Symmetry-Adapted Linear Combinations (SALCs)

SALC construction is the backbone of every modern qualitative MO diagram used to rationalize transition-metal spectroscopy, magnetism, and reactivity. In octahedral and square-planar complexes, the σ- and π-SALC analysis explains the ligand-field splitting pattern that underlies UV-vis spectra used routinely to assign oxidation states and coordination geometries in synthetic and bioinorganic chemistry — for instance, the difference between a strong-field low-spin Fe(II) heme center and a high-spin Fe(III) resting state in cytochromes is a direct consequence of how σ- and π-donor/acceptor SALCs from the porphyrin and axial ligands interact with the iron d-orbital set. In organometallic chemistry, cyclopentadienyl and arene π-SALCs (constructed exactly as in Example 3) are the starting point for understanding the bonding in ferrocene, metallocenes generally, and η⁶-arene "piano-stool" complexes used as catalysts in olefin polymerization and asymmetric hydrogenation; the same Fourier-transform-on-a-ring logic extends directly to fullerene and nanotube frontier-orbital analysis in materials chemistry. In cluster chemistry and cage compounds — boranes, carboranes, and Zintl ions — Wade's rules and skeletal electron counting are underpinned by exactly this SALC machinery applied to vertex orbitals of a polyhedral cage, determining how many cluster bonding MOs a given deltahedron can support. The identical mathematical procedure, applied not to orbitals but to Cartesian displacement vectors on each atom, generates the symmetry-classified normal modes used in vibrational spectroscopy — the projection operator does not care whether the basis functions are electron orbitals or atomic displacement vectors, which is precisely why the SALC formalism developed here feeds forward, essentially unchanged in its mathematics, into Vibrational Modes and Reducible Representations for interpreting infrared and Raman spectra of coordination compounds. Instrumentally, SALC-based selection rules (an integral vanishes unless the product of representations contains the totally symmetric representation) are the rigorous justification behind IR/Raman mutual exclusion rules used daily to assign vibrational spectra of everything from metal carbonyls to solid-state materials, and the same group-theoretic overlap argument underlies computational quantum chemistry packages, which use symmetry-adapted basis functions to block-diagonalize the secular equation and dramatically reduce computational cost in electronic structure calculations.

Ask the tutor: How do the σ- and π-SALC sets combine to produce the full ligand-field splitting diagram (Δ_o or Δ_t) for an octahedral or tetrahedral complex with π-donor or π-acceptor ligands?

Picture the six fluorine σ-orbitals of SF₆ sitting on the six vertices of an octahedron, each pointing its lobe of maximum electron density inward toward sulfur. As a basis set for representing operations of the point group O_h, these six orbitals form a six-dimensional reducible representation Γ_σ. Applying the reduction formula (counting how many orbitals are left unmoved by each symmetry operation, weighted by the phase change if any) gives

This says the six ligand orbitals sort into one totally symmetric combination (all six in phase), a doubly degenerate pair transforming like $d_{z^2}, d_{x^2-y^2}$, and a triply degenerate set transforming like $p_x, p_y, p_z$. Sulfur's valence orbitals are 3s (A₁g), 3p (T₁u), and — in the older, now largely retired hypervalent-bonding picture — 3d (E_g + T₂g). Immediately, without any further calculation, the symmetry analysis tells us: the A₁g ligand SALC can only bond with the sulfur 3s orbital; the T₁u ligand SALCs can only bond with the three sulfur 3p orbitals; the E_g ligand SALCs can only bond with an E_g-symmetry central orbital (sulfur d_{z²}/d_{x²-y²}, in the extended basis) or, in the modern picture, remain essentially nonbonding/purely ligand-based since sulfur has no valence d orbital genuinely available at low energy. Crucially, there is no T₂g combination in Γ_σ at all — the ligand σ set simply does not contain any orbital of that symmetry, so sulfur d_{xy}, d_{xz}, d_{yz} (T₂g) are strictly nonbonding with respect to σ donation, regardless of energy matching. The projection operator is the machine that turns "Γ_σ contains A₁g + E_g + T₁u" into actual, explicit combinations such as

It works by summing, over every operation R in the group, the effect of R on one reference orbital, weighted by the character of R in the target irreducible representation. Operations that belong to the correct "symmetry channel" reinforce; operations that belong to the wrong channel cancel by destructive interference — exactly as a Fourier component isolates one frequency from a signal by weighted summation over a periodic basis.

The E pair deserves a second look, because it is generated differently from the A₁ combination and illustrates the central subtlety of projecting onto degenerate representations: a first projection gives one member of the pair mixed with junk that must be removed, and the second member must be obtained by projecting a different reference orbital and then orthogonalizing against the first.

The starting point is always the reducible representation of the ligand basis, obtained by counting unshifted orbitals with sign/phase weighting under each class of the point group:

where: the sum runs only over basis orbitals whose position is unmoved by operation $R$ (orbitals that move to a different atom contribute zero to the trace); the "contribution per orbital" is $+1$ for an orbital symmetric under $R$ (e.g. an s-type σ orbital, unchanged by any proper or improper operation), or $\pm1$ depending on how a directional orbital (like a p-orbital lobe) transforms under a reflection or rotation that leaves the atom fixed but may invert the lobe's phase. Reading: $\chi_{\text{basis}}(R)$ is nothing but the ordinary character (trace) of the representation matrix for $R$ in the ligand basis; because basis orbitals on different atoms never mix under a symmetry operation of the molecule (an operation either maps atom $k$ to itself or to a different atom, never partially), the matrix is automatically block-diagonal with $1\times1$ blocks (or $\pm1$ blocks) on the diagonal for unmoved atoms and zero rows/columns for moved atoms — which is exactly why this simple counting rule works instead of requiring a full matrix multiplication. The reduction formula (assumed known) then gives the number of times irreducible representation $i$ appears:

where: $h$ is the order of the group, $g(R)$ the number of operations in the class of $R$, $\chi^{(i)}(R)$ the character of irreducible representation $i$ for that class. Reading: this is an inner product on the space of class functions — $n_i$ is literally the "overlap" of the reducible representation with the pure irreducible representation $i$, and because irreducible representations form an orthonormal basis for class functions (a consequence of the Great Orthogonality Theorem), this overlap integral always returns a non-negative integer, which is the multiplicity of $i$ in the reducible representation. The projection operator, the actual object that builds explicit orbital combinations rather than just counting them, is

where: $l_i$ is the dimension of irreducible representation $i$ (1 for A/B labels, 2 for E, 3 for T), $h$ the group order, $\chi^{(i)}(R)$ the character table entry for $R$ in representation $i$, and $\hat{R}$ the operator that carries out the symmetry operation $R$ on whatever function it is applied to (moving basis orbitals to the positions their images occupy after the operation). Reading: applied to any reference basis orbital $\phi_1$, $\hat{P}^{(i)}\phi_1$ returns (up to normalization) the component of $\phi_1$ that transforms as representation $i$ — it is a symmetry filter, exactly analogous to a projection operator in ordinary linear algebra picking out the component of a vector along a given eigenspace, except here the "eigenspace" is an irreducible representation of the group rather than an eigenvalue of a single operator.

Example 1 — Foundational: the two ligand SALCs of a linear AB₂ molecule (D∞h, or its finite-order H₂O-type analogue with C₂ᵥ collapsed to the simplest possible case, treated first as D∞h/O for clean arithmetic, then bridged to the finite-group habit). For linear XY₂ (point group $D_{\infty h}$, but treat via the well-defined finite subgroup $C_{2v}$ result which every general chemistry course actually uses first), consider two ligand σ orbitals $\sigma_1,\sigma_2$ on either side of a central atom in $C_{2v}$ symmetry ($E, C_2, \sigma_v(xz), \sigma_v'(yz)$, with $C_2$ swapping the two ligands and both mirrors either fixing or swapping them depending on convention — take the standard convention where the molecular plane is $xz$, so $\sigma_v(xz)$ fixes both ligands and $C_2$ together with $\sigma_v'(yz)$ swaps them). Reducible representation: $\chi(E)=2$ (both fixed), $\chi(C_2)=0$ (both swapped), $\chi(\sigma_v(xz))=2$ (both fixed, lying in-plane), $\chi(\sigma_v'(yz))=0$ (both swapped). Reduction against $C_{2v}$ character table ($A_1=(1,1,1,1)$, $B_1=(1,-1,1,-1)$):

so $\Gamma_\sigma = A_1 + B_1$. Projecting $A_1$: $\hat P^{(A_1)}\sigma_1 = \frac{1}{4}[\sigma_1 + \sigma_2 + \sigma_1 + \sigma_2] = \frac12(\sigma_1+\sigma_2)$, normalized:

Projecting $B_1$: $\hat P^{(B_1)}\sigma_1 = \frac14[\sigma_1 - \sigma_2 + \sigma_1 - \sigma_2] = \frac12(\sigma_1-\sigma_2)$, normalized:

These match central-atom $s$ (or $p_z$, both $A_1$ in $C_{2v}$ with this axis convention) and $p_x$ ($B_1$) respectively — reproducing the familiar three-center MO picture of a linear/bent AB₂ fragment, with the crucial point that the out-of-phase combination $\psi(B_1)$ has zero net overlap with any $A_1$-symmetry central orbital, however good the energy match, purely by symmetry.

Example 2 — Applied: σ-SALCs for square-planar PtCl₄²⁻ (D₄h), locating which platinum orbitals can bond. Four chloride σ orbitals sit at the vertices of a square, point group $D_{4h}$ (order $h=16$; classes $E, 2C_4, C_2, 2C_2', 2C_2'', i, 2S_4, \sigma_h, 2\sigma_v, 2\sigma_d$). Unmoved-orbital count per class: $E$: 4; $C_4$: 0 (all four cycle); $C_2$ (the principal axis, $=C_4^2$): 0 (opposite pairs swap — actually for a square, $C_2$ about the principal axis sends each ligand to the diagonally opposite one, so 0 fixed); $C_2'$ (through two opposite ligands): 2 fixed; $C_2''$ (through opposite edge midpoints): 0 fixed; $i$: 0 fixed (inversion swaps opposite ligands); $S_4$: 0; $\sigma_h$ (molecular plane, contains all four ligands): 4 fixed; $\sigma_v$ (through opposite ligands, i.e. contains the $C_2'$ axes): 2 fixed; $\sigma_d$ (through edge midpoints): 0 fixed.

across classes $(E, 2C_4, C_2, 2C_2', 2C_2'', i, 2S_4, \sigma_h, 2\sigma_v, 2\sigma_d)$. Reducing against the $D_{4h}$ table (order $h=16$) picks out $A_{1g}, B_{1g}, E_u$ — the standard textbook result for a square-planar ML₄ σ-only ligand set:

(dimension check: $1+1+2=4$ ✓). Explicit projection (parallel to Examples 1 and the $T_d$ case above, omitted stroke-by-stroke for space) yields

(with ligands numbered sequentially around the square, ligand 1 and 3 trans to each other along $x$, ligands 2 and 4 trans along $y$). Matching to platinum valence orbitals by Mulliken label: $A_{1g}$ matches $6s$ (and $5d_{z^2}$, also $A_{1g}$ in $D_{4h}$); $B_{1g}$ matches $5d_{x^2-y^2}$ — the orbital whose large lobes point directly at all four ligands, explaining why $d_{x^2-y^2}$ is pushed to very high energy in square-planar crystal-field splitting, exactly the antibonding partner of this $B_{1g}$ SALC; $E_u$ matches the degenerate pair $(6p_x, 6p_y)$. No SALC of $A_{2g}, B_{2g}, E_g,$ or any ungerade-except-$E_u$ label appears, so $5d_{xy}$ ($B_{2g}$) and $5d_{xz},5d_{yz}$ ($E_g$) are strictly σ-nonbonding — they may still engage in π-bonding with ligand π-symmetric SALCs (a separate reducible representation built from ligand p-orbitals perpendicular to the Pt–Cl bonds), but never with these σ SALCs, by symmetry alone.

Example 3 — Exam/competition level: π-SALCs of the cyclopentadienyl ligand (C₅ᵥ local treatment via the free-ion D₅h ring), used in ferrocene-type sandwich bonding, worked with explicit complex-to-real recombination for the degenerate pairs. The cyclopentadienyl anion Cp⁻ has five carbon $p_z$ orbitals perpendicular to the ring plane, point group $D_{5h}$ (order $h=20$). The reducible representation for the five $p_z$ orbitals, using the convention that a $\sigma_h$ (the ring plane itself) sends every $p_z$ orbital to minus itself (since $p_z$ is antisymmetric with respect to reflection through the plane containing the ring): Classes of $D_{5h}$: $E, 2C_5, 2C_5^2, 5C_2, \sigma_h, 2S_5, 2S_5^3, 5\sigma_v$. Unmoved orbitals with sign: $E$: 5 orbitals, each $+1$, $\chi=5$. $C_5, C_5^2$: no atom fixed (5-fold rotation with 5 atoms, none coincide with themselves unless multiple of $2\pi$), $\chi=0$. $C_2$ (in-plane axes through one atom and the opposite edge-midpoint): 1 atom fixed, but $C_2$ rotates the $p_z$ orbital on that atom by 180° about an axis in the nodal plane, flipping its sign — contributes $-1$; the other four atoms move, contribute 0: $\chi=-1$. $\sigma_h$: all 5 atoms fixed in-plane, but each $p_z$ orbital flips sign under this reflection: $\chi=5\times(-1)=-5$. $S_5,S_5^3$ (rotation-reflection, combines a $C_5$-type rotation, which fixes no atom, with the $\sigma_h$ sign flip): $\chi=0$ (no atom fixed regardless of sign). $\sigma_v$ (vertical planes each containing one C and bisecting the opposite edge): fixes one atom's position, but $\sigma_v$ is a vertical mirror, and $p_z$ (perpendicular to the ring, i.e. lying in every vertical plane) is symmetric under a vertical reflection that contains the $z$-axis through that atom, so contributes $+1$ for the fixed atom: $\chi = 1$.

across $(E, 2C_5, 2C_5^2, 5C_2, \sigma_h, 2S_5, 2S_5^3, 5\sigma_v)$. Reducing against $D_{5h}$ (irreducible representations relevant here carry a subscript $''$ for antisymmetric-under-$\sigma_h$, matching the $p_z$ character): the standard, checkable result is

(dimension check $1+2+2=5$ ✓; every representation here is a double-prime type, consistent with $p_z$ being odd under $\sigma_h$ — a mismatch with any single-prime metal orbital would forbid bonding regardless of energy). Projection of the totally symmetric ($A_2''$) combination onto reference orbital $p_{z,1}$, using $\chi^{(A_2'')}=(1,1,1,-1,-1,-1,-1,-1)$ (characters for $A_2''$: symmetric under rotations, antisymmetric under $\sigma_h$ and vertical mirrors — matching the $''$ label and the $A_2$-type sign pattern under $C_2$/$\sigma_v$):

— all five $p_z$ orbitals fully in phase, the classic "all-bonding" π MO of Cp⁻, matching a metal $s$ or $p_z$/$d_{z^2}$ orbital of the appropriate parity in the sandwich complex. The $E_1''$ pair (one nodal plane, matching metal $(p_x,p_y)$ or $(d_{xz},d_{yz})$ depending on overall molecular symmetry in the full sandwich) and $E_2''$ pair (two nodal planes, matching metal $(d_{x^2-y^2},d_{xy})$) are generated by projecting with $\chi^{(E_1'')}$ and $\chi^{(E_2'')}$ respectively; because $D_5$ has real characters only for the $A$ and (via pairing complex-conjugate 1-D irreducible representations) genuinely 2-D $E_1,E_2$ blocks, the practical route is to build complex combinations first,

(this is exactly a discrete Fourier transform over the 5 ring positions — $n=0$ gives $\psi(A_2'')$ above), and then recombine the complex-conjugate pairs $n=\pm1$ and $n=\pm2$ into real, orthogonal SALCs:

with $\cos72^\circ = 0.309$ and $\cos144^\circ = -0.809$ giving the fully explicit numerical combination (an asymmetric, non-round-number set of coefficients — precisely the signature of a genuine 5-fold system, unlike the clean $\pm1,\pm\tfrac12$ coefficients that appear in 4-fold or 6-fold rings). This Fourier-transform structure is the general pattern for any cyclic $C_n$ or $D_{nh}$ π system (also underlying benzene's $a_{2u}+e_{1g}+e_{2u}$ π-SALC set), and it is exactly the discrete analogue of the plane-wave sum that appears when treating a cyclic system as a 1-D "electron on a ring."

Limiting-case check for Example 3. As the ring size $n\to\infty$ (uniform continuous ring), the discrete Fourier sum $\phi_n=\frac{1}{\sqrt n}\sum_k e^{2\pi ink/n}p_{z,k}$ becomes the continuous Fourier basis $e^{im\theta}$ for angular momentum $m$ around the ring — recovering the free-electron-on-a-ring energy ladder $E_m \propto m^2$, with $m=0$ (the all-in-phase $A$-type combination) lowest, exactly mirroring the aromatic-stabilization ordering familiar from Hückel theory, confirming the SALC construction reduces correctly to the known continuum limit.

Bethe's original 1929 treatment predates the ligand-orbital SALC formalism by roughly two decades; it was primarily a crystal-field (purely electrostatic, no covalency) treatment of metal d-orbital splitting. The extension to a fully covalent, MO-theoretic treatment — building explicit symmetry-adapted combinations of ligand orbitals and mixing them with metal orbitals of matching symmetry — is due to the joint development of ligand-field theory through the 1950s, with major contributions from John Van Vleck (who bridged crystal-field electrostatics and molecular-orbital covalency in a single unified framework) and from the systematic application of projection-operator group theory to inorganic bonding problems in the textbooks and papers of F. Albert Cotton, whose "Chemical Applications of Group Theory" (1963) is the direct source of the modern, routinely taught projection-operator SALC procedure used throughout inorganic chemistry today.