advanced30 min studyLesson 454 in Mathematics

Riemann curvature and sectional curvature

Fix a consistent curvature sign, compute a coordinate-independent sectional quantity, and distinguish a nonzero connection coefficient from actual curvature.

On this page

What you’ll understand

  • State the module's curvature convention and sectional-curvature quotient.
  • Explain tensoriality and the role of metric symmetries.
  • Compute flat polar curvature and the scaling of constant-curvature metrics.

A moving coordinate frame can produce nonzero connection coefficients even in a flat plane. Curvature measures a deeper obstruction: covariant differentiation in two directions fails to commute after accounting for the vector fields' own commutator. Because authors choose opposite signs for this operator, a formula is meaningful only after its convention and sectional-curvature contraction are declared together. We keep one choice throughout this module.

Fix the sign before making comparisons

Use R(X,Y)Z=∇X∇YZ−∇Y∇XZ−∇[X,Y]ZR(X,Y)Z=\nabla_X\nabla_YZ-\nabla_Y\nabla_XZ-\nabla_{[X,Y]}Z. This expression is linear over smooth functions in each of X, Y, and Z, so its value at a point depends only on the three tangent vectors there. The bracket correction is essential to that tensoriality. With this convention, a constant-curvature space satisfies R(X,Y)Z=K(g(Y,Z)X−g(X,Z)Y)R(X,Y)Z=K(g(Y,Z)X-g(X,Z)Y), and the standard round sphere has positive K.

The tensor is antisymmetric in its first two inputs. Metric compatibility supplies additional symmetries after pairing its output with a fourth vector, and torsion freedom gives the first Bianchi identity. These restrict the independent components dramatically, but they do not make the raw Christoffel symbols tensorial. When consulting a source with the opposite convention, reverse its curvature operator and adjust all related formulas together; changing only the word positive would corrupt the Jacobi equation and comparison inequalities later.

R(X,Y)Z=∇X∇YZ−∇Y∇XZ−∇[X,Y]Z,K(span⁡{u,v})=g(R(u,v)v,u)g(u,u)g(v,v)−g(u,v)2R(X,Y)Z=\nabla_X\nabla_YZ-\nabla_Y\nabla_XZ-\nabla_{[X,Y]}Z,\qquad K(\operatorname{span}\{u,v\})=\frac{g(R(u,v)v,u)}{g(u,u)g(v,v)-g(u,v)^2}
The vectors u,v are linearly independent. This paired choice of operator sign and contraction makes the round sphere's sectional curvature positive.↔ Scroll wide equations horizontally.

Sectional curvature belongs to a tangent two-plane

The denominator in the sectional quotient is the squared area of the parallelogram spanned by u and v. Tensor symmetries show that changing the spanning basis multiplies numerator and denominator by the same squared determinant, leaving their ratio unchanged. For an orthonormal pair, the denominator is one. Sectional curvature therefore measures the selected two-dimensional tangent direction, not the lengths of the chosen coordinate basis vectors. In dimension two there is only one tangent two-plane, giving Gaussian curvature.

On a round sphere of radius ρ, every tangent two-plane has curvature 1/ρ21/\rho^2. More generally, multiplying a metric by c² leaves its Levi-Civita connection and curvature operator of type (1,3) unchanged, because the scaling is constant and cancels in the Christoffel formula. Pairing the curvature output gains c², while the area denominator gains c⁴, so sectional curvature is divided by c². Keeping track of tensor type prevents an apparent contradiction between an unchanged connection and changed numerical curvature.

g~=c2g(c>0)⟹∇~=∇,K~=c−2K\widetilde g=c^2g\quad(c>0)\quad\Longrightarrow\quad \widetilde\nabla=\nabla,\qquad \widetilde K=c^{-2}K
The scaling factor is constant. The curvature operator is unchanged, but sectional curvature includes metric normalization and therefore scales inversely with squared length.↔ Scroll wide equations horizontally.

Coordinates cannot manufacture curvature

In the Euclidean polar chart, coordinate fields commute and the connection coefficients from the preceding lesson are nonzero. Computing curvature requires differentiating those coefficients and combining the connection products. The terms cancel exactly. This cancellation reflects the underlying Euclidean geometry, even though the chart basis rotates and stretches. A singular expression at the polar origin is a failure of the chart, not evidence that the plane has a geometric curvature singularity.

Curvature is local, while some properties of geodesics depend on global topology. A flat cylinder and a flat plane both have zero sectional curvature but differ in their closed routes and global exponential maps. Conversely, a surface can be bent in a surrounding space without changing its intrinsic metric, as rolling a flat sheet into a cylinder illustrates locally. Intrinsic sectional curvature should therefore be distinguished from the ambient bending of an embedded curve or surface.

Curvature cancellation in polar coordinates

For the Euclidean polar metric, compute R(∂r,∂θ)∂θR(\partial_r,\partial_\theta)\partial_\theta using ∇θ∂θ=−r∂r\nabla_\theta\partial_\theta=-r\partial_r and ∇r∂θ=(1/r)∂θ\nabla_r\partial_\theta=(1/r)\partial_\theta.

  1. The first iterated derivative is ∇r(−r∂r)=−∂r\nabla_r(-r\partial_r)=-\partial_r, since ∇r∂r=0\nabla_r\partial_r=0.

  2. The second is ∇θ((1/r)∂θ)=(1/r)(−r∂r)=−∂r\nabla_\theta((1/r)\partial_\theta)=(1/r)(-r\partial_r)=-\partial_r, because θ-differentiation leaves r unchanged. Coordinate fields commute, so the bracket term is zero.

  3. Subtracting the two equal terms gives the zero vector. The sectional denominator for the pair is r²>0 away from the origin, and the numerator is zero. Thus the sectional curvature is zero despite the nonzero Christoffel symbols.

The result

The polar plane is intrinsically flat; the connection coefficients' derivative and product terms cancel.

Explore the model

Interactive lab

Interactive lab · matrix

Shear is not curvature

Pause and Reset, then set a = 1 and b = 0 to inspect the identity grid. Reset and set both a = 1 and b = 1. Predict a sheared grid with determinant one and straight lines still straight. Treat each chosen constant map as defining a pullback Euclidean metric AᵀA: both are flat even though their coordinate metric matrices differ. Playback interpolates linear drawings, not curvature changing over a manifold.

Shear is not curvatureThe selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. Current readings: Area scale · det A: 1.4; Singular values · σmax, σmin: 1.597, 0.876; Condition number · κ₂(A): 1.823. All parameters are adjustable with the labeled controls below.Matrix A with first row 1.4, 0.6 and second row zero, one; display interpolates between identity and A while final determinant remains 1.4.−4−2024−2−1012x · equal geometric scaleyA = [ 1.4 0.6 ; 0 1 ]Display blend: 0% · final values below
Interpolated geometryOriginal unit squareSelected final transform
Area scale · det A
1.4
Singular values · σmax, σmin
1.597, 0.876
Condition number · κ₂(A)
1.823
t = 0 s

The selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forA distorted coordinate grid or a nonidentity metric matrix at a point does not establish nonzero sectional curvature.

Check your understanding

Under constant metric rescaling g̃=9g, how does sectional curvature change with this module's convention?

Try it yourself

In a constant-curvature space with K=−1/4, let e₁,e₂ be orthonormal and choose u=2e₁, v=3e₂. Compute R(u,v)v, the sectional numerator, and the denominator. Verify the quotient.

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

Use R(u,v)v=K(g(v,v)u−g(u,v)v). Orthogonality removes the second term, but the final pairing with u supplies another factor.

Compare with the solution

Here g(v,v)=9g(v,v)=9 and g(u,v)=0g(u,v)=0, so R(u,v)v=(−1/4)(9)(2e1)=−(9/2)e1R(u,v)v=(-1/4)(9)(2e_1)=-(9/2)e_1. Pairing with u gives numerator minus nine. The denominator is 4⋅9−0=364\cdot9-0=36, so the quotient is −9/36=−1/4-9/36=-1/4, as required. Rescaling the spanning vectors has changed the raw numerator but not the curvature of their tangent two-plane. The calculation uses the declared operator convention and its matching contraction.

Take it with you

  • Curvature signs must be read together with the operator and sectional contraction conventions.
  • Sectional curvature is normalized by tangent area and is independent of spanning basis.
  • Constant metric scaling changes K by inverse squared length, while coordinate distortion alone creates no curvature.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.