Second (Differential) Bianchi Identity

Curvature is what is left over when you carry a vector around a closed loop and find it rotated on return — the Riemann tensor packages that rotation-per-unit-area at a single point. The algebraic symmetries, including the first Bianchi identity, are constraints on those numbers at one point. But curvature also varies from point to point, and that variation is not free: it must be laid down so consistently that curvature could be the curvature of some connection in the first place. The second, or differential, Bianchi identity is exactly that consistency demand, written as a differential constraint on the covariant derivatives of Riemann. It is not a property of any particular spacetime and carries no adjustable content — it holds identically for the curvature of any torsion-free, metric-compatible connection, forced by the definition of curvature itself, in the same way that a gradient field is automatically curl-free. The compact statement is ∇[λRμν]ρσ=0\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0, the totally antisymmetrized covariant derivative of the Riemann tensor. Its power is unlocked by contraction: trace it twice and it collapses to ∇μGμν=0\nabla^\mu G_{\mu\nu}=0, the automatically vanishing divergence of the Einstein tensor Gμν=Rμν−12R gμνG_{\mu\nu}=R_{\mu\nu}-\tfrac12 R\,g_{\mu\nu} — the hinge of general relativity. Einstein wanted field equations whose geometric side was divergence-free so that local energy–momentum conservation ∇μTμν=0\nabla^\mu T_{\mu\nu}=0 would follow as an identity rather than an extra postulate, and the twice-contracted Bianchi identity is exactly what makes GμνG_{\mu\nu}, not RμνR_{\mu\nu} alone, the right object. In the language of curvature two-forms the same statement is DR=0\mathrm{D}\mathcal{R}=0, dual to the source-free Maxwell law dF=0\mathrm{d}F=0 and the Yang–Mills identity DF=0\mathrm{D}F=0 — which is why Wheeler read it as the geometric echo of a topological truism: the boundary of a boundary is zero.

Key takeaways
  • The second (differential) Bianchi identity is ∇[λRμν]ρσ=0\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0 — the totally antisymmetrized covariant derivative of the Riemann tensor vanishes, cycling the derivative index with the first pair of curvature indices.
  • It is an identity, not a field equation: it holds in every metric-compatible, torsion-free geometry as the Jacobi identity for covariant derivatives, so it constrains curvature but adds no physics of its own.
  • Contract it once to get ∇μRμνρσ=∇ρRνσ−∇σRνρ\nabla^\mu R_{\mu\nu\rho\sigma}=\nabla_\rho R_{\nu\sigma}-\nabla_\sigma R_{\nu\rho}, and again to get ∇μRμν=12∇νR\nabla^\mu R_{\mu\nu}=\tfrac12\nabla_\nu R, which assembles into ∇μGμν=0\nabla^\mu G_{\mu\nu}=0.
  • That vanishing divergence is why Einstein's equations are consistent with ∇μTμν=0\nabla^\mu T_{\mu\nu}=0, and why four of the ten field equations are constraints — matching the four-fold diffeomorphism freedom.
  • In differential-forms language it is DR=dR+ω∧R−R∧ω=0\mathrm{D}\mathcal{R}=\mathrm{d}\mathcal{R}+\omega\wedge\mathcal{R}-\mathcal{R}\wedge\omega=0, the curvature analogue of the source-free half of Maxwell's equations and of the gauge-theory Bianchi identity DF=0\mathrm{D}F=0.
  • Key idea: the differential Bianchi identity ∇[λRμν]ρσ=0\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0 is the Jacobi identity for covariant derivatives written in components; contracting it twice forces ∇μGμν=0\nabla^\mu G_{\mu\nu}=0, which is the geometric reason Einstein's equations are compatible with local energy–momentum conservation.

Where this lives today

The Bianchi identity is load-bearing in the computations that produced modern gravitational-wave astronomy. In numerical relativity — which simulates colliding black holes to generate the waveform templates LIGO and Virgo match against their data — the ten Einstein equations split into six evolution equations and four constraints (the Hamiltonian and momentum constraints) that must hold on every spatial slice. The twice-contracted Bianchi identity is precisely the statement that if the constraints hold on an initial slice, the evolution preserves them: constraint violation propagates as a wave that, analytically, stays zero. Real simulations run in finite precision, so tiny violations creep in and can blow up; entire formulations (BSSN, generalized harmonic gauge) are engineered so that the Bianchi-guaranteed constraint propagation stays controlled rather than amplified — BSSN by making the evolution system strongly hyperbolic, while generalized harmonic gauge and the Z4c/CCZ4 variants add explicit constraint-damping terms. Without that structural identity there would be no stable long-term binary-black-hole evolutions and no template banks — and GW150914 in 2015 would have had nothing to be compared against. The same identity, in vacuum form ∇μCμνρσ=0\nabla^\mu C_{\mu\nu\rho\sigma}=0, lets the emitted curvature be decomposed into Weyl scalars and read off as the two polarizations of a gravitational wave at infinity.

The physical picture

Parallel-transport a vector around a small closed loop and it comes back rotated; the rotation, per unit enclosed area, is the curvature threading the loop. Now build a small three-dimensional cell — think of a tiny cube — out of six such faces. Transport the vector around the full boundary of that cube by traversing each face loop in turn. Because the cube is a closed solid, every internal edge is shared by exactly two faces and is walked once in each direction, so those contributions cancel in pairs. The total rotation around the boundary of the boundary must therefore be zero. Take that cancellation to the infinitesimal limit and it becomes a statement about how curvature changes from face to face — its covariant derivative — summed cyclically over the three edge-directions of the cell. That statement is the second Bianchi identity. Nothing about a specific spacetime enters; the identity is forced by the mere fact that curvature is built from a connection.

Figure
The boundary of a boundary is zero: transport around the six faces of a small 3-cell, every interior edge is walked twice in opposite senses and cancels, so the summed curvature holonomy vanishes — the infinitesimal limit is the antisymmetrized covariant derivative of Riemann equalling zero
∂∂ = 0 : the curvature holonomy summed over the faces of a cell cancels face loop shared edge: walked twice, opposite senses → cancels Σ (holonomy over the 6 faces) = 0 infinitesimal limit: ∇[λ Rμν]ρσ = 0

Read the same fact through fields and it becomes a kinship with electromagnetism. Curvature is the field strength of the connection, exactly as FμνF_{\mu\nu} is the field strength of the electromagnetic potential. The differential Bianchi identity says this field strength has no "source" in its antisymmetrized derivative — the gravitational analogue of ∇⋅B=0\nabla\cdot\mathbf{B}=0 and ∂[μFνρ]=0\partial_{[\mu}F_{\nu\rho]}=0, the half of Maxwell's equations that follows from F=dAF=\mathrm{d}A with no charges anywhere. And just as those source-free equations let the other half be sourced consistently by a conserved current, the Bianchi identity lets Einstein's equations be sourced consistently by a conserved stress-energy tensor. The whole payoff is one chain of contractions, shown below.

Figure
The contraction cascade: two traces of the differential Bianchi identity turn geometry into a conservation law — it collapses through the once- and twice-contracted forms to ∇^μ Gμν = 0, which makes ∇^μ Tμν = 0 automatic once Gμν = κ Tμν
Two contractions: from the geometry of curvature to a conservation law ∇[λ Rμν]ρσ = 0 differential (second) Bianchi identity contract λ with ρ ∇^μ Rμνρσ = ∇ρ Rνσ − ∇σ Rνρ once-contracted Bianchi identity contract ν with σ ∇^μ Rμν = ½ ∇ν R twice-contracted Bianchi identity subtract ½ ∇(g R) ∇^μ Gμν = 0 Einstein tensor is divergence-free — identically impose Gμν = κ Tμν ∇^μ Tμν = 0 local energy–momentum conservation, for free

The identity is best thought of as an integrability condition: it is exactly what is required for a candidate curvature to be the curvature of some connection, and it keeps the second-order field equations from over-determining the ten metric components. The four relations packaged in ∇μGμν=0\nabla^\mu G_{\mu\nu}=0 remove four would-be dynamical equations, leaving six — precisely matched to the four-function freedom of choosing coordinates. Without it there would be no consistent theory of gravity of Einstein's form at all.

The core equations

The identity has one content and several faces. Its most economical form packs the whole cyclic structure into an antisymmetrization bracket: the covariant derivative of Riemann, totally antisymmetrized over the derivative index and the first curvature pair, vanishes. The square bracket denotes the normalized sum over all orderings with signs, and the antisymmetry of Riemann in that first pair is what lets three indices, not four, carry the cycle.

∇[λRμν]ρσ=0(1)\nabla_{[\lambda} R_{\mu\nu]\rho\sigma}=0 \qquad (1)

where: RμνρσR_{\mu\nu\rho\sigma} — Riemann curvature tensor, fully covariant; ∇λ\nabla_\lambda — Levi-Civita (metric-compatible, torsion-free) covariant derivative; [ λμν ][\,\lambda\mu\nu\,] — normalized total antisymmetrization over the three enclosed indices, the two curvature indices μ,ν\mu,\nu plus the derivative index λ\lambda. Physical reading: the rate of change of curvature, wrung of all its symmetric part across three directions, has nothing left over. Because Riemann is antisymmetric in its first pair, the 3!=63!=6 signed terms in the bracket collapse into three, and the identity wears its more recognizable cyclic form. This is the shape that makes the "Jacobi identity" fingerprint visible: the derivative index and the first pair of curvature indices march around a cycle.

∇λRμνρσ+∇μRνλρσ+∇νRλμρσ=0(2)\nabla_\lambda R_{\mu\nu\rho\sigma}+\nabla_\mu R_{\nu\lambda\rho\sigma}+\nabla_\nu R_{\lambda\mu\rho\sigma}=0 \qquad (2)

where: the three terms are the cyclic permutations of λ,μ,ν\lambda,\mu,\nu with the second pair ρσ\rho\sigma held fixed. Physical reading: sum the covariant gradient of curvature around a closed cycle of three directions and it cancels, exactly as the holonomies around the faces of a cell cancel on the shared edges. The same identity, expressed in Cartan's language of curvature two-forms, is a single clean line: the exterior covariant derivative of the curvature form vanishes. This form makes the parallel with gauge theory exact and is the version that connects the identity to characteristic classes and topology.

DR≡dR+ω∧R−R∧ω=0(3)\mathrm{D}\mathcal{R}\equiv \mathrm{d}\mathcal{R}+\omega\wedge\mathcal{R}-\mathcal{R}\wedge\omega=0 \qquad (3)

where: Rab=12Rabμν dxμ∧dxν\mathcal{R}^a{}_b=\tfrac12 R^a{}_{b\mu\nu}\,\mathrm{d}x^\mu\wedge\mathrm{d}x^\nu — curvature two-form; ωab\omega^a{}_b — connection one-form; D\mathrm{D} — exterior covariant derivative. Physical reading: curvature is a "closed" object under the covariant exterior derivative, the direct analogue of dF=0\mathrm{d}F=0 for the abelian field strength and DF=0\mathrm{D}F=0 in Yang–Mills theory. The physics begins at the first contraction. Tracing the identity over one index of the derivative-cycle with one of the fixed pair yields a relation between the divergence of Riemann and the antisymmetrized gradient of the Ricci tensor. It already looks Maxwell-like, and in vacuum it becomes precisely a wave equation for curvature.

∇μRμνρσ=∇ρRνσ−∇σRνρ(4)\nabla^\mu R_{\mu\nu\rho\sigma}=\nabla_\rho R_{\nu\sigma}-\nabla_\sigma R_{\nu\rho} \qquad (4)

where: Rνσ=RμνμσR_{\nu\sigma}=R^\mu{}_{\nu\mu\sigma} — Ricci tensor; ∇μ=gμλ∇λ\nabla^\mu=g^{\mu\lambda}\nabla_\lambda — the divergence acting on the first index. Physical reading: the divergence of the full curvature is not free but is fixed by how the Ricci tensor varies — the "source" for the divergence of Riemann is the gradient of its own trace. A second trace closes the loop. Contracting equation (4) over the remaining free pair produces a relation between the divergence of the Ricci tensor and the gradient of the Ricci scalar — the twice-contracted Bianchi identity, and the linchpin of everything that follows.

∇μRμν=12 ∇νR(5)\nabla^\mu R_{\mu\nu}=\tfrac12\,\nabla_\nu R \qquad (5)

where: R=gνσRνσR=g^{\nu\sigma}R_{\nu\sigma} — Ricci scalar. Physical reading: the Ricci tensor is not divergence-free — its divergence equals half the gradient of the scalar curvature — which is exactly why Rμν=κTμνR_{\mu\nu}=\kappa T_{\mu\nu} cannot be the field equation, and why a corrected object is needed. That corrected object is the Einstein tensor, and the factor 12\tfrac12 in (5) is precisely what its trace-reversed combination is built to cancel. The result is the identity that gives general relativity its consistency.

∇μGμν=0,Gμν=Rμν−12R gμν(6)\nabla^\mu G_{\mu\nu}=0,\qquad G_{\mu\nu}=R_{\mu\nu}-\tfrac12 R\,g_{\mu\nu} \qquad (6)

where: GμνG_{\mu\nu} — Einstein tensor, the unique symmetric, divergence-free tensor built from the metric and its first two derivatives that is linear in the second derivatives. Physical reading: the geometric side of the field equations conserves itself automatically, so imposing Gμν=κTμνG_{\mu\nu}=\kappa T_{\mu\nu} makes ∇μTμν=0\nabla^\mu T_{\mu\nu}=0 a corollary rather than an extra law. Finally, feeding the Ricci decomposition of Riemann back into the once-contracted identity isolates the divergence of the Weyl (conformal) tensor, with the Ricci and scalar pieces playing the role of sources. In four dimensions the coefficients are fixed.

∇μCμνρσ=∇[ρ ⁣(Rσ]ν−16gσ]νR)(7)\nabla^\mu C_{\mu\nu\rho\sigma}=\nabla_{[\rho}\!\left(R_{\sigma]\nu}-\tfrac{1}{6}g_{\sigma]\nu}R\right) \qquad (7)

where: CμνρσC_{\mu\nu\rho\sigma} — Weyl conformal tensor, the totally trace-free part of Riemann. Physical reading: the divergence of the free, tidal part of curvature is sourced by the local matter (through Ricci); in vacuum the right-hand side vanishes and ∇μCμνρσ=0\nabla^\mu C_{\mu\nu\rho\sigma}=0 governs how tidal curvature — gravitational radiation — propagates.

Derivation AProof from normal coordinates and the commutation of partials

The cleanest proof shows the identity is not a deep geometric fact but the commuting of mixed partial derivatives dressed in tensor language. The strategy: evaluate everything at a single arbitrary point in Riemann normal coordinates, where the Christoffel symbols vanish (though their derivatives do not, so curvature survives), reduce covariant derivatives of Riemann to plain second partials of the connection, antisymmetrize, and watch the symmetric pair of partials annihilate the whole sum. Because both sides are tensors and the point was arbitrary, a result true there in one frame is true everywhere in every frame.

Step 1. At a point pp choose Riemann normal coordinates, so Γρμν∣p=0\Gamma^\rho{}_{\mu\nu}\big|_p=0 while ∂αΓρμν∣p≠0\partial_\alpha\Gamma^\rho{}_{\mu\nu}\big|_p\neq0. There the Riemann tensor loses its quadratic ΓΓ\Gamma\Gamma terms and reduces to Rρσμν=∂μΓρνσ−∂νΓρμσR^\rho{}_{\sigma\mu\nu}=\partial_\mu\Gamma^\rho{}_{\nu\sigma}-\partial_\nu\Gamma^\rho{}_{\mu\sigma}. Why: killing the connection at pp strips curvature down to a pure derivative of the connection, the only part that can survive one more differentiation.

Step 2. Differentiate once covariantly. Since Γ=0\Gamma=0 at pp, the covariant derivative there is just the partial, so ∇λRρσμν∣p=∂λ∂μΓρνσ−∂λ∂νΓρμσ\nabla_\lambda R^\rho{}_{\sigma\mu\nu}\big|_p=\partial_\lambda\partial_\mu\Gamma^\rho{}_{\nu\sigma}-\partial_\lambda\partial_\nu\Gamma^\rho{}_{\mu\sigma}. Why: at the origin of normal coordinates the distinction between ∇\nabla and ∂\partial collapses for this one evaluation, converting a covariant statement into an ordinary calculus statement.

Step 3. Antisymmetrize over the three lower indices λ,μ,ν\lambda,\mu,\nu, holding σ\sigma and the upper ρ\rho fixed. Each surviving term has the schematic form ∂[λ∂μΓρν]σ\partial_{[\lambda}\partial_{\mu}\Gamma^\rho{}_{\nu]\sigma}: a second partial derivative, symmetric in its two derivative slots, sitting inside a total antisymmetrization of those very slots. Why: the antisymmetrization is precisely over the indices that the double partial treats symmetrically, setting up an automatic cancellation.

Step 4. A symmetric object antisymmetrized vanishes: ∂[λ∂μΓρν]σ=0\partial_{[\lambda}\partial_{\mu}\Gamma^\rho{}_{\nu]\sigma}=0 because ∂λ∂μ=∂μ∂λ\partial_\lambda\partial_\mu=\partial_\mu\partial_\lambda. Both terms of Step 2 die, so ∇[λRρ∣σ∣μν]∣p=0\nabla_{[\lambda}R^\rho{}_{|\sigma|\mu\nu]}\big|_p=0. Why: this is the whole content — the commuting of partial derivatives is the Jacobi identity for the connection, and it is what the Bianchi identity secretly is.

Step 5. The left-hand side is a tensor and the equation holds at the arbitrary point pp; a tensor equation valid at every point in one frame is valid in all frames. Lower ρ\rho with the metric (permissible since ∇g=0\nabla g=0) to reach the boxed identity ∇λRμνρσ+∇μRνλρσ+∇νRλμρσ=0\nabla_\lambda R_{\mu\nu\rho\sigma}+\nabla_\mu R_{\nu\lambda\rho\sigma}+\nabla_\nu R_{\lambda\mu\rho\sigma}=0. Why: normal coordinates are a scaffold, not an assumption — tensoriality promotes the local, single-frame result to a universal law.

Go deeper · Derivation B · 1 minThe differential-forms route via Cartan's second structure equation
Derivation BThe differential-forms route via Cartan's second structure equation

This derivation reaches the identity in three lines by exterior calculus and makes transparent that Bianchi is a corollary of d2=0\mathrm{d}^2=0 — the same nilpotency of the exterior derivative that underlies "the boundary of a boundary is zero." The goal: start from Cartan's second structure equation, which defines the curvature two-form from the connection one-form, hit it with an exterior derivative, and let nilpotency and the anticommuting of one-forms do the work, arriving at DR=0\mathrm{D}\mathcal{R}=0 with the connection pieces assembling into an exterior covariant derivative.

Step 1. Begin from the second structure equation R=dω+ω∧ω\mathcal{R}=\mathrm{d}\omega+\omega\wedge\omega and apply d\mathrm{d} to both sides: dR=d(dω)+d(ω∧ω)\mathrm{d}\mathcal{R}=\mathrm{d}(\mathrm{d}\omega)+\mathrm{d}(\omega\wedge\omega). Why: differentiating the definition of curvature is the only move available, and it is guaranteed to expose structure because d\mathrm{d} is nilpotent.

Step 2. Use d(dω)=0\mathrm{d}(\mathrm{d}\omega)=0 and the graded Leibniz rule on the product of one-forms, d(ω∧ω)=dω∧ω−ω∧dω\mathrm{d}(\omega\wedge\omega)=\mathrm{d}\omega\wedge\omega-\omega\wedge\mathrm{d}\omega, to get dR=dω∧ω−ω∧dω\mathrm{d}\mathcal{R}=\mathrm{d}\omega\wedge\omega-\omega\wedge\mathrm{d}\omega. Why: the first term dies by nilpotency; the sign in the second comes from moving d\mathrm{d} past a one-form, which anticommutes.

Step 3. Eliminate dω\mathrm{d}\omega using the structure equation, dω=R−ω∧ω\mathrm{d}\omega=\mathcal{R}-\omega\wedge\omega: substituting gives dR=(R−ω∧ω)∧ω−ω∧(R−ω∧ω)\mathrm{d}\mathcal{R}=(\mathcal{R}-\omega\wedge\omega)\wedge\omega-\omega\wedge(\mathcal{R}-\omega\wedge\omega). Why: re-expressing dω\mathrm{d}\omega back in terms of curvature is what will let the pure-connection cubic terms cancel and the curvature terms survive.

Step 4. The two ω∧ω∧ω\omega\wedge\omega\wedge\omega terms are identical and cancel, leaving dR=R∧ω−ω∧R\mathrm{d}\mathcal{R}=\mathcal{R}\wedge\omega-\omega\wedge\mathcal{R}. Rearranged, dR+ω∧R−R∧ω=0\mathrm{d}\mathcal{R}+\omega\wedge\mathcal{R}-\mathcal{R}\wedge\omega=0, which is exactly DR=0\mathrm{D}\mathcal{R}=0. Why: the leftover terms are precisely the connection corrections that upgrade the ordinary exterior derivative d\mathrm{d} to the covariant one D\mathrm{D}, so the vanishing is the covariant statement — valid in any frame, including a non-holonomic one.

Go deeper · Derivation C · 2 minTwo contractions to the conservation law ∇^μ Gμν = 0
Derivation CTwo contractions to the conservation law ∇^μ Gμν = 0

The single most consequential thing one does with the Bianchi identity is trace it twice, turning a fact about the geometry of curvature into the automatic conservation of the Einstein tensor. The goal: carry out both contractions explicitly, careful with the antisymmetries, so that ∇μGμν=0\nabla^\mu G_{\mu\nu}=0 emerges with its factor of 12\tfrac12 transparently traced to the metric contraction — and to show that the cancellation is no accident of four dimensions but holds in every dimension above two.

Step 1. Write the cyclic identity with the second pair up front, ∇λRρσμν+∇μRρσνλ+∇νRρσλμ=0\nabla_\lambda R_{\rho\sigma\mu\nu}+\nabla_\mu R_{\rho\sigma\nu\lambda}+\nabla_\nu R_{\rho\sigma\lambda\mu}=0, and contract with gρμg^{\rho\mu}, pulling the metric through the derivatives since ∇g=0\nabla g=0. Why: metric compatibility is what lets a contraction commute with ∇\nabla, turning traces of the identity into identities among the traced tensors.

Step 2. Evaluate the three traced terms using Rμσμν=RσνR^\mu{}_{\sigma\mu\nu}=R_{\sigma\nu} and the antisymmetry Rρσνλ=−RρσλνR_{\rho\sigma\nu\lambda}=-R_{\rho\sigma\lambda\nu}. They collapse to the once-contracted identity ∇μRμσνλ=∇νRσλ−∇λRσν\nabla^\mu R_{\mu\sigma\nu\lambda}=\nabla_\nu R_{\sigma\lambda}-\nabla_\lambda R_{\sigma\nu}, equivalently equation (4). Why: each trace either lands on a Ricci tensor or, on the middle term, on the divergence of Riemann, and the antisymmetries fix all the signs.

Step 3. Contract this once-contracted identity a second time with gσνg^{\sigma\nu}. Using gσνRσν=Rg^{\sigma\nu}R_{\sigma\nu}=R and gσν∇μRμσνλ=∇μRμλg^{\sigma\nu}\nabla^\mu R_{\mu\sigma\nu\lambda}=\nabla^\mu R_{\mu\lambda} (after one antisymmetry), the three terms give ∇λR−2 ∇μRμλ=0\nabla_\lambda R-2\,\nabla^\mu R_{\mu\lambda}=0. Why: the middle term is traced onto a Ricci divergence and appears with the same structure as one of the outer terms, which is what produces the factor of two.

Step 4. Solve for the Ricci divergence: ∇μRμλ=12∇λR\nabla^\mu R_{\mu\lambda}=\tfrac12\nabla_\lambda R, the twice-contracted Bianchi identity, equation (5). Why: this exposes the obstruction — the Ricci tensor alone is not conserved, its divergence being half the gradient of the scalar curvature.

Step 5. Take the divergence of the trace-reversed combination: ∇μ ⁣(Rμλ−12gμλR)=12∇λR−12∇λR=0\nabla^\mu\!\left(R_{\mu\lambda}-\tfrac12 g_{\mu\lambda}R\right)=\tfrac12\nabla_\lambda R-\tfrac12\nabla_\lambda R=0. The 12gμλR\tfrac12 g_{\mu\lambda}R term contributes exactly 12∇λR\tfrac12\nabla_\lambda R and cancels the Ricci divergence, giving ∇μGμλ=0\nabla^\mu G_{\mu\lambda}=0. Why: the coefficient 12\tfrac12 in the Einstein tensor is tuned to precisely this cancellation, which is why GμνG_{\mu\nu}, not RμνR_{\mu\nu}, is the geometric object that can equal a conserved source — and the cancellation is dimension-independent even though the trace gσνgσν=ng^{\sigma\nu}g_{\sigma\nu}=n that appears mid-proof is not.

Limiting cases

How the identity specializes or generalizes across the settings that matter most:

CaseResultMeaning
Flat spacetimeRμνρσ=0⇒∇[λRμν]ρσ=0R_{\mu\nu\rho\sigma}=0\Rightarrow\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0 triviallyin Minkowski space the identity holds vacuously; its content lives only where curvature varies
Vacuum (Ricci-flat)∇μCμνρσ=0\nabla^\mu C_{\mu\nu\rho\sigma}=0with Rμν=0R_{\mu\nu}=0 Riemann equals Weyl, and Bianchi becomes a source-free wave-like law for CC — the propagation of gravitational radiation
Maximally symmetric / Einstein spaceRμν=Rngμν⇒∇μR=0R_{\mu\nu}=\tfrac{R}{n}g_{\mu\nu}\Rightarrow\nabla_\mu R=0 for n>2n>2contraction forces the scalar curvature to be constant (Schur's theorem)
Connection with torsion∇[λRμν]ρσ=∇[λTμν]ρ+…\nabla_{[\lambda}R_{\mu\nu]}{}^{\rho}{}_{\sigma}=\nabla_{[\lambda}T_{\mu\nu]}{}^{\rho}+\dotsfor a general affine connection the identity acquires torsion terms; the clean form needs Tρμν=0T^{\rho}{}_{\mu\nu}=0
Gauge theory (Yang–Mills)DF=dF+A∧F+F∧A=0\mathrm{D}F=\mathrm{d}F+A\wedge F+F\wedge A=0the exact structural analogue: the curvature of any principal connection obeys a Bianchi identity, the geometric origin of source-free field equations

Worked examples

1 · Schur's theorem: isotropy forces constant curvature

A spacetime is an "Einstein space" if its Ricci tensor is proportional to the metric, Rμν=f gμνR_{\mu\nu}=f\,g_{\mu\nu} for some scalar function ff — the natural expression of local isotropy. Use the twice-contracted Bianchi identity to show that in dimension n>2n>2 the function ff cannot vary from point to point, so the scalar curvature is constant. This is Schur's theorem, and it is the reason the maximally symmetric cosmological models have a genuinely constant curvature radius.

Step 1. Take the trace of Rμν=f gμνR_{\mu\nu}=f\,g_{\mu\nu} with gμνg^{\mu\nu}. Since gμνgμν=ng^{\mu\nu}g_{\mu\nu}=n, this gives R=nfR=n f, so f=R/nf=R/n and the Einstein-space condition reads Rμν=RngμνR_{\mu\nu}=\tfrac{R}{n}g_{\mu\nu}.

Step 2. Take the divergence of both sides with ∇μ\nabla^\mu. On the right, metric compatibility lets ∇μ\nabla^\mu pass through gμνg_{\mu\nu}, so ∇μ ⁣(Rngμν)=1n∇νR\nabla^\mu\!\left(\tfrac{R}{n}g_{\mu\nu}\right)=\tfrac1n\nabla_\nu R. On the left, the twice-contracted Bianchi identity (5) gives ∇μRμν=12∇νR\nabla^\mu R_{\mu\nu}=\tfrac12\nabla_\nu R. Equating: 12∇νR=1n∇νR\tfrac12\nabla_\nu R=\tfrac1n\nabla_\nu R.

Step 3. Rearrange to (12−1n)∇νR=0\left(\tfrac12-\tfrac1n\right)\nabla_\nu R=0, i.e. n−22n ∇νR=0\dfrac{n-2}{2n}\,\nabla_\nu R=0. For any dimension n>2n>2 the prefactor is nonzero, so ∇νR=0\nabla_\nu R=0 — the scalar curvature is constant. Feeding this back, Rμν=RngμνR_{\mu\nu}=\tfrac{R}{n}g_{\mu\nu} with RR constant, so the full Ricci tensor is covariantly constant too.

Step 4. Check the boundary case. In n=2n=2 the prefactor n−22n\tfrac{n-2}{2n} vanishes and the argument gives no constraint — consistent with the fact that every 2-surface is trivially an Einstein space with RR free to vary (e.g. a generic curved sheet). In our four-dimensional world n=4n=4 gives n−22n=14≠0\tfrac{n-2}{2n}=\tfrac14\neq0, so isotropy of the Ricci tensor at every point of a 4-spacetime rigidly implies constant RR — the mathematical backbone of the de Sitter and anti–de Sitter geometries.

2 · The Bianchi identity delivers the cosmological continuity equation

In a spatially flat Friedmann–Lemaître–Robertson–Walker universe with metric ds2=−dt2+a(t)2 δijdxidxj\mathrm{d}s^2=-\mathrm{d}t^2+a(t)^2\,\delta_{ij}\mathrm{d}x^i\mathrm{d}x^j and a perfect fluid Tμν=diag(−ρ,p,p,p)T^\mu{}_\nu=\mathrm{diag}(-\rho,p,p,p), show that ∇μGμν=0\nabla^\mu G_{\mu\nu}=0 forces ρ˙+3H(ρ+p)=0\dot\rho+3H(\rho+p)=0 with H=a˙/aH=\dot a/a. Then read off how matter and radiation dilute with expansion.

Step 1. Because ∇μGμν=0\nabla^\mu G_{\mu\nu}=0 is an identity, the field equations Gμν=8πG TμνG_{\mu\nu}=8\pi G\,T_{\mu\nu} immediately give ∇μTμν=0\nabla^\mu T_{\mu\nu}=0 with no further work — the conservation of the source is inherited from the geometry. We only have to evaluate the ν=0\nu=0 (time) component of ∇μTμ0=0\nabla_\mu T^\mu{}_0=0.

Step 2. Expand the covariant divergence: ∇μTμ0=∂μTμ0+ΓμμλTλ0−Γλμ0Tμλ\nabla_\mu T^\mu{}_0=\partial_\mu T^\mu{}_0+\Gamma^\mu{}_{\mu\lambda}T^\lambda{}_0-\Gamma^\lambda{}_{\mu0}T^\mu{}_\lambda. For flat FLRW the needed Christoffels are Γij0=H δij\Gamma^i{}_{j0}=H\,\delta^i{}_j and Γμμ0=∂0ln⁡−g=3H\Gamma^\mu{}_{\mu 0}=\partial_0\ln\sqrt{-g}=3H, since −g=a3\sqrt{-g}=a^3.

Step 3. Insert T00=−ρT^0{}_0=-\rho, Tij=p δijT^i{}_j=p\,\delta^i{}_j: the time derivative gives −ρ˙-\dot\rho, the trace term gives 3H(−ρ)3H(-\rho), and the last term gives −3Hp-3Hp. Collecting, ∇μTμ0=−ρ˙−3Hρ−3Hp=0\nabla_\mu T^\mu{}_0=-\dot\rho-3H\rho-3Hp=0.

Step 4. Flip the sign to reach the standard form: ρ˙+3H(ρ+p)=0\dot\rho+3H(\rho+p)=0 — the cosmological continuity equation, obtained here purely from the Bianchi identity plus the field equations, with no independent conservation postulate.

Step 5. Read off the dilution laws with an equation of state p=wρp=w\rho. Then ρ˙/ρ=−3(1+w)a˙/a\dot\rho/\rho=-3(1+w)\dot a/a integrates to ρ∝a−3(1+w)\rho\propto a^{-3(1+w)}. Pressureless matter (w=0w=0) dilutes as ρ∝a−3\rho\propto a^{-3} — pure volume dilution. Radiation (w=13w=\tfrac13) dilutes as ρ∝a−4\rho\propto a^{-4} — an extra factor of aa from the cosmological redshift of each quantum. A cosmological constant (w=−1w=-1) gives ρ=const\rho=\text{const} — vacuum energy that does not dilute at all. Every one of these familiar scalings is, at root, a consequence of ∇μGμν=0\nabla^\mu G_{\mu\nu}=0.

Physics of the portrait

The identity carries the name of a geometer who was not the first to find it, in a story typical of nineteenth-century differential geometry — where the same structural truth was written down repeatedly before its physical weight was understood.

Origins · 1 minLuigi Bianchi and how this was discovered
LB
Luigi Bianchi
1856–1928 · Italian differential geometer at the Scuola Normale, Pisa

Luigi Bianchi was a leading figure of the Italian school of differential geometry, a student of Enrico Betti and Ulisse Dini at the Scuola Normale Superiore in Pisa, where he spent almost his entire career working on the theory of surfaces, transformation groups, and non-Euclidean geometry. The differential curvature identities that bear his name appear in his work around 1902 — yet, as often happened, they had been written earlier: Aurel Voss obtained them in 1880, and Gregorio Ricci-Curbastro, co-founder with Levi-Civita of the tensor calculus Einstein would later need, was aware of such relations. For two decades they were pure geometry with no evident purpose. Their moment came with general relativity: when Einstein and Hilbert reached the field equations in 1915, the question of why the geometric side should be automatically divergence-free — and hence compatible with conservation of energy and momentum — was answered by the twice-contracted form of Bianchi's identity. An obscure differential relation became the structural keystone of the most successful theory of gravity. Bianchi, who died in 1928, lived to see his identities move from the margins of geometry to the centre of physics, though the fully modern reading — Cartan's forms and Wheeler's "boundary of a boundary" — was still being assembled.

History
1880
— Aurel Voss writes down the differential curvature identities in a study of Riemannian geometry
1902
— Luigi Bianchi publishes the identities that come to bear his name
1915
— Einstein and Hilbert independently arrive at the field equations Gμν = κ Tμν
1917
— the contracted Bianchi identity ∇^μ Gμν = 0 is recognized as the guarantee of energy–momentum conservation
1922
— Cartan recasts curvature and its identity in the language of differential forms
1973
— Misner, Thorne and Wheeler popularize the boundary-of-a-boundary reading in Gravitation

Going further

References. Misner, Thorne & Wheeler, Gravitation, §§13.5, 15.3 (the boundary of a boundary) · Wald, General Relativity, §3.2 · Carroll, Spacetime and Geometry, §3.6 · Hartle, Gravity, §21.2 · Schutz, A First Course in General Relativity, §6.5 · Nakahara, Geometry, Topology and Physics, §7.4 (differential-forms version). Several internal consistencies are worth noting. The antisymmetrized double partial of the connection vanishes in normal coordinates, which is the heart of the proof of ∇[λRμν]ρσ=0\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0. In the double contraction to ∇μGμν=0\nabla^\mu G_{\mu\nu}=0, the factor 12\tfrac12 tracks the metric trace gσνgσν=ng^{\sigma\nu}g_{\sigma\nu}=n, while the cancellation itself is dimension-independent. The Weyl-divergence relation (7) is trace-free under gνρg^{\nu\rho} in n=4n=4, and it is this trace-freedom that fixes the coefficients 12\tfrac12 and 16\tfrac16 appearing there. Schur's theorem follows from the prefactor n−22n∇νR=0\tfrac{n-2}{2n}\nabla_\nu R=0, and the FLRW continuity equation ρ˙+3H(ρ+p)=0\dot\rho+3H(\rho+p)=0 yields the dilution laws ρ∝a−3(1+w)\rho\propto a^{-3(1+w)}. No numerical constants beyond standard tensor traces enter any of these results.

Prerequisites. Riemann Curvature Tensor · Curvature as the Commutator of Covariant Derivatives · Algebraic Symmetries of the Riemann Tensor · First (Algebraic) Bianchi Identity. Leads to. Contracted Bianchi Identities · Einstein Tensor · Weyl Conformal Tensor · Ricci Decomposition of the Riemann Tensor.

Ask the tutor: Ask me to walk through the double contraction index by index, or to show exactly how the four contracted Bianchi identities become the constraint equations of the initial-value problem.

Is the second Bianchi identity just the first one with a derivative attached?

No — they are structurally different. The first (algebraic) identity, Rρ[σμν]=0R_{\rho[\sigma\mu\nu]}=0, has no derivatives and cyclically permutes the last three indices of a single Riemann tensor; it constrains the curvature at one point. The second (differential) identity, ∇[λRμν]ρσ=0\nabla_{[\lambda}R_{\mu\nu]\rho\sigma}=0, carries one covariant derivative and cyclically permutes the derivative index with the first pair of curvature indices; it relates curvature at neighbouring points. The first governs how curvature is built at a point, the second how it may vary from point to point — and they come from different sources, the antisymmetry of the connection versus the commutation of derivatives (the Jacobi identity).

Why is a torsion-free connection needed for the simple form of the identity?

The clean proof leans on two facts that require zero torsion. First, in Riemann normal coordinates the Christoffel symbols vanish at a point, which is available only for a symmetric connection Γρμν=Γρνμ\Gamma^\rho{}_{\mu\nu}=\Gamma^\rho{}_{\nu\mu}. Second, after antisymmetrizing, the surviving terms are second partial derivatives whose two derivative slots are symmetric — and it is precisely the symmetry of the connection in its lower indices that makes the would-be extra pieces vanish. If the connection has torsion, Tρμν=Γρμν−Γρνμ≠0T^\rho{}_{\mu\nu}=\Gamma^\rho{}_{\mu\nu}-\Gamma^\rho{}_{\nu\mu}\neq0, then those pieces survive and the identity acquires extra terms involving ∇T\nabla T and T TT\,T. The Levi-Civita connection of general relativity is torsion-free by construction, so the simple form applies throughout GR.

Why can't the field equation just be Rμν = κ Tμν?

Because that equation is inconsistent with energy–momentum conservation. Matter obeys ∇μTμν=0\nabla^\mu T_{\mu\nu}=0, so taking the divergence of Rμν=κTμνR_{\mu\nu}=\kappa T_{\mu\nu} would demand ∇μRμν=0\nabla^\mu R_{\mu\nu}=0. But the twice-contracted Bianchi identity says ∇μRμν=12∇νR\nabla^\mu R_{\mu\nu}=\tfrac12\nabla_\nu R, which is generally not zero. Forcing it to vanish would impose ∇νR=0\nabla_\nu R=0 — constant scalar curvature everywhere — an unphysical extra constraint no realistic matter distribution would satisfy. The cure is to source the trace-reversed combination: Gμν=Rμν−12RgμνG_{\mu\nu}=R_{\mu\nu}-\tfrac12 R g_{\mu\nu} satisfies ∇μGμν=0\nabla^\mu G_{\mu\nu}=0 identically, so Gμν=κTμνG_{\mu\nu}=\kappa T_{\mu\nu} is consistent with conservation for any matter. This is exactly why the Einstein tensor, and not the Ricci tensor, sits on the left of the field equations.

How does the identity count the degrees of freedom in Einstein's equations?

The metric has ten independent components, and Einstein's equations are ten second-order partial differential equations for them. If all ten were genuinely dynamical, the metric would be over-determined and the theory would have no coordinate freedom. The four identities packaged in ∇μGμν=0\nabla^\mu G_{\mu\nu}=0 are differential relations among the ten equations, meaning only six are independent evolution equations; the remaining four are constraints on the initial data. Six dynamical equations for ten components leaves a four-function ambiguity — exactly the freedom to choose four spacetime coordinates (the diffeomorphism gauge freedom). So the Bianchi identity is what reconciles ten field equations with a well-posed initial-value problem and with general covariance.

What does "the boundary of a boundary is zero" actually mean here?

It is Wheeler's mnemonic for the topological fact that the boundary operator applied twice gives nothing: the boundary of a solid region is a closed surface, and a closed surface has no boundary of its own. Translate curvature into the holonomy — the rotation a vector picks up — around the faces of a small three-dimensional cell, and sum those rotations over the cell's boundary. Every internal face is shared by two neighbours and traversed once in each direction, so the contributions cancel exactly. In the infinitesimal limit that cancellation is the differential Bianchi identity. The same nilpotency ∂∂=0\partial\partial=0 appears in the forms language as d2=0\mathrm{d}^2=0 — what makes the Cartan derivation work and ties the identity to the closed characteristic forms of topology.

In what sense is the Bianchi identity the gravitational analogue of Maxwell's equations?

Curvature is the field strength of the connection, just as FμνF_{\mu\nu} is the field strength of the potential AμA_\mu. Whenever a field strength is built as the "curl" of a potential it automatically satisfies a source-free identity: F=dAF=\mathrm{d}A forces dF=0\mathrm{d}F=0, i.e. ∂[μFνρ]=0\partial_{[\mu}F_{\nu\rho]}=0, which encodes ∇⋅B=0\nabla\cdot\mathbf{B}=0 and Faraday's law. The differential Bianchi identity DR=0\mathrm{D}\mathcal{R}=0 is the exact non-abelian analogue for the Levi-Civita connection, and DF=0\mathrm{D}F=0 is the version for a Yang–Mills field. In each case Bianchi is the half of the field equations that is automatic, leaving the other half free to be sourced by a conserved current or stress-energy — one of the deepest structural links between gravity and gauge theory.

Before this

Where this leads