07.07.08 · representation-theory / compact-lie

Crystallographic point groups, space groups, and the crystallographic restriction theorem

shipped3 tiersLean: none

Anchor (Master): Sternberg *Group Theory and Physics* §4; Hahn (ed.) *International Tables for Crystallography Vol. A*; Schwarzenberger *N-dimensional Crystallography*; Hiller 1986 (*Amer. Math. Monthly* 93, 765–779) on the crystallographic groups via Bieberbach's theorems

Intuition Beginner

A crystal is matter arranged so that the same local pattern repeats forever in a regular grid. Salt, quartz, and snowflakes all have this property: pick up the pattern, slide it by one repeat distance, and it lands exactly on itself. The set of all rigid motions that leave the whole infinite pattern looking the same is the crystal's symmetry group, and crystallography is the study of which such groups can occur.

Two kinds of symmetry combine in a crystal. First there are the slides, called translations, which march along the repeating grid. Second there are the rotations and reflections that fix one point — the kinds of symmetry a single snowflake has. A snowflake looks the same after a sixth of a turn. A square tile looks the same after a quarter turn. The deep fact, found by counting, is that a repeating grid can only support a few of these rotational symmetries.

Here is the surprise this unit explains. A wallpaper or a crystal can have a rotation by one half, one third, one quarter, or one sixth of a turn. It can never have a rotation by one fifth of a turn. You cannot tile a flat floor with regular pentagons, and the same arithmetic forbids fivefold rotational symmetry in any crystal. This rule, the crystallographic restriction, is the reason there are exactly a handful of crystal symmetry types rather than infinitely many.

Visual Beginner

The picture shows two repeating dot patterns side by side. On the left, a grid where each cell is a square: it carries fourfold rotation, since a quarter turn about a lattice point maps the grid onto itself. On the right, a grid of equilateral triangles: it carries sixfold rotation about each vertex. Between them sits a failed attempt to build a fivefold grid, where regular pentagons leave gaps and never close up into a repeating pattern.

The takeaway from the picture is that the grid itself decides which rotations are allowed. A rotation is permitted only when turning the whole infinite grid sends every dot to another dot. Squares and triangles pass this test; pentagons fail it, which is why fivefold symmetry is missing from the list of crystal rotations.

Worked example Beginner

Take the square lattice: all points in the plane whose two coordinates are whole numbers, like $(0, 0)$ , $(1, 0)$ , $(0, 1)$ , $(2, 3)$ . Ask which rotations about the origin send this grid onto itself.

Step 1. A quarter turn (90 degrees) sends the point $(1, 0)$ to $(0, 1)$ and sends $(0, 1)$ to $(- 1, 0)$ . Both landing points have whole-number coordinates, so they are again grid points. Checking a few more points the same way, every grid point lands on a grid point. The quarter turn is a symmetry.

Step 2. A half turn (180 degrees) sends $(1, 0)$ to $(- 1, 0)$ and $(2, 3)$ to $(- 2, - 3)$ . Again whole-number coordinates, so the half turn is a symmetry too.

Step 3. Now try a fifth of a turn (72 degrees). The point $(1, 0)$ lands at $(cos 7 2^{\circ}, sin 7 2^{\circ})$ , which is about $(0.309, 0.951)$ . These are not whole numbers, and no amount of bookkeeping fixes that: the landing point is simply not on the grid.

Step 4. The same failure happens for any rotation whose angle is not a whole number of degrees out of the allowed set. Working out which angles keep every grid point on the grid leaves only turns by a half, a third, a quarter, or a sixth, plus the do-nothing rotation.

What this tells us: the grid forces the rotation count. A symmetry of a lattice must carry grid points to grid points, and that single requirement rules out fivefold and sevenfold turns while keeping the familiar two-, three-, four-, and sixfold ones.

Check your understanding Beginner

Formal definition Intermediate+

Work in Euclidean space $R^{n}$ with its group of rigid motions, the Euclidean group $E (n) = R^{n} ⋊ O (n)$ . An element is a pair $(t, A)$ acting by $x \mapsto A x + t$ , with composition $(t_{1}, A_{1}) (t_{2}, A_{2}) = (t_{1} + A_{1} t_{2}, A_{1} A_{2})$ . The homomorphism $(t, A) \mapsto A$ onto $O (n)$ is the linear part or point part; its kernel is the subgroup of pure translations, identified with $R^{n}$ .

A lattice $T \subset R^{n}$ is a discrete subgroup of rank $n$ , that is $T = Z a_{1} \oplus \dots \oplus Z a_{n}$ for a basis $a_{1}, \dots, a_{n}$ of $R^{n}$ . A crystallographic group (space group) is a subgroup $S \subseteq E (n)$ that is discrete and whose translation subgroup $S \cap R^{n}$ is a full-rank lattice $T$ with $S / T$ finite. The image $P$ of $S$ under the linear-part map is the point group of $S$ ; it is a finite subgroup of $O (n)$ that preserves $T$ , so $P \subseteq Aut (T) ≅ G L_{n} (Z)$ once a lattice basis is chosen. The defining exact sequence is

1 ⟶ T ⟶ S ⟶ P ⟶ 1.

A point group $P \subset O (n)$ is crystallographic when it stabilises some lattice; equivalently, in a suitable basis every element of $P$ has integer matrix entries. A Bravais lattice is a lattice classified up to the arithmetic type of its full point group $Aut (T) \cap O (n)$ (its holohedry); in dimension three there are $14$ such types. The arithmetic classification refines point groups by their conjugacy class inside $G L_{n} (Z)$ rather than inside $O (n)$ .

The extension is symmorphic when the sequence splits, so that $S ≅ T ⋊ P$ and every point-group operation can be realised about a single common origin. It is non-symmorphic when the sequence does not split: some point-group operation appears only when combined with a fractional lattice translation, producing screw axes (a rotation followed by a translation along the axis) and glide planes (a reflection followed by a translation in the mirror plane). Two space groups are equivalent when conjugate by an affine map; this is the relation under which the count $230$ is taken in dimension three. The classification up to orientation-preserving affine maps gives $11$ enantiomorphic pairs, raising an alternative count to $219$ .

Counterexamples to common slips

A finite subgroup of $O (3)$ need not be crystallographic. The icosahedral group $I_{h}$ of order $120$ contains a fivefold rotation, so it fixes no lattice and is none of the $32$ crystal classes — even though it is a perfectly good finite symmetry group of a molecule (buckminsterfullerene) or a quasicrystal.
The point group of a space group is not generally a subgroup of the space group. In a non-symmorphic group such as the diamond space group $F d \overset{ˉ}{3} m$ , a screw or glide operation projects to a point-group element whose pure-rotation lift is not itself a symmetry of the crystal; the point group lives only in the quotient $S / T$ .
"Bravais lattice" classifies lattices, not crystals. The number $14$ counts lattice types by holohedry; the finer datum of what decorates each lattice point is what multiplies $14$ Bravais lattices and $32$ point groups up to the $230$ space groups.

Key theorem with proof Intermediate+

Theorem (crystallographic restriction). Let $T \subset R^{2}$ or $R^{3}$ be a lattice and let $A \in O (n)$ be an orthogonal map of finite order $m$ that preserves $T$ . Then the order $m$ of any rotation appearing in the point group satisfies $m \in {1, 2, 3, 4, 6}$ .

The argument follows Sternberg ^[§4]. Because $A$ preserves the lattice $T$ , choosing a basis $a_{1}, \dots, a_{n}$ of $T$ expresses $A$ as an integer matrix $M \in G L_{n} (Z)$ : each $A a_{i}$ is again in $T$ , hence an integer combination of the basis. The trace of a matrix is unchanged by a change of basis, so

tr (A) = tr (M) \in Z .

Now compute the trace in an orthonormal eigenbasis instead. A rotation in the plane by angle $θ$ has trace $2 cos θ$ . In three dimensions, a proper rotation of $R^{3}$ fixes an axis and rotates the orthogonal plane by some angle $θ$ , giving eigenvalues $1, e^{i θ}, e^{- i θ}$ and trace $1 + 2 cos θ$ . An improper element (a rotoreflection) has trace $- 1 + 2 cos θ$ . In every case the trace equals $\pm 1 + 2 cos θ$ for the planar case absorbed into the same formula, and the integrality just established forces

2 cos θ \in Z, hence cos θ \in {- 1, - \frac{1}{2}, 0, \frac{1}{2}, 1} .

These five cosine values correspond to $θ \in {18 0^{\circ}, 12 0^{\circ}, 9 0^{\circ}, 6 0^{\circ}, 0^{\circ}}$ , that is to rotation angles $θ = 2 π / m$ with $m \in {2, 3, 4, 6, 1}$ . No other order is possible, because any other $m$ would give a value of $2 cos (2 π / m)$ strictly between two consecutive integers. In particular $m = 5$ gives $2 cos 7 2^{\circ} \approx 0.618$ , which is not an integer, so fivefold symmetry is excluded. The same integrality argument runs verbatim for an improper isometry, and runs in any dimension to constrain the eigenvalue angles, though the full list of admissible orders grows in dimension four and beyond.

Bridge. The restriction theorem is the foundational reason the crystal classes form a finite list, and it builds toward the full enumeration that this unit develops at Master tier. The integrality of the trace is exactly the bridge between two descriptions of the same operator: $A$ lives in the continuous group $O (n)$ , where its trace is the analytic quantity $\pm 1 + 2 cos θ$ , and $A$ also lives in the discrete group $G L_{n} (Z)$ , where its trace is forced to be an integer. Putting these together pins the rotation angle to a finite set. This is exactly the same tension between a compact Lie group and a discrete lattice inside it that appears again in the band theory of solids, where the translation lattice's character group is the Brillouin torus and the point group acts on it; it identifies the abstract finite point group with a concrete subgroup of $G L_{n} (Z)$ , and that identification is what makes the classification a finite computation rather than a continuous one. The central insight carried forward is that lattice-compatibility, not abstract group structure, is the binding constraint: the icosahedral group is a fine finite group but fixes no lattice, and the restriction theorem is precisely the obstruction.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Let $A \in G L_{2} (Z)$ have finite order. Show that its characteristic polynomial $t^{2} - (tr A) t + det A$ has $det A = \pm 1$ and $tr A \in {- 2, - 1, 0, 1, 2}$ , and list the possible orders.

Hint

A finite-order integer matrix is conjugate over $R$ to an orthogonal matrix, so its eigenvalues lie on the unit circle. Use that $det A = \pm 1$ for an integer matrix of finite order and that the trace is an integer between $- 2$ and $2$ .

Answer

A finite-order element of $G L_{2} (Z)$ is conjugate over $C$ to a diagonal matrix of roots of unity, so its eigenvalues $λ, \overset{ˉ}{λ}$ lie on the unit circle and $det A = λ \overset{ˉ}{λ} = ∣ λ ∣^{2} = 1$ for the rotation case, or $det A = - 1$ for the reflection case. The trace $λ + \overset{ˉ}{λ} = 2 cos θ$ is a real integer in $[- 2, 2]$ , hence $tr A \in {- 2, - 1, 0, 1, 2}$ . These give $θ = 18 0^{\circ}, 12 0^{\circ}, 9 0^{\circ}, 6 0^{\circ}, 0^{\circ}$ , that is orders $2, 3, 4, 6, 1$ for the proper case (and order $2$ for reflections). Rubric: full credit for the determinant, the trace range, and the order list.

Exercise 4 (medium, short-answer).

Explain why the splitting of $1 \to T \to S \to P \to 1$ is equivalent to the existence of a single point in $R^{n}$ fixed (as a set of axes) by lifts of every element of $P$ .

Hint

A splitting is a homomorphic section $P \to S$ . Realise each $P$ -element as a motion fixing a common origin.

Answer

A splitting is a homomorphism $s : P \to S$ with linear part the identity on $P$ . Each $s (p)$ is a motion $x \mapsto A_{p} x + t_{p}$ whose linear part is $A_{p}$ ; that $s$ is a homomorphism means the fractional translations $t_{p}$ compose consistently, which lets one solve for a common point $x_{0}$ with $A_{p} x_{0} + t_{p} = x_{0}$ for all $p$ simultaneously (averaging over $P$ ). Conjugating by the translation taking the origin to $x_{0}$ makes every $s (p)$ a pure point operation $x \mapsto A_{p} x$ about that origin, so $S = T ⋊ P$ is symmorphic. Conversely a common fixed origin yields the section. Rubric: full credit for the section-equals-common-origin equivalence with the averaging step.

Exercise 5 (medium, short-answer).

The space group of a crystal contains a screw operation: a $9 0^{\circ}$ rotation about the $z$ -axis followed by a translation of one quarter of the lattice period along $z$ , written $4_{1}$ . Show that applying it four times yields a pure lattice translation, and identify that translation.

Hint

Compose the screw with itself; the rotations multiply to the identity after four steps while the quarter translations accumulate.

Answer

Write the screw as $g = (\frac{1}{4} c \overset{z}{^}, R)$ where $R$ is the $9 0^{\circ}$ rotation about $z$ and $c$ is the lattice period. Then $g^{4} = (\frac{1}{4} c \overset{z}{^} + R (\frac{1}{4} c \overset{z}{^}) + R^{2} (\frac{1}{4} c \overset{z}{^}) + R^{3} (\frac{1}{4} c \overset{z}{^}), R^{4})$ . Since $R$ fixes $\overset{z}{^}$ , each term contributes $\frac{1}{4} c \overset{z}{^}$ , and $R^{4} = I$ , so $g^{4} = (c \overset{z}{^}, I)$ , a pure translation by one full lattice period $c$ along $z$ . The screw operation is genuinely non-symmorphic: $g$ itself is in $S$ but the bare rotation $R$ is not. Rubric: full credit for $g^{4} = (c \overset{z}{^}, I)$ and the non-symmorphic conclusion.

Exercise 6 (hard, short-answer).

Prove that there are exactly four wallpaper-compatible point groups containing a rotation of order $3$ or $6$ acting on a two-dimensional lattice, and identify the lattice they force.

Hint

A threefold or sixfold rotation forces the lattice to be hexagonal. Then enumerate the finite subgroups of $O (2)$ containing such a rotation that also preserve that lattice.

Answer

A rotation of order $3$ sends a shortest lattice vector $v$ to $R v$ and $R^{2} v$ ; these three vectors have equal length and pairwise $12 0^{\circ}$ angles, so $v$ and $R v$ generate a hexagonal lattice (rhombic unit cell with $6 0^{\circ}$ angle). The same lattice supports order $6$ , since $R_{60} = R_{120}^{- 1} \cdot R_{180}$ and $R_{180}$ always preserves a lattice. The point groups compatible with the hexagonal lattice and containing a threefold or sixfold rotation are $C_{3}$ (order $3$ ), $C_{6}$ (order $6$ ), and their dihedral extensions $D_{3}$ and $D_{6}$ (adding a mirror line) — four groups. (In the standard wallpaper enumeration these seed the plane groups $p 3, p 3 m 1, p 31 m, p 6, p 6 m$ , the multiplicity beyond four coming from the two inequivalent mirror placements for $D_{3}$ .) Rubric: full credit for forcing the hexagonal lattice plus the list $C_{3}, C_{6}, D_{3}, D_{6}$ .

Exercise 7 (hard, short-answer).

Show that the number of symmorphic space groups in three dimensions is governed by pairs (Bravais lattice, point group), and explain why this count is $73$ rather than $32 \times 14$ .

Hint

A symmorphic group is determined by an arithmetic crystal class: a point group together with a lattice it preserves, taken up to conjugacy in $G L_{3} (Z)$ . Not every point group is compatible with every Bravais lattice.

Answer

A symmorphic space group is $T ⋊ P$ , so it is fixed by the pair consisting of the lattice type $T$ and the point group $P \subseteq Aut (T) \cap O (3)$ , taken up to arithmetic equivalence (conjugacy of the pair in $G L_{3} (Z)$ ). This pair is exactly an arithmetic crystal class, and there are $73$ of them. The naive product $32 \times 14 = 448$ overcounts wildly because most point groups preserve only certain Bravais lattices — a fourfold point group needs a tetragonal or cubic lattice, not a triclinic one — and because different lattice centrings of the same holohedry that admit the same point-group action are identified. The $73$ arithmetic classes give the $73$ symmorphic groups; the remaining $230 - 73 = 157$ space groups are non-symmorphic, built by adding screw and glide translations classified by $H^{2} (P; T)$ . Rubric: full credit for the arithmetic-crystal-class identification and the compatibility explanation.

Advanced results Master

The classification organises into a tower of successive refinements, each a finite step. The base is the set of finite subgroups of $O (3)$ : the cyclic $C_{n}$ , dihedral $D_{n}$ , and the three exceptional polyhedral groups $T, O, I$ together with their reflection-augmented and rotoreflection variants. Imposing lattice-compatibility through the crystallographic restriction prunes this list to the $32$ crystallographic point groups (geometric crystal classes), since $C_{5}, C_{7}, \dots$ and the icosahedral families are removed and only $n \in {1, 2, 3, 4, 6}$ survive. These $32$ groups partition into $7$ crystal systems (triclinic, monoclinic, orthorhombic, tetragonal, trigonal, hexagonal, cubic) by holohedry.

Refining point groups by their conjugacy class inside $G L_{3} (Z)$ rather than $O (3)$ gives the $32$ geometric classes their arithmetic splitting: a point group together with the lattice it acts on, up to integer change of basis, is an arithmetic crystal class, and there are $73$ of these. The lattice types themselves, classified by holohedry, are the $14$ Bravais lattices, refining the $7$ systems by centring (primitive $P$ , body-centred $I$ , face-centred $F$ , base-centred $C$ ).

Theorem (extension classification of space groups). Fix an arithmetic crystal class, that is a finite $P \subset G L_{n} (Z)$ acting on a lattice $T ≅ Z^{n}$ . The equivalence classes of space groups with this point group and lattice are in bijection with the orbits of the second group-cohomology $H^{2} (P; T)$ under the action of the normaliser $N_{G L_{n} (Z)} (P) \times Aut (T)^{P}$ . The class $0 \in H^{2} (P; T)$ is the symmorphic split extension; non-zero classes encode the obligatory fractional translations producing screw axes and glide planes. In dimension three the total over all arithmetic classes is $230$ , of which $73$ are symmorphic ( $H^{2}$ -class zero) and $157$ are non-symmorphic.

Theorem (Bieberbach I; Sternberg §4 frames it; Bieberbach 1911). Every $n$ -dimensional crystallographic group $S$ contains a normal free abelian subgroup of rank $n$ and finite index, namely its translation lattice $T$ , and $T$ is the unique maximal abelian normal subgroup of finite index. This is what makes the point group $P = S / T$ a well-defined finite group intrinsic to $S$ , independent of any embedding. It resolves the first part of Hilbert's eighteenth problem: the point group is the abstract obstruction to a space group being a lattice.

Theorem (Bieberbach II; Bieberbach 1912). In each dimension $n$ there are finitely many crystallographic groups up to affine conjugacy; equivalently, two crystallographic groups abstractly isomorphic as groups are conjugate by an affine transformation of $R^{n}$ . The finiteness is the second half of Hilbert's eighteenth problem. The counts are $1, 17, 230, 4894, \dots$ in dimensions $1, 2, 3, 4$ , the dimension-two value being the $17$ wallpaper groups and the dimension-four value $4894$ (or $4783$ up to the finer affine equivalence) computed by Brown, Bülow, Neubüser, Wondratschek, and Zassenhaus in 1978.

Theorem (Zassenhaus algorithm). The enumeration of $n$ -dimensional space groups is a finite, effective computation: enumerate the finite subgroups of $G L_{n} (Z)$ up to conjugacy (the arithmetic crystal classes), and for each compute $H^{2} (P; Z^{n})$ and the normaliser action. Zassenhaus's 1948 reduction turns Bieberbach's abstract finiteness into the algorithm that produced the dimension-four table and underlies the modern crystallographic software in the International Tables.

Synthesis. The crystallographic restriction is the foundational reason the entire tower terminates in finite numbers, and it is exactly the integrality constraint that converts a continuous classification problem into a discrete one. Putting these refinements together, the count $230$ is not a coincidence of three-dimensional geometry but the value in dimension three of a uniform machine: finite subgroups of $O (n)$ pruned to $G L_{n} (Z)$ by lattice-compatibility (the $32$ point groups), refined arithmetically (the $73$ classes), and extended by the cohomology $H^{2} (P; T)$ that records screw and glide translations (the $157$ non-symmorphic groups). The central insight is that a space group is an extension $1 \to T \to S \to P \to 1$ of a finite point group by a translation lattice, and this single structural fact identifies the symmorphic groups with the zero cohomology class and the non-symmorphic groups with the non-zero classes. Bieberbach's theorems are the bridge that makes the point group $P$ intrinsic and the count finite in every dimension, and the Zassenhaus algorithm is the effective form of that bridge. The pattern recurs whenever a discrete subgroup sits inside a Lie group: the same extension-by-a-lattice structure governs the band representations of solids, where $P$ acts on the Brillouin torus $T$ , and the same cohomological obstruction reappears as the factor systems Tinkham uses to reduce electronic band structure by space-group symmetry.

Full proof set Master

Proposition (the trace integrality forces $m \in {1, 2, 3, 4, 6}$ in dimensions two and three), proof. Let $A \in O (n)$ with $n \in {2, 3}$ preserve a lattice $T$ and have finite order. Fixing a $Z$ -basis of $T$ writes $A$ as $M \in G L_{n} (Z)$ , so $tr A = tr M \in Z$ . Over $R$ , $A$ is an orthogonal map; its eigenvalues are $1$ (an even number of times in the rotation case, possibly with a single $- 1$ in the improper case) together with conjugate pairs $e^{\pm i θ}$ . In dimension two the trace is $2 cos θ$ (proper) or $0$ (a reflection, eigenvalues $\pm 1$ ). In dimension three a proper rotation has eigenvalues $1, e^{\pm i θ}$ and trace $1 + 2 cos θ$ , while an improper element has eigenvalues $- 1, e^{\pm i θ}$ and trace $- 1 + 2 cos θ$ . In each case $2 cos θ$ differs from an integer ( $tr A$ , or $tr A \mp 1$ ) by an integer, so $2 cos θ \in Z$ . Since $- 1 \leq cos θ \leq 1$ , this gives $2 cos θ \in {- 2, - 1, 0, 1, 2}$ and $cos θ \in {- 1, - \frac{1}{2}, 0, \frac{1}{2}, 1}$ . The corresponding $θ = 2 π / m$ have $m \in {2, 3, 4, 6, 1}$ . Any other $m$ produces $2 cos (2 π / m) \in / Z$ : for $m = 5$ , $2 cos 7 2^{\circ} = \frac{5 - 1}{2} \approx 0.618$ ; for $m \geq 7$ , $2 cos (2 π / m)$ lies strictly in $(1, 2)$ . $□$

Proposition (the point group preserves the lattice and is finite), proof. Let $S$ be a space group with translation lattice $T = S \cap R^{n}$ and point group $P = π (S)$ under the linear-part map $π$ . For $A \in P$ choose a lift $(t, A) \in S$ . Conjugation by $(t, A)$ sends a pure translation $(τ, I)$ to $(A τ, I)$ , and since $T$ is normal in $S$ (it is the kernel of $π$ ), $A τ \in T$ whenever $τ \in T$ . Thus $A$ maps $T$ to $T$ , so $P \subseteq Aut (T) ≅ G L_{n} (Z)$ . Finiteness: $P ≅ S / T$ , which is finite by the discreteness hypothesis (a discrete subgroup of $E (n)$ with full-rank translation lattice has finite point group, since $P$ is a discrete subgroup of the compact group $O (n)$ and a discrete subgroup of a compact group is finite). $□$

Proposition (symmorphic $⟺$ vanishing extension class), proof. The extension $1 \to T \to S \to P \to 1$ with $T$ abelian is classified up to equivalence by a class in $H^{2} (P; T)$ , where $P$ acts on $T$ through its inclusion in $G L_{n} (Z)$ . Concretely, choosing a set-theoretic section $s : P \to S$ with linear part the identity, the failure of $s$ to be a homomorphism is the factor system $f (p, q) = s (p) s (q) s (pq)^{- 1} \in T$ , a $2$ -cocycle. The extension splits — that is, admits a homomorphic section, hence $S ≅ T ⋊ P$ — exactly when $f$ is a coboundary, that is when its class in $H^{2} (P; T)$ vanishes. A homomorphic section provides a common origin fixed by all the $A_{p}$ acting linearly (solve $A_{p} x_{0} + t_{p} = x_{0}$ by averaging over $P$ , using that $\sum_{p} t_{p}$ is $P$ -equivariant), so the geometric statement "all point operations share a common centre" coincides with the cohomological statement "the class is zero". $□$

Proposition (a fivefold axis obstructs lattice symmetry, sharpened), proof. Suppose $A \in O (3)$ has order $5$ and preserves a lattice $T$ . By the first proposition $tr A \in Z$ , but a genuine order- $5$ rotation has trace $1 + 2 cos 7 2^{\circ} = 1 + \frac{5 - 1}{2} = \frac{1 + 5}{2}$ , the golden ratio, which is irrational. The contradiction shows no order- $5$ orthogonal map preserves any three-dimensional lattice. The same computation in dimension two gives $tr A = 2 cos 7 2^{\circ} = \frac{5 - 1}{2}$ , again irrational. This is why crystals forbid fivefold symmetry, and why the long-range order in the fivefold-symmetric quasicrystals discovered by Shechtman in 1982 cannot be periodic — they are aperiodic tilings, not lattices. $□$

Connections Master

Group action, orbit-stabiliser, class equation 01.02.03. A space group acts on $R^{n}$ by isometries, and the point group is the stabiliser-quotient structure made precise: the translation lattice is the orbit of the origin under translations, and the point group $P = S / T$ acts on the orbit space. The orbit-stabiliser counting that organises Wyckoff positions in the International Tables is the same class-equation bookkeeping introduced there, applied to the action of a finite point group on the unit cell.
Symmetry and group theory in chemistry 16.02.01. The molecular point groups assigned to molecules in chemistry are exactly the finite subgroups of $O (3)$ classified here, and the $32$ crystallographic point groups are the lattice-compatible subset of them. This unit supplies the math-side justification for why a molecular crystal's site symmetry is restricted to the crystal classes, closing the dependency where the chemistry unit invokes point-group character tables without deriving the crystallographic restriction.
Compact Lie group representation 07.07.01. The orthogonal group $O (n)$ is a compact Lie group, and the point group is a finite subgroup of it; the crystallographic restriction is the statement that lattice-preservation cuts the continuous group down to finitely many conjugacy classes of finite subgroups. The character theory of the finite point group used to reduce a crystal's vibrational or electronic representation is the finite-group specialisation of the compact-group representation theory developed there.
Character of a representation 07.01.03. The trace that drives the entire restriction theorem is the character of the defining three-dimensional representation of the point group evaluated on a rotation, $χ (R) = \pm 1 + 2 cos θ$ . The integrality of this character value over the lattice basis is the load-bearing computation, and the same character formula reappears in the reduction of the displacement representation that classifies normal modes.

Historical & philosophical context Master

The classification of crystal symmetry grew from mineralogy into group theory across the nineteenth century. Auguste Bravais established in 1850 that there are exactly $14$ lattice types in three dimensions, correcting an earlier count by Frankenheim, in Mémoire sur les systèmes formés par des points distribués régulièrement sur un plan ou dans l'espace (J. École Polytech. 19, 1–128) ^{[source pending]}. The enumeration of the $32$ crystal classes was assembled by Hessel (1830) and independently by Gadolin (1867). The decisive step — the full list of $230$ space groups — was achieved independently and almost simultaneously by Evgraf Fedorov in Russia ^{[source pending]} and Arthur Schoenflies in Germany ^{[source pending]}, both in 1891, with Barlow reaching the same count shortly after; their correspondence reconciled small discrepancies in each other's lists and is one of the cleanest cases of independent multiple discovery in mathematics.

The abstract structural theory came two decades later. Ludwig Bieberbach proved in 1911–1912, answering the first two parts of Hilbert's eighteenth problem, that every crystallographic group in every dimension contains a finite-index lattice of translations and that there are finitely many such groups up to affine equivalence ^{[source pending]}. This recast crystallography as the cohomology of finite-group extensions of free abelian groups, the form Hans Zassenhaus turned into an algorithm in 1948 and the form Sternberg presents in Group Theory and Physics (Cambridge, 1994) ^{[source pending]}, where the restriction theorem appears as a trace computation in the integer matrix representation of the point group. The dimension-four enumeration, $4894$ groups, was completed by Brown, Bülow, Neubüser, Wondratschek, and Zassenhaus in 1978. The supposed impossibility of fivefold symmetry was overturned in a different sense by Dan Shechtman's 1982 discovery of quasicrystals (Phys. Rev. Lett. 53, 1984), aperiodic structures with sharp fivefold diffraction that satisfy the restriction theorem precisely by not being lattices, for which Shechtman received the 2011 Nobel Prize in Chemistry.

Bibliography Master

@article{Bravais1850,
  author  = {Bravais, Auguste},
  title   = {M{\'e}moire sur les syst{\`e}mes form{\'e}s par des points distribu{\'e}s r{\'e}guli{\`e}rement sur un plan ou dans l'espace},
  journal = {Journal de l'{\'E}cole Polytechnique},
  volume  = {19},
  year    = {1850},
  pages   = {1--128}
}

@book{Fedorov1891,
  author    = {Fedorov, Evgraf S.},
  title     = {Simmetriya pravil'nykh sistem figur [Symmetry of Regular Systems of Figures]},
  publisher = {Zapiski Imperatorskogo S.-Peterburgskogo Mineralogicheskogo Obshchestva},
  volume    = {28},
  year      = {1891},
  pages     = {1--146}
}

@book{Schoenflies1891,
  author    = {Schoenflies, Arthur M.},
  title     = {Krystallsysteme und Krystallstructur},
  publisher = {Teubner},
  address   = {Leipzig},
  year      = {1891}
}

@article{BieberbachI1911,
  author  = {Bieberbach, Ludwig},
  title   = {{\"U}ber die Bewegungsgruppen der Euklidischen R{\"a}ume I},
  journal = {Mathematische Annalen},
  volume  = {70},
  year    = {1911},
  pages   = {297--336}
}

@article{BieberbachII1912,
  author  = {Bieberbach, Ludwig},
  title   = {{\"U}ber die Bewegungsgruppen der Euklidischen R{\"a}ume II},
  journal = {Mathematische Annalen},
  volume  = {72},
  year    = {1912},
  pages   = {400--412}
}

@article{Zassenhaus1948,
  author  = {Zassenhaus, Hans},
  title   = {{\"U}ber einen Algorithmus zur Bestimmung der Raumgruppen},
  journal = {Commentarii Mathematici Helvetici},
  volume  = {21},
  year    = {1948},
  pages   = {117--141}
}

@book{Sternberg1994,
  author    = {Sternberg, Shlomo},
  title     = {Group Theory and Physics},
  publisher = {Cambridge University Press},
  address   = {Cambridge},
  year      = {1994}
}

@book{ITA2016,
  editor    = {Aroyo, Mois I.},
  title     = {International Tables for Crystallography, Volume A: Space-Group Symmetry},
  edition   = {6},
  publisher = {Wiley},
  year      = {2016}
}

@article{BrownBulowNeubuser1978,
  author  = {Brown, Harold and B{\"u}low, Rolf and Neub{\"u}ser, Joachim and Wondratschek, Hans and Zassenhaus, Hans},
  title   = {Crystallographic Groups of Four-Dimensional Space},
  journal = {Wiley-Interscience Monographs},
  year    = {1978}
}

@article{Shechtman1984,
  author  = {Shechtman, Dan and Blech, Ilan and Gratias, Denis and Cahn, John W.},
  title   = {Metallic Phase with Long-Range Orientational Order and No Translational Symmetry},
  journal = {Physical Review Letters},
  volume  = {53},
  year    = {1984},
  pages   = {1951--1953}
}

@article{Hiller1986,
  author  = {Hiller, Howard},
  title   = {Crystallography and Cohomology of Groups},
  journal = {American Mathematical Monthly},
  volume  = {93},
  year    = {1986},
  pages   = {765--779}
}

Prerequisites

01.02.03
07.01.01
07.01.03
07.07.01
16.02.01

Tier anchors

beginner: Sternberg *Group Theory and Physics* §1.6 (finite subgroups of the rotation group) and §4 informal opening on lattices; Crash Course / 3Blue1Brown wallpaper-symmetry pictures
intermediate: Sternberg *Group Theory and Physics* §4 (crystallographic point groups, Bravais lattices, the restriction theorem); Tinkham *Group Theory and Quantum Mechanics* §9 (space groups)
master: Sternberg *Group Theory and Physics* §4; Hahn (ed.) *International Tables for Crystallography Vol. A*; Schwarzenberger *N-dimensional Crystallography*; Hiller 1986 (*Amer. Math. Monthly* 93, 765–779) on the crystallographic groups via Bieberbach's theorems

References

Sternberg — Group Theory and Physics · §1.6 (finite subgroups of O(3) and SO(3)), §4 (crystallographic point groups, the 32 crystal classes, Bravais lattices, space groups and the count 230)
Hahn (ed.) — International Tables for Crystallography, Volume A: Space-Group Symmetry · Part 1–7; the symmetry diagrams and generators of all 230 space groups, Hermann–Mauguin symbols, Wyckoff positions
Tinkham — Group Theory and Quantum Mechanics · §9 (space groups, factor systems, the Bloch / little-group reduction of band structure by translation symmetry)
Bieberbach — Über die Bewegungsgruppen der Euklidischen Räume I, II · Math. Ann. 70 (1911) 297–336 and 72 (1912) 400–412; the structure and finiteness theorems for crystallographic groups in every dimension (Hilbert's 18th problem)
Schoenflies — Krystallsysteme und Krystallstructur · Teubner 1891; one of the two independent enumerations of the 230 space groups
Fedorov — Симметрія правильныхъ системъ фигуръ (Symmetry of regular systems of figures) · Zap. Imp. S.-Peterb. Mineral. Obshch. 28 (1891) 1–146; the other independent enumeration of the 230 space groups

Estimated time

beginner: 16m
intermediate: 38m
master: 80m