Trace theorems for functions of bounded variation in metric spaces (Q1707936)

In this paper, the authors prove existence of traces of functions of bounded variation on the boundary of a certain class of domains in metric measure spaces equipped with a doubling measure supporting a 1-Poincaré inequality. \(L^1\) estimates of the trace functions are obtained without requiring the knowledge of the function in the exterior of the domain. A Maz'ya-type inequality for functions of bounded variation that vanish on a set of positive capacity is also obtained. Between the main results obtained in the paper which are too many to be presented here, we have selected the following. Theorem. Assume that \(\Omega\subset X\) is a bounded domain that supports a \((1,1)\)-Poincaré inequality, that there is a constant \(c_m>0\) such that the measure density condition \[ \mu(B(x,r)\cap\Omega)\geq c_m\mu(B(x,r)), \] holds for all \(x\in\partial\Omega\) and \(r\in (0,\operatorname{diam}(\Omega))\), and that \[ \mathcal H(B(x,r)\cap\partial\Omega)\leq C_{\partial\Omega}\mu(B(x,r))/r, \] for all \(x\in\partial\Omega\) and \(r\in(0,2\operatorname{diam}(\Omega))\). Then, there is a constant \(C_T=C_T(C_d,C_P,\lambda,\Omega)>0\), such that \[ \int_{\partial\Omega}| Tu| d\mathcal H\leq C_T\left(\int_\Omega| u| d\mu+\| Du\|(\Omega)\right). \] The concepts in the above statement are defined as follows. \((X,d,\mu)\) is a complete metric space equipped with a Borel regular outer measure \(\mu\), which measure is assumed to be doubling, namely there exists a constant \(C_d>0\), such that \[ 0<\mu(B(x,2r))\leq C_d\mu(B(x,r))<+\infty, \] for every ball \(B(x,r)\) with center \(x\in X\) and radius \(r>0\). The total variation of \(u\in BV(\Omega)\) is defined by \[ \| Du\|(\Omega)=\inf\{\lim\inf_{i\to\infty}\int_\Omega g_{u_i}d\mu;u_i\in\mathrm{Lip}_{\mathrm{loc}}(\Omega), u_i\to u\text{ in }\mathrm{Lip}_{\mathrm{loc}}(\Omega)\}, \] where each \(g_{u_i}\) is a 1-weak upper gradient of \(u_i\), that is each \(g_{u_i}\) is a nonnegative Borel function on \(\Omega\) such that for all curves \(\gamma\in\Omega\), we have \(| u(x)-u(y)|\leq_\gamma gds\), with \(x\) and \(y\) the endpoints of \(\gamma\). One says that \(\Omega\) supports a \((1,1)\)-Poincaré, if there exist the constants \(C_P>0\) and \(\lambda\geq 1\), such that for every \(B(x,r)\), \(u\in L^1_ {\mathrm{loc}}(\Omega)\) and \(g\) upper gradient of \(u\), we have \[ \frac{1}{\mu(B(x,r))}\int_{B(x,r)}| u-u_{B(x,r)}| d\mu\leq C_P\cdot r\cdot\frac{1}{\mu(B(x,\lambda r))}\int_{B(x,\lambda r)}gd\mu, \] with \(u_{B(x,r)}=\frac{1}{\mu(B(x,r))}\int_{B(x,r)}ud\mu\). The codimension 1 Hausdorff measure of \(A\subset\Omega\) is defined by \(\mathcal H(A)=\lim_{R\to 0}\mathcal H_{\mathcal R}\), where \[ \mathcal H_R(A)=\inf\{\sum\limits_{i=1}^\infty\mu(B(x_i,r_i))/r_i;A\subset\bigcup\limits_{i=1}^\infty B(x_i,r_i),r_i\leq R\}. \] A number \(Tu(x)\) is the trace of \(u\) at \(x\in\partial\Omega\) if we have \(\frac{1}{\mu(B(x,r)\cap\Omega)}\int_{\Omega\cap B(x,r)}| u_Tu| d\mu=0\).