Perovskite and Spinel Structures

Perovskite oxides are the structural chassis of an enormous swath of modern functional-materials chemistry. Halide perovskites such as methylammonium lead iodide (CH₃NH₃PbI₃, an ABX₃ compound with A = organic cation, B = Pb²⁺, X = I⁻) power the fastest-improving class of thin-film photovoltaics, having risen from single-digit to over 25% laboratory power-conversion efficiency within about a decade — and the same tolerance-factor reasoning developed for oxide perovskites is used, with appropriately re-scaled ionic/molecular radii, to screen candidate halide perovskite compositions computationally before synthesis. Oxide perovskites such as BaTiO₃ and Pb(Zr,Ti)O₃ (PZT) remain the industry-standard ferroelectric and piezoelectric ceramics in multilayer ceramic capacitors, ultrasound transducers, and actuators, with their functional behaviour traced directly to the tolerance-factor-driven cation-displacement distortion discussed above. High-temperature cuprate superconductors (YBa₂Cu₃O₇₋δ and relatives) and colossal-magnetoresistance manganites (La₁₋ₓCaₓMnO₃) are both derived from the perovskite lattice by chemical substitution on the A and B sites, exploiting the structure's tolerance for aliovalent doping while retaining the corner-sharing octahedral framework. Spinels are equally central to solid-state technology. LiCoO₂-related and LiMn₂O₄ spinel-type materials are cathode compounds in lithium-ion batteries, where lithium ions occupy octahedral sites within a close-packed oxide framework and cycle in and out during charge/discharge; the spinel LiMn₂O₄ specifically exploits the same tetrahedral/octahedral site bookkeeping developed here, with Li⁺ in tetrahedral holes and Mn³⁺/Mn⁴⁺ in the octahedral framework. Magnetite (Fe₃O₄), the inverse spinel worked out in detail above, remains the archetypal ferrimagnet: its two crystallographically and magnetically distinct octahedral and tetrahedral iron sublattices couple antiferromagnetically to each other while each sublattice orders ferromagnetically, giving the net moment exploited in magnetic recording media, biomedical MRI contrast agents (as superparamagnetic iron oxide nanoparticles), and — historically — the lodestone compass. Cobalt- and nickel-ferrite spinels (CoFe₂O₄, NiFe₂O₄) are workhorse soft and hard ferrite materials in inductor cores, microwave circulators, and magnetostrictive sensors, their inversion parameters (and hence magnetic properties) tunable by exactly the CFSE arguments used above. Finally, the same site-occupancy logic — asking which cation prefers tetrahedral versus octahedral coordination in a close-packed oxide lattice — reappears whenever real materials deviate from ideal stoichiometry, foreshadowing the broader subject of Point Defects in Ionic Solids, where antisite defects and cation disorder in exactly these perovskite and spinel lattices control ionic conductivity, catalytic activity, and mechanical properties.

Ask the tutor: How does the choice of normal versus inverse spinel structure change the predicted magnetic ordering and net magnetic moment in a ferrimagnet like Fe₃O₄?

Start with perovskite. Take the ideal cubic structure of CaTiO₃, formula ABX₃ with A = Ca²⁺, B = Ti⁴⁺, X = O²⁻. Place the B cation at the corners of a cube, the A cation at the body centre, and X anions at the face centres. Equivalently — and more usefully — put the B cation at the origin of a cube, oxide ions at the midpoints of the twelve cube edges, and the A cation at the cube centre. In this second description each B cation is surrounded by six oxide ions at the face-centres of its own coordination octahedron, and adjacent BX₆ octahedra share every corner with a neighbour, building a rigid three-dimensional corner-sharing octahedral network. The A cation sits in the twelve-coordinate cage left in the middle of eight such octahedra, coordinated to twelve X ions in a cuboctahedral arrangement. This is the structure of the mineral perovskite, CaTiO₃, and of a vast family of technologically important oxides. The close-packing view makes the "three ions in one lattice" idea precise. If you look only at the A and X ions, they together form a cubic close-packed array (an A ion and three X ions per unit cell face, mimicking the 1:3 ratio of ABX₃): A occupies one-quarter of the close-packed sites and X the other three-quarters, exactly as required by stoichiometry, and they pack together as if they were a single kind of nearly-equal-sized sphere. The small, highly charged B cation then occupies one-quarter of the octahedral holes of this close-packed A+X array — specifically the holes surrounded entirely by six X ions (never by A), which is why the B–X octahedra are so regular and so strongly bonded. This is a direct generalisation of the rock-salt-type filling of octahedral holes, except that here only some of the octahedral holes (those all-X) are filled, and the rest (holes with some A neighbours) are empty. Whether this idealised cubic arrangement survives depends on a simple geometric ratio, the Goldschmidt tolerance factor, discussed below. Most real perovskites are slightly distorted — tetragonal in BaTiO₃ below 393 K, orthorhombic in GaFeO₃-type tilted structures — because A is slightly too small or too large for a perfect fit; the corner-sharing octahedra tilt cooperatively to close the gap, without breaking any B–X bonds, a soft "rotate the rigid octahedra" distortion mode that is central to perovskite ferroelectricity and to the enormous compositional flexibility of the family (it tolerates A- and B-site substitution across almost the whole periodic table, hence its use as a template for high-Tc superconductors, ferroelectrics, and multiferroics). Now turn to spinel, AB₂X₄, the structure of MgAl₂O₄ and of a huge family of oxides written with a divalent A and trivalent B (2,3-spinels), though 4,2- and 6,1-charge combinations also exist. Here the anions X (almost always O²⁻ or S²⁻) form a cubic close-packed lattice on their own — a face-centred cubic array, exactly as in rock salt — and the cations are then distributed over the interstitial holes of that single close-packed sublattice. A close-packed array of N anions generates N octahedral holes and 2N tetrahedral holes; per formula unit AB₂X₄ (with 4 X ions), that means 4 octahedral holes and 8 tetrahedral holes available, but only 1 A and 2 B cations to place. The spinel structure fills exactly one-eighth of the tetrahedral holes and exactly one-half of the octahedral holes — never more, so that every filled site's cations are well separated (avoiding face-sharing polyhedra, which would bring cations into unfavourably close, high-repulsion contact). In the normal spinel arrangement, written A(Td)B₂O₄, the two A²⁺ ions per unit-cell formula sit in tetrahedral holes and the four B³⁺ ions sit in octahedral holes (numbers scaled to the cubic unit cell, which contains 8 formula units, 32 O²⁻, 8 tetrahedral cations and 16 octahedral cations). In the inverse spinel arrangement, written B(Td)ABO₄, one B³⁺ ion per formula unit moves into a tetrahedral hole, displacing the A²⁺ ion, which relocates into an octahedral hole alongside the remaining B³⁺. The octahedral sublattice in an inverse spinel therefore contains a 1:1 mixture of A and B cations, often in a random or partially ordered arrangement. Whether a given AB₂O₄ compound is normal, inverse, or an intermediate ("partially inverse," parameterised by an inversion parameter λ), is decided by the energetics worked out in the derivations below — principally the octahedral-site crystal field stabilisation energy of each cation's d-electron configuration.

Goldschmidt tolerance factor (perovskite geometric stability):

where: $r_A, r_B, r_X$ are the ionic radii (Shannon radii, matched to the correct coordination numbers — A is 12-coordinate, B is 6-coordinate, X is the anion) of the A-site cation, B-site cation, and anion respectively. Reading: $t$ compares the A–X contact distance along a cube face diagonal ($r_A+r_X$, scaled by $\sqrt2$ because the face diagonal of a cube of edge $a$ is $a\sqrt2$) to the B–X contact distance along a cube edge ($r_B + r_X = a/2$, so the edge is $a = 2(r_B+r_X)$). A perfectly cubic structure requires the A cation to exactly fit the cuboctahedral cage created by the BX₆ octahedral framework; $t=1$ is that exact geometric fit. Empirically, cubic or near-cubic perovskites occur for $0.9 \lesssim t \lesssim 1.0$; $t$ slightly below 1 (A too small) drives cooperative octahedral tilting (tetragonal, orthorhombic, rhombohedral distortions); $t$ well below about 0.71–0.8 or above about 1.0–1.1 destabilises the perovskite topology altogether in favour of other structures (ilmenite, hexagonal perovskite polytypes).

Octahedral vs tetrahedral site preference energy (spinel electronic stability), defined per cation as the extra stabilisation gained by occupying an octahedral rather than a tetrahedral hole:

where: $\mathrm{CFSE_{oct}}$ and $\mathrm{CFSE_{tet}}$ are the crystal field stabilisation energies of the same $d^n$ configuration in octahedral and tetrahedral crystal fields respectively (using $\Delta_{\mathrm{tet}} \approx \tfrac{4}{9}\Delta_{\mathrm{oct}}$ for the same metal–ligand distance and charge). Reading: cations with large octahedral CFSE and small tetrahedral CFSE (notably $d^3$ Cr³⁺, low-spin $d^6$ Fe²⁺/Co³⁺, and $d^8$ Ni²⁺) strongly prefer octahedral sites; cations with $d^0$, high-spin $d^5$, or $d^{10}$ configurations have zero CFSE in either field and show essentially no electronic preference, leaving the choice to be decided by size and Madelung (electrostatic) considerations instead.

Overall spinel free energy (schematic form used to rationalise normal vs inverse structures) sums the octahedral-site preferences of the two competing cations across the two possible arrangements:

where: the factor of 2 counts the two B³⁺ ions per formula unit that must move between sites when the structure inverts, and $\Delta E_{\text{Madelung}}$ is a lattice-electrostatic correction (typically smaller than the CFSE term but not always negligible, and it generally favours placing the higher-charge cation on the site with more nearest anion neighbours). Reading: if $\Delta_{\text{oct-pref}}(B^{3+})$ is much larger than $\Delta_{\text{oct-pref}}(A^{2+})$ — i.e. B³⁺ "wants" the octahedral site far more badly than A²⁺ does — moving B³⁺ into that site (even at the cost of displacing A²⁺ to tetrahedral) lowers the energy, i.e. $\Delta E_{\text{normal}\to\text{inverse}} < 0$ and the inverse structure is favoured; if the two cations have comparable or A-favouring octahedral preference, the normal structure survives.

Example 1 (foundational): counting holes and confirming spinel stoichiometry. A cubic close-packed (fcc) array of anions with $N$ ions per unit cell generates $N$ octahedral holes and $2N$ tetrahedral holes (standard result from Tetrahedral and Octahedral Holes). The conventional spinel unit cell contains 32 O²⁻ ions ($N=32$), so it contains $32$ octahedral holes and $64$ tetrahedral holes. The cell also contains 8 formula units of AB₂O₄, i.e. 8 A cations and 16 B cations, for 24 cations total. Check the stated fractional occupancies:

Both match the "1/8 Td + 1/2 Oh" rule quoted above, and the anion count is self-consistent: charge balance for a 2,3-spinel requires $2(1) + 3(2) = 8 = 2\times$(number of O per formula unit, i.e. $4\times2=8$), confirming AB₂O₄ is electroneutral, $2+6=8$. This is the arithmetic skeleton every spinel formula must satisfy regardless of normal/inverse detail.

Example 2 (applied): does BaTiO₃ satisfy the tolerance-factor window, and what distortion is expected? Use Shannon radii: $r_{\mathrm{Ba}^{2+}}(\mathrm{XII}) = 1.61$ Å, $r_{\mathrm{Ti}^{4+}}(\mathrm{VI}) = 0.605$ Å, $r_{\mathrm{O}^{2-}} \approx 1.40$ Å. Then

This is slightly above 1, meaning the Ba²⁺ ion is a little too large for a perfectly relaxed cuboctahedral cage; rather than tilting the octahedra (the response typically seen for $t<1$), the structure instead responds by a ferroelectric off-centre displacement of the small, highly polarisable Ti⁴⁺ ion within its oxygen octahedron, coupled to a small tetragonal elongation of the cell — precisely the distortion responsible for BaTiO₃'s room-temperature ferroelectricity below its Curie temperature of 393 K. Compare directly with CaTiO₃'s $t\approx0.97$ (Example above): CaTiO₃, with $t<1$, distorts by octahedral tilting (GdFeO₃-type, non-polar), while BaTiO₃, with $t>1$, distorts by cation displacement (ferroelectric) — the same tolerance-factor formalism correctly differentiates two qualitatively different distortion mechanisms depending on the sign of $t-1$.

Example 3 (exam/competition level): predicting the inversion parameter trend across a spinel series using CFSE. Consider the ferrite spinel series $\mathrm{M^{2+}Fe^{3+}_2O_4}$ and rank $\mathrm{NiFe_2O_4}$, $\mathrm{CoFe_2O_4}$, $\mathrm{MnFe_2O_4}$ by expected degree of inversion, using high-spin CFSE values (all M²⁺ and Fe³�hd⁺ taken high-spin, appropriate for these weak/moderate-field oxide hosts). Fe³⁺ is $d^5$ high-spin in all three compounds and — as derived above — has $\Delta_{\text{oct-pref}}(\mathrm{Fe}^{3+}) = 0$ regardless of host, so the comparison reduces entirely to each M²⁺ ion's own octahedral site preference, using the same $t_{2g}/e_g$ (or $e/t_2$) counting as before with $\Delta_t = \tfrac49\Delta_o$: For Ni²⁺ ($d^8$): octahedral configuration $t_{2g}^6 e_g^2$ gives $\mathrm{CFSE_{oct}} = -\left(6\times\tfrac25 - 2\times\tfrac35\right)\Delta_o = -\tfrac{6}{5}\Delta_o$; tetrahedral configuration $e^4t_2^4$ gives $\mathrm{CFSE_{tet}} = -\left(4\times\tfrac35-4\times\tfrac25\right)\Delta_t = -\tfrac45\Delta_t = -\tfrac45\times\tfrac49\Delta_o = -\tfrac{16}{45}\Delta_o$. So

the largest-magnitude octahedral preference of the three, correctly predicting NiFe₂O₄ should be strongly (essentially completely) inverse — matching the experimental result that NiFe₂O₄ is a fully inverse spinel, technologically important as a soft ferrite. For Co²⁺ ($d^7$): octahedral $t_{2g}^5e_g^2$ gives $\mathrm{CFSE_{oct}} = -\left(5\times\tfrac25-2\times\tfrac35\right)\Delta_o = -\tfrac{4}{5}\Delta_o$; tetrahedral $e^4t_2^3$ gives $\mathrm{CFSE_{tet}} = -\left(4\times\tfrac35-3\times\tfrac25\right)\Delta_t = -\tfrac65\Delta_t = -\tfrac65\times\tfrac49\Delta_o = -\tfrac{8}{15}\Delta_o$. So

a smaller magnitude than Ni²⁺ but still favouring octahedral occupancy, consistent with CoFe₂O₄ being observed experimentally as predominantly (though not always perfectly) inverse. For Mn²⁺ ($d^5$, high spin): as with Fe³⁺, a half-filled high-spin $d^5$ configuration has one electron per orbital in either geometry, so $\Delta_{\text{oct-pref}}(\mathrm{Mn}^{2+}) = 0$, exactly like Fe³⁺; with neither cation showing an electronic preference, MnFe₂O₄ is predicted to be close to the normal limit, its cation distribution decided by the (much weaker) size/Madelung terms — matching the experimental observation that MnFe₂O₄ is only weakly/partially inverse (inversion parameter around 0.2, far closer to normal than NiFe₂O₄'s near-total inversion). The predicted order of inversion tendency, $\mathrm{Ni}>\mathrm{Co}\gg\mathrm{Mn}$, tracks the calculated $|\Delta_{\text{oct-pref}}|$ ordering $0.844\,\Delta_o > 0.267\,\Delta_o \gg 0$ exactly.

Victor Moritz Goldschmidt, the Norwegian-born mineralogist and geochemist who systematised ionic radii and the tolerance factor concept, is the natural focus for the perovskite half of this story. Working in Oslo in the 1920s, Goldschmidt built one of the first extensive tables of ionic radii from X-ray crystal structure data, and used them to formulate general packing rules for how ions of given size and charge combine into stable lattices — rules that became the foundation of structural (geochemical) crystallography. His 1926 paper on the crystal structure of perovskite-type compounds introduced the size-ratio criterion that today bears his name, and his broader radius-ratio framework — extending the same logic used for Zinc Blende, Wurtzite and Rutile Structures — remains the standard first-pass tool for predicting whether a given ABX₃ combination will adopt the perovskite structure at all.