Authors: Mark Green and Stanley Osher
Published in: Asian Journal of Mathematics, 1999
Abstract
In this paper, we obtain some new inequalities for integrals of convex functions of the curvature (resp. Wulff curvature) of convex plane curves. We also show that the difference between the two sides of our inequalities are monotone decreasing as the region flows under the unit-speed outward normal (resp. Wulff) flow.
Introduction
The classical isoperimetric inequality ($L^2 \ge 4\pi A$) is a cornerstone of geometric analysis, establishing that the circle minimizes perimeter for a given area. However, characterizing the "roundness" or geometric efficiency of a shape often requires looking beyond simple area-perimeter relationships. In their 1999 paper, Mark Green and Stanley Osher extend these classical concepts into the realm of partial differential equations (PDEs) and materials science, providing a unified framework for generating new isoperimetric inequalities and describing how they evolve over time.
The Geometry of Growth: Steiner Polynomials
The core of the authors' approach lies in the Steiner polynomial. When a convex region $K$ in the plane expands outward at a constant unit speed, its area at time $t$ can be expressed as a quadratic polynomial:
$$A(t) = A + Lt + \pi t^2$$
where $A$ is the initial area and $L$ is the initial perimeter. Green and Osher focus on the roots of this polynomial, $t_1$ and $t_2$. While these roots occur at negative values of $t$—representing a theoretical point where the shape would collapse—they carry profound geometric information. The authors demonstrate that these roots are deeply linked to the shape's inradius and outradius via Bonnesen-style inequalities, serving as benchmarks for how much a shape deviates from a perfect disk.
A Unified Framework for Curvature Inequalities
Prior to this work, specific inequalities existed for certain curvature integrals, such as Gage’s inequality for $\int k^2 ds$. Green and Osher generalized this by developing a "two-piece" version of Jensen’s inequality combined with symmetrization arguments. Their central result (Theorem 0.1) provides a template: for any convex function $F$, the integral of the radius of curvature $\rho$ is bounded by the roots of the Steiner polynomial:
$$\frac{1}{2\pi} \int F(\rho) d\theta \ge \frac{1}{2}(F(-t_1) + F(-t_2))$$
By substituting different convex functions into this formula, the authors derived an infinite family of inequalities, including new bounds for higher powers of curvature ($k^3, k^4$) and transcendental functions. For example, they established a concrete bound for the third power of curvature:
$$\int_{\partial K} k^3 ds \ge \frac{L^2\pi - 2A\pi^2}{A^2}$$
Curvature Entropy and Wulff Flows
Beyond Euclidean geometry, the paper addresses anisotropic growth—where a shape grows at different speeds in different directions. This is modeled via the Wulff flow, which is essential for understanding crystal growth in materials science. The authors show that their framework holds even in these non-Euclidean settings, replacing the standard circle with the "Wulff shape" as the ideal state.
One of the most striking concepts introduced is the "entropy of the curvature." The authors prove an entropy inequality:
$$\int k \log(k \sqrt{A/\pi}) ds \ge 0$$
This suggests that curvature distribution has an inherent order that can be quantified and tracked.
Monotonicity and the Isoperimetric Defect
A critical question in geometric flows is whether a shape becomes more or less regular as it evolves. Green and Osher prove that the "isoperimetric defect"—the gap between the two sides of their inequalities—is monotone decreasing under both outward normal and Wulff flows.
This means that as a convex shape grows, it does not merely increase in size; it fundamentally trends toward its ideal state (a disk in the Euclidean case or a Wulff shape in the anisotropic case). This monotonicity provides a mathematical proof for the "rounding" effect observed in physical processes like flame front propagation or crystal precipitation.
Conclusion and Limitations
While the paper provides a powerful toolkit for 2D convex geometry, it notes specific boundaries for future research. The results are strictly dependent on the convexity of the curves; the framework fails for non-convex shapes where the support functions behave differently. Additionally, while the authors suggest the approach could be fruitful for higher-dimensional convex bodies (3D and beyond), the proofs in this text remain situated in the plane.
Ultimately, Green and Osher’s work bridges the gap between static geometric properties and the dynamic evolution of shapes, offering a rigorous mathematical foundation for understanding how growth leads to symmetry.
Summary updated