Contracted Bianchi Identities

The second Bianchi identity is a differential statement about the Riemann tensor — a cyclic constraint on how curvature changes from point to point — that holds for any metric-compatible, torsion-free connection. On its own it is a rich but unwieldy object with three free antisymmetrized indices. Its decisive role in gravitation appears only when it is contracted with the inverse metric. Each contraction spends one curvature index and buys a derivative relation among the lower-rank curvature objects: the first contraction ties the divergence of Riemann to gradients of the Ricci tensor, and the second ties the divergence of Ricci to the gradient of the Ricci scalar. Carry the contractions through and a single line falls out that is arguably the most consequential corollary in classical general relativity: the divergence of the Einstein tensor vanishes identically, $\nabla^{\mu}G_{\mu\nu}=0$ . This is not a field equation, not a property of any matter model, and not a gauge condition — it is a geometric identity built into the fabric of Riemannian (Lorentzian) geometry. It is exactly what let Einstein and Hilbert write a tensor field equation whose left-hand side is conserved by construction, so that setting it proportional to a conserved stress-energy tensor is consistent rather than contradictory. The identity also carries a counting message. Because $\nabla^{\mu}G_{\mu\nu}=0$ is four equations, only six of the ten Einstein equations are genuinely dynamical; the other four are constraints. That is the precise gravitational analogue of $\partial_{\mu}{}^{*}F^{\mu\nu}=0$ in electromagnetism, and it is the identity that the diffeomorphism (gauge) invariance of the theory forces on it — by Noether's second theorem the gauge freedom is the source of the identity, not its consequence. Throughout we assume the Levi-Civita connection, so that $g^{\mu\nu}$ passes freely through $\nabla$ because $\nabla_{\alpha}g_{\mu\nu}=0$ .

Key takeaways

The second Bianchi identity $\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0$ admits two successive metric contractions, each collapsing one curvature index into a derivative relation among lower-rank objects.
The once-contracted identity is $\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla_{\rho}R_{\nu\sigma}-\nabla_{\sigma}R_{\nu\rho}$ , linking the divergence of Riemann to gradients of the Ricci tensor.
The twice-contracted identity is $\nabla^{\mu}R_{\mu\nu}=\tfrac12\nabla_{\nu}R$ , equivalently $\nabla^{\mu}(R_{\mu\nu}-\tfrac12 g_{\mu\nu}R)=0$ .
Hence the Einstein tensor $G_{\mu\nu}=R_{\mu\nu}-\tfrac12 g_{\mu\nu}R$ is automatically divergence-free, identically and without invoking any field equation.
This is the geometric guarantee that $G_{\mu\nu}=8\pi G\,T_{\mu\nu}$ enforces local energy-momentum conservation $\nabla^{\mu}T_{\mu\nu}=0$ .
The four identities also turn four of the ten Einstein equations into constraints, leaving six dynamical equations — matching the four coordinate (gauge) functions.
Key idea: contracting $\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0$ once gives $\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla_{\rho}R_{\nu\sigma}-\nabla_{\sigma}R_{\nu\rho}$ ; contracting again gives $\nabla^{\mu}R_{\mu\nu}=\tfrac12\nabla_{\nu}R$ , which rearranges to the identically conserved Einstein tensor $\nabla^{\mu}G_{\mu\nu}=0$ .

The core insight

Contracting the differential Bianchi identity $\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0$ once gives $\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla_{\rho}R_{\nu\sigma}-\nabla_{\sigma}R_{\nu\rho}$ ; contracting again gives $\nabla^{\mu}R_{\mu\nu}=\tfrac12\nabla_{\nu}R$ , which rearranges to the identically divergence-free Einstein tensor $\nabla^{\mu}G_{\mu\nu}=0$ .

Where this lives today

The most vivid modern home of the contracted Bianchi identities is numerical relativity — the computational engine behind gravitational-wave astronomy. When LIGO and Virgo announced GW150914 in 2015, the first direct detection of a binary black-hole merger, the signal was interpreted by matching it against waveform templates computed from full numerical solutions of the Einstein equations. Those simulations split the ten field equations exactly as the constraint-counting derivation prescribes: six evolution equations march the metric forward in time, while the four $G^{0}{}_{\nu}=8\pi T^{0}{}_{\nu}$ are the Hamiltonian and momentum constraints, fixed on the initial slice. The contracted Bianchi identity guarantees that if the constraints hold initially, the evolution preserves them — but only exactly. In floating-point arithmetic they drift, and the growth of "constraint violation" is the central health monitor of every BSSN or generalized-harmonic code; constraint-damping schemes are engineered precisely to suppress it. Beyond simulation, the identity underwrites everyday cosmology: the Friedmann acceleration equation is never solved independently because it follows from the first Friedmann equation plus the continuity equation of Example 1, itself a contracted-Bianchi consequence. And in the search for physics beyond general relativity, the identity is a design constraint — any proposed modified-gravity Lagrangian must yield field equations with an identically conserved left-hand side, which is why Lovelock's theorem (that $G_{\mu\nu}+\Lambda g_{\mu\nu}$ is essentially unique in four dimensions) so tightly boxes in the space of viable theories.

The physical picture

The uncontracted Bianchi identity is a bookkeeping law about loops: it says that if you tally how the curvature two-form changes as you slide an infinitesimal loop around a closed circuit, the cyclic sum cancels. This is the geometric cousin of "the boundary of a boundary is zero" — the same topological reflex that makes $\partial_{[\alpha}F_{\beta\gamma]}=0$ true for the electromagnetic field. Contraction asks a narrower question of this loop law: not the full cyclic cancellation, but the net divergence left over when you sum along one loop direction. The remarkable answer is that this residual flux is never free — the geometry forces it to equal the gradient of a lower-rank curvature object. Do it once and the divergence of the full Riemann tensor is pinned to derivatives of the Ricci tensor. Do it a second time and the divergence of the Ricci tensor is pinned to half the gradient of the scalar curvature. At that point there are no free index pairs left on the curvature to contract, and the chain terminates. The endpoint is a specific combination — Ricci minus half the metric times the scalar — for which even this last residual flux cancels exactly. That combination is the Einstein tensor, and its divergence-free character is not imposed but discovered.

Figure

The two contractions of the second Bianchi identity form a cascade: contracting once turns the loop law into a divergence of the Riemann tensor, a second contraction gives ∇^μ R μν = ½ ∇ν R, and one rearrangement lands on the identically divergence-free Einstein tensor ∇^μ G μν = 0

Why does any of this matter for physics? Because a field theory of gravity must equate a geometric tensor to a source, and the source — energy and momentum — is locally conserved, $\nabla^{\mu}T_{\mu\nu}=0$ . If the geometric side were not automatically divergence-free, the equation would over-determine the metric and be inconsistent the moment matter moved. The contracted Bianchi identity is the escape: it hands us a geometric tensor whose divergence is guaranteed to vanish, so the field equations are consistent by construction. The very same four identities that guarantee this consistency also remove four of the ten equations from the roster of evolution equations, turning them into constraints on the initial data. That is the second figure: an accounting of degrees of freedom that must — and does — balance.

Figure

The identity ∇^μ G μν = 0 is four equations, and it balances the ledger of general relativity: of the ten Einstein equations four are constraints and six are dynamical, matching the ten metric functions minus four coordinate choices, so 10 = 6 + 4 on both sides — but the six metric functions left after fixing coordinates are not six physical degrees of freedom: the four constraints bind them too, and four-dimensional general relativity propagates only two, the two gravitational-wave polarizations

The core equations

Everything begins with the differential (second) Bianchi identity, a property of the Riemann tensor of any Levi-Civita connection. It is a total antisymmetrization over the covariant-derivative index and the first index pair of Riemann, and it is equivalent to a three-term cyclic sum. Nothing physical has yet been assumed — only the geometry of a metric connection.

\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0\quad\Longleftrightarrow\quad \nabla_{\lambda}R_{\mu\nu\rho\sigma}+\nabla_{\mu}R_{\nu\lambda\rho\sigma}+\nabla_{\nu}R_{\lambda\mu\rho\sigma}=0 \qquad (1)

where: $R_{\mu\nu\rho\sigma}$ — Riemann curvature tensor, all indices lowered; $\nabla_\lambda$ — Levi-Civita covariant derivative; $[\lambda\mu\nu]$ — total antisymmetrization over the derivative index and the first index pair. Reading: the way curvature varies is not arbitrary — its antisymmetrized gradient cancels cyclically, the differential shadow of "the boundary of a boundary is zero." The first contraction spends one metric factor to sum the derivative index against a member of the second index pair. The antisymmetry of Riemann in each pair does the rest: one of the three terms dies outright and the other two become gradients of the Ricci tensor. This is the once-contracted Bianchi identity.

\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla_{\rho}R_{\nu\sigma}-\nabla_{\sigma}R_{\nu\rho} \qquad (2)

where: $R_{\nu\sigma}=R^{\alpha}{}_{\nu\alpha\sigma}$ — Ricci tensor; $\nabla^{\mu}=g^{\mu\lambda}\nabla_\lambda$ — the raised covariant derivative, with the metric passed through because $\nabla_\alpha g_{\mu\nu}=0$ . Reading: the divergence of the full curvature tensor is not independent data — it is fixed entirely by how the Ricci tensor varies from point to point. Contract a second time, now pairing the two remaining free indices, and the divergence of the Ricci tensor is pinned to the gradient of the Ricci scalar. The factor of one half is the crucial fingerprint of this identity — it is exactly the coefficient that will make the Einstein tensor conserved.

\nabla^{\mu}R_{\mu\nu}=\tfrac{1}{2}\nabla_{\nu}R \qquad (3)

where: $R=g^{\mu\nu}R_{\mu\nu}$ — Ricci scalar; $\nabla_\nu R$ — its gradient. Reading: the Ricci tensor cannot have an arbitrary divergence; it is locked to half the gradient of its own trace. Rearrange equation (3) into the divergence of a single combination and the right-hand side cancels against itself. The Einstein tensor is that combination, and its vanishing divergence is an identity — true for every metric, with no physics assumed.

G_{\mu\nu}\equiv R_{\mu\nu}-\tfrac{1}{2}g_{\mu\nu}R,\qquad \nabla^{\mu}G_{\mu\nu}=0 \qquad (4)

where: $G_{\mu\nu}$ — Einstein tensor, symmetric and built from the metric and its first two derivatives; $g_{\mu\nu}$ — spacetime metric. Reading: subtract exactly one half the trace term and the residual flux left over from the second contraction cancels — this specific coefficient, and no other, produces a conserved tensor. The identity survives the addition of any constant multiple of the metric, because the metric is itself covariantly constant. Equating the resulting conserved tensor to the stress-energy source then forces local conservation of energy and momentum — a consequence of the field equations, not an extra postulate.

\nabla^{\mu}\big(G_{\mu\nu}+\Lambda g_{\mu\nu}\big)=0,\qquad G_{\mu\nu}+\Lambda g_{\mu\nu}=8\pi G\,T_{\mu\nu}\ \Rightarrow\ \nabla^{\mu}T_{\mu\nu}=0 \qquad (5)

where: $\Lambda$ — cosmological constant; $T_{\mu\nu}$ — stress-energy tensor; $G$ — Newton's constant in units with $c=1$ . Reading: because $\nabla_\alpha g_{\mu\nu}=0$ , the term $\Lambda g_{\mu\nu}$ rides along for free, and taking the divergence of both sides converts the geometric identity into the physical conservation law. In vacuum the Ricci tensor vanishes, so the right-hand side of the once-contracted identity (2) collapses, and the divergence of the trace-free part of curvature — the Weyl tensor — vanishes too. These are the source-free, Maxwell-like propagation equations that carry gravitational radiation.

R_{\mu\nu}=0\ \Rightarrow\ \nabla^{\mu}C_{\mu\nu\rho\sigma}=0 \qquad (6)

where: $C_{\mu\nu\rho\sigma}$ — Weyl conformal tensor, the trace-free piece of Riemann. Reading: strip out the Ricci content and what remains of the contracted Bianchi identity is a wave equation for the tidal field — the mathematical seed of gravitational-wave propagation.

Derivation AThe once-contracted Bianchi identity

The goal is to turn the fully antisymmetrized second Bianchi identity into a clean statement about the divergence of the Riemann tensor. We contract the covariant-derivative index against one member of the second index pair using the inverse metric, let metric compatibility carry the metric through the derivative, and then use the antisymmetry of Riemann in its first index pair together with the definition of the Ricci tensor to collapse two of the three terms — leaving a difference of Ricci gradients, which one final symmetry rewrites into the standard form.

Step 1. Write the cyclic form of the identity, $\nabla_{\lambda}R_{\mu\nu\rho\sigma}+\nabla_{\mu}R_{\nu\lambda\rho\sigma}+\nabla_{\nu}R_{\lambda\mu\rho\sigma}=0$ , and multiply through by $g^{\lambda\rho}$ , contracting the derivative index $\lambda$ against the third index $\rho$ . Why: contracting the derivative index into the second pair — rather than into its own first pair — is what produces Ricci tensors instead of a vanishing self-contraction, so the choice of which index to raise is the whole game.

Step 2. The first term becomes $g^{\lambda\rho}\nabla_{\lambda}R_{\mu\nu\rho\sigma}=\nabla^{\rho}R_{\mu\nu\rho\sigma}$ , the divergence of Riemann on its third slot. Why: metric compatibility, $\nabla_\alpha g_{\mu\nu}=0$ , lets the constant metric slide inside the derivative so that $g^{\lambda\rho}\nabla_\lambda=\nabla^\rho$ .

Step 3. The second term is $g^{\lambda\rho}\nabla_{\mu}R_{\nu\lambda\rho\sigma}=\nabla_{\mu}\big(g^{\lambda\rho}R_{\nu\lambda\rho\sigma}\big)$ ; using antisymmetry in the first pair, $R_{\nu\lambda\rho\sigma}=-R_{\lambda\nu\rho\sigma}$ , the contraction gives $-R_{\nu\sigma}$ , so the term is $-\nabla_{\mu}R_{\nu\sigma}$ . Why: the sign is decisive and is exactly where the base derivation stumbled — the minus comes from swapping the first pair before recognizing the standard Ricci contraction $g^{\lambda\rho}R_{\lambda\nu\rho\sigma}=R_{\nu\sigma}$ .

Step 4. The third term is $g^{\lambda\rho}\nabla_{\nu}R_{\lambda\mu\rho\sigma}=\nabla_{\nu}R_{\mu\sigma}$ directly, since here the contraction $g^{\lambda\rho}R_{\lambda\mu\rho\sigma}$ is already the Ricci tensor $R_{\mu\sigma}$ with no sign flip. Why: the index layout of the third term already matches the Ricci pattern (first and third slots contracted), so it contributes with a plus sign.

Step 5. Collect the three results: $\nabla^{\rho}R_{\mu\nu\rho\sigma}-\nabla_{\mu}R_{\nu\sigma}+\nabla_{\nu}R_{\mu\sigma}=0$ , i.e. $\nabla^{\rho}R_{\mu\nu\rho\sigma}=\nabla_{\mu}R_{\nu\sigma}-\nabla_{\nu}R_{\mu\sigma}$ . Why: this is already the once-contracted identity, merely with the divergence sitting on the third index of Riemann rather than the first.

Step 6. Use the pair-exchange symmetry $R_{\mu\nu\rho\sigma}=R_{\rho\sigma\mu\nu}$ to move the divergence onto the first slot and relabel the free indices, giving the canonical form $\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla_{\rho}R_{\nu\sigma}-\nabla_{\sigma}R_{\nu\rho}$ . Why: pair symmetry plus the symmetry of the Ricci tensor cosmetically rearranges the same content into the standard textbook layout — the physics of Step 5 is unchanged.

Go deeper · Derivation B · 2 minFrom the second contraction to a conserved Einstein tensor

Derivation BFrom the second contraction to a conserved Einstein tensor

The goal is to contract the once-contracted identity a second time to obtain $\nabla^{\mu}R_{\mu\nu}=\tfrac12\nabla_{\nu}R$ , with its characteristic factor of one half, and then to feed that result into the definition of the Einstein tensor and watch the divergence cancel identically. The delicate point is a single sign on the left-hand side of the second contraction — get it right and the factor of two appears cleanly; get it wrong (as the migrated draft did) and the equation degenerates into a tautology.

Step 1. Start from $\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla_{\rho}R_{\nu\sigma}-\nabla_{\sigma}R_{\nu\rho}$ and contract with $g^{\nu\rho}$ , pairing the first free index against the third. Why: this is the only remaining pair on the curvature tensor, so the second contraction is forced — there is nothing else to sum over.

Step 2. On the left, $g^{\nu\rho}\nabla^{\mu}R_{\mu\nu\rho\sigma}=\nabla^{\mu}\big(g^{\nu\rho}R_{\mu\nu\rho\sigma}\big)$ ; antisymmetry in the first pair gives $g^{\nu\rho}R_{\mu\nu\rho\sigma}=-g^{\nu\rho}R_{\nu\mu\rho\sigma}=-R_{\mu\sigma}$ , so the left side is $-\nabla^{\mu}R_{\mu\sigma}$ . Why: this minus sign is the crux — contracting the second-and-third slots of Riemann yields minus the Ricci tensor, and dropping it is what made the base article's Step 2 collapse to $0=-\nabla_\sigma R$ .

Step 3. On the right, the first term gives $g^{\nu\rho}\nabla_{\rho}R_{\nu\sigma}=\nabla^{\nu}R_{\nu\sigma}=\nabla^{\mu}R_{\mu\sigma}$ , and the second gives $g^{\nu\rho}\nabla_{\sigma}R_{\nu\rho}=\nabla_{\sigma}\big(g^{\nu\rho}R_{\nu\rho}\big)=\nabla_{\sigma}R$ . Why: the first is just a relabelled divergence of Ricci; the second uses that the metric trace of the Ricci tensor is the Ricci scalar, $g^{\nu\rho}R_{\nu\rho}=R$ .

Step 4. Assemble: $-\nabla^{\mu}R_{\mu\sigma}=\nabla^{\mu}R_{\mu\sigma}-\nabla_{\sigma}R$ , hence $2\nabla^{\mu}R_{\mu\sigma}=\nabla_{\sigma}R$ , i.e. $\nabla^{\mu}R_{\mu\nu}=\tfrac12\nabla_{\nu}R$ . Why: the factor of one half is not a convention — it is forced by the two independent contractions of Riemann both collapsing onto the same Ricci structure, and it is convention-independent across sign choices for the Riemann tensor.

Step 5. Take the divergence of the Einstein tensor: $\nabla^{\mu}G_{\mu\nu}=\nabla^{\mu}R_{\mu\nu}-\tfrac12\nabla^{\mu}(g_{\mu\nu}R)$ , and since $\nabla_\alpha g_{\mu\nu}=0$ the trace term is $\tfrac12 g_{\mu\nu}\nabla^{\mu}R=\tfrac12\nabla_{\nu}R$ . Why: linearity of the covariant derivative and metric compatibility isolate the one piece that can carry the scalar gradient.

Step 6. Substitute the twice-contracted identity: $\nabla^{\mu}G_{\mu\nu}=\tfrac12\nabla_{\nu}R-\tfrac12\nabla_{\nu}R=0$ . Why: the two halves are numerically identical for every metric, so the cancellation is an identity and not a dynamical statement — no field equation has been used anywhere in the chain.

Go deeper · Derivation C · 2 minWhy four identities mirror the gauge freedom of general relativity

Derivation CWhy four identities mirror the gauge freedom of general relativity

The goal is to show that the four identities $\nabla^{\mu}G_{\mu\nu}=0$ do more than guarantee conservation — they demote four of the ten Einstein equations from evolution equations to constraints on the initial data, exactly matching the four-fold freedom to relabel coordinates. We expand the covariant divergence into partial derivatives plus Christoffel terms, isolate the time derivative of the mixed components, and count derivative orders to expose which equations can and cannot propagate data forward in time.

Step 1. Write the identity with one index raised and split off the time direction, $\nabla_{\mu}G^{\mu}{}_{\nu}=\partial_{0}G^{0}{}_{\nu}+\partial_{i}G^{i}{}_{\nu}+(\Gamma\text{-terms})=0$ , so that $\partial_{0}G^{0}{}_{\nu}=-\partial_{i}G^{i}{}_{\nu}-(\Gamma\text{-terms})$ . Why: isolating the time derivative of the mixed time components is what reveals their derivative order, and hence whether they can serve as evolution equations.

Step 2. Note the generic order counting: any component of $G_{\mu\nu}$ contains up to two derivatives of the metric, so $G_{\mu\nu}\sim\partial^2 g$ . If $G^{0}{}_{\nu}$ contained a second time derivative $\partial_0^2 g$ , then $\partial_0 G^{0}{}_{\nu}$ would carry $\partial_0^3 g$ . Why: the highest time-derivative order is the diagnostic that separates a genuine evolution equation (second order in time) from a constraint (no second time derivative).

Step 3. But Step 1 equates $\partial_0 G^{0}{}_{\nu}$ to spatial derivatives of the other components, which contain at most $\partial_0^2 g$ — never $\partial_0^3 g$ . The only way this is consistent is if $G^{0}{}_{\nu}$ themselves contain no second time derivative, i.e. at most $\partial_0 g$ . Why: the identity algebraically forbids third time derivatives from appearing, which forces the four $G^0{}_\nu$ down to first order in time — they are constraints, fixed by the initial data on a slice.

Step 4. Count: of the ten equations $G_{\mu\nu}=8\pi G\,T_{\mu\nu}$ , the four $G^{0}{}_{\nu}=8\pi G\,T^{0}{}_{\nu}$ are the Hamiltonian and momentum constraints, leaving six dynamical equations; separately, the metric has ten components but four coordinate functions are freely choosable, leaving six metric fields to evolve — the counts $10-4=6$ agree on both sides. Those six are not six physical degrees of freedom: the four constraints bind them as well, and four-dimensional general relativity carries just two local propagating degrees of freedom, the two gravitational-wave polarizations. Why: the exact matching of six dynamical equations to six evolving metric components, with four constraints mirrored by four gauge functions, is the signature of a gauge theory of diffeomorphisms, and the contracted Bianchi identity is what enforces it. *The constraints are preserved under evolution precisely because $\nabla^\mu G_{\mu\nu}=0$ .*

Limiting cases

The contracted identities are robust, but their consequences shift when the geometry or the connection is specialized. The table below reads off what each structural assumption does to the chain.

Case	Result	Meaning
Einstein space, $R_{\mu\nu}=\tfrac{R}{n}g_{\mu\nu}$	$\nabla_\mu R=0$ for $n\ge 3$	The twice-contracted identity forces $(\tfrac1n-\tfrac12)\nabla_\nu R=0$ ; for dimension $n\ge3$ the coefficient is nonzero, so the scalar curvature is constant — Schur's lemma.
Vacuum with $\Lambda$ , $R_{\mu\nu}=\Lambda g_{\mu\nu}$	$\nabla^\mu(G_{\mu\nu}+\Lambda g_{\mu\nu})=0$ trivially	The $\Lambda g_{\mu\nu}$ term is covariantly constant, so it may be added to the field equations without spoiling conservation — the cosmological constant is "free."
Vacuum, $R_{\mu\nu}=0$	$\nabla^\mu C_{\mu\nu\rho\sigma}=0$	The once-contracted identity loses its source term; the Weyl tensor obeys a Maxwell-like propagation law, the basis of gravitational-wave analysis.
Connection with torsion	$\nabla^\mu G_{\mu\nu}\ne 0$ in general	In Einstein–Cartan theory the antisymmetrized derivatives pick up torsion terms; the clean conservation law is modified and must be replaced by a torsion-corrected identity.
Lovelock / higher curvature	$\nabla^\mu\big(G_{\mu\nu}+\alpha\,\mathcal{G}_{\mu\nu}\big)=0$	The Gauss–Bonnet (Lovelock) tensor $\mathcal{G}_{\mu\nu}$ is the unique higher-order term that is still identically divergence-free, via a generalized contracted Bianchi identity.

Common trapThe identity ∇^μ G μν = 0 already proves matter is conserved

It does not — and this is the single most common misreading. The statement $\nabla^{\mu}G_{\mu\nu}=0$ is a geometric identity, true for every metric whatsoever, including metrics that solve no field equation and describe no physical spacetime. It says nothing at all about matter. Local conservation of energy-momentum, $\nabla^{\mu}T_{\mu\nu}=0$ , follows only after you impose the Einstein field equation $G_{\mu\nu}=8\pi G\,T_{\mu\nu}$ and take its divergence: the left side vanishes by the identity, so the right side must too. Reverse the logic and you get a subtle but real error — believing that geometry alone constrains the source. Two further cautions ride along: metric compatibility, $\nabla_\alpha g_{\mu\nu}=0$ , is what licenses moving $g^{\mu\nu}$ through every derivative in the contractions, and a torsion-free connection is what makes the antisymmetrized Bianchi identity hold in the first place. Drop either assumption and the identity acquires extra source terms.

Worked examples

1 · Cosmological continuity from the contracted Bianchi identity (fully worked)

In a spatially flat FLRW universe with scale factor $a(t)$ and Hubble rate $H=\dot a/a$ , a perfect fluid has $T^{\mu}{}_{\nu}=\mathrm{diag}(-\rho,p,p,p)$ . Use the conservation law that the contracted Bianchi identity guarantees to find how the energy density evolves, then evaluate it for radiation, matter, and a cosmological constant.

Step 1. The field equations plus $\nabla^{\mu}G_{\mu\nu}=0$ force $\nabla_{\mu}T^{\mu}{}_{\nu}=0$ . Its time component is $\partial_{0}T^{0}{}_{0}+\Gamma^{\mu}{}_{\mu 0}T^{0}{}_{0}-\Gamma^{\lambda}{}_{\mu 0}T^{\mu}{}_{\lambda}=0$ . Using the FLRW Christoffels $\Gamma^{i}{}_{i0}=\dot a/a$ (three spatial values summing to $3H$ ) and $\Gamma^{i}{}_{j0}=H\delta^i_j$ , this reduces to the fluid continuity equation $\dot\rho+3H(\rho+p)=0$ .

Step 2. Introduce the equation of state $p=w\rho$ with $w$ constant. Then $\dot\rho/\rho=-3(1+w)\,\dot a/a$ , which integrates to $\rho\propto a^{-3(1+w)}$ . This single formula is the entire dilution history of any barotropic component, and it is a direct corollary of the contracted Bianchi identity.

Step 3. Evaluate the exponent. Matter (pressureless dust, $w=0$ ): $\rho\propto a^{-3}$ . Radiation ( $w=\tfrac13$ ): $\rho\propto a^{-4}$ . Cosmological constant ( $w=-1$ ): $\rho\propto a^{0}=\text{const}$ , consistent with $\Lambda g_{\mu\nu}$ being covariantly constant.

Step 4. Put in numbers. If the universe expands by a factor of two, $a\to 2a$ , then matter density falls by $2^{3}=\mathbf{8}$ , radiation density by $2^{4}=\mathbf{16}$ (the extra factor is the redshift of each photon's energy), and the vacuum density is unchanged. This is why radiation, once dominant, is diluted away fastest and the universe passes from radiation domination to matter domination to $\Lambda$ domination — an ordering dictated entirely by the conservation law that geometry hands us for free.

2 · Schur's lemma: an Einstein space has constant curvature (semi-numeric)

Suppose a spacetime satisfies the Einstein-space condition $R_{\mu\nu}=\lambda(x)\,g_{\mu\nu}$ for some scalar field $\lambda(x)$ . Show that in four dimensions $\lambda$ — and hence $R$ — must actually be constant.

Step 1. Take the metric trace: $R=g^{\mu\nu}R_{\mu\nu}=\lambda\,g^{\mu\nu}g_{\mu\nu}=n\lambda$ , so $\lambda=R/n$ . In $n=4$ dimensions, $\lambda=R/4$ .

Step 2. Take the divergence of the ansatz: $\nabla^{\mu}R_{\mu\nu}=\nabla^{\mu}(\lambda g_{\mu\nu})=g_{\mu\nu}\nabla^{\mu}\lambda=\nabla_{\nu}\lambda=\tfrac1n\nabla_{\nu}R$ . The metric slides out because $\nabla_\alpha g_{\mu\nu}=0$ .

Step 3. But the twice-contracted Bianchi identity independently demands $\nabla^{\mu}R_{\mu\nu}=\tfrac12\nabla_{\nu}R$ . Equate the two: $\big(\tfrac1n-\tfrac12\big)\nabla_{\nu}R=0$ . In $n=4$ the coefficient is $\tfrac14-\tfrac12=\mathbf{-\tfrac14}\ne 0$ .

Step 4. A nonzero coefficient forces $\nabla_{\nu}R=0$ , so $R$ is constant and therefore $\lambda=R/4$ is constant. The conclusion fails only in $n=2$ , where the coefficient $\tfrac12-\tfrac12=0$ vanishes identically — which is exactly why two-dimensional gravity is degenerate. The result, Schur's lemma, is a one-line consequence of the twice-contracted identity and underpins the classification of maximally symmetric spaces (de Sitter, anti-de Sitter, Minkowski).

Physics of the portrait

The contracted Bianchi identities sit at the confluence of two nineteenth-century programs — the absolute differential calculus of Ricci and Levi-Civita, and the differential geometry of curvature that Riemann founded — and a twentieth-century physics revolution that needed exactly this identity to exist. The person whose name the parent identity carries mapped the terrain a decade before anyone knew physics would require it.

Origins · 1 minLuigi Bianchi and how this was discovered

Luigi Bianchi

1856–1928 · differential geometry and the identities of the curvature tensor

Luigi Bianchi was an Italian geometer, a student of Enrico Betti and Ulisse Dini at the Scuola Normale Superiore in Pisa, where he later held the chair of analytic geometry for most of his career. He worked at the heart of the Italian school of differential geometry alongside Gregorio Ricci-Curbastro and Tullio Levi-Civita, whose "absolute differential calculus" — what we now call tensor analysis — provided the language in which the curvature identities are naturally expressed. In his lectures on differential geometry, published in 1902, Bianchi stated the differential identities of the Riemann tensor that now bear his name, $\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0$ , though the same relations had been written earlier by Aurel Voss in 1880 and were known in some form to Ricci; the attribution is one of history's characteristic imperfections. Bianchi could not have anticipated that his identity would become the structural keystone of a physical theory of gravitation. When Einstein and Hilbert, working largely independently in 1915, arrived at the field equations, they needed precisely a geometric tensor whose divergence vanishes identically so that it could be set equal to the conserved stress-energy of matter. The twice-contracted form of Bianchi's identity is exactly that guarantee. Einstein initially derived the contracted relation by hand, unaware of its pedigree in the geometry literature; only later was it recognized as a corollary of Bianchi's 1902 result. The identity thus links a purely mathematical statement about how curvature varies to the deepest consistency requirement of general relativity — that spacetime geometry and its matter source speak the same conservation law.

History

1880

— Voss writes the differential identities of the curvature tensor

1902

— Bianchi publishes them in his lectures, now the second Bianchi identity

1915

— Einstein and Hilbert reach the field equations Gμν = 8πTμν

1917

— Einstein adds Λgμν, still divergence-free by the identity

1922

— Friedmann's cosmology; continuity follows from ∇μTμν = 0

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §15.3 and Box 15.2 · Wald, General Relativity, §3.2, eqs. (3.2.14)–(3.2.16) and §10.2 · Carroll, Spacetime and Geometry, §3.7, eqs. (3.94)–(3.98) · Schutz, A First Course in General Relativity, §6.5 · Hartle, Gravity, curvature-identity appendix. The key checkpoints in the derivation chain are: (1) the once-contracted identity $\nabla^\mu R_{\mu\nu\rho\sigma}=\nabla_\rho R_{\nu\sigma}-\nabla_\sigma R_{\nu\rho}$ , obtained by contracting the antisymmetrized identity with $g^{\lambda\rho}$ and using pair symmetry; (2) the sign on the left of the second contraction, $g^{\nu\rho}R_{\mu\nu\rho\sigma}=-R_{\mu\sigma}$ , which yields $2\nabla^\mu R_{\mu\nu}=\nabla_\nu R$ ; (3) the algebraic cancellation $\nabla^\mu G_{\mu\nu}=0$ ; (4) the continuity result $\rho\propto a^{-3(1+w)}$ giving $a^{-3}$ , $a^{-4}$ , and const for $w=0,\tfrac13,-1$ ; (5) the Schur coefficient $\tfrac14-\tfrac12=-\tfrac14$ in four dimensions. No numerical constants beyond the coupling $8\pi G$ enter.

Prerequisites. Second (Differential) Bianchi Identity · Riemann Curvature Tensor · Algebraic Symmetries of the Riemann Tensor · Ricci Tensor · Ricci Scalar. Leads to. Einstein Tensor · Ricci Decomposition of the Riemann Tensor · Weyl Conformal Tensor.

Ask the tutor: Ask me to walk through the index gymnastics of either contraction line by line, or to show how the four constraints reappear in the ADM 3+1 split.

Why exactly two contractions — why not stop at one or continue to three?

The Riemann tensor has four indices, but its symmetries mean only two independent metric contractions produce something new. The first contraction reduces Riemann to the Ricci tensor (two indices); the second reduces Ricci to the Ricci scalar (zero free curvature indices). At that point there is nothing left on the curvature to contract — any further metric factor would just multiply the scalar. Stopping at one contraction gives the useful but non-conserved relation $\nabla^\mu R_{\mu\nu\rho\sigma}=\nabla_\rho R_{\nu\sigma}-\nabla_\sigma R_{\nu\rho}$ ; only the second contraction produces the divergence-of-Einstein statement. So "two" is not a choice — it is where the index structure of Riemann naturally terminates the chain.

What role does metric compatibility actually play in the derivation?

It is what lets you move the inverse metric $g^{\mu\nu}$ freely in and out of the covariant derivative. Every contraction step writes something like $g^{\lambda\rho}\nabla_\lambda(\cdots)=\nabla^\rho(\cdots)$ , and that manipulation is legal only because $\nabla_\alpha g_{\mu\nu}=0$ . If the connection were merely torsion-free but not metric-compatible (a connection with non-metricity), the metric would carry its own derivative through each step and the clean identities would acquire extra terms proportional to $\nabla g$ . Metric compatibility is one of the two defining properties of the Levi-Civita connection, and it is silently used in essentially every line of the contraction — it is why the whole calculation is "just index gymnastics" rather than a differential slog.

How is this the same as the Bianchi identity of electromagnetism?

In electromagnetism the homogeneous Maxwell equations are the statement $\partial_{[\alpha}F_{\beta\gamma]}=0$ , which follows because $F=dA$ and $d^2=0$ — "the boundary of a boundary is zero." That identity automatically implies $\partial_\mu{}^{*}F^{\mu\nu}=0$ , a conservation law for the dual field with no source. The contracted Bianchi identities are the gravitational analogue: $\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0$ is the geometric " $d^2=0$ " for curvature, and contracting it yields $\nabla^\mu G_{\mu\nu}=0$ , a source-free conservation law built into the geometry. In both theories the identity is what makes the inhomogeneous (source) equation consistent with a conserved current — $\partial_\mu F^{\mu\nu}=J^\nu$ needs $\partial_\nu J^\nu=0$ , and $G_{\mu\nu}=8\pi G\,T_{\mu\nu}$ needs $\nabla^\mu T_{\mu\nu}=0$ .

Does the cosmological constant spoil the conservation of the Einstein tensor?

No — and the reason is exactly why $\Lambda$ was allowed in the first place. The Einstein tensor is divergence-free by the twice-contracted identity, and the extra term $\Lambda g_{\mu\nu}$ is also divergence-free because the metric is covariantly constant: $\nabla^\mu(\Lambda g_{\mu\nu})=\Lambda\,\nabla^\mu g_{\mu\nu}=0$ . So $G_{\mu\nu}+\Lambda g_{\mu\nu}$ is conserved for any constant $\Lambda$ , and one can legally set it equal to $8\pi G\,T_{\mu\nu}$ . This is precisely the loophole Einstein exploited in 1917 to build a static universe. It also means the cosmological constant is indistinguishable, at the level of the conservation law, from a vacuum energy $T^{\rm vac}_{\mu\nu}=-\rho_{\rm vac}g_{\mu\nu}$ with $w=-1$ — which is why dark energy and $\Lambda$ are so hard to tell apart.

Why does a single sign error break the whole twice-contracted derivation?

Because the second contraction produces the same object — the divergence of the Ricci tensor — on both sides of the equation, and only the relative sign decides whether they add or cancel. The left side is $g^{\nu\rho}\nabla^\mu R_{\mu\nu\rho\sigma}=-\nabla^\mu R_{\mu\sigma}$ ; the first right-hand term is $+\nabla^\mu R_{\mu\sigma}$ . With the correct minus on the left you get $-\nabla^\mu R_{\mu\sigma}=\nabla^\mu R_{\mu\sigma}-\nabla_\sigma R$ , hence $2\nabla^\mu R_{\mu\sigma}=\nabla_\sigma R$ . Drop that minus and you get the tautology $\nabla^\mu R_{\mu\sigma}=\nabla^\mu R_{\mu\sigma}-\nabla_\sigma R$ , i.e. the false statement $\nabla_\sigma R=0$ for every metric. The sign comes from the antisymmetry of Riemann in its first index pair, $R_{\mu\nu\rho\sigma}=-R_{\nu\mu\rho\sigma}$ , and getting it right is the entire content of the calculation.

Is the divergence-free property special to the Einstein tensor, or do other tensors share it?

In four dimensions, and among tensors built algebraically from the metric, its first, and its second derivatives, Lovelock's theorem says the only symmetric divergence-free candidates are $G_{\mu\nu}$ and $g_{\mu\nu}$ — so the general conserved combination is $aG_{\mu\nu}+bg_{\mu\nu}$ , which is exactly $G_{\mu\nu}+\Lambda g_{\mu\nu}$ up to scaling. This near-uniqueness is why general relativity is so rigid. If you allow higher derivatives or more dimensions, additional divergence-free tensors appear — most famously the Gauss–Bonnet (Lovelock) tensor $\mathcal{G}_{\mu\nu}$ , which is identically conserved in dimension five and above via an extended contracted Bianchi identity. So the property is not unique in the abstract, but under the physical constraints of four dimensions and second-order equations it very nearly is — and that is a large part of why Einstein's equations look inevitable.