Bridging Carbonyls and Polynuclear Clusters
Look at Co₂(CO)₈ in solution. You see two carbonyl stretching patterns fighting for attention. There is a set of sharp bands near 2075–2020 cm⁻¹. There is also a softer one near 1857 cm⁻¹. This sits almost 200 wavenumbers below any terminal M–C≡O. That low band is not noise. It is the fingerprint of a carbon monoxide molecule lying sideways across a metal–metal bond. It donates electron density to two metals simultaneously. This single spectroscopic clue opens up a huge swath of chemistry. It leads to the polynuclear metal carbonyl clusters. These clusters bridge molecular organometallics and metallic/oxide surfaces. They also serve as structural models for chemisorbed CO on heterogeneous catalysts. Bridging carbonyls matter for good reason. They are not a minor structural curiosity. They are diagnostic of, and often essential to, metal–metal bonding. A μ₂-CO ligand simultaneously stabilizes a short M–M contact. It does this by providing a low-lying acceptor orbital that spans both metals. It also reports on the electron density at the cluster core. This happens via its stretching frequency. Clusters built from face- and edge-bridged M–M frameworks are the test bed for electron-counting rules. Examples include Fe₃(CO)₁₂, Co₄(CO)₁₂, Rh₆(CO)₁₆, and Os₅(CO)₁₆. These rules let chemists predict a cluster skeleton's shape. It might be a triangle, a tetrahedron, an octahedron, or something more exotic. The big picture connects three things already familiar from Metal Carbonyl Complexes and Structures and Infrared Spectroscopy of Carbonyl Ligands. These are the π-backbonding picture of terminal M–CO, the counting of valence electrons at a single metal center, and the vibrational signature of C≡O bond order. Here those ideas get extended further. They now cover two-metal (μ₂) and three-metal (μ₃) bridges. They also cover direct M–M bonds treated as two-electron two-center contacts. Finally, they extend to cluster-wide electron-counting schemes. These include the effective atomic number (EAN/18-electron) rule applied atom-by-atom. They also include the total valence electron (TVE) count. This count is compared against Wade–Mingos polyhedral skeletal electron pair (PSEP) predictions for more electron-deficient cages. This article builds the bonding picture for μ₂- and μ₃-CO from group-theoretic first principles. It derives the electron-counting formulas used for small clusters: dimer, triangle, tetrahedron, octahedron. It works three fully quantitative examples: Co₂(CO)₈, Fe₃(CO)₁₂, Rh₆(CO)₁₆. It closes with the scope limits and common counting traps. It also traces the historical route from Roland Hieber's early polynuclear carbonyls to Kenneth Wade's electron-counting insight. This story leads naturally into the wider family of Isoelectronic and Related Pi-Acceptor Ligands.
- A bridging CO (μ₂ or μ₃) donates a lone pair and accepts back-donation from two or three metals at once, lowering ν(CO) by 100–200 cm⁻¹ relative to terminal CO.
- The bonding picture: CO's filled π* combines symmetrically with metal d-orbitals across the M–M edge or face, while the CO lone pair (σ) donates into an empty metal hybrid — a four-electron, three-(or four-)center interaction.
- Electron counting in clusters uses the same 18-electron logic per metal, but M–M bonds and bridging ligands must be allocated consistently; the total valence electron (TVE) count for the whole cluster is the reliable macroscopic check.
- Bridging CO acts as a 2-electron donor overall (as does terminal CO); it is the distribution of that pair between two/three metals, not the electron count, that differs from terminal binding.
- For larger, more open cages the 18-electron-per-metal rule breaks down; Wade's rules (PSEP counting) take over, predicting closo/nido/arachno cluster shapes from skeletal electron pairs.
- Key idea: a bridging carbonyl is a π-acceptor ligand delocalized over an M–M(–M) unit, and correct cluster electron counting requires bookkeeping M–M bonds and bridge electrons exactly once each.
Where this lives today
Polynuclear carbonyl clusters are far more than a counting exercise — they are structural and mechanistic proxies for the active sites of heterogeneous catalysts. A μ₃-CO capping a triangular Ru₃ or Os₃ face is a molecular model for CO chemisorbed on a metallic (111) surface facet, and comparisons between cluster ν(CO) values and the vibrational spectra of CO adsorbed on Pt, Rh, or Ru single-crystal surfaces (measured by reflection–absorption IR spectroscopy, RAIRS) have been used for decades to assign surface CO binding sites (atop, bridge, three-fold hollow) by analogy with the well-defined bridging modes discussed here. Rhodium and cobalt carbonyl clusters related structurally to Co₄(CO)₁₂ and Rh₆(CO)₁₆ are relevant to industrial hydroformylation and Fischer–Tropsch chemistry, where mechanistic proposals invoke cluster fragmentation to mononuclear active species, cluster-surface intermediates with bridging acyl and CO ligands, and CO insertion steps whose transition states resemble the bridging geometries seen crystallographically in stable clusters. Iridium and osmium clusters (e.g., derivatives of Ir₄(CO)₁₂ and Os₃(CO)₁₂) serve as precursors for supported bimetallic and monometallic nanoparticle catalysts, where controlled thermal or oxidative decomposition of a well-defined molecular cluster onto an oxide support is used to generate uniform, sub-nanometer metal particles — "cluster-derived catalysts" — whose activity and selectivity can sometimes be traced back to the nuclearity and ligand arrangement of the parent cluster. Beyond catalysis, mixed-metal and heterometallic carbonyl clusters (containing, e.g., Fe/Ru/Os or Rh/Pt combinations) are used as synthetic entry points to nanoscale metal chalcogenide and phosphide clusters, and as precursors in chemical vapor deposition routes to thin metallic films. The fundamental electron-counting logic developed here — TVE, per-metal 18-electron bookkeeping for small close-shell clusters, and Wade–Mingos PSEP counting for larger or more open cages — remains the standard first-pass tool used by working inorganic chemists to rationalize a newly reported cluster's geometry before committing to full computational (DFT) structure prediction, and it dovetails directly with the electron-counting logic used for other strong π-acceptor ligand families such as those covered under Isoelectronic and Related Pi-Acceptor Ligands.
Ask the tutor: How would the electron count and predicted geometry change if [HFe₄(CO)₁₃]⁻ were oxidized by one electron instead of reduced further?
The physical picture
Terminal carbonyl bonding is already familiar: CO donates its 5σ lone pair (largely carbon-centered, weakly C–O antibonding) into an empty metal d/s/p hybrid, while filled metal d(π) orbitals back-donate into CO's empty 2π orbital. This synergic donor–acceptor pair is why ν(CO) drops from 2143 cm⁻¹ (free CO) to roughly 2120–1850 cm⁻¹ in terminal complexes, tracking the electron richness of the metal. When CO bridges an M–M edge (μ₂-CO), the same 2π orbital set — doubly degenerate in free CO, split into two orthogonal π orbitals in a bound ligand — now has a orbital of the right symmetry to overlap with a <strong>symmetric combination of d-orbitals on both metals simultaneously</strong>. Concretely, take the CO axis perpendicular to the M–M vector, sitting above the midpoint. One member of the π(CO) pair lies in the M–C–M plane; it forms a three-center, two-electron-acceptor interaction: metal(1) dπ + metal(2) dπ → CO π, all three orbitals of the same local symmetry combining constructively. Electron density flows from the metal–metal bonding manifold into CO π, weakening the C–O bond further than a single terminal M→CO backdonation would, and this is why ν(CO)bridging < ν(CO)terminal. The carbon lone pair (5σ) plays a complementary role: it can still donate into an empty metal orbital, but geometrically it donates into an orbital that has substantial density on both metals (again a symmetric combination), making the μ₂-CO a genuine three-center, four-electron bonding motif overall (2 electrons in, formally, the σ-donor interaction; the extra backbonding density further stabilizes the M–M bond itself). This is why bridging CO is frequently associated with, but not strictly required for, a formal metal–metal bond: it "staples" two metals together electronically as well as geometrically, shortening M–M distances by roughly 0.1–0.15 Å relative to unbridged analogues with the same formal bond order. Extending to μ₃-CO (face-capping, as seen in Rh₆(CO)₁₆ and many Ru₃/Os₃ clusters), the carbon sits above a triangular M₃ face, and now the relevant metal-orbital combination is a totally symmetric (a₁-like) combination of three dπ orbitals overlapping with the CO π* system in C₃ᵥ local symmetry. The ν(CO) drops further still, often into the 1730–1620 cm⁻¹ range, because the antibonding electron density is now spread — and reinforced by backdonation — over three metal centers.
The core ideas
Electron count at a bridging CO — the ligand remains a net 2-electron donor no matter how many metals it touches, exactly as for terminal CO. What differs is how those two electrons (plus the backdonated density) are shared:
where: the bracket denotes the CO lying across the M–M edge, donating its σ lone pair into a metal-centered orbital that has amplitude on both metals, and receiving π backdonation from a symmetric combination of both metals' d-orbitals.
Reading: a bridging CO does not "give" extra electrons to the cluster; it is still one CO, one C≡O(→O) unit, contributing the same 10 valence electrons total (4 from C, 6 from O) that it always has. Electron-counting conventions differ only in how the 2 donor electrons are assigned to individual metals for the purposes of computing each metal's individual 18-electron count — the ionic method assigns CO as a neutral 2e donor to the metal it is closest to, or 1e to each in symmetric bridging; the total-cluster count is unaffected by this choice.
Total valence electron (TVE) count for a cluster:
where: "group electrons of metal" means the d-electron count of the neutral metal in group-electron convention (e.g., Fe = 8, Co = 9, Rh = 9, Os = 8), each terminal or bridging CO donates 2 electrons regardless of hapticity, and charge is subtracted if the cluster is a cation (added if anionic).
Reading: TVE is a global invariant — it does not care about the internal distribution of bonds and bridges. It is the safest first check on any cluster formula, and it feeds directly into the skeletal-electron-pair count used in Wade's rules for cages that are too electron-poor for an 18-electron-per-metal description.
Effective atomic number rule extended to clusters (per-metal 18e count):
where: is the d-electron count of metal , counts CO ligands attributed to metal (bridging COs typically contributed as 1 electron per metal when shared over 2 metals, i.e. the "2 electrons split evenly"), and is the number of two-center M–M bonds from metal to its neighbors (each M–M bond contributes 1 electron to each partner, i.e. 2 electrons shared per bond).
Reading: each metal aims for EAN = 18 (electron configuration of the next noble gas), exactly the same target as in Metal Carbonyl Complexes and Structures — clusters simply add M–M bonds and bridge-sharing as additional electron sources per metal, on top of the terminal-ligand donation already familiar from mononuclear carbonyls.
Wade's rule (PSEP) skeletal electron count, needed once clusters become too electron-poor or geometrically non-close-packed for the 18e-per-metal picture (introduced here, developed fully elsewhere):
where: is the number of vertex metal atoms in the cluster core, and 12 is subtracted per metal because each vertex metal is assumed to use 3 orbitals (and hence 6 of its 9 valence orbitals' worth of electrons, contributing 12 electrons) for exo-bonding to its own terminal ligands, reserving the rest for cluster bonding.
Reading: SEP = n+1 predicts a closo (complete deltahedron) shape, SEP = n+2 a nido (one vertex removed) shape, and so on — this is the cluster analogue of borane electron counting, and it is the tool that eventually replaces "count to 18 at every metal" once clusters get large or unsaturated.
Goal. Show why a CO ligand bridging two metals experiences more π backdonation into its 2π* orbital than the same CO bound terminally to one metal — hence a lower C–O force constant and lower ν(CO).
Step 1 — Set up local symmetry. Place two metal atoms M₁, M₂ symmetrically about a mirror plane, with the bridging CO's carbon on that mirror plane, C and O both in the M₁–C–M₂ plane, C–O axis roughly perpendicular to the M₁–M₂ vector. The relevant point group for the M₂CO core is C₂ᵥ (or approximately so). CO's doubly degenerate π* pair, which transforms as a single irreducible representation (e) in the local C∞ᵥ symmetry of isolated CO, splits into two components on binding: one lying in the M–C–M plane (transforms as a₁ or b₁ depending on convention, symmetric with respect to the mirror) and one perpendicular to that plane (transforms as b₂ or a₂, antisymmetric).
Step 2 — Identify the symmetric metal-orbital combination. On each metal, a dπ orbital oriented toward the bridging carbon has the correct local symmetry to overlap with the in-plane π(CO) component. Because the two metals are related by the mirror plane, their individual dπ orbitals combine into a symmetric (bonding-type, same sign) combination and an antisymmetric combination . Only the <strong>symmetric</strong> combination has the correct symmetry (matches the in-plane π component) to overlap with π(CO); the antisymmetric combination is symmetry-forbidden from mixing with this π component (near-zero overlap by symmetry).
Step 3 — Compare overlap magnitudes. In a terminal M–CO unit, one metal dπ orbital overlaps with one π(CO) lobe — a standard 2-center interaction. In the bridging case, the symmetric combination has amplitude reinforcing constructively on both sides of the carbon, so the net spatial overlap integral with the in-plane π(CO) orbital is larger in magnitude than a single M–CO overlap (each metal contributes constructively rather than the interaction being confined to one M–C vector). This is a three-center, two-(or more)-electron bonding interaction analogous to the bridging hydride/B–H–B 3c-2e bond familiar from borane chemistry, but here it is a π-type acceptor interaction rather than a σ-donor one.
Step 4 — Consequence for the C–O bond. Greater metal-to-ligand electron density transferred into π(CO) (an orbital that is C–O antibonding) means greater weakening of the C–O bond order. Using the standard force-constant correlation with backdonation (as established via the Cotton–Kraihanzel and related treatments referenced in the study of terminal carbonyls), a larger population of π(CO) directly lowers the C–O stretching force constant , and hence
drops. Because the μ₂ interaction pools backdonation from two electron-rich metal centers into the same π* orbital, the total backdonation received by a bridging CO typically exceeds that received by any single terminal CO on the same cluster, even though each individual M–C(bridge) contact may be weaker than a typical terminal M–C bond. The result, confirmed by every structurally characterized case, is
with μ₃-CO going lower still (1730–1620 cm⁻¹) because now three metals' worth of dπ density can combine symmetrically (an a₁ representation in local C₃ᵥ symmetry) to donate into the CO π* system.
Limiting-case check. If the two metals were pulled infinitely far apart while somehow keeping the same C–O–M₂ geometry (an unphysical but useful limit), the symmetric combination would have negligible overlap with a π*(CO) orbital centered near only one metal, and the system would smoothly reduce to ordinary terminal bonding at that one metal — recovering , exactly as it must, since a "bridge" to an infinitely distant second metal is physically just a terminal ligand.
Go deeper · Derivation B · 2 minTotal valence electron counting for a triangular cluster: deriving the TVE–M–M bond relationship for Fe₃(CO)₁₂
Goal. Derive, from first principles of the 18-electron rule applied to each metal, the number of Fe–Fe bonds required in a triangular Fe₃ cluster carrying only carbonyl ligands, and show this is consistent with the observed structure (two bridging + ten terminal CO, one bridged edge).
Step 1 — Write the total valence electron count. Iron is group 8, contributing 8 valence electrons per atom in the neutral-atom convention. There are 12 CO ligands total (bridging or terminal — doesn't matter for the global count, since every CO donates 2 electrons regardless of hapticity):
Step 2 — Determine electrons needed for a triangular metal skeleton if each metal individually reaches 18. For 3 metal centers each independently satisfying the 18-electron rule, the total electron requirement, counting each M–M bonding electron pair once from each partner's perspective, is
because each M–M bond is shared between exactly two metals: when you sum "18 electrons at each metal" over all three metals, every M–M bonding electron pair has been counted twice (once at each end), so it must be subtracted once per bond to avoid double counting. Rearranged,
Step 3 — Solve for the number of M–M bonds using the TVE found in Step 1.
Step 4 — Interpret geometrically. Three M–M bonds among three metal atoms means every pair is bonded: a closed triangle of Fe–Fe bonds, each metal bonded to both others. This matches the crystallographically determined structure of Fe₃(CO)₁₂ — a triangle of iron atoms, each of the three Fe–Fe edges present, with one edge additionally bridged by two μ₂-CO ligands (giving that edge extra stabilization and a noticeably shorter Fe–Fe distance, 2.56 Å, versus 2.68 Å for the two unbridged edges) and the remaining ten CO ligands terminal (four on each of the two "wingtip" Fe atoms' equivalent, distributed as required by the low-temperature crystal structure, three terminal on the bridged pair each and three on the unbridged apex).
Limiting-case check. If instead one imagined a cluster with the same 48 TVE but attempted to force zero M–M bonds, Step 3's algebra would require , a contradiction — telling us immediately that three isolated, non-bonded 18-electron Fe(CO)ₓ fragments cannot be assembled from only 48 electrons and 12 COs; some M–M bonding is mandatory for electron sufficiency here, exactly as observed. Conversely, for an electron count of 54 (as in three genuinely independent, non-interacting 18-electron ML fragments), the formula predicts zero M–M bonds — consistent with, e.g., three separate mononuclear 18-electron carbonyls that happen not to be clustered at all.
Worked examples
Example 1 — Foundational: Co₂(CO)₈, verifying the 18-electron count with bridging CO and a Co–Co bond. Co is group 9 (9 valence electrons per neutral atom). The molecule is Co₂(CO)₈: six terminal CO and two bridging CO (μ₂) in the solid-state / low-temperature-solution C₂ᵥ isomer, with one Co–Co bond. Total valence electrons (global check):
For two metals each reaching 18 electrons with exactly one M–M bond connecting them,
These match, confirming one Co–Co bond is required and sufficient. Now check electron count at a single cobalt atom explicitly, say Co(1), which in the bridged structure bears 3 terminal CO plus a half-share of each of the two bridging CO plus the Co–Co bond:
That comes to 19, not 18 — this is the well-known subtlety of the "1-electron-per-metal" convention for symmetric bridges: it slightly overcounts unless the bridging COs are instead treated as normal 2-electron donors to one metal in the ionic/donor-pair convention with the M–M bond compensating. Using the alternative, fully self-consistent convention (each bridging CO assigned as a full 2e donor to one Co, with the other Co's shortfall made up entirely by the M–M bond order), both cobalts still average to 18 across the molecule, and the important, convention-independent result is the global one: TVE = 34, one Co–Co bond, both metals collectively 18-electron in the standard sense. The bonding/counting subtlety here is well known and is one of the reasons chemists prefer the global TVE check over insisting each atom individually and unambiguously totals to exactly 18 in every possible bookkeeping scheme. As a limiting case: the minor, rapidly-interconverting unbridged D₃d isomer of Co₂(CO)₈ has 4 terminal CO on each Co and one Co–Co bond, no bridges:
Again the "per-atom = 19" arithmetic artifact appears whenever a metal-metal bond electron is added on top of an already-fully-counted d⁹ + 4×CO = 17-electron fragment that then needs only 1 more electron (the bond) to reach 18 — the correct per-atom target when one M–M bond is present is in fact 18, obtained as . This example is a caution: bookkeeping conventions for "how many electrons does an M–M bond contribute to each atom" (1, matching a normal 2-electron 2-center bond split evenly) versus "how many does a bridging CO contribute to each metal" (commonly 1 electron per metal for a symmetric 2-electron bridge) must be applied consistently; mixing conventions is the single most common source of an "off-by-one" cluster electron count in student work.
Example 2 — Applied: Rh₆(CO)₁₆, an octahedral cluster with μ₃-CO capping alternate faces. Rh is group 9. Rh₆(CO)₁₆ consists of an octahedron of six Rh atoms, twelve M–M edges (an octahedron has 12 edges), 12 terminal CO (2 per Rh, one pointing outward from each vertex — actually the known structure has each Rh bearing 2 terminal CO) and 4 face-capping μ₃-CO (capping four of the octahedron's eight triangular faces, in an alternating pattern reminiscent of a tetrahedral subset of faces). Global TVE:
For an octahedral M₆ cluster with each metal individually 18-electron and with total M–M bonds (here all 12 edges of the octahedron are metal-metal bonded):
This does not match 86 — a genuinely useful discrepancy, because it signals that an octahedral M₆ cluster is not simply "every edge is a classical 2-electron 2-center bond, and every metal independently 18-electron." Applying Wade's rules instead (appropriate once the naive per-metal-18e bookkeeping fails to close), each Rh vertex uses 12 electrons for its own exo-cluster bonding (2 terminal CO × 2e = 4e reserved directly as external ligand bonds is only part of the accounting; the Wade convention reserves 12 electrons total per vertex for non-skeletal use), leaving:
Seven skeletal electron pairs for six vertices means — the closo condition — correctly predicting a closed octahedral deltahedron, exactly the observed Rh₆ geometry. This example shows explicitly why, once clusters exceed a handful of metals or adopt close-packed polyhedral cores, the Wade PSEP count (built from the same TVE) is the reliable predictive tool, superseding the atom-by-atom 18-electron bookkeeping that worked cleanly for the dimer and the triangle.
Example 3 — Exam/competition level: predicting whether [HFe₄(CO)₁₃]⁻ prefers a butterfly (nido) or closed-tetrahedral (closo) Fe₄ core. Consider the anion [HFe₄(CO)₁₃]⁻, a real, structurally characterized cluster. Iron: group 8, 4 atoms. Ligands: 13 CO (each a 2e donor regardless of terminal/bridging) plus one hydride (μ-H, a 1-electron donor in the neutral covalent counting convention, equivalent to a 2-electron donor as H⁻ in the ionic convention — here we use the neutral/covalent convention throughout for internal consistency, treating H as a 1-electron ligand and Fe as neutral atoms). The overall charge is −1, contributing one extra electron to the total count.
Now apply the Wade PSEP formula with vertex metals:
Six skeletal electron pairs for four vertices: compare against (closo, expected shape = tetrahedron) and (nido, expected shape = one vertex removed from a trigonal bipyramid, i.e. a "butterfly" Fe₄ arrangement, since a 5-vertex closo deltahedron — a trigonal bipyramid — missing one vertex leaves a 4-atom butterfly). Since SEP = 6 = n+2, the prediction is nido, i.e. a butterfly Fe₄ core, not a closed tetrahedron. This matches the experimentally determined structure: [HFe₄(CO)₁₃]⁻ indeed adopts a butterfly arrangement of four iron atoms (two "hinge" Fe atoms bonded to each other and to both "wingtip" Fe atoms, but the two wingtip Fe atoms not bonded to each other — five Fe–Fe bonds total rather than the six needed for a closed tetrahedron), with the bridging hydride typically found on one of the open (wingtip-wingtip non-bonded) edges or a face of the open cluster. As a limiting-case sanity check: if the hydride and the extra negative charge were both removed, i.e. consider the neutral, hypothetical "Fe₄(CO)₁₃" fragment, TVE would drop by 2 (losing the 1e from H and the 1e from the anionic charge) to 58, giving
predicting a closo tetrahedron instead — illustrating concretely how a single added electron pair (from protonation/reduction) can flip a cluster's preferred geometry from closed tetrahedral to open butterfly, exactly the kind of skeletal-electron-pair-driven structural change Wade's rules were designed to rationalize.
Limiting cases, scope & common traps
| Situation | What happens | Correct treatment |
|---|---|---|
| Small, close-packed clusters (M₂, M₃, M₄ tetrahedra) with all metals individually 18-electron | Per-atom EAN/18e counting works cleanly; TVE = 18n − 2(M–M bonds) | Use per-metal 18e count as primary tool; cross-check with global TVE |
| Larger or more open polyhedra (M₅, M₆ octahedra, butterflies, etc.) | Naive "every edge = 1 bond, every metal = 18e" often fails or is ambiguous (too many possible bond-count assignments) | Switch to Wade PSEP counting: SEP = ½(TVE − 12n); compare to n+1 (closo), n+2 (nido), n+3 (arachno) |
| Symmetric μ₂-CO bridging two equivalent metals | Ligand still donates only 2 electrons total; splitting "1e to each metal" is a bookkeeping convention, not a physical doubling of donation | Always check the global TVE; never let per-atom conventions imply the ligand donates more electrons overall than a terminal CO |
| μ₃-CO capping a triangular face | Same 2-electron total donation, now distributed over 3 metals; ν(CO) drops further (1730–1620 cm⁻¹) due to 3-center π-backbonding | Use IR shift as diagnostic for hapticity, not for electron count |
| Bridging vs terminal CO fluxionality (e.g., Co₂(CO)₈, Fe₃(CO)₁₂) | Bridging and terminal CO's interconvert rapidly on the NMR timescale even though IR (faster timescale) resolves distinct bridging/terminal bands | Don't assume a single static structure from solution ¹³C NMR alone; combine with low-temperature IR or solid-state XRD |
| High negative charge / hydride-rich clusters | Extra electrons can convert a closo to a nido/arachno shape (cluster "opens up") | Always add charge and hydride donor electrons explicitly into TVE before computing SEP |
| Clusters with interstitial (encapsulated) atoms, e.g., carbide clusters [Ru₆C(CO)₁₇] | The interstitial atom contributes its full valence electron count to the TVE, but occupies no true "vertex" position | Add interstitial atom's electrons to TVE; do NOT count it as one of the vertices in the Wade formula |
The chemist behind it
The systematic understanding of bridging carbonyls and cluster electron counting owes its central organizing idea — extending borane-style skeletal electron pair counting to transition-metal carbonyl clusters — to Kenneth Wade, working at Durham University in the 1970s. Wade had already built his reputation on borane and carborane cluster bonding, recognizing that BₙHₙ²⁻ cages followed a simple pattern relating the number of skeletal electron pairs to the polyhedral shape (closo, nido, arachno). The insight that transition-metal carbonyl clusters — chemically very different from boron hydrides on the surface, but topologically analogous cage compounds — obey essentially the same skeletal electron pair arithmetic (once each metal vertex's "exo" bonding electrons are subtracted off) was a landmark unification across what had been two separate subfields of inorganic chemistry.
Origins · 1 minKenneth Wade and how this was discovered
Why does bridging carbonyl geometry shorten the metal-metal distance?
A bridging CO provides an additional bonding pathway between the two metals: its π* orbital, populated by backdonation from a symmetric combination of both metals' dπ orbitals, creates extra electron density concentrated in the M–C–M region. This supplements whatever direct M–M σ-bonding interaction already exists, effectively adding bonding character along the M···M vector even beyond the formal metal–metal bond order. Experimentally, bridged M–M distances typically run 0.1–0.15 Å shorter than comparable unbridged M–M distances with the same nominal bond order — visible directly by comparing the bridged (2.52 Å) and unbridged (2.70 Å) Co–Co distances in the two isomers of Co₂(CO)₈. The bridge is best thought of as an electronic "staple" reinforcing the M–M bond, not merely a geometric spacer.
Is a bridging CO required whenever there is a metal-metal bond, or vice versa?
Neither implication holds strictly. Many well-characterized metal–metal bonded clusters and dimers (Mn₂(CO)₁₀, Re₂(CO)₁₀, the D₃d isomer of Co₂(CO)₈) have no bridging carbonyls at all — the M–M bond is a conventional two-center, two-electron σ bond with all CO ligands terminal. Conversely, bridging carbonyls can occasionally appear in geometries without a formal M–M bond, though this is rarer and usually signals a longer, weaker M···M contact. Bridging CO is best regarded as a common but not obligatory structural accompaniment to short M–M bonds, favored particularly by later transition metals in lower oxidation states with moderate steric bulk around the metal.
How do you decide, without a crystal structure, whether a cluster's CO ligands are terminal or bridging?
Infrared spectroscopy is the fastest diagnostic: terminal CO gives sharp bands in the 2120–1850 cm⁻¹ region, while bridging μ₂-CO gives distinctly separated bands in the 1850–1750 cm⁻¹ region, and μ₃-CO lower still (1730–1620 cm⁻¹). Because these ranges rarely overlap, a quick solution or solid-state IR spectrum, referenced against known standards, is usually sufficient for a first structural hypothesis, exactly the same logic developed for terminal carbonyls in the study of Infrared Spectroscopy of Carbonyl Ligands. Definitive confirmation, especially of the precise bridging pattern and cluster shape, still requires single-crystal X-ray diffraction, since IR alone cannot distinguish, for example, which specific edge of a triangle is bridged.
Why does the 18-electron rule sometimes fail for large clusters, forcing the switch to Wade's rules?
The 18-electron rule assumes each metal center behaves like an independent mononuclear fragment using nine valence orbitals (one s, three p, five d) to build bonds to its immediate ligand set, including its M–M neighbors. In small, close-packed clusters (dimers, triangles, tetrahedra) this bookkeeping is unambiguous because the number of M–M bonds equals the number of geometrically obvious edges. In larger or more open polyhedra, the "obvious" edge count and the electron-count-derived bond order can disagree, or multiple bonding descriptions become equally valid, because cluster (skeletal) bonding becomes delocalized over the whole polyhedral core rather than localized on individual two-center edges. Wade's rules sidestep this ambiguity entirely by counting total skeletal electron pairs and comparing to n+1/n+2/n+3, without ever needing to assign a specific bond order to each individual edge.
Does the choice of "ionic" versus "neutral/covalent" electron-counting convention change the final predicted structure?
No — both conventions, applied consistently, must give the same total valence electron count and hence the same Wade PSEP prediction and the same overall cluster electron count, because they are just different ways of partitioning the same total electrons between metal and ligand. The neutral (covalent) convention treats CO as a neutral 2-electron donor and hydride as a neutral 1-electron donor with the metal kept in its neutral, zero-oxidation-state electron count; the ionic convention treats hydride as H⁻ (2-electron donor) with the metal correspondingly assigned a formally higher oxidation state (and correspondingly fewer d-electrons). The two conventions must and do agree on the final TVE, and hence on the final answer, provided you do not mix them mid-calculation — which is precisely the trap flagged in the counting-traps table.
Why is Rh₆(CO)₁₆'s TVE-based per-metal count (84 electrons) different from its actual TVE (86), and what does that 2-electron discrepancy mean?
The mismatch is the signal that the naive assumption "every visible M–M edge is a classical, localized 2-electron 2-center bond, and every metal independently reaches exactly 18 electrons" is not the correct bonding description for this particular octahedral cage. Octahedral M₆ clusters, like octahedral boranes, generally have delocalized cluster bonding that cannot be partitioned into a simple edge-by-edge Lewis structure with each metal at a clean 18-electron count; the "extra" 2 electrons are exactly what is needed to push the cluster into Wade's SEP = n+1 closo-electron count (7 pairs for 6 vertices) rather than an idealized but ultimately non-self-consistent localized-bond count of 6 pairs. This is a textbook illustration of why the polyhedral skeletal electron pair method, not per-atom 18-electron bookkeeping, is the right tool once a cluster's cage is closed and its bonding is genuinely delocalized.
What happens to a cluster's electron count and shape when it is reduced or protonated?
Adding an electron (reduction) or a proton plus its bonding electron pair (protonation to form a hydride) both increase the total valence electron count, which directly increases the skeletal electron pair count computed from SEP = ½(TVE − 12n). Because SEP = n+1 gives closo, n+2 gives nido, and n+3 gives arachno structures, adding electron pairs systematically "opens up" a cluster core — a closed deltahedron can become a nido (vertex-missing) shape, and a nido shape can open further to arachno, exactly as demonstrated by comparing the neutral, hypothetical "Fe₄(CO)₁₃" (closo tetrahedron prediction) to the actual anionic hydride cluster [HFe₄(CO)₁₃]⁻ (nido butterfly prediction), which differ by exactly the two electrons contributed by the extra hydride and negative charge.