Past exam of the mathematics course of the University of Cambridge / 2022 / iii / Paper 154 (Introduction to Non-Linear Analysis)

Choose a test function $\chi$ that equals one on the unit ball and vanishes outside the ball of radius two. Sinceintegration by parts givesThe first term on the right is supported where $1\le |x|\le 2$, where $u^2\le |x{|}^{\alpha }{u}^2$. The Cauchy-Schwarz inequality and Young inequality bound the second term byAfter absorbing the first integral,

This estimate also proves completeness. Indeed, a Cauchy sequence in $\Sigma$ is Cauchy in $H^1$ locally, while its gradients and the functions $|x{|}^{\alpha /2}{u}_n$ converge in $L^2$. The limits agree locally with a function $u$, so $u\in \Sigma$ and $u_n\to u$ in the energy norm. Thus $\Sigma$ is a Hilbert space; it is the confining-potential energy space for $V(x)=|x{|}^{\alpha }$.

On the unit ball, part 1 controls the $L^2$ norm. Outside it, $|x{|}^{\alpha }\ge 1$, soHence $\Vert u{\Vert }_2\le C\Vert u{\Vert }_{\Sigma }$, proving that the embedding $\Sigma \hookrightarrow L^2$ is continuous.

For compactness, let $(u_n)$ be bounded in $\Sigma$. On each ball, it is bounded in $H^1$, so the Rellich-Kondrachov compactness theorem gives a subsequence convergent in local $L^2$. The tail estimateis uniform in $n$ and tends to zero as $R\to \infty$. A diagonal argument therefore gives convergence in all of $L^2$. This is the compact embedding of a confining-potential energy space.

For fixed $f\in L^2$, the functionalis bounded on $\Sigma$ by the continuous embedding from part 2:The Riesz representation theorem gives a unique $T(f)\in \Sigma$ such that $⟨v,T(f){⟩}_{\Sigma }={\ell }_f(v)$ for every $v\in \Sigma$. It also givesso the linear operator $T:L^2\to \Sigma$ is bounded.

As a map into $L^2$, $T$ factors asThe first arrow is bounded by part 3 and the second is the compact embedding from part 2; hence $T:L^2\to L^2$ is a compact operator.

Writing $u=T(f)$ in the defining identity givesIn particular this holds for every test function $v$, so the definition of a distributional derivative yields

Minimize $\Vert u{\Vert }_{\Sigma }^2$ subject to $\Vert u{\Vert }_2=1$. A minimizing sequence is bounded in $\Sigma$, and part 2 supplies a subsequence converging strongly in $L^2$ and weakly in $\Sigma$. The constraint survives the strong convergence, while weak lower semicontinuity of the norm shows that the limit $\psi$ attains the minimum. Since $\Vert |\psi |{\Vert }_2=\Vert \psi {\Vert }_2$ and $|\nabla |\psi ||\le |\nabla \psi |$ almost everywhere, we may take $\psi \ge 0$.

The Lagrange multiplier equation iswhere testing with $v=\psi$ shows that $\lambda =\Vert \psi {\Vert }_{\Sigma }^2>0$. ThusThis is the ground-state eigenfunction of a confining Schrödinger operator.

Set $p=2+4/d$. For $u_{a,\lambda }(x)=au(\lambda x)$, the change of variables $y=\lambda x$ givesandThe factors cancel, so the Weinstein functional satisfies $J(u_{a,\lambda })=J(u)$.

The Gagliardo-Nirenberg interpolation inequality givesConsequently $J(u)\ge C_d^{-1}$ for every nonzero $u\in H^1$, and therefore $I>0$.

By the two invariances from part 1, normalize a minimizing sequence $(u_n)$ so thatReplacing $u_n$ by $|u_n|$ and then by its symmetric decreasing rearrangement preserves its $L^2$ and $L^p$ norms and does not increase the gradient norm. A harmless dilation restores the normalization, so we may take the sequence nonnegative, radial, and radially decreasing.

The sequence has a weakly convergent subsequence in $H^1$. Radial compactness and the fixed scale give strong convergence in $L^p$; in particular the limit is nonzero because $\Vert u_n{\Vert }_p^p\to I^{-1}$. Weak lower semicontinuity of the $L^2$ and gradient norms then shows that the limit attains the infimum. This is the existence of a Weinstein-functional minimizer.

WriteDifferentiating $\log J(u+th)$ at $t=0$ for a real test function $h$ givesAfter integration by parts,Rescaling the dependent and independent variables reduces this Euler-Lagrange equation to the ground-state equation for $Q$. The uniqueness of its positive radial solution and the equality cases in rearrangement show that all minimizers areThis is the classification of Weinstein-functional minimizers.

The Pohozaev identity for the mass-critical NLS ground state givesSince $Q$ is a minimizer,

The definition of the infimum $I$ givesSubstituting the value of $I$ from part 4 into the energy yieldsThis is coercivity below the NLS ground-state mass.

Both Mass conservation for the nonlinear Schrödinger equation and Energy conservation for the nonlinear Schrödinger equation hold throughout the maximal lifespan. Since $\Vert u_0{\Vert }_2<\Vert Q{\Vert }_2$, the Sharp Gagliardo-Nirenberg inequality gives the uniform estimateThus the gradient norm stays bounded. The Blowup alternative for the nonlinear Schrödinger equation rules out a finite endpoint of the lifespan in either time direction, so the solution is global.

For real $h\in \mathcal{D}({\mathbb{R}}^d)$, differentiation under the integral givesThe ground-state equation $\Delta Q-Q+Q^{1+4/d}=0$ and integration by parts reduce this to

In particular, along the amplitude direction $h=Q$ the derivative is $-\Vert Q{\Vert }_2^2<0$. Hence $u_0=(1+\delta )Q$ has negative energy for every sufficiently small $\delta >0$, whilewhen $\delta$ is chosen small enough. The ground state has finite variance, so negative-energy blowup for the mass-critical focusing nonlinear Schrödinger equation shows that the corresponding solution blows up in finite time.

Under the mass-critical spatial scalingthe $L^2$ norm is invariant and the gradient norm is multiplied by ${\lambda }_n$. Therefore chooseFinite-time blowup and the Blowup alternative for the nonlinear Schrödinger equation imply $\Vert \nabla u(t_n){\Vert }_2\to \infty$, so ${\lambda }_n\to 0$.

The energy has scaling degree two:where the second equality uses Energy conservation for the nonlinear Schrödinger equation.

The profile decomposition modulo translations says that every bounded sequence $(w_n)$ in $H^1({\mathbb{R}}^d)$ has, after passage to a subsequence,where the translation parameters are asymptotically orthogonal,the squared $L^2$ and gradient norms decouple,and the remainder vanishes in every subcritical norm:

Part 3 and Mass conservation for the nonlinear Schrödinger equation giveThe last relation and the energy formula implywhere the final equality follows from the Pohozaev identity for the mass-critical NLS ground state. Thus $(v_n)$ is a minimizing sequence for the Weinstein functional with the same normalization as $Q$.

Apply the profile decomposition from part 4. The Sharp Gagliardo-Nirenberg inequality bounds each profile by the product of its gradient energy and its mass to the power $2/d$. Since the total mass is exactly $\Vert Q{\Vert }_2^2$, any split into two nonzero profiles would make the limiting inequality strict. Hence precisely one profile carries all the mass and gradient energy. The norm decouplings then make the remainder converge strongly to zero in $H^1$. For suitable translations $x_n$,and the Sobolev embedding theorem gives the required strong convergence in $L^{2+4/d}$. This is the compactness of a mass-critical minimizing sequence.