Domain of Dependence and Finite Speed

Reading the geometry hidden in d'Alembert

The previous guide handed us d'Alembert's formula, the exact solution of the wave equation u_tt = c^2 u_xx on the whole line. With initial displacement u(x,0) = f(x) and initial velocity u_t(x,0) = g(x), it reads u(x,t) = (1/2)[ f(x - c t) + f(x + c t) ] + (1/(2c)) integral of g from x - c t to x + c t. We treated it as an answer to compute with. Now we ask a sharper question: when you stand at one point (x,t) and read off that number, exactly which pieces of the starting data did you consult? The formula already tells you — you just have to look at the inputs.

Stare at where the data actually enters. The displacement f is only ever sampled at the two endpoints x - c t and x + c t. The velocity g is integrated, but only across the closed interval [x - c t, x + c t] — nothing outside it contributes a thing. So the value at (x,t) is built entirely from the initial data on that one finite segment of the x-axis, a segment of width 2 c t centred on x. Move your observation point and the segment slides and stretches, but it always stays finite. This segment is the domain of dependence of the point (x,t).

The backward triangle: what your past can reach

Draw spacetime with x across and t up. From the point (x,t) send two lines downward to the x-axis: one with slope corresponding to speed c going left, one going right. They land at x - c t and x + c t — exactly the endpoints of the domain of dependence. Together with the segment they cut off on the axis, those two lines bound a triangle hanging beneath (x,t). Everything that can possibly influence the value there lives inside this backward triangle, and nothing outside it can. That is the whole content of finite speed, drawn as a picture.

                (x, t)            observation point
                  /\
                 /  \             slopes are +-1/c
                /    \            (left- and right-going
               /      \           characteristics)
              /        \
  ----[------+----------+------]----------------> x   (t = 0)
      x-ct                x+ct
      <------ width 2ct ------>
        the domain of dependence

A small worked check makes it concrete. Take c = 2 and ask for the solution at the point x = 5, t = 3. The domain of dependence runs from 5 - 2·3 = -1 to 5 + 2·3 = 11, a segment of width 12 = 2 c t. If you change the initial pluck f or push g only out near x = 50, the solution at (5, 3) does not flinch — that disturbance is a full 39 units of x away and has had only 3 units of time, so at speed 2 it could travel at most 6 units. It simply has not arrived yet.

The forward triangle: what you influence

Now flip the picture. Pick a single point x0 on the initial line and ask the reverse question: which future points (x,t) does its data reach? Point (x,t) sees x0 exactly when x0 lies in its domain of dependence, i.e. when x - c t <= x0 <= x + c t, which rearranges to |x - x0| <= c t. That is a forward triangle opening upward from x0, with the same characteristic slopes. Inside it the data at x0 can be felt; outside it cannot. This upward wedge is the range of influence of the point x0.

Domain of dependence and range of influence are two readings of one fact, the same pair of characteristic lines seen from opposite ends. The first looks down from a spacetime point and asks what it inherits; the second looks up from an initial point and asks what it bequeaths. Both are bounded by the lines x ± c t = const, and that bound is precisely the wave speed c expressed as the steepest slope a signal can take in the (x,t) plane. No information outruns its own characteristic.

Why this is the opposite of the heat equation

Climb back to the heat rung for a moment and the contrast is stark. The solution of the heat equation u_t = k u_xx at any point and any positive time is an integral of the initial data against the heat kernel, a bell curve that is strictly positive everywhere on the line. So heating a single cold spot anywhere — even a mile away — raises the temperature here instantly, by an immeasurably tiny but genuinely nonzero amount. The heat equation has infinite propagation speed and no domain of dependence: its honest answer is "everything depends on everything."

The wave equation could not be more different, and the difference is exactly the type. Both are second-order and linear, but the heat equation is parabolic and the wave equation is hyperbolic — and this is one of the sharpest payoffs of the finite-versus-infinite speed distinction you met when we classified equations. Hyperbolic equations have real characteristics and therefore a finite wave speed and a strict horizon; parabolic ones have none, and gossip travels everywhere at once. The same machinery — write the solution as data hitting a kernel or a formula — exposes opposite physics.

There is a second contrast worth stating plainly, because beginners often expect smoothing from any diffusion-shaped equation. The heat kernel is infinitely smooth, so even spiky initial data becomes silky instantly — that is its smoothing effect. The wave equation does the opposite: d'Alembert just translates and superposes f and g, so a corner stays a corner, a jump stays a jump. A waveform you launch keeps its shape (in 1D) and its sharp features forever. Finite speed and no smoothing are two faces of the same hyperbolic coin.

Why finite speed matters: well-posedness and a checklist

Finite speed is not just a pretty picture; it is what makes the problem trustworthy. Because each (x,t) leans on only a bounded chunk of data, a small change in the initial pluck f or push g — confined to some region — changes the solution only inside that region's range of influence, and by a controlled amount. That is the heart of continuous dependence on data, the third leg of being a well-posed problem alongside existence and uniqueness. The domain of dependence even gives a clean proof of uniqueness: if two solutions share data on a base segment, they must agree throughout the triangle above it.

1. Locate the point. Fix the spacetime point (x,t) you care about, and recall the speed c from the equation u_tt = c^2 u_xx.
2. Drop the two characteristics. From (x,t) trace the lines x - c t and x + c t down to the t = 0 axis; they land at x - c t and x + c t.
3. Read off the base segment. The interval [x - c t, x + c t], of width 2 c t, is the domain of dependence — the only data that matters.
4. Flip for influence. To find what a source point x0 affects, open the same triangle upward: all (x,t) with |x - x0| <= c t form its range of influence.

Be honest about the boundaries of this clean story. The crisp triangles above are the 1D picture; in higher dimensions the backward set becomes the characteristic cone, and in odd dimensions like 3D something even sharper happens — the dependence collapses onto the cone's surface rather than filling its interior, which is Huygens' principle and the subject of the final guide in this rung. Finite speed itself survives in every dimension and is a robust feature of well-behaved hyperbolic equations; but the simple flat-shape, no-distortion travelling wave is special to one space dimension, so do not carry that part upstairs.