Killing Vectors and Conserved Quantities in Schwarzschild

Two numbers do almost all the work in Schwarzschild orbital mechanics: the specific energy $E$ and the specific angular momentum $L$ . This article is about why they are constant, and what they mean. The answer is one of the deepest ideas in geometry — that a continuous symmetry of spacetime is not a convenience but a conservation law. A Killing vector field $K^\mu$ points along a direction in which the metric is literally unchanged, an infinitesimal isometry encoded by Killing's equation $\nabla_\mu K_\nu + \nabla_\nu K_\mu = 0$ . The payoff is a one-line theorem: along any geodesic with tangent $\dot x^\mu$ , the scalar $K_\mu \dot x^\mu$ is conserved — symmetry becomes a constant of motion, exactly as ignorable coordinates in Lagrangian mechanics produce conserved momenta. It is Noether's theorem in geometric dress. Schwarzschild spacetime carries the maximal symmetry compatible with a static, asymptotically flat vacuum: independent of $t$ and of $\phi$ , and spherically symmetric, with a four-dimensional isometry group $\mathbb{R}_t \times SO(3)$ . The time-translation vector $\xi = \partial_t$ delivers the conserved energy $E = -\xi_\mu \dot x^\mu$ ; the axial vector $\eta = \partial_\phi$ delivers the angular momentum $L = \eta_\mu \dot x^\mu$ . Spherical symmetry lets us rotate any single orbit into the equatorial plane $\theta = \pi/2$ , and with $E$ , $L$ , and the normalization $g_{\mu\nu}\dot x^\mu \dot x^\nu = -\epsilon$ fixed, the geodesic problem collapses to one first-order radial equation of energy-conservation form. Every downstream prediction — the effective potential, the ISCO at $r=6GM$ , the photon sphere, perihelion precession, gravitational redshift — is built on these two constants and the reduction they make possible.

Key takeaways

Schwarzschild spacetime is static and spherically symmetric, carrying four Killing vector fields: the timelike $\xi^\mu=(\partial_t)^\mu$ and the three rotational generators of $SO(3)$ .
Each Killing vector $K^\mu$ supplies a quantity $K_\mu \dot x^\mu$ conserved along every geodesic, because $\nabla_{(\mu}K_{\nu)}=0$ kills the only surviving term.
Time-translation gives the conserved specific energy $E=(1-2GM/r)\,\dot t$ ; axial rotation gives the specific angular momentum $L=r^2\sin^2\theta\,\dot\phi$ .
Spherical symmetry confines each orbit to a plane ( $\theta=\pi/2$ ), reducing the dynamics to one radial equation $\dot r^2 = E^2-(1-2GM/r)(\epsilon+L^2/r^2)$ governed by an effective potential.
$E$ and $L$ are global quantities defined by the asymptotic symmetries — energy at infinity, not what a local hovering observer measures; the two differ by exactly the redshift factor $\sqrt{1-2GM/r}$ .
These two constants are the gateway to the effective potential, the ISCO at $r=6GM$ (accretion efficiency $\approx 5.7\%$ ), the photon sphere at $r=3GM$ , and Mercury's $43''$ per century.
Key idea: every Killing vector $K^\mu$ of a spacetime gives a conserved $K_\mu \dot x^\mu$ along geodesics; in Schwarzschild the timelike $\partial_t$ gives the energy $E$ and the axial $\partial_\phi$ gives the angular momentum $L$ , reducing orbital dynamics to a single radial energy equation.

The core insight

Every Killing vector $K^\mu$ of a spacetime gives a conserved quantity $K_\mu \dot x^\mu$ along geodesics; in Schwarzschild the timelike Killing vector $\partial_t$ gives the energy $E$ and the axial Killing vector $\partial_\phi$ gives the angular momentum $L$ , reducing orbital dynamics to a single radial energy equation.

Where this lives today

The two constants of this article are the daily working tools of relativistic astrophysics. In X-ray binaries and active galactic nuclei, matter spirals toward a black hole through a thin disk, and the ISCO energy $E=\sqrt{8/9}$ sets the $\sim5.7\%$ radiative efficiency that lets accreting holes outshine whole galaxies as quasars — a Kerr hole reaches toward $\sim42\%$ , and fitting the disk continuum to this efficiency is one of the main ways astronomers measure black-hole spin. The photon-sphere impact parameter $b_c=3\sqrt3\,GM$ is the radius of the dark "shadow" the Event Horizon Telescope resolved around M87\ in 2019 and Sagittarius A\ in 2022; its angular size is a direct readout of $GM/c^2$ over distance, a test of null-geodesic structure built entirely on $L/E$ . The relativistic correction to $E$ and $L$ produces the $43''$ -per-century anomaly in Mercury's perihelion, and the same constants for photons give the light-bending and Shapiro delay that deep-space navigation must correct for. And in gravitational-wave astronomy a binary's inspiral is a quasi-adiabatic sequence of orbits losing $E$ and $L$ to radiation — the waveform LIGO matches is a ledger of how these two Killing charges bleed away.

The physical picture

A Killing vector marks a direction in which the geometry "looks the same." Slide the whole Schwarzschild geometry along the time direction and nothing changes — the metric carries no $t$ — so no agency does work along that direction and the conjugate momentum, the energy, cannot change. Rotate about the polar axis and again nothing changes, fixing the axial angular momentum. This is Noether's theorem told geometrically: the conserved quantity comes from a symmetry of spacetime itself, not of some external potential. The first figure shows the two symmetries side by side, each producing its constant.

Figure

Two symmetries, two constants — the rotational Killing vector ∂φ makes L = r² dφ/dτ conserved, and the time-translation vector ∂t makes E = (1 − 2GM/r) dt/dτ conserved

Spherical symmetry does two jobs at once. It collapses the three rotational constants into one conserved magnitude — the total angular momentum — and it lets you rotate the frame so any orbit lies flat in the equatorial plane $\theta=\pi/2$ , where it stays for all time. The motion is then effectively $1+1$ dimensional: a particle moving radially in an effective potential built from $L$ , with $E$ the projection of the four-momentum onto the timelike Killing vector. The second figure is the payoff: the effective potential whose shape encodes every kind of orbit.

Figure

The effective potential V(r) for a massive particle with specific angular momentum L = 4GM (units G = M = c = 1), showing a stable circular orbit at the minimum r = 12GM and an unstable one at the maximum r = 4GM — as L drops to 2√3 GM these merge at the ISCO r = 6GM

The core equations

An isometry leaves the metric invariant. Made infinitesimal, this means the Lie derivative of $g_{\mu\nu}$ along the generating vector vanishes; converting partials to covariant derivatives turns that condition into Killing's equation — the definition of a Killing vector field.

\nabla_\mu K_\nu + \nabla_\nu K_\mu = 0 \qquad (1)

where: $K_\mu$ — Killing covector field generating a continuous isometry; $\nabla_\mu$ — Levi-Civita covariant derivative. Reading: only the symmetric part of $\nabla_\mu K_\nu$ must vanish; the antisymmetric part is unconstrained and encodes the local "rotation" of the flow. The reason a Killing vector matters for orbits is a single theorem: contract the geodesic tangent with a Killing vector and the resulting scalar is constant along the geodesic. The proof below needs both Killing's equation and the geodesic equation — one alone is not enough.

\frac{\mathrm{d}}{\mathrm{d}\lambda}\left(K_\mu \dot x^\mu\right) = 0 \qquad (2)

where: $\dot x^\mu = \mathrm{d}x^\mu/\mathrm{d}\lambda$ — geodesic tangent; $\lambda$ — affine parameter, equal to proper time $\tau$ for a massive particle. Reading: each independent symmetry hands you one first integral of the geodesic equation. To apply the theorem we need the metric and its symmetries explicitly. The Schwarzschild line element is static (no $t$ ), axisymmetric (no $\phi$ ), and spherically symmetric.

\mathrm{d}s^2 = -\left(1-\frac{2GM}{r}\right)\mathrm{d}t^2 + \left(1-\frac{2GM}{r}\right)^{-1}\mathrm{d}r^2 + r^2\left(\mathrm{d}\theta^2 + \sin^2\theta\,\mathrm{d}\phi^2\right) \qquad (3)

where: $M$ — mass of the central body; $r,\theta,\phi$ — Schwarzschild radial and angular coordinates; $t$ — asymptotic time coordinate. Reading: the components depend on $r,\theta$ only, so $\partial_t$ and $\partial_\phi$ are manifest symmetry directions. Because the metric is independent of $t$ and $\phi$ , the basis vectors along those directions are Killing vectors — the two that generate the orbital constants. (The full isometry group has four generators; the other two rotations are spent choosing the orbital plane.)

\xi^\mu = (\partial_t)^\mu = (1,0,0,0), \qquad \eta^\mu = (\partial_\phi)^\mu = (0,0,0,1) \qquad (4)

where: $\xi^\mu$ — timelike Killing vector, norm $\xi_\mu\xi^\mu = -(1-2GM/r)$ ; $\eta^\mu$ — axial Killing vector (rotation about the polar axis). Reading: $\xi$ is timelike only for $r>2GM$ , null at the horizon, spacelike inside. Contracting $\xi$ with the four-velocity gives the conserved energy per unit rest mass; the minus sign makes $E$ positive for future-directed motion outside the horizon.

E = -\xi_\mu \dot x^\mu = \left(1-\frac{2GM}{r}\right)\frac{\mathrm{d}t}{\mathrm{d}\tau} \qquad (5)

where: $E$ — specific energy (per unit rest mass), the value booked by an observer at infinity; $\tau$ — proper time. Reading: at large $r$ , $E\to 1 + \tfrac12 v^2 - GM/r$ , the Newtonian energy plus the rest-mass term $1$ . Contracting $\eta$ with the four-velocity gives the conserved angular momentum per unit mass; in the equatorial plane $\sin\theta=1$ and it takes its Keplerian form.

L = \eta_\mu \dot x^\mu = r^2 \sin^2\theta\,\frac{\mathrm{d}\phi}{\mathrm{d}\tau} \;\xrightarrow{\;\theta=\pi/2\;}\; r^2 \frac{\mathrm{d}\phi}{\mathrm{d}\tau} \qquad (6)

where: $L$ — specific angular momentum; $\theta=\pi/2$ — equatorial plane fixed by spherical symmetry. Reading: $L$ is the relativistic version of Kepler's $r^2\dot\phi$ . The last ingredient is the mass-shell normalization of the four-velocity, which fixes the scale and separates massive particles from photons.

g_{\mu\nu}\dot x^\mu \dot x^\nu = -\epsilon, \qquad \epsilon = \begin{cases}1 & \text{timelike (massive)}\\ 0 & \text{null (photon)}\end{cases} \qquad (7)

where: $\epsilon=1$ for a massive particle parametrized by proper time, $0$ for a photon on an affine parameter. Reading: not an extra assumption — it is the definition of proper-time (or affine) parametrization. Feeding $E$ , $L$ , and the normalization into one another eliminates $\dot t$ and $\dot\phi$ , leaving a single radial equation with the structure of Newtonian energy conservation.

\frac{1}{2}\left(\frac{\mathrm{d}r}{\mathrm{d}\tau}\right)^2 + V_{\rm eff}(r) = \frac{1}{2}E^2,\qquad V_{\rm eff}(r)=\frac{1}{2}\left(1-\frac{2GM}{r}\right)\left(\epsilon + \frac{L^2}{r^2}\right) \qquad (8)

where: $V_{\rm eff}$ — effective potential from the conserved $L$ and normalization $\epsilon$ . Reading: turning points are where $\dot r = 0$ ( $V_{\rm eff}=\tfrac12 E^2$ ); extrema of $V_{\rm eff}$ are circular orbits, stable at minima, unstable at maxima.

Derivation AFrom a symmetry to a constant of motion

The goal of this derivation is to prove the central theorem that makes the whole subject work: contracting the tangent of a geodesic with any Killing vector produces a scalar that does not change along the curve. We differentiate $K_\mu \dot x^\mu$ and watch two terms fall away for two independent reasons — one because Killing's equation kills the symmetric part of $\nabla K$ , the other because the geodesic equation zeroes the acceleration. Both conditions are needed, and seeing exactly where each enters is the whole insight.

Step 1. Write the rate of change of the scalar as a directional covariant derivative, $\dfrac{\mathrm{d}}{\mathrm{d}\lambda}(K_\mu \dot x^\mu) = \dot x^\nu \nabla_\nu (K_\mu \dot x^\mu)$ . Why: for a scalar the derivative along the curve equals the covariant derivative contracted with the tangent, so we may use the Leibniz rule of $\nabla$ .

Step 2. Apply the product rule: $\dot x^\nu \nabla_\nu (K_\mu \dot x^\mu) = \dot x^\nu \dot x^\mu \nabla_\nu K_\mu + K_\mu\, \dot x^\nu \nabla_\nu \dot x^\mu$ . Why: this splits the change into a piece probing how $K$ varies and a piece probing how the tangent accelerates — governed by two different equations.

Step 3. The second term vanishes by the geodesic equation, $\dot x^\nu \nabla_\nu \dot x^\mu = 0$ . Why: a geodesic has zero four-acceleration; on a non-geodesic this term survives and $K_\mu\dot x^\mu$ is not conserved.

Step 4. The first term vanishes by Killing's equation. Since $\dot x^\nu \dot x^\mu$ is symmetric, only the symmetric part of $\nabla_\nu K_\mu$ contributes: $\dot x^\nu \dot x^\mu \nabla_\nu K_\mu = \dot x^\nu \dot x^\mu \nabla_{(\nu} K_{\mu)} = 0$ . Why: a symmetric object contracted with an antisymmetric one gives zero, and Killing's equation states exactly $\nabla_{(\mu}K_{\nu)}=0$ . Both terms gone, $\mathrm{d}(K_\mu\dot x^\mu)/\mathrm{d}\lambda = 0$ . The result holds for any Killing vector on any spacetime.

Go deeper · Derivation B · 1 minEnergy and angular momentum from the geodesic Lagrangian

Derivation BEnergy and angular momentum from the geodesic Lagrangian

The goal here is to obtain the explicit forms $E=(1-2GM/r)\dot t$ and $L=r^2\sin^2\theta\,\dot\phi$ by the fastest route — the geodesic Lagrangian — and to show these conjugate momenta are exactly the Killing contractions $\xi_\mu\dot x^\mu$ and $\eta_\mu\dot x^\mu$ , so the "ignorable coordinate" picture and the "Killing vector" picture are the same statement written twice.

Step 1. Take the geodesic Lagrangian $\mathcal{L}=\tfrac12 g_{\mu\nu}\dot x^\mu \dot x^\nu$ , which in Schwarzschild reads $\mathcal{L}=\tfrac12[-(1-2GM/r)\dot t^2 + (1-2GM/r)^{-1}\dot r^2 + r^2\dot\theta^2 + r^2\sin^2\theta\,\dot\phi^2]$ , with the dot meaning $\mathrm{d}/\mathrm{d}\tau$ . Why: extremizing this $\mathcal{L}$ reproduces the geodesic equation, and its metric coefficients visibly contain no $t$ and no $\phi$ .

Step 2. Since $\partial\mathcal{L}/\partial t = 0$ , Euler–Lagrange gives $\dfrac{\mathrm{d}}{\mathrm{d}\tau}\dfrac{\partial\mathcal{L}}{\partial\dot t}=0$ , so $p_t=\partial\mathcal{L}/\partial\dot t = -(1-2GM/r)\dot t$ is constant. Define $E\equiv -p_t$ . Why: an ignorable coordinate makes its conjugate momentum constant; the minus sign makes $E>0$ for future-directed geodesics outside the horizon.

Step 3. Likewise $\partial\mathcal{L}/\partial\phi = 0$ gives the conserved $p_\phi = \partial\mathcal{L}/\partial\dot\phi = r^2\sin^2\theta\,\dot\phi \equiv L$ . Why: axisymmetry is the second ignorable coordinate, its momentum the axial angular momentum.

Step 4. Identify the momenta with Killing contractions: lowering an index, $\xi_\mu\dot x^\mu = g_{tt}\dot t = -(1-2GM/r)\dot t = p_t$ and $\eta_\mu\dot x^\mu = g_{\phi\phi}\dot\phi = r^2\sin^2\theta\,\dot\phi = p_\phi$ . Why: the "conserved momentum of an ignorable coordinate" is exactly "the four-velocity contracted with the Killing vector," so the two derivations coincide. Hence $E=(1-2GM/r)\dot t$ and $L=r^2\sin^2\theta\,\dot\phi$ . The sign convention $E=-\xi\cdot u$ gives $E>0$ outside the horizon.

Go deeper · Derivation C · 1 minCollapsing the orbit to one radial equation

Derivation CCollapsing the orbit to one radial equation

The goal of this final derivation is to show how the two constants plus the mass-shell condition reduce a four-dimensional geodesic problem to a single first-order equation for $r(\tau)$ — the reduction advertised in the key idea and drawn in the second figure. We flatten the orbit into the equatorial plane, then substitute the conservation laws into the normalization to eliminate $\dot t$ and $\dot\phi$ , arriving at the effective-potential energy equation.

Step 1. The polar Euler–Lagrange equation is $\dfrac{\mathrm{d}}{\mathrm{d}\tau}(r^2\dot\theta)=r^2\sin\theta\cos\theta\,\dot\phi^2$ . Setting $\theta=\pi/2,\ \dot\theta=0$ gives $\ddot\theta=0$ , so an orbit started in the plane stays there; by $SO(3)$ symmetry any orbit can be rotated into $\theta=\pi/2$ . Why: this removes $\theta$ and lets us write $L=r^2\dot\phi$ with $\sin\theta=1$ .

Step 2. In the equatorial plane the normalization (7) becomes $-(1-2GM/r)\dot t^2 + (1-2GM/r)^{-1}\dot r^2 + r^2\dot\phi^2 = -\epsilon$ . Why: the one scalar constraint tying the surviving velocities together.

Step 3. Solve the conservation laws (5) and (6) for the ignorable velocities, $\dot t = \dfrac{E}{1-2GM/r}$ and $\dot\phi = \dfrac{L}{r^2}$ , and substitute. Why: this trades the unknown $\dot t,\dot\phi$ for the constants $E,L$ , leaving $r$ as the only dynamical coordinate.

Step 4. The constraint becomes $-\dfrac{E^2}{1-2GM/r} + \dfrac{\dot r^2}{1-2GM/r} + \dfrac{L^2}{r^2} = -\epsilon$ . Multiply by $(1-2GM/r)$ and rearrange to get $\dot r^2 = E^2 - \left(1-\dfrac{2GM}{r}\right)\left(\epsilon + \dfrac{L^2}{r^2}\right)$ . Why: dividing by two casts it as $\tfrac12\dot r^2 + V_{\rm eff} = \tfrac12 E^2$ — a nonrelativistic particle of energy $\tfrac12 E^2$ in the potential $V_{\rm eff}$ . The four-dimensional geodesic is now a one-dimensional energy problem. The planar reduction is standard for any spherically symmetric spacetime.

Limiting cases

Case	Result	Meaning
Newtonian / weak field	$E \approx 1 + \tfrac{1}{2}v^2 - \dfrac{GM}{r}$	At large $r$ , small $v$ , $E$ (per unit mass, $c=1$ ) is rest energy plus the Newtonian conserved energy; $L\to r^2\dot\phi$ is ordinary specific angular momentum.
At the horizon	$\xi_\mu\xi^\mu = -(1-2GM/r)\to 0$	The timelike Killing vector becomes null at $r=2GM$ ; surface gravity $\kappa$ is defined there, and "static-observer energy" ceases to exist.
Photon (null) limit	$b\equiv \dfrac{L}{E},\quad b_c = 3\sqrt3\,GM$	Only the impact parameter $b=L/E$ survives affine rescaling; $b_c=3\sqrt3\,GM$ sets the photon sphere at $r=3GM$ .
Static observer	$E_{\rm loc} = \dfrac{E}{\sqrt{1-2GM/r}}$	The local energy exceeds $E$ and diverges as $r\to 2GM$ — the blueshift of a photon falling in from infinity.

Common trapThe conserved E is not the energy a hovering observer measures

The constant $E=(1-2GM/r)\dot t$ is the energy per unit mass relative to infinity — what a distant observer books, conserved because $\xi=\partial_t$ is an asymptotic symmetry. A static observer hovering at radius $r$ measures instead $E_{\rm loc}=E/\sqrt{1-2GM/r}$ , larger by the blueshift factor; the two coincide only as $r\to\infty$ , and the gap between them is the gravitational redshift. A second, sharper trap: $\xi=\partial_t$ is timelike only for $r>2GM$ . At the horizon its norm $-(1-2GM/r)$ vanishes and inside it turns spacelike, so the quantity called "energy" loses that meaning — which is why the sign of $E$ behaves so strangely near and inside a black hole.

Worked examples

1 · Redshift straight from the Killing-vector norm (confidence builder)

A static beacon hovers at $r=6GM$ and emits light at proper wavelength $\lambda_{\rm em}=500\ \mathrm{nm}$ ; a distant astronomer at $r\to\infty$ receives it. Using only the norm of the timelike Killing vector, find the received wavelength.

Step 1. A photon's conserved energy $E=-\xi_\mu p^\mu$ is the same at emission and reception because $\xi$ is a symmetry. Its locally measured frequency is $E$ divided by the local norm of $\xi$ : $\omega_{\rm loc}\propto E/\sqrt{-\xi_\mu\xi^\mu} = E/\sqrt{1-2GM/r}$ .

Step 2. Take the ratio of received to emitted frequency. At infinity $\sqrt{1-2GM/r}\to 1$ ; at the beacon $\sqrt{1-2GM/6GM}=\sqrt{2/3}=\mathbf{0.8165}$ . Hence $\dfrac{\omega_{\rm rec}}{\omega_{\rm em}}=\sqrt{1-2GM/r_{\rm em}}=0.8165$ .

Step 3. Wavelength scales inversely with frequency, so $\lambda_{\rm rec}=500/0.8165 = \mathbf{612\ nm}$ — the light climbs out of the well and reddens from green toward orange. The conserved $E$ never changed; only the locally measured frequency did, by the factor $\sqrt{1-2GM/r} = 0.8165$ .

2 · The ISCO, its constants, and the efficiency of an accretion disk (fully worked)

Circular orbits are the extrema of $V_{\rm eff}$ . For Schwarzschild the standard results (recalled from the circular-orbit article) are $L^2 = \dfrac{GM\,r^2}{r-3GM}$ and $E^2 = \dfrac{(r-2GM)^2}{r(r-3GM)}$ . Evaluate the ISCO at $r=6GM$ , and turn its energy into the maximum efficiency of a thin accretion disk. Work in $G=c=M=1$ units.

Step 1. Angular momentum at $r=6$ : $L^2 = \dfrac{r^2}{r-3} = \dfrac{36}{3} = 12$ , so $L = 2\sqrt3 \approx \mathbf{3.464}$ (units $GM/c$ ). This is the minimum angular momentum for which a stable circular orbit exists — below it the minimum and maximum of $V_{\rm eff}$ merge and the well disappears.

Step 2. Energy at $r=6$ : $E^2 = \dfrac{(r-2)^2}{r(r-3)} = \dfrac{16}{18} = \dfrac{8}{9}$ , so $E = \dfrac{2\sqrt2}{3} \approx \mathbf{0.9428}$ (per unit rest mass).

Step 3. A particle spiraling inward through near-circular orbits radiates the difference between its rest energy at infinity ( $E=1$ ) and its energy at the ISCO. The fractional binding energy is $1-E = 1 - 0.9428 = \mathbf{0.0572}$ .

Step 4. Therefore up to $\approx \mathbf{5.7\%}$ of the rest-mass energy of accreted matter can be radiated before it plunges through the ISCO — the canonical efficiency of a Schwarzschild accretion disk, versus $\sim0.7\%$ for hydrogen fusion. A non-rotating black hole is a $\sim$ 8-times-more-efficient engine than the Sun's core, and the number falls straight out of the conserved $E$ at $r=6GM$ .

Physics of the portrait

The idea that a symmetry of space itself could be the origin of a conservation law was born in nineteenth-century differential geometry, decades before there was a spacetime for it to act on — and the man whose equation now labels every isometry never saw a metric of gravity.

Origins · 1 minWilhelm Killing and how this was discovered

Wilhelm Killing

1847–1923 · geometry of continuous symmetry

Wilhelm Karl Joseph Killing was a German mathematician who spent the most productive years of his career not at a great research university but teaching at the Lyceum Hosianum in Braunsberg, a Catholic seminary college in East Prussia, carrying a heavy teaching load alongside first-rate research. In the 1880s he set out to classify all the ways a geometry can be continuously deformed into itself, and in doing so independently created the theory of what we now call Lie algebras — the "Killing form," the root systems, and the classification of simple algebras that Élie Cartan would later complete and make famous. In his 1892 work on space forms he wrote down the condition for an infinitesimal transformation to preserve a metric — the vanishing of the symmetrized derivative of the generating field, $\nabla_{(\mu}K_{\nu)}=0$ . When general relativity arrived a quarter-century later and physicists needed a coordinate-free way to say "this spacetime has a symmetry," Killing's equation was waiting, and the vector fields that solve it became "Killing vectors." Every conserved energy and angular momentum in a stationary or axisymmetric spacetime — from Schwarzschild to Kerr to the expanding universe — is a contraction with one of his vectors: pure geometry supplying, in advance, exactly the language physics would need.

History

1892

— Killing writes the isometry condition, the vanishing symmetrized derivative of a vector field, in his study of space forms

1915

— Einstein completes the field equations of general relativity

1916

— Schwarzschild solves them in vacuum, finding a static, spherically symmetric metric whose symmetries encode E and L

1918

— Noether proves that every continuous symmetry of an action yields a conserved quantity

1923

— Birkhoff shows any spherically symmetric vacuum is necessarily static, guaranteeing the timelike Killing vector even for a collapsing star

1968

— Carter finds a hidden Killing-tensor constant for the rotating Kerr metric, extending the symmetry program beyond Killing vectors

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §25.2–25.5 · Wald, General Relativity, §6.2–6.3 and Appendix C.3 (Killing vectors and conserved quantities) · Carroll, Spacetime and Geometry, §5.4 · Hartle, Gravity, Ch. 8–9 · Schutz, A First Course in General Relativity, Ch. 11. The key relations used throughout this article are: the constants $E=(1-2GM/r)\dot t$ , $L=r^2\sin^2\theta\,\dot\phi$ , and the radial equation $\dot r^2=E^2-(1-2GM/r)(\epsilon+L^2/r^2)$ . The circular-orbit relations $L^2=GMr^2/(r-3GM)$ and $E^2=(r-2GM)^2/[r(r-3GM)]$ yield, at the ISCO $r=6GM$ , the values $L=2\sqrt3\,GM$ , $E=2\sqrt2/3$ , and a binding fraction $1-E=0.0572$ ( $5.7\%$ ). The photon-sphere impact parameter is $b_c=3\sqrt3\,GM$ at $r=3GM$ .

Prerequisites. The Schwarzschild Metric · Geodesic Equations in Schwarzschild Spacetime · Birkhoff's Theorem. Leads to. Effective Potential for Massive Particle Orbits · Circular Orbits and Their Stability · Innermost Stable Circular Orbit (ISCO) · Photon Orbits and the Photon Sphere · Gravitational Redshift in Schwarzschild Spacetime.

Ask the tutor: Ask me to derive the ISCO from $V_{\rm eff}$ , work Mercury's $43''$ precession from the same $E$ and $L$ , or show how Carter's constant rescues integrability for a spinning Kerr black hole.

What exactly is a Killing vector, in plain words?

It is an arrow field marking a direction in which the geometry does not change. Slide the entire spacetime a tiny amount along a Killing vector field and every distance and angle comes back identical — the metric is invariant under that flow, a condition written as Killing's equation $\nabla_\mu K_\nu + \nabla_\nu K_\mu = 0$ . In Schwarzschild, sliding along $\partial_t$ leaves the metric alone (it has no $t$ in it) and rotating about $\partial_\phi$ leaves it alone (no $\phi$ ), so both are Killing vectors. The value of the concept is that it is coordinate-free: it names a symmetry as a geometric object, so the conserved quantities it produces are genuine invariants, not artefacts of a lucky coordinate choice.

Why do we need both Killing's equation and the geodesic equation for conservation?

Because the change in $K_\mu\dot x^\mu$ splits cleanly into two pieces. One, $\dot x^\mu\dot x^\nu\nabla_\mu K_\nu$ , asks how the Killing vector varies; it vanishes because $\dot x^\mu\dot x^\nu$ is symmetric and Killing's equation kills the symmetric part of $\nabla K$ . The other, $K_\mu\dot x^\nu\nabla_\nu\dot x^\mu$ , asks how the tangent accelerates; it vanishes only because a geodesic has zero acceleration. Remove either condition and one term survives: a non-geodesic (a rocket firing its engines) does not conserve $K_\mu\dot x^\mu$ even in a symmetric spacetime, and a curve in a spacetime with no symmetry has no such quantity at all. Conservation is the coincidence of a symmetric geometry and a free-falling path.

Is the specific energy E the same as the total energy of the particle?

It is the energy per unit rest mass, measured relative to infinity. Multiply $E$ by the rest mass $m$ to get the conserved energy an observer at infinity attributes to the particle; the value includes the rest-mass contribution, which is why $E\to1$ (not $0$ ) for a particle at rest far away. It is emphatically not the energy a local hovering observer reads — that observer measures $E/\sqrt{1-2GM/r}$ , blueshifted by sitting deep in the well. The distinction is the whole content of gravitational redshift, and why "energy" in curved spacetime always needs the qualifier "with respect to which observer."

Why does spherical symmetry let us just set θ = π/2?

Two facts combine. First, the polar equation of motion shows $\theta=\pi/2$ with $\dot\theta=0$ is a consistent solution: a particle starting in the equatorial plane and moving within it has vanishing polar acceleration $\ddot\theta$ and never drifts out. Second, spherical symmetry ( $SO(3)$ invariance) means the physics does not care which plane you call "equatorial," so any single orbit can be rotated so its initial position and velocity lie in $\theta=\pi/2$ . Since one orbit only ever explores a single plane, aligning that plane with the equator costs nothing — the Newtonian statement that a central-force orbit is planar, promoted to curved spacetime.

Why do photons depend only on b = L/E and not on E and L separately?

Null geodesics carry a freedom timelike ones do not: rescaling the affine parameter, $\lambda\to a\lambda$ , gives a physically identical light ray. Under it the tangent $\dot x^\mu$ scales by $1/a$ , so both $E=-\xi\cdot\dot x$ and $L=\eta\cdot\dot x$ scale together — but their ratio $b=L/E$ is invariant. Since the shape of the ray, its deflection, and whether it is captured cannot depend on an arbitrary parametrization, they depend only on $b$ . Physically $b$ is the impact parameter, the perpendicular offset of the incoming ray measured at infinity, and the critical value $b_c=3\sqrt3\,GM$ divides captured from escaping light — the edge of the black-hole shadow.

Do these conserved quantities still exist inside the horizon or for a spinning black hole?

The Killing vectors persist — $\partial_t$ and $\partial_\phi$ remain Killing vectors everywhere the coordinates are valid — so $E$ and $L$ are still constants of motion inside $r=2GM$ . What changes is their interpretation: $\partial_t$ becomes spacelike, so $E$ is no longer any observer's energy, and $r$ itself becomes timelike, forcing infall. For a rotating (Kerr) hole there is still a timelike-at-infinity $\partial_t$ and an axial $\partial_\phi$ , giving conserved $E$ and $L$ , but spherical symmetry is lost, so a third constant — Carter's constant, from a hidden Killing tensor rather than a Killing vector — is needed to make the geodesics fully integrable. Killing's geometric idea simply generalizes.