Melville Feynman was a uniform-salesman who had, by his own account, wanted to be a scientist and was determined his son would be one. He walked Richard through Principia Mathematica as a game - not the equations, the habit - asking always: what's really going on here? How would you know if you were fooling yourself?

In the Rockaway accent and unhurried Brooklyn cadence he would keep all his life he repeated the lesson his father had given him: distrust the name of the thing; ask what it does.

MIT for undergraduate, Princeton for graduate. At Princeton he worked under John Wheeler. The dissertation - "The Principle of Least Action in Quantum Mechanics" - is where the path integral begins. Wheeler once said of him: "He's another Dirac - only this time human."

It was at Princeton, too, that he met Arline Greenbaum. He married her in 1942 on the ferry to Staten Island. She was already ill with tuberculosis; the marriage was partly to care for her.

He went to Los Alamos at twenty-five, a junior theorist under Bethe. Oppenheimer had the famously unruly men to run and Feynman was among the unruliest - constantly poking at the mesa's security theatre. He cracked the filing-cabinet locks. Then the Mosler safes that held the atomic-bomb secrets.

Arline was in the Presbyterian sanatorium in Albuquerque, two hundred miles south. He would hitchhike down on weekends. On June 16, 1945, while he was with her, she died. He did not cry until weeks later, passing a dress in a shop window he thought she might have liked.

Three weeks after her death, he was at Trinity. A bright flash from the desert; he watched without goggles - the one man at the site to do so - on the reasoning that the windshield of a truck would stop the ultraviolet. "It was purple," he said afterward. He was twenty-seven.

After the war, Cornell for three years, then Caltech for the rest of his life. He turned down Princeton, the Institute for Advanced Study, and a standing invitation from almost everywhere else. Pasadena suited him: the weather, the proximity of open country, a department that let him lecture when he felt like it.

From 1961 to 1963 he taught the first- and second-year physics sequence at Caltech. The lectures were recorded and transcribed into three red volumes that remain in print. Sales of the Lectures, royalties on an introductory course, have kept him comfortably alive on library shelves.

He did much of his work in a diner called Gianonni's and later at a strip-club called the Papagayo Room in Pasadena, where the owner kept a booth open for him. He sketched equations on place-mats. The place was raided. He went to testify: yes, he used it as an office. Yes, others did too. The raid failed.

In the 1970s he painted his Dodge Maxivan with Feynman diagrams. The plates read QANTUM.

At the Shelter Island conference of 1948 - a small, hothouse affair on Long Island - Feynman, Schwinger, and Tomonaga each presented a different way of making quantum electrodynamics finite. The math agreed. Dyson showed the three were the same theory. Feynman's version used pictures.

The diagrams are a bookkeeping device - they enumerate the terms of a perturbation expansion - but they came into physics as a way of thinking about what happens in a scattering event: electrons travelling, photons exchanged, vertices as small dots. They remain how the field is taught.

He taught. Not only the canonical three-volume lectures but freshman seminars, invited talks, a commencement that became the best-known commencement ever given at Caltech: "Cargo Cult Science," 1974.

On the morning of 28 January 1986, the space shuttle Challenger broke apart 73 seconds after launch. All seven aboard died. The Rogers Commission was convened, and President Reagan appointed Feynman to it - against his wife Gweneth's better judgment; he was, by then, visibly ill.

At the televised hearing on 11 February 1986, Feynman dropped a rubber O-ring into a glass of ice water during a break, clamped it for a minute, then pulled it out stiff. "I believe this has some significance for our problem." The ring, cold, had lost its resilience. The booster seals on the morning of launch had been cold.

Appendix F is his minority report. He fought to get it included at all. It remains, three decades later, the plainest public document ever written about how large organisations lie to themselves.

He had been sick since the late 1970s: a rare cancer of the abdomen, Waldenström's. Surgeries, remissions. Through all of it, with his friend Ralph Leighton, he was trying to travel to the Tuvan Autonomous Soviet Socialist Republic - a small Turkic-speaking country in the southern Siberian taiga. Something about the name of its capital, Kyzyl, had caught him.

The travel papers arrived the week after he died.

He went into UCLA Medical Center on 3 February 1988 with an intestinal hemorrhage. He refused further surgery. On 15 February, sinking, he briefly woke, looked at those around him, and said what he said.