Motivation
A moving particle can carry a fixed scalar value while the volume occupied by a collection of particles changes. A density must respond to that volume change if the amount inside the moving collection is to stay constant. These are different transport laws. Their distinction disappears for an incompressible flow but matters whenever the flow expands or compresses space.
Characteristics therefore need more than a path formula. We also need the Jacobian of the map from initial particle labels to current positions. That determinant tells us which quantities are conserved and which norms may grow. It will also let us extend classical solutions to rough integrable data and prove uniqueness in a precise weak class.
Definitions
Let be , with . The velocity itself may be unbounded; the derivative bound gives at most linear growth. Let its autonomous flow and Jacobian be
The divergence bound satisfies for the Euclidean operator norm. We distinguish
Definition 1.2.1
(Advected value, conservative density and weak solution)
The first equation transports a value attached to each particle; the second transports density per unit spatial volume. For compactly supported initial data, a classical solution means a space-time function satisfying its equation and initial trace pointwise.
For , the weak class on a finite time interval is , with the initial value attained in norm. For every smooth compactly supported test function vanishing near terminal time, the respective weak identities are
All spatial integrals without a displayed domain are over the whole space. The two different adjoint expressions are essential.
Theorem/Proposition
Theorem 1.2.2
(The Jacobian supplies the density correction)
The flow exists for every real time and is a diffeomorphism, with inverse . Its determinant is positive and satisfies
For compactly supported data , the unique classical solutions are
Both supports are the image of the initial support under the flow.
Theorem 1.2.3
( evolution and weak well-posedness)
Define the two operators on measurable data by
For their exact norm formulas are
Consequently, with ,
These bounds hold also for , and the advected supremum norm is exactly preserved. Each family is a group. For finite it is strongly continuous and supplies the unique weak solution in the declared class. The supremum endpoint assertion concerns bounded operators; a weak class and strong continuity at that endpoint are not claimed here. In particular the density preserves signed total mass and its norm whenever the initial data are integrable.
Theorem 1.2.4
(A stationary weight produces an invariant measure)
Suppose a positive function satisfies . Then
for every finite whenever that norm is finite. If an advected scalar is multiplied by this weight, , it becomes a conservative density with initial density . A positive invariant weight is an additional hypothesis, not something assumed to exist for every velocity.
Proof sketch or proof idea
Global paths and changing volume
Proof
(Global flow, Liouville identity and the classical formulas)
Bounded spatial derivative makes the velocity globally Lipschitz and bounds its size by a constant plus a constant times the distance from zero. Grönwall prevents escape to infinity in finite time, both forward and backward. ODE uniqueness gives the composition law and inverse flow. Differentiating in the particle label gives
This derivative is justified by state difference quotients and their linear variational equations. Applying the same argument once more uses the continuous second derivative of the velocity and gives the stated flow regularity. Only bounds on compact finite-time trajectory families are needed for that second step. Differentiating the determinant by multilinearity of its columns in this matrix equation gives the Liouville identity; this calculation does not require assuming invertibility first:
Solving this scalar equation proves the exponential formula and its bounds. Along a path, the advected value has zero derivative. Expanding the divergence in the density equation instead gives a scalar value equation with the additional divergence coefficient. Multiplication by the Jacobian cancels it, proving the two formulas and uniqueness. The inverse maps are defined everywhere, so every observation point has one initial footpoint. Compact initial supports stay in a common compact set on each finite time slab.
The Jacobian therefore has a direct geometric reading.
Concept lensGeometric
A value stays fixed; an amount includes volume
An infinitesimal initial volume is multiplied by the flow Jacobian. Keeping its particle value fixed does not keep its integral fixed when that volume changes. A conserved density divides its value by precisely this factor. The positive determinant also shows that nonnegative initial density remains nonnegative.
Change of variables controls the norms
Proof
(Norm identities, group law and strong continuity)
Use the current position as the image of the initial label in the spatial integral. Its volume element is the positive Jacobian times the initial element. The advected scalar contributes no additional factor; the density contributes its reciprocal before taking the power. This gives the two exact norm identities and their bounds. Bijectivity gives preservation of the scalar essential supremum. The density bound at that endpoint follows from the reciprocal Jacobian estimate.
The flow composition law and its determinant chain rule give the two group laws. These formulas are well defined on equivalence classes: a diffeomorphism and its inverse preserve null sets locally, hence on a countable covering of the whole space.
For finite exponent, first take smooth compactly supported data. As time tends to zero, flow maps and Jacobians converge uniformly on the relevant compact sets, so the solutions converge in norm to the data. For either operator family write ; on its operator norm is bounded by . Choose a smooth compactly supported approximation to arbitrary data. The three-term estimate is
Make the first and third terms each smaller than one third of the desired tolerance by choosing the approximation. Then make the middle term smaller than one third by taking time close to zero. This proves strong continuity for all finite-exponent data. The group law gives continuity at every time. Density approximation and the same uniform bounds pass the classical equations to the two weak identities, proving existence in the declared class.
Backward tests prove weak uniqueness
Proof
(Uniqueness by backward adjoint tests)
Fix a terminal time and a smooth compactly supported terminal test . For the density equation use the backward test
for the advective equation use
They solve the corresponding homogeneous adjoint equations with terminal value equal to the chosen test. This follows from the already proved classical formulas with time measured from the terminal slice. Their supports lie in a common compact set. The density test is continuously differentiable in space and time. For the second test this regularity also uses the continuously differentiable Jacobian: its spatial derivative uses the second spatial derivative of the flow, and its time derivative is given by the Jacobian equation. This explains our second-derivative velocity hypothesis. The weak identities extend to compactly supported continuously differentiable tests by smooth approximation in function and first derivatives.
Write for the relevant solution and for its backward test. Take and set . The time cutoff avoids both endpoints, so smooth approximation of this test has no initial-boundary issue. The adjoint equation cancels all terms except the cutoff derivative, and the initial term vanishes. Thus
The pairing is continuous, by norm continuity of the solution and uniform continuity of the compactly supported test. Having zero distributional derivative, it is constant, including at the initial and terminal times. The initial trace fixes the terminal pairing for every terminal test, proving uniqueness. The argument specifies the weak class rather than claiming uniqueness among arbitrary distributions without time traces.
Each step of this argument uses a different hypothesis.
Proof X-ray
Why the two backward tests are different
The proof has three moves. First, duality exchanges the laws: the adjoint of the conservative operator is the advective one and conversely, so the density is tested against a transported value, and the advected scalar against a transported density carrying . Using for the advective equation would leave the residual in , and the pairing would no longer be constant unless the flow preserves volume. Second, regularity of the tests is where enters: is only because is , which needs second derivatives of the flow. Third, the cutoff removes both endpoint terms, so the identity reduces to . Only the declared class makes continuous up to and , and that continuity is what carries the initial trace to the terminal pairing.
Weighted invariance
Proof
(The weighted Jacobian is constant along paths)
Differentiate the proposed weight times the Jacobian along a path. Its derivative is the divergence of the weighted velocity, evaluated on that path, times the Jacobian. The assumed stationary equation makes it zero. Change of variables now proves the weighted norm identity. Finally, the product rule shows that multiplying an advected scalar by the stationary weight solves the conservative equation. Its initial density changes by the same multiplication.
Worked examples
Worked example 1.2.5
(Uniform expansion transports values and dilutes density)
In one dimension let , with real . Then
For positive rate, the scalar profile occupies a growing interval without changing its carried values, whereas the density amplitude decreases. The scalar integral is multiplied by the expansion factor; the density integral stays fixed. Negative rate reverses the geometric effect and concentrates density. The velocity is unbounded, but its bounded derivative gives a global flow.
Worked example 1.2.6
(An incompressible shear can build large gradients)
For , the flow is with determinant one. Both transport laws coincide:
The Lebesgue value norms in the theorem are preserved. Nevertheless, for smooth compactly supported data with ,
The profile becomes increasingly tilted and develops large derivatives without changing its square-integrable value norm. Volume preservation is not a bound on every derivative.
Worked example 1.2.7
(A varying speed preserves a weighted spatial measure)
On the line take . Define the travel-time coordinate
The speed stays between one and three, so the coordinate is increasing and onto the real line. Its reciprocal speed is a positive stationary weight. Advected values preserve the integral of every finite power against that weight. For initial density , write ; then
Equivalently, speed times density is an advected value, with initial value equal to speed times the initial density. This weighted description accounts for the different residence times in slow and fast regions.
Common mistakes
Common mistake
(Conserved mass does not freeze the maximum density)
For , a nonnegative compact initial profile gives . Its total mass is unchanged, while its maximum grows by the factor . A shrinking occupied interval balances that increase. Conservation of an integral does not imply preservation of the supremum norm.
Counterexample mode
The endpoint supremum-norm bound does not imply strong continuity
For constant unit velocity on the line and , the solution is . For the difference has absolute value one on ; for it does on . These intervals have positive measure, so
The operator is an isometry but does not converge strongly to the identity on this space. The finite-exponent continuity theorem deliberately excludes this endpoint.
The growth hypothesis on is also not removable, even for a divergence-free field.
Counterexample mode
Zero divergence alone does not provide a global flow
The smooth field has zero divergence but unbounded derivative,
On the maximal time interval containing zero, specified by , its flow is
The derivative matrix of the flow is triangular, with reciprocal diagonal factors, giving the displayed determinant. For escape occurs at time ; for it occurs at that negative time in the past. Local volume preservation remains true, but the globally defined diffeomorphisms required for our whole-space norm argument do not exist. A divergence estimate cannot replace the growth hypothesis that guaranteed complete paths.
Summary
The flow Jacobian distinguishes transported values from conserved density. Change of variables gives exact norm identities; density approximation and backward adjoint tests give weak well-posedness for finite integrability exponents. An invariant positive weight can restore isometries for compressible scalar transport, while strong continuity at the supremum endpoint and global path existence require separate attention.
Exercises
Quick check
(For with a real constant matrix , find both transport solutions and their exact norm factors. Explain why the trace controls these value norms even when also shears the profile.)
Solution
The flow is the matrix exponential, and its determinant is the exponential of the trace. Thus
For finite exponent their norm factors are respectively
At the supremum endpoint the scalar factor is one and the density factor is the exponential of negative trace times time. For nonzero data these are exact norm ratios for all real times, sharper than the general absolute-divergence bounds using . Shear can change shape and derivatives while leaving the determinant, and hence these value-norm formulas, unchanged.
Quick check
(For smooth compactly supported initial data, derive the energy derivatives of the advected scalar and the conservative density. Explain why compression can increase the density’s norm.)
Solution
Integration by parts is valid because supports remain compact on finite slabs. The two identities are
The additional divergence term in the density equation reverses the sign. Negative divergence where the density is supported can increase its energy even though its total mass remains fixed. Bounding divergence in absolute value recovers the corresponding square-integrable norm bounds.
Quick check
(For , finite and initial density , prove that is finite and constant. Which familiar conservation law occurs at ?)
Solution
Set . For classical solutions the product rule gives . For all initial data in the stated integrability class, the explicit flow formula gives the stronger operator identity
The stationary weight is , since . Apply the weighted scalar theorem directly to the transported initial value . It gives conservation of
The speed lies between one and three, so the weighted integrals are comparable to the ordinary finite-power integrals and are finite for the stated data. This proves the identity for weak solutions without differentiating them. At exponent one it is conservation of the density's absolute integral. For nonnegative data it is total mass; signed total mass is conserved as well.
Quick check
(For initial data and a fixed nonzero time, suppose either transport solution has a spatial representative. Use the inverse flow to show that the initial data already have a representative. Does strong continuity in finite imply smoothing?)
Solution
If the scalar solution has a continuously differentiable representative at that time, compose it with the forward flow to recover such a representative of the initial data. For density the inverse relation is
The flow is twice continuously differentiable and its positive Jacobian is continuously differentiable, so the product is again continuously differentiable. The identities hold almost everywhere and therefore give representatives of the original equivalence class. Norm continuity in time describes a different property; it does not imply new spatial regularity.