First Applications: The Vibrating String & the Heated Rod

Two everyday objects, one hidden equation each

Pin a guitar string down at both ends, pull it aside, and let go — it springs into a singing shimmer. Separately, heat the middle of a metal bar and walk away — the warmth spreads, the hot spot cools, and given time the whole bar drifts toward a flat temperature. These look like utterly different stories: one rings, the other goes quiet. Yet they are governed by almost the same mathematics, and Fourier series is the single tool that cracks both open. This guide is where the abstract claim of this rung — 'any signal is a sum of harmonics' — finally earns its keep on real physics.

Each object hides a partial differential equation — an equation relating not an ordinary derivative but several partial derivatives of one unknown function of two variables, here position x and time t. The vibrating string obeys the wave equation, written d^2u/dt^2 = c^2 d^2u/dx^2 (often abbreviated u_tt = c^2 u_xx), where u(x,t) is the sideways displacement of the string at point x and instant t, and c is the wave speed set by tension and mass. The heated rod obeys the heat equation, du/dt = k d^2u/dx^2 (u_t = k u_xx), where u(x,t) is now temperature and k is the thermal diffusivity. Two physical worlds, two PDEs that differ by a single subtle thing: one time derivative versus two.

Two more facts pin each problem down to a single answer. First, the boundary conditions: both ends are clamped. The string is tied at x = 0 and x = L so it cannot move there, u(0,t) = u(L,t) = 0; the rod's ends are held at zero temperature, the same equations. Second, the initial conditions: the shape you start with at t = 0 — the plucked profile of the string, or the starting temperature pattern of the rod. The PDE plus the boundary conditions plus the initial state make a boundary-value problem, and that package has exactly one solution. Our whole job is to find it.

Separation of variables: the engine that makes Fourier necessary

The master move is separation of variables: gamble that the answer factors into a part depending only on space times a part depending only on time, u(x,t) = X(x) T(t). This is a guess, not a law — it cannot represent every function of two variables — but it is exactly the right guess for these clean, linear problems with these simple boundaries. Substitute it into the PDE and a small miracle happens: all the x-stuff piles up on one side, all the t-stuff on the other.

1. Substitute u = X(x) T(t) into the PDE. For the rod: X T' = k X'' T. Divide through by k X T to get T'/(k T) = X''/X.
2. Argue the separation constant. The left side depends only on t, the right only on x, yet they are equal for all x and t — so each must equal one and the same constant, call it -lambda. One PDE in two variables has split into two ordinary differential equations.
3. Solve the space equation X'' + lambda X = 0 with the clamped-end conditions X(0) = X(L) = 0. Only special values lambda = (n pi / L)^2 give a nonzero solution, and that solution is X_n(x) = sin(n pi x / L). The boundaries hand-pick the harmonics.
4. Solve the matching time equation for each n, then superpose all the building blocks u_n(x,t) = X_n(x) T_n(t) into one infinite sum, and use the initial shape to pin the coefficients — which is precisely a Fourier series.

Look at what just happened in step 3, because it is the secret of the whole rung. We did not choose to use sines — the clamped boundaries forced them on us. A function that is zero at both x = 0 and x = L must be built from sin(n pi x / L); cosines refuse, since cos(0) is not zero. The boundary conditions are picking the orthogonal family for us, the very sine basis whose Euler-Fourier formulas you learned earlier. This is why Fourier series and these PDEs are inseparable: the physics of a clamped object literally selects the harmonics, and Fourier's recipe is the only way to find how much of each you need.

The string: standing waves that ring

Now finish the string. Its space part is X_n(x) = sin(n pi x / L) as derived. Its time part comes from the wave equation's two time derivatives, giving T_n'' + (c n pi / L)^2 T_n = 0 — the equation of a simple oscillator. So T_n(t) is a sine-and-cosine wobble in time at frequency omega_n = c n pi / L. Each building block u_n = sin(n pi x / L) [A_n cos(omega_n t) + B_n sin(omega_n t)] is a standing wave: a fixed sine-arch shape in space whose height pulses up and down in time, never travelling, just breathing in place.

The full motion superposes all of them: u(x,t) = sum over n of sin(n pi x / L)[A_n cos(omega_n t) + B_n sin(omega_n t)]. The slowest mode n = 1 sets the pitch you hear — the string's fundamental frequency omega_1 = c pi / L — and the faster modes n = 2, 3, ... are its overtones, the harmonics that give a violin and a flute their different timbres on the same note. To fix the amounts A_n, set t = 0 and demand the shape match the plucked profile f(x): then f(x) = sum A_n sin(n pi x / L), which is just the Fourier sine series of the initial shape. The A_n are its Fourier coefficients; the B_n come the same way from the initial velocity.

VIBRATING STRING -- length L, ends fixed, plucked into shape f(x), released from rest

  PDE:        u_tt = c^2 u_xx
  Boundary:   u(0,t) = u(L,t) = 0          (tied at both ends)
  Initial:    u(x,0) = f(x),  u_t(x,0) = 0  (plucked, then let go)

  Separate u = X(x) T(t)  ->   X'' + lambda X = 0 ,  T'' + c^2 lambda T = 0
  Boundaries force         X_n = sin(n pi x / L),  lambda_n = (n pi / L)^2
  Time part (released from rest) ->  T_n = A_n cos(omega_n t),  omega_n = c n pi / L

  Solution (superpose the standing waves):
      u(x,t) = sum_{n>=1} A_n sin(n pi x / L) cos(c n pi t / L)

  Coefficients = Fourier sine series of the pluck:
      A_n = (2/L) integral_0^L f(x) sin(n pi x / L) dx

The rod: the same modes, but fading instead of ringing

Now the rod, and watch where it diverges from the string. The space part is identical — X_n(x) = sin(n pi x / L), the very same sine modes, because the clamped boundaries are the same. The only change is the time equation. The heat equation has one time derivative, not two, so separation gives T_n' = -k (n pi / L)^2 T_n — a first-order equation whose solution is a decaying exponential, T_n(t) = exp(-k (n pi / L)^2 t). One derivative versus two is the whole difference between music and silence: the string's modes oscillate forever, the rod's modes simply melt away.

Superpose and you get u(x,t) = sum over n of B_n sin(n pi x / L) exp(-k (n pi / L)^2 t). Read the exponent: the decay rate is k(n pi / L)^2, which grows like n-squared. So the high harmonics — the sharp, wiggly, fine-detail modes — die fastest, while the slow n = 1 hump lingers longest. That is exactly why heat smooths things out: every jagged feature is high-harmonic, and high harmonics evaporate almost instantly, leaving the bar smooth and bland long before it is uniformly cold. To find the B_n, again set t = 0: the initial temperature f(x) must equal sum B_n sin(n pi x / L), so the B_n are once more Fourier sine coefficients of the starting shape.

Why it works, where it stops, and where it leads

The logic that lets us add infinitely many standing waves and still solve the PDE is the superposition principle: because both equations are linear, any sum of solutions is again a solution. We solve one harmonic in isolation — trivial, since a single mode is just an oscillator or a decaying exponential — and then trust that superposing them all gives the true motion, with the initial shape's Fourier coefficients as the mixing weights. Solve-one-mode-then-superpose is the same divide-and-conquer you met with linear ODEs, now lifted to functions of space and time. Fourier series is precisely the machine that does the dividing.

Step back and savour what you can now do. Two of the great equations of physics — wave and heat — yield to one disciplined routine: separate the variables, let the boundaries hand you the sine modes, solve each mode's tiny time equation, and let the initial shape's Fourier coefficients weight the sum. The string rings because its modes oscillate; the rod smooths because its modes decay; and the single number that distinguishes them is whether time enters once or twice. From here the road climbs in two directions. When the geometry stops being a flat interval — a vibrating drumhead, heat in a cylinder — the sine modes give way to other orthogonal families like the expansions over special functions you have met, all governed by the same project-onto-a-basis idea. And when the object is infinite rather than clamped, the discrete harmonics blur into a continuum and the Fourier series becomes the Fourier transform. You have just watched an abstract series become the language of the physical world.