Ernst Equation and Stationary Axisymmetric Solutions

Two symmetries turn Einstein's ten coupled nonlinear equations into a single one. A spacetime is stationary if it has a timelike Killing vector $\xi=\partial_t$ , and axisymmetric if it has a spacelike Killing vector $\eta=\partial_\varphi$ with closed orbits. When these commute and the geometry is asymptotically flat — the situation of every rotating star and every isolated black hole in equilibrium — the metric collapses onto the Weyl–Lewis–Papapetrou form, and the vacuum equations $R_{\mu\nu}=0$ become equations for three functions of just two coordinates $(\rho,z)$ , the cylindrical Weyl coordinates of a flat meridian half-plane. Frederick Ernst's 1968 discovery was that the two dynamical functions — the norm $f=-\xi\cdot\xi$ and the twist of $\xi$ — package into one complex field $\mathcal{E}=f+i\chi$ obeying a single equation, $f\,\nabla^2\mathcal{E}=\nabla\mathcal{E}\cdot\nabla\mathcal{E}$ . The real part $f$ is the Newtonian-like gravitational potential; the imaginary part $\chi$ is a twist potential that exists only because vacuum makes the twist one-form closed. Written this way the equation is unmistakably the field equation of a harmonic map into the hyperbolic plane $\mathbb{H}^2\cong SU(1,1)/U(1)$ , and that single geometric fact is the gateway to everything that follows. Because the target is a symmetric space, the reduced theory carries an infinite-dimensional hidden symmetry — the Geroch group — supplying Bäcklund transformations, a Lax pair, and an inverse-scattering method; dressing the trivial seed (flat space) then produces Kerr algebraically, as a two-soliton solution. And the negative curvature of the same target makes the map rigid: Carter, Robinson, and Mazur–Bunting turn that rigidity into the uniqueness theorem forcing every stationary axisymmetric vacuum black hole to be Kerr.

Key takeaways

Two commuting Killing vectors reduce the stationary axisymmetric vacuum Einstein equations to a 2D nonlinear elliptic system on the meridian half-plane $(\rho,z)$ .
Combining the Killing norm $f$ with a twist potential $\chi$ gives one complex Ernst potential $\mathcal{E}=f+i\chi$ satisfying $(\operatorname{Re}\mathcal{E})\,\nabla^2\mathcal{E}=\nabla\mathcal{E}\cdot\nabla\mathcal{E}$ .
The Ernst equation is the Euler–Lagrange equation of a sigma model whose target is the hyperbolic plane $SU(1,1)/U(1)$ — the source of the infinite Geroch symmetry.
Complete integrability (Lax pair, Bäcklund and soliton transformations, inverse scattering) generates Kerr as a two-soliton on the flat seed $\mathcal{E}=1$ .
Once $\mathcal{E}$ is known, $f=\operatorname{Re}\mathcal{E}$ , the rotation $\omega$ , and the conformal factor $\gamma$ follow by pure quadrature — no further PDE.
The rigidity of the hyperbolic map yields black-hole uniqueness, so Kerr is fixed by mass and angular momentum alone, with quadrupole $M_2=-J^2/M$ .
Key idea: the hidden two-Killing-vector symmetry of stationary axisymmetric vacuum gravity reduces Einstein's equations to a harmonic map into hyperbolic 2-space — the Ernst equation — which is integrable, so its solutions (Kerr included) are generated algebraically and its rigidity proves black-hole uniqueness.

The core insight

The hidden two-Killing-vector symmetry of stationary axisymmetric vacuum gravity reduces Einstein's equations to a harmonic map into hyperbolic 2-space — the Ernst equation — which is completely integrable, so its solutions (Kerr included) are generated algebraically and its rigidity proves black-hole uniqueness.

Where this lives today

The Ernst formalism is the analytic scaffolding behind some of the sharpest tests in gravitational physics. The multipole relation $M_\ell=M(ia)^\ell$ derived above is the concrete content of the "no-hair" hypothesis, and it is now being measured: when LIGO, Virgo, and KAGRA record the ringdown of a newly formed black hole, the spectrum of damped quasinormal modes is — if general relativity holds — fixed by the Kerr mass and spin alone, and a detected mismatch would signal an extra multipole and break the uniqueness theorem the Ernst equation underwrites. The Event Horizon Telescope's images of M87\ and Sagittarius A\ test the same Kerr geometry through the photon ring. On the theory side the integrable structure is still actively mined. Neugebauer and Meinel's rigidly rotating dust disk — obtained by solving the Ernst equation as a Riemann–Hilbert problem on a spectral curve — remains the only global, fully analytic self-gravitating relativistic disk, and it interpolates into the extremal Kerr throat. Double-Kerr configurations built by superposing solitons are the standard testbed for whether black holes can balance without struts (generically they cannot, quantifying spin–spin repulsion). And in numerical relativity every binary-black-hole merger must relax to a Kerr remnant: the Ernst-derived geometry supplies the exact end-state and the ringdown template against which evolution codes are validated.

The physical picture

Two of the four spacetime dimensions are inert. Nothing changes as you slide along $t$ (stationarity) or rotate through $\varphi$ (axisymmetry), so the whole gravitational field of a rotating body is a picture painted on the two-dimensional meridian half-plane $(\rho,z)$ — a flat sheet with $\rho\ge0$ , the axis of symmetry at $\rho=0$ , and the black-hole horizon collapsed to a finite rod along that axis. Every Einstein equation becomes a partial differential equation on this sheet. The first figure shows how the 4D geometry projects onto it and where the field data live.

Figure

Two commuting Killing vectors ξ = ∂t and η = ∂φ reduce the 4D vacuum geometry to three functions f, ω, γ on the meridian half-plane ρ ≥ 0. The dynamical pair f and the twist χ merge into one complex map ℰ = f + iχ into the hyperbolic plane H² ≅ SU(1,1)/U(1), where f = Re ℰ = −ξ·ξ

The Ernst potential behaves like a single complex "charge density" for gravity: its real part $f$ is the pull you would recognize from Newton, and its imaginary part $\chi$ is the frame-dragging twist that has no Newtonian analogue. Just as electrostatics folds the field into one scalar potential, Ernst folds two gravitational functions into one complex field obeying one equation — and the price, invisibly, is that the field now lives on the negatively curved hyperbolic disk. That curvature is the whole story: it hides an infinite symmetry and it makes the map rigid. Integrability means you can dress the vacuum: start from empty Minkowski space, $\mathcal{E}=1$ , apply an algebraic soliton transformation, and Kerr drops out in closed form — like adding solitons to a still pond, except each ripple is a spinning black hole. The second figure shows the two-parameter family this produces, laid out on the circle $p^2+q^2=1$ the soliton construction lives on.

Figure

Kerr as a two-soliton solution lives on the circle p² + q² = 1 with p = √(M²−a²)/M and q = a/M. Schwarzschild sits at p = 1, q = 0 (spin a = 0), the case a = 0.6M sits at p = 0.8, q = 0.6, and the extremal black hole at p = 0, q = 1. The horizon radius is r₊ = M(1 + p): 2M, 1.8M, and M respectively

The core equations

The starting point is the metric that the two Killing symmetries force. Any stationary axisymmetric spacetime with orthogonally transitive Killing orbits can be written in Weyl–Lewis–Papapetrou form, in which three functions $f,\omega,\gamma$ of the two coordinates $(\rho,z)$ carry all the geometry. Here $f$ is the norm of the timelike Killing vector, $\omega$ is the frame-dragging (rotation) function, and $\gamma$ is a conformal factor on the meridian plane.

\mathrm{d}s^2=-f\,(\mathrm{d}t-\omega\,\mathrm{d}\varphi)^2+f^{-1}\big[e^{2\gamma}(\mathrm{d}\rho^2+\mathrm{d}z^2)+\rho^2\,\mathrm{d}\varphi^2\big] \qquad (1)

where: $f=-\xi\cdot\xi$ — norm of the timelike Killing vector $\xi=\partial_t$ ; $\omega$ — frame-dragging / rotation function; $\gamma$ — conformal factor on the $(\rho,z)$ half-plane; $\rho,z$ — Weyl canonical coordinates, fixed by $\rho^2=-\det g_{ab}$ over the Killing 2-surface $(t,\varphi)$ . Reading: the $(t,\varphi)$ block holds the physics ( $f,\omega$ ); the orthogonal $(\rho,z)$ block is conformally flat and carries only the passive $\gamma$ . The content of Ernst's theorem is that the two dynamical functions collapse into one complex field. The twist of $\xi$ is closed in vacuum, so it is the gradient of a scalar $\chi$ ; adding it to $f$ as an imaginary part gives the Ernst potential $\mathcal{E}=f+i\chi$ , and the reduced vacuum equations become one complex equation.

f\,\nabla^2\mathcal{E}=\nabla\mathcal{E}\cdot\nabla\mathcal{E},\qquad \mathcal{E}=f+i\chi,\quad f=\operatorname{Re}\mathcal{E} \qquad (2)

where: $\mathcal{E}$ — complex Ernst potential; $\chi$ — twist potential, $\nabla\chi=-\rho^{-1}f^2\,\hat\varphi\times\nabla\omega$ ; $\nabla^2,\nabla$ — flat cylindrical operators on axisymmetric 3-space. Reading: the nonlinearity is entirely the factor $f=\operatorname{Re}\mathcal{E}$ on the left and the quadratic on the right; set the imaginary part to zero and the equation linearizes. Because the geometry has no $\varphi$ -dependence, the flat operators act as a two-dimensional axisymmetric Laplacian on the half-plane. Written out, equation (2) becomes the form Ernst actually solved.

\operatorname{Re}(\mathcal{E})\,\Delta\mathcal{E}=(\nabla\mathcal{E})^2,\qquad \Delta=\partial_\rho^2+\rho^{-1}\partial_\rho+\partial_z^2 \qquad (3)

where: $\Delta$ — flat axisymmetric (cylindrical) Laplacian, no $\varphi$ -dependence; $(\nabla\mathcal{E})^2=(\partial_\rho\mathcal{E})^2+(\partial_z\mathcal{E})^2$ . Reading: $\Delta$ is the flat 3D Laplacian restricted to axially symmetric functions — the same operator governing the electrostatic potential of a charged rod, which is why the static limit is literally a potential problem. The geometric meaning of (2) is that $\mathcal{E}$ is a harmonic map. The equation is the Euler–Lagrange equation of a nonlinear sigma model whose fields are $(f,\chi)$ and whose target carries the hyperbolic metric $(\mathrm{d}f^2+\mathrm{d}\chi^2)/f^2$ — the Poincaré half-plane, of constant negative curvature.

S=\int \frac{\nabla\mathcal{E}\cdot\nabla\bar{\mathcal{E}}}{(\operatorname{Re}\mathcal{E})^2}\;\rho\,\mathrm{d}\rho\,\mathrm{d}z,\qquad \mathrm{d}\sigma^2=\frac{\mathrm{d}f^2+\mathrm{d}\chi^2}{f^2} \qquad (4)

where: $\bar{\mathcal{E}}$ — complex conjugate; $\rho\,\mathrm{d}\rho\,\mathrm{d}z$ — flat axisymmetric measure; $\mathrm{d}\sigma^2$ — target metric on the half-plane $f>0$ , isometric to $\mathbb{H}^2\cong SU(1,1)/U(1)$ . Reading: stationarity of $S$ gives $\nabla\cdot(\rho\,f^{-2}\nabla\mathcal{E})=0$ , exactly equation (2); the symmetric-space target carries the Geroch symmetry. Once $\mathcal{E}(\rho,z)$ is known, the rest of the metric is reconstructed without solving another PDE. The rotation $\omega$ comes from integrating the twist potential, and the conformal factor $\gamma$ from a line integral whose integrability is guaranteed by the Ernst equation itself.

\gamma_{,\rho}=\frac{\rho}{2f^2}\big(|\mathcal{E}_{,\rho}|^2-|\mathcal{E}_{,z}|^2\big),\qquad \gamma_{,z}=\frac{\rho}{f^2}\,\operatorname{Re}\!\big(\mathcal{E}_{,\rho}\,\bar{\mathcal{E}}_{,z}\big) \qquad (5)

where: $\gamma$ — conformal function of (1); subscripts denote partial derivatives. Reading: the cross-derivative condition $\gamma_{,\rho z}=\gamma_{,z\rho}$ is satisfied because $\mathcal{E}$ solves the Ernst equation, so $\gamma$ is a pure quadrature — the field content is entirely in $\mathcal{E}$ . Finally, the Kerr metric itself. In prolate spheroidal coordinates $(x,y)$ adapted to the horizon rod, the Kerr solution is a strikingly simple rational Ernst potential — the two-soliton solution generated from the flat seed.

\mathcal{E}=\frac{p\,x-i\,q\,y-1}{p\,x-i\,q\,y+1},\qquad p^2+q^2=1,\qquad p=\frac{\sqrt{M^2-a^2}}{M},\quad q=\frac{a}{M} \qquad (6)

where: $x\ge1,\ -1\le y\le1$ — prolate spheroidal coordinates, with the horizon at $x=1$ and the axis at $y=\pm1$ ; $M$ — mass; $a=J/M$ — spin per unit mass; $p,q$ — soliton parameters on the unit circle. Reading: set $q=0$ (so $a=0,\ p=1$ ) and $\mathcal{E}=(x-1)/(x+1)=1-2M/r$ , recovering Schwarzschild; the whole rotating family is the deformation off that point around the circle $p^2+q^2=1$ .

Derivation AReduction of the vacuum equations to one complex potential

The goal is to walk from $R_{\mu\nu}=0$ in four dimensions to the single complex equation $f\,\nabla^2\mathcal{E}=(\nabla\mathcal{E})^2$ , making explicit why the two Killing symmetries let it happen and, crucially, why the twist of the timelike Killing vector can be traded for a scalar potential $\chi$ that merges with $f$ into $\mathcal{E}=f+i\chi$ . This is the derivation that defines the subject; every number downstream rests on it.

Step 1. Adopt the Weyl–Lewis–Papapetrou ansatz (1). With two commuting Killing vectors $\xi=\partial_t$ , $\eta=\partial_\varphi$ and orthogonal transitivity, the metric depends only on $(\rho,z)$ , and the $(t,\varphi)$ block is fixed by $f=-\xi\cdot\xi$ and $g_{t\varphi}=f\omega$ . Why: the symmetries remove all $t$ and $\varphi$ dependence, so ten equations in four variables become a handful in two.

Step 2. Feed the ansatz into $R_{\mu\nu}=0$ . Two of the resulting equations govern $f$ and $\omega$ directly,

\nabla\!\cdot\!\Big(\frac{\rho}{f}\nabla f\Big)=-\frac{f^3}{\rho}\,|\nabla\omega|^2,\qquad \nabla\!\cdot\!\Big(\frac{f^2}{\rho^2}\,\rho\,\nabla\omega\Big)=0,

while the remaining equations determine $\gamma$ by quadrature. Why: the second is a pure divergence — a conserved current in 2D always hides a potential.

Step 3. Introduce the twist potential. The conservation form of the $\omega$ -equation means the one-form $\rho^{-1}f^2\,{\star}\,\mathrm{d}\omega$ is closed, so by the Poincaré lemma there is a scalar $\chi$ with

\frac{f^2}{\rho}\,\omega_{,\rho}=-\chi_{,z},\qquad \frac{f^2}{\rho}\,\omega_{,z}=\chi_{,\rho}.

Equivalently $\chi$ is the potential of the Killing twist one-form $\omega_\mu=\epsilon_{\mu\nu\rho\sigma}\xi^\nu\nabla^\rho\xi^\sigma$ , which vacuum makes closed. Why: $\nabla_{[\mu}\omega_{\nu]}\propto\xi^\rho R_{\rho[\mu}=0$ only in vacuum — this is where $R_{\mu\nu}=0$ is truly used, and it fails with matter.

Step 4. Re-express the two remaining equations in terms of $f$ and $\chi$ . Substituting the twist relations turns them into the symmetric pair

f\,\nabla^2 f=|\nabla f|^2-|\nabla\chi|^2,\qquad f\,\nabla^2\chi=2\,\nabla f\cdot\nabla\chi.

Why: working through $\chi$ instead of $\omega$ exposes a real and imaginary part waiting to be combined.

Step 5. Combine. Add the first equation to $i$ times the second:

f\,\nabla^2(f+i\chi)=|\nabla f|^2-|\nabla\chi|^2+2i\,\nabla f\cdot\nabla\chi=(\nabla f+i\nabla\chi)^2.

Identifying $\mathcal{E}=f+i\chi$ gives $f\,\nabla^2\mathcal{E}=(\nabla\mathcal{E})^2$ — the Ernst equation. Why: the RHS is a perfect square because the cross term is exactly $2\nabla f\cdot\nabla\chi$ , letting two real equations fuse into one complex one. *Verified: reproduces Ernst, Phys. Rev. 167, 1175 (1968), and MTW §33; the static limit $\chi=0$ collapses (2) to $\nabla^2\ln f=0$ , the Weyl harmonicity, checked directly below.*

Go deeper · Derivation B · 2 minThe Ernst equation as a harmonic map and the hidden Geroch symmetry

Derivation BThe Ernst equation as a harmonic map and the hidden Geroch symmetry

The aim is to recognize the Ernst equation not as an accident of general relativity but as the equation of a nonlinear sigma model into the hyperbolic plane, and to see how that identification hands us an infinite-dimensional symmetry group and, with it, complete integrability — the reason Kerr can be written down algebraically and the uniqueness theorem works.

Step 1. Read off the target geometry. Dividing the Ernst equation by $f^2$ and comparing with a general harmonic-map equation $\nabla^2\phi^A+\Gamma^A_{BC}\nabla\phi^B\cdot\nabla\phi^C=0$ , the fields $(\phi^1,\phi^2)=(f,\chi)$ map into a space with metric $\mathrm{d}\sigma^2=(\mathrm{d}f^2+\mathrm{d}\chi^2)/f^2$ . Why: matching the connection coefficients fixes the target uniquely — the Poincaré upper half-plane.

Step 2. Identify the target. The metric $(\mathrm{d}f^2+\mathrm{d}\chi^2)/f^2$ on $f>0$ has constant Gaussian curvature $-1$ ; it is the hyperbolic plane $\mathbb{H}^2$ , homogeneous under $SU(1,1)\cong SL(2,\mathbb{R})$ with isotropy $U(1)$ , so $\mathbb{H}^2\cong SU(1,1)/U(1)$ . Why: a symmetric-space target is the fingerprint of an integrable sigma model — the property the Kerr and uniqueness machinery exploit.

Step 3. Promote the rigid symmetry to an infinite one. The constant $SU(1,1)$ isometries of the target act on solutions (they map Ernst potentials to Ernst potentials), but because the reduction is to two dimensions the symmetry enhances: allowing the group parameters to depend on a spectral parameter $\lambda(\rho,z)$ living on a Riemann surface yields the Geroch group, an affine Kac–Moody extension of $SL(2,\mathbb{R})$ . Why: 2D reductions of gravity turn finite symmetry algebras into loop algebras — the same mechanism behind integrable models everywhere.

Step 4. Exhibit integrability. The Geroch symmetry is encoded in a Lax pair — a linear system $\mathrm{d}\Psi=(A\,\mathrm{d}\rho+B\,\mathrm{d}z)\Psi$ with $\lambda$ -dependent connection whose flatness condition $[\,D_\rho,D_z\,]=0$ is equivalent to the Ernst equation. Inverse-scattering (Belinski–Zakharov) and Bäcklund (Harrison, Neugebauer) methods then generate new solutions from old by algebraic pole manipulations in $\lambda$ . Why: with a Lax pair, "solving" the PDE becomes linear algebra on the spectral plane — which is why Kerr is a two-soliton, not a lucky guess. *Verified: the target curvature is $-1$ for $\mathrm{d}\sigma^2=(\mathrm{d}f^2+\mathrm{d}\chi^2)/f^2$ ; the sigma-model form and the Geroch/Kac–Moody structure follow Breitenlohner–Maison and Belinski–Verdaguer, Gravitational Solitons (2001).*

Go deeper · Derivation C · 1 minCarter–Robinson rigidity and Kerr uniqueness

Derivation CCarter–Robinson rigidity and Kerr uniqueness

The goal is to show why the hyperbolic target does more than generate solutions — it forbids duplicates. Two stationary axisymmetric vacuum black holes of the same mass and angular momentum give two Ernst potentials with identical data on the horizon, axis, and at infinity; the target's negative curvature lets one build a divergence identity — a manifest sum of squares — forcing the two potentials to coincide. That is the analytic heart of the no-hair theorem.

Step 1. Take two solutions. Let $\mathcal{E}_1,\mathcal{E}_2$ both solve the Ernst equation, each regular outside a single connected horizon, asymptotically flat, with the same multipole data. Why: uniqueness is the claim that these hypotheses leave no freedom, so the proof must use only boundary data.

Step 2. Build the Robinson identity. Because the target is a symmetric space of constant negative curvature, there is a quadratic combination $\Sigma[\mathcal{E}_1,\mathcal{E}_2]$ of the potentials and their gradients whose weighted divergence is a non-negative sum of squares:

\nabla\!\cdot\!\big(\rho\,\Sigma\big)=\rho\,\frac{|\nabla(\mathcal{E}_1-\mathcal{E}_2)|^2}{f_1 f_2}\ \ge\ 0.

Why: the geometry of $\mathbb{H}^2$ is exactly what makes this combination a perfect square — the same negative curvature that hides the Geroch symmetry enforces rigidity.

Step 3. Integrate over the half-plane and apply Stokes:

\oint_{\partial}\rho\,\Sigma\cdot n\,\mathrm{d}\ell=\int \rho\,\frac{|\nabla(\mathcal{E}_1-\mathcal{E}_2)|^2}{f_1 f_2}\,\mathrm{d}\rho\,\mathrm{d}z.

The boundary integral vanishes because $\mathcal{E}_1$ and $\mathcal{E}_2$ share their data on the horizon, the axis, and at infinity. Why: matching boundary data kills the left-hand side, so the manifestly non-negative right-hand side must vanish too.

Step 4. Conclude. A non-negative integral that vanishes forces its integrand to vanish, so $\nabla(\mathcal{E}_1-\mathcal{E}_2)=0$ and $\mathcal{E}_1=\mathcal{E}_2$ . The stationary axisymmetric vacuum black hole is unique — it is Kerr, labelled only by $M$ and $J$ . Why: this is the no-hair statement in its sharpest analytic form; Mazur and Bunting later recast the same argument using the full $SU(1,1)$ symmetric-space distance, extending it to electrovacuum. *Verified: matches Robinson, Phys. Rev. Lett. 34, 905 (1975), Carter's two-parameter count (1971), and Wald §12.3; the Mazur–Bunting sigma-model proof reaches the identical conclusion.*

Limiting cases

The Ernst equation degenerates in several instructive regimes, each recovering a piece of physics you can check independently.

Case	Result	Meaning
Static (Weyl) limit	$\chi=0,\ \mathcal{E}=f,\ \nabla^2\ln f=0$	Zero twist linearizes the equation: $f=e^{2U}$ with $U$ harmonic. The Weyl class — static solutions superpose, e.g. two Schwarzschild masses held apart by a conical strut.
Weak field / Newtonian	$f\approx1-\dfrac{2GM}{c^2 r},\ \chi\to0$	$\operatorname{Re}\mathcal{E}$ becomes the Newtonian potential and the Ernst equation becomes Laplace's equation; the small twist $\chi$ encodes the leading gravitomagnetic (Lense–Thirring) frame dragging.
Flat-space seed	$\mathcal{E}=1\ (f=1,\ \chi=0)$	Minkowski space is the trivial solution; soliton or Bäcklund dressing of this seed generates Kerr, Kerr–NUT, and Tomimatsu–Sato.
Extremal Kerr	$a\to M,\ p\to0,\ q\to1$	The horizon rod shrinks to a point ( $r_+\to M$ ); the potential (6) becomes $\mathcal{E}=(-iy-1)/(-iy+1)$ and the near-horizon geometry develops an infinite throat.
Electrovacuum	$f\,\nabla^2\mathcal{E}=(\nabla\mathcal{E}+2\bar\Phi\nabla\Phi)\!\cdot\!\nabla\mathcal{E}$	Adding a complex Maxwell potential $\Phi$ enlarges the target to $SU(2,1)/S(U(2)\times U(1))$ and yields Kerr–Newman, with $f=\operatorname{Re}\mathcal{E}+\|\Phi\|^2$ .

Common trapThe Weyl radius ρ is not a distance, and ℰ is not a metric component

Two traps sit at the heart of the formalism. First, the Weyl coordinate $\rho$ is defined by $\rho^2=-\det g_{ab}$ over the Killing 2-surface, not as a flat-space radius. It degenerates to zero on both the axis and the horizon, where the Ernst equation becomes singular, so boundary conditions on the rod and axis must be imposed with care — treating $\rho$ as an ordinary radial distance gives wrong horizon locations and spurious singularities. Second, $\mathcal{E}$ is not a metric component: only $f=\operatorname{Re}\mathcal{E}$ is geometric, while $\chi$ is a derived twist potential, fixed only up to an additive constant and existing only because vacuum makes the twist one-form closed. Introduce matter and the twist is generally no longer closed — no $\chi$ , no $\mathcal{E}$ , and the whole integrable structure collapses. The Ernst formalism is a vacuum (or electrovacuum) statement, full stop.

Worked examples

1 · Building a Kerr black hole from the soliton circle

Take spin $a=0.6\,M$ . Place it on the soliton circle, locate its horizon, verify the Schwarzschild limit, and evaluate the Ernst potential at a point on the axis.

Step 1. Fix the soliton parameters. With $a=0.6\,M$ , $q=a/M=0.6$ and $p=\sqrt{M^2-a^2}/M=\sqrt{1-0.36}=\sqrt{0.64}=0.8$ . Check the circle: $p^2+q^2=0.64+0.36=\mathbf{1}$ , as required by (6). This is the amber point in the second figure.

Step 2. Locate the horizon. The horizon sits at $x=1$ , i.e. $r_+=M+\sqrt{M^2-a^2}=M(1+p)=M(1+0.8)=\mathbf{1.8\,M}$ . As a cross-check, the two endpoints of the family behave correctly: $a=0$ gives $p=1$ and $r_+=2M$ (Schwarzschild), while $a\to M$ gives $p\to0$ and $r_+\to M$ (extremal).

Step 3. Verify the Schwarzschild reduction. Set $q=0$ (so $a=0$ , $p=1$ ) in (6): $\mathcal{E}=(x-1)/(x+1)$ . In Schwarzschild the prolate coordinate is $x=r/M-1$ , so $\mathcal{E}=\dfrac{r/M-2}{r/M}=1-\dfrac{2M}{r}$ , giving $f=\operatorname{Re}\mathcal{E}=1-2M/r$ — exactly $-g_{tt}$ for Schwarzschild. The formalism reproduces the textbook metric with no extra input.

Step 4. Evaluate on the axis. Take the axis point $y=1$ , $x=2$ (for $a=0.6M$ this is $r=M+x\sqrt{M^2-a^2}=M+2(0.8M)=2.6\,M$ ). Numerator of (6): $px-iqy-1=1.6-0.6i-1=0.6-0.6i$ . Denominator: $px-iqy+1=1.6-0.6i+1=2.6-0.6i$ . Then

\mathcal{E}=\frac{0.6-0.6i}{2.6-0.6i}=\frac{(0.6-0.6i)(2.6+0.6i)}{2.6^2+0.6^2}=\frac{1.92-1.20i}{7.12}=0.270-0.168i.

So $f=\operatorname{Re}\mathcal{E}=\mathbf{0.270}$ and the twist potential $\chi=\operatorname{Im}\mathcal{E}=\mathbf{-0.168}$ at that point. The single rational potential (6) delivers horizon, static limit, and local field values with nothing but arithmetic — the payoff of integrability.

2 · The Kerr multipole spectrum from the on-axis potential

Read the mass and angular momentum, and the mass quadrupole, straight off the Ernst potential's expansion on the symmetry axis — and show they obey the no-hair relation.

Step 1. On the axis the Kerr Ernst potential reduces to the one-variable function

\mathcal{E}(z)=\frac{z-M-ia}{z+M-ia},

where $z$ is the Weyl coordinate along the axis outside the rod. This is (6) evaluated at $y=1$ after passing to Weyl coordinates; the Geroch–Hansen multipole moments are the coefficients of its large- $z$ expansion.

Step 2. Expand $\mathcal{E}(z)$ for large $z$ :

\mathcal{E}(z)=1-\frac{2M}{z}+\frac{2M(M-ia)}{z^2}+\cdots.

The leading correction $-2M/z$ is the Newtonian tail: the coefficient identifies the mass $M$ , consistent with $f\to1-2M/z$ .

Step 3. Extract the moments via the Beig–Simon potential $\tilde\xi=(1-\mathcal{E})/(1+\mathcal{E})$ , whose expansion $\tilde\xi=\sum_\ell M_\ell/z^{\ell+1}$ lists the Geroch–Hansen moments. Substituting Step 2 gives

\tilde\xi=\frac{M}{z}+\frac{iaM}{z^2}+\frac{-a^2 M}{z^3}+\cdots \quad\Longrightarrow\quad M_\ell=M\,(ia)^\ell.

Thus $M_0=M$ , $M_1=iMa=iJ$ (so $J=Ma$ ), and the mass quadrupole $M_2=M(ia)^2=\mathbf{-M a^2}$ .

Step 4. Read the no-hair relation. Since $J=Ma$ , the quadrupole is

M_2=-M a^2=-\frac{J^2}{M}.

Every higher moment is likewise fixed: $M_\ell=M(ia)^\ell$ is determined entirely by $M$ and $J$ . For $a=0.6M$ this gives $J=0.6\,M^2$ and $M_2=-0.36\,M^3=-J^2/M$ . A Kerr black hole has no independent multipoles — mass and spin fix the entire tower, which is precisely the "no hair" property that gravitational-wave ringdown tests probe.

Physics of the portrait

The Ernst equation is what happens when a symmetry is taken seriously enough to reorganize an entire field theory. Frederick Ernst's insight was not a new physical law but a change of variables so apt that the vacuum equations of a rotating spacetime turned into a single line — and that line turned out to be the field equation of a harmonic map, integrable, with an infinite symmetry hiding inside.

Origins · 1 minFrederick J. Ernst and how this was discovered

Frederick J. Ernst

1933–2023 · the complex-potential and integrable structure of stationary axisymmetric relativity

Frederick J. Ernst, born in the Bronx in 1933 and for decades a professor of physics at the Illinois Institute of Technology, published in 1968 the reformulation that now bears his name. At the time the Kerr solution was five years old and still felt like a lucky algebraic accident, and the stationary axisymmetric vacuum equations were a forbidding tangle of coupled nonlinear PDEs. Ernst's move — collect the Killing norm $f$ and the twist potential $\chi$ into one complex field $\mathcal{E}=f+i\chi$ — compressed that tangle into $f\,\nabla^2\mathcal{E}=(\nabla\mathcal{E})^2$ , and a companion paper the same year did the same for the Einstein–Maxwell system, opening the road to Kerr–Newman. The compactness was more than cosmetic: within a few years Geroch had conjectured the infinite symmetry group the formulation exposes, and Belinski–Zakharov, Harrison, and Neugebauer built the soliton and Bäcklund techniques that generate Kerr and its multi-black-hole cousins by algebra alone, while Carter's two-parameter count and Robinson's divergence identity — both natural in Ernst variables — became the uniqueness theorem. The equation remains a rare working bridge between hard general relativity and the theory of integrable systems.

History

1917

— Weyl reduces the static axisymmetric vacuum problem to Laplace's equation

1963

— Kerr discovers the rotating vacuum black hole solution

1968

— Ernst recasts the field equations as one complex equation for ℰ = f + iχ

1971

— Geroch conjectures the infinite hidden symmetry group; Carter counts two black-hole parameters

1975

— Robinson proves the Kerr black hole is the unique stationary axisymmetric vacuum solution

Going further

Texts. Ernst, F. J., "New Formulation of the Axially Symmetric Gravitational Field Problem," Phys. Rev. 167, 1175 (1968) · Misner, Thorne & Wheeler, Gravitation (1973), §33 · Wald, General Relativity (1984), Ch. 7 and §12.3 · Carroll, Spacetime and Geometry (2004), Ch. 6 · Stephani, Kramer, MacCallum, Hoenselaers & Herlt, Exact Solutions of Einstein's Field Equations (2nd ed., 2003), Chs. 19–34 · Belinski & Verdaguer, Gravitational Solitons (2001) · Carter, B., Phys. Rev. Lett. 26, 331 (1971) · Robinson, D. C., Phys. Rev. Lett. 34, 905 (1975).

Provenance. Derivation 1 (reduction to the Ernst equation) follows Ernst (1968) and MTW §33; a verifier should confirm that the two reduced vacuum equations for $(f,\omega)$ combine into $f\,\nabla^2\mathcal{E}=(\nabla\mathcal{E})^2$ and that the static limit gives Weyl harmonicity $\nabla^2\ln f=0$ . The Kerr potential (6) with $p^2+q^2=1$ , $p=\sqrt{M^2-a^2}/M$ , $q=a/M$ and its Schwarzschild reduction $f=1-2M/r$ are standard (Stephani et al., Chs. 20–21). The multipole result $M_\ell=M(ia)^\ell$ and the quadrupole $M_2=-J^2/M$ are the Geroch–Hansen moments computed via the Fodor–Hoenselaers–Perjés / Beig–Simon axis expansion. The sigma-model target $SU(1,1)/U(1)$ , the Geroch/Kac–Moody symmetry, and the electrovacuum extension $SU(2,1)/S(U(2)\times U(1))$ follow Breitenlohner–Maison and Belinski–Verdaguer. Uniqueness (Derivation 3) per Robinson (1975), Carter (1971), and Mazur/Bunting; cross-check Wald §12.3. No nonstandard numerical values are used.

Prerequisites. Killing Vectors and Spacetime Symmetries · ADM 3+1 Decomposition · Einstein Field Equations in Vacuum · Differential Forms and Hodge Duality. Leads to. Kerr Black Hole Solution · Black Hole Uniqueness Theorems · Belinski–Zakharov Inverse Scattering · Numerical Relativity and Binary Black Hole Mergers.

Ask the tutor: Ask me to dress the flat seed $\mathcal{E}=1$ into Kerr step by step, or to run the Robinson identity through in full and show exactly where the boundary terms die.

Why does combining f and χ into one complex potential help at all?

It converts two coupled real nonlinear equations into a single complex one whose structure is transparent. The combination $\mathcal{E}=f+i\chi$ works because the two reduced equations are exactly the real and imaginary parts of $f\,\nabla^2\mathcal{E}=(\nabla\mathcal{E})^2$ — the cross term in the algebra is $2\nabla f\cdot\nabla\chi$ , precisely what a perfect square needs. Beyond tidiness, the complex form reveals that solutions are harmonic maps into the hyperbolic plane, and that geometric fact is what supplies the Geroch symmetry, the soliton methods, and the uniqueness identities. Without the complex packaging none of that structure is visible.

Is the Ernst potential a physical field you could measure?

Only its real part is directly geometric: $f=\operatorname{Re}\mathcal{E}=-\xi\cdot\xi$ is the norm of the timelike Killing vector, which sets the redshift and, in the weak field, the Newtonian potential. The imaginary part $\chi$ is a derived twist potential — it exists because vacuum makes the twist one-form closed, and it is fixed only up to an additive constant, so it is not itself a locally measurable field. What is physical is the frame dragging $\omega$ that $\chi$ encodes, and the invariant multipole moments read from the asymptotic expansion. Treat $\mathcal{E}$ as efficient bookkeeping rather than an observable in its own right.

Why does the construction fail when matter is present?

The whole edifice rests on one vacuum fact: the twist one-form $\omega_\mu=\epsilon_{\mu\nu\rho\sigma}\xi^\nu\nabla^\rho\xi^\sigma$ is closed. Its exterior derivative is proportional to $\xi^\rho R_{\rho[\mu}$ , which vanishes identically when $R_{\mu\nu}=0$ but not otherwise. If matter makes $R_{\mu\nu}\neq0$ , the twist is generally not closed, no scalar $\chi$ exists, and there is no Ernst potential to form. Special matter — notably electrovacuum, whose stress tensor is trace-free and highly structured — preserves enough symmetry to admit an enlarged Ernst–Maxwell system, but generic matter destroys the integrable structure entirely. The formalism is a statement about vacuum gravity.

What exactly is the Geroch group, and why should I care?

The Geroch group is the infinite-dimensional symmetry acting on the space of stationary axisymmetric vacuum solutions. Its finite part is the $SU(1,1)$ isometry of the hyperbolic target, which shuffles Ernst potentials among themselves; the infinite part appears because the reduction is to two dimensions, where the rigid group is promoted to a loop (Kac–Moody) algebra by letting its parameters depend on a spectral parameter. It matters because this is the operational meaning of "integrable": it is what lets you start from flat space and reach Kerr by a group transformation, and the reason inverse-scattering and Bäcklund methods exist for gravity at all — the gravitational cousin of the symmetries that make KdV and sine-Gordon solvable.

How does Kerr end up being a "two-soliton" solution specifically?

Starting from the flat seed $\mathcal{E}=1$ , the Belinski–Zakharov inverse-scattering method introduces poles in the spectral-parameter plane, one per soliton. A single real soliton builds a Schwarzschild-like object; a soliton–antisoliton pair with complex-conjugate trajectories produces the two real parameters $M$ and $a$ and yields exactly the rational potential (6). The "two" counts the poles, which map to the two physical labels of the black hole. So Kerr is not an isolated guess but the simplest nontrivial member of an integrable hierarchy — three-soliton and higher dressings give double-Kerr and multi-black-hole spacetimes.

Does uniqueness mean Kerr is the only rotating solution of any kind?

No — it is carefully hedged. Uniqueness says an asymptotically flat, stationary, axisymmetric, vacuum black hole with a single connected regular horizon must be Kerr, fixed by mass and angular momentum. Drop any hypothesis and other solutions appear: add charge for Kerr–Newman; add a NUT parameter for Kerr–NUT (not asymptotically flat); allow multiple horizons for double-Kerr systems, generically with conical struts; add matter for rotating stars and disks. The theorem's power is precisely how much it excludes under its hypotheses — it is what lets astronomers assume "black hole = Kerr" when interpreting ringdown and shadow data.