Curvature

1 Fundamental notions of curvature

1.1 Intuitive geometric meaning

Curvature measures how a geometric object bends relative to an idealized linear model. For a curve in the plane, “zero curvature” corresponds to being locally straight; larger curvature indicates stronger turning. For a surface, curvature quantifies how the surface deviates from being locally flat in different directions. In both cases, curvature converts qualitative bending into quantitative invariants.

1.2 Curvature of curves vs. curvature of surfaces

Curvature for curves is typically a single scalar (often signed in the plane) at each point, describing how the tangent direction changes with distance. For surfaces, curvature is richer: it depends on direction within the tangent plane and is captured by quantities such as principal curvatures and the second fundamental form. Thus, a curve has one “bending rate,” while a surface has a directional bending profile.

1.3 Local vs. global curvature

Curvature is defined through local behavior—derivatives or infinitesimal variation near a point. Nevertheless, global geometry can be influenced by local curvature via accumulation effects. For instance, the total turning of a plane curve relates to its overall shape, and integrating curvature over a surface can constrain global properties. Still, curvature itself is inherently a local concept.

2 Curvature of plane curves

2.1 Arc length parameterization

2.1.1 Tangent angle and turning rate

For a sufficiently smooth plane curve parameterized by arc length \(s\), the unit tangent vector \(T(s)\) changes along the curve. One convenient description uses the tangent angle \(\theta(s)\), defined so that \(T(s)=(\cos\theta(s),\sin\theta(s))\). The curvature is then the rate of change of this angle with respect to arc length: \[ \kappa(s)=\frac{d\theta}{ds}. \] This ties curvature directly to the “turning rate” per unit distance along the curve.

2.1.2 Signed curvature and orientation

Signed curvature distinguishes turning left versus right relative to a chosen orientation. When the tangent angle increases in the counterclockwise sense, the curvature is positive; if it decreases, it is negative. The sign convention depends on the orientation of the curve and the plane, but once fixed it yields consistent geometric information.

2.2 Formulae for curvature

2.2.1 Curvature in terms of derivatives of a parametrization

For a regular plane curve given by a parameter \(t\mapsto (x(t),y(t))\), curvature can be expressed using first and second derivatives. A common formula is \[

\kappa(t)=\frac{	x'(t)y''(t)-y'(t)x''(t)	}{\bigl(x'(t)^2+y'(t)^2\bigr)^{3/2}},

\] with an optional sign determined by the orientation convention. The denominator normalizes by the speed cubed, ensuring that the result corresponds to arc-length behavior.

2.2.2 Curvature using first and second fundamental forms in 1D form

In the plane-curve setting, the “fundamental form” viewpoint can be reduced to one-dimensional differential geometry: the first fundamental form corresponds to the induced metric along the curve (essentially \(ds^2\)), while the second fundamental form captures how the normal component changes. In practice, this yields formulas equivalent to the derivative-based curvature expressions, but phrased in geometric terms involving unit normal fields and normal acceleration.

2.3 Curvature for graphs \(y = f(x)\)

2.3.1 Standard curvature expression

If a curve is represented as a graph \((x,f(x))\) and is sufficiently smooth, curvature at \(x\) can be written as \[ \kappa(x)=\frac{f''(x)}{(1+(f'(x))^2)^{3/2}}, \] with sign reflecting whether the graph bends upward or downward relative to the chosen orientation. This formula follows from differentiating the unit tangent vector and accounting for arc-length scaling.

2.3.2 Examples and typical shapes

• Line: If \(f(x)=ax+b\), then \(f''(x)=0\), so curvature vanishes everywhere.
• Parabola: For \(f(x)=x^2\), curvature is nonzero and varies with \(x\), becoming smaller as the slope grows.
• Circle as a graph (local representation): Curvature is constant in magnitude on a circle; a graph expression captures this constant locally when the representation is valid (away from vertical tangents).

3 Curvature of space curves

3.1 Frenet–Serret framework

3.1.1 Tangent, normal, and binormal vectors

For a regular space curve parameterized by arc length \(s\), the Frenet–Serret theory introduces an orthonormal moving frame:

• \(T(s)\): unit tangent direction,
• \(N(s)\): principal normal, pointing toward the direction of fastest turning,
• \(B(s)\): binormal, completing a right-handed triad via \(B=T\times N\).

These vectors capture both bending within the osculating plane and twisting out of it.

3.1.2 Definition of curvature from Frenet–Serret equations

Curvature is defined through how \(T\) changes with arc length: \[ \frac{dT}{ds}=\kappa(s)\,N(s). \] Equivalently, \(\kappa(s)=\left\lVert \frac{dT}{ds}\right\rVert\). This generalizes the planar turning-rate idea: curvature measures the magnitude of the tangent’s derivative per unit arc length.

3.2 Alternative expressions for curvature

3.2.1 Cross-product derivative formula

For a curve \(\mathbf{r}(t)\) with velocity \(\mathbf{r}'(t)\) and acceleration \(\mathbf{r}''(t)\), curvature can be computed (for regular points where \(\mathbf{r}'(t)\neq 0\)) by \[ \kappa(t)=\frac{\lVert \mathbf{r}'(t)\times \mathbf{r}''(t)\rVert}{\lVert \mathbf{r}'(t)\rVert^{3}}. \] This expression avoids building the Frenet frame explicitly and remains invariant under reparameterizations.

3.2.2 Reparameterization invariance

Although curvature formulas may use \(t\), the resulting value corresponds to geometric bending and does not depend on the specific parameterization, provided the curve is regular and differentiability assumptions hold. This invariance is essential: the shape of the curve determines curvature, not the speed at which the curve is traversed.

4 Curvature for general parametrized curves

4.1 Reparameterization effects

If a curve is given by a parameter that is not arc length, derivatives change in a way that can distort naive formulas. The correct curvature remains invariant after compensating for speed factors, typically involving powers of \(\|\mathbf{r}'\|\) to match arc-length differentiation. Hence, curvature is a geometric quantity, not a coordinate artifact.

4.2 Regularity requirements and breakdown points

Curvature formulas usually require:

• Regularity: the velocity \(\mathbf{r}'\) should not vanish (so there is a well-defined tangent direction).
• Differentiability: at least second derivatives are needed for curvature defined via second-order variation.

At points where regularity fails—such as where the curve stops or becomes non-smooth in a way that breaks differentiability—curvature may be undefined or must be treated using weaker notions (e.g., generalized curvature).

4.3 Curvature of non-smooth curves (overview)

For curves with corners, cusps, or other non-differentiable points, curvature cannot be described purely by classical derivatives at the singularities. One approach is to consider curvature as a measure concentrated at such points, while other approaches use smoothing or discrete curvature approximations. These generalized viewpoints aim to preserve the idea of “turning” while accommodating irregular geometry.

5 Curvature of surfaces

5.1 Normal vectors and local geometry

A regular surface has a well-defined unit normal direction at each smooth point. Local geometry can be studied by how the normal varies across nearby points and how tangent directions map under differentiation. These mechanisms link curvature to normal derivatives and to the interaction between the surface and its ambient space.

5.2 Second fundamental form

5.2.1 Principal curvatures

The second fundamental form encodes how the surface bends in each tangent direction. For a given point, it can be diagonalized relative to the tangent plane, yielding two principal curvatures \(\kappa_1\) and \(\kappa_2\). They represent maximal and minimal normal bending rates among all directions in the tangent plane.

5.2.2 Shape operator and eigenvalues

The shape operator (also called the Weingarten map) is a linear transformation on the tangent plane defined through the derivative of the unit normal. Its eigenvalues are the principal curvatures, and its eigenvectors give the principal directions. This operator-centric view provides a coordinate-free method to characterize curvature behavior.

5.3 Mean curvature and Gaussian curvature

5.3.1 Definitions and relations to principal curvatures

Two central scalar curvature invariants are:

• Mean curvature \(H=\frac{1}{2}(\kappa_1+\kappa_2)\),
• Gaussian curvature \(K=\kappa_1\kappa_2\).

These are built from the principal curvatures, so they inherit geometric meaning even when principal directions are not explicitly constructed.

5.3.2 Interpretation via local area/angle distortion

Gaussian curvature relates to how geodesics behave and how angles and areas distort under the surface’s intrinsic geometry. Roughly, positive \(K\) corresponds to local “spherical” behavior, negative \(K\) to “saddle-like” behavior, and zero \(K\) to locally developable behavior (in the classical smooth setting).

6 Intrinsic vs. extrinsic curvature

6.1 Intrinsic geometry and Gauss’s Theorema Egregium (statement level)

Intrinsic curvature describes a surface using only measurements within the surface, such as lengths, angles, and geodesics, without referring to how the surface sits in space. Gauss’s Theorema Egregium asserts that Gaussian curvature is determined entirely by intrinsic geometry. In other words, bending a surface in space may change its extrinsic form while leaving \(K\) unchanged.

6.2 Extrinsic geometry and embedding dependence

Extrinsic curvature depends on the ambient space and the embedding of the surface. The second fundamental form, principal curvatures, and mean curvature can vary when the same intrinsic surface is realized in a different way in space. This dependence reflects that normal directions and normal derivatives are embedding-sensitive.

6.3 Examples of how they differ

A classic illustration is that a surface can be bent without stretching—preserving intrinsic distances—yet its normal behavior changes. Gaussian curvature remains fixed under such isometric deformations, while mean curvature and the second fundamental form can change significantly.

7 Curvature tensors in higher-dimensional settings

7.1 Riemannian curvature and sectional curvature

7.1.1 Curvature operator on tangent spaces

In Riemannian geometry, curvature is encoded by the Riemann curvature tensor. It can be regarded (at a point) as defining how vectors are rotated when parallel transported around infinitesimal loops. This leads to a geometric measure of noncommutativity of parallel transport, represented by an operator acting on tangent spaces.

7.2 Scalar and Ricci curvature (overview)

From the Riemann curvature tensor one forms derived contractions:

• Ricci curvature, obtained by tracing over one pair of indices, summarizes how volume elements expand or contract in various directions.
• Scalar curvature, the full trace, provides a single number reflecting average curvature.

These invariants are central in geometric analysis because they influence volume growth, heat flow, and compactness-type results.

8 Curvature in analysis and differential operators

8.1 Laplace–Beltrami operator and curvature terms

On a curved space (such as a Riemannian manifold), the Laplace–Beltrami operator generalizes the ordinary Laplacian. When studying related operators—especially in the presence of additional bundles or in heat kernel expansions—curvature contributes correction terms. This shows how geometry affects analytic behavior, even for differential equations defined on the manifold.

8.2 Geometric measure considerations (overview)

Curvature also appears in geometric measure theory through generalized notions of curvature for sets and varifolds, where smooth surfaces are approximated by generalized objects. In these contexts, curvature relates to first and second variations—how quantities like area change under perturbations—extending curvature ideas beyond classical differentiable settings.

8.3 Curvature estimates and regularity themes (overview)

Many analysis results are driven by estimates that bound curvature or curvature-related quantities. Such bounds can imply regularity: solutions may become smoother, singularities may be controlled, or convergence of geometric approximations may be established. The specific mechanisms vary, but curvature acts as a key quantitative parameter.

9 Curvature and approximation of shapes

9.1 Local quadratic approximation

9.1.1 Osculating circle and osculating sphere

For curves, curvature determines the best local circular model. The osculating circle at a point has the same tangent direction and curvature, matching second-order behavior along the curve (when expressed appropriately). For surfaces, an analogous idea uses an osculating sphere (or quadratic surface model) derived from principal curvatures, providing a local “second-order” approximation of shape.

9.2 Numerical estimation of curvature (overview)

In applications, curvature is often estimated from sampled points. Typical approaches use local polynomial fitting, circle fitting, or discrete differential geometry to approximate derivatives and then compute curvature from those approximations. Accuracy depends on sampling density, noise level, and the regularity of the underlying shape; robust methods try to stabilize derivative estimates.

10 Curvature theorems and applications (non-controversial, mathematical)

10.1 Gauss map and related constructions (overview)

The Gauss map sends each point on a surface to the unit normal direction, viewed on the unit sphere. It provides a geometric bridge between how normals spread and how the surface bends. The differential of the Gauss map is related to the shape operator, making curvature accessible through mapping properties.

10.2 Differential geometry identities (overview)

A number of classical identities relate curvature quantities, derivatives, and the second fundamental form. Examples include structure equations that govern how the frame fields vary and compatibility conditions ensuring consistent curvature behavior. These identities underpin many computations and conceptual links between curvature invariants.

10.3 Curvature-driven shape evolution (overview)

Geometric flows evolve shapes so that curvature guides the motion. One intuitive example is motion by mean curvature, where surfaces adjust locally in the direction that tends to reduce area at a rate governed by mean curvature. Such flows connect curvature to dynamical processes and to the formation or smoothing of features, studied in a largely mathematical framework.

11 Common examples and computations

11.1 Circles and helices

• Circle: A circle of radius \(R\) has constant curvature \(\kappa=1/R\) (up to sign in the planar oriented setting).
• Helix: A space curve shaped like a spring can have constant curvature and torsion. Curvature reflects the radius of the helix and how quickly the tangent turns as one moves along the curve.

11.2 Spheres and cylinders

• Sphere: Principal curvatures are equal everywhere; for radius \(R\), \(\kappa_1=\kappa_2=1/R\). Consequently \(H=1/R\) and \(K=1/R^2\).
• Cylinder: One principal curvature is zero (along the axis direction), while the other equals \(1/R\), where \(R\) is the cylinder radius. This yields \(K=0\) and a mean curvature depending on \(R\).

11.3 Graph surfaces and simple parameterizations

For a surface given as a graph \(z=f(x,y)\), curvature can be computed using the normal vector and derivatives of \(f\). In many elementary cases, such as paraboloids or planes, explicit formulas for mean and Gaussian curvature are available, illustrating how curvature depends on second derivatives and on the slope of the graph.

12 Pathologies, edge cases, and regularity

12.1 Points of zero curvature

Zero curvature indicates locally linear behavior for curves or locally developable behavior for surfaces in the relevant sense. For planar curves, \(\kappa=0\) corresponds to no change in tangent direction to first order beyond straightness. For surfaces, vanishing of certain curvature invariants may indicate flatness along directions or that the surface is developable, depending on which quantity is zero.

12.2 Cusps, corners, and lack of differentiability (overview)

At non-smooth points, tangent directions may fail to be well-defined continuously, or second derivatives may not exist. Classical curvature can thus become undefined. In generalized treatments, one may assign curvature concentrated at singularities, capturing the net turning or angle deficit without requiring smooth differentiability.

12.3 Dependence on parametrization smoothness

Even if a geometric object is visually smooth, a chosen parametrization might not be sufficiently differentiable for classical curvature formulas. Curvature computations often assume a minimum differentiability class (such as twice differentiability for second-derivative formulas). Ensuring consistent smoothness conditions is crucial when applying curvature definitions computationally or in proofs.

13 Notation, units, and conventions

13.1 Sign conventions for signed curvature

In planar curves, signed curvature depends on orientation: it distinguishes leftward versus rightward turning. Different texts may use opposite sign conventions based on whether positive corresponds to clockwise or counterclockwise rotation, or on the direction of the chosen normal. The magnitude of curvature is invariant; only the sign changes with convention.

13.2 Units and dimensional analysis

Curvature has units of inverse length. If coordinates carry units (e.g., meters), then curvature computed from derivatives must scale accordingly, because curvature measures bending per unit distance. This dimensional viewpoint helps check formula consistency in both analytic and numerical contexts.

13.3 Coordinate and frame choices

Curvature can be expressed using different coordinate systems or moving frames. While the specific formulas may change, properly defined curvature quantities should match under coordinate transformations. In higher dimensions, choosing frames to diagonalize or simplify the shape operator can make principal curvatures and related invariants more transparent.