Symmetry Groups in Action

Learning Objectives

Section 1 gave us the abstract definition of a group. We met $D_3$, the symmetries of a triangle, and we verified the four axioms by counting. Beautiful — but $D_3$ is a toy. The real prize is the rest of the crystallographic landscape: water, ammonia, methane, the cubic rocksalt crystal, the hexagonal layers of graphene, and every other structure you will compute in VASP. Every one of them carries a finite group of point symmetries, and we now have the vocabulary to name, count, and use those groups.

By the end of this section you will be able to:

1. Recognise the five building-block symmetry operations of a finite point group: $E$, $C_n$, $\sigma$, $i$, and $S_n$.
2. Read and write a Schoenflies symbol (e.g. $C_{2v}$, $D_{3h}$, $O_h$) and predict the order $|G|$ from the symbol alone.
3. Identify a small set of generators and explain how a few axes plus a few mirrors build every other operation in the group through closure.
4. Walk through three real molecules — water, ammonia, methane — and, for each, list every operation in its point group and play with the symmetry elements live in 3D.
5. Connect this geometric picture to a concrete VASP/spglib output: read the international symbol, the point group, and the order of the group from one Python call on a NaCl primitive cell.

The big idea: abstract group theory becomes useful the moment you stop asking “does this set form a group?” and start asking “which group is it?”. The Schoenflies notation is the answer to the second question, and it is the language every crystallography textbook, every DFT code, and every spectroscopy paper writes in.

From Abstract to Concrete

Imagine handing the four axioms of section 1 to a chemist who has never seen them. They would shrug: “this is just bookkeeping. Show me a molecule.” The conversion from axioms to molecules proceeds in three steps, all of them quick, all of them non-negotiable.

1. Pick the universe. Stand a finite, rigid object in empty space — a molecule, a unit cell, a polyhedron. Its centroid stays fixed. We will only consider operations that fix this centre point. (Translations are deferred to the next chapter, where they join the point group to make a space group.)
2. Find every isometry that maps the object to itself. An isometry is a rigid motion: a rotation about an axis through the centroid, a reflection across a plane through the centroid, an inversion through the centroid, or any composition of those. The axioms of section 1 guarantee these isometries form a group under composition.
3. Name the group. Among the infinitely many possible finite point groups, the geometry of 3D space allows surprisingly few. Counting only those that are compatible with crystal translations leaves exactly $32$ — the famous 32 crystallographic point groups. Each has a Schoenflies symbol. Memorising the dozen or so most common is the same kind of investment as memorising the periodic table's first row: you will use them every day.

The Five Building Blocks

Every operation in every finite point group falls into one of five types. Memorise these once and you will read any character table or VASP OUTCAR symmetry block at a glance.

The determinant column is the same proper/improper split we met at the end of section 1: $+1$ for operations that preserve handedness (a right-hand glove stays a right-hand glove), $-1$ for operations that flip it (a right-hand glove becomes a left-hand glove). The order column gives the smallest number of times you have to apply the operation before you return to the identity.

E — the Identity

The trivial “do nothing” operation. Every group contains exactly one identity element by axiom G3. As a 3×3 matrix it is the identity matrix; its determinant is $+1$; its order is 1.

Cₙ — Proper Rotations

A $C_n$ axis is a line through the centroid such that rotation by $360^\circ / n$ maps the object to itself. The single axis generates $n - 1$ non-trivial rotations (the angles $360^\circ k / n$ for $k = 1, 2, \ldots, n-1$) plus the identity, for a total of $n$ elements in the cyclic subgroup it generates. So a single $C_3$ axis gives you the subgroup $\{E, C_3, C_3^2\}$ — three operations from one axis.

Crystals admit only $n = 1, 2, 3, 4, 6$ (the famous crystallographic restriction theorem, proved in chapter 1). Quasicrystals can host $C_5$ and $C_8$, but classical crystals cannot.

σ — Mirror Reflections

A mirror plane $\sigma$ is a plane through the centroid such that reflecting across it maps the object to itself. Every mirror has determinant $-1$; applying it twice returns the identity, so its order is 2. Mirrors come in three flavours, distinguished by their orientation relative to the principal axis:

The distinction between $\sigma_v$ and $\sigma_d$ is bookkeeping that pays off when you read character tables: $\sigma_d$ sits in a different conjugacy class from $\sigma_v$ in groups like $D_{4h}$, and it transforms differently under the irreducible representations.

i — Inversion

Inversion takes the point at $(x, y, z)$ to the point at $(-x, -y, -z)$. A crystal has inversion symmetry — is centrosymmetric — if every atom at $\mathbf{r}$ has an identical atom at $-\mathbf{r}$. Centrosymmetry is decisive for: piezoelectricity (forbidden in centrosymmetric crystals), second-harmonic generation (also forbidden), and selection rules in IR/Raman spectra. As a matrix,

Of the 32 crystallographic point groups, exactly 11 contain inversion. They are called the Laue classes and they govern which Bragg reflections you observe in a powder XRD pattern.

Sₙ — Improper Rotations

An improper rotation $S_n$ means: rotate by $360^\circ / n$, then reflect across the plane perpendicular to the rotation axis. It is one operation, not two — although you can decompose it as $S_n = \sigma_h \cdot C_n$ for analysis. Two special cases collapse into the building blocks we have already met:

• $S_1 = \sigma_h$ — a rotation by 360° is just the identity, so an $S_1$ is the same as a single mirror.
• $S_2 = i$ — rotating by 180° and then reflecting across the perpendicular plane sends $(x, y, z)$ through two axis flips to $(-x, -y, -z)$, which is exactly inversion.

Genuinely new improper rotations begin at $S_3, S_4, S_6$ — and in crystals only those three orders are allowed (same crystallographic restriction). The order of $S_n$ as an element is $n$ if $n$ is even and $2n$ if $n$ is odd, so $S_4$ has order 4 but $S_3$ has order 6.

Schoenflies Notation

Schoenflies symbols are short, mnemonic, and almost entirely compositional: a one- or two-letter root names the rotation skeleton, and a subscript adds the mirror/inversion decoration. Memorise the table below and you can already name 90% of the groups you will encounter in this book.

A handful of regularities make the table painless to remember:

• $C$ means “cyclic” — driven by a single axis. $D$ means “dihedral” — that axis plus a ring of perpendicular $C_2$s, like the spokes of a wheel. $T$ and $O$ are the two highly symmetric polyhedra (tetrahedron and octahedron/cube).
• The subscript tells you which mirrors are present: $v$ for vertical, $h$ for horizontal, $d$ for dihedral.
• Adding one mirror or inversion doubles $|G|$. That is why $|D_{nh}| = 4n$ while $|D_n| = 2n$: the horizontal mirror brings a second “copy” of every existing operation along.
• The cubic groups $T_d, O_h$ sit at the top of the symmetry hierarchy — and almost every common crystal lives in one of them. NaCl, KCl, LiF, MgO, all III-V semiconductors with zincblende structure, diamond, silicon, germanium, copper, gold, silver — every face-centred-cubic compound has $O_h$.

Interactive Point-Group Explorer

Time to see these groups. The explorer below shows three molecules in 3D — water, ammonia, and methane. For each molecule you can toggle the rotation axes (color-coded by $n$) and the mirror planes (translucent amber discs). The right-hand panel shows the Schoenflies symbol, the order $|G|$, and the breakdown of operations by type. Drag to rotate, scroll to zoom.

A few observations to make before you read on:

1. Switch to water. Turn on rotation axes — exactly one red $C_2$ axis, vertical, bisecting the H–O–H angle. Turn on mirrors — two amber discs, both vertical, one containing all three atoms (the molecular plane) and one perpendicular to it.
2. Switch to ammonia. One green $C_3$ axis through the nitrogen. Three mirror planes — each contains the $C_3$ axis and one N–H bond. $|G| = 6$ matches $D_3$ from §1, but the planes here are $\sigma_v$ rather than the in-plane reflections of the abstract triangle: same order, structurally identical, geometrically different incarnation.
3. Switch to methane. Turn axes on and watch the forest light up: four green $C_3$ axes (one through each C–H bond) and three red $C_2$ axes along the cartesian directions. With 6 mirror planes added, you can already enumerate $1 + 8 + 3 + 6 = 18$ operations. The remaining six are the $S_4$ improper rotations, which we cannot draw cleanly as static elements but live along the same axes as the $C_2$s.

Generators — the DNA of a Group

Look at the breakdown for methane: 24 operations. Drawing 24 things sounds tedious, and in fact we did not — we drew four $C_3$ axes, three $C_2$ axes, and six mirrors, total 13 symmetry elements. The remaining 11 operations come for free: closure (axiom G1) forces them into existence the moment we declare the first 13. This is what mathematicians mean by a generating set.

A subset $S \subseteq G$ is a generating set for $G$ if every element of $G$ can be written as a product of finitely many elements of $S$ and their inverses. We write $G = \langle S \rangle$, read “$G$ generated by $S$”.

The pattern is striking: two generators usually suffice for a finite point group, three are enough for any of the 32 crystallographic ones. This is why a real molecule's symmetry can be verified by inspecting only a handful of axes and mirrors — closure does the rest of the work for you.

Three Worked Examples

Pull the explorer above into the same field of view as this section and switch molecules as you read. The aim is to see each operation while you read its symbol.

Water — C₂ᵥ, |G| = 4

Water's three atoms lie in a plane, with the oxygen at the apex of an obtuse triangle. Place the $C_2$ axis along the bisector of the H–O–H angle. The four operations are:

Two of the operations swap the H atoms; the other two leave them fixed. As 3×3 matrices in the convention “$y$ is the $C_2$ direction, $xy$ is the molecular plane”:

Notice that $C_2 = \sigma_v \cdot \sigma_v'$ — the rotation is the composition of the two mirrors. So the group is generated by any two of $\{C_2, \sigma_v, \sigma_v'\}$; the third is redundant. $C_{2v}$ is abelian — composing any two of these elements gives the same answer in either order, because their matrices are all diagonal.

Ammonia — C₃ᵥ, |G| = 6

Ammonia is a low pyramid: nitrogen at the apex, three hydrogens forming the triangular base. The $C_3$ axis runs through the nitrogen perpendicular to the base. The six operations are:

• $E$ — identity.
• $C_3$ — rotate +120° about the vertical axis. Permutes the three H's cyclically.
• $C_3^2$ — rotate +240°. Inverse of $C_3$.
• Three vertical mirrors $\sigma_v^{(1)}, \sigma_v^{(2)}, \sigma_v^{(3)}$ — each containing the $C_3$ axis and one N–H bond, fixing that bond and swapping the other two.

The order is 6 — the same as $D_3$ from §1 — and indeed $C_{3v}$ and $D_3$ are the same abstract group. Two groups are isomorphic when their multiplication tables agree up to relabelling. Geometrically: $C_{3v}$ is a 3D pyramid, $D_3$ is a 2D triangle, but algebraically they are indistinguishable. This is your first taste of an important principle: the same abstract group can sit in many different geometric realisations, and group theory cares only about the algebra.

Methane — Tₐ, |G| = 24

Methane ($\text{CH}_4$) is the showpiece. The carbon sits at the centre of a regular tetrahedron with a hydrogen at each vertex. The 24 operations of $T_d$ come from five distinct classes:

Add the counts: $1 + 8 + 3 + 6 + 6 = 24$. Two generators suffice — for instance, one $C_3$ axis and one $S_4$ axis. The other 22 operations follow by composition.

The 32 Crystallographic Point Groups

We have built up the language; now the punchline. Of the infinitely many ways to arrange axes and mirrors in 3D, exactly 32 point groups are compatible with a periodic crystal lattice. The compatibility condition is the crystallographic restriction theorem: only $C_n$ with $n \in \{1, 2, 3, 4, 6\}$ are allowed, because no other rotation order tiles space periodically.

These 32 groups partition into 7 crystal systems by the maximum allowed rotation order:

Notice the ceiling: $|O_h| = 48$ is the largest order any crystallographic point group can have. NaCl, diamond, copper, silicon — every cubic crystal you compute in VASP — sits in $O_h$ or one of its subgroups. The practical consequence is the kind of speedup we cited at the top of the chapter: VASP can shrink a Brillouin-zone sum by up to a factor of 48 simply by recognising $O_h$ from your POSCAR.

VASP Connection — Reading the Point Group

Time to make the language concrete. The script below gives spglib a primitive NaCl cell and asks for the symmetry dataset. It prints the international symbol, the space-group number, the Hermann-Mauguin point-group symbol, and the order $|G|$ — the four numbers you will read whenever VASP analyses a structure. This is exactly the line of the OUTCAR that begins “Found 48 space group operations”.

Click any line of the code below to see, on the left, what each variable holds, exactly which arguments each library function consumes, and the value at that line. Watch especially line 27, where we walk through every one of the 48 rotation matrices and classify each by determinant.

      u     v     w
0   0.0   0.0   0.0    (Na)
1   0.5   0.5   0.5    (Cl)

International symbol : Fm-3m
Space group number   : 225
Point group (HM)     : m-3m
Order |G|            : 48
  proper   (E, C_n)        : 24
  improper (sigma, i, S_n) : 24

1import numpy as np
2import spglib
3
4# NaCl rocksalt (space group Fm-3m, #225) — primitive FCC cell.
5a = 5.64                              # lattice constant in Å
6lattice = np.array([
7    [0.0,   0.5*a, 0.5*a],
8    [0.5*a, 0.0,   0.5*a],
9    [0.5*a, 0.5*a, 0.0  ],
10])
11positions = np.array([
12    [0.0, 0.0, 0.0],                  # Na at origin
13    [0.5, 0.5, 0.5],                  # Cl at (1/2, 1/2, 1/2)
14])
15numbers = [11, 17]                    # Z(Na)=11, Z(Cl)=17
16cell = (lattice, positions, numbers)
17
18# One call extracts the space-group + point-group.
19ds = spglib.get_symmetry_dataset(cell, symprec=1e-5)
20
21print(f"International symbol : {ds['international']}")
22print(f"Space group number   : {ds['number']}")
23print(f"Point group (HM)     : {ds['pointgroup']}")
24print(f"Order |G|            : {len(ds['rotations'])}")
25
26# Split each operation into proper / improper by determinant.
27proper = sum(1 for R in ds['rotations']
28             if int(round(np.linalg.det(R))) == +1)
29improper = len(ds['rotations']) - proper
30print(f"  proper   (E, C_n)        : {proper}")
31print(f"  improper (sigma, i, S_n) : {improper}")

The expected console output:

International symbol : Fm-3m
Space group number   : 225
Point group (HM)     : m-3m
Order |G|            : 48
  proper   (E, C_n)        : 24
  improper (sigma, i, S_n) : 24

Read this output back to the table at the top of the previous section: $|G| = 48$ is the maximum any crystallographic point group can have, the proper/improper split is 24/24 (NaCl is centrosymmetric, so half the operations contain a reflection), and the Hermann-Mauguin symbol $m\bar{3}m$ translates one-to-one into the Schoenflies symbol $O_h$. Every cubic crystal in this book — silicon, germanium, gold, NaCl, MgO — will produce essentially this same line.

Summary

1. Every finite molecule or unit cell carries a point group — the set of isometries that fix the centroid and map the object to itself. Closure is automatic, by axiom G1.
2. Five symmetry building blocks generate every point group: $E$, $C_n$, $\sigma$, $i$, $S_n$. Their determinants split them into proper ($+1$) and improper ($-1$).
3. Schoenflies notation labels a point group by its generators: $C_n$, $C_{nv}$, $C_{nh}$, $D_n$, $D_{nh}$, $D_{nd}$, $T$, $T_d$, $T_h$, $O$, $O_h$. Counting generators predicts $|G|$.
4. The crystallographic restriction admits only $n \in \{1, 2, 3, 4, 6\}$ for $C_n$, which whittles the infinite zoo of possible point groups down to exactly 32. They partition into 7 crystal systems, with $O_h$ as the highest-order group at $|G| = 48$.
5. Three molecules anchor the typology: $\text{H}_2\text{O} \to C_{2v}$ with 4 operations, $\text{NH}_3 \to C_{3v}$ with 6, $\text{CH}_4 \to T_d$ with 24.
6. spglib reads a (lattice, positions, numbers) tuple and reports the international symbol, the space-group number, the Hermann-Mauguin point group, and the full list of rotation matrices in one call. For NaCl: Fm-3m, #225, $m\bar{3}m \equiv O_h$, $|G| = 48$.

Looking ahead. We now have a vocabulary of point groups and can spot them on real molecules and crystals. The next section asks: how does each operation act on a function — say, an electron orbital or a phonon mode? The answer is a 3×3 matrix per operation, called a representation — and the rest of the chapter (sections 3–7) extracts every meaningful physical statement from those matrices alone.