Stability in the Liouville theorem on Heisenberg groups (Q2455316): Difference between revisions

Let \(H^n\) be the Heisenberg group which can be identified with \(R^{2n+1}\). Any point \(x\in H^n\) can be represented as \((z,t)\in C^n \times R\) where \(z=(x_1+i x_{n+1},x_2+ix_{n+2},\dots,x_n+ix_{2n}) \in C^n\) and \(t=x_{2n+1}\in R\). The group law is then defined as \[ x \cdot y=(z,t)\cdot(w,s)=(z+w,t+s+2\Im(z,w)). \] The vector fields \[ \begin{aligned} X_i &=\frac{\partial}{\partial x_i}+2x_{i+n}\frac{\partial} {\partial x_{2n+1}},\quad i=1,2,\dots,n,\\ X_{i+n}&= \frac{\partial}{\partial x_{i+n}}-2x_i\frac{\partial}{\partial x_{2n+1}},\quad i=1,2,\dots,n,\\ X_{2n+1}&= \frac{\partial}{\partial x_{2n+1}} \end{aligned} \] form the standard basis in the Lie algebra \(H_n\). The norm \(\rho(x)=(|z|^4+t^2)^{1/4}\) determines a Heisenberg metric according to the rule \(\rho(x,y)=\rho(x^{-1}\cdot y)\) for any two points \(x,y\in H^n\). Let \(B=B(x,r)=\{y\in H^n\mid\rho(x,y) <r\}\) be a ball of radius \(r>0\) centered at the point \(x\in H^n\), and let \(v=2n+2\) be the homogeneous dimension. Let \(\Omega\) be a domain in \(H^n\). The Sobolev space \(W^1_q(\Omega)\), where \(1 \leq q\leq\infty\), consists of locally integrable functions \(f:\Omega\to R\) that have generalized derivatives \(X_if\) for \(i=1,2,\dots,2n\) and \(\|f\|_{W^1_q (\Omega)}=\|f\|_{q,\Omega}+\|\nabla_Lf\|_{q,\Omega}<\infty\) where \(\nabla_Lf= (X_1f,X_2f,\dots,X_{2n}f)\). If \(f\in W_q^1(U)\) for every open set \(U\) with \(\overline U\subset\Omega\), then \(f\) is said to be of class \(W^1_{q,\text{loc}} (\Omega)\). A mapping \(f:\Omega\to H^n\), where \(\Omega\subset H^n\), belongs to the class \(W^1_{q,\text{loc}} (\Omega,H^n)\) if the following conditions hold: (a) for any \(y\in H^n\), the function \([f]_y(x)=\rho(f(x),y)\) belongs to the class \(W^1_{q, \text{loc}}(\Omega)\); (b) the family of functions \(\{\nabla_L [f]_y \}_{y\in H^n}\) has a majorant in \(L_{q,\text{loc}}(\Omega)\); that is, there exists a function \(g\in L_{q,\text{loc}}(\Omega)\) such that it does not depend on \(y\) and \(|\nabla_L[f]_y(x)|\leq g(x)\) for almost all \(x\in \Omega\). For any mapping \(f\) of Sobolev class, the formal horizontal differential \(D_hf=(X_i f_j)_{i,j=1,2,\dots,2n}\) generates a homomorphism \(Df:H_n\to H_n\) of the Lie algebra. The determinant of the matrix \(Df(x)\) is called the (formal) Jacobian of the mapping \(f\) and denoted by \(J(x,f)\). Let \(f:U\to H^n\) be a mapping of class \(W^1_{v,\text{loc}}(U,H^n)\) defined on an open set \(U\subset H^n\). We say that \(f\) is a mapping with bounded distortion if there exists a constant \(K\geq 1\) such that \(|D_hf(x)|^v\leq KJ(x,f)\) almost everywhere on \(U\). The least constant \(K\) is called external distortion coefficient of the mapping \(f\) and denoted \(K_0(f)\). The number \(K_0(f)^{\frac{1} {n+1}}\) is called the (linear) distortion coefficient of \(f\) and denoted by \(K(f)\). Let \(\text{SU}(1,n+1)=\{g\in \text{SU}(n+2):\langle gy,gy\rangle= \langle y,y\rangle\}\) where \(\langle y,y\rangle=y_0 \overline y_0-\sum^{n+1}_{i=1}y_i\overline y_i\). The group \(\text{SU}(1,n+1)\), together with the reflection \((z,t)\mapsto(\overline z,-t)\), generates the group \(M_n\) of Möbius transformations. The main result of this paper is the following theorem on local stability in the Sobolev norm. Theorem 1. There exist positive constants \(A_0\) and \(\varepsilon_0\) such that, for every \(n>1\), any mapping \(f:U \to H^n\) with bounded distortion \(K= K(f)\leq 1+\varepsilon_0\) belongs to \(W^1_{p, \text{loc}}(U,H^n)\) for all \(p\in[v,\frac{A_0}{\sqrt{K-1}})\). Moreover, for any ball \(B=B(a,r)\) such that \(B(a,8r)\subset U\), there exists a mapping \(\varphi\in M_n\) such that \(\varphi\neq\infty\) on \(B(a,\frac{10r}{3})\), \[ \rho(\varphi^{-1} \circ f(x),x)\leq C_1r\sqrt{K-1}\text{ for all }x\in B\biggl(a,\frac{8r}{9}\biggr), \] \[ \|D_h (\varphi^{-1}\circ f)-I\|_{p,B}\leq C_2r^{\frac{v}{p}}(K-1)^{\frac{v+p}{2p}}, \] \[ \|D_hf-D_h\varphi\|_{p,B}\leq C_3\sqrt{K-1}\|D_h\varphi\|_{p,B}, \] for any number \(p\in[v,\frac{A_0(1-\delta)}{\sqrt{K-1}})\), where \(\delta>0\). Moreover, there exists a number \(A_1\in(v,\frac{A_0}{\sqrt{K-1}})\) such that \[ \|D_h (\varphi^{-1}\circ f)-I\|_{p,B}\leq C_4r^{v/p}(K-1) \] for all \(p\in [v, A_1)\). The constants \(C_1,\dots, C_4\) depend only on \(n,p\), and \(\delta\). Theorem 1 can be used to prove stability in the Sobolev norm on John domains. A domain \(U\subset H^n\) is called a John domain with inner radius \(\alpha\) and outer radius \(\beta\), or a domain of class \(J(\alpha,\beta)\), where \(0<\alpha\leq \beta<\infty\), if there exists a point \(p_0\in U\) such that any other point \(p \in U\) can be joined with \(p_0\) in \(U\) by a rectifiable curve \(\gamma(s)\), where \(0\leq s\leq l\leq \beta\), \(\gamma(0)=p\), \(\gamma(l)=p_0\) and \(\rho(\gamma(s), \partial u)\geq\frac{\alpha}{l}s\) for all \(s\in[0,l]\). Theorem 2. Let \(U\) be a domain of class \(J(\alpha,\beta)\) in \(H^n\), where \(n>1\). Then there exist positive numbers \(A_2,A_3\) such that for any mapping \(f:U \to H^n\) with bounded distortion \(K=K(f)\), there exists a mapping \(\varphi\in M_n\) for which the following conditions hold: (a) if \(K<1 +A_2(\frac{\alpha}{\beta})^2\), then \(\rho(\varphi^{-1}\circ f(x),x)\leq C_1\frac{\beta^2}{\alpha}\sqrt{K-1}\) for all \(x\in U\); (b) if \(K<1 +\varepsilon_0\) and \(p\in[v,\min\{\frac{A_3}{K-1} (\frac{\alpha} {\beta})^{v+3},\frac{A_0(1-\delta)}{\sqrt{K-1}}\})\), where \(\delta >0\), then \[ \int_U|D_h(\varphi^{-1}\circ f)(x)-I|^p\,dx\leq C_2\beta^v(K-1)^{\frac {v+p}{2}}, \] \(\int_U|D_h(\varphi^{-1}\circ f)(x)-I|^p\,dx\leq C_3\beta^v(K-1)^p\) provided that \(p<A_1\). The constant \(C_1\) depends only on \(n\), the constants \(C_2\) and \(C_3\) depend on \(\frac {\beta}{\alpha}\), \(n,p\), and \(\delta\), the constants \(\varepsilon_0, A_0\) and \(A_1\) are the same as in Theorem 1. For John domains in the group \(H^1\), the following theorem is valid. Theorem 3. Let \(U\) be a domain of class \(J(\alpha,\beta)\) in \(H^1\). Then there exists an \(\varepsilon=\varepsilon(\frac{\beta}{\alpha})>0\) and functions \(\lambda_i:[0, \varepsilon)\to[0,\infty)\) such that \(\lambda_i(t)\to 0\) as \(t\to 0\) for \(i=0,1\) and, for a mapping \(f:U\to H^1\) with bounded distortion \(K=K(f)<1+\varepsilon\), there exists a mapping \(\varphi\in M_1\) such that \(\rho(\varphi^{-1}\circ f(x), x)\leq\frac{\beta^2} {\alpha}\lambda_0(K-1)\) for all \(x\in U\) and \(\int_U|D_h (\varphi^{-1} \circ f)(x)-I|^pdx\leq C\beta^4(\lambda_1(K-1))^p\) for all \(p\in[4,\min \{\frac{E_1}{(\lambda_0(K-1))^2}(\frac{\alpha}{\beta})^7,\frac{E_2} {\lambda_1(K-1)}\})\). The constants \(E_1\) and \(E_2\) are positive, and \(C\) depends on \(\frac{\alpha}{\beta}\) and \(p\).