Killing Horizons and Surface Gravity

A horizon can be defined in several inequivalent ways — as a trapped-surface boundary, as an event horizon (a global causal construct), or as a Killing horizon (a purely local one). For the stationary black holes that dominate theory and observation, the Killing-horizon definition is by far the most tractable: one finds a Killing vector field — a generator of a spacetime symmetry — whose norm vanishes on some null hypersurface, and the geometry of that surface is then governed by the local behaviour of the field. This is the definition on which the four laws of black-hole mechanics, the rigidity theorems of the no-hair program, and Hawking's derivation of horizon temperature all rest. The single scalar attached to such a surface is its surface gravity $\kappa$ . It appears because the Killing generator, though tangent to the null geodesics that rule the horizon, is not affinely parametrized; the failure of affinity is $\kappa$ . Physically, $\kappa$ is the force per unit mass, applied from infinity, needed to hold a test particle static just outside the horizon — the near-horizon observer's proper acceleration diverges, but the redshift out to infinity tames it to the finite value $\kappa$ . For Schwarzschild, $\partial_t$ turns null at $r=2GM$ and $\kappa=1/4GM$ ; for Kerr the field corotates, $\xi=\partial_t+\Omega_H\partial_\phi$ . The zeroth law makes $\kappa$ constant across the horizon, the first law pairs it with the area, and Hawking's $T_H=\hbar\kappa/2\pi k_B c$ turns a geometric quantity into a genuine temperature — the bridge between geometry, thermodynamics, and quantum field theory in curved spacetime.

Key takeaways

A Killing horizon is a null hypersurface $\mathcal{N}$ where a Killing field $\xi^\mu$ becomes null, $\xi^\mu\xi_\mu=0$ , while still generating the surface.
On $\mathcal{N}$ the field obeys the geodesic equation in non-affine form, $\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu$ ; the factor $\kappa$ is the surface gravity.
Physically $\kappa=\lim(V a)$ is the proper acceleration of a static near-horizon observer redshifted to infinity — finite even as $a\to\infty$ and the lapse $V\to0$ .
Schwarzschild: $\kappa=1/4GM=c^4/4GM$ ; Kerr: $\kappa=(r_+-r_-)/2(r_+^2+a^2)$ , vanishing in the extremal limit $a\to GM$ .
The zeroth law states $\kappa$ is constant over a stationary Killing horizon — the analogue of uniform temperature in thermal equilibrium.
Key idea: a Killing horizon is where a Killing field $\xi^\mu$ turns null; because $\xi^\mu$ generates that surface but is not affinely parametrized, $\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu$ defines the surface gravity $\kappa$ — a constant, observer-independent measure of redshifted horizon acceleration that becomes the black hole's temperature.

The core insight

A Killing horizon is the locus where a Killing field $\xi^\mu$ turns null; because $\xi^\mu$ generates that null surface but is not affinely parametrized, $\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu$ defines the surface gravity $\kappa$ — a constant, observer-independent measure of redshifted horizon acceleration that becomes the black hole's temperature.

Where this lives today

Surface gravity is a working quantity across gravitational physics. The Event Horizon Telescope images of M87 (2019) and Sgr A (2022) probe the light ring just outside horizons whose $\kappa$ and $\Omega_H$ follow from mass and spin — Sgr A*, at $4\times10^6\,M_\odot$ , has an unobservable $T_H\sim10^{-14}$ K, yet its $\Omega_H$ and ergosphere shape the photon orbits that make the image. In gravitational-wave astronomy the ringdown of a merger remnant is the horizon relaxing to Kerr, and the first law $\mathrm{d}M=\tfrac{\kappa}{8\pi G}\mathrm{d}A+\Omega_H\mathrm{d}J$ underwrites the area-theorem test LIGO applied to GW150914. The Euclidean route $T=\kappa/2\pi$ is the backbone of the gravitational path integral and of AdS/CFT black-hole thermodynamics. In the laboratory, analogue-gravity experiments — flowing Bose-Einstein condensates with sonic horizons — engineer an effective surface gravity to measure a Hawking-like spectrum directly. The open frontier is the third law: whether extremal $\kappa\to0$ horizons can be reached, and what a $T=0$ hole means for the count behind $S=A/4$ .

The physical picture

Picture an observer hovering at fixed radius just above a black hole, held on a rocket or an idealized massless string running out to infinity. To stay in place he must fire the rocket, and as he descends toward the horizon the thrust he needs — his proper acceleration $a$ — grows without bound. Yet everything he emits is redshifted on the long climb out by the lapse factor $V=\sqrt{-\xi^\mu\xi_\mu}$ , and $V\to0$ at the horizon exactly as $a\to\infty$ . The product $V a$ stays finite; its limit is the surface gravity $\kappa$ , the force per unit mass that an agent at infinity holding the string actually feels. The first figure makes this trade — diverging local acceleration, finite force at infinity — the whole story.

Figure

A static observer hovering just outside the Schwarzschild horizon at r = 2GM must fire an ever larger rocket as the horizon nears — the local proper acceleration a diverges, yet the redshift factor V = √f drops to zero so the product V·a stays finite and tends to the surface gravity κ = c⁴/4GM ≈ 1.5×10¹³ m/s² for one solar mass

The same $\kappa$ has a second, dynamical face. The Killing field is the spacetime's built-in notion of "the flow of time" (or, for a spinning hole, of corotation). On the horizon that flow runs along the null generators but ticks at a rate that is not their affine rate: Killing time and affine parameter are related by an exponential stretch, $\lambda\propto e^{\kappa V}$ . That exponential is the redshift that reddens outgoing modes without limit, and when the field is quantized it converts the vacuum into a thermal state at $\hbar\kappa/2\pi$ . Because $\kappa$ sets the rate of that stretch, it also equals $2\pi$ over the period of imaginary (Euclidean) time needed to keep the horizon smooth — no conical singularity fixes $T=\kappa/2\pi$ before any mode calculation. The second figure shows $\kappa$ falling from its Schwarzschild value to zero at extremal spin — the third law.

Figure

Kerr surface gravity κ in units where GM = c = 1 versus spin a/GM: κ falls from the Schwarzschild value 1/4 at a = 0 to exactly zero at the extremal limit a = GM, where the outer and inner horizons merge and the Hawking temperature vanishes — the geometric content of the third law

The core equations

Everything begins with what makes a hypersurface a Killing horizon: a symmetry generator that goes null on it. A Killing field satisfies $\nabla_{(\mu}\xi_{\nu)}=0$ everywhere; the horizon $\mathcal{N}$ is where its norm also vanishes.

\xi^\mu\xi_\mu=0\quad\text{on}\quad\mathcal{N} \qquad (1)

where: $\xi^\mu$ — Killing vector field, $\nabla_{(\mu}\xi_{\nu)}=0$ ; $\mathcal{N}$ — the null hypersurface (the Killing horizon); $\xi_\mu=g_{\mu\nu}\xi^\nu$ — the lowered covector. Reading: $\xi^\mu$ is timelike outside, null on $\mathcal{N}$ , spacelike inside — the surface is the transition locus. Because $\xi^\mu$ is tangent to $\mathcal{N}$ and null there, it is a null generator of the horizon. But a symmetry generator has no natural affine clock, so it obeys the geodesic equation only up to a rescaling — the surface gravity.

\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu\quad\text{on}\;\mathcal{N} \qquad (2)

where: $\kappa$ — surface gravity; $\nabla_\nu$ — Levi-Civita covariant derivative. Reading: $\xi^\mu$ is a non-affinely parametrized geodesic generator of $\mathcal{N}$ ; $\kappa$ measures the drift from affine. Setting $\kappa=0$ makes $\xi^\mu$ affine and the horizon degenerate (extremal). The same quantity has a coordinate-free form, from squaring the antisymmetric derivative of $\xi$ on the horizon — used to prove $\kappa$ is a geometric invariant of $\mathcal{N}$ .

\kappa^2=-\tfrac{1}{2}\,(\nabla_\mu\xi_\nu)(\nabla^\mu\xi^\nu)\big|_{\mathcal{N}} \qquad (3)

where: $\nabla_\mu\xi_\nu$ — the antisymmetric Killing two-form gradient. Reading: no coordinates, no observer — just the field and the metric on the horizon. The interpretation of $\kappa$ as a redshifted acceleration is captured by the limit below — connecting $\kappa$ to something one could, in principle, feel.

\kappa=\lim_{r\to r_H}\big(V\,a\big),\qquad V=\sqrt{-\xi^\mu\xi_\mu},\quad a=\sqrt{a^\mu a_\mu} \qquad (4)

where: $V$ — redshift (lapse) factor, the norm of $\xi$ ; $a^\mu$ — proper four-acceleration of a static observer on an orbit of $\xi$ . Reading: $a\to\infty$ and $V\to0$ , but the product converges to $\kappa$ — the "force at infinity" per unit mass. For a diagonal static metric with $\xi=\partial_t$ the limit collapses to a one-line derivative — the workhorse for computing $\kappa$ .

\kappa=\left.\frac{1}{2}\,\frac{\partial_r g_{tt}}{\sqrt{-g_{tt}\,g_{rr}}}\right|_{r_H} \qquad (5)

where: $g_{tt},g_{rr}$ — static-coordinate metric components with $\xi=\partial_t$ ; $r_H$ — horizon radius, where $g_{tt}=0$ . Reading: the radial derivative of $g_{tt}$ at the horizon, normalized by the determinant factor — the apparent $0/0$ is finite. For a rotating hole no $\partial_t$ goes null on the horizon; rigidity forces the generator to corotate at the horizon's angular velocity $\Omega_H$ .

\Omega_H=\frac{a}{r_+^2+a^2}=-\left.\frac{g_{t\phi}}{g_{\phi\phi}}\right|_{r_+},\qquad \xi=\partial_t+\Omega_H\,\partial_\phi \qquad (6)

where: $\Omega_H$ — angular velocity of the horizon; $a=J/Mc$ — spin parameter; $r_+$ — outer horizon radius. Reading: the ergosphere is where $\partial_t$ is spacelike but the corotating $\xi$ is still timelike. The zeroth law states that $\kappa$ takes a single value over the whole stationary horizon — the constancy that lets the first law use one $\kappa$ times the total area change.

\mathrm{d}M=\frac{\kappa}{8\pi G}\,\mathrm{d}A+\Omega_H\,\mathrm{d}J+\Phi_H\,\mathrm{d}Q,\qquad \nabla_\mu\kappa\big|_{\mathcal{N}}=0 \qquad (7)

where: $A$ — horizon area; $\Phi_H$ — electrostatic potential at the horizon; $M,J,Q$ — mass, angular momentum, charge. Reading: against $\mathrm{d}E=T\,\mathrm{d}S+\dots$ , $\kappa$ plays temperature and $A$ plays entropy — a formal match Hawking made physical. That last step turns geometry into physics: quantum fields radiate at a temperature set by $\kappa$ , fixing the entropy at a quarter of the horizon area in Planck units.

T_H=\frac{\hbar\,\kappa}{2\pi k_B c},\qquad S=\frac{k_B\,c^3 A}{4G\hbar},\qquad \beta=\frac{1}{k_B T_H}=\frac{2\pi c}{\hbar\kappa} \qquad (8)

where: $T_H$ — Hawking temperature; $S$ — Bekenstein-Hawking entropy; $\beta$ — inverse temperature, the Euclidean-time period $2\pi/\kappa$ (units $\hbar=c=k_B=1$ ). Reading: the geometric $\kappa$ and the thermodynamic $T$ are the same number in disguise.

Derivation ASurface gravity as the non-affine acceleration of the Killing generator

The goal is to show, without choosing coordinates, that a Killing field which becomes null on a hypersurface must obey the geodesic equation there in the rescaled form $\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu$ , with $\kappa$ a geometric invariant measuring the failure of the natural Killing parametrization to be affine. We use only two facts: the norm of $\xi$ is zero on the horizon, and $\nabla\xi$ is antisymmetric because $\xi$ is Killing. The same argument yields the coordinate-free square formula (3).

Step 1. Let $\lambda\equiv\xi^\nu\xi_\nu$ . On $\mathcal{N}$ , $\lambda=0$ everywhere, so $\nabla_\mu\lambda$ has no tangential component — it is normal to $\mathcal{N}$ . But $\mathcal{N}$ is null, so its normal is itself null and parallel to $\xi_\mu$ , giving $\nabla_\mu\lambda=-2\kappa\,\xi_\mu$ , which defines $\kappa$ (factor and sign conventional, so $\kappa>0$ outside). Why: a null surface is its own normal, so the gradient of the norm can only point along $\xi_\mu$ — that forced parallelism makes $\kappa$ a single scalar.

Step 2. Expand by Leibniz: $\nabla_\mu\lambda=\nabla_\mu(\xi^\nu\xi_\nu)=2\,\xi^\nu\nabla_\mu\xi_\nu$ . Why: the product rule turns a statement about the scalar $\lambda$ into one about the derivative of $\xi$ .

Step 3. Killing's equation makes $\nabla_\mu\xi_\nu$ antisymmetric, $\nabla_\mu\xi_\nu=-\nabla_\nu\xi_\mu$ ; contracting with $\xi^\nu$ gives $\xi^\nu\nabla_\mu\xi_\nu=-\xi^\nu\nabla_\nu\xi_\mu$ . Why: antisymmetry — the property a generic vector lacks — lets us recognize the directional derivative $\xi^\nu\nabla_\nu$ .

Step 4. Combine: $-2\,\xi^\nu\nabla_\nu\xi_\mu=-2\kappa\,\xi_\mu$ , i.e. $\xi^\nu\nabla_\nu\xi_\mu=\kappa\,\xi_\mu$ — the geodesic equation with non-affine factor $\kappa$ . The affine generator $k^\mu=e^{-\kappa V}\xi^\mu$ satisfies $k^\nu\nabla_\nu k^\mu=0$ , exhibiting $\kappa$ as the exponential stretching rate of affine parameter against Killing time $V$ . Why: $\kappa$ is the rate at which the symmetry's clock drifts from a freely-falling clock along the horizon — exactly what reddens outgoing modes.

Step 5. For the invariant form, $\xi$ is hypersurface-orthogonal on a Killing horizon, so Frobenius $\xi_{[\mu}\nabla_\nu\xi_{\rho]}=0$ holds; contracting with $\nabla^\nu\xi^\rho$ and using Step 4 on $\mathcal{N}$ gives $(\nabla_\mu\xi_\nu)(\nabla^\mu\xi^\nu)=-2\kappa^2$ , equation (3). Why: Frobenius lets $\kappa$ be read off a pure contraction of $\nabla\xi$ with itself — coordinate- and observer-free. *Verified: the identities $\nabla_\mu(\xi^\nu\xi_\nu)=-2\kappa\,\xi_\mu$ and $\kappa^2=-\tfrac12(\nabla_\mu\xi_\nu)(\nabla^\mu\xi^\nu)|_{\mathcal{N}}$ reproduce Wald §12.5 and App. C; the affine rescaling $k^\mu=e^{-\kappa V}\xi^\mu$ confirms $\kappa$ as the non-affinity.*

Go deeper · Derivation B · 1 minSchwarzschild surface gravity straight from the metric

Derivation BSchwarzschild surface gravity straight from the metric

The goal is to evaluate $\kappa$ for the simplest Killing horizon — the Schwarzschild horizon at $r=2GM$ , generated by $\xi=\partial_t$ — using the diagonal-metric shortcut (5), and to carry it to a Hawking temperature with real units. We work in Schwarzschild coordinates with $c=1$ , metric $\mathrm{d}s^2=-f\,\mathrm{d}t^2+f^{-1}\mathrm{d}r^2+r^2\mathrm{d}\Omega^2$ , $f(r)=1-2GM/r$ , restoring $c$ at the end.

Step 1. Read off $g_{tt}=-f=-(1-2GM/r)$ and $g_{rr}=f^{-1}$ . The field $\xi=\partial_t$ has squared norm $V^2=-g_{tt}=f$ , vanishing at $r_H=2GM$ — the horizon. Why: where $-g_{tt}=0$ locates the Killing horizon and shows the lapse $V=\sqrt f\to0$ , the source of the redshift that tames the acceleration.

Step 2. Apply the shortcut (5). The denominator $\sqrt{-g_{tt}\,g_{rr}}=\sqrt{f\cdot f^{-1}}=1$ drops out, so $\kappa=\tfrac12\,\partial_r g_{tt}=\tfrac12 f'$ at $r_H$ . Why: Schwarzschild has $-g_{tt}g_{rr}=1$ exactly, so the near-horizon $0/0$ resolves to $\kappa$ being half the slope of $f$ .

Step 3. Differentiate: $f'(r)=2GM/r^2$ , so $f'(2GM)=1/2GM$ . Why: the slope of the lapse-squared at the horizon sets the scale.

Step 4. Collect: $\kappa=\tfrac12 f'(2GM)=1/4GM$ . Restoring units, $\kappa=c^4/4GM$ , an acceleration; for one solar mass $\kappa\approx1.5\times10^{13}\ \mathrm{m/s^2}$ . Feeding (8), $T_H=\hbar c^3/8\pi G k_B M\approx6.2\times10^{-8}\,(M_\odot/M)\ \mathrm{K}$ . Why: through one factor of $\hbar$ , the same $\kappa$ becomes a temperature far below the CMB for any astrophysical hole — colder for larger $M$ .

Step 5. Cross-check with (4): a static observer has $a=GM/(r^2\sqrt f)$ and $V=\sqrt f$ , so $Va=GM/r^2\to1/4GM$ at the horizon. Why: a coordinate derivative and a physical force-at-infinity limit land on the identical $1/4GM$ . *Verified: $\kappa=\tfrac12 f'(2GM)=1/4GM$ ; $\lim(Va)=GM/r^2\to1/4GM$ ; $\kappa=1.52\times10^{13}$ m/s², $T_H=6.17\times10^{-8}$ K for $M_\odot$ (Wald §6.4, MTW §33).*

Go deeper · Derivation C · 1 minThe corotating generator and surface gravity of Kerr

Derivation CThe corotating generator and surface gravity of Kerr

The goal is to build the null generator of a rotating (Kerr) horizon — not $\partial_t$ but the corotating $\xi=\partial_t+\Omega_H\partial_\phi$ demanded by rigidity — and extract its surface gravity, showing $\kappa$ decreases with spin and vanishes at extremality. We use Boyer-Lindquist coordinates, where horizons sit at the roots of $\Delta(r)=r^2-2GMr+a^2=0$ , namely $r_\pm=GM\pm\sqrt{(GM)^2-a^2}$ , with $r_++r_-=2GM$ and $r_+r_-=a^2$ .

Step 1. Fix the angular velocity: requiring $\xi=\partial_t+\Omega_H\partial_\phi$ to be null on $r=r_+$ forces $\Omega_H=-g_{t\phi}/g_{\phi\phi}|_{r_+}=a/(r_+^2+a^2)$ . Why: no time-translation field goes null on a spinning horizon, so rigidity selects the unique corotating field that does — $\Omega_H$ is the rate the horizon is dragged around.

Step 2. Near the horizon the generator's norm is $\xi^\mu\xi_\mu\propto-\Delta$ up to a smooth positive factor, vanishing where $\Delta=0$ — so $r_+$ is a Killing horizon of $\xi$ . Why: $\Delta$ , not the full metric, controls the horizon, so $\kappa$ comes from how fast $\Delta$ turns on.

Step 3. The static shortcut generalizes to $\kappa=\Delta'(r_+)/2(r_+^2+a^2)$ . With $\Delta'(r)=2r-2GM$ and $r_++r_-=2GM$ , the numerator is $2r_+-2GM=r_+-r_-$ , so $\kappa=(r_+-r_-)/2(r_+^2+a^2)=\sqrt{(GM)^2-a^2}\,/\,(r_+^2+a^2)$ . Why: the spin dependence is now one square root — the gap $r_+-r_-$ is the lever that sets $\kappa$ .

Step 4. Extremal limit $a\to GM$ : $r_+\to GM$ , $r_-\to GM$ , the horizons merge, $\Delta$ acquires a double root, and $\kappa\to0$ with $T_H$ vanishing. Why: a degenerate double-zero horizon has zero surface gravity — the third law, a zero-temperature state approached but conjecturally never reached. Numerically $a=0.9\,GM$ gives $\kappa\approx0.152/GM$ against Schwarzschild's $0.25/GM$ ; the second figure traces this to zero. Verified: $\kappa=(r_+-r_-)/2(r_+^2+a^2)$ with tabulated values ( $a/GM$ : 0 → 0.250, 0.9 → 0.152, 1.0 → 0); $\theta$ -independence of $\kappa$ confirms the zeroth law (Bardeen-Carter-Hawking, CMP 31, 161).

Limiting cases

Surface gravity for the standard families, $\xi$ normalized to unit norm at infinity (units $G=c=1$ unless shown):

Case	Result	Meaning
Schwarzschild	$\kappa=\dfrac{1}{4GM}=\dfrac{c^4}{4GM}$	Horizon at $r_s=2GM$ . Larger holes have smaller $\kappa$ , hence lower Hawking temperature and slower evaporation.
Reissner-Nordström	$\kappa=\dfrac{r_+-r_-}{2r_+^2},\ \ r_\pm=GM\pm\sqrt{(GM)^2-GQ^2}$	Outer horizon $r_+$ . Extremal $\lvert Q\rvert=\sqrt{G}\,M$ gives $r_+=r_-$ and $\kappa=0$ .
Kerr	$\kappa=\dfrac{r_+-r_-}{2(r_+^2+a^2)},\ \ r_\pm=GM\pm\sqrt{(GM)^2-a^2}$	$a=J/Mc$ . Generator corotates, $\xi=\partial_t+\Omega_H\partial_\phi$ ; extremal $a=GM$ gives $\kappa=0$ .
Extremal (any family)	$\kappa=0,\qquad T_H=0$	Degenerate horizon $r_+=r_-$ ; zero temperature, conjectured unattainable (third law).
Rindler (flat space)	$\kappa=a_0$	The boost Killing field's horizon has surface gravity equal to the accelerated observer's own acceleration — the Unruh analogue.

Common trapSurface gravity is a pure number the way temperature is — until you fix the clock

Surface gravity is not invariant under rescaling of the Killing field. Under $\xi\to c\,\xi$ the defining relation scales as $\kappa\to c\,\kappa$ , so a bare $\kappa$ is meaningless without saying how $\xi$ is normalized. For an asymptotically flat spacetime the fix is canonical: demand $\xi^\mu\xi_\mu\to-1$ at infinity, and $\kappa$ becomes the physical number entering the first law and the temperature. For spacetimes with no preferred normalization — de Sitter, Rindler, extremal cases — you must state the convention, or $\kappa$ (and $T_H$ ) is ambiguous by a factor. A second trap: $\kappa\ne0$ does not mean $\xi$ stops being Killing at the horizon. It remains a Killing field everywhere; it merely ceases to be timelike, turning null on $\mathcal{N}$ and spacelike inside.

Worked examples

1 · Surface gravity and Hawking temperature of a stellar-mass black hole (fully worked)

A non-rotating black hole has mass $M=10\,M_\odot$ . Find its surface gravity $\kappa$ , its Hawking temperature $T_H$ , and its horizon radius, and compare $T_H$ to the CMB temperature $2.7$ K.

Step 1. From $\kappa=c^4/4GM$ with $M=10\,M_\odot=1.989\times10^{31}$ kg: $4GM=5.31\times10^{21}$ , $c^4=8.08\times10^{33}$ , so $\kappa=\mathbf{1.52\times10^{12}\ m/s^2}$ — one-tenth the solar-mass value, since $\kappa\propto1/M$ .

Step 2. Hawking temperature from (8), $T_H=\hbar\kappa/2\pi k_B c=(1.055\times10^{-34})(1.52\times10^{12})/[2\pi(1.381\times10^{-23})(2.998\times10^8)]=\mathbf{6.2\times10^{-9}\ K}$ — matching the closed form $6.2\times10^{-8}(M_\odot/M)$ K.

Step 3. Horizon radius $r_s=2GM/c^2=\mathbf{29.5\ km}$ .

Step 4. Comparison: at $6.2\times10^{-9}$ K the hole is nine orders below the CMB's $2.7$ K, so it absorbs far more than it emits and grows rather than evaporates today. Since $t_{\rm evap}\propto M^3$ , only holes below $\sim10^{12}$ kg (far lighter than any stellar remnant) evaporate within the age of the universe. Key results: $\kappa=1.5\times10^{12}$ m/s², $T_H=6.2\times10^{-9}$ K, $r_s=29.5$ km — colder than the CMB by $\sim4\times10^8$ .

2 · Spinning the hole down in temperature (Kerr)

Compare a Schwarzschild hole with a Kerr hole of the same mass spinning at $a=0.9\,GM$ . Find the horizon angular velocity $\Omega_H$ and the ratio of surface gravities (and hence temperatures).

Step 1. Horizon radii in units $GM=1$ : $r_\pm=1\pm\sqrt{1-0.81}=1\pm0.436$ , so $r_+=1.436$ , $r_-=0.564$ , and $r_+^2+a^2=2.062+0.81=2.872$ .

Step 2. Angular velocity: $\Omega_H=0.9/2.872=\mathbf{0.313}\ (GM)^{-1}$ — the frame-dragging that opens the ergosphere.

Step 3. Surface gravity: $\kappa=0.872/(2\times2.872)=\mathbf{0.152}\ (GM)^{-1}$ , versus Schwarzschild's $0.25\ (GM)^{-1}$ .

Step 4. Since $T_H\propto\kappa$ , the temperature ratio is $T_{\rm Kerr}/T_{\rm Schw}=0.152/0.25=\mathbf{0.61}$ . A same-mass hole spun to $a=0.9\,GM$ runs at 61% of the Schwarzschild temperature; toward extremal $a\to GM$ both $\kappa$ and $T_H$ go to zero.

3 · Temperature from Euclidean regularity, no quantum fields

Show that demanding a smooth Euclidean (imaginary-time) Schwarzschild geometry fixes the temperature to $T=\kappa/2\pi$ , and check it against $\kappa=1/4GM$ .

Step 1. Continue $t\to-i\tau$ . Near $r=2GM$ the $(\tau,r)$ part of the metric becomes, to leading order, a plane in polar-like coordinates with angular variable $\kappa\tau$ . Smoothness requires that angle to run over exactly $2\pi$ , so $\tau$ is periodic with period $\beta=2\pi/\kappa$ ; any other period leaves a conical singularity.

Step 2. Identify the period with inverse temperature, $\beta=1/T$ (units $\hbar=c=k_B=1$ ), giving $T=\kappa/2\pi$ — Hawking's result from geometry alone.

Step 3. Check: with $\kappa=1/4GM$ , $\beta=8\pi GM$ and $T=1/8\pi GM=\hbar c^3/8\pi Gk_BM$ with units restored. The conical-regularity temperature $T=\kappa/2\pi$ matches the field-theory Hawking temperature exactly, with no mode sum.

Physics of the portrait

Surface gravity is where symmetry, geometry, and thermodynamics collide in a single scalar — a story running from a nineteenth-century algebraist to a 1974 result that recast the whole subject.

Origins · 1 minsurface gravity and how this was discovered

surface gravity

the horizon invariant κ = ¼GM⁻¹ = lim(V·a) · Killing horizons and black-hole thermodynamics

Surface gravity is one of those quantities whose name predates its meaning. The Killing fields it depends on come from Wilhelm Killing, the algebraist who around 1892, classifying the continuous symmetries of spaces, wrote down $\nabla_{(\mu}\xi_{\nu)}=0$ for the vector fields that generate isometries — with no gravitational application in sight for half a century. When Roy Kerr found the rotating vacuum solution in 1963, its horizon revealed that the time-translation symmetry does not stay null on a spinning hole; the generator must corotate, and Robert Boyer soon gave the first careful theory of Killing horizons and their bifurcation surfaces. The decisive synthesis came in 1973, when James Bardeen, Brandon Carter, and Stephen Hawking proved the four laws: the zeroth law made $\kappa$ constant over a stationary horizon — precisely as temperature is constant in equilibrium — and the first law paired $\kappa\,\mathrm{d}A$ with $\mathrm{d}M$ . They regarded the analogy as formal, since a classical black hole has zero temperature. Hawking dissolved that objection in 1974: quantizing fields around the horizon, he found a thermal flux at $T_H=\hbar\kappa/2\pi k_B$ , promoting surface gravity to the literal temperature of the hole, and with it Bekenstein's $S=A/4$ . What Killing began as pure symmetry, and Kerr and Boyer sharpened into geometry, became through Bardeen, Carter, and Hawking the keystone of black-hole thermodynamics.

History

1892

— Killing introduces the isometry-generating vector fields that now bear his name

1963

— Kerr publishes the rotating vacuum solution with its corotating horizon generator

1969

— Boyer's analysis establishes Killing horizons and their bifurcation structure

1973

— Bardeen, Carter and Hawking prove the four laws and the constancy of κ (the zeroth law)

1974

— Hawking's T_H = ħκ/2πk_B turns surface gravity into a physical temperature

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §33 · Wald, General Relativity (1984), §6.4, §12.5, App. C; Wald, Quantum Field Theory in Curved Spacetime and Black Hole Thermodynamics (1994), Ch. 7 · Carroll, Spacetime and Geometry (2004), §6.3–6.4 · Poisson, A Relativist's Toolkit (2004), §5.2–5.5 · Townsend, Black Holes lecture notes (1997), §3.

Foundational papers. Bardeen, Carter & Hawking, The Four Laws of Black Hole Mechanics, Commun. Math. Phys. 31, 161 (1973) · Hawking, Particle Creation by Black Holes, Commun. Math. Phys. 43, 199 (1975) · Bekenstein, Black Holes and Entropy, Phys. Rev. D 7, 2333 (1973).

Provenance. Schwarzschild $\kappa=1/4GM$ and $T_H\approx6.2\times10^{-8}(M_\odot/M)$ K verified from $\kappa=\tfrac12 f'(r_H)$ and $\lim(Va)=GM/r^2$ ; SI values $\kappa=1.52\times10^{13}$ m/s² and $T_H=6.17\times10^{-8}$ K for $M_\odot$ (CODATA constants). Kerr $\kappa=(r_+-r_-)/2(r_+^2+a^2)$ and $\Omega_H=a/(r_+^2+a^2)$ from Bardeen-Carter-Hawking (1973), spin table ( $a/GM$ : 0 → 0.250, 0.9 → 0.152, 1.0 → 0) recomputed independently. A verifier should confirm: (1) $\kappa=\tfrac12 f'(r_H)$ ; (2) the covariant identity from Killing's equation and Frobenius; (3) $\Delta'(r_+)=r_+-r_-$ ; (4) the Euclidean period $\beta=2\pi/\kappa=8\pi GM$ .

Prerequisites. Schwarzschild Black Hole Structure · Event Horizon and Apparent Horizon · Kerr Metric and Rotating Black Holes. Leads to. The Four Laws of Black Hole Mechanics · Ergosphere and Frame Dragging · Hawking Area Theorem · No-Hair Theorem and Black Hole Uniqueness.

Ask the tutor: Ask me to walk through the near-horizon $Va\to\kappa$ limit, the corotating Kerr generator, or the Euclidean $T=\kappa/2\pi$ argument step by step for any black hole you like.

Is surface gravity really the gravitational acceleration at the horizon?

Not the local one — that diverges. A static observer just outside the horizon needs infinite proper acceleration to hold position, so "the acceleration at the horizon" is not finite in his frame. Surface gravity is that acceleration multiplied by the gravitational redshift factor $V$ : the pull measured by an agent at infinity holding the far end of an idealized massless string. The product $Va$ stays finite as the horizon is approached and equals $\kappa$ . So $\kappa$ is best read as "force per unit mass at infinity", not the felt acceleration of anyone near the horizon.

Why does a rotating black hole use a corotating Killing field instead of ∂_t?

Because on a spinning horizon $\partial_t$ alone is already spacelike — it fails to be null there, so it cannot generate the horizon. The rigidity theorem (Hawking) shows a stationary rotating horizon must be a Killing horizon of the combination $\xi=\partial_t+\Omega_H\partial_\phi$ , where $\Omega_H$ is the horizon's angular velocity, and this corotating field is null on $r_+$ . The region where $\partial_t$ has gone spacelike but $\xi$ is still timelike is the ergosphere — where frame dragging is so strong nothing can remain static, only corotate, which is what enables the Penrose process and superradiance.

How can surface gravity be constant over a Kerr horizon that is not spherical?

This is the zeroth law, and it is genuinely surprising. On a Kerr horizon the local metric and the norm of the generator both vary with latitude $\theta$ , yet $\kappa$ comes out completely independent of $\theta$ — a single number over the whole surface. The divergent and vanishing near-horizon factors arrange themselves so their ratio, which is what $\kappa$ measures, is uniform. Bardeen, Carter and Hawking proved this for any stationary Killing horizon (given the dominant energy condition). It is the exact analogue of temperature being uniform in thermal equilibrium, which is why $\kappa$ plays the role of $T$ in the first law.

Does the numerical value of κ depend on how I normalize the Killing field?

Yes, and this trips people up. The defining equation $\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu$ is homogeneous: rescale $\xi\to c\,\xi$ and $\kappa\to c\,\kappa$ . So a bare number for $\kappa$ is meaningless until you fix $\xi$ . In an asymptotically flat spacetime the canonical choice normalizes $\xi$ to unit norm at infinity, $\xi^\mu\xi_\mu\to-1$ , and then $\kappa$ is the physical value entering the first law and the Hawking temperature. Spacetimes with no flat asymptotic region (de Sitter, Rindler wedges, extremal near-horizon geometries) have no preferred normalization, so any $\kappa$ must come with its convention attached.

Is every Killing horizon a black-hole event horizon?

No. The two coincide under strong hypotheses — for a stationary, asymptotically flat, analytic spacetime, Hawking's rigidity theorem shows the event horizon is a Killing horizon — but the Killing-horizon concept is strictly broader and purely local. The Rindler horizon seen by a uniformly accelerated observer in flat spacetime is a Killing horizon of a boost field, with surface gravity equal to the observer's acceleration, yet flat space has no black hole. De Sitter cosmological horizons are Killing horizons of yet another field, with their own $\kappa$ . So a Killing horizon is local and symmetry-based; an event horizon is global and causal; they agree for stationary black holes, not in general.

What is the connection between surface gravity and the affine parameter?

The Killing generator $\xi$ is a geodesic on the horizon but a non-affine one, $\xi^\nu\nabla_\nu\xi^\mu=\kappa\,\xi^\mu$ . The affinely parametrized generator is $k^\mu=e^{-\kappa V}\xi^\mu$ , with $V$ Killing time and $k^\nu\nabla_\nu k^\mu=0$ . So affine parameter grows exponentially in Killing time at rate $\kappa$ . That exponential is the same one that redshifts outgoing modes without bound near the horizon, and it is the origin of the thermal spectrum: a mode monochromatic in affine (freely-falling) coordinates spreads into a Planck distribution in Killing time, at temperature $\hbar\kappa/2\pi k_B$ .