One equation, two machines
Hang a mass on a spring, pull it down, let go. Newton's second law says mass times acceleration equals the forces, and there are three: the spring pulling back in proportion to how far you have stretched it (-k x), a drag that fights motion in proportion to speed (-c dx/dt), and whatever you push with from outside (F(t)). Collect them and you get m d^2x/dt^2 + c dx/dt + k x = F(t). That is the whole of this guide — a single second-order linear equation, with constant coefficients, the kind whose homogeneous part you already tamed with the characteristic equation earlier in this rung.
Now build a circuit instead: a coil (inductance L), a resistor R, and a capacitor C in a loop, driven by a voltage V(t). Add the voltage drops with Kirchhoff's law and you get L d^2q/dt^2 + R dq/dt + q/C = V(t), where q is the charge on the capacitor. Look closely — it is the *same equation*. The mass m became inductance L, the drag c became resistance R, the spring stiffness k became 1/C, the displacement x became charge q. This is not a loose analogy; the two systems are governed by identical mathematics, so every fact you learn about the spring is simultaneously a fact about the circuit. That economy is exactly why linear ODEs are worth studying so carefully.
The spring left alone: free oscillation
Start with no drag and no push: m d^2x/dt^2 + k x = 0. Rearranged, d^2x/dt^2 = -(k/m) x — the acceleration always points back toward the center, harder the farther out you are. The characteristic equation m r^2 + k = 0 has purely imaginary roots r = plus-or-minus i omega_0, with omega_0 = sqrt(k/m), and the solution is x(t) = A cos(omega_0 t) + B sin(omega_0 t), a pure sinusoid that swings forever. The constant omega_0 is the natural frequency — the rhythm the system rings at when you pluck it and walk away. A stiffer spring or a lighter mass rings faster; that single number is the system's signature.
It is worth seeing why a real exponential and a cosine are the same animal here. When the characteristic roots came out complex, r = plus-or-minus i omega_0, the natural building block was the complex exponential e^{i omega_0 t}. Euler's formula e^{i theta} = cos(theta) + i sin(theta) unpacks it into a cosine and a sine, so the two real solutions cos(omega_0 t) and sin(omega_0 t) are just the real and imaginary parts of one rotating complex number. Physicists and engineers keep the complex form to the very end and take the real part only at the last step — it makes the algebra of amplitude and phase almost trivial, as we will see when forcing arrives.
The two constants A and B are exactly the freedom a second-order equation always carries — recall from earlier that the order counts how many starting facts you must supply. Here they are fixed by where the mass starts and how fast it is moving, x(0) and dx/dt(0). Equivalently, write the solution as a single shifted cosine x(t) = R cos(omega_0 t - phi); then R is the amplitude (how far it swings) and phi is the phase (where in the cycle it begins). Energy sloshes back and forth between the stretched spring (potential) and the moving mass (kinetic), the total never changing, because nothing yet drains it.
Damping: where the energy goes
Real springs slow down; real circuits lose heat. Put the drag back in: m d^2x/dt^2 + c dx/dt + k x = 0, the damped harmonic oscillator. The characteristic equation is m r^2 + c r + k = 0, and its discriminant c^2 - 4 m k decides everything about the system's character. The quadratic formula gives r = (-c plus-or-minus sqrt(c^2 - 4mk)) / (2m), and three qualitatively different things can happen depending on the sign under that square root.
- Underdamped (c^2 < 4mk): the roots are complex, r = -gamma plus-or-minus i omega_d with gamma = c/(2m). The solution is x(t) = R e^{-gamma t} cos(omega_d t - phi) — it still oscillates, but inside a shrinking envelope e^{-gamma t}. A plucked guitar string, a child's slowing swing. Note the ringing frequency omega_d = sqrt(omega_0^2 - gamma^2) is slightly *lower* than the natural omega_0; damping makes it swing a touch slower as well as die away.
- Critically damped (c^2 = 4mk): the discriminant vanishes and the two roots merge into one repeated root r = -c/(2m). A repeated root forces a second solution that carries an extra factor of t, so x(t) = (A + B t) e^{-gamma t}. The mass slides home without overshooting, and crucially it does so in the *shortest possible time* — which is why good door closers and car shock absorbers are tuned right here, at the knife's edge between oscillating and crawling.
- Overdamped (c^2 > 4mk): both roots are real and negative, so x(t) = A e^{r1 t} + B e^{r2 t} decays with no oscillation at all — the system oozes back to rest like a spoon in honey, slower than the critical case. A screen door with too strong a closer that takes an age to shut.
Whatever the regime, notice the unifying fact: with any positive damping the homogeneous solution carries a decaying factor e^{-gamma t} and therefore *fades to zero* as time runs on. This x_h(t) is called the transient — it remembers the initial conditions, rings briefly, and then is gone. Hold that thought, because it is the key to understanding what a steady push does to the system.
Driving it: transient versus steady state
Now push the mass with a rhythm of your own: take a sinusoidal force F(t) = F_0 cos(omega t), driven at *your* frequency omega, which need not match the system's natural omega_0. The full solution is the homogeneous transient plus a forced particular piece, x(t) = x_h(t) + x_p(t). To find x_p the cleanest tool is the method of undetermined coefficients: since the right-hand side is a sinusoid at frequency omega, guess that x_p is a sinusoid at the *same* frequency, x_p(t) = R cos(omega t - phi), and solve for the amplitude R and phase lag phi by substituting back in.
Here is where the split pays off. The transient x_h dies away (we just saw it carries e^{-gamma t}), so after a while only x_p survives. That surviving piece is the steady state: the system has forgotten how it started and now simply swings along *at the driving frequency omega*, not its own omega_0. The takeaway is genuinely useful — drive any damped linear system long enough and it locks onto your rhythm. The early messy motion, where the system's natural ringing fights your push, is the transient; the clean eventual rhythm is the steady state.
The amplitude of the steady state is the prize. Substituting the guess gives R = F_0 / sqrt((k - m omega^2)^2 + (c omega)^2). Read this as a function of the driving frequency omega: it tells you, push by push, how big a response you get for a given push. The ratio of output amplitude to input — the transfer function — is the heart of how engineers think about every filter, antenna, and shock mount. And look where that denominator gets small: when m omega^2 approaches k, i.e. omega approaches omega_0 = sqrt(k/m). That is the doorway to resonance.
Resonance, and the gentler cousin: beats
Drive a system right at its natural frequency, omega = omega_0, and the spring agrees with your push at every instant: you add energy in the same direction it is already moving, push after push after push. With light damping the steady-state amplitude R balloons to roughly F_0/(c omega_0) — large, and growing without bound as the damping c shrinks toward zero. This is [[calc-resonance|resonance]]: a small periodic force, applied at exactly the right rhythm, builds an enormous response. It is how a tiny radio signal at exactly the dial frequency is picked out from the air while all others are ignored, how a singer's note shatters a wine glass, and the mechanism blamed for the Tacoma Narrows bridge's wild oscillations before it collapsed.
Now drive it *close to* but not exactly at the natural frequency — omega near omega_0, with negligible damping. Superpose the natural ringing and the forced response, two cosines of nearly equal frequency, and a trigonometric identity turns their sum into a fast oscillation at the average frequency wrapped inside a slow envelope at the *difference* frequency (omega_0 - omega)/2. The amplitude swells and fades, swells and fades. These are [[beats|beats]]: the throbbing waxing and waning you hear when two slightly mistuned guitar strings sound together, the slow wah-wah you tune away by making the beats slower and slower until they vanish and the strings agree. Beats are resonance's gentler cousin — the system trading energy back and forth with the drive because it cannot quite keep in step.
Putting it together, and where it leads
Step back and the whole story is one picture. The complete motion is always x(t) = x_h(t) + x_p(t): a transient that remembers your starting conditions and (with damping) fades, plus a steady response that endures as long as the drive does. Free, damped, driven, resonant, beating — these are not five separate topics but five readings of the same equation, distinguished by what you switch on. The diagram below maps the territory at a glance.
m x'' + c x' + k x = F(t) omega_0 = sqrt(k/m) gamma = c/(2m)
F = 0, c = 0 free x = R cos(omega_0 t - phi) rings forever
F = 0, c > 0 transient x = R e^{-gamma t} cos(omega_d t - phi) fades
underdamped c^2 < 4mk (oscillates, decaying)
critical c^2 = 4mk (fastest return, no overshoot)
overdamped c^2 > 4mk (creeps back, no oscillation)
F = F0 cos(omega t) x = [transient] + [steady state at omega]
amplitude R = F0 / sqrt((k - m omega^2)^2 + (c omega)^2)
omega ~ omega_0 R blows up -> RESONANCE
omega near omega_0, c~0 slow envelope -> BEATSThere is an even slicker route to all of this that a later guide develops in full: the Laplace transform turns the whole differential equation, initial conditions and all, into ordinary algebra in a variable s, where the transfer function 1/(m s^2 + c s + k) appears directly and resonance shows up as poles creeping toward the imaginary axis. For now, the hand-method picture you have built is the one to keep: guess a sinusoid, read off amplitude and phase, watch the denominator.
And the next rung up generalizes the whole thing: couple two springs, or a chain of them, and one scalar equation becomes a coupled *system* of linear ODEs. The natural frequencies you found here split into several normal modes, each ringing at its own rate, and the elegant tool for solving them all at once is the matrix exponential e^{A t} — the subject of the guide that follows. Everything in this guide is the one-dimensional seed of that larger flower.