Interior Geometry and Radial Infall to the Singularity

The Schwarzschild line element carries a factor $(1-2GM/r)$ that changes sign at $r=2GM$ . Outside, it is positive and everything is ordinary: $t$ is a time, $r$ is a place, and a rocket can hold you at fixed radius forever. Cross the horizon and the factor turns negative — and with it the whole causal furniture rearranges. The component $g_{tt}$ becomes positive and $g_{rr}$ negative, the coordinate-level signature of something physical: $r$ has stopped being a spatial direction and become a time, one that runs only downhill toward $r=0$ . Once $r$ is a timelike coordinate, its decrease is no more optional than the arrival of next Tuesday. Every future-directed worldline — the burning rocket's as much as the falling stone's — has $\mathrm{d}r<0$ , because "the future" now points toward smaller $r$ . The singularity is therefore not a place you might steer around; it is a spacelike hypersurface lying across the future of every interior event, a moment in time that arrives no matter what you do. Firing engines only chooses which timelike line inside the cone you ride, and every line ends at $r=0$ . This is what "infall is inevitable" means, stated exactly, and it sharpens into a number: the longest proper time any observer can buy between horizon and singularity is $\pi GM/c^3$ — about fifteen microseconds for a solar mass, a minute for the hole at the Galactic centre. At the end waits a genuine curvature singularity, where the Kretschmann invariant diverges as $r^{-6}$ and tidal forces shred without bound. The interior of Schwarzschild is where the geometry stops being a stage and becomes the plot.

Key takeaways

Inside the horizon $(r<2GM)$ the sign of $(1-2GM/r)$ flips: $g_{tt}$ becomes positive and $g_{rr}$ negative, so $r$ is a timelike coordinate and $t$ is spacelike — space and time swap roles.
Because $r$ is timelike and decreasing toward the future is forced, reaching $r=0$ is as unavoidable as reaching tomorrow: no rocket, rope, or signal can hold or reverse it.
The surface $r=0$ is not a point in space but a spacelike hypersurface in the observer's future, and a true curvature singularity where $R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}=48G^2M^2/r^6$ diverges.
The maximum proper time from horizon to singularity is $\tau_{\max}=\pi GM/c^3\approx15.5\,(M/M_\odot)\,\mu\mathrm{s}$ , realized by the observer momentarily at rest at the horizon; any thrust or inward speed shortens it.
Light cones tip fully over inside: both edges of every future cone point to smaller $r$ , the coordinate-free statement that $r=0$ lies to the future of every interior event, making the horizon a one-way membrane.
Key idea: inside the horizon the signs of $g_{tt}$ and $g_{rr}$ flip, $r$ becomes a timelike coordinate that only decreases, and $r=0$ becomes a spacelike surface in your future — so hitting the singularity is as inevitable as the passage of time, with a maximum survival of $\pi GM/c^3$ .

The core insight

Inside the horizon the signs of $g_{tt}$ and $g_{rr}$ flip: $r$ becomes timelike and can only decrease toward the future, so $r=0$ is a spacelike hypersurface lying across the future of every interior event. Hitting the singularity is as inevitable as the passage of time, and the maximum proper time from horizon to singularity is $\pi GM/c^3$ .

Where this lives today

The interior geometry sits beneath the era's flagship black-hole observations. The Event Horizon Telescope's images of M87\ (2019) and Sagittarius A\ (2022) resolve structure at a few Schwarzschild radii, and the bright ring and central shadow are set by the same $2GM$ that flips the signature just within it. LIGO–Virgo–KAGRA's binary-merger catalog probes horizons forming and ringing down, with ringdown frequencies governed by the near-horizon spacetime the interior analysis completes. Tidal disruption events — stars shredded by supermassive holes — are the observable face of the spaghettification above, the disruption radius set by the $2GM/r^3$ law; for a Sun-like star and a $10^6\,M_\odot$ hole, disruption occurs outside the horizon as a months-long flare, while larger holes swallow the star whole. And the singularity, where $K\to\infty$ , marks where classical general relativity signs off — the open problem that quantum-gravity candidates (loop-cosmology bounces, string-theoretic fuzzballs) are built to resolve, replacing the $r=0$ divergence while leaving the exterior untouched.

The physical picture

Outside the horizon, gravity pulls you inward but you can resist: fire an engine hard enough and you hover, because $r$ is a location and "staying put" is a legitimate worldline. Inside, that option evaporates — $r$ has become a clock, and clocks only run down. Asking to hold $r$ fixed inside is asking to sit at a fixed instant while your proper time advances, a contradiction rather than an engineering challenge. The singularity at $r=0$ is simply the last tick of that clock. The cleanest way to see it is the light cones. Far outside, a future cone opens symmetrically: you may move to larger or smaller $r$ . Approaching the horizon the cones tip inward; at $r=2GM$ the outgoing edge stands vertical — outgoing light barely holds station and cannot climb. Inside, the cone has tipped past vertical, so both edges lean toward smaller $r$ . Since your worldline must stay inside your future cone, every allowed direction reduces $r$ . The first figure tracks one worldline through this tipping — a one-way funnel in time, not in space.

Figure

Light cones tip across the Schwarzschild horizon at r = 2GM. Outside, a cone opens both ways in r. At the horizon the outgoing edge stands vertical. Inside, both edges point toward smaller r, so every future-directed worldline (green) must fall to the spacelike singularity at r = 0

How long does the plunge last? Not long, and — counterintuitively — struggling makes it worse. The most leisurely descent drifts across the horizon at zero speed and simply falls; any inward velocity or thrust tightens the trajectory against the tipped cone and steals proper time. Integrating the free-fall equation for that marginal case gives a clean, finite $\tau_{\max}=\pi GM/c^3$ that scales with the mass. The second figure is what an infalling observer actually experiences: areal radius versus the proper time left on the clock, from horizon to annihilation.

Figure

The longest possible plunge: areal radius r versus proper time τ for an observer who crosses the horizon at rest and free-falls to r = 0. The curve is the free-fall cycloid r = GM(1 + cos η), and the total elapsed proper time is exactly πGM/c³ ≈ 15.5 μs for one solar mass

The core equations

The whole story starts from one line element and the single sign it carries. Written in Schwarzschild coordinates with $c=1$ , the geometry outside a mass $M$ is static and spherically symmetric, and the factor $f(r)=1-2GM/r$ multiplies the time part while its inverse multiplies the radial part.

\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\,\mathrm{d}\Omega^2 \qquad (1)

where: $r$ — areal radius, defined so a sphere of that label has area $4\pi r^2$ (spacelike for $r>2GM$ , timelike for $r<2GM$ ); $t$ — Schwarzschild time (timelike outside, spacelike inside); $M$ — geometric mass, with the horizon at $r_s=2GM$ ; $\mathrm{d}\Omega^2=\mathrm{d}\theta^2+\sin^2\theta\,\mathrm{d}\varphi^2$ — the unit 2-sphere. Reading it: the signature is $(-,+,+,+)$ , and whichever coordinate carries the minus sign is the local time direction. Isolating the two diagonal blocks that flip makes the swap explicit. The signature never changes; only which coordinate owns the timelike minus sign does.

g_{tt}=-\left(1-\frac{2GM}{r}\right),\qquad g_{rr}=\left(1-\frac{2GM}{r}\right)^{-1} \qquad (2)

where: $g_{tt}$ — negative for $r>2GM$ , zero at the horizon, positive for $r<2GM$ ; $g_{rr}$ — positive outside, divergent at the horizon, negative inside. Reading it: $g_{tt}>0$ inside means $\partial_t$ is spacelike there, so no static observer exists; $g_{rr}<0$ means $\partial_r$ is timelike, so $r$ is a clock. The reversal at $r=2GM$ is a property of the chart, not the geometry. The coordinate-independent question — is anything physically singular there? — is answered by a scalar built from the full Riemann tensor, the Kretschmann invariant.

K=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}=\frac{48\,G^2M^2}{r^6} \qquad (3)

where: $R_{\mu\nu\rho\sigma}$ — Riemann curvature tensor; $r$ — areal radius. Reading it: $K$ is finite at the horizon, $K(2GM)=3/(4G^4M^4)$ , so nothing physical breaks there; but $K\to\infty$ as $r\to0$ , a statement every observer agrees on, so $r=0$ is a true edge of the manifold. To follow a faller across the horizon we need a conserved quantity. Since the metric is independent of $t$ , the Killing vector $\partial_t$ gives an energy per unit rest mass constant along every geodesic — valid inside too, even though $\partial_t$ is no longer timelike there.

\tilde E=\left(1-\frac{2GM}{r}\right)\frac{\mathrm{d}t}{\mathrm{d}\tau} \qquad (4)

where: $\tilde E$ — energy per unit rest mass, conserved along geodesics; $\tilde E=1$ for a particle released from rest at infinity, $\tilde E=\sqrt{1-2GM/R}$ for one released from rest at radius $R$ ; $\tau$ — proper time. Reading it: $\tilde E$ is a constant of motion set by the initial drop, and it is the only free parameter in radial infall. Feeding $\tilde E$ into the normalization $g_{\mu\nu}u^\mu u^\nu=-1$ collapses radial motion to one first-order equation. The factor $(1-2GM/r)$ cancels, leaving an expression finite and real straight through the horizon.

\left(\frac{\mathrm{d}r}{\mathrm{d}\tau}\right)^2=\tilde E^2-\left(1-\frac{2GM}{r}\right)=\tilde E^2-1+\frac{2GM}{r} \qquad (5)

where: $\mathrm{d}r/\mathrm{d}\tau$ — proper-time rate of change of areal radius, real and nonzero for a faller at every $r>0$ . Reading it: nothing misbehaves at $r=2GM$ — the algebraic proof that the horizon is smooth to cross; only $r=0$ blows up the right side. Finally, the norm of the static Killing vector is the invariant that marks the flip: where it changes sign defines a Killing horizon.

\partial_t\cdot\partial_t=g_{tt}=-\left(1-\frac{2GM}{r}\right)\;\begin{cases}<0 & r>2GM\ (\text{timelike})\\[2pt]=0 & r=2GM\\[2pt]>0 & r<2GM\ (\text{spacelike})\end{cases} \qquad (6)

where: $\partial_t$ — the static Killing vector, timelike outside and spacelike inside. Reading it: a static observer would have to move along $\partial_t$ ; once that vector is spacelike, "staying at fixed $r$ " is a spacelike (superluminal) curve, forbidden for any real body.

Derivation AThe radial infall equation, regular across the horizon

The goal is to reduce the radial-infall geodesic problem to one clean first-order equation for $\mathrm{d}r/\mathrm{d}\tau$ , and to show it is finite at $r=2GM$ even though $g_{rr}$ diverges there. This is the engine behind everything else in the article: it supplies the proper-time integral for the survival-time calculation, and its right-hand side is what stays positive down to $r=0$ , forcing the fall. We use the two ingredients a geodesic always gives — the conserved energy from time-translation symmetry and the four-velocity normalization — and let the horizon factor cancel of its own accord.

Step 1. Restrict to radial motion, $\dot\theta=\dot\varphi=0$ , and write $g_{\mu\nu}u^\mu u^\nu=-1$ (dot is $\mathrm{d}/\mathrm{d}\tau$ ): $-\left(1-\tfrac{2GM}{r}\right)\dot t^2+\left(1-\tfrac{2GM}{r}\right)^{-1}\dot r^2=-1$ . Why: the normalization is the one algebraic constraint every timelike worldline obeys, valid in any chart on both sides of the horizon.

Step 2. Insert the conserved energy $\dot t=\tilde E\,(1-2GM/r)^{-1}$ into the first term, which becomes $-\tilde E^2(1-2GM/r)^{-1}$ . Why: trading the unknown $\dot t$ for the constant $\tilde E$ removes one variable using the metric's symmetry — exactly what a Killing vector is for.

Step 3. Multiply the whole equation through by the factor $(1-2GM/r)$ and rearrange. The radial term's inverse factor is cancelled by the multiplier, the time term loses its inverse factor, and the right side $-1$ picks up one factor of $(1-2GM/r)$ : the result is $\dot r^2=\tilde E^2-\left(1-\tfrac{2GM}{r}\right)$ . Why: this cancellation is the crux — the very factor that blows up in $g_{rr}$ is algebraically removed, so the physical rate $\dot r$ is finite at $r=2GM$ where the coordinate description looked catastrophic.

Step 4. Read off the signs. As $\dot r^2=\tilde E^2-1+2GM/r$ , the right side is strictly positive for all $0<r<\infty$ when $\tilde E\ge1$ , and for a marginal drop it stays positive right down to $r=0$ , where it diverges. Why: a strictly positive $\dot r^2$ means the radial velocity never vanishes on the way in — the seed of inevitability developed in the third derivation. *Verified: reproduces the standard radial-geodesic result of Carroll §5.6 and MTW §25.5; the horizon factor cancels analytically, leaving $\dot r^2=\tilde E^2-1+2GM/r$ finite at $r=2GM$ .*

Go deeper · Derivation B · 2 minLongest survival: the proper time from horizon to singularity

Derivation BLongest survival: the proper time from horizon to singularity

The goal is to compute exactly the maximum proper time an observer can experience between crossing $r=2GM$ and being destroyed at $r=0$ , and to prove it is $\pi GM/c^3$ , achieved not by thrusting but by the gentlest descent — drifting through the horizon at zero radial speed. This is the quantitative heart of "infall is inevitable": inevitability alone would allow a long fall, but here we bound the clock. The tool is the radial infall equation, integrated as a proper-time integral with the integrand made as large as physics permits.

Step 1. Express proper time as an integral over radius. Taking $\dot r<0$ on the way in, $\tau=\displaystyle\int_0^{2GM}\frac{\mathrm{d}r}{|\dot r|}=\int_0^{2GM}\frac{\mathrm{d}r}{\sqrt{\tilde E^2-1+2GM/r}}$ . The elapsed proper time is largest when the integrand is largest, i.e. when the denominator is smallest, i.e. when $\tilde E^2$ is as small as it can be. Why: more speed means less time on the clock; to maximize survival we must minimize the descent speed, which means minimizing the conserved energy.

Step 2. Find the smallest allowed $\tilde E$ . A particle that just barely reaches the horizon from rest satisfies $\tilde E=\sqrt{1-2GM/R}\to0$ as its release radius $R\to2GM$ . Setting $\tilde E=0$ gives the extremal integrand and $\dot r^2=2GM/r-1$ . Why: $\tilde E=0$ is the boundary of the allowed range for a body that crosses the horizon with vanishing speed; any real infall has $\tilde E\ge0$ , so this is the genuine maximizer, not merely a stationary point.

Step 3. Substitute the free-fall cycloid $r=GM(1-\cos\eta)=2GM\sin^2(\eta/2)$ , for which $2GM/r-1=\cot^2(\eta/2)$ and $\mathrm{d}r=GM\sin\eta\,\mathrm{d}\eta$ . The integrand collapses: $\dfrac{\mathrm{d}r}{\sqrt{2GM/r-1}}=GM\sin\eta\,\tan(\eta/2)\,\mathrm{d}\eta=GM(1-\cos\eta)\,\mathrm{d}\eta$ . Why: the cycloid is the exact free-fall solution, and it is precisely the substitution that turns the square root into an elementary trigonometric integrand.

Step 4. Integrate over the half-cycle from horizon $(\eta=\pi,\ r=2GM)$ to centre $(\eta=0,\ r=0)$ : $\tau_{\max}=GM\displaystyle\int_0^{\pi}(1-\cos\eta)\,\mathrm{d}\eta=GM\big[\eta-\sin\eta\big]_0^{\pi}=\pi GM$ , i.e. $\pi GM/c^3$ in units. Why: the $\sin\eta$ term vanishes at both endpoints, leaving the clean $\pi GM$ — the coordinate-invariant maximum survival time, independent of maneuvering.

Step 5. Put in numbers. With $GM_\odot/c^3=4.93\,\mu\mathrm{s}$ , the maximum is $\tau_{\max}=\pi(4.93)\approx15.5\,\mu\mathrm{s}$ per solar mass, scaling linearly with $M$ . Why: the linear scaling is what makes supermassive holes locally survivable for far longer than stellar ones, and it is the number every worked example below builds on. *Verified: matches MTW §31.4 and Carroll §5.7; the cycloid $r=\tfrac12 r_{\max}(1+\cos\eta)$ with $r_{\max}=2GM$ gives $\Delta\tau=\pi GM$ , and $\pi GM_\odot/c^3=\pi\times4.93\,\mu\mathrm{s}=15.5\,\mu\mathrm{s}$ by direct evaluation.*

Go deeper · Derivation C · 1 minWhy r equals zero is unavoidable inside the horizon

Derivation CWhy r equals zero is unavoidable inside the horizon

The goal is to prove that inevitability is not rhetoric but a theorem about causal structure: for any future-directed timelike worldline in $r<2GM$ — accelerated or geodesic, powered by any engine — the areal radius $r$ must strictly decrease and reach zero in finite proper time. The argument uses only the interior metric signature and the requirement that the four-velocity be timelike and future-directed, converting "the cones tip over" into an inequality with no escape clause.

Step 1. Write the normalization $g_{\mu\nu}u^\mu u^\nu=-1$ using the interior form of the metric, in which $(2GM/r-1)>0$ : $\left(\tfrac{2GM}{r}-1\right)^{-1}\dot r^2-\left(\tfrac{2GM}{r}-1\right)\dot t^2-r^2\big(\dot\theta^2+\sin^2\theta\,\dot\varphi^2\big)=1$ . Why: writing the factor as the manifestly positive $(2GM/r-1)$ exposes that the $\dot r^2$ term now carries the timelike role, which is the whole content of the space-time swap.

Step 2. Isolate the radial term: $\left(\tfrac{2GM}{r}-1\right)^{-1}\dot r^2=1+\left(\tfrac{2GM}{r}-1\right)\dot t^2+r^2\big(\dot\theta^2+\sin^2\theta\,\dot\varphi^2\big)$ . Every term on the right is non-negative for $r<2GM$ . Why: because each right-hand term is a square times a positive coefficient, the left side is bounded below by $1$ regardless of how the observer distributes motion among $t$ , $\theta$ , $\varphi$ — thrust cannot make it smaller.

Step 3. Multiply by $(2GM/r-1)$ to get $\dot r^2\ge \tfrac{2GM}{r}-1>0$ for all $r<2GM$ . Hence $\dot r$ never vanishes inside the horizon, so $r$ is monotonic along the worldline; future-directedness fixes the sign as $\dot r<0$ . Why: a quantity whose derivative never vanishes and never changes sign is strictly monotone — this is the rigorous replacement for "you can't turn around", and it holds for accelerated observers too since we never assumed a geodesic.

Step 4. Bound the remaining time. From any interior radius $r_0$ , $\tau\le\displaystyle\int_0^{r_0}\frac{\mathrm{d}r}{\sqrt{2GM/r-1}}<\infty$ : the integral converges, so $r=0$ is reached in finite proper time no matter what. Why: convergence turns "r decreases" into "r hits zero soon" — the singularity is not merely approached but attained, within the previous derivation's $\pi GM/c^3$ . *Verified: consistent with Wald §6.4 — the future-pointing condition plus the interior signature forces $r$ monotone decreasing — and with the global Kruskal–Szekeres picture, in which the two spacelike singularity lines bound the interior to the future.*

Limiting cases

How the geometry and the infall behave from far outside to the singularity:

Case	Result	Meaning
Far exterior $r\gg2GM$	$g_{tt}\to-1,\ g_{rr}\to+1$	Nearly flat spacetime; $r$ is an ordinary spatial radius, static observers exist, Newtonian gravity is recovered.
Horizon $r\to2GM$	$g_{tt}\to0,\ g_{rr}\to\infty$ , $K=3/(4G^4M^4)$	Coordinate (not curvature) singularity; removable in Kruskal–Szekeres or Eddington–Finkelstein charts; a faller crosses smoothly in finite $\tau$ .
Interior $0<r<2GM$	$g_{tt}>0,\ g_{rr}<0$	Signature preserved but roles swapped: $r$ timelike, $t$ spacelike; no static observers; $\mathrm{d}r<0$ forced to the future.
Marginal drop $\tilde E\to0$	$\tau_{\max}=\pi GM/c^3$	Longest possible survival, horizon to singularity; realized by crossing the horizon at rest, $\approx15.5\,(M/M_\odot)\,\mu\mathrm{s}$ .
Singularity $r\to0$	$K\sim48G^2M^2/r^6\to\infty$	Genuine curvature singularity; unbounded tidal stretching; spacelike future endpoint of every interior causal curve.

Common trapThe interior is a place you could hover in with a big enough rocket

It is not. There are no static observers anywhere inside the horizon, because $\partial_t$ is spacelike there — a worldline of fixed $r$ would be a spacelike (faster-than-light) curve no massive body can follow. The "radial distance" $r$ has become a clock running down, so thrust cannot stop the fall any more than a rocket can stop next Tuesday; it only changes which timelike line inside the future cone you ride, and every such line ends at $r=0$ . Two traps ride along. First, do not confuse the removable coordinate singularity at $r=2GM$ (finite curvature, $K=3/(4G^4M^4)$ ) with the genuine curvature singularity at $r=0$ (invariants diverge as $r^{-6}$ ). Second, firing engines shortens survival, not extends it, because any added radial speed or energy raises $\tilde E$ and shrinks the proper-time integral. The record holder does nothing at all.

Worked examples

1 · Maximum survival time across three black holes (fully worked)

How long, at most, does an observer live between crossing the horizon and hitting the singularity of (a) a $10\,M_\odot$ stellar black hole, (b) Sagittarius A\, the $4.3\times10^6\,M_\odot$ hole at the Galactic centre, and (c) M87\, the $6.5\times10^9\,M_\odot$ giant imaged by the Event Horizon Telescope?

Step 1. Use the maximum from Derivation 2, $\tau_{\max}=\pi GM/c^3$ , with the per-solar-mass constant $GM_\odot/c^3=4.93\,\mu\mathrm{s}$ , so $\tau_{\max}=15.5\,(M/M_\odot)\,\mu\mathrm{s}$ . The relation is exactly linear in $M$ , so each case is one multiplication.

Step 2. Stellar hole: $\tau_{\max}=15.5\,\mu\mathrm{s}\times10=\mathbf{155\,\mu s}$ — far too brief to notice, let alone react to.

Step 3. Sagittarius A\*: $\tau_{\max}=15.5\,\mu\mathrm{s}\times4.3\times10^6\approx\mathbf{67\ seconds}$ — about a minute between the horizon and oblivion.

Step 4. M87\*: $\tau_{\max}=15.5\,\mu\mathrm{s}\times6.5\times10^9\approx1.0\times10^5\ \mathrm{s}\approx\mathbf{28\ hours}$ . The larger the hole, the longer the interior — survival grows linearly with $M$ , from microseconds to more than a day. In every case, no engine can extend the number: thrust only shortens it.

2 · Tidal shredding: where spaghettification actually happens

Radial tidal stretching per unit length is governed by the geodesic-deviation eigenvalue $2GM/r^3$ . Evaluate it at the horizon of a $10\,M_\odot$ hole and of Sagittarius A\, and say where an unprotected human is torn apart in each case.*

Step 1. At the horizon $r_s=2GM/c^2$ , the radial tidal gradient is $2GM/r_s^3=c^6/(4G^2M^2)$ , scaling as $1/M^2$ — so bigger holes have gentler horizons. Evaluate this one formula twice.

Step 2. Stellar hole ( $10\,M_\odot$ ): $2GM/r_s^3\approx1.0\times10^{8}\ \mathrm{s^{-2}}$ . Across a $\sim2\,\mathrm{m}$ body the differential acceleration is $\approx2\times10^{8}\,\mathrm{m/s^2}\approx\mathbf{2\times10^{7}\,g}$ — lethal, and in fact you are shredded outside the horizon, well before reaching $r_s$ .

Step 3. Sagittarius A\* ( $4.3\times10^6\,M_\odot$ ): the same gradient is $\approx6\times10^{-4}\,\mathrm{s^{-2}}$ , giving $\approx1\times10^{-3}\,\mathrm{m/s^2}\approx\mathbf{10^{-4}\,g}$ across your body — utterly imperceptible. You cross the supermassive horizon completely intact.

Step 4. The two lethal moments decouple with mass. For the stellar hole, tides kill you before the horizon; for the supermassive one, you cross unharmed and are spaghettified only deep inside, near $r=0$ , within the $\sim67\,\mathrm{s}$ the previous example allotted. The singularity is unavoidable either way — only the manner and timing of destruction differ.

Physics of the portrait

Understanding the interior demanded a leap most founders resisted: taking seriously that a star could collapse through its own horizon to a singularity. The person who first computed that continued collapse was J. Robert Oppenheimer.

Origins · 1 minJ. Robert Oppenheimer and how this was discovered

J. Robert Oppenheimer

1904–1967 · continued gravitational collapse and the interior of a collapsing star

Oppenheimer is remembered by the public for Los Alamos, but among relativists his most enduring work is a pair of 1939 papers with his students. In "On Massive Neutron Cores" (with George Volkoff, extending Richard Tolman) he found the maximum mass a cold neutron star can support — the Tolman–Oppenheimer–Volkoff limit — establishing that heavy enough remnants have nothing left to hold them up. He then asked the dreaded next question. In "On Continued Gravitational Contraction," with Hartland Snyder, he modeled a pressureless dust sphere collapsing through its Schwarzschild radius. The result was the first honest description of the black-hole interior: to a distant observer the star freezes and reddens at the horizon, but to an observer riding the surface the collapse continues smoothly across $r=2GM$ and reaches the singularity in finite proper time — exactly the space-time swap and inevitable infall this article makes precise. Published on 1 September 1939, it was buried by the outbreak of war and Wheeler's early skepticism; only in the 1960s — with Finkelstein's one-way membrane, Kruskal and Szekeres's maximal extension, and Penrose's singularity theorem — was the interior Oppenheimer and Snyder had glimpsed fully understood. The modern picture of what lies beyond a horizon begins with his calculation.

History

1916

— Schwarzschild publishes the vacuum solution outside a spherical mass

1939

— Oppenheimer and Snyder compute continued gravitational collapse to a singularity

1958

— Finkelstein interprets r = 2GM as a one-way causal membrane

1960

— Kruskal and Szekeres give the maximal analytic extension of the geometry

1965

— Penrose proves the first singularity theorem, making infall generic

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §31.4–31.6 (interior and free fall) · Wald, General Relativity, §6.4 (Schwarzschild interior and global structure) · Carroll, Spacetime and Geometry, §5.6–5.7 (radial geodesics, Kruskal extension) · Hartle, Gravity, ch. 12 · Schutz, A First Course in General Relativity, ch. 11.

Provenance. The line element, the sign reversal of $g_{tt},g_{rr}$ , and $K=48G^2M^2/r^6$ are standard (Carroll §5.6, MTW §31); $\dot r^2=\tilde E^2-1+2GM/r$ follows from normalization plus the $\partial_t$ -Killing energy. A verifier should: (1) confirm the horizon factor cancels in Derivation 1; (2) recompute $\tau_{\max}=\pi GM$ from the cycloid $r=\tfrac12 r_{\max}(1+\cos\eta)$ , $r_{\max}=2GM$ ; (3) check $GM_\odot/c^3=4.93\,\mu\mathrm{s}$ , hence $\pi GM_\odot/c^3\approx15.5\,\mu\mathrm{s}$ (an earlier draft's $6.57\,\mu\mathrm{s}$ was an error, corrected here); (4) verify the tidal gradient $c^6/(4G^2M^2)\approx10^8\,\mathrm{s^{-2}}$ at a $10\,M_\odot$ horizon. Masses (Sgr A\ $4.3\times10^6\,M_\odot$ , M87\ $6.5\times10^9\,M_\odot$ ) are standard.

Prerequisites. The Schwarzschild Metric · Schwarzschild Radius and the Geometric Mass · Geodesic Equations in Schwarzschild Spacetime · Killing Vectors and Conserved Quantities in Schwarzschild. Leads to. Birkhoff's Theorem · Effective Potential for Massive Particle Orbits · Gravitational Redshift in Schwarzschild Spacetime.

Ask the tutor: Curious what it would actually feel like to fall in — the tides, the tipping cones, the last microseconds — ask me to walk you through the plunge step by step.

Does space really "turn into time" inside a black hole, or is that just loose talk?

It is precise, not loose — with one clarification. The signature is always $(-,+,+,+)$ ; nothing about the count of time and space directions changes. What changes is which coordinate carries the timelike minus sign. Outside, $g_{tt}<0$ , so $t$ is time and $r$ is spatial. Inside, $g_{tt}>0$ and $g_{rr}<0$ , so now $r$ is the direction along which proper time accumulates and $t$ labels a spatial direction. Because a timelike coordinate can only run one way toward the future, $r$ can only decrease. So "space turns into time" means exactly that the roles of $t$ and $r$ swap — and that swap is what makes reaching $r=0$ inevitable.

If I fell in, could I fire my rockets to buy more time before the singularity?

No — and worse, thrusting shortens your remaining time. The maximum, $\pi GM/c^3$ , is achieved by the observer who does nothing: crosses at zero radial speed and free-falls. Any engine burn, in any direction, adds energy and radial motion, raising the conserved $\tilde E$ and shrinking the proper-time integral $\int \mathrm{d}r/|\dot r|$ . Intuitively, the singularity is a moment in your future, and rushing around inside the light cone only gets you there sooner — exactly as accelerating hard in flat space brings your future forward relative to a coasting twin. The way to live longest inside a black hole is to relax completely.

How can r = 0 be a "moment in time" rather than a place at the centre?

Because inside the horizon $r$ is a timelike coordinate. A surface of constant $r$ there is a surface of constant time — a spacelike hypersurface, the same kind of object as "the instant 3 p.m. everywhere". The surface $r=0$ is the last such instant, lying to the future of every interior event. This is why you cannot point to it or orbit it: no spatial direction leads to $r=0$ , only the forward direction of time. On a Kruskal or Penrose diagram the singularity is drawn as a horizontal (spacelike) line across the top, not a vertical world-line at the centre — precisely because it is a time, not a place.

Why is the survival time so short — only microseconds for a stellar black hole?

Because the only length scale is the Schwarzschild radius $2GM/c^2$ , and its time scale $2GM/c^3$ is under ten microseconds for a solar mass. The maximum interior time $\pi GM/c^3\approx15.5\,\mu\mathrm{s}\times(M/M_\odot)$ is just this scale times $\pi/2$ , growing linearly with mass: a stellar hole gives microseconds, Sagittarius A\* about a minute, a $10^9\,M_\odot$ giant more than a day. The brevity for small holes really says that stellar black holes are physically tiny — a few kilometres across — so light itself crosses them in microseconds, and no massive observer can take longer to fall the same distance.

Is the singularity at r = 0 where the collapsed star's matter sits?

No — this is a persistent misconception. In the idealized Schwarzschild picture the matter is not sitting at $r=0$ waiting; the singularity is in the future, not at a central point in space. In the Oppenheimer–Snyder collapse the star's surface itself falls to $r=0$ and the singularity forms only at the end. For an eternal Schwarzschild hole there is no matter at all — pure vacuum curvature, with $r=0$ a future boundary of spacetime rather than a place occupied by anything. Picturing the singularity as "a dense dot at the centre" imports a spatial, Newtonian image that the timelike character of $r$ specifically forbids.

Does the interior sign flip mean the horizon is a special local place?

No — and that is the subtle point. The sign flip of $g_{tt}$ and $g_{rr}$ is a feature of the Schwarzschild chart, not of the geometry at $r=2GM$ . A freely falling observer crossing the horizon measures nothing unusual locally: curvature is finite, tides are gentle for a large hole, no experiment pins down the crossing moment. The horizon is a global feature — the boundary of the region from which no signal escapes — defined by the whole future of the spacetime, not by local measurements. The components misbehave there only because static Schwarzschild coordinates are tied to hovering observers who cease to exist at $r=2GM$ ; in Eddington–Finkelstein or Kruskal–Szekeres coordinates the crossing is manifestly smooth.