Anatomy of Character Tables

Character tables are not a museum piece of nineteenth-century mathematics; they are the daily working tool of anyone assigning vibrational spectra. Infrared and Raman spectroscopists use the linear and quadratic function columns constantly: a peak's presence or absence in an IR spectrum, or its presence or absence in a Raman spectrum of the same compound, is diagnosed by checking which Mulliken symmetry species the corresponding normal mode belongs to and whether that species appears in the $x,y,z$ column (IR) or the quadratic column (Raman). This is exactly how vibrational spectroscopists distinguish cis from trans isomers of square-planar and octahedral metal complexes — the two isomers belong to different point groups ($C_{2v}$ versus $D_{4h}$, for instance), predicting different numbers of IR- and Raman-active carbonyl stretches, a classic diagnostic used routinely for metal carbonyl complexes like $\text{M(CO)}_4\text{X}_2$ isomers. In ligand-field and crystal-field theory, the character table for the relevant point group ($O_h$ for a regular octahedral complex, $T_d$ for tetrahedral) is what predicts that the five d-orbitals split into $e_g + t_{2g}$ (octahedral) or $e + t_2$ (tetrahedral) sets, directly from the quadratic function columns of those tables. This single piece of group theory underlies the entire qualitative structure of crystal-field splitting diagrams used to rationalize colors, magnetism, and reaction mechanisms of transition-metal complexes throughout inorganic chemistry. In molecular orbital theory, Symmetry-Adapted Linear Combinations of ligand orbitals — essential for building qualitative and semi-quantitative MO diagrams of everything from water to ferrocene to metal-cluster compounds — are constructed using exactly the reduction-formula machinery whose foundation is the character table's rows and columns. Computational chemistry packages (Gaussian, ORCA, ADF, and others) exploit point-group symmetry, read directly from these tables, to block-diagonalize Fock and Hamiltonian matrices, cutting computational cost enormously for symmetric molecules — a direct industrial and academic payoff of the abstract orthogonality relations underlying the table. Beyond molecules, character tables of space groups (the crystallographic extension of point groups) underlie the interpretation of X-ray diffraction selection rules, phonon dispersion curves in solid-state physics, and the classification of electronic bands in semiconductors — all descendants of the same Mulliken-inspired bookkeeping applied to the translational symmetry of a periodic lattice rather than a single finite molecule.

Ask the tutor: Can you walk through constructing the reducible representation for the four N–H bonds of ammonia and reducing it using the $C_{3v}$ character table?

Picture the water molecule, $C_{2v}$ symmetry, sitting with its $C_2$ axis along $z$ and its two hydrogens symmetric about the $xz$ and $yz$ planes. The point group has four symmetry operations: the identity $E$, the $C_2$ rotation, and two mirror planes $\sigma_v(xz)$ and $\sigma_v'(yz)$. Because $C_{2v}$ is abelian (every operation commutes with every other), each operation is in a class by itself — four operations, four classes, four columns in the table. Each of these operations, acting on any basis (a set of orbitals, bond vectors, or displacement coordinates), can be represented by a matrix, and the trace of that matrix is the operation's character for that particular representation. The character table simply tabulates these traces for the smallest possible ("irreducible") representations — the four ways, in this case, that something can transform under $C_{2v}$ without those four ways further splitting into smaller pieces. The rows of the table are the irreps themselves, and they always carry a Mulliken symbol as a name tag. For $C_{2v}$ these are $A_1, A_2, B_1, B_2$. The letter tells you the character under the principal rotation axis: $A$ means symmetric (character $+1$) under $C_2$, $B$ means antisymmetric (character $-1$). The subscript then distinguishes behavior under a secondary symmetry element — here, the subscript 1 or 2 records symmetric or antisymmetric behavior under $\sigma_v(xz)$. This is not arbitrary bookkeeping: it is a compressed way of writing down the entire character row, because the row is exactly what the label describes. The rightmost columns of the table — often boxed off or set in a different font in printed tables — list the linear functions $x, y, z$, the rotations $R_x, R_y, R_z$, and the quadratic functions $x^2, y^2, z^2, xy, xz, yz$ (or combinations like $x^2 - y^2$ and $x^2+y^2$). These entries tell you which irrep a given Cartesian function belongs to. Since the $p_z$ orbital transforms exactly as the coordinate $z$, and the $d_{xy}$ orbital transforms exactly as the product function $xy$, this column is a direct dictionary from abstract symmetry labels to real atomic orbitals. It is also how you read off IR activity (a vibration is IR-active if it shares a symmetry label with $x$, $y$, or $z$, since the transition dipole operator transforms as these) and Raman activity (shares a label with a quadratic function, since polarizability is a quadratic quantity).

Now contrast this with a group that has genuine degeneracy, such as $C_{3v}$ (ammonia). Here the three-fold axis generates operations $C_3$ and $C_3^2$ that are conjugate to each other (related by a similarity transformation within the group), so they are lumped into a single class "$2C_3$." Likewise the three mirror planes form a single class "$3\sigma_v$." The irreps are $A_1$, $A_2$, and $E$; the $E$ irrep has character $2$ under the identity, signaling that it is genuinely two-dimensional — no single number describes how a $(p_x, p_y)$ pair transforms under $C_3$, because $C_3$ mixes $p_x$ and $p_y$ into each other. This is the structural reason degenerate orbitals (like the $e$ set in a trigonal field, or $t_2$ and $e$ in $T_d$) appear in inorganic chemistry: symmetry, not accident, forces certain orbitals to remain equal in energy.

The character of an operation $R$ in representation $\Gamma$ is defined as the trace of its representative matrix:

where: $D_\Gamma(R)$ is the matrix representing operation $R$ in representation $\Gamma$, and the sum runs over the diagonal elements. Reading: the character is a single number per operation per representation; because trace is invariant under change of basis, all operations in the same conjugacy class (related by $D(R') = D(S)D(R)D(S)^{-1}$ for some $S$ in the group) automatically share the same character, which is exactly why a class needs only one column. The dimension of an irrep — and hence its degeneracy — is read directly from the identity column, since $D(E)$ is always the identity matrix of size $\ell \times \ell$:

where: $\ell_\Gamma$ is the dimension of irrep $\Gamma$ (1 for A/B species, 2 for E species, 3 for T species). Reading: you can tell instantly, without knowing anything else about the group, whether an irrep describes a single non-degenerate state (A or B) or a set of orbitals forced to stay degenerate by symmetry (E, doubly; T, triply). The great orthogonality theorem, which is the mathematical machine underlying the whole table, states that irrep matrix elements obey

where: $h$ is the order of the group (total number of symmetry operations), $\ell_i$ is the dimension of irrep $i$, and the Kronecker deltas enforce that different irreps (and different matrix positions within one irrep) are mutually orthogonal. Reading: this is the deep reason character tables have the elegant orthogonality properties used every day in group theory — rows (irreps) are orthogonal vectors when weighted by class size, and so are columns (classes), when treated as vectors in the space of group elements. Taking traces of the orthogonality theorem collapses it to the character orthogonality relation actually used in practice:

where: the sum runs over classes $C_k$, $n_k$ is the number of operations in class $k$, and $\chi_i, \chi_j$ are the characters of irreps $i$ and $j$. Reading: this is precisely the row-orthogonality property that lets you check a character table for self-consistency, and it is also the identity that the Reduction Formula is built from when decomposing a reducible representation into irreps. Finally, a simple but powerful consistency check: the sum of the squares of the irrep dimensions must equal the order of the group,

where: the sum runs over all irreps $i$ of the group. Reading: this equation is a fast sanity check on any character table you write down or look up — if the squared dimensions don't add up to the number of operations in the group, either an irrep is missing or a dimension is wrong.

Example 1 — Foundational: reading $NH_3$'s $C_{3v}$ table to place its lone pair and N–H bonds. The $C_{3v}$ character table (order $h=6$) reads:

under classes $(E, 2C_3, 3\sigma_v)$, with $z$ and $x^2+y^2, z^2$ in $A_1$; $R_z$ in $A_2$; and $(x,y)$, $(R_x,R_y)$, $(xz,yz)$, $(x^2-y^2,xy)$ all in $E$. The nitrogen $2p_z$ orbital points along the $C_3$ axis, so by the same logic as Step 2 of the first derivation (unchanged by every operation of the group, since the axis maps to itself and the mirror planes all contain the axis), it transforms as $A_1$. The nitrogen $2p_x, 2p_y$ pair, lying perpendicular to the axis, gets rotated into linear combinations of each other by $C_3$ — this is exactly the signature of a two-dimensional irrep — so it transforms as $E$. Checking dimension consistency: $\ell(A_1)^2 + \ell(A_2)^2 + \ell(E)^2 = 1+1+4 = 6 = h$. ✓ This matches the general identity $\sum_i \ell_i^2 = h$ exactly. Since the lone pair sits mostly in an $A_1$-symmetry hybrid (mixing $2s$ and $2p_z$, both $A_1$), and the three N–H bonds transform as $A_1 \oplus E$ (one totally symmetric combination plus a degenerate pair), the bonding picture that emerges — one $A_1$ bonding set plus one doubly degenerate $E$ bonding set — is exactly the result used when building Symmetry-Adapted Linear Combinations for ammonia's molecular orbitals.

Example 2 — Applied: is the asymmetric stretch of a bent $XY_2$ molecule ($C_{2v}$, bond angle 104.5°, e.g. a molecule with unequal-looking geometry like $\text{ClO}_2$'s local frame) IR active? Take a generic bent triatomic in $C_{2v}$ with the $C_2$ axis along $z$ and the molecular plane as $\sigma_v(xz)$. The three normal modes are symmetric stretch, bend, and asymmetric stretch. Constructing the reducible representation for atomic Cartesian displacements (3 atoms × 3 coordinates = 9), one obtains $\chi_{\text{red}} = (9, -1, 1, 3)$ under $(E, C_2, \sigma_v, \sigma_v')$ for a molecule with non-collinear, non-equivalent bond lengths of, say, 1.47 Å and 1.47 Å with unequal angle contributions in the reducible trace (the exact number depends only on how many atoms are unmoved by each operation, times the contribution per unmoved atom — a geometric fact independent of bond length). Applying the reduction formula from the second derivation with $h=4$ and $n_k = 1$ for every class in $C_{2v}$:

So $\Gamma_{\text{total}} = 3A_1 \oplus A_2 \oplus 2B_1 \oplus 3B_2$ (summing dimensions: $3+1+2+3=9$ ✓, matching the total basis size). Subtracting the three translations ($A_1+B_1+B_2$, since these are $z$, $x$, $y$ respectively from Example-level lookup) and three rotations ($A_2 + B_1 + B_2$, from $R_z, R_y, R_x$) leaves the vibrational representation:

Three vibrations total (as expected for $3N-6 = 3$), split as $2A_1 + B_2$: the symmetric stretch and the bend are both $A_1$ (totally symmetric, so both must appear together with breathing-type character), and the asymmetric stretch is $B_2$. Checking IR activity: the table shows $z$ transforms as $A_1$, and $y$ transforms as $B_2$ — both symmetry species appear in the linear function column, so all three vibrations are IR active, with the asymmetric stretch specifically active because it shares its symmetry label with the $y$ component of the dipole operator.

Example 3 — Exam/competition level: using character orthogonality to catch a fabricated character table. Suppose a student proposes the following (incorrect) table for a group of order $h=8$ claimed to be $C_{4v}$, with classes $(E, 2C_4, C_2, 2\sigma_v, 2\sigma_d)$:

Check this using the dimension identity first: $1+1+1+1+4 = 8 = h$. ✓ That passes. Now check the deeper character orthogonality condition, Step 2 of the reduction-formula derivation, applied to the $E$ row against itself:

This equals $h=8$ correctly (self-orthogonality requires this sum to equal $h$), so this particular row seems fine at first glance — but check $E$ against $A_1$ (should be zero, by the orthogonality of distinct irreps):

That's also fine. The actual error is more subtle: the genuine $C_{4v}$ character for $E$ under $C_2$ is $\chi_E(C_2) = -2$, which the fabricated table does have correctly — so this particular fabrication happens to be right. This is the pedagogical point of the exercise: orthogonality checks are necessary but not automatically sufficient to catch every possible typo (a table could satisfy all orthogonality sums yet mislabel which basis functions belong where, or could accidentally satisfy sums with a genuinely wrong character in one slot compensated elsewhere). The robust way to derive $\chi_E(C_2)=-2$ independently is to note that in the two-dimensional $E$ representation, $C_2 = C_4^2$, and if $C_4$ is represented by a $90°$ rotation matrix with $\chi_E(C_4)=2\cos(90°)=0$, then $C_2$ corresponds to a $180°$ rotation with $\chi_E(C_2) = 2\cos(180°) = -2$, consistent with the general formula $\chi_E(C_n^k) = 2\cos(2\pi k/n)$ for the two-dimensional irrep built from rotation matrices — confirming the entry by direct construction rather than by orthogonality bookkeeping alone.

The character table as chemists use it descends from the mathematical theory of group representations developed in the late nineteenth and early twentieth centuries by Ferdinand Georg Frobenius, Issai Schur, and William Burnside, but the specific labeling convention every chemistry student memorizes — the Mulliken symbols $A$, $B$, $E$, $T$, with subscripts and $g/u$ superscripts — comes from the American physical chemist Robert S. Mulliken, who in 1955 published the systematic notation that unified the chaotic labeling schemes different spectroscopists had been using for molecular electronic states and vibrations. Mulliken's genius was not inventing new mathematics but recognizing that chemists needed a standardized, physically transparent naming convention layered on top of the already-existing mathematical classification of irreducible representations, so that a symmetry label could be read off a table and immediately connected to a real orbital or vibrational mode without re-deriving representation matrices each time. Mulliken had already won recognition for his pioneering work founding molecular orbital theory itself — describing bonding using delocalized orbitals spread over the whole molecule rather than localized electron-pair bonds — for which he received the Nobel Prize in Chemistry in 1966. His notational system for symmetry species was a natural extension of that molecular-orbital worldview: if electrons occupy orbitals that are inherently properties of the whole molecule, then those orbitals must be classified by the whole molecule's symmetry, and a universal labeling scheme was needed to make this classification communicable across the field. The convention he proposed — $A$ and $B$ for one-dimensional irreps (symmetric or antisymmetric under the principal rotation), $E$ for doubly degenerate, $T$ (originally $F$) for triply degenerate, with $g/u$ subscripts for centrosymmetric groups and primes for reflection behavior — is precisely the system embedded in every character table used today.