Curvature — Riemann, Ricci, and Scalar

Curvature quantifies the failure of a Riemannian or pseudo-Riemannian manifold to be flat. The Riemann curvature tensor

R^{ρ}_{σ μν}

is the complete local invariant measuring how vectors fail to return to themselves after parallel transport around an infinitesimal loop; its successive traces yield the Ricci tensor

R_{μν}

(volume distortion along geodesics) and the scalar curvature

R

(a single number per point). Applies wherever a connection exists: differential geometry, general relativity, geometric analysis, and PDE. Developed by Riemann (1854), Christoffel, Ricci-Curbastro and Levi-Civita (1900), crowned by Einstein's field equations (1915).

Given a smooth manifold $M$ with a metric $g$ , the Levi-Civita connection $\nabla$ provides a unique torsion-free, metric-compatible way to differentiate vector fields. The Riemann curvature tensor is the obstruction to $\nabla$ behaving like ordinary differentiation: it measures the non-commutativity of second covariant derivatives, equivalently the holonomy of parallel transport around small loops. Curvature matters because it is the local, coordinate-independent invariant distinguishing genuinely curved spaces from flat ones — a sphere cannot be flattened onto a plane precisely because $R \neq = 0$ . The Ricci tensor and scalar curvature are partial contractions carrying less information but with profound geometric and physical meaning, governing volume growth, geodesic focusing, and ultimately the gravitational field. A surprising fact: in dimension 2 and 3 the full Riemann tensor is algebraically determined by Ricci and scalar curvature, but in dimension $\geq 4$ a genuinely new piece — the Weyl tensor — survives, encoding tidal forces and gravitational radiation in vacuum.

The Core Insight

Curvature is the commutator of covariant derivatives. Parallel transport that depends on the path is the geometric signature of curvature; algebraically this is captured by

[\nabla_{μ}, \nabla_{ν}] V^{ρ} = R^{ρ}_{σ μν} V^{σ}

. Everything else — Ricci, scalar, sectional, Gauss curvature, Einstein equations, geodesic deviation — is a contraction, specialization, or trace of this single object.

Formal Statement

Levi-Civita connection

On a (pseudo-)Riemannian manifold $(M, g)$ there is a unique affine connection $\nabla$ that is torsion-free ( $\nabla_{X} Y - \nabla_{Y} X = [X, Y]$ ) and metric-compatible ( $\nabla g = 0$ ). In coordinates its Christoffel symbols are

Γ_{μν}^{λ} = \frac{1}{2} g^{λσ} (\partial_{μ} g_{σ ν} + \partial_{ν} g_{σ μ} - \partial_{σ} g_{μν}) .

⟳write metric compatibility for the three cyclic permutations

\partial_{μ} g_{ν σ}

, add two and subtract the third, then use symmetry

Γ_{μν}^{λ} = Γ_{ν μ}^{λ}

from vanishing torsion.

Here $g_{μν}$ are metric components, $g^{μν}$ its inverse, and $\partial_{μ} = \partial / \partial x^{μ}$ . The $Γ$ 's are not tensor components.

Riemann curvature tensor

The Riemann tensor is defined by the failure of covariant derivatives to commute. For a vector field $V$ ,

R^{ρ}_{σ μν} V^{σ} = (\nabla_{μ} \nabla_{ν} - \nabla_{ν} \nabla_{μ}) V^{ρ},

with the explicit coordinate expression

R^{ρ}_{σ μν} = \partial_{μ} Γ_{ν σ}^{ρ} - \partial_{ν} Γ_{μ σ}^{ρ} + Γ_{μ λ}^{ρ} Γ_{ν σ}^{λ} - Γ_{ν λ}^{ρ} Γ_{μ σ}^{λ} .

⟳expand

\nabla_{μ} (\nabla_{ν} V^{ρ})

treating

\nabla_{ν} V^{ρ}

as a

(1, 1)

-tensor, antisymmetrize in

μν

; the

\partial\partial V

terms cancel and the connection terms organize into [3].

Lowering the upper index, $R_{ρ σ μν} = g_{ρ α} R^{α}_{σ μν}$ , exposes the algebraic symmetries.

Algebraic symmetries of the Riemann tensor

The fully covariant Riemann tensor satisfies:

R_{ρ σ μν} = - R_{σ ρ μν} = - R_{ρ σ ν μ} = R_{μν ρ σ}, R_{ρ [σ μν]} = 0.

⟳antisymmetry in the last pair is built into [2]; antisymmetry in the first pair follows from

\nabla g = 0

applied to

\nabla_{[μ} \nabla_{ν]} g_{ρ σ} = 0

; the first Bianchi identity

R_{ρ [σ μν]} = 0

follows from torsion-freeness via the Jacobi identity for

\nabla

In $n$ dimensions the number of independent components is $\frac{1}{12} n^{2} (n^{2} - 1)$ : this is $1$ for $n = 2$ , $6$ for $n = 3$ , $20$ for $n = 4$ .

Differential (second) Bianchi identity

\nabla_{[λ} R_{ρ σ] μν} = 0 ⟺ \nabla_{λ} R_{ρ σ μν} + \nabla_{μ} R_{ρ σ ν λ} + \nabla_{ν} R_{ρ σ λ μ} = 0.

⟳work in normal coordinates at a point where

Γ = 0

(but

\partial Γ \neq = 0

); then

R

is linear in

\partial Γ

and the cyclic sum of

\partial

-derivatives telescopes to zero.

Contracting [5] twice yields the contracted Bianchi identity $\nabla^{μ} G_{μν} = 0$ — the rock on which Einstein's equations rest.

Ricci tensor and scalar curvature

The Ricci tensor is the trace of Riemann over the first and third indices:

R_{μν} = R^{λ}_{μ λ ν} = g^{λσ} R_{λ μ σ ν} .

⟳the symmetries [4] guarantee this is the only independent nontrivial single trace (other contractions vanish or reproduce

\pm R_{μν}

). Symmetry

R_{μν} = R_{ν μ}

follows from the pair-swap symmetry in [4].

The scalar curvature is the full trace:

R = g^{μν} R_{μν} = g^{μν} g^{λσ} R_{λ μ σ ν} .

⟳contract [6] with the inverse metric; this is the unique scalar invariant linear in second derivatives of

g

The trace-free Einstein tensor $G_{μν} = R_{μν} - \frac{1}{2} R g_{μν}$ is divergence-free by [5].

Sectional curvature

For a 2-plane $Π = span (u, v) \subset T_{p} M$ , the sectional curvature is

K (Π) = \frac{R ( u , v , v , u )}{g ( u , u ) g ( v , v ) - g ( u , v ) ^{2}} = \frac{R _{ρ σ μν} u ^{ρ} v ^{σ} v ^{μ} u ^{ν}}{∣ u ∣ ^{2} ∣ v ∣ ^{2} - ⟨ u , v ⟩ ^{2}} .

⟳the denominator is the squared area of the parallelogram spanned by

u, v

; the numerator is the Gaussian curvature of the surface swept by geodesics tangent to

Π

(Gauss's Theorema Egregium in tensor form).

Knowing $K (Π)$ for all planes at $p$ determines the full Riemann tensor at $p$ .

Parallel transport around a loop on a sphere reveals curvature as holonomy: the vector returns rotated by an angle equal to the enclosed area times $K$.

Why It Holds

Riemann tensor as the commutator — full derivation

We show that $(\nabla_{μ} \nabla_{ν} - \nabla_{ν} \nabla_{μ}) V^{ρ}$ is tensorial in $V$ and equals $R^{ρ}_{σ μν} V^{σ}$ with [3].

Step 1. Compute

\nabla_{ν} V^{ρ} = \partial_{ν} V^{ρ} + Γ_{ν σ}^{ρ} V^{σ}

. Denote this

(1, 1)

-tensor

W^{ρ}_{ν}

. Step 2. Differentiate again, treating

W

as a

(1, 1)

-tensor:

\nabla_{μ} W^{ρ}_{ν} = \partial_{μ} W^{ρ}_{ν} + Γ_{μ λ}^{ρ} W^{λ}_{ν} - Γ_{μν}^{λ} W^{ρ}_{λ} .

Step 3. Substitute

W^{ρ}_{ν} = \partial_{ν} V^{ρ} + Γ_{ν σ}^{ρ} V^{σ}

and expand. The result contains:

\partial_{μ} \partial_{ν} V^{ρ}

, terms

\partial (Γ V)

, and

ΓΓ V

, plus a

Γ_{μν}^{λ}

term symmetric in

μν

. Step 4. Antisymmetrize in

μ \leftrightarrow ν

. The symmetric double-derivative

\partial_{μ} \partial_{ν} V^{ρ}

cancels. The symmetric term

Γ_{μν}^{λ} \nabla_{λ} V^{ρ}

cancels (since

Γ_{μν}^{λ} = Γ_{ν μ}^{λ}

, this is where torsion-freeness enters). The first-derivative-of-

V

terms cancel by symmetry of

Γ

in lower indices. Step 5. What survives multiplies

V^{σ}

\partial_{μ} Γ_{ν σ}^{ρ} - \partial_{ν} Γ_{μ σ}^{ρ} + Γ_{μ λ}^{ρ} Γ_{ν σ}^{λ} - Γ_{ν λ}^{ρ} Γ_{μ σ}^{λ} = R^{ρ}_{σ μν} .

Since the LHS of [2] is manifestly tensorial (a difference of tensors) and acts $C^{\infty}$ -linearly on $V$ , $R^{ρ}_{σ μν}$ is a genuine $(1, 3)$ -tensor despite being built from non-tensorial $Γ$ 's. $■$

Geodesic deviation: Ricci controls volume focusing

Let $γ_{s} (t)$ be a one-parameter family of geodesics with tangent $T^{μ} = \overset{γ}{˙}^{μ}$ and separation (Jacobi) field $S^{μ} = \partial γ^{μ} / \partial s$ .

Step 1. Geodesic equation:

T^{ν} \nabla_{ν} T^{μ} = 0

. Since

[T, S] = 0

, one has

T^{ν} \nabla_{ν} S^{μ} = S^{ν} \nabla_{ν} T^{μ}

. Step 2. Compute the relative acceleration

A^{μ} = T^{ν} \nabla_{ν} (T^{λ} \nabla_{λ} S^{μ})

. Commuting derivatives introduces Riemann:

\frac{D ^{2} S ^{μ}}{d t ^{2}} = - R^{μ}_{ν ρ σ} T^{ν} S^{ρ} T^{σ} .

⟳replace

T \nabla S

S \nabla T

, commute

T \nabla S \nabla T \to S \nabla T \nabla T + R (\dots)

; the geodesic term vanishes leaving Riemann.

Step 3. For a bundle of geodesics with cross-sectional volume

δ V

, the expansion

θ = \nabla_{μ} T^{μ}

obeys the Raychaudhuri equation:

\frac{d θ}{d t} = - \frac{1}{n - 1} θ^{2} - σ_{μν} σ^{μν} + ω_{μν} ω^{μν} - R_{μν} T^{μ} T^{ν} .

⟳take the trace of the evolution of

B_{μν} = \nabla_{ν} T_{μ}

, split into trace (

θ

), symmetric trace-free (shear

σ

), antisymmetric (rotation

ω

); the Riemann term contracts to

R_{μν} T^{μ} T^{ν}

Thus $R_{μν} T^{μ} T^{ν} > 0$ (focusing) forces geodesics to converge — the geometric content of the Ricci tensor and the engine of singularity theorems. $■$

Bernhard Riemann | 1826–1866

In his 1854 Göttingen Habilitationsvortrag "Über die Hypothesen, welche der Geometrie zu Grunde liegen" — delivered before Gauss — Riemann founded $n$ -dimensional differential geometry, introduced the metric $d s^{2} = \sum g_{ij} d x^{i} d x^{j}$ , and gave the curvature in terms of what we now call sectional curvature, generalizing Gauss's surface theory to arbitrary dimension. He understood curvature as the obstruction to local flatness and even speculated that physical space might be curved and that geometry should be determined by the matter within it — anticipating general relativity by sixty years. He died at 39 of tuberculosis, leaving the tensorial machinery to be completed by Christoffel, Ricci, and Levi-Civita.

Building Intuition

Picture an ant walking a small closed loop while keeping a spear pointing "as straight as possible" (parallel transport). On a flat plane the spear returns unchanged. On a sphere it comes back rotated — the rotation angle equals the enclosed area times the Gaussian curvature. The Riemann tensor is the infinitesimal version of this rotation: feed it two vectors ( $μ, ν$ ) spanning the loop and it returns the linear map ( $ρ, σ$ ) that rotates transported vectors.

The Ricci tensor answers a coarser question: if I shoot a small ball of test particles along geodesics in direction $T$ , does the ball's volume shrink or grow? $R_{μν} T^{μ} T^{ν}$ is the focusing rate. The scalar curvature $R$ is coarser still: it compares the volume of a small geodesic ball to a Euclidean ball of the same radius:

\frac{Vol ( B _{r} )}{Vol _{Eucl} ( B _{r} )} = 1 - \frac{R}{6 ( n + 2 )} r^{2} + O (r^{4}) .

Positive

R

⇒ balls are smaller than Euclidean (sphere-like crowding); negative

R

⇒ larger (hyperbolic spreading).

Worked Examples

The round 2-sphere of radius $a$

Given: Metric

d s^{2} = a^{2} (d θ^{2} + sin^{2} θ d ϕ^{2})

, so

g_{θ θ} = a^{2}

g_{ϕϕ} = a^{2} sin^{2} θ

. Step 1 — Christoffels. Nonzero ones:

Γ_{ϕϕ}^{θ} = - sin θ cos θ

Γ_{θ ϕ}^{ϕ} = Γ_{ϕθ}^{ϕ} = cot θ

. Step 2 — Riemann. Using [3]:

R^{θ}_{ϕθ ϕ} = \partial_{θ} Γ_{ϕϕ}^{θ} - Γ_{ϕϕ}^{θ} Γ_{ϕθ}^{ϕ} = (- cos 2 θ) - (- sin θ cos θ) (cot θ) = sin^{2} θ .

Lower:

R_{θ ϕθ ϕ} = g_{θ θ} R^{θ}_{ϕθ ϕ} = a^{2} sin^{2} θ .

Step 3 — Sectional/Gauss curvature. Denominator

= g_{θ θ} g_{ϕϕ} = a^{4} sin^{2} θ

, so

K = \frac{a ^{2} sin ^{2} θ}{a ^{4} sin ^{2} θ} = \frac{1}{a ^{2}} .

Step 4 — Ricci and scalar.

R_{θ θ} = g^{ϕϕ} R_{ϕθ ϕθ} = \frac{1}{a ^{2} s i n ^{2} θ} \cdot a^{2} sin^{2} θ = 1

; similarly

R_{ϕϕ} = sin^{2} θ

. Thus

R_{μν} = \frac{1}{a ^{2}} g_{μν}

(Einstein metric). Scalar:

R = g^{μν} R_{μν} = 2 K = \frac{2}{a ^{2}} [units: (length)^{- 2}]

For $a = 1$ m, $R = 2 m^{- 2}$ , constant everywhere — as it must be by symmetry.

Schwarzschild scalar curvature is zero — but Riemann is not

Given: Schwarzschild vacuum metric (

c = 1

d s^{2} = - (1 - \frac{2 GM}{r}) d t^{2} + (1 - \frac{2 GM}{r})^{- 1} d r^{2} + r^{2} d Ω^{2}

. Step 1 — Vacuum ⇒ Ricci-flat. Einstein's vacuum equation gives

R_{μν} = 0

, hence

R = 0

everywhere outside

r = 2 GM

. No scalar curvature whatsoever. Step 2 — Riemann is nonzero. Tidal forces are real. A representative component:

R^{\hat{t}}_{\overset{r}{^} \hat{t} \overset{r}{^}} = - \frac{2 GM}{r ^{3}} (orthonormal frame) .

Step 3 — Curvature invariant (Kretschmann scalar). The diagnostic that survives:

K_{Kr} = R_{ρ σ μν} R^{ρ σ μν} = \frac{48 G ^{2} M ^{2}}{r ^{6}} .

Step 4 — Interpretation. At the horizon

r = 2 GM

K_{Kr} = \frac{48 G ^{2} M ^{2}}{( 2 GM ) ^{6}} = \frac{3}{4 G ^{4} M ^{4}}

— finite, so the horizon is a coordinate (not curvature) singularity. As

r \to 0

K_{Kr} \to \infty

: a true singularity.

R = 0, R_{μν} = 0, R_{ρ σ μν} R^{ρ σ μν} = \frac{48 G ^{2} M ^{2}}{r ^{6}} m^{- 4} (geometrized)

For a solar-mass black hole ( $GM / c^{2} \approx 1.5$ km), at $r = 10$ km the Kretschmann scalar is $\sim 48 (1.5)^{2} /1 0^{6} k m^{- 4} \approx 1.1 \times 1 0^{- 4} k m^{- 4}$ .

Common Misconceptions

Ricci-flat is not flat.

R_{μν} = 0

allows nonzero Riemann tensor (Schwarzschild, gravitational waves, all Ricci-flat Kähler/Calabi–Yau manifolds). Only

R^{ρ}_{σ μν} \equiv 0

means flat. In dimension

\leq 3

the Weyl tensor vanishes identically so Ricci-flat does imply flat — but never in

n \geq 4

. Scalar curvature is weak.

R

is a single function; it loses enormous information. Two very different geometries can share the same

R

. The Yamabe problem (always achievable: conformally rescale to constant

R

) shows how floppy scalar curvature is compared to Ricci. $Γ$ is not a tensor. Christoffel symbols transform inhomogeneously; you cannot conclude flatness from

Γ = 0

at one point (always achievable via normal coordinates). Flatness requires the tensor

R

to vanish. A flat torus has zero curvature but nontrivial holonomy/topology. Take

R^{2} / Z^{2}

with the inherited Euclidean metric. Here

R^{ρ}_{σ μν} \equiv 0

: locally indistinguishable from the plane, all sectional curvatures vanish, parallel transport around contractible loops is trivial. Yet the torus is globally distinct from the plane, and parallel transport around a non-contractible loop is the identity but the loop cannot be shrunk. Lesson: curvature is a local invariant; it does not see global topology directly (that is the business of Chern–Gauss–Bonnet, which integrates curvature against topology). Conversely, the cylinder

S^{1} \times R

is intrinsically flat (

K = 0

) despite looking "curved" in

R^{3}

— extrinsic bending is invisible to the intrinsic Riemann tensor (Theorema Egregium).

Connections

The Riemann tensor is the curvature 2-form $Ω = d ω + ω \land ω$ of the Levi-Civita Connection expressed in a frame, linking to Principal Bundles and Chern–Weil Theory where characteristic classes (Euler, Pontryagin, Chern) are built from curvature polynomials. The decomposition $Riemann = Weyl \oplus Ricci-traceless \oplus Scalar$ under the orthogonal group relates to Representation Theory of $O (n)$ . Sectional curvature bounds connect to Comparison Geometry (Rauch, Toponogov, Bishop–Gromov) and to Ricci Flow, Hamilton and Perelman's heat-equation-like evolution $\partial_{t} g = - 2 Ric$ that proved the Poincaré Conjecture. Via Geodesic Deviation and the Raychaudhuri equation, Ricci curvature feeds the Singularity Theorems of Penrose and Hawking. The scalar curvature is the Lagrangian density of Einstein–Hilbert Action, whose variation yields the Einstein Field Equations.

Real-World Applications

General Relativity / GPS. The Einstein equations $G_{μν} = 8 π G T_{μν}$ equate curvature to energy-momentum; GPS satellite clocks must be corrected for spacetime curvature (gravitational time dilation $\sim 38 μ s$ /day) or positioning errors accumulate at $\sim 10$ km/day.
Gravitational-wave detection (LIGO/Virgo). Detected strains $h \sim 1 0^{- 21}$ are oscillations of the Weyl (vacuum Riemann) tensor; the detector arms measure geodesic deviation [9] directly.
Cosmology. The Friedmann equations come from Ricci/scalar curvature of the FLRW metric; the spatial curvature parameter $k \in {- 1, 0, + 1}$ is constrained by CMB data to $∣ Ω_{k} ∣ < 0.005$ (nearly flat universe).
Black-hole astrophysics & EHT imaging. Photon orbits and shadow size are governed by Schwarzschild/Kerr curvature; the 2019 M87* image tests strong-field Riemann curvature.
Computer graphics & geometry processing. Discrete Gaussian/mean curvature drives mesh smoothing, remeshing, and the discrete Laplace–Beltrami operator.
Machine learning / information geometry. The Fisher information metric endows statistical manifolds with curvature; natural-gradient methods exploit Ricci-like quantities, and curvature of loss landscapes guides optimization.
Manifold learning & data science. Ricci curvature on graphs (Ollivier–Ricci) detects community structure and bottlenecks in networks.

Historical Development

1827 — Gauss's Disquisitiones generales circa superficies curvas introduces Gaussian curvature and the Theorema Egregium: curvature is intrinsic.
1854 — Riemann's Habilitationsvortrag generalizes to $n$ dimensions, defining the metric and sectional curvature; lays the foundation for the curvature tensor.
1869 — Elwin Christoffel introduces the symbols $Γ_{μν}^{λ}$ and the covariant differentiation that bears his name; effectively writes down the four-index curvature tensor.
1888–1896 — Lipschitz and Beltrami develop related curvature expressions.
1900 — Ricci-Curbastro and Levi-Civita publish Méthodes de calcul différentiel absolu, codifying tensor calculus; the Ricci tensor is named.
1913–1915 — Einstein, guided by Grossmann (and racing Hilbert), arrives at the field equations $G_{μν} = 8 π G T_{μν}$ ; the contracted Bianchi identity $\nabla^{μ} G_{μν} = 0$ guarantees energy-momentum conservation.
1917 — Levi-Civita and Schouten formalize parallel transport, clarifying curvature as holonomy.
1925–1926 — Élie Cartan reformulates curvature via moving frames and differential forms ( $Ω = d ω + ω \land ω$ ), unifying it with Lie-group geometry.
1956–1963 — Decomposition by Weyl/Singer–Thorpe; comparison geometry (Rauch, Berger, Toponogov); Hawking–Penrose singularity theorems (1965–70) use Ricci curvature.
1982–2003 — Hamilton's Ricci flow and Perelman's resolution of the Poincaré and geometrization conjectures place curvature evolution at the center of modern geometry.

Subtleties & Limits

Prescribing curvature. Which functions arise as the scalar curvature of a metric in a given conformal class? The Yamabe problem (solved by Trudinger–Aubin–Schoen) addresses constant

R

; the broader Kazdan–Warner obstructions limit which

R

are achievable. Prescribing the full Ricci tensor (the Ricci-prescription / DeTurck problem) is far harder and only locally understood. Positive mass and scalar curvature. The relationship between nonnegative scalar curvature, minimal surfaces, and the positive mass theorem (Schoen–Yau, Witten) remains an active frontier, with recent dimensional-restriction results and the Gromov dihedral/torus rigidity conjectures only partly resolved. Singularities of Ricci flow. Classifying finite-time singularities (Type I/II, ancient solutions, the structure of

κ

-solutions) in dimension

\geq 4

is open; the topology one can extract from curvature pinching beyond Brendle–Schoen's differentiable sphere theorem is incomplete. Quantum gravity. What replaces the smooth Riemann tensor at the Planck scale — whether spacetime curvature is fundamental or emergent (entanglement, causal sets, spin foams) — is the deepest open question touching this material.