The crystallographic restriction
A snowflake has six-fold symmetry: turn it by
60^\circ and it drops back onto itself, six times in a full circle.
Wallpaper, brick walls, honeycombs and salt crystals repeat with two-, three-, four- and
six-fold symmetry too. But hunt through every crystal that has ever been catalogued and you will
never find one with clean five-fold symmetry — nor seven-fold,
nor eight-fold. Nature seems to have banned them.
This is not an accident or a gap in the collection. It is a theorem. Any pattern
that repeats — that has translational symmetry, so a copy of it can be slid along
and land exactly on itself — is allowed only the rotation orders
2, 3, 4 and
6. Five-fold symmetry is mathematically impossible for a
repeating pattern. That single fact is the hinge this whole strand turns on.
What "periodic" means
A periodic pattern — a tessellation
of the plane, or the atoms in a crystal — is one built from a repeating lattice:
pick any two directions in which the pattern slides onto itself, and the whole thing is just one
tile stamped over and over on that grid. The set of all those slide-vectors forms the
lattice, and this is exactly what a crystal
is made of.
Now suppose the pattern also has a
rotational symmetry of
order n — turning it by
\tfrac{360^\circ}{n} about some centre leaves it unchanged. The
astonishing thing is that the lattice and the rotation cannot get along for just any
n. Requiring both at once forces
n into a tiny list.
The proof — a shortest arrow can't be beaten
Here is the classic argument, and it is beautifully short. Among all the slide-vectors of the
lattice, pick a shortest one; call its length a. Draw
it as an arrow from one lattice point A to a neighbour
B.
-
If the pattern has an n-fold rotation, then rotating that arrow by
\theta = \tfrac{360^\circ}{n} about A must
land on another lattice arrow, ending at a lattice point
B'. Rotating the reverse arrow about B
gives a lattice point A'.
-
Because differences of lattice vectors are themselves lattice vectors, the segment from
A' to B' is also a lattice
vector — so its length must be either 0 or at least
a (nothing shorter than the shortest is allowed).
-
A little trigonometry gives that length as a\,(2\cos\theta - 1).
Demanding it be 0 or \ge a pins
\cos\theta down so tightly that
2\cos\theta is forced to be a whole number.
-
A periodic pattern can have an n-fold rotation
only if 2\cos\!\left(\tfrac{2\pi}{n}\right) is an
integer.
-
The only whole values 2\cos\theta can take are
-2, -1, 0, 1, 2, and these correspond to exactly
n = 1, 2, 3, 4, 6.
-
So the only rotational symmetries a repeating pattern may possess are of
order 2, 3, 4 and 6.
Worked example 1 — testing each order
Let's just try the condition 2\cos\!\left(\tfrac{2\pi}{n}\right) on
each candidate and see which give a whole number:
- n = 2:\ 2\cos 180^\circ = -2 — an integer. Allowed.
- n = 3:\ 2\cos 120^\circ = -1 — an integer. Allowed.
- n = 4:\ 2\cos 90^\circ = 0 — an integer. Allowed.
- n = 6:\ 2\cos 60^\circ = 1 — an integer. Allowed.
-
n = 5:\ 2\cos 72^\circ \approx 0.618 — not a whole
number. Forbidden.
The lone failure at n = 5 is the whole story. (Try
n = 7: 2\cos\tfrac{360^\circ}{7} \approx 1.247,
again not whole — every order except 2, 3, 4, 6 falls at exactly this
hurdle.) The value 0.618 is no ordinary number, by the way: it is
\tfrac{1}{\varphi}, the reciprocal of the golden ratio
— the same irrational number that haunts five-fold shapes everywhere.
Worked example 2 — pentagons leave a gap
There is a second way to see why five is forbidden, with no trigonometry at all. To tile
the plane you must fit shapes snugly around every point, and the angles meeting at a point must add
to exactly 360^\circ. A regular pentagon's interior angle is
108^\circ. So:
- Three pentagons at a point give 3 \times 108^\circ = 324^\circ —
that leaves a 360^\circ - 324^\circ = 36^\circ gap.
- A fourth pentagon needs 108^\circ but only
36^\circ is free — it would overlap.
Because 108^\circ does not divide
360^\circ a whole number of times, regular pentagons can never surround
a point cleanly — so they can never be the repeating cell of a periodic tiling. Squares
(4 \times 90^\circ), triangles
(6 \times 60^\circ) and hexagons
(3 \times 120^\circ) all fit perfectly; pentagons alone are locked out.
Step through the figure below and watch the gap appear.
Allowed and forbidden centres on a lattice
On a triangular lattice, the six nearest neighbours of any point sit at the corners of a regular
hexagon, so a 60^\circ turn carries every dot onto another dot — order
6 works (and with it orders 2 and
3). Then watch a five-armed star try the same trick: its arms fall
between the dots, never on them.
See the pentagon gap
Three regular pentagons packed around a point use
324^\circ, and the leftover
36^\circ wedge is highlighted. There is simply no way to close the
ring.
For most of the twentieth century the crystallographic restriction was treated as iron law: no
crystal could show five-fold symmetry, full stop. Then in 1982 the materials
scientist Dan Shechtman shot a beam of electrons through a rapidly cooled
aluminium–manganese alloy and got a diffraction pattern with ten bright spots in a ring — clean,
sharp five-fold symmetry. Colleagues told him it was impossible; one handed him a
textbook and suggested he re-read it.
He was right and the textbook was incomplete. The material was ordered but not
periodic — a quasicrystal. It obeys the theorem by escaping its one assumption.
Shechtman won the 2011 Nobel Prize in Chemistry for the discovery.
The theorem forbids five-fold symmetry only for periodic patterns — ones with a
repeating lattice. That assumption is doing all the work. Drop it, and five-fold symmetry
becomes not just possible but famous:
-
Penrose tilings
cover the whole plane with five-fold symmetry using two tile shapes and never
repeat.
-
Quasicrystals
are the physical version: real matter with sharp five-fold diffraction and no lattice.
So the crystallographic restriction is exactly what makes Penrose tilings and quasicrystals
remarkable, not what makes them impossible. They are the beautiful exception that
only exists because they threw away the repeating grid.
The one idea to keep
-
A periodic pattern (one with a repeating lattice) may have rotational
symmetry only of order 2, 3, 4 or 6.
-
Five-fold (and seven-fold, eight-fold, …) symmetry is impossible for a
periodic pattern, because 2\cos\!\left(\tfrac{2\pi}{n}\right) is not
a whole number for those n.
-
Give up periodicity and five-fold symmetry returns — that is precisely what Penrose tilings
and quasicrystals do.