On April 8, 1982, Dan Shechtman looked at a diffraction pattern from an aluminum-manganese alloy and wrote three words in his notebook: 10 fold ???
Three question marks. That's the honest part.
The pattern he saw had tenfold rotational symmetry, which crystallography said was impossible. Not unlikely — impossible. There's a theorem about it. Crystals are periodic: the atoms arrange in a unit cell, the unit cell repeats through space, and that periodicity constrains what symmetries you can have. You can tile a floor with equilateral triangles, squares, or hexagons — they fit edge to edge, they repeat, no gaps. Pentagons don't tile. They leave gaps. So five-fold symmetry, and by extension tenfold, was provably incompatible with how crystals worked.
Shechtman knew this. He spent the rest of the afternoon looking for an explanation that would fit. Crystal twinning is a known phenomenon — two differently oriented crystals grown together can produce a combined diffraction pattern that mimics symmetries neither crystal actually has. He cycled through electron microscopy techniques trying to find the twins that had to be there. His group head, John Cahn, told him to go away — these were twins and not terribly interesting.
The twins weren't there.
He waited two years to publish. The paper came out in 1984 in Physical Review Letters. Within weeks, Paul Steinhardt and Dov Levine published a response coining the word "quasicrystal" and providing the mathematical framework that explained what Shechtman had found. Linus Pauling, a two-time Nobel laureate and probably the most famous chemist alive, declared: "There are no quasicrystals, only quasi-scientists." He proposed his own twinning explanation and kept proposing it until he died in 1994, without updating.
Here's the thing about the theorem: it was right. Periodic structures cannot have five-fold symmetry. The error wasn't in the theorem — the error was in assuming everything solid and crystalline was periodic. Quasicrystals are ordered but not periodic. They have long-range orientational order — every atom knows which way it's pointing relative to every other atom — but no repeating unit cell. They're aperiodic.
Roger Penrose had worked out the mathematics of aperiodic tiling in 1974, eight years before Shechtman's aluminum-manganese alloy. Two tile shapes — a thick and thin rhombus — that cover the plane completely, never repeat, and have perfect fivefold symmetry throughout. The math was there waiting. The matter just hadn't shown up yet, or no one had looked for it in the right way.
In 1991, the International Union of Crystallography did something unusual. They didn't dismiss Shechtman's results or quietly fold them into a footnote. They changed the definition of crystal. A crystal is now any solid that produces a sharp diffraction pattern — periodic or not. The category expanded. The theorem stayed true within its original domain, and the map got redrawn around the territory.
What I keep returning to: the theorem was valid, the conclusion from it was wrong, and those are different problems. "Fivefold symmetry is impossible in periodic structures" doesn't logically extend to "fivefold symmetry is impossible in all structures." The gap between those two sentences is where nine years of scientific resistance lived — and where one notebook entry with three question marks was already sitting.
The question marks are the interesting part. Shechtman saw what was there, wrote what he saw, and flagged his own uncertainty. He didn't write "impossible" or "twinning artifact." He wrote what he observed and marked it unresolved. That's a small thing to notice, but it seems significant.
Pauling was maximally certain and wrong. Shechtman wrote question marks and was right. That's probably not a universal rule. But it's something to think about when the observation and the theorem are in conflict — which one is the load-bearing object, and which is the assumption hiding inside the other.