1 Definition and Intuition
Mean curvature is a scalar quantity assigned to a surface at each point, summarizing how the surface bends around that point. For a surface embedded in three-dimensional Euclidean space, it arises naturally from considering how the surface curves in different normal directions and then averaging the resulting curvature rates.
1.1 Principal curvatures
At a point on a smooth surface, there are typically special tangent directions—called principal directions—along which the normal curvature is extremal. The corresponding normal curvatures are the principal curvatures, usually denoted \(k_1\) and \(k_2\). These numbers describe how quickly the surface bends when one moves along the surface in the principal directions.
1.2 Average curvature (mean curvature)
Mean curvature is commonly defined as the average of the principal curvatures: \[ H=\frac{1}{2}(k_1+k_2). \] The factor \(1/2\) is a convention chosen so that \(H\) matches standard formulas in differential geometry and variational theory. Conceptually, \(H\) measures the balance between bending “outward” and “inward” as detected from the two principal curvatures.
1.3 Sign conventions and orientation
The sign of mean curvature depends on the choice of unit normal vector. Reversing the orientation of the normal flips the signs of \(k_1\), \(k_2\), and hence \(H\). Because different fields sometimes adopt different normal conventions, formulas are often presented with explicit orientation statements, or they are written in terms of geometric vectors (such as a mean curvature vector) to reduce ambiguity.
1.4 Geometric interpretation via normal sections
Another way to view mean curvature is through normal sections: take planes perpendicular to a chosen tangent direction, intersect the surface, and examine the curvature of the resulting curve at the point of contact. When the chosen tangent direction varies, the normal curvature changes; at principal directions it reaches its extremes. Mean curvature then corresponds to the average of these extremal curvatures, providing a single number that reflects the overall local bending rate.
2 Mean Curvature for Surfaces
For analytic work, mean curvature is expressed using local differential-geometric data derived from how the surface is embedded and how it inherits metric properties from the ambient space.
2.1 Graphs and local formulas
A frequently used setting is when a surface is given as a graph \(z=u(x,y)\). In that case, mean curvature can be written in terms of first and second derivatives of \(u\), producing a concrete formula useful for computations. This “graph form” connects mean curvature directly to the differential operator that generalizes the divergence of a normalized gradient.
2.2 Parameterized surfaces
More generally, a surface can be parameterized by an immersion \[ X(u,v) \in \mathbb{R}^3. \] One computes tangent vectors \(X_u, X_v\), builds the induced metric, and then evaluates how the surface bends by analyzing second derivatives of \(X\) projected onto the normal direction. Mean curvature emerges from these geometric quantities without requiring the surface to be a global graph.
2.3 First and second fundamental forms
The first fundamental form captures the intrinsic metric of the surface and is determined by dot products of tangent vectors: \[ I = \begin{pmatrix} E & F \\ F & G \end{pmatrix}. \] The second fundamental form records how the surface bends in the normal direction and is formed from second derivatives projected onto the unit normal: \[ II = \begin{pmatrix} e & f \\ f & g \end{pmatrix}. \] In standard coordinates, mean curvature can be expressed in terms of \(E,F,G\) and \(e,f,g\). This formulation clarifies which parts of the computation depend on metric structure versus bending data.
2.4 Relationship to the Laplace–Beltrami operator
The Laplace–Beltrami operator is the surface analogue of the Laplacian, built from the induced metric. Mean curvature is linked to the Laplace–Beltrami operator acting on the embedding map: in many treatments, the surface mean curvature vector is (up to sign conventions) the Laplacian of the position vector constrained to the surface. This connection underlies both geometric analysis and numerical schemes.
3 Curvature Calculations
Mean curvature can be calculated in multiple equivalent ways, depending on whether one starts from an explicit parametrization, from intrinsic curvature operators, or from implicit representations.
3.1 Curvature from an explicit parametrization
Given \(X(u,v)\), one computes:
- the unit normal \(n\),
- the coefficients of the first and second fundamental forms, and
- the principal-curvature combination that yields mean curvature.
Operationally, mean curvature is often computed via an expression involving the first and second fundamental form matrices rather than explicitly computing \(k_1\) and \(k_2\).
3.2 Using shape operator (Weingarten map)
The shape operator describes how the unit normal changes as one moves along the surface. Denoting it by \(S\), its eigenvalues are the principal curvatures. In this framework, mean curvature is half the trace of \(S\): \[ H=\frac{1}{2}\operatorname{tr}(S), \] again with sign depending on the chosen normal. This approach is conceptually clean because it packages curvature information into a linear map on the tangent space.
3.3 Computing mean curvature for common surfaces
For standard shapes, symmetry typically allows direct evaluation of principal curvatures, producing simple closed forms for mean curvature.
3.3.1 Sphere
For a sphere of radius \(R\), both principal curvatures are equal to \(1/R\) (with sign determined by normal orientation). Therefore, \[ H=\frac{1}{R} \] under the conventional outward-normal sign choice (and \(-1/R\) for the opposite orientation).
3.3.2 Cylinder
For a right circular cylinder of radius \(R\), one principal curvature is \(1/R\) while the other is \(0\) because the surface is flat along the axial direction. Hence, \[ H=\frac{1}{2R} \] (with sign determined by normal orientation).
3.3.3 Plane
For a plane, both principal curvatures are zero everywhere. Consequently, \[ H=0, \] reflecting that the plane has no bending in any direction.
3.4 Examples from implicit surfaces
If a surface is represented implicitly by an equation \(F(x,y,z)=0\), one can compute its unit normal from \(\nabla F\) and then express mean curvature in terms of first and second derivatives of \(F\). This method is widely used because many applications start from level sets rather than explicit parametrizations, and it avoids parametrization singularities.
4 Special Classes of Surfaces
Some surface families are characterized by specific behavior of mean curvature, leading to classification results or simplifying properties in analysis.
4.1 Minimal surfaces (H = 0)
Minimal surfaces are defined by vanishing mean curvature \(H=0\). They are critical points of the area functional, meaning small normal variations do not change area to first order. Classic examples include the plane and certain nontrivial surfaces found by solving the associated minimal surface equation.
4.2 Constant mean curvature surfaces
Constant mean curvature surfaces satisfy \(H=\text{constant}\) over the entire surface. This condition balances the tendency of area minimization with a compensating constraint (often interpreted as modeling objects with uniform pressure difference in physical analogies). Spherical caps and Delaunay-type surfaces are typical examples in the rotationally symmetric setting.
4.3 Totally umbilic surfaces
A surface is totally umbilic if its principal curvatures are equal at every point: \(k_1=k_2\). Such surfaces include spheres and planes (in the appropriate orientations and embeddings). Umbilicity implies that curvature is the same in all tangent directions, and mean curvature then becomes directly tied to that shared principal value.
4.4 Surfaces with symmetry and reduced computation
Symmetry can reduce the complexity of computing mean curvature. For instance, rotational symmetry can reduce the problem to a curve in a plane, translating surface curvature computation into curvature of a profile curve. Similarly, translational symmetry can simplify the principal curvature structure, often yielding one principal curvature as a known function and the other as constrained by geometry.
5 Variational Principles
Mean curvature is deeply connected to variational problems, where one studies how geometric functionals respond to smooth perturbations.
5.1 Area functional and Euler–Lagrange equations
Consider the area functional \(A\) that assigns to each surface its surface area. When a surface is perturbed by a small normal displacement, the first-order change in area yields an Euler–Lagrange equation. The resulting condition for a stationary surface is precisely \(H=0\), linking minimal surfaces to variational calculus.
5.2 Mean curvature as a first variation
More explicitly, the first variation of area is proportional to the integral of mean curvature against the normal component of the variation. Thus mean curvature functions as the “gradient” of area with respect to normal deformations, making it an effective diagnostic tool: where \(H\) is large in magnitude, area changes rapidly under normal motion.
5.3 Prescribed mean curvature problems
In many settings, one seeks surfaces whose mean curvature equals a specified function \(H(x)\). Such prescribed mean curvature problems generalize minimal surface equations by replacing the equilibrium condition \(H=0\) with a target curvature profile. This framework is central in geometric analysis and generates a family of nonlinear partial differential equations.
5.4 Stability and second variation (overview)
Beyond first variation, the second variation measures whether a stationary surface is stable (area increases for small perturbations) or unstable. The second variation involves additional operators derived from geometry, including terms related to the squared norm of the second fundamental form and the ambient Laplace–Beltrami structure. Although details depend on context, the general viewpoint is that stability is governed by how mean curvature-driven forces respond to perturbations.
6 Mean Curvature Flow
Mean curvature flow describes an evolution of surfaces where points move in the normal direction at a speed proportional to mean curvature.
6.1 Basic evolution equation
In its simplest form for a moving surface \(S(t)\), the normal velocity \(V_n\) satisfies \[ V_n = -H, \] with sign depending on convention. The negative sign is often chosen so that regions with positive mean curvature move inward, promoting smoothing behavior analogous to heat diffusion.
6.2 Interpretation as gradient descent (conceptual)
Mean curvature flow can be interpreted as a gradient descent process for the area functional: the surface moves in the direction that decreases area most efficiently in the normal direction. This variational interpretation explains why mean curvature flow tends to reduce irregularities, at least until singular behavior arises.
6.3 Singularities and qualitative behavior (overview)
As the surface evolves, it may develop singularities where curvature becomes unbounded or the surface topology changes. Typical singularity models involve shrinking spheres or neck-like regions that resemble cones. A major theme in geometric flow is understanding which singularities can occur and how to continue the flow through or beyond them, often via rescaling and limiting arguments.
6.4 Applications and toy examples
Toy examples illustrate qualitative behavior: a shrinking sphere collapses in finite time while remaining symmetric; a cylinder under mean curvature flow tends to shrink and can develop pinch-like features in certain geometries. In applications, mean curvature flow is used conceptually in smoothing, segmentation ideas, and geometric regularization, frequently through discretizations and approximations.
7 Coordinate Systems and Extensions
Different formulations express mean curvature using coordinate-dependent objects, while geometric quantities remain invariant under appropriate changes. Extensions clarify how mean curvature generalizes to higher dimensions and higher codimensions.
7.1 Mean curvature in different parametrizations
Mean curvature is intrinsic to the embedded surface geometry, even though its computational expressions vary with coordinates. Switching parametrizations changes the formulas’ intermediate terms—metrics, Christoffel symbols, or Jacobians—but the resulting scalar (with consistent sign convention) remains the same. This coordinate invariance ensures that mean curvature is a meaningful geometric descriptor rather than an artifact of a chosen chart.
7.2 Boundary terms and regularity considerations (overview)
When variational principles involve surfaces with boundary, integration by parts introduces boundary terms. Conditions on the boundary—such as fixing the boundary curve, prescribing contact angles, or imposing natural boundary constraints—determine the correct Euler–Lagrange equations. Regularity issues also arise: solutions may develop singularities or lose smoothness, particularly for nonlinear curvature-driven equations.
7.3 Higher-dimensional generalizations (mean curvature vector)
For a hypersurface embedded in \(\mathbb{R}^{n+1}\), mean curvature generalizes through the mean curvature vector \( \mathbf{H} \), which points in the normal direction and encodes the average of principal curvatures across \(n\) tangent directions. The scalar mean curvature is then often obtained by pairing \( \mathbf{H} \) with a unit normal, collapsing the vector information to a single number when the codimension is one.
7.4 Codimension and the role of normal vectors
When the surface has codimension greater than one, there is not a single distinguished normal direction. Instead, there is a normal bundle, and the curvature information distributes across multiple normal directions. In this setting, mean curvature is best described by the mean curvature vector, which remains coordinate-free and captures how the surface bends in the ambient space as a whole.
8 Discrete and Computational Approaches
Practical computation often requires discretizing surfaces, since many applications use triangular meshes, point clouds, or level-set grids rather than smooth parametrizations.
8.1 Discrete mean curvature on meshes (conceptual)
On a mesh, mean curvature is approximated by quantities associated with vertices, edges, or faces. One goal is to mimic the geometric meaning of mean curvature as an averaging of bending effects. Different discretizations target different properties: some aim to preserve consistency with known smooth limits; others prioritize numerical stability or speed.
8.2 Numerical estimation methods
Numerical methods include finite-difference approximations on implicit grids, variational discretizations based on energy principles, and geometric estimators that relate curvature to angle defects or local normal variations. Accuracy depends on mesh quality, resolution, and how normal directions are estimated or smoothed prior to curvature computation.
8.3 Curvature filtering and smoothing (overview)
Because discrete curvature estimates can be noisy—especially on irregular meshes—computational workflows often apply smoothing or regularization. These steps may include denoising normals, applying neighborhood averaging, or evolving the mesh under discrete curvature flow. The intent is to reduce spurious oscillations while retaining meaningful geometric features.
8.4 Visualization techniques for mean curvature fields
Mean curvature values are frequently visualized using color maps on the mesh, where sign and magnitude are encoded via contrasting hues and brightness. For interpretability, practitioners may normalize by local scale, clip extreme values, or combine mean curvature with related measures (such as Gaussian curvature or principal curvature estimates) to better understand surface shape behavior in different regions.