18-Electron Rule

The octet rule, grown five orbitals bigger

A carbon atom wants eight electrons because it has exactly four valence orbitals — one 2s and three 2p — and two electrons per orbital makes eight. That is the octet rule. A transition metal has the same s and p orbitals plus five d orbitals, for nine valence orbitals in total. Fill all nine and you hold 9 × 2 = 18 electrons. That is the 18-electron rule: a transition metal reaches a closed, noble-gas-like shell when the sum of its own d electrons and the electrons donated by its ligands equals 18.

The "noble gas" the metal is mimicking is the one at the end of its period. A first-row metal at 18 electrons has the valence configuration of krypton; a second-row metal mimics xenon; a third-row metal mimics radon. Because of this, the rule's older name is the effective atomic number (EAN) rule — Sidgwick noticed in the 1920s that stable metal carbonyls always pushed the metal's effective electron count up to the next noble gas.

The physical reason an 18-electron count is special is molecular-orbital, not magic. In an octahedral ML₆ complex the nine metal orbitals combine with ligand orbitals to give six bonding/non-bonding levels (the three t₂g and the bonding combinations) below a gap, and the antibonding eg* and higher levels above it. Eighteen electrons exactly fill everything bonding and non-bonding while leaving the strongly antibonding orbitals empty. Add a nineteenth electron and it must go into an antibonding orbital, which destabilizes the complex; leave one out at seventeen and you have an unpaired electron in a bonding region and a reactive radical.

Two ways to count, one answer

Electron counting looks intimidating because there are two accepted methods, and chemists switch between them mid-conversation. They always give the same total — they are two bookkeeping conventions for the same molecule.

Covalent (neutral-atom) method. Treat the metal as a neutral atom and count all of its valence electrons (= its group number). Treat each ligand as a neutral fragment and count how many electrons that neutral radical donates.

Total = (group number of M) + Σ (electrons donated by each neutral ligand) + (overall charge correction)
   charge correction: subtract the complex's charge (e.g. −1 for a cation, +1 for an anion)

Ionic (donor-pair) method. Assign the metal an oxidation state, count its d electrons at that charge, then add 2 electrons for every ligand treated as a closed-shell donor (an anion or a neutral lone-pair donor).

Total = (d electrons of M at its oxidation state) + Σ (2 per donor ligand)

The key is to stay inside one method per calculation. The donation numbers differ between the methods for charged ligands, which is exactly why mixing them produces wrong answers:

Ligand	Covalent (neutral) donation	Ionic donation	Type
CO, PR₃, NH₃, H₂O	2	2	L (neutral 2-e donor)
H, Cl, Br, CH₃	1	2 (as H⁻, Cl⁻…)	X (1-e covalent / anionic)
η⁵-Cp (cyclopentadienyl)	5	6 (as Cp⁻)	L₂X
η⁶-benzene	6	6	L₃
=O, =NR (terminal oxo/imido)	2	4 (as O²⁻, NR²⁻)	X₂
η²-alkene (C=C)	2	2	L
NO (linear)	3	2 (as NO⁺)	LX

Notice the neutral and ionic columns agree for CO, alkenes, and benzene (L-type ligands), and disagree for halides, hydride, and Cp (X-type and mixed ligands). The disagreement is paid back when you also account for oxidation state and overall charge, so the grand total matches.

Worked example: ferrocene counts to 18 either way

Ferrocene, Fe(η⁵-C₅H₅)₂, is the molecule that launched organometallic chemistry. Counting it both ways is the canonical sanity check.

COVALENT (neutral-atom) method
  Fe atom (group 8):                 8
  2 × η⁵-Cp• (neutral, 5 e each):   10
  overall charge of molecule (0):    0
  ────────────────────────────────────
  TOTAL                             18  ✓

IONIC (donor-pair) method
  Fe²⁺ has d-electron count:         6   (group 8 minus 2 for the +2 state)
  2 × Cp⁻ (6 e each as anion):      12
  ────────────────────────────────────
  TOTAL                             18  ✓

Both routes land on 18, which is why ferrocene is famously robust — it sublimes at 100 °C, survives boiling in air, and is oxidized only reversibly to the deep-blue 17-electron ferrocenium cation [Fe(Cp)₂]⁺. That one-electron oxidation removes a non-bonding electron, and the resulting 17-electron species is paramagnetic and a clean reference electrode couple at +0.40 V vs SHE — chemists use the ferrocene/ferrocenium couple as an internal voltammetry standard precisely because the 18 ⇌ 17 electron transfer is so reversible.

Contrast the neighbours. Cobaltocene, Co(Cp)₂, has 19 electrons — one in an antibonding e₁g* orbital — so it is a powerful one-electron reductant (E ≈ −1.3 V vs ferrocene) that readily gives up that electron to become the 18-electron cobaltocenium cation. Nickelocene has 20 electrons, two of them antibonding, and is even more reactive, readily losing a ring to relieve the over-count.

Metal carbonyls: the rule's home turf

Binary metal carbonyls are where the rule is sharpest, because CO is a strong π-acceptor that stabilises a full d shell and donates a clean 2 electrons. The number of CO ligands a neutral metal can carry is set by the arithmetic of reaching 18.

Complex	Group	Metal e⁻	CO (×2)	M–M / charge	Total
Ni(CO)₄	10	10	4 × 2 = 8	—	18 ✓
Fe(CO)₅	8	8	5 × 2 = 10	—	18 ✓
Cr(CO)₆	6	6	6 × 2 = 12	—	18 ✓
Mn₂(CO)₁₀	7	7	5 × 2 = 10	+1 (M–M bond)	18 ✓
Co₂(CO)₈	9	9	4 × 2 = 8	+1 (M–M bond)	18 ✓
V(CO)₆	5	5	6 × 2 = 12	—	17 ✗
[V(CO)₆]⁻	5	5 + 1	6 × 2 = 12	−1 charge	18 ✓

The odd-electron metals — Mn (group 7), Co (group 9) — cannot reach 18 as a mononuclear M(CO)ₓ because every CO adds an even 2 to an odd metal count. Their solution is a metal–metal bond: each M–M bond contributes 1 electron to each metal, converting 17 into 18. So Mn forms Mn₂(CO)₁₀ with a single Mn–Mn bond, and Co forms Co₂(CO)₈. The rule literally predicts dimerization, the bridging-CO structures, and the M–M bond order before you ever run a synthesis.

Vanadium is the famous holdout. V(CO)₆ has 17 electrons and could dimerize — but the six CO ligands are so bulky that a V₂(CO)₁₂ with a V–V bond would be too crowded, so V(CO)₆ remains a rare, isolable, stable 17-electron radical. It is dark green-black, sublimes, and is reduced effortlessly to the 18-electron [V(CO)₆]⁻. Even the exception advertises the rule: the molecule's defining property is that it is one electron short.

Where 16 ⇌ 18 powers catalysis

If 18 electrons were the only allowed count, transition-metal catalysis would be impossible, because a saturated complex has no vacant site to bind a substrate. The working insight is that 16-electron and 18-electron species interconvert, and most homogeneous catalytic cycles oscillate between them. The 16-electron form is coordinatively unsaturated — it has an empty orbital ready to grab a molecule.

The cycle of Wilkinson's catalyst, RhCl(PPh₃)₃, for alkene hydrogenation is the textbook demonstration. Rhodium is group 9; in the +1 state it is d⁸, and the three phosphines plus chloride give a square-planar 16-electron complex. Each catalytic step changes the count by exactly two:

RhCl(PPh3)3            16 e⁻   (Rh(I), d8, square planar)
  − PPh3  (dissociation)
RhCl(PPh3)2            14 e⁻   active species, very unsaturated
  + H2    (oxidative addition)         +2 e⁻, +2 oxidation state
RhCl(H)2(PPh3)2        16 e⁻   (Rh(III), d6)
  + alkene (coordination)              +2 e⁻
RhCl(H)2(alkene)(PPh3)2  18 e⁻ saturated
  migratory insertion → alkyl hydride  (count unchanged, 16 e⁻ after)
  reductive elimination → alkane       −2 e⁻, regenerate Rh(I)
RhCl(PPh3)2            14 e⁻   back to start

Oxidative addition raises the count by 2 and the oxidation state by 2; reductive elimination lowers both by 2. Ligand association adds 2, dissociation removes 2. Counting electrons around the loop tells you which elementary steps are even possible: you cannot do an oxidative addition on an 18-electron complex, so a ligand must leave first. This is why "the resting state is 18, the active state is 16" is the single most useful sentence in organometallic catalysis. The same 16/18 shuttle drives hydroformylation (HCo(CO)₄), the Monsanto acetic-acid process ([Rh(CO)₂I₂]⁻), and palladium cross-coupling.

When the rule bends or breaks

The 18-electron rule is a strong predictor for the middle and late transition metals with strong-field, π-accepting ligands (CO, alkenes, phosphines). It weakens systematically outside that zone, and knowing where it fails is as useful as knowing where it holds.

• Early transition metals are often electron-deficient. Group 3–6 metals simply don't have enough d electrons, and their large radii leave no room for enough ligands. WMe₆ has 12 electrons, TiCl₄ has 8, and [TaF₇]²⁻ has 14. They are perfectly stable because the empty orbitals are high in energy and there is nothing energetically cheap to fill them with.
• Square-planar d⁸ complexes stop at 16. Late metals in the +2/+1 d⁸ configuration — Rh(I), Ir(I), Ni(II), Pd(II), Pt(II), Au(III) — adopt a square-planar geometry in which the p_z orbital perpendicular to the plane is too high to use. With only eight usable orbitals, the closed-shell count is 16, not 18. This is why Wilkinson's catalyst (16), Vaska's complex IrCl(CO)(PPh₃)₂ (16), and cisplatin PtCl₂(NH₃)₂ (16) are stable as drawn.
• Strong-field, low-coordinate, and bulky systems. Steric bulk can cap the ligand count below what 18 requires; conversely, very strong π-donors can make a complex perfectly content at 16.
• Lanthanides and actinides ignore it. The 4f and 5f electrons are core-like and don't bond, so f-block "complexes" are governed by sterics and ionic radius, not an electron count.

A clean rule of thumb: the 18-electron rule is most reliable for octahedral complexes of groups 6–10 with π-acceptor ligands, reliable-with-an-asterisk (16 e⁻) for square-planar d⁸, and merely advisory for early metals and f-block elements.

18-electron rule vs the octet rule

	Octet rule	18-electron rule
Applies to	Main-group (s/p block)	Transition metals (d block)
Valence orbitals	4 (1 s + 3 p)	9 (1 s + 3 p + 5 d)
Closed-shell count	8 electrons	18 electrons
Noble gas mimicked	Ne, Ar (period end)	Kr, Xe, Rn (period end)
Common stable exceptions	BF₃ (6), SF₆ (12), NO (odd)	16 e⁻ square-planar d⁸; early-metal deficiency
Reliability	Very high for C, N, O, F	High for mid/late d-block with π-acceptors
What an over-count costs	Hypervalency / expanded shell	Electrons forced into antibonding orbitals
Predicts	Lewis structures, bonding capacity	Stoichiometry, M–M bonds, catalytic steps

Both rules are the same idea — fill the valence orbitals to a closed, noble-gas shell — counted over a different number of orbitals. And both have honest exceptions: just as boron is happy with a sextet in BF₃, square-planar platinum is happy with 16.

Where the count does real work

• Predicting whether a complex can even be made. Before attempting a synthesis, chemists count electrons. A proposed neutral Mn(CO)₅ (17 e⁻) won't exist as a monomer — it dimerizes to Mn₂(CO)₁₀. A target with a 20-electron count is a red flag that you've drawn too many ligands.
• Reading a catalytic mechanism. Every arrow in a published organometallic cycle is annotated with an electron count. If a step claims oxidative addition to an 18-electron complex, the mechanism is wrong — something must dissociate first to make a 16-electron intermediate.
• Industrial catalysis. Hydroformylation (the "oxo process," ~10 million tonnes of aldehydes per year) runs through HCo(CO)₄ and HCo(CO)₃ — an 18/16 pair. The Monsanto and Cativa acetic-acid processes cycle a rhodium or iridium center between 16 and 18 electrons through CO and methyl migrations.
• Materials and stability. The robustness of ferrocene (18 e⁻) made it a building block for ferrocene-based polymers, redox sensors, and even an anti-knock fuel additive; its 18-electron closed shell is what lets it survive conditions that destroy most organometallics.

Common misconceptions and pitfalls

• Mixing the two counting methods. The most common error. If you count Fe as a neutral group-8 atom (8), you must treat Cp as a neutral 5-electron donor — not as Cp⁻ (6). Pick a method and stay in it for the whole molecule.
• Forgetting the overall charge. A cation has fewer electrons, an anion more. In the covalent method, subtract the charge: [Mn(CO)₆]⁺ is 7 + 12 − 1 = 18, not 19. In the ionic method the charge is already baked into the oxidation state.
• Treating it as a law instead of a guide. Stable 16-electron square-planar complexes and electron-deficient early metals are real and abundant. The rule is a powerful heuristic, strongest for π-acceptor mid/late d-block complexes — not an inviolable law like charge balance.
• Miscounting hapticity. An η⁵-Cp donates 5 (neutral) but an η¹-Cp (σ-bonded) donates only 1. The same ligand can change its electron count by ring-slipping, which is exactly how an 18-electron complex opens a site without losing a ligand entirely.
• Assuming bridging ligands count the same. A terminal CO donates 2 to one metal; a μ₂-bridging CO donates 1 to each of two metals. Get the bridging mode wrong and the whole cluster count collapses.
• Confusing oxidation state with d-electron count. Oxidation state is a formalism for the metal's charge; the d-electron count is group number minus oxidation state. Only the d-electron count goes into the 18-electron sum, and only in the ionic method.