on the Kenig-Merle critical element method

Overview

Kenig and Merle introduced a concentration-compactness and rigidity method for the energy-critical focusing nonlinear Schrödinger equation (Kenig & Merle, 2006). Below the ground-state threshold, radial solutions scatter. On the other side of the threshold, finite-time blow-up can occur.

The method has become a standard architecture for critical dispersive equations: assume scattering fails, extract a minimal nonscattering critical element, prove its orbit is compact modulo symmetries, and then rule it out by a rigidity identity.

Energy-Critical Focusing NLS

For spatial dimensions $3\le d\le 5$, consider

$$i{\partial }_{t}u+\Delta u+|u{|}^{\frac{4}{d-2}}u=0,u(0)=u_0\in {\dot{H}}^1({\mathbb{R}}^d).$$

The scaling

$$u_{\lambda }(t,x)={\lambda }^{\frac{d-2}{2}}u({\lambda }^{2}t,\lambda x)$$

leaves the homogeneous Sobolev norm invariant:

$$\Vert \nabla u_{\lambda }(0){\Vert }_{L^2}=\Vert \nabla u_0{\Vert }_{L^2}.$$

This is why the equation is energy-critical.

The conserved energy is

$$E(u)=\frac{1}{2}{\int }_{{\mathbb{R}}^d}|\nabla u{|}^{2}dx-\frac{d-2}{2d}{\int }_{{\mathbb{R}}^d}|u{|}^{\frac{2d}{d-2}}dx.$$

The focusing sign makes the potential term negative. Large concentration can therefore drive collapse.

Ground State Threshold

The stationary ground state $W$ solves

$$-\Delta W=|W{|}^{\frac{4}{d-2}}W.$$

It is the optimizer for the sharp Sobolev inequality

$$\Vert f{\Vert }_{L^{\frac{2d}{d-2}}}\le C_d\Vert \nabla f{\Vert }_{L^2}.$$

This makes $W$ the natural boundary between scattering and blow-up behavior.

Kenig-Merle threshold theorem

Assume $u_0$ is radial and

$$E(u_0)<E(W),\Vert \nabla u_0{\Vert }_{L^2}<\Vert \nabla W{\Vert }_{L^2}.$$

Then the solution exists globally and scatters in both time directions.

If instead $E(u_0)<E(W)$ but

$$\Vert \nabla u_0{\Vert }_{L^2}>\Vert \nabla W{\Vert }_{L^2},$$

then finite-time blow-up follows under additional hypotheses such as finite variance, with radial arguments covering the threshold setting treated in the paper.

Scattering means that there exist free Schrödinger solutions $e^{it\Delta }{u}_{\pm }$ such that

$$\Vert u(t)-e^{it\Delta }{u}_{\pm }{\Vert }_{{\dot{H}}^1}\to 0$$

as $t\to \pm \infty$.

Coercivity Below the Ground State

The first key lemma is variational. If

$$E(u)<E(W),\Vert \nabla u{\Vert }_2<\Vert \nabla W{\Vert }_2,$$

then the nonlinear part cannot overpower the kinetic energy. More precisely, the sharp Sobolev inequality implies a positive coercivity gap:

$$\int |\nabla u{|}^2-\int |u{|}^{\frac{2d}{d-2}}\ge c\int |\nabla u{|}^2.$$

This prevents the virial functional from having the wrong sign and keeps the solution in the scattering side of the threshold as long as it exists.

The same variational analysis shows that data with gradient norm above $W$ remain above the threshold. The ground state is therefore not just an example; it divides phase space into invariant regions.

Critical Element

Assume the theorem is false. Then there is a smallest energy level below $E(W)$ at which scattering fails. Choose a sequence of nonscattering solutions with energies decreasing to this critical energy.

The profile decomposition for bounded sequences in ${\dot{H}}^1$ expresses initial data as a sum of decoupled concentrating profiles:

$$u_{0,n}=\sum _{j=1}^J{\lambda }_{j,n}^{-\frac{d-2}{2}}{\varphi }^j\left(\frac{x-x_{j,n}}{{\lambda }_{j,n}}\right)+r_n^J,$$

with asymptotic orthogonality of scales, centers, and times. Energy decouples among the profiles.

The minimality of the critical energy forces only one nonlinear profile to survive. If two profiles carried positive energy, each would lie below the critical threshold and scatter; perturbation theory would then imply scattering of the original sequence, a contradiction.

Critical element

The surviving solution $u_c$ is a minimal nonscattering solution. Its orbit is precompact modulo the scaling symmetry: there exists a scale function $\lambda (t)>0$ such that

$$\left\{\lambda (t{)}^{-\frac{d-2}{2}}{u}_c\left(t,\frac{x}{\lambda (t)}\right):t\in I\right\}$$

has compact closure in ${\dot{H}}^1$.

This compactness property is much stronger than boundedness. It says the solution is a coherent object that neither disperses nor splits into separated bubbles.

Rigidity

The rigidity step proves that no nonzero compact critical element can exist. For radial solutions, compactness modulo scaling implies spatial localization: for every $\epsilon >0$, there is $R$ such that most of the kinetic energy lies in

$$|x|\le R/\lambda (t)$$

for all $t$.

One then uses a localized virial identity. For a radial cutoff $a_R(x)$ approximating $|x{|}^2$, define

$$V_R(t)=\int a_R(x)|u_c(t,x){|}^{2}dx.$$

Differentiating twice gives, schematically,

$$V_R^{\prime \prime }(t)=8\left(\int |\nabla u_c{|}^2-\int |u_c{|}^{\frac{2d}{d-2}}\right)+\text{cutoff errors}.$$

Coercivity makes the main term positive. Compactness makes the cutoff errors small. Hence $V_R^{\prime \prime }(t)$ is uniformly positive over long intervals.

But $V_R(t)$ grows at most quadratically with a coefficient controlled by the localized mass and energy. The compactness properties force a contradiction when the virial convexity is integrated over time. Therefore the critical element cannot exist, and scattering must hold.

Structural Legacy

The proof has a modular form that reappears in wave maps, critical wave equations, Schrödinger maps, and other dispersive models:

1. a sharp variational threshold,
2. a profile decomposition,
3. perturbation theory for approximate solutions,
4. extraction of a minimal compact counterexample,
5. a monotonicity or virial argument excluding compact dynamics.

The point is not merely that the theorem solves one NLS problem. It gives a robust method for turning failure of scattering into a rigid object, then using the PDE’s conservation laws to forbid that object.

Strichartz Norms and Perturbative Stability

Scattering is measured in a scale-critical spacetime norm, typically a Strichartz norm such as

$$\Vert u{\Vert }_{L_{t,x}^{\frac{2(d+2)}{d-2}}(I\times {\mathbb{R}}^d)}.$$

If this norm is finite on the lifespan, the solution scatters. The perturbation theorem says that an approximate solution with small error in the dual Strichartz space stays close to a true solution, provided the critical norm is controlled.

This perturbation theorem is what makes profile decomposition useful. After decomposing initial data into profiles, one evolves each profile nonlinearly, adds the evolutions, and treats the interaction error perturbatively. Orthogonality of scales and centers makes the error small.

Profile Decomposition and Energy Decoupling

A bounded sequence in ${\dot{H}}^1$ can fail to be compact only through the symmetries of the equation: scaling, translation, and time translation under the linear flow. The profile decomposition extracts these failures one at a time. The remainder has vanishing linear Strichartz norm, so it is irrelevant to scattering.

Energy decoupling is essential:

$$E(u_{0,n})=\sum _{j=1}^{J}E({\varphi }^j)+E(r_n^J)+o_n(1)+o_J(1).$$

At the minimal nonscattering energy, this forces exactly one bad profile. If there were two, each would have lower energy and would scatter, and perturbation theory would make the original sequence scatter as well.

Proof: Compactness from Minimality

Let $E_c$ be the infimum of energies below $E(W)$ for which scattering fails. Choose nonscattering data $u_{0,n}$ with $E(u_{0,n})\downarrow E_c$. The linear profile decomposition writes $u_{0,n}$ as a sum of profiles with asymptotically orthogonal scales, spatial centers, and time shifts, plus a remainder whose free evolution has vanishing critical Strichartz norm.

Each profile has an associated nonlinear evolution. Energy decoupling gives

$$E(u_{0,n})=\sum _{j}E({\varphi }^j)+E(r_n^J)+o_n(1)+o_J(1),$$

and the sharp Sobolev coercivity keeps all nonzero profiles on the subthreshold side. If every nonlinear profile had energy strictly below $E_c$, then by definition each profile would scatter. Orthogonality of the parameters would make the sum of the scattering profiles an approximate solution, and the long-time perturbation theorem would force $u_n$ itself to scatter. This contradicts the choice of $u_{0,n}$.

Hence exactly one profile carries the critical obstruction. Its nonlinear evolution is the critical element $u_c$. To prove compactness of its orbit, repeat the same profile-decomposition argument on an arbitrary sequence of times $t_n$. If the rescaled data $u_c(t_n)$ had two nontrivial profiles, or a nontrivial dispersive remainder, the same lower-energy scattering and perturbation argument would contradict minimality. After applying the scaling symmetry, every time sequence has a convergent subsequence in ${\dot{H}}^1$.

Minimality therefore converts qualitative failure into almost periodicity modulo scaling. Rigidity arguments are powerful precisely because they act on this much smaller class of solutions: a compact critical element cannot disperse, split, or radiate without violating the definition of $E_c$.

Localized Virial Mechanics

The exact virial identity for finite-variance solutions uses

$$V(t)=\int |x{|}^2|u(t,x){|}^{2}dx.$$

Energy-critical solutions need not have finite variance, so Kenig and Merle use a localized weight $a_R(x)$ that agrees with $|x{|}^2$ for $|x|\le R$ and flattens outside. Differentiating produces the same main expression as the formal virial identity, plus annular error terms involving $\nabla u$ and the critical nonlinearity where $|x|\sim R$.

For the critical element, compactness modulo scaling gives uniform localization in the scale-normalized variables: for every $\epsilon >0$ there is $R_0$ such that the kinetic and critical Lebesgue tails outside $R_0/\lambda (t)$ are smaller than $\epsilon$ for every $t$ in the lifespan. Radial Sobolev estimates convert this compactness into control of the nonlinear annular errors. Choosing the virial cutoff much larger than the concentration scale makes the error terms smaller than the coercive main term.

The coercive term has a fixed sign because the solution remains below the ground-state threshold and below the gradient norm of $W$. Thus the localized virial functional is forced to be uniformly convex on long time intervals after the cutoff errors are absorbed. On the other hand, the same localization bounds the size and first derivative of the localized virial quantity in terms of the cutoff scale and conserved energy. Integrating a uniformly positive second derivative over a sufficiently long interval contradicts those bounds.

This is where compactness becomes rigidity. The critical element is localized enough for virial convexity to see it, but any nonzero solution with that convexity must eventually escape the compact regime. The only compact subthreshold solution left by the rigidity argument is the zero solution, contradicting nonscattering.

Transferable Mechanisms

Kenig and Merle used the same concentration-compactness and rigidity architecture for the energy-critical focusing wave equation (Kenig & Merle, 2008). This confirms that the method is not tied to the Schrödinger flow; it applies when a critical scaling, a sharp variational threshold, a profile decomposition, and a rigidity identity are all available.

In other dispersive problems, the practical checklist is therefore precise. One first identifies the scale-invariant norm and the threshold object. One then proves perturbative stability and profile decomposition at that scale. If a minimal bad solution can be forced to have compact orbit modulo symmetries, a virial, Morawetz, monotonicity, or channel-of-energy argument may rule it out.

Links

• on interpolation theorems --- Strichartz and Sobolev estimates are part of the functional-analytic background behind scattering theory.
• on Grönwall’s inequality --- perturbative stability estimates use nonlinear continuity arguments whose scalar model is Grönwall control.
• on Hardy inequalities --- threshold variational estimates often rely on sharp functional inequalities and coercivity, similar in spirit to Hardy-type bounds.

References

🐻  Kenig, C.E. & Merle, F. 2006. Global Well-Posedness, Scattering and Blow-Up for the Energy-Critical, Focusing, Non-Linear Schrodinger Equation in the Radial Case. Inventiones Mathematicae 166(3), 645–675.

🐻  Kenig, C.E. & Merle, F. 2008. Global Well-Posedness, Scattering and Blow-Up for the Energy-Critical Focusing Non-Linear Wave Equation. Acta Mathematica 201(2), 147–212.