Wavevector Transport on an Affine Incompressible Background

Good—last lesson established that a shared-phase wave has no self-advection because its induced velocity is tangent to its own wavefronts. That cancellation remains useful only if the background flow carries the wavefronts consistently. Otherwise the phase develops a large residual, of size proportional to the high frequency.

We now derive the law that prevents this. For an affine incompressible background velocity

$$u_{\mathrm{old}}(x,t)=D(t)x,\mathrm{tr}D(t)=0,$$

the wavevector must satisfy

$$\boxed{\dot{\zeta }(t)=-D(t{)}^T\zeta (t).}$$

This is the wavevector transport equation. It says that wavevectors are transported by the inverse transpose of the fluid deformation, rather than as ordinary material vectors. We will derive it directly, interpret it through characteristics, and see how incompressible strain can make frequency grow rapidly.

---

Transporting a phase, not merely a scalar amplitude

Consider the temperature-wave component from the previous lessons:

$$\vartheta (x,t)=\Theta (t)\sin q(x,t),q(x,t)=\lambda \zeta (t)\cdot x.$$

Here:

• $\lambda$ is a fixed large base frequency;
• $\zeta (t)$ is the evolving wavevector;
• $\lambda \zeta (t)$ is the physical phase gradient;
• $\Theta (t)$ is the scalar amplitude.

The old affine flow $u_{\mathrm{old}}=Dx$ is the local background supplied by earlier layers. The immediate aim is not to make $\vartheta$ a complete solution of the temperature equation: the new wave is deliberately coupled to the old temperature gradient, and that coupling will determine its amplitude later. The narrower aim is to ensure that the carrier phase is transported correctly by the background.

For a passively transported scalar, the relevant operator is the material derivative

$$D_t={\partial }_t+u_{\mathrm{old}}\cdot \nabla .$$

Apply it to $\vartheta$. First,

$${\partial }_t\vartheta =\dot{\Theta }\sin q+\lambda \Theta (\dot{\zeta }\cdot x)\cos q.$$

Next, because

$$\nabla \vartheta =\lambda \Theta \zeta \cos q,$$

we obtain

$$u_{\mathrm{old}}\cdot \nabla \vartheta =(Dx)\cdot \left(\lambda \Theta \zeta \cos q\right).$$

Use the elementary transpose identity

$$(Dx)\cdot \zeta =x\cdot (D^T\zeta ).$$

Therefore,

$$u_{\mathrm{old}}\cdot \nabla \vartheta =\lambda \Theta (D^T\zeta \cdot x)\cos q.$$

Combining the two pieces gives

$$D_t\vartheta =\dot{\Theta }\sin q+\lambda \Theta (\dot{\zeta }+D^T\zeta )\cdot x\cos q.$$

The second term is the dangerous one. It has a factor $\lambda$, so at high frequency even a modest mismatch in the phase law creates a leading-order error. To remove it, impose

$$\boxed{\dot{\zeta }+D^T\zeta =0.}$$

Equivalently,

$$\boxed{\dot{\zeta }=-D^T\zeta .}$$

With that choice,

$$D_t\vartheta =\dot{\Theta }\sin q.$$

Thus the affine flow no longer distorts the phase incorrectly. Only the amplitude remains to be determined by the deliberately retained Boussinesq interactions in the next lesson.

A useful separation is:

Quantity	Governing issue
Wavevector $\zeta$	Geometric transport by the affine flow
Amplitude $\Theta$	Coupling to the old temperature gradient and buoyancy
Self-advection	Already cancelled by phase–velocity orthogonality

---

Why the transpose is unavoidable

The transpose can look like a technical artefact of rewriting a dot product, but it reflects the geometric nature of a wavevector.

A material point is a vector-like object. Under the affine flow,

$$\dot{X}(t)=D(t)X(t).$$

By contrast, $\zeta$ is the normal to a family of wavefronts. Normals are covectors: they act on displacements through the scalar pairing $\zeta \cdot x$. When spatial positions are deformed, the covectors that define constant phase surfaces must transform by the inverse transpose.

The phase should be unchanged along a background trajectory:

$$\frac{d}{dt}q(X(t),t)=0.$$

Indeed,

$$\frac{d}{dt}(\zeta (t)\cdot X(t))=\dot{\zeta }\cdot X+\zeta \cdot \dot{X}.$$

Since $\dot{X}=DX$,

$$\frac{d}{dt}(\zeta \cdot X)=(\dot{\zeta }+D^T\zeta )\cdot X.$$

So phase constancy along every trajectory is exactly equivalent to

$$\dot{\zeta }=-D^T\zeta .$$

This gives a compact rule worth retaining:

Particles follow $D$; phase gradients follow $-D^T$.

Treating $\zeta$ as if it obeyed $\dot{\zeta }=D\zeta$ would generally fail to keep the phase constant along particle paths.

---

The deformation-gradient form

Let $F(t)$ be the deformation matrix of the affine background, defined by

$$\dot{F}(t)=D(t)F(t),F(0)=I.$$

A particle initially at $a$ is located at

$$X(t)=F(t)a.$$

The solution to the wavevector equation is

$$\boxed{\zeta (t)=F(t{)}^{-T}\zeta (0).}$$

To see this directly, require the phase to retain its initial value on each material point:

$$\zeta (t)\cdot X(t)=\zeta (0)\cdot a.$$

Substituting $X(t)=F(t)a$ gives

$$\zeta (t)\cdot F(t)a=\zeta (0)\cdot a.$$

Since this must hold for every initial position $a$,

$$F(t{)}^T\zeta (t)=\zeta (0),$$

and hence

$$\zeta (t)=F(t{)}^{-T}\zeta (0).$$

This is the most useful conceptual formulation. The flow map $F$ transports the material; the inverse transpose $F^{-T}$ transports the normals to material surfaces.

---

Incompressibility permits directional frequency growth

For the affine background to be incompressible,

$$\nabla \cdot (Dx)=\mathrm{tr}D=0.$$

Equivalently, the deformation preserves area:

$$\det F(t)=\exp \left({\int }_0^t\mathrm{tr}D(s)ds\right)=1.$$

Area preservation does not mean every length is preserved. In two dimensions, if one direction is compressed, another must be stretched by the reciprocal amount. The wavevector grows precisely when the background compresses physical space in the direction normal to the wavefronts.

Take the elementary hyperbolic strain

$$D=\left(\begin{array}{cc}-a& 0\\ 0& a\end{array}\right),a>0.$$

It is incompressible because its trace is zero. A particle evolves as

$$X_1(t)=e^{-at}{X}_1(0),X_2(t)=e^{at}{X}_2(0).$$

Thus physical space is compressed in the first coordinate and stretched in the second.

The wavevector equation is

$$\dot{\zeta }=-D^T\zeta =\left(\begin{array}{cc}a& 0\\ 0& -a\end{array}\right)\zeta .$$

Therefore,

$${\zeta }_1(t)=e^{at}{\zeta }_1(0),{\zeta }_2(t)=e^{-at}{\zeta }_2(0).$$

If the initial wavevector is normal to the first coordinate direction,

$$\zeta (0)=\left(\begin{array}{c}1\\ 0\end{array}\right),$$

then

$$\zeta (t)=\left(\begin{array}{c}{e}^{at}\\ 0\end{array}\right).$$

The phase becomes

$$q(x,t)=\lambda e^{at}{x}_1.$$

The spatial wavelength in the $x_1$ direction is consequently proportional to

$$\frac{1}{\lambda e^{at}}.$$

The affine flow is squeezing wavefronts closer together. It does not create more oscillations by itself; rather, it compresses a fixed material pattern into progressively smaller physical scales.

This is the elementary low-to-high frequency mechanism underlying the later layer construction.

---

Strain changes wavevector length; rotation changes orientation

Every matrix $D$ has a symmetric and antisymmetric part:

$$D=S+A,$$

where

$$S=\frac{D+D^T}{2},A=\frac{D-D^T}{2}.$$

The symmetric part $S$ is strain; the antisymmetric part $A$ is rigid rotation. From

$$\dot{\zeta }=-D^T\zeta ,$$

we can calculate the change in wavevector magnitude:

$$\frac{d}{dt}|\zeta {|}^2=2\zeta \cdot \dot{\zeta }=-2{\zeta }^T{D}^T\zeta =-2{\zeta }^{T}S\zeta .$$

The antisymmetric part disappears because

$${\zeta }^{T}A\zeta =0.$$

Hence:

• strain can amplify or attenuate $|\zeta |$, depending on orientation;
• rigid rotation changes the direction of $\zeta$ but not its length.

For example, if

$$D=\gamma J,J=\left(\begin{array}{cc}0& -1\\ 1& 0\end{array}\right),$$

then $D^T=-D$, and

$$\dot{\zeta }=\gamma J\zeta .$$

The wavevector simply rotates with constant magnitude. This distinction will matter in Module 2: controlled rotation can reset a layer into a favorable orientation, while compression is what supplies gradient gain.

---

A small connection to the curated paper

The Chen–Hou paper does not use this plane-wave ansatz as its main presentation, but its dynamic-rescaling formulation contains a related transport term: an affine radial drift of the form $c_{l}x\cdot \nabla$. Read the following passage as a comparison point. In the paper, this drift comes from changing coordinates to a dynamically rescaled frame; it is not itself the incompressible physical affine flow considered above. Still, it illustrates why an affine transport term must be handled explicitly when tracking concentrating spatial structure.

For the present construction, the key point is more local and geometric: once the older flow is approximated by $D(t)x$ inside an affine core, the new oscillatory layer must choose its wavevector according to $\dot{\zeta }=-D^T\zeta$.

---

A useful template for later calculations

Whenever a new layer has phase

$$q(x,t)=\lambda \zeta (t)\cdot x,$$

on a background

$$u_{\mathrm{old}}(x,t)=D(t)x,$$

the calculation has three steps:

1. Differentiate the phase in time:

   $${\partial }_{t}q=\lambda \dot{\zeta }\cdot x.$$

2. Apply background advection:

   $$u_{\mathrm{old}}\cdot \nabla q=\lambda (D^T\zeta )\cdot x.$$

3. Set their sum to zero, yielding

   $$\dot{\zeta }=-D^T\zeta .$$

If the affine field includes a spatially constant translation,

$$u_{\mathrm{old}}(x,t)=D(t)x+b(t),$$

the wavevector equation is unchanged. One merely adds a time-dependent scalar phase offset to account for translation. The linear part $D(t)$, which determines deformation, is what governs $\zeta$.

---

Takeaways

The wavevector transport law on an affine incompressible background is

$$\boxed{\dot{\zeta }=-D^T\zeta .}$$

It comes from requiring the high-frequency carrier phase

$$\lambda \zeta (t)\cdot x$$

to remain constant along trajectories of the background flow. Equivalently, if $F$ is the deformation gradient of that flow, then

$$\zeta (t)=F(t{)}^{-T}\zeta (0).$$

The inverse transpose is essential because wavevectors are normals to wavefronts, not material displacement vectors. Incompressibility preserves area, but hyperbolic strain can compress the wavefront-normal direction exponentially, making $|\zeta |$ and thus the physical frequency grow exponentially.

Next, we will use this phase law to isolate the remaining leading interactions and derive the coupled ODEs for the temperature and vorticity amplitudes.