Waves & Oscillations

The Anharmonic Oscillator

When the restoring force stops obeying Hooke's law — F = -kx - βx³

An anharmonic oscillator is a vibrating system whose restoring force deviates from Hooke's law F = -kx, acquiring higher-order terms (e.g. F = -kx - βx³) so that its potential energy is non-parabolic. That single change breaks the perfect regularity of simple harmonic motion: the oscillation frequency becomes amplitude-dependent, the response sprouts harmonics (2ω, 3ω, …) and mixing tones, and an asymmetric well produces thermal expansion. The Duffing equation ẍ + 2γẋ + ω₀²x + βx³ = F₀cos(ωt), introduced by Georg Duffing in 1918, is the canonical model.

  • Restoring forceF = -kx - βx³ (Hooke + cubic)
  • PotentialU = ½kx² + ⅓αx³ + ¼βx⁴ + …
  • Duffing equationẍ + 2γẋ + ω₀²x + βx³ = F₀cos(ωt)
  • Frequency shiftω(A) ≈ ω₀(1 + 3βA²/8ω₀²)
  • Key signatureamplitude-dependent period (not isochronous)
  • Physical origin ofthermal expansion, overtones, distortion

Interactive visualization

Press play, or step through manually. The visualization is yours to drive — try it before reading on.

Open visualization fullscreen ↗

Watch the 60-second explainer

A condensed visual walkthrough — narrated, captioned, under a minute.

Definition

Start with the workhorse of introductory physics — the simple harmonic oscillator. Its restoring force is strictly linear:

F = -kx     (Hooke's law)     U(x) = ½kx²

The potential is a perfect parabola, and the resulting motion is a single clean sinusoid whose angular frequency ω₀ = √(k/m) does not depend on how large the swing is. This special property is called isochronism, and it is a mathematical accident of the parabola.

An anharmonic oscillator is anything for which this is not exactly true. Expand any real potential in a Taylor series about its minimum at x = 0:

U(x) = ½kx² + (1/3)αx³ + (1/4)βx⁴ + …
F(x) = -dU/dx = -kx - αx² - βx³ - …

The terms beyond ½kx² are the anharmonic terms. The cubic (αx³) is an asymmetric correction — it tilts the well so one side is steeper than the other. The quartic (βx⁴) is a symmetric correction — it stiffens (β > 0, "hardening") or softens (β < 0, "softening") the well while keeping it symmetric. Because these terms grow faster than x, they are invisible at small amplitude and dominate at large amplitude.

Why anharmonicity matters

Simple harmonic motion is a lie we tell ourselves at small amplitude. The moment you look closely — or push hard — nature is anharmonic. This one idea unifies a startling range of physics:

  • Thermal expansion. A symmetric parabola cannot expand a solid when heated; the cubic term in the interatomic potential is why bridges, rails and pistons grow with temperature.
  • Molecular spectroscopy. The convergence of vibrational energy levels toward a dissociation limit, and the appearance of overtone bands, are direct fingerprints of the anharmonic Morse potential.
  • Nonlinear optics. Electrons in an anharmonic potential re-radiate at doubled frequency — the basis of second-harmonic generation and the green laser pointer.
  • Audio distortion. Overdriven loudspeaker cones and amplifiers add harmonics and intermodulation tones because their restoring behavior is no longer linear.
  • Chaos. The driven double-well Duffing oscillator is one of the simplest systems to exhibit deterministic chaos and a strange attractor.
  • MEMS and precision clocks. Micro-mechanical resonators become anharmonic at large drive, limiting their frequency stability and inviting the amplitude "jump" phenomenon.

How it works, step by step

1. A parabola makes the period constant. For F = -kx the equation of motion mẍ = -kx has the solution x(t) = A cos(ω₀t + φ) with ω₀ = √(k/m). Notice A cancels out of ω₀ — double the amplitude and the period is unchanged. That is isochronism.

2. A quartic well breaks isochronism. Add βx³ to the force (a symmetric Duffing potential U = ½kx² + ¼βx⁴). Now the effective stiffness felt by the mass depends on where it is. Averaged over a cycle, the frequency shifts with amplitude A. Multiple-scales / averaging perturbation theory gives, to leading order:

ω(A) ≈ ω₀ · ( 1 + (3β A²) / (8 ω₀²) )

For a hardening spring (β > 0) the frequency rises with amplitude; for a softening spring (β < 0, the pendulum case) it falls.

3. A cubic well makes the motion asymmetric. With an αx³ term the mass spends more time on the softer side of the well, so its time-averaged position <x> drifts away from x = 0. Heat the oscillator (increase its energy) and <x> drifts further — the microscopic origin of thermal expansion.

4. The response is no longer a pure sine. Because the restoring force distorts the motion, driving at frequency ω produces output at ω, 2ω, 3ω, … The Fourier spectrum sprouts harmonics whose amplitudes grow with drive strength.

5. Two tones mix. Drive at ω₁ and ω₂ simultaneously and the nonlinearity generates combination frequencies. A quadratic term (αx²) produces sum/difference tones ω₁ ± ω₂; a cubic term (βx³) produces ω₁ ± 2ω₂ and 2ω₁ ± ω₂. These are heterodyne and intermodulation products.

6. The resonance peak bends and jumps. Solve the driven, damped Duffing equation and the amplitude-vs-frequency curve leans over (to higher ω for hardening, lower ω for softening). Over a range of drive frequencies three steady-state amplitudes coexist; as you sweep the drive the response snaps discontinuously between branches — a hysteretic amplitude jump with no counterpart in linear resonance.

The Duffing equation and its resonance

The complete damped, driven anharmonic oscillator is written compactly as the Duffing equation:

ẍ + 2γ ẋ + ω₀² x + β x³ = F₀ cos(ω t)

with symbols and SI units:

SymbolMeaningUnits
xdisplacement from equilibriumm
ẋ, ẍvelocity, accelerationm/s, m/s²
γdamping rate (½ the coefficient of ẋ)s⁻¹
ω₀ = √(k/m)natural (small-amplitude) angular frequencyrad/s
βcubic nonlinearity per unit mass (+ hardening, − softening)m⁻² s⁻²
F₀drive acceleration amplitude (force per unit mass)m/s²
ωdrive angular frequencyrad/s

Set β = 0 and this collapses to the ordinary driven, damped harmonic oscillator. Keep β ≠ 0 and you get the bent backbone curve, the amplitude jump, and — if ω₀² < 0 so the potential has two wells — period-doubling routes to chaos. Duffing published this in his 1918 monograph Erzwungene Schwingungen bei veränderlicher Eigenfrequenz ("Forced oscillations with variable natural frequency").

Harmonic vs. anharmonic at a glance

PropertyHarmonic (SHO)Anharmonic
Restoring forceF = -kx (exactly linear)F = -kx - αx² - βx³ - … (nonlinear)
Potential shapePerfect parabola U = ½kx²Non-parabolic (tilted / stiffened / softened)
Period vs. amplitudeIndependent (isochronous)Amplitude-dependent
WaveformPure sinusoid, single frequencyDistorted — contains 2ω, 3ω, …
Two-tone driveJust ω₁ and ω₂Sum/difference & intermodulation tones
Resonance curveSymmetric Lorentzian peakBent peak, hysteretic jump
Quantum levelsEvenly spaced ℏω(n+½)Converging (Morse): ℏω(n+½) − ℏωxₑ(n+½)²
Thermal expansionExactly zeroNonzero (from asymmetric term)
Long-time behaviorAlways regularCan be chaotic (double well, driven)

Worked example — the pendulum as a softening oscillator

The simple pendulum is the most familiar anharmonic oscillator. Its exact equation of motion is

θ̈ + (g/L) sin θ = 0

Expanding sin θ ≈ θ − θ³/6 shows the restoring "force" gains a negative cubic term — a softening spring. Because of it, the period lengthens with amplitude θ₀ according to the series

T(θ₀) = T₀ · ( 1 + (1/16)θ₀² + (11/3072)θ₀⁴ + … )

where T₀ = 2π√(L/g) is the small-angle period. Plug in numbers for a few amplitudes:

Amplitude θ₀Fractional period increase
5° (0.087 rad)+0.048%
10° (0.175 rad)+0.19%
23° (0.40 rad)+1.0%
45° (0.785 rad)+3.99%
90° (1.571 rad)+18.0%

This is exactly the anharmonic effect Christiaan Huygens confronted in the 1650s–1670s: because a wide-swinging pendulum runs slow, it is not perfectly isochronous, which spoils timekeeping. His cycloidal cheeks were an engineering fix that forced the bob onto a truly isochronous (tautochrone) path — an early acknowledgment that the real pendulum is anharmonic.

Worked example — thermal expansion from an asymmetric well

Model an interatomic bond by U(x) = ½kx² − gx³ (a Morse/Lennard-Jones potential expanded to third order, with g > 0 making the well softer on the stretch side). Using classical statistical mechanics, the thermal average displacement at temperature T is, to leading order,

⟨x⟩ ≈ (3 g / k²) · k_B T

where k_B = 1.381 × 10⁻²³ J/K. The average bond length grows linearly with temperature, and the linear thermal-expansion coefficient α_L = (1/L)(d⟨x⟩/dT) comes out positive and roughly constant — matching observation. Representative values: aluminum 23 × 10⁻⁶ K⁻¹, steel 12 × 10⁻⁶ K⁻¹, fused silica 0.5 × 10⁻⁶ K⁻¹, and the anomalous Invar alloy ≈ 1 × 10⁻⁶ K⁻¹. Crucially, if you drop the −gx³ term, ⟨x⟩ = 0 for all T and the expansion coefficient is exactly zero — a purely harmonic solid never expands.

Quantum anharmonicity and molecular spectra

Quantize a real bond and the evenly spaced ladder of the quantum harmonic oscillator, Eₙ = ℏω(n + ½), gives way to the Morse spectrum:

Eₙ = ℏω(n + ½) − ℏω xₑ (n + ½)²

The anharmonicity constant xₑ is small and positive, so successive levels crowd together and eventually reach a dissociation limit rather than continuing forever. This is why infrared and Raman spectra show overtone bands (n = 0 → 2, 0 → 3) at frequencies slightly below exact integer multiples of the fundamental, and why the fundamental itself is a hair below the harmonic value. For hydrogen H₂ the vibrational constant is ω ≈ 4401 cm⁻¹ and ωxₑ ≈ 121 cm⁻¹; for carbon monoxide CO, ω ≈ 2170 cm⁻¹ and ωxₑ ≈ 13 cm⁻¹.

Common misconceptions

  • "Anharmonic just means nonlinear damping." No — anharmonicity is a nonlinear restoring force (a non-parabolic potential). Nonlinear damping (like drag ∝ ẋ²) is a separate effect. A frictionless anharmonic oscillator is still anharmonic.
  • "Real springs obey Hooke's law until they break." Hooke's law is only the first term of a Taylor expansion; springs deviate continuously, and at large strain a coil stiffens (coil binding) then plastically yields past its elastic limit.
  • "Amplitude-dependent frequency violates conservation of energy." It does not. Energy is perfectly conserved in an undriven, undamped anharmonic oscillator; the period simply depends on the total energy (equivalently, on amplitude).
  • "A symmetric anharmonic term causes thermal expansion." Only the asymmetric (odd-power, e.g. cubic) terms shift ⟨x⟩ and drive expansion. A purely quartic, symmetric well changes the frequency but keeps ⟨x⟩ = 0, so it does not expand.
  • "Harmonics in the output mean the drive contained them." A single-frequency drive of a linear system gives a single-frequency output. Harmonics in the response are generated by the nonlinearity itself — evidence the system is anharmonic.
  • "The Duffing amplitude jump is a resonance like any other." The jump is a genuine bistability with hysteresis: the amplitude you get depends on the sweep direction. Linear resonance has a single, direction-independent Lorentzian peak.

Where anharmonicity shows up

  • Solid-state physics. Phonon–phonon scattering (a purely anharmonic process) gives crystals finite thermal conductivity and causes thermal expansion; harmonic crystals would have infinite conductivity and zero expansion.
  • Nonlinear optics. The anharmonic electronic potential produces second- and third-harmonic generation, sum-frequency mixing, and the optical Kerr effect.
  • Molecular and IR spectroscopy. Morse anharmonicity sets overtone positions and dissociation energies.
  • Acoustics and audio. Total-harmonic-distortion (THD) and intermodulation distortion in speakers, strings and reeds are anharmonic effects.
  • Nonlinear dynamics and chaos. The driven double-well Duffing oscillator is a textbook route to a strange attractor.
  • MEMS/NEMS resonators and clocks. Drive them hard and they exhibit the Duffing bent peak and amplitude jump, limiting stability.
  • Josephson junctions. A current-biased junction behaves as an anharmonic oscillator; its non-uniform level spacing is what makes a superconducting qubit addressable.

Frequently asked questions

What makes an oscillator anharmonic?

An oscillator is anharmonic whenever its restoring force is not exactly proportional to displacement — that is, when it violates Hooke's law F = -kx. Equivalently, its potential energy U(x) is not a pure parabola. Real potentials, when Taylor-expanded about the minimum, read U(x) = ½kx² + (1/3)αx³ + (1/4)βx⁴ + …; the cubic, quartic and higher terms are the anharmonic terms. As long as the amplitude is tiny these terms are negligible and the motion looks simple-harmonic, but at finite amplitude they reshape the motion.

Why does the frequency of an anharmonic oscillator depend on amplitude?

In a pure harmonic oscillator the parabola makes the period isochronous — independent of amplitude — because the restoring force scales exactly with x. Add a quartic term and the effective stiffness changes with how far the mass swings. For a hardening spring (β > 0) the frequency rises with amplitude; for a softening spring (β < 0) it falls. Perturbation theory gives ω(A) ≈ ω₀(1 + 3βA²/8ω₀²) for the Duffing potential. The simple pendulum is a familiar softening case: its period grows by about 1% at a 23° amplitude.

What is the Duffing equation?

The Duffing equation is the canonical damped, driven anharmonic oscillator: ẍ + 2γẋ + ω₀²x + βx³ = F₀cos(ωt). The βx³ term is a cubic (odd) nonlinearity; β > 0 stiffens the spring, β < 0 softens it. Georg Duffing introduced it in 1918. It is famous for a bent resonance peak, a hysteretic amplitude jump as you sweep the drive frequency, and — with a double-well potential — chaotic motion and a strange attractor.

How does anharmonicity cause thermal expansion?

A purely harmonic (symmetric parabolic) interatomic potential would give zero thermal expansion — the time-averaged bond length stays at the minimum no matter how hot it gets. Real interatomic potentials, like the Lennard-Jones or Morse potential, are asymmetric: the repulsive wall is steeper than the attractive tail. As temperature rises and the atom samples higher-energy parts of the well, its average position shifts outward, so the material expands. The cubic term αx³ in U(x) is the direct origin of thermal expansion; expansion coefficients are typically 10⁻⁵ to 10⁻⁶ per kelvin for solids.

What are harmonics and frequency mixing in an anharmonic system?

Drive an anharmonic oscillator at a single frequency ω and the nonlinear restoring force distorts the response, so the output contains 2ω, 3ω and higher harmonics. Drive it at two frequencies ω₁ and ω₂ and a quadratic nonlinearity produces sum and difference tones ω₁ ± ω₂, while a cubic term produces ω₁ ± 2ω₂ and 2ω₁ ± ω₂. This is exactly how nonlinear optics generates second-harmonic (frequency-doubled) light and how intermodulation distortion appears in overdriven amplifiers and loudspeakers.

Are real springs anharmonic?

Yes. Hooke's law is only the leading-order term of a Taylor expansion, so every real spring becomes nonlinear once you stretch it far enough. A coil spring stiffens as its coils bind and eventually plastically deforms past its elastic limit; rubber bands and biological tissue stiffen dramatically at large strain (strain-hardening). No physical restoring force is perfectly linear over all displacements — linearity is an excellent small-amplitude approximation, not an exact law.

How does anharmonicity change a molecule's vibrational energy levels?

A harmonic oscillator has equally spaced quantum levels Eₙ = ℏω(n + ½). Real molecular bonds follow the Morse potential, which crowds the levels together as energy rises: Eₙ = ℏω(n + ½) − ℏωxₑ(n + ½)², where xₑ is the anharmonicity constant. This convergence is why overtone bands in infrared spectra sit slightly below integer multiples of the fundamental, and why the levels pile up to a dissociation limit instead of continuing forever. For H₂, ω ≈ 4401 cm⁻¹ and ωxₑ ≈ 121 cm⁻¹.