Ten-fold Peaks from Al-Mn: How Diffraction Redefined the Crystal — Epoche B2
Research note — the problem as it stood For most of the twentieth century every textbook said the same thing: a crystal is a single unit cell — a small box containing a few atoms — copied over and over on a periodic lattice, the way a wallpaper motif repeats across a wall. The lattice is the set of positions at which the copies sit; "periodic" means that shifting the whole structure by one cell length lands every atom exactly on top of another atom. From this one assumption a hard theorem follows: if a pattern repeats periodically in space, its rotational symmetry can only be of order 1, 2, 3, 4 or 6. This is the crystallographic restriction theorem . Five-fold and ten-fold rotations are excluded. So five-fold symmetry in a crystal was not merely rare; it was thought to be geometrically impossible. Because the theorem is the hinge of the whole story, it is worth seeing why it is true. Why only 1, 2, 3, 4 and 6 The everyday intuition comes from tiling a bathroom floor. Squares, equilateral triangles and regular hexagons tile the plane; regular pentagons do not. The interior angle of a regular pentagon is $108^\circ$, and around any corner point the angles of the tiles meeting there must sum to exactly $360^\circ$: three pentagons give $324^\circ$ and leave a gap, four give $432^\circ$ and overlap. No arrangement closes up. The real proof is only slightly harder, and it plants a seed that will sprout later. Suppose a lattice is unchanged by a rotation $R$. A symmetry of the lattice sends lattice points to lattice points, so if we write $R$ as a matrix using the lattice's own basis vectors as coordinates, every entry of that matrix is a whole number — the image of a basis vector is some integer combination of basis vectors. Now use one fact from linear algebra: the trace of a matrix (the sum of its diagonal entries) does not change when you change basis. In an ordinary orthogonal basis, a rotation by angle $\theta$ in the plane has trace $2\cos\theta$. In the lattice basis the trace is a sum of integers, hence an integer. Therefore $2\cos\theta$ must be an integer. Since $\cos\theta$ lies between $-1$ and $1$, the only possibilities are $2\cos\theta = -2, -1, 0, 1, 2$, giving $\theta = 180^\circ, 120^\circ, 90^\circ, 60^\circ, 0^\circ$ — precisely the rotations of order 2, 3, 4, 6 and 1. A five-fold rotation has $\theta = 72^\circ$ and $2\cos 72^\circ = 0.618\ldots$, which is not an integer. That failing number is no stranger: $0.618\ldots = \tau - 1$, where $\tau = \tfrac{1+\sqrt5}{2} \approx 1.618$ is the golden ratio. The very quantity that forbids five-fold symmetry in a periodic lattice will return as the signature of the structure that achieves it. 8 April 1982 — the observation Electron diffraction works because electrons travel as waves. A beam is fired through a thin sample; each atom scatters a wavelet; in most directions the wavelets arrive out of step and cancel, but in special directions the path differences are whole numbers of wavelengths, the wavelets add in phase, and a detector far away records a bright spot. The geometry of the spots is therefore a direct readout of the geometry of the atomic arrangement. Crucially, the spots are sharp only if atoms separated by thousands of atomic spacings still scatter in step — sharp spots mean long-range order . A glass, whose atoms have only local, short-range arrangement, produces no spots at all, just broad diffuse rings. Dan Shechtman, working at the US National Bureau of Standards, was studying aluminium–manganese alloys solidified by melt spinning — squirting molten metal onto a spinning cold wheel, which freezes it at around a million degrees per second. The purpose of such violent cooling is to trap arrangements of atoms that slow cooling would erase. On 8 April 1982 he recorded a diffraction pattern with ten sharp spots arranged in a perfect ten-fold ring; his notebook entry reads "10 Fold???". One clarification about the number ten: a diffraction pattern is always symmetric under inversion — the intensity in direction $\mathbf{q}$ equals that in $-\mathbf{q}$, a result known as Friedel's law — so a five-fold axis in the material necessarily shows up as ten spots on the screen. The pattern thus announced five-fold rotational symmetry combined with long-range order: sharp spots that the restriction theorem says a periodic structure cannot produce. The mundane suspects Extraordinary data attract ordinary explanations first, and rightly so. A measurement artefact. Instrumental faults produce smeared, irreproducible features. These peaks stayed sharp, appeared in sample after sample, and survived changes of instrument and operator. Multiple twinning. Twinning is the intergrowth of several ordinary crystals in symmetric orientations; their superposed diffraction patterns can fake a symmetry that no single grain possesses — five cubic crystals arranged like orange segments could in principle mimic ten spots. This was the serious objection, pressed for years by Linus Pauling, twice a Nobel laureate, who maintained that the patterns came from twinned cubic crystals (Pauling, 1985). It was ruled out by two tests. First, shrinking the electron beam so that it sampled ever smaller regions — down to single grains — still produced the full ten-fold pattern, whereas a patchwork of twins would eventually show its individual patches. Second, and decisively, tilting the sample mapped out the symmetry in three dimensions. Shechtman found five-fold axes, three-fold axes and two-fold axes at the precise mutual angles of an icosahedron — the twenty-faced Platonic solid — including adjacent five-fold axes separated by $\arctan 2 \approx 63.4^\circ$ (Shechtman, Blech, Gratias & Cahn, 1984). A patchwork can fake spots in one viewing direction, but it cannot conspire to present one consistent icosahedral symmetry from every direction at once. Later, slowly grown and thermodynamically stable quasicrystals such as aluminium–copper–iron gave peaks as sh