1. Madrid

The International Congress of Mathematicians opened in Madrid on August 22, 2006, with more than 4,500 participants. King Juan Carlos I presided over the opening ceremony. John Ball, president of the International Mathematical Union, announced four Fields medalists: Andrei Okounkov, Grigori Perelman, Terence Tao, and Wendelin Werner.

Three came forward. Perelman had declined.

The physical object already existed: a medal of fourteen-carat gold, 63.5 millimetres across and weighing approximately 169 grams. Archimedes appeared on one side. The reverse bore a Latin inscription honouring outstanding mathematical writings, together with the sphere and cylinder traditionally associated with his tomb. The accompanying award was 15,000 Canadian dollars. Contemporary reporting said the medal would remain available if Perelman changed his mind.

Ball had known before the ceremony. In June, he travelled to St Petersburg to discuss the award with Perelman. Later accounts described roughly ten hours of conversation over two days. The possibilities included attending to receive it, accepting without attending, or refusing altogether. The important point is not a theatrical reconstruction of the room: the refusal was considered and communicated before Madrid, not improvised as a public performance at the ceremony.

Reported explanations distinguished the validity of the work from additional recognition. If the proof was correct, why should an award be necessary? Ball also described a more complicated estrangement from the profession and a reluctance to become its public emblem. These are accounts transmitted through interviews, not a signed statement listing every reason.

The official citation was careful. It recognized Perelman’s contributions to geometry and his revolutionary insights into the analytic and geometric structure of Ricci flow. It did not use Poincaré’s name, though newspapers and the accompanying discussion did. The IMU’s record continues to name Perelman as a 2006 medalist and records that he declined acceptance.

Meanwhile, readers were overwhelming the server that held his papers. Their curiosity concerned three manuscripts that had never appeared as articles in a refereed journal. A medal celebrating mathematical writings had been awarded largely for files anyone could download.

This did not mean that refereeing had become unnecessary or that putting a claim online made it true. It meant that several activities often bundled together by journals—circulation, priority, expert checking, and public recognition—had taken different routes. The proof had become accepted without its author agreeing to every institution through which acceptance was expressed.

2. Three Files

The first manuscript appeared on November 11, 2002: thirty-nine pages on the entropy formula for Ricci flow and its geometric applications. Soon afterward, Perelman sent its abstract to a small group of mathematicians, including Richard Hamilton.

The second appeared on March 10, 2003. Its twenty-two pages concerned Ricci flow with surgery on three-manifolds. The third, posted on July 17, comprised seven pages on finite extinction time for certain three-manifolds. Together they made sixty-eight pages.

The abstracts did not announce a conquest of Poincaré. Hamilton’s program, however, was explicitly present. The first paper described geometric consequences of new monotonic quantities and a route toward geometrization. The second acknowledged that one earlier assertion about volume and eventual smoothness had not been justified, while explaining that the other conclusions did not depend on it. The third credited an argument from Hamilton’s work. The tone resembled a technical progress report more than the announcement of a century’s end.

Perelman did not submit these papers to a journal. Their home was arXiv, begun by Paul Ginsparg at Los Alamos in August 1991 and moved with him to Cornell in 2001. The project’s modest beginnings—a small workstation and expectations of a comparatively limited annual flow of submissions—gave little indication of the infrastructure it would become.

A journal could distribute work, establish a public date, and arrange specialist assessment. ArXiv separated those functions. Moderation screened submissions for suitability, but was not peer review. A timestamp documented an accessible version; it did not certify the argument. Publication schedules also meant that circulation was rapid rather than literally instantaneous in every case.

A contrast had appeared in April 2002, when Martin Dunwoody posted a short proposed proof of the Poincaré conjecture on his university webpage. An essential step failed, and a revised title acquired a question mark. Online access had made the argument available for scrutiny, including scrutiny by its author. It had supplied no guarantee.

Perelman’s reported willingness to let others correct and develop his work fits that distinction. Making the files public allowed the result to be judged without requiring him to control every later version of its explanation.

In April 2003, he spent about a month lecturing in the United States. MIT scheduled talks on April 7, 9, and 11, at 4:30 in the afternoon, on Ricci flow and the geometrization of three-manifolds. He also visited Stony Brook, Princeton, and Columbia. Job offers did not keep him there. He returned to the St Petersburg branch of the Steklov Institute, and the third paper appeared that summer.

The files were brief. Their public examination would not be.

3. A Rubber Band

The Clay Mathematics Institute’s familiar illustration begins with an apple and a doughnut. A loop lying on the apple’s surface can slide and contract to a point without leaving the surface. On a doughnut, a loop circling the central opening cannot always do so.

For a connected closed surface—compact and without boundary—having every loop contractible characterizes the two-dimensional sphere. Poincaré asked whether an analogous condition characterized the three-dimensional sphere. This sphere is not the solid ball in ordinary space. It is a three-dimensional manifold, naturally described as the boundary of a four-dimensional ball. The question concerns its intrinsic topology, not the shape of a familiar object sitting on a table.

Poincaré reached the right question after an earlier criterion proved insufficient. In 1900, he had suggested that matching the sphere’s homology would identify a three-sphere. In 1904, he constructed a counterexample: a space whose homological measurements resembled those of a sphere but whose fundamental group was nontrivial. One description assembles it by gluing two genus-two handlebodies, the solid counterparts of double-handled surfaces.

Homology and the fundamental group detect different information. The counterexample made that difference unavoidable. Poincaré then asked about a closed three-manifold in which every loop could contract. He ended the discussion by observing that pursuing the question would take him too far.

It took others almost a century. Higher-dimensional versions were settled earlier, through work in the 1960s and, for dimension four in the topological category, the 1980s. Here the formulation matters: those generalizations concern homotopy spheres, not merely all simply connected manifolds in higher dimensions. Simple connectivity alone does not identify a sphere once the dimension rises. Nor should the topological four-dimensional theorem be confused with the smooth four-dimensional question.

In May 2000, Clay included the three-dimensional conjecture among seven Millennium Prize Problems, with a million dollars attached to each.

Perelman arrived by recognizable academic stages, although at remarkable speed. At sixteen, he earned a perfect score at the 1982 International Mathematical Olympiad in Budapest. He studied in Leningrad, completed his doctorate in 1990, and worked at the local Steklov branch. In 1992 he began visits to NYU’s Courant Institute and Stony Brook, followed by a Miller fellowship at Berkeley from 1993 to 1995.

A four-page paper in 1994 proved the soul conjecture, open for more than twenty years. An invited address at that year’s congress in Zurich followed. He returned to St Petersburg in 1995.

During his American years he encountered Hamilton, twenty-three years his senior, after a lecture at the Institute for Advanced Study. Accounts of their meeting emphasize patience, openness, and willingness to discuss unpublished ideas. The contrast with later disputes is powerful, but it remains a remembered encounter, not evidence that every other mathematician behaved differently.

4. Surgery

Hamilton introduced Ricci flow in 1982. Rather than studying a fixed geometric metric, he allowed it to evolve according to curvature. The analogy with heat spreading through metal is useful: a differential equation redistributes something unevenly concentrated. But geometry is nonlinear, and the analogy should not suggest that every region simply becomes flatter.

Hamilton proved that a compact three-dimensional manifold with positive Ricci curvature evolves, after normalization to control its scale, toward constant positive sectional curvature. Such a manifold has spherical geometry; it need not be the sphere itself rather than a spherical quotient. The simple description “it becomes round” suppresses both the normalization and the topological qualification.

For a general initial metric, the flow can develop singularities. A neck may become thinner while curvature grows without bound, preventing the ordinary smooth solution from continuing. Hamilton’s program proposed surgery: recognize suitable necks, cut them, attach controlled caps, and resume the flow.

The operation raises difficult questions. Can the developing singular regions be classified well enough to know where cutting is legitimate? Can the parameters be chosen so that surgeries do not accumulate infinitely within a finite time interval? Will estimates survive the changes introduced by each cut?

Perelman supplied decisive new tools, including entropy and reduced-volume quantities, noncollapsing estimates, and control of the local geometry near singularities. Those tools allowed Hamilton’s program to pass obstacles that had resisted years of work.

For the Poincaré case, finite extinction provides a further simplification. A Ricci flow with appropriately controlled surgery becomes extinct in finite time for the relevant class of manifolds. Tracking which topological pieces have been removed allows one to reconstruct the possibilities for the starting manifold. Under the Poincaré hypotheses, the classification forces the original manifold to be a three-sphere. One does not infer that conclusion merely from watching a picture shrink.

Hamilton’s role is written into Perelman’s own abstracts. It is not a concession invented after the prize dispute. Equally, recognizing the program’s origin does not diminish the new ideas required to make it succeed.

Why, then, did sixty-eight pages need more than a thousand pages of later exposition? Compression is one answer, though not a complete description of every subsequent contribution. Perelman often provided the central construction or estimate while leaving expert readers to supply substantial intermediate arguments. Surgery parameters, limiting processes, and the passage from local information to a global conclusion all needed careful elaboration.

The expositors did real mathematical work. That does not make the original proof merely a conjectural suggestion. It means that the distance between an inventor’s concise argument and a community’s reliable understanding can itself require years of research.

5. Hundreds of Pages

Bruce Kleiner and John Lott began posting notes on Perelman’s papers in 2003. Readers could watch the exposition grow and change. A consolidated version appeared on arXiv in May 2006 and was published in Geometry & Topology in 2008, at roughly 270 pages.

In June 2006, Huai-Dong Cao and Xi-Ping Zhu published a 328-page article in the Asian Journal of Mathematics. John Morgan and Gang Tian posted their account in July; it became a book of more than five hundred pages in 2007.

These were not three unrelated discoveries of the same theorem. They developed the Hamilton–Perelman approach, explaining arguments, checking details, and sometimes supplying different local treatments. Nor did all three cover precisely the same scope: Morgan and Tian’s 2007 book concentrated on Ricci flow and the Poincaré conjecture, while the broader geometrization story required further work and exposition.

Acceptance accumulated rather than occurring in a single vote. Morgan lectured in Madrid on August 24 about the conjecture as solved. Science selected the achievement as its 2006 breakthrough of the year. Subsequent corrections to expository texts did not turn this process into a fiction; they were part of the continuing work of making a large proof dependable.

The Clay award also required more than the existence of three uploaded files. Its rules involved publication and a waiting period, as well as acceptance by the mathematical community. The refereed Morgan–Tian monograph and other detailed treatments contributed to the record on which Clay acted. It would be too simple to say that somebody else’s book mechanically replaced a missing journal submission and automatically triggered a prize.

The Cao–Zhu article generated a separate controversy about attribution. Its original title announced a complete proof of the Poincaré and geometrization conjectures, followed by a subtitle identifying Hamilton–Perelman theory. Some readers understood the wording as implying that the decisive completion belonged to the expositors.

There was also a specific, verifiable omission. Kleiner and Lott complained that an argument drawn from their earlier online notes had not been credited. Cao and Zhu acknowledged the omission and apologized. They explained that notes made while studying the earlier exposition had later been incorporated without their source being recognized.

The December 3 revision changed the title to “Hamilton-Perelman’s Proof of the Poincaré Conjecture and the Geometrization Conjecture.” Its introduction identified the attribution correction and explained that the new title and abstract were intended to make the allocation of credit clear. The authors said the mathematical body otherwise remained essentially unchanged.

That documentary sequence is firm: complaint, acknowledgment, apology, revised attribution. It does not, by itself, establish every allegation made about motives or the journal’s editorial process. Correcting a source omission and judging an entire professional conflict are different tasks.

6. Becoming Visible

A large part of the public account of Perelman’s refusal comes from interviews conducted by Sylvia Nasar and David Gruber in 2006. Their New Yorker article appeared online immediately before the Madrid congress and in the August 28 issue.

It reported dissatisfaction with professional ethics, unease about compulsory public visibility, and a sense that continued participation demanded compromises. Perelman had left his Steklov position in December 2005. His mathematical withdrawal and his refusal of the medal were placed within a wider conflict about credit and authority. That broad narrative is influential; it is not an unmediated record of his mind.

The article also portrayed Shing-Tung Yau’s role in the controversy. Yau objected strongly. On September 18, his lawyers demanded correction and an apology, alleging distortion and defamation. Hamilton publicly supported Yau and objected to what he considered an unfair portrayal. Daniel Stroock and Michael Anderson also challenged the handling of statements attributed to them. The magazine defended its reporting; the threatened litigation did not become a lawsuit.

The dispute cannot be responsibly resolved by choosing whichever account produces the most satisfying villain. Yau defended the value of detailed exposition and denied trying to appropriate Perelman’s achievement. Others believed public descriptions had diminished Perelman’s contribution. The acknowledged omission in Cao and Zhu’s article is evidence about that omission. It cannot bear the weight of proving every participant’s intentions.

There is a second problem with the quotations that now circulate independently of their sources. A remark can acquire the authority of a first-person declaration while losing the circumstances in which it was recorded. Contemporary newspaper accounts reported Perelman emphasizing that his published calculations were what he had to offer the public. Later stories repeated more vivid formulations about money, fame, heroes, and being displayed like an animal. Some were attributed retrospectively to 2006 without a recoverable original interview setting.

They should not be stitched into a single imagined manifesto. Translation, selection, editing, and republication stand between the speaker and the reader. Nor does declining interviews license strangers to fill the silence with a diagnosis or a romantic biography.

This chapter is part of the same problem. The person who declined public display has become an inexhaustible subject for public explanation.

The evidence has an uneven shape. On the mathematical side are dated files, versions, page counts, references, and corrections. On the side of refusal are a small number of reported conversations and institutional notices. A four-thousand-year-old account on clay may survive as a directly inspectable object. An explanation given in a room twenty years ago may survive only through a newspaper.

That asymmetry does not make all personal testimony worthless. It means our confidence should be matched to the kind of evidence we actually possess.

7. A Million Dollars

In 2006, refusing the Fields Medal did not yet amount to refusing the Millennium Prize. Ball later reported that Perelman would discuss the matter with Clay if an award were made. Other contemporary reporting likewise presented the decision as unresolved.

Clay announced its award on March 18, 2010. The first Millennium Prize recognized the solution of the Poincaré conjecture. A two-day meeting in Paris began on June 8. Perelman did not attend.

On July 1, Interfax reported that he had informed Clay about a week earlier that he would not accept. In the remarks attributed to him, he described dissatisfaction with decisions made by the organized mathematical community and emphasized that Hamilton’s contribution was no smaller than his own. These remain reported remarks rather than a detailed signed account identifying every disputed decision.

The emphasis differed from the earlier proposition that a correct proof needed no additional recognition. Now the allocation of recognition itself was foregrounded. It would nevertheless be an overreach to infer a complete theory of fair credit from a few sentences.

The two prizes had different rules. Fields medal eligibility required that a candidate’s fortieth birthday not fall before January 1 of the award year. Hamilton, born in January 1943, was sixty-three in 2006 and could not qualify. Perelman turned forty that June; 2006 was his final eligible congress. A newspaper’s shorthand “under forty” can obscure this calendar rule.

The Millennium Prize had no comparable age restriction. Its 2010 award named Perelman alone. Hamilton had received other recognition, including a Clay Research Award in 2003, and later the Steele and Shaw prizes. He taught at Columbia from 1998 until his death on September 29, 2024, aged eighty-one, as the university’s obituary records. None of those facts settles how Perelman believed the specific award should have been allocated.

One of the most widely circulated purported explanations is especially weakly sourced: the claim that someone who knew how to control the universe had no need of a million dollars. It emerged in 2011 through a purported filmmaker’s account of a conversation. Contemporary challenges questioned the intermediary’s credentials and identified errors in the surrounding story. There is no comparably secure record of Perelman delivering the line.

The sentence is memorable partly because it supplies exactly the character audiences expect: the otherworldly genius dismissing ordinary money. That narrative convenience is a reason to demand stronger evidence, not a substitute for it. It should be described as an unverified attribution, not used as the chapter’s explanation of a real person.

8. Two Ledgers

Perelman was not the first mathematician to refuse a major prize. In April 1988, the Royal Swedish Academy of Sciences announced the Crafoord Prize jointly for Alexander Grothendieck and his former student Pierre Deligne. The shared award amounted to 1.6 million Swedish kronor, approximately $270,000 at the conversion used in the contemporary announcement.

Grothendieck replied on April 19. Here, unlike in much of the Perelman story, a signed letter survives. Le Monde published a version on May 4; editorial changes displeased him.

His reasons combined material sufficiency, a judgment about mathematical value, and an indictment of professional conduct. His salary and approaching pension, he said, already met his family’s needs. Prestige and surplus resources were distributed unequally. The vitality of work done decades earlier should be measured by the ideas growing from it, not by a belated honour. He also believed accepting would endorse a scientific culture whose ethics he condemned. These were Grothendieck’s judgments, not independently established descriptions of every colleague.

Perelman had already declined a European Mathematical Society prize in 1996, without leaving a comparable public explanation. He was the first to decline the Fields Medal.

Prizes can also change what recipients subsequently do. George Borjas and Kirk Doran compared Fields medalists with high-achieving contemporaries who did not receive the medal. Their study, published in 2015 after earlier working-paper versions, found lower subsequent measured output among winners and more movement into unfamiliar topics. Their estimates attributed roughly half the output decline to this increased intellectual experimentation.

The study concerns patterns in observational data, not the motives of every medalist. Publication counts, citations, and student numbers do not exhaust mathematical value. Exploring a difficult new field may reduce short-term output while enlarging what a researcher can understand. The important point here is narrower: a prize is not necessarily a passive label attached after the work is finished.

Mathematical validity and public reward follow different rules. A proof should remain valid regardless of its author’s identity. A prize must identify a recipient, invite a person to a ceremony, and sometimes transfer money. Usually those systems coexist without open collision. In 2006 and 2010, one person made their difference impossible to ignore.

The institutions ultimately kept separate records. The IMU lists Perelman and his refusal. Clay recognizes the conjecture as solved. The unaccepted money funded a Poincaré Chair, jointly established with the Institut Henri Poincaré, supporting early-career mathematicians for visits of six months to a year over a five-year funding period. Its first appointments began in 2014.

Clay explained the purpose by recalling Perelman’s own opportunity at Berkeley’s Miller Institute in 1993–95: time for a promising mathematician to develop ideas. The chair announcement thus redirected the reward toward conditions in which future work might become possible.

Those earlier years had offered something other than a medal or a million dollars: time, freedom, and colleagues willing to discuss mathematics.

The three files remain publicly available. No payment and no distinguished identity are required to read them. Whether a piece of work is correct, and what its maker receives in exchange, belong in different ledgers.

On September 11, 2026, twenty-five Fields medalists would sign a declaration asking another version of the question: when mathematical results can be produced with increasing speed, what is the activity ultimately for?