Controlled Rotation: Resetting Vorticity While Preserving Temperature-Gradient Gain

Hello. This lesson begins the second module, where the single-wave amplification mechanism is turned into a many-layer construction. The key engineering problem is easy to state: a layer can acquire a very large temperature gradient, but that same gradient drives vorticity and can make the layer’s induced velocity too large to keep under control. A controlled rotation separates those two effects. It cancels the layer’s accumulated vorticity while preserving the short wavelength, hence preserving the large temperature gradient.

We will work in the local affine-wave model. It is not yet the full localized construction; the next lessons address how infinitely many such layers are scheduled, localized, and corrected.

---

The two quantities that must be disentangled

For the inviscid two-dimensional Boussinesq system,

$${\partial }_t\omega +\mathbf{u}\cdot \nabla \omega ={\partial }_x\theta ,{\partial }_t\theta +\mathbf{u}\cdot \nabla \theta =0,\mathbf{u}={\nabla }^{\perp }(-\Delta {)}^{-1}\omega ,$$

the temperature $\theta$ is transported, while vorticity $\omega$ is forced by the horizontal temperature derivative ${\partial }_x\theta$.

Differentiating the temperature transport equation gives

$$({\partial }_t+\mathbf{u}\cdot \nabla )\nabla \theta =-(\nabla \mathbf{u}{)}^T\nabla \theta .$$

Thus a deformation of the flow can make $\nabla \theta$ large even though the amplitude of $\theta$ remains constant along trajectories. In particular, a temperature wave can become increasingly fine-scale without becoming large in $L^{\infty }$.

The construction must track two distinct pieces of information:

Quantity	What it measures	Why it matters
$|\nabla \theta |$	the strength of the stored fine scale	its divergence is the desired singular behavior
${\partial }_x\theta$	the horizontal component of that gradient	it is the instantaneous source for $\omega$
$\omega$	the accumulated effect of past forcing	it generates velocity and can disrupt later layers

The important distinction is that ${\partial }_x\theta$ is a component of $\nabla \theta$, while $|\nabla \theta |$ is its magnitude. Rotation can make the former small or sign-changing without reducing the latter.

---

A single wave on an affine background

Near the core of one layer, idealize the low-frequency background velocity by an affine incompressible field

$$\mathbf{U}(t,x)=A(t)x,\mathrm{tr}A(t)=0.$$

For the moment, treat this as the velocity experienced by the wave and omit the wave’s own small induced velocity. Let the temperature and vorticity layer be

$$\theta (t,x)=a\sin (k(t)\cdot x),\omega (t,x)=b(t)\cos (k(t)\cdot x).$$

Here:

• $a$ is the temperature amplitude;
• $k(t)$ is the physical wavevector;
• $K(t)=|k(t)|$ is the frequency;
• $b(t)$ is the vorticity amplitude.

The sine/cosine offset is simply forced by differentiation: differentiating a sine temperature wave produces a cosine vorticity source. Both fields nevertheless depend on the same phase $k(t)\cdot x$. Consequently, their own Biot–Savart velocity is perpendicular to $k$, so it does not advect that phase at leading order.

For the phase to be transported by the affine field, require

$$\frac{d}{dt}(k(t)\cdot x(t))=0$$

along trajectories satisfying $\dot{x}=A(t)x$. This yields the familiar covector transport law

$$\dot{k}(t)=-A(t{)}^{T}k(t).$$

This equation is the wave-level version of the gradient equation above: the wavevector and the temperature gradient are acted on by the transpose deformation matrix.

Substituting the ansatz into the vorticity equation gives

$$\dot{b}(t)=a{k}_1(t),$$

where $k_1$ is the horizontal component of the wavevector. Hence

$$b(t)=b(t_0)+a{\int }_{t_0}^t{k}_1(s)ds.$$

Meanwhile,

$$\nabla \theta (t,x)=ak(t)\cos (k(t)\cdot x),$$

so its maximal size is

$$G(t)=a|k(t)|=aK(t).$$

This is the central bookkeeping identity:

$$\boxed{\text{gradient size is }aK,\text{vorticity is the signed time integral of }a{k}_1.}$$

A large gradient does not by itself imply large vorticity. It implies the potential for large vorticity forcing, but the realized vorticity depends on the signed history of the horizontal component.

---

What the rotation controls

Write the wavevector in polar form:

$$k(t)=K(t)\left(\begin{array}{c}\cos \varphi (t)\\ \sin \varphi (t)\end{array}\right).$$

Then the forcing and gradient magnitude become

$$\dot{b}(t)=aK(t)\cos \varphi (t),G(t)=aK(t).$$

The angle $\varphi$ controls the source; the magnitude $K$ controls the gradient gain.

Suppose an earlier strain stage has already made $K$ very large. During the reset stage, we want to rotate the vector while keeping $K$ fixed, or at least comparable to its large value. The simplest idealized choice is a rigid rotation.

Define

$$J=\left(\begin{array}{cc}0& -1\\ 1& 0\end{array}\right),A(t)=\dot{\varphi }(t)J.$$

Since $\mathrm{tr}J=0$, this is incompressible. Moreover, $A^T=-\dot{\varphi }J$, so the wavevector equation becomes

$$\dot{k}=-A^{T}k=\dot{\varphi }Jk.$$

That is exactly the equation for a vector of fixed length rotating with angle $\varphi (t)$. Thus

$$\frac{d}{dt}K(t)=0$$

through the pure-rotation phase. The gradient gain $G=aK$ is retained exactly.

At the same time,

$$\dot{b}=aK\cos \varphi .$$

As $\varphi$ moves through directions with $\cos \varphi <0$, the Boussinesq forcing has the opposite sign and removes vorticity that was accumulated earlier.

---

Resetting the vorticity amplitude

Suppose a layer enters the rotation stage at time $t_s$ with

$$b(t_s)=b_s,K(t_s)=K_{\ast }.$$

During a rotation with $K(t)=K_{\ast }$, the exit amplitude is

$$b(t_r)=b_s+a{K}_{\ast }{\int }_{t_s}^{t_r}\cos \varphi (t)dt.$$

To reset the layer, choose the angular protocol so that

$${\int }_{t_s}^{t_r}\cos \varphi (t)dt=-\frac{b_s}{a{K}_{\ast }}.$$

Then

$$b(t_r)=0.$$

This is only one scalar condition, while the rotation schedule $\varphi (t)$ has considerable freedom. A conceptual protocol is:

1. Rotate the wavevector into the left half-plane, near $\varphi =\pi$, where $k_1$ is negative.
2. Spend enough controlled time there for the negative vorticity forcing to cancel the pre-existing $b_s$.
3. End with the wavevector vertical, at $\varphi =\pi /2$, or close to vertical.

The final orientation is useful because

$$k_1(t_r)=K_{\ast }\cos \left(\frac{\pi }{2}\right)=0.$$

Therefore,

$${\partial }_x\theta (t_r,x)=0$$

for the ideal wave. The layer is not merely reset at one instant: if subsequent evolution maintains that near-vertical orientation, it is also made dormant as a leading-order vorticity source.

The layer has retained

$$|\nabla \theta {|}_{\mathrm{amp}}=a{K}_{\ast },$$

but it has lost its leading-order ability to drive vorticity through ${\theta }_x$.

---

Why reset is essential for layer assembly

The velocity produced by a Fourier vorticity mode has a favorable frequency factor. From Biot–Savart,

$$|{\mathbf{u}}_{\mathrm{layer}}{|}_{\mathrm{amp}}\asymp \frac{|b|}{K},|\nabla {\mathbf{u}}_{\mathrm{layer}}{|}_{\mathrm{amp}}\asymp |b|.$$

So when $b$ is reset to zero, the layer’s induced velocity and velocity gradient disappear at leading order. Yet its temperature wave remains:

$$\theta =a\sin (k\cdot x),|\theta {|}_{\mathrm{amp}}=a,|\nabla \theta {|}_{\mathrm{amp}}=a{K}_{\ast }.$$

This is the desired asymmetry:

Before reset	After reset
$K_{\ast }$ is large	$K_{\ast }$ remains large
$|\nabla \theta |$ is large	$|\nabla \theta |$ remains large
$b$ may be large	$b$ is made small or zero
the layer can produce appreciable velocity	its leading-order velocity is removed
${\partial }_x\theta$ may actively force vorticity	choosing $k$ vertical makes ${\partial }_x\theta$ vanish

Without this maneuver, every previously amplified layer would retain appreciable vorticity. Summing many layers would then create a large low-frequency velocity or strain field, destroying the carefully controlled background in which the next layer is meant to amplify.

With reset, an old layer becomes a kind of stored derivative reservoir: it contributes a large temperature gradient to the final singular behavior, but it no longer substantially drives the dynamics that builds subsequent layers.

---

What “retaining the gain” does and does not mean

There are three useful qualifications.

First, the retained quantity is principally the size of the gradient, not necessarily the horizontal derivative. At the end of the rotation,

$$|\nabla \theta |\text{ is large},|{\partial }_x\theta |\text{ can be zero}.$$

Second, this is not a cancellation of the temperature wave itself. Its amplitude $a$ remains bounded and nonzero; only the orientation of its high-frequency oscillation has changed.

Third, the pure rigid rotation is an ideal local model. In a full construction, the rotation is implemented within an affine core, then spatially cut off and accompanied by corrections. Those operations introduce residual terms. The point of this calculation is not that a globally rigid rotation is the final solution, but that the key cancellation has a robust geometric origin: vorticity remembers a signed projection history, whereas gradient growth remembers frequency magnitude.

---

A compact mental model

When following the later construction, keep these four statements together:

1. Strain raises the frequency $K$, producing a large gradient $aK$.
2. Boussinesq forcing sees only the horizontal projection $aK\cos \varphi$.
3. Vorticity is the time-integrated signed forcing, $b=b_s+a\int K\cos \varphi dt$.
4. A designed rotation can make that integral cancel while preserving $K$.

The reset is therefore neither dissipation nor a loss of the fine-scale structure. It is a controlled geometric cancellation that makes high derivatives accumulate without allowing every old layer to remain dynamically active.

---

Key takeaways

• In a wave layer, $|\nabla \theta {|}_{\mathrm{amp}}=a|k|$, while vorticity obeys $\dot{b}=a{k}_1$.
• The vorticity amplitude depends on the signed temporal integral of the wavevector’s horizontal component.
• A rigid incompressible rotation keeps $|k|$, and hence the temperature-gradient gain, fixed.
• By rotating through directions with negative $k_1$, one can cancel previously accumulated vorticity.
• Ending with $k$ nearly vertical makes the retained large gradient largely invisible to the source ${\partial }_x\theta$, leaving the layer dormant at leading order.
• This reset prevents old high-frequency layers from accumulating into an uncontrollable velocity field.

Next, we will turn this one-layer mechanism into a toy infinite schedule: infinitely many layers will fit before a finite terminal time, their gradients will diverge, and their field amplitudes will remain bounded.