Fermi Resonance: When Two Vibrations Trade Identities

The accidental degeneracy that broke the harmonic picture

Start with the harmonic oscillator, the model that underwrites every introductory vibrational spectrum. Each normal mode is an independent ladder of evenly spaced rungs, and the only allowed transitions are Δv = ±1 in one mode at a time. Overtones (Δv = 2) and combination bands are strictly forbidden and carry zero intensity. Real molecules violate this in two distinct ways. The diagonal violation is ordinary anharmonicity: the potential is not a perfect parabola, so ladder rungs crowd together as you climb, and overtones become weakly allowed. The off-diagonal violation — the one Fermi identified — is a coupling that mixes two different vibrational states when they happen to lie at nearly the same energy.

Fermi resonance is precisely that off-diagonal, near-degenerate mixing. It requires two things to coincide. First, an accidental energy match: a fundamental of one mode must fall close to an overtone (2νᵢ) or a combination band (νᵢ + νⱼ) of others. "Accidental" is the operative word — nothing in the molecule's symmetry forces the coincidence; it is a numerical fluke of the force field. Second, a symmetry match: the two states must transform as the same irreducible representation of the molecular point group, because only then can a term in the Hamiltonian connect them.

When both conditions hold, the two zero-order states are no longer good eigenstates. Quantum mechanics mixes them into two new states that are neither pure fundamental nor pure overtone. The energies repel — the upper level goes higher, the lower goes lower — and, crucially, intensity redistributes. This is why Fermi resonance is often described as two vibrations trading identities: after mixing, you cannot point to one measured band and call it "the stretch" and the other "the overtone." Each is a blend.

The two-level secular equation: repulsion and intensity borrowing

The mathematics is the same 2×2 problem that governs bonding/antibonding MO formation and avoided crossings everywhere in chemistry — which is exactly why it is so intuitive once you have seen it once. Let the two zero-order (unperturbed) states be |a⟩, the fundamental, at energy Eₐ, and |b⟩, the overtone/combination, at energy E_b. The perturbation Hamiltonian has an off-diagonal matrix element W = ⟨a|Ĥ′|b⟩. Diagonalizing the 2×2 block

| Eₐ   W |
| W   E_b |

gives the observed levels

E± = (Eₐ + E_b)/2 ± ½·√(δ² + 4W²),   where δ = Eₐ − E_b.

So the splitting between the two bands you actually measure is ΔE = √(δ² + 4W²), which is always larger than the bare separation |δ|. The states repel. In the perfectly resonant limit δ = 0 (exact accidental degeneracy) the splitting reaches its minimum value of 2W, the eigenstates become the symmetric and antisymmetric combinations (|a⟩ ± |b⟩)/√2, and the mixing is 50/50 — maximal identity exchange.

The intensity story follows directly. The overtone |b⟩ has essentially no transition dipole of its own; the fundamental |a⟩ carries all the oscillator strength. After mixing, each eigenstate contains a share of |a⟩, so each carries a share of the intensity. Write the mixing angle θ with tan 2θ = 2W/δ; the upper and lower bands then carry intensities proportional to cos²θ and sin²θ. At δ = 0, θ = 45° and the intensity splits evenly — a formerly invisible overtone now stands as tall as a fundamental. This is intensity borrowing, and it is the experimental fingerprint that distinguishes Fermi resonance from a coincidental pile-up of two independent bands.

The CO₂ dyad: Fermi's original worked example

Carbon dioxide (point group D∞h, linear, 3N−5 = 4 vibrational modes) is the textbook case and the one Fermi solved. Its modes are the symmetric stretch ν₁ (σg⁺), the doubly degenerate bend ν₂ (πu) near 667 cm⁻¹, and the antisymmetric stretch ν₃ (σu⁺) near 2349 cm⁻¹. The bare symmetric stretch is expected around 1330–1340 cm⁻¹. Its first overtone 2ν₂ lands at roughly 2 × 667 ≈ 1334 cm⁻¹ — an almost exact accidental degeneracy with ν₁.

Now the symmetry test. The overtone of a πu mode contains the symmetric product [πu × πu] = σg⁺ + δg. The σg⁺ component of 2ν₂ transforms identically to ν₁ (also σg⁺). Same irreducible representation, near-identical energy: the two states mix strongly. The result is the famous Fermi dyad — two intense Raman lines at approximately 1388 and 1285 cm⁻¹ (historically reported by Fermi as ~1388.3 and ~1285.5 cm⁻¹), symmetrically flanking the ~1337 cm⁻¹ position where the unperturbed states would have coincided. The δg component of 2ν₂, having no σg⁺ partner to mix with, stays put and weak.

Two consequences seal the interpretation. First, both dyad members are strong in the Raman (σg⁺ modes are Raman-active in D∞h, and the overtone has borrowed stretch character), and both are IR-silent — consistent with a mutual-exclusion, centrosymmetric molecule. Second, the splitting is large (~103 cm⁻¹ from the endpoints above; ~103–107 cm⁻¹ across sources) precisely because δ ≈ 0 puts the system at maximal resonance; the coupling matrix element extracted from the dyad is on the order of W ≈ 50 cm⁻¹. Fermi resonance in CO₂ is not a curiosity — it dominates the molecule's Raman signature and is a workhorse thermometry probe in combustion diagnostics via coherent anti-Stokes Raman scattering (CARS).

Where the coupling comes from: cubic anharmonicity

What is W physically? It is the matrix element of the anharmonic part of the potential energy, the terms beyond the harmonic parabola. Expand the potential in the normal coordinates qᵢ: V = ½Σωᵢqᵢ² + Σ kᵢⱼₖ qᵢqⱼqₖ + …. The leading anharmonic correction is cubic. For a Fermi resonance between fundamental ν₁ and overtone 2ν₂, the responsible term is k₁₂₂ q₁q₂². Its matrix element between |v₁=1, v₂=0⟩ and |v₁=0, v₂=2⟩ is nonzero (harmonic-oscillator ladder operators connect states differing by Δv₁ = 1 and Δv₂ = 2), giving W ∝ k₁₂₂ with a numerical prefactor from the oscillator wavefunctions.

This immediately explains the symmetry rule from group theory. The coupling term must be totally symmetric for it to appear in the Hamiltonian, so the product of the two states' symmetries with the operator must contain the totally symmetric representation. Equivalently, the two coupled states must share an irreducible representation. It also explains why Fermi resonances are selective: only mode pairs with a large cubic coupling constant and a small energy gap resonate strongly. A pair can be nearly degenerate yet barely interact if the relevant kᵢⱼₖ is tiny, or interact modestly across a larger gap if the coupling is strong.

There is a hierarchy worth naming. When a fundamental couples to a combination band via a cubic term k₁₂₃ q₁q₂q₃ you get the same 2×2 physics with a different coupling constant. When three or more zero-order states cluster together — common in polyatomics — the 2×2 secular equation generalizes to an n×n matrix (a Fermi polyad), and the observed bands are eigenvectors spread across several basis states. The higher CO₂ polyads built on the antisymmetric stretch — the (10⁰1, 02⁰1, …) manifold that layers ν₃ onto the resonant ν₁/2ν₂ pair — are a standard example, and unraveling such polyads is central to the effective-Hamiltonian analyses that produce spectroscopic databases like HITRAN.

Everyday sightings: aldehydes, benzene, and the tells that give it away

Fermi resonance is not exotic. The most-taught organic example is the aldehyde C–H stretch doublet. The aldehydic ν(C–H) fundamental falls near ~2800 cm⁻¹, and the first overtone of the C–H in-plane bend (2 × δ, with δ ≈ 1390 cm⁻¹) lands nearby at ~2780 cm⁻¹. The two mix and split into the diagnostic pair at roughly 2830 and 2730 cm⁻¹ — a reliable IR signature that flags an aldehyde and is used by practicing spectroscopists precisely because it is so distinctive. Remove the resonance partner and the doublet collapses.

Other classic cases:

• Benzene ~3070 / ~3050 cm⁻¹ region — the aromatic C–H stretch mixes with overtones/combinations of ring modes, producing the structured group of bands rather than a single line.
• Ketone and ester carbonyls — a ν(C=O) fundamental can borrow from a nearby combination band, broadening or doubling the carbonyl feature and complicating quantitative assignment.
• Water and ice — the ν₁/2ν₂ interplay of H₂O contributes to the complex O–H stretch envelope; the Raman spectrum of liquid water shows structure that anharmonic coupling helps explain.
• Carbon tetrachloride (CCl₄) — the symmetric stretch ν₁ near 459 cm⁻¹ resonates with the ν₂ + ν₄ combination band, splitting the strong Raman line into the classic ~459/~449 cm⁻¹ doublet in this tetrahedral molecule.

How do you diagnose it experimentally rather than assume it? The tells are unambiguous: (1) an unexpectedly strong band where only a weak overtone was predicted; (2) a splitting larger than the bare frequency gap; (3) an isotope test — substituting ²H, ¹³C, or ¹⁸O shifts the two zero-order states by different amounts, detuning δ. As δ grows, the resonance switches off: the split shrinks toward |δ|, and the borrowed intensity drains back into the fundamental. That differential response to isotopic or solvent perturbation is the cleanest proof that you are looking at mixed states, not two coincidental fundamentals.

Limits, subtleties, and the modern computational view

Fermi resonance is a near-degenerate perturbation problem, and that is exactly where naive perturbation theory fails. Second-order vibrational perturbation theory (VPT2), the standard workhorse for computing anharmonic frequencies, contains energy denominators of the form 1/(Eₐ − E_b) — equivalently 1/(ωᵢ − 2ωⱼ) when written with the bending fundamental frequency ωⱼ rather than the overtone energy E_b. When a Fermi resonance drives that denominator toward zero, the correction diverges — VPT2 produces nonsensical, exploding frequencies. The rigorous fix is to identify the resonant pairs, remove those terms from the perturbative sum, and instead diagonalize the coupled block variationally. Modern anharmonic codes (e.g., the GVPT2 / deperturbation-plus-resonance treatment implemented in packages like Gaussian and CFOUR) do exactly this, and getting the resonance list right is often the difference between a useful and a useless computed spectrum.

Several subtleties deserve care. Detuning matters more than coupling near resonance. Because ΔE = √(δ² + 4W²) is flat in δ around δ = 0, a strongly resonant pair is remarkably insensitive to small changes; but push δ past ~2W and the system rapidly reverts to "normal." This nonlinearity is why temperature, pressure, hydrogen bonding, and phase changes can switch Fermi resonances on and off — a solvent that shifts a fundamental by 20 cm⁻¹ can visibly redistribute intensity between the dyad members. Fermi resonance is distinct from Darling–Dennison resonance, which couples states differing by two quanta in each of two modes (a quartic k₁₁₂₂ term, Δv = ±2/∓2) and is prominent among high stretching overtones of H₂O and CH-containing molecules. Both are anharmonic couplings, but of different order and selection rules.

Historically, Fermi's 1931 paper sat inside the very first years of quantum spectroscopy — Raman's effect was reported in 1928, and Fermi supplied the theoretical machinery to interpret CO₂ almost immediately. The idea proved far larger than the molecule that inspired it: the same 2×2 near-degenerate mixing recurs as vibronic coupling, avoided crossings on potential surfaces, and coupled-mode effects throughout molecular physics. Whenever two states of matching symmetry drift into near-degeneracy and a term in the Hamiltonian links them, they repel and mix — Fermi resonance is simply the vibrational face of that universal quantum rule.

Fermi resonance versus ordinary anharmonicity — two ways vibrational levels shift, and why only one scrambles state identity

Feature	Ordinary anharmonicity	Fermi resonance
Energy match required	No — always present	Yes — near-degeneracy of a fundamental with an overtone/combination
Symmetry condition	None	Interacting states must belong to the same irreducible representation
Effect on energies	Uniformly lowers overtone spacings (−2xₑ per quantum)	Extra mutual repulsion: levels split apart beyond the diagonal gap
Effect on intensity	Overtones stay weak (10⁻²–10⁻³ of fundamental)	Weak overtone borrows intensity; two bands of comparable strength appear
Eigenstate character	Nearly pure (one quantum number)	Mixed — each observed band is a superposition of both zero-order states
Sensitivity to isotope/solvent	Weak	Strong — small frequency shifts detune the resonance dramatically