From Special to General Relativity: The Path to the Field Equations

In 1902 the Italian geometer Luigi Bianchi published an identity about the curvature of a Riemannian manifold. It has four indices, it holds for every metric that has ever been written down or ever will be, and it had nothing whatever to do with physics. Contract it twice and it says that one particular combination of curvatures — and, up to scale, only that one — has identically vanishing divergence. Thirteen years later Einstein needed exactly such a combination, because energy and momentum are conserved, $\nabla^\mu T_{\mu\nu} = 0$ , and whatever curvature he set equal to $T_{\mu\nu}$ had to be divergence-free or the equation would contradict itself. There is only one candidate. The geometry had been waiting. That is where this chapter ends and general relativity begins — and the boundary is sharp. Special relativity handles acceleration perfectly well. Rindler observers, horizons, position-dependent clock rates, the Unruh bath: all of it lives in flat Minkowski spacetime with $R^{\rho}{}_{\sigma\mu\nu} = 0$ at every event. What special relativity cannot host is tides — the relative acceleration of neighbouring free particles that no change of coordinates removes. That one irreducible object is the whole of the difference. This article does not derive the field equations — nobody can, from inside special relativity. What it does is show you why the answer has to look the way it does, and then check it against Newton.

Key takeaways

Signature $(+,-,-,-)$ : $\mathrm{d}s^2 = c^2\mathrm{d}t^2 - \mathrm{d}x^2$ , $p\cdot p = +m^2c^2$ , timelike positive. Weak static field: $g_{00} = 1 + 2\Phi/c^2$ — with a plus. (The base for this page prints $\eta_{\mu\nu} = \mathrm{diag}(-1,1,1,1)$ and $g_{00} = -(1+2\Phi/c^2)$ : that is the opposite convention, and it is corrected throughout.)
What SR cannot do. Not acceleration — SR does acceleration fine. Tides. The tidal tensor $E_{ij} = \partial_i\partial_j\Phi$ of `aq-sr-equivalence-principle` is removable by no coordinate change, and in flat spacetime it is exactly zero. A theory of gravity must therefore make the metric dynamical.
What this article PROVES. That in the weak, static, slow-motion limit geodesic deviation reproduces Newton's tidal tensor exactly — $\;R^{i}{}_{0j0} = \frac{1}{c^2}\partial_i\partial_j\Phi\;$ , all nine components, computed symbolically — hence $R_{00} = \nabla^2\Phi/c^2$ ; and that this forces the coupling to be $8\pi G/c^4$ and nothing else.
And about the left-hand side: $R_{\mu\nu} = \kappa T_{\mu\nu}$ is impossible — its divergence plus the contracted Bianchi identity forces $T = \text{const}$ everywhere, so no localised mass could exist. $G_{\mu\nu} = R_{\mu\nu} - \tfrac12 g_{\mu\nu}R$ is the unique fix: $\nabla^\mu(R_{\mu\nu} - a\,g_{\mu\nu}R) = (\tfrac12 - a)\nabla_\nu R$ vanishes identically only at $a = \tfrac12$ .
What it does NOT prove, and does not pretend to. The field equations themselves. $G_{\mu\nu} + \Lambda g_{\mu\nu} = \frac{8\pi G}{c^4}T_{\mu\nu}$ is stated, with attribution (Einstein, 25 November 1915). Schwarzschild, geodesics, light bending, gravitational waves are the general-relativity track's — forward-linked, not annexed.
Key idea: gravity is not a force in spacetime, it is the shape of spacetime — and the reason you cannot bolt it onto special relativity is not that SR fears acceleration, but that flat spacetime has no tidal tensor to offer. Make the metric a field, demand only that it be conserved like its source, and the geometry hands you back an equation that Bianchi had already half-written in 1902.

The core insight

Every accelerated-frame effect in this chapter — the clock gradient, the horizon, the redshift, the thermal bath — was obtained by changing coordinates on flat spacetime, and can be undone by changing back. Tides cannot. The relative acceleration of two freely falling particles is a tensor, and a tensor that is zero in one chart is zero in all of them. So "is there gravity here?" has a coordinate-independent answer: a number you compute, not a frame you choose. General relativity is the theory in which that number is allowed to be non-zero.

Where this lives today

The equation is now an instrument. On 14 September 2015 the two LIGO detectors recorded a strain of about $10^{-21}$ as two black holes of roughly $36$ and $29$ solar masses merged; the waveform was matched against templates computed by solving equation (5) numerically. That the templates fit is a test of the equation in exactly the strong-field, dynamical regime where the Newtonian limit above is worthless.

And it is an engineering constraint. Every satellite-navigation fix depends on the $g_{00}$ of Step 1. `aq-sr-equivalence-principle` carries the GPS timing budget and its conventions; the point to take from this page is narrower — the correction is computed from a metric component that differs from unity in the ninth decimal place, and without it the position solution drifts by kilometres per day.

The physical picture

Take the equivalence principle at its word, as `aq-sr-equivalence-principle` sets it out: a freely falling laboratory is an inertial laboratory. Special relativity then hands you, free, every effect an accelerating observer can see. Clocks at different heights tick at different rates — `aq-sr-rindler-metric` derives it, and the Riemann tensor of that metric is zero. There is a horizon — `aq-sr-rindler-horizon` maps it, in flat spacetime. Light climbing a tower is redshifted by $gh/c^2$ — `aq-sr-gravitational-redshift` gets the Pound–Rebka number with no field equation in sight. So where does it stop? It stops the moment your laboratory is big. Two apples released side by side in a falling lift do not stay side by side: both fall toward the Earth's centre, so they converge. Released one above the other, they separate, because the lower one is pulled harder. Neither effect can be transformed away, because the transformation that removes the field at one point is not the one that removes it at the other. What survives is the tidal tensor — owned by `aq-sr-equivalence-principle`, which computes its eigenvalues $(-2GM/r^3,\, +GM/r^3,\, +GM/r^3)$ , notes that it is traceless in vacuum, and turns it into the expiry inequality that gives a $1\,\mathrm{m}$ laboratory on Earth $0.81\ \mathrm{s}$ of inertial life. `aq-sr-tidal-forces-equivalence-limits` pushes it further. Both are recalled here; neither is re-derived. Here is the crux, and it is a counting argument as much as a physical one. The tidal tensor is a second derivative of the potential. A coordinate change at a point can set the metric to $\eta_{\mu\nu}$ and kill its first derivatives — that is exactly what a freely falling frame is, and it is why the equivalence principle works. It cannot kill the second derivatives: there are simply not enough functions. In four dimensions the second derivatives of the metric have $100$ independent components; a coordinate change can dispose of $80$ of them; $20$ survive. Those twenty are the Riemann tensor, and `aq-sr-failure-global-inertial-frames` counts them. Twenty numbers that no observer can make go away. Gravity is those numbers. The argument now closes on itself. Curvature is the field, so the metric is the potential; the potential obeys a field equation; and the source is energy, which is a tensor. So there is a tensor on the right, a tensor built from second derivatives of the metric on the left, and it must be conserved, because $T_{\mu\nu}$ is. That is very nearly enough to determine it.

Figure

PUNCHLINE — the boundary between SR and GR is not acceleration, it is TIDES. LEFT: the Rindler rocket. Accelerating hard, clocks running at different rates, a horizon below — and two released test masses drift along straight, parallel lines. The Riemann tensor is ZERO at every event. Special relativity handles ALL of this. RIGHT: a lab falling toward a real mass. Same equivalence principle, same freely-falling frame — but the two test masses CONVERGE, because the field lines converge. That convergence is the tidal tensor, it is a TENSOR, and a tensor that is non-zero in one chart is non-zero in every chart. No coordinate transformation removes it. That is the whole of the difference, and it is why the metric has to become a dynamical field

The core equations

The metric stops being a fixed backdrop and becomes the field:

\mathrm{d}s^2 = g_{\mu\nu}(x)\,\mathrm{d}x^\mu \mathrm{d}x^\nu, \qquad g_{\mu\nu} \to \eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1) \ \ \text{locally} \qquad (1)

where: $g_{\mu\nu}(x)$ has ten independent components. Physical reading: at any single event one can always choose coordinates making $g_{\mu\nu} = \eta_{\mu\nu}$ with $\partial_\lambda g_{\mu\nu} = 0$ — that is the freely falling frame: the equivalence principle, in tensor calculus. What one cannot do is make $\partial_\lambda\partial_\rho g_{\mu\nu}$ vanish. Everything follows from that one sentence. The connection and the curvature built from it:

\Gamma^{\rho}{}_{\mu\nu} = \tfrac12 g^{\rho\sigma}\big(\partial_\mu g_{\sigma\nu} + \partial_\nu g_{\sigma\mu} - \partial_\sigma g_{\mu\nu}\big), \qquad R^{\rho}{}_{\sigma\mu\nu} = \partial_\mu\Gamma^{\rho}{}_{\nu\sigma} - \partial_\nu\Gamma^{\rho}{}_{\mu\sigma} + \Gamma^{\rho}{}_{\mu\lambda}\Gamma^{\lambda}{}_{\nu\sigma} - \Gamma^{\rho}{}_{\nu\lambda}\Gamma^{\lambda}{}_{\mu\sigma} \qquad (2)

where: $\Gamma$ is not a tensor and $R$ is — which is the entire difference between a coordinate effect and a physical one. Rindler coordinates on flat spacetime have non-zero $\Gamma$ (so do polar coordinates on a flat plane) and zero $R$ . `aq-sr-rindler-metric` computes both. Free fall, and what happens to two freely falling particles:

\frac{\mathrm{d}^2x^\rho}{\mathrm{d}\tau^2} + \Gamma^{\rho}{}_{\mu\nu}\frac{\mathrm{d}x^\mu}{\mathrm{d}\tau}\frac{\mathrm{d}x^\nu}{\mathrm{d}\tau} = 0, \qquad\qquad \boxed{\;\frac{D^2\xi^\mu}{\mathrm{d}\tau^2} = -R^{\mu}{}_{\nu\rho\sigma}\,u^\nu\,\xi^\rho\,u^\sigma\;} \qquad (3)

where: $\xi^\mu$ is the separation of two neighbouring geodesics. Physical reading: the geodesic equation can always be made trivial at a point by choosing the falling frame — one particle feels nothing, ever. The deviation equation cannot. It is the observable statement of gravity, and its coefficient is the Riemann tensor. Two contractions, and the one combination that is conserved:

R_{\mu\nu} = R^{\rho}{}_{\mu\rho\nu}, \qquad R = g^{\mu\nu}R_{\mu\nu}, \qquad \boxed{\;G_{\mu\nu} \equiv R_{\mu\nu} - \tfrac12 g_{\mu\nu}R, \qquad \nabla^\mu G_{\mu\nu} = 0\ \ \text{identically}\;} \qquad (4)

where: the last equality is the twice-contracted Bianchi identity, true of every metric, proved by nobody for this purpose. Physical reading: this is the hinge of the whole subject. $\nabla^\mu G_{\mu\nu} = 0$ is not a field equation, not a law of nature and not an assumption — it is a theorem of geometry. And $\nabla^\mu T_{\mu\nu} = 0$ is conservation of energy and momentum. Two vanishing divergences, and the equation writes itself.

\boxed{\;G_{\mu\nu} + \Lambda g_{\mu\nu} = \frac{8\pi G}{c^4}\,T_{\mu\nu}\;} \qquad (5)

where: $T_{\mu\nu}$ is the stress–energy tensor, $\Lambda$ the cosmological constant. STATED, NOT DERIVED — Einstein, Sitzungsberichte der Preussischen Akademie, 25 November 1915, 844–847 ( $\Lambda$ added 1917). Nothing on this page derives (5) and nothing can: it is a postulate, warranted by a century of experiment. What the next two cards establish is that the left-hand side could hardly have been anything else, and that $8\pi G/c^4$ is not a choice.

Building the equation you are not allowed to derive

Derivation AThe Newtonian limit — geodesic deviation IS Newton's tidal tensor, and it pins the coupling constant to 8 pi G over c to the fourth

A new theory earns its licence by reproducing the old one where the old one worked. The goal here is to run the weak-field, static, slow-motion limit all the way down and watch Poisson's equation come out — not as a plausibility argument but as a computation, with the signature carried explicitly, because the base for this page performs the whole calculation in the opposite convention. The pay-off is bigger than a sanity check: the constant $8\pi G/c^4$ in equation (5) was never Einstein's to choose. This limit forces it.

Step 1 — read the metric off the redshift. A clock at rest at fixed position has $\mathrm{d}s^2 = g_{00}\,(c\,\mathrm{d}t)^2$ , so its proper time is $\mathrm{d}\tau = \sqrt{g_{00}}\,\mathrm{d}t$ . `aq-sr-gravitational-time-dilation-heuristic` gets the rate ratio between two heights from the equivalence principle alone: $\mathrm{d}\tau/\mathrm{d}t = 1 + \Phi/c^2$ to first order. Squaring,

\boxed{\,g_{00} = 1 + \frac{2\Phi}{c^2}\,}

Why: note the sign. In $(+,-,-,-)$ a stationary clock must have $\mathrm{d}s^2 > 0$ , so $g_{00}$ is positive, near $+1$ . The base prints $g_{00} = -(1+2\Phi/c^2)$ and $\mathrm{d}\tau = \sqrt{-g_{00}}\,\mathrm{d}t$ — correct in $(-,+,+,+)$ , wrong here, and it hands a stationary clock an imaginary proper time. At the Earth's surface $2\Phi/c^2 = -1.39\times10^{-9}$ : the metric departs from Minkowski's in the ninth decimal place, and that is the whole of terrestrial gravity.

Step 2 — linearise and compute the curvature. Take the static weak field $g_{\mu\nu} = \eta_{\mu\nu} + h_{\mu\nu}$ with $h_{00} = 2\Phi/c^2$ , $h_{ij} = +2\Phi\delta_{ij}/c^2$ (in our signature $g_{ij} = -(1 - 2\Phi/c^2)\delta_{ij}$ ), and grind (2) to first order in $\Phi$ . Every time derivative dies. The result, for all nine spatial index pairs:

\boxed{\;R^{i}{}_{0j0} = \frac{1}{c^{2}}\,\partial_i\partial_j\Phi\;}

Why: this is the article's owned computation, and it is done symbolically in the verification script rather than quoted — all nine components, from the definition of the Riemann tensor, in $(+,-,-,-)$ . Only $h_{00}$ contributes at this order; the spatial perturbation drops out of $R^{i}{}_{0j0}$ entirely, which is why the Newtonian limit never sees it and why Newton could not have discovered light bending.

Step 3 — feed it to geodesic deviation. In the slow-motion limit $u^\mu \approx (c,0,0,0)$ , so only $\nu = \sigma = 0$ survives on the right of (3):

\frac{\mathrm{d}^2\xi^i}{\mathrm{d}t^2} = -R^{i}{}_{0j0}\,c^2\,\xi^j = -\big(\partial_i\partial_j\Phi\big)\,\xi^j .

Why: that is Newton's tidal equation, letter for letter. The object on the right is precisely the tidal tensor $E_{ij} = \partial_i\partial_j\Phi$ that `aq-sr-equivalence-principle` owns — eigenvalues $(-2GM/r^3, +GM/r^3, +GM/r^3)$ , stretch along the radius, squeeze across. We did not re-derive it; we watched general relativity hand it back. Curvature and tidal force are the same object.

Step 4 — contract, and pin the constant. Contract Step 2 on $i = j$ : $R_{00} = R^{i}{}_{0i0} = \nabla^2\Phi/c^2$ . Now take (5) with $\Lambda = 0$ in its trace-reversed form, $R_{\mu\nu} = \frac{8\pi G}{c^4}\big(T_{\mu\nu} - \tfrac12 g_{\mu\nu}T\big)$ . For dust at rest, $T_{00} = \rho c^2$ and the trace is $T = \rho c^2$ , so the right-hand side is $\frac{8\pi G}{c^4}\big(\rho c^2 - \tfrac12\rho c^2\big) = \frac{4\pi G\rho}{c^2}$ . Equate:

\frac{1}{c^{2}}\nabla^{2}\Phi = \frac{4\pi G}{c^{2}}\rho \qquad\Longrightarrow\qquad \boxed{\;\nabla^{2}\Phi = 4\pi G\rho\;} \qquad \textbf{∎}

Why: the coupling was never free. With $\kappa$ unknown this limit gives $\nabla^2\Phi = \tfrac12\kappa\rho c^4$ , and matching Poisson forces $\kappa = 8\pi G/c^4$ — uniquely. Note where the $8\pi$ comes from: $4\pi$ from Poisson, and a factor $2$ from the trace reversal — that is, from the $\tfrac12$ in the Einstein tensor. The two halves of the equation know about each other. *Verified symbolically in verify_aq-sr-special-to-general-relativity.py: sympy builds $g_{\mu\nu} = \mathrm{diag}(1+2\Phi/c^2,\ -(1-2\Phi/c^2)\delta_{ij})$ with $\Phi$ an arbitrary function of $(x,y,z)$ , computes every Christoffel symbol and every $R^{i}{}_{0j0}$ from the definition, linearises, and confirms all nine equal $\partial_i\partial_j\Phi/c^2$ ; contracts to $R_{00} = \nabla^2\Phi/c^2$ ; and solves the trace-reversed field equation for the coupling, returning $8\pi G/c^4$ exactly.*

Go deeper · Derivation B · 2 minWhy the Einstein tensor and not the Ricci tensor — a three-line proof that the obvious equation forbids stars

Derivation BWhy the Einstein tensor and not the Ricci tensor — a three-line proof that the obvious equation forbids stars

The natural guess, and the one Einstein spent two years on, is $R_{\mu\nu} = \kappa T_{\mu\nu}$ : curvature on the left, matter on the right, one index pair each, dimensions fine. It is wrong, and not by a subtlety — it can be killed in three lines, and the demonstration is worth more than the answer, because it shows the structure of the field equations is forced by conservation of energy alone. This card gives the disproof, then shows the repair is unique.

Step 1 — write down the only geometric fact you need. The twice-contracted Bianchi identity, equation (4):

\nabla^\mu\Big(R_{\mu\nu} - \tfrac12 g_{\mu\nu}R\Big) = 0 \qquad\Longleftrightarrow\qquad \nabla^\mu R_{\mu\nu} = \tfrac12\,\nabla_\nu R .

Why: this is a theorem, not a physical law — true of every metric on every manifold, with no reference to matter, gravity or dynamics. Bianchi published it in 1902; Voss had it in 1880 and Ricci around 1889. It sat in the geometry literature for over a decade before a physicist needed it.

Step 2 — take the divergence of the naive equation. Suppose $R_{\mu\nu} = \kappa T_{\mu\nu}$ . Local conservation of energy and momentum says $\nabla^\mu T_{\mu\nu} = 0$ — this is not negotiable; it is what "the source is conserved" means. Then $\nabla^\mu R_{\mu\nu} = \kappa\,\nabla^\mu T_{\mu\nu} = 0$ , and Step 1 turns that into

\tfrac12\,\nabla_\nu R = 0 \qquad\Longrightarrow\qquad R = \text{constant throughout spacetime}.

Why: nothing has been assumed about the matter, and the conclusion is already alarming — the Ricci scalar, a local and measurable curvature, would have to be the same number everywhere in the universe.

Step 3 — trace the naive equation and collect the wreckage. Contracting $R_{\mu\nu} = \kappa T_{\mu\nu}$ with $g^{\mu\nu}$ gives $R = \kappa T$ . With $R$ constant, $T$ is constant. For dust $T = \rho c^2$ , so

\rho = \text{constant everywhere.} \qquad \textbf{∎}

Why: the naive field equation forbids a star. It forbids a planet, a proton and an empty room; it admits only a uniform unchanging dust filling all of space. The equation is not defective, it is dead. And notice what killed it: not experiment, not the Newtonian limit, not aesthetics. Conservation of energy killed it, with a nineteenth-century identity as the weapon.

Step 4 — the repair, and its uniqueness. Look for the fix in the family $R_{\mu\nu} - a\,g_{\mu\nu}R = \kappa T_{\mu\nu}$ . Taking the divergence with Step 1:

\nabla^\mu\big(R_{\mu\nu} - a\,g_{\mu\nu}R\big) = \tfrac12\nabla_\nu R - a\,\nabla_\nu R = \Big(\tfrac12 - a\Big)\nabla_\nu R .

This vanishes for arbitrary $R$ if and only if $a = \tfrac12$ . Why: the $\tfrac12$ in the Einstein tensor is not a convention. It is the unique number for which the left-hand side is automatically conserved — hence the unique number for which the field equation does not contradict the conservation of its own source. Add $\Lambda g_{\mu\nu}$ if you like: $\nabla^\mu g_{\mu\nu} = 0$ by metric compatibility, so it costs nothing. That is exactly why $\Lambda$ can be there, and why nothing in the structure forbids it. *Verified symbolically in verify_aq-sr-special-to-general-relativity.py: the identity $\nabla^\mu(R_{\mu\nu} - a\,g_{\mu\nu}R) = (\tfrac12 - a)\nabla_\nu R$ is confirmed on a symbolic metric; $a = \tfrac12$ is shown to be the unique root; and the $R_{\mu\nu} = \kappa T_{\mu\nu}$ chain $\Rightarrow R = \text{const} \Rightarrow T = \text{const}$ is executed step by step.*

Common trapThis article does NOT derive the field equations, and neither does any other article in this track

Equation (5) is stated with attribution and tested against Newton. It is not derived here, and it cannot be: no amount of special relativity, however cleverly deployed, produces a field equation for the metric. What is established above is strictly this — the source must be $T_{\mu\nu}$ ; the left-hand side must be divergence-free; $G_{\mu\nu}$ is the unique second-derivative curvature tensor with that property; and the coupling is then forced to $8\pi G/c^4$ by the Newtonian limit. That is a strong constraint, and it is not a derivation. Einstein's own route was three years of trial, error and one retracted theory.

Forward-linked, not annexed — the general-relativity track owns: the Schwarzschild solution, black holes, Mercury's perihelion, light deflection and lensing, gravitational waves, cosmology, and the field equations proper. This chapter's job is to show you exactly where special relativity stops. It stops here.

Common trapThe signature, and four cards in the base that are written in the other one

Our lock is $(+,-,-,-)$ : $\mathrm{d}s^2 = c^2\mathrm{d}t^2 - \mathrm{d}x^2$ , timelike positive. The AllQFT base for this page uses the opposite convention in four places — its overview prints $\eta_{\mu\nu} = \mathrm{diag}(-1,1,1,1)$ ; its limits card prints $g_{00} \approx -(1 + 2\Phi/c^2)$ ; its time-dilation derivation scales proper time as $\sqrt{-g_{00}}$ ; its geodesic card extremises $\int\sqrt{-g_{\mu\nu}\mathrm{d}x^\mu\mathrm{d}x^\nu}$ . Each is right in $(-,+,+,+)$ and a sign error here. (The lane's GATE 0 audit recorded this base as carrying zero hard defects; these four were missed.) The four-second test: ask what a clock at rest does. In $(+,-,-,-)$ it must have $\mathrm{d}s^2 > 0$ . If your $g_{00}$ is negative, your stationary clock is spacelike and its proper time imaginary — you have imported another convention's formula into this one's notation. Re-derive; never transcribe.

Figure

The coupling constant, drawn to scale — and why it took a star to notice gravity. The bar is log₁₀ of the energy density needed to produce a given spacetime curvature, from G_μν = (8πG/c⁴) T_μν with 8πG/c⁴ = 2.0766 × 10⁻⁴³ N⁻¹. To curve spacetime by 1 m⁻² you must supply 4.82 × 10⁴² J/m³. Marked, left to right: the strongest steady laboratory magnetic field (~10⁷ J/m³), the core of the Sun (~10¹⁴), a neutron star (~10³³) — and the requirement, four orders of magnitude beyond even that. THE PUNCHLINE IS THE GAP: 8πG/c⁴ is an INVERSE FORCE, and its reciprocal c⁴/8πG ≈ 4.8 × 10⁴² newtons is the stiffness of spacetime. Gravity is not weak because its charge is small. It is weak because the thing it bends is the most rigid object in physics — and it wins anyway, because energy density never comes with a minus sign and so gravity never screens

Limiting cases

Case	Result	Meaning
$R^{\rho}{}_{\sigma\mu\nu} = 0$	$g_{\mu\nu} \to \eta_{\mu\nu}$ globally	GR is special relativity. All of this chapter's accelerated-frame physics lives here
$\Phi/c^2 \ll 1$ , static, slow	$g_{00} = 1 + 2\Phi/c^2$ ; $\nabla^2\Phi = 4\pi G\rho$	Newton, recovered — and the coupling forced to $8\pi G/c^4$
neighbourhood of that point	$\partial\partial g \neq 0$ : 20 components	never removable. This is gravity, and it is why the principle is local
$T_{\mu\nu} = 0$ , $\Lambda = 0$	$R_{\mu\nu} = 0$ , not $R^{\rho}{}_{\sigma\mu\nu} = 0$	vacuum is not flat. Gravitational waves and Schwarzschild live in this gap — GR track

Worked examples

1 · How stiff is spacetime?

Equation (5) reads "curvature = constant × energy density". Put numbers on the constant and ask what it costs to bend spacetime by an appreciable amount.

Step 1 — identify. $G_{\mu\nu}$ has dimensions of (length) $^{-2}$ , being second derivatives of a dimensionless metric; $T_{\mu\nu}$ is an energy density, $\mathrm{J\,m^{-3}}$ . So $8\pi G/c^4$ carries units $\mathrm{m^{-2}}/(\mathrm{J\,m^{-3}}) = \mathrm{N^{-1}}$ : the coupling constant of gravity is an inverse force.

Step 2 — the boundary number. With $G = 6.67430\times10^{-11}$ and $c = 2.99792458\times10^{8}$ :

\frac{8\pi G}{c^{4}} = \frac{1.67743\times10^{-9}}{8.07761\times10^{33}} = \mathbf{2.0766\times10^{-43}\ N^{-1}} .

Step 3 — invert it. To produce a curvature of $1\ \mathrm{m^{-2}}$ — spacetime bent on the scale of a metre — you need an energy density of $1/(2.0766\times10^{-43}) = \mathbf{4.82\times10^{42}\ J\,m^{-3}}$ . Laboratory sources fall some forty orders of magnitude short.

Step 4 — what it means. Spacetime is the stiffest thing in physics. The reciprocal of the coupling, $c^4/8\pi G \approx 4.8\times10^{42}\ \mathrm{N}$ , is a force so large that the entire electromagnetic content of a laboratory is not a rounding error on it. This is why gravity is the weakest force and yet the one that shapes the universe: feeble per unit source, and the only one that never cancels — there is no negative energy density to screen it. It is also why the Earth's metric departs from $\eta_{\mu\nu}$ only in the ninth decimal place, $2\Phi/c^2 = -1.39\times10^{-9}$ , and yet holds you to the floor.

2 · Newton, from the top

Take the field equation on trust for one paragraph and get the inverse-square law back out — including the numerical factor.

Step 1 — identify. Outside a static spherical mass $M$ , the source vanishes: $T_{\mu\nu} = 0$ . Take $\Lambda = 0$ . The trace-reversed equation gives $R_{\mu\nu} = 0$ , so certainly $R_{00} = 0$ .

Step 2 — apply the limit. By the first derivation, $R_{00} = \nabla^2\Phi/c^2$ in the weak static field. So $\nabla^2\Phi = 0$ in vacuum — Laplace's equation, whose spherically symmetric decaying solution is $\Phi = -GM/r$ , with $GM$ fixed by matching to the interior, where $\nabla^2\Phi = 4\pi G\rho$ integrates to $\oint\nabla\Phi\cdot\mathrm{d}\mathbf{A} = 4\pi GM$ .

Step 3 — the boundary number. So $g_{00} = 1 - 2GM/(rc^2)$ . At the Earth's surface, $GM_\oplus = 3.986\times10^{14}\ \mathrm{m^3\,s^{-2}}$ and $R_\oplus = 6.371\times10^{6}\ \mathrm{m}$ :

\frac{2GM_\oplus}{R_\oplus c^{2}} = \frac{2\times3.986\times10^{14}}{6.371\times10^{6}\times8.98755\times10^{16}} = \mathbf{1.39\times10^{-9}} .

A clock at sea level runs slow by $6.96\times10^{-10}$ — about $22\ \mathrm{ms}$ per year — relative to one at infinity.

Step 4 — the boundary, and the honesty. That $g_{00}$ is the first two terms of the Schwarzschild metric, and it is as far as this article is entitled to go. The exact solution, the horizon at $r = 2GM/c^2$ , the perihelion advance and the bending of light all need the full field equations solved exactly, not their Newtonian limit — and they belong to the general-relativity track. What we have shown is that the limit is clean: general relativity contains Newton, with the right coefficient, and it does so because the coupling was fixed to make it so. A theory that reproduces its predecessor is not thereby confirmed. It is merely permitted to continue.

Physics of the portrait

Origins · 2 minEinstein field equations and how this was discovered

Einstein field equations

1907–1915 · the happiest thought · the Entwurf · Göttingen · four Wednesdays in November

The eight-year detour. In 1907, writing a review for the Jahrbuch der Radioaktivität und Elektronik, Einstein had what he later called the happiest thought of his life: a man in free fall does not feel his own weight. From it he extracted gravitational time dilation and the redshift — the content of `aq-sr-rindler-metric` — eight years before he had a field equation, with nothing but special relativity and an accelerating box. The remaining eight years were spent learning that the box is not enough. The wrong turn. With Marcel Grossmann he arrived in 1913 at the Entwurf theory, whose field equations were not generally covariant, and defended it with the "hole argument" — convincing himself general covariance was inconsistent with determinism. It cost him two years. Four Wednesdays. In November 1915 he presented to the Prussian Academy on four consecutive Wednesdays, correcting himself as he went. The fourth, on 25 November 1915 — "Die Feldgleichungen der Gravitation", Sitzungsberichte, 844–847 — is the equation. David Hilbert, in Göttingen and in correspondence with him, presented his own variational derivation on 20 November; the priority question has been argued over ever since, and the physics is Einstein's. What was already waiting. The identity guaranteeing $\nabla^\mu G_{\mu\nu} = 0$ — the one thing that makes the equation consistent with the conservation of its own source — had been in the geometry literature since Voss (1880), was found again by Ricci (about 1889), and was published by Luigi Bianchi in 1902. No physicist commissioned it. Einstein reached $G_{\mu\nu}$ by other means; the geometry had got there first.

History

1880

— Aurel Voss obtains the differential identity for the Riemann tensor later attributed to Bianchi; Ricci finds it again around 1889 and forgets it.

1902

— Luigi Bianchi publishes the identities that bear his name. Their twice-contracted form makes one combination of curvatures identically divergence-free. No physical application is in view.

1915

— In November Einstein presents four papers to the Prussian Academy on successive Wednesdays; the fourth, on 25 November, gives the field equations in their final form (Sitzungsberichte, 844-847). Hilbert presents a variational derivation in Göttingen on 20 November.

1916

— Einstein's Annalen der Physik review, Die Grundlage der allgemeinen Relativitätstheorie, sets out the completed theory; Schwarzschild solves the equations exactly within months. De Sitter derives the geodetic precession of a gyroscope in orbit.

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §§17.1–17.4 · Wald, General Relativity, §4.3 · Hartle, Gravity, chs. 21–22 · Rindler, Relativity: Special, General and Cosmological, 2nd ed., ch. 10 · Carroll, Spacetime and Geometry, ch. 4. Primary: Einstein, Sitzungsber. Preuss. Akad. Wiss. (25 Nov 1915), 844–847; Einstein, Ann. Phys. 49, 769 (1916).

Provenance. Signature $(+,-,-,-)$ . Every equation and number above is recomputed in verify_aq-sr-special-to-general-relativity.py, which also reads this file and asserts each printed value against what it computed. Base defects repaired: (i) the base is mostly-plus in four places — $\eta_{\mu\nu} = \mathrm{diag}(-1,1,1,1)$ (overview), $g_{00} \approx -(1+2\Phi/c^2)$ (limits), $\sqrt{-g_{00}}$ (time-dilation derivation), $\sqrt{-g_{\mu\nu}\mathrm{d}x^\mu\mathrm{d}x^\nu}$ (geodesic card); the lane's GATE 0 audit recorded this base as carrying zero hard defects and missed all four. (ii) Its Pound–Rebka figure of $2.5\times10^{-15}$ is loose — `aq-sr-gravitational-redshift` owns that number and prints $2.46\times10^{-15}$ . (iii) It gives the GPS budget as " $\sim45$ "/" $\sim7$ " $\mu\mathrm{s}$ /day; the corpus ruling is $+45.8$ / $-7.2$ /net $+38.6$ and `aq-sr-equivalence-principle` owns the framing, so this page does not print a third version.

Scope. The equivalence principle, the tidal tensor $E_{ij}$ and the expiry inequality are owned by `aq-sr-equivalence-principle`; the count of $20$ by `aq-sr-failure-global-inertial-frames`; the flat-space clock gradient and horizon by `aq-sr-rindler-metric` and `aq-sr-rindler-horizon`. All recalled, none re-derived. Schwarzschild, geodesics, light bending, black holes and gravitational waves are the general-relativity track's — forward-linked, deliberately not claimed here.

Builds on. The Equivalence Principle · Tidal Forces and the Limits of the Equivalence Principle · Gravitational Time Dilation · The Rindler Metric. Leads to. Gravitational Redshift · The Rindler Horizon · the general-relativity track.

If special relativity handles acceleration, why is there a general relativity at all?

Because gravity is not acceleration. Every accelerated-frame effect in this chapter — the clock gradient, the horizon, the redshift, the Unruh bath — comes from changing coordinates on flat spacetime, and can be undone by changing back; `aq-sr-rindler-metric` computes the Riemann tensor and finds zero. What no change of coordinates can undo is the tidal field: two apples released side by side in a falling lift converge. That convergence is a tensor, and a tensor non-zero in one chart is non-zero in all of them. General relativity is not needed for an accelerating rocket. It is needed for the twenty numbers that survive after you have spent all your coordinate freedom.

Did Einstein derive the field equations, or guess them?

Neither word is right, and the honest answer is instructive. He constrained them, hard, over three years — general covariance, the Newtonian limit, conservation of energy, the vanishing of the source in vacuum — and found the unique tensor meeting the constraints, having first spent two years on a non-covariant theory (the Entwurf) defended with an argument he later saw was wrong. He did not reach $G_{\mu\nu}$ via the Bianchi identity; the recognition that the identity guarantees the consistency he had been fighting for came afterwards. What this page reconstructs is not Einstein's path but the logic that makes the destination inevitable — a different, and for a reader more useful, thing.

Why does the cosmological constant not spoil the conservation argument?

Because the metric is covariantly constant: $\nabla^\mu g_{\mu\nu} = 0$ , identically, by the definition of the Levi-Civita connection. Adding $\Lambda g_{\mu\nu}$ to a divergence-free left-hand side leaves it divergence-free, for any constant $\Lambda$ . This is why $\Lambda$ is allowed — and why nothing in the structure tells you its value, or even that it should be small. Einstein put it in for a static universe in 1917 and regretted it; observation put it back in 1998. The consistency argument that kills $R_{\mu\nu} = \kappa T_{\mu\nu}$ has nothing to say against $\Lambda$ , and that silence is one of the deepest open problems in physics.

Vacuum means R_uv = 0. Does that mean empty space is flat?

No, and the distinction is where most of general relativity lives. $R_{\mu\nu} = 0$ is a contraction of the full Riemann tensor, and in four dimensions the contraction throws information away: $R_{\mu\nu}$ has $10$ independent components, $R^{\rho}{}_{\sigma\mu\nu}$ has $20$ . The other $10$ — the Weyl tensor — are completely unconstrained by the vacuum equation. That is the room in which the Schwarzschild geometry outside a star, the tidal field that stretches you into a black hole, and gravitational waves through empty space all live. Newtonian gravity has no analogue: there, $\nabla^2\Phi = 0$ in vacuum and the field outside a mass is fixed entirely by the mass. In general relativity, empty space carries curvature of its own, travelling at $c$ .

Where exactly does the factor 8 pi come from?

From two places, worth separating. The $4\pi$ is Newton's: the $4\pi$ of $\nabla^2\Phi = 4\pi G\rho$ , itself the surface area of the unit sphere showing up in Gauss's law. The extra factor $2$ comes from the trace reversal — from the $\tfrac12$ inside the Einstein tensor. Write the field equation as $R_{\mu\nu} = \frac{8\pi G}{c^4}(T_{\mu\nu} - \tfrac12 g_{\mu\nu}T)$ , put dust on the right, and the $00$ component gives $\rho c^2 - \tfrac12\rho c^2 = \tfrac12\rho c^2$ ; that halving is exactly what turns $8\pi$ into the $4\pi$ Newton needs. So the $\tfrac12$ that conservation forces into $G_{\mu\nu}$ and the factor $2$ the Newtonian limit needs in the coupling are the same $\tfrac12$ , appearing twice. Not a coincidence — the equation being consistent with itself.

Why twenty components? Where does the number come from?

Count. The second derivatives of the metric $\partial_\lambda\partial_\rho g_{\mu\nu}$ have $10$ (for $g_{\mu\nu}$ ) $\times\ 10$ (for the symmetric pair $\lambda\rho$ ) $= 100$ independent numbers at a point. A coordinate change $x \to x'(x)$ can be Taylor-expanded, and its third-order coefficients are what act on the second derivatives of the metric: there are $4 \times 20 = 80$ of them. So $100 - 80 = 20$ numbers cannot be touched. Independently, the Riemann tensor's own symmetries in $n$ dimensions leave $n^2(n^2-1)/12$ components, which at $n = 4$ is $16 \times 15/12 = 20$ . The two counts agree, and that agreement is the theorem: the irreducible content of the second derivatives of the metric is the Riemann tensor. `aq-sr-failure-global-inertial-frames` does the counting properly.