A New Foundation

An address in Edinburgh

In August 1958, Alexander Grothendieck addressed the International Congress in Edinburgh on The Cohomology Theory of Abstract Algebraic Varieties. The title looked back to a problem already old enough to possess its own vocabulary. Algebraic geometers wanted cohomology theories that would extract global invariants from varieties and, in particular, support André Weil’s program for varieties over finite fields. The address also looked past the category of objects named in its title. Its sixteen printed pages sketched a geometry in which the ambient language would have to be enlarged before the desired cohomology could be built. The talk was not the first appearance of sheaves, homological algebra, or arithmetic geometry. It stated that these strands now required a common foundation.11Grothendieck, “The Cohomology Theory of Abstract Algebraic Varieties”.

“Foundation” can mislead here. It may suggest a retreat from research to first principles, as though Grothendieck paused the real work in order to decide what objects properly existed. The movement was almost the reverse. The problems had multiplied faster than the available language. Riemann–Roch had shown how a functorial reformulation could expose the structure behind a classical theorem. Sheaf cohomology had made local-to-global reasoning systematic. Weil’s conjectures demanded a cohomology suitable for algebraic varieties in positive characteristic. Families, specialization, infinitesimal deformation, and reduction modulo primes kept pressing beyond the comfortable boundaries of varieties over one fixed field. A new foundation was not to stand beneath this work like a stone platform. It was to pass through the work like a grammar.22Alexander Grothendieck and Jean Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas,” Publications Mathématiques de l’IHÉS 4 (1960), https://webusers.imj-prg.fr/~leila.schneps/grothendieckcircle/EGA/EGAI.pdf; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The Edinburgh paper should therefore be read as an intermediate document. It did not contain the finished architecture of the coming decade. Some claims remained programs; some proposed routes would later change; several subjects announced for one treatise would migrate into seminars and different publications. The printed argument nevertheless enlarged the scale of the program: it arranged objects, maps, and operations so that a large family of results could be asked in one idiom. This is a statement about the document’s scope, not an inference about Grothendieck’s unrecorded intention.33Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck, “The Cohomology Theory of Abstract Algebraic Varieties”.

That ambition had a date-stamped practical consequence. On October 17, 1958, Grothendieck wrote to Jean-Pierre Serre that he had begun preparing short papers and commentaries for Jean Dieudonné, who, he said, had made a good start on writing up schemes. Serre’s editorial note in the published correspondence is unambiguous: the writing of EGA had just begun. Grothendieck hoped that the first four chapters could be drafted by spring and published by Léon Motchane. The timetable proved wildly optimistic, but the letter catches the project before later monumentality hardened around it. There were notes, a collaborator beginning to write, a publisher-institution taking shape, and a plan still fluid enough to absorb a projected joint paper with Serre.44Grothendieck and Serre, Correspondance Grothendieck–Serre, letter of October 17, 1958, pp. 73–75.

The projected book was already nicknamed the multiplodocus, a comic name proportionate to its threatened size. Humor did not make the labor light. The foundation would require definitions stable under repeated use; proofs whose hypotheses survived base change; notations that could carry hundreds of pages; and a publication system able to issue chapters while the whole remained unfinished. Its making joined conceptual invention to patient redaction. The join matters historically. Grothendieck supplied the governing program, new constructions, and much of the mathematical substance. Dieudonné turned plans, lectures, notes, and continuing exchanges into extended systematic exposition, checking and organizing the material as the collaboration advanced. Serre questioned formulations, supplied corrections, contributed earlier theorems and an alternative mathematical temperament, but did not coauthor EGA. Motchane made an institute and a journal available. None of those roles is adequately described by the single-author image.

What had to be inherited

The opening of EGA I, published in 1960, names its debts more candidly than many later accounts do. The dedication is to Oscar Zariski and André Weil. The introduction calls Serre’s Faisceaux algébriques cohérents (FAC) fundamental and says that it introduced more than one young reader, including one of EGA’s authors, to algebraic geometry. It discusses Claude Chevalley’s local-ring schemes and Masayoshi Nagata’s extension of algebraic geometry over Dedekind domains. It lists commutative algebra, homological algebra, sheaf theory, and functorial language among the reader’s essential equipment. The account begins from an existing mathematical network.55Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 5–8.

EGA I identifies both a problem and a prior standard in Weil’s work. Its introduction treats his 1946 Foundations of Algebraic Geometry as an unavoidable antecedent, while the bibliography records his finite-field program. The same pages credit Zariski’s local and birational methods, describe Chevalley’s local-ring schemes as naturally extensible, and place Nagata’s work over Dedekind domains among the routes beyond geometry over one field. This wording follows EGA’s own account rather than assigning the same role to each predecessor.66Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas,” 7–8.

Serre’s position was especially close. FAC, published in 1955, used coherent algebraic sheaves and their cohomology on algebraic varieties. GAGA, published the following year, compared algebraic and analytic geometry over the complex numbers. These papers did not define Grothendieck schemes. They established facts and habits without which the scheme-theoretic program is difficult to imagine: the Zariski topology could carry useful sheaves; cohomology could turn local algebra into global information; the relation between algebraic and analytic categories could be formulated through functors and proved by exact comparison. Grothendieck extended this terrain, but he did not retrospectively author it.77Serre, “Faisceaux algébriques cohérents”; Serre, “Géométrie algébrique et géométrie analytique”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 6–7.

The exchange between Serre and Grothendieck was not a relay in which one finished and the other started. Their published correspondence records proposals, objections, corrections, and divergent judgments. Serre might ask when the multiplodocus would actually appear, point to an error in a draft, or resist a claim that a new formalism had already dissolved an old difficulty. Grothendieck might absorb the correction, reject the emphasis, or place the question inside a still larger program. In 1959, for example, Serre was reading and commenting on chapters concerning projective morphisms while Grothendieck was simultaneously pursuing descent and existence theorems. The traffic ran both ways, but the projects remained distinct.88Grothendieck and Serre, Correspondance Grothendieck–Serre, letters of October 25 and 31 and November 5–15, 1959, pp. 96–117.

The word “Bourbaki” also belongs in this inheritance, provided it is not used as a magic cause. Bourbaki had made structural organization, explicit mappings, and reusable foundations part of a powerful collective practice. Dieudonné was a central Bourbaki author; Grothendieck participated in the group during these years; Serre and Chevalley were deeply involved. The Bourbaki seminar supplied a rapid public channel for results that could not wait for EGA’s long chapters. Yet EGA was not simply Bourbaki applied to geometry. Its driving questions came from active algebraic geometry, and Grothendieck’s readiness to alter categories of objects went beyond a preference for polished axioms. The relation was one of tools, personnel, standards, and tensions—not identity.

One can see this in the tone of EGA I. It expects the reader to know large parts of modern algebra, yet it claims in principle to presuppose no particular knowledge of algebraic geometry. It even warns that an overly exclusive habit of the birational viewpoint can hinder entry into the new methods. That sentence is polemical, but it is not a verdict that earlier geometry was obsolete. Birational geometry had generated indispensable problems and techniques; Zariski’s influence appears throughout the new work. The warning concerns exclusivity. A foundation adequate to fibers, nilpotents, arithmetic bases, and cohomology could not make the generic point of an irreducible variety the only privileged place from which geometry was understood.99Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, p. 5.

What the Edinburgh paper required

The Edinburgh paper did not yet present the mature scheme definition that belongs to Chapter 26. Its historical force lies elsewhere: it made several previously separate demands appear in one program. A cohomology for varieties over finite fields would need functoriality, products, cycle classes, comparison with familiar theories where comparison made sense, and a way to follow arithmetic endomorphisms. Each demand carried prerequisites that the short congress address could name without fully supplying.1010Grothendieck, “The Cohomology Theory of Abstract Algebraic Varieties”.

The paper therefore reads as a requirements document for later foundations. It asks for constructions stable enough to be reused across varieties and fields, while leaving open how the objects and maps should be defined. That distinction matters. A requirement can direct research before its solution exists, but it does not contain the later solution in finished form. The paper’s value is programmatic and documentary, not prophetic.

EGA I later points readers back to the Edinburgh address as preliminary orientation. The cross-reference establishes continuity between the public program and the treatise, yet it also records a change of genre. The congress paper compresses aims into sixteen pages. EGA expands definitions, dependencies, and proofs over hundreds of pages. The movement from one text to the other is itself the chapter’s subject.

The 1958 address also fixes a boundary on retrospective credit. Weil’s conjectures supplied a major target; Serre’s coherent sheaves and comparison results supplied examples of what sheaf methods could achieve; commutative and homological algebra supplied languages already in circulation. Grothendieck’s later architecture reorganized these materials, but the address did not erase their prior authorship.1111Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 5–8; Grothendieck, “The Cohomology Theory of Abstract Algebraic Varieties”.

The introduction as a map of prerequisites

EGA I’s introduction is unusually explicit about what readers must bring. It names commutative algebra, homological algebra, sheaf theory, and functorial language. The list is not a neutral syllabus. It shows that the proposed foundation depended on several modern subjects that had only recently received standard treatments.1212Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, p. 5.

The introduction also distinguishes prerequisite from ownership. Cartan–Eilenberg and Godement are cited for homological algebra and sheaves. Serre’s FAC is called fundamental. Chevalley’s local-ring schemes and Nagata’s work over Dedekind domains are acknowledged as antecedents. Zariski and Weil appear as unavoidable influences. These sentences make a single-origin story impossible without diminishing Grothendieck’s role in the synthesis.

What changed was the scale at which the tools were made to cooperate. A theorem in one setting could be reformulated so that its hypotheses survived a change of base; a local algebra construction could enter a global sheaf; an existence question could be expressed through a functor. Chapters 26–28 own the full explanations of scheme, relative geometry, and representability. Here the introduction supplies only the documentary fact that EGA was built to make such traffic systematic.

The historical terminology remains part of that evidence. EGA’s use of préschéma and schéma does not coincide with later English convention, and the shift must be signaled when a period statement is quoted. The detailed definition and examples are deferred to Chapter 26. The present chapter needs only the consequence: the treatise was introducing a language whose own vocabulary had not yet stabilized into modern textbook form.1313Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas,” I.

A program before its definitions

The printed sequence of proposed chapters provides a more reliable precision window than repeating the later definitions in advance. The plan moves from the language of schemes to global morphisms, coherent cohomology, local study, construction, descent, group schemes, differential questions, fundamental groups, duality, intersections, Picard schemes, and Weil cohomology. It is a map of dependencies and aspirations, not a table of results already possessed.1414Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 6–7.

The order explains why the foundation became so large. Definitions of objects were followed by classes of maps; those maps supported cohomology; cohomology and formal geometry entered existence theorems; existence questions led to descent and parameter spaces. The plan’s later fragmentation across EGA, the Bourbaki exposés, and SGA shows that the map changed while it was being followed.

This documentary sequence also clarifies Grothendieck’s contribution without assigning every later theorem to him. He set an unusually broad architecture, proved central results, and directed programs in which other mathematicians wrote, corrected, transmitted, and extended the theory. Dieudonné’s redaction and Serre’s criticism were not secondary decorations on a finished design; the correspondence shows them operating while the design was still moving.1515Grothendieck and Serre, Correspondance Grothendieck–Serre, 73, 96–117.

The full mathematical ladders now begin where the chapter plan assigns them. Chapter 26 explains the scheme, Chapter 27 the relative map and base change, and Chapter 28 the functorial existence problem. This chapter retains the dated need, the textual sequence, and the people who converted a program into publishable mathematics.

Writing in collaboration

The title page of EGA I is a small historical document. It attributes the work to A. Grothendieck and identifies J. Dieudonné as collaborator in its redaction. Bibliographies often cite them together, reasonably recognizing the sustained collaboration, but the title page preserves an asymmetry. Grothendieck directed the mathematical program and supplied core ideas, drafts, lectures, and decisions. Dieudonné undertook the extended writing and organization through which that program became a treatise.1616Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, title page.

The correspondence gives unusually concrete evidence. Grothendieck supplied papers, comments, and notes as Dieudonné drafted the opening material. By late 1959, the first chapters had moved to the printer, work on Chapter II continued, and Grothendieck described later portions as ready in note form. Serre returned detailed criticism on projective and quasi-projective morphisms; his later annotations report that several corrections entered EGA. The sequence shows iterative work with distinct roles in conceptual direction, checking, redaction, and criticism.1717Grothendieck and Serre, Correspondance Grothendieck–Serre, 73, 101–17.

Dieudonné’s contribution is often diminished by compliments that sound generous: extraordinary speed, phenomenal memory, tireless capacity. Those qualities mattered, but they can reduce intellectual labor to horsepower. To organize a new language across hundreds of propositions is to decide sequence, isolate dependencies, normalize notation, expose gaps, and turn an insight that works in a creator’s hands into an argument that another mathematician can verify. The surviving title page and letters warrant calling Dieudonné collaborator and rédacteur. They do not warrant assigning every lemma sentence by sentence without manuscript comparison.

Nor should Serre’s role be inflated into hidden coauthorship. His FAC and GAGA papers were direct mathematical sources. His letters and readings influenced formulations and caught errors. Grothendieck proposed that Serre write historical notes for EGA and, at moments, suggested joint treatment of particular subjects; those plans did not become coauthored EGA chapters. The distinction allows the relationship to appear in its actual strength. Serre could remain an independent mathematician, sometimes skeptical of the scale or timing of Grothendieck’s plans, while functioning as their most exacting early interlocutor.

This ecology also included people not present on the EGA title page. Chevalley’s conversations inspired an October 1958 note for Bourbaki. Pierre Cartier’s work on inseparable descent would be named in Grothendieck’s 1959 descent lecture. Roger Godement’s sheaf theory and Cartan–Eilenberg’s homological algebra supplied standard references. Samuel and Zariski’s commutative algebra stood in while Bourbaki’s volumes were still being prepared. Attribution becomes less dramatic when multiplied in this way, but more truthful. A foundation is precisely where inherited work becomes easiest to mistake for the natural order of things.

An institute before a building

The project acquired an institutional address at the same moment. Léon Motchane founded the Institut des Hautes Études Scientifiques in 1958, inspired by the Institute for Advanced Study in Princeton. IHÉS’s institutional profile ties Dieudonné’s appointment to Grothendieck’s recruitment and gives Grothendieck’s permanent-professor range as 1958–1970. Chapter 29 examines that institutional account and its archival limits.1818Institut des Hautes Études Scientifiques, “Alexander Grothendieck, Permanent Professor from 1958 to 1970”; Institut des Hautes Études Scientifiques, “The Foundation”.

The later phrase “IHÉS at Bures” compresses chronology. Through 1961 the mathematical section worked from borrowed Fondation Thiers space in Paris while Motchane prepared Bois-Marie near Orsay. The institute occupied the Bures property in 1962. Chapter 29 treats the move; here the relevant point is that appointments, publication, visitors, and a long research program preceded the permanent site.1919Institut des Hautes Études Scientifiques, “The Scientific Heritage of IHÉS,” 1.

The distinction matters because the landscape of Bures can too easily become a retrospective stage set. There was no 1959 gathering beneath trees at Bois-Marie that this chapter is entitled to describe. There were rooms in Paris, correspondence across the Atlantic, lectures at Bourbaki, pages moving between Grothendieck and Dieudonné, and an institute trying to establish credibility. EGA I’s title page locates the institute at 5 Rond-Point Bugeaud in Paris XVIe, not on a country estate.2020Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, title page.

In 1959 Motchane and Dieudonné also established Publications Mathématiques de l’IHÉS, with Dieudonné as its first editor-in-chief. The journal did more than display the institute’s name. It provided a vehicle suited to work whose length and serial development sat awkwardly within conventional article schedules. EGA I appeared as number 4 in 1960; EGA II would appear as number 8 in 1961. The institute, journal, and treatise grew together.2121Institut des Hautes Études Scientifiques, “The *Publications Mathématiques de l’IHES*: History of the Review,” https://www.ihes.fr/en/library/les-publications-mathematiques-de-lihes/; Numdam, “*Publications Mathématiques de l’IHÉS*: Volume Index,” https://www.numdam.org/item/PMIHES/.

This arrangement did not remove external standards. A journal issue remained a published mathematical object, open to correction and scrutiny. Grothendieck’s letters show drafts moving to press before the whole plan was settled. Later fascicles revised terminology and supplemented earlier chapters. The serial form exposed a tension characteristic of the program: it wanted foundations stable enough for generations of use, yet it had to publish while discovery was active. The authors answered by numbering tightly, cross-referencing aggressively, and declaring chapters open to later additions.

A plan larger than its first book

The October 1958 letter projected a sequence whose early chapters would cover the language, cohomology, covering spaces, group schemes, intersections, duality, Riemann–Roch, abelian varieties, and Weil cohomology. EGA I’s published introduction offered a revised thirteen-chapter plan. Even that plan did not predict the final distribution of texts. EGA as published stopped during Chapter IV. Fundamental groups and étale coverings appeared through SGA 1; group schemes through SGA 3; other themes migrated elsewhere or remained incomplete.2222Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 6–7; Grothendieck and Serre, Correspondance Grothendieck–Serre, 73.

It would be wrong to treat the discrepancy as simple failure. The initial plan acted as a research map. Writing one region exposed the need for new machinery in another; seminars could develop material faster and more collectively than the treatise; proofs grew. But it would be equally wrong to say that the whole plan was achieved “in spirit.” Some promised chapters were never written, some conjectural routes failed, and some existence problems resisted. A historically responsible narrative keeps program, theorem, and later fulfillment separate.2323Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Grothendieck’s schedules outran the calendar. In October 1958 he hoped for four chapters by the following spring; a year later, his letter projected several more for 1960. Serre’s editorial notes identify chapters that never appeared. Those inaccurate predictions document a rapidly expanding set of obligations rather than execution of a fixed, fully visible plan.

The word Éléments invited comparison with Euclid and Bourbaki, while the first sentence promised a treatise on foundations. Yet EGA was not intended as an elementary textbook in the modern pedagogical sense. Its “elements” were generative units. The authors wanted definitions and propositions from which research-level geometry could be rebuilt. A reader would encounter no leisurely historical survey; EGA I explicitly disclaimed competence to provide even a summary history and said it would cite only references especially useful for comprehension or for major origins.2424Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 7–8.

That disclaimer made the mathematical text efficient, but it creates work for biography. A formal theory naturally suppresses the contingent path by which it formed. Once $\operatorname{Spec}A$, fiber product, proper morphism, and representable functor have become standard, they seem arranged by necessity. The letters and lectures restore alternatives, jokes, overconfident schedules, failed inclusions, and named predecessors. They show not only what the foundation says but why its authors needed it when they did.

Freedom and its discipline

Grothendieck’s foundational turn is sometimes narrated as a preference for maximum abstraction. That phrase misses the selective character of the program. He did not generalize every variable indiscriminately. He sought notions that behaved well under operations central to geometry. A property of morphisms became valuable if it was local in an appropriate sense, stable under composition, or preserved by base change. A category became useful if fiber products existed and sheaves could be studied on it. A definition succeeded when it converted a recurring proof into a formal consequence without emptying the theorem of content.

The resulting prose can look impersonal because EGA suppresses discovery in favor of dependency. Definitions precede lemmas, which support propositions, which recur as cited tools hundreds of pages later. But the emotional temperature of the letters is different. There Grothendieck speaks of a technique coming into view, adds chapters to an already enormous plan, and responds impatiently to obstacles. One should not infer private states beyond the words. It is enough to observe a contrast between the controlled public treatise and the rapidly proliferating program documented in correspondence.2525Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Foundation was also a bid for collective freedom. If a language could be made dependable, students and collaborators would not have to reenact its invention before using it. The seminar could distribute problems. A theorem proved in relative form could serve many fibers. A construction represented by a scheme could be pulled back by any test scheme. This freedom was not egalitarian in every institutional or personal sense; Grothendieck exercised unusual authority over the program, and EGA’s difficulty restricted its immediate readership. Mathematically, however, the aim was to make methods transferable.

One aspect of the ambition was to enlarge and clarify the surrounding theory until a resistant statement became accessible. That description must still account for labor. Generality did not typeset chapters, repair errors, or decide which hypotheses survived. Dieudonné’s redaction, Serre’s criticism, predecessors’ theorems, and the institute’s publication apparatus helped turn the method into usable mathematics.2626Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

What the foundation did not claim

The new language did not prove that classical algebraic geometry had been misguided. It incorporated classical varieties and recovered their established results under suitable hypotheses. It did not make computation unnecessary; affine calculations, local algebra, and explicit equations remained indispensable. It did not make counterexamples disappear; a broader category often revealed new pathologies. It did not turn every moduli problem into a scheme; automorphisms and quotients would soon show where representability failed. It did not by itself produce a Weil cohomology satisfying all desired properties.2727Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Nor did it establish one uniquely compulsory ontology for all geometry. Schemes became extraordinarily successful because they balanced algebraic generality with workable local structure. Later algebraic spaces, stacks, formal schemes, and derived objects would enlarge the setting for problems schemes could not directly solve. To call those later objects “already schemes in disguise” would diminish the specificity of both achievements. The foundation of 1958–1960 was open-ended in practice even when its rhetoric was comprehensive.2828Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Its novelty should not be understated either. Allowing spectra of arbitrary commutative rings as affine pieces, retaining nilpotents, building geometry over arbitrary bases, and insisting on functorial constructions changed the subject’s center of gravity. Arithmetic and geometry could occupy one category. A family was no longer an accessory to its fibers. Infinitesimal neighborhoods became legitimate geometric objects. The language made questions visible that a collection of separate classical theories had obscured.

This balance—inheritance without reduction, novelty without creation myth—is the right scale for the period. Grothendieck’s contribution was not to utter the word “scheme” over an empty field. It was to recognize how a constellation of algebra, topology, and geometry could be reorganized, to drive that reorganization through a formidable program, and to test the language against major research problems. Dieudonné’s contribution extended far beyond copying. It was to make the program into sustained mathematical writing and to carry editorial responsibility at the institution publishing it. Serre’s contribution was not to stand as disciple or opponent. He supplied essential prior mathematics and remained an independent critic and correspondent. Motchane’s contribution was not mathematical authorship. It was to construct conditions in which a project of unusual length and risk could proceed.

A dated sequence of tests

The proposed foundation was tested in public documents rather than in one retrospective manifesto. The Edinburgh address set the cohomological target in 1958. The May 1959 Bourbaki exposé joined schemes to formal geometry. The December 1959 exposé treated faithfully flat descent. The February 1960 exposé opened with representable and pro-representable functors before presenting formal existence for modules. Together these dates show the program moving from language toward construction.2929Alexandre Grothendieck, “Géométrie formelle et géométrie algébrique,” Exposé 182, Séminaire Bourbaki, vol. 5 (1958–1960), 1960, 193–220, https://www.numdam.org/item/SB_1958-1960__5__193_0/; Alexandre Grothendieck, “Technique de descente et théorèmes d’existence en géométrie algébrique. I. Généralités. Descente par morphismes fidèlement plats,” Exposé 190, Séminaire Bourbaki, vol. 5 (1958–1960), 1960, 299–327, https://www.numdam.org/item/SB_1958-1960__5__299_0/; Alexandre Grothendieck, “Technique de descente et théorèmes d’existence en géométrie algébrique. II. Le théorème d’existence en théorie formelle des modules,” Exposé 195, Séminaire Bourbaki, vol. 5 (1958–1960), 1960, 369–90, https://www.numdam.org/item/SB_1958-1960__5__369_0/; Grothendieck, “The Cohomology Theory of Abstract Algebraic Varieties”.

The sequence matters more here than a second explanation of each concept. Every lecture took an inherited tool and asked whether it remained usable in the broader setting. Formal geometry required nilpotent thickenings to be admitted rather than discarded. Descent required a category in which base change and overlap data could be stated exactly. Representability required a functorial formulation of the classification problem. Chapters 26–28 explain those objects and conditions; Chapter 25 records when they entered the working program.

The lectures also show that generality was accountable to theorems. A new noun was insufficient. The framework had to produce fiber products, sheaves, classes of morphisms, formal comparison, descent, and existence results. This book uses FGA retrospectively for these dated Bourbaki exposés; the acronym was not the contemporary title of one 1958–1961 book. The individual dates distinguish an announced method from a theorem already printed.

The correspondence adds a private but still documentary clock. On October 31, 1959, Grothendieck linked descent and formal geometry with the characterization of functors of points; on November 5 he described the technical difficulties of the projected existence theorem and the obstruction caused by automorphisms. These letters establish problems under active investigation. They do not license the later language of stacks or Artin representability as knowledge already complete in 1959.3030Grothendieck and Serre, Correspondance Grothendieck–Serre, 101–10.

Formal geometry in the publication sequence

The May 1959 exposé begins with schemes because the later formal-geometry argument needs them. That order is documentary evidence about function: the language was introduced in service of a construction problem, not displayed as an isolated philosophical object. The full definition, examples, and infinitesimal calculations belong to Chapter 26.3131Grothendieck, “Géométrie formelle et géométrie algébrique,” 193–220.

The February 1960 exposé changes emphasis again. Its opening pages place representable and pro-representable functors before the existence theorem. The order shows a method being assembled: state the universal problem, analyze compatible formal data, and ask when an algebraic object exists. Chapter 28 supplies the precise variance and examples. Here the relevant fact is the sequence on the printed page.3232Grothendieck, “Technique de descente et théorèmes d’existence en géométrie algébrique. II. Le théorème d’existence en théorie formelle des modules,” A.1–2, pp. 369–370.

This sequence also disciplines attribution. Formal completion, category theory, Hilbert and Chow parameter methods, Weil’s descent, and Cartier’s inseparable case had separate histories. EGA and the Bourbaki exposés made them cooperate, but cooperation does not transfer every ingredient to one author. The Stacks Project history attributes the lemma now called Yoneda’s only in bounded terms because the 1954 primary paper has not been examined for this account.

The documents end with limits rather than total victory. Automorphisms could block a fine moduli scheme; compatible formal data needed hypotheses before algebraization; quotient existence remained a separate theorem. Later algebraic spaces and stacks answered some of these limits, but those later solutions retain their own dates and authors. The foundation’s 1958–1960 achievement was to state the obstacles in a language capable of supporting proofs.

The documents of 1958–1960

It is possible to reconstruct a restrained view of the work without inventing a room or a conversation. The documents supply their own sequence. The Edinburgh proceedings preserve the public program. The October 1958 letter preserves the hand-over of notes and commentaries to Dieudonné and the hope of publication through Motchane. The May 1959 Bourbaki text presents schemes and formal geometry. The institute’s history records that early seminars used two rooms at the Fondation Thiers. The October and November 1959 letters discuss chapters going to press, Serre’s criticisms, descent, and existence. The December 1959 and February 1960 Bourbaki texts publish parts of that existence strategy. The EGA I title page fixes the institutional address and the collaboration formula.3333Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck, “Géométrie formelle et géométrie algébrique”; Grothendieck, “Technique de descente et théorèmes d’existence en géométrie algébrique. I. Généralités. Descente par morphismes fidèlement plats”; Grothendieck, “Technique de descente et théorèmes d’existence en géométrie algébrique. II. Le théorème d’existence en théorie formelle des modules”; Grothendieck and Serre, Correspondance Grothendieck–Serre; Institut des Hautes Études Scientifiques, “The Scientific Heritage of IHÉS”.

What cannot responsibly be added is equally important. The sources used here do not record the weather on a drafting day, the arrangement of papers on Dieudonné’s desk, the tone in which Motchane proposed a publication date, or the private thought with which Grothendieck first wrote $\operatorname{Spec}A$. Such details would make a scene vivid by making it false. The drama lies in the changing plan and in the traceable allocation of work.3434Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

One letter fixes the pace. On October 31, 1959, Grothendieck reported that the opening EGA material was in press, shifted attention from a Bourbaki obligation on categories to a general existence theorem, and linked descent and formal geometry with functors $T\mapsto\operatorname{Hom}_S(T,X)$. The report connects the relative base, the functor of points, and a representing object without proving the projected general criterion.3535Grothendieck and Serre, Correspondance Grothendieck–Serre, 101.

Serre’s replies prevent the plan from appearing inevitable. He asked after publication, tested statements, and sent technical corrections. Some of Grothendieck’s conjectural assertions in the correspondence were later corrected in editorial notes. The letters thus offer more than celebratory anticipation. They show mathematical reliability being produced through disagreement and revision.

Dieudonné’s editorial work at Publications Mathématiques connects the mathematical and institutional desks. As first editor-in-chief from 1959, he helped establish the journal while collaborating on the institute’s largest serial work. It is tempting to convert this convergence into a tale of perfect organization. The delayed and redistributed EGA plan argues otherwise. The arrangement was productive because it could publish unfinished architecture, not because it eliminated incompletion.

The address at Rond-Point Bugeaud gives this mobile work a temporary center. Correspondence connected Paris to Princeton and Harvard; Bourbaki meetings and seminars supplied other sites; the coming Bures estate would provide a permanent one. The “foundation” was therefore geographically distributed before the institute occupied the house later associated with it.

Reception before inevitability

By the time later textbooks taught schemes as the standard foundation of algebraic geometry, the controversy of unfamiliarity had largely disappeared from the definition. In 1959 it had not. Grothendieck’s Bourbaki exposition says that geometers had persisted in excluding nilpotents and that this attitude had seriously obstructed natural methods. The polemical wording proves that the expanded objects were not yet obvious.3636Grothendieck, “Géométrie formelle et géométrie algébrique,” 193.

Resistance need not be caricatured as fear of abstraction. A working algebraic geometer could reasonably ask what a general ring contributed to a problem about varieties, whether nonclosed points deserved geometric status, whether nilpotents clarified or obscured intuition, and when hundreds of foundational pages would produce a theorem unavailable by established means. Different research programs impose different opportunity costs. Scheme theory prevailed not because such questions were foolish but because it answered many of them through use.3737Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Nor was acceptance instantaneous. Terminology changed; expositions multiplied; younger mathematicians learned through seminars and drafts; major theorems demonstrated the language’s reach. The EGA and SGA corpus itself became difficult to navigate. A foundation can unify a field while raising the entrance cost for individuals. That double effect belongs to the history.

The philosophical lesson is consequently modest. “Greater generality” is not a sufficient account of progress. The relevant generality was structured by morphisms and constrained by examples. It preserved classical cases, admitted arithmetic and infinitesimal ones, and supported constructions. Where it failed—especially for some quotients and moduli problems—the failure became visible enough to motivate a further enlargement.

Drafts, schedules, and the difference between a plan and a book

The correspondence gives the EGA project a scale that the finished fascicle conceals. In October 1958 Grothendieck expected several opening chapters to be ready within months. In October 1959 he again projected rapid progress through material that would later be redistributed, delayed, or left unwritten in the announced form. Serre’s editorial annotations identify some of those divergences. The schedule is evidence of an active plan, not evidence that the later corpus already existed in outline at every level.3838Grothendieck and Serre, Correspondance Grothendieck–Serre, 73, 96–117.

The division of labor appears through verbs. Grothendieck sent notes, wrote commentaries, lectured, and described chapters ahead. Dieudonné drafted, organized, and moved material toward the printer. Serre read, objected, and returned corrections. Motchane supplied a publication route through the new institute. These roles overlap, but the sources do not support replacing them with one collective author or one solitary author.

The letters also show why publication dates cannot be treated as dates of invention. A definition could circulate in notes before a fascicle appeared. A chapter could reach the printer while later sections remained provisional. A seminar exposé could state a theorem whose full systematic home had not yet been written. The history therefore needs several clocks: discussion, draft, lecture, press, and published object.

EGA I’s introduction formalized one version of the program by listing thirteen chapters. That list was already different from the sequence sketched in the October 1958 letter. The differences are not editorial noise. They show the program responding to mathematical dependencies: construction and descent moved forward, while fundamental groups, group schemes, duality, and cohomology grew into seminar programs of their own.3939Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”, introduction, pp. 6–7; Grothendieck and Serre, Correspondance Grothendieck–Serre, 73.

This moving plan also changes how incompletion should be described. The missing chapters are not invisible completed books, and the published fascicles are not fragments without independent value. Each issued part became a working reference. The unissued plan remained a map of intended relations, some of which were realized in other forms and some of which remained open.

The distinction protects later authorship. A topic named in an EGA outline does not become Grothendieck’s completed theorem when another mathematician later proves it. A seminar directed by Grothendieck does not transfer every exposé to him. A correction inserted by Dieudonné or Serre is not erased because the title page uses an asymmetric credit line. The documentary object under discussion must determine the verb.

For the reader, this means that 1960 is neither a beginning nor an endpoint. It is the year in which one substantial layer became public under stable pagination. The preceding letters and exposés explain how it formed; the later fascicles and seminars show what it could not yet contain. That middle position is the appropriate close to a chapter about foundation.

The same discipline governs the IHÉS chronology. Recruitment, acceptance, formal appointment, effective start, and the later institutional service range may carry different dates. The current profile gives 1958–1970, while an adjacent secondary account gives March 1959 for a formal appointment. Without an opened personnel record, Part V preserves the distinction instead of forcing the dates into one event.4040Institut des Hautes Études Scientifiques, “Alexander Grothendieck, Permanent Professor from 1958 to 1970”.

The result is less dramatic than a single founding moment and more informative. A program was announced, revised, divided among people and genres, and made progressively citable. The first fascicle is the point at which that process produced a durable object, not the instant at which all of its mathematics came into being.

A second consequence concerns genre. The ICM paper, private correspondence, Bourbaki exposé, EGA fascicle, and institutional profile answer different historical questions. The congress paper states a public program. A letter records what its author reported to one correspondent on one date. An exposé fixes a mathematical treatment for an expert audience. EGA supplies a numbered reference architecture. An institutional profile records the institution’s later account of appointments and identity. None can substitute for all the others.

Reading them together prevents two opposite errors. The first treats the published treatise as though every definition and proof appeared at the moment of first conception. The second treats drafts and letters as though aspiration were already theorem. The dates and genres instead show a sequence of proposal, criticism, redaction, proof, and publication.

They also distribute uncertainty. The EGA title page directly supports its credit formula. The correspondence directly supports the exchange of notes and corrections. The IHÉS profile supports the service range it currently reports. The exact contract date remains unknown because none of those documents is the personnel contract. A bounded biography can be confident about the first three statements and explicit about the fourth gap.

This method keeps the chapter from becoming either a technical primer or an institutional legend. Its task is to show how mathematical and material records converged before the first fascicle appeared, then pass the concepts to the chapters that own their definitions.

The source hierarchy follows the same logic. Primary mathematical texts support what they print; later histories can synthesize influence but cannot silently replace a missing primary locator. Official institutional pages are authoritative for their own current claims, not for every detail of an earlier negotiation. Modern mathematical references verify definitions without establishing historical priority. Keeping those functions separate makes the notes more demanding, but it also lets a later integrator revise one claim without destabilizing the chapter’s entire chronology.

The Edinburgh program in its own tense

The 1958 congress paper is best read in the future tense in which much of it was conceived. Grothendieck surveys cohomology theories for abstract algebraic varieties and the formal properties one wants: functoriality, exact sequences, products, cycle classes, and comparison with known theories where available. The larger aim is a cohomology capable of entering Weil’s conjectures.4141Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The paper’s existence at an international congress matters. A program stated there asks a broad mathematical public to judge not one proof but a direction. It gathers existing results, names missing constructions, and recruits attention. The subsequent scheme language should not be written backward into every sentence as a finished apparatus. The address stands between the Riemann–Roch and Tôhoku achievements of the previous Part and the EGA machinery of this one.4242Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Its English title uses “abstract algebraic varieties,” not “schemes.” EGA I later cites the address as a preliminary overview for readers approaching the treatise. The shift in terminology records development rather than concealment: the cohomological problem helped force the category in which it could be treated.

The printed proceedings appeared in 1960, while the congress itself took place in 1958. A bibliography that lists only 1960 can make the public program look contemporary with EGA I rather than prior to it. This chapter therefore attaches both dates whenever chronology matters.

At Edinburgh, Grothendieck was not yet an IHÉS author presenting an established school. The institute had only just been created. The later alignment of congress program, permanent position, journal, and treatise was still being assembled. The address belongs to the moment when research direction and institution met.

One should also resist measuring the lecture solely by what later succeeded. A program can contain routes abandoned, formulations replaced, and confident expectations revised. Its historical force lies in coordinating work under uncertainty. EGA reworked the Edinburgh plan and changed the ground on which it was pursued.

The bibliography as an intellectual map

EGA I’s bibliography is short enough to read as a map. Cartan and Chevalley’s 1955–1956 seminar, Cartan–Eilenberg’s Homological Algebra, Godement’s sheaf theory, Grothendieck’s Tôhoku paper and Edinburgh address, Nagata’s work, Samuel and Zariski’s commutative algebra, Serre’s FAC and GAGA, Weil’s foundations and finite-field paper, and Zariski’s work appear among the twenty-two entries.4343Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The list does not prove that every cited work contributed equally or in the same way. References can provide prerequisites, direct influences, contrast, or specific theorems. EGA’s introduction helps distinguish them. It calls FAC fundamental, says Chevalley’s local-ring schemes admit a natural extension, identifies Nagata’s development over Dedekind domains, and names Zariski and Weil as unavoidable influences.4444Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The bibliography also shows how young the technical synthesis was. Godement’s book appeared in 1958. Cartan–Eilenberg’s in 1956. Grothendieck’s Tôhoku paper in 1957. The foundation drew on tools whose standard expositions were nearly contemporaneous with the project.

This recency complicates the word “foundation.” EGA was not codifying a century-old consensus from a safe distance. It was installing recently developed homological and sheaf methods inside an active reconstruction of geometry. Some readers would have been learning prerequisites almost as quickly as the treatise used them.

The authors’ claim to presuppose no particular algebraic geometry should therefore not be confused with accessibility to beginners. They expected command of commutative algebra, homological algebra, sheaf theory, and functorial language. The foundation shifted what beginning meant.

The listed works also keep national narratives porous. American, French, Japanese, and international publication venues appear; Weil wrote across institutional worlds; Nagata’s results entered a Parisian treatise; Cartan–Eilenberg’s Princeton book supplied homological language. The “French school” depended on a transnational mathematical archive.

Missing names should be interpreted cautiously. A concise technical bibliography is not a census of everyone who shaped the subject. EGA explicitly disclaims a historical survey. Absence can reflect scope, citation practice, or indirect transmission. This is why bibliography and historiography cannot be identical.

The word “natural”

Grothendieck repeatedly defended methods as natural. In the 1959 scheme lecture, excluding nilpotents had obstructed “truly natural” methods. In later accounts, naturality can sound like an intuitive radiance: the right concept appears inevitable once seen.4545Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Within the mathematics, naturality has a stricter meaning. A construction is natural when it is compatible with morphisms and does not depend on arbitrary choices. A map attached to $X$ should commute with pullback along $X'\to X$. Isomorphic inputs should give coherently isomorphic outputs. Coordinates may calculate the construction but should not define its identity.4646Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The functorial meaning does not exhaust Grothendieck’s rhetoric. He also used “natural” evaluatively, for concepts that made diverse theorems align. Yet functorial compatibility supplies a test against pure taste.

Nilpotents offer an example. Keeping them is natural not because infinitesimals are aesthetically mysterious, but because tensor products, quotients, and scheme-theoretic intersections produce them. Automatically reducing after every construction inserts a nonfunctorial loss of information into geometry.

Relative morphisms offer another. Properness is universally closed because closedness stable under every base change is the behavior applications need. The stronger definition can look less immediate and more natural at once: less tied to one picture, more compatible with the category.

Naturality does not mean uniqueness of historical path. Other foundations of algebraic geometry were possible and useful. Nor does it mean every generalization is natural. A broader definition that destroys finiteness or fails under basic operations may be formally general and mathematically sterile.

The biography should therefore translate “natural” into evidence: which diagrams commute, which constructions become internal, which theorems unify, and which counterexamples disappear or appear. Otherwise the word merely repeats admiration.

Teaching a language before a community spoke it

A foundational text creates a pedagogical paradox. It must use the language while teaching it. EGA answers through relentless internal reference and by assuming neighboring modern subjects. The Bourbaki exposés answer differently: they present concentrated pieces to mathematicians already sharing much of the background.4747Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The seminar audience could interrupt, compare the new definition with an old theorem, and carry a proof into another lecture. The printed EGA reader encounters the final dependency chain. These were complementary routes of adoption.4848Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Serre’s role as reader is especially valuable. Because he already possessed deep algebraic geometry, his objections tested whether new terminology clarified rather than merely renamed. His corrections on projective morphisms show the foundation being audited by someone not responsible for its redaction.

Dieudonné’s prose also taught by consistency. Repeated forms of definition, proposition, corollary, and reference created habits. A reader who survived the initial density acquired a language in which later claims became shorter.

Adoption was not universal or instant. Researchers could continue to solve problems in classical language. Some found the scale forbidding. The later dominance of schemes can hide a period when learning them was a choice with real cost.

The institute reduced that cost for a concentrated group. Seminars, visitors, drafts, and access to authors supplied apprenticeship that the printed page alone could not. The new foundation was textual and social at once.

This social fact does not determine mathematical truth. It helps explain diffusion. A correct but unread language does not reorganize a field; a community must test, teach, and apply it.

Forecasting by chapter title

The October 1958 letter and the 1960 introduction both forecast future chapters. Comparing them shows the plan changing. Covering spaces, group schemes, intersections, duality, Riemann–Roch, abelian varieties, and Weil cohomology shift position and scope. New construction chapters intervene.4949Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Chapter titles served as research commitments. To reserve a chapter for descent was to claim that descent would become general enough to organize. To reserve one for the fundamental group was to claim that the scheme category could support a satisfactory theory. To end with Weil cohomology was to declare the ultimate test.5050Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

But a reserved title is not a result. The distinction matters acutely for later credit. SGA 1 developed the fundamental group through many exposés. SGA 3 involved multiple authors on group schemes. Étale cohomology required a collective program. The 1958 outline demonstrates Grothendieck’s direction, not sole prior possession of all subsequent proofs.

Forecasting also imposed pressure. Each new concept had to fit the earlier language, and each difficult proof could enlarge the prerequisites. The treatise expanded partly because the map exposed missing roads.

The plan’s revisions show responsiveness. Formal geometry, representability, quotients, and Hilbert schemes became indispensable construction techniques. A rigid commitment to the first outline would have been less responsive than changing it.

The foundation’s future orientation is therefore neither prophecy nor mere planning. It is mathematical action through organized incompletion.

Foundation by theorem

A definition becomes foundational when enough theorems make it dependable. For schemes, this meant more than showing that spectra could be glued. Fiber products had to exist. Quasi-coherent modules had to localize and glue. Immersions, finiteness, and separation had to behave under composition and base change.5151Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

Each proof reduced the amount of rebuilding required in later work. Once the tensor-product formula for affine fiber products was globalized, a new family could use it without reopening foundations. Once closed subschemes corresponded to quasi-coherent ideals, infinitesimal neighborhoods became systematic.5252Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

This accumulation explains EGA’s scale better than the phrase “love of abstraction.” A short manifesto could announce schemes; it could not certify their use through cohomology, descent, and construction.

The certification was never final. Some statements needed correction, hypotheses evolved, and later foundations broadened beyond Noetherian schemes. Dependability came from explicit proofs and corrigible references, not immunity to revision.

Grothendieck’s role was to select a category and drive its consequences across a research program. Dieudonné’s role was essential to turning those consequences into a chain another mathematician could audit. Serre’s corrections show the audit operating while the chain was built.

The foundation was therefore an achievement in mathematical trust. Readers accepted new objects because operations returned the right classical cases, new cases produced theorems, and errors could be located by number. Trust rested on work visible in the text.

This is also why a biography cannot replace explanation with the word “vision.” Vision names the direction. Foundation names the proved infrastructure that allowed others to travel.

The scale of novelty

Novelty can reside in a definition, a theorem, a synthesis, or a change of working scale. Grothendieck’s foundational achievement involved all four, but not every ingredient at once.5353Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The prime spectrum was rooted in commutative algebra. Local rings, sheaves, cohomology, and categories had prior histories. Scheme theory supplied new scope and binding. Formal existence, descent, and Hilbert representability supplied theorems that justified the binding.5454Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

The change of scale may be the hardest contribution to name. Instead of solving one classification problem, Grothendieck sought the conditions under which classification problems become representable. Instead of choosing one field, he made base change ordinary. Instead of tolerating infinitesimals as calculations, he made them objects.

This scale created danger. A program large enough to absorb many predecessors can make them look like fragments awaiting their organizer. They were complete mathematicians pursuing their own questions. The chapter’s source markers therefore attach inherited results to names even when EGA unifies them.

Scale also creates retrospective inevitability. Once textbooks begin with schemes, the older boundaries seem arbitrary. In 1958, changing them required persuasion, proofs, and a community willing to learn.

The new foundation should be judged neither as vocabulary nor revelation. It was an engineered mathematical environment whose components came from many places and whose design bore Grothendieck’s unusual reach.

That formulation leaves room for revision. Later spaces and categories would show where schemes were too narrow. Their success does not refute the foundation; it demonstrates that a living foundation can expose its own boundary.

Credit for the foundation

At the chapter boundary, the allocation remains exact. Grothendieck directed the new scheme-theoretic program and supplied its central constructions. Dieudonné collaborated in the redaction and organization of EGA and edited its institutional journal. Serre supplied foundational theorems, criticism, and correspondence without becoming an EGA coauthor. Weil, Zariski, Chevalley, Nagata, Cartan, Eilenberg, and Godement belong to the inherited mathematical field in distinct ways. Motchane supplied institutional conditions, not mathematical authorship. These statements are deliberately narrower than a single-author summary. They can be supported by title pages, dated letters, introductions, and publications, and they leave manuscript-level questions open for later archival review. Their cooperation explains the foundation better than either single-author myth or anonymous collectivity, both misleading frames.5555Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”; Grothendieck and Serre, Correspondance Grothendieck–Serre.

One final distinction governs the chapter. A foundation can be historically new without being logically first. Grothendieck did not have to invent rings, sheaves, or categories for their combination to change algebraic geometry. Conversely, assembling older ingredients does not make every consequence his alone. The relevant evidence is dated and granular: which definition was stated, which theorem was proved, which chapter was written, which correction was supplied, and which institution published it. That discipline leaves the achievement large. In three years a congress program, correspondence, Bourbaki lectures, an institute, a journal, and the first EGA fascicle converged. They did not converge automatically. The new foundation emerged because mathematical design, redaction, criticism, and institutional risk reinforced one another while remaining identifiable forms of work. That convergence, rather than any isolated anecdote, is the documented event this chapter calls a new foundation and the reason its many contributors must remain visible throughout the whole book.

The evidentiary distinction also changes the chapter’s grammar. A dated letter supports what Grothendieck reported on that date, not every later consequence of the report. EGA I supports its own stated debts, not a complete history of the subject. The institutional profile supports what IHÉS now says about recruitment, while the unopened contract leaves the exact effective date unknown. These limits do not weaken the narrative. They keep separate a contemporary document, a later editorial annotation, an institutional account, and a bounded historical inference. The chapter therefore uses direct past tense for dated actions, present tense for what extant texts state, and qualified language for interpretations of collaboration or influence. That calibration is especially important in a foundational period, when later standard terminology can make unfinished plans look settled and later success can make one participant appear to possess results that still required other authors, seminars, and proofs. That method also prevents a generic statement about influence from silently becoming a claim of sole invention. Where the evidence shows only that a text cites, names, or criticizes another work, the prose says exactly that. No attribution is strengthened merely to give a transition force or a conclusion cleaner symmetry.

That wording keeps the chapter’s confidence proportional to its sources and leaves later specialists room to correct the synthesis without dismantling its chronology for future review.

The first fascicle

EGA I was deposited in the third quarter of 1960. The fascicle ran from page 5 to page 228 of Publications Mathématiques de l’IHÉS number 4. Its title, Le langage des schémas, announced that language itself was the subject. The contents moved through affine schemes, preschemes and schemes, products, subspaces, questions of separation, finiteness conditions, rational maps, sheaves of modules, quasi-coherent and coherent sheaves, and formal schemes. The organization made the chapter both dictionary and machine.5656Grothendieck and Dieudonné, “Éléments de géométrie algébrique I: Le langage des schémas”.

The publication date marks an achievement, not a clean beginning. By 1960 the relevant ideas had appeared in lectures, letters, and work already under way. It did not mark completion: two further fascicles followed during 1961, first-edition definitions later changed, and planned material moved into seminars. Chapter 30 gives the publication record.

A reader opening the fascicle encountered austerity. There was no biographical preface, no scene of discovery, and no claim to have broken single-handedly with the past. There was a dedication, a statement of purpose, a demanding list of prerequisites, a plan, acknowledgments of direct influences, and then definitions. This restraint is itself evidence. Later celebration often supplies a heroic voice that the foundational document does not use.

The title The Language of Schemes also places a useful limit on philosophical interpretation. Language can enable thought without determining every thought. It can make relations expressible without deciding which theorems are true. Grothendieck’s scheme language would guide an immense program, but the results still required hypotheses, constructions, and proofs. A language is not a substitute for geometry; it is one of the places where geometry becomes possible.

By the end of 1960, then, “a new foundation” meant several things at once: a category of geometric objects built from rings, a relative syntax organized around morphisms, a functorial discipline, a serial treatise, and an institutional collaboration. The elements were not yet settled into the later textbook picture. Their power lay precisely in motion. The next step is to look closely at the object whose name came to stand for the whole change—not as a mystical enlargement of space, but as a rigorously constructed answer to what geometry should remember.

The program’s next requirement was a precise account of the scheme itself.