Technopedia

The technologist’s encyclopedia

Coupled oscillations

Coupled oscillations describe two or more oscillators that interact, so that the motion of one affects the other. The system’s behavior decomposes into normal modes: independent patterns of motion, each oscillating at its own frequency. This is relevant to molecular vibrations, phonons in crystals, coupled pendulums, and LC filter networks — anywhere energy sloshes between degrees of freedom.

Authoritative sources

Two coupled oscillators

The canonical setup: two masses m1m_1 and m2m_2 attached to fixed walls by springs of stiffness k1k_1 and k2k_2, and coupled to each other by a spring of stiffness kck_c:

m₁ m₂ k kc k x₁ x₂

Newton’s second law on each mass gives two coupled equations, where each x¨i\ddot{x}_i depends on both displacements through the coupling term kcxjk_c x_j:

m1x¨1+(k1+kc)x1−kcx2=0m_1\ddot{x}_1 + (k_1 + k_c)x_1 - k_c x_2 = 0 m2x¨2+(k2+kc)x2−kcx1=0m_2\ddot{x}_2 + (k_2 + k_c)x_2 - k_c x_1 = 0
Details

Matrix form. With x=(x1,x2)T\mathbf{x} = (x_1, x_2)^T, the equations become Mx¨+Kx=0\mathbf{M}\ddot{\mathbf{x}} + \mathbf{K}\mathbf{x} = 0, where M\mathbf{M} is the diagonal mass matrix and K\mathbf{K} the stiffness matrix. Trying a solution x(t)=e eiωt\mathbf{x}(t) = \mathbf{e}\,e^{i\omega t} gives the eigenvalue problem M−1K e=ω2e\mathbf{M}^{-1}\mathbf{K}\,\mathbf{e} = \omega^2\mathbf{e}: the squares of the normal-mode frequencies are the eigenvalues of M−1K\mathbf{M}^{-1}\mathbf{K} (equivalently, the frequencies are the square roots of those eigenvalues), and the eigenvectors give the mode patterns. For NN coupled oscillators, M−1K\mathbf{M}^{-1}\mathbf{K} is N×NN \times N and there are exactly NN normal modes.

Symmetric case. For m1=m2=mm_1 = m_2 = m and k1=k2=kk_1 = k_2 = k, the eigenvectors are (1,1)(1, 1) and (1,−1)(1, -1), giving the two normal frequencies:

ω+=kmω−=k+2kcm\omega_+ = \sqrt{\frac{k}{m}} \qquad \omega_- = \sqrt{\frac{k + 2k_c}{m}}

The in-phase mode (x1=x2x_1 = x_2, eigenvector (1,1)(1,1)): the coupling spring is never stretched, so it exerts no force — each mass oscillates as if the other were absent, at the bare frequency k/m\sqrt{k/m}. The out-of-phase mode (x1=−x2x_1 = -x_2, eigenvector (1,−1)(1,-1)): the coupling spring is stretched twice as much as either outer spring per unit displacement, adding 2kc2k_c to the effective stiffness, so the frequency rises to (k+2kc)/m\sqrt{(k+2k_c)/m}.

Example problem

Take m1=2m_1 = 2 kg, m2=1m_2 = 1 kg, k1=k2=0k_1 = k_2 = 0 (masses connected only to each other by kc=10k_c = 10 N/m, free to translate). Show that one normal mode has ω=0\omega = 0. Solution: the stiffness matrix is (kc−kc−kckc)\begin{pmatrix} k_c & -k_c \\ -k_c & k_c \end{pmatrix}; the vector (1,1)(1,1) is an eigenvector with eigenvalue 0 — the whole pair translates freely, with no restoring force. This zero-frequency mode is the discrete analogue of a sound wave’s long-wavelength limit.

Normal modes

A normal mode is a pattern in which all parts oscillate at the same frequency with fixed relative amplitudes. Any motion of the system — however complicated — is a superposition of the normal modes:

xi(t)=∑nAn ei(n)cos⁡(ωnt+ϕn)x_i(t) = \sum_{n} A_n\, e_i^{(n)} \cos(\omega_n t + \phi_n)

where ei(n)e_i^{(n)} is the ii-th component of the nn-th eigenvector (the mode pattern) and An,ϕnA_n, \phi_n are set by the initial conditions.

For the symmetric two-mass system, the two mode patterns are:

In-phase: x₁ = x₂, frequency ω₊ both masses move together: coupling spring unstretched masses move oppositely: coupling spring stretched, higher ω
Details

Why eigenvectors? In a mode, the ratio x1/x2x_1/x_2 stays constant in time — that is the defining property of an eigenvector. Only then can both masses share a single cos⁡(ωnt)\cos(\omega_n t) factor. Any other motion mixes the two modes, so the ratio changes in time.

Orthogonality. Modes with different frequencies are orthogonal with respect to the mass matrix: ∑iMi ei(n)ei(m)=0\sum_i M_i\, e_i^{(n)} e_i^{(m)} = 0 for n≠mn \neq m. This lets you project any initial condition onto the modes independently, exactly like a Fourier decomposition (see Wave propagation).

Physical picture. In the in-phase mode the coupling spring does no work, so the mode cannot exchange energy with anything — it oscillates in isolation. In the out-of-phase mode the coupling spring alternately stores and releases energy, adding stiffness and raising the frequency. Coupling always splits the frequencies apart: degenerate identical oscillators become ω+\omega_+ and ω−>ω+\omega_- \gt \omega_+.

Example problem

A symmetric system with k=25k = 25 N/m, kc=3k_c = 3 N/m, m=1m = 1 kg starts at rest with x1=2x_1 = 2 cm, x2=0x_2 = 0. Decompose into modes. Solution: (2,0)=(1,1)+(1,−1)(2, 0) = (1,1) + (1,-1), so A+=A−=2/2A_+ = A_- = 2/\sqrt{2} cm in normalized eigenvectors — equal parts of each mode. Since the mode frequencies are close (ω+=5\omega_+ = 5 rad/s, ω−=31≈5.57\omega_- = \sqrt{31} \approx 5.57 rad/s), the two mode superpositions drift in and out of phase, producing beats (next section).

Energy exchange and beats

If the excitation is not a pure mode — say mass 1 is displaced by AA and released from rest while mass 2 is held at 0 — the motion is an equal superposition of both modes:

x1(t)=A2[cos⁡(ω+t)+cos⁡(ω−t)]x_1(t) = \frac{A}{2}\bigl[\cos(\omega_+ t) + \cos(\omega_- t)\bigr] x2(t)=A2[cos⁡(ω+t)−cos⁡(ω−t)]x_2(t) = \frac{A}{2}\bigl[\cos(\omega_+ t) - \cos(\omega_- t)\bigr] t = π/(ω₋−ω₊): all energy on m₂ t = 2π/(ω₋−ω₊): back on m₁ x₁(t) envelope ±A cos[(ω₋−ω₊)t/2] t
Details

Sum-to-product form. Applying the identities cos⁡α+cos⁡β=2cos⁡α−β2cos⁡α+β2\cos\alpha + \cos\beta = 2\cos\frac{\alpha-\beta}{2}\cos\frac{\alpha+\beta}{2} and cos⁡α−cos⁡β=−2sin⁡α−β2sin⁡α+β2\cos\alpha - \cos\beta = -2\sin\frac{\alpha-\beta}{2}\sin\frac{\alpha+\beta}{2}:

x1(t)=Acos⁡ ⁣(ω−−ω+2 t)cos⁡ ⁣(ω++ω−2 t)x_1(t) = A\cos\!\Bigl(\frac{\omega_- - \omega_+}{2}\,t\Bigr)\cos\!\Bigl(\frac{\omega_+ + \omega_-}{2}\,t\Bigr) x2(t)=Asin⁡ ⁣(ω−−ω+2 t)sin⁡ ⁣(ω++ω−2 t)x_2(t) = A\sin\!\Bigl(\frac{\omega_- - \omega_+}{2}\,t\Bigr)\sin\!\Bigl(\frac{\omega_+ + \omega_-}{2}\,t\Bigr)

Each mass oscillates rapidly at the average frequency (ω++ω−)/2(\omega_+ + \omega_-)/2, under a slowly varying envelope of angular frequency (ω−−ω+)/2(\omega_- - \omega_+)/2. The two envelopes are in quadrature: when ∣cos⁡∣|\cos| is maximal for mass 1, ∣sin⁡∣|\sin| is zero for mass 2.

Beat frequency. The envelope amplitude repeats at the beat frequency

fbeat=∣ω−−ω+∣2πf_{beat} = \frac{|\omega_- - \omega_+|}{2\pi}

— the same formula as the beats heard between two detuned tuning forks, and the same interference mathematics as in Wave propagation.

Energy transfer period. The energy on each mass is proportional to the envelope squared, cos⁡2 ⁣(ω−−ω+2t)\cos^2\!\bigl(\frac{\omega_- - \omega_+}{2}t\bigr) and sin⁡2 ⁣(ω−−ω+2t)\sin^2\!\bigl(\frac{\omega_- - \omega_+}{2}t\bigr). Squaring doubles the angular frequency, so the energy sloshes completely from mass 1 to mass 2 and back with period

Ttransfer=2πω−−ω+=1fbeatT_{transfer} = \frac{2\pi}{\omega_- - \omega_+} = \frac{1}{f_{beat}}

At t=π/(ω−−ω+)t = \pi/(\omega_- - \omega_+) (half period), mass 1 is momentarily at rest with zero energy and mass 2 carries all of it; at t=2π/(ω−−ω+)t = 2\pi/(\omega_- - \omega_+) the energy is back on mass 1. Total energy is conserved — the coupling spring is the pipe through which it flows.

Example problem

Two identical pendulums (m=0.5m = 0.5 kg each, L=1L = 1 m, so ω0=g/L≈3.13\omega_0 = \sqrt{g/L} \approx 3.13 rad/s) are coupled by a weak spring that raises the antisymmetric frequency to ω−=3.33\omega_- = 3.33 rad/s. One pendulum is released from rest at 55 cm. How long until the other pendulum swings at full amplitude? Solution: Ttransfer=2π/(3.33−3.13)=2π/0.20≈31T_{transfer} = 2\pi/(3.33 - 3.13) = 2\pi/0.20 \approx 31 s. The beat frequency is fbeat=0.20/(2π)≈0.032f_{beat} = 0.20/(2\pi) \approx 0.032 Hz — slow energy exchange is the signature of weak coupling.

Extension to N oscillators and waves

Couple NN identical masses (spacing aa) with springs kk to their neighbors, plus a weak coupling kck_c (with ϵ=kc/k\epsilon = k_c/k), and the NN normal frequencies form a band indexed by a wavenumber kn=nπ/(Na)k_n = n\pi/(N a):

ω(k)=ω01+2ϵ(1−cos⁡(ka)),ω0=km\omega(k) = \omega_0\sqrt{1 + 2\epsilon\bigl(1 - \cos(ka)\bigr)}, \qquad \omega_0 = \sqrt{\frac{k}{m}} chain of N coupled masses k = 0 k = π/a ω₀ ωₘₐₓ ka ω(k) band: N discrete modes fill this curve N discrete modes fill this curve

In the limit N→∞N \to \infty, the discrete modes fill the curve densely and the system is described by the wave equation — a continuous medium instead of a chain of masses.

Details

The bridge from particles to fields. The ladder is:

Where the same band structure appears. Atoms in a crystal (phonon dispersion — the k=0k = 0 acoustic mode is the “in-phase” mode, and the speed of sound is its slope), coupled LC circuits (filter passbands), a guitar string (harmonics are the normal modes), and a drum membrane (2D modes). In every case, the continuum limit of coupled oscillators is a field.

Example problem

For a chain with ϵ=0.01\epsilon = 0.01 (weak coupling) and a=1a = 1 nm, compare ω\omega at k=π/(100a)k = \pi/(100a) to ω0\omega_0. Solution: 1−cos⁡(0.01π)=1−cos⁡(0.0314)≈4.93×10−41 - \cos(0.01\pi) = 1 - \cos(0.0314) \approx 4.93 \times 10^{-4}, so ω≈ω01+2×0.01×4.93×10−4≈ω0(1+4.9×10−6)\omega \approx \omega_0\sqrt{1 + 2 \times 0.01 \times 4.93 \times 10^{-4}} \approx \omega_0(1 + 4.9 \times 10^{-6}). Long-wavelength modes barely feel the coupling — which is why sound in a crystal propagates almost dispersionlessly at low frequency.

Quizz

Problem. Two identical masses (m=1m = 1 kg) are connected to fixed walls by springs of stiffness k=100k = 100 N/m and to each other by a coupling spring kc=50k_c = 50 N/m. Mass 1 is displaced by 10 cm and released from rest, with mass 2 held at its equilibrium position and then released.

  1. Find the two normal-mode frequencies.
  2. Find the beat frequency and the time for the energy to transfer completely to mass 2.
  3. Sketch (or describe) the displacement of each mass at that moment.
Solution

1. Normal-mode frequencies. For the symmetric case:

  • In-phase: ω+=k/m=100/1=10\omega_+ = \sqrt{k/m} = \sqrt{100/1} = 10 rad/s.
  • Out-of-phase: ω−=(k+2kc)/m=(100+100)/1=200≈14.14\omega_- = \sqrt{(k + 2k_c)/m} = \sqrt{(100 + 100)/1} = \sqrt{200} \approx 14.14 rad/s.

2. Beat frequency and transfer time. The initial condition (A,0)(A, 0) is an equal superposition of the two modes, so the envelope beats at

fbeat=∣ω−−ω+∣2π=14.14−102π=4.146.28≈0.66 Hzf_{beat} = \frac{|\omega_- - \omega_+|}{2\pi} = \frac{14.14 - 10}{2\pi} = \frac{4.14}{6.28} \approx 0.66\ \text{Hz}

The energy is proportional to the envelope squared, so it transfers completely from mass 1 to mass 2 in half the beat period:

ttransfer=πω−−ω+=π4.14≈0.76 st_{transfer} = \frac{\pi}{\omega_- - \omega_+} = \frac{\pi}{4.14} \approx 0.76\ \text{s}

and returns to mass 1 after a full period Ttransfer=2π/(ω−−ω+)≈1.52T_{transfer} = 2\pi/(\omega_- - \omega_+) \approx 1.52 s.

3. State at t≈0.76t \approx 0.76 s. Mass 1 is momentarily at rest at its equilibrium position (its envelope cos⁡(ω−−ω+2t)=cos⁡(π/2)=0\cos\bigl(\frac{\omega_- - \omega_+}{2}t\bigr) = \cos(\pi/2) = 0), while mass 2 reaches its maximum displacement of 10 cm — all the energy has crossed the coupling spring. The two masses then play the motion back in reverse, returning the energy to mass 1 at t≈1.52t \approx 1.52 s.