Two-Horizon Surface Gravities and Nonequilibrium

A stationary de Sitter black hole generally has two simple outer boundaries with different acceleration scales: the black-hole horizon and the cosmological horizon. One Killing field defines signed surface gravities on both, so their ratio is meaningful even though the overall normalization is not canonical. In uncharged Schwarzschild–de Sitter, equal magnitudes force the horizons to coalesce. Charge can instead produce a distinct lukewarm branch, while rotation adds angular-potential compatibility conditions.

Helpful background. Exact Solutions of Einstein’s Equations supplies static and charged de Sitter solutions; Causality and Global Structure supplies the observer patch; The Einstein Equation supplies the $\Lambda\gt0$ field equations; and Black-Hole Geometry and Mechanics supplies Killing-horizon normalization.

One generator and two signed surface gravities

For Schwarzschild–de Sitter, write

$$f(r)=1-\frac{2m}{r}-\frac{r^2}{\ell^2}, \qquad \ell^2=\frac{3}{\Lambda}.$$

The same static-patch Killing field $T=\partial_t$ generates both simple horizons. Its signed surface gravities are

$$\kappa_b=\frac12\left(\frac1{r_b}-\frac{3r_b}{\ell^2}\right)\gt0,$$ $$\kappa_c=\frac12\left(\frac1{r_c}-\frac{3r_c}{\ell^2}\right)\lt0.$$

The signs reflect opposite boundary orientations of the interval $r_b\lt r\lt r_c$. The local positive scales are $\kappa_b$ and $|\kappa_c|$. If $T$ is multiplied by a positive constant $C$, both signed values are multiplied by $C$. Therefore no common normalization can turn unequal magnitudes into equal ones.

One sometimes normalizes $T$ at the radius $r_g$ where a static geodesic has zero acceleration. Since $f'(r_g)=0$,

$$r_g^3=m\ell^2, \qquad T_{\mathrm{unit}}=\frac{T}{\sqrt{f(r_g)}}.$$

This prescription gives unit norm at one preferred orbit inside the patch, but it rescales both horizons identically. It changes neither the ratio $\kappa_b/|\kappa_c|$ nor the conclusion about equality. The normalization dependence follows from the Killing-horizon definition in Bardeen, Carter, and Hawking 1973, pp. 164–166.

Equal magnitudes in the uncharged family

Assume first that the roots are distinct. Equality of the two positive scales requires

$$\frac1{r_b}-\frac{3r_b}{\ell^2} =\frac{3r_c}{\ell^2}-\frac1{r_c},$$

and hence

$$r_br_c=\frac{\ell^2}{3}.$$

Both roots must also yield the same mass parameter:

$$m=\frac{r_b}{2}\left(1-\frac{r_b^2}{\ell^2}\right) =\frac{r_c}{2}\left(1-\frac{r_c^2}{\ell^2}\right).$$

After factoring the difference $r_c-r_b$, the distinct-root assumption gives

$$r_b^2+r_br_c+r_c^2=\ell^2.$$

Combining the last two relations yields

$$(r_c-r_b)^2=0,$$

contradicting distinctness. Thus uncharged Schwarzschild–de Sitter has no nondegenerate equal-magnitude pair. Equality occurs only at the Nariai point $r_b=r_c=\ell/\sqrt3$, where both $\kappa$ values vanish relative to the original $T$. A rescaled near-horizon time gives a finite Nariai acceleration scale, but that is a limiting generator rather than evidence for a generic two-horizon equilibrium.

Euclidean regularity and the geometric obstruction

Near a simple horizon, analytic continuation $t=-i\tau$ gives a polar metric whose absence of a conical defect requires period

$$\beta_h=\frac{2\pi}{|\kappa_h|}.$$

One Euclidean time cannot have two different periods. Consequently, generic Schwarzschild–de Sitter data leave a conical defect at one horizon if the other is made smooth. This is a sharp geometric obstruction to a single regular Euclidean section.

The obstruction should not be overstated. The Lorentzian metric remains an exact stationary solution when the magnitudes differ. Nor does the classical calculation by itself specify a quantum state or prove a direction of energy flow. Those questions depend on state choice, field content, and backreaction and belong to QFT in Curved Spacetime. Here “nonequilibrium” means that the two horizon scales do not define one common smooth periodicity with the declared generator.

The same local analysis in rotating metrics also includes an angular identification. Smoothness must make the Euclidean time and azimuthal shifts compatible at both endpoints. Gibbons, Lü, Page, and Pope 2005, §5, especially eqs. (5.2)–(5.7) shows why matching $|\kappa|$ alone is then insufficient: the horizon angular velocities must satisfy the same global identifications.

Gravitational redshift cannot repair the mismatch either. A static observer at one radius $r$ has four-velocity $u=T/\sqrt{f(r)}$. Any horizon frequency defined relative to $T$ is converted to that observer’s proper time by the same factor $1/\sqrt{f(r)}$. Thus the two local scales at that observer obey

$$\frac{\kappa_b/\sqrt{f(r)}} {|\kappa_c|/\sqrt{f(r)}} =\frac{\kappa_b}{|\kappa_c|}.$$

Moving the observer changes both redshifted values but not their ratio at the same event. Comparing values assigned by two different observers without transporting them to one normalization merely reintroduces the invalid two-rescaling trick.

The Nariai limit makes this point especially clear. As the roots coalesce, the maximum of $f$ between them tends to zero, so the static-geodesic normalization $T/\sqrt{f(r_g)}$ diverges while the original surface gravities vanish. A finite limiting scale emerges only after rescaling the coordinates and taking the geometry’s limit together. Limits of normalized generators need not equal the normalization of the limiting generator.

The charged lukewarm exception

For Reissner–Nordström–de Sitter, use the geometric charge length $q$ and

$$f(r)=1-\frac{2m}{r}+\frac{q^2}{r^2}-\frac{r^2}{\ell^2}.$$

On the branch $q^2=m^2$,

$$f(r)=\left(1-\frac{m}{r}\right)^2-\frac{r^2}{\ell^2}.$$

The outer black-hole and cosmological roots satisfy

$$r^2-\ell r+m\ell=0,$$

so $r_b+r_c=\ell$. At either root,

$$\kappa_h=\frac1\ell-\frac{2r_h}{\ell^2},$$

which immediately gives $\kappa_c=-\kappa_b$. The horizons are distinct for the appropriate subextremal range, yet their magnitudes agree. This is the lukewarm family classified in Romans 1992, §§2–3.

Charge therefore changes the conclusion rather than refuting the uncharged derivation. A complete equilibrium statement must also compare electric potentials and specify the ensemble or boundary condition. Equal surface gravity is necessary for a common Euclidean period but does not erase other chemical or rotational data.

Diagnose an invalid normalization trick

Suppose $|\kappa_b|\ne|\kappa_c|$. A calculation rescales the generator by $C_b$ at the black-hole horizon and by a different $C_c$ at the cosmological horizon so that the reported values agree. Has it produced one equilibrium normalization? When can distinct horizons genuinely have equal magnitudes?

Solution

No. A single stationary patch has one global Killing field. Multiplying it by one constant rescales both surface gravities, leaving their ratio unchanged. Using $C_bT$ at one boundary and $C_cT$ at the other compares two different Hamiltonian evolutions and cannot define one Euclidean time or one global normalization.

In uncharged Schwarzschild–de Sitter, genuine equality forces the Nariai double root. In charged Reissner–Nordström–de Sitter, the lukewarm branch $q^2=m^2$ can have distinct $r_b$ and $r_c$ with $\kappa_c=-\kappa_b$. Rotating cases require angular identifications in addition to equal magnitudes.

Nariai and Ultracold Limits next resolves the geometry retained when two or three horizons coalesce.

References

• Bardeen, James M., Brandon Carter, and Stephen W. Hawking. “The Four Laws of Black Hole Mechanics.” Communications in Mathematical Physics 31 (1973): 161–170. DOI.
• Gibbons, Gary W., Hong Lü, Don N. Page, and Christopher N. Pope. “The General Kerr–de Sitter Metrics in All Dimensions.” Journal of Geometry and Physics 53 (2005): 49–73. DOI. Open manuscript.
• Romans, Larry J. “Supersymmetric, Cold and Lukewarm Black Holes in Cosmological Einstein–Maxwell Theory.” Nuclear Physics B 383 (1992): 395–415. DOI. Open manuscript.