---
title: Anisotropic Outer Minkowski Content
url: https://www.emergentmind.com/topics/anisotropic-outer-minkowski-content
type: topic
---

# Anisotropic Outer Minkowski Content

Searching arXiv for the cited papers to ground the article in current literature.
Anisotropic outer Minkowski content is a first-order parallel-volume functional obtained by replacing Euclidean outer tubular neighborhoods with Minkowski sums by a fixed convex body \(C\). In the codimension-one setting developed by Chambolle, Lussardi, and Novaga, it measures the asymptotic outer volume gain \(\frac{|E+\varepsilon C|-|E|}{\varepsilon}\) of a set \(E\subset \mathbb R^n\) and, under the same hypotheses that guarantee the classical outer Minkowski content equals perimeter, converges to the anisotropic perimeter \(\int_{\partial^*E} h_C(\nu_E)\,d\mathscr H^{n-1}\) [1203.5190]. Subsequent work has clarified the dependence on the representative of \(E\), the relation with two-sided boundary Minkowski content, the role of the complement, extensions to lower-dimensional structuring elements and lower-dimensional rectifiable sets, and the distinction between metric anisotropy induced by a convex body and directional anisotropy encoded by support measures or curvature-direction measures [2508.08156], [2504.03339], [2601.22681], [2509.06545].

## 1. Geometric definition and anisotropic data

The anisotropy is encoded by a fixed closed convex body \(C\subset \mathbb R^n\), bounded and containing \(0\) in its interior [1203.5190]. Its support function is
\[
h_C(\nu)=\sup_{x\in C} x\cdot \nu,
\]
and its polar function is
\[
h_C^\circ(x):=\sup_{h_C(\nu)\le 1} x\cdot \nu.
\]
Both are convex, positively one-homogeneous, and Lipschitz, and satisfy
\[
C=\{h_C^\circ\le 1\}.
\]
Because \(0\in \operatorname{int}(C)\), there exist \(0<a<b\) such that
\[
B(0,a)\subseteq C\subseteq B(0,b),
\]
hence
\[
a|\nu|\, \le\, h_C(\nu)\,\le\, b|\nu|,\qquad \frac{1}{b}|x|\,\le\, h^\circ_C(x)\,\le\, \frac{1}{a}|x|.
\]
These estimates place the anisotropic theory within the same coercive framework as the Euclidean one [1203.5190].

For a measurable set \(E\subset \Omega\), the raw anisotropic outer \(\varepsilon,C\)-Minkowski content is
\[
\mathcal {SM}_{\varepsilon,C}^0(E;\Omega)\ :=\ \frac{1}{\varepsilon} \big(|\Omega\cap(E+\varepsilon C)|-|E|\big),
\]
where \(E+\varepsilon C=\{x+y:x\in E,\ y\in \varepsilon C\}\) is the Minkowski sum [1203.5190]. In sufficiently regular situations this equals
\[
\mathcal{SM}_{\varepsilon,C}^0(E)=\frac{|E+\varepsilon C\setminus E|}{\varepsilon}.
\]
This is a one-sided content: it measures only outward growth, in contrast with the two-sided Minkowski content of the boundary [2508.08156].

A central subtlety is that \(\mathcal {SM}_{\varepsilon,C}^0\) is sensitive to modifications on null sets [1203.5190]. To remove this defect, the robust set-functional version is defined through
\[
F_{\varepsilon, C}(u;\Omega)\ :=\ \frac{1}{\varepsilon}\int_{\Omega}\big( \operatorname*{ess\,sup}_{\Omega\cap (x-\varepsilon C)}u-u(x)\big)\,dx,
\]
and
\[
\mathcal {SM}_{\varepsilon,C}(E;\Omega)\ :=\ F_{\varepsilon, C}(\chi_E;\Omega).
\]
For measurable \(E\),
\[
\mathcal {SM}_{\varepsilon,C}(E;\Omega) = \min_{|E'\triangle E|=0} \mathcal {SM}_{\varepsilon,C}^0(E';\Omega) = \mathcal {SM}_{\varepsilon,C}^0(E^1;\Omega) = \mathcal {SM}_{\varepsilon,C}^0(\Omega\setminus E^0;\Omega),
\]
where \(E^1\) and \(E^0\) are the density-one and density-zero points of \(E\) [1203.5190]. This identifies the measure-theoretically canonical representative.

The expected limit is the anisotropic perimeter
\[
h_C(E;\Omega) \ :=\ \begin{cases}
\displaystyle \int_{\partial^* E\cap\Omega} h_C(\nu_E(x))\,d\mathscr H^{n-1}(x) & \textrm{ if $E$ has finite perimeter in $\Omega$},\\
+\infty &\textrm{ else,}
\end{cases}
\]
with \(\partial^*E\) the reduced boundary and \(\nu_E\) the measure-theoretic outer unit normal [1203.5190]. When \(C=B(0,1)\), \(h_C(\nu)=|\nu|\), and the anisotropic perimeter reduces to the classical perimeter.

## 2. Relation to classical Minkowski content and perimeter

The classical outer Minkowski content of a compact set \(E\subset \mathbb R^n\) is
\[
SM(E)\ :=\ \lim_{\varepsilon \to0^+}\frac{|\{x \in \mathbb R^n : dist(x,E)\leq \varepsilon\}\setminus E|}{\varepsilon},
\]
whenever the limit exists [1203.5190]. Ambrosio–Colesanti–Villa proved that if \(E\) has finite perimeter and the Minkowski content of \(\partial E\) exists and equals the perimeter, then \(SM(E)\) exists and equals \(\operatorname{Per}(E)\) [1203.5190]. In the notation used there, the sufficient condition is
\[
\lim_{\varepsilon\to 0}\frac{|\{x\in \Omega \,:\, dist(x,\partial^* E)\le \varepsilon\}|}{2\varepsilon} \ =\ Per (E;\Omega).
\]

The anisotropic theory preserves this structural dependence on the isotropic hypothesis. The central pointwise theorem states that if \(E\) is a finite-perimeter set such that
\[
\lim_{\varepsilon \to 0} \mathcal {SM}_{\varepsilon,B(0,1)}^0(E;\Omega)\ =\ Per(E;\Omega),
\]
then
\[
\lim_{\varepsilon \to 0} \mathcal {SM}_{\varepsilon,C}^0(E;\Omega)\ =\ h_C(E;\Omega)
\]
[1203.5190]. Thus the anisotropic outer Minkowski content exists on the same class of sets for which the isotropic outer Minkowski content exists and equals perimeter, and its limit is exactly the anisotropic surface energy.

For convex or smooth sets, this conclusion is consistent with Steiner-type expansions. In particular, for convex \(E\),
\[
|E+\varepsilon C|=|E|+ \varepsilon\, h_C(E;\Omega)+O(\varepsilon^2),
\]
so the first variation coefficient is already the anisotropic perimeter [1203.5190]. The codimension-one theory can therefore be understood as extending the familiar first-order convex expansion to broad geometric-measure-theoretic classes of nonconvex finite-perimeter sets.

A common misconception is that anisotropic outer content should exist for arbitrary measurable sets merely because the support function \(h_C\) is well defined. The codimension-one theory does not make such a claim. Existence is tied to finite perimeter and, for the pointwise identification of the raw content, to the same isotropic regularity condition needed in the Euclidean case [1203.5190].

## 3. Variational formulation, coarea structure, and \(\Gamma\)-convergence

The anisotropic outer Minkowski content is also a variational approximation scheme. A generalized coarea formula holds:
\[
F_{\varepsilon,C}(u;\Omega)\ =\ \int_{-\infty}^\infty \mathcal {SM}_{\varepsilon,C}(\{u>s\};\Omega)\,ds,
\]
which makes the set-functional version the level-set counterpart of a nonlocal anisotropic total variation approximation [1203.5190].

The main variational theorem states that, as \(\varepsilon\to 0\), both \(\mathcal {SM}_{\varepsilon,C}\) and \(\mathcal {SM}_{\varepsilon,C}^0\) \(\Gamma\)-converge in \(L^1_{\rm loc}(\Omega)\) to
\[
h_C(E;\Omega)=\begin{cases}
\displaystyle \int_{\partial^* E\cap\Omega} h_C(\nu_E(x))\,d\mathscr H^{n-1}(x) & \textrm{if \(E\) has finite perimeter in \(\Omega\)},\\
+\infty & \textrm{else.}
\end{cases}
\]
Moreover, if \(\sup_{\varepsilon>0}\mathcal {SM}_{\varepsilon,C}(E_\varepsilon;\Omega)<+\infty\), then, up to subsequences, \(E_\varepsilon^1\) converges in \(L^1_{\rm loc}(\Omega)\) to some set \(E\) [1203.5190]. This identifies anisotropic perimeter as the unique variational limit of anisotropic outer parallel-volume growth.

The functional analogue is equally explicit. As \(\varepsilon\to 0\), \(F_{\varepsilon,C}\) \(\Gamma\)-converges to
\[
TV_{-C}(u;\Omega) \ :=\ \begin{cases}
\displaystyle \int_{\Omega} h_C(-Du)& \textrm{ if }u\in BV(\Omega),\\
+\infty &\textrm{ else.}
\end{cases}
\]
[1203.5190]. This places anisotropic outer Minkowski content within the theory of anisotropic \(BV\) energies, rather than solely within convex geometry.

The proof mechanism combines an anisotropic distance function
\[
dist_C(x,E)\ :=\ \inf_{y\in E}\,h^\circ_C(x-y),
\]
with the identity
\[
h_C(\nabla d)=1 \quad\text{a.e. in } \{d>0\},
\]
and Reshetnyak lower semicontinuity [1203.5190]. This reveals that the anisotropy acts through the gauge \(h_C^\circ\) in the tubular neighborhoods and through \(h_C\) in the limiting surface density.

## 4. Boundary content, symmetrization, and anisotropy-independence phenomena

The outer content is one-sided, whereas boundary Minkowski content is two-sided. For a finite-perimeter set \(E\), the \(C\)-anisotropic Minkowski content of the topological boundary is defined by
\[
\mathcal M_C(\partial E;\Omega)= \lim_{\varepsilon\to0_+} \frac{1}{2\varepsilon}\lambda^n\Big(\big((\partial E\cap\Omega)\oplus \varepsilon C\big)\cap\Omega\Big),
\]
when the limit exists [2508.08156]. Because \(h_C\) need not be even, the expected limit is not \(\operatorname{Per}_{h_C}(E;\Omega)\) alone, but
\[
\frac12\big(\operatorname{Per}_{h_C}(E;\Omega)+\operatorname{Per}_{h_C}(\Omega\setminus E;\Omega)\big)
=
\frac12\int_{\Omega\cap\partial^*E}\big(h_C(\nu_E)+h_C(-\nu_E)\big)\, d\mathcal H^{n-1}.
\]
This averaging reflects the fact that a two-sided boundary neighborhood sees both orientations [2508.08156].

The symmetrized content already appears in the 2012 codimension-one paper:
\[
\mathcal M_{\varepsilon,C}(E;\Omega)\ := \ \frac{\mathcal {SM}_{\varepsilon,C}(E;\Omega)+\mathcal {SM}_{\varepsilon,C}(\Omega\setminus E;\Omega)}{2},
\]
and under the isotropic hypothesis one has
\[
\lim_{\varepsilon \to 0} \mathcal M_{\varepsilon,C}^0(E;\Omega)=\int_{\partial^* E} \frac{h_C(\nu_E(x))+h_C(-\nu_E(x))}{2}\,d\mathscr H^{n-1}(x)
\]
[1203.5190]. This is the natural codimension-one anisotropic analogue of the classical Minkowski content of a hypersurface.

A major refinement was obtained in "Existence of Anisotropic Minkowski Content" [2508.08156]. For a set \(E\) of finite perimeter and any two convex bodies \(C,C'\in\mathcal C_0^n\), the existence of the anisotropic Minkowski content of \(\partial E\) with the expected limit for \(C\) is equivalent to the corresponding existence statement for \(C'\). In particular,
\[
\mathcal M(\partial E;\Omega)=\operatorname{Per}(E;\Omega)
\]
if and only if
\[
\mathcal M_C(\partial E;\Omega) = \frac12\big(\operatorname{Per}_{h_C}(E;\Omega)+\operatorname{Per}_{h_C}(\Omega\setminus E;\Omega)\big)
\]
[2508.08156]. This establishes anisotropy-independence for the boundary content of finite-perimeter sets.

For outer content, the same paper proves a corollary requiring both \(E\) and its complement. If \(\overline E\subseteq \Omega\), then
\[
\mathcal{SM}_C(E;\Omega)=\operatorname{Per}_{h_C}(E;\Omega)
\quad\text{and}\quad
\mathcal{SM}_C(\Omega\setminus E;\Omega)=\operatorname{Per}_{h_C}(\Omega\setminus E;\Omega)
\]
hold for one anisotropy \(C\) if and only if they hold for every anisotropy, in particular for the Euclidean one [2508.08156]. A common misunderstanding is that existence for \(E\) alone should suffice. The paper explicitly shows this is false in general.

## 5. Extensions to lower-dimensional structuring elements and lower-dimensional sets

The convex body defining the anisotropy need not be full-dimensional. "On the (outer) Minkowski content with lower-dimensional structuring element" studies the outer \(Q\)-Minkowski content
\[
SM_Q(A)=\lim_{r\to 0^+}\frac{\lambda_n((A+rQ)\setminus A)}{r},
\]
for compact convex \(Q\subset\mathbb R^n\), possibly with \(\dim Q<n\) [2504.03339]. The associated anisotropic perimeter is
\[
P_Q(A)=\int_{S^{n-1}} h_{Q\cup\{0\}}(v)\,S_{n-1}(A;dv),
\]
which reduces to the ordinary perimeter when \(Q=B^n\) [2504.03339].

The paper proves a general lower bound
\[
\liminf_{r\to 0^+}\frac{\lambda_n((A+rQ)\setminus A)}{r}\ge P_Q(A),
\]
and shows that if the isotropic outer Minkowski content exists and equals \(P(A)\), then \(SM_Q(A)=P_Q(A)\) for any nonempty compact \(Q\) [2504.03339]. Its main novelty is that when \(Q\) lies in a \(k\)-dimensional subspace \(L\), a weaker AFP-condition relative to \(L\) is sufficient for existence. This reflects that a lower-dimensional dilation probes only selected directions.

The same directional principle governs lower-dimensional anisotropic Minkowski content for rectifiable sets. "Anisotropic Minkowski Content for Countably \(\mathcal H^k\)-rectifiable Sets" defines
\[
\mathcal M^{s}_{r,C}(E)\coloneq \frac{\lambda^n(E\oplus rC)}{\omega_{n-s}\,r^{\,n-s}},
\]
and proves that for a compact \(k\)-rectifiable set \(S\),
\[
\mathcal M^k_C(S)= \frac{1}{\omega_{n-k}} \int_S \mathcal H^{n-k}(C_x)\,d\mathcal H^k(x),
\qquad
C_x:=P_{(\mathrm T_x^k S)^\perp}(C)
\]
[2601.22681]. Thus the density is the \((n-k)\)-dimensional volume of the projection of \(C\) onto the normal space of \(S\). In codimension one this reduces to
\[
\mathcal M^{n-1}_C(S) = \frac12\int_S\big(h_C(\nu_S)+h_C(-\nu_S)\big)\,d\mathcal H^{n-1},
\]
recovering the anisotropic boundary content formula [2601.22681].

The 2026 paper also shows that for full-dimensional \(C\), validity of the representation formula for one \(C\) implies validity for every full-dimensional \(C\) [2601.22681]. By contrast, for lower-dimensional \(C\), dependence on the choice of \(C\) is genuine. This sharpens the distinction between anisotropy-independence phenomena in full-dimensional codimension-one theory and the genuinely directional behavior of lower-dimensional structuring bodies.

## 6. Directional refinements, tensor-valued analogues, and fractal variants

Anisotropic outer Minkowski content can be generalized in two different directions. One is gauge anisotropy, where Euclidean balls are replaced by a convex body \(C\). The other is directional or tensorial refinement of outer parallel coefficients. "Minkowski Tensors of Anisotropic Spatial Structure" develops the latter viewpoint through local Steiner formulas and support measures [1009.2340].

For a convex body \(K\subset\mathbb R^d\), the local outer parallel set
\[
\mathcal{M}_\epsilon (K,\eta):=\{\mathbf{x}\in \mathbb R^d\setminus K:d_K(\mathbf{x})\le \epsilon,(\mathbf{p}_K(\mathbf{x}),\mathbf{n}_K(\mathbf{x}))\in\eta\}
\]
satisfies
\[
V_d(\mathcal{M}_\epsilon(K,\eta))=\sum_{\nu=0}^{d-1}\epsilon^{d-\nu}\kappa_{d-\nu}\Lambda_\nu(K,\eta),
\]
where \(\Lambda_\nu(K,\eta)\) are support measures [1009.2340]. Because \(\eta\subset \mathbb R^d\times S^{d-1}\) may restrict the normal direction, these are already directional outer-content coefficients. Minkowski tensors arise by integrating \(\mathbf{x}^r\mathbf{n}^s\) against \(\Lambda_\nu\), thereby producing tensor-valued anisotropic refinements of scalar outer parallel-volume coefficients [1009.2340].

A different extension appears in fractal geometry. "Fractal curvatures and Minkowski content of self-conformal sets" studies Cesàro-averaged limits of rescaled outer parallel sets \(F_r\) and refines them to curvature-direction measures on \(F\times S^{d-1}\) [1211.3421]. Here the anisotropy is not induced by a non-Euclidean gauge \(C\), but by directional dependence on normals and orthogonal cocycles. This suggests that anisotropic outer Minkowski content has at least two mathematically distinct meanings: gauge anisotropy via \(E+rC\), and directional anisotropy via measures on position-normal space.

Lower-dimensional anisotropic outer content for compact sets is studied in "Anisotropic lower-dimensional Minkowski content and \(\mathcal S\)-content" [2509.06545]. For compact \(E\in\mathcal K^n\),
\[
\mathcal{SM}^s_{r,C}(E):=\frac{V_{E,C}(r)-V_{E,C}(0)}{r^{\,n-s}},
\qquad
V_{E,C}(r)=\lambda^n(E\oplus rC),
\]
defines the \(s\)-dimensional anisotropic outer Minkowski content [2509.06545]. The paper proves that \(V_{E,C}\) is of Kneser type of order \(n\), yielding inequalities
\[
\mathcal S_C^s(E)_* \le \mathcal{SM}_C^s(E)_* \le \mathcal{SM}_C^s(E)^* \le \mathcal S_C^s(E)^*.
\]
For null sets, the same relations hold with ordinary anisotropic Minkowski content \(\mathcal M_C^s\) in place of \(\mathcal{SM}_C^s\) [2509.06545]. The Sierpiński gasket example shows that lower and upper anisotropic Minkowski and \(\mathcal S\)-contents may all differ, so no general equality principle survives in the fractal regime.

## 7. Scope, proof mechanisms, and open directions

The codimension-one theory rests on reduced boundaries, blow-up to half-spaces, and measure-theoretic normals rather than topological boundaries [1203.5190]. In the pointwise theorem, the measures
\[
\mu_\varepsilon:=\frac{1}{\varepsilon}\left( \chi_{E+\varepsilon C}-\chi_E\right)\mathscr L^n
\]
are shown to concentrate on \(\partial^*E\), and one identifies their density with \(h_C(\nu_E)\) by combining a lower bound from \(\Gamma\)-convergence with a flatness argument driven by the isotropic outer-content hypothesis [1203.5190]. This makes the reduced boundary the correct geometric support of the limit.

Two scope restrictions recur throughout the literature. First, raw outer content depends on the representative of \(E\), so measure-theoretic representatives such as \(E^1\) or \(\Omega\setminus E^0\) are indispensable [1203.5190], [2508.08156]. Second, pointwise existence statements are not available for arbitrary measurable sets or arbitrary finite-perimeter sets; they require hypotheses equivalent or comparable to those guaranteeing the Euclidean result [1203.5190], [2508.08156].

Several later works suggest broader directions. The 2025 anisotropy-independence theorem indicates that for boundary Minkowski content of finite-perimeter sets, the existence question is fundamentally isotropy-independent [2508.08156]. The lower-dimensional papers show, however, that once one leaves the full-dimensional codimension-one setting, dependence on the structuring body can be sharp and discontinuous [2504.03339], [2601.22681]. This suggests that the codimension-one equivalence phenomenon is exceptional rather than universal.

A plausible implication is that anisotropic outer Minkowski content should be regarded not as a single invariant but as a family of asymptotic functionals whose behavior depends strongly on which aspect of anisotropy is being encoded: ambient gauge, structuring-body dimension, directional filtering, tensorial weighting, or capillary boundary geometry. The literature supports this differentiation. Gauge anisotropy in finite-perimeter codimension one leads to anisotropic perimeter [1203.5190]; directional refinement leads to support measures and Minkowski tensors [1009.2340]; self-conformal directional theory leads to curvature-direction measures rather than a non-Euclidean gauge content [1211.3421]; and lower-dimensional theories replace support functions by projected volumes of \(C\) on normal spaces [2601.22681].

In its most established form, anisotropic outer Minkowski content is therefore the first-order outer parallel-volume growth under Minkowski enlargement by a convex body \(C\), with limit
\[
\int_{\partial^*E} h_C(\nu_E)\,d\mathscr H^{n-1}
\]
for the same class of finite-perimeter sets that support the classical Euclidean theorem [1203.5190]. Subsequent developments refine this statement by separating one-sided from two-sided content, clarifying anisotropy-independence, extending the theory to lower-dimensional settings, and embedding it into broader integral-geometric and fractal frameworks [2508.08156], [2504.03339], [2601.22681], [2509.06545], [1009.2340], [1211.3421].

Source: https://www.emergentmind.com/topics/anisotropic-outer-minkowski-content