Energy Transfer Between Coupled Oscillators

In 1665, confined to his room by a minor illness, Christiaan Huygens stared at two of his pendulum clocks hanging from the same wooden beam and noticed something eerie: no matter how he started them, within half an hour their pendulums settled into perfect anti-phase lockstep. He called it an "odd kind of sympathy." The clocks were talking to each other — tiny vibrations carried through the beam were shuttling energy back and forth between them. Huygens had discovered, two centuries early, the physics of energy transfer between coupled oscillators. Take any two identical oscillators — pendulums, masses on springs, LC circuits — and connect them weakly. Start one moving while the other rests. Then wait. The first oscillator slowly dies down while the second, untouched, swings up from nothing; then the flow reverses, and the energy sloshes back. The whole exchange is governed by a single small number: the frequency splitting $\delta$ between the system's two normal modes. Energy migrates completely in a time $T_{\mathrm{transfer}} = \pi/\delta$ , and it migrates completely only if the two oscillators are perfectly in tune. That last clause is the deep one. The same mathematics — beating between two nearly degenerate modes — reappears as Rabi oscillations of a quantum two-level atom, the design equation of radio-frequency filters, energy hopping between vibrating bonds in a molecule, and the two-qubit gates of a superconducting quantum computer. Learn the pendulums and you have learned them all.

Key takeaways

Two identical oscillators joined by a weak coupling no longer oscillate independently: the true free motions are the symmetric and antisymmetric normal modes, with slightly different frequencies $\omega_+$ and $\omega_-$ .
Starting all the energy in oscillator 1 excites both modes equally; the small splitting $\delta = \omega_- - \omega_+$ makes them drift out of phase, and the energy sloshes fully into oscillator 2 and back — beating.
The masses wiggle fast, at the mean carrier frequency $\bar\omega = (\omega_+ + \omega_-)/2$ ; the amplitude envelope modulates slowly, at $\delta/2$ ; complete transfer takes $T_{\mathrm{transfer}} = \pi/\delta$ .
Total energy is conserved — only its partition oscillates: $E_1 = E_0\cos^2(\delta t/2)$ , $E_2 = E_0\sin^2(\delta t/2)$ .
Complete transfer demands identical (resonant) oscillators. Detuning $\Delta$ caps the transferred fraction at $f_{\max} = \Omega_c^2/[\Omega_c^2 + (\Delta/2)^2] < 1$ — the classical Rabi formula.
The identical mathematics governs quantum Rabi flopping, coupled LC filters, molecular vibrational energy flow, and two-qubit iSWAP gates.
Key idea: decompose the coupled motion into normal modes — the slow beat at the mode splitting $\delta = \omega_- - \omega_+$ carries the energy back and forth, with complete transfer only at exact resonance.

The core insight

Decompose the coupled motion into normal modes; the slow beat at the mode-splitting frequency $\delta = \omega_- - \omega_+$ governs the back-and-forth energy flow, with complete transfer only at exact resonance.

Where this lives today

The purest modern incarnation of this article is sitting in a dilution refrigerator. In a superconducting quantum processor, two qubits coupled by a tunable element are exactly the detuned-oscillator problem (8): bring the qubits to resonance and excitation sloshes between them at the coupling rate; wait precisely half a slosh, $t = \pi/(2\Omega_c)$ , and you have executed an iSWAP gate — the energy-transfer beat, frozen at its midpoint, is a native two-qubit operation. Detune the qubits and the exchange shuts off as the Lorentzian (9) collapses — which is precisely how the gate is turned off. Gate calibration sheets at IBM, Google, and Rigetti are, in effect, measurements of $\Omega_{\mathrm{eff}}$ and $f_{\max}$ . The same physics runs at every scale. Wireless chargers push power through the resonant coupling of two coils, engineered so the "transfer time" beats the dissipation time — the $\gamma \ll \delta$ condition of the limiting-cases table as a design spec. In photonics, two evanescently coupled waveguides swap light along the propagation direction (distance playing the role of time), the basis of directional couplers in every lithium-niobate modulator. And in chemistry, Fermi resonances shuttle vibrational energy between near-degenerate molecular modes, setting the speed of intramolecular energy redistribution that decides which bond of an excited molecule breaks first.

The physical picture

Hang two identical pendulums from a common support and join them with a soft, floppy spring. Pull pendulum 1 aside, hold pendulum 2 dead still, and release. For the first few swings nothing seems to happen — pendulum 1 swings, pendulum 2 barely quivers. But watch for a minute. Swing by swing, pendulum 1 loses amplitude and pendulum 2 gains it, until — startlingly — pendulum 1 hangs motionless and pendulum 2 has stolen the entire motion. Then the theft reverses. The energy is a fluid sloshing between two tanks connected by a thin pipe, and the narrowness of the pipe — the weakness of the coupling — sets the slow sloshing rate.

Figure

The demonstration — two identical pendulums, a weak spring, and energy that will not stay put

Why does the energy move at all? Because "pendulum 1 swinging, pendulum 2 still" is not a natural motion of the coupled system. The coupled pair has exactly two natural motions — its normal modes. In the symmetric mode both pendulums swing together, in phase, and the coupling spring between them never stretches; it might as well not exist, so this mode keeps the bare frequency $\omega_+ = \omega_0$ . In the antisymmetric mode they swing in perfect opposition, working the spring hard on every cycle, which stiffens the restoring force and raises the frequency to $\omega_- > \omega_+$ . A localized excitation — all the motion in pendulum 1 — is secretly a 50/50 mixture of these two modes. And a mixture of two clocks ticking at slightly different rates cannot hold its shape: the modes drift apart in phase at the difference frequency $\delta = \omega_- - \omega_+$ , and where they initially added constructively in pendulum 1, half a beat later they add constructively in pendulum 2 instead. The pattern of energy has migrated even though neither mode has changed its amplitude at all. The result, traced in time, is the signature waveform of this whole subject — a fast carrier oscillation trapped inside a slow envelope:

Figure

Beating — each mass oscillates fast at the carrier frequency while its envelope pulses slowly at the mode splitting

Keep the two frequencies straight from the start, because conflating them is the classic error. The carrier — the visible wiggle — runs at the mean frequency $\bar\omega = (\omega_+ + \omega_-)/2 \approx \omega_0$ , fast. The envelope — the slow swelling and dying of each oscillator's amplitude — runs at $\delta/2$ , slow. Energy, which goes as amplitude squared, completes its journey from oscillator 1 to oscillator 2 in half an envelope period: $T_{\mathrm{transfer}} = \pi/\delta$ . Weaken the coupling and $\delta$ shrinks, the pipe narrows, and the slosh takes longer — but it still, eventually, completes. What can stop complete transfer is not weak coupling but detuning: make the two oscillators even slightly unequal and part of the energy stays locked in the first one forever. That resonance condition is the punchline of this article.

Two modes, one beat

This article rides on the normal-mode machinery built in Coupled oscillators — there you will find the general eigenvalue problem, and in Normal modes its full anatomy. Here we need only the simplest case: two equal masses $m$ , each on its own spring of stiffness $k$ , joined to each other by a weak spring $\kappa \ll k$ . Newton's second law for each mass reads

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 from equilibrium; $\kappa(x_1 - x_2)$ — the coupling force, proportional to the stretch of the connecting spring, pulling the masses toward each other's displacement. Each equation would be ordinary SHM if not for the last term, which feeds motion of one mass into the dynamics of the other. That cross-feed is the pipe through which energy will flow. The system decouples the moment you stop watching the masses and start watching their sum and difference. Define the normal coordinates

\eta_+ = x_1 + x_2, \qquad \eta_- = x_1 - x_2 \qquad (2)

where: $\eta_+$ — the symmetric (center-of-mass-like) coordinate, blind to the coupling spring, which never stretches when the masses move together; $\eta_-$ — the antisymmetric (stretch) coordinate, which works the coupling spring twice over. Adding the two equations of motion kills the coupling terms; subtracting them doubles the coupling. Each combination obeys its own independent SHM, with frequencies

\omega_+ = \sqrt{\frac{k}{m}}, \qquad \omega_- = \sqrt{\frac{k + 2\kappa}{m}} \qquad (3)

where: $\omega_+$ — the symmetric-mode frequency, exactly the bare $\omega_0 = \sqrt{k/m}$ because the coupling spring is a spectator; $\omega_-$ — the antisymmetric-mode frequency, raised because opposing motion stretches the coupling spring by $2x$ and adds $2\kappa$ of effective stiffness. The mode that ignores the spring is slower; the mode that fights it is faster. Everything about energy transfer hides in how close these two frequencies are. For weak coupling, expand the square root to first order in $\kappa/k$ :

\delta \equiv \omega_- - \omega_+ \approx \omega_0\,\frac{\kappa}{k} = \frac{\kappa}{\sqrt{mk}} \qquad (4)

where: $\delta$ — the mode splitting, linear in the coupling stiffness. Halve the coupling and you halve the splitting; the beat slows in exact proportion. This single small number is the metronome of the entire energy exchange — nothing else about the coupling survives into the final answer.

Derivation AFrom localized start to the beating solution

We are deriving the celebrated beating solution — $x_1 = A\cos(\delta t/2)\cos(\bar\omega t)$ , $x_2 = A\sin(\delta t/2)\sin(\bar\omega t)$ — for two identical weakly coupled oscillators released with all the motion in oscillator 1. The strategy is the standard normal-mode route: write the general motion as a superposition of the two independent mode oscillations, fix the four constants from the localized initial conditions, then reorganize the resulting sum of two cosines with sum-to-product identities so that the fast and slow time scales separate explicitly. The step to watch is the initial condition: a localized start forces the two modes to enter with equal amplitudes — that equality is precisely what makes later perfect cancellation (and hence complete transfer) possible, and it is exactly what detuning will destroy.

Step 1. Write the general solution as independent SHM in each normal coordinate:

\eta_+ = C_+\cos(\omega_+ t + \phi_+), \qquad \eta_- = C_-\cos(\omega_- t + \phi_-)

Why: equations (1) added and subtracted give $\ddot\eta_\pm = -\omega_\pm^2\,\eta_\pm$ — two uncoupled oscillators, each with its own amplitude and phase to be fixed by initial data.

Step 2. Invert (2) to return to the physical coordinates:

x_1 = \tfrac12(\eta_+ + \eta_-), \qquad x_2 = \tfrac12(\eta_+ - \eta_-)

Step 3. Impose the localized start: $x_1(0) = A$ , $x_2(0) = 0$ , $\dot x_1(0) = \dot x_2(0) = 0$ . In mode language $\eta_+(0) = \eta_-(0) = A$ with zero initial velocities, so $C_+ = C_- = A$ and $\phi_\pm = 0$ . Why the equal weights matter: "all the motion in oscillator 1" is not a mode — it is the fifty-fifty superposition of both modes. Each mode alone spreads its motion over both masses equally; only their interference localizes it.

Step 4. Substitute back:

x_1 = \frac{A}{2}\left[\cos\omega_+ t + \cos\omega_- t\right], \qquad x_2 = \frac{A}{2}\left[\cos\omega_+ t - \cos\omega_- t\right]

Why the signs differ: mass 1 carries the modes in phase, mass 2 carries them with opposite signs — so wherever the two cosines agree, mass 1 moves and mass 2 stands still, and vice versa.

Step 5. Apply the sum-to-product identities with $\bar\omega = (\omega_+ + \omega_-)/2$ and $\delta = \omega_- - \omega_+$ :

x_1(t) = A\cos\!\Big(\frac{\delta t}{2}\Big)\cos(\bar\omega t), \qquad x_2(t) = A\sin\!\Big(\frac{\delta t}{2}\Big)\sin(\bar\omega t) \qquad (5)

where: the second factor in each product is the fast carrier at $\bar\omega$ ; the first is the slow envelope at $\delta/2$ . The envelopes are a cosine and a sine of the same slow argument — when one is full the other is empty, in perfect complement. Verified: at $\delta \to 0$ (no coupling) the solution reduces to $x_1 = A\cos\bar\omega t$ , $x_2 = 0$ — independent oscillators, as it must. Matches Taylor ch. 11 and Morin's coupled-pendulum result.

Equation (5) is worth reading slowly. Oscillator 1 starts at full amplitude ( $\cos 0 = 1$ ) and dies to exactly zero when $\delta t/2 = \pi/2$ ; at that same instant oscillator 2's envelope $\sin(\delta t/2)$ reaches its maximum. The instant of total silence in oscillator 1 is the instant of total uproar in oscillator 2. Note also the carrier phases: $x_2$ oscillates as a sine against $x_1$ 's cosine — the receiving oscillator runs a quarter cycle behind the donor, which is the microscopic mechanism of the transfer: each fast cycle, oscillator 1's spring force does a little net positive work on oscillator 2 through the coupling, precisely because of that 90° lag.

The energy ledger

Go deeper · Derivation B · 1 minEnergy partition and its conservation

Derivation BEnergy partition and its conservation

We are deriving the energy bookkeeping of the beat — $E_1 = E_0\cos^2(\delta t/2)$ , $E_2 = E_0\sin^2(\delta t/2)$ , with constant total — from the beating solution (5). The route is the two-timescale argument: because $\delta \ll \bar\omega$ , each envelope is effectively frozen over any single fast cycle, so each oscillator behaves at every instant like an ordinary SHM whose amplitude happens to be the current envelope value; the SHM energy formula $E = \tfrac12 m\omega^2 a^2$ from Energy in harmonic motion then applies cycle by cycle. Watch the small print: these are cycle-averaged energies, and the tiny energy parked in the coupling spring (of order $\kappa/k$ relative to the total) is neglected — both approximations are controlled by the weak-coupling condition.

Step 1. Read off the slowly varying amplitudes from (5):

a_1(t) = A\cos\!\Big(\frac{\delta t}{2}\Big), \qquad a_2(t) = A\sin\!\Big(\frac{\delta t}{2}\Big)

Why this is legitimate: over one fast period $2\pi/\bar\omega$ the envelope argument advances by only $\pi\delta/\bar\omega \ll 1$ — the envelope is constant to first approximation, so each mass executes clean SHM at the current amplitude.

Step 2. Apply the SHM energy formula to each oscillator, with $\omega_0 \approx \bar\omega$ for weak coupling:

E_i = \tfrac12\,m\,\omega_0^2\,a_i^2

Step 3. Insert the envelopes and define $E_0 = \tfrac12 m\omega_0^2 A^2$ , the energy the system started with:

E_1(t) = E_0\cos^2\!\Big(\frac{\delta t}{2}\Big), \qquad E_2(t) = E_0\sin^2\!\Big(\frac{\delta t}{2}\Big) \qquad (6)

Step 4. Check the sum: $E_1 + E_2 = E_0[\cos^2 + \sin^2] = E_0$ , constant. Why the partition oscillates at $\delta$ , not $\delta/2$ : $\cos^2(\delta t/2) = \tfrac12[1 + \cos\delta t]$ — squaring the envelope doubles its frequency, so the energy completes a full out-and-back cycle in $2\pi/\delta$ and a one-way trip in half that. *Verified: at $t = \pi/\delta$ , $E_1 = 0$ and $E_2 = E_0$ — complete transfer, consistent with Landau-Lifshitz §23-24.*

The one-way travel time is the number an experimenter actually measures with a stopwatch:

T_{\mathrm{transfer}} = \frac{\pi}{\delta} = \frac{\pi\sqrt{mk}}{\kappa} \qquad (7)

where: the second form uses the weak-coupling splitting (4). Read it as an impedance statement: stiffer coupling moves energy faster, in inverse proportion. For the classroom pendulum demo, $T_{\mathrm{transfer}}$ is tens of seconds — dozens of fast swings — which is exactly what makes the effect so theatrical: the audience has time to notice that pendulum 1 has stopped. And note what is not conserved: nothing about oscillator 1 individually. Its energy runs from $E_0$ to zero and back with no half-measures. Only the total — and, separately, the energy in each normal mode — stays constant. Energy conservation lives in mode space; the drama in position space is pure interference.

Detuning, resonance, and the classical Rabi formula

Now break the symmetry. Let the two oscillators have slightly different bare frequencies, $\omega_1$ and $\omega_2 = \omega_1 + \Delta$ — two pendulums of slightly unequal length, two LC circuits with mismatched capacitors. Everything above came from the equal-weight superposition forced by the localized start; detuning unbalances the weights, and the energy exchange becomes both faster and shallower:

Figure

Energy exchange at resonance and off resonance — detuning makes the slosh faster but forever incomplete

The cleanest way to see it is to strip out the fast carrier entirely and follow only the envelopes. Write each oscillator's motion as $x_j = \mathrm{Re}\,[b_j(t)\,e^{i\bar\omega t}]$ , where the complex amplitudes $b_j$ vary slowly compared with the carrier — the same complex-exponential bookkeeping that tames every oscillation problem. Substituting into the coupled equations and dropping second derivatives of the slow envelopes gives

i\dot b_1 = \frac{\Delta}{2}\,b_1 + \Omega_c\,b_2, \qquad i\dot b_2 = -\frac{\Delta}{2}\,b_2 + \Omega_c\,b_1 \qquad (8)

where: $\Delta = \omega_2 - \omega_1$ — the detuning between the bare frequencies; $\Omega_c$ — the coupling rate, equal to $\delta/2$ for our spring-coupled pair at resonance. Anyone who has met the Schrödinger equation for a two-level atom should feel déjà vu: equation (8) is that equation, with the mode amplitudes as the "wavefunction" and the detuning and coupling as the level splitting and drive. Classical beats and quantum Rabi flopping are one calculation wearing two costumes.

Go deeper · Derivation C · 2 minIncomplete transfer under detuning

Derivation CIncomplete transfer under detuning

We are deriving the maximum fraction of energy that can ever reach oscillator 2 when the oscillators are detuned by $\Delta$ — the classical Rabi formula $f_{\max} = \Omega_c^2/[\Omega_c^2 + (\Delta/2)^2]$ . The route: solve the linear envelope system (8) with everything starting in oscillator 1, which is cleanest in terms of the system's eigenfrequency; the answer will show that detuning both speeds up the exchange and caps it below completeness. Watch the structure of the result — the cap is a Lorentzian in the detuning, the universal resonance line shape, and it does not care about the sign of $\Delta$ : too long a pendulum blocks transfer exactly as effectively as too short.

Step 1. Look for normal solutions $b_j \propto e^{-i\lambda t}$ of the envelope system. The 2×2 eigenvalue problem gives

\lambda_\pm = \pm\,\Omega_{\mathrm{eff}}, \qquad \Omega_{\mathrm{eff}} = \sqrt{\Omega_c^2 + \left(\frac{\Delta}{2}\right)^2}

Why $\Omega_{\mathrm{eff}}$ grows with detuning: the eigenvalues of the matrix with diagonal $\pm\Delta/2$ and off-diagonal $\Omega_c$ sit at plus-or-minus the quadrature sum — detuning pushes the two envelope eigenfrequencies apart, so the exchange oscillation actually gets faster.

Step 2. Solve with the localized initial condition $b_1(0) = 1$ , $b_2(0) = 0$ . Projecting onto the two eigenvectors and recombining:

|b_2(t)|^2 = \frac{\Omega_c^2}{\Omega_{\mathrm{eff}}^2}\,\sin^2(\Omega_{\mathrm{eff}}\,t)

Why the prefactor appears: with detuning the two eigenvectors are no longer the symmetric and antisymmetric combinations — they lean toward the oscillators they are closer to in frequency. The localized start then projects unequally onto them, and unequal amplitudes can never fully cancel in oscillator 1 or fully add in oscillator 2.

Step 3. Read off the peak of the sine-squared:

f_{\max} = \frac{\Omega_c^2}{\Omega_c^2 + (\Delta/2)^2} \qquad (9)

where: $f_{\max}$ — the largest fraction of the initial energy that ever visits oscillator 2, reached at $t = \pi/(2\Omega_{\mathrm{eff}})$ . It is a Lorentzian of width $2\Omega_c$ in the detuning: match the oscillators to within their coupling rate and transfer is nearly complete; detune by much more and the energy barely leaks.

Step 4. Check the resonant limit: $\Delta \to 0$ gives $\Omega_{\mathrm{eff}} = \Omega_c = \delta/2$ and $|b_2|^2 = \sin^2(\delta t/2)$ — exactly the energy partition (6), with complete transfer at $t = \pi/\delta$ . Why this consistency check matters: it fixes the identification $\Omega_c = \delta/2$ and confirms that the envelope formalism contains everything the exact solution knew. *Verified: identical in form to the two-level Rabi flopping formula; reduces to (6) at zero detuning; symmetric under $\Delta \to -\Delta$ as required.*

Which equation answers which question

you want	use	notes
the two mode frequencies	$\omega_+ = \sqrt{k/m}$ , $\omega_- = \sqrt{(k+2\kappa)/m}$ , eq. (3)	symmetric mode ignores the coupling spring
the beat/splitting rate	$\delta \approx \kappa/\sqrt{mk}$ , eq. (4)	linear in coupling; the master small number
the full motion from a localized start	eq. (5)	carrier at $\bar\omega$ , envelope at $\delta/2$
energy in each oscillator vs time	$E_1 = E_0\cos^2(\delta t/2)$ , $E_2 = E_0\sin^2(\delta t/2)$ , eq. (6)	cycle-averaged; total constant
time for complete transfer	$T_{\mathrm{transfer}} = \pi/\delta$ , eq. (7)	half the envelope period
exchange with detuning	$\Omega_{\mathrm{eff}}$ and $f_{\max}$ , eqs. (8)-(9)	faster but shallower; Lorentzian cap

Limiting cases

Case	Result	Meaning
weak coupling, identical oscillators	$\delta \ll \omega_0$ : clean beats, 100% transfer in $\pi/\delta$	the textbook regime — many fast cycles per slosh
zero coupling	$\delta \to 0$ , $T_{\mathrm{transfer}} \to \infty$	independent oscillators; energy never moves
strong coupling	$\kappa \sim k$ : $\omega_\pm$ far apart	no two-timescale separation — just two distinct modes, no "beating" story
detuned oscillators	$f_{\max} = \dfrac{\Omega_c^2}{\Omega_c^2 + (\Delta/2)^2} < 1$	exchange faster ( $\Omega_{\mathrm{eff}}$ ) but forever incomplete
light damping $\gamma \ll \delta$	many transfer cycles before decay	beats visible, slowly dying — see Damped harmonic oscillator
heavy damping $\gamma \gg \delta$	energy dissipates before one slosh completes	beating suppressed entirely

Common trapThe masses wiggle fast, so the energy must move fast

No — and this is the error that ruins most first encounters with beats. The visible oscillation runs at the carrier frequency $\bar\omega \approx \omega_0$ , but the energy moves at the far slower envelope rate set by the splitting $\delta = \omega_- - \omega_+ \approx \kappa/\sqrt{mk}$ , which for weak coupling is tiny. In the classroom demo the pendulums swing once a second while the energy takes half a minute to change hands: two timescales, three orders of magnitude apart, in one system. A second trap hides in the factor of 2: the amplitude envelope oscillates at $\delta/2$ , but the energy $\propto$ envelope² oscillates at $\delta$ — so complete transfer takes $\pi/\delta$ , not $2\pi/\delta$ . And the third trap is resonance: 100% transfer is special to identical oscillators. Any detuning $\Delta$ caps the exchange at $\Omega_c^2/[\Omega_c^2 + (\Delta/2)^2]$ — your slightly mismatched lab pendulums will slosh energy shallowly and quickly, never completely.

Worked examples

1 · The classroom demo, by the numbers (confidence builder)

Two carts of mass $m = 0.50\ \mathrm{kg}$ each sit on separate springs of stiffness $k = 20\ \mathrm{N/m}$ and are joined by a weak spring $\kappa = 0.80\ \mathrm{N/m}$ . Cart 1 is displaced and released with cart 2 at rest. Find the two mode frequencies, the beat splitting, the complete-transfer time, and the number of fast oscillations per transfer.

Step 1. Bare and symmetric-mode frequency: $\omega_+ = \omega_0 = \sqrt{k/m} = \sqrt{20/0.50} = 6.32\ \mathrm{rad/s}$ .

Step 2. Antisymmetric mode: $\omega_- = \sqrt{(k + 2\kappa)/m} = \sqrt{21.6/0.50} = \sqrt{43.2} = 6.57\ \mathrm{rad/s}$ .

Step 3. Splitting, exact and approximate: $\delta = 6.573 - 6.325 = \boxed{0.248\ \mathrm{rad/s}}$ ; the weak-coupling formula gives $\kappa/\sqrt{mk} = 0.80/\sqrt{10} = 0.253\ \mathrm{rad/s}$ — within 2%, confirming the linearization.

Step 4. Transfer time: $T_{\mathrm{transfer}} = \pi/\delta = \pi/0.248 = \boxed{12.7\ \mathrm{s}}$ .

Step 5. Fast cycles per transfer: the carrier period is $2\pi/\bar\omega = 2\pi/6.449 = 0.974\ \mathrm{s}$ , so the energy needs $12.7/0.974 \approx \boxed{13\ \text{oscillations}}$ to change hands — slow enough to watch, fast enough for a lecture.

2 · Detuning kills the demo (exam level)

Two carts on springs, $m = 0.20\ \mathrm{kg}$ and $k = 8.0\ \mathrm{N/m}$ each, coupled by $\kappa = 0.40\ \mathrm{N/m}$ . (a) Find the transfer time when the carts are identical. (b) A student swaps cart 2 for one of mass $0.21\ \mathrm{kg}$ — a 5% error. What fraction of the energy now reaches cart 2, and when does the exchange peak?

Step 1. Identical case: $\omega_0 = \sqrt{8.0/0.20} = 6.32\ \mathrm{rad/s}$ , $\omega_- = \sqrt{8.8/0.20} = 6.63\ \mathrm{rad/s}$ , so $\delta = 0.309\ \mathrm{rad/s}$ and $T_{\mathrm{transfer}} = \pi/0.309 = \boxed{10.2\ \mathrm{s}}$ .

Step 2. The detuning: cart 2's new bare frequency is $\omega_2 = \sqrt{8.0/0.21} = 6.17\ \mathrm{rad/s}$ , so $\Delta = \omega_2 - \omega_1 = 6.172 - 6.325 = -0.152\ \mathrm{rad/s}$ . Only the magnitude matters. A 5% mass error has produced a detuning of the same size as the whole coupling effect.

Step 3. Coupling rate: at resonance the exchange formula must reproduce Step 1, which fixes $\Omega_c = \delta/2 = 0.154\ \mathrm{rad/s}$ .

Step 4. Transferred fraction from (9): $f_{\max} = \dfrac{(0.154)^2}{(0.154)^2 + (0.076)^2} = \dfrac{0.0238}{0.0238 + 0.0058} = \boxed{0.80}$ — a fifth of the energy never leaves cart 1, from a 5% mass mismatch.

Step 5. Peak time: $\Omega_{\mathrm{eff}} = \sqrt{(0.154)^2 + (0.076)^2} = 0.172\ \mathrm{rad/s}$ , so the maximum occurs at $t = \pi/(2\Omega_{\mathrm{eff}}) = \boxed{9.1\ \mathrm{s}}$ — earlier than the resonant 10.2 s. Detuning speeds the exchange up while making it incomplete, exactly the Lorentzian trade-off of equation (9).

Origins · 1 minEnrico Fermi and how this was discovered

Enrico Fermi

1901–1954 · Rome → Chicago, Nobel Prize 1938

Fermi appears in this story for what may be the most consequential "failed" calculation in mechanics. In 1953, with Pasta, Ulam, and programmer Mary Tsingou at Los Alamos, he set the MANIAC computer loose on a chain of 64 oscillators with weak nonlinear coupling, expecting the energy — started in one mode — to spread and equipartition among all the modes, the founding assumption of statistical mechanics. It refused. The energy sloshed among a handful of low modes and then, astonishingly, reassembled almost perfectly in the initial mode: recurrence, not thermalization. The Fermi-Pasta-Ulam problem, published after Fermi's death in 1955, is the direct many-body descendant of this article's two-oscillator slosh — and its resolution seeded two fields, solitons and chaos theory (via the KAM theorem). Fermi called it "a minor discovery." It was the first numerical experiment in the history of physics.

History

1665

— Christiaan Huygens, ill in bed, observes his two pendulum clocks on a common beam locking into anti-phase — the first recorded coupled-oscillator energy exchange.

1877

— Lord Rayleigh's Theory of Sound systematizes beats and coupled vibration, giving the sum-to-product analysis used to this day.

1937

— Isidor Rabi derives the flopping formula for a two-level system in an oscillating field — mathematically the detuned energy-transfer problem, now in quantum dress.

1955

— The Fermi-Pasta-Ulam-Tsingou report reveals that weakly nonlinearly coupled oscillators refuse to thermalize, launching numerical physics, solitons, and chaos.

2007

— An MIT group led by Marin Soljačić transfers 60 W across 2 m via resonantly coupled coils, turning the beat formula into wireless power engineering.

Going further

Texts. Taylor, Classical Mechanics, ch. 11 (Coupled Oscillators and Normal Modes) · Morin, Introduction to Classical Mechanics, coupled-oscillators chapter · French, Vibrations and Waves, ch. 5 · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., ch. 6 · Landau & Lifshitz, Mechanics, §23-24.

Provenance. Verifier should rederive the normal-mode frequencies $\omega_+=\sqrt{k/m}$ , $\omega_-=\sqrt{(k+2\kappa)/m}$ and the splitting $\delta\approx\kappa/\sqrt{mk}$ from the coupled equations (Taylor ch. 11, Morin coupled-oscillators chapter); confirm the beating solution and energy partition $E_1=E_0\cos^2(\delta t/2)$ against Landau-Lifshitz §23-24; and check the detuned transfer fraction $f_{\max}=\Omega_c^2/(\Omega_c^2+(\Delta/2)^2)$ against the standard Rabi formula. All expressions are symbolic textbook results; no fitted numbers used. Worked-example numbers on this page ( $\delta = 0.248\ \mathrm{rad/s}$ and $T = 12.7$ s for the carts; $f_{\max} = 0.80$ and $t_{\mathrm{peak}} = 9.1$ s for the detuned pair) are illustrative arithmetic recomputed from the boxed formulas.

Builds on: Small oscillations about equilibrium · Simple harmonic motion · Amplitude, phase, and angular frequency · Energy in simple harmonic motion · Complex exponential representation of oscillations. Leads to: Resonance · Driven oscillations · Anharmonic pendulum and amplitude dependence · Damped harmonic oscillator.

Why does starting only oscillator 1 excite both normal modes, and why equally?

Because "oscillator 1 displaced, oscillator 2 at rest" is not a shape the system can hold — each normal mode is a pattern spread over both masses. To write the localized start in mode language you must solve $\eta_+ = x_1 + x_2 = A$ and $\eta_- = x_1 - x_2 = A$ : both modes enter with amplitude $A$ , exactly equal. Physically, the symmetric mode wants to move both masses together and the antisymmetric wants to move them oppositely; only the fifty-fifty sum moves mass 1 while cancelling at mass 2. The equality is not a coincidence of this example — any strictly localized excitation of two identical coupled oscillators projects equally onto the two modes, and that equal weighting is precisely what allows the later perfect cancellation that empties oscillator 1 completely.

Is the beat frequency δ or δ/2? Different books seem to disagree

Both, depending on what you track — this is a bookkeeping convention, not a physics disagreement. The amplitude envelope of oscillator 1 is $\cos(\delta t/2)$ , oscillating at angular frequency $\delta/2$ . But the energy goes as the envelope squared, $\cos^2(\delta t/2) = \tfrac12[1 + \cos\delta t]$ , which oscillates at $\delta$ — squaring doubles the frequency because the energy cannot tell a positive envelope from a negative one. What the ear hears in acoustic beats, and what the eye sees in the pendulum demo, is the loudness — the energy — so the perceived beat rate is $\delta$ , and one complete one-way energy transfer takes $\pi/\delta$ . Books quoting $\delta/2$ are describing the envelope; books quoting $\delta$ are describing what you observe.

Where exactly does the energy sit while it is in transit between the oscillators?

Almost all of it is always in one oscillator or the other — the coupling spring itself holds only a fraction of order $\kappa/k$ of the total, negligible by the weak-coupling assumption. Mid-transfer, when $\delta t/2 = \pi/4$ , each oscillator simply holds half the energy: $E_1 = E_2 = E_0/2$ . The transfer mechanism is ordinary work: oscillator 2 lags oscillator 1 by a quarter carrier cycle (the $\sin$ against $\cos$ in equation (5)), and with that phase relationship the coupling force on mass 2 points along mass 2's velocity more often than against it — positive net work every cycle. When the energy flows back, the lag reverses. There is no mysterious reservoir; just a weak force with the right phase, applied patiently over many cycles.

How is this "the classical analogue of Rabi oscillations"? What maps to what?

Write the slowly varying envelope equations (8) next to the Schrödinger equation for a two-level atom in a resonant field and they are term-for-term identical: the envelope pair $(b_1, b_2)$ plays the two-level amplitudes, the detuning $\Delta$ plays the atom-field frequency mismatch, and the coupling rate $\Omega_c$ plays the Rabi frequency. "Probability of finding the atom excited" becomes "fraction of energy in oscillator 2," and both obey $|b_2|^2 = (\Omega_c^2/\Omega_{\mathrm{eff}}^2)\sin^2(\Omega_{\mathrm{eff}}t)$ . The analogy is exact because both systems are two coupled modes evolving linearly. What does not map is interpretation: the classical $|b_2|^2$ is a definite energy fraction you could measure continuously, while the quantum version is a probability revealed only on measurement. The mathematics is shared; the ontology is not.

Do the beats survive damping, and how many transfers can I actually see in a real demo?

The criterion is the race between the damping rate $\gamma$ and the splitting $\delta$ . Each oscillator loses energy as $e^{-2\gamma t}$ (amplitude as $e^{-\gamma t}$ ), so the number of visible transfer cycles is roughly $N \sim \delta/(2\pi\gamma)$ . Lecture pendulums with $Q \sim 500$ and a beat period of thirty seconds show two or three clean sloshes before the whole picture fades. If $\gamma \gg \delta$ the energy dies before completing even one trip — the beating regime disappears, which is why you cannot demonstrate this with heavily damped oscillators like door closers. Engineers state the same condition positively: for wireless power transfer to work, the coupling rate must beat the loss rate, $\Omega_c \gg \gamma$ — the "strong coupling regime" that every cavity-QED and optomechanics paper advertises is exactly this inequality.

Can I speed up the transfer by making the coupling spring stiffer? What breaks if I overdo it?

Yes — $T_{\mathrm{transfer}} = \pi\sqrt{mk}/\kappa$ , so doubling $\kappa$ halves the transfer time. What breaks at large $\kappa$ is the story, not the mathematics. The two-timescale picture — fast carrier, slow envelope — requires $\delta \ll \omega_0$ , i.e. $\kappa \ll k$ . Push $\kappa$ toward $k$ and the "beat" becomes as fast as the oscillation itself: the system no longer looks like energy sloshing between two recognizable oscillators, but simply like two very different normal modes doing their own thing. The exact normal-mode solution remains valid at any coupling strength; it is the interpretation — "the energy is in oscillator 1 now" — that loses meaning when the modes are far apart, because on any timescale you can define, both masses are moving comparably.

My two "identical" lab pendulums never hand over all the energy. How close is close enough?

Equation (9) is your tolerance spec. Complete transfer needs the detuning small compared with the coupling: $f_{\max} > 0.95$ requires $|\Delta| \lesssim 0.46\,\Omega_c$ , i.e. the bare frequencies matched to well within the mode splitting itself. Since $\delta/\omega_0 \approx \kappa/k$ is maybe a percent in a typical demo, you need the pendulum frequencies matched to a fraction of a percent — for $\omega \propto L^{-1/2}$ , lengths matched to a few parts in a thousand, millimetres on a metre. That is why the demo is fussy: worked example 2 shows a 5% mass mismatch already trapping 20% of the energy. The practical recipe is to tune one pendulum's length while watching the first oscillator's minimum amplitude and minimizing it — you are experimentally zeroing $\Delta$ using $f_{\max}$ as the signal.

Does the energy REALLY vanish from oscillator 1 at the beat node, or is a little always left?

For truly identical, undamped oscillators the zero is exact — at $t = \pi/\delta$ the envelope $\cos(\delta t/2)$ passes through zero and oscillator 1 is instantaneously at rest at its equilibrium point with zero velocity: no kinetic, no potential energy. (Its envelope then goes negative, which just means the returning oscillation is phase-flipped.) In any real experiment three things blur the perfect null: residual detuning caps the transfer via $f_{\max} < 1$ and leaves oscillator 1 wiggling at the node; damping shrinks the whole picture; and the small energy resident in the coupling spring, of order $\kappa/k$ , never belongs to either oscillator. But these are perturbations. The interference argument — two equal-amplitude modes in antiphase at mass 1 — cancels exactly, and the cleaner the experiment, the deader the pendulum at the node.

Energy Transfer Between Coupled Oscillators

Where this lives today

The physical picture

Two modes, one beat

The energy ledger

Detuning, resonance, and the classical Rabi formula

Which equation answers which question

Limiting cases

Worked examples

1 · The classroom demo, by the numbers (confidence builder)

2 · Detuning kills the demo (exam level)

Going further

Before this

Where this leads