Motivation
Finite propagation says that distant data cannot affect an observation before a wave has had time to travel there. It does not say that an observation becomes quiet once the first disturbance has passed. Whether a signal has a tail depends on more than its propagation speed. Even the constant-coefficient wave equation behaves differently in two and three spatial dimensions.
In three dimensions, the solution can be constructed from averages on an expanding sphere. In two dimensions, projecting that sphere gives a weighted average over an entire disk. A compact initial disturbance can therefore remain inside the region sampled by a two-dimensional observer long after the leading front has passed. This geometric difference produces a precise contrast between eventual silence and a decaying tail.
We derive both formulas, verify their initial traces and prove uniqueness by energy on a shrinking ball. Compared with the one-dimensional formula, the new feature is the geometry of the averaging surface and what survives its projection. We discuss dimensions two and three, without claiming a classification of every dimension or every wave operator.
Definitions
All data and solutions here are real-valued. Let and . Consider
Our classical class consists of functions up to time zero on every finite closed time strip, with one-sided temporal derivatives there. In particular the initial traces of and are and . No boundedness or finite whole-space energy is required for the uniqueness argument below; its energies are over bounded balls.
Definition 4.3.1
(Spherical mean in three dimensions)
For , put
Here is the unit sphere, with area . For positive radius this is equivalently the integral over the sphere centered at , divided by its area. Define .
The spatial mean and the velocity propagator are different objects. Multiplication by time and, for displacement data, a time derivative will be essential. Averages alone do not automatically solve the wave equation.
Theorem/Proposition
Theorem 4.3.2
(Kirchhoff construction in three dimensions)
If and , the unique classical solution is
For positive time, the same formula is
The derivative acts on the whole bracket, including its moving sphere and prefactor. The unit-sphere version defines the regular extension at time zero.
Theorem 4.3.3
(Poisson construction in two dimensions by descent)
If and , define
Then the unique classical solution is . For positive time,
The edge singularity is integrable. Derivatives are justified using the fixed unit disk, not by casually differentiating a singular moving boundary.
Theorem 4.3.4
(Cone uniqueness and dimension-dependent clearing)
In either dimension, if two classical solutions have identical initial displacement and velocity throughout the open ball , they have the same value at . In particular the classical Cauchy solution is unique, and compactly supported data propagate no faster than .
In three dimensions, if both data are supported in , the solution is zero wherever
Thus every fixed observer sees zero after . In two dimensions, zero displacement and a nonnegative nonzero smooth compactly supported velocity instead give a strictly positive solution at every fixed observer for all sufficiently large times. This does not violate finite propagation.
Proof sketch or proof idea
Proof
(Spherical means solve the three-dimensional problem)
The spherical mean satisfies a radial identity. Differentiating in radius and using the divergence theorem gives
Differentiate the ball integral by radius, and divide by its positive squared radius. The result is
The curvature term distinguishes a spherical average from a one-dimensional traveling profile. It is removed by multiplying the mean by radius, or equivalently by time along an expanding sphere. Indeed,
The angular formula controls the apparent singularity at zero. Symmetry gives the average of equal to zero and the average of equal to . Taylor expansion and differentiation of the fixed angular integral imply
For the displacement contribution we use and differentiate its wave equation once in time; the stated third derivatives justify this operation. Adding gives the prescribed displacement and velocity. Derivatives through order two of the resulting expression extend to time zero by the same fixed-sphere integrals and the stated data regularity. All estimates are local in space, so unbounded smooth data cause no difficulty on the compact spheres involved.
The second moment also identifies the initial acceleration explicitly:
Indeed, write . The product rule gives , and the second moment gives . The initial acceleration thus agrees with the PDE, rather than being an unspecified boundary trace.
Proof X-ray
Two corrections hidden in Kirchhoff's formula
The radial identity alone gives , which is not a wave equation: the curvature term is the obstacle. The key move is multiplication by time. With , the product rule gives , and the extra first-derivative term is exactly the curvature term evaluated at . The same factor makes start from zero displacement with velocity , while the vanishing first angular moment makes its initial acceleration zero. The displacement term then needs no new computation: since the coefficients do not depend on time, is again a solution, with displacement and velocity . This last step is where the extra derivative in is spent.
Proof
(Descent to two dimensions)
For descent, extend two-dimensional data independently of a third coordinate and apply the construction just proved. Each hemisphere projects onto the unit disk. Its surface element becomes , and the two hemispheres supply a factor of two. Thus the spherical velocity propagator is exactly . In polar coordinates the disk weight has integral , so it is integrable and gives for constant data. Substitution yields the physical-disk formula with its factor .
On the fixed disk, differentiating acts on smooth data and powers of time, leaving the same integrable weight. It is locally dominated on bounded observation/time regions, including time zero. The constructed three-dimensional expression is independent of the observation coordinate , so and its equation reduces to the two-dimensional wave equation. Therefore the descent gives a classical two-dimensional solution and its initial traces. No spatial decay in the fictitious third direction is needed: the three-dimensional construction and the local uniqueness proof do not assume finite global energy.
Concept lensGeometric
A sphere projects onto a filled disk
A point on either hemisphere projects to a point inside the disk. The surface becomes steep near its equator, producing the larger weight near the disk edge. Points throughout the disk still contribute, not only its boundary. This projection explains how two dimensions can retain a tail while sharing the same finite wave speed with three dimensions.
Proof
(Energy on a shrinking cone, clearing and the two-dimensional tail)
For uniqueness, let be a difference with zero data on the relevant initial ball. For , integrate over the ball of radius centered at :
The moving-boundary differentiation and the PDE give
Here is the outward unit normal and the tangential gradient is the part orthogonal to it. The initial energy is zero because initial velocity and the gradient of the zero displacement vanish inside the open base ball; its boundary sphere has zero volume measure. Continuous derivative traces on the compact cone and convergence of the ball radii give continuity of energy at initial time. Classical regularity on the closed smaller balls justifies the positive-time integrations above. Nonnegativity therefore forces zero energy throughout the smaller balls. In particular before the apex, so integration along that vertical line and continuity at the apex give . This proves local dependence and uniqueness without a whole-space energy integral.
Finite propagation follows immediately from the initial ball statement. The stronger three-dimensional assertion uses the representation: it samples , its gradient, and on the sphere of radius . Under the strict displayed inequality that sphere stays a positive distance from their supports. The positive margin persists for nearby times, so the displacement mean and its time derivative vanish, as does the velocity term. Dependence on the gradient matters: knowing only the values of the displacement on a single sphere does not determine its radial derivative there. This sphere-local dependence and the resulting clearing are the form of Huygens' principle established here.
In two dimensions with zero displacement, the physical-disk kernel is strictly positive. Once its disk contains the compact velocity support, a nonnegative velocity positive on some open set gives a positive integral. More precisely, writing its total integral as , at every fixed observation point
To justify this, on the fixed compact support the weight after multiplication by time is . It tends uniformly to a constant and is uniformly bounded for sufficiently large time. This proves the limit and the tail. Signed data can have cancellation; the positive-tail statement is not asserted for every nonzero velocity.
Worked examples
Worked example 4.3.5
(Polynomial data check the dimension-dependent constants)
Take zero velocity and . In three dimensions its spherical mean is , so the formula gives . In two dimensions descent gives . Directly, , and both answers satisfy the equation and data. These smooth unbounded examples are legitimate in the local classical class, though they do not have finite whole-space energy.
Worked example 4.3.6
(The same radial velocity has different late behavior)
Choose a smooth nonnegative radial velocity , compactly supported in a ball of radius , and not identically zero; take zero displacement. Assume the radial profile defines a smooth function at the origin, for example by making it constant near zero. At the center in three dimensions,
which is zero after the wave sphere leaves the support. In two dimensions,
For this integral still contains the nonzero initial velocity and is positive. The distinction concerns which part of the backward ball is sampled, rather than a different leading-front speed.
Worked example 4.3.7
(A plane mode from a spherical average)
In three dimensions take , , with . Choosing polar direction along shows
The angular average reduces to one half of the integral of the corresponding cosine over the interval from minus one to one. Multiplying by time and differentiating removes the frequency divisor. The mode oscillates instead of decaying. In two dimensions, direct differentiation of the displayed plane-mode answer verifies the PDE and initial data, so uniqueness identifies the answer there as well; this last step is a direct verification rather than an evaluation of the disk integral. Finite-mass or compact-support conclusions do not apply to these data.
Counterexample mode
A spherical average alone is not the wave propagator
For the quadratic displacement in three dimensions, simply taking gives second time derivative but spatial Laplacian multiplied by wave speed squared equal to . It fails the PDE. The derivative of time times the mean in Kirchhoff's formula is essential, not a normalization that may be dropped.
The second misreading keeps the correct formula but draws a conclusion that finite speed does not support.
Counterexample mode
Finite speed does not imply eventual silence
The nonnegative compact velocity in two dimensions gives a positive solution at the center for every sufficiently large time. Its signal travels no faster than the wave speed: the initial support is already within the backward disk at those times. A statement about the outside of that disk cannot prohibit contributions from its interior. Therefore the stronger three-dimensional clearing statement cannot be inferred from finite propagation alone.
Common mistakes
Common mistake
(Differentiating the disk formula as if its boundary were fixed)
Keep the surface area, the average and the velocity propagator distinct. Changing a sphere to a disk changes its weight and constants. The integrable disk-edge singularity does not justify differentiating its singular physical-coordinate formula as though the boundary were fixed. The fixed-unit-disk representation supplies the needed differentiation argument.
Common mistake
(Stretching Huygens' principle beyond what was proved)
Huygens' principle here refers to the homogeneous constant-coefficient three-dimensional equation. We have not proved it for added lower-order terms, forcing, boundaries, variable coefficients or every odd dimension. Nor does the local energy argument assume that each smooth solution has finite global energy. Its shrinking bounded balls are precisely what avoid that extra hypothesis.
Summary
The spherical-mean radial identity constructs the three-dimensional wave propagator with its correct time factors. Descent turns its sphere into a weighted two-dimensional disk. Cone energy proves uniqueness and finite propagation in both dimensions, while the explicit kernels show a sharper difference: compact three-dimensional data eventually clear each observer, whereas positive compact two-dimensional velocity can leave an inverse-time tail.
Exercises
Quick check
(In three dimensions solve the zero-velocity problem with displacement , where and , and check the initial traces and the equation.)
Spherical averages preserve affine functions and add the squared radius to the quadratic part.
Solution
The solution is . The affine Laplacian is zero and the quadratic Laplacian is six, so both sides of the PDE equal . Initial displacement is the prescribed polynomial and initial velocity is zero. The angular first moments vanish, which explains the unchanged affine terms.
Quick check
(In three dimensions take zero displacement and a smooth radial velocity supported in the shell , with . Determine when the center can receive a signal and give its value.)
At the center the entire observation sphere has the same radial value.
Solution
Writing the radial profile as , we have . It vanishes when or , so a signal can occur only during . Even inside that permitted interval the value can vanish if the chosen profile does. Support containment gives a possible window, not a guarantee of a nonzero signal at every permitted time.
Quick check
(For a compactly supported two-dimensional velocity with integral and zero displacement, prove that at each fixed . Is positivity needed for that limit?)
For large time the support is inside the disk and its rescaled weight is uniformly bounded.
Solution
At fixed and sufficiently large time, all points of the velocity support satisfy , and
Choose with the velocity support in . Once , the displayed denominator is at least . The bounded support also makes it tend uniformly to one. The integrable function then controls the integral. Positivity is unnecessary for this limit, but was needed for the earlier persistent-positive-tail conclusion. When the signed integral is zero, this leading coefficient vanishes; the argument does not conclude that the entire signal is zero.
Quick check
(In three dimensions take the positive displacement with , and zero velocity. Compute the center value and show that it eventually becomes negative.)
Differentiate time times the radial mean; compare the conclusion with the heat equation's positivity property.
Solution
The mean at the center is , hence
It is negative for . The data are smooth and positive, but displacement propagation contains a time derivative of an average, which need not preserve positivity. This contrasts with the positive averaging operator for heat; it is compatible with wave uniqueness and finite speed.