A streak of dye on a moving river
You have climbed up through single-variable calculus, met partial derivatives, and solved ordinary differential equations. Now an unknown depends on two things at once — position x and time t — and we write it u(x,t). Before any of the famous equations of this subject, start with the gentlest one of all. Picture a long, thin river with a streak of dye floating on it. The water slides steadily to the right and the dye neither spreads nor fades. A photo taken one second from now looks exactly like the photo now, only shifted a little downstream. That picture is the transport equation, and the whole of this rung grows out of it.
Let u(x,t) be the amount of dye at position x and time t. The transport equation says it is purely carried along: u_t + c u_x = 0, where c is the (constant) flow speed. Read the two terms against each other. u_t is how fast the dye amount changes at a fixed point as time ticks; u_x is how steeply the dye profile rises or falls across space. The equation insists these exactly cancel — the local change in time is nothing but the spatial slope being swept past you at speed c. Nothing is created, destroyed, or smeared; the pattern is only relocated. Because the medium does the carrying, this is also called the advection equation.
The answer is a travelling wave
Here is the punchline, and it is beautiful: the solution is u(x,t) = f(x - c t), where f is the starting profile, set by the initial condition u(x,0) = f(x). The graph of u at time t is the graph of f slid bodily to the right by the distance c t — same height, same width, same shape, just moved. This is a travelling-wave solution: a frozen profile gliding across space at the steady speed c. Everything about u sits inside the single combined variable x - c t, often called the characteristic variable. That combination is the seed of the next guide, the method of characteristics.
Let us check it really works, because seeing the cancellation cements the idea. If u(x,t) = f(x - c t), then by the chain rule u_t = -c f'(x - c t) and u_x = f'(x - c t). Plug into the equation: u_t + c u_x = -c f' + c f' = 0. It vanishes for any differentiable f whatsoever. That last word is the honest and important part: a single ODE pins its solution down with a few constants, but a PDE leaves a whole arbitrary function free. Here every shape f is a general solution until you say what the dye looks like at t = 0. The data is what selects one f from the infinitely many.
u_t + c u_x = 0, u(x,0) = f(x) --> u(x,t) = f(x - c t) t = 0: f( . ) bump centred at x = 0 t = 1: f( . - c ) same bump, now at x = c t = 3: f( . - 3c ) same bump, now at x = 3c
A tiny worked example
Take c = 2 and a single Gaussian bump as initial data, u(x,0) = e^(-x^2), centred at x = 0. The recipe says u(x,t) = f(x - 2t) = e^(-(x - 2t)^2). Now read off snapshots. At t = 1 the peak — where the exponent is zero — sits at x - 2 = 0, that is x = 2. At t = 3 it sits at x = 6. The bump keeps its height of 1 and its width unchanged at every instant; it simply rides the current at speed 2. That is the entire behaviour of the equation, made concrete.
It is worth tracing one path by hand. Suppose you walk along the curve x(t) = x0 + c t, keeping pace with the flow. The dye value you see, U(t) = u(x(t), t), changes at rate dU/dt = u_t + (dx/dt) u_x = u_t + c u_x, which the equation makes exactly zero. So u is constant along each such line — these are the characteristic lines x - c t = constant — and that single observation is the whole answer in disguise. This is the seed the next guide grows into a method for far harder equations. For the linear first-order equation here, the lines are fixed straight tracks, drawn before the solution is even known.
Where it sits, and what to watch for
Be careful not to confuse u_t + c u_x = 0 with the second-order wave equation u_tt = c^2 u_xx that you will meet a few rungs up. The transport equation is one factor of it — it carries a signal in only one direction, while the wave equation carries left-movers and right-movers at once. Transport is the simplest example of a finite-speed, non-smoothing process: a sharp corner in f stays a sharp corner forever, just relocated. That is the opposite of the diffusion equation u_t = k u_xx, which blurs every corner instantly. Honest framing: transport preserves features; diffusion erases them.
Two honest caveats. First, the clean answer u(x,t) = f(x - c t) relies on c being a constant. If the flow speed varies, c(x,t), or — much more dramatically — if the speed depends on u itself, the tidy straight tracks bend, fast parts overtake slow parts, and the smooth picture can break down into a shock. That genuinely nonlinear story is the climax of this rung. Second, a corner in f gives a u that satisfies the equation everywhere except along that one moving corner; making sense of such non-smooth solutions is exactly why later rungs build weak and distributional solutions. For now, enjoy the cleanest case.