Algebraic Curvature Symmetries and Decomposition

Scrollable formula. Use Left and Right arrow keys to reveal hidden content.

Scrollable table. Use Left and Right arrow keys to reveal additional columns.

Scrollable figure. Use Left and Right arrow keys to inspect hidden content.

For a Levi–Civita connection, curvature is far more structured than an arbitrary rank-four tensor. Its pair antisymmetries, pair exchange symmetry, and algebraic Bianchi identity reduce the independent components to $n^2(n^2-1)/12$ ; traces then split those components into Weyl, trace-free Ricci, and scalar parts. The trace-free Weyl tensor is precisely what survives when every contraction is removed.

Required background. Affine Connections and Covariant Derivatives supplies the general connection and tensor-derivative framework; The Levi-Civita Connection supplies the torsion-free metric specialization used throughout.

Helpful background. Calculus and Linear Algebra for Relativity supplies linear maps, bilinear forms, and index notation; the traces and dimension counts are developed here. Exterior Algebra and Differential Forms supplies the two-form and wedge algebra used in the curvature-operator comparison.

Algebraic symmetries of Levi–Civita curvature

Lower the first index of the curvature tensor. Metric compatibility, torsion freedom, and the commutator convention imply

\begin{aligned} R_{abcd}&=-R_{bacd}, & R_{abcd}&=-R_{abdc},\\ R_{abcd}&=R_{cdab}, & R_{a[bcd]}&=0. \end{aligned}

Here brackets denote antisymmetrization with unit weight. The first two relations say that $R$ is a bilinear form on $\Lambda^2T_pM$ ; pair exchange makes that form symmetric. The final relation is the algebraic, or first, Bianchi identity. These properties require the Levi–Civita hypotheses. For a general affine connection with torsion or nonmetricity, the corresponding identities acquire extra terms and should not be imported unchanged.

One route to the pair-exchange symmetry is to combine metric compatibility, $g(R(X,Y)Z,W)=-g(Z,R(X,Y)W)$ , with the algebraic Bianchi identity. A complete proof from the connection axioms is given in Calegari 2022, Proposition 7.1.3, pp. 53–54.

Counting independent components

Let $N=n(n-1)/2$ be the dimension of $\Lambda^2T_pM$ . A symmetric bilinear form on that space has $N(N+1)/2$ components. Once the pair symmetries hold, the independent Bianchi constraints correspond to choices of four distinct indices: the fully antisymmetric part must vanish. This removes the dimension of $\Lambda^4T_pM$ , namely $\binom n4$ . Hence

\frac{N(N+1)}{2}-\binom n4 =\frac{n^2(n^2-1)}{12}.

This gives $1$ , $6$ , and $20$ independent curvature components in dimensions two, three, and four. The count is an important check on any decomposition: in four dimensions, the Weyl tensor has $10$ components, trace-free Ricci has $9$ , and the scalar part has $1$ , so $20=10+9+1$ .

The low-dimensional limits are informative. In two dimensions the scalar curvature determines all of $R_{abcd}$ . In three dimensions the Weyl tensor vanishes identically, so Ricci determines Riemann algebraically. Only for $n\ge4$ can a nonzero trace-free Weyl tensor carry curvature not recoverable from Ricci.

Weyl–Ricci–scalar decomposition

Define the Schouten tensor for $n\ge3$ by

P_{ab}=\frac{1}{n-2} \left(R_{ab}-\frac{R}{2(n-1)}g_{ab}\right).

For symmetric two-tensors $A$ and $B$ , their Kulkarni–Nomizu product is

(A\mathbin{\owedge}B)_{abcd} =A_{ac}B_{bd}-A_{ad}B_{bc}-A_{bc}B_{ad}+A_{bd}B_{ac}.

Then

R_{abcd}=C_{abcd}+(g\mathbin{\owedge}P)_{abcd},

where every contraction of $C_{abcd}$ vanishes. Expanding $P$ gives

\begin{aligned} R_{abcd}={}&C_{abcd} +\frac{1}{n-2} \left(g_{ac}R_{bd}-g_{ad}R_{bc}-g_{bc}R_{ad}+g_{bd}R_{ac}\right)\\ &-\frac{R}{(n-1)(n-2)} \left(g_{ac}g_{bd}-g_{ad}g_{bc}\right). \end{aligned}

The coefficients follow from taking a trace, rather than from guesswork. Direct contraction gives

g^{ac}(g\mathbin{\owedge}P)_{abcd} =(n-2)P_{bd}+(\operatorname{tr}_gP)g_{bd}.

Demanding that this be $R_{bd}$ first fixes $\operatorname{tr}_gP=R/[2(n-1)]$ and then gives the displayed Schouten tensor. Thus $C=Riem-g\mathbin{\owedge}P$ has zero trace by construction. The other possible traces vanish by its pair symmetries. This verifies both the normalization and the reconstruction. The representation-theoretic organization is developed in Calegari 2022, §7.2, pp. 54–55, PDF.

Constant curvature and what traces miss

If every nondegenerate two-plane at a point has the same sectional curvature $K$ , then

R_{abcd}=K(g_{ac}g_{bd}-g_{ad}g_{bc}).

Contraction gives $R_{ab}=(n-1)Kg_{ab}$ and $R=n(n-1)K$ , while $C_{abcd}=0$ for $n\ge3$ . If this isotropic form holds throughout a connected manifold of dimension $n\gt2$ , the Bianchi identity forces $K$ to be constant. The round sphere has $K\gt0$ and verifies all signs. Yet $C=0$ does not generally mean flatness: constant-curvature spaces with $K\ne0$ are conformally flat but curved. Conversely, $R_{ab}=0$ does not force $R_{abcd}=0$ in dimensions four and above, because Weyl curvature may remain.

This page establishes an algebraic decomposition at one point. It does not classify concrete spacetimes or assign dynamical meaning to any summand. Derivatives of curvature, causal alignment, and frame normalization supply additional data when local equivalence is at issue.

The decomposition is orthogonal in the natural representation-theoretic sense, but “trace free” should not be read as “physically negligible.” In four dimensions the Weyl part contains half of the algebraically independent curvature components. It is the part unchanged, with one index raised, by conformal rescaling in dimensions above three, whereas the Schouten part carries the change in Ricci traces. This is an algebraic statement about conformal geometry, not yet a statement about radiation or tidal measurement.

The signature does not change the component count: it changes the orthogonal group and the causal classification of directions, but the polynomial index symmetries have the same dimensions. What does change is the behavior of contractions and canonical forms. A self-adjoint operator in an indefinite inner product need not be diagonalizable over the reals, so Riemannian eigenvalue intuition is insufficient for Lorentzian classification. That is one reason normalized frame data become necessary later in the chapter.

A reliable decomposition check has three layers. First contract the proposed Weyl tensor on every allowed pair and verify zero. Second reconstruct Riemann from $C$ , $R_{ab}$ , and $R$ and confirm all pair symmetries. Third test low-dimensional limits: $C$ must vanish identically in $n=3$ , while the Schouten formula itself is singular in $n=2$ and must be replaced by $R_{abcd}=\tfrac12R(g_{ac}g_{bd}-g_{ad}g_{bc})$ . These are structural limits, not optional simplifications.

Exercises

Verify the four-dimensional component split from the general formulas.

Solution

Riemann has $4^2(4^2-1)/12=20$ components. A symmetric Ricci tensor has $4(5)/2=10$ , decomposing into $9$ trace-free components and one scalar. The remainder has $20-10=10$ components and is the Weyl tensor. Thus $20=10+9+1$ in Weyl, trace-free Ricci, and scalar order.

References

Danny Calegari, Introduction to Differential Geometry, preliminary version, May 26, 2022, University of Chicago. PDF.