Relativity: Black Holes

Relativity » Graduate Topics » Black Holes

Black Holes

A black hole is a region of spacetime from which no causal signal can reach distant observers; its boundary is the event horizon. This page covers the stationary black-hole solutions of general relativity — Schwarzschild (mass only), Reissner–Nordström (mass and charge) and Kerr (mass and spin) — together with their horizons, orbits and singularities, the Penrose diagrams that summarize their causal structure, the laws of black-hole thermodynamics, the information paradox, and the present observational evidence. It assumes General Relativity and uses the language of Tensor Formalism.

Conventions. Geometric units $G = c = 1$ throughout unless SI factors are shown explicitly, so mass, length and time share a dimension ($M_\odot \approx 1.48\ \text{km} \approx 4.93\ \mu\text{s}$). Signature $(-,+,+,+)$; $d\Omega^2 = d\theta^2 + \sin^2\theta\, d\phi^2$ is the unit 2-sphere metric. Charge $Q$ is in geometrized Gaussian units and spin enters through $a = J/M$. To restore SI units substitute $M \to GM/c^2$, $Q^2 \to GQ^2/(4\pi\varepsilon_0 c^4)$ and $a \to J/(Mc)$.

The stationary family at a glance

Astrophysical black holes are described, to excellent approximation, by the Kerr metric: any net charge is rapidly neutralized by surrounding plasma, while angular momentum is inherited from the collapsing matter and is hard to shed.

The Schwarzschild Solution

Karl Schwarzschild found the static, spherically symmetric vacuum solution in 1916, weeks after Einstein published the field equations:

By Birkhoff’s theorem this is the unique spherically symmetric vacuum solution, and it is automatically static. It therefore describes the exterior of any non-rotating spherical body — a star, a planet, or a black hole — even one that is pulsating or collapsing radially, since spherically symmetric motion cannot radiate gravitational waves. For $r \gg 2M$ the metric approaches Minkowski space and $g_{tt} \approx -(1 + 2\Phi)$ with Newtonian potential $\Phi = -M/r$.

Schwarzschild radius and event horizon

The metric factor $1 - 2M/r$ vanishes at the Schwarzschild radius

For the Earth $r_s \approx 9\ \text{mm}$. A body compressed inside its Schwarzschild radius must collapse, and the sphere $r = r_s$ becomes its event horizon: the boundary of the set of events that can send signals to future null infinity.

The horizon is a coordinate singularity, not a physical one. The curvature invariant (Kretschmann scalar)

is finite at $r = 2M$, and an infalling observer crosses the horizon in finite proper time with nothing locally remarkable happening. The divergence of $g_{rr}$ reflects only the failure of the static coordinates, much as longitude fails at the poles.

Tidal forces at the horizon. The radial tidal acceleration across a body of length $L$ is $\Delta a \approx 2GML/r^3$. Evaluated at $r = r_s$ it scales as $1/M^2$:

Large black holes can therefore be entered intact; small ones “spaghettify” infalling matter well outside the horizon.

Orbits, the photon sphere and the ISCO

Geodesics in Schwarzschild conserve energy $E = (1 - 2M/r)\,dt/d\tau$ and angular momentum $L = r^2\,d\phi/d\tau$ per unit mass. Radial motion then reduces to a one-dimensional problem,

with $\epsilon = 1$ for massive particles and $\epsilon = 0$ for light. The extra $-2ML^2/r^3$ term relative to Newtonian gravity is responsible for perihelion precession and, near the hole, for two characteristic radii:

• Photon sphere, $r = 3M$: the unstable circular orbit of light. Photons with impact parameter $b < b_c = 3\sqrt{3}\,M$ are captured, so a black hole backlit by a luminous source casts a shadow of angular radius $\approx 3\sqrt{3}\,M/D$ — about $2.6$ times the horizon’s coordinate size.
• Innermost stable circular orbit (ISCO), $r = 6M$: circular orbits exist for massive particles down to $r = 3M$, but they are unstable inside $6M$. The ISCO sets the inner edge of a thin accretion disk and hence the radiative efficiency $\eta = 1 - E_{\text{ISCO}} = 1 - \sqrt{8/9} \approx 5.7\%$ — the fraction of rest-mass energy radiated by gas spiralling in to the ISCO. Nuclear fusion, by comparison, releases $0.7\%$.

Rotation shifts these radii substantially (see Kerr orbits below).

Eddington–Finkelstein and Kruskal–Szekeres coordinates

Define the tortoise coordinate and the ingoing null coordinate

In ingoing Eddington–Finkelstein coordinates $(v, r, \theta, \phi)$,

which is regular at $r = 2M$ ($\det g = -r^4\sin^2\theta$). Ingoing radial light rays are the lines $v = \text{const}$; the outgoing family tilts over and, for $r < 2M$, also moves to smaller $r$ — the horizon is where outgoing light stands still.

Kruskal–Szekeres coordinates $(T, X)$ cover the entire maximal analytic extension. In the exterior region ($r > 2M$):

Radial light rays travel at $45^\circ$, the horizon is the pair of null lines $X = \pm T$, and the singularity $r = 0$ is the hyperbola $T^2 - X^2 = 1$. The extension has four regions: our exterior (I), the black-hole interior (II), a time-reversed white hole (IV), and a second exterior (III) joined to ours through the Einstein–Rosen bridge, a non-traversable wormhole. Regions III and IV are artefacts of an eternal black hole; a hole formed by collapse has only regions I and II (see Penrose diagrams).

Formation and the singularity theorems

Oppenheimer and Snyder (1939) showed that a homogeneous pressureless ball collapses through its horizon to $r = 0$ in finite proper time. Whether this depended on the idealized symmetry was settled by Penrose’s singularity theorem (1965): if spacetime contains a trapped surface (a closed 2-surface from which both ingoing and outgoing light converge), the null energy condition holds, and space has a non-compact Cauchy surface, then spacetime is null-geodesically incomplete — some light ray ends after finite affine parameter. Hawking and Penrose extended the argument in 1970. The theorems predict that singularities form, not what they are like; this work earned Penrose half of the 2020 Nobel Prize in Physics.

The central singularity

At $r = 0$ the Kretschmann scalar diverges: this is a genuine curvature singularity where tidal forces become infinite and classical general relativity stops making predictions. Inside the horizon $g_{tt}$ and $g_{rr}$ change sign, so $r$ is a time coordinate: decreasing $r$ is as unavoidable as the passage of time, and the singularity is a moment in every interior observer’s future, not a place that can be avoided.

Worked example: time to the singularity. For radial free fall from rest at infinity ($E = 1$), $dr/d\tau = -\sqrt{2M/r}$, so the proper time from horizon to singularity is

The longest possible proper time inside is $\pi M$, achieved by falling freely from rest at the horizon; firing rockets in any direction only shortens it. In SI units $\pi GM/c^3 \approx 1.5\times10^{-4}\ \text{s}$ for $M = 10\,M_\odot$ and about $66\ \text{s}$ for Sagittarius A* ($4.3\times10^6\,M_\odot$).

The Reissner–Nordström Solution

Adding electric charge gives the Reissner–Nordström (RN) solution, the unique static, spherically symmetric solution of the Einstein–Maxwell equations:

with potential $A_\mu\,dx^\mu = -(Q/r)\,dt$. The roots of $f(r) = 0$ are

• Outer (event) horizon $r_+$: the surface of no return.
• Inner (Cauchy) horizon $r_-$: beyond it the future is no longer determined by initial data, because the singularity is timelike and can inject arbitrary information. Radiation falling in behind an observer is infinitely blueshifted at $r_-$, driving mass inflation (Poisson–Israel, 1990); the smooth RN interior is not expected to survive in realistic collapse. Whether the resulting singularity is strong enough to enforce strong cosmic censorship is still debated, and the answer is known to depend on the matter content and on a positive cosmological constant.

Cosmic censorship. Penrose’s weak cosmic censorship conjecture asserts that generic gravitational collapse of reasonable matter never produces a singularity visible from infinity. It is unproven; fine-tuned counterexamples exist (e.g. critical collapse at the threshold of black-hole formation), but attempts to overcharge or overspin a black hole past extremality by throwing in test particles fail once self-force and back-reaction are included. Extremality behaves as an unreachable limit, in parallel with the third law below.

The Kerr Solution

Roy Kerr found the rotating vacuum solution in 1963. In Boyer–Lindquist coordinates,

The off-diagonal $dt\,d\phi$ term encodes frame dragging: zero-angular-momentum observers rotate with angular velocity $\omega = -g_{t\phi}/g_{\phi\phi}$ relative to infinity. Setting $a = 0$ recovers Schwarzschild; setting $M = 0$ gives flat space in oblate spheroidal coordinates. Unlike Schwarzschild, Kerr is not the unique exterior of a rotating star — real stars carry higher multipole moments — but it is the unique end state of collapse (see No-hair theorem).

Horizons and the ergosphere

The horizons are the roots of $\Delta = 0$,

which exist only for $a \le M$ (the Kerr bound, $J \le M^2$ in geometric units or $J \le GM^2/c$ in SI). The static limit or ergosurface is where $g_{tt} = 0$:

It touches the horizon at the poles and bulges out to $r = 2M$ at the equator. The region between it and $r_+$ is the ergosphere. There $\partial_t$ is spacelike, so no observer can remain at rest relative to distant stars — every worldline must co-rotate — yet escape to infinity is still possible, because the event horizon has not been crossed.

The Penrose process and superradiance

Because the Killing vector $\xi_{(t)} = \partial_t$ is spacelike in the ergosphere, a particle there can have negative conserved energy $E = -\xi_{(t)}\cdot p < 0$ as measured from infinity. In the Penrose process (1969) a particle entering the ergosphere splits in two; one fragment falls through the horizon on a negative-energy orbit and the other escapes with more energy than the original. The difference comes from the hole’s rotational energy, and the hole spins down.

The wave analogue is superradiance: a mode $e^{-i\omega t + im\phi}$ scattering off the hole is amplified when

where $\Omega_H$ is the angular velocity of the horizon. If a massive bosonic field has a Compton wavelength comparable to the hole’s size, superradiant modes can be trapped and grow exponentially into a boson cloud, spinning the hole down. The observed spins of stellar and supermassive black holes are therefore used to exclude ranges of ultralight axion-like particle masses.

Irreducible mass

Energy extraction cannot reduce the horizon area (see the area theorem). Christodoulou (1970) showed that this leaves an irreducible mass

For a maximally rotating hole ($J = M^2$), $M_{\text{irr}} = M/\sqrt{2}$, so up to $1 - 1/\sqrt{2} \approx 29\%$ of the mass-energy is extractable rotational energy. Magnetic analogues of the Penrose process — above all the Blandford–Znajek mechanism (1977), in which magnetic field lines threading the horizon are wound up by frame dragging — are the leading explanation for the relativistic jets of active galactic nuclei and are the standard engine in general-relativistic MHD simulations of M87*.

Kerr orbits and radiative efficiency

Rotation pulls prograde orbits inward and pushes retrograde orbits outward:

(For $a \to M$ the Boyer–Lindquist radii coincide, but the proper distances between horizon, photon orbit and ISCO remain finite and distinct.) Because the ISCO depends so strongly on spin, X-ray continuum fitting and relativistically broadened iron-line spectroscopy of the inner disk are the main electromagnetic spin measurements. Thorne (1974) showed that photon capture limits disk-fed spin-up to about $a \approx 0.998\,M$.

The ring singularity

The curvature singularity sits where $\rho^2 = 0$: $r = 0$ and $\theta = \pi/2$. In Kerr–Schild coordinates this is a ring of radius $a$ in the equatorial plane. The analytic extension continues through the disk bounded by the ring into a region of negative $r$ that contains closed timelike curves. As with the RN inner horizon, this interior structure is believed to be destroyed by mass inflation in any realistic collapse.

The no-hair theorem

The uniqueness theorems of Israel (1967), Carter (1971), Hawking (1972) and Robinson (1975) establish that a stationary, asymptotically flat, non-singular-outside-the-horizon black hole in Einstein–Maxwell theory is a member of the Kerr–Newman family, fixed by three numbers:

All other information about the collapsed matter — composition, shape, baryon number — is radiated away as gravitational and electromagnetic waves (the ringdown) or hidden behind the horizon. A concrete, testable consequence: every multipole moment of a Kerr hole is fixed by $M$ and $a$, e.g. the mass quadrupole is $Q_2 = -Ma^2$, and every ringdown quasinormal-mode frequency is a function of $(M, a)$ alone. Measuring two or more modes and checking that they imply the same $(M, a)$ — black-hole spectroscopy — is a direct test of the Kerr hypothesis. The theorem assumes four-dimensional Einstein–Maxwell theory; with other matter fields (Yang–Mills, some scalar couplings) “hairy” black holes exist.

Penrose Diagrams

A Penrose (conformal) diagram compactifies spacetime by a conformal rescaling $\tilde g_{\mu\nu} = \Omega^2 g_{\mu\nu}$ that brings infinity to a finite distance while preserving null geodesics. With spherical symmetry suppressed (each point is a 2-sphere), radial light rays run at $45^\circ$ and the whole causal structure can be read off by eye. The boundary of an asymptotically flat spacetime consists of:

The Schwarzschild diagram makes three facts visible:

• The event horizon is the boundary of the causal past of $\mathscr{I}^+$. Once inside region II, every future-directed ($\le 45^\circ$ from vertical) path ends on the singularity.
• The singularity is spacelike (horizontal): it is a moment of time, not a location.
• The eternal solution contains a white hole (IV) and a second exterior (III). A black hole formed by stellar collapse has neither: its diagram is region I and II with the collapsing star’s surface replacing the left half.

For Reissner–Nordström and Kerr the singularity is timelike (vertical) and the maximally extended diagram is an infinite tower of alternating exterior, between-horizon and inner regions. An observer could in principle avoid the singularity and pass through the Cauchy horizon into another universe — precisely the structure that mass inflation is expected to destroy.

Reading rules: light moves at $45^\circ$; your causal future is the upward-opening $45^\circ$ wedge; a horizon is a null line crossable in one direction only; a horizontal singularity is unavoidable, a vertical one is avoidable in principle.

Black-Hole Thermodynamics

In the early 1970s Bardeen, Carter and Hawking proved four laws of black-hole mechanics with the same structure as thermodynamics. Bekenstein (1972) argued that the analogy must be literal — otherwise dropping entropy into a black hole would violate the second law — and Hawking’s discovery of black-hole radiation (1974) fixed the constants.

The four laws

Here $\kappa$ is the surface gravity — the force per unit mass, measured at infinity, needed to hold a particle static just outside the horizon — $A$ the horizon area, $\Omega_H$ the horizon angular velocity and $\Phi_H$ the horizon electric potential. For Kerr–Newman

reducing to $\kappa = 1/4M$ and $A = 16\pi M^2$ for Schwarzschild, where one checks directly that $dM = (\kappa/8\pi)\,dA$. Integrating the first law with Euler’s theorem for homogeneous functions gives the Smarr formula

The area theorem was tested directly in 2025: in the gravitational-wave event GW250114 the final horizon area inferred from the ringdown exceeded the summed areas of the two initial black holes, as required, at a significance above $4\sigma$.

Hawking temperature

Quantizing matter fields on a collapsing black-hole background, Hawking (1974) found that the horizon emits a thermal flux at the Hawking temperature

The temperature is inversely proportional to mass. Every astrophysical black hole is far colder than the $2.7\ \text{K}$ cosmic microwave background and today absorbs more than it emits. The spectrum at infinity is a black body modified by greybody factors, the frequency-dependent transmission through the curvature potential around the hole.

Origin of the radiation. The picture of virtual pairs split at the horizon is a heuristic. The precise statement is that the notion of particle is observer-dependent: the quantum state that is regular (vacuum-like) for freely falling observers at the horizon is thermally populated for static observers far away. The same effect in flat space is the Unruh effect — a uniformly accelerated observer in the Minkowski vacuum sees a thermal bath at $T_U = \hbar a/(2\pi c k_B)$ — and $\kappa$ plays the role of the acceleration. Hawking radiation has not been observed from any astrophysical black hole, but its kinematics have been reproduced in analogue systems; in Bose–Einstein-condensate “sonic black holes” Steinhauer’s group reported spontaneous Hawking phonons with a thermal spectrum at the predicted temperature (2016, 2019).

Bekenstein–Hawking entropy

With $T_H$ fixed, the first law determines the Bekenstein–Hawking entropy

The entropy is one quarter of the horizon area in Planck units. Ordinary thermodynamic entropy is extensive, scaling with volume; black-hole entropy scales with area. This is the basis of the holographic principle — the maximum information content of a region is bounded by its boundary area — made precise in AdS/CFT (see Toward Quantum Gravity). For certain supersymmetric extremal black holes in string theory, Strominger and Vafa (1996) counted the microstates and reproduced $A/4$ exactly.

Worked example: a stellar black hole. For $M = 10\,M_\odot$, $r_s \approx 2.95\times10^4\ \text{m}$, $A = 4\pi r_s^2 \approx 1.1\times10^{10}\ \text{m}^2$, and $\ell_P^2 \approx 2.6\times10^{-70}\ \text{m}^2$:

The progenitor star had an entropy of order $10^{58}\,k_B$; collapse increased it by some twenty orders of magnitude. The supermassive black holes dominate the entropy budget of the observable universe.

Evaporation and lifetime

Losing energy raises $T_H$, so black holes have negative heat capacity and evaporation runs away. Treating the hole as a black body of area $\propto M^2$ at $T_H \propto 1/M$ gives $dM/dt \propto -M^{-2}$ and a finite lifetime. For emission of photons only,

with order-unity corrections once greybody factors and all particle species light enough to be emitted are included. Primordial black holes with initial mass near $5\times10^{11}\ \text{kg}$ would be completing their evaporation today, ending in a burst of gamma rays; searches by Fermi-LAT and ground-based Cherenkov and water-Cherenkov observatories have not found such bursts and bound their local rate. The final Planck-scale stage lies outside known physics.

The Information Paradox

If a black hole formed from matter in a pure quantum state evaporates completely into exactly thermal radiation, the final state is mixed. Unitary quantum evolution never maps a pure state to a mixed one, so either information is lost — a modification of quantum mechanics — or Hawking’s semiclassical calculation fails somewhere it was expected to be reliable. This is the black-hole information paradox (Hawking, 1976).

The sharpest form is an entanglement argument:

1. No drama. By the equivalence principle, an infalling observer sees vacuum at a large horizon, which requires each outgoing Hawking mode to be entangled with a partner mode just inside.
2. Unitarity. For the final radiation to be pure, late Hawking quanta must be entangled with the earlier radiation.
3. Monogamy of entanglement. A system cannot be maximally entangled with two others at once.

Page (1993) showed that unitarity requires the entanglement entropy of the radiation to follow the Page curve: rising while the hole is young, peaking at the Page time (roughly when the hole has lost half its initial Bekenstein–Hawking entropy), then falling to zero. Hawking’s calculation instead gives a monotonically increasing entropy.

The island computations show that the semiclassical gravitational path integral, evaluated correctly, already knows about unitarity. What they do not provide is a detailed account of how information is encoded in individual Hawking quanta, nor a derivation for realistic four-dimensional, asymptotically flat black holes. The problem remains a principal testing ground for quantum gravity; see Toward Quantum Gravity and String Theory.

Observational Status

Black holes have gone from mathematical curiosities to precisely measured astrophysical objects.

Next steps. Ground-based detectors plan a further observing period in 2026 and the O5 run from around late 2027. ESA’s LISA mission, adopted in 2024 and targeted for launch in the mid-2030s, will observe mergers of $10^5$–$10^7\,M_\odot$ black holes and extreme-mass-ratio inspirals that map the Kerr geometry orbit by orbit. Next-generation ground detectors (Einstein Telescope, Cosmic Explorer) and space-extended very-long-baseline interferometry are aimed at precision black-hole spectroscopy and movies of horizon-scale plasma.

Continue Reading

• Previous: Tensor Formalism & the Field Equations — the curvature machinery these solutions satisfy.
• Next: Gravitational Waves — linearized gravity, binary inspiral, and the ringdown of newly formed black holes.
• Up: Graduate Topics — the relativity deep-dive hub.

See Also

• General Relativity — the equivalence principle, the field equations, and the classic tests.
• Relativistic Cosmology — horizons in expanding spacetimes and de Sitter temperature.
• Toward Quantum Gravity — holography, AdS/CFT, and the programmes that aim to resolve the singularity.
• String Theory — microscopic state counting that reproduces the Bekenstein–Hawking entropy.
• Quantum Field Theory — the field theory behind Hawking and Unruh radiation.
• Thermodynamics — the four laws that black holes mirror.
• Physics Hub — browse all physics topics.