Crystal Field Theory: Basic Model and Assumptions

Take a colourless aqueous solution of Na₂SO₄ and a colourless solution of NiCl₂. Mix them, and nothing happens. But dissolve NiCl₂ alone in water and you get a vivid green solution. Add ammonia and it turns deep blue-violet. The metal ion has not changed its oxidation state, its electron count, or its identity. Only the arrangement of ligands around it has changed. Something about surrounding a transition-metal ion with a specific geometry of charged or dipolar groups is reorganizing its electrons. This reorganization absorbs specific wavelengths of visible light. Crystal field theory (CFT) is the model built to explain exactly this. It treats the ligands as nothing more than point charges (or point dipoles) arranged around a metal ion. It asks how that electrostatic environment reorganizes the energies of the metal's five d orbitals. This matters because colour, magnetism, and much of the thermodynamic stability of transition-metal complexes trace back to the same root cause. The five degenerate d orbitals of a free ion split into two or more energy groups once ligands are attached. CFT is deliberately the simplest possible model of this splitting — purely electrostatic, no covalency, no orbital overlap. Yet it predicts, with an accuracy that still surprises people, the number of split levels and their relative ordering. Once parametrized, it even predicts quantitative trends in color and magnetism across the transition series. It is the entry point to essentially every discussion of transition-metal spectroscopy and magnetochemistry. It also sets up the machinery — $\Delta_o$ , $\Delta_t$ , high spin/low spin, Jahn–Teller distortion — used throughout Coordination Chemistry: Structure & Bonding. The big picture is this: a free transition-metal ion has five d orbitals that are degenerate by spherical symmetry. Bring up six ligands to form an octahedron, or four ligands to form a tetrahedron or a square plane. The spherical symmetry is then broken. The d orbitals no longer all interact equally with the ligand point charges. Orbitals pointing directly at the ligands are pushed up in energy more than orbitals pointing between the ligands. The pattern of splitting is fixed entirely by symmetry — group theory guarantees the number and degeneracies of the resulting levels. The size of the splitting, however, is fixed by the electrostatics — the charge, the metal–ligand distance, and the radial extent of the d orbitals. This article builds the model from the ground up. It covers the physical assumptions CFT makes, and the electrostatic Hamiltonian for a point-charge field. It explains how spherical symmetry is broken by an octahedral or tetrahedral arrangement of charges. It also explains why the d orbitals split into two groups, rather than remaining degenerate or splitting into five distinct levels. It stops just short of deriving the actual splitting pattern and magnitude for the octahedron. That quantitative step, together with $\Delta_o$ , high-spin/low-spin crossover, and crystal field stabilization energy, is the subject of Crystal Field Theory: Octahedral Splitting.

Key takeaways

CFT models ligands as point negative charges (or point dipoles) that create a purely electrostatic field around a metal ion — no covalent metal–ligand bonding is assumed.
In a free ion the five d orbitals are degenerate; any non-spherical arrangement of surrounding charges necessarily lifts this degeneracy, by symmetry alone.
The pattern of splitting (how many groups, their degeneracies) is fixed by the point group of the ligand arrangement; the magnitude of splitting is an electrostatic energy scale set by charge, distance, and radial wavefunction.
Orbitals whose lobes point directly at ligand charges are destabilized more than orbitals whose lobes point between ligands — this asymmetry is the entire origin of "crystal field splitting."
The barycenter (weighted average) of the d orbital energies is preserved: splitting redistributes energy among the d orbitals without changing their average, by construction of the electrostatic perturbation.
CFT is known to be physically incomplete — real M–L bonds have real covalency — yet it reproduces the symmetry of splitting patterns exactly and the qualitative size trends astonishingly well.
Key idea: treating ligands as point charges in a definite geometry is enough, by symmetry alone, to lift the five-fold degeneracy of the d orbitals into two or more distinct energy levels.

The core insight

Crystal field theory replaces the full quantum-mechanical metal–ligand bond with an electrostatic point-charge (or point-dipole) perturbation. The resulting splitting of the d orbitals is a direct, calculable consequence of how each d-orbital's angular shape overlaps in space with the positions of the surrounding point charges.

Where this lives today

Crystal field theory's basic electrostatic model, despite its known limitations, remains the working intuition behind an enormous range of modern chemistry and materials science. In geochemistry and mineralogy, crystal field stabilization energy explains why certain transition-metal ions (Cr³⁺, Ni²⁺) strongly prefer octahedral sites in mineral lattices such as spinel and olivine, directly controlling trace-element partitioning between magma and crystallizing minerals — a foundational tool in igneous petrology. In materials science, the same electrostatic splitting logic (extended to more exotic coordination environments) underlies the design of phosphors and laser gain media, most famously the ruby laser, where Cr³⁺ substituted octahedrally into Al₂O₃ produces the characteristic red fluorescence exploited by Maiman's original 1960 laser. In bioinorganic chemistry, crystal-field-style reasoning about high-spin versus low-spin iron centers explains the spin-state switching in hemoglobin as it binds O₂, and the site-selective incorporation of Zn²+, Cu²+, and Fe²+/Fe³+ into metalloenzyme active sites is rationalized in the first instance using CFSE arguments before more detailed density-functional calculations are brought in. In analytical and industrial chemistry, the colours of transition-metal pigments and gemstones (the green of emerald and the red of ruby both come from Cr³⁺ in an octahedral oxide field, with the difference in colour arising entirely from differences in the crystal field strength imposed by the surrounding lattice) are taught and rationalized using exactly the splitting diagram developed here. And pedagogically, crystal field theory remains the mandatory first framework students learn before moving to the more complete but more mathematically demanding ligand field theory and molecular orbital treatments of transition-metal bonding — the topic picked up immediately in the discussion of quantitative octahedral splitting, $\Delta_o$ , and the spectrochemical series.

Ask the tutor: How does the size of the octahedral splitting parameter Δo actually depend quantitatively on the ligand's charge and the metal–ligand bond distance?

The physical picture

Start from the free gaseous ion, say Ni²⁺, isolated in a vacuum with nothing else around it. Its d electrons occupy five degenerate orbitals — $d_{xy}$ , $d_{xz}$ , $d_{yz}$ , $d_{x^2-y^2}$ , $d_{z^2}$ — all at exactly the same energy because the potential the electron feels (from the nucleus and the other core electrons) is spherically symmetric. Group theory guarantees this: in the full rotation group, all five d orbitals transform as a single irreducible representation, so nothing distinguishes one from another energetically. Now bring up six point negative charges (idealized "ligands") along the $\pm x$ , $\pm y$ , $\pm z$ axes, forming a perfect octahedron. The potential felt by a d electron is no longer spherical — it now has the (lower) symmetry of the octahedron, point group $O_h$ . Immediately, symmetry alone tells us the fivefold degeneracy cannot survive: in $O_h$ , the d orbitals split into two sets, a doubly-degenerate set (labelled $e_g$ ) and a triply-degenerate set ( $t_{2g}$ ). This is not a numerical coincidence; it is dictated by the reduction of the d-orbital representation under $O_h$ symmetry into its irreducible components. CFT's job is to explain why one set goes up and the other goes down, and by how much. The physical reason is simple electrostatic repulsion between the negatively-charged ligands and the negatively-charged electron cloud of each d orbital. The $d_{z^2}$ orbital has large lobes pointing directly along $+z$ and $-z$ , straight at two of the six ligands. The $d_{x^2-y^2}$ orbital has four lobes pointing directly along $\pm x$ and $\pm y$ , straight at the other four ligands. Both orbitals therefore have substantial electron density concentrated exactly where the point charges sit, and an electron in either orbital is strongly repelled — these two orbitals are raised in energy and grouped together as $e_g$ . By contrast, $d_{xy}$ , $d_{xz}$ , and $d_{yz}$ have lobes that point between the axes — between the ligands, not at them. An electron in one of these three orbitals experiences much less repulsion because the electron density is smaller in the directions where the point charges lie. These three orbitals are lowered relative to $e_g$ and are grouped as $t_{2g}$ . Crucially, CFT keeps all the electrostatic energy raised by the ligand field's average (spherical) component separate from the splitting itself: the average energy of all five d orbitals in the field is raised uniformly (because the ligands are, on average, negative charges near the metal, repelling all d electrons somewhat) — call this energy $E_{sph}$ — and then the non-spherical part of the field further separates $e_g$ from $t_{2g}$ around that shifted average, called the barycenter. The two effects are separable because the octahedral potential can be expanded in spherical harmonics; the $\ell=0$ term shifts everything uniformly, while the $\ell=4$ term (the lowest nonzero multipole compatible with octahedral symmetry) does the actual splitting.

Figure

Free ion vs. octahedral field: the five-fold degeneracy is lifted

For a tetrahedral arrangement of four ligands (point group $T_d$ ), the geometry is entirely different: no ligand sits on an axis through an orbital lobe, and the closest approach of ligand charges to orbital lobes is reversed relative to the octahedron. The result — derived carefully in Crystal Field Theory: Octahedral Splitting — is that the splitting pattern inverts: the triply-degenerate set ( $t_2$ , analogous in character to $t_{2g}$ ) goes up, and the doubly-degenerate set ( $e$ ) goes down, and the total splitting $\Delta_t$ is intrinsically smaller because there are only four charges instead of six and none of them point directly at an orbital lobe.

Figure

Ligand positions relative to d-orbital lobes: why e_g and t_2g split oppositely

The core ideas

The starting point of CFT is a Hamiltonian for a d electron in the field of $n$ point charges $-q$ (the ligands) located at positions $\mathbf{R}_i$ around the metal ion. The perturbation to the electron's potential energy is

V_{CF}(\mathbf{r}) = \sum_{i=1}^{n} \frac{(-e)(-q)}{4\pi\varepsilon_0 \,|\mathbf{r} - \mathbf{R}_i|}

where:

$\mathbf{r}$ is the position of the d electron relative to the metal nucleus,
$\mathbf{R}_i$ is the position of the $i$ -th ligand point charge,
$-e$ is the electron charge and $-q$ the (formal) ligand point charge,
$\varepsilon_0$ is the vacuum permittivity.

Reading it: this is nothing but Coulomb's law, summed over every ligand, evaluated at the position of a d electron rather than at a fixed point — it is the entire physical content of CFT. Every subsequent result (splitting pattern, $\Delta_o$ , $\Delta_t$ , the spectrochemical series taken as an empirical scale) comes from evaluating matrix elements of this one operator between d orbitals. Because $|\mathbf{r}-\mathbf{R}_i|^{-1}$ can be expanded in spherical harmonics (the standard multipole expansion), $V_{CF}$ decomposes as

V_{CF}(\mathbf{r}) = \sum_{k=0}^{\infty}\sum_{m=-k}^{k} A_{km}\, r^{k}\, Y_{km}(\theta,\phi)

where:

$A_{km}$ are coefficients fixed by the ligand charges and geometry,
$Y_{km}$ are spherical harmonics,
only terms with $k \le 4$ have nonzero matrix elements between d orbitals ( $\ell=2$ ), because the matrix element $\langle d|Y_{km}|d\rangle$ vanishes by the triangle rule unless $k \le 2\ell = 4$ .

Reading it: the $k=0$ term is the spherically-symmetric part of the field — it raises all five d orbitals equally and does no splitting. The $k=4$ term is the lowest-order term whose angular shape is compatible with cubic (octahedral or tetrahedral) symmetry, and it is entirely responsible for the $e_g$ / $t_{2g}$ (or $e$ / $t_2$ ) splitting. Odd- $k$ terms vanish identically for a centrosymmetric arrangement like an octahedron, which is also why octahedral complexes have (Laporte-forbidden) weak d–d absorption bands rather than fully allowed ones — a fact rationalized entirely by this same symmetry argument. The barycenter rule — that splitting conserves the average d-orbital energy — follows because the $k=4$ perturbation is traceless over the d-orbital manifold:

\sum_{i=1}^{5} \langle d_i | V_{CF}^{(k=4)} | d_i \rangle = 0

where the sum runs over the five d orbitals in any convenient basis. Reading it: whatever energy is gained by lowering $t_{2g}$ must be exactly compensated by raising $e_g}$ (weighted by degeneracy), because a traceless operator's eigenvalues must sum to zero. This immediately yields the famous $3:2$ weighting used throughout crystal field stabilization energy calculations: with three $t_{2g}$ orbitals lowered by $x$ each and two $e_g$ orbitals raised by $y$ each, tracelessness requires $3x = 2y$ .

Derivation AWhy five d orbitals split into exactly two groups under an octahedral field

Goal. Show, using only symmetry, that an octahedral arrangement of six identical point charges must split the d orbitals into a 3-fold and a 2-fold set — never five distinct levels, never a different partition — and confirm this is consistent with the electrostatic picture above.

Step 1 — identify the symmetry of the perturbation. Six point charges at $(\pm a,0,0)$ , $(0,\pm a,0)$ , $(0,0,\pm a)$ define an octahedral arrangement whose symmetry operations (rotations, reflections, improper rotations that leave the six points as a set invariant) form the point group $O_h$ . The electrostatic potential $V_{CF}(\mathbf r)$ built from these six charges must itself be invariant under every operation of $O_h$ — it cannot have lower symmetry than its source charges.

Step 2 — decompose the reducible representation of the five d orbitals under $O_h$ . The five d orbitals transform, under the parent rotation group, as the representation $D^{(2)}$ . Restricting this representation to the operations of $O_h$ and computing characters for each class (E, 8C₃, 6C₂, 6C₄, 3C₂′, i, 6S₄, 8S₆, 3σₕ, 6σd) gives a reducible representation with characters $\{5, -1, 1, -1, 1, 5, -1, -1, 1, 1\}$ (using the standard $O_h$ character table convention). Applying the reduction formula

n_j = \frac{1}{h}\sum_{R} \chi(R)\,\chi_j(R)

where $h=48$ is the order of $O_h$ and $\chi_j$ are the irreducible characters, gives exactly $n_{E_g}=1$ and $n_{T_{2g}}=1$ , and zero for every other irreducible representation. In words: the reducible $D^{(2)}$ representation of the five d orbitals decomposes exactly as $E_g \oplus T_{2g}$ under $O_h$ — nothing more, nothing less. This is a rigorous group-theoretic fact, independent of any detail of the electrostatics; it fixes the pattern (a 2-set and a 3-set, no other partition possible) before any energy has even been calculated.

Step 3 — assign which set is which using the electrostatic argument. Symmetry alone cannot say which set (the 2-fold or the 3-fold) lies higher; that requires evaluating the sign and magnitude of $\langle d_i|V_{CF}|d_i\rangle$ explicitly for a representative orbital of each set. Take $d_{z^2}$ (an $E_g$ member) versus $d_{xy}$ (a $T_{2g}$ member). The angular parts are

d_{z^2} \propto (3\cos^2\theta - 1), \qquad d_{xy}\propto \sin^2\theta \sin 2\phi

Evaluating the electron density $|d_{z^2}|^2$ at the six ligand directions $(\theta,\phi) = (0,-), (\pi,-), (\pi/2,0),(\pi/2,\pi/2),(\pi/2,\pi),(\pi/2,3\pi/2)$ shows that $|d_{z^2}|^2$ is large exactly at the poles ( $\theta=0,\pi$ , where two ligands sit) and appreciable in the equatorial plane too (where four more ligands sit), because $(3\cos^2\theta-1)^2$ is large at $\theta=0$ and moderate at $\theta=\pi/2$ . Doing the same for $d_{xy}$ shows its density vanishes exactly along all six ligand directions (all lie along the Cartesian axes, where $\sin 2\phi = 0$ or $\sin\theta=0$ ) and is maximal only along the diagonals, between ligands. Consequently the Coulomb repulsion energy is larger for $E_g$ orbitals than for $T_{2g}$ orbitals, i.e.

\langle d_{z^2}|V_{CF}|d_{z^2}\rangle > \langle d_{xy}|V_{CF}|d_{xy}\rangle

so $E_g$ lies above $T_{2g}$ — matching the familiar picture with $e_g$ higher, $t_{2g}$ lower.

Step 4 — sanity check via the trace condition. Since the $k=4$ perturbation is traceless over the five-orbital manifold (Step of the core-ideas section above), the two energies obtained in Step 3 must satisfy $3\,\varepsilon(t_{2g}) + 2\,\varepsilon(e_g) = 0$ when measured from the barycenter. Placing $t_{2g}$ at $-0.4\,\Delta_o$ and $e_g$ at $+0.6\,\Delta_o$ satisfies this exactly: $3(-0.4)+2(0.6) = -1.2+1.2=0$ . This reproduces the standard " $0.4\Delta_o$ / $0.6\Delta_o$ " splitting bookkeeping used everywhere in CFT stabilization-energy calculations, and confirms internal consistency between the group-theoretic pattern (Step 2), the electrostatic ordering (Step 3), and the barycenter constraint.

Go deeper · Derivation B · 2 minBarycenter conservation forces the 3:2 (or 2:3) energy-weighting rule

Derivation BBarycenter conservation forces the 3:2 (or 2:3) energy-weighting rule

Goal. Derive, from first principles of the point-charge Hamiltonian, why the two split levels must be weighted in exactly a 3-to-2 ratio (for octahedral/tetrahedral splitting into 3-fold and 2-fold sets) rather than any other ratio — a result used quantitatively for every crystal field stabilization energy (CFSE) calculation.

Step 1 — set up the trace of the perturbation over the d manifold. The total first-order energy shift of all five d orbitals, summed, is

\sum_{i=1}^5 \Delta E_i = \sum_{i=1}^5 \langle d_i | V_{CF} | d_i \rangle = \mathrm{Tr}(V_{CF})

evaluated in the d-orbital subspace. Because $V_{CF}$ restricted to $\ell=2$ states contains only the spherical ( $k=0$ ) part and the ( $k=4$ ) part (Step 2 in the core-ideas discussion — all other multipoles have zero matrix elements between $\ell=2$ functions), we can write $V_{CF} = V^{(0)} + V^{(4)}$ acting on the d subspace.

Step 2 — isolate the $k=4$ (splitting) contribution. The $k=0$ term is a constant times the identity operator on the d subspace (it cannot distinguish orbitals — that is what "spherically symmetric" means), so it contributes equally, and only rigidly, to every d orbital; it does not split anything, it only shifts the whole d block up by some amount $E_{sph}$ . The splitting comes entirely from $V^{(4)}$ , and it is a standard result of the theory of spherical tensor operators that a rank-4 tensor operator acting within a single $\ell = 2$ shell is traceless:

\mathrm{Tr}\big(V^{(4)}\big) = \sum_{i=1}^{5}\langle d_i|V^{(4)}|d_i\rangle = 0

This is because the trace of any operator is basis-independent, and $V^{(4)}$ has no component along the identity (rank-0) operator by construction — a rank-4 spherical tensor is orthogonal to the rank-0 (scalar) operator under the trace inner product.

Step 3 — apply this to the $O_h$ result of Step 2 above. We already established that the five d orbitals split as $E_g$ (2-fold, energy $\varepsilon_e$ relative to barycenter) and $T_{2g}$ (3-fold, energy $\varepsilon_t$ relative to barycenter). The trace condition becomes

2\,\varepsilon_e + 3\,\varepsilon_t = 0

Step 4 — solve for the ratio and fix the scale using $\Delta_o$ . Define $\Delta_o \equiv \varepsilon_e - \varepsilon_t$ (the total splitting). Combining with $2\varepsilon_e = -3\varepsilon_t$ :

\varepsilon_e = \frac{3}{5}\Delta_o, \qquad \varepsilon_t = -\frac{2}{5}\Delta_o

Check: $\varepsilon_e - \varepsilon_t = \tfrac{3}{5}\Delta_o + \tfrac{2}{5}\Delta_o = \Delta_o$ ✓, and $2\varepsilon_e+3\varepsilon_t = \tfrac{6}{5}\Delta_o - \tfrac{6}{5}\Delta_o = 0$ ✓. These are exactly the familiar $+0.6\Delta_o$ ( $e_g$ ) and $-0.4\Delta_o$ ( $t_{2g}$ ) values. The same algebra, applied to a tetrahedral field where the pattern inverts (2-fold set lower, 3-fold set higher, called $e$ and $t_2$ ), gives $\varepsilon_{t_2} = +\tfrac{3}{5}\Delta_t$ and $\varepsilon_e = -\tfrac{2}{5}\Delta_t$ — the mirror image, because the same trace-zero constraint applies regardless of which set the electrostatics happens to push up.

Worked examples

Example 1 (foundational): confirming the barycenter with an explicit electron count. Consider a high-spin d⁵ ion (e.g. Mn²⁺ or Fe³⁺) in an octahedral field, with all five electrons unpaired (one in each of the five d orbitals: three in $t_{2g}$ , two in $e_g$ ). Compute the total electronic energy relative to the (hypothetical) spherical-field average, and confirm it is exactly zero — i.e., a half-filled shell feels no net crystal field stabilization. Each $t_{2g}$ electron contributes $-0.4\Delta_o$ and each $e_g}$ electron contributes $+0.6\Delta_o$ . With 3 electrons in $t_{2g}$ and 2 in $e_g$ :

E_{CF} = 3(-0.4\Delta_o) + 2(+0.6\Delta_o) = -1.2\Delta_o + 1.2\Delta_o = 0

As expected from the trace/barycenter rule: a symmetrically (half-)filled configuration always has zero net crystal field stabilization energy, because it samples $t_{2g}$ and $e_g}$ in exactly the ratio (3:2) that the trace condition requires to cancel. This is precisely why high-spin d⁵ complexes (like $[\mathrm{Mn(H_2O)_6}]^{2+}$ ) are notoriously pale in colour and weakly affected by ligand field strength in terms of stabilization energy — though they still show (very weak, spin-forbidden) d–d transitions.

Example 2 (applied): asymmetric CFSE for a non-half-filled, non-symmetric configuration. Take a low-spin d⁶ octahedral complex such as $[\mathrm{Co(NH_3)_6}]^{3+}$ , with electron configuration $t_{2g}^6 e_g^0$ (all six electrons paired into the lower set — a genuinely asymmetric filling, useful precisely because it is not the trivial symmetric case). Using the values from the derivation above:

E_{CF} = 6(-0.4\Delta_o) + 0(+0.6\Delta_o) = -2.4\,\Delta_o

Compare this with the same d⁶ ion in a high-spin configuration $t_{2g}^4 e_g^2$ (as would occur with a weak-field ligand):

E_{CF} = 4(-0.4\Delta_o) + 2(+0.6\Delta_o) = -1.6\Delta_o + 1.2\Delta_o = -0.4\,\Delta_o

The low-spin configuration is stabilized by a full $2.0\,\Delta_o$ more than the high-spin one (ignoring, for the moment, the pairing energy cost, which is treated quantitatively once $\Delta_o$ is compared against the mean pairing energy $P$ — the subject taken up in Crystal Field Theory: Octahedral Splitting). This asymmetric comparison — $-2.4\Delta_o$ vs. $-0.4\Delta_o$ , not some contrived equal split — is exactly the kind of quantitative leverage CFT provides even at this "basic model" stage: a purely electrostatic bookkeeping scheme correctly predicts that filling the lower set preferentially is always more stabilizing, in proportion to $\Delta_o$ .

Example 3 (exam / competition level): reconciling geometry-dependent splitting patterns using only the trace rule. A d⁷ ion (e.g. Co²⁺) is placed, in two separate experiments, into (a) an octahedral field with $\Delta_o = 9,300\ \mathrm{cm^{-1}}$ (weak field, high spin: $t_{2g}^5 e_g^2$ ) and (b) a tetrahedral field with $\Delta_t = 4,100\ \mathrm{cm^{-1}}$ (necessarily high spin, filling $e^4 t_2^3$ since tetrahedral fields are essentially never strong enough to force pairing). Compute the CFSE in each geometry, in $\mathrm{cm^{-1}}$ , and compare. For the octahedral case, using $\varepsilon_{t_{2g}} = -0.4\Delta_o$ , $\varepsilon_{e_g}=+0.6\Delta_o$ and configuration $t_{2g}^5 e_g^2$ :

E_{CF}(\mathrm{oct}) = \big[5(-0.4) + 2(0.6)\big]\Delta_o = (-2.0+1.2)\Delta_o = -0.8\,\Delta_o

E_{CF}(\mathrm{oct}) = -0.8 \times 9{,}300\ \mathrm{cm^{-1}} = -7{,}440\ \mathrm{cm^{-1}}

For the tetrahedral case the pattern is inverted (trace rule applied to $T_d$ : the 2-fold $e$ set sits at $-\tfrac{3}{5}\Delta_t$ and the 3-fold $t_2$ set sits at $+\tfrac{2}{5}\Delta_t$ — check the trace: $2(-\tfrac35)+3(\tfrac25) = -\tfrac65+\tfrac65=0$ ✓). With configuration $e^4 t_2^3$ :

E_{CF}(\mathrm{tet}) = \Big[4\Big(-\tfrac{3}{5}\Big) + 3\Big(\tfrac{2}{5}\Big)\Big]\Delta_t = \Big(-\tfrac{12}{5}+\tfrac{6}{5}\Big)\Delta_t = -\tfrac{6}{5}\Delta_t = -1.2\,\Delta_t

E_{CF}(\mathrm{tet}) = -1.2 \times 4{,}100\ \mathrm{cm^{-1}} = -4{,}920\ \mathrm{cm^{-1}}

So even though the fractional stabilization coefficient is larger in magnitude for the tetrahedral case ( $1.2$ vs. $0.8$ , in units of the respective splitting parameter), the absolute stabilization is smaller in tetrahedral geometry ( $-4{,}920\ \mathrm{cm^{-1}}$ vs. $-7{,}440\ \mathrm{cm^{-1}}$ ) because $\Delta_t \ll \Delta_o$ for chemically comparable ligands (a robust empirical/electrostatic result, since $\Delta_t \approx \tfrac{4}{9}\Delta_o$ for the same metal, ligand, and metal–ligand distance, reflecting both the smaller number of ligands and their non-axial approach). This is exactly why octahedral coordination, all else equal, is almost always more crystal-field-stabilized than tetrahedral coordination for a given metal/ligand pair — one of the qualitative structural preferences CFT gets right from pure electrostatics, before any covalency is even invoked.

Limiting cases, scope & common traps

Situation	What CFT predicts	What actually limits/extends it
Spherical field (isolated ion, or a hypothetical "12-coordinate" perfectly spherical shell)	No splitting; five d orbitals stay degenerate	Correctly recovered: $k=4$ term vanishes for a truly spherical charge shell
Free ion, gas phase, no ligands	d orbitals degenerate, $\Delta=0$	Matches atomic spectroscopy: term energies from Racah parameters alone
$d^0$ or $d^{10}$ configurations	Zero net CFSE regardless of geometry (empty or completely full set)	Correct — no splitting-dependent color, consistent with $\mathrm{Zn^{2+}}$ , $\mathrm{Sc^{3+}}$ being colourless
$d^5$ high spin	Zero net CFSE (barycenter argument, Example 1)	Matches the pale colour and unusually weak field-sensitivity of high-spin Mn(II)/Fe(III) complexes
Highly covalent ligands (CO, CN⁻, alkenes)	CFT (point charge) underestimates the splitting badly	Requires ligand field theory / MO treatment with π back-bonding — CFT's ionic assumption fails outright here
Very ionic, hard ligands (F⁻, H₂O, O²⁻)	CFT works reasonably well quantitatively	Electrostatics genuinely dominates; closest to CFT's assumed regime
Jahn–Teller-active configurations (e.g. high-spin d⁴, d⁹)	Simple CFT (undistorted geometry) predicts degenerate ground states	Real molecules distort to remove the degeneracy — requires an extension beyond the rigid point-charge/fixed-geometry model
Tetrahedral vs. octahedral, same ligand/metal	$\Delta_t \approx \frac{4}{9}\Delta_o$	Well-verified trend; absolute CFSE always smaller for tetrahedral for a given ligand

Common trapMisconception: "CFT says the metal–ligand bond is ionic, so it must be a bad approximation for everything."

CFT's ionic, no-covalency assumption is a genuine idealization — real metal–ligand bonds virtually always have some covalent character, and for π-acceptor ligands (CO, CN⁻, phosphines) that covalent character dominates the bonding energetics entirely. But the symmetry results of CFT (which orbitals group together, how many levels, the barycenter/trace rule, the qualitative ordering $e_g > t_{2g}$ for octahedral fields) do not actually depend on the bonding being ionic — they depend only on the point-group symmetry of the ligand arrangement. This is why the (more sophisticated, covalency-including) ligand field theory framework reproduces the same pattern of split levels as CFT, just with a reinterpreted and typically larger splitting energy. CFT gets the qualitative geometry right for reasons that have nothing to do with whether the ionic picture is quantitatively true.

The chemist behind it

Crystal field theory's electrostatic model was developed not by a chemist studying coordination compounds directly, but by physicists working on the magnetic and optical properties of ions embedded in ionic crystals — a problem of practically identical mathematical structure to a metal ion surrounded by ligands.

Origins · 1 minHans Bethe and how this was discovered

Hans Bethe

1906–2005 · Crystal-field theory in solid-state and nuclear physics

Hans Bethe formulated the original point-charge theory of splitting in crystalline electric fields in 1929, analyzing how the energy levels of a paramagnetic ion embedded in an ionic crystal lattice are split by the electrostatic field of surrounding lattice ions. His treatment — using exactly the group-theoretic reduction of the d (or f) orbital representation under the local site symmetry, combined with the electrostatic point-charge potential — is the direct mathematical ancestor of what chemists now call crystal field theory. Bethe went on to a career centered on nuclear and particle physics (he won the 1967 Nobel Prize in Physics for his theory of stellar nucleosynthesis), but his 1929 crystal-field paper, written while working in Sommerfeld's group in Munich, was rediscovered and adapted by chemists in the 1950s as inorganic chemists sought a quantitative explanation for transition-metal spectra and magnetism.

History

- 1929 — Hans Bethe publishes the point-charge theory of energy-level splitting for ions in crystalline electric fields.

- 1932 — John H. Van Vleck extends the theory and clarifies its application to magnetic susceptibilities of transition-metal ions.

- 1951–1953 — Leslie Orgel, and independently others, popularize crystal field theory among chemists to explain transition-metal complex colors and thermodynamic stabilities (introducing what became known as Orgel diagrams).

- 1950s — the recognized inadequacy of the pure ionic/point-charge picture (it cannot explain nephelauxetic effects or the strength of π-acceptor ligands) drives development of the more general ligand field theory, incorporating covalency via molecular orbital theory.

- 1960s–present — crystal field theory remains the standard first-pass teaching model for d–d spectra, magnetism, and structural preferences (Jahn–Teller distortions, site preferences in spinels), even as quantitative work uses full ligand field / DFT treatments.

Does crystal field theory require the ligands to actually be point charges, or can it handle neutral ligands like water or ammonia?

CFT can be extended to neutral ligands by treating them as point dipoles rather than point charges — a water molecule's oxygen lone pair, for instance, is modeled as a negative end of a dipole pointing at the metal. The angular dependence of the splitting (which orbitals go up, which go down, and in what pattern) comes out identically whether you use point charges or point dipoles, because both produce an electrostatic potential with the same point-group symmetry around the metal. Only the absolute magnitude of the splitting differs between the two models. This is one reason CFT, despite its crude "point charge" language, still works reasonably well even for complexes with electrically neutral ligands.

If crystal field theory ignores covalency, why does it correctly predict the ordering of ligand field strengths (the spectrochemical series)?

It does not predict the spectrochemical series from first principles — the ordering (I⁻ < Br⁻ < Cl⁻ < F⁻ < H₂O < NH₃ < en < CN⁻ ≈ CO, roughly) is an empirical scale, not something CFT derives from pure point-charge electrostatics. In fact, a naive electrostatic model predicts the wrong order in places (it would rank F⁻, a hard, small, highly charged anion, as an extremely strong-field ligand, and cannot explain why CN⁻ and CO, weakly charged or neutral, produce the largest splittings of all). The correct explanation for the strong-field end of the series requires π-acceptor back-bonding, a genuinely covalent, molecular-orbital effect that lies entirely outside the electrostatic point-charge model. CFT correctly predicts the pattern of splitting for any given ligand arrangement, but the ranking of ligand strengths needs the fuller ligand field / MO treatment.

Why does the crystal field splitting conserve the average (barycenter) energy of the d orbitals?

Mathematically, this follows because the part of the electrostatic potential responsible for splitting (the lowest-order term compatible with the ligand geometry's symmetry, formally a rank-4 spherical tensor for octahedral or tetrahedral fields) has zero trace when evaluated over the five d orbitals — its eigenvalues, weighted by degeneracy, must sum to zero. Physically, this says that redistributing electron density among orbitals that already integrate to give a spherically-averaged shell cannot, by itself, change the average energy of that shell; only the genuinely spherical part of the ligand field (which raises all orbitals equally) shifts the overall average, and that part is deliberately excluded when people quote "the splitting."

Is crystal field theory the same thing as ligand field theory?

No, though the terms are sometimes used loosely and interchangeably in introductory contexts. Crystal field theory is specifically the purely electrostatic, point-charge (or point-dipole) model with no covalent metal–ligand bonding built in. Ligand field theory is the more general and more accurate successor that incorporates covalency using molecular orbital theory, while retaining CFT's symmetry-derived splitting pattern (the same $e_g$ / $t_{2g}$ labels, for instance) as a starting classification. In practice, chemists often use "crystal field splitting parameter" ( $\Delta_o$ ) as shorthand even when the underlying physical picture being invoked is really the ligand field (covalent) one — a terminological holdover from CFT's historical priority.

Why doesn't crystal field theory split the d orbitals into five completely separate energy levels?

Because the splitting is dictated by the symmetry of the ligand arrangement, and for a highly symmetric arrangement like a perfect octahedron or tetrahedron, several d orbitals remain related to each other by the symmetry operations of that point group (rotations that map the ligand positions onto themselves also map certain d orbitals onto linear combinations of each other). Orbitals related by a symmetry operation of the point group must remain exactly degenerate — this is a rigorous consequence of group representation theory, not an approximation. Only in point groups with very low symmetry (no rotation axes higher than 2-fold, for instance) can all five d orbitals become genuinely nondegenerate.

Does crystal field theory apply only to first-row (3d) transition metals, or also to 4d, 5d metals and lanthanides/actinides?

The electrostatic symmetry argument itself is completely general and applies to any set of d orbitals (or, with minor modification, f orbitals for lanthanides/actinides) surrounded by ligands in any geometry. In practice, CFT is least accurate for 4d and 5d metals, where relativistic effects and much stronger covalent/π-backbonding interactions make the pure ionic picture a poorer starting approximation, and for lanthanides, where the f orbitals are so contracted and shielded by outer s and p electrons that the crystal field splitting is very small compared to spin-orbit coupling, and the whole hierarchy of interactions is reordered relative to the d-block case.

What experimental evidence first convinced chemists that d orbitals actually split as CFT predicts?

The single most direct evidence is the visible-light absorption spectra of transition-metal complexes: octahedral Ti³⁺ complexes (a simple d¹ ion, with only one electron and therefore only one possible transition) show a single broad absorption band whose energy, converted from wavelength, matches $\Delta_o$ directly and whose band sometimes shows a shoulder attributable to Jahn–Teller distortion of the excited state — exactly the single $t_{2g}\to e_g$ transition CFT predicts for a one-electron case. More broadly, the systematic trends in color, magnetic moment (high-spin vs. low-spin electron counting), and thermodynamic stability (lattice energies, hydration enthalpies showing the "double-humped" curve across the first transition series) all matched the CFSE predictions quantitatively enough, using a single fitted $\Delta_o$ per complex, to establish the model's validity well before more sophisticated MO calculations became feasible.