The \(\varepsilon - \varepsilon^\beta\) property in the isoperimetric problem with double density, and the regularity of isoperimetric sets (Q783318)

From MaRDI portal
scientific article
Language Label Description Also known as
English
The \(\varepsilon - \varepsilon^\beta\) property in the isoperimetric problem with double density, and the regularity of isoperimetric sets
scientific article

    Statements

    The \(\varepsilon - \varepsilon^\beta\) property in the isoperimetric problem with double density, and the regularity of isoperimetric sets (English)
    0 references
    0 references
    0 references
    0 references
    12 August 2020
    0 references
    Consider two lower semi-continuous locally summable functions \(f:\mathbb{R}^N\to\mathbb{R}^+\) and \(h:\mathbb{R}^N\times\mathbb{S}^{N-1}\to\mathbb{R}^+\) (called densities). The weighted volume and perimeter of a Borel set \(E\subset\mathbb{R}^N\) are defined, respectively, as \[ |E|=\int_Ef(x)d x\quad \mathrm{and} \quad P(E)=\int_{\partial^*E}h(x,\nu_E(x))d\mathcal{H}^{N-1}(x), \] where \(\partial^*E\) is the reduced boundary of every set \(E\) of locally finite perimeter, \(\mathcal{H}^N\) is the Hausdorff measure in \(\mathbb{R}^N\), and \(\nu_E(x)\in\mathbb{S}^{N-1}\) is the outer normal at \(x\in\partial^*E\). If \(h(x,\nu)=f(x)\) for every \(x\in\mathbb{R}^N\) and \(\nu\in\mathbb{S}^{N-1}\), then the problem reduces to the single density. The isoperimetric problem with single density is a wide generalization of the classical Euclidean isoperimetric problem, and it has been deeply studied in the last decades. The case of double density is yet a further important generalization, since many of the possible applications correspond to two different densities. The simplest example is given by Riemannian manifolds, which locally behave as \(\mathbb{R}^N\) with double density, where \(f\) is the norm of the Riemannian metric, and \(h\) is its derivative. The isoperimetric problem consists in finding sets \(E\) of minimal perimeter, if they exist, among those of fixed volume. Existence, boundedness and regularity of isoperimetric sets are the three main problems. A fundamental tool to study the isoperimetric problem is the classical \(\varepsilon-\varepsilon\) regularity property of Almgren, which basically says that one can locally modify a set \(E\) by changing its volume by a small, positive or negative, quantity, while increasing the perimeter by at most a quantity \(C|\varepsilon|\). In [J. Reine Angew. Math. 728, 65--103 (2017; Zbl 1369.49061)], \textit{E. Cinti} and \textit{A. Pratelli} introduced the so-called \(\varepsilon-\varepsilon^\beta\) property. If \(E\subset\mathbb R^N\) is a set of locally finite perimeter, \(0\le\beta\le 1\), and \(C>0\), then \(E\) is said to satisfy the \(\varepsilon-\varepsilon^\beta\) property with constant \(C\) relative to the densities \(f\) and \(h\) if there exist a ball \(B\) with \(\mathcal{H}^{N-1}(B\cap\partial^*E)>0\) and a constant \(\overline\varepsilon>0\) such that, for every \(|\varepsilon|<\overline\varepsilon\), there is a set \(F\subset\mathbb R^N\) satisfying \[ F\Delta E\subset\!\subset B, \quad |F|-|E|=\varepsilon, \quad P(F)-P(E)\le C|\varepsilon|^\beta. \] In the case of single density, the authors have shown that whenever \(f\) is Hölder continuous with some exponent \(0\le\alpha\le 1\), then the \(\varepsilon-\varepsilon^\beta\) property holds for every set of locally finite perimeter for some \(\beta=\beta(N,\alpha)\). Moreover, if \(f\) is locally bounded and continuous, the \(\varepsilon-\varepsilon^\beta\) property holds with \(\beta=\frac{N-1}{N}\) and with any positive constant \(C\). As a consequence of the \(\varepsilon-\varepsilon^\beta\) property, boundedness and \(C^{1,\eta}\) regularity of isoperimetric sets for some \(\eta=\eta(N,\alpha)\) were obtained. In this paper, the authors extend the above results to the general case of a double density. They show that if \(f\) ang \(h\) are bounded, \(h\) is locally \(\alpha\)-Hölder in spatial variable for some \(0\le\alpha\le 1\), and \(E\subseteq\mathbb{R}^N\) is a set of finite perimeter, then \(E\) satisfies the \({\varepsilon-\varepsilon^\beta}\) property with some \(C\), where \(\beta\) is \[ \beta=\beta(\alpha,N)=\frac{\alpha+(N-1)(1-\alpha)}{\alpha+N(1-\alpha)}. \] Next, they show that if there exists a constant \(M>0\) such that \(\frac{1}{M}\le f(x)\le M\), \(\frac{1}{M}\le h(x,\nu)\le M\) for all \(x\in\mathbb{R}^N\) and \(\nu\in\mathbb{S}^{N-1}\), and \(E\) is an isoperimetric set satisfying the \(\varepsilon-\varepsilon^\beta\) property with \(\beta=\frac{N-1}{N}\) and with an arbitrary small constant \({C>0}\), then \(E\) is bounded. A set \(E\subseteq\mathbb{R}^N\) of locally finite perimeter is said to be quasi-minimal if there exists a constant \(K > 0\) such that \(P_{\mathrm{Eucl}}(E,B_r(x))\le Kr^{N-1}\) for every ball \(B_r(x)\), and \(E\) is called a \(\omega\)-minimal set for some continuous and increasing function \(\omega:\mathbb{R}^+\to\mathbb{R}^+\) with \(\omega(0)=0\), if \(P_{\mathrm{Eucl}}(E,B_r(x))\le P_{\mathrm{Eucl}}(H,B_r(x))+\omega(r)r^{N-1}\) for every ball \(B_r(x)\) and every set \(H\Delta E\subset\!\subset B_r(x)\). The set \(E\) is called porous if there exists a constant \(\delta >0\) such that, for every \(x\in\partial E\) and every \(r > 0\), there are two balls \(B_1,B_2\subseteq B_r(x)\), both with radius \(\delta r\) such that \(B_1\subseteq E\) and \(B_2\subseteq \mathbb{R}^N\setminus E\). The authors show that if a set \(E\) of locally finite perimeter is locally quasi-minimal, then it is porous and \(\partial E=\partial^*E\) up to \(\mathcal{H}^{N-1}\)-negligible sets. Additionally, if it is locally \(\omega\)-minimal with \(w(r)=Cr^{2\eta}\) for some \(\eta\in(0,\frac12]\), then \(\partial E\) is \(C^{1,\eta}\). Finally, as a consequence, the authors prove that if \(f\) and \(h\) are locally bounded, then any isoperimetric set is porous, and its reduced boundary coincides \(\mathcal{H}^{N-1}\)-a.e. with its topological boundary. In addition, if \(h=h(x)\) is locally \(\alpha\)-Hölder for some \(\alpha\in (0,1]\), then \(\partial E\) is \(C^{1,\eta}\), where \(\eta=\eta(N,\alpha)=\frac{\alpha}{2N(1-\alpha)+2\alpha}\).
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    \(\varepsilon-\varepsilon^\beta\) property
    0 references
    isoperimetric problem
    0 references
    anisotropic perimeter
    0 references
    Finsler surface energy
    0 references
    boundedness
    0 references
    regularity
    0 references
    0 references
    0 references