Tortoise Coordinate and Radial Null Geodesics

The Schwarzschild line element in its textbook coordinates $(t,r,\theta,\phi)$ carries a scar at the Schwarzschild radius $r=2GM$ : the component $g_{tt}=-(1-2GM/r)$ vanishes and $g_{rr}=(1-2GM/r)^{-1}$ blows up. Yet every curvature invariant there — the Kretschmann scalar $R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}=48G^2M^2/r^6$ chief among them — stays perfectly finite, so nothing physical is wrong. What has broken is the map, not the territory. The tortoise coordinate is the sharpest instrument for exposing that breakage and putting the geometry into the form later tools need. The construction is almost embarrassingly simple: rescale the radial coordinate so its differential absorbs the offending metric factor, $\mathrm{d}r_*=\mathrm{d}r/f$ with $f=1-2GM/r$ . Two payoffs follow at once and organize this whole article. First, the two-dimensional radial block becomes conformally flat, $\mathrm{d}s^2_{(2)}=f(-\mathrm{d}t^2+\mathrm{d}r_*^2)$ , so radial light rays are exact $\pm45^\circ$ lines $t=\pm r_*+\text{const}$ and the causal structure is laid bare — the horizon simply recedes to $r_*\to-\infty$ . Second, any field separated on the background turns its radial equation into a flat one-dimensional wave equation with a potential bump, $\partial_t^2\psi-\partial_{r_*}^2\psi+V_\ell\psi=0$ , the launching pad for black-hole scattering, quasinormal ringdown, and greybody factors. Wheeler's whimsical name recalls Zeno: in $r_*$ a photon forever approaches the horizon and never overtakes it, like Achilles and the tortoise.

Key takeaways

The tortoise coordinate $r_*$ is the radial reparametrization defined by $\mathrm{d}r_*=(1-2GM/r)^{-1}\mathrm{d}r$ , integrating to $r_*=r+2GM\ln|r/2GM-1|$ .
It stretches the exterior $r\in(2GM,\infty)$ onto the full line $r_*\in(-\infty,+\infty)$ , pushing the event horizon out to $r_*\to-\infty$ .
In the $(t,r_*)$ plane radial null geodesics become straight $\pm45^\circ$ lines $t=\pm r_*+\text{const}$ , so light cones look exactly Minkowskian and "infinite coordinate time to reach the horizon" is exposed as a slicing artefact.
It flattens the radial wave operator into the Regge–Wheeler master form $\partial_t^2\psi-\partial_{r_*}^2\psi+V_\ell\psi=0$ , a scattering problem with a potential peaked near the photon sphere $r=3GM$ .
The null coordinates $u=t-r_*$ and $v=t+r_*$ built from it seed the Eddington–Finkelstein and Kruskal–Szekeres charts that remove the coordinate singularity.
Key idea: replacing $r$ by $r_*$ , defined through $\mathrm{d}r_*=\mathrm{d}r/f$ with $f=1-2GM/r$ , makes the $t$ – $r$ block conformally flat, $\mathrm{d}s^2_{(2)}=f(-\mathrm{d}t^2+\mathrm{d}r_*^2)$ , so radial null cones are exactly $\pm45^\circ$ and the horizon recedes to $r_*=-\infty$ .

The core insight

Replacing $r$ by $r_*$ , defined through $\mathrm{d}r_*=\mathrm{d}r/f$ with $f=1-2GM/r$ , makes the $t$ – $r$ block of the Schwarzschild metric conformally flat, $\mathrm{d}s^2_{(2)}=f(-\mathrm{d}t^2+\mathrm{d}r_*^2)$ , so radial null cones are exactly $\pm45^\circ$ and the horizon recedes to $r_*=-\infty$ .

Where this lives today

The tortoise coordinate stopped being a pedagogical curiosity the moment gravitational-wave astronomy became real. When two black holes merge, the remnant settles by radiating damped sinusoids — the quasinormal modes — which are precisely the eigenfrequencies of the scattering problem (7) with outgoing conditions at $r_*\to+\infty$ and purely ingoing conditions at the horizon $r_*\to-\infty$ . For a non-rotating hole the fundamental $\ell=2$ mode sits at $2GM\omega/c^3\approx0.3737-0.0890\,i$ , and for GW150914 — a remnant of roughly $62\,M_\odot$ — this becomes a ringdown tone near $250\ \mathrm{Hz}$ decaying in about $4\ \mathrm{ms}$ , the numbers LIGO fit in 2015 to test the black-hole hypothesis. The eikonal limit makes the geometry vivid: the real frequency is $\ell$ times the photon-sphere orbital frequency $\Omega_c=c^3/(3\sqrt{3}\,GM)$ and the damping is the Lyapunov exponent of that unstable orbit, so the barrier at $r=3GM$ and the ringing of a real merger are the same physics. The transmission through that barrier is the greybody factor reshaping Hawking's spectrum, and the near-horizon exponential (9) is what makes the Kruskal chart analytic. The construction generalizes intact: any static spherically symmetric lapse $f(r)$ defines its own $r_*=\int\mathrm{d}r/f$ , with Reissner–Nordström pushing each of its two horizons to a tortoise infinity.

The physical picture

Near the horizon a static observer's clock runs slow relative to bookkeeper time $t$ by the factor $\sqrt{f}$ , and the coordinate machinery measuring radial distance stretches correspondingly. The tortoise coordinate is engineered to soak up exactly that stretching: it takes the thin shell of ordinary radius just outside $r=2GM$ and unrolls it into an infinitely long throat. What was a finite region where the coordinates misbehave becomes an infinite region where they behave impeccably — a finite-but-singular chart traded for an infinite-but-smooth one, and the smoothness lets radial light cones spring back open to $45^\circ$ everywhere. A useful second reading of $r_*$ is as a "light-travel coordinate". The extra coordinate time a radial photon burns climbing out of the well near the horizon is booked, one for one, as extra $r_*$ -length. Because the well is logarithmically deep, that length runs to infinity — the same fact a distant astronomer reports as an infalling beacon freezing and reddening away without ever quite crossing. The first figure turns this into arithmetic, plotting $r_*(r)$ so the divergence is something you read off a curve rather than take on faith. In the wave picture the payoff is different but no less clean. A field on a curved background normally drags a tangle of metric factors through its equation of motion. Switching to $r_*$ sweeps every one of them into a single localized potential $V_\ell(r_*)$ perched near the photon sphere and leaves the flat one-dimensional wave operator behind. The problem collapses to "a wave on an infinite flat line scattering off a hill", which is why black-hole perturbation theory borrows the whole toolbox of one-dimensional quantum scattering — barriers, transmission coefficients, resonances — wholesale.

Figure

The tortoise map r_ = r + 2GM ln(r/2GM − 1) plotted in units of 2GM: the curve dives to −∞ at the horizon r = 2GM and merges with the 45° asymptote r_ ≈ r far away, and at the photon sphere r = 3GM it reads r_*/2GM = 0.807

Figure

Radial light in the (t, r_) plane: ingoing and outgoing null rays are exact ±45° straight lines t = ± r_ + const, so the light cones stay undistorted while the horizon recedes to r_* → −∞ and an infalling signal takes t → ∞ to fade out

The core equations

Everything begins with the Schwarzschild exterior geometry, the unique static spherically symmetric vacuum solution. Written in geometric units $c=1$ it isolates the lapse function $f(r)=1-2GM/r$ , whose single zero at $r=2GM$ is the source of all the coordinate trouble the tortoise coordinate is built to cure.

\mathrm{d}s^2 = -\Big(1-\frac{2GM}{r}\Big)\mathrm{d}t^2 + \Big(1-\frac{2GM}{r}\Big)^{-1}\mathrm{d}r^2 + r^2\,\mathrm{d}\Omega^2 \qquad (1)

where: $G,M$ — Newton's constant and the hole's mass (geometric units, $c=1$ ); $r$ — areal radius, so a sphere at fixed $r$ has area $4\pi r^2$ ; $\mathrm{d}\Omega^2=\mathrm{d}\theta^2+\sin^2\theta\,\mathrm{d}\phi^2$ — unit-sphere metric; $f(r)\equiv1-2GM/r$ — lapse function, vanishing at the horizon. Reading: the whole pathology lives in the $tt$ and $rr$ entries; the angular part is regular. The defining relation is the one design choice in the entire subject. We demand a new radial variable whose differential cancels the diverging $g_{rr}$ , i.e. one built from $\mathrm{d}r/f$ .

\frac{\mathrm{d}r_*}{\mathrm{d}r} = \frac{1}{f(r)} = \Big(1-\frac{2GM}{r}\Big)^{-1} \qquad (2)

where: $r_*$ — the tortoise (Regge–Wheeler) radial coordinate. Reading: because $1/f\to\infty$ as $r\to2GM^+$ , an infinitesimal step in $r$ near the horizon costs an unbounded step in $r_*$ — the near-horizon shell is being stretched without limit. Integrating (2) in closed form gives the coordinate itself. The logarithm is the fingerprint of the simple zero of $f$ , and its argument vanishes exactly at the horizon.

r_* = r + 2GM\,\ln\!\left|\frac{r}{2GM} - 1\right| + \text{const} \qquad (3)

where: $\ln$ — natural logarithm; the argument $\to0$ at $r=2GM$ , so $r_*\to-\infty$ there; $\text{const}$ — integration constant, conventionally set to zero. Reading: at large $r$ the logarithm is a slow correction and $r_*\approx r$ ; at the horizon it diverges logarithmically. Feeding $\mathrm{d}r_*^2=f^{-2}\mathrm{d}r^2$ back into the radial two-metric produces the result that makes the causal structure manifest: a Minkowski two-metric times an overall conformal factor $f$ .

\mathrm{d}s^2_{(2)} = \Big(1-\frac{2GM}{r}\Big)\big(-\mathrm{d}t^2 + \mathrm{d}r_*^2\big) \qquad (4)

where: $r=r(r_*)$ — now an implicit function of $r_*$ through (3). Reading: a positive conformal factor rescales lengths but leaves null cones untouched, so light in this block behaves exactly as in flat spacetime. Setting $\mathrm{d}s^2_{(2)}=0$ for radial light and dividing out the conformal factor collapses the null condition to unit slope. Radial null geodesics are straight lines.

t = \pm\, r_* + \text{const} \qquad (5)

where: the $+$ branch is an outgoing ray ( $r_*$ rising with $t$ ); the $-$ branch is an ingoing ray ( $r_*$ falling with $t$ ). Reading: plotted against $r_*$ , every radial light ray is an exact $\pm45^\circ$ line — the light cones never tip over, no matter how close to the horizon. From the two null families one names the retarded and advanced null coordinates. These are the raw material of every regular chart that follows.

u = t - r_*,\qquad v = t + r_* \qquad (6)

where: $u$ — retarded time, constant along an outgoing ray; $v$ — advanced time, constant along an ingoing ray. Reading: holding $v$ fixed labels ingoing wavefronts and yields the horizon-crossing Eddington–Finkelstein chart; combining $u$ and $v$ exponentially yields Kruskal–Szekeres. For wave propagation the coordinate earns its keep a second time. Separating a scalar field on the background and trading $r$ for $r_*$ turns the radial dynamics into a flat one-dimensional wave equation carrying an effective potential.

\frac{\partial^2 \psi}{\partial t^2} - \frac{\partial^2 \psi}{\partial r_*^2} + V_\ell(r)\,\psi = 0 \qquad (7)

where: $\psi$ — radial mode function, with the field written $\Phi=\sum_{\ell m}\frac{\psi(t,r)}{r}Y_{\ell m}$ ; $V_\ell$ — effective potential; $\ell$ — angular-momentum multipole. Reading: this is the canonical form of one-dimensional scattering, with $r_*$ running over the entire real line. The potential collects every metric factor into one localized barrier. For a massless scalar it reads as below; the same equation with the curvature term altered describes gravitational (Regge–Wheeler / Zerilli) perturbations.

V_\ell(r) = \Big(1-\frac{2GM}{r}\Big)\!\left[\frac{\ell(\ell+1)}{r^2} + \frac{2GM}{r^3}\right] \qquad (8)

where: $\ell(\ell+1)/r^2$ — centrifugal barrier; $2GM/r^3$ — spin-0 curvature term, replaced by $-6GM/r^3$ for axial gravitational perturbations. Reading: $V_\ell$ vanishes at both ends ( $r_*\to\pm\infty$ ) and peaks near the photon sphere $r\approx3GM$ — the barrier waves must tunnel through. Finally, near the horizon the closed form (3) inverts to an exponential relation between ordinary and tortoise radius, controlled by the surface gravity. This single exponential is the analytic key to the Kruskal extension and to horizon thermodynamics.

\kappa = \frac{1}{4GM},\qquad r - 2GM \;\propto\; e^{\,r_*/(2GM)} = e^{\,2\kappa r_*} \qquad (9)

where: $\kappa$ — surface gravity of the Schwarzschild horizon. Reading: the depth scale $2GM=1/(2\kappa)$ that sets the logarithm in (3) is the same $\kappa$ that fixes the Hawking temperature $T_H=\kappa/2\pi=1/(8\pi GM)$ .

Derivation AIntegrating the defining ODE to the logarithmic tortoise coordinate

The goal is to take the one-line definition $\mathrm{d}r_*/\mathrm{d}r=1/f$ and integrate it in closed form, so that the abstract demand "cancel the diverging metric factor" becomes the concrete function $r_*=r+2GM\ln|r/2GM-1|$ , and so its logarithmic divergence at the horizon and asymptotic agreement with $r$ at infinity both fall out explicitly rather than being asserted. The calculation is elementary partial fractions, and we finish by differentiating back to confirm nothing was inverted incorrectly.

Step 1. Clear the fraction in the integrand: $\dfrac{1}{f}=\dfrac{1}{1-2GM/r}=\dfrac{r}{r-2GM}$ , multiplying top and bottom by $r$ . Why: this exposes a rational function whose numerator and denominator differ only by a constant, which is exactly the shape polynomial division simplifies.

Step 2. Do the division: $\dfrac{r}{r-2GM}=1+\dfrac{2GM}{r-2GM}$ , splitting off the constant piece. Why: the "1" will integrate to $r$ and reproduce the far-field behaviour, while the remainder carries all the horizon structure into a single logarithmic term.

Step 3. Integrate term by term, $r_*=\displaystyle\int\!\Big(1+\frac{2GM}{r-2GM}\Big)\mathrm{d}r = r + 2GM\ln|r-2GM| + C$ , and absorb the constant $-2GM\ln(2GM)$ into $C$ to make the logarithm's argument dimensionless, giving $r_*=r+2GM\ln|r/2GM-1|$ . Why: a dimensionless argument is what lets the same formula be read in geometric units or in kilometres, and it is the standard convention in MTW and Carroll.

Step 4. Read off the two limits. As $r\to2GM^+$ the argument $\to0^+$ and $\ln\to-\infty$ , so $r_*\to-\infty$ ; as $r\to\infty$ the logarithm grows only like $\ln r$ while the leading $r$ dominates, so $r_*\to r$ . Why: these two limits are the entire content of the coordinate — an infinite throat at the horizon and Minkowski agreement at infinity — and they are now proven, not assumed. *As a check, differentiating the closed form gives $\mathrm{d}r_*/\mathrm{d}r = 1 + 2GM/(r-2GM) = r/(r-2GM) = 1/f$ , recovering the defining ODE exactly.*

Go deeper · Derivation B · 1 minFlattening the scalar wave equation into a one-dimensional scattering problem

Derivation BFlattening the scalar wave equation into a one-dimensional scattering problem

The goal is to show that a massless scalar obeying $\Box\Phi=0$ on the Schwarzschild background, once separated in spherical harmonics and re-expressed in the tortoise coordinate, loses every explicit metric factor from its radial dynamics and lands on the flat wave operator $\partial_t^2-\partial_{r_*}^2$ plus a localized potential $V_\ell$ — the statement that black-hole perturbations are nothing but one-dimensional waves scattering off a barrier. The engine of the simplification is the single identity $\partial_{r_*}=f\,\partial_r$ , the defining relation read as an operator.

Step 1. Write the curved d'Alembertian $\Box\Phi=\dfrac{1}{\sqrt{-g}}\partial_\mu(\sqrt{-g}\,g^{\mu\nu}\partial_\nu\Phi)=0$ with $\sqrt{-g}=r^2\sin\theta$ , $g^{tt}=-1/f$ , and $g^{rr}=f$ . Why: these three metric ingredients are all the geometry the wave equation ever sees, and each will be disposed of in turn.

Step 2. Separate variables as $\Phi=\sum_{\ell m}\dfrac{\psi(t,r)}{r}\,Y_{\ell m}(\theta,\phi)$ and use $\nabla^2_{S^2}Y_{\ell m}=-\ell(\ell+1)Y_{\ell m}$ . Why: the spherical harmonics diagonalize the angular Laplacian, converting the two angular derivatives into the algebraic centrifugal term $\ell(\ell+1)/r^2$ , and the $1/r$ prefactor is chosen precisely so the radial equation will have no first-derivative term.

Step 3. Apply the operator identity $f\,\partial_r=\partial_{r_*}$ , which follows from $\mathrm{d}r_*/\mathrm{d}r=1/f$ , to the radial part. The $g^{tt}=-1/f$ factor multiplies the whole equation by $f$ , turning the time term into a bare $\partial_t^2\psi$ and the radial term into $\partial_{r_*}^2\psi$ . Why: this is the pivotal move — the same $1/f$ that made the coordinates singular is exactly what the change of variable cancels, leaving a flat operator.

Step 4. Collect the leftover algebraic pieces into $V_\ell(r)=f\big[\ell(\ell+1)/r^2+f'/r\big]$ and evaluate $f'=2GM/r^2$ to get $V_\ell=(1-2GM/r)\big[\ell(\ell+1)/r^2+2GM/r^3\big]$ . Why: every surviving metric factor has been swept into this one function of position, so the dynamics reduce to $\partial_t^2\psi-\partial_{r_*}^2\psi+V_\ell\psi=0$ on the full line $r_*\in(-\infty,\infty)$ . *As a check on the limits: $V_\ell\to\ell(\ell+1)/r^2$ (the flat centrifugal barrier) as $M\to0$ , and $V_\ell\to0$ at both $r_*\to+\infty$ (since $f\to1$ but $1/r^2\to0$ ) and $r_*\to-\infty$ (since $f\to0$ ) — the standard free-scattering asymptotics confirming the potential vanishes at both ends of the line.*

Go deeper · Derivation C · 1 minWhy coordinate time to reach the horizon diverges

Derivation CWhy coordinate time to reach the horizon diverges

The goal is to show quantitatively that a radially infalling photon takes an infinite amount of Schwarzschild coordinate time $t$ to reach the horizon — the origin of the famous "frozen star" appearance — and to expose that this divergence is nothing but the divergence of $r_*$ we already found, hence a statement about the chart, not about anything a falling observer experiences. We integrate the ingoing null condition and watch $r_*$ supply the infinity transparently.

Step 1. For an ingoing radial photon the null condition (5) reads $t=-r_*+\text{const}$ , so $\dfrac{\mathrm{d}t}{\mathrm{d}r}=-\dfrac{1}{f}$ , the slope being set by the same $1/f$ that defines $r_*$ . Why: the coordinate time elapsed is therefore literally an integral of $\mathrm{d}r/f$ , i.e. a difference of tortoise coordinates.

Step 2. Integrate inward from a starting radius $r_1$ to a shell at $r=2GM+\epsilon$ : $\Delta t=\displaystyle\int_{2GM+\epsilon}^{r_1}\frac{\mathrm{d}r}{f}=r_*(r_1)-r_*(2GM+\epsilon)$ . Why: the elapsed coordinate time is exactly the tortoise separation between the two radii — the coordinate was built to make this identity hold.

Step 3. Take the shell to the horizon. Since $r_*(2GM+\epsilon)\approx2GM\ln(\epsilon/2GM)\to-\infty$ as $\epsilon\to0$ , the elapsed time $\Delta t\to+\infty$ : the photon never arrives at $r=2GM$ in coordinate time. Why: the logarithmic depth of the tortoise throat is precisely the infinite coordinate delay a distant bookkeeper assigns to the crossing. *Note the contrast: an infalling observer logs a finite affine parameter and finite proper time across the horizon; the divergence sits only in $t$ and $r_*$ , confirming that $r=2GM$ is a coordinate singularity, not a curvature singularity.*

Limiting cases

Behaviour of $r_*=r+2GM\ln|r/2GM-1|$ and the surrounding machinery in the regimes that matter:

Case	Result	Meaning
Far field $r\gg2GM$	$r_*\approx r+2GM\ln(r/2GM)\to r$	the logarithm is a slow correction; $r_*$ tracks $r$ and the spacetime is asymptotically Minkowski, so cones are already nearly $45^\circ$
Near horizon $r\to2GM^+$	$r_*\approx2GM\ln\!\big(\tfrac{r-2GM}{2GM}\big)\to-\infty$	logarithmic divergence: the horizon lives at $r_*=-\infty$ , reached only as $t\to+\infty$ along ingoing null rays
Flat-space limit $M\to0$	$r_*\to r,\ f\to1$	the tortoise coordinate degenerates to the ordinary radius; the throat vanishes and Minkowski is recovered
Surface-gravity scale	$r-2GM\propto e^{2\kappa r_*},\ \kappa=1/4GM$	the logarithmic depth is set by $2GM=1/2\kappa$ , tying $r_*$ to the temperature $T_H=\kappa/2\pi$
Eikonal potential $\ell\to\infty$	$V_\ell\to f\,\ell(\ell+1)/r^2$ , peak at $r=3GM$	the barrier localizes on the photon sphere, linking scattering to the unstable light orbit

Worked examples

1 · Reading the tortoise map, point by point (confidence builder)

Work in units of the Schwarzschild radius, $x\equiv r/2GM$ , so that $r_*/2GM=x+\ln(x-1)$ . Tabulate $r_*$ at the photon sphere, at $r=4GM$ , and at two radii approaching the horizon, and interpret the spacing.

Step 1. Photon sphere, $r=3GM$ so $x=1.5$ : $r_*/2GM=1.5+\ln(0.5)=1.5-0.693=\mathbf{0.807}$ . A modest positive number — the barrier that governs wave scattering sits at a perfectly ordinary tortoise location, only $0.8$ throat-widths outside $r_*=0$ .

Step 2. At $r=4GM$ , $x=2$ : $r_*/2GM=2+\ln(1)=\mathbf{2.000}$ , since the logarithm's argument is exactly $1$ . This is the tidy reference point where $r_*$ and $r-2GM$ momentarily coincide in these units.

Step 3. Approaching the horizon, $x=1.01$ then $x=1.001$ : $r_*/2GM=1.01+\ln(0.01)=1.01-4.605=\mathbf{-3.60}$ , and $1.001+\ln(0.001)=1.001-6.908=\mathbf{-5.91}$ . Each factor-of-ten step inward in $(r-2GM)$ drives $r_*$ down by $2GM\ln10\approx2.30\times2GM$ — a fixed increment, never a final value.

Step 4. Interpretation: the ordinary shell between $x=1.5$ and $x=1.001$ is a thin sliver, $\Delta r=0.499\times2GM$ , but it maps to a tortoise span $\Delta r_*/2GM=0.807-(-5.91)=\mathbf{6.71}$ — more than thirteen times wider. The near-horizon shell is stretched without bound, and no finite $r_*$ ever labels the horizon itself.

2 · The frozen beacon: coordinate delay for a 10 M⊙ hole

A radio beacon hovers just outside the horizon of a $10\,M_\odot$ black hole and drifts slowly inward. By how much does the coordinate-time delay of its signals to a distant station grow each time its height above the horizon shrinks by a factor of ten? Give a number in milliseconds.

Step 1. Fix the scale. The Schwarzschild radius is $2GM/c^2=2\times1.477\ \mathrm{km}\times10=\mathbf{29.5\ km}$ . Restoring $c$ , a tortoise length $\Delta r_*$ corresponds to a coordinate-time delay $\Delta t=\Delta r_*/c$ .

Step 2. From Example 1, dropping the beacon's height $(r-2GM)$ by one decade lowers $r_*$ by the fixed increment $\Delta r_*=2GM\ln10$ . Numerically $\Delta r_*=29.5\ \mathrm{km}\times2.303=\mathbf{67.9\ km}$ .

Step 3. Convert to time: $\Delta t=\Delta r_*/c=67.9\ \mathrm{km}/(2.998\times10^5\ \mathrm{km\,s^{-1}})=2.27\times10^{-4}\ \mathrm{s}=\mathbf{0.227\ ms}$ per decade of approach.

Step 4. Interpretation: the delay grows by the same $0.23$ ms every time the beacon halves-and-halves-again its gap to the horizon — an arithmetic march with no last term, so the total delay diverges logarithmically. A distant observer therefore never sees the beacon cross; its signals simply arrive ever later and, by the companion redshift factor $\sqrt{f}$ , ever fainter — the "frozen star" is this logarithm in disguise.

Physics of the portrait

The tortoise coordinate is not the discovery of a single person but the distillation of forty years of struggle to read the Schwarzschild horizon correctly — from a metric that looked singular in 1916 to a smooth one-way membrane by 1960, with $r_*$ as the connective tissue at every step.

Origins · 1 mintortoise coordinate and how this was discovered

tortoise coordinate

the radial map defined by dr_/dr = 1/f that carries the horizon out to r_ = −∞ · Schwarzschild causal structure and black-hole perturbation theory

When Karl Schwarzschild wrote his vacuum solution in 1916, the surface $r=2GM$ seemed a genuine edge of the world, and for two decades most relativists — Einstein among them — treated it as a physical singularity. The rehabilitation came in pieces. Arthur Eddington, in 1924, quietly rewrote the metric in a coordinate that stayed finite across $r=2GM$ , without grasping what he had done. The wave-mechanical face arrived first: in 1957 Tullio Regge and John Archibald Wheeler, studying whether a black hole is stable against perturbations, introduced $r_*$ to fold the metric factors out of their master equation — and it was Wheeler, the century's great coiner of physics names (later "black hole" itself), who dubbed it "tortoise" after Zeno's paradox, since light in it forever approaches but never overtakes the horizon. A year later David Finkelstein read $v=t+r_*$ as a coordinate crossing the horizon cleanly, revealing a causal one-way membrane rather than a singularity. In 1960 Kruskal and Szekeres exponentiated $u$ and $v$ into the maximal analytic extension, and by 1970 Zerilli had completed the polar perturbation theory. Through all of it $r_*$ is the pivot from which the regular charts, the stability analysis, and quasinormal-mode theory all grow.

History

1916

— Schwarzschild publishes the static spherically symmetric vacuum solution

1924

— Eddington writes a chart that stays regular across r = 2GM

1957

— Regge and Wheeler introduce r_* and the perturbation master equation, and Wheeler names it the tortoise coordinate

1958

— Finkelstein reads v = t + r_* as an ingoing horizon-crossing coordinate

1960

— Kruskal and Szekeres exponentiate u and v into the maximal analytic extension

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §25.5 and Box 31.2 · Wald, General Relativity (1984), §6.3–6.4 · Carroll, Spacetime and Geometry (2004), §5.6–5.7 · Schutz, A First Course in General Relativity, 2nd ed., Ch. 11 · Hartle, Gravity (2003), Ch. 12 · Chandrasekhar, The Mathematical Theory of Black Holes (1983), Ch. 4. All results are in geometric units $c=1$ . Key values to keep in mind: differentiating $r_*=r+2GM\ln|r/2GM-1|$ returns $\mathrm{d}r_*/\mathrm{d}r=1/f$ ; the conformal radial metric is $f(-\mathrm{d}t^2+\mathrm{d}r_*^2)$ ; the scalar potential is $V_\ell=f[\ell(\ell+1)/r^2+2GM/r^3]$ with the axial gravitational potential using $-6GM/r^3$ ; surface gravity $\kappa=1/4GM$ and Hawking temperature $T_H=1/(8\pi GM)$ ; tortoise-map values $r_*/2GM=0.807$ at $r=3GM$ and a per-decade coordinate delay of $2GM\ln10/c=0.227$ ms for a $10\,M_\odot$ hole; and the fundamental $\ell=2$ quasinormal frequency $2GM\omega\approx0.3737-0.0890\,i$ .

Prerequisites. The Schwarzschild Metric · Schwarzschild Radius and the Geometric Mass · Geodesic Equations in Schwarzschild Spacetime. Leads to. Eddington–Finkelstein Coordinates · Kruskal–Szekeres Coordinates · Gravitational Redshift in Schwarzschild Spacetime · Photon Orbits and the Photon Sphere.

Ask the tutor: Ask me to draw the $(t,r_*)$ light-cone diagram, integrate $\mathrm{d}r_*/\mathrm{d}r=1/f$ step by step, or trace how $r_*$ grows into the Eddington–Finkelstein and Kruskal–Szekeres charts.

Why is it called the "tortoise" coordinate?

The name is John Wheeler's, from the 1957 Regge–Wheeler perturbation paper, and it points at Zeno's paradox of Achilles and the tortoise. In the tortoise coordinate an ingoing light ray obeys $t=-r_*+\text{const}$ , and since the horizon sits at $r_*=-\infty$ , the ray's $r_*$ marches downward forever without ever reaching bottom — just as Achilles, in Zeno's telling, keeps halving his lead and never quite passes the tortoise. It is a coordinate joke with real content: the "never arrives" is a genuine feature of Schwarzschild bookkeeper time, even though the photon crosses in finite proper terms.

Does the tortoise coordinate actually remove the singularity at r = 2GM?

Not by itself. The change $r\to r_*$ still leaves an overall conformal factor $f$ that vanishes at the horizon, so the chart $(t,r_*)$ never reaches $r=2GM$ at all — it maps the horizon to $r_*=-\infty$ and refuses to cover it. What $r_*$ does is expose the coordinate-singular structure so cleanly that the cure becomes obvious: form $u=t-r_*$ and $v=t+r_*$ and swap one of them for $t$ . The advanced chart $(v,r)$ is manifestly regular across the future horizon, and exponentiating both $u$ and $v$ gives Kruskal–Szekeres, regular everywhere. So the tortoise coordinate is the diagnostic and stepping stone, not the final regular chart.

Is r_* a physical distance you could measure with a ruler?

No, and conflating it with proper distance is the classic trap. Proper radial distance is $\int\mathrm{d}r/\sqrt{f}$ , which stays finite down to the horizon; the tortoise coordinate is $\int\mathrm{d}r/f$ , which diverges there because of the extra half-power of $f$ . The two agree far away, where $f\to1$ , but near the horizon they part company completely. What $r_*$ measures is better called light-travel or conformal distance: it is the variable in which radial light moves at unit coordinate speed, so equal $r_*$ -intervals are equal light-crossing intervals in coordinate time. The infinite tortoise throat is a statement about light and clocks, not about meter sticks.

How does r_* turn a curved-space wave equation into a flat one?

Through the single operator identity $\partial_{r_*}=f\,\partial_r$ , just the defining relation read backwards. In the scalar d'Alembertian on the Schwarzschild background the troublesome factors are the $-1/f$ in $g^{tt}$ and the $f$ in $g^{rr}$ . Substituting the mode ansatz $\Phi=(\psi/r)Y_{\ell m}$ and multiplying through by $f$ makes the time term a clean $\partial_t^2\psi$ , while $f\,\partial_r(f\,\partial_r)$ becomes exactly $\partial_{r_*}^2\psi$ . Everything else — the angular centrifugal term and the derivative of $f$ — collects into a single potential $V_\ell(r_*)$ . The result is a textbook one-dimensional wave equation on the infinite line, which is why ringing black holes read like an exercise in quantum scattering off a barrier.

Where does the effective potential come from, and why does it live at the photon sphere?

It is the residue of the metric's curvature once the flat wave operator has been extracted: $V_\ell=f[\ell(\ell+1)/r^2+2GM/r^3]$ for a scalar, with the second term changed for gravitational modes. The $\ell(\ell+1)/r^2$ piece is an ordinary centrifugal barrier — waves with angular momentum are held away from the center — and multiplying by $f$ makes it vanish at both the horizon and infinity, so it must peak in between. For large $\ell$ that peak sits at $r=3GM$ , the photon sphere, because the barrier's location is fixed by where null geodesics can orbit. Physically the potential is the same light ring that traps photons on unstable orbits, which is why its geometry sets both quasinormal frequencies and greybody factors.

Why does light take infinite coordinate time to reach the horizon if it crosses in finite proper time?

Because the two times answer different questions. Schwarzschild coordinate time $t$ is kept by static observers infinitely far away — the "bookkeeper" who never falls in. To that bookkeeper, signals from an infalling body arrive ever later and ever more redshifted, and in the limit they cease: the crossing is pushed to $t=+\infty$ , which the tortoise coordinate registers as $r_*\to-\infty$ . The infalling body's own wristwatch tells a different story; its proper time $\tau$ ticks off a finite interval to the horizon and continues smoothly past it. No contradiction — one clock belongs to a distant static frame the horizon renders unable to keep up, the other rides with the faller. The "freeze" is a fact about the distant frame, not the traveler.