Chapter 1
15 min read
Section 5 of 70

Symmetry Operations

The Architecture of Crystals — Real Space

What Symmetry Means for Crystals

A crystal looks the same after certain geometric transformations — rotations, reflections, inversions, and combinations thereof. Each such transformation is called a symmetry operation. The full set of symmetry operations that maps a crystal onto itself defines the crystal's symmetry group, which in turn dictates its physical properties: optical activity, piezoelectricity, allowed vibrational modes, and even the electronic band structure.

Computationally, symmetry is a powerful tool for efficiency. If a crystal has 48 symmetry operations, VASP can reduce the number of k-points in a Brillouin zone calculation by up to a factor of 48, cutting the computational cost dramatically.

Intuition: A symmetry operation is any action that, if performed on the crystal while you look away, leaves you unable to tell anything happened. The crystal is indistinguishable before and after the operation.

The Identity Operation (E)

The simplest symmetry operation is the identity, denoted EE (from the German Einheit, meaning unity). It does nothing — every atom stays where it is:

E:rrE: \mathbf{r} \rightarrow \mathbf{r}

While trivial, the identity is mathematically essential. It serves as the identity element of the symmetry group, satisfying Eg=gE=gE \cdot g = g \cdot E = g for every operation gg. Every crystal, no matter how asymmetric, possesses at least this one symmetry operation.

In matrix form, the identity is the 3×33 \times 3 identity matrix:

E=(100010001)E = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{pmatrix}

Rotation Operations

A proper rotation CnC_n rotates the crystal by 2πn\frac{2\pi}{n} radians (or 360n\frac{360^\circ}{n}) about an axis. The integer nn is called the order of the rotation.

OperationAngleDescription
C1360 degFull rotation = identity (E)
C2180 degHalf turn
C3120 degOne-third turn
C490 degQuarter turn
C660 degOne-sixth turn

Repeated application of CnC_n generates a series of operations: Cn,Cn2,Cn3,,Cnn=EC_n, C_n^2, C_n^3, \ldots, C_n^n = E. For example, C4C_4 generates the set {C4,C42=C2,C43,C44=E}\{C_4, C_4^2 = C_2, C_4^3, C_4^4 = E\}.

Rotation Matrices About the z-Axis

A rotation by angle θ\theta about the zz-axis has the matrix:

Rz(θ)=(cosθsinθ0sinθcosθ0001)R_z(\theta) = \begin{pmatrix} \cos\theta & -\sin\theta & 0 \\ \sin\theta & \cos\theta & 0 \\ 0 & 0 & 1 \end{pmatrix}

The specific matrices for the crystallographic rotations about z are:

C2C_2 (180 deg rotation about z)

C2=(100010001)C_2 = \begin{pmatrix} -1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & 1 \end{pmatrix}

C3C_3 (120 deg rotation about z)

C3=(1232032120001)C_3 = \begin{pmatrix} -\frac{1}{2} & -\frac{\sqrt{3}}{2} & 0 \\[6pt] \frac{\sqrt{3}}{2} & -\frac{1}{2} & 0 \\[6pt] 0 & 0 & 1 \end{pmatrix}

C4C_4 (90 deg rotation about z)

C4=(010100001)C_4 = \begin{pmatrix} 0 & -1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1 \end{pmatrix}

C6C_6 (60 deg rotation about z)

C6=(1232032120001)C_6 = \begin{pmatrix} \frac{1}{2} & -\frac{\sqrt{3}}{2} & 0 \\[6pt] \frac{\sqrt{3}}{2} & \frac{1}{2} & 0 \\[6pt] 0 & 0 & 1 \end{pmatrix}

Determinant Check

Every proper rotation matrix has determinant +1+1. This is the quickest way to distinguish rotations from improper operations (which have determinant 1-1).

The Crystallographic Restriction Theorem

Not every rotation is compatible with a periodic lattice. The crystallographic restriction theorem states that only 1-fold, 2-fold, 3-fold, 4-fold, and 6-fold rotations are compatible with translational periodicity in three dimensions.

n{1,2,3,4,6}n \in \{1, 2, 3, 4, 6\}

Why No 5-Fold or 7-Fold?

The proof is elegant. Consider a rotation CnC_n acting on a lattice point R\mathbf{R}. Because the lattice is periodic, the rotated point must also be a lattice point. In matrix form, when the rotation is expressed in the lattice basis (fractional coordinates), all matrix elements must be integers. The trace of the rotation matrix satisfies:

tr(Cn)=1+2cos(2πn)Z\text{tr}(C_n) = 1 + 2\cos\left(\frac{2\pi}{n}\right) \in \mathbb{Z}

Since cos(2π/n)\cos(2\pi/n) must yield an integer trace, only the values cos(2π/n){1,12,0,12,1}\cos(2\pi/n) \in \{-1, -\frac{1}{2}, 0, \frac{1}{2}, 1\} are allowed, giving n{1,2,3,4,6}n \in \{1, 2, 3, 4, 6\}.

nAngle (deg)cos(2pi/n)TraceAllowed?
136013Yes
2180-1-1Yes
3120-1/20Yes
49001Yes
5720.309...1.618...No
6601/22Yes
751.40.623...2.247...No
8450.707...2.414...No

Quasicrystals

Five-fold symmetry does appear in nature — in quasicrystals, discovered by Dan Shechtman in 1984 (Nobel Prize 2011). However, quasicrystals are not periodic in the traditional sense: they possess long-range order without translational periodicity.

Reflection (Mirror Planes)

A reflection (denoted σ\sigma) maps every point to its mirror image through a plane. Mirror planes are classified by their orientation relative to the principal rotation axis:

SymbolNameDescription
sigma_hHorizontalPerpendicular to principal axis
sigma_vVerticalContains the principal axis
sigma_dDihedralBisects two C2 axes

The matrix for reflection through the xyxy-plane (a horizontal mirror σh\sigma_h) is:

σh=(100010001)\sigma_h = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & -1 \end{pmatrix}

Similarly, reflection through the xzxz-plane flips the yy-coordinate:

σxz=(100010001)\sigma_{xz} = \begin{pmatrix} 1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & 1 \end{pmatrix}

A reflection has determinant 1-1 and is its own inverse: σ2=E\sigma^2 = E. Physically, applying a mirror reflection twice returns every atom to its original position.

Chirality

A molecule or crystal that lacks any mirror planes or improper rotations is called chiral. Chiral crystals can exhibit optical activity — they rotate the plane of polarised light. This is determined entirely by the symmetry operations present (or absent) in the point group.

Inversion

The inversion operation (denoted ii) maps every point r\mathbf{r} to r-\mathbf{r} through a centre of symmetry (called the inversion centre or centre of symmetry):

i:(xyz)(xyz)i: \begin{pmatrix} x \\ y \\ z \end{pmatrix} \rightarrow \begin{pmatrix} -x \\ -y \\ -z \end{pmatrix}

The matrix representation is simply:

i=(100010001)=Ii = \begin{pmatrix} -1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & -1 \end{pmatrix} = -\mathbf{I}

Like reflection, inversion is its own inverse: i2=Ei^2 = E. The determinant of the inversion matrix is (1)3=1(-1)^3 = -1, classifying it as an improper operation.

Physical Significance

The presence or absence of inversion symmetry has profound physical consequences:

  • Centrosymmetric crystals (with inversion) cannot exhibit piezoelectricity or second-harmonic generation.
  • Non-centrosymmetric crystals (without inversion) can be piezoelectric. Zinc blende CdSe is a classic example — it lacks inversion symmetry and is therefore piezoelectric.
  • In X-ray diffraction, Friedel's law (F(hkl)=F(hˉkˉlˉ)|F(hkl)| = |F(\bar{h}\bar{k}\bar{l})|) holds exactly only for centrosymmetric crystals.

Improper Rotations

An improper rotation SnS_n is a compound operation: a rotation CnC_n followed by a reflection σh\sigma_h through the plane perpendicular to the rotation axis:

Sn=σhCnS_n = \sigma_h \cdot C_n

Special cases connect to operations we already know:

OperationEquivalent toExplanation
S1sigma_h360 deg rotation (= E) followed by reflection = just a reflection
S2i (inversion)180 deg rotation + reflection = inversion
S3C3 + sigma_h120 deg rotation + reflection
S4C4 + sigma_h90 deg rotation + reflection (irreducible)
S6C6 + sigma_h = C3 + i60 deg rotation + reflection

The matrix for S4S_4 about the z-axis is:

S4=σhC4=(100010001)(010100001)=(010100001)S_4 = \sigma_h \cdot C_4 = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & -1 \end{pmatrix} \begin{pmatrix} 0 & -1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1 \end{pmatrix} = \begin{pmatrix} 0 & -1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & -1 \end{pmatrix}

The determinant is 1-1, confirming that S4S_4 is an improper operation.

Classification Rule

All symmetry operations fall into two classes: proper (det = +1: rotations and identity) and improper (det = -1: reflections, inversion, and improper rotations). This classification is encoded in the determinant of the matrix representation.

Matrix Representation

Every symmetry operation can be represented as a 3×33 \times 3 matrix W\mathbf{W} acting on position vectors. For a point symmetry operation (no translational component), the transformed position is:

r=Wr\mathbf{r}' = \mathbf{W} \cdot \mathbf{r}

The matrices W\mathbf{W} for all crystallographic point symmetry operations share these properties:

  • Orthogonal: WTW=I\mathbf{W}^T \mathbf{W} = \mathbf{I} (lengths and angles are preserved)
  • Determinant: det(W)=+1\det(\mathbf{W}) = +1 for proper rotations, det(W)=1\det(\mathbf{W}) = -1 for improper operations
  • Integer entries in fractional basis: When expressed in the lattice vector basis, the matrix elements are integers (0 or ±1 for cubic systems)

Complete Set for the Cubic System

Here is a summary of key symmetry matrices in Cartesian coordinates:

OperationMatrix (diagonal or key elements)det
E (identity)diag(1, 1, 1)+1
i (inversion)diag(-1, -1, -1)-1
C2 about zdiag(-1, -1, 1)+1
C4 about z[[0,-1,0],[1,0,0],[0,0,1]]+1
sigma_h (xy plane)diag(1, 1, -1)-1
sigma_v (xz plane)diag(1, -1, 1)-1
S4 about z[[0,-1,0],[1,0,0],[0,0,-1]]-1
🐍python
1import numpy as np
2
3# Define fundamental symmetry matrices
4E = np.eye(3)                          # Identity
5i_inv = -np.eye(3)                     # Inversion
6C2z = np.diag([-1, -1, 1])            # C2 about z
7C4z = np.array([[0,-1,0],[1,0,0],[0,0,1]], dtype=float)  # C4 about z
8sigma_h = np.diag([1, 1, -1])         # Mirror in xy-plane
9sigma_v = np.diag([1, -1, 1])         # Mirror in xz-plane
10S4z = sigma_h @ C4z                    # Improper rotation S4 about z
11
12# Verify properties
13for name, W in [("E", E), ("i", i_inv), ("C2z", C2z), ("C4z", C4z),
14                ("sigma_h", sigma_h), ("S4z", S4z)]:
15    det = np.linalg.det(W)
16    orth = np.allclose(W.T @ W, np.eye(3))
17    print(f"{name:8s}: det = {det:+.0f}, orthogonal = {orth}")

Combining Operations

Symmetry operations can be composed (applied one after another) by multiplying their matrices. If operation AA is applied first and then operation BB, the combined operation is:

C=BAC = B \cdot A

Note the order: BAB \cdot A means "apply AA first, then BB." Matrix multiplication is generally not commutative, so ABBAA \cdot B \neq B \cdot A in general.

Example: Two C4 Rotations

Applying C4C_4 about z twice should give C2C_2 about z:

C42=C4C4=(010100001)(010100001)=(100010001)=C2C_4^2 = C_4 \cdot C_4 = \begin{pmatrix} 0 & -1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} 0 & -1 & 0 \\ 1 & 0 & 0 \\ 0 & 0 & 1 \end{pmatrix} = \begin{pmatrix} -1 & 0 & 0 \\ 0 & -1 & 0 \\ 0 & 0 & 1 \end{pmatrix} = C_2

Group Multiplication Table

For a small set of operations, we can build a group multiplication table (also called a Cayley table) that shows the result of every pairwise composition. Here is the table for the group {E,C2,σv,σv}\{E, C_2, \sigma_v, \sigma_v'\} (the point group C2vC_{2v}):

First \ SecondEC2sigma_vsigma_v'
EEC2sigma_vsigma_v'
C2C2Esigma_v'sigma_v
sigma_vsigma_vsigma_v'EC2
sigma_v'sigma_v'sigma_vC2E

This table satisfies all four group axioms: closure (every product is in the set), associativity, identity (E), and inverses (each element appears exactly once in every row and column).


VASP Connection: Symmetry Detection

VASP automatically detects symmetry operations at the start of every calculation. The key parameter controlling this behaviour is the ISYM tag in the INCAR file:

ISYM valueBehaviour
0Symmetry completely switched off
1Use symmetry (default for non-US-PP)
2Use symmetry, more efficient (default for PAW/US-PP)
3Force symmetry, do not check positions
-1Use symmetry only to construct initial charge density

INCAR Example

📝text
1# INCAR: Symmetry settings
2ISYM   = 2    # Use symmetry (recommended for PAW)
3SYMPREC = 1E-5  # Tolerance for symmetry detection (angstroms)
4
5# For defect calculations where symmetry is broken:
6# ISYM = 0   # Turn off symmetry completely

Reading Symmetry from OUTCAR

After a VASP run, the detected symmetry operations are printed near the top of the OUTCAR file. Look for lines like:

📝text
1Found 24 space group operations (of which 24 are symmorphic)
2
3 irot       det(A)        alpha          n_x          n_y          n_z
4    1     1.000000     0.000000     1.000000     0.000000     0.000000
5    2     1.000000   120.000000     0.577350     0.577350     0.577350
6    3     1.000000   240.000000     0.577350     0.577350     0.577350
7   ...

Each line gives the determinant (proper vs improper), the rotation angle, and the rotation axis. This information tells you exactly which point group VASP has identified for your structure.

SYMPREC Sensitivity

If your atomic positions are slightly off (e.g., from a molecular dynamics snapshot), VASP may detect fewer symmetry operations than expected. Increasing SYMPREC can help, but setting it too large may force incorrect symmetry. The default of 10510^{-5} Å is usually appropriate for well-relaxed structures.

Checking Symmetry with Python

The spglib library can identify space groups and symmetry operations from atomic positions before you even run VASP. This is extremely useful for verifying that your POSCAR has the expected symmetry.
🐍python
1import spglib
2import numpy as np
3
4# CdSe zinc blende structure
5lattice = np.diag([6.077, 6.077, 6.077])
6positions = [
7    [0.000, 0.000, 0.000],  # Cd
8    [0.500, 0.500, 0.000],  # Cd
9    [0.500, 0.000, 0.500],  # Cd
10    [0.000, 0.500, 0.500],  # Cd
11    [0.250, 0.250, 0.250],  # Se
12    [0.750, 0.750, 0.250],  # Se
13    [0.750, 0.250, 0.750],  # Se
14    [0.250, 0.750, 0.750],  # Se
15]
16numbers = [48, 48, 48, 48, 34, 34, 34, 34]  # Cd=48, Se=34
17
18cell = (lattice, positions, numbers)
19sym = spglib.get_symmetry(cell)
20print(f"Number of symmetry operations: {len(sym['rotations'])}")
21# Output: 24 (for zinc blende Td symmetry)
22
23spg = spglib.get_spacegroup(cell)
24print(f"Space group: {spg}")
25# Output: F-43m (216)

Summary

Symmetry operations are the building blocks of crystal symmetry. Every operation that maps a crystal onto itself can be expressed as a 3×33 \times 3 orthogonal matrix, and the collection of all such operations forms a mathematical group.

  1. Identity (E): Does nothing; the mandatory trivial symmetry.
  2. Proper rotations (CnC_n): Rotation by 2π/n2\pi/n; only n=1,2,3,4,6n = 1, 2, 3, 4, 6 are compatible with lattice periodicity.
  3. Reflections (σ\sigma): Mirror planes; classified as horizontal, vertical, or dihedral.
  4. Inversion (i): Maps rr\mathbf{r} \to -\mathbf{r}; its presence determines centrosymmetry and rules out piezoelectricity.
  5. Improper rotations (SnS_n): Compound rotation + reflection; special cases include S1=σS_1 = \sigma and S2=iS_2 = i.
  6. Matrix representation: All operations are orthogonal matrices with det = ±1, enabling computational manipulation.
  7. VASP uses symmetry (ISYM tag) to reduce computational cost by exploiting equivalent k-points and equivalent atoms.
Looking Ahead: Individual symmetry operations are ingredients. In the next section, we will combine them into point groups — complete sets of operations that describe the macroscopic symmetry of a crystal. The 32 crystallographic point groups classify every crystal into one of 7 crystal systems.
Loading comments...