Coupled Oscillators

In February 1665, Christiaan Huygens lay ill in bed, watching two of his own pendulum clocks hanging from a common wooden beam. Within half an hour, no matter how he restarted them, the two pendulums fell into perfect anti-phase step — swinging toward each other and away from each other in exact synchrony. He called it une sympathie merveilleuse, an odd sympathy, and correctly guessed the cause: imperceptible vibrations traveling through the beam. Each clock was nudging the other. The two oscillators were not independent; they were coupled. Coupling is the single most consequential thing you can do to a pair of oscillators. Connect two identical pendulums by even the feeblest spring and neither pendulum owns its motion any more: strike one and its energy drains, swing by swing, into its neighbour — completely — then returns, in an endless slow exchange whose period is set by the splitting between the system's two new special frequencies. Those two frequencies belong to the two motions that survive coupling unchanged: the symmetric swing, which never stretches the spring, and the antisymmetric swing, which works it hardest. One frequency has become two — the classical seed of every avoided crossing, every bonding/antibonding pair of molecular levels, every two-qubit energy swap. This article works the two-body problem completely and concretely: the coupled equations of motion, the sum-and-difference trick that unties them, the energy slosh and its period, and the sharp condition — identical frequencies — for the transfer to be complete. Master this 2×2 case and the general machinery of Normal Modes becomes a story you already know.

Key takeaways

Two oscillators joined by any interaction have linked equations of motion: the force on each depends on the other's position, so neither moves independently.
For two identical oscillators the coupled equations untie by pure symmetry: the sum $s = x_1 + x_2$ and difference $d = x_1 - x_2$ each obey a simple harmonic equation, with frequencies $\omega_s = \sqrt{k/m}$ (coupling spring never stretches) and $\omega_a = \sqrt{(k+2\kappa)/m}$ (coupling spring fully engaged).
Coupling splits the single frequency into two: for weak coupling the splitting is $\Delta\omega \approx \kappa/m\omega_0$ — proportional to the coupling strength.
Strike one oscillator and both modes are excited equally; their slow drift out of step passes the energy completely to the neighbour in time $\pi/\Delta\omega$ and returns it at $T_{\text{ex}} = 2\pi/\Delta\omega$ . The slosh is two-frequency beating, seen from the energy side.
Complete transfer requires identical oscillators. Detune the pair and the peak fraction of energy exchanged collapses to $F = g^2/(g^2+\Delta^2)$ — a 1% mismatch can block almost all of it.
The same 2×2 physics underlies Huygens' sympathetic clocks, coupled LC circuits, wireless power transfer, and the two-qubit energy swap.
Key idea: coupling two identical oscillators splits their common frequency into a symmetric and an antisymmetric mode, and the small splitting between these two pure tones sets the slow, complete, back-and-forth exchange of energy — which detuning destroys.

The core insight

Coupling two identical oscillators splits their common frequency into two — a symmetric mode the coupler cannot feel and a stiffer antisymmetric mode — and everything the pair does, including the complete slow sloshing of energy from one to the other, is the superposition of these two pure tones drifting in and out of step.

Where this lives today

The 2×2 coupled oscillator is now a workhorse of quantum engineering. In superconducting quantum processors, two transmon qubits coupled through a bus resonator are literally this article's system: tune the qubits into resonance with each other and a single excitation sloshes back and forth — a quantum energy exchange completing in tens of nanoseconds — and stopping the slosh exactly at the halfway point ( $t = \pi/\Delta\omega$ , the complete-handoff time) implements the iSWAP two-qubit gate. Detune the qubits and equation (6) turns the exchange off, which is precisely how such gates are switched: the Lorentzian gate $F = g^2/(g^2+\Delta^2)$ is an operating principle, run billions of times per second in 2026-era processors. The avoided crossing of equation (5) is what a spectroscopist actually sees when calibrating the chip. The same physics powers resonant wireless charging — a transmitter coil and a receiver coil are two tuned LC oscillators whose magnetic coupling plays the role of $\kappa$ , transferring power efficiently only when both are tuned to the same frequency — and explains why a piano's strings sing back only their own sung note, why turbine blades must never share a natural frequency with their neighbours, and why molecular bonds form: two identical atomic levels, coupled, split into bonding and antibonding pairs, with the electron settling into the classical system's "symmetric mode."

The physical picture

Hang two identical pendulums side by side and join them with a weak spring. There are exactly two ways to start them so that the motion stays simple forever. Pull both aside the same way and release: they swing in unison, the spring between them keeping a constant length, never stretched, never pushing — as far as the dynamics is concerned the spring might as well not exist, and each pendulum swings at its bare frequency. Pull them apart in opposite directions and release: now the spring is stretched and compressed as violently as the geometry allows, adding its restoring force to gravity's, and the pair swings against each other at a distinctly higher frequency. These two special motions never change their shape; each is a pure tone.

Figure

The two pure tones of a coupled pair — the symmetric swing ignores the spring, the antisymmetric swing works it hardest

Now do the interesting experiment: hold pendulum 2 still, pull pendulum 1 aside, and release both. This start is neither pure tone — it is an exact 50/50 mixture of the two. At first the two modes agree on pendulum 1 and cancel on pendulum 2, so pendulum 1 swings alone. But the antisymmetric mode runs slightly fast, and cycle after cycle the two tones drift out of step. Wherever they cancel, a pendulum stands still; wherever they agree, it swings hard. The result is the signature phenomenon of weak coupling: the first pendulum's swing dies away to nothing while the second, untouched by any hand, swings up to the full original amplitude — and then the exchange runs in reverse.

Figure

Strike one pendulum and watch the energy slosh — complete transfer at half the exchange period, complete return at the full one

Read the two traces together. Each pendulum's swing is a fast oscillation at the average of the two mode frequencies, wrapped in a slow envelope; the two envelopes are exactly complementary, because whatever energy leaves one pendulum arrives at the other. The weaker the coupling spring, the closer the two mode frequencies, and the more majestically slow the exchange — Huygens' clocks, coupled only through faint beam vibrations, took many hundreds of swings to trade their energy. The whole phenomenon lives in one number: the frequency splitting $\Delta\omega = \omega_a - \omega_s$ .

Two carts, three springs: the cleanest coupled system

The standard laboratory version replaces pendulums with two carts of mass $m$ on a track, each tied to a wall by a spring of stiffness $k$ , and tied to each other by a weaker coupling spring $\kappa$ . Newton's second law for each cart collects two forces: its own wall spring, and the coupling spring, which pulls with a force proportional to the relative displacement — that relative term is the coupling, the mathematical signature of two problems refusing to separate:

m\ddot{x}_1 = -k x_1 - \kappa\,(x_1 - x_2), \qquad m\ddot{x}_2 = -k x_2 - \kappa\,(x_2 - x_1) \qquad (1)

where: $x_1, x_2$ — displacements of the two carts from their equilibrium positions; $k$ — stiffness of each wall spring; $\kappa$ — stiffness of the coupling spring; $m$ — each cart's mass. The coupling term vanishes whenever $x_1 = x_2$ : the middle spring only acts when the carts move relative to each other. Neither equation can be solved alone — $x_2$ appears in the equation for $x_1$ and vice versa. For two pendulums of length $L$ joined at the bobs, the same equations hold with $k \to mg/L$ ; everything below carries over with that single substitution.

Derivation ADecoupling by sum and difference coordinates

We are solving the coupled pair (1) exactly — finding the two special frequencies of the two-cart system and the shapes of motion that go with them. The strategy is the symmetry trick that makes the 2×2 case a two-line calculation: instead of the individual positions $x_1$ and $x_2$ , follow their sum and their difference. Because the system is mirror-symmetric (swap the carts and nothing changes), these combinations must evolve independently — the equations literally fall apart when added and subtracted. Watch for where the factor of 2 enters the antisymmetric frequency: each cart feels the coupling spring stretched by twice its own displacement when the carts move oppositely. The general machinery for systems without this symmetry is the eigenvalue problem, treated in The Normal-Mode Eigenvalue Problem — here symmetry hands us the answer directly.

Step 1. Add the two equations of (1). The coupling terms are equal and opposite, $-\kappa(x_1-x_2)$ and $-\kappa(x_2-x_1)$ , so they cancel identically:

m(\ddot{x}_1 + \ddot{x}_2) = -k\,(x_1 + x_2)

Why the cancellation is guaranteed: the coupling spring pushes the carts with equal and opposite forces (Newton's third law), so it can never change the motion of the pair's center — the sum coordinate is blind to the coupling.

Step 2. Define $s = x_1 + x_2$ . Then $\ddot{s} = -(k/m)\,s$ — a free simple harmonic oscillator with frequency $\omega_s = \sqrt{k/m}$ , the bare single-cart frequency. Why this is the symmetric mode: if the motion is pure $s$ (so $d=0$ , i.e. $x_1 = x_2$ always), the carts move in unison and the middle spring keeps constant length — the picture on the left of the hero figure.

Step 3. Subtract the two equations. Now the wall terms give $-k(x_1 - x_2)$ while the coupling terms reinforce: $-\kappa(x_1-x_2) + \kappa(x_2-x_1)\cdot (-1) = -2\kappa(x_1-x_2)$ . With $d = x_1 - x_2$ :

m\ddot{d} = -(k + 2\kappa)\,d

Why the 2: when the carts move oppositely by $\pm d/2$ , the middle spring's length changes by the full $d$ — each cart feels the spring stretched by twice its own displacement, so the coupling contributes $2\kappa$ to the effective stiffness per cart.

Step 4. Read off the two frequencies and their shapes:

\omega_s = \sqrt{\frac{k}{m}} \quad \text{with shape } (1,1), \qquad \omega_a = \sqrt{\frac{k + 2\kappa}{m}} \quad \text{with shape } (1,-1)

The symmetric mode $(1,1)$ oscillates at the bare frequency because the coupler never engages; the antisymmetric mode $(1,-1)$ is stiffer, hence faster. These are the system's two normal modes — the concept the Normal Modes article develops in general.

Step 5. Reconstruct the original coordinates: $x_1 = \tfrac12(s + d)$ and $x_2 = \tfrac12(s - d)$ . Since $s$ and $d$ are each free oscillators, the most general motion of the pair is

x_1(t) = \tfrac12 s(t) + \tfrac12 d(t), \qquad x_2(t) = \tfrac12 s(t) - \tfrac12 d(t)

with $s(t)$ and $d(t)$ sinusoids at $\omega_s$ and $\omega_a$ whose four constants come from the four initial conditions. Two coupled equations have become two independent ones — solved. As a quick check: setting $\kappa = 0$ collapses both frequencies to $\sqrt{k/m}$ — two independent carts, as they must be.

The two frequencies deserve to be displayed on their own, because their difference runs everything that follows. For a weak coupling spring, expand the square root: $\omega_a = \omega_s\sqrt{1 + 2\kappa/k} \approx \omega_s(1 + \kappa/k)$ , so the splitting is linear in the coupling:

\Delta\omega \;=\; \omega_a - \omega_s \;\approx\; \omega_0\,\frac{\kappa}{k} \;=\; \frac{\kappa}{m\,\omega_0} \qquad (2)

where: $\omega_0 = \sqrt{k/m}$ — the bare frequency; $\kappa$ — the coupling stiffness. This is the level splitting: before coupling, the pair had one frequency (twice over); after coupling, two distinct frequencies straddle the old one's job, separated in proportion to the coupling strength. Halve the coupling spring's stiffness and the two tones move twice as close together. Keep this formula in view — it is the direct classical ancestor of the split molecular levels and qubit avoided crossings pictured below.

Figure

Coupling splits one frequency into two — and if the oscillators are detuned, the two branches refuse to cross

The right-hand panel already contains chemistry and quantum engineering in embryo. Two identical atomic orbitals coupled by bond overlap split into a lower (bonding) and higher (antibonding) level — the exact analogue of the symmetric and antisymmetric mode, with the splitting set by the coupling. Sweep one oscillator's frequency through the other's and the two branches repel rather than cross, with a minimum gap fixed by the coupling strength: an avoided crossing, measured every day in the spectroscopy of coupled superconducting qubits.

The energy slosh

Now for the phenomenon Huygens saw. Strike cart 1 (or pendulum 1) while its partner rests. In mode language the initial state $x_1 = A$ , $x_2 = 0$ means $s(0) = A$ and $d(0) = A$ : both modes excited, in equal strength. Everything that follows is the two modes drifting out of step.

Go deeper · Derivation B · 2 minThe slosh period from two drifting tones

Derivation BThe slosh period from two drifting tones

We derive the complete time history of both oscillators after striking only the first one, and extract the headline result: energy is handed over completely in time $\pi/\Delta\omega$ and returns at the exchange period $T_{\text{ex}} = 2\pi/\Delta\omega$ . The route is pure superposition — write the struck start as an equal mixture of the two modes, let each mode run at its own frequency, and recombine. The one identity needed is the sum-to-product formula for cosines, which repackages "two nearby tones" as "one carrier times one slow envelope"; the Beats article develops that identity as a phenomenon in its own right — here we need its energy consequence. Watch the factor of two at the end: the amplitude envelope repeats only after $4\pi/\Delta\omega$ , but the energy exchange is complete twice per envelope cycle.

Step 1. Impose the initial conditions $x_1(0) = A$ , $x_2(0) = 0$ , both at rest. In sum and difference coordinates: $s(0) = A$ , $d(0) = A$ , $\dot{s}(0) = \dot{d}(0) = 0$ . Each normal coordinate then evolves as a free cosine at its own frequency:

s(t) = A\cos\omega_s t, \qquad d(t) = A\cos\omega_a t

Why both modes are excited equally: the strike state $(A, 0)$ is not a mode shape — it is exactly half $(A,A)$ plus half $(A,-A)$ , a 50/50 superposition of the symmetric and antisymmetric patterns.

Step 2. Recombine into the physical coordinates using $x_1 = \tfrac12(s+d)$ , $x_2 = \tfrac12(s-d)$ :

x_1(t) = \frac{A}{2}\big(\cos\omega_s t + \cos\omega_a t\big), \qquad x_2(t) = \frac{A}{2}\big(\cos\omega_s t - \cos\omega_a t\big)

Why $x_2$ starts silent: at $t=0$ the two cosines are equal, so their difference vanishes — the modes cancel exactly on the unstruck cart. The cancellation cannot last, because the two cosines run at different rates.

Step 3. Apply the sum-to-product identities with the average $\bar\omega = (\omega_s + \omega_a)/2$ and the splitting $\Delta\omega = \omega_a - \omega_s$ :

x_1(t) = A\cos\!\Big(\frac{\Delta\omega\,t}{2}\Big)\cos(\bar\omega t), \qquad x_2(t) = A\sin\!\Big(\frac{\Delta\omega\,t}{2}\Big)\sin(\bar\omega t) \qquad (3)

where: $\bar\omega$ — the fast carrier frequency both oscillators actually display; $\Delta\omega/2$ — the slow envelope frequency. Each oscillator swings at the average of the two mode frequencies, inside an envelope that breathes at half their difference — and the two envelopes, $\cos$ and $\sin$ , are exactly complementary.

Step 4. Track the energy. Averaged over the fast carrier, each oscillator's energy is proportional to its envelope squared: $E_1 \propto \cos^2(\Delta\omega\,t/2)$ and $E_2 \propto \sin^2(\Delta\omega\,t/2)$ , with $E_1 + E_2$ constant. The first pendulum stands completely still when its envelope first vanishes, at $\Delta\omega\,t/2 = \pi/2$ , and all the energy returns home when the squared envelope completes a full cycle:

t_{\text{transfer}} = \frac{\pi}{\Delta\omega}, \qquad T_{\text{ex}} = \frac{2\pi}{\Delta\omega} \;\approx\; \frac{2\pi m \omega_0}{\kappa} \qquad (4)

where: $t_{\text{transfer}}$ — time for the first complete handoff of energy; $T_{\text{ex}}$ — the full exchange (out-and-back) period; the last form uses the weak-coupling splitting (2). The number of carrier swings per exchange is $\bar\omega/\Delta\omega \approx k/\kappa$ — the ratio of the spring constants literally counts the swings. Two quick checks: as $\kappa \to 0$ , $\Delta\omega \to 0$ and $T_{\text{ex}} \to \infty$ — uncoupled oscillators never trade energy; and at $t = \pi/\Delta\omega$ equation (3) gives $|x_2| = A$ at carrier peaks, the full original amplitude, consistent with energy conservation.

Notice what was not needed: no driving force, no resonance curve, no damping. Yet the transfer is as dramatic as any driven resonance — pendulum 2 is brought from rest to full amplitude by a force that is tiny at every instant. The trick is the same one that powers Resonance: the weak coupling force stays persistently in phase with pendulum 2's velocity for hundreds of swings, and tiny work done always with the right sign compounds without limit. Resonant energy transfer is resonance with the loudspeaker replaced by a sibling oscillator.

Detuning: why the transfer demands identical twins

The compounding argument has a hidden assumption: the push must stay in phase, which requires the two oscillators to keep time identically. Make pendulum 2 slightly longer or shorter and the phase agreement drifts; energy begins to flow back before the handoff completes. The question "how much energy gets across?" has a beautifully sharp answer.

Go deeper · Derivation C · 2 minDetuning blocks the resonant transfer

Derivation CDetuning blocks the resonant transfer

We derive the peak fraction of energy that a struck oscillator can hand to a detuned partner, and find the Lorentzian law $F = g^2/(g^2 + \Delta^2)$ : the moment the detuning exceeds the coupling, the transfer collapses. The setup is two pendulums with different natural frequencies $\omega_1, \omega_2$ , joined by the same spring $\kappa$ ; the strategy is the mode ansatz on the 2×2 system, tracked through a mixing angle that measures how much the modes still resemble "symmetric" and "antisymmetric". This calculation is the classical twin of the two-level problem in quantum mechanics — same algebra, same avoided crossing, same Lorentzian. The subtlety to watch: detuning $\Delta$ and coupling $g$ both live in units of frequency squared here; comparing them is comparing spring stiffnesses, not raw frequencies.

Step 1. Write the equations of motion with $g = \kappa/m$ and absorb each oscillator's share of the coupling spring into shifted frequencies $\nu_1^2 = \omega_1^2 + g$ , $\nu_2^2 = \omega_2^2 + g$ :

\ddot{x}_1 = -\nu_1^2\,x_1 + g\,x_2, \qquad \ddot{x}_2 = -\nu_2^2\,x_2 + g\,x_1

Why shift the frequencies: each oscillator feels the coupling spring even when its partner is frozen; splitting that self-effect from the true cross-coupling $g\,x_2$ isolates the term responsible for energy transfer.

Step 2. Insert the mode ansatz $x_i = a_i\cos\omega t$ , in which both coordinates share one frequency. The pair of algebraic equations $(\nu_1^2 - \omega^2)a_1 = g\,a_2$ and $(\nu_2^2 - \omega^2)a_2 = g\,a_1$ is consistent only if their product holds:

(\nu_1^2 - \omega^2)(\nu_2^2 - \omega^2) = g^2

Why a product condition: each equation fixes the amplitude ratio $a_2/a_1$ ; a mode exists only when both fix the same ratio, which is exactly the compatibility condition above.

Step 3. Solve the quadratic in $\omega^2$ . With the mean $\mu = (\nu_1^2 + \nu_2^2)/2$ and the detuning $\Delta = (\nu_1^2 - \nu_2^2)/2$ :

\omega_\pm^2 = \mu \pm \sqrt{\Delta^2 + g^2} \qquad (5)

where: $\Delta$ — half the mismatch of the (shifted) squared frequencies; $g = \kappa/m$ — the coupling in the same units. The splitting between the two modes is $2\sqrt{\Delta^2 + g^2}$ : never smaller than $2g$ , even at perfect tuning — this square root is the avoided crossing in the level diagram above. At $\Delta = 0$ it reproduces the identical-oscillator splitting; at large detuning the modes approach the two bare frequencies.

Step 4. Find the mode shapes through the mixing angle $\theta$ defined by $\tan 2\theta = g/\Delta$ . The two mode shapes are $(\cos\theta, \sin\theta)$ and $(-\sin\theta, \cos\theta)$ : at zero detuning $\theta = 45°$ and the modes are the fully shared symmetric/antisymmetric patterns; at large detuning $\theta \to 0$ and each mode retreats into a single pendulum. Why the retreat kills the transfer: a mode localized on pendulum 1 carries its energy nowhere — sharing the energy requires modes that genuinely straddle both oscillators.

Step 5. Strike pendulum 1 and decompose $(A, 0)$ onto the two modes: the coefficients are $A\cos\theta$ and $-A\sin\theta$ . Recombining exactly as in the slosh derivation, the unstruck pendulum oscillates with envelope amplitude $A\sin 2\theta\,$ , reached when the two modes drift fully out of step. Its peak energy fraction is the square:

F \;=\; \sin^2 2\theta \;=\; \frac{g^2}{g^2 + \Delta^2} \qquad (6)

where: $F$ — the maximum fraction of the initial energy that ever resides in pendulum 2. Perfectly tuned twins ( $\Delta = 0$ ) trade everything, $F = 1$ , no matter how weak the coupling — weakness only slows the exchange. Detune by exactly the coupling strength ( $\Delta = g$ ) and only half gets across; detune by a few times $g$ and the energy essentially stays home. Two limiting checks: at $\Delta = 0$ , $F = 1$ recovers the complete transfer of the identical case; as $g \to 0$ at fixed detuning, $F \to g^2/\Delta^2 \to 0$ , matching perturbation-theory intuition. The formula is the classical analogue of the two-level Rabi transfer probability.

Figure

The Lorentzian gatekeeper — peak energy transferred versus detuning, in units of the coupling

This Lorentzian gate is why "sympathetic" phenomena feel almost magical in their selectivity. A piano string sings when you sing its note at it and ignores every other; a wireless charging coil delivers power efficiently only to a receiver tuned to its frequency; and Huygens' two clocks could talk to each other only because they were built as near-perfect twins. Coupling opens the channel; matching decides whether anything flows through it.

Limiting cases

Case	Result	Meaning
zero coupling, $\kappa = 0$	$\omega_a = \omega_s = \omega_0$ , $T_{\text{ex}} \to \infty$	two independent oscillators; degenerate frequency; energy never transfers
weak coupling, $\kappa \ll k$	$\Delta\omega \approx \kappa/m\omega_0$ , complete slosh over $\approx k/\kappa$ swings	near-degenerate modes; slow, total energy exchange — the Huygens regime
strong coupling, $\kappa \gg k$	$\omega_a \approx \sqrt{2\kappa/m} \gg \omega_s = \omega_0$	modes widely split; motion visibly decomposes into fast relative + slow collective parts
detuned pair, $\Delta \gg g$	$F \approx g^2/\Delta^2 \to 0$ , modes localize	each oscillator keeps its own motion; only a trickle of energy leaks across
perfectly tuned, $\Delta = 0$	$F = 1$ for any $g > 0$	complete transfer guaranteed; coupling strength only sets the tempo $T_{\text{ex}} = 2\pi/\Delta\omega$

Common trap“The energy comes back after one full envelope period, 4π/Δω”

The amplitude envelope $\cos(\Delta\omega\,t/2)$ does take $4\pi/\Delta\omega$ to repeat — but energy does not care about the envelope's sign. The energy in each oscillator goes as the envelope squared, $\cos^2(\Delta\omega\,t/2)$ , which repeats twice as fast. So the first complete handoff happens at $t = \pi/\Delta\omega$ , and the energy is fully back home at $T_{\text{ex}} = 2\pi/\Delta\omega$ — at which moment the struck pendulum swings with its original amplitude but possibly opposite phase, which no energy measurement notices. The same factor of 2 appears in Beats: a two-tone sound wobbles in loudness at $\Delta\omega$ , the difference frequency, not at the envelope frequency $\Delta\omega/2$ . In an exam, decide first whether the question asks about amplitude or energy, then pick the period.

Worked examples

1 · Two carts and a weak coupling spring (confidence builder)

Two $0.50\ \mathrm{kg}$ carts sit on a frictionless track, each tied to a wall by a spring of stiffness $k = 200\ \mathrm{N/m}$ and to each other by a spring $\kappa = 8.0\ \mathrm{N/m}$ . Cart 1 is displaced $6.0\ \mathrm{cm}$ and both carts released from rest. Find the two mode frequencies, the splitting, the time for cart 2 to acquire all the energy, and the number of oscillations that takes.

Step 1. Symmetric mode: $\omega_s = \sqrt{k/m} = \sqrt{200/0.50} = 20.0\ \mathrm{rad/s}$ .

Step 2. Antisymmetric mode: $\omega_a = \sqrt{(k+2\kappa)/m} = \sqrt{216/0.50} = \sqrt{432} = 20.78\ \mathrm{rad/s}$ .

Step 3. Splitting: $\Delta\omega = 0.78\ \mathrm{rad/s}$ . The weak-coupling shortcut (2) predicts $\kappa/m\omega_0 = 8.0/(0.50 \times 20.0) = 0.80\ \mathrm{rad/s}$ — high by only 2%, so the approximation is already excellent at $\kappa/k = 4\%$ .

Step 4. First complete handoff from (4): $t_{\text{transfer}} = \pi/\Delta\omega = \pi/0.78 = \boxed{4.0\ \mathrm{s}}$ ; the full exchange period is $T_{\text{ex}} = 8.0\ \mathrm{s}$ .

Step 5. The carrier runs at $\bar\omega = 20.39\ \mathrm{rad/s}$ , period $0.308\ \mathrm{s}$ , so the handoff takes $4.0/0.308 \approx 13$ oscillations, and the round trip $\approx 26$ — matching the estimate $k/\kappa = 200/8 = 25$ swings per exchange. Throughout, the total energy stays at $E = \tfrac12(k+\kappa)A^2 = \tfrac12(208)(0.060)^2 = \boxed{0.37\ \mathrm{J}}$ ; only its address changes.

2 · Huygens' clocks, tuned and mistuned (exam level)

Two pendulum clocks have bobs of mass $m = 1.5\ \mathrm{kg}$ on rods of length $L = 0.994\ \mathrm{m}$ (a seconds pendulum: $T = 2.00\ \mathrm{s}$ ). Their coupling through the shared beam is modeled as a weak spring $\kappa = 0.030\ \mathrm{N/m}$ joining the bobs. (a) Find the mode frequencies, the exchange period, and the number of swings per exchange. (b) One clock's rod is 1.0% longer, $L_2 = 1.004\ \mathrm{m}$ . What fraction of the energy can now transfer?

Step 1. Bare frequency: $\omega_0^2 = g/L = 9.81/0.994 = 9.869\ \mathrm{s^{-2}}$ , so $\omega_0 = 3.142\ \mathrm{rad/s}$ and indeed $T = 2\pi/\omega_0 = 2.00\ \mathrm{s}$ .

Step 2. Coupling parameter: $g_c = \kappa/m = 0.030/1.5 = 0.020\ \mathrm{s^{-2}}$ . Antisymmetric mode: $\omega_a = \sqrt{\omega_0^2 + 2g_c} = \sqrt{9.909} = 3.148\ \mathrm{rad/s}$ ; symmetric mode stays at $\omega_s = 3.142\ \mathrm{rad/s}$ .

Step 3. Splitting: $\Delta\omega = \omega_a - \omega_s = 0.00637\ \mathrm{rad/s}$ (the shortcut $g_c/\omega_0 = 0.020/3.142 = 0.00637$ agrees). Exchange period from (4): $T_{\text{ex}} = 2\pi/0.00637 = \boxed{986\ \mathrm{s} \approx 16\ \mathrm{minutes}}$ — about $493$ full swings per exchange, matching $m\omega_0^2/\kappa = mg/L\kappa = 493$ . A patient phenomenon: exactly the sluggish, majestic energy creep Huygens watched from his sickbed.

Step 4. Now detune. The longer clock has $\omega_2^2 = 9.81/1.004 = 9.771\ \mathrm{s^{-2}}$ . Shifted frequencies: $\nu_1^2 = 9.869 + 0.020 = 9.889$ , $\nu_2^2 = 9.771 + 0.020 = 9.791$ . Detuning: $\Delta = (\nu_1^2 - \nu_2^2)/2 = 0.049\ \mathrm{s^{-2}}$ — already $2.5$ times the coupling $g_c = 0.020\ \mathrm{s^{-2}}$ .

Step 5. Transfer fraction from (6): $F = g_c^2/(g_c^2 + \Delta^2) = (0.020)^2/\big((0.020)^2 + (0.049)^2\big) = 0.00040/0.00280 = \boxed{0.14}$ . A 1% length mismatch blocks 86% of the energy transfer. This is why sympathetic effects demand near-identical partners — and why Huygens, the best clockmaker alive, was uniquely equipped to stumble on them.

Origins · 1 minChristiaan Huygens and how this was discovered

Christiaan Huygens

1629–1695 · The Hague and Paris

Christiaan Huygens invented the pendulum clock in 1656 — improving timekeeping accuracy overnight from a quarter-hour to seconds per day — and then, in February 1665, made the first recorded observation of coupled oscillators while confined to bed by illness. Watching two of his clocks hang from a common beam, he saw their pendulums fall into exact anti-phase lockstep within half an hour, and correctly diagnosed the cause as imperceptible motions of the beam coupling the two: an explanation, not just an observation, decades ahead of the mathematics needed to formalize it. His Horologium Oscillatorium (1673) — analyzing pendulum dynamics, the cycloidal isochrone, and centrifugal force — stands beside Newton's Principia among the founding texts of mechanics. Huygens also discovered Titan, explained Saturn's rings, and built the wave theory of light whose interference ideas are, fittingly, the same superposition physics that makes his clocks trade energy.

History

1657

— Huygens patents the pendulum clock, built by Salomon Coster in The Hague; accuracy improves a hundredfold.

1665

— Confined to bed, Huygens observes the "odd sympathy" of two clocks on a common beam and attributes it to imperceptible beam vibrations — the first documented coupled-oscillator experiment.

1673

— Horologium Oscillatorium publishes the full theory of pendulum motion, the mathematical foundation the sympathy phenomenon would eventually rest on.

1877

— Rayleigh's Theory of Sound treats coupled vibrations, mode splitting, and beats systematically, giving Huygens' observation its modern linear-theory backbone.

2002

— Bennett, Schatz, Rockwood and Wiesenfeld reproduce Huygens' experiment with modern instrumentation at Georgia Tech, confirming beam-mediated coupling and mapping when anti-phase locking occurs.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 4 (Oscillations) and the coupled-oscillator / normal-mode sections · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 6 (Oscillations) · Landau & Lifshitz, Mechanics (Vol. 1), §23–24 (Oscillations of systems with several degrees of freedom) · Taylor, Classical Mechanics, Ch. 11 (Coupled Oscillators and Normal Modes) All results are analytic and contain no empirical constants. The equations of motion yield $M\ddot{x}=-Kx$ from the quadratic Lagrangian; add/subtract decoupling confirms $\omega_+=\sqrt{k/m}$ and $\omega_-=\sqrt{(k+2\kappa)/m}$ ; the beating envelope follows from sum-to-product identities; and the periodic-chain extension gives the lattice dispersion $\omega(q)=2\sqrt{k/m}|\sin(qa/2)|$ . Worked-example numbers on this page ( $\omega_s = 20.0$ , $\omega_a = 20.78$ rad/s, $T_{\text{ex}} = 8.0$ s, $E = 0.37$ J; $\omega_0 = 3.142$ rad/s, $T_{\text{ex}} = 986$ s, $F = 0.14$ at 1% detuning) are illustrative arithmetic recomputed from the boxed formulas, as is the detuned-transfer Lorentzian $F = g^2/(g^2+\Delta^2)$ , which follows from the two-mode mixing angle.

Builds on: Simple Harmonic Motion · Small Oscillations About Equilibrium · Complex Exponential Representation of Oscillations · Energy in Simple Harmonic Motion Leads to: Resonance · Driven Oscillations · Damped Harmonic Oscillator · The Simple Pendulum

If each normal mode conserves its own energy forever, how can energy possibly move from pendulum 1 to pendulum 2?

Both bookkeepings are correct — they just track different things. The symmetric mode's energy and the antisymmetric mode's energy are separately constant from the moment of the strike; nothing ever flows between modes. But "pendulum 1's energy" is not a mode energy: writing $x_1 = \tfrac12(s+d)$ , its energy contains cross terms between the two modes, and those interference terms oscillate at the slow difference frequency as the modes drift in and out of step. The slosh is pure interference: the modes are steady, but where their sum happens to concentrate its amplitude sweeps back and forth between the two pendulums. Per-mode accounting shows constancy; per-object accounting shows the transfer — same physics, two ledgers.

After I strike pendulum 1, does it swing at its natural frequency

\omega_0

Not quite — and the small discrepancy is the whole phenomenon. The struck pendulum's motion, equation (3), is a carrier at $\bar\omega = (\omega_s + \omega_a)/2$ , the average of the two mode frequencies, which for the pendulum-plus-spring system sits slightly above $\omega_0$ . The frequencies genuinely present in the motion are only $\omega_s$ and $\omega_a$ ; the strike excites both, and no single sinusoid at $\omega_0$ appears anywhere. Watch a struck pendulum casually and it looks like plain SHM at roughly the bare frequency — but wait $\sim k/\kappa$ swings and the envelope betrays the two-tone truth. The individual pendulum's "own frequency" stops being a motion of the system the moment you couple it; only the mode frequencies are physically realizable pure tones.

Is the energy slosh the same thing as beats?

Same identity, different observable. Beats, as developed in the Beats article, is what two nearby frequencies do when superposed at one place: a carrier at the average frequency, an envelope at half the difference. The coupled-oscillator slosh is exactly that superposition — the two frequencies are the modes, and each pendulum displays the two-tone sum. What coupling adds is the anti-correlation: pendulum 1 carries the $\cos$ envelope while pendulum 2 carries the $\sin$ envelope, so the loudness wobble of beats becomes a physical hand-off of energy between two objects. You can hear beats from two independent tuning forks; you need coupling for the forks to actually exchange their vibrational energy.

Why does the symmetric mode have exactly the bare frequency? Doesn't the coupling spring add stiffness to everything?

A spring only pushes when its length changes. In the symmetric mode the two masses move in identical copies — equal displacements at every instant — so the coupling spring translates rigidly, keeping exactly its natural length, exerting zero force forever. The mode is dynamically invisible to the coupler, and each mass moves under its wall spring alone: frequency $\sqrt{k/m}$ , untouched. This is a general and useful principle: any motion that respects a symmetry of the coupling escapes it. In the CO $_2$ molecule the same logic makes the symmetric stretch blind to certain interactions, and in circuit pairs the "common mode" passes through couplings that savage the "differential mode." Whenever a coupled system has a mode at a suspiciously unshifted frequency, look for the motion the coupler cannot see.

How weak can the coupling be and still give complete energy transfer?

Arbitrarily weak — with two conditions attached. Equation (6) says tuned twins ( $\Delta = 0$ ) reach $F = 1$ for any nonzero coupling; weakness only stretches the timetable, $T_{\text{ex}} = 2\pi/\Delta\omega \propto 1/\kappa$ . The first catch is tuning: the transfer stays near-complete only while the detuning stays below the coupling, $\Delta \lesssim g$ , and a weaker coupling makes that tolerance window proportionally narrower — the real reason feeble couplings rarely produce dramatic transfer outside a lab. The second catch is damping (ignored in this article): the exchange must complete before friction spends the energy, so you need $T_{\text{ex}}$ shorter than the decay time, i.e. a frequency splitting larger than the linewidth. Engineers call that the strong-coupling condition, and it decides whether wireless chargers and qubit buses actually work.

Was Huygens' clock synchronization just the beating described here?

No — and the distinction is worth internalizing. Linear coupled-oscillator theory, everything on this page, predicts energy sloshing back and forth forever; it can never make two oscillators lock into a fixed phase relationship, because superposition preserves both mode amplitudes. Huygens' clocks did something stronger: they settled into permanent anti-phase and stayed there. That requires the two ingredients this article deliberately excludes — energy input (the escapement mechanism kicking each pendulum every swing) and dissipation — which make the system nonlinear and let it forget initial conditions. The linear physics still sets the stage: the beam coupling and near-identical frequencies open the communication channel, and the antisymmetric mode's special status (it moves the beam least) is why anti-phase wins. But true synchronization is a nonlinear limit-cycle phenomenon, the opening chapter of a field that now covers firefly flashes, cardiac pacemakers, and power-grid stability.

What happens with three or more coupled oscillators — can I still use sum and difference?

Sum and difference is the two-body special case of a general recipe. With $N$ coupled oscillators there are always exactly $N$ special combinations of the coordinates that evolve independently — the normal coordinates — but for $N \ge 3$ , or for two unequal masses, guessing them from symmetry usually fails and you need the systematic machinery: write the equations as a matrix problem and solve $\det(K - \omega^2 M) = 0$ for the mode frequencies and shapes. That machinery lives in The Normal-Mode Eigenvalue Problem, and the conceptual picture — every motion as a superposition of fixed-shape, single-frequency patterns — is the subject of Normal Modes. The two-cart case you now own is the template: everything there generalizes, with eigenvectors replacing "sum" and "difference."

Coupled Oscillators

Where this lives today

The physical picture

Two carts, three springs: the cleanest coupled system

The energy slosh

Detuning: why the transfer demands identical twins

Limiting cases

Worked examples

1 · Two carts and a weak coupling spring (confidence builder)

2 · Huygens' clocks, tuned and mistuned (exam level)

Going further

Before this

Where this leads