advanced39 min studyLesson 139 in Physics

Hamiltonian perturbation and resonance

Average weak anharmonicity, recognize small denominators, and distinguish coherent addition from resonant energy transfer.

On this page

What you’ll understand

  • Compute a first-order averaged frequency shift.
  • Identify the denominator that makes a Fourier perturbation resonant.
  • State the timescale and model limitations of an averaged prediction.

A weak correction to an integrable Hamiltonian often produces a weak correction to its motion, but that statement can fail near resonance or over a long enough time. Action-angle coordinates reveal why. Most angle-dependent terms oscillate and average away at leading order. A resonant term evolves slowly instead, so its effects accumulate. The smallness of a coefficient must therefore be judged together with frequency relations, action scales, and the interval over which a prediction is needed.

Separate a solvable part from a controlled correction

Write H as an integrable part depending only on actions plus a small angle-dependent correction. The unperturbed frequencies are derivatives of the integrable Hamiltonian. Expand a periodic correction in Fourier modes of the angles. A nonresonant oscillatory mode can often be removed to first order by a near-identity canonical transformation, leaving its angle average as the leading change in the effective Hamiltonian. The transformation must remain small throughout the domain where it is used.

A numerical coefficient called ε is not by itself a sufficient smallness test. For a quartic correction to a harmonic oscillator, its energy relative to the quadratic energy grows with amplitude. The same ε can be negligible for one orbit and dominant for another. Always identify the dimensionless combination involving the action or amplitude that controls the approximation. An expansion that is excellent for energy can also accumulate a large phase error after many cycles if its frequency correction is ignored.

H(J,θ)=H0(J)+ϵH1(J,θ),H‾=H0+ϵ⟨H1⟩θ+O(ϵ2)H(J,\theta)=H_0(J)+\epsilon H_1(J,\theta),\qquad \overline H=H_0+\epsilon\langle H_1\rangle_\theta+O(\epsilon^2)
This first-order averaged form assumes a valid nonresonant near-identity construction and a perturbative domain. Error control over time requires additional estimates.↔ Scroll wide equations horizontally.

A quartic oscillator acquires an amplitude-dependent frequency

Use normalized mass and harmonic angular frequency equal to one, with a positive weak quartic term εq⁴/4. The unperturbed substitution q equal to the square root of twice J times sine of θ converts that term into εJ² times sine to the fourth power. Its angle average is three eighths of εJ². Differentiating the averaged Hamiltonian with respect to J predicts a frequency that increases with action for a positive quartic correction.

This result describes the leading anharmonic frequency shift, not an exact sinusoidal solution. The true orbit contains waveform corrections as well. In the averaged Hamiltonian, J denotes the canonical action after the near-identity averaging and agrees with harmonic action at zeroth order. Substituting zeroth-order harmonic action into the leading frequency correction changes only higher-order terms, but the linear energy term requires the first-order action conversion for first-order energy accuracy. Using initial amplitude A in the frequency shift therefore permits J approximately A squared over two. The expansion requires εA squared to be small.

H=p2+q22+ϵq44,H‾=J+3ϵJ28,ωeff≈1+3ϵJ4H=\frac{p^2+q^2}{2}+\frac{\epsilon q^4}{4},\qquad \overline H=J+\frac{3\epsilon J^2}{8},\qquad \omega_{\mathrm{eff}}\approx1+\frac{3\epsilon J}{4}
The averaged normal form keeps first-order terms and uses ⟨sin⁴θ⟩ = 3/8. J denotes the averaged canonical action, coinciding with harmonic action at zeroth order; the domain requires εJ small.↔ Scroll wide equations horizontally.

Small denominators identify resonance

A Fourier term with angle combination k dotted into θ oscillates at k dotted into the unperturbed frequency vector. Removing it perturbatively involves division by that combination. Near zero, the proposed transformation is no longer small even if ε is small. With periodic external forcing, the relevant combinations also include integer multiples of the drive frequency. The correct response is to retain the slowly varying resonant angle and derive a reduced resonant Hamiltonian rather than force a divergent nonresonant formula.

Resonance can exchange actions between modes, create islands, or change long-term stability, depending on the Hamiltonian. It is not identical to constructive interference of two prescribed waves. Equal-frequency waves of fixed amplitude add coherently but do not by themselves model energy supplied by a drive or transferred between dynamical oscillators. A lesson diagram should expose that distinction: phase alignment illustrates a slowly varying relative phase, while actual resonance dynamics needs coupled evolution equations and an energy-accounting model.

k⋅ω≈0ork⋅ω−ℓΩ≈0\mathbf k\cdot\boldsymbol\omega\approx0\quad\text{or}\quad \mathbf k\cdot\boldsymbol\omega-\ell\Omega\approx0
Here k and ℓ are integers, ω are internal angular frequencies, and Ω is an external drive frequency. Near resonance the corresponding slow term must be treated separately.↔ Scroll wide equations horizontally.

Estimate a weak quartic frequency shift

For the normalized quartic oscillator take ε = 0.02 and an averaged canonical action J = 2. Estimate the averaged energy and angular frequency, keeping only first-order terms.

  1. The controlling combination is ϵJ=0.04\epsilon J=0.04, small compared with one. The averaged quartic energy correction is 3ϵJ2/8=3(0.02)(4)/8=0.033\epsilon J^2/8=3(0.02)(4)/8=0.03.

  2. The averaged Hamiltonian value is therefore H‾≈2+0.03=2.03\overline H\approx2+0.03=2.03. This is the first-order averaged value, not a claim that the instantaneous quartic energy is independent of angle in the unperturbed variables.

  3. Differentiate before substituting: ωeff≈1+3ϵJ/4=1.03\omega_{\mathrm{eff}}\approx1+3\epsilon J/4=1.03. Over ten unperturbed time units the leading extra phase is about 0.3 radians, illustrating how a small frequency correction accumulates.

The result

The first-order averaged energy is 2.03 and angular frequency is 1.03. The positive quartic term stiffens this oscillator at larger action.

Explore the model

Interactive lab

Interactive lab · waves

Inspect phase coherence without inventing a resonance solver

Set amplitude ratio to 1. Predict that phase 0° doubles the sinusoidal amplitude and phase 180° cancels it, then compare the sum during playback. The two components have the same fixed frequency, so their relative phase stays fixed. This lab illustrates coherent addition only: it has phase and amplitude-ratio controls, no frequency detuning, no dynamical energy exchange, and no secularly growing driven oscillator.

Inspect phase coherence without inventing a resonance solverTwo waves share the same frequency and wave number. Playback advances t in sin(2x − t) + a sin(2x − t + phase). Current readings: Sum at x = 0: 0.866; Resultant amplitude: 1.732; Phase difference: 60°. All parameters are adjustable with the labeled controls below.−3.14−1.5701.573.14−3.4−1.701.73.4x · radiansAmplitude
SumFirst componentSecond component
Sum at x = 0
0.866
Resultant amplitude
1.732
Phase difference
60°
t = 0 s

Two waves share the same frequency and wave number. Playback advances t in sin(2x − t) + a sin(2x − t + phase). Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forA slowly varying phase is central to resonance analysis, but adding prescribed waves is not itself resonance dynamics.

Check your understanding

Why can a Fourier term invalidate a nonresonant perturbation expansion despite a small ε?

Try it yourself

For the same normalized oscillator, express the leading angular-frequency shift in terms of the unperturbed amplitude A, then estimate it for ε = 0.01 and A = 2.

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

Use J ≈ A²/2 only to the perturbative order needed.

Compare with the solution

Substitution gives ωeff≈1+3ϵA2/8\omega_{\mathrm{eff}}\approx1+3\epsilon A^2/8. With the stated values, the shift is 3(0.01)(4)/8=0.0153(0.01)(4)/8=0.015, so the estimate is 1.015. The controlling amplitude combination is εA² = 0.04. This leading result does not include higher harmonics or an all-time phase-error guarantee; those require a more detailed perturbation calculation.

Take it with you

  • Perturbative smallness depends on orbit scales as well as a coefficient.
  • Angle averaging predicts leading frequency shifts under stated assumptions.
  • Near resonance, small denominators require retaining slow dynamical terms.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.