Sharp weighted Young's inequalities and Moser-Trudinger inequalities on Heisenberg type groups and Grushin spaces (Q2498497)

The author gives sharp weighted Moser-Trudinger inequalities for \(x\)-symmetric functions on the Grushin plane and for first-layer symmetric functions on groups of Heisenberg type. Let \(X=\partial_{x}\) and \(Y=x\partial_{y}\) be the Grushin system on \({\mathbb R}^2\). We denote by \(\nabla_{G}f=(Xf, Yf)\) the associated horizontal gradient. Theorem 1. Let \(\alpha \geq 1\) and let \(w(x,y)=|x|^{\alpha}\). Set \(A(\alpha)=(3+\alpha)B(\frac{1}{2}, \frac{2+\alpha}{2})^{1/(2+\alpha)}\). Then there exists a constant \(C_{0}\) such that \[ \sup_{D,f}\frac{1}{\int_{D}w}\int_{D}\exp \Biggl\{ A(\alpha) \biggl(\frac{|f|} {||\nabla_{G}f||_{L^{3+\alpha}(D,w)}} \biggr)^{(3+\alpha)'} \Biggr\} w \leq C_{0},\tag{1} \] where the supremum is taken over all \(x\)-symmetric domains \(D \subset {\mathbb R}^2\) with finite area and all nonconstant \(x\)-symmetric functions \(f \in HW_{0}^{1, 3+\alpha}(D,w)\). Moreover, the supremum in \({\text (1)}\) is infinite if \(A(\alpha)\) is replaced by any strictly larger value. Here we have denoted by \(HW_{0}^{1,p}(D, w)\) the horizontal weighted Sobolev class, that is, the completion of \(C_{0}^{\infty}(D)\) in the norm \(||f||_{L^{p}(D,w)} + ||\nabla_{G}f||_{L^{p}(D,w)}\) and \(Q' =Q/(Q-1)\). Theorem 2. Let \(G \) be a Heisenberg type group with horizontal layer \(V_{1} \subset {\mathcal G}\) of dimension \(k_{1} \geq 2\) and nonhorizontal layer \(V_{2} \subset {\mathcal G}\) of dimension \(k_{2}\). Let \(w(g)=|x_{1}(g)|^{-a}, \;\;0 \leq a \leq k_{1}-2\). Set \(A(G, w)=(Q-a)c_{Q-2a}^{1/(Q-a-1)}\) where \(Q=k_{1}+2k_{2}\) is the homogeneous dimension of \(G\), \(m=k_{1}+k_{2}\) is the topological dimension of \(G\) and \[ c_{p}=\frac{2\pi^{m/2}\Gamma((k_{1}+p)/4)}{4^{k_{2}}\Gamma(k_{1}/2) \Gamma((Q+p)/4)} \] is the integral of the \(p\)-th power of \(|\nabla_{G}N|\) over the unit sphere \(\{ N=1\}\) relative to a certain canonically defined surface area measure, where \(N\) is the standard homogeneous norm in the group. Then there exists a constant \(C_{0}\) such that \[ \sup_{\Omega, F}\frac{1}{\int_{\Omega}w}\int_{\Omega}\exp\Biggl\{ A(G,w) \biggl(\frac{|F|}{||\nabla_{G}F||_{L^{Q-a}(\Omega,w)}} \biggr)^{(Q-a)'}\Biggr\} w \leq C_{0},\tag{2} \] where the supremum is taken over all first-layer symmetric domains \(\Omega \subset G\) with finite Haar measure and all nonconstant first-layer symmetric functions \(F \in HW_{0}^{1, Q-a}(\Omega, w)\). The supremum in \({\text (2)}\) becomes infinite if \(A(G,w)\) is replaced by any strictly larger value. Here the function \(x_{1}: G \rightarrow {\mathbb R}^{k_{1}}\) in Theorem 2 assignes to each group element \(g \in G\) its horizontal component. A function \(U: G \rightarrow {\mathbb R}\) is called first-layer symmetric if \(U(g)\) depends only on \(|x_{1}(g)|\) as well as the values of the higher-layer components of \(g\), and a domain \(\Omega\) is called \textit{first-layer symmetric} if its characteristic function is first-layer symmetric. To prove these theorems the author establishes weighted Young's inequalities on arbitrary Carnot groups. The proofs use some sharp estimates for hypergeometric functions.