Large-time asymptotics of solutions to the Kramers-Fokker-Planck equation with a short-range potential (Q2018325)

The paper describes the asymptotic behavior of the solution to the Kramers-Fokker-Planck equation written as \(\partial _{t}u(t;x,v)+Pu(t;x,v)=0\) in \((0,+\infty )\times \mathbb{R}^{n}\times \mathbb{R}^{n}\) where \(n\geq 1\) and \(P\) is the Kramers-Fokker-Planck operator defined as \(P=-\Delta _{v}+ \frac{1}{4}|v|^{2}-\frac{n}{2}+v\cdot \nabla _{x}-\nabla V(x)\cdot \nabla _{v}\), \(V\) being a \(C^{1}\) potential which satisfies \(\int_{\mathbb{R} ^{n}}e^{-V(x)}dx=1\) and \(|V(x)|+\left\langle x\right\rangle |\nabla V(x)|\leq C\left\langle x\right\rangle ^{-\rho }\) for some \(\rho \geq -1\) and with \(\left\langle x\right\rangle =(1+|x|^{2})^{1/2}\). The main result of the paper proves the existence of \(C>0\) such that the spectrum \(\sigma (P) \) of \(P\) satisfies \(\sigma P)\cap \{z;|\mathfrak{I}z|>C,\mathfrak{R}z\leq \frac{1}{C}|\mathfrak{I}z|^{1/2}\}=\emptyset \) and the resolvent \( R(z)=(P-z)^{-1}\) satisfies \(\left\| R(z)\right\| \leq \frac{C}{ |z|^{1/2}}\) and \(\left\| (1-\Delta _{v}+v^{2})^{1/2}R(z)\right\| \leq \frac{C}{|z|^{1/4}}\) for \(|\mathfrak{I}z|>C\) and \(\mathfrak{R}z\leq \frac{1 }{C}|\mathfrak{I}z|^{1/2}\). In the case where \(n=3\) and \(\rho >1\) then the operator \(P\) has no eigenvalue in a neighborhood of 0 and the semigroup \(S(t) \) associated to \(-P\) satisfies \(\left\| S(t)\right\| _{\mathcal{L} ^{2,s}\rightarrow \mathcal{L}^{2,s}}\leq C_{s}t^{-3/2}\) for every \(t>0\), where \(\mathcal{L}^{2,s}=L^{2}(\mathbb{R}_{x,v}^{2n};\left\langle x\right\rangle ^{2s}dxdv)\). The author first studies the free Kramers-Fokker-Planck operator defined as \(P_{0}=v\cdot \nabla _{x}-\Delta _{v}+\frac{1}{4}|v|^{2}-\frac{n}{2}\) and the semigroup associated to \(P_{0}\) . For the proof of the main theorem, the author essentially writes \( (P-z)R_{0}(z)=1+R_{0}(z)W=1+G_{0}W+O(|z|^{\epsilon })\) in the space \( \mathcal{B}(0,-s;0,-s)\) of continuous linear operators from \(\mathcal{H} ^{0,s}\) to \(\mathcal{H}^{0,-s}\), for \(1<s<(1+\rho )/2\) and \(z\) near 0 and non real, where \(R_{0}\) is the resolvent of \(P_{0}\) and \(W=-\nabla V(x)\cdot \nabla _{v}\). He proves that \(G_{0}W\) is a compact operator in \(\mathcal{L} ^{2,-s}\) for \(1<s<(1+\rho )/2\) and that \(\ker _{\mathcal{L} ^{2,-s}}(1+G_{0}W)=\{0\}\).