The Robin-Day Classification: Mixed-Valence Compounds and the Creutz-Taube Ion

What 'mixed valence' means and why Robin and Day needed three boxes

A mixed-valence compound contains a single element present in two (or more) formal oxidation states at crystallographically or chemically distinct sites — the same metal wearing two hats. The phenomenon is ancient in a practical sense: Prussian blue, Fe₄[Fe(CN)₆]₃, has been a pigment since 1704 and owes its intense colour to iron existing simultaneously as Fe(II) (low-spin, bound to carbon of cyanide) and Fe(III) (high-spin, bound to nitrogen). The deep blue is not a d-d transition of either ion — it is an intervalence transition in which an electron is optically excited (photoinduced) from Fe(II) to Fe(III).

In 1967 Melvin B. Robin and Peter Day reviewed the whole zoo of such species and proposed a three-tier classification keyed to a single physical variable: the degree of electronic communication between the two redox sites, quantified by the electronic coupling matrix element H_ab (also written V or H_AB). Their scheme is not about how many oxidation states are present but about how well the sites talk to each other relative to the energy cost of moving charge:

• Class I — the sites are effectively insulated (H_ab ≈ 0). Valences are fully trapped, integer, and often sit in different coordination geometries. There is no intervalence band; the compound behaves as a physical mixture of two ordinary complexes.
• Class II — weak but non-zero coupling (0 < 2H_ab < λ). Valences remain distinguishable, but a low-energy optical band appears: the intervalence charge-transfer (IVCT) or metal-to-metal charge-transfer band. Thermal electron transfer between sites is now possible.
• Class III — strong coupling (2H_ab ≥ λ). The extra electron is fully delocalized over both sites; each metal carries a fractional, averaged oxidation state, and the concept of a localized 'valence' breaks down.

The genius of the scheme is that it maps a continuum of coupling onto three qualitatively different regimes, each with its own spectroscopic fingerprint. The remaining sections show that the boundaries are governed by the competition between coupling H_ab and reorganization energy λ — precisely the language of Marcus-Hush electron-transfer theory.

The two-state potential surface: Marcus-Hush and the origin of the IVCT band

Robin and Day gave the phenomenology; Noel Hush (1967, building on Rudolph Marcus's electron-transfer theory) gave the quantitative model. Consider two redox sites A and B and a single mobile electron. Draw the electronic energy against a single generalized nuclear coordinate that interpolates between 'electron on A' and 'electron on B' (mostly metal-ligand bond lengths and solvent polarization). Each diabatic (non-interacting) state is a parabola of force constant f; for a symmetric system A-A the two parabolas are identical wells offset along the coordinate, crossing at the midpoint.

The vertical energy gap between the two diabatic parabolas measured at the equilibrium geometry of one well is the reorganization energy λ, split into an inner-sphere part λᵢ (metal-ligand bond compression/elongation) and an outer-sphere part λₒ (reorientation of solvent dipoles). Turning on electronic coupling H_ab mixes the two diabatic states; the crossing becomes an avoided crossing, and the adiabatic ground surface acquires the classic double-well (Class II) or single-well (Class III) shape depending on the size of H_ab relative to λ.

For a symmetric Class II system, Hush derived the central results that make the theory testable. The optical IVCT band maximum equals the total reorganization energy: E_op = ν̃_max = λ. The thermal barrier to electron transfer is ΔG‡ = λ/4 − H_ab + H_ab²/λ (the −H_ab term is the resonance stabilization at the crossing). And the coupling itself can be extracted directly from the IVCT band using the Hush formula:

H_ab = (0.0206/d) · √(ε_max · ν̃_max · Δν̃₁/₂) (cm⁻¹), where ε_max is the molar absorptivity (M⁻¹cm⁻¹), ν̃_max and the full width at half maximum Δν̃₁/₂ are in cm⁻¹, and d is the effective donor-acceptor distance in Å. Because a Class II band arises from a thermally equilibrated distribution of solvent configurations, its predicted Gaussian half-width is Δν̃₁/₂ = √(2310 · ν̃_max) at 298 K — a diagnostic that a genuine Class II band should obey and a delocalized Class III band should violate (Class III bands are much narrower than this Gaussian limit and cut off sharply on the low-energy side).

Diagnostics: how you actually tell the classes apart

The classification is only as useful as the experiments that assign a compound to a box. Several handles, used together, do the job:

• IVCT band shape and solvent dependence. A Class II band is broad, roughly Gaussian, obeys the √(2310·ν̃_max) width, and — crucially — its energy shifts with solvent because E_op = λ contains the outer-sphere λₒ, which depends on the Pekar factor γ = (1/n² − 1/εₛ) (n = refractive index, εₛ = static dielectric constant). A Class III band is narrow, asymmetric (steep low-energy edge), and solvent-independent, because a fully delocalized electron has no solvent trapping to reorganize.
• Comproportionation constant K_c. From cyclic voltammetry, the separation of the two sequential one-electron waves ΔE₁/₂ gives K_c = exp(ΔE₁/₂·F/RT) = 10^(ΔE₁/₂/0.0592 V) at 298 K for the equilibrium [2,2] + [3,3] ⇌ 2[2,3]. A large K_c signals thermodynamic stability of the mixed-valence state but is only partly electronic — it also folds in electrostatics, statistics (a factor of 4), and inductive effects, so a big ΔE₁/₂ does not by itself prove strong coupling.
• Vibrational spectroscopy (IR/Raman) on the intrinsic timescale. Because IR probes on a ~10⁻¹³ s timescale, faster than most thermal hopping, it can resolve two distinct oxidation-state environments even when they interconvert. A localized (Class II) M-CO or bridging-ligand mode appears as two bands (one per oxidation state); a delocalized (Class III) system shows a single averaged band. This is the classic tool for distinguishing Class II from Class III when the optical picture is ambiguous.
• Structure and EPR g-anisotropy. Crystallographically inequivalent metal sites and localized, anisotropic EPR signals argue for trapping; equivalent sites and averaged/isotropic responses argue for delocalization.

No single measurement is decisive because each samples a different timescale (optical ~10⁻¹⁵ s, IR ~10⁻¹³ s, EPR ~10⁻⁹ s, Mössbauer ~10⁻⁷ s, X-ray ~10⁻¹⁸ s but time-averaged over the exposure). A compound can look 'delocalized' to a fast probe and 'localized' to a slower one — the essence of the notoriously hard Class II–III borderline, where 2H_ab ≈ λ and the double well is barely resolved.

The Creutz-Taube ion, worked through with real numbers

The Creutz-Taube ion, [(NH₃)₅Ru(μ-pyrazine)Ru(NH₃)₅]⁵⁺, prepared by Carol Creutz and Henry Taube (communication 1969; full paper J. Am. Chem. Soc. 1973, 95, 1086), is the most studied mixed-valence molecule ever made. Each ruthenium is octahedral, six-coordinate: five ammines plus one pyrazine nitrogen. In the fully oxidized [3,3] form both are Ru(III) (low-spin d⁵); fully reduced [2,2] both are Ru(II) (low-spin d⁶). The mixed-valence [2,3] state, formally Ru(II)/Ru(III), is the object of the fight.

The electrochemistry is textbook. Cyclic voltammetry shows two reversible one-electron waves separated by ΔE₁/₂ ≈ 0.39 V, giving a comproportionation constant K_c = 10^(0.39/0.0592) ≈ 4×10⁶ — the mixed-valence ion is strongly favoured over a 1:1 mixture of [2,2] and [3,3]. Its IVCT band lies in the near-IR at ν̃_max ≈ 6370 cm⁻¹ (about 1570 nm) with ε ≈ 5000 M⁻¹cm⁻¹. Feeding a localized (Class II) interpretation into the Hush formula with d ≈ 6.9 Å gives an electronic coupling of order H_ab ≈ 3000–4000 cm⁻¹ — large enough that 2H_ab is comparable to λ, placing the ion right on the Class II–III boundary.

Three observations pushed the consensus toward Class III (delocalized) for the pyrazine-bridged ion: (1) the IVCT band is much narrower than the Class II Gaussian √(2310·6370) ≈ 3840 cm⁻¹ prediction and is solvent-independent; (2) the near-IR band shows a sharp low-energy cutoff characteristic of a delocalized transition; and (3) intervalence-band and vibrational analyses (notably by Meyer, Hush, and later resonance-Raman and near-IR work) are most consistent with a single, symmetric ground-state minimum. Yet the assignment was fiercely debated for ~20 years precisely because different techniques on different timescales gave different answers — a warning that the Robin-Day boxes are idealized limits of a continuum. Replacing pyrazine with the longer 4,4'-bipyridine bridge, [(NH₃)₅Ru(μ-4,4'-bpy)Ru(NH₃)₅]⁵⁺, drops H_ab, widens and solvent-sensitizes the IVCT band, and yields an unambiguous Class II ion — showing the classification tuning smoothly with bridge length.

Limits, subtleties, and where the two-state picture breaks

The elegance of Robin-Day plus Hush comes from a deliberately minimal two-state, single-mode, symmetric model. Real systems strain every assumption, and knowing the failure modes is what separates a competent assignment from a naïve one.

• The bridge is not innocent. The two-state model treats the bridge as a passive tunnelling barrier (superexchange), but for conjugated or redox-active bridges (pyrazine, viologens, polyenes) the bridge orbitals mix strongly and can even carry the odd electron themselves. Then a proper description needs a three-state (donor-bridge-acceptor) model, and 'is the metal Ru(2.5)?' becomes ill-posed. Brunschwig, Creutz, and Sutin (Chem. Soc. Rev. 2002) recast the borderline explicitly in these terms.
• Class II–III borderline. Exactly at 2H_ab = λ the ground-state double well flattens to a single minimum; here band shape, solvent dependence, and IR all give contradictory or intermediate answers. This is not merely experimental noise — it is the physical crossover, and the Creutz-Taube ion lives in it.
• Asymmetric systems. When the two sites differ (different metals or ligands), a driving force ΔG° enters: E_op = λ + ΔG°, so the IVCT energy no longer equals λ, and Hush's symmetric formulas must be modified. K_c likewise acquires a thermodynamic asymmetry term.
• Vibronic coupling and PKS theory. The Piepho-Krausz-Schatz (PKS, 1978) vibronic model treats the coupling of the electronic states to a symmetric bridging vibration explicitly and reproduces band shapes across the Class II→III transition that the purely electronic Hush picture cannot. It shows the 'valence' can be dynamically averaged by the vibration even when H_ab is modest.
• Reorganization energy is solvent- and temperature-dependent. λₒ scales with the Pekar factor γ; cooling or changing solvent can push a molecule across the class boundary. Some biferrocenium salts are famously valence-tautomeric, switching between trapped and delocalized behaviour with temperature and even with the counterion and crystal packing (Hendrickson's work), proving the classes are not fixed molecular labels but conditions-dependent states.

In short, Robin-Day is a map, not a territory: superb for organizing intuition, but the interesting chemistry lives on its borders.

Why it matters: from Marcus theory to molecular electronics

Mixed-valence chemistry is the cleanest experimental testbed for Marcus-Hush electron-transfer theory, and that connection earned Rudolph Marcus the 1992 Nobel Prize in Chemistry and Henry Taube the 1983 Nobel Prize (for mechanisms of electron-transfer reactions, including the inner-sphere pathway he established). A Class II mixed-valence ion is, in effect, a self-exchange reaction with the two partners covalently tethered: the intramolecular electron transfer it undergoes is the same event that, in the bimolecular world, controls corrosion, respiration, and photosynthesis. By reading λ straight off the IVCT band (E_op = λ) and H_ab off its intensity, one can predict the thermal electron-transfer rate through k_et = κ·ν_n·exp(−ΔG‡/RT) with ΔG‡ = λ/4 in the weak-coupling limit — a rare case where an optical spectrum directly forecasts a rate constant.

Biologically, mixed-valence metal clusters are ubiquitous electron conduits: the [2Fe-2S] and [4Fe-4S] ferredoxins, the CuA site of cytochrome c oxidase (a genuinely delocalized Class III Cu(1.5)-Cu(1.5) pair whose narrow near-IR band is the enzymatic analogue of the Creutz-Taube spectrum), and the oxygen-evolving Mn₄CaO₅ cluster of Photosystem II all exploit valence delocalization to move electrons with minimal reorganization. Double-exchange coupling — Zener's mechanism, in which delocalization of an itinerant electron aligns core spins ferromagnetically — is the magnetic face of the same physics and underlies mixed-valence manganite magnetism.

Technologically, the field has become the molecular-scale playground for molecular electronics and switching. A mixed-valence dimer is a two-site 'wire' whose conductance is set by H_ab; tuning bridge length, conjugation, and redox potentials lets chemists build molecular rectifiers, memory elements, and quantum-dot-cellular-automata cells. Electrochromic materials (Prussian-blue displays, ruthenium polymers) switch colour precisely by turning intervalence bands on and off. The through-bond distance dependence of H_ab — which decays roughly exponentially, H_ab ∝ exp(−βr/2) with attenuation β characteristic of the bridge — is now the design rule for how far and how fast charge will travel across a molecule, from the ferredoxins of a cell to a candidate single-molecule transistor.

The three Robin-Day classes contrasted by electronic coupling, spectroscopy, and physical behavior.

Property	Class I (trapped)	Class II (weakly coupled)	Class III (delocalized)
Electronic coupling H_ab	≈ 0 (2H_ab ≪ λ)	0 < 2H_ab < λ	2H_ab ≥ λ
Valences	Distinct, integer, trapped	Distinct but exchanging	Averaged, non-integer
IVCT band	Absent (or vibronic only)	Broad, solvent-dependent, Gaussian	Narrow, intense, solvent-independent
Sites' symmetry	Inequivalent geometries	Two minima on ground surface	Single minimum, equivalent sites
Example	GaCl₂ = Ga⁺[GaCl₄]⁻	Biferrocenium; many Ru dimers	Creutz-Taube ion (widely held)