1The splitting of the d orbitals
In a free transition-metal ion the five d orbitals are degenerate. When six ligands are placed on the ±x, ±y and ±z axes, this degeneracy is lifted, because the orbitals are oriented differently with respect to the ligands. Two of them, dz² and dx²−y², have lobes directed along the axes, towards the ligands. The other three, dxy, dxz and dyz, have lobes directed between the axes.
The orbitals directed towards the ligands are raised in energy relative to the other three. In an octahedral field the d orbitals therefore separate into two sets, labelled according to their symmetry: the doubly degenerate eg set at higher energy and the triply degenerate t2g set at lower energy. The energy separation between the two sets is the ligand-field splitting parameter, Δₒ. Because the mean energy of the five orbitals is unchanged by the splitting (the barycentre rule), the two eg orbitals lie 0.6 Δₒ above the barycentre and the three t2g orbitals 0.4 Δₒ below it.
Crystal field, ligand field and angular overlap models. Crystal field theory treats the ligands as point charges that repel the d electrons. It reproduces the splitting patterns correctly, because these are determined by symmetry. Ligand field theory takes covalency into account: the “d orbitals” of a splitting diagram are in fact the metal-based antibonding orbitals of the complex, and the energy of each depends on the extent of its overlap with the ligand orbitals. The angular overlap model, which is used for all the calculations on this page, expresses this overlap as a sum of contributions from the individual metal–ligand bonds. Each ligand raises the energy of a given d orbital by eσFσ² + eπFπ², where Fσ and Fπ are angular overlap factors. For an octahedron this gives E(eg) = 3eσ and E(t2g) = 4eπ, so that Δₒ = 3eσ − 4eπ.
Factors that determine the magnitude of Δ
- The ligand. Ligands can be arranged in a spectrochemical series of increasing Δ, from the halides (weak field) through water and ammonia to cyanide and CO (strong field). The origin of this order lies in the σ- and π-bonding properties of the ligands, which are discussed in Section 2.
- The oxidation state of the metal. A higher oxidation state gives shorter, more covalent metal–ligand bonds and a larger Δ. For example, Δₒ ≈ 10 400 cm⁻¹ for [Fe(H₂O)₆]²⁺ but ≈ 14 000 cm⁻¹ for [Fe(H₂O)₆]³⁺.
- The period. The 4d and 5d orbitals are more extended than the 3d orbitals and overlap more strongly with the ligand orbitals, so Δ increases by roughly 30–50 % on descending a group.
- The metal–ligand distance. Δ decreases steeply as the bond length increases, approximately as r⁻⁵. This dependence can be examined with the bond-length slider in the workbench. It also explains why applied pressure can change the spin state of a complex.
- The coordination geometry. Four ligands in a tetrahedral arrangement give Δt ≈ 4/9 Δₒ, and the order of the two sets is inverted. The ratio can be checked by selecting Tetrahedron in the workbench.
Ligand-field splittings are usually quoted in cm⁻¹. For reference, 1 eV ≈ 8 066 cm⁻¹ ≈ 96.5 kJ mol⁻¹, and a splitting of 20 000 cm⁻¹ corresponds to light of wavelength 500 nm. The units menu at the top of the page converts all energies on the page to the selected unit.
2Orbital overlap and metal–ligand bonding
The interaction between a metal orbital and a ligand orbital depends on their overlap, that is, on the extent to which their lobes coincide in space with the same or opposite phase (sign). Three cases can be distinguished.
- σ overlap occurs end-on, along the metal–ligand axis: the donor orbital of the ligand is directed along the bond towards a lobe of the metal orbital. In an octahedron, dz² and dx²−y² overlap with the ligands in this way.
- π overlap occurs side-on: a ligand p orbital perpendicular to the bond lies parallel to a lobe of the metal orbital. Each t2g orbital overlaps in this way with four of the six ligands.
- Zero net overlap results when the ligand lies on a nodal plane of the metal orbital, so that the positive and negative contributions cancel exactly. An example is dxy with ligand σ orbitals located on the x and y axes.
Where the overlap is non-zero, the two orbitals combine to form two molecular orbitals. The in-phase combination is bonding: it lies below both parent orbitals, is predominantly ligand in character, and accommodates the ligand lone pairs. The out-of-phase combination is antibonding: it lies above both parent orbitals and is predominantly metal in character. The eg set of a crystal-field diagram corresponds to this σ* combination. In the absence of π bonding, its energy above the t2g set is a measure of the strength of the metal–ligand σ interaction.
The ligand orbitals are drawn with a size proportional to their overlap with the selected d orbital; coral and teal denote the two phases. In the bonding combination each ligand lobe has the same phase as the metal lobe it faces, and in the antibonding combination the ligand phases are reversed. The d electrons occupy the antibonding combination.
π bonding and the spectrochemical series
σ donation alone raises the energy of the eg set and leaves the t2g set non-bonding. π interactions then modify the energy of the t2g set. A π-donor ligand (F⁻, Cl⁻, O²⁻) has filled p orbitals lying below the metal d orbitals; their interaction raises the t2g set and reduces Δₒ. A π-acceptor ligand (CO, CN⁻) has empty π* orbitals lying above the metal d orbitals; their interaction lowers the t2g set and increases Δₒ. In the angular overlap model the distinction appears as the sign of eπ, which is positive for π donors and negative for π acceptors.
3High-spin and low-spin configurations
Electrons occupy the lowest available orbitals first, but two energy terms compete. Placing a second electron in an orbital that is already singly occupied costs the pairing energy P, which arises from the additional Coulomb repulsion and the loss of exchange stabilisation; the latter is the origin of Hund’s rule. Placing the electron in an eg orbital instead costs Δₒ.
For octahedral d¹–d³ and d⁸–d¹⁰ ions there is only one possible ground-state configuration. For d⁴–d⁷ there are two. The high-spin configuration, with the maximum number of unpaired electrons, is favoured when Δₒ < P; the low-spin configuration, in which the electrons pair in the t2g set, is favoured when Δₒ > P. This can be seen in the workbench by selecting d⁶ and the Auto spin setting and then varying Δ through the value of P: the configuration changes abruptly when the two are equal.
- 3d ions lie close to the crossover. Weak-field ligands give high-spin complexes, whereas CN⁻, CO and bipyridine give low-spin complexes.
- 4d and 5d ions are almost always low spin, because Δ is larger and the more diffuse orbitals reduce the pairing energy.
- Spin crossover occurs when Δ ≈ P. Because the eg electrons are antibonding, the high-spin form has longer metal–ligand bonds (by about 0.2 Å for Fe²⁺–N). Pressure therefore favours the low-spin state, whereas an increase in temperature favours the high-spin state, which has the greater vibrational entropy. The LaCoO₃ example lies close to this crossover.
The spin state can be determined from magnetic measurements. When the orbital angular momentum is quenched, the magnetic moment is close to the spin-only value, √(n(n+2)) μB, which is given in the workbench readout. Ions with orbitally degenerate t2g ground states show additional orbital contributions to the moment; the reason is discussed in Section 7.
The workbench uses a one-electron description with a single pairing energy. The energies of real many-electron states depend also on the interelectronic repulsion, which is described by the Racah parameters B and C. A complete treatment requires Tanabe–Sugano diagrams, which are needed for the assignment of d–d spectra.
4The Jahn–Teller effect
The Jahn–Teller theorem states that a non-linear molecule with an orbitally degenerate ground state will undergo a distortion that removes the degeneracy. In ligand-field terms, if a degenerate set of orbitals is unevenly occupied, a distortion that splits the set allows more electrons to occupy the orbital that is lowered in energy than the orbital that is raised, so that the total energy decreases.
Splitting under a tetragonal distortion
A tetragonal elongation along z, which gives two long axial bonds and four slightly shorter equatorial bonds, lowers the energy of every orbital with a z component, because the axial ligands with which these orbitals overlap have moved further away. The orbitals lying in the xy plane are raised in energy. A compression has the opposite effect. Each set splits about its own barycentre:
- eg splits into dz² (a1g) and dx²−y² (b1g), separated by δ₁; each orbital moves by δ₁/2.
- t2g splits into the pair dxz, dyz (eg) and dxy (b2g), separated by δ₂; the pair moves by δ₂/3 and dxy by 2δ₂/3.
- In general δ₁ > δ₂, because the eg orbitals are σ-bonding and therefore more sensitive to changes in bond length than the t2g orbitals, which take part only in π bonding.
The effect on the total energy depends on how the electrons are distributed. A set with the same number of electrons in each of its orbitals gains no energy from the splitting, because the stabilisation of the lowered orbitals is exactly cancelled by the destabilisation of the raised ones. The orbitals of such a set split as a consequence of the distortion but make no net contribution to driving it. Only an unevenly occupied set provides a driving force. The workbench readout lists the energy contribution of each set separately, so that this can be verified for any dⁿ configuration.
Two examples. In high-spin Mn³⁺ (d⁴, t2g³eg¹) the single eg electron drives a strong elongation and occupies the lowered dz² orbital; the three t2g electrons, one in each orbital, make no net contribution. In high-spin Fe²⁺ (d⁶, t2g⁴eg²) the sixth electron, which is paired in the t2g set, drives a weak distortion. In the simplest model a compression is preferred, because the electron pair can then occupy the single orbital, dxy, which is lowered by 2δ₂/3. An elongation would place the pair in the dxz/dyz set, which is lowered by only δ₂/3 and remains degenerate. The two eg electrons make no net contribution.
Strong and weak Jahn–Teller effects
Uneven occupation of the eg set (high-spin d⁴, low-spin d⁷, d⁹) produces large, static distortions that are readily observed crystallographically; the long and short bonds typically differ by 0.2–0.4 Å. Uneven occupation of the t2g set (d¹, d², low-spin d⁴ and d⁵, high-spin d⁶ and d⁷) produces small distortions, which are often dynamic or are masked by spin–orbit coupling, crystal packing or covalency. The configurations d³, high-spin d⁵, low-spin d⁶, d⁸ and d¹⁰ are Jahn–Teller inactive.
The Jahn–Teller energy surface
For an ion with an unevenly occupied eg set, the relevant distortions are the two components, Q₃ and Q₂, of the eg vibrational mode of the octahedron. To first order the energy depends only on the amplitude of the distortion, ρ = √(Q₂² + Q₃²), and not on its direction, φ. The minimum energy therefore lies along a circular trough surrounding a central maximum, a surface often described as a “Mexican hat”. Higher-order terms warp the trough. Both bond anharmonicity (a bond is more easily stretched than compressed) and 4s–3dz² mixing favour the three elongated structures at φ = 0°, 120° and 240°, which is why Cu²⁺ and Mn³⁺ complexes are almost always elongated rather than compressed. The warping can be examined in the Jahn–Teller surface tab by selecting d⁹ and switching the two options under Model settings on and off.
If the barriers between the three minima are small compared with the thermal energy, the complex interconverts rapidly between them and appears undistorted on average; this is the dynamic Jahn–Teller effect. In a crystal, by contrast, the strain fields of neighbouring sites can fix each site in a particular minimum and so produce an ordered pattern of distortions. This cooperative Jahn–Teller effect is discussed in Section 5.
A related phenomenon, the second-order (pseudo-) Jahn–Teller effect, can distort molecules with non-degenerate ground states when an excited state lies at low energy. It is responsible for the off-centre displacement of d⁰ Ti⁴⁺ in ferroelectric BaTiO₃ and for the stereochemically active lone pairs of Pb²⁺ and Bi³⁺.
5Molecular complexes and extended solids
In a molecular complex the coordination polyhedron is an isolated unit. In an extended solid, by contrast, the polyhedra share corners, edges or faces with their neighbours. This connectivity has several important consequences, which are summarised in the table below.
| Discrete complex | Extended solid | |
|---|---|---|
| Energy levels | Discrete orbitals; Δ obtained directly from d–d spectra. | Levels broaden into t2g and eg bands of width W through metal–ligand–metal electron transfer (hopping). The eg bands (σ) are usually wider than the t2g bands (π). |
| Jahn–Teller distortions | Each molecule distorts independently, and the distortion may be dynamic. | Shared ligands couple the distortions of neighbouring sites, giving cooperative Jahn–Teller order and orbital ordering, often with a structural phase transition. |
| Ligand environment | Determined by the choice of ligands. | Determined by the crystal structure: connectivity, octahedral tilts, A-site cations and the Madelung potential. The site symmetry is often lower than Oh (for example D3d in corundum and in layered oxides). |
| Conductivity | Molecular solids are generally insulators. | Band theory predicts a metal for a partly filled band, but electron correlation (U) can instead give a Mott–Hubbard or charge-transfer insulator. |
| Oxidation states | Usually well defined. | May be ambiguous because of covalency. When ΔCT is small or negative, holes reside partly on oxygen (ligand holes), as in the cuprates and the RNiO₃ nickelates. |
| Spin state | Spin crossover can be controlled through the choice of ligands. | Spin-state transitions are coupled to the lattice (LaCoO₃); pressure-induced transitions occur in the Earth’s lower mantle (Fe²⁺ in ferropericlase). |
| Site preference | Not applicable. | Differences in LFSE help to determine whether cations occupy octahedral or tetrahedral sites, as in normal and inverse spinels (Fe₃O₄ is inverse, Mn₃O₄ normal) and in the strong octahedral site preference of Cr³⁺. |
Cooperative Jahn–Teller distortions
Consider a Mn³⁺ ion in a perovskite. If its octahedron is elongated along x, the oxygen atom that it shares with its neighbour along x moves away from it and therefore towards that neighbour. The neighbouring ion now has a short bond along x, and the distortion most compatible with this is an elongation along y. The long axes therefore alternate from site to site in an antiferrodistortive pattern, and the occupied eg orbital alternates between d3x²−r²-like and d3y²−r²-like forms. This arrangement is an example of orbital ordering. If instead all the long axes are parallel (a ferrodistortive arrangement, as in Mn₃O₄), every octahedron is elongated in the same direction, and the crystal as a whole becomes tetragonal rather than cubic.
Bond colours: amber, long; blue, short. In the random arrangement, red marks shared oxygen atoms at which the requirements of the two neighbouring sites conflict.
From levels to bands: why NiO is an insulator
In a solid, electrons can transfer between neighbouring metal sites through the shared ligands, and each crystal-field level broadens into a band of width W. Simple band theory predicts that any material with a partly filled band is a metal. This prediction is correct for SrVO₃ and ReO₃ but fails for NiO, CoO and La₂CuO₄, which have partly filled d bands yet are insulators with band gaps of several electronvolts.
The explanation lies in the on-site Coulomb repulsion U, the energy required to place an electron on a site that is already occupied. When U > W, the band splits into a lower and an upper Hubbard band separated by a gap, and the material is a Mott–Hubbard insulator. In the late transition-metal oxides a second energy must also be considered: the charge-transfer energy ΔCT required to transfer an electron from O 2p to the metal. If ΔCT < U, the lowest-energy excitation is from O 2p to metal d, and the material is a charge-transfer insulator. This is the basis of the Zaanen–Sawatzky–Allen classification. It can be explored in the Bands tab of the workbench: for a description resembling NiO, select the NiO example and set W ≈ 1.5 eV, U ≈ 8 eV and ΔCT ≈ 5.5 eV.
Crystal-field concepts remain useful in solids, although Δ then refers to the separation between the t2g- and eg-derived bands in the density of states rather than to the gap between two discrete levels. The labels t2g and eg remain meaningful even when the d bands are several electronvolts wide.
6Other coordination geometries
Each coordination geometry gives a characteristic splitting pattern. The workbench calculates all of them with the same angular overlap parameters, so that the different geometries can be compared directly.
- Tetrahedral (Td)
- The four ligands occupy alternate corners of a cube. No d orbital points directly at a ligand, so the splitting is smaller (Δt ≈ 4/9 Δₒ) and the order is inverted, with the e set (dz², dx²−y²) below the t₂ set (dxy, dxz, dyz). The tetrahedron has no centre of inversion, so the labels carry no g or u subscript and the d–d transitions are more intense than in octahedral complexes. Tetrahedral complexes are almost always high spin. Tetrahedral d⁹ ions such as CuCl₄²⁻ are Jahn–Teller active and are flattened towards a square-planar geometry; this can be modelled with the flattening slider.
- Square planar (D4h)
- The square plane is the limiting geometry obtained when both axial ligands of an octahedron are removed. The dx²−y² orbital lies far above the other four. Eight electrons exactly fill the four lower orbitals, which is why strong-field d⁸ ions (Ni²⁺ with CN⁻, and nearly all complexes of Pd²⁺, Pt²⁺ and Au³⁺) are square planar and diamagnetic. The dz² orbital lies unusually low because of 4s–3dz² mixing; when this mixing is switched off under Model settings, dz² rises.
- Square pyramidal (C4v)
- Removal of one axial ligand gives a geometry intermediate between the octahedron and the square plane. It is common for Cu²⁺ and for V⁴⁺ in vanadyl complexes.
- Trigonal distortion (D3d)
- A trigonal distortion is an elongation or compression of the octahedron along one of its threefold axes (a body diagonal of the cube). Such distortions occur in the corundum structure (Al₂O₃, Cr₂O₃), in layered oxides with edge-sharing octahedra (LiCoO₂, NaₓCoO₂), and in honeycomb compounds such as the iridates and RuCl₃. The eg set remains degenerate, but t2g splits into a1g (a dz²-like orbital directed along the threefold axis) and egπ. The effect can be seen with the trigonal slider and the [111] view.
The symmetry labels depend on the point group. In D4h, dz² becomes a1g, dx²−y² b1g, dxy b2g, and the dxz/dyz pair eg. Note that this eg set derives from the octahedral t2g set, not from the octahedral eg set. The workbench labels each level according to the current point group, and correlation lines connect each level to the level of the regular geometry from which it derives.
7Spin–orbit coupling and heavy elements
An electron moving in the electric field of the nucleus experiences, in its own frame of reference, a magnetic field. This field interacts with the magnetic moment associated with the electron spin, so that the spin is coupled to the orbital motion. The interaction is known as spin–orbit coupling and is described by the Hamiltonian HSO = ζ l·s.
Spin–orbit coupling in the heavier elements
Spin–orbit coupling increases rapidly with atomic number, but whether it is the dominant interaction depends on the energy scale with which it is compared. For a hydrogen-like ion ζ is proportional to Z⁴. For valence electrons, screening weakens this dependence, but ζ still increases steeply down a group, from roughly 0.01–0.1 eV for the 3d metals to 0.1–0.2 eV for 4d and 0.3–0.6 eV for 5d. However, Δₒ also increases, to 3–5 eV in 5d oxides, so that even for iridium the crystal-field splitting remains the largest term. The important change is that ζ becomes comparable to the smaller energy scales: the splitting of t2g by low-symmetry distortions, the t2g bandwidth and the Hund’s coupling. Relative to these, spin–orbit coupling is decisive. Only for the 4f elements does spin–orbit coupling exceed the crystal-field splitting, because the 4f shell is shielded by the filled 5s and 5p shells and experiences ligand-field splittings of at most a few tens of meV.
Comparison with crystal-field splitting
- The coordinates on which it acts. The crystal field acts only on the spatial part of the wavefunction. The spin is unaffected, and each level can hold one spin-up and one spin-down electron. Spin–orbit coupling acts on the spatial and spin coordinates together, so its eigenstates are spin-orbitals, for example combinations of dxz↑ and dyz↓ with complex coefficients. It is then no longer possible to assign an electron to a single real orbital such as dxy.
- The labelling of the levels. Crystal-field levels are labelled by irreducible representations of the point group (t2g, eg); spin–orbit levels are labelled by the total angular momentum quantum number j. In an octahedral field the t2g set behaves as an l = 1 shell with the sign of the coupling reversed, so spin–orbit coupling splits its six spin-orbitals into a jeff = 3/2 quartet, lowered by ζ/2, and a jeff = ½ doublet, raised by ζ. The two are separated by 3ζ/2.
- Kramers degeneracy. Neither a crystal field nor spin–orbit coupling can split a Kramers doublet; only a magnetic field can do so. The jeff = ½ doublet therefore has no remaining orbital degeneracy, and spin–orbit coupling removes the Jahn–Teller instability that a low-spin t2g⁵ ion would otherwise show.
- Quenching of orbital angular momentum. Real orbitals carry no net orbital angular momentum (quenching), so the magnetic moments of most 3d ions are close to the spin-only values. The eg pair cannot carry orbital angular momentum at all, whereas the t2g set retains an effective l = 1. Ions with orbitally degenerate T ground terms (for example high-spin Co²⁺ and Fe²⁺) therefore show appreciable orbital contributions to the moment, and significant magnetic anisotropy, even in the 3d series.
The lines show the energies of the t2g spin-orbital levels, each of which is a Kramers doublet, coloured from blue (jeff = 3/2 character) to violet (jeff = ½ character). Click a level in the diagram at the right of the plot to display its electron density. Filled circles represent electrons, placed in the low-spin arrangement appropriate for 4d and 5d ions.
The jeff = ½ state is central to much current research on compounds of the heavier transition metals. In Sr₂IrO₄, low-spin Ir⁴⁺ (5d⁵) has a filled jeff = 3/2 quartet and a single electron in the jeff = ½ doublet. The resulting half-filled band is narrow, and a moderate U is sufficient to split it, giving a spin–orbit Mott insulator. On honeycomb lattices with 90° metal–ligand–metal bond angles (Na₂IrO₃, α-RuCl₃), the same states give rise to bond-dependent Kitaev exchange, which is the most widely studied route to a quantum spin liquid. When a tetragonal or trigonal field comparable to ζ is introduced in the model above, the states lose their pure jeff = ½ character; the extent to which this occurs in real materials is an active area of research.
Glossary
About the model
All orbital energies are calculated with the angular overlap model, with eσ and eπ proportional to r⁻⁵. Their values are determined by Δₒ and by the eπ/eσ ratio of the chosen ligand, and an optional d–s mixing term (eds = ¼ eσ) can be included. Level energies are given relative to the barycentre. The Jahn–Teller energy surface is the sum of the d-electron energy, measured from the barycentre, and an elastic energy with one Morse (or harmonic) potential for each bond; the spin state is held fixed at the selected value. The band model uses semicircular densities of states with a Hubbard splitting. It is a schematic representation of the Zaanen–Sawatzky–Allen scheme, not a band-structure calculation. The spin–orbit model diagonalises ζ l·s, together with a tetragonal or trigonal field, in the basis of the six t2g spin-orbitals. The parameters of the examples are representative literature values, rounded for teaching purposes; they have not been fitted to experimental data.
Further reading
- B. N. Figgis and M. A. Hitchman, Ligand Field Theory and Its Applications, Wiley-VCH (2000).
- T. A. Albright, J. K. Burdett and M.-H. Whangbo, Orbital Interactions in Chemistry, 2nd ed., Wiley (2013).
- D. I. Khomskii, Transition Metal Compounds, Cambridge University Press (2014).
- P. A. Cox, Transition Metal Oxides: An Introduction to Their Electronic Structure and Properties, Oxford University Press (1992).
- I. B. Bersuker, The Jahn–Teller Effect, Cambridge University Press (2006).
- K. I. Kugel and D. I. Khomskii, “The Jahn–Teller effect and magnetism: transition metal compounds”, Sov. Phys. Usp. 25, 231 (1982).
- J. Zaanen, G. A. Sawatzky and J. W. Allen, Phys. Rev. Lett. 55, 418 (1985).
- B. J. Kim et al., Science 323, 1329 (2009); G. Jackeli and G. Khaliullin, Phys. Rev. Lett. 102, 017205 (2009).