early. Working knowledge and stabilized publication were both real, but they were not the same.

Mathematically, the visible architecture was extensive. Schemes and relative morphisms had a substantial printed foundation. The algebraic fundamental group, faithfully flat descent, group schemes, Grothendieck topologies, topoi, and étale cohomology had active seminar forms. The ℓ-adic and trace program was being developed. Yet several later syntheses were not finished. The general Weil Riemann hypothesis remained open. The standard conjectures had not appeared in their later printed formulation. The six-operation formalism had not reached every form in which subsequent generations would use the phrase.

The seminar’s achievement should therefore be described in the present progressive, not the retrospective perfect. An architecture was being built. Some rooms were usable; some plans changed; some connections became visible only when later mathematicians entered. The 1966 Fields Medal would recognize a revolution already plain to the profession. It could not convert the open doors into finished walls.

This temporal stance changes the portrait of Grothendieck. He was not standing above a completed system, surveying what others would merely apply. He was inside a collective construction, directing parts of it, writing another part with Dieudonné, corresponding about weaknesses, and moving toward questions the existing language could not yet answer. Recognition and incompletion coexisted.

Apprenticeship by obligation

A research seminar teaches differently from a course. A course normally begins with a body of knowledge whose order has been chosen for learners. An SGA exposé could begin with a question whose correct statement was still part of the research. The participant responsible for it had to learn enough background to decide what could be asserted, and enough of the surrounding program to make the assertion usable.

This produced apprenticeship by obligation. Responsibility for a written exposé forced a young mathematician to convert understanding into public structure. Definitions had to match earlier installments. A proof sketch had to be completed or its gap named. Examples had to obey the relative hypotheses. The learner became a contributor because the program needed a reliable node, not because a curriculum declared the training finished.

Grothendieck’s later witnesses describe his ability to recall prerequisites while moving quickly. That combination would have mattered in such a setting. Recall gave newcomers handles; speed kept the frontier visible. The testimony does not allow a transcript of any lecture, but it helps explain how formidable formalism could function orally before the books existed.

The obligation ran both ways. A director assigning a topic had to make its relation to the whole intelligible. If the expositor could not state the needed theorem, the program might have exposed a real missing foundation. Teaching was therefore a test of architecture. Confusion was not always a failure of the student; sometimes it was evidence that the road had not been built.

This training model also created a distinctive form of confidence. A participant who had written one exposé knew a region in depth and knew how it connected to EGA or another SGA installment. No one needed to master all volumes before doing research. The architecture permitted local expertise while the common language kept it interoperable.

Interoperability did not eliminate hierarchy. The director often selected the region and judged the treatment. A young expositor’s freedom operated inside an agenda already shaped by Grothendieck. The opportunity could be generous and the authority immense. Both conditions belong to an accurate account of apprenticeship.

A publication system outside the journal article

The seminar created a form between speech and monograph. A journal article normally isolates a result, limits preliminaries, and assigns a compact author list. SGA could spend an exposé on foundations, another on a comparison theorem, and a third on an application, while maintaining one program across them. The unit of publication was the architecture.

This form suited Grothendieck’s mathematics because general statements generate long dependency chains. It also suited a period when the field’s language was changing too quickly for a finished encyclopedia to precede research. Fascicles could circulate before a definitive volume. A later redaction could incorporate corrections without forcing every result into a separate paper.

The costs were real. Scarce fascicles privileged those near major libraries and seminar networks. A citation might point to notes difficult to obtain. Version differences could complicate verification. The eventual LNM books improved access for one generation while arriving too late to document live chronology transparently.

The SGA form also blurred conventional authorship. A program director’s name could dominate catalogs even when individual exposés had authors. Library metadata sometimes simplifies title-page roles. Modern databases may list a redactor as an author or omit contributors. Historical work must return to the physical title and table of contents rather than trust one normalized record.

At the same time, the form preserved more names than an unsigned school tradition would have done. Exposé bylines remain available for correction of the broad phrase “Grothendieck proved.” The document contains the evidence of its own collectivity, provided the reader looks below the series title.

SGA was therefore not an accidental container for results. It was a technology of research communication adapted to a foundational program. The gap between seminar and book is one of its design features and one of its historical hazards.

Counterexamples as members of the seminar

Generality advances through counterexamples as much as through definitions. A proposed statement stable over fields might fail over a nonreduced base. A property true for proper morphisms might collapse without properness. A quotient relation might have no scheme quotient. Such examples forced the seminar to identify the actual boundary of a theorem.

The scheme language multiplied these tests. Nilpotent elements, special fibers, and arbitrary base change brought cases classical variety theory could avoid. Critics could see this as pathological expansion. For Grothendieck, the difficult cases often revealed which formulation was natural. A theorem that survived them was more portable; a theorem that failed learned its hypotheses.

The seminar’s collective structure made counterexample work especially valuable. The director might see a universal pattern; another participant might know a local ring or group scheme that resisted it. Neither contribution was secondary. The general theorem acquired truth through the encounter.

This is another reason not to narrate the program as one mind unfolding a plan. A plan can specify where to look. It cannot contain every obstruction discovered by people working there. The architecture changed in response to its inhabitants.

The revised books often hide this drama by presenting the corrected theorem without the failed version. For mathematics that is a virtue. For history it means the counterexample may survive only in correspondence, early notes, or testimony. Where those sources have not been checked, the biography should not invent an instructive failure merely because one would fit the method.

Still, the polished statements bear indirect evidence. Their elaborate hypotheses record pressures. A condition such as quasi-compactness, constructibility, or separatedness is often the scar of a case that had to be controlled. Reading hypotheses historically can reveal work even when the conversation has vanished.

The seminar’s present tense

For participants, SGA numbers did not denote a complete shelf. SGA 1 could be a body of recent notes while SGA 3 was under way. EGA references could point to fascicles just printed. A result later assigned to SGA 4 might be known through an exposé or a visitor’s account. The program existed in the present tense of uneven access.

That condition changed how authority worked. Grothendieck’s oral statement or letter could carry weight before a citable proof was available. Trust in the speaker and network bridged the publication gap. Later redaction converted personal authority into documentary authority, but not instantly.

The conversion was ethically important. Mathematics intends proofs to be checkable independently of who states them. A seminar at the frontier necessarily operates with provisional trust. Producing the book repays that trust by making dependencies public. The enormous editorial labor was part of the field’s claim to objectivity.

It also allowed dissent to become durable. Once a statement was printed, another mathematician could identify a gap without having attended or belonged to the circle. The common language expanded the community of possible critics. Generality and publication together decentralized the work they initially concentrated.

This temporal movement—from personal communication to shared reference—offers a better account of Grothendieck’s influence than charisma alone. Charisma can attract an audience. A reference system enables that audience to become a field.

The limits of the seminar form

No production system is neutral. A theme large enough for SGA could absorb years of work and make alternative problems seem peripheral. The demand for maximal generality could delay a usable theorem. A participant skilled in concrete calculation might find the reference chains alienating. The seminar’s success should not be measured only by the descendants who adopted its language.

Some mathematicians learned the theory through more concise papers or later textbooks rather than the original volumes. That does not invalidate SGA, but it shows that creation and pedagogy can require different forms. The first architecture may be broader than the shortest route through it.

Grothendieck’s method also depended unusually on his sustained attention. The programs cohered because one director carried relations across years. That concentration was powerful and fragile. If his priorities changed, or if the institutional arrangement failed, no committee could automatically reproduce the same direction.

The unfinished EGA and delayed SGA editions reveal the fragility without erasing success. Total foundations continually receded as mathematics grew. The seminar could redirect work faster than the treatise could absorb it. Eventually the modular corpus, not the completed encyclopedia, became the enduring form.

This limit prepares the later biography. Grothendieck’s departure from IHES in 1970 would interrupt one institutional system, yet the languages developed there would continue through collaborators. The architecture had become independent enough to survive its principal director. In 1966 that test had not yet arrived.

The reader the seminar required

SGA did not imagine a reader who would absorb each page in isolation. It required someone willing to move among exposés, EGA references, and earlier theories of sheaves and categories. Reading became navigation. A proof’s meaning included the route by which its hypotheses entered.

This demand trained a new kind of algebraic geometer. Fluency meant more than calculating examples; it meant recognizing universal properties, changing bases, and moving between a site and its sheaf category. The reader learned to ask whether an operation was natural before asking for coordinates.

The training could exclude. Long prerequisite chains favored people with access to the seminar or patient mentors. Later textbooks would build shorter entrances. The original difficulty should not be romanticized as proof that only exceptional readers deserved the theory.

Yet the demanding reader was also empowered. Once inside the language, a mathematician could transport results across fields and test general statements on new bases. The architecture repaid apprenticeship with mobility.

Grothendieck’s seminar thus produced readers as well as theorems. Its lasting community consisted partly of people trained to see a problem as a missing functor, descent statement, or compatibility. That change of habit was one of the decade’s least countable results.

A seminar as an action

The noun seminar names a program, but the historical object was an action repeated across people and media. A question was exposed, pulled into a common language, tested, written, corrected, and exposed again. The later volume is the fixed trace of those verbs. Reading only the finished noun loses the movement that made it reliable.

Grothendieck’s central achievement in this form was not ownership of every step. It was the sustained capacity to keep the steps directed toward an architecture while allowing local results to alter the plan. A rigid plan would have broken under new examples; an undirected conversation would not have produced SGA.

The seminar’s history therefore lies between event and book. It cannot be recreated as a scene from title pages, and it cannot be reduced to bibliography. The recoverable truth is a disciplined process whose dates, names, and mathematical interfaces remain on paper.

Overlapping programs

The numbering SGA 3 and SGA 4 suggests a clean succession that the printed dates do not support. SGA 3 carries the years 1962–64, while SGA 4 carries 1963–64. During their overlap, group schemes and the foundations of topoi and étale cohomology belonged to concurrent programs rather than consecutive completed courses.3232Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

The overlap mattered mathematically. Group schemes supplied examples of torsors, descent, and arithmetic variation; sites and topoi supplied a setting in which such objects and their cohomology could be studied locally. Neither subject was a chapter waiting for the other to finish. Their interaction helped determine which general statements were needed.3333Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

It also mattered socially. Demazure shared direction of SGA 3, while Artin and Verdier shared direction of SGA 4 with Grothendieck. The concurrent dates therefore record concurrent centers of responsibility. A numbered-series summary that names only Grothendieck loses information present on the title pages.3434Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

Publication later removed the overlap from view. The SGA 3 books appeared together in 1970; the SGA 4 books appeared in 1972–73. Shelved by series number, they look sequential. Read by seminar date, they show a research environment carrying several fronts at once.3535Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

This concurrency sets a limit on causal narration. A concept used in both programs may have moved through talks, notes, or correspondence, but title dates alone do not establish its route. The record supports overlap and named direction; finer influence requires exposé-level evidence.

The overlap also changes how workload is interpreted. Two seminar labels covering the same year do not prove that every exposé was delivered on a fixed weekly schedule or that the same audience attended both. They establish concurrent named programs, not a reconstructed calendar.3636Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

This bounded conclusion is historically useful. It explains why ideas about group schemes, descent, and cohomology could develop in contact while preserving the distinct title-page credits. It also leaves room for original fascicles to refine the sequence without forcing later editorial order back onto 1963–64.

The concurrent programs also explain why a single “school” label remains insufficient. Shared vocabulary linked the work, but different directors, exposé authors, and publication schedules created separate lines of responsibility. The seminar’s unity was organized rather than homogeneous.3737Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

The next question

By the end of this sequence, cohomology was no longer one invariant attached to one kind of space. It was becoming an architecture of sites, sheaves, derived functors, supports, base change, duality, and traces. Yet one elementary-sounding question kept returning beneath the machinery: whether an object that existed after passage to a cover, with compatible copies wherever they could be compared, came from an object over the original base.

That question had already occupied Grothendieck’s Bourbaki exposés from 1959 through 1962. It entered SGA 1, moduli constructions, group schemes, and the definition of new topologies. It was at once technical and philosophical. To descend an object is to decide whether local presentations are merely compatible descriptions or pieces of a genuine whole. The answer depends on the kind of object, the cover, and the strength of effectivity available. No slogan can replace those hypotheses.

The seminar supplied the social analogue but not the proof. Many people could carry pieces of a program; bylines and dates preserved their differences; a common language made comparison possible. Its mathematical counterpart was descent. Before étale cohomology could count points or topoi could redefine space, geometry needed exact conditions for passing from compatible data on a cover back to the original base.