on critical SQG and nonlocal maximum principles

Overview

Kiselev, Nazarov, and Volberg proved global well-posedness for the critical two-dimensional dissipative surface quasi-geostrophic equation by an elementary but powerful nonlocal maximum principle (Kiselev, Nazarov & Volberg, 2006). The proof preserves a carefully designed modulus of continuity under the full nonlinear evolution.

The paper is important because it shows how a critical drift-diffusion equation can be controlled pointwise, without relying only on energy estimates.

The Critical SQG Equation

The dissipative SQG equation is

$${\partial }_t\theta +u\cdot \nabla \theta +\Lambda \theta =0,$$

where

$$u=R^{\perp }\theta =(-R_2\theta ,R_1\theta )$$

and

$$\Lambda =(-\Delta {)}^{1/2}.$$

The Riesz transforms make the velocity divergence-free:

$$\nabla \cdot u=0.$$

Thus the transport term rearranges the scalar $\theta$, while $\Lambda \theta$ gives nonlocal dissipation.

The scaling is

$${\theta }_{\lambda }(t,x)=\theta (\lambda t,\lambda x).$$

Under this scaling, the drift and dissipation have the same order. This is the critical case. If the dissipation were stronger, classical smoothing would dominate. If weaker, possible singularity formation becomes much harder to exclude.

Moduli of Continuity

A modulus of continuity is an increasing concave function

$$\omega :[0,\infty )\to [0,\infty )$$

with $\omega (0)=0$. A function $\theta$ obeys this modulus if

$$|\theta (x)-\theta (y)|\le \omega (|x-y|)$$

for all $x,y$.

The strategy is to choose $\omega$ so that if the initial data obeys the modulus, the solution obeys it forever. A uniform modulus with finite slope at the origin gives a uniform gradient bound:

$$\Vert \nabla \theta (t){\Vert }_{L^{\infty }}\le {\omega }^{\prime }(0).$$

Once the gradient remains bounded, smoothness can be continued.

Breakthrough scenario

Suppose the modulus fails for the first time at $t=t_{\ast }$ and points $x\ne y$. Then

$$\theta (t_{\ast },x)-\theta (t_{\ast },y)=\omega (|x-y|)$$

after choosing signs. At this first contact, the time derivative of the left side must be nonnegative. The proof shows it is actually negative.

Drift Versus Dissipation

Let

$$\xi =|x-y|.$$

At a breakthrough pair, the transport part contributes

$$(u(x)-u(y))\cdot \nabla \theta$$

and is bounded by

$$\Omega (\xi ){\omega }^{\prime }(\xi ),$$

where $\Omega$ is a modulus of continuity for the velocity induced by the scalar modulus $\omega$.

Because $u=R^{\perp }\theta$ is nonlocal, $\Omega$ depends on all scales of $\omega$. A typical bound has the form

$$\Omega (\xi )\lesssim {\int }_0^{\xi }\frac{\omega (\eta )}{\eta }d\eta +\xi {\int }_{\xi }^{\infty }\frac{\omega (\eta )}{{\eta }^2}d\eta .$$

The first term comes from nearby singular-integral contributions; the second comes from far-field effects.

The dissipation contributes a negative quantity

$$D(\xi )<0.$$

For the half-Laplacian, this term is also nonlocal and can be estimated directly from the modulus. Concavity of $\omega$ ensures that the symmetric second difference of $\theta$ is negative in the right averaged sense.

The desired inequality is

$$\Omega (\xi ){\omega }^{\prime }(\xi )+D(\xi )<0$$

for every possible breakthrough distance $\xi$.

Construction of the Modulus

The modulus is built with two competing features.

At small scales, it must have finite slope so it gives a gradient bound, but it must be curved enough that dissipation dominates the drift. A model behavior is

$$\omega (\xi )=\xi -{\xi }^{3/2}$$

near zero, modified to satisfy the exact estimates.

At large scales, the modulus grows slowly. This prevents the velocity modulus $\Omega$ from becoming too large through the far-field integral. The final modulus is concave, unbounded or suitably large for the data after scaling, and tuned so that the nonlocal inequality above holds at all scales.

Proof: Preservation of the Modulus

Start with smooth initial data. The equation is invariant under simultaneous scaling of time and space, so one first dilates the initial datum until it obeys the chosen modulus strictly:

$$|{\theta }_0(x)-{\theta }_0(y)|<\omega (|x-y|)$$

for every $x\ne y$. The strictness matters because it gives an open time interval on which the inequality is preserved. If the modulus is eventually lost, continuity in time gives a first time $t_{\ast }$ and a pair $x\ne y$ with

$$\theta (t_{\ast },x)-\theta (t_{\ast },y)=\omega (\xi ),\xi =|x-y|.$$

The sign has been chosen so that the larger value is at $x$.

At that pair, the equation gives

$$\frac{d}{dt}(\theta (x)-\theta (y))=-(u(x)\cdot \nabla \theta (x)-u(y)\cdot \nabla \theta (y))-(\Lambda \theta (x)-\Lambda \theta (y)).$$

Since equality occurs at distance $\xi$, the function

$$z\mapsto \theta (t_{\ast },z)-\theta (t_{\ast },y)-\omega (|z-y|)$$

has a maximum at $z=x$, and the analogous one-point comparison holds at $y$. This bounds the relevant directional derivatives by ${\omega }^{\prime }(\xi )$. The drift term is therefore no larger than

$$|u(x)-u(y)|{\omega }^{\prime }(\xi )\le \Omega (\xi ){\omega }^{\prime }(\xi ),$$

where the Riesz-transform estimate gives

$$\Omega (\xi )=A\left({\int }_0^{\xi }\frac{\omega (\eta )}{\eta }d\eta +\xi {\int }_{\xi }^{\infty }\frac{\omega (\eta )}{{\eta }^2}d\eta \right)$$

up to an absolute constant $A$.

The half-Laplacian is then tested against the same two-point information. Using the Poisson semigroup representation, the dissipative contribution is bounded above by the explicit negative quantity

$$D(\xi )=\frac{1}{\pi }{\int }_0^{\xi /2}\frac{\omega (\xi +2\eta )+\omega (\xi -2\eta )-2\omega (\xi )}{{\eta }^2}d\eta$$ $$+\frac{1}{\pi }{\int }_{\xi /2}^{\infty }\frac{\omega (2\eta +\xi )-\omega (2\eta -\xi )-2\omega (\xi )}{{\eta }^2}d\eta .$$

Concavity makes the first integral negative, and the second records the fact that the two values at $x$ and $y$ are already separated by the full allowed amount.

The modulus is constructed so that

$$\Omega (\xi ){\omega }^{\prime }(\xi )+D(\xi )<0$$

for every $\xi >0$. Hence the time derivative of the left side at a first contact is strictly negative, contradicting the assumption that equality was being reached from below. The modulus is preserved for all time, and its finite slope at the origin gives the global gradient bound.

Criticality of the Argument

The method is precisely matched to the critical equation. The drift is as singular as the scalar, since Riesz transforms preserve scale. The half-Laplacian gives exactly one derivative of dissipation. A purely local maximum principle would not see enough smoothing. The nonlocal modulus proof extracts the missing information from the integral structure of $\Lambda$.

Scope

The argument gives a clean proof for critical dissipative SQG. Supercritical variants, where the dissipation is ${\Lambda }^{\alpha }$ with $\alpha <1$, do not satisfy the same balance, and the modulus inequality cannot be made to close in this direct form.

Continuation Criterion and the Role of Lipschitz Control

For smooth SQG solutions, breakdown can occur only if the gradient becomes uncontrollable. A standard continuation criterion is governed by quantities such as

$${\int }_0^T\Vert \nabla \theta (t){\Vert }_{L^{\infty }}dt.$$

Thus a preserved modulus of continuity with finite slope at the origin gives precisely the missing a priori estimate. The modulus argument is not an auxiliary regularity improvement; it closes the global existence proof.

The Riesz transform creates the main obstruction. If $\theta$ has modulus $\omega$, then $u=R^{\perp }\theta$ is not Lipschitz with the same modulus. The singular integral estimate produces the larger modulus $\Omega$, with contributions from scales below and above $\xi$. This is why the proof uses a specially curved modulus instead of a simple Lipschitz bound.

Breakthrough Calculation in Detail

At the first breakthrough pair $x,y$, set $\xi =|x-y|$ and choose coordinates so that the difference quotient is aligned with $x-y$. Concavity of $\omega$ implies that the directional derivatives at $x$ and $y$ are bounded by ${\omega }^{\prime }(\xi )$ in the relevant direction. The transport contribution is therefore no larger than

$$|u(x)-u(y)|{\omega }^{\prime }(\xi )\le \Omega (\xi ){\omega }^{\prime }(\xi ).$$

The velocity estimate is the first genuinely nonlocal point. The kernel of the Riesz transform is singular at the origin but has cancellation. Splitting the integral defining $u(x)-u(y)$ into $|z|\le \xi$ and $|z|>\xi$ gives two different contributions. The near part is controlled by the local oscillation $\omega (\eta )$ at scales $\eta \le \xi$. The far part uses cancellation between the kernels centered at $x$ and $y$, producing an extra factor $\xi /\eta$ and hence the tail integral in $\Omega$.

For the dissipative part, the half-Laplacian can be written as a principal-value integral of second differences. At a breakthrough point, translated values satisfy

$$\theta (x+h)-\theta (y+h)\le \omega (\xi )$$

and

$$\theta (x+h)-\theta (y-h)\le \omega (|x-y+2h|),$$

with analogous inequalities for the reflected translations. Pairing opposite increments in the singular integral converts these inequalities into the two integral terms defining $D(\xi )$. The first term measures curvature of $\omega$ at scales below $\xi$; the second measures the deficit created by comparing the saturated pair to points farther apart.

The proof is the construction of $\omega$ so that

$$\Omega (\xi ){\omega }^{\prime }(\xi )+D(\xi )<0$$

for both small and large $\xi$.

At small scales, the paper uses a modulus with leading behavior

$$\omega (\xi )=\xi -{\xi }^{3/2}.$$

Its derivative stays finite at the origin, but its second derivative is singularly negative. This makes the dissipative second-difference term dominate the transport term. At large scales, $\omega$ is continued with very slow growth, for instance through a derivative of the form

$${\omega }^{\prime }(\xi )=\frac{\gamma }{\xi \left(4+\log (\xi /\delta )\right)}$$

after a small matching scale $\delta$. This slow tail prevents the Riesz-transform modulus $\Omega$ from accumulating too much far-field velocity. The modulus is therefore a two-scale object designed around the exact kernels of both $R^{\perp }$ and $\Lambda$.

Criticality Through Scaling

The equation

$${\partial }_t\theta +u\cdot \nabla \theta +\Lambda \theta =0$$

is invariant under

$$\theta (t,x)\mapsto \theta (\lambda t,\lambda x).$$

The $L^{\infty }$ norm is invariant under this scaling, and the maximum principle gives

$$\Vert \theta (t){\Vert }_{L^{\infty }}\le \Vert {\theta }_0{\Vert }_{L^{\infty }}.$$

However, $L^{\infty }$ control alone does not control the velocity gradient, because Riesz transforms do not map $L^{\infty }$ to $L^{\infty }$.

The modulus method fills exactly this gap. It promotes a qualitative maximum principle into a quantitative two-point maximum principle. Instead of asking whether values of $\theta$ stay bounded, it asks whether differences of values remain bounded by a scale-dependent function $\omega (\xi )$. This is why the method is stronger than energy estimates at the critical scaling.

Transferable Mechanisms

The critical SQG problem also admits a De Giorgi-style regularity approach, developed by Caffarelli and Vasseur for drift-diffusion equations with fractional diffusion (Caffarelli & Vasseur, 2010). That parallel route validates the broader principle that critical nonlocal dissipation can be converted into regularity, but the correct mechanism may be either level-set energy decay or a pointwise modulus.

Constantin and Vicol later developed nonlinear maximum principles for dissipative nonlocal operators, extending the pointwise philosophy behind the modulus argument to other active scalar and fluid models (Constantin & Vicol, 2012). For new equations, the useful question is whether the nonlocal dissipation can be made to dominate a nonlinear drift at exactly the scale where a breakthrough scenario would occur.

Links

• on Moser iteration --- both arguments turn a maximum-principle philosophy into quantitative regularity, but SQG needs a nonlocal two-point version.
• on Hardy inequalities --- nonlocal operators and scale-critical estimates often require controlling singular kernels across all scales.
• on interpolation theorems --- critical drift-diffusion analysis repeatedly balances norms that scale in different but compatible ways.

References

🐻  Caffarelli, L. & Vasseur, A. 2010. Drift Diffusion Equations with Fractional Diffusion and the Quasi-Geostrophic Equation. Annals of Mathematics 171(3), 1903–1930.

🐻  Constantin, P. & Vicol, V. 2012. Nonlinear Maximum Principles for Dissipative Linear Nonlocal Operators and Applications. Geometric and Functional Analysis 22(5), 1289–1321.

🐻  Kiselev, A., Nazarov, F. & Volberg, A. 2006. Global Well-Posedness for the Critical 2D Dissipative Quasi-Geostrophic Equation. Inventiones Mathematicae 167(3), 445–453.