Motivation
A string needs both its starting shape and its starting velocity. Two strings can occupy the same straight position while one is moving upward and the other is motionless. Their later displacements differ, so initial position alone cannot determine a wave. This is the first consequence of having a second derivative in time.
A second consequence is a finite speed of communication. A disturbance at a distant point cannot influence an observation point immediately. For the one-dimensional wave equation, the relevant geometry is a cone in space and time. Its base selects the initial data that matter, and its interior selects the forcing that can reach the observation point. The formula for a forced wave records this geometry exactly, rather than merely suggesting it in a picture.
We first construct the solution and then prove the finite-cone statement a second way using energy. The two arguments answer different questions. The formula computes a displacement; the energy calculation explains why information outside a characteristic boundary cannot enter the region it controls. Duhamel's formula then describes a source as a succession of velocity increments, each propagated for the time remaining.
Definitions
Fix a wave speed and a time horizon . The forced Cauchy problem on the line is
The function is displacement, is initial velocity, and supplies acceleration in this normalization. In this note a classical solution is up to the closed strip: all derivatives of total order at most two extend continuously to its temporal endpoints, with temporal derivatives taken from within the strip. In particular the initial traces are and . We use , and . These convenient assumptions justify the differentiations below; no bound on spatial growth is needed for the finite integrals defining a solution at one point.
The lines and are characteristics. They are the paths of right-moving and left-moving signals. For a point , with , its closed backward cone is
At time zero the cone has base ; at time it closes at the observation point. A domain-of-dependence statement specifies which data must agree to make two solutions agree at that point. A support statement is a special case obtained by comparing a solution with the zero solution.
For classical solutions whose spatial support is contained in a fixed compact interval throughout , define
This class makes the integrations by parts below immediate. The same energy theory can be extended to less regular finite-energy data, but that extension is not assumed here. The expression measures velocity and spatial deformation; an initial displacement condition is still part of specifying the physical state.
Theorem/Proposition
Theorem 4.1.1
(Classical solution with a source)
Under the stated regularity assumptions, the unique classical solution on is
The first two terms are d'Alembert's formula. The last term is Duhamel's formula for zero initial displacement and velocity.
Theorem 4.1.2
(Finite domain of dependence)
Suppose two classical solutions have identical displacement and velocity data on , and their source functions agree on . Then the solutions agree throughout that cone, including its vertex. In particular, if and vanish outside for , then at time the solution vanishes outside .
Both kinds of initial data occur in this statement. The source condition is also essential: zero initial data do not prevent a source from creating a disturbance later. To prove agreement throughout the cone from the formula, note that every smaller backward cone based at a point of the original cone is contained in it, by the triangle inequality.
Theorem 4.1.3
(Energy supplied by a force)
For a real classical solution in the compact-support class above, including the continuous derivative traces at initial time,
With zero source, energy is constant. Applied to a difference of two solutions, the inequality controls changes in velocity and spatial derivative using differences in their initial state and source.
The spatially periodic examples later in the note need not have finite total energy on the whole line. They are covered by the pointwise solution theorem, which uses finite intervals, rather than by this particular global energy statement. This separation prevents a valid explicit example from silently acquiring an inapplicable integral estimate.
Proof sketch or proof idea
Proof
(Characteristic coordinates and differentiation of Duhamel’s term)
For the homogeneous equation, use characteristic coordinates and , writing . The differential operator becomes
Thus a homogeneous solution is a sum . The initial conditions give and . Solving for the derivatives and integrating gives the first two terms of the formula. Any homogeneous solution with zero initial displacement and velocity is zero, so the representation also proves uniqueness without an assumption at spatial infinity.
Now let denote the double integral in the theorem. Its inner interval shrinks to length zero when . Differentiation therefore yields
Differentiating once more produces a new endpoint contribution:
The integral term equals . Hence , while . All differentiations take place over compact intervals within a fixed backward cone, where the continuous derivatives are bounded. Adding the homogeneous solution proves existence. Subtracting two candidate solutions leaves the already solved homogeneous zero-data problem, proving uniqueness.
Concept lensStructural
A source adds velocity increments
Let be the homogeneous solution with zero initial displacement and initial velocity . Then Duhamel's term is . The slice at time contributes a velocity increment, which travels for the remaining time . The facts and explain why the source appears when the integral is differentiated twice, rather than once.
The formula already proves finite dependence, but a local energy proof shows what fixes the speed.
Proof
(Energy in a shrinking interval)
Let be a difference whose source is zero inside the backward cone and whose initial data are zero on its base. For , set
and let be one half of the integral of from to . Differentiating this moving-interval integral gives an interior flux term and the two endpoint-motion terms. Since and , completing squares gives
The initial energy is zero because both initial velocity and the derivative of the zero initial displacement vanish on the base. Nonnegativity makes identically zero. Thus both derivatives vanish in each open shrinking interval. Along the central vertical line, ; integrating from the initial value zero and then taking the limit to the vertex proves . Repeating the argument at smaller cones gives the full dependence statement.
Proof X-ray
Why a shrinking interval proves finite speed
Global energy conservation alone does not specify where a disturbance can travel. The local argument puts the boundary on characteristic lines. Their speeds turn each endpoint contribution into a negative square, so energy cannot enter this shrinking region from outside. Zero derivative energy then gives zero time variation; the initial displacement fixes the remaining constant. That last condition must not be dropped merely because the energy contains derivatives.
Proof
(Forced energy estimate)
For the global forced-energy estimate, differentiate , substitute the equation, and integrate the spatial second derivative by parts. Compact support removes the boundary term, leaving . Cauchy–Schwarz gives . Dividing by is unsafe when the initial energy is zero. Instead differentiate for ; its derivative is at most . Integrate and let . This proves the estimate, including the important zero-initial-energy case.
Worked examples
Worked example 4.1.4
(Selecting a single direction of travel)
Let be and set , and . Substituting in d'Alembert's formula gives
The initial velocity cancels the left-moving part. A concrete compactly supported choice is for and zero otherwise. The function and its first two derivatives join continuously to zero at both endpoints. Its support at time is . Specifying a shape alone would instead split it equally into two oppositely moving copies when the initial velocity is zero.
Worked example 4.1.5
(Finite speed does not mean that only the fronts are nonzero)
Take , and for , with elsewhere. This velocity is , and
If and , the interval contains the whole velocity support. Consequently in that central region. Outside it is zero. The disturbance has a finite outer edge, yet behind the fronts there is a nonzero displacement plateau. Its derivatives vanish in the plateau, so this description is consistent with energy traveling outward. A finite-speed theorem does not assert that every point returns to zero immediately after a front passes.
Worked example 4.1.6
(A bounded resonant force can produce growing displacement)
Let , put , and take zero initial data with . A sine mode with zero displacement and velocity profile evolves as . Duhamel therefore gives
For the integral, use ; the second term integrates to zero. The forcing has absolute value at most one, but the displacement at suitable points and times grows linearly. Resonance accumulates the repeated input. The example is spatially periodic and is not being assigned a finite whole-line energy.
Counterexample mode
Initial displacement alone gives neither uniqueness nor a height bound
With and initial displacement zero, both and solve the wave equation. Their initial velocities differ. Thus a maximum principle bounding all later displacement by the initial displacement is false, and a uniqueness claim omitting velocity is false for the same elementary reason. The repaired initial-value problem prescribes both and ; its stability estimates control that complete state.
The second counterexample concerns the source rather than the initial state.
Counterexample mode
A source can act where no initial wave was present
Zero initial displacement and velocity with the everywhere-present source give . The solution becomes nonzero everywhere at positive time. This does not violate finite speed: the source is already present at every location, including inside every backward cone. A support conclusion based only on initial data requires the source to vanish, or requires its support to be included in the cone calculation.
Common mistakes
Common mistake
(The factor and the limits change together)
The factor and the expanding integration limits belong together. Changing the speed changes both. The time derivative of the velocity integral contains endpoint terms; omitting them loses the initial velocity and usually destroys the equation. For forcing, the inner interval uses , not , because a contribution created at time has less time to travel.
Common mistake
(A permitted contribution need not be visible)
Finite propagation is a restriction on which data can matter, not a claim that every permitted contribution is nonzero. Cancellation can occur, and a smooth source may vanish along the first geometrically permitted arrival. The cone gives an upper bound on communication speed; it does not guarantee a visible signal at every boundary point of the cone.
Common mistake
(Conserved energy is not constant displacement)
Conserved energy does not imply that the displacement at each point is constant. Energy may move from one region to another, and the same total can be distributed differently between velocity and deformation. Forced energy is not generally conserved: the sign and size of describe work supplied to or removed from the wave.
Summary
The wave state consists of displacement and velocity. d'Alembert's formula propagates that state along two characteristic directions, while Duhamel's formula adds source contributions from the backward cone. The shrinking-cone energy identity proves finite speed independently of the explicit representation. A global energy estimate controls the derivative state when its integrability and support hypotheses hold; it does not replace those hypotheses or a displacement datum.
Exercises
Quick check
(Solve the unforced problem with and , where . Verify both initial conditions and describe its time behavior.)
Combine the two traveling terms.
Solution
The cosine addition identity gives . At zero it equals and its time derivative is zero. Direct differentiation gives . This is an oscillating standing mode, with no decay factor; it is different from the exponential damping of a heat mode.
Quick check
(Suppose the initial data are zero and a classical source vanishes before time and outside the spatial interval , where . For , prove that the solution vanishes outside .)
Check whether a backward cone can meet the source support.
Solution
If , then for every the left endpoint satisfies . The source integral is therefore zero. The same argument on the left handles . Contributions at vanish by assumption, and the initial terms vanish as well. This proves the stated support containment without asserting nonzero values throughout the allowed interval.
Quick check
(With zero initial data and spatially constant source , evaluate Duhamel's formula and check the resulting ordinary differential equation.)
The inner interval has length .
Solution
Duhamel reduces to . It has , and , so it satisfies the forced PDE. The cancellation of the speed comes from multiplying the interval length by .
Quick check
(Two classical compact-support waves have identical initial data and sources differing by at most in norm at each time. Bound the energy of their difference, and explain why initial displacement still matters in concluding equality when .)
Apply the energy estimate to the difference, starting from zero derivative energy.
Solution
For the difference , and the theorem gives , hence . If , both and vanish, so the difference is a constant. The common initial displacement makes that constant zero; in the stated compact-support class, a nonzero constant is also excluded by support. The derivative-energy expression itself does not encode an absolute displacement level.