Coupling of Matter to Gravity

Once you know how matter behaves in the flat spacetime of special relativity, general relativity poses a single sharp question: how does that same matter behave when spacetime is curved? The answer is the minimal coupling prescription — a recipe so compact it fits on one line. Wherever the flat-space Lagrangian carries the Minkowski metric $\eta_{\mu\nu}$ , write the dynamical metric $g_{\mu\nu}$ ; wherever it carries a partial derivative $\partial_\mu$ , write the Levi-Civita covariant derivative $\nabla_\mu$ ; and replace the flat volume element $\mathrm{d}^4x$ by the invariant $\sqrt{-g}\,\mathrm{d}^4x$ . What comes out is a scalar action invariant under every smooth coordinate change — a diffeomorphism. This one recipe is the action-level statement of the Einstein equivalence principle. At any point $p$ you can choose Riemann normal coordinates in which $g_{\mu\nu}(p)=\eta_{\mu\nu}$ and $\partial_\lambda g_{\mu\nu}(p)=0$ , so $\nabla_\mu\to\partial_\mu$ and the matter equations locally read exactly as in special relativity. Gravity lives not at a point but in how the geometry changes from point to point — in the connection $\Gamma^\lambda_{\mu\nu}$ and ultimately the Riemann tensor, which no coordinate choice can erase. The coupling runs both ways. The very action that dictates how matter moves also sources Einstein's equations: vary the total action $S=\tfrac{1}{2\kappa}\int R\sqrt{-g}\,\mathrm{d}^4x+S_m$ with respect to $g^{\mu\nu}$ and out drops $G_{\mu\nu}=\kappa T_{\mu\nu}$ , with the stress-energy tensor born as the metric-variation of $S_m$ . Geometry tells matter how to move; matter tells geometry how to curve. The glue is diffeomorphism invariance, which hands you $\nabla_\mu T^{\mu\nu}=0$ for free, as a Noether identity rather than a separate law. And a warning is built in: "minimal" is a choice, not a theorem, because once two covariant derivatives meet, curvature has nowhere to hide.

Key takeaways

Minimal coupling promotes any special-relativistic matter theory to curved spacetime by three replacements — $\eta_{\mu\nu}\to g_{\mu\nu}$ , $\partial_\mu\to\nabla_\mu$ , $\mathrm{d}^4x\to\sqrt{-g}\,\mathrm{d}^4x$ — while demanding the action stay a generally covariant scalar.
It is the action-level equivalence principle: in a local inertial frame ( $g_{\mu\nu}\to\eta_{\mu\nu}$ , $\Gamma^\lambda_{\mu\nu}\to0$ at a point) the physics is exactly special-relativistic, so gravity vanishes locally.
The covariant matter action defines the stress-energy tensor by its metric-variation, $T_{\mu\nu}=-\tfrac{2}{\sqrt{-g}}\,\delta S_m/\delta g^{\mu\nu}$ — automatically symmetric, and the source in $G_{\mu\nu}=\kappa T_{\mu\nu}$ with $\kappa=8\pi G/c^4\approx2.08\times10^{-43}\ \mathrm{s^2\,kg^{-1}\,m^{-1}}$ .
Diffeomorphism invariance of $S_m$ forces $\nabla_\mu T^{\mu\nu}=0$ on-shell — a Noether identity, not an extra postulate, and exactly what the contracted Bianchi identity $\nabla^\mu G_{\mu\nu}=0$ demands.
Minimal is not unique: curvature terms like $\xi R\phi^2$ are equally covariant and share the flat limit, and $[\nabla_\mu,\nabla_\nu]V^\rho=R^\rho{}_{\sigma\mu\nu}V^\sigma$ makes the derivative ordering matter once curvature is present.
Key idea: Replace $\eta_{\mu\nu}\to g_{\mu\nu}$ , $\partial_\mu\to\nabla_\mu$ , $\mathrm{d}^4x\to\sqrt{-g}\,\mathrm{d}^4x$ in a flat-space scalar action; the resulting diffeomorphism-invariant $S_m[\phi,g]$ both governs matter dynamics and defines $T_{\mu\nu}$ , with $\nabla_\mu T^{\mu\nu}=0$ following automatically from coordinate invariance.

Where this lives today

Matter-gravity coupling is the working engine of relativistic astrophysics and cosmology. The Tolman-Oppenheimer-Volkoff equation — minimal coupling of a perfect fluid, with $\nabla_\mu T^{\mu\nu}=0$ supplying hydrostatic balance — sets the maximum mass of a neutron star, a number sharpened by the 2017 merger GW170817, where relativistic hydrodynamics coupled to gravity had to reproduce both the gravitational-wave chirp and the kilonova. In numerical relativity the coupled Einstein-matter system is evolved directly on supercomputers to supply the waveform templates LIGO and Virgo match against. In cosmology the minimally coupled scalar of Example 1 is the inflaton: a slowly rolling field with $\dot\phi^2\ll V$ gives $p\approx-\rho$ and drives accelerated expansion, and the choice between minimal and non-minimal ( $\xi R\phi^2$ ) coupling distinguishes competing models — Higgs inflation lives or dies on the value of $\xi$ , which Planck's constraints test directly. At the frontier, the failure of minimal coupling for spinors and the ordering ambiguity in curved space are entry points to Einstein-Cartan gravity and to the open problem of coupling quantum fields to dynamical geometry.

The physical picture

Take your favourite flat-space field theory — Maxwell, a scalar, a fluid — and ask how a freely falling observer describes it in a gravitational field. By the equivalence principle their local laboratory is indistinguishable from one drifting in deep space, so their equations must read special-relativistically. Minimal coupling is the unique-up-to-curvature way to stitch all these local inertial frames into one covariant law: the "comma-goes-to-semicolon rule" of MTW, where every partial derivative (comma) becomes a covariant derivative (semicolon) and $\eta\to g$ . The corrections it introduces — Christoffel symbols — are exactly the ones that vanish in a local inertial frame, which is why the recipe is honest. But the deeper content is that the same action does two jobs. It is the arena where matter moves, and it is the source term that curves spacetime. The first figure makes that two-way street the whole picture, and pins the strength of the coupling to a real number.

Figure

The coupling runs both ways: one action S = S_EH + S_m ties geometry to matter. Varying it by the metric gives G_μν = κ T_μν with κ = 8πG/c⁴ ≈ 2.08×10⁻⁴³ s² kg⁻¹ m⁻¹ — geometry tells matter how to move, matter tells geometry how to curve

The recipe itself is best seen as a substitution acting on a Lagrangian, together with the promise that the corrections it adds switch off in a local inertial frame. The second figure shows the three replacements side by side with the manifold they act on: a curved surface carrying, at each point, a tangent flat patch where the connection vanishes and the metric is Minkowski. That tangent patch is the equivalence principle drawn as geometry — and the weak-field number beside it shows just how gently curvature enters at the surface of the Earth.

Figure

Minimal coupling as three substitutions η→g, ∂→∇, d⁴x→√−g d⁴x, glued to a curved manifold that carries a flat local inertial frame at every point p where Γ→0 and g→η. In the weak field g₀₀ = −(1 + 2Φ/c²); at Earth's surface 2Φ/c² ≈ −1.4×10⁻⁹

Why does demanding a scalar action buy so much? Because general covariance is enormously restrictive: it ties the matter dynamics, the definition of energy-momentum, and the conservation law into one self-consistent package, so you never separately postulate $\nabla_\mu T^{\mu\nu}=0$ — it falls out. The ambiguity in "minimal" is the counterweight: whenever two derivatives meet, $[\nabla_\mu,\nabla_\nu]\neq0$ , and the recipe quietly makes a choice that different theories make differently, agreeing only where spacetime is flat.

The core equations

The recipe is three replacements, applied to a flat-space scalar Lagrangian and nothing more. Each replacement is the minimal covariant object that reduces to its flat counterpart in a local inertial frame.

\eta_{\mu\nu}\to g_{\mu\nu},\qquad \partial_\mu\to\nabla_\mu,\qquad \mathrm{d}^4x\to\sqrt{-g}\,\mathrm{d}^4x \qquad (1)

where: $\eta_{\mu\nu}$ — Minkowski metric, $\mathrm{diag}(-1,1,1,1)$ ; $g_{\mu\nu}$ — dynamical curved-spacetime metric; $\nabla_\mu$ — Levi-Civita covariant derivative (metric-compatible, torsion-free); $\sqrt{-g}$ — square root of minus the metric determinant, the invariant volume density. Read it as: keep the flat theory's form, but let the geometry become dynamical. The covariant derivative is what carries the gravitational information. On a vector it adds a connection term built from first derivatives of the metric — the Christoffel symbols — and it is those symbols, not the metric value, that encode the local gravitational field.

\nabla_\mu V^\nu=\partial_\mu V^\nu+\Gamma^\nu_{\mu\lambda}V^\lambda,\qquad \Gamma^\lambda_{\mu\nu}=\tfrac12 g^{\lambda\sigma}\big(\partial_\mu g_{\sigma\nu}+\partial_\nu g_{\sigma\mu}-\partial_\sigma g_{\mu\nu}\big) \qquad (2)

where: $\Gamma^\lambda_{\mu\nu}$ — Christoffel symbols of the second kind (Levi-Civita connection); $V^\nu$ — the vector field being differentiated; $\partial_\mu$ — ordinary partial derivative. Because $\Gamma$ is built from $\partial g$ , it can always be made to vanish at a chosen point — that is the local inertial frame. Applying (1) to any flat matter Lagrangian yields the covariant matter action: a single diffeomorphism-invariant scalar functional of the fields and the metric. This object is the hinge of the whole subject.

S_m[\phi,g]=\int \mathcal{L}_m\big(\phi,\nabla\phi,g_{\mu\nu}\big)\,\sqrt{-g}\,\mathrm{d}^4x \qquad (3)

where: $S_m$ — matter action, a diffeomorphism-invariant scalar; $\mathcal{L}_m$ — matter Lagrangian (a scalar, so $\mathcal{L}_m\sqrt{-g}$ is a density); $\phi$ — schematic matter fields of any tensor type. The metric appears both explicitly in $\mathcal{L}_m$ and in the volume measure. Because the metric sits inside $S_m$ , differentiating the action with respect to it must produce a physical response — and that response is precisely the stress-energy tensor. This is the Hilbert definition, and it is automatically symmetric because $g^{\mu\nu}$ is.

T_{\mu\nu}=-\frac{2}{\sqrt{-g}}\frac{\delta S_m}{\delta g^{\mu\nu}}=-2\frac{\partial\mathcal{L}_m}{\partial g^{\mu\nu}}+g_{\mu\nu}\mathcal{L}_m \qquad (4)

where: $T_{\mu\nu}$ — symmetric (Hilbert) stress-energy tensor; $\delta S_m/\delta g^{\mu\nu}$ — functional derivative of $S_m$ with respect to the inverse metric. The factor $-2$ is fixed by convention so the source enters Einstein's equations as $+\kappa T_{\mu\nu}$ . The canonical example is a single scalar field: its minimally coupled Lagrangian is the flat one with $\eta\to g$ and (since a scalar has $\nabla_\mu\phi=\partial_\mu\phi$ ) no Christoffel corrections in the kinetic term at all. Its field equation is the flat wave equation with the curved d'Alembertian.

\mathcal{L}_{\text{scalar}}=-\tfrac12 g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi-V(\phi),\qquad \Box\phi-V'(\phi)=0,\quad \Box\phi=\frac{1}{\sqrt{-g}}\partial_\mu\!\big(\sqrt{-g}\,g^{\mu\nu}\partial_\nu\phi\big) \qquad (5)

where: $\phi$ — real scalar field; $V(\phi)$ — potential, e.g. $\tfrac12 m^2\phi^2$ ; $\Box$ — curved-space d'Alembertian (Laplace-Beltrami operator); $V'$ — $\mathrm{d}V/\mathrm{d}\phi$ . The $\sqrt{-g}$ in $\Box$ is exactly what makes the wave operator covariant. Electromagnetism is the instructive case where minimal coupling does less than you might fear: the field strength is a curl, and the symmetric Christoffel terms cancel, so $F_{\mu\nu}$ is untouched and Maxwell's source equation simply gains covariant derivatives.

\mathcal{L}_{\text{EM}}=-\tfrac14 g^{\mu\alpha}g^{\nu\beta}F_{\mu\nu}F_{\alpha\beta},\qquad F_{\mu\nu}=\nabla_\mu A_\nu-\nabla_\nu A_\mu=\partial_\mu A_\nu-\partial_\nu A_\mu \qquad (6)

where: $A_\mu$ — electromagnetic 4-potential; $F_{\mu\nu}$ — field strength, metric-independent because the connection is torsion-free ( $\Gamma^\lambda_{\mu\nu}=\Gamma^\lambda_{\nu\mu}$ ). Only the raising of indices, via $g^{\mu\alpha}g^{\nu\beta}$ , brings the geometry in. Finally, the covariant conservation law. It is not a separate axiom bolted on — it is a Noether identity forced by the diffeomorphism invariance of $S_m$ , and it is exactly what the coupled field equations require through the contracted Bianchi identity $\nabla^\mu G_{\mu\nu}=0$ .

\nabla_\mu T^{\mu\nu}=0 \qquad (7)

where: $\nabla_\mu T^{\mu\nu}$ — covariant divergence of the stress-energy tensor. It holds whenever the matter fields satisfy their own equations of motion (on-shell), and it is the curved-space replacement for $\partial_\mu T^{\mu\nu}=0$ .

Derivation AThe Hilbert stress-energy tensor from metric variation

The goal is to show that a single, well-defined operation on the covariant matter action — differentiating it with respect to the inverse metric — produces a symmetric, conserved stress-energy tensor with no further input, and to expose the one piece of machinery ( $\delta\sqrt{-g}$ ) that makes the calculation go. We vary $S_m$ at fixed matter fields, track the explicit metric dependence and the volume dependence separately, and read off $T_{\mu\nu}$ as the coefficient of $\delta g^{\mu\nu}$ .

Step 1. Vary the volume density. Jacobi's formula for the derivative of a determinant gives $\delta g=g\,g^{\mu\nu}\delta g_{\mu\nu}$ , and since $g^{\mu\nu}\delta g_{\mu\nu}=-g_{\mu\nu}\delta g^{\mu\nu}$ , one finds $\delta\sqrt{-g}=-\tfrac12\sqrt{-g}\,g_{\mu\nu}\,\delta g^{\mu\nu}$ . Why: the measure itself depends on the metric, so a full metric variation of the action cannot ignore it; this term is what supplies the $+g_{\mu\nu}\mathcal{L}_m$ piece of the tensor.

Step 2. Collect the variation of the action. Writing $\delta(\mathcal{L}_m\sqrt{-g})=\big(\partial\mathcal{L}_m/\partial g^{\mu\nu}\big)\delta g^{\mu\nu}\sqrt{-g}+\mathcal{L}_m\,\delta\sqrt{-g}$ and substituting Step 1 gives $\delta S_m=\int\big(\partial\mathcal{L}_m/\partial g^{\mu\nu}-\tfrac12 g_{\mu\nu}\mathcal{L}_m\big)\delta g^{\mu\nu}\sqrt{-g}\,\mathrm{d}^4x$ . Why: the two metric-dependences — explicit in the Lagrangian, implicit in the measure — combine into a single coefficient of $\delta g^{\mu\nu}$ , which is all a stress-energy tensor can be.

Step 3. Read off the tensor. Defining $T_{\mu\nu}\equiv-\tfrac{2}{\sqrt{-g}}\,\delta S_m/\delta g^{\mu\nu}$ yields $T_{\mu\nu}=-2\,\partial\mathcal{L}_m/\partial g^{\mu\nu}+g_{\mu\nu}\mathcal{L}_m$ . Because $g^{\mu\nu}=g^{\nu\mu}$ , the functional derivative is symmetric in $\mu\nu$ , so $T_{\mu\nu}$ is symmetric with no symmetrization by hand. Why: this is the decisive advantage over the canonical Noether tensor of flat field theory, which is generally neither symmetric nor gauge-invariant; the Hilbert tensor is both, precisely because it couples to the symmetric metric.

Step 4. Note what fixes the constant. The factor $-2$ is chosen so that when $S_m$ is added to the Einstein-Hilbert action and the total is stationary, the matter term appears as $+\kappa T_{\mu\nu}$ on the right of $G_{\mu\nu}=\kappa T_{\mu\nu}$ ; any other constant would merely rescale $\kappa$ . Why: the normalization is not physics, it is bookkeeping that lines up the Hilbert definition with the Newtonian value $\kappa=8\pi G/c^4$ .

Go deeper · Derivation B · 1 minCovariant conservation as a Noether identity

Derivation BCovariant conservation as a Noether identity

The aim is to prove that $\nabla_\mu T^{\mu\nu}=0$ is not an independent postulate but a mathematical consequence of the diffeomorphism invariance of the matter action, holding automatically whenever the matter fields are on-shell. We generate a metric variation from an infinitesimal coordinate change, insert the Hilbert definition of $T_{\mu\nu}$ , integrate by parts, and use the arbitrariness of the coordinate shift to force the divergence to vanish.

Step 1. Under an infinitesimal diffeomorphism $x^\mu\to x^\mu+\xi^\mu$ , the metric changes by a Lie derivative, $\delta g_{\mu\nu}=\nabla_\mu\xi_\nu+\nabla_\nu\xi_\mu$ , so on the inverse metric $\delta g^{\mu\nu}=-(\nabla^\mu\xi^\nu+\nabla^\nu\xi^\mu)$ . Why: a diffeomorphism is a relabelling of points, and a scalar action cannot change under it — this is the symmetry we are about to cash in.

Step 2. Evaluate the induced change in $S_m$ . On-shell the field-variation terms vanish by the matter equations of motion, leaving only the metric piece, $\delta S_m=-\tfrac12\int T_{\mu\nu}\,\delta g^{\mu\nu}\sqrt{-g}\,\mathrm{d}^4x$ , and this must equal zero because $S_m$ is a scalar. Why: invariance of the action plus the equations of motion isolates the pure geometric response, which is where the conservation law is hiding.

Step 3. Substitute Step 1 and integrate by parts. Using the symmetry of $T_{\mu\nu}$ , $0=\int T_{\mu\nu}\nabla^\mu\xi^\nu\sqrt{-g}\,\mathrm{d}^4x=-\int(\nabla^\mu T_{\mu\nu})\,\xi^\nu\sqrt{-g}\,\mathrm{d}^4x$ , discarding the boundary term for compactly supported $\xi^\nu$ . Why: moving the derivative off $\xi$ and onto $T$ turns a statement about the action into a statement about the divergence of the tensor.

Step 4. Because $\xi^\nu$ is an arbitrary vector field, the integrand must vanish pointwise: $\nabla_\mu T^{\mu\nu}=0$ . Why: this is the Noether identity for local (gauge-like) symmetries — general covariance guarantees conservation, exactly matching the contracted Bianchi identity $\nabla^\mu G_{\mu\nu}=0$ that the geometry side already obeys.

Go deeper · Derivation C · 1 minEinstein's equations from the coupled action

Derivation CEinstein's equations from the coupled action

The purpose here is to close the loop: to show that varying the sum of the gravitational and matter actions by the metric yields Einstein's equations with the Hilbert stress-energy tensor as the source, so that the same $S_m$ which governs matter also curves spacetime. We add the Einstein-Hilbert term to $S_m$ , vary each piece with respect to $g^{\mu\nu}$ , impose stationarity, and rearrange into the standard form.

Step 1. Vary the gravitational part. The variation of $\int R\sqrt{-g}\,\mathrm{d}^4x$ gives $\int(R_{\mu\nu}-\tfrac12 R g_{\mu\nu})\delta g^{\mu\nu}\sqrt{-g}\,\mathrm{d}^4x$ ; the $\delta R_{\mu\nu}$ contribution is a total covariant divergence collected by the Gibbons-Hawking-York boundary term. Why: only the Einstein tensor $G_{\mu\nu}=R_{\mu\nu}-\tfrac12 R g_{\mu\nu}$ survives in the bulk, which is why the geometry side of the field equations is exactly $G_{\mu\nu}$ .

Step 2. Vary the matter part. By the Hilbert definition, $\delta S_m=-\tfrac12\int T_{\mu\nu}\,\delta g^{\mu\nu}\sqrt{-g}\,\mathrm{d}^4x$ . Why: this is precisely the tensor built in the first derivation, now appearing as the matter response to a metric variation — the source.

Step 3. Impose stationarity. Setting $\delta S=0$ for arbitrary $\delta g^{\mu\nu}$ gives $\tfrac{1}{2\kappa}(R_{\mu\nu}-\tfrac12 R g_{\mu\nu})-\tfrac12 T_{\mu\nu}=0$ . Why: the principle of least action for the coupled system is the single statement that produces both the geometry and its coupling to matter at once.

Step 4. Rearrange to $G_{\mu\nu}=\kappa T_{\mu\nu}$ with $\kappa=8\pi G/c^4$ . Taking the covariant divergence, the contracted Bianchi identity $\nabla^\mu G_{\mu\nu}=0$ then reproduces $\nabla^\mu T_{\mu\nu}=0$ — the same law the matter Noether identity gave independently. Why: the two conservation statements agreeing is the internal consistency check that makes the coupling well-posed rather than overdetermined.

Limiting cases

Minimal coupling is pinned down by what it must reduce to, and it shows its seams where curvature has no flat analogue.

Case	Result	Meaning
Local inertial frame	$g_{\mu\nu}\to\eta_{\mu\nu},\ \Gamma^\lambda_{\mu\nu}\to0,\ \nabla_\mu\to\partial_\mu$	At any point, Riemann normal coordinates make the metric Minkowski and the connection vanish; the covariant matter Lagrangian becomes its special-relativistic form — the equivalence principle realized.
Newtonian / weak field	$g_{00}\simeq-(1+2\Phi/c^2),\ T^{00}\simeq\rho c^2$	For slow matter in a weak static field the coupled point-particle action gives $\ddot x^i=-\partial_i\Phi$ , recovering Newton with $\Phi$ the potential; $\kappa=8\pi G/c^4$ is fixed here.
Non-minimal coupling	$\mathcal{L}=-\tfrac12 g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi-\tfrac12 m^2\phi^2-\tfrac12\xi R\phi^2$	Adding $\xi R\phi^2$ is equally covariant and shares the flat limit; $\xi=\tfrac16$ gives conformal coupling in 4D. Minimal ( $\xi=0$ ) is a choice, not a theorem.
Spinor fields	$\mathcal{L}_{\text{Dirac}}=\bar\psi\big(i\gamma^a e_a^\mu D_\mu-mc/\hbar\big)\psi,\ D_\mu=\partial_\mu+\tfrac14\omega_{\mu ab}\gamma^{ab}$	Spinors cannot couple to $g_{\mu\nu}$ directly; one needs a tetrad $e_a^\mu$ and spin connection $\omega_{\mu ab}$ . Minimal coupling here means the spinor-covariant derivative built from $\omega$ .

Common trapMinimal coupling is unique and unambiguous

It is neither. First, "minimal" means only that you use $g_{\mu\nu}$ , $\nabla_\mu$ , and $\sqrt{-g}$ with no explicit curvature tensors — but explicit-curvature terms like $\xi R\phi^2$ are perfectly covariant and reduce to the same flat theory, so they are legitimate alternatives that minimal coupling simply declines to include. Second, and more subtly, the recipe is genuinely ambiguous beyond first derivatives: promoting $\partial_\mu\partial_\nu\to\nabla_\mu\nabla_\nu$ hides an ordering choice, because $[\nabla_\mu,\nabla_\nu]V^\rho=R^\rho{}_{\sigma\mu\nu}V^\sigma$ differs from zero by curvature. Two "minimally coupled" theories that disagree only by such a curvature term are different physical theories that happen to coincide in flat space. Third, the prescription is clean for tensor fields but fails outright for spinors: Dirac fields have no minimal coupling using $g_{\mu\nu}$ alone and require the tetrad and spin connection. Treat "minimal" as a deliberate simplicity convention, not a derivation.

Worked examples

1 · The stress-energy of a minimally coupled scalar field (fully worked)

Take the minimally coupled real scalar $\mathcal{L}_m=-\tfrac12 g^{\alpha\beta}\partial_\alpha\phi\,\partial_\beta\phi-V(\phi)$ . Compute its Hilbert stress-energy tensor and confirm the energy density is positive in the local inertial frame.

Step 1. Isolate the explicit metric dependence. Only the kinetic term carries $g^{\alpha\beta}$ , so $\partial\mathcal{L}_m/\partial g^{\mu\nu}=-\tfrac12\partial_\mu\phi\,\partial_\nu\phi$ ; the potential $V(\phi)$ has none. This single term is all the metric variation sees.

Step 2. Insert into the Hilbert definition (4), $T_{\mu\nu}=-2\,\partial\mathcal{L}_m/\partial g^{\mu\nu}+g_{\mu\nu}\mathcal{L}_m$ , giving $T_{\mu\nu}=\partial_\mu\phi\,\partial_\nu\phi+g_{\mu\nu}\big(-\tfrac12(\partial\phi)^2-V\big)$ , where $(\partial\phi)^2\equiv g^{\alpha\beta}\partial_\alpha\phi\,\partial_\beta\phi$ . The result is manifestly symmetric.

T_{\mu\nu}=\partial_\mu\phi\,\partial_\nu\phi-g_{\mu\nu}\Big(\tfrac12 g^{\alpha\beta}\partial_\alpha\phi\,\partial_\beta\phi+V(\phi)\Big)

Step 3. Evaluate the energy density in a local inertial frame ( $g_{\mu\nu}\to\eta_{\mu\nu}$ , signature $-+++$ ). Then $(\partial\phi)^2=-\dot\phi^2+|\nabla\phi|^2$ and $g_{00}=-1$ , so $T_{00}=\dot\phi^2-(-1)\big(\tfrac12(-\dot\phi^2+|\nabla\phi|^2)+V\big)=\tfrac12\dot\phi^2+\tfrac12|\nabla\phi|^2+V(\phi)$ . $T_{00}=\tfrac12\dot\phi^2+\tfrac12|\nabla\phi|^2+V(\phi)\ge0$ whenever $V\ge0$ — kinetic plus gradient plus potential energy, exactly as expected.

Step 4. Sanity-check the flat limit. Setting $g_{\mu\nu}\to\eta_{\mu\nu}$ throughout reproduces the textbook special-relativistic scalar stress-energy (Carroll Eq. 1.114); the curved expression is its minimal-coupling promotion, and covariant conservation $\nabla_\mu T^{\mu\nu}=0$ is equivalent on-shell to the field equation $\Box\phi=V'(\phi)$ . The same tensor sources the Friedmann equations for a cosmological scalar.

2 · The gravitational coupling constant and the Newtonian limit (fully numeric)

Fix the constant $\kappa$ in $G_{\mu\nu}=\kappa T_{\mu\nu}$ by matching to Newtonian gravity, evaluate it numerically, and read off the size of the metric correction at the Earth's surface.

Step 1. In the weak, static, slow-motion limit the $00$ -component of $G_{\mu\nu}=\kappa T_{\mu\nu}$ reduces to Poisson's equation. Writing $g_{00}=-(1+2\Phi/c^2)$ and $T^{00}\simeq\rho c^2$ , the field equation becomes $\nabla^2\Phi=\tfrac12\kappa\rho c^4$ . Matching to Newton's $\nabla^2\Phi=4\pi G\rho$ forces $\tfrac12\kappa c^4=4\pi G$ , i.e. $\kappa=8\pi G/c^4$ .

Step 2. Put in numbers with $G=6.674\times10^{-11}\ \mathrm{m^3\,kg^{-1}\,s^{-2}}$ and $c=2.998\times10^8\ \mathrm{m/s}$ , so $c^4=8.08\times10^{33}\ \mathrm{m^4\,s^{-4}}$ . Then $\kappa=8\pi(6.674\times10^{-11})/(8.08\times10^{33})=$ $2.08\times10^{-43}\ \mathrm{s^2\,kg^{-1}\,m^{-1}}$ . The staggering smallness of $\kappa$ is why enormous masses produce feeble curvature.

Step 3. Estimate the metric correction at Earth's surface. With $\Phi=-GM_\oplus/R_\oplus$ , $M_\oplus=5.97\times10^{24}\ \mathrm{kg}$ , $R_\oplus=6.37\times10^6\ \mathrm{m}$ , we get $\Phi=-6.26\times10^7\ \mathrm{m^2/s^2}$ , hence $\Phi/c^2=-6.96\times10^{-10}$ and $2\Phi/c^2\approx-1.4\times10^{-9}$ . The deviation of $g_{00}$ from its flat value $-1$ is about one part in a billion.

Step 4. Interpret. That $10^{-9}$ is the gravitational time-dilation and redshift of clocks at the ground versus infinity — small, but far larger than modern optical-clock precision ( $\sim10^{-18}$ ), and the reason GPS must correct for it. Minimal coupling plus the tiny $\kappa$ turns a one-part-in- $10^9$ metric ripple into an everyday engineering correction.

Physics of the portrait

The coupling of matter to gravity is Einstein's central achievement made precise: the demand that the laws of physics take the same form in every frame, freely falling or not, is what forces matter to talk to geometry at all.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879-1955 · the equivalence principle and how matter sources curvature

Einstein's route to coupling matter and gravity began with what he called the happiest thought of his life: a person in free fall does not feel their own weight. Elevated to a principle in 1907, this equivalence of gravity and acceleration told him gravity could be transformed away locally — exactly the statement that in a local inertial frame the connection vanishes and physics is special-relativistic. Over the next eight years he sought field equations making matter the source of curvature while respecting general covariance, the demand that the laws hold in arbitrary coordinates. In November 1915 he reached $G_{\mu\nu}=\kappa T_{\mu\nu}$ , the stress-energy tensor on the right as the density and flux of energy-momentum, the Einstein tensor on the left as pure geometry. The same month David Hilbert obtained the equations from a variational principle, showing that one scalar action $S_{\text{EH}}+S_m$ contains both the dynamics of matter and its gravitational source. The synthesis Wheeler later distilled — matter tells spacetime how to curve, spacetime tells matter how to move — is Einstein's content: the coupling is not a force added to geometry, it is geometry responding to whatever carries energy and momentum.

History

1907

— Einstein formulates the equivalence principle, the seed of local inertial frames

1915

— Einstein presents G_μν = κ T_μν; Hilbert derives it from the action δ(S_EH + S_m) = 0

1916

— Einstein publishes the generally covariant theory of general relativity

1918

— Noether proves that local coordinate invariance forces the conservation identity ∇_μ T^μν = 0

1940

— Belinfante and Rosenfeld relate the canonical and symmetric Hilbert stress-energy tensors

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §16.2-16.3 (comma-goes-to-semicolon rule), §21.3 · Wald, General Relativity (1984), §4.3 and Appendix E (variational $T_{\mu\nu}$ ) · Carroll, Spacetime and Geometry (2004), §4.1-4.3, §5.1 · Hartle, Gravity (2003), Ch. 22 · Schutz, A First Course in General Relativity (2nd ed., 2009), Ch. 7. Key results to keep in mind: the volume-element identity $\delta\sqrt{-g}=-\tfrac12\sqrt{-g}\,g_{\mu\nu}\delta g^{\mu\nu}$ (from Jacobi's formula) underlies the Hilbert definition; the scalar stress-energy in the flat limit gives $T_{00}=\tfrac12\dot\phi^2+\tfrac12|\nabla\phi|^2+V$ ; the covariant-derivative commutator is $[\nabla_\mu,\nabla_\nu]V^\rho=R^\rho{}_{\sigma\mu\nu}V^\sigma$ ; and the coupling constant evaluates to $\kappa=8\pi G/c^4\approx2.08\times10^{-43}\ \mathrm{s^2\,kg^{-1}\,m^{-1}}$ , with the weak-field metric correction at Earth's surface $2\Phi/c^2\approx-1.4\times10^{-9}$ .

Prerequisites. The Stress-Energy Tensor · Variational Definition of Stress-Energy · The Einstein-Hilbert Action. Leads to. The Einstein Field Equations · Local Conservation of Energy-Momentum · Perfect Fluid Stress-Energy · Variational Derivation of the Field Equations.

Ask the tutor: Ask me to walk through the metric variation of any matter Lagrangian — scalar, electromagnetic, or perfect fluid — and watch its Hilbert stress-energy tensor and covariant conservation law fall out step by step.

Is minimal coupling the same thing as the equivalence principle?

They are two faces of one idea. The equivalence principle says that in a small enough freely falling laboratory, gravity is undetectable and physics is special-relativistic. Minimal coupling is that statement written at the level of the action: by replacing $\eta\to g$ , $\partial\to\nabla$ , $\mathrm{d}^4x\to\sqrt{-g}\,\mathrm{d}^4x$ , you build a Lagrangian whose corrections — the Christoffel symbols — all vanish at any chosen point in a local inertial frame, recovering the flat theory there. So minimal coupling is a concrete recipe that guarantees the equivalence principle holds for whatever field you feed in. The two part ways only in the fine print: the equivalence principle is a physical statement about local experiments, while minimal coupling is a specific mathematical prescription that could be modified (by curvature terms) without violating any experiment done in a sufficiently small region.

Why define the stress-energy tensor by varying the metric instead of using Noether's theorem?

The Noether procedure applied to spacetime translations gives the canonical stress-energy tensor, which has two defects: it is generally not symmetric, and for gauge fields like electromagnetism it is not even gauge-invariant. Neither can sit on the right-hand side of $G_{\mu\nu}=\kappa T_{\mu\nu}$ , whose left side is symmetric and geometric. The Hilbert definition $T_{\mu\nu}=-\tfrac{2}{\sqrt{-g}}\,\delta S_m/\delta g^{\mu\nu}$ cures both automatically: because the metric is symmetric, the tensor is symmetric, and because it comes from a diffeomorphism-invariant action, it is gauge-invariant and covariantly conserved. The two definitions differ by a total divergence (the Belinfante-Rosenfeld improvement term) and give the same total energy and momentum, but only the Hilbert tensor is the correct local source of gravity.

Does minimal coupling work for every kind of field?

It works cleanly for scalars and tensors of any rank, because those objects have coordinate indices that the covariant derivative and the metric know how to handle. It fails for spinors. A Dirac field transforms under the spin group, not under the general linear group acting on coordinate indices, so there is simply no way to build its covariant derivative from Christoffel symbols. The fix is to introduce a local orthonormal frame — a tetrad or vierbein $e^a_\mu$ — and a spin connection $\omega_{\mu ab}$ , and to minimally couple through the spinor-covariant derivative $D_\mu=\partial_\mu+\tfrac14\omega_{\mu ab}\gamma^{ab}$ . This tetrad formalism is strictly more general than the metric one, and it is the natural entry point to theories with torsion such as Einstein-Cartan gravity.

What exactly makes minimal coupling ambiguous?

The ambiguity appears only when two or more derivatives act. Minimal coupling tells you to replace $\partial_\mu\to\nabla_\mu$ , but $\partial_\mu\partial_\nu$ is symmetric in $\mu\nu$ while $\nabla_\mu\nabla_\nu$ is not, because $[\nabla_\mu,\nabla_\nu]V^\rho=R^\rho{}_{\sigma\mu\nu}V^\sigma$ . So a flat term $\partial_\mu\partial_\nu$ could be promoted to $\nabla_\mu\nabla_\nu$ , to $\nabla_\nu\nabla_\mu$ , or to their symmetric average, and these differ by curvature — a quantity with no flat-space template to guide the choice. Adding an explicit term proportional to $R$ or $R_{\mu\nu}$ leaves the flat limit untouched but changes the curved theory. This is why "minimal" must be pinned down as: a fixed derivative ordering plus zero explicit-curvature couplings. It is a convention adopted for simplicity, and experiment must decide whether nature respects it.

How does one action give both the matter equations and Einstein's equations?

The total action $S=\tfrac{1}{2\kappa}\int R\sqrt{-g}\,\mathrm{d}^4x+S_m[\phi,g]$ depends on two kinds of variables: the matter fields $\phi$ and the metric $g_{\mu\nu}$ . Stationarity under variations of $\phi$ (with $g$ fixed) gives the matter equations of motion — the wave equation for a scalar, Maxwell's for the photon. Stationarity under variations of $g^{\mu\nu}$ (with $\phi$ fixed) gives $G_{\mu\nu}=\kappa T_{\mu\nu}$ , with $T_{\mu\nu}$ arising from the metric-dependence of $S_m$ . Both come from one variational principle, which is why the coupling is consistent: the same $S_m$ that dictates how matter evolves also determines how it curves spacetime. The law $\nabla_\mu T^{\mu\nu}=0$ links the two — from the geometry side by the Bianchi identity, from the matter side by diffeomorphism invariance.

Why is the gravitational coupling constant κ so small, and does it matter?

$\kappa=8\pi G/c^4\approx2.08\times10^{-43}\ \mathrm{s^2\,kg^{-1}\,m^{-1}}$ is tiny because $G$ is tiny and $c^4$ is enormous, so an immense energy density is needed to produce even small curvature: the right side of $G_{\mu\nu}=\kappa T_{\mu\nu}$ is heavily suppressed. This is why gravity is by far the weakest interaction and why the metric near the Earth departs from flat by only about one part in a billion (Example 2). In practice the smallness of $\kappa$ is what lets us treat spacetime as flat for almost all laboratory physics, and it is also why detecting gravitational waves required measuring length changes of order $10^{-21}$ . The weakness of the coupling is a central quantitative fact about our universe, not a flaw in the theory.