Figure 5.
Introduction
The space of black holes in higher-dimensional vacuum gravity is richer than its four-dimensional counterpart. Independent rotation planes and non-spherical horizon topology allow geometrically distinct solutions to carry the same mass and angular momenta. In five dimensions, the Myers–Perry black hole [1 ] , the singly spinning black ring [2 ] and the doubly spinning Pomeransky–Sen’kov ring [3 ] provide explicit examples of this richer structure.
Disconnected horizons enlarge this space further. Black Saturn consists of a spherical black hole surrounded by a black ring. Its singly spinning solution exhibits continuous non-uniqueness and frame dragging between the two components [7 ] . Other exact configurations include concentric black di-rings, whose rings lie in the same plane [8 , 9 ] , and orthogonal or bicycling black rings, whose circular directions lie in different planes [10 , 11 ] . Constructions and phase-space arguments also extend to configurations with several rings, with or without a central hole [8 , 10 , 11 , 12 ] . These solutions show how the freedom to distribute mass and angular momentum among several horizons produces non-uniqueness at fixed asymptotic charges.
For stationary five-dimensional vacuum spacetimes with two commuting axial symmetries, the Einstein equations reduce to an integrable system on a two-dimensional orbit space [21 , 25 ] . Inverse scattering then generates exact solutions by algebraic transformations of a simpler seed metric [25 , 26 ] . This method underlies the constructions of the doubly spinning ring and singly spinning Saturn [3 , 7 , 27 ] . Extending the Saturn solution to independent rotation in both planes on each horizon is nevertheless difficult. The resulting expressions must satisfy the axis identifications, conical balance and horizon regularity conditions simultaneously.
The second spin is particularly important for extremality. Both the five-dimensional Myers–Perry black hole and the balanced doubly spinning ring require nonzero angular momenta in both planes to reach a regular extremal horizon of finite area [1 , 3 ] . These examples motivate retaining both spins on each component of a Saturn. The blackfold construction already provides perturbative evidence for a thin extremal doubly spinning ring surrounding an extremal doubly spinning central hole [14 ] . An exact solution is needed to determine how this configuration extends beyond the thin-ring approximation and to study its full exterior geometry.
In this paper we construct a six-parameter family of doubly spinning black Saturns in five-dimensional vacuum gravity. We extend the inverse-scattering construction of the doubly spinning ring [27 ] to a seed with the rod structure appropriate to a ring surrounding a spherical hole, realising the Saturn generalisation proposed there. We use AI tools to assist with the symbolic manipulations and numerical checks. We exhibit balanced configurations in which both horizons are extremal and have finite area, with a regular, causal exterior in the parameter domains established below. To our knowledge, this gives the first regular exact solution with standard asymptotic flatness and disconnected extremal horizons of topologies S 3 S^3 S 3 and S 1 × S 2 S^1 \times S^2 S 1 × S 2 in vacuum gravity.
We organise the paper as follows. § 2 presents the metric, its parameters and the balance conditions. § 3 studies the geometry, thermodynamics and nonextremal regularity. § 4 constructs the doubly extremal metric and its horizon extensions. § 5 discusses phase diagrams, comparisons with blackfolds and frame dragging. We conclude in § 6 . Appendix § 8 derives the single-object and singly spinning limits, together with the two metrics in which only one horizon is extremal, including the required parameter scalings and coordinate continuations. Appendix § 9 gives the explicit metric coefficients and thermodynamic formulas, followed by the regularity criteria and the domains covered by our checks. Appendix § 10 derives the blackfold balance equations, mutual-equilibrium solutions and thermodynamics, and explains their comparison with the exact solution. Appendix § 11 proves the obstruction to mutual equilibrium for the doubly extremal family of § 4 , and distinguishes the exact boundary exclusions from the evidence supplied by wider numerical searches.
The doubly-spinning Saturn metric
The construction of the doubly-spinning black Saturn has three ingredients. We first generate a stationary vacuum metric with two candidate horizon rods. We then choose the angular identifications and impose balance, so that the axes describe smooth rotation axes without conical defects. Finally, we write the result in scalar functions suitable for reading off its geometry.
Geometry and inverse-scattering construction
We seek a five-dimensional solution of vacuum Einstein gravity, R μ ν = 0 R_{\mu\nu} = 0 R μν = 0 , with one stationary and two commuting rotational Killing fields. The rotations act in the ring plane and in the transverse plane. Away from the axes and horizons, generalised Weyl coordinates separate the three Killing directions from the two-dimensional orbit space [20 , 21 , 22 ] :
d s 2 = G a b d x a d x b + f ( d ρ 2 + d z 2 ) , x a = ( t , ψ , ϕ ) , det G = − ρ 2 . (1) \mathrm{d}s^2 = G_{ab}\mathrm{d}x^a\mathrm{d}x^b + f(\mathrm{d}\rho^2 + \mathrm{d}z^2), \qquad x^a = (t,\psi,\phi), \qquad\det G = -\rho^2. \tag*{(1)} d s 2 = G ab d x a d x b + f ( d ρ 2 + d z 2 ) , x a = ( t , ψ , ϕ ) , det G = − ρ 2 . (1)
Here G G G is a symmetric 3 × 3 3 \times3 3 × 3 matrix containing the time and angular components, while f f f is the conformal factor of the orbit-space metric. The condition det G = − ρ 2 \det G = -\rho^2 det G = − ρ 2 fixes the Weyl radial coordinate. The coordinates ( t , ψ , ϕ ) (t,\psi,\phi) ( t , ψ , ϕ ) are the Killing coordinates used in the construction. Their physical normalisation and angular periods will be fixed in § 2.2 .
The boundary ρ = 0 \rho= 0 ρ = 0 is divided into intervals called rods [21 , 22 ] . Rod structure is the one-dimensional instance of the more general domain structure [23 , 24 ] . On each open rod, a linear combination of Killing fields lies in the kernel of G G G . A spacelike kernel is a shrinking angular circle while a horizon kernel is its null generator. Our target is the sequence axis-ring horizon-axis-spherical horizon-axis shown in Figure 1 (PDF p. 9) . Two additional points facilitate the construction and will subsequently be made smooth.
Figure 1 (PDF p. 9). Physical rods after auxiliary-point removal. The horizons have generators χ h = ∂ T + Ω A d φ A \chi_{\mathrm{h}}=\partial_T+\Omega_A d\varphi^A χ h = ∂ T + Ω A d φ A . The points a 1 , a 6 a_1,a_6 a 1 , a 6 are smooth auxiliary seams, not physical turning points. Rod lengths in this drawing are schematic.
The vacuum equations for G G G reduce to the integrable matrix equation [21 , 25 ]
∂ ρ ( ρ G − 1 ∂ ρ G ) + ∂ z ( ρ G − 1 ∂ z G ) = 0. (2) \partial_{\rho}\left(\rho G^{-1}\partial_{\rho}G\right)+\partial_{z}\left(\rho G^{-1}\partial_{z}G\right)=0. \tag*{(2)} ∂ ρ ( ρ G − 1 ∂ ρ G ) + ∂ z ( ρ G − 1 ∂ z G ) = 0. (2)
Once G G G is known, define U = ρ ( ∂ ρ G ) G − 1 U=\rho(\partial_{\rho}G)G^{-1} U = ρ ( ∂ ρ G ) G − 1 and V = ρ ( ∂ z G ) G − 1 V=\rho(\partial_{z}G)G^{-1} V = ρ ( ∂ z G ) G − 1 . The remaining Einstein equations are [21 , 25 ]
∂ ρ log f = − 1 ρ + 1 4 ρ Tr ( U 2 − V 2 ) , ∂ z log f = 1 2 ρ Tr ( U V ) . (3) \partial_{\rho}\log f=-\frac{1}{\rho}+\frac{1}{4\rho}\operatorname{Tr}(U^2-V^2),\qquad\partial_{z}\log f=\frac{1}{2\rho}\operatorname{Tr}(UV). \tag*{(3)} ∂ ρ log f = − ρ 1 + 4 ρ 1 Tr ( U 2 − V 2 ) , ∂ z log f = 2 ρ 1 Tr ( U V ) . (3)
Their compatibility follows from (2) . Integrating these equations determines f f f up to a multiplicative constant. We use the Belinski–Zakharov inverse-scattering method [26 ] , in its higher-dimensional form [25 , 27 ] , to generate G G G from a simpler diagonal solution called the seed. The method replaces the nonlinear matrix equation by an auxiliary linear system, then transforms a known solution of that system algebraically. The resulting metric solves the vacuum field equations.
To proceed further, we choose six ordered points and their associated pole functions
0 = a 1 < a 2 < a 3 < a 4 < a 5 < a 6 , u i = ρ 2 + ( z − a i ) 2 − ( z − a i ) , v i = − ρ 2 u i . (4) 0=a_1<a_2<a_3<a_4<a_5<a_6,\qquad u_i=\sqrt{\rho^2+(z-a_i)^2}-(z-a_i),\qquad v_i=-\frac{\rho^2}{u_i}. \tag*{(4)} 0 = a 1 < a 2 < a 3 < a 4 < a 5 < a 6 , u i = ρ 2 + ( z − a i ) 2 − ( z − a i ) , v i = − u i ρ 2 . (4)
Fixing a 1 = 0 a_1=0 a 1 = 0 chooses the origin of z z z without removing a physical parameter. For ρ > 0 \rho>0 ρ > 0 , u i u_i u i and v i v_i v i are the positive and negative roots of λ 2 + 2 ( z − a i ) λ − ρ 2 = 0 \lambda^2+2(z-a_i)\lambda-\rho^2=0 λ 2 + 2 ( z − a i ) λ − ρ 2 = 0 . They encode the rod endpoints and make the metric functions algebraic. We define x = ρ 2 x=\rho^2 x = ρ 2 and d i j = a i − a j d_{ij}=a_i-a_j d ij = a i − a j and choose the diagonal seed to be
G s = diag ( − u 1 u 4 u 3 u 6 , u 2 u 5 u 1 , x u 3 u 6 u 2 u 4 u 5 ) , (5) G_s=\operatorname{diag}\left(-\frac{u_1u_4}{u_3u_6},\frac{u_2u_5}{u_1},\frac{x u_3u_6}{u_2u_4u_5}\right), \tag*{(5)} G s = diag ( − u 3 u 6 u 1 u 4 , u 1 u 2 u 5 , u 2 u 4 u 5 x u 3 u 6 ) , (5)
whose determinant is − x -x − x . The powers of the u i u_i u i specify which Killing components vanish or diverge on each axis interval. The seed is an intermediate solution and regularity is imposed on the final metric.
To introduce rotation while retaining det G = − x \det G=-x det G = − x , we use the two-step soliton transformation of [25 , 27 ] . First, we remove six poles with polarisations along single Killing directions. Then we insert the same poles with polarisations that mix those directions. A polarisation is a constant vector specifying the combination of Killing directions used in a soliton transformation. The first step gives a diagonal matrix H H H while the second generates the time-angle and angle-angle components in the metric responsible for rotation. Because the pole locations are the same in both steps, their determinant factors cancel.
The six removed poles and their trivial directions are
Table 4.
The auxiliary linear system uses a complex variable λ \lambda λ , called the spectral parameter, in addition to ρ \rho ρ , z z z . Its matrix solution Ψ 0 ( λ , ρ , z ) \Psi_0(\lambda,\rho,z) Ψ 0 ( λ , ρ , z ) encodes H H H through Ψ 0 ( 0 , ρ , z ) = H \Psi_0(0,\rho,z)=H Ψ 0 ( 0 , ρ , z ) = H . The pole functions u i u_i u i , v i v_i v i specify the moving poles of the soliton transformation in this auxiliary variable [25 , 26 ] . The quantities needed below are
H = diag ( − u 1 u 3 u 4 u 6 , − u 1 u 2 u 5 x , − u 2 u 3 u 5 u 4 u 6 ) (6) H=\operatorname{diag}\left(-\frac{u_1u_3}{u_4u_6},-\frac{u_1u_2u_5}{x},-\frac{u_2u_3u_5}{u_4u_6}\right) \tag*{(6)} H = diag ( − u 4 u 6 u 1 u 3 , − x u 1 u 2 u 5 , − u 4 u 6 u 2 u 3 u 5 ) (6)
P b ( λ ) = x − 2 ( z − b ) λ − λ 2 , T ( λ ) = ( u 4 − λ ) ( u 6 − λ ) , b = − 1. (7) P_b(\lambda)=x-2(z-b)\lambda-\lambda^2,\qquad\mathcal{T}(\lambda)=(u_4-\lambda)(u_6-\lambda),\qquad b=-1. \tag*{(7)} P b ( λ ) = x − 2 ( z − b ) λ − λ 2 , T ( λ ) = ( u 4 − λ ) ( u 6 − λ ) , b = − 1. (7)
We place b b b to the left of a 1 = 0 a_1=0 a 1 = 0 and choose the unit of the Weyl coordinates so that a 1 − b = 1 a_1-b=1 a 1 − b = 1 , hence b = − 1 b=-1 b = − 1 . This auxiliary separation is not a physical rod length and fixes none of the gaps a i + 1 − a i a_{i+1}-a_i a i + 1 − a i . A different negative b b b gives the same family after redefining the polarisation constants. The factor P b P_b P b is a convenient way to represent the seed factor x x x in the auxiliary system, since P b ( 0 ) = x P_b(0)=x P b ( 0 ) = x . Its freedom is a choice of representation for Ψ 0 \Psi_0 Ψ 0 , accompanied by a redefinition of the constant polarisations. Indeed, at a pole, P b ( u i ) = 2 ( b − a i ) u i P_b(u_i)=2(b-a_i)u_i P b ( u i ) = 2 ( b − a i ) u i , so changing b b b only rescales the corresponding constant coefficients. Keeping b b b to the left of all a i a_i a i avoids a zero of this factor. For this choice a diagonal generating matrix is
( Ψ 0 ) t t = − ( u 1 − λ ) ( u 3 − λ ) T ( λ ) , ( Ψ 0 ) ψ ψ = − ( u 1 − λ ) ( u 2 − λ ) ( u 5 − λ ) P b ( λ ) , ( Ψ 0 ) ϕ ϕ = − ( u 2 − λ ) ( u 3 − λ ) ( u 5 − λ ) T ( λ ) , ( Ψ 0 ) a b = 0 ( a ≠ b ) . (8) \begin{aligned}
(\Psi_0)_{tt}&=-\frac{(u_1-\lambda)(u_3-\lambda)}{\mathcal{T}(\lambda)},&
(\Psi_0)_{\psi\psi}&=-\frac{(u_1-\lambda)(u_2-\lambda)(u_5-\lambda)}{P_b(\lambda)},\\
(\Psi_0)_{\phi\phi}&=-\frac{(u_2-\lambda)(u_3-\lambda)(u_5-\lambda)}{\mathcal{T}(\lambda)},&
(\Psi_0)_{ab}&=0\quad(a\ne b). \tag*{(8)}
\end{aligned} ( Ψ 0 ) tt ( Ψ 0 ) ϕϕ = − T ( λ ) ( u 1 − λ ) ( u 3 − λ ) , = − T ( λ ) ( u 2 − λ ) ( u 3 − λ ) ( u 5 − λ ) , ( Ψ 0 ) ψ ψ ( Ψ 0 ) ab = − P b ( λ ) ( u 1 − λ ) ( u 2 − λ ) ( u 5 − λ ) , = 0 ( a = b ) . (8)
This matrix satisfies the auxiliary linear equations, usually called the Lax system [25 , 26 , 27 ] . Their compatibility is (2) : this is why transforming the generating matrix produces another vacuum solution. Inserting the six poles with new polarisations then gives the following position-dependent rows m i m_i m i . They contain the result of solving the auxiliary system once and for all. Evaluating the final metric only requires these explicit rows.
The constants ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) (C_1,C_2,C_3,B_4,B_5,C_6) ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) are polarisation parameters. Let T i = T ( u i ) \mathcal{T}_i=\mathcal{T}(u_i) T i = T ( u i ) , T ^ i = T ( v i ) \widehat{\mathcal{T}}_i=\mathcal{T}(v_i) T i = T ( v i ) and P i = P b ( u i ) P_i=P_b(u_i) P i = P b ( u i ) , then
m 1 = ( C 1 T 1 u 3 − u 1 , P 1 ( u 2 − u 1 ) ( u 5 − u 1 ) , 0 ) , m 2 = ( 0 , C 2 P 2 ( u 1 − u 2 ) ( u 5 − u 2 ) , T 2 ( u 3 − u 2 ) ( u 5 − u 2 ) ) , m 3 = ( T 3 u 1 − u 3 , 0 , C 3 T 3 ( u 2 − u 3 ) ( u 5 − u 3 ) ) , m 4 = ( − T ^ 4 ( u 1 − v 4 ) ( u 3 − v 4 ) , 0 , − B 4 T ^ 4 ( u 2 − v 4 ) ( u 3 − v 4 ) ( u 5 − v 4 ) ) , m 5 = ( 0 , B 5 P 5 ( u 1 − u 5 ) ( u 2 − u 5 ) , T 5 ( u 2 − u 5 ) ( u 3 − u 5 ) ) , m 6 = ( − C 6 T ^ 6 ( u 1 − v 6 ) ( u 3 − v 6 ) , 0 , − T ^ 6 ( u 2 − v 6 ) ( u 3 − v 6 ) ( u 5 − v 6 ) ) . (9) \begin{aligned}
m_1&=\left(\frac{C_1\mathcal{T}_1}{u_3-u_1},\frac{P_1}{(u_2-u_1)(u_5-u_1)},0\right),\\
m_2&=\left(0,\frac{C_2P_2}{(u_1-u_2)(u_5-u_2)},\frac{\mathcal{T}_2}{(u_3-u_2)(u_5-u_2)}\right),\\
m_3&=\left(\frac{\mathcal{T}_3}{u_1-u_3},0,\frac{C_3\mathcal{T}_3}{(u_2-u_3)(u_5-u_3)}\right),\\
m_4&=\left(-\frac{\widehat{\mathcal{T}}_4}{(u_1-v_4)(u_3-v_4)},0,-\frac{B_4\widehat{\mathcal{T}}_4}{(u_2-v_4)(u_3-v_4)(u_5-v_4)}\right),\\
m_5&=\left(0,\frac{B_5P_5}{(u_1-u_5)(u_2-u_5)},\frac{\mathcal{T}_5}{(u_2-u_5)(u_3-u_5)}\right),\\
m_6&=\left(-\frac{C_6\widehat{\mathcal{T}}_6}{(u_1-v_6)(u_3-v_6)},0,-\frac{\widehat{\mathcal{T}}_6}{(u_2-v_6)(u_3-v_6)(u_5-v_6)}\right). \tag*{(9)}
\end{aligned} m 1 m 2 m 3 m 4 m 5 m 6 = ( u 3 − u 1 C 1 T 1 , ( u 2 − u 1 ) ( u 5 − u 1 ) P 1 , 0 ) , = ( 0 , ( u 1 − u 2 ) ( u 5 − u 2 ) C 2 P 2 , ( u 3 − u 2 ) ( u 5 − u 2 ) T 2 ) , = ( u 1 − u 3 T 3 , 0 , ( u 2 − u 3 ) ( u 5 − u 3 ) C 3 T 3 ) , = ( − ( u 1 − v 4 ) ( u 3 − v 4 ) T 4 , 0 , − ( u 2 − v 4 ) ( u 3 − v 4 ) ( u 5 − v 4 ) B 4 T 4 ) , = ( 0 , ( u 1 − u 5 ) ( u 2 − u 5 ) B 5 P 5 , ( u 2 − u 5 ) ( u 3 − u 5 ) T 5 ) , = ( − ( u 1 − v 6 ) ( u 3 − v 6 ) C 6 T 6 , 0 , − ( u 2 − v 6 ) ( u 3 − v 6 ) ( u 5 − v 6 ) T 6 ) . (9)
The soliton ansatz has only six pole residues. The equations fixing those residues form a finite linear system, whose coefficients are Γ i j \Gamma_{ij} Γ ij below. Solving it gives the correction to $H.
With ( p i ) = ( u 1 , u 2 , u 3 , v 4 , u 5 , v 6 ) (p_i)=(u_1,u_2,u_3,v_4,u_5,v_6) ( p i ) = ( u 1 , u 2 , u 3 , v 4 , u 5 , v 6 ) , we can form the symmetric 6 × 6 6 \times6 6 × 6 matrix Γ \Gamma Γ and the 3 × 6 3 \times6 3 × 6 matrix K K K according to:
Γ i j = m i H m j T x + p i p j , K a i = H a a ( m i ) a p i , G = H − K Γ − 1 K T . (10) \Gamma_{ij}=\frac{m_i H m_j^{\mathsf{T}}}{x+p_i p_j},\qquad K_{ai}=\frac{H_{aa}(m_i)_a}{p_i},\qquad G=H-K\Gamma^{-1}K^{\mathsf{T}}. \tag*{(10)} Γ ij = x + p i p j m i H m j T , K ai = p i H aa ( m i ) a , G = H − K Γ − 1 K T . (10)
The term K Γ − 1 K T K\Gamma^{-1}K^{\mathsf{T}} K Γ − 1 K T incorporates the interaction of all six poles. Together with the rows above, (10) specifies every time and angular component.
It remains to determine f f f , the coefficient of d ρ 2 + d z 2 d\rho^2+dz^2 d ρ 2 + d z 2 in (2.1. This coefficient fixes proper distances in the orbit space and enters the balance and temperature formulas. Let f s f_s f s be the corresponding coefficient for the diagonal seed G s G_s G s . Integrating (2.3 with G = G s G=G_s G = G s gives f s f_s f s as a product of R i j = x + u i u j R_{ij}=x+u_i u_j R ij = x + u i u j . To state the exponents compactly, define w i w_i w i so that its three entries are the powers of u i u_i u i in the t t , ψ ψ , ϕ ϕ tt,\psi\psi,\phi\phi tt , ψ ψ , ϕϕ seed components. For example u 1 u_1 u 1 has powers ( 1 , − 1 , 0 ) (1,-1,0) ( 1 , − 1 , 0 ) . The six exponent vectors are
w 1 = ( 1 , − 1 , 0 ) , w 2 = w 5 = ( 0 , 1 , − 1 ) , w 3 = w 6 = ( − 1 , 0 , 1 ) , w 4 = ( 1 , 0 , − 1 ) . (11) w_1=(1,-1,0),\qquad w_2=w_5=(0,1,-1),\qquad w_3=w_6=(-1,0,1),\qquad w_4=(1,0,-1). \tag*{(11)} w 1 = ( 1 , − 1 , 0 ) , w 2 = w 5 = ( 0 , 1 , − 1 ) , w 3 = w 6 = ( − 1 , 0 , 1 ) , w 4 = ( 1 , 0 , − 1 ) . (11)
The seed coefficient and the coefficient of the final metric are respectively
f s = u 2 u 5 u 1 ∏ i R i i − 1 ∏ i < j R i j − w i ⋅ w j , f = f s det Γ ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) det Γ ( 0 , 0 , 0 , 0 , 0 , 0 ) . (12) f_s=\frac{u_2u_5}{u_1}\prod_i R_{ii}^{-1}\prod_{i<j}R_{ij}^{-w_i\cdot w_j},\qquad
f=f_s\frac{\det\Gamma(C_1,C_2,C_3,B_4,B_5,C_6)}{\det\Gamma(0,0,0,0,0,0)}. \tag*{(12)} f s = u 1 u 2 u 5 i ∏ R ii − 1 i < j ∏ R ij − w i ⋅ w j , f = f s det Γ ( 0 , 0 , 0 , 0 , 0 , 0 ) det Γ ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) . (12)
We set the multiplicative constant in f s f_s f s to one as an initial normalisation convention. A different positive constant rescales proper distances and is accounted for when fixing angular periods and the overall length scale in sections 2.2 and 2.4. In the denominator of (12) , all six polarisation constants are zero. This is the transformation that inserts the poles along the original single Killing directions and therefore recovers the seed. The determinant ratio measures the change in f f f when those original polarisations are replaced by the rotating ones.
Equations (10) and (12) now specify the full vacuum metric wherever their denominators are nonzero. The next step is to select parameters and angular identifications for which its axes and horizons are regular.
Physical rods, normalisation and balance
An exact vacuum metric that results from this procedure is not necessarily a regular black hole spacetime. We must remove the two auxiliary rod points, fix the physical angular circles, and ensure that the finite axis between the horizons has a smooth identification. These requirements have different roles. First, a 1 a_1 a 1 and a 6 a_6 a 6 are auxiliary points, not ends of physical horizons. Requiring their apparent turning points to disappear fixes two polarisation parameters. On the choice of signs used here,
C 1 = − d 31 2 d 21 d 41 d 51 d 61 , C 6 = d 61 2 d 62 d 65 . (13) C_1=-\sqrt{\frac{d_{31}}{2d_{21}d_{41}d_{51}d_{61}}},\qquad C_6=\sqrt{\frac{d_{61}}{2d_{62}d_{65}}}. \tag*{(13)} C 1 = − 2 d 21 d 41 d 51 d 61 d 31 , C 6 = 2 d 62 d 65 d 61 . (13)
In fact the complete metric is invariant under each root involution u g ↦ − x / u g u_g\mapsto-x/u_g u g ↦ − x / u g , g = 1 , 6 g=1,6 g = 1 , 6 , after imposing these squared relations. Odd dependence on the auxiliary square roots therefore cancels. Regularity at their centres additionally requires the nonzero denominator and signature conditions discussed below.
Second, the outer axes determine the angular identifications. Let v L , v R v_L,v_R v L , v R be their kernels in the original ( t , ψ , ϕ ) (t,\psi,\phi) ( t , ψ , ϕ ) coordinates, normalised by v L ϕ = 1 v_L^\phi=1 v L ϕ = 1 and v R ψ = 1 v_R^\psi=1 v R ψ = 1 . If an orbit of v i v_i v i is provisionally given period 2 π 2\pi 2 π , its circumference divided by its proper radius approaches 2 π q i 2\pi\sqrt{q_i} 2 π q i , with q i q_i q i as below. Thus the smooth 2 π 2\pi 2 π -periodic generator is v i / q i v_i/\sqrt{q_i} v i / q i , and the physical Killing basis is
q i = lim ρ → 0 v i T G v i ρ 2 f , A = ( e t , v R q R , v L q L ) , δ = det A > 0. (14) q_i=\lim_{\rho\to0}\frac{v_i^{\mathsf T}Gv_i}{\rho^2f},\qquad A=\left(e_t,\frac{v_R}{\sqrt{q_R}},\frac{v_L}{\sqrt{q_L}}\right),\qquad\delta=\det A>0. \tag*{(14)} q i = ρ → 0 lim ρ 2 f v i T G v i , A = ( e t , q R v R , q L v L ) , δ = det A > 0. (14)
For the branch considered here q L , q R > 0 q_L,q_R>0 q L , q R > 0 . The physical metric is
d s A 2 = G a b d X a d X b + f ( d ρ 2 + d z 2 ) , G = A T G A , X = ( T , Ψ , Φ ) , det G = − δ 2 ρ 2 . (15) ds_A^2=\mathcal{G}_{ab}dX^a dX^b+f(d\rho^2+dz^2),\qquad\mathcal{G}=A^{\mathsf T}GA,\qquad X=(T,\Psi,\Phi),\qquad\det\mathcal{G}=-\delta^2\rho^2. \tag*{(15)} d s A 2 = G ab d X a d X b + f ( d ρ 2 + d z 2 ) , G = A T G A , X = ( T , Ψ , Φ ) , det G = − δ 2 ρ 2 . (15)
The columns of A A A mean x a = A a b X b x^a=A^a{}_bX^b x a = A a b X b locally. Globally we take Ψ \Psi Ψ and Φ \Phi Φ to be independently 2 π 2\pi 2 π -periodic at fixed T T T . The time components of the outer-axis kernels are absorbed into the definition of the angular generators. The stationary vector remains ∂ T = ∂ t \partial_T=\partial_t ∂ T = ∂ t and has unit norm at infinity. Since the basis change multiplies the Killing determinant by δ 2 \delta^2 δ 2 , the canonical Weyl coordinates for this same metric are ρ ˉ = δ ρ \bar{\rho}=\delta\rho ρ ˉ = δ ρ and z ˉ = δ ( z − z 0 ) \bar{z}=\delta(z-z_0) z ˉ = δ ( z − z 0 ) , with base factor f / δ 2 f/\delta^2 f / δ 2 .
Third, the finite axis a 3 < z < a 4 a_3<z<a_4 a 3 < z < a 4 must shrink the same primitive circle as the lower axis. Here “primitive” means a cycle traversed once, not an integer multiple of that cycle. With v M t = 1 v_M^t=1 v M t = 1 , matching the time and ψ \psi ψ components fixes its direction, and matching q M = q L q_M=q_L q M = q L fixes its circumference-to-radius ratio:
B = ( v M t − v L t , v M ψ − v L ψ , q M / q L − 1 ) = 0. (16) \mathcal{B}=\left(v_M^t-v_L^t,v_M^\psi-v_L^\psi,q_M/q_L-1\right)=0. \tag*{(16)} B = ( v M t − v L t , v M ψ − v L ψ , q M / q L − 1 ) = 0. (16)
The first two equations align the degenerating cycles while the third is mechanical balance, excluding a conical strut or membrane that would otherwise support the ring. These conditions are algebraic. To see how they are obtained, we expand G = G 0 + O ( ρ 2 ) G=G_0+O(\rho^2) G = G 0 + O ( ρ 2 ) and f = f 0 + O ( ρ 2 ) f=f_0+O(\rho^2) f = f 0 + O ( ρ 2 ) on an open spacelike rod. If the kernel is normalised by v k = 1 v^k=1 v k = 1 and I I I denotes the other two Killing directions, solving G 0 v = 0 G_0v=0 G 0 v = 0 gives
v I = − G 0 , I I − 1 G 0 , I k , q = − 1 f 0 det G 0 , I I . (17) v^I=-G_{0,II}^{-1}G_{0,Ik},\qquad q=-\frac{1}{f_0\det G_{0,II}}. \tag*{(17)} v I = − G 0 , I I − 1 G 0 , I k , q = − f 0 det G 0 , I I 1 . (17)
For the second identity, the determinant in the basis ( v , e I ) (v,e_I) ( v , e I ) is ( v T G v ) det G 0 , I I + O ( x 2 ) = − x (v^{\mathsf T}Gv)\det G_{0,II}+O(x^2)=-x ( v T G v ) det G 0 , I I + O ( x 2 ) = − x . Dividing by x f 0 xf_0 x f 0 gives q q q , the square of the circumference-to-radius ratio divided by 2 π 2\pi 2 π . Thus q = 1 q=1 q = 1 means that a 2 π 2\pi 2 π -periodic angle has no conical deficit or excess. In subsection B.3 we give finite algebraic formulas for the rod coefficients, so these relations also specify how to evaluate (16) .
After removing the two auxiliary points the physical rod diagram is shown in Figure 1 (PDF p. 9) . Equal adjacent cycles on the ring give S 1 × S 2 S^1\times S^2 S 1 × S 2 topology while the different primitive cycles bordering the central rod give S 3 S^3 S 3 . The outer cycles generate the standard angular lattice, so the asymptotic end has S 3 S^3 S 3 , with no lens-space quotient.
Explicit scalar form
The matrix construction is compact and explains the origin of the solution. To read the line element directly, it is useful to package the same components into the scalar-function form familiar from the singly spinning Saturn [7 ] . The second spin requires two rotational potentials and a mixed angular coefficient. We fix the positive auxiliary function
P = R 34 2 R 12 R 23 , R i j = ρ 2 + u i u j , (18) P=R_{34}^{2}R_{12}R_{23}, \qquad R_{ij}=\rho^{2}+u_i u_j, \tag*{(18)} P = R 34 2 R 12 R 23 , R ij = ρ 2 + u i u j , (18)
and introduce seven scalar functions H x , H y , ω ψ , ω ϕ , V ψ ψ , V ψ ϕ , V ϕ ϕ H_x,H_y,\omega_\psi,\omega_\phi,V_{\psi\psi},V_{\psi\phi},V_{\phi\phi} H x , H y , ω ψ , ω ϕ , V ψ ψ , V ψ ϕ , V ϕϕ . Their complete definitions as finite sums of scalar products are given in (B.8)–(B.10) in subsection B.1. The ancillary Mathematica file supplies all seven expanded numerator polynomials and an evaluator, with the symbol dictionary given in the same appendix. In the original Killing coordinates ( t , ψ , ϕ ) (t,\psi,\phi) ( t , ψ , ϕ ) of (1) , before imposing the physical angular periods, the full line element is
d s 2 = 1 H x [ − H y d t 2 − 2 ω ψ d t d ψ − 2 ω ϕ d t d ϕ + V ψ ψ d ψ 2 + 2 V ψ ϕ d ψ d ϕ + V ϕ ϕ d ϕ 2 ] + H x P ( d ρ 2 + d z 2 ) . (19) ds^2=\frac{1}{H_x}\left[-H_y dt^2-2\omega_\psi dt d\psi-2\omega_\phi dt d\phi+V_{\psi\psi}d\psi^2+2V_{\psi\phi}d\psi d\phi+V_{\phi\phi}d\phi^2\right]+H_xP(d\rho^2+dz^2). \tag*{(19)} d s 2 = H x 1 [ − H y d t 2 − 2 ω ψ d t d ψ − 2 ω ϕ d t d ϕ + V ψ ψ d ψ 2 + 2 V ψ ϕ d ψ d ϕ + V ϕϕ d ϕ 2 ] + H x P ( d ρ 2 + d z 2 ) . (19)
Thus (19) explicitly displays all six Killing-metric components and the base factor. The functions are rational in the pole functions and polynomial in the six polarisation constants, with a common, positive, polarisation-independent denominator specified in (B.11).
Another closed form of the metric is possible and isolates the stationary one-form by completing the square in d t dt d t . The angular coefficients left over from this operation are expressed through
F ψ ψ = H y V ψ ψ + ω ψ 2 H x 2 , F ψ ϕ = H y V ψ ϕ + ω ψ ω ϕ H x 2 , F ϕ ϕ = H y V ϕ ϕ + ω ϕ 2 H x 2 . (20) F_{\psi\psi}=\frac{H_yV_{\psi\psi}+\omega_\psi^2}{H_x^2}, \qquad F_{\psi\phi}=\frac{H_yV_{\psi\phi}+\omega_\psi\omega_\phi}{H_x^2}, \qquad F_{\phi\phi}=\frac{H_yV_{\phi\phi}+\omega_\phi^2}{H_x^2}. \tag*{(20)} F ψ ψ = H x 2 H y V ψ ψ + ω ψ 2 , F ψ ϕ = H x 2 H y V ψ ϕ + ω ψ ω ϕ , F ϕϕ = H x 2 H y V ϕϕ + ω ϕ 2 . (20)
The constants in (14) are q A = A A t q_A=A^t_A q A = A A t and b i A = A A i b_{iA}=A^i_A b i A = A A i , with i = ψ , ϕ i=\psi,\phi i = ψ , ϕ , A = Ψ , Φ A=\Psi,\Phi A = Ψ , Φ . They implement d t = d T + q A d φ A dt=dT+q_A d\varphi^A d t = d T + q A d φ A and d x i = b i A d φ A dx^i=b_{iA}d\varphi^A d x i = b i A d φ A . Defining
ω A = b ψ A ω ψ + b ϕ A ω ϕ , \omega_A=b_{\psi A}\omega_\psi+b_{\phi A}\omega_\phi, ω A = b ψ A ω ψ + b ϕ A ω ϕ ,
F A B = b ψ A b ψ B F ψ ψ + ( b ψ A b ϕ B + b ϕ A b ψ B ) F ψ ϕ + b ϕ A b ϕ B F ϕ ϕ , (21) \mathcal{F}_{AB}=b_{\psi A}b_{\psi B}F_{\psi\psi}+(b_{\psi A}b_{\phi B}+b_{\phi A}b_{\psi B})F_{\psi\phi}+b_{\phi A}b_{\phi B}F_{\phi\phi}, \tag*{(21)} F A B = b ψ A b ψ B F ψ ψ + ( b ψ A b ϕB + b ϕ A b ψ B ) F ψ ϕ + b ϕ A b ϕB F ϕϕ , (21)
the physically normalised metric becomes
d s A 2 = − H y H x [ d T + ( ω Ψ H y + q Ψ ) d Ψ + ( ω Φ H y + q Φ ) d Φ ] 2 + H x [ P ( d ρ 2 + d z 2 ) + F Ψ Ψ d Ψ 2 + 2 F Ψ Φ d Ψ d Φ + F Φ Φ d Φ 2 H y ] . (22) ds_A^2=-\frac{H_y}{H_x}\left[\mathrm{d}T+\left(\frac{\omega_\Psi}{H_y}+q_\Psi\right)\mathrm{d}\Psi+\left(\frac{\omega_\Phi}{H_y}+q_\Phi\right)\mathrm{d}\Phi\right]^2+H_x\left[P(\mathrm{d}\rho^2+\mathrm{d}z^2)+\frac{F_{\Psi\Psi}\mathrm{d}\Psi^2+2F_{\Psi\Phi}\mathrm{d}\Psi\mathrm{d}\Phi+F_{\Phi\Phi}\mathrm{d}\Phi^2}{H_y}\right]. \tag*{(22)} d s A 2 = − H x H y [ d T + ( H y ω Ψ + q Ψ ) d Ψ + ( H y ω Φ + q Φ ) d Φ ] 2 + H x [ P ( d ρ 2 + d z 2 ) + H y F ΨΨ d Ψ 2 + 2 F ΨΦ d Ψ d Φ + F ΦΦ d Φ 2 ] . (22)
The physical angles Ψ \Psi Ψ and Φ \Phi Φ both have period 2 π 2\pi 2 π . The constants q Ψ , q Φ q_\Psi,q_\Phi q Ψ , q Φ are time shifts, distinct from the cone factors q L , q R q_L,q_R q L , q R . Subsection B.1 also gives scalar algebraic formulas for these constants and b i A b_{iA} b i A . To obtain a balanced Saturn, we impose both the auxiliary-point conditions (13) and the axis conditions (16) .
The stationary Killing field becomes null at an ergosurface: g T T = − H y / H x = 0 g_{TT}=-H_y/H_x=0 g T T = − H y / H x = 0 , so H y = 0 H_y=0 H y = 0 when H x H_x H x is finite and nonzero. The factors 1 / H y 1/H_y 1/ H y in the completed-square form (22) might then appear singular. They arise only from completing the square: expanding it cancels these denominators and recovers (19) , followed by the constant change of coordinates. The latter form is therefore the appropriate expression at an ergosurface. One useful algebraic consistency identity is
F Ψ Ψ F Φ Φ − F Ψ Φ 2 = ρ 2 H y H x . (23) F_{\Psi\Psi}F_{\Phi\Phi}-F_{\Psi\Phi}^2=\frac{\rho^2H_y}{H_x}. \tag*{(23)} F ΨΨ F ΦΦ − F ΨΦ 2 = H x ρ 2 H y . (23)
Indeed, the completed-square form gives det G = − ( H x / H y ) det ( F i j ) \det G=-(H_x/H_y)\det(F_{ij}) det G = − ( H x / H y ) det ( F ij ) , so (23) is precisely the Weyl constraint det G = − ρ 2 \det G=-\rho^2 det G = − ρ 2 expressed in scalar functions. It provides a check on their relative normalisation. Smoothness at H y = 0 H_y=0 H y = 0 must still be checked in the component form, where no division by H y H_y H y occurs.
The choice (18) makes the comparison with the singly spinning metric particularly direct. Under the degeneration and endpoint map of appendix § 8 .§ 3 , the scalar identities are
H x ⟶ H x E F , H y ⟶ H y E F , P ⟶ P E F , ω Ψ ⟶ − ω ψ E F , ω Φ ⟶ 0 , V Ψ Φ ⟶ 0 , F Ψ Ψ ⟶ G y E F , F Ψ Φ ⟶ 0 , F Φ Φ ⟶ H y E F H x E F G x E F . (24) \begin{aligned}
H_x&\longrightarrow H_x^{\mathrm{EF}}, & H_y&\longrightarrow H_y^{\mathrm{EF}}, & P&\longrightarrow P^{\mathrm{EF}},\\
\omega_\Psi&\longrightarrow-\omega_\psi^{\mathrm{EF}}, & \omega_\Phi&\longrightarrow0, & V_{\Psi\Phi}&\longrightarrow0,\\
F_{\Psi\Psi}&\longrightarrow G_y^{\mathrm{EF}}, & F_{\Psi\Phi}&\longrightarrow0, & F_{\Phi\Phi}&\longrightarrow\frac{H_y^{\mathrm{EF}}}{H_x^{\mathrm{EF}}}G_x^{\mathrm{EF}}. \tag*{(24)}
\end{aligned} H x ω Ψ F ΨΨ ⟶ H x EF , ⟶ − ω ψ EF , ⟶ G y EF , H y ω Φ F ΨΦ ⟶ H y EF , ⟶ 0 , ⟶ 0 , P V ΨΦ F ΦΦ ⟶ P EF , ⟶ 0 , ⟶ H x EF H y EF G x EF . (24)
The common angular reflection and canonical normalisation already specified in that appendix then give precisely the singly spinning Black Saturn line element [7 ] .
Independent parameters and scale conventions
After setting a 1 = 0 a_1=0 a 1 = 0 , the construction has five endpoint positions and six polarisation constants. The preceding conditions reduce the count:
Table 2.1.
Therefore, a convenient local choice of the 6 free parameters is
p = ( g 1 , g 2 , g 3 , g 4 , g 5 , C 2 ) , g i = a i + 1 − a i > 0. (25) p = (g_1,g_2,g_3,g_4,g_5,C_2), \qquad g_i = a_{i+1} - a_i > 0. \tag*{(25)} p = ( g 1 , g 2 , g 3 , g 4 , g 5 , C 2 ) , g i = a i + 1 − a i > 0. (25)
The gaps g 2 g_2 g 2 and g 4 g_4 g 4 are the ring and central-hole rod lengths, and g 3 g_3 g 3 is the intervening axis length in Weyl coordinates. The remaining gaps g 1 , g 5 g_1,g_5 g 1 , g 5 locate the auxiliary points and affect the final metric even after those points have been made smooth.
For given p p p , (13) fixes C 1 , C 6 C_1,C_6 C 1 , C 6 , and we solve (16) for ( C 3 , B 4 , B 5 ) (C_3,B_4,B_5) ( C 3 , B 4 , B 5 ) . These three quantities are locally smooth functions of p p p when the derivative of the three balance equations with respect to them is invertible. This condition is verified near the reference configuration in subsection B.4.2. Section 3.5 relates the six parameters p p p to the two horizon masses and four horizon angular momenta. One parameter sets the overall length scale of the entire hole-ring configuration and fixing it leaves five dimensionless parameters specifying shape and rotation.
We introduce two scale conventions. The metric d s A 2 ds_A^2 d s A 2 has the chosen 2 π 2\pi 2 π angular periods. Replacing ( ρ , z ) (\rho,z) ( ρ , z ) by ( ρ ~ , z ~ ) (\tilde{\rho},\tilde{z}) ( ρ ~ , z ~ ) is a coordinate change of the same geometry. For comparison with limiting metrics we also use a constant rescaling of the entire solution. We define
B = A diag ( 1 , δ − 1 / 2 , δ − 1 / 2 ) , G ˉ = B T G B , f ˉ = f / δ , det G ˉ = − ρ 2 , (26) B = A\operatorname{diag}(1,\delta^{-1/2},\delta^{-1/2}), \qquad\bar{G} = B^TGB, \qquad\bar{f} = f/\delta, \qquad\det\bar{G} = -\rho^2, \tag*{(26)} B = A diag ( 1 , δ − 1/2 , δ − 1/2 ) , G ˉ = B T GB , f ˉ = f / δ , det G ˉ = − ρ 2 , (26)
so that it gives d s c a n 2 = δ − 1 d s A 2 ds_{\mathrm{can}}^2 = \delta^{-1}ds_A^2 d s can 2 = δ − 1 d s A 2 after T A = δ T c a n T_A = \sqrt{\delta}\,T_{\mathrm{can}} T A = δ T can , with the angular periods unchanged. We use d s A 2 ds_A^2 d s A 2 for the nonextremal geometry and the reductions to Myers-Perry and Pomeransky-Sen'kov described in appendices A.1 and A.2. We use d s c a n 2 ds_{\mathrm{can}}^2 d s can 2 for the extremal analysis and for the analytic reduction to the singly spinning Saturn of [7 ] in appendix A.3. The corresponding thermodynamic conversion is given after the charges have been defined, in section 3.3.
Geometry, thermodynamics and nonextremal regularity
We now compute the temperatures, entropies and angular velocities of both horizons, the ADM charges at infinity and the individual horizon Komar charges. We explain the resulting Smarr relations and first law, construct coordinates through the horizons, and state sufficient conditions for a regular causal exterior. All formulas in this section use d s A 2 ds_A^2 d s A 2 and the asymptotic normalisation of section 2.2.
Horizon geometry, temperature and entropy
The rods [ a 2 , a 3 ] [a_2,a_3] [ a 2 , a 3 ] and [ a 4 , a 5 ] [a_4,a_5] [ a 4 , a 5 ] represent the ring (h = R h = \mathrm{R} h = R ) and central hole (h = H h = \mathrm{H} h = H ). Their generators are
χ h = ∂ T + Ω h A ∂ φ A , ( φ A ) = ( Ψ , Φ ) , (27) \chi_h = \partial_T + \Omega_h^A\partial_{\varphi^A}, \qquad(\varphi^A) = (\Psi,\Phi), \tag*{(27)} χ h = ∂ T + Ω h A ∂ φ A , ( φ A ) = ( Ψ , Φ ) , (27)
where the repeated angular indices are summed. Both sets of angular velocities refer to the common asymptotic frame fixed in section 2.2.
The generator lies in the kernel of the Killing block, which fixes its angular velocities. We can evaluate the leading rod coefficients (G [ s ] G^{[s]} G [ s ] , f [ s ] f^{[s]} f [ s ] ) defined in (B.37) and (B.39), at an interior point of the horizon interval, with s = 2 s = 2 s = 2 for the ring and s = 4 s = 4 s = 4 for the hole. Their parameter-only expressions are given in subsection B.3. Setting I = ( ψ , ϕ ) I=(\psi,\phi) I = ( ψ , ϕ ) , the kernel and the scalar q h q_h q h are
v h = ( 1 − ( G I I [ s ] ) − 1 G I t [ s ] ) , q h = − 1 f [ s ] det G I I [ s ] , w h = A − 1 v h . (28) v_h=\begin{pmatrix}1\\-(G_{II}^{[s]})^{-1}G_{It}^{[s]}\end{pmatrix},\qquad q_h=-\frac{1}{f^{[s]}\det G_{II}^{[s]}},\qquad w_h=A^{-1}v_h. \tag*{(28)} v h = ( 1 − ( G I I [ s ] ) − 1 G I t [ s ] ) , q h = − f [ s ] det G I I [ s ] 1 , w h = A − 1 v h . (28)
Here s = 2 s=2 s = 2 labels the ring and s = 4 s=4 s = 4 the central hole. The raw kernel v h v_h v h is chosen to have t t t component one. Applying A − 1 A^{-1} A − 1 expresses it in the physical coordinates. Dividing the result by w h T w_h^T w h T gives χ h \chi_h χ h , whose T T T component is one. The invariant q h q_h q h is the horizon analogue of the axis cone factor: it is negative, and its magnitude determines the surface gravity. In corotating coordinates the leading time–radial metric has the Rindler form: − κ h 2 R ⊥ 2 d T 2 + d R ⊥ 2 -\kappa_h^2R_\perp^2\mathrm{d}T^2+\mathrm{d}R_\perp^2 − κ h 2 R ⊥ 2 d T 2 + d R ⊥ 2 , with proper radial distance R ⊥ = f 0 ρ R_\perp=\sqrt{f_0}\rho R ⊥ = f 0 ρ at fixed z z z . Hence G ( χ h , χ h ) = − κ h 2 f 0 ρ 2 + O ( ρ 4 ) \mathcal{G}(\chi_h,\chi_h)=-\kappa_h^2f_0\rho^2+O(\rho^4) G ( χ h , χ h ) = − κ h 2 f 0 ρ 2 + O ( ρ 4 ) . The exact rod data make the surface gravity κ h \kappa_h κ h and angular velocities Ω h A \Omega_h^A Ω h A constant on each component. The Hawking temperature is T h = κ h / ( 2 π ) T_h=\kappa_h/(2\pi) T h = κ h / ( 2 π ) :
κ h = − q h ∣ w h T ∣ , T h = − q h 2 π ∣ w h T ∣ , Ω h A = w h A w h T . (29) \kappa_h=\frac{\sqrt{-q_h}}{\lvert w_h^T\rvert},\qquad T_h=\frac{\sqrt{-q_h}}{2\pi\lvert w_h^T\rvert},\qquad\Omega_h^A=\frac{w_h^A}{w_h^T}. \tag*{(29)} κ h = ∣ w h T ∣ − q h , T h = 2 π ∣ w h T ∣ − q h , Ω h A = w h T w h A . (29)
To obtain the entropy, we set λ A B = G A B \lambda_{AB}=\mathcal{G}_{AB} λ A B = G A B and use a subscript zero for the horizon limit. A spatial horizon section has area
A h = ( 2 π ) 2 ∫ a − a + f 0 det λ 0 d z . (30) A_h=(2\pi)^2\int_{a_-}^{a_+}\sqrt{f_0\det\lambda_0}\,\mathrm{d}z. \tag*{(30)} A h = ( 2 π ) 2 ∫ a − a + f 0 det λ 0 d z . (30)
The determinant condition gives κ h 2 f 0 det λ 0 = δ 2 \kappa_h^2f_0\det\lambda_0=\delta^2 κ h 2 f 0 det λ 0 = δ 2 . Since κ h \kappa_h κ h is constant, the area density is simply δ / κ h \delta/\kappa_h δ / κ h . Writing ℓ R = a 3 − a 2 \ell_R=a_3-a_2 ℓ R = a 3 − a 2 , ℓ H = a 5 − a 4 \ell_H=a_5-a_4 ℓ H = a 5 − a 4 and using S h = A h / ( 4 G 5 ) S_h=A_h/(4G_5) S h = A h / ( 4 G 5 ) , we obtain
A h = 4 π 2 δ ℓ h κ h , S h = π 2 δ ℓ h ∣ w h T ∣ G 5 − q h , T h S h = π δ ℓ h 2 G 5 . (31) A_h=\frac{4\pi^2\delta\ell_h}{\kappa_h},\qquad S_h=\frac{\pi^2\delta\ell_h\lvert w_h^T\rvert}{G_5\sqrt{-q_h}},\qquad T_hS_h=\frac{\pi\delta\ell_h}{2G_5}. \tag*{(31)} A h = κ h 4 π 2 δ ℓ h , S h = G 5 − q h π 2 δ ℓ h ∣ w h T ∣ , T h S h = 2 G 5 π δ ℓ h . (31)
The total entropy is the sum of the two horizon entropies, S t o t = S R + S H S_{\mathrm{tot}}=S_R+S_H S tot = S R + S H . Hence
S t o t = π 2 δ G 5 ( a 3 − a 2 κ R + a 5 − a 4 κ H ) . (32) S_{\mathrm{tot}}=\frac{\pi^2\delta}{G_5}\left(\frac{a_3-a_2}{\kappa_R}+\frac{a_5-a_4}{\kappa_H}\right). \tag*{(32)} S tot = G 5 π 2 δ ( κ R a 3 − a 2 + κ H a 5 − a 4 ) . (32)
These formulas apply to nonextremal rods. At extremality both ℓ h \ell_h ℓ h and κ h \kappa_h κ h vanish and their ratio is replaced by data from the limiting horizon geometry in § 4.5 .
Smooth coordinates across each nonextremal horizon
The divergence of the radial metric in Weyl coordinates at a horizon is a coordinate effect if it can be removed by a nonsingular extension. We construct that extension separately for the ring and central black hole. For either horizon, we suppress its label h h h , set x = ρ 2 x=\rho^2 x = ρ 2 and θ A = φ A − Ω A T \theta^A=\varphi^A-\Omega^AT θ A = φ A − Ω A T , and define the actual metric coefficients
λ A B = G A B , β A = G ( χ , ∂ φ A ) x , α = − G ( χ , χ ) x , a = α − κ 2 f x . (33) \lambda_{AB}=\mathcal{G}_{AB},\qquad\beta_A=\frac{\mathcal{G}(\chi,\partial_{\varphi^A})}{x},\qquad\alpha=-\frac{\mathcal{G}(\chi,\chi)}{x},\qquad a=\frac{\alpha-\kappa^2f}{x}. \tag*{(33)} λ A B = G A B , β A = x G ( χ , ∂ φ A ) , α = − x G ( χ , χ ) , a = x α − κ 2 f . (33)
On the certified branch these are analytic in ( x , z ) (x,z) ( x , z ) along each open horizon. Bounded Laurent expressions and the exact kernel remove the first apparent poles. The horizon identity α ( 0 , z ) = κ 2 f ( 0 , z ) \alpha(0,z)=\kappa^2f(0,z) α ( 0 , z ) = κ 2 f ( 0 , z ) removes the last. The metric becomes
d s 2 = − x α d T 2 + 2 x β A d T d θ A + λ A B d θ A d θ B + f ( d x 2 4 x + d z 2 ) . (34) ds^2=-x\alpha\,dT^2+2x\beta_A\,dT\,d\theta^A+\lambda_{AB}\,d\theta^A d\theta^B+f\left(\frac{dx^2}{4x}+dz^2\right). \tag*{(34)} d s 2 = − xα d T 2 + 2 x β A d T d θ A + λ A B d θ A d θ B + f ( 4 x d x 2 + d z 2 ) . (34)
With an arbitrary constant x ∗ > 0 x_\ast>0 x ∗ > 0 , the future coordinate v = T + ( 2 κ ) − 1 log ( x / x ∗ ) v=T+(2\kappa)^{-1}\log(x/x_\ast) v = T + ( 2 κ ) − 1 log ( x / x ∗ ) gives
d s 2 = − x α d v 2 + α κ d v d x − α 4 κ 2 d x 2 + f d z 2 + 2 x β A d v d θ A − β A κ d x d θ A + λ A B d θ A d θ B , det g = − δ 2 f 2 4 . (35) \begin{aligned}
ds^2={}&-x\alpha\,dv^2+\frac{\alpha}{\kappa}\,dv\,dx-\frac{\alpha}{4\kappa^2}\,dx^2+f\,dz^2+2x\beta_A\,dv\,d\theta^A\\
&-\frac{\beta_A}{\kappa}\,dx\,d\theta^A+\lambda_{AB}\,d\theta^A d\theta^B,\qquad\det g=-\frac{\delta^2f^2}{4}. \tag*{(35)}
\end{aligned} d s 2 = − xα d v 2 + κ α d v d x − 4 κ 2 α d x 2 + f d z 2 + 2 x β A d v d θ A − κ β A d x d θ A + λ A B d θ A d θ B , det g = − 4 δ 2 f 2 . (35)
Thus the extension through x = 0 x=0 x = 0 is analytic and nondegenerate. The outgoing chart reverses the logarithm and the two mixed radial signs. The cancellation is explicit: substituting d T = d v − d x / ( 2 κ x ) dT=dv-dx/(2\kappa x) d T = d v − d x / ( 2 κ x ) combines the singular radial terms into − ( α − κ 2 f ) d x 2 / ( 4 κ 2 x ) -(\alpha-\kappa^2f)dx^2/(4\kappa^2x) − ( α − κ 2 f ) d x 2 / ( 4 κ 2 x ) , which is finite by the horizon identity. When f f f is finite and positive, the determinant is nonzero and the chart extends across x = 0 x=0 x = 0 to negative x x x .
The horizons can be extended separately even when their κ h \kappa_h κ h and Ω h A \Omega_h^A Ω h A differ. Subsection B.4.3 gives the two-sided Kruskal coordinates and joint horizon-axis coordinates at all four poles.
Asymptotic flatness and thermodynamic charges
Here we exhibit the asymptotically Minkowskian coordinates and then read off the total mass and angular momenta. Use the physical Killing coordinates ( T , Ψ , Φ ) (T,\Psi,\Phi) ( T , Ψ , Φ ) of (15) , with both angles of period 2 π 2\pi 2 π and δ = det A > 0 \delta=\det A>0 δ = det A > 0 . We set the irrelevant origin shift z 0 z_0 z 0 to zero and define
ρ = r 2 2 δ sin 2 θ , z = r 2 2 δ cos 2 θ , 0 ≤ θ ≤ π 2 , \rho=\frac{r^2}{2\delta}\sin2\theta,\qquad z=\frac{r^2}{2\delta}\cos2\theta,\qquad0\leq\theta\leq\frac{\pi}{2}, ρ = 2 δ r 2 sin 2 θ , z = 2 δ r 2 cos 2 θ , 0 ≤ θ ≤ 2 π ,
R W = ρ 2 + z 2 = r 2 2 δ , c = z R W = cos 2 θ . (36) R_W=\sqrt{\rho^2+z^2}=\frac{r^2}{2\delta},\qquad c=\frac{z}{R_W}=\cos2\theta. \tag*{(36)} R W = ρ 2 + z 2 = 2 δ r 2 , c = R W z = cos 2 θ . (36)
In the canonical Weyl coordinates ρ ^ = δ ρ \hat{\rho}=\delta\rho ρ ^ = δ ρ and z ^ = δ z \hat{z}=\delta z z ^ = δ z , this is the usual five-dimensional spherical-coordinate transformation [21 ] . The factor δ \delta δ is essential when using the unrescaled metric d s A 2 ds_A^2 d s A 2 .
To see the limit directly, we expand the pole functions as u i = ( 1 − c ) ( R W + a i ) + O ( R W − 1 ) u_i=(1-c)(R_W+a_i)+O(R_W^{-1}) u i = ( 1 − c ) ( R W + a i ) + O ( R W − 1 ) and substitute in (10) –(12) , or equivalently in the scalar expressions. After the constant change of Killing coordinates by A A A , the leading coefficients are
f = δ 2 R W [ 1 + O ( R W − 1 ) ] , G T T = − 1 + O ( R W − 1 ) , G Ψ Ψ = δ R W ( 1 − c ) + O ( 1 ) , G Φ Φ = δ R W ( 1 + c ) + O ( 1 ) , G T Ψ , G T Φ = O ( R W − 1 ) , G Ψ Φ = O ( 1 ) . (37) \begin{aligned}
f&=\frac{\delta}{2R_W}\left[1+O(R_W^{-1})\right],\qquad G_{TT}=-1+O(R_W^{-1}),\\
G_{\Psi\Psi}&=\delta R_W(1-c)+O(1),\qquad G_{\Phi\Phi}=\delta R_W(1+c)+O(1),\\
G_{T\Psi},G_{T\Phi}&=O(R_W^{-1}),\qquad G_{\Psi\Phi}=O(1). \tag*{(37)}
\end{aligned} f G ΨΨ G T Ψ , G T Φ = 2 R W δ [ 1 + O ( R W − 1 ) ] , G T T = − 1 + O ( R W − 1 ) , = δ R W ( 1 − c ) + O ( 1 ) , G ΦΦ = δ R W ( 1 + c ) + O ( 1 ) , = O ( R W − 1 ) , G ΨΦ = O ( 1 ) . (37)
Here the time normalisation and the flat spatial normalisation are both consequences of the metric. In particular, the leading coefficient f 1 f_1 f 1 in f = f 1 / R W + ⋯ f=f_1/R_W+\cdots f = f 1 / R W + ⋯ is angle independent. Normalising each outer axis to period 2 π 2\pi 2 π makes the two leading angular coefficients 2 f 1 R W ( 1 + c ) 2f_1R_W(1+c) 2 f 1 R W ( 1 + c ) . The determinant condition det G = − δ 2 ρ 2 \det G=-\delta^2\rho^2 det G = − δ 2 ρ 2 then gives 2 f 1 = δ 2f_1=\delta 2 f 1 = δ on the branch f 1 > 0 f_1>0 f 1 > 0 . The leading time-angular constants vanish in this same outer-axis basis. We give the algebraic coefficient check in B.3.4.
The orbit-space part transforms exactly according to
d ρ 2 + d z 2 = r 2 δ 2 ( d r 2 + r 2 d θ 2 ) . (38) d\rho^2+dz^2=\frac{r^2}{\delta^2}(dr^2+r^2d\theta^2). \tag*{(38)} d ρ 2 + d z 2 = δ 2 r 2 ( d r 2 + r 2 d θ 2 ) . (38)
Combining this identity with (37) gives
d s A 2 = d s f l a t 2 + h , d s f l a t 2 = − d T 2 + d r 2 + r 2 ( d θ 2 + sin 2 θ d Ψ 2 + cos 2 θ d Φ 2 ) . (39) ds_A^2=ds_{\mathrm{flat}}^2+h,\qquad ds_{\mathrm{flat}}^2=-dT^2+dr^2+r^2\left(d\theta^2+\sin^2\theta\,d\Psi^2+\cos^2\theta\,d\Phi^2\right). \tag*{(39)} d s A 2 = d s flat 2 + h , d s flat 2 = − d T 2 + d r 2 + r 2 ( d θ 2 + sin 2 θ d Ψ 2 + cos 2 θ d Φ 2 ) . (39)
The correction has components h μ ν = O ( r − 2 ) h_{\mu\nu}=O(r^{-2}) h μν = O ( r − 2 ) in the asymptotically Cartesian coordinates
( X 1 , X 2 ) = r sin θ ( cos Ψ , sin Ψ ) , ( X 3 , X 4 ) = r cos θ ( cos Φ , sin Φ ) . (40) (X_1,X_2)=r\sin\theta(\cos\Psi,\sin\Psi),\qquad(X_3,X_4)=r\cos\theta(\cos\Phi,\sin\Phi). \tag*{(40)} ( X 1 , X 2 ) = r sin θ ( cos Ψ , sin Ψ ) , ( X 3 , X 4 ) = r cos θ ( cos Φ , sin Φ ) . (40)
Their first spatial derivatives are O ( r − 3 ) O(r^{-3}) O ( r − 3 ) . These coordinates also remove the polar-coordinate degeneracies at θ = 0 , π / 2 \theta=0,\pi/2 θ = 0 , π /2 and the outer-axis conditions supply the smooth matching there. Together with the independent 2 π 2\pi 2 π identifications, this gives the standard S 3 S^3 S 3 asymptotic metric.
Thermodynamic charges
The leading time components of the metric beyond flat space define the ADM charges:
G T T = − 1 + 8 G 5 M 3 π r 2 + O ( r − 4 ) , G_{TT}=-1+\frac{8G_5M}{3\pi r^2}+O(r^{-4}), G T T = − 1 + 3 π r 2 8 G 5 M + O ( r − 4 ) ,
G T Ψ = − 4 G 5 J Ψ π r 2 sin 2 θ + O ( r − 4 ) , G T Φ = − 4 G 5 J Φ π r 2 cos 2 θ + O ( r − 4 ) . (41) G_{T\Psi}=-\frac{4G_5J_\Psi}{\pi r^2}\sin^2\theta+O(r^{-4}),\qquad G_{T\Phi}=-\frac{4G_5J_\Phi}{\pi r^2}\cos^2\theta+O(r^{-4}). \tag*{(41)} G T Ψ = − π r 2 4 G 5 J Ψ sin 2 θ + O ( r − 4 ) , G T Φ = − π r 2 4 G 5 J Φ cos 2 θ + O ( r − 4 ) . (41)
Equivalently,
M = lim R W → ∞ 3 π δ R W 4 G 5 ( G T T + 1 ) , J Ψ , Φ = − lim R W → ∞ π δ R W G 5 ( 1 ∓ c ) G T Ψ , T Φ . (42) M=\lim_{R_W\to\infty}\frac{3\pi\delta R_W}{4G_5}(G_{TT}+1),\qquad J_{\Psi,\Phi}=-\lim_{R_W\to\infty}\frac{\pi\delta R_W}{G_5(1\mp c)}G_{T\Psi,T\Phi}. \tag*{(42)} M = R W → ∞ lim 4 G 5 3 π δ R W ( G T T + 1 ) , J Ψ , Φ = − R W → ∞ lim G 5 ( 1 ∓ c ) π δ R W G T Ψ , T Φ . (42)
The ADM quantities describe the entire spacetime. To distinguish the ring and central hole, we integrate the same stationary and rotational Killing fields over a section H h \mathcal{H}_h H h of each horizon. With outward orientation, their Komar charges are
M h = − 3 32 π G 5 ∫ H h ⋆ d ξ ♭ , J h A = 1 16 π G 5 ∫ H h ⋆ d m A ♭ , ξ = ∂ T , m A = ∂ φ A . (43) M_h=-\frac{3}{32\pi G_5}\int_{\mathcal{H}_h}\star d\xi^\flat,\qquad J_{hA}=\frac{1}{16\pi G_5}\int_{\mathcal{H}_h}\star dm_A^\flat,\qquad\xi=\partial_T,\quad m_A=\partial_{\varphi^A}. \tag*{(43)} M h = − 32 π G 5 3 ∫ H h ⋆ d ξ ♭ , J h A = 16 π G 5 1 ∫ H h ⋆ d m A ♭ , ξ = ∂ T , m A = ∂ φ A . (43)
These reduce to one-dimensional integrals of the Weyl matrix current:
Q a ( z ) = lim ρ → 0 ρ ( G − 1 ∂ ρ G ) T a , Q_a(z)=\lim_{\rho\to0}\rho(G^{-1}\partial_\rho G)^T{}_a, Q a ( z ) = ρ → 0 lim ρ ( G − 1 ∂ ρ G ) T a ,
M h = 3 π 8 G 5 ∫ a − a + Q T d z , J h A = − π 4 G 5 ∫ a − a + Q A d z . (44) M_h=\frac{3\pi}{8G_5}\int_{a_-}^{a_+}Q_T\,dz,\qquad J_{hA}=-\frac{\pi}{4G_5}\int_{a_-}^{a_+}Q_A\,dz. \tag*{(44)} M h = 8 G 5 3 π ∫ a − a + Q T d z , J h A = − 4 G 5 π ∫ a − a + Q A d z . (44)
The reduction to one dimension integrates out the two periodic angles. More usefully, the vacuum matrix equation (2) makes the current locally integrable: there is a matrix potential P \mathcal{P} P with
∂ z P = ρ G − 1 ∂ ρ G , ∂ ρ P = − ρ G − 1 ∂ z G . (45) \partial_z\mathcal{P}=\rho G^{-1}\partial_\rho G,\qquad\partial_\rho\mathcal{P}=-\rho G^{-1}\partial_zG. \tag*{(45)} ∂ z P = ρ G − 1 ∂ ρ G , ∂ ρ P = − ρ G − 1 ∂ z G . (45)
A horizon integral is consequently an endpoint difference. For this solution P \mathcal{P} P is algebraic in the pole functions, endpoints and polarisation constants, and its explicit expression is (B.30) in subsection B.3.1. Under the change from ( t , ψ , ϕ ) (t,\psi,\phi) ( t , ψ , ϕ ) to the normalised coordinates ( T , Ψ , Φ ) (T,\Psi,\Phi) ( T , Ψ , Φ ) in (14) , it becomes A − 1 P A A^{-1}\mathcal{P}A A − 1 P A . Its T T T row is constant on each smooth spacelike axis, so only three axis values enter the charges.
Let P [ 0 ] \mathcal{P}^{[0]} P [ 0 ] , P [ 3 ] \mathcal{P}^{[3]} P [ 3 ] , P [ 6 ] \mathcal{P}^{[6]} P [ 6 ] denote its algebraic values on the lower, middle and upper axes, respectively. Subsection B.3 gives these values explicitly as ratios of finite determinant coefficients, with no unevaluated axis limit. Define the three-component row
p a [ s ] = ( A − 1 P [ s ] A ) a T , Δ R p = p [ 3 ] − p [ 0 ] , Δ H p = p [ 6 ] − p [ 3 ] , Δ ∞ p = p [ 6 ] − p [ 0 ] . (46) p^{[s]}_{a}=\left(A^{-1}\mathcal{P}^{[s]}A\right)^{T}_{a},\qquad\Delta_{\mathrm{R}}p=p^{[3]}-p^{[0]},\qquad\Delta_{\mathrm{H}}p=p^{[6]}-p^{[3]},\qquad\Delta_{\infty}p=p^{[6]}-p^{[0]}. \tag*{(46)} p a [ s ] = ( A − 1 P [ s ] A ) a T , Δ R p = p [ 3 ] − p [ 0 ] , Δ H p = p [ 6 ] − p [ 3 ] , Δ ∞ p = p [ 6 ] − p [ 0 ] . (46)
All ADM and individual Komar charges are consequently
M h = 3 π δ 8 G 5 Δ H p T , J h A = − π δ 4 G 5 Δ H p A , M = 3 π δ 8 G 5 Δ ∞ p T , J A = − π δ 4 G 5 Δ ∞ p A , A = Ψ , Φ . (47) \begin{aligned}
M_{\mathrm{h}}&=\frac{3\pi\delta}{8G_{5}}\Delta_{\mathrm{H}}p_{T},\qquad J_{\mathrm{h}A}=-\frac{\pi\delta}{4G_{5}}\Delta_{\mathrm{H}}p_{A},\\
M&=\frac{3\pi\delta}{8G_{5}}\Delta_{\infty}p_{T},\qquad J_{A}=-\frac{\pi\delta}{4G_{5}}\Delta_{\infty}p_{A},\qquad A=\Psi,\Phi. \tag*{(47)}
\end{aligned} M h M = 8 G 5 3 π δ Δ H p T , J h A = − 4 G 5 π δ Δ H p A , = 8 G 5 3 π δ Δ ∞ p T , J A = − 4 G 5 π δ Δ ∞ p A , A = Ψ , Φ. (47)
Adding the two differences cancels the common middle-axis value p [ 3 ] p^{[3]} p [ 3 ] , leaving only the difference between the outer axes. This explains additivity directly in the algebraic formulas. It also follows geometrically. In vacuum, d ⋆ d k ♭ = 0 \mathrm{d}\star\mathrm{d}k^{\flat}=0 d ⋆ d k ♭ = 0 for any Killing field k k k . Stokes’ theorem on a spatial exterior bounded by infinity and the two horizons gives
M = M R + M H , J A = J R A + J H A . (48) M=M_{\mathrm{R}}+M_{\mathrm{H}},\qquad J_{A}=J_{\mathrm{R}A}+J_{\mathrm{H}A}. \tag*{(48)} M = M R + M H , J A = J R A + J H A . (48)
This requires a regular exterior with no additional boundary sources. The potential jumps make the same conservation law explicit. We also checked this conservation law by extracting the charges independently at infinity and on the horizons (footnote 1).
Scale dependence of the thermodynamic quantities. The change from d s A 2 \mathrm{d}s^{2}_{A} d s A 2 to d s c a n 2 \mathrm{d}s^{2}_{\mathrm{can}} d s can 2 in (26) gives
M c a n = M A δ , ( J c a n , A c a n ) = ( J A , A A ) δ 3 / 2 , ( κ c a n , Ω c a n ) = δ ( κ A , Ω A ) . (49) M_{\mathrm{can}}=\frac{M_{A}}{\delta},\qquad(J_{\mathrm{can}},\mathcal{A}_{\mathrm{can}})=\frac{(J_{A},\mathcal{A}_{A})}{\delta^{3/2}},\qquad(\kappa_{\mathrm{can}},\Omega_{\mathrm{can}})=\sqrt{\delta}(\kappa_{A},\Omega_{A}). \tag*{(49)} M can = δ M A , ( J can , A can ) = δ 3/2 ( J A , A A ) , ( κ can , Ω can ) = δ ( κ A , Ω A ) . (49)
This operation changes the scale of all proper lengths by L = δ − 1 / 2 L=\delta^{-1/2} L = δ − 1/2 . More generally, multiplying a vacuum line element by L 2 L^{2} L 2 , and rescaling time so that its norm at infinity remains − 1 -1 − 1 , produces a geometrically similar configuration. At fixed G 5 G_{5} G 5 , its mass is L 2 M L^{2}M L 2 M , its spins and horizon areas are L 3 J A L^{3}J_{A} L 3 J A and L 3 A h L^{3}\mathcal{A}_{h} L 3 A h , and its temperatures and angular velocities are T h / L T_{h}/L T h / L and Ω h A / L \Omega^{A}_{h}/L Ω h A / L . Setting L = δ − 1 / 2 L=\delta^{-1/2} L = δ − 1/2 gives (49) and the entropy changes by the same factor as area. The same factors apply separately to each horizon. The dimensionless spins J A / M 3 / 2 J_{A}/M^{3/2} J A / M 3/2 and areas A h / M 3 / 2 \mathcal{A}_{h}/M^{3/2} A h / M 3/2 , with the fixed G 5 G_{5} G 5 factors in (79) , are unchanged by this conversion.
Smarr relations, the first law and independent charges
The individual Komar charges satisfy
2 3 M h = T h S h + Ω h A J h A . (50) \frac{2}{3}M_{h}=T_{h}S_{h}+\Omega^{A}_{h}J_{hA}. \tag*{(50)} 3 2 M h = T h S h + Ω h A J h A . (50)
To derive this relation, insert ξ = χ h − Ω h A m A \xi=\chi_h-\Omega_h^{A}m_A ξ = χ h − Ω h A m A into the mass integral (43) . With our orientation, ∫ H h ∗ d χ h b = − 2 κ h A h \int_{\mathcal{H}_h} *\mathrm{d}\chi_h^b=-2\kappa_h A_h ∫ H h ∗ d χ h b = − 2 κ h A h , while the remaining terms are Ω h A \Omega_h^{A} Ω h A times the angular-momentum integrals. The coefficients in (43) then give M h = 3 κ h A h / ( 16 π G 5 ) + ( 3 / 2 ) Ω h A J h A M_h=3\kappa_h A_h/(16\pi G_5)+(3/2)\Omega_h^{A}J_{hA} M h = 3 κ h A h / ( 16 π G 5 ) + ( 3/2 ) Ω h A J h A , which is (50) . This derives the identity without defining the mass from the Smarr relation. Summing the two identities and using (48) gives the total Smarr relation.
For variations between balanced stationary solutions with the same asymptotic normalisation, the boundary-term derivation of black hole mechanics [12 , 28 ] gives
d M = ∑ h = R , H ( T h d S h + Ω h A d J h A ) . (51) \mathrm{d}M=\sum_{h=\mathrm{R},\mathrm{H}}\left(T_h\mathrm{d}S_h+\Omega_h^{A}\mathrm{d}J_{hA}\right). \tag*{(51)} d M = h = R , H ∑ ( T h d S h + Ω h A d J h A ) . (51)
Each horizon contributes its own boundary term, with its own intensive variables. This is the two-spin form of disconnected horizon mechanics in [12 ] . The varied parameters must continue to satisfy the axis conditions (16) otherwise a conical defect may appear and its stress would contribute an additional work term. As a check, a uniform change of length gives d M = 2 M d log L \mathrm{d}M=2M\mathrm{d}\log L d M = 2 M d log L and ( d S h , d J h A ) = 3 ( S h , J h A ) d log L (\mathrm{d}S_h,\mathrm{d}J_{hA})=3(S_h,J_{hA})\mathrm{d}\log L ( d S h , d J h A ) = 3 ( S h , J h A ) d log L . Equation (51) then reproduces the summed Smarr relation.
The algebraic Smarr identity for a component is not an isolated hole differential law for its Komar mass. Changing a Saturn parameter changes the interaction as well as the horizons. In general, d M h − T h d S h − Ω h A d J h A \mathrm{d}M_h-T_h\mathrm{d}S_h-\Omega_h^{A}\mathrm{d}J_{hA} d M h − T h d S h − Ω h A d J h A is nonzero and only the sum vanishes. Mutual thermal equilibrium imposes the further requirements T R = T H T_{\mathrm{R}}=T_{\mathrm{H}} T R = T H and Ω R A = Ω H A \Omega_{\mathrm{R}}^{A}=\Omega_{\mathrm{H}}^{A} Ω R A = Ω H A , studied in section 5.2.
The physical content of the parameter count can be tested with the six horizon quantities
Q h = ( M R , J R Ψ , J R Φ , M H , J H Ψ , J H Φ ) . (52) Q_h=(M_{\mathrm{R}},J_{\mathrm{R}\Psi},J_{\mathrm{R}\Phi},M_{\mathrm{H}},J_{\mathrm{H}\Psi},J_{\mathrm{H}\Phi}). \tag*{(52)} Q h = ( M R , J R Ψ , J R Φ , M H , J H Ψ , J H Φ ) . (52)
At the reference solution, we compute the 6 × 6 6\times6 6 × 6 derivative matrix of these quantities with respect to p p p in (25) . The computed matrix is invertible and remains so under refinement of the derivatives. Thus, numerically, small changes of the six parameters can independently change the six horizon charges. Only their three sums are measured at infinity. We checked these derivatives under step refinement, together with the first law (footnote 1).
Axes, causality and the tested parameter domain
Three kinds of points require separate treatment: the open exterior, the rotation axes, and the horizons including their poles. In the open exterior the metric coefficients must be smooth and nondegenerate. On an axis the shrinking angular circle must close with circumference 2 π 2\pi 2 π times its proper radius. At a horizon the coordinates of section 3.2 must give a smooth extension. At the auxiliary points a 1 , a 6 a_1,a_6 a 1 , a 6 , the square roots introduced by the construction cancel by (13) and the remaining denominators must also stay nonzero so that this cancellation leaves finite metric components.
For example, on a Φ \Phi Φ axis the no-conical-defect condition is g Φ Φ = ρ 2 f 0 + O ( ρ 4 ) g_{\Phi\Phi}=\rho^2 f_0+O(\rho^4) g ΦΦ = ρ 2 f 0 + O ( ρ 4 ) . The part of the metric perpendicular to the axis is then f 0 ( d ρ 2 + ρ 2 d Φ 2 ) + ⋯ f_0(\mathrm{d}\rho^2+\rho^2\mathrm{d}\Phi^2)+\cdots f 0 ( d ρ 2 + ρ 2 d Φ 2 ) + ⋯ , which is smooth in ( ρ cos Φ , ρ sin Φ ) (\rho\cos\Phi,\rho\sin\Phi) ( ρ cos Φ , ρ sin Φ ) for f 0 > 0 f_0>0 f 0 > 0 and a 2 π 2\pi 2 π -periodic Φ \Phi Φ . The axis matching of section 2.2 ensures that this angle runs once around the shrinking circle. Where a horizon meets an axis, subsection B.4.3 uses coordinates regular in both directions at once.
To exclude closed timelike curves, it is insufficient to check the Ψ \Psi Ψ and Φ \Phi Φ circles separately: a linear combination could still be timelike. Instead, we consider their full 2 × 2 2 \times2 2 × 2 metric Λ = ( G A B ) A , B = Ψ , Φ \Lambda= (G_{AB})_{A,B=\Psi,\Phi} Λ = ( G A B ) A , B = Ψ , Φ . At every finite off-axis exterior point, sufficient conditions are finite smooth coefficients and
f > 0 , G Ψ Ψ > 0 , det Λ > 0. (53) f > 0,\qquad G_{\Psi\Psi} > 0,\qquad\det\Lambda> 0. \tag*{(53)} f > 0 , G ΨΨ > 0 , det Λ > 0. (53)
The last two inequalities make the angular block positive definite, so no angular combination is timelike. The base is positive and the Killing determinant negative, giving Lorentzian signature. More strongly, the cofactor formula for the inverse metric gives
g T T = − det Λ δ 2 ρ 2 < 0. (54) g^{TT} = -\frac{\det\Lambda}{\delta^2\rho^2} < 0. \tag*{(54)} g T T = − δ 2 ρ 2 det Λ < 0. (54)
The angular identifications leave T T T single-valued, and (54) says that its gradient is timelike. Thus T T T increases strictly along every future causal curve, which rules out closed causal curves and establishes stable causality. An ergoregion has g T T > 0 g_{TT} > 0 g T T > 0 , but this concerns the stationary vector rather than the gradient of T T T and does not contradict g T T < 0 g^{TT} < 0 g T T < 0 . At an axis a circle has zero size, so the strict angular inequalities are replaced by the smooth Cartesian-axis conditions above. This combination of metric extension and a global time function follows the strategy used for singly spinning Saturns [29 , 30 ] .
Subsection B.4 writes (53) as explicit algebraic inequalities for the six-parameter Saturn metric. We check that they are obeyed at least in a certain region of parameter space, together with regularity of all finite boundary components. A broader numerical scan varies every parameter by up to 30% around the reference configuration and finds no regularity or causality violation. These results support a regular nonextremal regime but a complete proof of causality and regularity in the entire exterior is beyond the scope of this work. These checks are given in appendices B.4.2 and B.4.6. For the double-extremal examples in the next section, the bounds do cover the entire exterior.
Two extremal horizons
The black Saturn metric admits a limit in which both horizons are extremal. We construct the limiting metric, give coordinates that extend through its horizons and compute its thermodynamic quantities. Regularity and causality are established for the parameter interval and examples specified in subsection B.4.4. Appendix § 8 .§ 4 treats the two other possibilities: an extremal central hole surrounded by a nonextremal ring, and an extremal ring surrounding a nonextremal hole.
A correlated degeneration of the metric
For a nonextremal horizon, (31) relates its Weyl rod length to κ h A h \kappa_h A_h κ h A h . A finite-area extremal horizon therefore becomes a point in the orbit space. The geometry near that point can nevertheless contain an entire compact horizon, at the end of an infinite spatial throat. To retain this geometry we must scale the rod lengths, polarisations and normalisation together.
We fix λ > 0 \lambda> 0 λ > 0 and introduce finite constants ( Z , c , d , h , v ∗ , w , r 0 , β ) (Z,c,d,h,v_{*},w,r_{0},\beta) ( Z , c , d , h , v ∗ , w , r 0 , β ) , with Z , d , h , r 0 > 0 Z,d,h,r_{0} > 0 Z , d , h , r 0 > 0 and β ≠ 0 \beta\ne0 β = 0 . It is useful to define
s = r 0 4 + λ 2 , r + = s + λ , z c = Z − 1 2 ( s − r 0 2 ) , a 2 = c λ , a 3 = c λ + d λ 2 , a 4 = z c − λ / 2 , a 5 = z c + λ / 2 , a 6 = a 5 + h / λ , a 1 = 0. (55) \begin{aligned}
s &= \sqrt{r_{0}^{4}+\lambda^{2}}, & r_{+} &= \sqrt{s+\lambda}, & z_{c} &= Z-\frac{1}{2}(s-r_{0}^{2}),\\
a_{2} &= c\lambda, & a_{3} &= c\lambda+d\lambda^{2}, & a_{4} &= z_{c}-\lambda/2,\\
a_{5} &= z_{c}+\lambda/2, & a_{6} &= a_{5}+h/\lambda, & a_{1} &= 0. \tag*{(55)}
\end{aligned} s a 2 a 5 = r 0 4 + λ 2 , = c λ , = z c + λ /2 , r + a 3 a 6 = s + λ , = c λ + d λ 2 , = a 5 + h / λ , z c a 4 a 1 = Z − 2 1 ( s − r 0 2 ) , = z c − λ /2 , = 0. (55)
As λ → 0 + \lambda\to0^{+} λ → 0 + , the ring and hole rod lengths vanish as d λ 2 d\lambda^{2} d λ 2 and λ \lambda λ , while their positions approach 0 and Z Z Z . Thus Z Z Z is the limiting separation in Weyl coordinates and d d d controls the relative collapse rate. The auxiliary endpoint a 6 a_{6} a 6 moves to infinity as h / λ h/\lambda h / λ . The remaining constants specify how the rotation parameters approach this boundary of the family. We keep the auxiliary-point conditions (13) and set
α = − r 0 β r + , b 0 = β r 0 r + , B 4 = 2 λ − 2 C 6 λ b 0 C 6 2 λ + b 0 , B 5 = 2 λ ( a 6 − a 5 ) C 6 α a 5 + 1 , η = 1 − C 1 C 2 C 3 = c ( 1 + Z − 1 ) λ + w λ 2 , C 2 = 1 − η C 1 C 3 . (56) \begin{aligned}
\alpha&= -\frac{r_{0}}{\beta r_{+}}, & b_{0} &= \frac{\beta r_{0}}{r_{+}},\\
B_{4} &= \frac{\sqrt{2\lambda}-2C_{6}\lambda b_{0}}{C_{6}\sqrt{2\lambda}+b_{0}}, & B_{5} &= \frac{\sqrt{2\lambda}(a_{6}-a_{5})C_{6}\alpha}{a_{5}+1},\\
\eta&= 1-C_{1}C_{2}C_{3}=c(1+Z^{-1})\lambda+w\lambda^{2}, & C_{2} &= \frac{1-\eta}{C_{1}C_{3}}. \tag*{(56)}
\end{aligned} α B 4 η = − β r + r 0 , = C 6 2 λ + b 0 2 λ − 2 C 6 λ b 0 , = 1 − C 1 C 2 C 3 = c ( 1 + Z − 1 ) λ + w λ 2 , b 0 B 5 C 2 = r + β r 0 , = a 5 + 1 2 λ ( a 6 − a 5 ) C 6 α , = C 1 C 3 1 − η . (56)
It is convenient to write the limiting parameter as a square, so that the expressions can be expanded in integer powers of τ \tau τ :
λ = τ 2 / 2 , h = H ∗ 2 / 4 , C 3 = v ∗ τ . (57) \lambda=\tau^{2}/2,\qquad h=H_{*}^{2}/4,\qquad C_{3}=v_{*}\tau. \tag*{(57)} λ = τ 2 /2 , h = H ∗ 2 /4 , C 3 = v ∗ τ . (57)
The + 1 +1 + 1 in (56) comes from the choice b = − 1 b=-1 b = − 1 in P b P_b P b of section 2. Equations (55) –(57) specify how both the rod endpoints and the Belinski–Zakharov polarisation constants ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) (C_{1},C_{2},C_{3},B_{4},B_{5},C_{6}) ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) vary as λ \lambda λ tends to zero. The latter constants are the coefficients in the dressing vectors of section 2 and they control the mixing of time and the two rotational Killing directions. Thus the prescription fixes the rotation data as well as the collapsing horizon rods.
The relation for η \eta η cancels the leading pole in the dressed Killing block. This is why the polarisation constants must be tuned together with the rod lengths. The algebraic cancellation is displayed in subsection B.2.
The original Killing coordinates also become poorly normalised in this limit. Before extracting the finite metric we apply a τ \tau τ -dependent linear change of those coordinates, with matrix
u b = 1 2 ( τ − 1 + τ ) , v b = 1 2 ( τ − 1 − τ ) , C = ( 1 − v b / C 6 − u b / C 6 0 u b v b 0 v b u b ) , det C = 1. (58) u_b=\frac{1}{2}(\tau^{-1}+\tau),\qquad v_b=\frac{1}{2}(\tau^{-1}-\tau),\qquad C=\begin{pmatrix}1-v_b/C_6 & -u_b/C_6\\ 0 & u_b & v_b\\ 0 & v_b & u_b\end{pmatrix},\qquad\det C=1. \tag*{(58)} u b = 2 1 ( τ − 1 + τ ) , v b = 2 1 ( τ − 1 − τ ) , C = 1 − v b / C 6 0 0 − u b / C 6 u b v b v b u b , det C = 1. (58)
We substitute (55) –(57) in the metric, apply this matrix, and expand in τ \tau τ . The notation [ τ 0 ] [\tau^{0}] [ τ 0 ] means the coefficient independent of τ \tau τ after cancelling common powers. The limiting metric is
G e = [ τ 0 ] ( C T G C ) , f e = [ τ 0 ] ( f / λ 3 ) . (59) G_{\mathrm e}=[\tau^{0}](C^{\mathrm T}GC),\qquad f_{\mathrm e}=[\tau^{0}](f/\lambda^{3}). \tag*{(59)} G e = [ τ 0 ] ( C T GC ) , f e = [ τ 0 ] ( f / λ 3 ) . (59)
The multiplication of f f f by a parameter-dependent constant is an allowed integration-constant choice before normalisation. We then recompute the outer kernels and cone factors of ( G ‾ e , f ‾ e ) (\overline{G}_e,\overline{f}_e) ( G e , f e ) , and apply the canonical normalisation (26) . We denote the resulting metric ( G ‾ e , f ‾ e ) (\overline{G}_e,\overline{f}_e) ( G e , f e ) and have det G ‾ e = − ρ 2 \det\overline{G}_e=-\rho^2 det G e = − ρ 2 . Equations (9) –(12) and (55) –(59) are a finite algebraic specification of the exact extremal spacetime.
Balance is imposed on this limiting metric by the same axis conditions as in § 2.2 . The auxiliary path used to calculate its coefficients need not itself be balanced at every positive λ \lambda λ . Subsection B.2 evaluates (59) algebraically: it gives the finite scalar polynomials whose ratios are the metric components at λ = 0 \lambda=0 λ = 0 . This provides a formula that can be used directly.
For the regularity analysis it is useful to replace the unbounded half-plane ( ρ , z ) (\rho,z) ( ρ , z ) by a square. Define the distances to the two horizon points and the coordinates
u = R R + R H − Z R R + R H + Z , y = R R − R H + Z 2 Z , R R = ρ 2 + z 2 , R H = ρ 2 + ( z − Z ) 2 . (60) u=\frac{R_{\mathrm{R}}+R_{\mathrm{H}}-Z}{R_{\mathrm{R}}+R_{\mathrm{H}}+Z},\qquad y=\frac{R_{\mathrm{R}}-R_{\mathrm{H}}+Z}{2Z},\qquad R_{\mathrm{R}}=\sqrt{\rho^2+z^2},\quad R_{\mathrm{H}}=\sqrt{\rho^2+(z-Z)^2}. \tag*{(60)} u = R R + R H + Z R R + R H − Z , y = 2 Z R R − R H + Z , R R = ρ 2 + z 2 , R H = ρ 2 + ( z − Z ) 2 . (60)
The finite off-axis exterior is 0 < u , y < 1 0<u,y<1 0 < u , y < 1 . The three axes form the edges u = 0 u=0 u = 0 and y = 0 , 1 y=0,1 y = 0 , 1 while the horizons are the corners ( 0 , 0 ) (0,0) ( 0 , 0 ) and ( 0 , 1 ) (0,1) ( 0 , 1 ) , and spatial infinity is u = 1 u=1 u = 1 . On the subfamily studied below the metric is rational in u , y u,y u , y , allowing uniform polynomial sign bounds over the whole exterior.
Explicit scalar form at zero temperature
The correlated limit also has a scalar-function presentation. In the Killing coordinates obtained after the transformation C C C , and before normalising the angular periods, define seven functions K , Y , W ψ , W ϕ , V ψ ψ , V ψ ϕ , V ϕ ϕ K,Y,W_\psi,W_\phi,V_{\psi\psi},V_{\psi\phi},V_{\phi\phi} K , Y , W ψ , W ϕ , V ψ ψ , V ψ ϕ , V ϕϕ by the finite polynomial formulas of subsection B.2. Then
d s e 2 = 1 K [ − Y d t 2 − 2 W ψ d t d ψ − 2 W ϕ d t d ϕ + V ψ ψ d ψ 2 + 2 V ψ ϕ d ψ d ϕ + V ϕ ϕ d ϕ 2 ] + K ( d ρ 2 + d z 2 ) . (61) ds_e^2=\frac{1}{K}\left[-Y\,dt^2-2W_\psi\,dt\,d\psi-2W_\phi\,dt\,d\phi+V_{\psi\psi}\,d\psi^2+2V_{\psi\phi}\,d\psi\,d\phi+V_{\phi\phi}\,d\phi^2\right]+K(d\rho^2+dz^2). \tag*{(61)} d s e 2 = K 1 [ − Y d t 2 − 2 W ψ d t d ψ − 2 W ϕ d t d ϕ + V ψ ψ d ψ 2 + 2 V ψ ϕ d ψ d ϕ + V ϕϕ d ϕ 2 ] + K ( d ρ 2 + d z 2 ) . (61)
Here the scalar convention sets the analogue of P P P to one: K = f e K=f_e K = f e , Y = − f e ( G e ) t t Y=-f_e(G_e)_{tt} Y = − f e ( G e ) tt , W i = − f e ( G e ) t i W_i=-f_e(G_e)_{ti} W i = − f e ( G e ) t i and V i j = f e ( G e ) i j V_{ij}=f_e(G_e)_{ij} V ij = f e ( G e ) ij . These identities identify the functions and their explicit evaluation does not require constructing G e G_e G e . With i , j = ψ , ϕ i,j=\psi,\phi i , j = ψ , ϕ , put
J i j = Y V i j + W i W j K 2 . (62) J_{ij}=\frac{YV_{ij}+W_iW_j}{K^2}. \tag*{(62)} J ij = K 2 Y V ij + W i W j . (62)
The doubly spinning extremal metric can thus be written as
d s e 2 = − Y K ( d t + W ψ Y d ψ + W ϕ Y d ϕ ) 2 + K [ d ρ 2 + d z 2 + J ψ ψ d ψ 2 + 2 J ψ ϕ d ψ d ϕ + J ϕ ϕ d ϕ 2 Y ] . (63) ds_e^2=-\frac{Y}{K}\left(dt+\frac{W_\psi}{Y}d\psi+\frac{W_\phi}{Y}d\phi\right)^2+K\left[d\rho^2+dz^2+\frac{J_{\psi\psi}d\psi^2+2J_{\psi\phi}d\psi\,d\phi+J_{\phi\phi}d\phi^2}{Y}\right]. \tag*{(63)} d s e 2 = − K Y ( d t + Y W ψ d ψ + Y W ϕ d ϕ ) 2 + K [ d ρ 2 + d z 2 + Y J ψ ψ d ψ 2 + 2 J ψ ϕ d ψ d ϕ + J ϕϕ d ϕ 2 ] . (63)
As in the nonextremal case, Y = 0 Y=0 Y = 0 is an ergosurface rather than a singularity caused by completing the square. At such a surface use (61) , whose coefficients do not divide by Y Y Y . In particular, J ψ ψ J ϕ ϕ − J ψ ϕ 2 = ρ 2 Y / K J_{\psi\psi}J_{\phi\phi}-J_{\psi\phi}^2=\rho^2Y/K J ψ ψ J ϕϕ − J ψ ϕ 2 = ρ 2 Y / K .
For the physical canonical metric, let B e B_e B e be the constant basis obtained by the outer-axis normalisation and canonical rescaling specified after (59) . Write d t = d T + q A d φ A dt=dT+q_A d\varphi^A d t = d T + q A d φ A , d x i = b A i d φ A dx^i=b^i_A d\varphi^A d x i = b A i d φ A , where ( φ 1 , φ 2 ) = ( Ψ , Φ ) (\varphi^1,\varphi^2)=(\Psi,\Phi) ( φ 1 , φ 2 ) = ( Ψ , Φ ) have period 2 π 2\pi 2 π , and let δ e \delta_e δ e be the determinant of the outer-axis basis before canonical rescaling. The physical scalar functions are explicitly
K ˉ = K δ e , Y ˉ = Y δ e , W ˉ A = Y q A + b A i W i δ e , \bar{K}=\frac{K}{\delta_e},\qquad\bar{Y}=\frac{Y}{\delta_e},\qquad\bar{W}_A=\frac{Yq_A+b^i_AW_i}{\delta_e}, K ˉ = δ e K , Y ˉ = δ e Y , W ˉ A = δ e Y q A + b A i W i ,
V ˉ A B = b A i b B j V i j − q A b B i W i − q B b A i W i − Y q A q B δ e . (64) \bar{V}_{AB}=\frac{b^i_A b^j_B V_{ij}-q_A b^i_B W_i-q_B b^i_A W_i-Yq_Aq_B}{\delta_e}. \tag*{(64)} V ˉ A B = δ e b A i b B j V ij − q A b B i W i − q B b A i W i − Y q A q B . (64)
Replacing all functions in (61) –(63) by their barred versions gives ( G ˉ e , f ˉ e ) (\bar{G}_e,\bar{f}_e) ( G ˉ e , f ˉ e ) .
Subsection B.2 retains every parameter of the correlated limit before balance is imposed. It expresses the seven functions by finite polynomial arithmetic, so the metric can be evaluated directly at zero temperature. The ancillary Mathematica file implements these components. Subsection B.4.4 specifies a balanced parameter interval and further examples of regular doubly extremal black Saturns, together with the bounds used to establish their regularity.
Coordinates through the degenerate horizons
Although both horizon rods have collapsed, their neighbourhoods can be resolved by polar coordinates about the points z H = 0 , Z z_H=0,Z z H = 0 , Z . We first work with R > 0 R>0 R > 0 in the exterior, then continue R R R through zero to obtain an interior coordinate:
ρ = R sin θ , z = z H + R cos θ . (65) \rho=R\sin\theta,\qquad z=z_H+R\cos\theta. \tag*{(65)} ρ = R sin θ , z = z H + R cos θ . (65)
The choice of square-root branch is essential. The root centred at the chosen horizon must continue as the signed coordinate R R R . Replacing it by ∣ R ∣ \lvert R\rvert ∣ R ∣ would create a cusp at R = 0 R=0 R = 0 . At the ring the two roots in (60) are ( R , Z 2 − 2 Z R cos θ + R 2 ) \left(R,\sqrt{Z^2-2ZR\cos\theta+R^2}\right) ( R , Z 2 − 2 Z R cos θ + R 2 ) , and at the hole they are ( Z 2 + 2 Z R cos θ + R 2 , R ) \left(\sqrt{Z^2+2ZR\cos\theta+R^2},R\right) ( Z 2 + 2 Z R cos θ + R 2 , R ) . The other root is analytic for ∣ R ∣ < Z \lvert R\rvert<Z ∣ R ∣ < Z .
In canonical coordinates set
F = R 2 f ˉ e , a = sin 2 θ det Λ , β A = ( Λ − 1 ) A B ( G ˉ e ) T B , φ A = Φ A − Ω H A T , k A = β A + Ω H A R . (66) F=R^2\bar{f}_e,\qquad a=\frac{\sin^2\theta}{\det\Lambda},\qquad\beta^A=(\Lambda^{-1})^{AB}(\bar{G}_e)_{TB},\qquad\varphi^A=\Phi^A-\Omega^A_H T,\qquad k^A=\frac{\beta^A+\Omega^A_H}{R}. \tag*{(66)} F = R 2 f ˉ e , a = det Λ sin 2 θ , β A = ( Λ − 1 ) A B ( G ˉ e ) T B , φ A = Φ A − Ω H A T , k A = R β A + Ω H A . (66)
Here Φ A = ( Ψ , Φ ) \Phi^A=(\Psi,\Phi) Φ A = ( Ψ , Φ ) and Λ \Lambda Λ is the canonical angular block. Completing the angular square yields
d s 2 = − R 2 a d T 2 + F ( d R 2 R 2 + d θ 2 ) + Λ A B ( d φ A + R k A d T ) ( d φ B + R k B d T ) . (67) ds^2=-R^2a\,dT^2+F\left(\frac{dR^2}{R^2}+d\theta^2\right)+\Lambda_{AB}(d\varphi^A+Rk^A dT)(d\varphi^B+Rk^B dT). \tag*{(67)} d s 2 = − R 2 a d T 2 + F ( R 2 d R 2 + d θ 2 ) + Λ A B ( d φ A + R k A d T ) ( d φ B + R k B d T ) . (67)
Write each coefficient as a Taylor expansion in signed R R R , for example F = F 0 + R F 1 + O ( R 2 ) F=F_0+RF_1+O(R^2) F = F 0 + R F 1 + O ( R 2 ) . Direct expansion of the metric shows that Ω H A \Omega^A_H Ω H A , k 0 A k^A_0 k 0 A and
C 2 = a 0 F 0 > 0 , T 1 = a 1 / C 2 − F 1 2 C F 0 (68) C^2=\frac{a_0}{F_0}>0,\qquad T_1=\frac{a_1/C^2-F_1}{2CF_0} \tag*{(68)} C 2 = F 0 a 0 > 0 , T 1 = 2 C F 0 a 1 / C 2 − F 1 (68)
are independent of θ \theta θ on each horizon, with C > 0 C>0 C > 0 . These constants determine the coordinate shifts needed to remove the 1 / R 2 1/R^2 1/ R 2 and 1 / R 1/R 1/ R radial poles. Since they do not depend on angle, the shifts define actual coordinates:
T = v + 1 C R + T 1 log ∣ R ∣ , φ A = χ A + k 0 A C log ∣ R ∣ . (69) T=v+\frac{1}{CR}+T_1\log\lvert R\rvert,\qquad\varphi^A=\chi^A+\frac{k^A_0}{C}\log\lvert R\rvert. \tag*{(69)} T = v + C R 1 + T 1 log ∣ R ∣ , φ A = χ A + C k 0 A log ∣ R ∣ . (69)
This pole-cancellation construction is related to the extremal extensions in [33 , 34 ] . With
U R = − 1 C R 2 + T 1 R , Q A = k 0 A − k A C R + T 1 k A , (70) U_R=-\frac{1}{CR^2}+\frac{T_1}{R},\qquad Q^A=\frac{k_0^A-k^A}{CR}+T_1k^A, \tag*{(70)} U R = − C R 2 1 + R T 1 , Q A = C R k 0 A − k A + T 1 k A , (70)
the transformed metric is
d s 2 = − R 2 a ( d v + U R d R ) 2 + F R 2 d R 2 + F d θ 2 + Λ A B ( d χ A + R k A d v + Q A d R ) ( d χ B + R k B d v + Q B d R ) . (71) ds^2=-R^2a(\mathrm{d}v+U_R\mathrm{d}R)^2+\frac{F}{R^2}\mathrm{d}R^2+F\mathrm{d}\theta^2
+\Lambda_{AB}(\mathrm{d}\chi^A+Rk^A\mathrm{d}v+Q^A\mathrm{d}R)(\mathrm{d}\chi^B+Rk^B\mathrm{d}v+Q^B\mathrm{d}R). \tag*{(71)} d s 2 = − R 2 a ( d v + U R d R ) 2 + R 2 F d R 2 + F d θ 2 + Λ A B ( d χ A + R k A d v + Q A d R ) ( d χ B + R k B d v + Q B d R ) . (71)
Since k − k 0 = O ( R ) k-k_0=O(R) k − k 0 = O ( R ) , Q Q Q is analytic. The apparently singular radial coefficient reduces to
F − a / C 2 R 2 + 2 a T 1 C R − a T 1 2 , (72) \frac{F-a/C^2}{R^2}+\frac{2aT_1}{CR}-aT_1^2, \tag*{(72)} R 2 F − a / C 2 + C R 2 a T 1 − a T 1 2 , (72)
whose R − 2 R^{-2} R − 2 and R − 1 R^{-1} R − 1 terms vanish by (68) . The resulting analytic metric obeys
g v R ∣ R = 0 = C F 0 ≠ 0 , det g ( v , R , θ , χ 1 , χ 2 ) = − F 2 sin 2 θ . (73) g_{vR}|_{R=0}=CF_0\ne0,\qquad\det g_{(v,R,\theta,\chi^1,\chi^2)}=-F^2\sin^2\theta. \tag*{(73)} g v R ∣ R = 0 = C F 0 = 0 , det g ( v , R , θ , χ 1 , χ 2 ) = − F 2 sin 2 θ . (73)
Thus R = 0 R=0 R = 0 is a regular null hypersurface in ( v , R , θ , χ A ) (v,R,\theta,\chi^A) ( v , R , θ , χ A ) . Reversing the radial shifts gives an outgoing chart. The generator ∂ v \partial_v ∂ v has vanishing surface gravity, as required for an extremal horizon.
The coordinates above still have the usual polar-coordinate singularity at θ = 0 , π \theta=0,\pi θ = 0 , π . To include the poles of the compact horizon, replace the shrinking angle by Cartesian coordinates. Put x p = 2 ( 1 − ∣ cos θ ∣ ) x_p=\sqrt{2(1-|\cos\theta|)} x p = 2 ( 1 − ∣ cos θ ∣ ) , X = x p cos χ d e g X=x_p\cos\chi_{\mathrm{deg}} X = x p cos χ deg , Y = x p sin χ d e g Y=x_p\sin\chi_{\mathrm{deg}} Y = x p sin χ deg . The shrinking cycle is Φ \Phi Φ at both ring poles and the lower hole pole, and Ψ \Psi Ψ at the upper hole pole. Exact kernel and unit-cone identities, together with joint analyticity in ( R , cos θ ) (R,\cos\theta) ( R , cos θ ) , remove the polar singularities even at R = 0 R=0 R = 0 . The Cartesian determinant is − F 2 -F^2 − F 2 , and the remaining circle has positive norm. The primitive angular lattice excludes an orbifold quotient.
Global exterior and interpretation of the horizons
For the balanced parameter interval of proposition B.1 and the seven roots in Table 3 (subsection B.4.4), the denominators of the exact rational metric stay nonzero throughout 0 < u , y < 1 0<u,y<1 0 < u , y < 1 , and the canonical versions of (53) hold there. After removing the known factors associated with a shrinking circle, the inequalities also hold on each open axis. The charts above cover the two focal corners. At infinity we use (36) with δ = 1 \delta=1 δ = 1 , because this metric is in the canonical convention. Expanding the limiting metric itself gives (37) and hence the flat metric (39) . The expansion must be made after taking the extremal limit: the auxiliary endpoint a 6 a_6 a 6 recedes to infinity, so an interchange of these limits cannot be assumed. The all-angle bounds described in subsection B.4 control the asymptotic end of this limiting metric. These statements establish a complete exterior cover for that interval and those individual roots. They do not establish regularity for arbitrary constants in the extremal ansatz.
Table 3. Two nonzero extremal horizon areas in the canonical scale at seven globally certified balance roots. Both surface gravities vanish exactly. Displayed areas are rounded from interval enclosures.
The resulting global time has timelike gradient. More explicitly, the lapse after completing the angular square is N T 2 = ρ 2 / det Λ > 0 N_T^2=\rho^2/\det\Lambda>0 N T 2 = ρ 2 / det Λ > 0 . Along an orbit-space path to the asymptotic end, choose d Φ A = − β A d T d\Phi^{A}=-\beta^{A}dT d Φ A = − β A d T and increase d T dT d T sufficiently to obtain a future timelike escaping curve. At a future horizon a future causal vector V V V satisfies g ( ∂ v , V ) = C F 0 V R ≤ 0 g(\partial_v,V)=CF_0V^{R}\leq0 g ( ∂ v , V ) = C F 0 V R ≤ 0 and cannot cross outward from the locally attached interior. This identifies the outer null boundaries in the chosen future extension. It would be interesting to pursue a maximal analytic extension.
Thermodynamics at double extremality
The constants in the regular canonical chart give the entropies without integrating a horizon section. Indeed, F 0 det Λ 0 = sin 2 θ / C h 2 F_0\det\Lambda_0=\sin^2\theta/C_h^2 F 0 det Λ 0 = sin 2 θ / C h 2 by (68) , and the two angles have period 2 π 2\pi 2 π . Consequently
T h = 0 , A h = 8 π 2 C h , S h = 2 π 2 G 5 C h , S t o t = 2 π 2 G 5 ( C R − 1 + C H − 1 ) . (74) T_h=0,\qquad A_h=\frac{8\pi^2}{C_h},\qquad S_h=\frac{2\pi^2}{G_5C_h},\qquad S_{\mathrm{tot}}=\frac{2\pi^2}{G_5}\left(C_R^{-1}+C_H^{-1}\right). \tag*{(74)} T h = 0 , A h = C h 8 π 2 , S h = G 5 C h 2 π 2 , S tot = G 5 2 π 2 ( C R − 1 + C H − 1 ) . (74)
Here C h C_h C h and Ω h A \Omega_h^A Ω h A are constants determined by the exact metric near the horizon. To extract them, expand a component at fixed θ \theta θ in the signed radial coordinate R R R of § 4.3 . We write [ R n ] Q [R^n]Q [ R n ] Q for the coefficient of R n R^n R n in this expansion: if Q = Q 0 + R Q 1 + O ( R 2 ) Q=Q_0+RQ_1+O(R^2) Q = Q 0 + R Q 1 + O ( R 2 ) , then [ R 0 ] Q = Q 0 [R^0]Q=Q_0 [ R 0 ] Q = Q 0 . In particular, f ˉ e \bar f_e f ˉ e has a second-order pole, but R 2 f ˉ e R^2\bar f_e R 2 f ˉ e has a finite constant coefficient. Since C h C_h C h and Ω h A \Omega_h^A Ω h A are independent of θ \theta θ , they may be computed at the equator θ = π / 2 \theta=\pi/2 θ = π /2 , where sin 2 θ = 1 \sin^2\theta=1 sin 2 θ = 1 :
L h = [ R 0 ] Λ , F h = [ R 0 ] ( R 2 f ˉ e ) , ( b h ) A = [ R 0 ] ( G ˉ e ) A T , L_h=[R^0]\Lambda,\qquad F_h=[R^0](R^2\bar f_e),\qquad(b_h)_A=[R^0](\bar G_e)_{AT}, L h = [ R 0 ] Λ , F h = [ R 0 ] ( R 2 f ˉ e ) , ( b h ) A = [ R 0 ] ( G ˉ e ) A T ,
C h = ( F h det L h ) − 1 / 2 , Ω h = − L h − 1 b h . (75) C_h=(F_h\det L_h)^{-1/2},\qquad\Omega_h=-L_h^{-1}b_h. \tag*{(75)} C h = ( F h det L h ) − 1/2 , Ω h = − L h − 1 b h . (75)
Here L h L_h L h is the 2 × 2 2\times2 2 × 2 angular metric, b h b_h b h is a two-component column with A ∈ { Ψ , Φ } A\in\{\Psi,\Phi\} A ∈ { Ψ , Φ } , and the positive root is chosen for C h C_h C h .
The central and ring extremal area-spin relations provide useful additional consistency checks:
A H = 8 π G 5 ∣ J H Ψ J H Φ ∣ , A R = 8 π G 5 ∣ J R Φ ∣ . (76) A_H=8\pi G_5\sqrt{\lvert J_{H\Psi}J_{H\Phi}\rvert},\qquad A_R=8\pi G_5\lvert J_{R\Phi}\rvert. \tag*{(76)} A H = 8 π G 5 ∣ J H Ψ J H Φ ∣ , A R = 8 π G 5 ∣ J R Φ ∣ . (76)
They agree with the spherical and ring extremal near-horizon relations [35 ] . They do not identify the whole interacting metric with a pair of isolated solutions. The horizon angular velocities, redshifts and Komar masses retain the interaction.
There is a short way to obtain the individual extremal charges. The area–spin relations (76) fix one ring spin and the product of the central spins, while conservation fixes the remaining combinations. Together with the independently determined ADM spins this determines all four horizon spins. The horizon Smarr relations then give the two Komar masses. On a connected regular branch fix σ R = sgn J R Φ \sigma_R=\operatorname{sgn}J_{R\Phi} σ R = sgn J R Φ and σ H = sgn ( J H Ψ J H Φ ) \sigma_H=\operatorname{sgn}(J_{H\Psi}J_{H\Phi}) σ H = sgn ( J H Ψ J H Φ ) . The near-horizon identities (76) and charge conservation give
J R Φ = σ R π G 5 C R , J H Φ = J Φ − σ R π G 5 C R , J_{R\Phi}=\frac{\sigma_R\pi}{G_5C_R},\qquad J_{H\Phi}=J_\Phi-\frac{\sigma_R\pi}{G_5C_R}, J R Φ = G 5 C R σ R π , J H Φ = J Φ − G 5 C R σ R π ,
J H Ψ = σ H π 2 G 5 2 C H 2 J H Φ , J R Ψ = J Ψ − J H Ψ , J_{H\Psi}=\frac{\sigma_H\pi^2}{G_5^2C_H^2J_{H\Phi}},\qquad J_{R\Psi}=J_\Psi-J_{H\Psi}, J H Ψ = G 5 2 C H 2 J H Φ σ H π 2 , J R Ψ = J Ψ − J H Ψ ,
M h = 3 2 ( Ω h Ψ J h Ψ + Ω h Φ J h Φ ) . (77) M_h=\frac{3}{2}\left(\Omega_h^\Psi J_{h\Psi}+\Omega_h^\Phi J_{h\Phi}\right). \tag*{(77)} M h = 2 3 ( Ω h Ψ J h Ψ + Ω h Φ J h Φ ) . (77)
A finite nonzero central area excludes J H Φ = 0 J_{H\Phi}=0 J H Φ = 0 . The signs are discrete branch data, not freely adjustable continuous parameters. In the angular orientation of the extremal phase plots they are ( σ R , σ H ) = ( + 1 , − 1 ) (\sigma_R,\sigma_H)=(+1,-1) ( σ R , σ H ) = ( + 1 , − 1 ) . Here the final equality uses the extremal horizon Smarr identity. Direct Komar surfaces provide an independent check of every local charge, including the masses. The ADM mass is obtained separately at infinity and agrees with M R + M H M_R+M_H M R + M H .
Explicitly, in the canonical convention put ξ = 1 / R W \xi=1/R_W ξ = 1/ R W , z = 0 z=0 z = 0 and ρ = 1 / ξ \rho=1/\xi ρ = 1/ ξ . The notation g T a ( 1 ) = [ ξ 1 ] ( G ˉ e ) T a g_{Ta}^{(1)}=[\xi^1](\bar{G}_e)_{Ta} g T a ( 1 ) = [ ξ 1 ] ( G ˉ e ) T a denotes the coefficient linear in inverse radius in the asymptotic expansion. In the asymptotic coordinates of (36) , z = 0 z=0 z = 0 corresponds to θ = π / 4 \theta=\pi/4 θ = π /4 (distinct from the local horizon polar angle used above). The ADM charges are
M = 3 π 4 G 5 g T T ( 1 ) , J Ψ = − π G 5 g T Ψ ( 1 ) , J Φ = − π G 5 g T Φ ( 1 ) . (78) M=\frac{3\pi}{4G_5}g_{TT}^{(1)},\qquad J_{\Psi}=-\frac{\pi}{G_5}g_{T\Psi}^{(1)},\qquad J_{\Phi}=-\frac{\pi}{G_5}g_{T\Phi}^{(1)}. \tag*{(78)} M = 4 G 5 3 π g T T ( 1 ) , J Ψ = − G 5 π g T Ψ ( 1 ) , J Φ = − G 5 π g T Φ ( 1 ) . (78)
The rational functions in (60) make this a finite polynomial calculation: at z = 0 z=0 z = 0 we substitute u = 1 − Z ξ + O ( ξ 2 ) u=1-Z\xi+O(\xi^2) u = 1 − Z ξ + O ( ξ 2 ) , y = 1 2 − Z ξ / 4 + O ( ξ 2 ) y=\frac{1}{2}-Z\xi/4+O(\xi^2) y = 2 1 − Z ξ /4 + O ( ξ 2 ) . After cancelling the common leading power, the linear coefficient of ( n 0 + n 1 ξ ) / ( d 0 + d 1 ξ ) (n_0+n_1\xi)/(d_0+d_1\xi) ( n 0 + n 1 ξ ) / ( d 0 + d 1 ξ ) is ( n 1 d 0 − n 0 d 1 ) / d 0 2 (n_1d_0-n_0d_1)/d_0^2 ( n 1 d 0 − n 0 d 1 ) / d 0 2 . Equations (75) –(78) therefore specify all extremal thermodynamic quantities.
Phase diagrams, blackfold predictions and frame dragging
The thermodynamic data allow three complementary comparisons. First, we compare solutions in mutual thermal and rotational equilibrium at fixed angular momentum in the second rotation plane. Next, we follow a double-extremal Saturn curve, whose two angular momenta both vary, and compare it with isolated black holes, rings and the blackfold approximation. Finally, we hold conserved charges fixed to study non-uniqueness and isolate frame dragging in the second rotation plane.
Reduced variables
At fixed ADM mass, entropy is proportional to total horizon area. Following the five-dimensional phase-diagram conventions of [2 , 7 , 12 ] , we remove the overall length scale by defining
j A = 27 π 32 G 5 J A M 3 / 2 , a h = 3 16 3 π A h ( G 5 M ) 3 / 2 , j_A=\sqrt{\frac{27\pi}{32G_5}}\frac{J_A}{M^{3/2}},\qquad a_h=\frac{3}{16}\sqrt{\frac{3}{\pi}}\frac{A_h}{(G_5M)^{3/2}}, j A = 32 G 5 27 π M 3/2 J A , a h = 16 3 π 3 ( G 5 M ) 3/2 A h ,
a t o t = a R + a H , ω h A = 8 G 5 M 3 π Ω h A . (79) a_{\mathrm{tot}}=a_R+a_H,\qquad\omega_{hA}=\sqrt{\frac{8G_5M}{3\pi}}\Omega_{hA}. \tag*{(79)} a tot = a R + a H , ω h A = 3 π 8 G 5 M Ω h A . (79)
Individual reduced Komar spins use the same total M M M . Since there are two ADM spins, a phase comparison requires a surface or fixed-spin slices. For a Myers-Perry (MP) black hole with parameters ( μ , a , b ) (\mu,a,b) ( μ , a , b ) , j Ψ = a / μ j_\Psi=a/\sqrt{\mu} j Ψ = a / μ , j Φ = b / μ j_\Phi=b/\sqrt{\mu} j Φ = b / μ , we have that
a M P = 2 1 − j Ψ 2 − j Φ 2 + [ 1 − ( j Ψ + j Φ ) 2 ] [ 1 − ( j Ψ − j Φ ) 2 ] , ∣ j Ψ ∣ + ∣ j Φ ∣ ≤ 1. (80) a_{\mathrm{MP}}=2\sqrt{1-j_\Psi^2-j_\Phi^2+\sqrt{\left[1-(j_\Psi+j_\Phi)^2\right]\left[1-(j_\Psi-j_\Phi)^2\right]}},\qquad|j_\Psi|+|j_\Phi|\leq1. \tag*{(80)} a MP = 2 1 − j Ψ 2 − j Φ 2 + [ 1 − ( j Ψ + j Φ ) 2 ] [ 1 − ( j Ψ − j Φ ) 2 ] , ∣ j Ψ ∣ + ∣ j Φ ∣ ≤ 1. (80)
For a Pomeransky-Sen’kov (PS) ring, we take Ψ \Psi Ψ as the ring plane and 0 ≤ ν < 1 0 \le\nu< 1 0 ≤ ν < 1 , 2 ν ≤ λ < 1 + ν 2\sqrt{\nu} \le\lambda< 1+\nu 2 ν ≤ λ < 1 + ν . The positive-spin sheet is
j Ψ = ( 1 + λ − 6 ν + λ ν + ν 2 ) 1 + λ + ν 2 2 λ ( 1 − ν ) 2 , j Φ = ( 1 − λ + ν ) ν ( 1 + λ + ν ) 2 λ ( 1 − ν ) 2 , a P S = 2 ( 1 + λ + ν ) ( 1 − λ + ν ) 3 / 2 λ ( 1 − ν ) 2 ( y h 2 − y h ) , y h = − 2 λ + λ 2 − 4 ν . (81) \begin{aligned}
j_{\Psi} &= \frac{(1+\lambda-6\nu+\lambda\nu+\nu^{2})\sqrt{1+\lambda+\nu}}{2\sqrt{2\lambda}(1-\nu)^{2}},\\
j_{\Phi} &= \frac{(1-\lambda+\nu)\sqrt{\nu(1+\lambda+\nu)}}{\sqrt{2\lambda}(1-\nu)^{2}},\\
a_{\mathrm{PS}} &= \frac{2(1+\lambda+\nu)(1-\lambda+\nu)^{3/2}}{\sqrt{\lambda}(1-\nu)^{2}(y_{h}^{2}-y_{h})}, \qquad y_{h}=-\frac{2}{\lambda+\sqrt{\lambda^{2}-4\nu}}. \tag*{(81)}
\end{aligned} j Ψ j Φ a PS = 2 2 λ ( 1 − ν ) 2 ( 1 + λ − 6 ν + λ ν + ν 2 ) 1 + λ + ν , = 2 λ ( 1 − ν ) 2 ( 1 − λ + ν ) ν ( 1 + λ + ν ) , = λ ( 1 − ν ) 2 ( y h 2 − y h ) 2 ( 1 + λ + ν ) ( 1 − λ + ν ) 3/2 , y h = − λ + λ 2 − 4 ν 2 . (81)
Both PS branches are included wherever the same pair of spins admits two rings. Reflection of either angular coordinate supplies other sign sectors.
Configurations in mutual thermal equilibrium
We now restrict the balanced black Saturn family to configurations in mutual thermodynamic equilibrium. Mutual equilibrium imposes three conditions in addition to balance:
κ R = κ H , Ω R Ψ = Ω H Ψ , Ω R Φ = Ω H Φ . (82) \kappa_{\mathrm{R}}=\kappa_{\mathrm{H}}, \qquad\Omega_{\mathrm{R}}^{\Psi}=\Omega_{\mathrm{H}}^{\Psi}, \qquad\Omega_{\mathrm{R}}^{\Phi}=\Omega_{\mathrm{H}}^{\Phi}. \tag*{(82)} κ R = κ H , Ω R Ψ = Ω H Ψ , Ω R Φ = Ω H Φ . (82)
The temperature equality prevents heat exchange, while the two angular-velocity equalities prevent rotational exchange in either plane. At zero temperature the first condition is automatic, but the two angular conditions remain nontrivial. The double-extremal solution of § 4 cannot satisfy both while retaining a regular axis between the horizons: appendix D proves this for the full finite parameter chart and every real sign sector. We provide strong numerical evidence that doubly spinning extremal black saturns cannot be in thermal equilibrium, which was already predicted in the blackfold regime [14 ] .
At nonzero temperature, solving these conditions within the balanced Saturn family leads to a branch approaching the auxiliary boundary a 6 = a 5 a_{6}=a_{5} a 6 = a 5 while both physical horizon rods retain nonzero length. Thus neither horizon becomes extremal at this boundary. A finite expression for the limiting metric follows by setting
a 6 = a 5 + ϵ 2 , B 5 = ϵ B ˉ 5 , C 6 = C ˉ 6 ϵ + O ( ϵ ) , C ˉ 6 2 = a 5 2 ( a 5 − a 2 ) . (83) a_{6}=a_{5}+\epsilon^{2}, \qquad B_{5}=\epsilon\bar{B}_{5}, \qquad C_{6}=\frac{\bar{C}_{6}}{\epsilon}+O(\epsilon), \qquad\bar{C}_{6}^{2}=\frac{a_{5}}{2(a_{5}-a_{2})}. \tag*{(83)} a 6 = a 5 + ϵ 2 , B 5 = ϵ B ˉ 5 , C 6 = ϵ C ˉ 6 + O ( ϵ ) , C ˉ 6 2 = 2 ( a 5 − a 2 ) a 5 . (83)
The other polarisations are finite. We collect the exact monomials of f f f and f G fG f G before taking the coefficient:
f ∗ = [ f ] ϵ 0 , G ∗ = [ f G ] ϵ 0 f ∗ . (84) f_{\ast}=[f]_{\epsilon^{0}}, \qquad G_{\ast}=\frac{[fG]_{\epsilon^{0}}}{f_{\ast}}. \tag*{(84)} f ∗ = [ f ] ϵ 0 , G ∗ = f ∗ [ f G ] ϵ 0 . (84)
The retained powers have nonnegative degrees, and u 6 = u 5 u_{6}=u_{5} u 6 = u 5 in the coefficient. This defines the complete limiting vacuum metric on its regular off-axis domain, including its conformal factor. For each fixed j Φ j_{\Phi} j Φ , we trace a one-parameter family using
q = log a 4 − a 3 a 2 − a 1 . q=\log\frac{a_{4}-a_{3}}{a_{2}-a_{1}}. q = log a 2 − a 1 a 4 − a 3 .
Here a 4 − a 3 > 0 a_{4}-a_{3}>0 a 4 − a 3 > 0 is the Weyl-coordinate gap between the ring and central-hole horizon rods. We use a 2 − a 1 = 1 a_{2}-a_{1}=1 a 2 − a 1 = 1 to fix the scale in the continuation. This gap is a coordinate parameter, not the proper distance between the horizons. At every step we adjust the remaining parameters to solve balance, the three equilibrium conditions and the chosen value of j Φ j_{\Phi} j Φ . Figure 2 shows the resulting slices at j Φ = 0.01 , 0.05 , 0.10 j_{\Phi}=0.01, 0.05, 0.10 j Φ = 0.01 , 0.05 , 0.10 . Points are connected in increasing q q q , so the curve continues smoothly even where j Ψ ( q ) j_{\Psi}(q) j Ψ ( q ) reverses direction. Near such a reversal, two distinct configurations can have the same j Ψ j_{\Psi} j Ψ and different areas. We checked the exterior and rod signs along this branch, providing numerical evidence for regularity.1
Figure 2. Mutual thermal equilibrium. Panels (a)–(c) compare total reduced area at fixed j Φ = 0.01 , 0.05 , 0.10 j_{\Phi} = 0.01, 0.05, 0.10 j Φ = 0.01 , 0.05 , 0.10 while (d) locates these sections in the two-spin plane, with colour denoting Saturn area. Purple: the auxiliary-boundary metric (84) , blue and green: Myers-Perry and Pomeransky–Sen’kov. Dashed orange and dotted green: leading blackfold Saturn and isolated-ring predictions, displayed only for j Ψ ≥ 1 j_{\Psi} \ge1 j Ψ ≥ 1 . Panel (c) is a qualitative comparison as the blackfold approximation is not strictly valid there. The hierarchy of scales and the larger mismatch at j Φ = 0.10 j_{\Phi} = 0.10 j Φ = 0.10 are discussed in appendix § 10 . The Saturn sections are numerical continuations, not a complete equilibrium surface.
In mutual equilibrium the first law simplifies to
d M = κ 8 π G 5 d ( A R + A H ) + Ω A d J A . (85) \mathrm{d}M=\frac{\kappa}{8\pi G_{5}}\mathrm{d}(A_{\mathrm{R}}+A_{\mathrm{H}})+\Omega_{\mathrm{A}}\mathrm{d}J_{\mathrm{A}}. \tag*{(85)} d M = 8 π G 5 κ d ( A R + A H ) + Ω A d J A . (85)
At fixed mass, this gives
d a t o t = − 2 2 ∑ A Ω A κ d j A . (86) \mathrm{d}a_{\mathrm{tot}}=-2\sqrt{2}\sum_{\mathrm{A}}\frac{\Omega_{\mathrm{A}}}{\kappa}\mathrm{d}j_{\mathrm{A}}. \tag*{(86)} d a tot = − 2 2 A ∑ κ Ω A d j A . (86)
Equivalently, along a fixed-j Φ j_{\Phi} j Φ curve,
d a t o t d q = − 2 2 Ω Ψ κ d j Ψ d q . \frac{\mathrm{d}a_{\mathrm{tot}}}{\mathrm{d}q}=-2\sqrt{2}\frac{\Omega_{\Psi}}{\kappa}\frac{\mathrm{d}j_{\Psi}}{\mathrm{d}q}. d q d a tot = − 2 2 κ Ω Ψ d q d j Ψ .
At a turning point of j Ψ ( q ) j_{\Psi}(q) j Ψ ( q ) the right-hand side vanishes, so a t o t ( q ) a_{\mathrm{tot}}(q) a tot ( q ) is stationary as well, provided Ω Ψ / κ \Omega_{\Psi}/\kappa Ω Ψ / κ is finite. If this ratio is also nonzero, the two stationary-point conditions are equivalent. This explains the paired area and spin extrema seen when the branch doubles back in the plot.
On the plotted equilibrium branches, the Saturn has a smaller total reduced area than the competing PS ring at the same two ADM spins, where both exist. This is consistent with the singly spinning equilibrium phase diagram of [7 , 12 ] . The leading nonextremal blackfold curves are derived in appendix C. They reproduce part of the thin-ring behaviour, while agreement deteriorates on the j Φ = 0.10 j_{\Phi}=0.10 j Φ = 0.10 slice. That appendix explains the scale hierarchy and endpoint sensitivity responsible for treating this slice as a qualitative comparison.
A double-extremal phase curve
The extremal comparison that we provide here is outside mutual rotational equilibrium. Both temperatures vanish, but the regular double-extremal family cannot have both angular velocities equal, as proved in appendix D. The blackfold curves are likewise non-equilibrium configurations in their controlled regime. Fix (B.56) and continue the exact coefficient metric by varying d d d and solving balance for ( c , H ∗ , v ∗ ) (c,H_{*},v_{*}) ( c , H ∗ , v ∗ ) . The sampled curve spans 0.00014 ≤ d ≤ 14 0.00014\le d\le14 0.00014 ≤ d ≤ 14 , with two positive finite areas at each point. Increasing d d d grows the ring from a small perturbation to a component with an area larger than the centre. Some representative numerical examples used to plot these curves are
Table 5.
ADM charges are obtained from the exact asymptotic coefficients, and areas from exact horizon coefficients. All these examples are globally regular.
For a meaningful blackfold comparison we must specify what is held fixed. At each exact Saturn point, we choose the isolated reference centre to have the same ( A H , J H Ψ , J H Φ ) (\mathcal{A}_{\mathrm{H}}, J_{\mathrm{H}\Psi}, J_{\mathrm{H}\Phi}) ( A H , J H Ψ , J H Φ ) . The central-horizon identity (76) is precisely the extremal MP area–spin relation, so the isolated reference centre is extremal. Matching the Weyl separation fixes R 2 = 2 Z + ( μ − a 2 − b 2 ) / 2 R^2 = 2Z_{+}(\mu-a^2-b^2)/2 R 2 = 2 Z + ( μ − a 2 − b 2 ) /2 , and the transverse potential fixes m = k / ( 2 ∣ Ω R Φ ∣ ) m = k/(2|\Omega^{\Phi}_{\mathrm{R}}|) m = k / ( 2∣ Ω R Φ ∣ ) and α = sign Ω R Φ \alpha= \operatorname{sign}\Omega^{\Phi}_{\mathrm{R}} α = sign Ω R Φ . Radial balance determines the orbital velocity providing a prediction for the curve of doubly-spinning extremal Saturns.
At the thin end m / R ≃ 1.02 × 10 − 6 m/\mathcal{R} \simeq1.02 \times10^{-6} m / R ≃ 1.02 × 1 0 − 6 , the relative discrepancies are 3.74 × 10 − 6 3.74 \times10^{-6} 3.74 × 1 0 − 6 in ring orbital spin, 2.05 × 10 − 6 2.05 \times10^{-6} 2.05 × 1 0 − 6 in transverse spin and area, and 2.49 × 10 − 7 2.49 \times10^{-7} 2.49 × 1 0 − 7 in the ADM mass increment. At d = 0.014 d = 0.014 d = 0.014 they are approximately one hundred times larger. Agreement improves with decreasing thickness, as expected from a leading blackfold expansion. At the thick end m / R ≃ 0.182 m/R \simeq0.182 m / R ≃ 0.182 , the orbital spin discrepancy is about 40% and is not surprising since this part of the blackfold line is an extrapolation beyond its regime of validity.
The isolated extremal boundaries are
M P : ∣ j Ψ ∣ + ∣ j Φ ∣ = 1 , a M P = 2 2 ∣ j Ψ j Φ ∣ , P S : j Ψ = 1 8 ∣ j Φ ∣ + ∣ j Φ ∣ , a P S = 2 2 ∣ j Φ ∣ , 0 < ∣ j Φ ∣ < 1 4 . (87) \begin{aligned}
\mathrm{MP}:&\ \lvert j_{\Psi}\rvert+ \lvert j_{\Phi}\rvert= 1, && a_{\mathrm{MP}} = 2\sqrt{2\lvert j_{\Psi}j_{\Phi}\rvert},\\
\mathrm{PS}:&\ j_{\Psi} = \frac{1}{8\lvert j_{\Phi}\rvert} + \lvert j_{\Phi}\rvert, && a_{\mathrm{PS}} = 2\sqrt{2}\lvert j_{\Phi}\rvert,\quad0 < \lvert j_{\Phi}\rvert< \frac{1}{4}. \tag*{(87)}
\end{aligned} MP : PS : ∣ j Ψ ∣ + ∣ j Φ ∣ = 1 , j Ψ = 8 ∣ j Φ ∣ 1 + ∣ j Φ ∣ , a MP = 2 2 ∣ j Ψ j Φ ∣ , a PS = 2 2 ∣ j Φ ∣ , 0 < ∣ j Φ ∣ < 4 1 . (87)
They are shown with the solid Saturn curve and the matched dashed blackfold curve in Figure 3 . The two-spin view in panel (c) shows which configurations can be compared at the same conserved charges. Panel (a) suppresses j Φ j_{\Phi} j Φ : two points at the same horizontal position can therefore have different transverse spins, even when their areas coincide. Likewise, panel (b) suppresses j Ψ j_{\Psi} j Ψ . In panel (c), two configurations have the same pair ( j Ψ , j Φ ) (j_{\Psi},j_{\Phi}) ( j Ψ , j Φ ) only if they occupy the same point. After fixing the common mass, an entropy comparison is meaningful only at such matching spin pairs. The Saturn curve crosses total j Φ = 0 j_{\Phi} = 0 j Φ = 0 by cancellation between the two horizon contributions, without either horizon losing extremality. At d = 0.014 d = 0.014 d = 0.014 , ∣ j Ψ ∣ + ∣ j Φ ∣ > 1 \lvert j_{\Psi}\rvert+ \lvert j_{\Phi}\rvert> 1 ∣ j Ψ ∣ + ∣ j Φ ∣ > 1 and ∣ j Φ ∣ > 1 / 4 \lvert j_{\Phi}\rvert> 1/4 ∣ j Φ ∣ > 1/4 , so this regular example from subsection B.4.4 has no regular isolated MP hole or balanced PS ring with the same ADM mass and both angular momenta. The MP bound is ∣ j Ψ ∣ + ∣ j Φ ∣ ≤ 1 \lvert j_{\Psi}\rvert+ \lvert j_{\Phi}\rvert\le1 ∣ j Ψ ∣ + ∣ j Φ ∣ ≤ 1 , whereas PS requires ∣ j Φ ∣ ≤ 1 / 4 \lvert j_{\Phi}\rvert\le1/4 ∣ j Φ ∣ ≤ 1/4 .
Figure 3. Double-extremal configurations. Panels (a) and (b) show area against each ADM spin, (c) shows the same configurations in the two-spin plane, (d) compares the blackfold prediction with the exact ring charges, area and ADM mass. Solid purple: the limiting Saturn metric, dashed orange: matched leading blackfold, blue and green: extremal MP and PS. The star marks d = 0.014 d = 0.014 d = 0.014 . Both spins vary along the Saturn curve, and the two horizons need not have equal angular velocities.
Non-uniqueness and dragging in the second plane
With six independent horizon charges but only three sums measured at infinity, one expects continuous non-uniqueness. We exhibit this by following a numerical family with fixed ADM charges (G 5 = 1 G_5 = 1 G 5 = 1 ):
( M , J Ψ , J Φ ) = ( 16.655554840 , 38.214561224 , 2.087501673 ) , ( j Ψ , j Φ ) = ( 0.915318098 , 0.05 ) , (88) \begin{aligned}
(M, J_{\Psi}, J_{\Phi}) &= (16.655554840, 38.214561224, 2.087501673), \\
(j_{\Psi}, j_{\Phi}) &= (0.915318098, 0.05), \tag*{(88)}
\end{aligned} ( M , J Ψ , J Φ ) ( j Ψ , j Φ ) = ( 16.655554840 , 38.214561224 , 2.087501673 ) , = ( 0.915318098 , 0.05 ) , (88)
while a t o t a_{\rm tot} a tot ranges from 0.696024735 to 0.746032225. At these charges the MP area is 1.078054506 and the two PS areas are 0.978641732 and 0.972337169. This particular Saturn path has smaller entropy than MP and PS black holes. The varying area gives direct numerical non-uniqueness at fixed conserved charges.
For the second-plane analogue of the original Saturn dragging experiment in [7 ] , we impose J H Φ = 0 J_{\mathrm{H}\Phi}=0 J H Φ = 0 while retaining the central orbital spin. Solving balance together with this zero-spin condition gives a configuration whose central horizon still rotates in the transverse plane relative to the nonrotating frame at infinity. For this example and Figure 4 , we reverse both angular coordinates relative to the A A A -normalised basis of § 2 : ( Ψ , Φ ) ↦ ( − Ψ , − Φ ) (\Psi,\Phi)\mapsto(-\Psi,-\Phi) ( Ψ , Φ ) ↦ ( − Ψ , − Φ ) . This reverses both J h A J_{\mathrm{h}A} J h A and Ω h A \Omega_{\mathrm{h}}^{A} Ω h A for each plane and leaves products Ω h A J h A \Omega_{\mathrm{h}}^{A}J_{\mathrm{h}A} Ω h A J h A unchanged. In this orientation the horizon charges are
Figure 4. Transverse frame dragging. Upper left: the central angular velocity changes as the ring transverse spin varies, with the other five Komar charges fixed and J H Φ = 0 J_{\mathrm{H}\Phi}=0 J H Φ = 0 . Upper right: on a separate fixed-rod continuation, a central counter-spin is needed to make ω H Φ = 0 \omega_{\mathrm{H}\Phi}=0 ω H Φ = 0 . Lower panel: the local zero-angular-momentum angular velocity approaches the central horizon value and decays to zero at infinity.
( M R , J R Ψ , J R Φ ) = ( 13.676724457 , 33.413964321 , 2.706293961 ) , ( M H , J H Ψ , J H Φ ) = ( 2.736435333 , 3.792434226 , 0 ) . (89) \begin{aligned}
(M_{\mathrm{R}},J_{\mathrm{R}\Psi},J_{\mathrm{R}\Phi})&=(13.676724457,33.413964321,2.706293961),\\
(M_{\mathrm{H}},J_{\mathrm{H}\Psi},J_{\mathrm{H}\Phi})&=(2.736435333,3.792434226,0). \tag*{(89)}
\end{aligned} ( M R , J R Ψ , J R Φ ) ( M H , J H Ψ , J H Φ ) = ( 13.676724457 , 33.413964321 , 2.706293961 ) , = ( 2.736435333 , 3.792434226 , 0 ) . (89)
Direct integration over the central horizon gives ∣ J H Φ ∣ < 10 − 24 \lvert J_{\mathrm{H}\Phi}\rvert<10^{-24} ∣ J H Φ ∣ < 1 0 − 24 , whereas
ω H Φ = 0.158748079001 ≠ 0. (90) \omega_{\mathrm{H}\Phi}=0.158748079001\ne0. \tag*{(90)} ω H Φ = 0.158748079001 = 0. (90)
For comparison, an isolated five-dimensional MP hole obeys
J Φ M P = π μ b 4 G 5 , Ω Φ M P = b r + 2 + b 2 . J_{\Phi}^{\mathrm{MP}}=\frac{\pi\mu b}{4G_{5}},\qquad\Omega_{\Phi}^{\mathrm{MP}}=\frac{b}{r_{+}^{2}+b^{2}}. J Φ MP = 4 G 5 π μ b , Ω Φ MP = r + 2 + b 2 b .
For positive mass and a regular horizon, J Φ M P = 0 J_{\Phi}^{\mathrm{MP}}=0 J Φ MP = 0 therefore implies b = 0 b=0 b = 0 and Ω Φ M P = 0 \Omega_{\Phi}^{\mathrm{MP}}=0 Ω Φ MP = 0 , even if the other spin is nonzero. For the central Saturn horizon, by contrast, J H Φ = 0 J_{\mathrm{H}\Phi}=0 J H Φ = 0 while Ω H Φ ≠ 0 \Omega_{\mathrm{H}}^{\Phi}\ne0 Ω H Φ = 0 . The surrounding rotating ring therefore induces transverse rotation relative to infinity even when the central horizon carries no transverse Komar angular momentum. This rotation is induced by the ring’s gravitational field.
To separate this interaction from changing the other horizon charges, we hold the remaining five Komar charges fixed and increase J R Φ J_{\mathrm{R}\Phi} J R Φ by up to 40%. The central transverse spin stays zero, while ω H Φ \omega_{\mathrm{H}\Phi} ω H Φ decreases from 0.158748079001 to 0.111864934012. Both masses, both ring-plane spins and J H Φ = 0 J_{\mathrm{H}\Phi}=0 J H Φ = 0 are held fixed, but balance requires the shape and separation to adjust. The decreasing angular velocity therefore measures the response along this constrained family, it does not assert that increasing ring spin at fixed geometry weakens dragging. Conversely, a separate continuation at fixed rod gaps, varying C 2 C_{2} C 2 and readjusting the balance parameters, reaches ω H Φ = 0 \omega_{\mathrm{H}\Phi}=0 ω H Φ = 0 with j H Φ = − 0.006885711854 ≠ 0 j_{\mathrm{H}\Phi}=-0.006885711854\ne0 j H Φ = − 0.006885711854 = 0 . Here a central counter-spin cancels the induced horizon rotation.
The local shift seen by a zero-angular-momentum observer is
Ω Z A M O A = − ( Λ − 1 ) A B G T B . (91) \Omega_{\mathrm{ZAMO}}^{A}=-(\Lambda^{-1})^{AB}G_{TB}. \tag*{(91)} Ω ZAMO A = − ( Λ − 1 ) A B G T B . (91)
This follows by setting the two angular covariant momenta p A = Λ A B ϕ ˙ B + G A T T ˙ p_{A}=\Lambda_{AB}\dot{\phi}^{B}+G_{AT}\dot{T} p A = Λ A B ϕ ˙ B + G A T T ˙ to zero and solving for ϕ ˙ A / T ˙ \dot{\phi}^{A}/\dot{T} ϕ ˙ A / T ˙ . Along the central equatorial Weyl profile, the transverse component approaches the horizon angular velocity (90) near the centre and vanishes at infinity.
Figure 4 shows three complementary tests. The upper left panel measures how the zero-spin central horizon responds as the ring spin varies with the other five Komar charges fixed. The upper right panel shows that zero central spin and zero central angular velocity occur at different points on the fixed-rod continuation. The lower panel exhibits the corresponding angular shift directly in the metric, interpolating between the rotating horizon and the nonrotating asymptotic frame. The plotted quantity is the dimensionless shift
ω Z A M O A = 8 G 5 M 3 π Ω Z A M O A , (92) \omega_{\mathrm{ZAMO}}^{A} = \sqrt{\frac{8G_{5}M}{3\pi}}\,\Omega_{\mathrm{ZAMO}}^{A}, \tag*{(92)} ω ZAMO A = 3 π 8 G 5 M Ω ZAMO A , (92)
using the total ADM mass M M M of each configuration. Thus uppercase Ω Z A M O A \Omega_{\mathrm{ZAMO}}^{A} Ω ZAMO A denotes the angular velocity in inverse-length units, while lowercase ω Z A M O A \omega_{\mathrm{ZAMO}}^{A} ω ZAMO A denotes its reduced value. Together these diagnostics distinguish the horizon’s intrinsic angular momentum from the angular velocity induced by its environment.
Discussion
We have constructed a six-parameter family of doubly spinning black Saturns in five-dimensional vacuum gravity, including balanced configurations with two finite-area extremal horizons and a regular, causal exterior in the domains established here. To our knowledge, these provide the first regular exact solutions with standard asymptotic flatness and disconnected extremal horizons of topologies S 3 S^3 S 3 and S 1 × S 2 S^1 \times S^2 S 1 × S 2 . Despite the existent inverse scattering method, without AI tools, this would have been a Herculean task: the seven scalar numerators in appendix § 9 .§ 1 contain 1643 monomials! Nevertheless, their compact representation makes it possible to compute thermodynamic charges, check regularity and causality, and study phase diagrams and transverse frame dragging.
The nonextremal family admits configurations in mutual thermodynamic equilibrium. For the doubly extremal metric of § 4 , however, equality of both horizon angular velocities is incompatible with a regular intervening axis. This exact obstruction agrees with the absence of mutual equilibrium in the thin-ring blackfold regime [14 ] . Appendix § 11 extends the obstruction to specified boundary limits and gives further numerical evidence from independently retuned parameters.
The blackfold approach also predicts extremal di-rings and bi-rings, with two disconnected ring horizons [14 ] . Within its regime of validity, extremal di-rings can attain mutual thermodynamic equilibrium. For extremal bi-rings, the matching conditions instead fall outside that regime. Exact inverse-scattering constructions would test these predictions beyond the thin-ring approximation and clarify the role of topology in the extremal phase diagram.
Appendix § 9 .§ 4 combines algebraic regularity conditions with interval bounds and numerical sampling. For the specified doubly extremal subfamily, the interval bounds cover the complete exterior and both horizon extensions. The wider nonextremal scans found no violations at tested points, but do not establish global regularity throughout the sampled parameter range. A natural next step is to extend the bounds to a full exterior cover on an open nonextremal parameter region and seek analytic proofs analogous to those for smoothness and stable causality of the singly spinning Saturn [29 , 30 ] .
Our local horizon coordinates provide a starting point for studying geodesics and the interior spacetime. Geodesic methods for doubly spinning rings [34 ] and the extension techniques developed for black rings [31 , 32 , 33 ] suggest useful approaches. A maximal analytic extension would determine the global causal structure beyond the locally extended horizons.
These solutions are also relevant to the semiclassical gravitational path integral, in which distinct geometries with the same boundary data can compete [13 ] . Disconnected-horizon configurations should be included when compatible with those data. A generic stationary Saturn has different temperatures and angular velocities on its two components, so it does not automatically admit a smooth Euclidean continuation with a single thermal identification. The subfamilies in mutual equilibrium are therefore especially relevant to comparisons at fixed temperature and angular potentials, while entropy comparisons at fixed conserved charges can be made more generally.
Acknowledgements
J. Armas is partly funded by the Dutch Institute for Emergent Phenomena (DIEP) cluster at the University of Amsterdam via the DIEP programme Foundations and Applications of Emergence (FAEME) and the national NWA-ORC 2023 consortium EAAS. The author used OpenAI Codex Astra 6 Pro, and Anthropic Claude Fable 5.1 to assist with manuscript revision, literature cross-checks, and review of mathematical arguments and analysis code. The author performed human checks on the AI-assisted material and takes responsibility for the final content and conclusions.
Black Saturn limits
This appendix discusses various limits of the general black Saturn solution in the main text. Removing one component yields Myers-Perry or Pomeransky-Sen’kov black hole, while removing transverse rotation yields the singly spinning Saturn. We also discuss in detail limits of the black saturn in which only one horizon is extremal.
Removing the ring: doubly spinning Myers–Perry
Let a 1 , a 2 , a 3 → 0 a_1,a_2,a_3 \to0 a 1 , a 2 , a 3 → 0 , put A = a 4 A=a_4 A = a 4 , B = a 5 B=a_5 B = a 5 , C = a 6 C=a_6 C = a 6 and L = B − A > 0 L=B-A>0 L = B − A > 0 , and denote the limiting polarizations by the same symbols. In the spectral gauge a 1 = 0 a_1=0 a 1 = 0 , define
η = 1 − C 1 C 2 C 3 , c = C 6 = 1 2 ( C − B ) , (93) \eta= 1-C_1C_2C_3,\qquad c=C_6=\frac{1}{\sqrt{2(C-B)}}, \tag*{(93)} η = 1 − C 1 C 2 C 3 , c = C 6 = 2 ( C − B ) 1 , (93)
h = B 4 , k = − B 5 ( B + 1 ) 2 ( A − B ) ( C − B ) , (94) h=B_4,\qquad k=-\frac{B_5(B+1)}{2(A-B)(C-B)}, \tag*{(94)} h = B 4 , k = − 2 ( A − B ) ( C − B ) B 5 ( B + 1 ) , (94)
α = 2 L k c , β = 2 L 1 − c h h + 2 c L , n = η h + 2 c L 2 c L , δ = n 2 ( 1 − α 2 β 2 ) . (95) \alpha=\sqrt{2L}\frac{k}{c},\qquad\beta=\sqrt{2L}\frac{1-ch}{h+2cL},\qquad n=\eta\frac{h+2cL}{2cL},\qquad\delta=n^2(1-\alpha^2\beta^2). \tag*{(95)} α = 2 L c k , β = 2 L h + 2 c L 1 − c h , n = η 2 c L h + 2 c L , δ = n 2 ( 1 − α 2 β 2 ) . (95)
We take the orientation branch n > 0 n>0 n > 0 and ∣ α β ∣ < 1 |\alpha\beta|<1 ∣ α β ∣ < 1 . The formula requires finite polarizations, η ≠ 0 \eta\ne0 η = 0 and h + 2 c L ≠ 0 h+2cL\ne0 h + 2 c L = 0 . The exact MP parameters are
m = 2 L n 2 ( 1 + α 2 ) ( 1 + β 2 ) , a = − 2 L n α , b = − 2 L n β . (96) m=2Ln^2(1+\alpha^2)(1+\beta^2),\qquad a=-\sqrt{2L}\,n\alpha,\qquad b=-\sqrt{2L}\,n\beta. \tag*{(96)} m = 2 L n 2 ( 1 + α 2 ) ( 1 + β 2 ) , a = − 2 L n α , b = − 2 L n β . (96)
Set τ = t + ϕ / c \tau=t+\phi/c τ = t + ϕ / c , ℓ = L / 2 \ell=\sqrt{L/2} ℓ = L /2 and
\begin{pmatrix}
\tau\\
\psi\\
\phi
\end{pmatrix}
=
\underbrace{\begin{pmatrix}
1 & 2\alpha\ell n & 2\beta\ell n\\
0 & n & -\alpha\beta n\\
0 & -\alpha\beta n & n
\end{pmatrix}}_{\mathcal{B}}
\begin{pmatrix}
T\\
\Psi\\
\Phi
\end{pmatrix},
\qquad
\rho_{\mathrm{MP}}=\delta\rho,\quad
z_{\mathrm{MP}}=\delta\left(z-\frac{A+B}{2}\right). \tag*{(97)}\end{pmatrix}
The determinant of this Killing-coordinate transformation is δ \delta δ while the preceding shift from t t t to τ \tau τ has unit determinant. We eliminate the disappearing poles before taking the limit. Block Gaussian elimination (the Schur complement) of the three coalescing left poles leaves the three right poles and a conformal multiplier η 2 \eta^2 η 2 . Put U = u A U=u_A U = u A , V = u B V=u_B V = u B , W = u C W=u_C W = u C and R U V = x + U V R_{UV}=x+UV R U V = x + U V . The auxiliary root satisfies
W 2 = ( V 2 − x V + 1 c 2 ) W + x . (98) W^2=\left(\frac{V^2-x}{V}+\frac{1}{c^2}\right)W+x. \tag*{(98)} W 2 = ( V V 2 − x + c 2 1 ) W + x . (98)
Eliminating its pole by a second Schur complement and making the time shift above leaves a two-pole metric G 2 = H 2 − N T Γ − 1 N G_2 = H_2 - N^{\mathsf T}\Gamma^{-1}N G 2 = H 2 − N T Γ − 1 N . With p = k / c p = k/c p = k / c and q = ( 1 − c h ) / ( h + 2 c L ) q = (1 - ch)/(h + 2cL) q = ( 1 − c h ) / ( h + 2 c L ) its data are
H 2 = diag ( − V / U , V , x / U ) , N = ( 1 0 q R U V / V 1 p R U V / V 0 ) , (99) H_2 = \operatorname{diag}(-V/U,V,x/U), \qquad N = \begin{pmatrix} 1 & 0 & qR_{UV}/V \\ 1 & pR_{UV}/V & 0 \end{pmatrix}, \tag*{(99)} H 2 = diag ( − V / U , V , x / U ) , N = ( 1 1 0 p R U V / V q R U V / V 0 ) , (99)
Γ 00 = U ( − x V + q 2 R U V 2 ) V 2 ( U 2 + x ) , Γ 11 = − U V + p 2 R U V 2 / V V 2 + x , Γ 01 = U U − V . (100) \Gamma_{00} = \frac{U(-xV+q^2R_{UV}^2)}{V^2(U^2+x)}, \qquad\Gamma_{11} = \frac{-UV+p^2R_{UV}^2/V}{V^2+x}, \qquad\Gamma_{01} = \frac{U}{U-V}. \tag*{(100)} Γ 00 = V 2 ( U 2 + x ) U ( − x V + q 2 R U V 2 ) , Γ 11 = V 2 + x − U V + p 2 R U V 2 / V , Γ 01 = U − V U . (100)
Every occurrence of W W W cancels exactly. Introduce
R A = x + ( z − A ) 2 , R B = x + ( z − B ) 2 , P = R A + R B L , ξ = R A − R B L , (101) R_A = \sqrt{x+(z-A)^2}, \qquad R_B = \sqrt{x+(z-B)^2}, \qquad P = \frac{R_A+R_B}{L}, \qquad\xi= \frac{R_A-R_B}{L}, \tag*{(101)} R A = x + ( z − A ) 2 , R B = x + ( z − B ) 2 , P = L R A + R B , ξ = L R A − R B , (101)
U = L 2 ( P − 1 ) ( 1 − ξ ) , V = L 2 ( P + 1 ) ( 1 − ξ ) , x = L 2 4 ( P 2 − 1 ) ( 1 − ξ 2 ) . (102) U = \frac{L}{2}(P-1)(1-\xi), \qquad V = \frac{L}{2}(P+1)(1-\xi), \qquad x = \frac{L^2}{4}(P^2-1)(1-\xi^2). \tag*{(102)} U = 2 L ( P − 1 ) ( 1 − ξ ) , V = 2 L ( P + 1 ) ( 1 − ξ ) , x = 4 L 2 ( P 2 − 1 ) ( 1 − ξ 2 ) . (102)
Then
H 2 = diag ( − P + 1 P − 1 , ℓ 2 ( P + 1 ) ( 1 − ξ ) , ℓ 2 ( P + 1 ) ( 1 + ξ ) ) , (103) H_2 = \operatorname{diag}\left(-\frac{P+1}{P-1},\ell^2(P+1)(1-\xi),\ell^2(P+1)(1+\xi)\right), \tag*{(103)} H 2 = diag ( − P − 1 P + 1 , ℓ 2 ( P + 1 ) ( 1 − ξ ) , ℓ 2 ( P + 1 ) ( 1 + ξ ) ) , (103)
N = ( 1 0 β ℓ ( P − 1 ) 1 α ℓ ( P − 1 ) 0 ) , (104) N = \begin{pmatrix} 1 & 0 & \beta\ell(P-1) \\ 1 & \alpha\ell(P-1) & 0 \end{pmatrix}, \tag*{(104)} N = ( 1 1 0 α ℓ ( P − 1 ) β ℓ ( P − 1 ) 0 ) , (104)
Γ 00 = ( P − 1 ) [ β 2 ( P − 1 ) − 1 − ξ ] 2 ( P + ξ ) , Γ 11 = ( P − 1 ) [ α 2 ( P − 1 ) − 1 + ξ ] 2 ( P − ξ ) , (105) \Gamma_{00} = \frac{(P-1)[\beta^2(P-1)-1-\xi]}{2(P+\xi)}, \qquad\Gamma_{11} = \frac{(P-1)[\alpha^2(P-1)-1+\xi]}{2(P-\xi)}, \tag*{(105)} Γ 00 = 2 ( P + ξ ) ( P − 1 ) [ β 2 ( P − 1 ) − 1 − ξ ] , Γ 11 = 2 ( P − ξ ) ( P − 1 ) [ α 2 ( P − 1 ) − 1 + ξ ] , (105)
Γ 01 = − P − 1 2 . (106) \Gamma_{01} = -\frac{P-1}{2}. \tag*{(106)} Γ 01 = − 2 P − 1 . (106)
Direct inversion of this 2 × 2 2 \times2 2 × 2 matrix gives
D = ( 1 − α 2 β 2 ) P + ( 1 + α 2 ) ( 1 + β 2 ) + ( α 2 − β 2 ) ξ , (107) \mathcal{D} = (1-\alpha^2\beta^2)P+(1+\alpha^2)(1+\beta^2)+(\alpha^2-\beta^2)\xi, \tag*{(107)} D = ( 1 − α 2 β 2 ) P + ( 1 + α 2 ) ( 1 + β 2 ) + ( α 2 − β 2 ) ξ , (107)
( G 2 ) τ τ = − 1 + 2 ( 1 + α 2 ) ( 1 + β 2 ) D , Σ = n 2 L D , (108) (G_2)_{\tau\tau} = -1+\frac{2(1+\alpha^2)(1+\beta^2)}{\mathcal{D}}, \qquad\Sigma= n^2LD, \tag*{(108)} ( G 2 ) τ τ = − 1 + D 2 ( 1 + α 2 ) ( 1 + β 2 ) , Σ = n 2 L D , (108)
r 2 = δ L P + 1 2 ( m − a 2 − b 2 ) , cos 2 θ = ξ . (109) r^2 = \delta LP+\frac{1}{2}(m-a^2-b^2), \qquad\cos2\theta= \xi. \tag*{(109)} r 2 = δ L P + 2 1 ( m − a 2 − b 2 ) , cos 2 θ = ξ . (109)
Substitution proves all six independent entries of B T G 2 B = G M P \mathcal{B}^{\mathsf T}G_2\mathcal{B}=G_{\mathrm{MP}} B T G 2 B = G MP . The conformal-factor determinant ratio gives, independently,
f ∗ = Σ 4 R A R B = δ 2 f M P ( ρ M P , z M P ) , ( m − ( a + b ) 2 ) ( m − ( a − b ) 2 ) = 4 δ 2 L 2 . (110) f_*=\frac{\Sigma}{4R_AR_B}=\delta^2f_{\mathrm{MP}}(\rho_{\mathrm{MP}},z_{\mathrm{MP}}), \qquad(m-(a+b)^2)(m-(a-b)^2)=4\delta^2L^2. \tag*{(110)} f ∗ = 4 R A R B Σ = δ 2 f MP ( ρ MP , z MP ) , ( m − ( a + b ) 2 ) ( m − ( a − b ) 2 ) = 4 δ 2 L 2 . (110)
The factor η 2 \eta^2 η 2 from the first Schur complement is included through n 2 n^2 n 2 . We recover the Myers–Perry metric [1 ] in Boyer–Lindquist coordinates:
d s M P 2 = − d T 2 + m Σ ( d T − a sin 2 θ d Ψ − b cos 2 θ d Φ ) 2 + Σ r 2 Δ d r 2 + Σ d θ 2 ds_{\mathrm{MP}}^2=-dT^2+\frac{m}{\Sigma}(dT-a\sin^2\theta\,d\Psi-b\cos^2\theta\,d\Phi)^2+\frac{\Sigma r^2}{\Delta}dr^2+\Sigma d\theta^2 d s MP 2 = − d T 2 + Σ m ( d T − a sin 2 θ d Ψ − b cos 2 θ d Φ ) 2 + Δ Σ r 2 d r 2 + Σ d θ 2
+ ( r 2 + a 2 ) sin 2 θ d Ψ 2 + ( r 2 + b 2 ) cos 2 θ d Φ 2 , (111) \qquad+(r^2+a^2)\sin^2\theta\,d\Psi^2+(r^2+b^2)\cos^2\theta\,d\Phi^2, \tag*{(111)} + ( r 2 + a 2 ) sin 2 θ d Ψ 2 + ( r 2 + b 2 ) cos 2 θ d Φ 2 , (111)
Σ = r 2 + a 2 cos 2 θ + b 2 sin 2 θ , Δ = ( r 2 + a 2 ) ( r 2 + b 2 ) − m r 2 . (112) \Sigma=r^2+a^2\cos^2\theta+b^2\sin^2\theta,\qquad\Delta=(r^2+a^2)(r^2+b^2)-mr^2. \tag*{(112)} Σ = r 2 + a 2 cos 2 θ + b 2 sin 2 θ , Δ = ( r 2 + a 2 ) ( r 2 + b 2 ) − m r 2 . (112)
The preceding algebra verifies every component of this metric and its Weyl base factor, without fitting m , a , b m,a,b m , a , b to ADM data. With D M P = ( m − ( a + b ) 2 ) ( m − ( a − b ) 2 ) D_{\mathrm{MP}}=\sqrt{\left(m-(a+b)^2\right)\left(m-(a-b)^2\right)} D MP = ( m − ( a + b ) 2 ) ( m − ( a − b ) 2 ) and r + 2 = ( m − a 2 − b 2 + D M P ) / 2 r_+^2=\left(m-a^2-b^2+D_{\mathrm{MP}}\right)/2 r + 2 = ( m − a 2 − b 2 + D MP ) /2 , the limiting charges and intensive variables are therefore
M M P = 3 π m 8 G 5 , J Ψ M P = π m a 4 G 5 , J Φ M P = π m b 4 G 5 , (113) M_{\mathrm{MP}}=\frac{3\pi m}{8G_5},\qquad J^{\mathrm{MP}}_{\Psi}=\frac{\pi ma}{4G_5},\qquad J^{\mathrm{MP}}_{\Phi}=\frac{\pi mb}{4G_5}, \tag*{(113)} M MP = 8 G 5 3 π m , J Ψ MP = 4 G 5 π ma , J Φ MP = 4 G 5 π mb , (113)
A M P = 2 π 2 m r + , S M P = π 2 m r + 2 G 5 , κ M P = D M P m r + , Ω A M P = ( a r + 2 + a 2 , b r + 2 + b 2 ) . (114) A_{\mathrm{MP}}=2\pi^2mr_+,\qquad S_{\mathrm{MP}}=\frac{\pi^2mr_+}{2G_5},\qquad\kappa_{\mathrm{MP}}=\frac{D_{\mathrm{MP}}}{mr_+},\qquad\Omega^{\mathrm{MP}}_A=\left(\frac{a}{r_+^2+a^2},\frac{b}{r_+^2+b^2}\right). \tag*{(114)} A MP = 2 π 2 m r + , S MP = 2 G 5 π 2 m r + , κ MP = m r + D MP , Ω A MP = ( r + 2 + a 2 a , r + 2 + b 2 b ) . (114)
The remaining horizon is the MP horizon, and its Komar integrals coincide with these ADM charges. The regular finite-area isolated extremal limit is
m = ( ∣ a ∣ + ∣ b ∣ ) 2 , r + 2 = ∣ a b ∣ > 0. (115) m=(|a|+|b|)^2,\qquad r_+^2=|ab|>0. \tag*{(115)} m = ( ∣ a ∣ + ∣ b ∣ ) 2 , r + 2 = ∣ ab ∣ > 0. (115)
Within the map above it can be reached by a correlated limit L → 0 L\to0 L → 0 , n ∼ L − 1 / 2 n\sim L^{-1/2} n ∼ L − 1/2 and ∣ α β ∣ → 1 |\alpha\beta|\to1 ∣ α β ∣ → 1 , keeping m , a , b m,a,b m , a , b finite. Indeed m − ( ∣ a ∣ + ∣ b ∣ ) 2 = 2 L n 2 ( 1 − ∣ α β ∣ ) 2 m-(|a|+|b|)^2=2Ln^2(1-|\alpha\beta|)^2 m − ( ∣ a ∣ + ∣ b ∣ ) 2 = 2 L n 2 ( 1 − ∣ α β ∣ ) 2 and D M P = 2 δ L D_{\mathrm{MP}}=2\delta L D MP = 2 δ L . This isolated limit is distinct from the interacting construction in section 4.
Removing the central hole: the Pomeransky–Sen'kov ring
Use a small parameter t > 0 t>0 t > 0 , with a 1 = 0 a_1=0 a 1 = 0 and
a 4 = A , a 5 = A + t 2 , a 6 = A + ( 1 + q ) t 2 , q > 0 , (116) a_4=A,\qquad a_5=A+t^2,\qquad a_6=A+(1+q)t^2,\qquad q>0, \tag*{(116)} a 4 = A , a 5 = A + t 2 , a 6 = A + ( 1 + q ) t 2 , q > 0 , (116)
B 4 = b 4 t + o ( t ) , B 5 = b 5 t 3 + o ( t 3 ) , C 6 = α 6 t + O ( t ) , α 6 2 = A 2 q ( A − a 2 ) . (117) B_4=b_4t+o(t),\qquad B_5=b_5t^3+o(t^3),\qquad C_6=\frac{\alpha_6}{t}+O(t),\qquad\alpha_6^2=\frac{A}{2q(A-a_2)}. \tag*{(117)} B 4 = b 4 t + o ( t ) , B 5 = b 5 t 3 + o ( t 3 ) , C 6 = t α 6 + O ( t ) , α 6 2 = 2 q ( A − a 2 ) A . (117)
Take b 5 ≠ 0 b_5\ne0 b 5 = 0 , K = 1 + q α 6 b 4 ≠ 0 \mathcal{K}=1+q\alpha_6b_4\ne0 K = 1 + q α 6 b 4 = 0 , and finite C 1 , C 2 , C 3 C_1,C_2,C_3 C 1 , C 2 , C 3 . We denote by G C H T G_{\mathrm{CHT}} G CHT and f C H T f_{\mathrm{CHT}} f CHT the Killing block and base coefficient of the unbalanced doubly spinning black ring given in [27 ] . We use the three-soliton representation of that solution, with four rod endpoints and seed constant one. The reduction gives
G 6 ⟶ G C H T , f 6 ⟶ K 2 f C H T , c 1 ′ = − A C 1 , c 2 ′ = − a 2 + 1 A − a 2 C 2 , c 3 ′ = C 3 . (118) G_6\longrightarrow G_{\mathrm{CHT}},\qquad f_6\longrightarrow\mathcal{K}^2f_{\mathrm{CHT}},\qquad c'_1=-AC_1,\qquad c'_2=-\frac{a_2+1}{A-a_2}C_2,\qquad c'_3=C_3. \tag*{(118)} G 6 ⟶ G CHT , f 6 ⟶ K 2 f CHT , c 1 ′ = − A C 1 , c 2 ′ = − A − a 2 a 2 + 1 C 2 , c 3 ′ = C 3 . (118)
We can see how the two merging poles cancel without inverting the full six-by-six matrix. We normalise rows 4 and 6 to time component one and row 5 to ψ \psi ψ component one. If U = u A U=u_A U = u A , R = x + U 2 R=x+U^2 R = x + U 2 and D 0 = 2 x U / R D_0=2xU/R D 0 = 2 xU / R , their leading forms are
m 4 = ( 1 , 0 , t b 4 r 4 ) , m 6 = ( 1 , 0 , t r 4 / α 6 ) , m 5 = ( 0 , 1 , t l 5 ) . (119) m_4=(1,0,tb_4r_4),\qquad m_6=(1,0,tr_4/\alpha_6),\qquad m_5=(0,1,tl_5). \tag*{(119)} m 4 = ( 1 , 0 , t b 4 r 4 ) , m 6 = ( 1 , 0 , t r 4 / α 6 ) , m 5 = ( 0 , 1 , t l 5 ) . (119)
In the basis ( m 6 − m 4 , m 5 ) (m_6-m_4,m_5) ( m 6 − m 4 , m 5 ) the limiting Gram block is
\begin{pmatrix} 0 & B \\ B & C_0 \end{pmatrix},\qquad B=\frac{r_4l_5H_{\phi\phi}}{q\alpha_6D_0}\mathcal{K},\qquad H_{\phi\phi}=-u_2u_3/U. \tag*{(120)}\end{pmatrix}
The difference row has zero limiting N N N and zero coupling to the other Gram rows. The lower-right entry of this block’s inverse is zero. Thus its Schur corrections to G G G and to the remaining Gram data vanish exactly. Its determinant is − B 2 -B^2 − B 2 . Tracking the raw row factors in both the numerator and the zero-polarization reference determinant leaves precisely K 2 \mathcal{K}^2 K 2 in the conformal-factor ratio. In particular, one must track those reference factors before setting polarisations to zero. The normalised row-5 expression is singular under that substitution.
Row 4 is now a trivial time anti-soliton at − x / U -x/U − x / U . Removing it gives
H C H T = diag ( u 1 u 3 / x , − u 1 u 2 U / x , − u 2 u 3 / U ) . (121) H_{\mathrm{CHT}}=\operatorname{diag}(u_1u_3/x,-u_1u_2U/x,-u_2u_3/U). \tag*{(121)} H CHT = diag ( u 1 u 3 / x , − u 1 u 2 U / x , − u 2 u 3 / U ) . (121)
It multiplies the time entry of row i i i by ( u i + x / U ) / ( u i − U ) (u_i+x/U)/(u_i-U) ( u i + x / U ) / ( u i − U ) . Rescale row 1 by − A -A − A . The resulting three rows are exactly the CHT rows with the parameters in (A.26). Coalescing the last three seed weights also gives the CHT seed conformal factor. These steps prove the stated reduction of both G G G and f f f .
Write s = c 2 ′ s=c'_2 s = c 2 ′ . Impose the CHT regularity and balance conditions
A ( A − a 3 ) = ( A − a 2 ) 2 , c 1 ′ = a 3 2 a 2 A , c 3 ′ = 2 c 1 ′ s ( a 3 − a 2 ) ( A − a 2 ) a 3 . (122) A(A-a_3)=(A-a_2)^2,\qquad c'_1=\sqrt{\frac{a_3}{2a_2A}},\qquad c'_3=\frac{2c'_1s(a_3-a_2)(A-a_2)}{a_3}. \tag*{(122)} A ( A − a 3 ) = ( A − a 2 ) 2 , c 1 ′ = 2 a 2 A a 3 , c 3 ′ = a 3 2 c 1 ′ s ( a 3 − a 2 ) ( A − a 2 ) . (122)
The CHT-to-PS transformation of [27 ] then supplies an analytic identification with the balanced PS metric. Explicitly, put
a ˉ = 1 2 ( a 2 + a 3 ) , c s = a 3 − a 2 2 A − a 2 − a 3 , β s = a 3 − a 2 a 2 c s 2 , Q = 1 + c s 2 − β s ( 1 − c s 2 ) . (123) \bar{a}=\frac{1}{2}(a_2+a_3),\qquad c_s=\frac{a_3-a_2}{2A-a_2-a_3},\qquad\beta_s=\frac{a_3-a_2}{a_2}c_s^2,\qquad Q=1+c_s^2-\beta_s(1-c_s^2). \tag*{(123)} a ˉ = 2 1 ( a 2 + a 3 ) , c s = 2 A − a 2 − a 3 a 3 − a 2 , β s = a 2 a 3 − a 2 c s 2 , Q = 1 + c s 2 − β s ( 1 − c s 2 ) . (123)
μ = 2 c s Q + Q 2 − 4 c s 2 , ν c = μ − c s 1 − c s μ . (124) \mu=\frac{2c_s}{Q+\sqrt{Q^2-4c_s^2}},\qquad\nu_c=\frac{\mu-c_s}{1-c_s\mu}. \tag*{(124)} μ = Q + Q 2 − 4 c s 2 2 c s , ν c = 1 − c s μ μ − c s . (124)
λ = μ + ν c , ν P S = μ ν c , k 2 = δ 0 ( A − a ˉ ) . (125) \lambda=\mu+\nu_c,\qquad\nu_{\mathrm{PS}}=\mu\nu_c,\qquad k^2=\delta_0(A-\bar{a}). \tag*{(125)} λ = μ + ν c , ν PS = μ ν c , k 2 = δ 0 ( A − a ˉ ) . (125)
Here δ 0 = K 2 δ C H T \delta_0=\mathcal{K}^2\delta_{\mathrm{CHT}} δ 0 = K 2 δ CHT , ρ P S = δ 0 ρ \rho_{\mathrm{PS}}=\delta_0\rho ρ PS = δ 0 ρ and z P S = δ 0 ( z − a ˉ ) z_{\mathrm{PS}}=\delta_0(z-\bar{a}) z PS = δ 0 ( z − a ˉ ) . The explicit ring-coordinate map is given below. Multiplication of f C H T f_{\mathrm{CHT}} f CHT by K 2 \mathcal{K}^2 K 2 multiplies both unit-period axis vectors by ∣ K ∣ |\mathcal{K}| ∣ K ∣ . Thus the physical Killing basis is A C H T diag ( 1 , ∣ K ∣ , ∣ K ∣ ) A_{\mathrm{CHT}}\operatorname{diag}(1,|\mathcal{K}|,|\mathcal{K}|) A CHT diag ( 1 , ∣ K ∣ , ∣ K ∣ ) .
The reduction fixes the limiting scale analytically. Selecting a particular path through balanced six-pole solutions additionally requires following the appropriate balance root.
In the PS angular convention set ( t , ψ , ϕ ) = ( T , − Ψ , − Φ ) (t,\psi,\phi)=(T,-\Psi,-\Phi) ( t , ψ , ϕ ) = ( T , − Ψ , − Φ ) and G ( ξ ) = ( 1 − ξ 2 ) ( 1 + λ ξ + ν ξ 2 ) G(\xi)=(1-\xi^2)(1+\lambda\xi+\nu\xi^2) G ( ξ ) = ( 1 − ξ 2 ) ( 1 + λ ξ + ν ξ 2 ) . The explicit map to ring coordinates is
ρ P S 2 = − 4 k 4 G ( x r ) G ( y r ) ( 1 − ν ) 2 ( x r − y r ) 4 , z P S = k 2 ( 1 − x r y r ) [ 2 + λ ( x r + y r ) + 2 ν x r y r ] ( 1 − ν ) ( x r − y r ) 2 . (126) \rho_{\mathrm{PS}}^2=\frac{-4k^4G(x_r)G(y_r)}{(1-\nu)^2(x_r-y_r)^4},\qquad z_{\mathrm{PS}}=\frac{k^2(1-x_ry_r)[2+\lambda(x_r+y_r)+2\nu x_ry_r]}{(1-\nu)(x_r-y_r)^2}. \tag*{(126)} ρ PS 2 = ( 1 − ν ) 2 ( x r − y r ) 4 − 4 k 4 G ( x r ) G ( y r ) , z PS = ( 1 − ν ) ( x r − y r ) 2 k 2 ( 1 − x r y r ) [ 2 + λ ( x r + y r ) + 2 ν x r y r ] . (126)
Here x r , y r x_r,y_r x r , y r are ring coordinates, distinct from the dressing variable x = ρ 2 x=\rho^2 x = ρ 2 . For brevity write them as X , Y X,Y X , Y in the next formulae. The target metric in these coordinates is [3 , 27 ]
d s P S 2 = − H ( Y , X ) H ( X , Y ) ( d t − ω ψ d ψ − ω ϕ d ϕ ) 2 − F ( X , Y ) H ( Y , X ) d ψ 2 − 2 J ( X , Y ) H ( Y , X ) d ψ d ϕ + F ( Y , X ) H ( Y , X ) d ϕ 2 + 2 k 2 H ( X , Y ) ( 1 − ν ) 2 ( X − Y ) 2 [ d X 2 G ( X ) − d Y 2 G ( Y ) ] , (127) ds_{\mathrm{PS}}^{2}=-\frac{H(Y,X)}{H(X,Y)}(dt-\omega_{\psi}d\psi-\omega_{\phi}d\phi)^{2}-\frac{F(X,Y)}{H(Y,X)}d\psi^{2}-\frac{2J(X,Y)}{H(Y,X)}d\psi d\phi+\frac{F(Y,X)}{H(Y,X)}d\phi^{2}+\frac{2k^{2}H(X,Y)}{(1-\nu)^{2}(X-Y)^{2}}\left[\frac{dX^{2}}{G(X)}-\frac{dY^{2}}{G(Y)}\right], \tag*{(127)} d s PS 2 = − H ( X , Y ) H ( Y , X ) ( d t − ω ψ d ψ − ω ϕ d ϕ ) 2 − H ( Y , X ) F ( X , Y ) d ψ 2 − H ( Y , X ) 2 J ( X , Y ) d ψ d ϕ + H ( Y , X ) F ( Y , X ) d ϕ 2 + ( 1 − ν ) 2 ( X − Y ) 2 2 k 2 H ( X , Y ) [ G ( X ) d X 2 − G ( Y ) d Y 2 ] , (127)
H ( X , Y ) = 1 + λ 2 − ν 2 + 2 λ ( X + ν Y ) ( 1 − ν X Y ) + ν ( 1 − λ 2 − ν 2 ) X 2 Y 2 , (128) H(X,Y)=1+\lambda^{2}-\nu^{2}+2\lambda(X+\nu Y)(1-\nu XY)+\nu(1-\lambda^{2}-\nu^{2})X^{2}Y^{2}, \tag*{(128)} H ( X , Y ) = 1 + λ 2 − ν 2 + 2 λ ( X + ν Y ) ( 1 − ν X Y ) + ν ( 1 − λ 2 − ν 2 ) X 2 Y 2 , (128)
J ( X , Y ) = 2 k 2 λ ν ( 1 − X 2 ) ( 1 − Y 2 ) ( 1 − ν ) 2 ( X − Y ) [ 1 + λ 2 − ν 2 + 2 λ ν ( X + Y ) − ν ( 1 − λ 2 − ν 2 ) X Y ] , (129) J(X,Y)=\frac{2k^{2}\lambda\sqrt{\nu}(1-X^{2})(1-Y^{2})}{(1-\nu)^{2}(X-Y)}\left[1+\lambda^{2}-\nu^{2}+2\lambda\nu(X+Y)-\nu(1-\lambda^{2}-\nu^{2})XY\right], \tag*{(129)} J ( X , Y ) = ( 1 − ν ) 2 ( X − Y ) 2 k 2 λ ν ( 1 − X 2 ) ( 1 − Y 2 ) [ 1 + λ 2 − ν 2 + 2 λ ν ( X + Y ) − ν ( 1 − λ 2 − ν 2 ) X Y ] , (129)
F ( X , Y ) = 2 k 2 ( 1 − ν ) 2 ( X − Y ) 2 { G ( X ) ( 1 − Y 2 ) Q 1 ( Y ) + G ( Y ) P ( X ) } , (130) F(X,Y)=\frac{2k^{2}}{(1-\nu)^{2}(X-Y)^{2}}\left\{G(X)(1-Y^{2})Q_{1}(Y)+G(Y)P(X)\right\}, \tag*{(130)} F ( X , Y ) = ( 1 − ν ) 2 ( X − Y ) 2 2 k 2 { G ( X ) ( 1 − Y 2 ) Q 1 ( Y ) + G ( Y ) P ( X ) } , (130)
q = ( 1 + ν ) [ ( 1 − ν ) 2 − λ 2 ] , Q 1 ( Y ) = q + λ ( 1 − λ 2 + 2 ν − 3 ν 2 ) Y , q=(1+\nu)\left[(1-\nu)^{2}-\lambda^{2}\right],\qquad Q_{1}(Y)=q+\lambda(1-\lambda^{2}+2\nu-3\nu^{2})Y, q = ( 1 + ν ) [ ( 1 − ν ) 2 − λ 2 ] , Q 1 ( Y ) = q + λ ( 1 − λ 2 + 2 ν − 3 ν 2 ) Y ,
P ( X ) = 2 λ 2 + λ [ ( 1 − ν ) 2 + λ 2 ] X + q X 2 + λ [ 1 − λ 2 − 3 ν 2 + 2 ν 3 ] X 3 + ν ( 1 − ν ) ( 1 − λ 2 − ν 2 ) X 4 , (131) P(X)=2\lambda^{2}+\lambda\left[(1-\nu)^{2}+\lambda^{2}\right]X+qX^{2}+\lambda\left[1-\lambda^{2}-3\nu^{2}+2\nu^{3}\right]X^{3}+\nu(1-\nu)(1-\lambda^{2}-\nu^{2})X^{4}, \tag*{(131)} P ( X ) = 2 λ 2 + λ [ ( 1 − ν ) 2 + λ 2 ] X + q X 2 + λ [ 1 − λ 2 − 3 ν 2 + 2 ν 3 ] X 3 + ν ( 1 − ν ) ( 1 − λ 2 − ν 2 ) X 4 , (131)
ω ψ = 2 k λ ( 1 + Y ) H ( Y , X ) 1 + λ + ν 1 − λ + ν [ 1 + λ − ν + 2 ν X ( 1 − Y ) + ν ( 1 − λ − ν ) X 2 Y ] , (132) \omega_{\psi}=\frac{2k\lambda(1+Y)}{H(Y,X)}\sqrt{\frac{1+\lambda+\nu}{1-\lambda+\nu}}\left[1+\lambda-\nu+2\nu X(1-Y)+\nu(1-\lambda-\nu)X^{2}Y\right], \tag*{(132)} ω ψ = H ( Y , X ) 2 k λ ( 1 + Y ) 1 − λ + ν 1 + λ + ν [ 1 + λ − ν + 2 ν X ( 1 − Y ) + ν ( 1 − λ − ν ) X 2 Y ] , (132)
ω ϕ = 2 k λ ν [ ( 1 + ν ) 2 − λ 2 ] H ( Y , X ) ( 1 − X 2 ) Y . (133) \omega_{\phi}=\frac{2k\lambda\sqrt{\nu\left[(1+\nu)^{2}-\lambda^{2}\right]}}{H(Y,X)}(1-X^{2})Y. \tag*{(133)} ω ϕ = H ( Y , X ) 2 k λ ν [ ( 1 + ν ) 2 − λ 2 ] ( 1 − X 2 ) Y . (133)
This also specifies the orbit-space metric completely. The comparison converts it to Weyl form using f P S = g X X / [ ( ∂ X ρ P S ) 2 + ( ∂ X z P S ) 2 ] f_{\mathrm{PS}}=g_{XX}/[(\partial_{X}\rho_{\mathrm{PS}})^{2}+(\partial_{X}z_{\mathrm{PS}})^{2}] f PS = g X X / [( ∂ X ρ PS ) 2 + ( ∂ X z PS ) 2 ] , and checks f S a t u r n → δ 0 2 f P S f_{\mathrm{Saturn}}\to\delta_{0}^{2}f_{\mathrm{PS}} f Saturn → δ 0 2 f PS .
The PS charges and area, in the Saturn angular orientation, are
M P S = 3 π k 2 λ G 5 ( 1 + ν − λ ) , (134) M_{\mathrm{PS}}=\frac{3\pi k^{2}\lambda}{G_{5}(1+\nu-\lambda)}, \tag*{(134)} M PS = G 5 ( 1 + ν − λ ) 3 π k 2 λ , (134)
J ψ P S = − 2 π k 3 λ ( 1 + ν ) 2 − λ 2 [ 1 + λ − 6 ν + λ ν + ν 2 ] G 5 ( 1 − ν ) 2 ( 1 + ν − λ ) 2 , (135) J_{\psi}^{\mathrm{PS}}=-\frac{2\pi k^{3}\lambda\sqrt{(1+\nu)^{2}-\lambda^{2}}[1+\lambda-6\nu+\lambda\nu+\nu^{2}]}{G_{5}(1-\nu)^{2}(1+\nu-\lambda)^{2}}, \tag*{(135)} J ψ PS = − G 5 ( 1 − ν ) 2 ( 1 + ν − λ ) 2 2 π k 3 λ ( 1 + ν ) 2 − λ 2 [ 1 + λ − 6 ν + λ ν + ν 2 ] , (135)
J ϕ P S = − 4 π k 3 λ ν [ ( 1 + ν ) 2 − λ 2 ] G 5 ( 1 − ν ) 2 ( 1 + ν − λ ) , (136) J_{\phi}^{\mathrm{PS}}=-\frac{4\pi k^{3}\lambda\sqrt{\nu[(1+\nu)^{2}-\lambda^{2}]}}{G_{5}(1-\nu)^{2}(1+\nu-\lambda)}, \tag*{(136)} J ϕ PS = − G 5 ( 1 − ν ) 2 ( 1 + ν − λ ) 4 π k 3 λ ν [( 1 + ν ) 2 − λ 2 ] , (136)
A P S = 32 π 2 k 3 λ ( 1 + λ + ν ) ( 1 − ν ) 2 ( y h − 1 − y h ) , y h = − λ + λ 2 − 4 ν 2 ν . (137) A_{\mathrm{PS}}=\frac{32\pi^{2}k^{3}\lambda(1+\lambda+\nu)}{(1-\nu)^{2}(y_{h}^{-1}-y_{h})},\qquad y_{h}=\frac{-\lambda+\sqrt{\lambda^{2}-4\nu}}{2\nu}. \tag*{(137)} A PS = ( 1 − ν ) 2 ( y h − 1 − y h ) 32 π 2 k 3 λ ( 1 + λ + ν ) , y h = 2 ν − λ + λ 2 − 4 ν . (137)
The ring domain is − 1 ≤ X ≤ 1 -1\le X\le1 − 1 ≤ X ≤ 1 , y h ≤ Y < − 1 y_{h}\le Y<-1 y h ≤ Y < − 1 , with 0 ≤ ν < 1 0\le\nu<1 0 ≤ ν < 1 , 2 ν ≤ λ < 1 + ν 2\sqrt{\nu}\le\lambda<1+\nu 2 ν ≤ λ < 1 + ν . These formulas include the extremal boundary by continuity, using the nonsingular expression y h = − 2 / ( λ + λ 2 − 4 ν ) y_{h}=-2/(\lambda+\sqrt{\lambda^{2}-4\nu}) y h = − 2/ ( λ + λ 2 − 4 ν ) . The horizon entropy is A P S / ( 4 G 5 ) A_{\mathrm{PS}}/(4G_{5}) A PS / ( 4 G 5 ) and the retained ring Komar integrals equal the ADM charges.
The singly spinning black Saturn
The six-pole metric admits the general singly spinning Saturn of [7 ] , including a spinning central hole, as an explicit auxiliary-pole boundary. “Singly spinning” here means one rotation plane. Both components can carry spin in that plane.
Parameters and balance
Keep the spectral gauge a 1 = 0 a_1 = 0 a 1 = 0 , b s p e c t r a l = − 1 b_{\mathrm{spectral}} = -1 b spectral = − 1 , and take
0 < a 2 < a 3 < a 4 < a 5 , a 6 = a 5 + ϵ 2 , ϵ → 0 + . (138) 0 < a_2 < a_3 < a_4 < a_5,\qquad a_6 = a_5 + \epsilon^2,\qquad\epsilon\to0^+. \tag*{(138)} 0 < a 2 < a 3 < a 4 < a 5 , a 6 = a 5 + ϵ 2 , ϵ → 0 + . (138)
B 5 = ϵ b + o ( ϵ ) , C 6 = γ / ϵ + O ( ϵ ) , γ 2 = a 5 2 ( a 5 − a 2 ) . (139) B_5 = \epsilon\,\mathfrak{b} + o(\epsilon),\qquad C_6 = \gamma/\epsilon+ O(\epsilon),\qquad\gamma^2 = \frac{a_5}{2(a_5-a_2)}. \tag*{(139)} B 5 = ϵ b + o ( ϵ ) , C 6 = γ / ϵ + O ( ϵ ) , γ 2 = 2 ( a 5 − a 2 ) a 5 . (139)
C 2 ⟶ 0 , C 3 ⟶ 0 , B 4 ⟶ 0 , C 1 ⟶ − a 3 2 a 2 a 4 a 5 2 . (140) C_2 \longrightarrow0,\qquad C_3 \longrightarrow0,\qquad B_4 \longrightarrow0,\qquad C_1 \longrightarrow-\sqrt{\frac{a_3}{2a_2a_4a_5^2}}. \tag*{(140)} C 2 ⟶ 0 , C 3 ⟶ 0 , B 4 ⟶ 0 , C 1 ⟶ − 2 a 2 a 4 a 5 2 a 3 . (140)
C 1 , C 6 C_1,C_6 C 1 , C 6 retain their exact phantom-removal values before taking the limit. The last three zero conditions are imposed on the finite auxiliary-limit metric. We can also approach them continuously while the other scaled variables remain finite. Both physical horizon intervals, [ a 2 , a 3 ] [a_2,a_3] [ a 2 , a 3 ] and [ a 4 , a 5 ] [a_4,a_5] [ a 4 , a 5 ] , remain nonzero. Only the auxiliary interval collapses. Setting those three polarizations to zero at a fixed nonzero auxiliary interval is not the reduction asserted here.
We use the label EF for the solution of [7 ] and denote its endpoints by A i A_i A i , to distinguish their ordering from ours. The complete map is
( A 1 , A 2 , A 3 , A 4 , A 5 ) = ( 0 , a 5 , a 4 , a 3 , a 2 ) . (141) (A_1,A_2,A_3,A_4,A_5)=(0,a_5,a_4,a_3,a_2). \tag*{(141)} ( A 1 , A 2 , A 3 , A 4 , A 5 ) = ( 0 , a 5 , a 4 , a 3 , a 2 ) . (141)
c 1 E F = − 2 a 4 a 5 C 1 = 2 a 3 a 4 a 2 , c 2 E F = − 2 ( a 5 + 1 ) ( a 5 − a 3 ) a 5 b γ . (142) c_1^{\mathrm{EF}}=-2a_4a_5C_1=\sqrt{\frac{2a_3a_4}{a_2}},\qquad c_2^{\mathrm{EF}}=-\frac{2(a_5+1)(a_5-a_3)}{a_5}\,\mathfrak{b}\gamma. \tag*{(142)} c 1 EF = − 2 a 4 a 5 C 1 = a 2 2 a 3 a 4 , c 2 EF = − a 5 2 ( a 5 + 1 ) ( a 5 − a 3 ) b γ . (142)
L 2 = a 5 , κ 1 = a 4 a 5 , κ 2 = a 3 a 5 , κ 3 = a 2 a 5 , c ˉ 2 = c 2 E F c 1 E F ( 1 − κ 2 ) . (143) L^2=a_5,\qquad\kappa_1=\frac{a_4}{a_5},\qquad\kappa_2=\frac{a_3}{a_5},\qquad\kappa_3=\frac{a_2}{a_5},\qquad\bar{c}_2=\frac{c_2^{\mathrm{EF}}}{c_1^{\mathrm{EF}}(1-\kappa_2)}. \tag*{(143)} L 2 = a 5 , κ 1 = a 5 a 4 , κ 2 = a 5 a 3 , κ 3 = a 5 a 2 , c ˉ 2 = c 1 EF ( 1 − κ 2 ) c 2 EF . (143)
Thus 0 < κ 3 < κ 2 < κ 1 < 1 0 < \kappa_3 < \kappa_2 < \kappa_1 < 1 0 < κ 3 < κ 2 < κ 1 < 1 . With N = 1 + κ 2 c ˉ 2 ≠ 0 N=1+\kappa_2\bar{c}_2\ne0 N = 1 + κ 2 c ˉ 2 = 0 , the asymptotic constants are
k E F = 1 ∣ N ∣ , q E F = c 1 E F c ˉ 2 N . (144) k_{\mathrm{EF}}=\frac{1}{\lvert N\rvert},\qquad q_{\mathrm{EF}}=\frac{c_1^{\mathrm{EF}}\bar{c}_2}{N}. \tag*{(144)} k EF = ∣ N ∣ 1 , q EF = N c 1 EF c ˉ 2 . (144)
The EF balance condition, fixing b \mathfrak{b} b through this map, is
c ˉ 2 = 1 κ 2 [ s ( κ 1 − κ 2 ) κ 1 ( 1 − κ 2 ) ( 1 − κ 3 ) ( κ 1 − κ 3 ) − 1 ] , s = ± 1. (145) \bar{c}_2=\frac{1}{\kappa_2}\left[\frac{s(\kappa_1-\kappa_2)}{\sqrt{\kappa_1(1-\kappa_2)(1-\kappa_3)(\kappa_1-\kappa_3)}}-1\right],\qquad s=\pm1. \tag*{(145)} c ˉ 2 = κ 2 1 [ κ 1 ( 1 − κ 2 ) ( 1 − κ 3 ) ( κ 1 − κ 3 ) s ( κ 1 − κ 2 ) − 1 ] , s = ± 1. (145)
The usual positive-Komar-mass branch is s = + 1 s=+1 s = + 1 . The metric identification itself holds before balance and on both balanced branches wherever the normalization is nonsingular.
Metric identity and normalisation
The permutation (A.49) defines the EF pole functions μ i \mu_i μ i . We use the functions H x , H y , P , G x , G y , ω ψ H_x,H_y,P,G_x,G_y,\omega_\psi H x , H y , P , G x , G y , ω ψ given in section 2 of [7 ] . In particular,
G x = ρ 2 μ 4 μ 3 μ 5 , G y = μ 3 μ 5 μ 4 , G_x=\frac{\rho^2\mu_4}{\mu_3\mu_5},\qquad G_y=\frac{\mu_3\mu_5}{\mu_4}, G x = μ 3 μ 5 ρ 2 μ 4 , G y = μ 4 μ 3 μ 5 ,
P = ( ρ 2 + μ 3 μ 4 ) 2 ( ρ 2 + μ 1 μ 5 ) ( ρ 2 + μ 4 μ 5 ) . (146) P=(\rho^2+\mu_3\mu_4)^2(\rho^2+\mu_1\mu_5)(\rho^2+\mu_4\mu_5). \tag*{(146)} P = ( ρ 2 + μ 3 μ 4 ) 2 ( ρ 2 + μ 1 μ 5 ) ( ρ 2 + μ 4 μ 5 ) . (146)
Exact cancellation of the Laurent monomials gives, in the intermediate Killing coordinates,
f r a w = H x P , ( G t t , G t ψ , G t ϕ , G ψ ψ , G ϕ ϕ ) = ( − H y H x ω ψ 2 H x , 0 , 0 , G x ) . (147) f_{\rm raw}=H_xP,\qquad(G_{tt},G_{t\psi},G_{t\phi},G_{\psi\psi},G_{\phi\phi})=\left(-\frac{H_y}{H_x}\frac{\omega_\psi^2}{H_x},0,0,G_x\right). \tag*{(147)} f raw = H x P , ( G tt , G t ψ , G tϕ , G ψ ψ , G ϕϕ ) = ( − H x H y H x ω ψ 2 , 0 , 0 , G x ) . (147)
G ψ ψ = H x G y H y − ω ψ 2 H x H y . (148) G_{\psi\psi}=\frac{H_xG_y}{H_y}-\frac{\omega_\psi^2}{H_xH_y}. \tag*{(148)} G ψ ψ = H y H x G y − H x H y ω ψ 2 . (148)
The last identity also follows from det G = − ρ 2 \det G=-\rho^2 det G = − ρ 2 and extends through H y = 0 H_y=0 H y = 0 by the regular rational expression for the metric. One can verify the identity by expanding the published EF functions and collecting pole monomials with
( u i − u j ) ( ρ 2 + u i u j ) = 2 ( a i − a j ) u i u j . (149) (u_i-u_j)(\rho^2+u_iu_j)=2(a_i-a_j)u_iu_j. \tag*{(149)} ( u i − u j ) ( ρ 2 + u i u j ) = 2 ( a i − a j ) u i u j . (149)
The coefficients agree with variable endpoints and polarisations.
In the canonical physical normalisation (26) the outer-axis matrix has δ = det A = N 2 \delta=\det A=N^2 δ = det A = N 2 , and its unit-determinant version is
B = ( 1 − q E F 0 0 1 0 0 0 1 ) , f c a n = f r a w / N 2 . (150) B=\begin{pmatrix}
1&-\mathord{q}_{\rm EF}&0\\
0&1&0\\
0&0&1
\end{pmatrix},\qquad f_{\rm can}=f_{\rm raw}/N^2. \tag*{(150)} B = 1 0 0 − q EF 1 0 0 0 1 , f can = f raw / N 2 . (150)
Then t E F = T t_{\rm EF}=T t EF = T , ψ E F = − Ψ \psi_{\rm EF}=-\Psi ψ EF = − Ψ , ϕ E F = Φ \phi_{\rm EF}=\Phi ϕ EF = Φ , with the same ρ , z \rho,z ρ , z , gives
d s 2 = − H y H x [ d t E F + ( ω ψ H y + q E F ) d ψ E F ] 2 + H x G y H y d ψ E F 2 + G x d ϕ E F 2 + k E F 2 H x P ( d ρ 2 + d z 2 ) . (151) ds^2=-\frac{H_y}{H_x}\left[dt_{\rm EF}+\left(\frac{\omega_\psi}{H_y}+q_{\rm EF}\right)d\psi_{\rm EF}\right]^2+\frac{H_xG_y}{H_y}d\psi_{\rm EF}^2+G_xd\phi_{\rm EF}^2+k_{\rm EF}^2H_xP(d\rho^2+dz^2). \tag*{(151)} d s 2 = − H x H y [ d t EF + ( H y ω ψ + q EF ) d ψ EF ] 2 + H y H x G y d ψ EF 2 + G x d ϕ EF 2 + k EF 2 H x P ( d ρ 2 + d z 2 ) . (151)
Equivalently, t r a w = t E F + q E F ψ E F t_{\rm raw}=t_{\rm EF}+q_{\rm EF}\psi_{\rm EF} t raw = t EF + q EF ψ EF and ψ r a w = − ψ E F \psi_{\rm raw}=-\psi_{\rm EF} ψ raw = − ψ EF . This fixes the angular-momentum sign convention and the conformal normalization explicitly.
Charges and entropy
Here write b = c ˉ 2 b=\bar{c}_2 b = c ˉ 2 only in the following thermodynamic formulas, and set G 5 = 1 G_5=1 G 5 = 1 . Define
H = κ 3 ( 1 − κ 1 ) + κ 1 κ 2 ( 1 − κ 2 ) ( 1 − κ 3 ) b 2 , (152) H=\kappa_3(1-\kappa_1)+\kappa_1\kappa_2(1-\kappa_2)(1-\kappa_3)b^2, \tag*{(152)} H = κ 3 ( 1 − κ 1 ) + κ 1 κ 2 ( 1 − κ 2 ) ( 1 − κ 3 ) b 2 , (152)
R = κ 3 − κ 3 ( κ 1 − κ 2 ) b + κ 1 κ 2 ( 1 − κ 3 ) b 2 , (153) R=\kappa_3-\kappa_3(\kappa_1-\kappa_2)b+\kappa_1\kappa_2(1-\kappa_3)b^2, \tag*{(153)} R = κ 3 − κ 3 ( κ 1 − κ 2 ) b + κ 1 κ 2 ( 1 − κ 3 ) b 2 , (153)
S = κ 3 − κ 2 ( κ 1 − κ 3 ) b + κ 1 κ 2 ( 1 − κ 2 ) b 2 . (154) S=\kappa_3-\kappa_2(\kappa_1-\kappa_3)b+\kappa_1\kappa_2(1-\kappa_2)b^2. \tag*{(154)} S = κ 3 − κ 2 ( κ 1 − κ 3 ) b + κ 1 κ 2 ( 1 − κ 2 ) b 2 . (154)
In the EF orientation the individual charges become exactly
M H = 3 π L 2 H 4 κ 3 N , M R = 3 π L 2 κ 2 [ 1 − ( 1 − κ 2 ) b ] R 4 κ 3 N 2 , (155) M_{\rm H}=\frac{3\pi L^2H}{4\kappa_3N},\qquad M_{\rm R}=\frac{3\pi L^2\kappa_2[1-(1-\kappa_2)b]R}{4\kappa_3N^2}, \tag*{(155)} M H = 4 κ 3 N 3 π L 2 H , M R = 4 κ 3 N 2 3 π L 2 κ 2 [ 1 − ( 1 − κ 2 ) b ] R , (155)
J H E F = − π L 3 b H κ 3 N 2 κ 1 κ 2 2 κ 3 , J R E F = π L 3 S R κ 3 N 3 κ 2 2 κ 1 κ 3 , (156) J_{\rm H}^{\rm EF}=-\frac{\pi L^3bH}{\kappa_3N^2}\sqrt{\frac{\kappa_1\kappa_2}{2\kappa_3}},\qquad J_{\rm R}^{\rm EF}=\frac{\pi L^3SR}{\kappa_3N^3}\sqrt{\frac{\kappa_2}{2\kappa_1\kappa_3}}, \tag*{(156)} J H EF = − κ 3 N 2 π L 3 b H 2 κ 3 κ 1 κ 2 , J R EF = κ 3 N 3 π L 3 S R 2 κ 1 κ 3 κ 2 , (156)
M A D M = M R + M H , J A D M E F = J R E F + J H E F . (157) M_{\rm ADM}=M_{\rm R}+M_{\rm H},\qquad J_{\rm ADM}^{\rm EF}=J_{\rm R}^{\rm EF}+J_{\rm H}^{\rm EF}. \tag*{(157)} M ADM = M R + M H , J ADM EF = J R EF + J H EF . (157)
All individual and ADM transverse angular momenta vanish, as do both transverse horizon angular velocities. Our canonical J Ψ J_{\Psi} J Ψ is minus the displayed EF J J J , owing to the single common reflection above. Restoring G 5 G_{5} G 5 divides every displayed mass and angular momentum by G 5 G_{5} G 5 . The areas are
A H = 4 π 2 L 3 2 ( 1 − κ 1 ) 3 ( 1 − κ 2 ) ( 1 − κ 3 ) κ 3 ( 1 − κ 1 ) N 2 H (158) A_{\mathrm{H}}=4\pi^{2}L^{3}\sqrt{\frac{2(1-\kappa_{1})^{3}}{(1-\kappa_{2})(1-\kappa_{3})\kappa_{3}(1-\kappa_{1})N^{2}}}\,H \tag*{(158)} A H = 4 π 2 L 3 ( 1 − κ 2 ) ( 1 − κ 3 ) κ 3 ( 1 − κ 1 ) N 2 2 ( 1 − κ 1 ) 3 H (158)
A R = 4 π 2 L 3 2 κ 2 ( κ 2 − κ 3 ) 3 κ 1 ( κ 1 − κ 3 ) ( 1 − κ 3 ) κ 3 N 2 R , S t o t a l = A R + A H 4 G 5 . (159) A_{\mathrm{R}}=4\pi^{2}L^{3}\sqrt{\frac{2\kappa_{2}(\kappa_{2}-\kappa_{3})^{3}}{\kappa_{1}(\kappa_{1}-\kappa_{3})(1-\kappa_{3})\kappa_{3}N^{2}}}\,R,\qquad S_{\mathrm{total}}=\frac{A_{\mathrm{R}}+A_{\mathrm{H}}}{4G_{5}}. \tag*{(159)} A R = 4 π 2 L 3 κ 1 ( κ 1 − κ 3 ) ( 1 − κ 3 ) κ 3 N 2 2 κ 2 ( κ 2 − κ 3 ) 3 R , S total = 4 G 5 A R + A H . (159)
These are precisely EF equations (3.26)–(3.40). The normalized metric isometry identifies the integration surfaces, Killing fields and volume form, so it identifies each horizon Komar integral, not just their sum. Direct asymptotic expansion independently reproduces the EF ADM formulas. Horizon quadratures at five balanced configurations, including both signs in (A.53), agree with the individual charges and areas to better than 5 × 10 − 31 5 \times10^{-31} 5 × 1 0 − 31 in normalised error.
To approach this limit through solutions that remain in balance, we fix ( a 1 , … , a 5 ) = ( 0 , 1 , 2 , 4 , 5 ) (a_{1},\ldots,a_{5})=(0,1,2,4,5) ( a 1 , … , a 5 ) = ( 0 , 1 , 2 , 4 , 5 ) and C 2 = 0 C_{2}=0 C 2 = 0 , and solve the axis equations (16) at each nonzero ϵ \epsilon ϵ . This requires corrections C 3 , B 4 = O ( ϵ ) C_{3},B_{4}=O(\epsilon) C 3 , B 4 = O ( ϵ ) . The transverse spins vanish as O ( ϵ ) O(\epsilon) O ( ϵ ) . Masses, original spins and areas converge as O ( ϵ 2 ) O(\epsilon^{2}) O ( ϵ 2 ) . Thus the limiting identities and the approach through exactly balanced finite-pole metrics are consistent.
Solutions with one extremal horizon
There are two distinct mixed-extremal boundaries: a zero-temperature ring surrounding a nonextremal hole, and a nonextremal ring surrounding a zero-temperature hole. We give the metric of each of these cases and regularity checks.
An extremal ring and a nonextremal central hole
Choose real constants
0 < a < b , 0 < A < B < C , k , q , u , v ∈ R , q ≠ 0. (160) 0<a<b,\qquad0<A<B<C,\qquad k,q,u,v\in\mathbb{R},\qquad q\ne0. \tag*{(160)} 0 < a < b , 0 < A < B < C , k , q , u , v ∈ R , q = 0. (160)
For an auxiliary parameter ϵ > 0 \epsilon>0 ϵ > 0 , prescribe
( a 1 , … , a 6 ) = ( 0 , a ϵ , b ϵ , A , B , C ) , (a_{1},\ldots,a_{6})=(0,a\epsilon,b\epsilon,A,B,C), ( a 1 , … , a 6 ) = ( 0 , a ϵ , b ϵ , A , B , C ) ,
C 1 = − b 2 a A B C , C 6 = C 2 ( C − a ϵ ) ( C − B ) , C_{1}=-\sqrt{\frac{b}{2aABC}},\qquad C_{6}=\sqrt{\frac{C}{2(C-a\epsilon)(C-B)}}, C 1 = − 2 a A B C b , C 6 = 2 ( C − a ϵ ) ( C − B ) C ,
C 2 = q , C 3 = 1 − k ϵ C 1 q , B 4 = u , B 5 = v . (161) C_{2}=q,\qquad C_{3}=\frac{1-k\epsilon}{C_{1}q},\qquad B_{4}=u,\qquad B_{5}=v. \tag*{(161)} C 2 = q , C 3 = C 1 q 1 − k ϵ , B 4 = u , B 5 = v . (161)
The values of C 1 C_{1} C 1 and C 6 C_{6} C 6 satisfy (13) , which remove the apparent turning points at the auxiliary endpoints a 1 a_{1} a 1 and a 6 a_{6} a 6 . The remaining tuning gives
1 − C 1 C 2 C 3 = k ϵ . (162) 1-C_{1}C_{2}C_{3}=k\epsilon. \tag*{(162)} 1 − C 1 C 2 C 3 = k ϵ . (162)
The ring rod shrinks to z = 0 z=0 z = 0 , while the hole retains the rod [A, B]. Equation (A.70) retains the ring’s finite-area throat. The common rescaling ( a , b , k ) ↦ s ( a , b , k ) (a,b,k)\mapsto s(a,b,k) ( a , b , k ) ↦ s ( a , b , k ) only reparametrises ϵ \epsilon ϵ , so we may set a = 1 a=1 a = 1 . Eight constants then remain before balance. Balance is imposed on the limiting metric. To explicitly construct the metric we use x = ρ 2 x=\rho^{2} x = ρ 2 , u i = x + ( z − a i ) 2 − ( z − a i ) u_i=\sqrt{x+(z-a_i)^2}-(z-a_i) u i = x + ( z − a i ) 2 − ( z − a i ) , the rows M M M and ratios r i r_i r i of (B.2)–(B.3), and
H = diag ( − u 1 u 3 u 4 u 6 , − u 1 u 2 u 5 x , − u 2 u 3 u 5 u 4 u 6 ) , H=\operatorname{diag}\left(-\frac{u_1u_3}{u_4u_6},-\frac{u_1u_2u_5}{x},-\frac{u_2u_3u_5}{u_4u_6}\right), H = diag ( − u 4 u 6 u 1 u 3 , − x u 1 u 2 u 5 , − u 4 u 6 u 2 u 3 u 5 ) ,
( p 1 , … , p 6 ) = ( u 1 , u 2 , u 3 , − x / u 4 , u 5 , − x / u 6 ) , (p_1,\ldots,p_6)=(u_1,u_2,u_3,-x/u_4,u_5,-x/u_6), ( p 1 , … , p 6 ) = ( u 1 , u 2 , u 3 , − x / u 4 , u 5 , − x / u 6 ) ,
Γ i j = M i H M j T x + p i p j , N i α = M i α H α α p i . (163) \Gamma_{ij}=\frac{M_iHM_j^{\mathsf T}}{x+p_ip_j},\qquad N_{i\alpha}=\frac{M_{i\alpha}H_{\alpha\alpha}}{p_i}. \tag*{(163)} Γ ij = x + p i p j M i H M j T , N i α = p i M i α H α α . (163)
All coefficients needed below have degree at most two in ϵ \epsilon ϵ . For a i = a i 0 + s i ϵ a_i=a_i^0+s_i\epsilon a i = a i 0 + s i ϵ , d i = z − a i 0 d_i=z-a_i^0 d i = z − a i 0 and D i = x + d i 2 D_i=\sqrt{x+d_i^2} D i = x + d i 2 , they follow from
u i = D i − d i + s i ( 1 − d i / D i ) ϵ + s i 2 x 2 D i 3 ϵ 2 + O ( ϵ 3 ) , u_i=D_i-d_i+s_i(1-d_i/D_i)\epsilon+\frac{s_i^2x}{2D_i^3}\epsilon^2+O(\epsilon^3), u i = D i − d i + s i ( 1 − d i / D i ) ϵ + 2 D i 3 s i 2 x ϵ 2 + O ( ϵ 3 ) ,
C 6 = 1 2 ( C − B ) ( 1 + a ϵ 2 C + 3 a 2 ϵ 2 8 C 2 ) + O ( ϵ 3 ) , ( s i ) = ( 0 , a , b , 0 , 0 , 0 ) . (164) C_6=\frac{1}{\sqrt{2(C-B)}}\left(1+\frac{a\epsilon}{2C}+\frac{3a^2\epsilon^2}{8C^2}\right)+O(\epsilon^3),\qquad(s_i)=(0,a,b,0,0,0). \tag*{(164)} C 6 = 2 ( C − B ) 1 ( 1 + 2 C a ϵ + 8 C 2 3 a 2 ϵ 2 ) + O ( ϵ 3 ) , ( s i ) = ( 0 , a , b , 0 , 0 , 0 ) . (164)
We cancel the endpoint differences in the ratios r i r_i r i before expanding. At degree zero we write
M 1 ( 0 ) = ( α , 1 , 0 ) , M 2 ( 0 ) = ( 0 , β , 1 ) , M 3 ( 0 ) = ( 1 , 0 , γ ) , α β γ = − 1. (165) M_1^{(0)}=(\alpha,1,0),\qquad M_2^{(0)}=(0,\beta,1),\qquad M_3^{(0)}=(1,0,\gamma),\qquad\alpha\beta\gamma=-1. \tag*{(165)} M 1 ( 0 ) = ( α , 1 , 0 ) , M 2 ( 0 ) = ( 0 , β , 1 ) , M 3 ( 0 ) = ( 1 , 0 , γ ) , α β γ = − 1. (165)
These three symbols are local row coefficients, not the parameters of § 4 . Let P P P be the identity except for P 3 i = ( − α − 1 , ( α β ) − 1 , 1 , 0 , 0 , 0 ) i P_{3i}=(-\alpha^{-1},(\alpha\beta)^{-1},1,0,0,0)_i P 3 i = ( − α − 1 , ( α β ) − 1 , 1 , 0 , 0 , 0 ) i , and let σ i = δ i 3 \sigma_i=\delta_{i3} σ i = δ i 3 . Hold P P P fixed when extracting coefficients:
Γ ^ i j = [ ϵ σ i + σ j ] ( P Γ P T ) i j , N ^ i α = [ ϵ σ i ] ( P N ) i α , M ^ i α = [ ϵ σ i ] ( P M ) i α . (166) \widehat{\Gamma}_{ij}=[\epsilon^{\sigma_i+\sigma_j}](P\Gamma P^{\mathsf T})_{ij},\qquad\widehat{N}_{i\alpha}=[\epsilon^{\sigma_i}](PN)_{i\alpha},\qquad\widehat{M}_{i\alpha}=[\epsilon^{\sigma_i}](PM)_{i\alpha}. \tag*{(166)} Γ ij = [ ϵ σ i + σ j ] ( P Γ P T ) ij , N i α = [ ϵ σ i ] ( P N ) i α , M i α = [ ϵ σ i ] ( P M ) i α . (166)
With Γ r e f \Gamma_{\mathrm{ref}} Γ ref formed from the same rows with all six polarisations zero, the raw limiting metric is
G ∗ = H ( 0 ) − N ^ T Γ ^ − 1 N ^ , f ∗ = f s e e d ( 0 ) det Γ ^ det Γ r e f ( 0 ) . (167) G_* = H^{(0)}-\widehat{N}^{\mathsf T}\widehat{\Gamma}^{-1}\widehat{N},\qquad f_* = f_{\mathrm{seed}}^{(0)}\frac{\det\widehat{\Gamma}}{\det\Gamma_{\mathrm{ref}}^{(0)}}. \tag*{(167)} G ∗ = H ( 0 ) − N T Γ − 1 N , f ∗ = f seed ( 0 ) det Γ ref ( 0 ) det Γ . (167)
For completeness, in these row conventions the seed factor is
f s e e d = u 2 u 5 u 1 ∏ i ≤ j ( x + u i u j ) − w i ⋅ w j / ( 1 + δ i j ) , f_{\mathrm{seed}}=\frac{u_2u_5}{u_1}\prod_{i\leq j}(x+u_iu_j)^{-\mathbf{w}_i\cdot\mathbf{w}_j/(1+\delta_{ij})}, f seed = u 1 u 2 u 5 i ≤ j ∏ ( x + u i u j ) − w i ⋅ w j / ( 1 + δ ij ) ,
w 1 = ( 1 , − 1 , 0 ) , w 2 = w 5 = ( 0 , 1 , − 1 ) , w 3 = w 6 = ( − 1 , 0 , 1 ) , w 4 = ( 1 , 0 , − 1 ) . (168) \mathbf{w}_1=(1,-1,0),\qquad\mathbf{w}_2=\mathbf{w}_5=(0,1,-1),\qquad\mathbf{w}_3=\mathbf{w}_6=(-1,0,1),\qquad\mathbf{w}_4=(1,0,-1). \tag*{(168)} w 1 = ( 1 , − 1 , 0 ) , w 2 = w 5 = ( 0 , 1 , − 1 ) , w 3 = w 6 = ( − 1 , 0 , 1 ) , w 4 = ( 1 , 0 , − 1 ) . (168)
The matrix inverse in (A.75) is finite and may equivalently be written as an adjugate divided by a determinant. The row transformation diag ( 1 , 1 , ϵ − 1 , 1 , 1 , 1 ) P \operatorname{diag}(1,1,\epsilon^{-1},1,1,1)P diag ( 1 , 1 , ϵ − 1 , 1 , 1 , 1 ) P explains the construction: the third row has a first-order zero and its diagonal Gram entry a second-order zero. Thus G → G ∗ G\to G_* G → G ∗ and f / ϵ 2 → f ∗ f/\epsilon^2\to f_* f / ϵ 2 → f ∗ . The constant rescaling of f f f is made before the physical axis normalisation. This procedure removes the singular matrix inversion that would result from setting the endpoints equal at the outset.
An extremal central hole and a nonextremal ring
To find this branch, we start near the double-extremal solution and increase the ring area at fixed four local angular momenta. Decreasing the central rod length then takes an auxiliary endpoint through infinity. Continuing the physical metric across this coordinate boundary is essential. In the original BZ representation C 1 , C 3 , B 4 , C 6 C_1,C_3,B_4,C_6 C 1 , C 3 , B 4 , C 6 become imaginary and the constant transformation W = diag ( 1 , i , i ) W=\operatorname{diag}(1,i,i) W = diag ( 1 , i , i ) of the Killing coordinates, together with f ↦ − f f\mapsto-f f ↦ − f , gives a real metric. The purely real formula below implements this continuation and preserves the continued signed spins. Use eight real constants
a < − 1 , C < U < U + ℓ < 0 , ℓ > 0 , c 2 , c 3 , b , w ∈ R , a<-1,\qquad C<U<U+\ell<0,\qquad\ell>0,\qquad c_2,c_3,b,w\in\mathbb{R}, a < − 1 , C < U < U + ℓ < 0 , ℓ > 0 , c 2 , c 3 , b , w ∈ R ,
γ = − U + ℓ 2 U a ( a + 1 ) C (169) \gamma=\sqrt{-\frac{U+\ell}{2Ua(a+1)C}} \tag*{(169)} γ = − 2 U a ( a + 1 ) C U + ℓ (169)
and prescribe the following path for sufficiently small L > 0 L>0 L > 0 :
( a 1 , … , a 6 ) = ( 0 , U , U + ℓ , a L , ( a + 1 ) L , C ) , (a_1,\ldots,a_6)=(0,U,U+\ell,aL,(a+1)L,C), ( a 1 , … , a 6 ) = ( 0 , U , U + ℓ , a L , ( a + 1 ) L , C ) ,
C 1 = i γ / L , C 2 = c 2 , C 3 = i c 3 , B 4 = i b , C_1=i\gamma/L,\qquad C_2=c_2,\qquad C_3=ic_3,\qquad B_4=ib, C 1 = iγ / L , C 2 = c 2 , C 3 = i c 3 , B 4 = ib ,
B 5 = L ( − b + w L ) 2 U γ , C 6 = i C 2 ( C − U ) ( C − ( a + 1 ) L ) . (170) B_5=\frac{L(-b+wL)}{2U\gamma},\qquad C_6=i\sqrt{\frac{C}{2(C-U)(C-(a+1)L)}}. \tag*{(170)} B 5 = 2 U γ L ( − b + w L ) , C 6 = i 2 ( C − U ) ( C − ( a + 1 ) L ) C . (170)
The ring occupies [ U , U + ℓ ] [U,U+\ell] [ U , U + ℓ ] . The central rod has length L L L and collapses to z = 0 z=0 z = 0 . Both phantom equations hold identically. The critical relation is
C 1 = i γ / L , B 4 + 2 U C 1 B 5 = i w L . (171) C_1=i\gamma/L,\qquad B_4+2UC_1B_5=iwL. \tag*{(171)} C 1 = iγ / L , B 4 + 2 U C 1 B 5 = i w L . (171)
Although B 5 B_5 B 5 tends to zero, its first and second coefficients survive in the limiting metric. At L ≠ 0 L\ne0 L = 0 these variables describe all nine parameters before balance. At L = 0 L=0 L = 0 they leave eight. We use the same r i r_i r i as before, with the endpoints in (A.78). The real rows and pre-dressing matrix are
M 1 = ( r 1 γ / L , − 1 , 0 ) , M 2 = ( 0 , r 2 c 2 , 1 ) , M 3 = ( 1 , 0 , r 3 c 3 ) , M 4 = ( 1 , 0 , r 4 b ) , M_1=(r_1\gamma/L,-1,0),\qquad M_2=(0,r_2c_2,1),\qquad M_3=(1,0,r_3c_3),\qquad M_4=(1,0,r_4b), M 1 = ( r 1 γ / L , − 1 , 0 ) , M 2 = ( 0 , r 2 c 2 , 1 ) , M 3 = ( 1 , 0 , r 3 c 3 ) , M 4 = ( 1 , 0 , r 4 b ) ,
M 5 = ( 0 , r 5 B 5 , 1 ) , M 6 = ( r 6 c 6 , 0 , − 1 ) , H = diag ( − u 1 u 3 u 4 u 6 , u 1 u 2 u 5 x , u 2 u 3 u 5 u 4 u 6 ) . (172) M_5=(0,r_5B_5,1),\qquad M_6=(r_6c_6,0,-1),\qquad H=\operatorname{diag}\left(-\frac{u_1u_3}{u_4u_6},\frac{u_1u_2u_5}{x},\frac{u_2u_3u_5}{u_4u_6}\right). \tag*{(172)} M 5 = ( 0 , r 5 B 5 , 1 ) , M 6 = ( r 6 c 6 , 0 , − 1 ) , H = diag ( − u 4 u 6 u 1 u 3 , x u 1 u 2 u 5 , u 4 u 6 u 2 u 3 u 5 ) . (172)
Here c 6 c_6 c 6 is the positive real square root in (A.78), without i i i . Form Γ , N \Gamma,N Γ , N with the same poles and formulas as in (A.71). Define
S = I + ( r 5 B 5 ) E 51 , det S = 1 , Γ ˉ = [ L 0 ] ( S Γ S T ) , N ˉ = [ L 0 ] ( S N ) , S=I+(r_5B_5)E_{51},\qquad\det S=1,\qquad\bar{\Gamma}=[L^0](S\Gamma S^{\mathsf{T}}),\qquad\bar{N}=[L^0](SN), S = I + ( r 5 B 5 ) E 51 , det S = 1 , Γ ˉ = [ L 0 ] ( S Γ S T ) , N ˉ = [ L 0 ] ( S N ) ,
M ˉ = [ L 0 ] ( S M ) , H 0 = [ L 0 ] H , P 0 = [ L 0 ] ( − f s e e d det Γ r e f ) . (173) \bar{M}=[L^0](SM),\qquad H_0=[L^0]H,\qquad P_0=[L^0]\left(-\frac{f_{\mathrm{seed}}}{\det\Gamma_{\mathrm{ref}}}\right). \tag*{(173)} M ˉ = [ L 0 ] ( S M ) , H 0 = [ L 0 ] H , P 0 = [ L 0 ] ( − det Γ ref f seed ) . (173)
E 51 E_{51} E 51 has one unit entry in row five, column one. The reference matrix sets all six polarisations to zero, retaining the row signs in (A.80). Coefficients are extracted after cancelling the known Laurent powers. The result is
G ∗ = H 0 − N ˉ T Γ ˉ − 1 N ˉ , f ∗ = P 0 det Γ ˉ . (174) G_* = H_0-\bar{N}^{\mathsf{T}}\bar{\Gamma}^{-1}\bar{N},\qquad f_* = P_0\det\bar{\Gamma}. \tag*{(174)} G ∗ = H 0 − N ˉ T Γ ˉ − 1 N ˉ , f ∗ = P 0 det Γ ˉ . (174)
This is a full metric at L = 0 L = 0 L = 0 . One can see why the tuning is required before inverting the Gram matrix. Take q = lim B 5 / L q = \lim B_{5}/L q = lim B 5 / L and, at L = 0 L = 0 L = 0 , denote ( u 1 , u 2 , u 3 , u 6 ) = ( h 0 , v 0 , t 0 , k 0 ) (u_{1}, u_{2}, u_{3}, u_{6}) = (h_{0}, v_{0}, t_{0}, k_{0}) ( u 1 , u 2 , u 3 , u 6 ) = ( h 0 , v 0 , t 0 , k 0 ) . Then
lim L → 0 L { Γ 54 + ( r 5 B 5 ) Γ 14 } = − v 0 t 0 ( x + h 0 2 ) 2 k 0 x ( x + h 0 v 0 ) ( b + 2 U γ q ) . (175) \lim_{L\to0} L\{\Gamma_{54} + (r_{5}B_{5})\Gamma_{14}\} = -\frac{v_{0}t_{0}(x+h_{0}^{2})}{2k_{0}x(x+h_{0}v_{0})}(b+2U\gamma q). \tag*{(175)} L → 0 lim L { Γ 54 + ( r 5 B 5 ) Γ 14 } = − 2 k 0 x ( x + h 0 v 0 ) v 0 t 0 ( x + h 0 2 ) ( b + 2 U γ q ) . (175)
It vanishes precisely for the leading relation in (A.79). The apparent quadratic pole in the ( 5 , 5 ) (5,5) ( 5 , 5 ) entry multiplies a second difference of the Cauchy kernel, whose zero is quadratic in p 5 − p 1 = O ( L ) p_{5}-p_{1}=O(L) p 5 − p 1 = O ( L ) . Thus the displayed row operation yields finite matrices while retaining the free subleading coefficient w w w .
Physical normalisation, balance and thermodynamic charges
We now discuss the physical metrics, absense of conical singularities and thermodynamics charges. The two raw metrics above must be put in a common asymptotically flat frame before interpreting their charges. In order to do so, on the outer axes we solve G ∗ v = 0 G_{*}v=0 G ∗ v = 0 , normalising v L ϕ = 1 v_{L}^{\phi}=1 v L ϕ = 1 and v R ψ = 1 v_{R}^{\psi}=1 v R ψ = 1 . For i , j i,j i , j complementary to the component fixed to one, we set
Q v = − [ f ∗ ( G ∗ i i G ∗ j j − G ∗ i j 2 ) ] − 1 , A = ( e t , v R / Q R , v L / Q L ) , δ = ∣ det A ∣ . (176) Q_{v}=-\left[f_{*}\left(G_{*ii}G_{*jj}-G_{*ij}^{2}\right)\right]^{-1},\qquad\mathcal{A}=(e_{t},v_{R}/\sqrt{Q_{R}},v_{L}/\sqrt{Q_{L}}),\qquad\delta=\lvert\det\mathcal{A}\rvert. \tag*{(176)} Q v = − [ f ∗ ( G ∗ ii G ∗ j j − G ∗ ij 2 ) ] − 1 , A = ( e t , v R / Q R , v L / Q L ) , δ = ∣ det A ∣ . (176)
All these are finite axis coefficients. For the extremal ring example we choose B = A diag ( 1 , δ − 1 / 2 , δ − 1 / 2 ) \mathcal{B}=\mathcal{A}\operatorname{diag}(1,\delta^{-1/2},\delta^{-1/2}) B = A diag ( 1 , δ − 1/2 , δ − 1/2 ) . For the continued extremal hole example we choose
B = ( e t , − A ψ / δ , − A ϕ / δ ) . (177) \mathcal{B}=(e_{t},-\mathcal{A}_{\psi}/\sqrt{\delta},-\mathcal{A}_{\phi}/\sqrt{\delta}). \tag*{(177)} B = ( e t , − A ψ / δ , − A ϕ / δ ) . (177)
This choice has determinant − 1 -1 − 1 and fixes the signs of both angular coordinates. Applying the basis change and rescaling in (26) to (A.75) or (A.82), we obtain the physical metric
d s 2 = ( G ˉ ) a b d y a d y b + f ˉ ( d ρ 2 + d z 2 ) , G ˉ = B T G ∗ B , f ˉ = f ∗ / δ , det G ˉ = − ρ 2 . (178) \mathrm{d}s^{2}=(\bar{G})_{ab}\mathrm{d}y^{a}\mathrm{d}y^{b}+\bar{f}(\mathrm{d}\rho^{2}+\mathrm{d}z^{2}),\qquad\bar{G}=\mathcal{B}^{T}G_{*}\mathcal{B},\qquad\bar{f}=f_{*}/\delta,\qquad\det\bar{G}=-\rho^{2}. \tag*{(178)} d s 2 = ( G ˉ ) ab d y a d y b + f ˉ ( d ρ 2 + d z 2 ) , G ˉ = B T G ∗ B , f ˉ = f ∗ / δ , det G ˉ = − ρ 2 . (178)
Both physical angles have period 2 π 2\pi 2 π . On the intermediate axis the three balance conditions require zero time component, the same angular direction as the left axis, and unit cone factor. Equivalently its canonically normalised rod vector is ( 0 , 0 , 1 ) (0,0,1) ( 0 , 0 , 1 ) , up to the fixed orientation. These conditions fix the angular identifications as well as remove a conical defect.
All local and ADM charges have finite algebraic expressions. Let ( K , V , M ) (K,V,M) ( K , V , M ) denote ( Γ ^ , N ^ , M ^ ) (\hat{\Gamma},\hat{N},\hat{M}) ( Γ ^ , N ^ , M ^ ) for the extremal ring, or ( Γ ˉ , N ˉ , M ˉ ) (\bar{\Gamma},\bar{N},\bar{M}) ( Γ ˉ , N ˉ , M ˉ ) for the extremal hole, and define
s = ( u 4 + u 6 − u 1 − u 3 , − u 1 − u 2 − u 5 − 4 z − 2 , u 4 + u 6 − u 2 − u 3 − u 5 − 2 z ) , s=\left(u_{4}+u_{6}-u_{1}-u_{3},-u_{1}-u_{2}-u_{5}-4z-2,u_{4}+u_{6}-u_{2}-u_{3}-u_{5}-2z\right), s = ( u 4 + u 6 − u 1 − u 3 , − u 1 − u 2 − u 5 − 4 z − 2 , u 4 + u 6 − u 2 − u 3 − u 5 − 2 z ) ,
Q ∗ = − ( V T K − 1 M ) T − diag ( s ( 0 ) ) , Q = B − 1 Q ∗ B . (179) \mathcal{Q}_{*}=-(V^{T}K^{-1}M)^{T}-\operatorname{diag}(s^{(0)}),\qquad\mathcal{Q}=\mathcal{B}^{-1}\mathcal{Q}_{*}\mathcal{B}. \tag*{(179)} Q ∗ = − ( V T K − 1 M ) T − diag ( s ( 0 ) ) , Q = B − 1 Q ∗ B . (179)
Here ( 0 ) (0) ( 0 ) means the constant coefficient in the respective limit. The time row of Q \mathcal{Q} Q is constant on each ordinary axis. Let Q − , Q 0 , Q + \mathcal{Q}_{-},\mathcal{Q}_{0},\mathcal{Q}_{+} Q − , Q 0 , Q + be its values to the left of the ring, between horizons, and to the right of the hole. Then
Δ Q R = Q 0 − Q − , Δ Q H = Q + − Q 0 , \Delta Q_{\mathrm{R}} = Q_0 - Q_-, \qquad\Delta Q_{\mathrm{H}} = Q_+ - Q_0, Δ Q R = Q 0 − Q − , Δ Q H = Q + − Q 0 ,
M i = 3 π 8 G 5 ( Δ Q i ) t , J i A = − π 4 G 5 ( Δ Q i ) A t , M_i = \frac{3\pi}{8G_5}(\Delta Q_i)^t, \qquad J_{iA} = -\frac{\pi}{4G_5}(\Delta Q_i)^t_A, M i = 8 G 5 3 π ( Δ Q i ) t , J i A = − 4 G 5 π ( Δ Q i ) A t ,
M = M R + M H , J A = J R A + J H A . (180) M = M_{\mathrm{R}} + M_{\mathrm{H}}, \qquad J_A = J_{\mathrm{R}A} + J_{\mathrm{H}A}. \tag*{(180)} M = M R + M H , J A = J R A + J H A . (180)
When we add the two horizon contributions, the intermediate-axis value Q 0 Q_0 Q 0 cancels: Δ Q R + Δ Q H = Q + − Q − \Delta Q_{\mathrm{R}}+\Delta Q_{\mathrm{H}} = Q_+ - Q_- Δ Q R + Δ Q H = Q + − Q − . This gives the final line and agrees with the charges at infinity. Axis values mean constant coefficients of the finite expressions: expand u i = ∣ z − a i ∣ − ( z − a i ) + x / ( 2 ∣ z − a i ∣ ) + ⋯ u_i = |z-a_i|-(z-a_i)+x/(2|z-a_i|)+\cdots u i = ∣ z − a i ∣ − ( z − a i ) + x / ( 2∣ z − a i ∣ ) + ⋯ , factor common powers of x x x , and extract x 0 x^0 x 0 .
For either non-extremal horizon, let Λ A B = G ˉ A B \Lambda_{AB}=\bar{G}_{AB} Λ A B = G ˉ A B and let ℓ i \ell_i ℓ i be its finite rod length (B − A B-A B − A for the hole, ℓ \ell ℓ for the ring). At any interior point of that rod,
Ω i A = − ( Λ − 1 ) A B G ˉ B t , κ i = ( f ˉ det Λ ) − 1 / 2 , T i = κ i 2 π , \Omega_i^A = -(\Lambda^{-1})^{AB}\bar{G}_{Bt}, \qquad\kappa_i = (\bar{f}\det\Lambda)^{-1/2}, \qquad T_i = \frac{\kappa_i}{2\pi}, Ω i A = − ( Λ − 1 ) A B G ˉ B t , κ i = ( f ˉ det Λ ) − 1/2 , T i = 2 π κ i ,
A i = 4 π 2 ℓ i κ i , S i = A i 4 G 5 . (181) \mathcal{A}_i = \frac{4\pi^2\ell_i}{\kappa_i}, \qquad S_i = \frac{\mathcal{A}_i}{4G_5}. \tag*{(181)} A i = κ i 4 π 2 ℓ i , S i = 4 G 5 A i . (181)
At the extremal point use ρ = R sin θ \rho=R\sin\theta ρ = R sin θ , z = z i + R cos θ z=z_i+R\cos\theta z = z i + R cos θ . With F 0 = [ R 0 ] ( R 2 f ˉ ) F_0=[R^0](R^2\bar{f}) F 0 = [ R 0 ] ( R 2 f ˉ ) and Λ 0 = [ R 0 ] Λ \Lambda_0=[R^0]\Lambda Λ 0 = [ R 0 ] Λ ,
T i = κ i = 0 , Ω i A = − [ R 0 ] ( Λ − 1 ) A B G ˉ B t , T_i=\kappa_i=0, \qquad\Omega_i^A=-[R^0](\Lambda^{-1})^{AB}\bar{G}_{Bt}, T i = κ i = 0 , Ω i A = − [ R 0 ] ( Λ − 1 ) A B G ˉ B t ,
A i = 4 π 2 ∫ 0 π F 0 det Λ 0 d θ , S t o t = A R + A H 4 G 5 . (182) \mathcal{A}_i=4\pi^2\int_0^\pi\sqrt{F_0\det\Lambda_0}\,d\theta, \qquad S_{\mathrm{tot}}=\frac{\mathcal{A}_{\mathrm{R}}+\mathcal{A}_{\mathrm{H}}}{4G_5}. \tag*{(182)} A i = 4 π 2 ∫ 0 π F 0 det Λ 0 d θ , S tot = 4 G 5 A R + A H . (182)
On the regular branches the density D i = F 0 det Λ 0 / sin θ D_i=\sqrt{F_0\det\Lambda_0}/\sin\theta D i = F 0 det Λ 0 / sin θ is constant, giving A i = 8 π 2 D i \mathcal{A}_i=8\pi^2D_i A i = 8 π 2 D i by finite horizon coefficients alone. Independent area–spin checks give
A R = 8 π G 5 ∣ J R Φ ∣ ( T R = 0 ) , A H = 8 π G 5 ∣ J H Ψ J H Φ ∣ ( T H = 0 ) . (183) \mathcal{A}_{\mathrm{R}}=8\pi G_5|J_{\mathrm{R}\Phi}| \quad(T_{\mathrm{R}}=0), \qquad\mathcal{A}_{\mathrm{H}}=8\pi G_5\sqrt{|J_{\mathrm{H}\Psi}J_{\mathrm{H}\Phi}|} \quad(T_{\mathrm{H}}=0). \tag*{(183)} A R = 8 π G 5 ∣ J R Φ ∣ ( T R = 0 ) , A H = 8 π G 5 ∣ J H Ψ J H Φ ∣ ( T H = 0 ) . (183)
The local Smarr relations and the first law along a smooth balanced branch read
2 3 M i = T i S i + Ω i A J i A , d M = T w a r m d S w a r m + ∑ i = R , H Ω i A d J i A . (184) \frac{2}{3}M_i=T_iS_i+\Omega_i^A J_{iA}, \qquad dM=T_{\mathrm{warm}}dS_{\mathrm{warm}}+\sum_{i=\mathrm{R,H}}\Omega_i^A dJ_{iA}. \tag*{(184)} 3 2 M i = T i S i + Ω i A J i A , d M = T warm d S warm + i = R , H ∑ Ω i A d J i A . (184)
where T w a r m T_{\mathrm{warm}} T warm and S w a r m S_{\mathrm{warm}} S warm are the temperature and entropy, respectively, of the non-extremal horizon. Since one temperature is strictly positive, neither mixed-extremal branch is in mutual thermal equilibrium. Exact metric components and thermodynamic checks are provided in the ancillary Mathematica file.
Balanced examples, horizon extension and validation
Here we give numerical examples for these configurations with a single extremal horizon. For the extremal ring case take
( a , b , A , B , C , k ) = ( 1 , 2 , 4 , 5 , 5.1 , 3.2 ) , (a,b,A,B,C,k)=(1,2,4,5,5.1,3.2), ( a , b , A , B , C , k ) = ( 1 , 2 , 4 , 5 , 5.1 , 3.2 ) ,
q = − 2.78734053234366822678 , u = 1.18604581758086428649 , q=-2.78734053234366822678, \qquad u=1.18604581758086428649, q = − 2.78734053234366822678 , u = 1.18604581758086428649 ,
v = 0.032580068573907557108. (185) v=0.032580068573907557108. \tag*{(185)} v = 0.032580068573907557108. (185)
For an extremal central hole with a nonextremal ring, we find the following parameters. Here ℓ = a 3 − a 2 \ell= a_3-a_2 ℓ = a 3 − a 2 is the ring’s Weyl rod length in (A.78). It is small but strictly positive and stays fixed when we take the central-hole extremal limit. The factor 10 − 5 10^{-5} 1 0 − 5 below is ordinary scientific notation, so ℓ ≃ 0.0000204936 \ell\simeq0.0000204936 ℓ ≃ 0.0000204936 in the chosen coordinate units:
a = − 3263801.766316202 , U = − 4.796412879959584 , ℓ = 2.049355872243065 × 10 − 5 , C = − 1608957.852727379 , c 2 = 85867.17426548942 , c 3 = − 0.02758903860699587 , b = − 0.002097626322389592 , w = 8273.594438716184. (186) \begin{aligned}
a &= -3263801.766316202, & U &= -4.796412879959584,\\
\ell&= 2.049355872243065 \times10^{-5}, & C &= -1608957.852727379,\\
c_2 &= 85867.17426548942, & c_3 &= -0.02758903860699587,\\
b &= -0.002097626322389592, & w &= 8273.594438716184. \tag*{(186)}
\end{aligned} a ℓ c 2 b = − 3263801.766316202 , = 2.049355872243065 × 1 0 − 5 , = 85867.17426548942 , = − 0.002097626322389592 , U C c 3 w = − 4.796412879959584 , = − 1608957.852727379 , = − 0.02758903860699587 , = 8273.594438716184. (186)
We obtained these examples by solving the balance equations numerically. The displayed parameters are rounded, while the calculations use the more precise values supplied in the ancillary file. This matters especially for (A.94), where large terms cancel in the metric. Substituting only the printed digits can therefore produce misleading errors.
Table 1. The example (A.93), with G 5 = 1 G_5 = 1 G 5 = 1 . The ADM charges are M = 14.2456841940 M = 14.2456841940 M = 14.2456841940 , J Ψ = − 30.1033189908 J_\Psi= -30.1033189908 J Ψ = − 30.1033189908 , J Φ = − 0.267456543440 J_\Phi= -0.267456543440 J Φ = − 0.267456543440 , and S t o t = 48.1603826378 S_{\rm tot} = 48.1603826378 S tot = 48.1603826378 .
Table 2. The example (A.94), with G 5 = 1 G_5 = 1 G 5 = 1 . The ADM charges are M = 2.690228342792 M = 2.690228342792 M = 2.690228342792 , J Ψ = − 1.637463396379 J_\Psi= -1.637463396379 J Ψ = − 1.637463396379 , J Φ = 1.078279373341 J_\Phi= 1.078279373341 J Φ = 1.078279373341 , and S t o t = 8.290024236659 S_{\rm tot} = 8.290024236659 S tot = 8.290024236659 .
We checked each example in two independent ways: we extracted the ADM charges at infinity and integrated the Komar currents around each horizon. The resulting sums agree to better than 10 − 32 10^{-32} 1 0 − 32 , and both horizons satisfy their Smarr relations. We also checked that the angular velocities are constant over each horizon and that the extremal area–spin relations hold. At 156 exterior points for the extremal-ring example and 244 for the extremal-hole example, we found f ˉ > 0 \bar{f} > 0 f ˉ > 0 , G ˉ Ψ Ψ > 0 \bar{G}_{\Psi\Psi} > 0 G ˉ ΨΨ > 0 and det Λ > 0 \det\Lambda> 0 det Λ > 0 . These tests support regularity and causality at the sampled points, but do not establish a uniform bound on the full exterior.
We can extend the coordinates beyond each extremal horizon. Near either extremal point write
d s 2 = − R 2 N d t 2 + Λ A B ( d ϕ A + β A d t ) ( d ϕ B + β B d t ) + F ( d R 2 R 2 + d θ 2 ) , N = sin 2 θ det Λ , F = R 2 f ˉ . (187) ds^2 = -R^2Ndt^2+\Lambda_{AB}(d\phi^A+\beta^A dt)(d\phi^B+\beta^B dt)+F\left(\frac{dR^2}{R^2}+d\theta^2\right), \qquad N=\frac{\sin^2\theta}{\det\Lambda}, \qquad F=R^2\bar{f}. \tag*{(187)} d s 2 = − R 2 N d t 2 + Λ A B ( d ϕ A + β A d t ) ( d ϕ B + β B d t ) + F ( R 2 d R 2 + d θ 2 ) , N = det Λ sin 2 θ , F = R 2 f ˉ . (187)
Let β A = − Ω i A + k A R + O ( R 2 ) \beta^A=-\Omega_i^A+k^A R+O(R^2) β A = − Ω i A + k A R + O ( R 2 ) , and define
c i n 2 = N 0 F 0 , t 1 = N 1 / c i n 2 − F 1 2 c i n F 0 . (188) c_{\mathrm{in}}^2=\frac{N_0}{F_0}, \qquad t_1=\frac{N_1/c_{\mathrm{in}}^2-F_1}{2c_{\mathrm{in}}F_0}. \tag*{(188)} c in 2 = F 0 N 0 , t 1 = 2 c in F 0 N 1 / c in 2 − F 1 . (188)
When c i n , t 1 , k A c_{\mathrm{in}},t_1,k^A c in , t 1 , k A are angularly constant, the transformation
d t = d v + ( − 1 c i n R 2 + t 1 R ) d R , d ( ϕ A − Ω i A t ) = d ψ A + k A c i n R d R (189) dt=dv+\left(-\frac{1}{c_{\mathrm{in}}R^2}+\frac{t_1}{R}\right)dR,\qquad d(\phi^A-\Omega_i^A t)=d\psi^A+\frac{k^A}{c_{\mathrm{in}}R}dR \tag*{(189)} d t = d v + ( − c in R 2 1 + R t 1 ) d R , d ( ϕ A − Ω i A t ) = d ψ A + c in R k A d R (189)
cancels the potentially divergent radial terms. This is the extremal construction used in section 4, closely related to the extensions in [33 , 34 ] . At both examples, evaluation at five horizon angles verifies that the constants required by (A.97) are angle independent and that the subleading radial poles cancel. Continuing the common roots u 1 , u 4 , u 5 u_1,u_4,u_5 u 1 , u 4 , u 5 on their signed analytic branches across R = 0 R=0 R = 0 , three two-sided angular checks give smooth limits with differences of order R R R and determinant − F 2 sin 2 θ -F^2\sin^2\theta − F 2 sin 2 θ . These are local extension checks. They do not construct a maximal interior extension.
Explicit metric coefficients, thermodynamics, regularity and checks
We give the formulas needed to evaluate the metric and its physical properties. In B.1 we write the general metric as scalar functions, and in B.2 we obtain their doubly extremal values. In B.3 we evaluate the metric on the rods and at infinity to compute the charges, temperatures and entropies. Finally, in B.4 we explain how we test conical balance, smoothness and causality, and state the domains covered by these tests.1
General metric coefficients
The matrix formula in section 2 can be expanded into scalar products without printing thousands of terms. We do this in three steps: assign each pole a Killing-direction label, evaluate the resulting Cauchy products, and sum over the finite set of label assignments. The final seven functions are listed in (B.10). The pole functions are (4) . We use x = ρ 2 x=\rho^2 x = ρ 2 , p = ( u 1 , u 2 , u 3 , − x / u 4 , u 5 , − x / u 6 ) p=(u_1,u_2,u_3,-x/u_4,u_5,-x/u_6) p = ( u 1 , u 2 , u 3 , − x / u 4 , u 5 , − x / u 6 ) , and the three scalar coefficients are
h t = − u 1 u 3 u 4 u 6 , h ψ = − u 1 u 2 u 5 x , h ϕ = − u 2 u 3 u 5 u 4 u 6 . (190) h_t=-\frac{u_1u_3}{u_4u_6},\qquad h_\psi=-\frac{u_1u_2u_5}{x},\qquad h_\phi=-\frac{u_2u_3u_5}{u_4u_6}. \tag*{(190)} h t = − u 4 u 6 u 1 u 3 , h ψ = − x u 1 u 2 u 5 , h ϕ = − u 4 u 6 u 2 u 3 u 5 . (190)
The normalised dressing rows give a compact set of scalar weights. We divide each row in (9) by the entry that remains nonzero when its polarisation constant vanishes. These rescalings depend on position and endpoints but not on the six polarisation constants. The resulting 6 × 3 6 \times3 6 × 3 matrix has rows
M 1 = ( C 1 r 1 , 1 , 0 ) , M 2 = ( 0 , C 2 r 2 , 1 ) , M 3 = ( 1 , 0 , C 3 r 3 ) , M 4 = ( 1 , 0 , B 4 r 4 ) , M 5 = ( 0 , B 5 r 5 , 1 ) , M 6 = ( C 6 r 6 , 0 , 1 ) , (191) M_1=(C_1r_1,1,0),\quad M_2=(0,C_2r_2,1),\quad M_3=(1,0,C_3r_3),
\\
M_4=(1,0,B_4r_4),\quad M_5=(0,B_5r_5,1),\quad M_6=(C_6r_6,0,1), \tag*{(191)} M 1 = ( C 1 r 1 , 1 , 0 ) , M 2 = ( 0 , C 2 r 2 , 1 ) , M 3 = ( 1 , 0 , C 3 r 3 ) , M 4 = ( 1 , 0 , B 4 r 4 ) , M 5 = ( 0 , B 5 r 5 , 1 ) , M 6 = ( C 6 r 6 , 0 , 1 ) , (191)
where
r 1 = − 4 d 41 d 61 d 21 d 51 d 31 ( a 1 + 1 ) u 2 2 u 4 u 5 u 6 R 13 u 3 R 14 R 16 R 12 R 15 , r_1=-\frac{4d_{41}d_{61}d_{21}d_{51}}{d_{31}(a_1+1)}\frac{u_2^2u_4u_5u_6R_{13}}{u_3R_{14}R_{16}R_{12}R_{15}}, r 1 = − d 31 ( a 1 + 1 ) 4 d 41 d 61 d 21 d 51 u 3 R 14 R 16 R 12 R 15 u 2 2 u 4 u 5 u 6 R 13 ,
r 2 = − ( a 2 + 1 ) d 32 2 d 12 d 42 d 62 u 3 R 12 R 24 R 26 u 1 u 2 u 4 u 6 R 23 , r_2=-\frac{(a_2+1)d_{32}}{2d_{12}d_{42}d_{62}}\frac{u_3R_{12}R_{24}R_{26}}{u_1u_2u_4u_6R_{23}}, r 2 = − 2 d 12 d 42 d 62 ( a 2 + 1 ) d 32 u 1 u 2 u 4 u 6 R 23 u 3 R 12 R 24 R 26 ,
r 3 = d 13 2 d 23 d 53 u 1 R 23 R 35 u 2 u 3 u 5 R 13 , r 4 = R 14 u 4 R 24 R 45 , r_3=\frac{d_{13}}{2d_{23}d_{53}}\frac{u_1R_{23}R_{35}}{u_2u_3u_5R_{13}},\qquad r_4=\frac{R_{14}u_4}{R_{24}R_{45}}, r 3 = 2 d 23 d 53 d 13 u 2 u 3 u 5 R 13 u 1 R 23 R 35 , r 4 = R 24 R 45 R 14 u 4 ,
r 5 = − ( a 5 + 1 ) d 35 2 d 15 d 45 d 65 u 3 R 15 R 45 R 56 u 1 u 4 u 5 u 6 R 35 , r 6 = R 26 R 56 u 6 R 16 . (192) r_5=-\frac{(a_5+1)d_{35}}{2d_{15}d_{45}d_{65}}\frac{u_3R_{15}R_{45}R_{56}}{u_1u_4u_5u_6R_{35}},\qquad r_6=\frac{R_{26}R_{56}}{u_6R_{16}}. \tag*{(192)} r 5 = − 2 d 15 d 45 d 65 ( a 5 + 1 ) d 35 u 1 u 4 u 5 u 6 R 35 u 3 R 15 R 45 R 56 , r 6 = u 6 R 16 R 26 R 56 . (192)
We rebuild Γ \Gamma Γ from these rows and put N i a = M i a H a a / p i N_{ia}=M_{ia}H_{aa}/p_i N ia = M ia H aa / p i . Both G G G and P \mathcal{P} P are invariant under the row rescaling. The ratio det Γ / det Γ 0 \det\Gamma/\det\Gamma_0 det Γ/ det Γ 0 is also unchanged when the same rescaling is used in its reference matrix, whose six polarisation constants are zero.
Finite scalar products and sums
We set b i a = M i a b_i^a=M_{ia} b i a = M ia using (B.2). These position-dependent weights differ from the constant angular basis coefficients b i A b_{iA} b i A in (21) . Each row has two possible nonzero entries, so choosing one entry from each of the six rows gives 2 6 = 64 2^6=64 2 6 = 64 assignments. Their set is A \mathcal{A} A , with the label order t < ψ < ϕ t<\psi<\phi t < ψ < ϕ :
A = { t , ψ } × { ψ , ϕ } × { t , ϕ } × { t , ϕ } × { ψ , ϕ } × { t , ϕ } . (193) \mathcal{A}=\{t,\psi\}\times\{\psi,\phi\}\times\{t,\phi\}\times\{t,\phi\}\times\{\psi,\phi\}\times\{t,\phi\}. \tag*{(193)} A = { t , ψ } × { ψ , ϕ } × { t , ϕ } × { t , ϕ } × { ψ , ϕ } × { t , ϕ } . (193)
For α ∈ A \alpha\in\mathcal{A} α ∈ A put
I a ( α ) = { i : α i = a } , χ α = ( − 1 ) ∑ i < j 1 α i > α j ∏ i = 1 6 b i α i , r a ( α ) = # { i : α i > a } . (194) I_a(\alpha)=\{i:\alpha_i=a\},\qquad\chi_\alpha=(-1)^{\sum_{i<j}\mathbf{1}_{\alpha_i>\alpha_j}}\prod_{i=1}^{6}b_i^{\alpha_i},\qquad r_a(\alpha)=\#\{i:\alpha_i>a\}. \tag*{(194)} I a ( α ) = { i : α i = a } , χ α = ( − 1 ) ∑ i < j 1 α i > α j i = 1 ∏ 6 b i α i , r a ( α ) = # { i : α i > a } . (194)
Here I a I_a I a groups the poles assigned to direction a a a , while the sign in χ α \chi_\alpha χ α records the permutations needed to group them. The integer r a ( α ) r_a(\alpha) r a ( α ) counts labels and is unrelated to the rational row functions r i r_i r i . For increasing index sets I , J ⊂ { 1 , … , 6 } I,J\subset\{1,\ldots,6\} I , J ⊂ { 1 , … , 6 } and ϵ , η ∈ { 0 , 1 } \epsilon,\eta\in\{0,1\} ϵ , η ∈ { 0 , 1 } satisfying n = ∣ I ∣ + ϵ = ∣ J ∣ + η n=|I|+\epsilon=|J|+\eta n = ∣ I ∣ + ϵ = ∣ J ∣ + η , define the scalar product
C ϵ η ( I , J ) = ( − x ) n ( n − 1 ) / 2 ∏ i < i ′ i , i ′ ∈ I ( p i ′ − p i ) ∏ j < j ′ j , j ′ ∈ J ( p j ′ − p j ) ∏ i ∈ I , j ∈ J ( x + p i p j ) ∏ j ∈ J p j ϵ ∏ i ∈ I p i η . (195) C_{\epsilon\eta}(I,J)=\frac{(-x)^{n(n-1)/2}\prod_{\substack{i<i'\\i,i'\in I}}(p_{i'}-p_i)\prod_{\substack{j<j'\\j,j'\in J}}(p_{j'}-p_j)}{\prod_{\substack{i\in I,j\in J}}(x+p_ip_j)\prod_{j\in J}p_j^\epsilon\prod_{i\in I}p_i^\eta}. \tag*{(195)} C ϵη ( I , J ) = ∏ i ∈ I , j ∈ J ( x + p i p j ) ∏ j ∈ J p j ϵ ∏ i ∈ I p i η ( − x ) n ( n − 1 ) /2 ∏ i < i ′ i , i ′ ∈ I ( p i ′ − p i ) ∏ j < j ′ j , j ′ ∈ J ( p j ′ − p j ) . (195)
Every empty product is one. In particular C 00 ( ∅ , ∅ ) = C 11 ( ∅ , ∅ ) = 1 C_{00}(\varnothing,\varnothing)=C_{11}(\varnothing,\varnothing)=1 C 00 ( ∅ , ∅ ) = C 11 ( ∅ , ∅ ) = 1 . All pole-pair denominators can be factored by (B.34). The differences are equally explicit:
p i − p j = { 2 ( a i − a j ) u i u j / R i j , i , j ∉ { 4 , 6 } , 2 x ( a i − a j ) / R i j , i , j ∈ { 4 , 6 } , R i j / u j , i ∉ { 4 , 6 } , j ∈ { 4 , 6 } , − R i j / u i , i ∈ { 4 , 6 } , j ∉ { 4 , 6 } . (196) p_i-p_j=
\begin{cases}
2(a_i-a_j)u_i u_j/R_{ij}, & i,j\notin\{4,6\},\\
2x(a_i-a_j)/R_{ij}, & i,j\in\{4,6\},\\
R_{ij}/u_j, & i\notin\{4,6\},\ j\in\{4,6\},\\
-R_{ij}/u_i, & i\in\{4,6\},\ j\notin\{4,6\}.
\end{cases} \tag*{(196)} p i − p j = ⎩ ⎨ ⎧ 2 ( a i − a j ) u i u j / R ij , 2 x ( a i − a j ) / R ij , R ij / u j , − R ij / u i , i , j ∈ / { 4 , 6 } , i , j ∈ { 4 , 6 } , i ∈ / { 4 , 6 } , j ∈ { 4 , 6 } , i ∈ { 4 , 6 } , j ∈ / { 4 , 6 } . (196)
The two scalar sums required for all seven functions are
S 0 = ∑ α , β ∈ A ∣ I k ( α ) ∣ = ∣ I k ( β ) ∣ ∀ k χ α χ β ∏ k = t , ψ , ϕ h k ∣ I k ( α ) ∣ C 00 ( I k ( α ) , I k ( β ) ) , S_0=\sum_{\substack{\alpha,\beta\in\mathcal{A}\\ |I_k(\alpha)|=|I_k(\beta)|\ \forall k}}\chi_\alpha\chi_\beta\prod_{k=t,\psi,\phi}h_k^{|I_k(\alpha)|}C_{00}\bigl(I_k(\alpha),I_k(\beta)\bigr), S 0 = α , β ∈ A ∣ I k ( α ) ∣ = ∣ I k ( β ) ∣ ∀ k ∑ χ α χ β k = t , ψ , ϕ ∏ h k ∣ I k ( α ) ∣ C 00 ( I k ( α ) , I k ( β ) ) ,
S a b = ∑ α , β ∈ A ∣ I k ( α ) ∣ + δ k a = ∣ I k ( β ) ∣ + δ k b ∀ k ( − 1 ) r a ( α ) + r b ( β ) χ α χ β ∏ k = t , ψ , ϕ h k ∣ I k ( α ) ∣ + δ k a C δ k a , δ k b ( I k ( α ) , I k ( β ) ) . (197) S_{ab}=\sum_{\substack{\alpha,\beta\in\mathcal{A}\\ |I_k(\alpha)|+\delta_{ka}=|I_k(\beta)|+\delta_{kb}\ \forall k}}(-1)^{r_a(\alpha)+r_b(\beta)}\chi_\alpha\chi_\beta\prod_{k=t,\psi,\phi}h_k^{|I_k(\alpha)|+\delta_{ka}}C_{\delta_{ka},\delta_{kb}}\bigl(I_k(\alpha),I_k(\beta)\bigr). \tag*{(197)} S ab = α , β ∈ A ∣ I k ( α ) ∣ + δ k a = ∣ I k ( β ) ∣ + δ k b ∀ k ∑ ( − 1 ) r a ( α ) + r b ( β ) χ α χ β k = t , ψ , ϕ ∏ h k ∣ I k ( α ) ∣ + δ k a C δ k a , δ k b ( I k ( α ) , I k ( β ) ) . (197)
The sums give explicit rational functions of the pole functions and parameters. Setting all six polarisation constants to zero gives the reference factor
S r e f = h t 2 h ψ 2 h ϕ 2 C 00 ( { 3 , 4 } , { 3 , 4 } ) C 00 ( { 1 } , { 1 } ) C 00 ( { 2 , 5 , 6 } , { 2 , 5 , 6 } ) , S_{\mathrm{ref}}=h_t^2h_\psi^2h_\phi^2C_{00}(\{3,4\},\{3,4\})C_{00}(\{1\},\{1\})C_{00}(\{2,5,6\},\{2,5,6\}), S ref = h t 2 h ψ 2 h ϕ 2 C 00 ({ 3 , 4 } , { 3 , 4 }) C 00 ({ 1 } , { 1 }) C 00 ({ 2 , 5 , 6 } , { 2 , 5 , 6 }) ,
L = f s P S r e f . (198) \mathcal{L}=\frac{f_s}{P S_{\mathrm{ref}}}. \tag*{(198)} L = P S ref f s . (198)
where the elementary product f s f_s f s is given in (12) and P P P in (18) . The seven desired functions are
H x = L S 0 , H y = − L S t t , H_x=\mathcal{L}S_0,\qquad H_y=-\mathcal{L}S_{tt}, H x = L S 0 , H y = − L S tt ,
ω ψ = − L S t ψ , ω ϕ = − L S t ϕ , \omega_\psi=-\mathcal{L}S_{t\psi},\qquad\omega_\phi=-\mathcal{L}S_{t\phi}, ω ψ = − L S t ψ , ω ϕ = − L S tϕ ,
V ψ ψ = L S ψ ψ , V ψ ϕ = L S ψ ϕ , V ϕ ϕ = L S ϕ ϕ . (199) V_{\psi\psi}=\mathcal{L}S_{\psi\psi},\qquad V_{\psi\phi}=\mathcal{L}S_{\psi\phi},\qquad V_{\phi\phi}=\mathcal{L}S_{\phi\phi}. \tag*{(199)} V ψ ψ = L S ψ ψ , V ψ ϕ = L S ψ ϕ , V ϕϕ = L S ϕϕ . (199)
This completes the scalar specification of the general metric. The product formula (B.6) follows from the Cauchy identity. Assigning each of the six factors its label t , ψ , ϕ t,\psi,\phi t , ψ , ϕ gives (B.8) by multilinearity. Consequently S 0 = det Γ S_0=\det\Gamma S 0 = det Γ , S a b / S 0 = G a b S_{ab}/S_0=G_{ab} S ab / S 0 = G ab , and S r e f = det Γ 0 S_{\mathrm{ref}}=\det\Gamma_0 S ref = det Γ 0 . These equalities establish the identity with the inverse-scattering construction but they are not needed to evaluate the scalar formulas.
Polynomial form. After collecting terms using (B.7), all seven functions share the positive monomial denominator
F = 4 u 1 3 u 2 3 u 3 4 u 4 4 u 5 3 u 6 4 ( d 42 d 52 d 54 d 63 ) 2 × R 12 R 23 R 25 R 34 3 R 36 R 46 ∏ 1 ≤ i ≤ j ≤ 6 R i j . (200) \mathcal{F}=4u_1^3u_2^3u_3^4u_4^4u_5^3u_6^4(d_{42}d_{52}d_{54}d_{63})^2
\times R_{12}R_{23}R_{25}R_{34}^3R_{36}R_{46}\prod_{1\leq i\leq j\leq6}R_{ij}. \tag*{(200)} F = 4 u 1 3 u 2 3 u 3 4 u 4 4 u 5 3 u 6 4 ( d 42 d 52 d 54 d 63 ) 2 × R 12 R 23 R 25 R 34 3 R 36 R 46 1 ≤ i ≤ j ≤ 6 ∏ R ij . (200)
Specifically,
( H x , H y , ω ψ , ω ϕ , V ψ ψ , V ψ ϕ , V ϕ ϕ ) = 1 F ( M x , M y , W ψ , W ϕ , V ψ ψ , V ψ ϕ , V ϕ ϕ ) . (201) (H_x,H_y,\omega_\psi,\omega_\phi,V_{\psi\psi},V_{\psi\phi},V_{\phi\phi})
=\frac{1}{\mathcal{F}}(\mathcal{M}_x,\mathcal{M}_y,\mathcal{W}_\psi,\mathcal{W}_\phi,\mathcal{V}_{\psi\psi},\mathcal{V}_{\psi\phi},\mathcal{V}_{\phi\phi}). \tag*{(201)} ( H x , H y , ω ψ , ω ϕ , V ψ ψ , V ψ ϕ , V ϕϕ ) = F 1 ( M x , M y , W ψ , W ϕ , V ψ ψ , V ψ ϕ , V ϕϕ ) . (201)
The numerators are integer-coefficient polynomials in x x x , u i u_i u i , R i j R_{ij} R ij , d i j d_{ij} d ij with i > j i>j i > j , a i + 1 a_i+1 a i + 1 , and ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) (C_1,C_2,C_3,B_4,B_5,C_6) ( C 1 , C 2 , C 3 , B 4 , B 5 , C 6 ) . Their coefficients are specified by the finite sums (B.8)–(B.10). The expanded polynomials are included in the ancillary Mathematica file.
Scalar outer-axis normalisation
We now normalise the two outer axes to primitive angles of period 2 π 2\pi 2 π , as in (14) . To evaluate the constants in (22) , we set
D = f = P H x , ( Q t t , Q t ψ , Q t ϕ , Q ψ ψ , Q ψ ϕ , Q ϕ ϕ ) = P ( − H y , − ω ψ , − ω ϕ , V ψ ψ , V ψ ϕ , V ϕ ϕ ) . (202) D=f=PH_x,\qquad(Q_{tt},Q_{t\psi},Q_{t\phi},Q_{\psi\psi},Q_{\psi\phi},Q_{\phi\phi})=P(-H_y,-\omega_\psi,-\omega_\phi,V_{\psi\psi},V_{\psi\phi},V_{\phi\phi}). \tag*{(202)} D = f = P H x , ( Q tt , Q t ψ , Q tϕ , Q ψ ψ , Q ψ ϕ , Q ϕϕ ) = P ( − H y , − ω ψ , − ω ϕ , V ψ ψ , V ψ ϕ , V ϕϕ ) . (202)
Here Q a b = f G a b Q_{ab}=fG_{ab} Q ab = f G ab are scalar functions and are unrelated to the Komar-current potential. Evaluate their exact constant coefficients at z L = a 1 − 1 z_L=a_1-1 z L = a 1 − 1 and z R = a 6 + 1 z_R=a_6+1 z R = a 6 + 1 . For a collected monomial, assign
ord s x = 1 , ord s u i = 1 i ≤ s , ord s R i j = 1 i ≤ s o r j ≤ s , s = 0 , 6. (203) \operatorname{ord}_s x=1,\qquad\operatorname{ord}_s u_i=1_{i\leq s},\qquad\operatorname{ord}_s R_{ij}=1_{i\leq s\ \mathrm{or}\ j\leq s},\qquad s=0,6. \tag*{(203)} ord s x = 1 , ord s u i = 1 i ≤ s , ord s R ij = 1 i ≤ s or j ≤ s , s = 0 , 6. (203)
All other factors have order zero. Every collected term of D D D , Q a b Q_{ab} Q ab has nonnegative order. Discard the positive-order terms and replace x x x by one, u i u_i u i by u ˉ i \bar{u}_i u ˉ i and R i j R_{ij} R ij by R ˉ i j \bar{R}_{ij} R ˉ ij from (B.33) and the leading coefficients following (B.37). This gives the exact values on the axes.
On the lower axis we obtain
D L = Q t t Q ψ ψ − Q t ψ 2 , q L = − D / D L , (204) D_L=Q_{tt}Q_{\psi\psi}-Q_{t\psi}^{2},\qquad q_L=-D/D_L, \tag*{(204)} D L = Q tt Q ψ ψ − Q t ψ 2 , q L = − D / D L , (204)
v L t = Q t ψ Q ψ ϕ − Q ψ ψ Q t ϕ D L , v L ψ = Q t ψ Q t ϕ − Q t t Q ψ ϕ D L , v L ϕ = 1. v_L^t=\frac{Q_{t\psi}Q_{\psi\phi}-Q_{\psi\psi}Q_{t\phi}}{D_L},\qquad v_L^\psi=\frac{Q_{t\psi}Q_{t\phi}-Q_{tt}Q_{\psi\phi}}{D_L},\qquad v_L^\phi=1. v L t = D L Q t ψ Q ψ ϕ − Q ψ ψ Q tϕ , v L ψ = D L Q t ψ Q tϕ − Q tt Q ψ ϕ , v L ϕ = 1.
On the upper axis the corresponding formulas are
D R = Q t t Q ϕ ϕ − Q t ϕ 2 , q R = − D / D R , (205) D_R=Q_{tt}Q_{\phi\phi}-Q_{t\phi}^{2},\qquad q_R=-D/D_R, \tag*{(205)} D R = Q tt Q ϕϕ − Q tϕ 2 , q R = − D / D R , (205)
v R t = Q t ϕ Q ψ ϕ − Q ϕ ϕ Q t ψ D R , v R ϕ = Q t ϕ Q t ψ − Q t t Q ψ ϕ D R , v R ψ = 1. v_R^t=\frac{Q_{t\phi}Q_{\psi\phi}-Q_{\phi\phi}Q_{t\psi}}{D_R},\qquad v_R^\phi=\frac{Q_{t\phi}Q_{t\psi}-Q_{tt}Q_{\psi\phi}}{D_R},\qquad v_R^\psi=1. v R t = D R Q tϕ Q ψ ϕ − Q ϕϕ Q t ψ , v R ϕ = D R Q tϕ Q t ψ − Q tt Q ψ ϕ , v R ψ = 1.
Use the appropriate axis coefficients in each line. Then
q Ψ = v R t / q R , q Φ = v L t / q L , q_\Psi=v_R^t/\sqrt{q_R},\qquad q_\Phi=v_L^t/\sqrt{q_L}, q Ψ = v R t / q R , q Φ = v L t / q L ,
b ψ Ψ = 1 / q R , b ϕ Ψ = v R ϕ / q R , b_{\psi\Psi}=1/\sqrt{q_R},\qquad b_{\phi\Psi}=v_R^\phi/\sqrt{q_R}, b ψ Ψ = 1/ q R , b ϕ Ψ = v R ϕ / q R ,
b ψ Φ = v L ψ / q L , b ϕ Φ = 1 / q L . (206) b_{\psi\Phi}=v_L^\psi/\sqrt{q_L},\qquad b_{\phi\Phi}=1/\sqrt{q_L}. \tag*{(206)} b ψ Φ = v L ψ / q L , b ϕ Φ = 1/ q L . (206)
This is the scalar version of (14) . In the canonical convention, replace q A , b i A q_A,b_{iA} q A , b i A by q A / δ , b i A / δ q_A/\sqrt{\delta},b_{iA}/\sqrt{\delta} q A / δ , b i A / δ and the base coefficient P P P by P / δ P/\delta P / δ in the physical line element.
Doubly extremal metric coefficients
We obtain the seven functions in (61) by expanding the general metric along the correlated limit of § 4 . The formulas below evaluate the metric directly at zero temperature. We keep ( Z , c , d , H ∗ , v ∗ , w , r 0 , β ) (Z,c,d,H_*,v_*,w,r_0,\beta) ( Z , c , d , H ∗ , v ∗ , w , r 0 , β ) free at this stage and impose nonzero denominators and balance subsequently. We write H = H ∗ H=H_* H = H ∗ and v = v ∗ v=v_* v = v ∗ .
To see why the relation for η \eta η in (56) is needed, put x = ρ 2 x=\rho^2 x = ρ 2 and let η = E λ + O ( λ 2 ) \eta=E\lambda+O(\lambda^2) η = E λ + O ( λ 2 ) . Eliminating the coupled pole residues in (10) leaves a coefficient proportional to
U 2 ( U − V ) 2 V β H ∗ 2 ( x + U 2 ) [ E − c ( 1 + Z − 1 ) ] , U = x + z 2 − z , V = x + ( z − Z ) 2 − ( z − Z ) . (207) \frac{U^2(U-V)}{2V\beta H_*^2(x+U^2)}\left[E-c\left(1+Z^{-1}\right)\right], \qquad U=\sqrt{x+z^2}-z,\quad V=\sqrt{x+(z-Z)^2}-(z-Z). \tag*{(207)} 2 V β H ∗ 2 ( x + U 2 ) U 2 ( U − V ) [ E − c ( 1 + Z − 1 ) ] , U = x + z 2 − z , V = x + ( z − Z ) 2 − ( z − Z ) . (207)
Since Z = ( V − U ) ( x + U V ) / ( 2 U V ) Z=(V-U)(x+UV)/(2UV) Z = ( V − U ) ( x + U V ) / ( 2 U V ) and V > U > 0 V>U>0 V > U > 0 off the axis, the prefactor is nonzero. Equation (56) therefore enforces a genuine cancellation required by this limiting construction. Thus the leading powers of λ \lambda λ alone do not determine the limit: the displayed subleading relation among the polarisations is also needed.
Taylor coefficients for the extremal limit
To obtain the extremal functions it suffices to keep eight coefficients, through seventh order, of an auxiliary variable ϵ = τ 2 \epsilon=\tau^2 ϵ = τ 2 . Terms of higher order cannot contribute to the finite metric. We express this truncation by working in the algebra P = R [ ϵ ] / ( ϵ 8 ) \mathcal{P}=\mathbb{R}[\epsilon]/(\epsilon^8) P = R [ ϵ ] / ( ϵ 8 ) , with coefficients rational in the parameters and in x , U , V x,U,V x , U , V . That is, multiply polynomials in the usual way and discard powers ϵ 8 \epsilon^8 ϵ 8 and higher. For a polynomial A A A with constant coefficient A 0 ≠ 0 A_0\ne0 A 0 = 0 , every integer power and every square root below is specified by the finite formula
B α ( A ) = A 0 α ∑ k = 0 7 ( α k ) ( A − A 0 A 0 ) k in P . (208) B_\alpha(A)=A_0^\alpha\sum_{k=0}^{7}\binom{\alpha}{k}\left(\frac{A-A_0}{A_0}\right)^k\quad\text{in }\mathcal{P}. \tag*{(208)} B α ( A ) = A 0 α k = 0 ∑ 7 ( k α ) ( A 0 A − A 0 ) k in P . (208)
All square-root arguments used below have constant entry one, fixing the branch. No differential or infinite-series operation is required. Set x = ρ 2 x=\rho^2 x = ρ 2 , use U , V U,V U , V from (B.18), and put
Z = ( V − U ) ( x + U V ) 2 U V , z = x − U 2 2 U . (209) Z=\frac{(V-U)(x+UV)}{2UV},\qquad z=\frac{x-U^2}{2U}. \tag*{(209)} Z = 2 U V ( V − U ) ( x + U V ) , z = 2 U x − U 2 . (209)
We first expand the endpoints:
S = r 0 2 B 1 / 2 ( 1 + ϵ 2 4 r 0 4 ) , R = B 1 / 2 ( S + ϵ / 2 r 0 2 ) , A = Z − S − r 0 2 2 − ϵ 4 , B = Z − S − r 0 2 2 + ϵ 4 , C = H 2 2 + ϵ B , ( a 1 , … , a 5 ) = ( 0 , c ϵ / 2 , c ϵ / 2 + d ϵ 2 / 4 , A , B ) . (210) \begin{aligned}
S&=r_0^2B_{1/2}\left(1+\frac{\epsilon^2}{4r_0^4}\right),&
R&=B_{1/2}\left(\frac{S+\epsilon/2}{r_0^2}\right),\\
A&=Z-\frac{S-r_0^2}{2}-\frac{\epsilon}{4},&
B&=Z-\frac{S-r_0^2}{2}+\frac{\epsilon}{4},\\
C&=\frac{H^2}{2}+\epsilon B,&
(a_1,\ldots,a_5)&=(0,c\epsilon/2,c\epsilon/2+d\epsilon^2/4,A,B). \tag*{(210)}
\end{aligned} S A C = r 0 2 B 1/2 ( 1 + 4 r 0 4 ϵ 2 ) , = Z − 2 S − r 0 2 − 4 ϵ , = 2 H 2 + ϵ B , R B ( a 1 , … , a 5 ) = B 1/2 ( r 0 2 S + ϵ /2 ) , = Z − 2 S − r 0 2 + 4 ϵ , = ( 0 , cϵ /2 , cϵ /2 + d ϵ 2 /4 , A , B ) . (210)
Here C C C is the finite factor of the receding endpoint after removing its divergent power, a 6 = ϵ − 1 C a_6=\epsilon^{-1}C a 6 = ϵ − 1 C and it is distinct from a polarisation constant. Define
b 0 = β / R , α = − 1 / ( β R ) , π 1 = − 1 Z B 1 / 2 ( ( 1 + d ϵ / ( 2 c ) ) Z 2 H 2 2 A B C ) , π 6 = 1 H B 1 / 2 ( C C − c ϵ 2 / 2 ) , π 2 = 1 − c ( 1 + Z − 1 ) ϵ / 2 − w ϵ 2 / 4 π 1 ν , π 3 = ν , π 4 = 1 − ϵ π 6 b 0 b 0 + ϵ π 6 , π 5 = H 2 π 6 α 2 ( B + 1 ) . (211) \begin{aligned}
b_0&=\beta/R,&
\alpha&=-1/(\beta R),\\
\pi_1&=-\frac{1}{Z}B_{1/2}\left(\frac{(1+d\epsilon/(2c))Z^2H^2}{2ABC}\right),&
\pi_6&=\frac{1}{H}B_{1/2}\left(\frac{C}{C-c\epsilon^2/2}\right),\\
\pi_2&=\frac{1-c(1+Z^{-1})\epsilon/2-w\epsilon^2/4}{\pi_1\nu},&
\pi_3&=\nu,\\
\pi_4&=\frac{1-\epsilon\pi_6b_0}{b_0+\epsilon\pi_6},&
\pi_5&=\frac{H^2\pi_6\alpha}{2(B+1)}. \tag*{(211)}
\end{aligned} b 0 π 1 π 2 π 4 = β / R , = − Z 1 B 1/2 ( 2 A B C ( 1 + d ϵ / ( 2 c )) Z 2 H 2 ) , = π 1 ν 1 − c ( 1 + Z − 1 ) ϵ /2 − w ϵ 2 /4 , = b 0 + ϵ π 6 1 − ϵ π 6 b 0 , α π 6 π 3 π 5 = − 1/ ( β R ) , = H 1 B 1/2 ( C − c ϵ 2 /2 C ) , = ν , = 2 ( B + 1 ) H 2 π 6 α . (211)
Every quotient here means multiplication by (B.19) with α = − 1 \alpha=-1 α = − 1 . If τ 2 = ϵ \tau^2=\epsilon τ 2 = ϵ , the six polarisation factors are products of the finite factors ( π 1 , … , π 6 ) (\pi_1,\ldots,\pi_6) ( π 1 , … , π 6 ) and the following powers of τ \tau τ
( ν C 1 , ν C 2 , ν C 3 , ν B 4 , ν B 5 , ν C 6 ) = ( 1 , − 2 , 1 , 1 , 0 , 1 ) . (212) (\nu_{C_1},\nu_{C_2},\nu_{C_3},\nu_{B_4},\nu_{B_5},\nu_{C_6})=(1,-2,1,1,0,1). \tag*{(212)} ( ν C 1 , ν C 2 , ν C 3 , ν B 4 , ν B 5 , ν C 6 ) = ( 1 , − 2 , 1 , 1 , 0 , 1 ) . (212)
The bookkeeping symbol τ \tau τ is not assigned a numerical value. For i ≤ 5 i\leq5 i ≤ 5 , put u i 0 = U u_i^0=U u i 0 = U , a i 0 = 0 a_i^0=0 a i 0 = 0 for i ≤ 3 i\leq3 i ≤ 3 and u i 0 = V u_i^0=V u i 0 = V , a i 0 = Z a_i^0=Z a i 0 = Z otherwise. With r i = ( x + ( u i 0 ) 2 ) / ( 2 u i 0 ) r_i=(x+(u_i^0)^2)/(2u_i^0) r i = ( x + ( u i 0 ) 2 ) / ( 2 u i 0 ) and Δ i = a i − a i 0 \Delta_i=a_i-a_i^0 Δ i = a i − a i 0 , set
u ^ i = u i 0 + Δ i + r i [ B 1 / 2 ( 1 − 2 ( z − a i 0 ) Δ i r i 2 + Δ i 2 r i 2 ) − 1 ] , \widehat{u}_i=u_i^0+\Delta_i+r_i\left[B_{1/2}\left(1-\frac{2(z-a_i^0)\Delta_i}{r_i^2}+\frac{\Delta_i^2}{r_i^2}\right)-1\right], u i = u i 0 + Δ i + r i [ B 1/2 ( 1 − r i 2 2 ( z − a i 0 ) Δ i + r i 2 Δ i 2 ) − 1 ] ,
D = C − ϵ z , u ^ 6 = D [ 1 + B 1 / 2 ( 1 + x ϵ 2 D 2 ) ] . (213) D=C-\epsilon z,\qquad\widehat{u}_6=D\left[1+B_{1/2}\left(1+\frac{x\epsilon^2}{D^2}\right)\right]. \tag*{(213)} D = C − ϵz , u 6 = D [ 1 + B 1/2 ( 1 + D 2 x ϵ 2 ) ] . (213)
Likewise the pole functions have finite factors u ^ i \widehat{u}_i u i and powers ν u i = 0 \nu_{u_i}=0 ν u i = 0 for i ≤ 5 i\leq5 i ≤ 5 , ν u 6 = − 2 \nu_{u_6}=-2 ν u 6 = − 2 . All remaining factors of subsection B.1 are now explicit:
Table 6 (PDF p. 50).
As before d i j = a i − a j d_{ij}=a_i-a_j d ij = a i − a j with i > j i>j i > j . The last two rows incorporate the singular boost s b = v b s_b=v_b s b = v b , c b = u b c_b=u_b c b = u b of section 4.
The seven finite scalar sums
For a monomial m = q ∏ a a n a m=q\prod_a a^{n_a} m = q ∏ a a n a from the collected general expressions, write each factor as a = τ ν a a ^ ( ϵ ) a=\tau^{\nu_a}\widehat{a}(\epsilon) a = τ ν a a ( ϵ ) , using the finite factor and power in the preceding table and (B.23). Put N = ( 6 − ∑ a n a ν a ) / 2 N=(6-\sum_a n_a\nu_a)/2 N = ( 6 − ∑ a n a ν a ) /2 and define the finite scalar operation
E ( m ) = q ∑ ∑ a k a = N 0 ≤ k a ≤ 7 ∏ a ( B n a ( a ^ ) ) k a . (214) \mathcal{E}(m)=q\sum_{\substack{\sum_a k_a=N\\0\leq k_a\leq7}}\prod_a\left(B_{n_a}(\widehat{a})\right)_{k_a}. \tag*{(214)} E ( m ) = q ∑ a k a = N 0 ≤ k a ≤ 7 ∑ a ∏ ( B n a ( a ) ) k a . (214)
The subscript selects a Taylor coefficient, and an empty sum is zero. We apply E \mathcal{E} E to each monomial and add the results. This extracts the coefficient that survives in the extremal metric. We use the general scalar functions of B.1 to form D = P H x \mathcal{D}=PH_x D = P H x and
( Q t t , Q t ψ , Q t ϕ , Q ψ ψ , Q ψ ϕ , Q ϕ ϕ ) = P ( − H y , − ω ψ , − ω ϕ , V ψ ψ , V ψ ϕ , V ϕ ϕ ) . (215) (Q_{tt},Q_{t\psi},Q_{t\phi},Q_{\psi\psi},Q_{\psi\phi},Q_{\phi\phi})=P(-H_y,-\omega_\psi,-\omega_\phi,V_{\psi\psi},V_{\psi\phi},V_{\phi\phi}). \tag*{(215)} ( Q tt , Q t ψ , Q tϕ , Q ψ ψ , Q ψ ϕ , Q ϕϕ ) = P ( − H y , − ω ψ , − ω ϕ , V ψ ψ , V ψ ϕ , V ϕϕ ) . (215)
These are the collected monomial expressions for f f f and f G a b fG_{ab} f G ab . The desired extremal functions, with every factor evaluated by the finite table above, are
K = 8 E ( D ) , Y = − 8 E ( Q t t ) , W ψ = 8 E ( s b Q t t / C 6 − c b Q t ψ − s b Q t ϕ ) , W ϕ = 8 E ( c b Q t t / C 6 − s b Q t ψ − c b Q t ϕ ) , V ψ ψ = 8 E ( s b 2 Q t t / C 6 2 − 2 s b c b Q t ψ / C 6 − 2 s b 2 Q t ϕ / C 6 + c b 2 Q ψ ψ + 2 s b c b Q ψ ϕ + s b 2 Q ϕ ϕ ) , V ψ ϕ = 8 E ( s b c b Q t t / C 6 2 − ( s b 2 + c b 2 ) Q t ψ / C 6 − 2 s b c b Q t ϕ / C 6 + s b c b Q ψ ψ + ( s b 2 + c b 2 ) Q ψ ϕ + s b c b Q ϕ ϕ ) , V ϕ ϕ = 8 E ( c b 2 Q t t / C 6 2 − 2 s b c b Q t ψ / C 6 − 2 c b 2 Q t ϕ / C 6 + s b 2 Q ψ ψ + 2 s b c b Q ψ ϕ + c b 2 Q ϕ ϕ ) . (216) \begin{aligned}
K &= 8\mathcal{E}(\mathcal{D}), \qquad Y = -8\mathcal{E}(Q_{tt}),\\
W_{\psi} &= 8\mathcal{E}\left(s_bQ_{tt}/C_6-c_bQ_{t\psi}-s_bQ_{t\phi}\right),\\
W_{\phi} &= 8\mathcal{E}\left(c_bQ_{tt}/C_6-s_bQ_{t\psi}-c_bQ_{t\phi}\right),\\
V_{\psi\psi} &= 8\mathcal{E}\left(s_b^2Q_{tt}/C_6^2-2s_bc_bQ_{t\psi}/C_6-2s_b^2Q_{t\phi}/C_6\right.\\
&\qquad\left.{}+c_b^2Q_{\psi\psi}+2s_bc_bQ_{\psi\phi}+s_b^2Q_{\phi\phi}\right),\\
V_{\psi\phi} &= 8\mathcal{E}\left(s_bc_bQ_{tt}/C_6^2-(s_b^2+c_b^2)Q_{t\psi}/C_6-2s_bc_bQ_{t\phi}/C_6\right.\\
&\qquad\left.{}+s_bc_bQ_{\psi\psi}+(s_b^2+c_b^2)Q_{\psi\phi}+s_bc_bQ_{\phi\phi}\right),\\
V_{\phi\phi} &= 8\mathcal{E}\left(c_b^2Q_{tt}/C_6^2-2s_bc_bQ_{t\psi}/C_6-2c_b^2Q_{t\phi}/C_6\right.\\
&\qquad\left.{}+s_b^2Q_{\psi\psi}+2s_bc_bQ_{\psi\phi}+c_b^2Q_{\phi\phi}\right). \tag*{(216)}
\end{aligned} K W ψ W ϕ V ψ ψ V ψ ϕ V ϕϕ = 8 E ( D ) , Y = − 8 E ( Q tt ) , = 8 E ( s b Q tt / C 6 − c b Q t ψ − s b Q tϕ ) , = 8 E ( c b Q tt / C 6 − s b Q t ψ − c b Q tϕ ) , = 8 E ( s b 2 Q tt / C 6 2 − 2 s b c b Q t ψ / C 6 − 2 s b 2 Q tϕ / C 6 + c b 2 Q ψ ψ + 2 s b c b Q ψ ϕ + s b 2 Q ϕϕ ) , = 8 E ( s b c b Q tt / C 6 2 − ( s b 2 + c b 2 ) Q t ψ / C 6 − 2 s b c b Q tϕ / C 6 + s b c b Q ψ ψ + ( s b 2 + c b 2 ) Q ψ ϕ + s b c b Q ϕϕ ) , = 8 E ( c b 2 Q tt / C 6 2 − 2 s b c b Q t ψ / C 6 − 2 c b 2 Q tϕ / C 6 + s b 2 Q ψ ψ + 2 s b c b Q ψ ϕ + c b 2 Q ϕϕ ) . (216)
The notation C 6 C_6 C 6 in these monomials means the polarisation factor with finite factor π 6 \pi_6 π 6 and power 1. It differs from the endpoint factor C C C . Their lowest powers of τ \tau τ are ( − 4 , − 4 , − 6 , − 6 , − 8 , − 8 , − 8 ) (-4,-4,-6,-6,-8,-8,-8) ( − 4 , − 4 , − 6 , − 6 , − 8 , − 8 , − 8 ) . Every contributing N N N is an integer between zero and seven, which proves that the stated finite polynomial degree suffices.
To see the identification with the earlier extremal construction, E \mathcal{E} E returns the τ 6 \tau^6 τ 6 entry of each scalar monomial. The displayed combinations are precisely the scalar expansion of f c T G C f^cTGC f c T GC , while λ − 3 = 8 τ − 6 \lambda^{-3}=8\tau^{-6} λ − 3 = 8 τ − 6 . Cancellation of the lower powers therefore gives K = f e K=f_e K = f e and the six densities used in § 4.2 .
The subfamily used for the global bounds. The ancillary Mathematica file evaluates these finite formulas for all parameters of the correlated ansatz. We then apply the physical normalisation (64) and impose balance. For Z = 5 Z=5 Z = 5 , r 0 = 37 / 50 r_0=37/50 r 0 = 37/50 , β = − 41 / 50 \beta=-41/50 β = − 41/50 , w = − 30 w=-30 w = − 30 , additional expanded polynomials retain ( u , y , c , H ∗ , v ∗ , d ) (u,y,c,H_*,v_*,d) ( u , y , c , H ∗ , v ∗ , d ) symbolically. These expanded expressions apply to this subfamily only.
Their common denominator, up to an overall positive constant, is
c 2 H 6 v 2 ( u y − y + 1 ) 4 ( u y − u − y ) 4 . (217) c^2H^6v^2(uy-y+1)^4(uy-u-y)^4. \tag*{(217)} c 2 H 6 v 2 ( u y − y + 1 ) 4 ( u y − u − y ) 4 . (217)
The denominator is nonzero for 0 < u , y < 1 0<u,y<1 0 < u , y < 1 and c H v ≠ 0 cHv\ne0 cH v = 0 . This statement concerns the scalar-density denominator. Regularity of the metric also requires the already studied zeros and boundary behaviour of K K K .
Analytic thermodynamic quantities
We compute the thermodynamic quantities in four steps. First, we write the current potential whose endpoint differences give the Komar charges. Second, we explain how to take finite values of the metric and potential on a rod, where direct substitution would give indeterminate ratios. Third, we evaluate these quantities on the three axes and two horizons. These values determine the charges, angular velocities, temperatures and entropies in (3.21)–(3.6). Finally, we extract the ADM charges independently at infinity. The formulas use only finite algebraic operations. Their denominators must be nonzero in the chosen chart, and balance and the regularity conditions must still be imposed.
Current potential and horizon charges
The matrix potential P \mathcal{P} P converts each current integral into a difference between its endpoint values, as in (45) . For this solution it can be written algebraically. We write M M M for the 6 × 3 6 \times3 6 × 3 matrix of dressing rows (only in this matrix formula), N = K T N = K^{\mathsf{T}} N = K T , and
s t = u 4 + u 6 − u 1 − u 3 , s ψ = − u 1 − u 2 − u 5 − 4 z − 2 , s ϕ = u 4 + u 6 − u 2 − u 3 − u 5 − 2 z . (218) \begin{aligned}
s_t &= u_4 + u_6 - u_1 - u_3,\\
s_\psi&= -u_1 - u_2 - u_5 - 4z - 2,\qquad s_\phi= u_4 + u_6 - u_2 - u_3 - u_5 - 2z. \tag*{(218)}
\end{aligned} s t s ψ = u 4 + u 6 − u 1 − u 3 , = − u 1 − u 2 − u 5 − 4 z − 2 , s ϕ = u 4 + u 6 − u 2 − u 3 − u 5 − 2 z . (218)
Then
P = − diag ( s t , s ψ , s ϕ ) − M T Γ − 1 N , ∂ z P = 2 x G − 1 ∂ x G , 2 ∂ x P = − G − 1 ∂ z G . (219) \mathcal{P} = -\operatorname{diag}(s_t,s_\psi,s_\phi) - M^{\mathsf{T}}\Gamma^{-1}N,\qquad\partial_z\mathcal{P} = 2xG^{-1}\partial_xG,\qquad2\partial_x\mathcal{P} = -G^{-1}\partial_zG. \tag*{(219)} P = − diag ( s t , s ψ , s ϕ ) − M T Γ − 1 N , ∂ z P = 2 x G − 1 ∂ x G , 2 ∂ x P = − G − 1 ∂ z G . (219)
These identities follow from the first inverse-spectral coefficient of the dressed generating matrix. We can also check them by differentiating (10) . Constant changes of Killing basis act by P ↦ A − 1 P A \mathcal{P} \mapsto A^{-1}\mathcal{P}A P ↦ A − 1 P A . The T T T row of this transformed potential is constant on each smooth spacelike axis.
Taking finite values on the rods
To resolve the indeterminate ratios on an axis, we use the normalised rows (B.2), factor the known powers of x = ρ 2 x = \rho^2 x = ρ 2 and retain the finite coefficients. For an analytic square matrix B ( x ) = B 0 + x B 1 + x 2 B 2 + ⋯ B(x) = B_0 + xB_1 + x^2B_2 + \cdots B ( x ) = B 0 + x B 1 + x 2 B 2 + ⋯ , define three explicit functions. Let B 0 [ j ← B k ] B_0[j \leftarrow B_k] B 0 [ j ← B k ] mean that only column j j j of B 0 B_0 B 0 is replaced by column j j j of B k B_k B k . Then
D 0 ( B ) = det B 0 , \mathcal{D}_0(B) = \det B_0, D 0 ( B ) = det B 0 ,
D 1 ( B ) = ∑ j det B 0 [ j ← B 1 ] , \mathcal{D}_1(B) = \sum_j \det B_0[j \leftarrow B_1], D 1 ( B ) = j ∑ det B 0 [ j ← B 1 ] ,
D 2 ( B ) = ∑ j det B 0 [ j ← B 2 ] + ∑ j < k det B 0 [ j ← B 1 , k ← B 1 ] . (220) \mathcal{D}_2(B) = \sum_j \det B_0[j \leftarrow B_2] + \sum_{j<k} \det B_0[j \leftarrow B_1,k \leftarrow B_1]. \tag*{(220)} D 2 ( B ) = j ∑ det B 0 [ j ← B 2 ] + j < k ∑ det B 0 [ j ← B 1 , k ← B 1 ] . (220)
By multilinearity these are exactly [ x n ] det B [x^n]\det B [ x n ] det B for n = 0 , 1 , 2 n = 0, 1, 2 n = 0 , 1 , 2 . We use the following evaluation positions:
z 0 = a 1 − 1 , z 2 = ( a 2 + a 3 ) / 2 , z 3 = ( a 3 + a 4 ) / 2 , z 4 = ( a 4 + a 5 ) / 2 , z 6 = a 6 + 1. (221) \begin{aligned}
z_0 &= a_1 - 1,\qquad z_2 = (a_2 + a_3)/2,\qquad z_3 = (a_3 + a_4)/2,\\
z_4 &= (a_4 + a_5)/2,\qquad z_6 = a_6 + 1. \tag*{(221)}
\end{aligned} z 0 z 4 = a 1 − 1 , z 2 = ( a 2 + a 3 ) /2 , z 3 = ( a 3 + a 4 ) /2 , = ( a 4 + a 5 ) /2 , z 6 = a 6 + 1. (221)
The unit offset is arbitrary in the fixed reference units. At z s z_s z s write D i = ∣ z s − a i ∣ > 0 D_i = |z_s-a_i| > 0 D i = ∣ z s − a i ∣ > 0 . The needed input coefficients are the finite binomial expressions
u i = 2 D i 1 i > s + ∑ n = 1 5 ( 1 / 2 n ) D i 1 − 2 n x n + O ( x 6 ) , R i j = x + u i u j . (222) u_i = 2D_i\,1_{i>s} + \sum_{n=1}^{5} \binom{1/2}{n}D_i^{1-2n}x^n + O(x^6),\qquad R_{ij} = x + u_i u_j. \tag*{(222)} u i = 2 D i 1 i > s + n = 1 ∑ 5 ( n 1/2 ) D i 1 − 2 n x n + O ( x 6 ) , R ij = x + u i u j . (222)
Known powers of x x x are factored before truncation and division. Although only quadratic coefficients of the final matrices are needed, keeping the five root coefficients in (B.33) preserves them through intermediate Laurent divisions. For pole-pair denominators use R i j R_{ij} R ij for two solitons, x R i j / ( u i u j ) xR_{ij}/(u_i u_j) x R ij / ( u i u j ) for two antisolitons, and, for a soliton at i i i and an antisoliton at j j j ,
x + p i p j = 2 x u i ( a j − a i ) R i j . (223) x+p_i p_j=\frac{2x u_i(a_j-a_i)}{R_{ij}}. \tag*{(223)} x + p i p j = R ij 2 x u i ( a j − a i ) . (223)
This exact identity avoids a subtraction of terms whose leading coefficients coincide.
Metric and potential on each rod
The index s s s labels the intervals containing the five positions z s z_s z s in (B.32). On each interval, the table below lists the powers of x x x needed to make the matrix entries finite. Form the diagonal matrices S = diag ( x α i ) S=\operatorname{diag}(x^{\alpha_i}) S = diag ( x α i ) , R = diag ( x β i ) R=\operatorname{diag}(x^{\beta_i}) R = diag ( x β i ) , and set
L = R − 1 S Γ S , U = R − 1 S N , V = S N , W = S M . (224) L=R^{-1}S\Gamma S,\quad U=R^{-1}SN,\quad V=SN,\quad W=SM. \tag*{(224)} L = R − 1 S Γ S , U = R − 1 S N , V = S N , W = S M . (224)
The powers and determinant orders are
Table 7.
Here det L = x d ( D d ( L ) + O ( x ) ) \det L=x^d(\mathcal{D}_d(L)+O(x)) det L = x d ( D d ( L ) + O ( x )) and D d ( L ) ≠ 0 \mathcal{D}_d(L)\ne0 D d ( L ) = 0 in this chart. The entries of L , U , W , x η V L,U,W,x^\eta V L , U , W , x η V and x η H x^\eta H x η H are analytic at x = 0 x=0 x = 0 . For a , b = t , ψ , ϕ a,b=t,\psi,\phi a , b = t , ψ , ϕ define two bordered matrices, of size 7 × 7 7\times7 7 × 7 ,
B a b = ( L U : b x η V : a T x η H a b ) , E a b = ( L U : b W : a T 0 ) . (225) B^{ab}=\begin{pmatrix} L & U_{:b}\\ x^\eta V_{:a}^{\mathrm T} & x^\eta H_{ab}\end{pmatrix},\qquad
E^{ab}=\begin{pmatrix} L & U_{:b}\\ W_{:a}^{\mathrm T} & 0\end{pmatrix}. \tag*{(225)} B ab = ( L x η V : a T U : b x η H ab ) , E ab = ( L W : a T U : b 0 ) . (225)
The desired metric and potential blocks are the explicit expressions
G a b [ s ] = D d + η ( B a b ) D d ( L ) , P a b [ s ] = D d ( E a b ) D d ( L ) − s a ( 0 ) δ a b . (226) G_{ab}^{[s]}=\frac{\mathcal{D}_{d+\eta}(B^{ab})}{\mathcal{D}_d(L)},\qquad
P_{ab}^{[s]}=\frac{\mathcal{D}_d(E^{ab})}{\mathcal{D}_d(L)}-s_a(0)\delta_{ab}. \tag*{(226)} G ab [ s ] = D d ( L ) D d + η ( B ab ) , P ab [ s ] = D d ( L ) D d ( E ab ) − s a ( 0 ) δ ab . (226)
Here the s a s_a s a are the three seed expressions in (B.30), evaluated by the constant root coefficients in (B.33). Notice that d + η ≤ 2 d+\eta\le2 d + η ≤ 2 for all five rows of the table.
To compute f [ s ] f^{[s]} f [ s ] , we remove the known powers of x x x from u i u_i u i and R i j R_{ij} R ij and retain their leading coefficients
u ˉ i = { [ 2 ( z s − a i ) ] − 1 , i ≤ s , 2 ( a i − z s ) , i > s , R ˉ i j = { 1 , i , j ≤ s , u ˉ i u ˉ j , i , j > s , 1 + u ˉ i u ˉ j , otherwise . (227) \bar{u}_i=
\begin{cases}
[2(z_s-a_i)]^{-1}, & i\le s,\\
2(a_i-z_s), & i>s,
\end{cases}
\qquad
\bar{R}_{ij}=
\begin{cases}
1, & i,j\le s,\\
\bar{u}_i\bar{u}_j, & i,j>s,\\
1+\bar{u}_i\bar{u}_j, & \text{otherwise}.
\end{cases} \tag*{(227)} u ˉ i = { [ 2 ( z s − a i ) ] − 1 , 2 ( a i − z s ) , i ≤ s , i > s , R ˉ ij = ⎩ ⎨ ⎧ 1 , u ˉ i u ˉ j , 1 + u ˉ i u ˉ j , i , j ≤ s , i , j > s , otherwise . (227)
The total power of x x x in f s f_s f s is zero on these rods. Thus
f s , 0 = u ˉ 2 u ˉ 5 u ˉ 1 ∏ i R ˉ i i − 1 ∏ i < j R ˉ i j − w i ⋅ w j , f [ s ] = f s , 0 D d ( L ) D d ( L 0 ) . (228) f_{s,0}=\frac{\bar{u}_2\bar{u}_5}{\bar{u}_1}\prod_i\bar{R}_{ii}^{-1}\prod_{i<j}\bar{R}_{ij}^{-w_i\cdot w_j},\qquad f^{[s]}=f_{s,0}\frac{\mathcal{D}_d(L)}{\mathcal{D}_d(L_0)}. \tag*{(228)} f s , 0 = u ˉ 1 u ˉ 2 u ˉ 5 i ∏ R ˉ ii − 1 i < j ∏ R ˉ ij − w i ⋅ w j , f [ s ] = f s , 0 D d ( L 0 ) D d ( L ) . (228)
L 0 L_0 L 0 is the same matrix with all six polarisation constants zero. The subscript s s s in f s , 0 f_{s,0} f s , 0 denotes the seed, not an extra free parameter. These formulas together with (17) fix the outer basis A A A , δ \delta δ , both horizon kernels, and every quantity in (47) , (29) , (31) and (32) .
The Schur determinant identity explains the cancellation:
det B a b det L = x n G a b , det E a b det L = − M T a Γ − 1 N : b . (229) \frac{\det B^{ab}}{\det L}=x^{n}G_{ab},\qquad\frac{\det E^{ab}}{\det L}=-M^{\mathsf{T}}{}_{a}\Gamma^{-1}N_{:b}. \tag*{(229)} det L det B ab = x n G ab , det L det E ab = − M T a Γ − 1 N : b . (229)
The powers in the table remove all entrywise poles. Exact lower-order coefficient cancellations then give (B.37). This also explains why a large expanded rational function has a short determinant representation. Individual charges use differences of the constant T T T rows on neighbouring axes, so the arbitrary positions on each connected axis do not enter the answer.
ADM charges at infinity
We check the ADM charges independently by expanding the metric at infinity. Let F a = f G t a F_a=fG_{ta} F a = f G t a , and put ξ = 1 / R W \xi=1/R_W ξ = 1/ R W , z = 0 z=0 z = 0 , x = ξ − 2 x=\xi^{-2} x = ξ − 2 . The collected Cauchy expansion of (10) –(12) is a finite sum of monomials in x x x , u i u_i u i , R i j R_{ij} R ij and endpoint/polarisation constants. For one monomial write its spatial part as x e x ∏ i u i e i ∏ i ≤ j R i j e i j x^{e_x}\prod_i u_i^{e_i}\prod_{i\leq j}R_{ij}^{e_{ij}} x e x ∏ i u i e i ∏ i ≤ j R ij e ij and its parameter coefficient as q q q . Define
n = − 2 e x − ∑ i e i − 2 ∑ i ≤ j e i j , W = q 2 ∑ e i j , n=-2e_x-\sum_i e_i-2\sum_{i\leq j}e_{ij},\qquad W=q\,2^{\sum e_{ij}}, n = − 2 e x − i ∑ e i − 2 i ≤ j ∑ e ij , W = q 2 ∑ e ij ,
L 1 = ∑ i e i a i + 1 2 ∑ i ≤ j e i j ( a i + a j ) . (230) L_1=\sum_i e_i a_i+\frac{1}{2}\sum_{i\leq j}e_{ij}(a_i+a_j). \tag*{(230)} L 1 = i ∑ e i a i + 2 1 i ≤ j ∑ e ij ( a i + a j ) . (230)
These follow from u i = ξ − 1 ( 1 + a i ξ + O ( ξ 2 ) ) u_i=\xi^{-1}(1+a_i\xi+O(\xi^2)) u i = ξ − 1 ( 1 + a i ξ + O ( ξ 2 )) and R i j = ξ − 2 ( 2 + ( a i + a j ) ξ + O ( ξ 2 ) ) R_{ij}=\xi^{-2}(2+(a_i+a_j)\xi+O(\xi^2)) R ij = ξ − 2 ( 2 + ( a i + a j ) ξ + O ( ξ 2 )) . For each of f f f , F t F_t F t , F ψ F_\psi F ψ , F ϕ F_\phi F ϕ all collected monomials have n ≥ 1 n\geq1 n ≥ 1 and hence
f = f 1 ξ + f 2 ξ 2 + O ( ξ 3 ) , F a = h a 1 ξ + h a 2 ξ 2 + O ( ξ 3 ) , f=f_1\xi+f_2\xi^2+O(\xi^3),\qquad F_a=h_{a1}\xi+h_{a2}\xi^2+O(\xi^3), f = f 1 ξ + f 2 ξ 2 + O ( ξ 3 ) , F a = h a 1 ξ + h a 2 ξ 2 + O ( ξ 3 ) ,
( c 1 , c 2 ) = ( ∑ n = 1 W , ∑ n = 1 W L 1 + ∑ n = 2 W ) . (231) (c_1,c_2)=\left(\sum_{n=1}W,\sum_{n=1}WL_1+\sum_{n=2}W\right). \tag*{(231)} ( c 1 , c 2 ) = ( n = 1 ∑ W , n = 1 ∑ W L 1 + n = 2 ∑ W ) . (231)
where ( c 1 , c 2 ) (c_1,c_2) ( c 1 , c 2 ) denotes the coefficient pair of the chosen function. Only the indicated monomials contribute. We set
b a = h a 2 f 1 − h a 1 f 2 f 1 2 , b ^ b = ∑ a b a A a b . (232) b_a=\frac{h_{a2}f_1-h_{a1}f_2}{f_1^2},\qquad\widehat{b}_b=\sum_a b_a A^a{}_b. \tag*{(232)} b a = f 1 2 h a 2 f 1 − h a 1 f 2 , b b = a ∑ b a A a b . (232)
Then the independent ADM expressions in the A A A convention are
M = 3 π δ 4 G 5 b ^ T , J Ψ = − π δ G 5 b ^ Ψ , J Φ = − π δ G 5 b ^ Φ . (233) M=\frac{3\pi\delta}{4G_5}\widehat{b}_T,\qquad J_\Psi=-\frac{\pi\delta}{G_5}\widehat{b}_\Psi,\qquad J_\Phi=-\frac{\pi\delta}{G_5}\widehat{b}_\Phi. \tag*{(233)} M = 4 G 5 3 π δ b T , J Ψ = − G 5 π δ b Ψ , J Φ = − G 5 π δ b Φ . (233)
Equation (47) is the more compact standalone formula, while the monomial form is useful as an independent implementation and conservation check.
The same expansion can be made at fixed z / R W z/R_W z / R W . The leading coefficients of f f f and f G t a fG_{t a} f G t a are independent of this angle, while those of the angular densities are affine in z / R W z/R_W z / R W . We also find [ R W − 1 ] ( f G t t ) = − [ R W − 1 ] f [R_W^{-1}](fG_{tt})=-[R_W^{-1}]f [ R W − 1 ] ( f G tt ) = − [ R W − 1 ] f . Together with the outer-axis normalisation, these identities give the flat leading metric in the coordinates of § 3.3 .
Regularity criteria and verified domains
We express the geometric criteria of § 3.6 as explicit sign tests. We first give formulas valid for the full six-parameter construction, then state which regions have been verified by interval bounds and which have only been sampled.
Exterior smoothness and causality
All elementary denominators in the raw dressing rows are nonzero for x > 0 x>0 x > 0 and ordered endpoints. In particular u i > 0 u_i>0 u i > 0 , x + u i u j > 0 x+u_i u_j>0 x + u i u j > 0 , and a mixed-pole denominator is
x + u s ( − x / u t ) = 2 x u s ( a t − a s ) x + u s u t ≠ 0. (234) x+u_s(-x/u_t)=\frac{2xu_s(a_t-a_s)}{x+u_su_t}\ne0. \tag*{(234)} x + u s ( − x / u t ) = x + u s u t 2 x u s ( a t − a s ) = 0. (234)
Let D = det Γ D=\det\Gamma D = det Γ and D 0 = det Γ 0 D_0=\det\Gamma_0 D 0 = det Γ 0 . The latter splits into Cauchy blocks on the groups { 3 , 4 } \{3,4\} { 3 , 4 } , { 1 } \{1\} { 1 } and { 2 , 5 , 6 } \{2,5,6\} { 2 , 5 , 6 } . Using the Cauchy product (B.6) and the negative diagonal entries of H H H , we find block signs − - − , − - − , + + + . Thus D 0 > 0 D_0>0 D 0 > 0 throughout the off-axis domain. Since f s > 0 f_s>0 f s > 0 , positivity of the physical base factor is equivalent to D > 0 D>0 D > 0 .
To express angular positivity without inverting the full metric, use the unit-determinant basis B B B in (26) and write H ′ = B T H B H'=B^{\mathsf T}HB H ′ = B T H B , L = B T K L=B^{\mathsf T}K L = B T K , where K K K is the 3 × 6 3\times6 3 × 6 matrix in (10) . Let L Ψ L_\Psi L Ψ denote its Ψ \Psi Ψ row and L a n g L_{\mathrm{ang}} L ang its two angular rows. Define
D Ψ = det ( Γ L Ψ T L Ψ H Ψ Ψ ′ ) , D T 2 = det ( Γ L a n g T L a n g H a n g ′ ) . (235) D_\Psi=\det\begin{pmatrix}\Gamma& L_\Psi^{\mathsf T}\\ L_\Psi& H'_{\Psi\Psi}\end{pmatrix},
\qquad
D_{T^2}=\det\begin{pmatrix}\Gamma& L_{\mathrm{ang}}^{\mathsf T}\\ L_{\mathrm{ang}} & H'_{\mathrm{ang}}\end{pmatrix}. \tag*{(235)} D Ψ = det ( Γ L Ψ L Ψ T H ΨΨ ′ ) , D T 2 = det ( Γ L ang L ang T H ang ′ ) . (235)
Schur complementation gives D Ψ = D G ‾ Ψ Ψ D_\Psi=D\overline{G}_{\Psi\Psi} D Ψ = D G ΨΨ and D T 2 = D det Λ D_{T^2}=D\det\Lambda D T 2 = D det Λ . An explicit sufficient set of exterior conditions is therefore
D > 0 , D Ψ > 0 , D T 2 > 0 for all x > 0 , z ∈ R . (236) D>0,\qquad D_\Psi>0,\qquad D_{T^2}>0\quad\text{for all }x>0,\ z\in\mathbb{R}. \tag*{(236)} D > 0 , D Ψ > 0 , D T 2 > 0 for all x > 0 , z ∈ R . (236)
These determinants have orders six, seven and eight and retain all six free parameters. They ensure a smooth Lorentzian metric off the axis and a positive angular block. In particular, g T T = − D T 2 / ( x D ) < 0 g^{TT}=-D_{T^2}/(xD)<0 g T T = − D T 2 / ( x D ) < 0 , so the physical time T T T is a time function and there are no closed causal curves in this domain. Conical balance is a separate boundary condition, given by (16) . The axes and horizons also require the regular coordinates described below. They are sufficient, rather than a minimal characterisation: removable cancellations could admit a different representation beyond a domain where this unreduced determinant stays nonzero.
The inequalities must hold at every exterior point. To express this requirement algebraically, we use u i 2 + 2 ( z − a i ) u i = x u_i^2+2(z-a_i)u_i=x u i 2 + 2 ( z − a i ) u i = x , u i > 0 u_i>0 u i > 0 , and positive square-root variables for the normalisation. Multiplying denominators by their squares preserves signs. This gives a finite system of polynomial equations and inequalities in the parameters and spatial variables, supplemented by the axis and horizon conditions. We have not eliminated the spatial variables to obtain a closed list of elementary parameter inequalities for all six parameters.
Balance roots and the nonextremal parameter domain
We impose the three balance equations (16) : the middle axis must have the same degenerating Killing direction and cone factor as the lower axis. Since both outer angles have period 2 π 2\pi 2 π , these conditions remove the remaining conical defect and align the angular identifications. We have verified balance in a nonextremal neighbourhood as follows. Write p = ( a 2 − a 1 , … , a 6 − a 5 , C 2 ) p=(a_2-a_1,\ldots,a_6-a_5,C_2) p = ( a 2 − a 1 , … , a 6 − a 5 , C 2 ) and p ∗ = ( 1 , 1 , 2 , 1 , 0.1 , − 0.1 ) p_*=(1,1,2,1,0.1,-0.1) p ∗ = ( 1 , 1 , 2 , 1 , 0.1 , − 0.1 ) . For
∣ p i − p ∗ i ∣ ≤ 10 − 9 ( i = 1 , … , 6 ) , (237) \lvert p_i-p_{*i}\rvert\leq10^{-9}\quad(i=1,\ldots,6), \tag*{(237)} ∣ p i − p ∗ i ∣ ≤ 1 0 − 9 ( i = 1 , … , 6 ) , (237)
the exact balance equations have a unique enclosed analytic root ( C 3 , B 4 , B 5 ) (C_3,B_4,B_5) ( C 3 , B 4 , B 5 ) near the reference values
( a 1 , … , a 6 ) = ( 0 , 1 , 2 , 4 , 5 , 5.1 ) , C 2 = − 0.1 , (a_1,\ldots,a_6)=(0,1,2,4,5,5.1),\qquad C_2=-0.1, ( a 1 , … , a 6 ) = ( 0 , 1 , 2 , 4 , 5 , 5.1 ) , C 2 = − 0.1 ,
C 3 ≃ 0.833176774260059 , B 4 ≃ 0.612383796741305 , C_3 \simeq0.833176774260059,\qquad B_4 \simeq0.612383796741305, C 3 ≃ 0.833176774260059 , B 4 ≃ 0.612383796741305 ,
B 5 ≃ 0.058168784076090. (238) B_5 \simeq0.058168784076090. \tag*{(238)} B 5 ≃ 0.058168784076090. (238)
These rounded values locate the root. We use the exact balance solution in the checks. Interval bounds prove regularity on every finite boundary component: both horizons, all four poles, all ordinary axes and both auxiliary points. Cone factors are unity for the primitive 2 π 2\pi 2 π cycles. The proof also covers a compact exterior rectangle, but does not cover the whole exterior for every parameter in this nonextremal box.
We use interval arithmetic to verify that the balance root exists for every parameter choice in (B.48). We enclose the three unknown polarisations and the balance Jacobian, including rounding errors, and apply the interval Newton method in its Krawczyk form [36 ] . The resulting map sends the proposed root box strictly inside itself, with a nonsingular Jacobian enclosure. This proves existence and uniqueness of the nearby root. A root-box radius of 10 − 8 10^{-8} 1 0 − 8 suffices here. The implicit function theorem then makes the three polarisations analytic functions of the six free parameters.
Nonextremal horizon and endpoint charts
For completeness, the ingoing extension (35) has a two-sided version. We use the coefficients α \alpha α , a a a , β A \beta_A β A and λ A B \lambda_{AB} λ A B of section 3.1. For both horizon branches simultaneously, let
U = − ρ e − κ T , V = ρ e κ T , x = − U V , ω = U d V − V d U . (239) U=-\rho e^{-\kappa T},\qquad V=\rho e^{\kappa T},\qquad x=-UV,\qquad\omega=U\,\mathrm{d}V-V\,\mathrm{d}U. \tag*{(239)} U = − ρ e − κ T , V = ρ e κ T , x = − U V , ω = U d V − V d U . (239)
The corresponding metric is
d s 2 = − f d U d V + f d z 2 + λ A B d θ A d θ B − a 4 κ ω 2 − β A κ ω d θ A , det g = − δ 2 f 2 4 κ 2 . (240) \mathrm{d}s^2=-f\,\mathrm{d}U\mathrm{d}V+f\,\mathrm{d}z^2+\lambda_{AB}\mathrm{d}\theta^A\mathrm{d}\theta^B-\frac{a}{4\kappa}\omega^2-\frac{\beta_A}{\kappa}\omega\mathrm{d}\theta^A,\qquad\det g=-\frac{\delta^2 f^2}{4\kappa^2}. \tag*{(240)} d s 2 = − f d U d V + f d z 2 + λ A B d θ A d θ B − 4 κ a ω 2 − κ β A ω d θ A , det g = − 4 κ 2 δ 2 f 2 . (240)
To check the axes, horizon endpoints and auxiliary points, we expand the exact scalar expressions for f f f and f G fG f G in powers of a local radial coordinate. This exposes cancellations at the boundary before any numerical evaluation. On each ordinary axis, positivity of the surviving block, the exact kernel and the unit cone identity give regular Cartesian polar tensors. At the two auxiliary seams, the root involutions remove the nonanalytic square-root branch and the certified denominators ensure bounded, nondegenerate coefficients.
For a physical endpoint a i a_i a i put y i = z − a i y_i=z-a_i y i = z − a i and
X = ρ 2 + y i 2 − σ y i , Y = ρ 2 + y i 2 + σ y i , x = X Y , F c = 1 4 ( X + Y ) f . (241) X=\sqrt{\rho^2+y_i^2}-\sigma y_i,\qquad Y=\sqrt{\rho^2+y_i^2}+\sigma y_i,\qquad x=XY,\qquad F_c=\frac{1}{4}(X+Y)f. \tag*{(241)} X = ρ 2 + y i 2 − σ y i , Y = ρ 2 + y i 2 + σ y i , x = X Y , F c = 4 1 ( X + Y ) f . (241)
Choose ( i , σ ) = ( 2 , + 1 ) , ( 3 , − 1 ) , ( 4 , + 1 ) , ( 5 , − 1 ) (i,\sigma)=(2,+1),(3,-1),(4,+1),(5,-1) ( i , σ ) = ( 2 , + 1 ) , ( 3 , − 1 ) , ( 4 , + 1 ) , ( 5 , − 1 ) , so that X = 0 X=0 X = 0 is the horizon and Y = 0 Y=0 Y = 0 the axis. With ϑ \vartheta ϑ the degenerating corotating angle and ζ \zeta ζ the surviving one, the analytic Killing entries have the factors
g T T = − X a , g T ζ = X b c , g T ϑ = X Y e c , g ϑ ϑ = Y c c , g ϑ ζ = Y d c , g ζ ζ = L c , a = 4 κ 2 F c + X a 1 , c c = 4 F c + Y h c . (242) \begin{aligned}
g_{TT}&=-Xa, & g_{T\zeta}&=Xb_c, & g_{T\vartheta}&=XYe_c,\\
g_{\vartheta\vartheta}&=Yc_c, & g_{\vartheta\zeta}&=Yd_c, & g_{\zeta\zeta}&=L_c,\\
a&=4\kappa^2F_c+Xa_1, & c_c&=4F_c+Yh_c. \tag*{(242)}
\end{aligned} g T T g ϑϑ a = − X a , = Y c c , = 4 κ 2 F c + X a 1 , g T ζ g ϑζ c c = X b c , = Y d c , = 4 F c + Y h c . g T ϑ g ζ ζ = X Y e c , = L c , (242)
The base metric is F c ( d X 2 / X + d Y 2 / Y ) F_c(\mathrm{d}X^2/X+\mathrm{d}Y^2/Y) F c ( d X 2 / X + d Y 2 / Y ) . We factor these powers in the exact expressions before evaluating X = Y = 0 X=Y=0 X = Y = 0 , so the cancellations hold at the axis–horizon junction as well as along each edge. We then set
U c = X e − κ T , V c = X e κ T , P = Y cos ϑ , Q = Y sin ϑ , (243) U_c=\sqrt{X}e^{-\kappa T},\qquad V_c=\sqrt{X}e^{\kappa T},\qquad P=\sqrt{Y}\cos\vartheta,\qquad Q=\sqrt{Y}\sin\vartheta, \tag*{(243)} U c = X e − κ T , V c = X e κ T , P = Y cos ϑ , Q = Y sin ϑ , (243)
and define ω c = U c d V c − V c d U c \omega_c=U_c\mathrm{d}V_c-V_c\mathrm{d}U_c ω c = U c d V c − V c d U c , η c = P d Q − Q d P \eta_c=P\mathrm{d}Q-Q\mathrm{d}P η c = P d Q − Q d P . The joint metric is
d s 2 = − 4 F c d U c d V c + 4 F c ( d P 2 + d Q 2 ) + L c d ζ 2 − a 1 4 κ 2 ω c 2 − b c κ ω c d ζ − e c κ ω c η c + h c η c 2 + 2 d c η c d ζ . (244) \begin{aligned}
\mathrm{d}s^2={}&-4F_c\mathrm{d}U_c\mathrm{d}V_c+4F_c(\mathrm{d}P^2+\mathrm{d}Q^2)+L_c\mathrm{d}\zeta^2-\frac{a_1}{4\kappa^2}\omega_c^2-\frac{b_c}{\kappa}\omega_c\mathrm{d}\zeta\\
&-\frac{e_c}{\kappa}\omega_c\eta_c+h_c\eta_c^2+2d_c\eta_c\mathrm{d}\zeta. \tag*{(244)}
\end{aligned} d s 2 = − 4 F c d U c d V c + 4 F c ( d P 2 + d Q 2 ) + L c d ζ 2 − 4 κ 2 a 1 ω c 2 − κ b c ω c d ζ − κ e c ω c η c + h c η c 2 + 2 d c η c d ζ . (244)
At the joint origin its determinant is − 64 F c 4 L c < 0 -64F_c^4L_c<0 − 64 F c 4 L c < 0 . All four F c , L c F_c,L_c F c , L c factors have positive interval enclosures on the small six-parameter box. The compact exterior certificate for that box covers only e 1 / 4 ≤ ρ ≤ e 3 / 4 e^{1/4}\leq\rho\leq e^{3/4} e 1/4 ≤ ρ ≤ e 3/4 , 2.75 ≤ z ≤ 3.25 2.75\leq z\leq3.25 2.75 ≤ z ≤ 3.25 , using 512 cells. It is not an exterior cover to infinity.
Balanced regular extremal examples
To exhibit regular solutions, fix four of the constants in the limiting ansatz:
Z = 5 , r 0 = 37 / 50 , β = − 41 / 50 , w = − 30 , (245) Z=5,\qquad r_0=37/50,\qquad\beta=-41/50,\qquad w=-30, \tag*{(245)} Z = 5 , r 0 = 37/50 , β = − 41/50 , w = − 30 , (245)
and solve the three exact limiting-axis balance equations for ( c , H ∗ , v ∗ ) (c,H_\ast,v_\ast) ( c , H ∗ , v ∗ ) . For ∣ d − 7 / 500 ∣ ≤ 10 − 12 |d-7/500|\leq10^{-12} ∣ d − 7/500∣ ≤ 1 0 − 12 the root is uniquely enclosed about
c ≃ 0.0048605562993129358582 , c\mathrel{\simeq}0.0048605562993129358582, c ≃ 0.0048605562993129358582 ,
H ∗ ≃ 3.1504581717070573003 , v ∗ ≃ − 16.1961121853210728223. (246) H_\ast\mathrel{\simeq}3.1504581717070573003,\qquad v_\ast\mathrel{\simeq}-16.1961121853210728223. \tag*{(246)} H ∗ ≃ 3.1504581717070573003 , v ∗ ≃ − 16.1961121853210728223. (246)
These decimals locate the branch. The metric is defined by the exact balance equations, whose solution is enclosed by interval arithmetic. The enclosure proves that for each d d d in this interval there is one nearby balanced solution and that it depends analytically on d d d . The verification method is given in the following subsection.
Proposition B.1. For (B.56) and ∣ d − 7 / 500 ∣ ≤ 10 − 12 \lvert d - 7/500\rvert\le10^{-12} ∣ d − 7/500 ∣ ≤ 1 0 − 12 , the exactly balanced coefficient metric has a smooth asymptotically flat vacuum exterior with two finite-area degenerate horizons of topologies S 1 × S 2 S^1 \times S^2 S 1 × S 2 and S 3 S^3 S 3 . Its exterior is stably causal and has no conical or orbifold defects or exterior curvature singularities. Both horizons, including their poles, have smooth local extensions. All proper lengths may be multiplied by an arbitrary positive scale.
We prove this by combining exact algebraic cancellations, axis balance and the local charts of section 4.3 with interval bounds over the whole exterior. The interval calculations bound the relevant functions at every point in each cell and include rounding errors. We describe the inequalities and their treatment at the boundaries below. The result covers this small one-dimensional parameter interval, together with the free overall scale. It does not establish regularity for arbitrary values of all constants in the extremal ansatz.
The same complete checks pass at seven separated values of d d d listed in Table 3 . At each value, interval arithmetic isolates the balance solution and verifies the same denominator, signature and causality inequalities throughout the exterior, including the axis, horizon and asymptotic limits. Their ring areas range from a very small probe to a component larger than the central horizon.
These examples establish the existence of a regular extremal central hole surrounded by a regular extremal ring.
For the parameters in proposition B.1 and Table 3 , interval bounds keep F 0 F_0 F 0 , C 2 C^2 C 2 and the required denominators positive for every polar angle, including the pole limits. They therefore establish a regular extension over each complete horizon. We also checked the signed-radius and polar extensions numerically (footnote 1).
Complete extremal exterior bounds
Here we verify the inequalities throughout the unbounded exterior. The compact coordinates (60) turn this task into a positivity test on a unit square, with the axes, horizons and infinity on its boundary. The inverse map is
ρ 2 = 4 Z 2 u y ( 1 − y ) ( 1 − u ) 2 , z = Z ( 1 + u ) y − u 1 − u , U = 2 Z u ( 1 − y ) 1 − u , V = 2 Z ( 1 − y ) 1 − u . (247) \rho^2 = \frac{4Z^2uy(1-y)}{(1-u)^2}, \qquad z = Z\frac{(1+u)y-u}{1-u}, \qquad U = \frac{2Zu(1-y)}{1-u}, \qquad V = \frac{2Z(1-y)}{1-u}. \tag*{(247)} ρ 2 = ( 1 − u ) 2 4 Z 2 u y ( 1 − y ) , z = Z 1 − u ( 1 + u ) y − u , U = 1 − u 2 Z u ( 1 − y ) , V = 1 − u 2 Z ( 1 − y ) . (247)
The axes are y = 0 , 1 y=0,1 y = 0 , 1 and u = 0 u=0 u = 0 , the two horizons are ( 0 , 0 ) , ( 0 , 1 ) (0,0),(0,1) ( 0 , 0 ) , ( 0 , 1 ) , and infinity is u = 1 u=1 u = 1 . For a common polynomial denominator Q e Q_e Q e , write G ˉ e , a b = N a b / Q e \bar{G}_{e,ab}=N_{ab}/Q_e G ˉ e , ab = N ab / Q e and f ˉ e = N f / Q e \bar{f}_e=N_f/Q_e f ˉ e = N f / Q e . Sign-independent sufficient polynomial conditions on the open square are
Q e 2 > 0 , N f Q e > 0 , N Ψ Ψ Q e > 0 , N Ψ Ψ N Φ Φ − N Ψ Φ 2 > 0. (248) Q_e^2 > 0, \qquad N_fQ_e > 0, \qquad N_{\Psi\Psi}Q_e > 0, \qquad N_{\Psi\Psi}N_{\Phi\Phi}-N_{\Psi\Phi}^2 > 0. \tag*{(248)} Q e 2 > 0 , N f Q e > 0 , N ΨΨ Q e > 0 , N ΨΨ N ΦΦ − N ΨΦ 2 > 0. (248)
Equivalent tests can use the denominator of each component separately to keep the polynomials smaller. On the boundary, known factors such as the squared radius of a shrinking circle must first be removed using the exact axis or horizon expansions.
We express each polynomial in a Bernstein basis on the square. The basis functions are nonnegative and sum to one, so positive lower bounds on all coefficients imply positivity at every point. We enclose the coefficients using outward-rounded Arb arithmetic [37 ] , keeping exact zeros distinct from intervals that contain zero. Near one horizon corner, a potentially negative mixed coefficient is bounded by positive neighbouring terms. After removing the known boundary zeros, the remaining terms give strict positivity in the interior. We check the signs of the remaining factors on each open axis. The exact limiting-axis kernels and cone factors fix the global angular lattice and time coordinate. Closed-angle horizon bounds verify the denominators, F 0 F_0 F 0 , C 2 C^2 C 2 , and det Λ 0 / [ s ( 1 − s ) ] \det\Lambda_0/[s(1-s)] det Λ 0 / [ s ( 1 − s )] with s = ( 1 + cos θ ) / 2 s=(1+\cos\theta)/2 s = ( 1 + cos θ ) /2 . The signed horizon and polar charts of § 4.3 then supply the corner cover. All-angle asymptotic expansions complete the proof at infinity.
These steps prove proposition B.1 and the analogous statement at each root in Table 3 . They use the exact balance roots, not rounded substitutions. They establish exterior regularity and local horizon extension.
Scope of the broad numerical search
We performed a wider numerical scan by varying all six parameters independently as
p = p ∗ ⊙ ( 1 + ξ 1 , … , 1 + ξ 6 ) , ∣ ξ i ∣ ≤ 0.30. (249) p = p_\ast\mathbin{\odot} (1+\xi_1,\ldots,1+\xi_6), \qquad|\xi_i| \le0.30. \tag*{(249)} p = p ∗ ⊙ ( 1 + ξ 1 , … , 1 + ξ 6 ) , ∣ ξ i ∣ ≤ 0.30. (249)
The scan comprises 321 balanced states: the centre, all 64 corners and 256 Halton points, with 4,410 exterior samples per state. Logarithmic radial and endpoint sampling resolves the rods and approaches the horizons, axes and infinity, while axis and cap tests use their algebraic limits. All sampled horizons have positive area and surface gravity, and all axis, cone and cap tests pass. Independent evaluations of the original BZ metric resolve ill-conditioned points and confirm all retained sign minima. The smallest sampled value of det Λ / ( G Ψ Ψ G Φ Φ ) \det\Lambda/(G_{\Psi\Psi}G_{\Phi\Phi}) det Λ/ ( G ΨΨ G ΦΦ ) is approximately 0.8768, with − g T T > 1 -g^{TT}>1 − g T T > 1 at every sample. The sampled reduced spins include both signs of the second ADM angular momentum. This provides numerical evidence for a regular nonextremal regime with all six parameters varying independently.
Blackfold construction of black Saturns
We construct a thin black ring around a Myers–Perry hole by bending a Kerr black string into a circle. We use the blackfold equations of [15 , 16 ] and the treatment of transverse rotation in [14 ] . Keeping the string temperature nonzero gives the thermal configurations shown in Figure 2 . Taking it to zero recovers the extremal construction of [14 ] .
We first impose radial force balance. We then compute the charges and, when required, impose equality of the two horizon temperatures and both angular velocities. These are leading blackfold predictions, whose range of validity we discuss below.
Embedding, conventions and local thermodynamics
Let the MP mass parameter be μ \mu μ , with spins a a a in the ring plane and b b b in the transverse plane. In the coordinates of [14 ] , we embed the string by
t = τ , r = R , θ = 0 , χ = φ , 0 ≤ φ < 2 π . (250) t=\tau,\qquad r=R,\qquad\theta=0,\qquad\chi=\varphi,\qquad0\leq\varphi<2\pi. \tag*{(250)} t = τ , r = R , θ = 0 , χ = φ , 0 ≤ φ < 2 π . (250)
The angle conventions are ( χ , ψ ) A H O = ( Ψ , Φ ) (\chi,\psi)_{\mathrm{AHO}}=(\Psi,\Phi) ( χ , ψ ) AHO = ( Ψ , Φ ) and their background mass parameter satisfies m A H O 2 = μ m_{\mathrm{AHO}}^{2}=\mu m AHO 2 = μ . Their potentials ( Ω , Ω ^ ) (\Omega,\widehat{\Omega}) ( Ω , Ω ) are our ( Ω R H , Ω R F ) (\Omega_{\mathrm{R}}^{\mathrm{H}},\Omega_{\mathrm{R}}^{\mathrm{F}}) ( Ω R H , Ω R F ) . Below, m m m denotes the local Kerr-string mass scale and must not be confused with m A H O m_{\mathrm{AHO}} m AHO . The induced metric on the string and its useful functions are
γ a b d σ a d σ b = − N 2 d τ 2 + R 2 ( d φ − ϖ d τ ) 2 , − γ = N R , \gamma_{ab}\mathrm{d}\sigma^{a}\mathrm{d}\sigma^{b}=-N^{2}\mathrm{d}\tau^{2}+\mathcal{R}^{2}(\mathrm{d}\varphi-\varpi\mathrm{d}\tau)^{2},\qquad\sqrt{-\gamma}=N\mathcal{R}, γ ab d σ a d σ b = − N 2 d τ 2 + R 2 ( d φ − ϖ d τ ) 2 , − γ = N R ,
Σ R = R 2 + b 2 , Δ R = ( R 2 + a 2 ) ( R 2 + b 2 ) − μ R 2 , \Sigma_{R}=R^{2}+b^{2},\qquad\Delta_{R}=(R^{2}+a^{2})(R^{2}+b^{2})-\mu R^{2}, Σ R = R 2 + b 2 , Δ R = ( R 2 + a 2 ) ( R 2 + b 2 ) − μ R 2 ,
R 2 = R 2 + a 2 + μ a 2 Σ R , N 2 = Δ R Σ R R 2 , ϖ = μ a Σ R R 2 . (251) \mathcal{R}^{2}=R^{2}+a^{2}+\frac{\mu a^{2}}{\Sigma_{R}},\qquad N^{2}=\frac{\Delta_{R}}{\Sigma_{R}\mathcal{R}^{2}},\qquad\varpi=\frac{\mu a}{\Sigma_{R}\mathcal{R}^{2}}. \tag*{(251)} R 2 = R 2 + a 2 + Σ R μ a 2 , N 2 = Σ R R 2 Δ R , ϖ = Σ R R 2 μ a . (251)
Thus R \mathcal{R} R is the proper radius of the circular orbit and ϖ \varpi ϖ is the background angular shift. For k = ∂ τ + Ω ∂ φ k=\partial_{\tau}+\Omega\partial_{\varphi} k = ∂ τ + Ω ∂ φ , put
v = R ( Ω − ϖ ) N , Γ = ( 1 − v 2 ) − 1 / 2 , k = ∣ k ∣ = N Γ . (252) v=\frac{\mathcal{R}(\Omega-\varpi)}{N},\qquad\Gamma=(1-v^{2})^{-1/2},\qquad k=|k|=\frac{N}{\Gamma}. \tag*{(252)} v = N R ( Ω − ϖ ) , Γ = ( 1 − v 2 ) − 1/2 , k = ∣ k ∣ = Γ N . (252)
Equivalently,
k 2 = 1 − Ω 2 ( R 2 + a 2 ) − μ ( 1 − a Ω ) 2 R 2 + b 2 . (253) k^{2}=1-\Omega^{2}(R^{2}+a^{2})-\frac{\mu(1-a\Omega)^{2}}{R^{2}+b^{2}}. \tag*{(253)} k 2 = 1 − Ω 2 ( R 2 + a 2 ) − R 2 + b 2 μ ( 1 − a Ω ) 2 . (253)
We require R > r + , H R>r_{+,\mathrm{H}} R > r + , H and k 2 > 0 k^{2}>0 k 2 > 0 . The local seed has Kerr rotation parameter b s = m α b_{\mathrm{s}}=m\alpha b s = m α , ∣ α ∣ ≤ 1 |\alpha|\leq1 ∣ α ∣ ≤ 1 , and horizon radius
σ = m 1 − α 2 , r s = m + σ . (254) \sigma=m\sqrt{1-\alpha^{2}},\qquad r_{\mathrm{s}}=m+\sigma. \tag*{(254)} σ = m 1 − α 2 , r s = m + σ . (254)
Its pressure, local temperature and transverse angular velocity are
P = − m 2 G 5 , T = σ 4 π m r s , ω l o c = α 2 r s . (255) P=-\frac{m}{2G_{5}},\qquad\mathcal{T}=\frac{\sigma}{4\pi m r_{\mathrm{s}}},\qquad\omega_{\mathrm{loc}}=\frac{\alpha}{2r_{\mathrm{s}}}. \tag*{(255)} P = − 2 G 5 m , T = 4 π m r s σ , ω loc = 2 r s α . (255)
In [14 ] the local horizon radius r 0 r_{0} r 0 and Kerr parameter b ^ \widehat{b} b are therefore ( r s , m α ) (r_{\mathrm{s}},m\alpha) ( r s , m α ) . Stationarity redshifts the local potentials as T = T / k \mathcal{T} = T/k T = T / k and ω l o c = Ω ^ / k \omega_{\mathrm{loc}}=\widehat{\Omega}/k ω loc = Ω / k . For κ = 2 π T \kappa=2\pi T κ = 2 π T and D = κ 2 + Ω ^ 2 D=\sqrt{\kappa^{2}+\widehat{\Omega}^{2}} D = κ 2 + Ω 2 , inverting these relations gives
α = Ω ^ D , m = k 2 ( D + κ ) , r s = k 2 D . (256) \alpha=\frac{\widehat{\Omega}}{D},\qquad m=\frac{k}{2(D+\kappa)},\qquad r_{\mathrm{s}}=\frac{k}{2D}. \tag*{(256)} α = D Ω , m = 2 ( D + κ ) k , r s = 2 D k . (256)
Here κ ≥ 0 \kappa\ge0 κ ≥ 0 and the sign of Ω ^ \hat{\Omega} Ω ^ fixes the transverse spin orientation.
Stationary action and mechanical balance
At leading order we use the stationary free-energy functional of [14 , 15 ] . For our circular embedding it is the pressure integral
F = − ∫ 0 2 π d φ − γ P = π N R m G 5 = π N R k 2 G 5 ( D + κ ) . (257) \mathcal{F}=-\int_{0}^{2\pi}\mathrm{d}\varphi\sqrt{-\gamma}\,P=\frac{\pi N R m}{G_{5}}=\frac{\pi N R k}{2G_{5}(D+\kappa)}. \tag*{(257)} F = − ∫ 0 2 π d φ − γ P = G 5 π N R m = 2 G 5 ( D + κ ) π N R k . (257)
At T = 0 T=0 T = 0 it becomes F = π N R k / ( 2 G 5 ∣ Ω ^ ∣ ) \mathcal{F}=\pi N R k/(2G_{5}|\hat{\Omega}|) F = π N R k / ( 2 G 5 ∣ Ω ^ ∣ ) , the leading extremal functional of [14 ] . Spin-connection and self-gravity corrections are omitted consistently at this order. We vary R R R while holding ( T , Ω , Ω ^ ) (T,\Omega,\hat{\Omega}) ( T , Ω , Ω ^ ) and ( μ , a , b ) (\mu,a,b) ( μ , a , b ) fixed. Since D + κ D+\kappa D + κ is then constant, the embedding equation is simply
d d R ( N R κ ) = 0. (258) \frac{\mathrm{d}}{\mathrm{d}R}(N R\kappa)=0. \tag*{(258)} d R d ( N R κ ) = 0. (258)
We see that radial balance is independent of the local Kerr spin at leading order. To solve it explicitly, define
K 00 = N ′ N , K 11 = − R ′ R , K 01 = R ω ′ 2 N , ′ = d d R . (259) K_{00}=\frac{N'}{N},\qquad K_{11}=-\frac{\mathcal{R}'}{\mathcal{R}},\qquad K_{01}=\frac{\mathcal{R}\omega'}{2N},\qquad{}'=\frac{\mathrm{d}}{\mathrm{d}R}. \tag*{(259)} K 00 = N N ′ , K 11 = − R R ′ , K 01 = 2 N R ω ′ , ′ = d R d . (259)
Using k = N 1 − v 2 k=N\sqrt{1-v^{2}} k = N 1 − v 2 in (C.9) gives
A v v 2 + 2 K 01 v + C v = 0 , A v = 2 K 11 − K 00 , C v = 2 K 00 − K 11 . (260) A_v v^2+2K_{01}v+C_v=0,\qquad A_v=2K_{11}-K_{00},\qquad C_v=2K_{00}-K_{11}. \tag*{(260)} A v v 2 + 2 K 01 v + C v = 0 , A v = 2 K 11 − K 00 , C v = 2 K 00 − K 11 . (260)
Thus the mechanically balanced branches are
v ± = − K 01 ± K 01 2 − A v C v A v , Ω ± = ω + N v ± R . (261) v_{\pm}=\frac{-K_{01}\pm\sqrt{K_{01}^{2}-A_vC_v}}{A_v},\qquad\Omega_{\pm}=\omega+\frac{N v_{\pm}}{\mathcal{R}}. \tag*{(261)} v ± = A v − K 01 ± K 01 2 − A v C v , Ω ± = ω + R N v ± . (261)
We keep only real roots with ∣ v ± ∣ < 1 |v_{\pm}|<1 ∣ v ± ∣ < 1 . If A v = 0 A_v=0 A v = 0 , we use the linear equation (C.11) instead. In flat space these reduce to v = ± 1 / 2 v=\pm1/\sqrt{2} v = ± 1/ 2 and Ω = ± 1 / ( 2 R ) \Omega=\pm1/(\sqrt{2}R) Ω = ± 1/ ( 2 R ) . Substitution of an extremal MP background gives the circular-ring balance equation used in section 3.2 of [14 ] . Choosing an extremal central background and κ > 0 \kappa>0 κ > 0 constructs an extremal hole with a nonextremal probe ring. Conversely, κ = 0 \kappa=0 κ = 0 in a nonextremal background constructs an extremal ring around a nonextremal hole.
Thermodynamic charges and the mass convention
On a stationary embedding, the free-energy derivatives do not contain a contribution from the change of R R R , because ∂ R F = 0 \partial_R\mathcal{F}=0 ∂ R F = 0 . Therefore
J R Ψ = − ∂ F ∂ Ω , J R Φ = − ∂ F ∂ Ω ^ , S R = − ∂ F ∂ T . (262) J_{R\Psi}=-\frac{\partial\mathcal{F}}{\partial\Omega},\qquad J_{R\Phi}=-\frac{\partial\mathcal{F}}{\partial\hat{\Omega}},\qquad S_R=-\frac{\partial\mathcal{F}}{\partial T}. \tag*{(262)} J R Ψ = − ∂ Ω ∂ F , J R Φ = − ∂ Ω ^ ∂ F , S R = − ∂ T ∂ F . (262)
with the background held fixed. Equations (C.7) and (C.8) give
J R Ψ = π m R 2 Γ 2 v G 5 , J R Φ = 2 π R Γ m 2 α G 5 . (263) J_{R\Psi}=\frac{\pi m R^{2}\Gamma^{2}v}{G_{5}},\qquad J_{R\Phi}=\frac{2\pi R\Gamma m^{2}\alpha}{G_{5}}. \tag*{(263)} J R Ψ = G 5 π m R 2 Γ 2 v , J R Φ = G 5 2 π R Γ m 2 α . (263)
S R = 2 π F D = 4 π 2 R Γ m r s G 5 , A R = 16 π 2 R Γ m r s . (264) S_R=\frac{2\pi\mathcal{F}}{D}=\frac{4\pi^{2}R\Gamma m r_s}{G_{5}},\qquad A_R=16\pi^{2}R\Gamma m r_s. \tag*{(264)} S R = D 2 π F = G 5 4 π 2 R Γ m r s , A R = 16 π 2 R Γ m r s . (264)
The intensive quantities can also be written in local-seed form:
κ R = κ σ 2 m r s , Ω R Φ = κ α 2 r s , Ω R Ψ = ϖ + N v R , T R = κ R 2 π . (265) \kappa_R = \frac{\kappa\sigma}{2mr_s},\qquad\Omega_R^\Phi= \frac{\kappa\alpha}{2r_s},\qquad\Omega_R^\Psi= \varpi+ \frac{Nv}{\mathcal{R}},\qquad T_R = \frac{\kappa_R}{2\pi}. \tag*{(265)} κ R = 2 m r s κσ , Ω R Φ = 2 r s κ α , Ω R Ψ = ϖ + R N v , T R = 2 π κ R . (265)
In particular T R S R + Ω R Φ J R Φ = F T_RS_R+\Omega_R^\Phi J_{R\Phi}=\mathcal{F} T R S R + Ω R Φ J R Φ = F . The energy conjugate to the background time at infinity is
E R = F + T R S R + Ω R Ψ J R Ψ + Ω R Φ J R Φ = 2 F + Ω R Ψ J R Ψ . (266) E_R=\mathcal{F}+T_RS_R+\Omega_R^\Psi J_{R\Psi}+\Omega_R^\Phi J_{R\Phi}=2\mathcal{F}+\Omega_R^\Psi J_{R\Psi}. \tag*{(266)} E R = F + T R S R + Ω R Ψ J R Ψ + Ω R Φ J R Φ = 2 F + Ω R Ψ J R Ψ . (266)
It is important to distinguish this probe energy from a horizon Komar mass in a background containing another gravitating object. Using the binding-tension convention of [14 , 17 ] , we have
T ^ = 2 E R − 3 ( T R S R + Ω R Ψ J R Ψ + Ω R Φ J R Φ ) = F − Ω R Ψ J R Ψ . (267) \widehat{T}=2E_R-3\left(T_RS_R+\Omega_R^\Psi J_{R\Psi}+\Omega_R^\Phi J_{R\Phi}\right)=\mathcal{F}-\Omega_R^\Psi J_{R\Psi}. \tag*{(267)} T = 2 E R − 3 ( T R S R + Ω R Ψ J R Ψ + Ω R Φ J R Φ ) = F − Ω R Ψ J R Ψ . (267)
Their intrinsic mass prescription M R , K = E R − T ^ / 2 M_{R,K}=E_R-\widehat{T}/2 M R , K = E R − T /2 then gives
M R , K = 3 2 ( F + Ω R Ψ J R Ψ ) , E R − M R , K = T ^ 2 , ℓ R = 2 N R σ . (268) M_{R,K}=\frac{3}{2}\left(\mathcal{F}+\Omega_R^\Psi J_{R\Psi}\right),\qquad E_R-M_{R,K}=\frac{\widehat{T}}{2},\qquad\ell_R=2NR\sigma. \tag*{(268)} M R , K = 2 3 ( F + Ω R Ψ J R Ψ ) , E R − M R , K = 2 T , ℓ R = 2 N R σ . (268)
Here ℓ R \ell_R ℓ R is the leading canonical Weyl rod length. The local Smarr relation follows immediately. The mass distinction and its relation to background response are discussed in [14 , 17 ] .
The entropy must be differentiated at finite temperature before taking κ → 0 \kappa\to0 κ → 0 . For Ω ^ ≠ 0 \widehat{\Omega}\ne0 Ω = 0 this limit is finite:
α = sgn Ω ^ , m = r s = κ 2 ∣ Ω ^ ∣ , A R = 8 π G 5 ∣ J R Φ ∣ . (269) \alpha=\operatorname{sgn}\widehat{\Omega},\qquad m=r_s=\frac{\kappa}{2|\widehat{\Omega}|},\qquad A_R=8\pi G_5|J_{R\Phi}|. \tag*{(269)} α = sgn Ω , m = r s = 2∣ Ω ∣ κ , A R = 8 π G 5 ∣ J R Φ ∣. (269)
Setting T = 0 T=0 T = 0 in the action before differentiating would lose this entropy.
Solutions in mutual thermal equilibrium
We now look for solutions of the blackfold construction that satisfy thermodynamic equilibrium. Write h = r + , H 2 h=r_{+,H}^2 h = r + , H 2 for the squared MP horizon radius. Its parameters and potentials satisfy
μ = ( h + a 2 ) ( h + b 2 ) h , h > ∣ a b ∣ , \mu=\frac{(h+a^2)(h+b^2)}{h},\qquad h>|ab|, μ = h ( h + a 2 ) ( h + b 2 ) , h > ∣ ab ∣ ,
κ H = h 2 − a 2 b 2 h ( h + a 2 ) ( h + b 2 ) , Ω H Ψ = a h + a 2 , Ω H Φ = b h + b 2 . (270) \kappa_H=\frac{h^2-a^2b^2}{\sqrt{h}(h+a^2)(h+b^2)},\qquad\Omega_H^\Psi=\frac{a}{h+a^2},\qquad\Omega_H^\Phi=\frac{b}{h+b^2}. \tag*{(270)} κ H = h ( h + a 2 ) ( h + b 2 ) h 2 − a 2 b 2 , Ω H Ψ = h + a 2 a , Ω H Φ = h + b 2 b . (270)
The strict inequality selects a nonextremal outer horizon. For mutual equilibrium set ( κ , Ω , Ω ^ ) = ( κ H , Ω H Ψ , Ω H Φ ) (\kappa,\Omega,\widehat{\Omega})=(\kappa_H,\Omega_H^\Psi,\Omega_H^\Phi) ( κ , Ω , Ω ) = ( κ H , Ω H Ψ , Ω H Φ ) . The transverse seed is then fixed:
D H = κ H 2 + ( Ω H Φ ) 2 , α = Ω H Φ D H , m = κ 2 ( D H + κ H ) . (271) D_H=\sqrt{\kappa_H^2+(\Omega_H^\Phi)^2},\qquad\alpha=\frac{\Omega_H^\Phi}{D_H},\qquad m=\frac{\kappa}{2(D_H+\kappa_H)}. \tag*{(271)} D H = κ H 2 + ( Ω H Φ ) 2 , α = D H Ω H Φ , m = 2 ( D H + κ H ) κ . (271)
The remaining radial equation has a particularly simple analytic form. Let x = R 2 x=R^2 x = R 2 , assume a ≠ 0 a\ne0 a = 0 , and define
x − = a 2 b 2 h , x L = h ( h + a 2 ) a 2 − b 2 . (272) x_-=\frac{a^2b^2}{h},\qquad x_L=\frac{h(h+a^2)}{a^2}-b^2. \tag*{(272)} x − = h a 2 b 2 , x L = a 2 h ( h + a 2 ) − b 2 . (272)
Then x − < h < x L x_{-}<h<x_{L} x − < h < x L , and direct substitution in (C.4) gives
κ H 2 = a 2 ( x − h ) ( x L − x ) ( h + a 2 ) 2 ( x + b 2 ) , ( N R κ H ) 2 = a 2 ( x − h ) 2 ( x − x − ) ( x L − x ) ( h + a 2 ) 2 ( x + b 2 ) 2 . (273) \kappa_{H}^{2}=\frac{a^{2}(x-h)(x_{L}-x)}{(h+a^{2})^{2}(x+b^{2})},\qquad(N R\kappa_{H})^{2}=\frac{a^{2}(x-h)^{2}(x-x_{-})(x_{L}-x)}{(h+a^{2})^{2}(x+b^{2})^{2}}. \tag*{(273)} κ H 2 = ( h + a 2 ) 2 ( x + b 2 ) a 2 ( x − h ) ( x L − x ) , ( N R κ H ) 2 = ( h + a 2 ) 2 ( x + b 2 ) 2 a 2 ( x − h ) 2 ( x − x − ) ( x L − x ) . (273)
Thus the orbit must lie between the horizon and the outer light surface, h < x < x L h<x<x_{L} h < x < x L . Taking the logarithmic derivative of the second expression reduces radial balance to
2 x − h + 1 x − x − − 1 x L − x − 2 x + b 2 = 0. (274) \frac{2}{x-h}+\frac{1}{x-x_{-}}-\frac{1}{x_{L}-x}-\frac{2}{x+b^{2}}=0. \tag*{(274)} x − h 2 + x − x − 1 − x L − x 1 − x + b 2 2 = 0. (274)
Equivalently, its physical root solves the cubic
2 ( h + b 2 ) ( x − x − ) ( x L − x ) + ( x L + x − − 2 x ) ( x − h ) ( x + b 2 ) = 0 , h < x < x L . (275) 2(h+b^{2})(x-x_{-})(x_{L}-x)+(x_{L}+x_{-}-2x)(x-h)(x+b^{2})=0,\qquad h<x<x_{L}. \tag*{(275)} 2 ( h + b 2 ) ( x − x − ) ( x L − x ) + ( x L + x − − 2 x ) ( x − h ) ( x + b 2 ) = 0 , h < x < x L . (275)
This gives an analytic specification of the equilibrium radius without a simultaneous fit of the three potentials. The left side of (C.25) decreases strictly from + ∞ +\infty + ∞ to − ∞ -\infty − ∞ : the first and last terms combine to 2 ( h + b 2 ) / [ ( x − h ) ( x + b 2 ) ] 2(h+b^{2})/[(x-h)(x+b^{2})] 2 ( h + b 2 ) / [( x − h ) ( x + b 2 )] , which decreases, and each remaining term also decreases. Hence there is exactly one physical radius for each nonextremal background with a ≠ 0 a\ne0 a = 0 at this order. Equations (C.22, (C.14) and (C.15) then determine its thermodynamics. The existence of this formal root does not ensure that the resulting ring is thin enough for the approximation as we shall see below.
This calculation also exposes the obstruction to finding doubly extremal black Saturns in thermodynamic equilibrium. For a finite-area extremal centre, h = ∣ a b ∣ > 0 h=|ab|>0 h = ∣ ab ∣ > 0 and x L = h x_{L}=h x L = h , so
κ H 2 = − a 2 ( x − h ) 2 ( h + a 2 ) 2 ( x + b 2 ) < 0 ( x > h ) . (276) \kappa_{H}^{2}=-\frac{a^{2}(x-h)^{2}}{(h+a^{2})^{2}(x+b^{2})}<0\qquad(x>h). \tag*{(276)} κ H 2 = − ( h + a 2 ) 2 ( x + b 2 ) a 2 ( x − h ) 2 < 0 ( x > h ) . (276)
A circular ring with the central hole’s orbital angular velocity is then spacelike outside the horizon. No separated, timelike ring in this leading MP-background construction can be in mutual equilibrium with that extremal centre. This agrees with the obstruction in section 3.5 of [14 ] . The same obstruction for the exact Saturn metric obtained in this paper is proved separately in appendix D.
Matching to exact Saturn charges
For comparison with an exact solution, we choose an isolated MP having the central horizon’s area and its two Komar angular momenta. With
X = A H 2 π 2 , p = 4 G 5 J H Ψ π , q = 4 G 5 J H Φ π , (277) X=\frac{\mathcal{A}_{H}}{2\pi^{2}},\qquad p=\frac{4G_{5}J_{H\Psi}}{\pi},\qquad q=\frac{4G_{5}J_{H\Phi}}{\pi}, \tag*{(277)} X = 2 π 2 A H , p = π 4 G 5 J H Ψ , q = π 4 G 5 J H Φ , (277)
its parameters are fixed algebraically:
μ 3 = ( X 2 + p 2 ) ( X 2 + q 2 ) X 2 , ( a , b , r + , H ) = ( p , q , X ) μ , M M P = 3 π μ 8 G 5 . (278) \mu^{3}=\frac{(X^{2}+p^{2})(X^{2}+q^{2})}{X^{2}},\qquad(a,b,r_{+,H})=\frac{(p,q,X)}{\mu},\qquad M_{\mathrm{MP}}=\frac{3\pi\mu}{8G_{5}}. \tag*{(278)} μ 3 = X 2 ( X 2 + p 2 ) ( X 2 + q 2 ) , ( a , b , r + , H ) = μ ( p , q , X ) , M MP = 8 G 5 3 π μ . (278)
We use M M P M_{\mathrm{MP}} M MP for the mass of this isolated reference hole. The quantity M H , K M_{H,K} M H , K instead comes from the Komar integral around the central horizon of the full Saturn, where the ring’s gravitational field is present. These masses need not agree, even though we have chosen the same central area and angular momenta.
We denote the blackfold ring’s probe energy by E R b f E^{\mathrm{bf}}_R E R bf and its horizon Komar mass by M R , K b f M^{\mathrm{bf}}_{R,K} M R , K bf . They are the quantities E R E_R E R and M R , K M_{R,K} M R , K defined above, with the superscript identifying the approximation. Holding the central entropy and spins fixed, we compare the masses through
M A D M − M M P ≃ E R b f , M H , K − M M P ≃ E R b f − M R , K b f . (279) \begin{aligned}
M_{\mathrm{ADM}}-M_{\mathrm{MP}} &\simeq E^{\mathrm{bf}}_R,\\
M_{\mathrm{H,K}}-M_{\mathrm{MP}} &\simeq E^{\mathrm{bf}}_R-M^{\mathrm{bf}}_{R,K}. \tag*{(279)}
\end{aligned} M ADM − M MP M H , K − M MP ≃ E R bf , ≃ E R bf − M R , K bf . (279)
The second relation follows from the first and the exact ADM–Komar sum (48) . Equivalently, the sum of the two physical horizon masses includes the central response relative to the isolated reference. Replacing that responding central mass by M M P M_{\mathrm{MP}} M MP while retaining only M R , K b f M^{\mathrm{bf}}_{R,K} M R , K bf would omit this contribution.
The canonical rod separation fixes the leading orbital radius through
d z = a 4 + a 5 − a 2 − a 3 2 , R 2 = 2 d z + μ − a 2 − b 2 2 . (280) d_z=\frac{a_4+a_5-a_2-a_3}{2},\qquad R^2=2d_z+\frac{\mu-a^2-b^2}{2}. \tag*{(280)} d z = 2 a 4 + a 5 − a 2 − a 3 , R 2 = 2 d z + 2 μ − a 2 − b 2 . (280)
The exact ring’s κ R , Ω R Φ \kappa_R,\Omega_R^\Phi κ R , Ω R Φ then determine its local m , α m,\alpha m , α by (C.7). Its orbital potential, Komar mass, spins and area provide independent comparisons, as does the total mass increment. Along a thin nonextremal sequence, decreasing the ring rod length from 10 − 4 10^{-4} 1 0 − 4 to 10 − 16 10^{-16} 1 0 − 16 reduces the relative orbital-spin error from 1.93 × 10 − 5 1.93 \times10^{-5} 1.93 × 1 0 − 5 to 1.93 × 10 − 17 1.93 \times10^{-17} 1.93 × 1 0 − 17 . For the extremal-hole example (A.94), the blackfold surface gravity differs from the exact ring value by 4.7 × 10 − 4 4.7 \times10^{-4} 4.7 × 1 0 − 4 in relative terms. Its thickness-to-radius and thickness-to-separation ratios are 1.04 × 10 − 4 1.04 \times10^{-4} 1.04 × 1 0 − 4 and 1.42 × 10 − 4 1.42 \times10^{-4} 1.42 × 1 0 − 4 , respectively. We discuss the doubly extremal comparison in section 5.
The thermal curves use the same reference ensemble: M b f = M M P + E R M^{\mathrm{bf}}=M_{\mathrm{MP}}+E_R M bf = M MP + E R , J A b f = J A M P + J R A J_A^{\mathrm{bf}}=J_A^{\mathrm{MP}}+J_{R A} J A bf = J A MP + J R A and S b f = S M P + S R S^{\mathrm{bf}}=S_{\mathrm{MP}}+S_R S bf = S MP + S R . These leading expressions do not include a higher-order calculation of the central deformation or ring self-field. The reference MP quantities are J A M P = π μ ( a , b ) / ( 4 G 5 ) J_A^{\mathrm{MP}}=\pi\mu(a,b)/(4G_5) J A MP = π μ ( a , b ) / ( 4 G 5 ) and S M P = π 2 μ h / ( 2 G 5 ) S_{\mathrm{MP}}=\pi^2\mu\sqrt{h}/(2G_5) S MP = π 2 μ h / ( 2 G 5 ) .
Validity and the plotted comparisons
Here we discuss the validity of the blackfold curves presented in the phase diagrams. The local string thickness must be small compared with the bending radius, background-curvature radii and separation from the central horizon. Transverse spin-connection corrections must also be small compared with the local rotation scale [14 ] . We monitor 2 m / R 2m/R 2 m / R and 2 m / ( R − r + , H ) 2m/(R-r_{+,H}) 2 m / ( R − r + , H ) , where the latter uses a coordinate separation proxy. A small ring-to-centre mass ratio is not required by itself: a distant ring can be heavier than the centre while producing a weak potential there. For thermal matching one generically has m ∼ L H m\sim L_H m ∼ L H , E R / M M P ∼ R / L H E_R/M_{\mathrm{MP}}\sim R/L_H E R / M MP ∼ R / L H , but G 5 E R / R 2 ∼ m / R ≪ 1 G_5E_R/R^2\sim m/R\ll1 G 5 E R / R 2 ∼ m / R ≪ 1 , where L H L_H L H is a central length scale. The local scale ratios and omitted response terms must be assessed together. The flat-background limit supplies a useful analytic check. For positive spins the leading isolated-ring relations are
α = 8 j Ψ j Φ , a r i n g , b f = 2 4 j Ψ ( 1 + 1 − 64 j Ψ 2 j Φ 2 ) , 0 ≤ 8 j Ψ j Φ ≤ 1. (281) \alpha=8j_\Psi j_\Phi,\qquad a_{\mathrm{ring,bf}}=\frac{\sqrt{2}}{4j_\Psi}\left(1+\sqrt{1-64j_\Psi^2j_\Phi^2}\right),\qquad0\leq8j_\Psi j_\Phi\leq1. \tag*{(281)} α = 8 j Ψ j Φ , a ring , bf = 4 j Ψ 2 ( 1 + 1 − 64 j Ψ 2 j Φ 2 ) , 0 ≤ 8 j Ψ j Φ ≤ 1. (281)
Since m / R = 1 / ( 16 j ˙ Ψ 2 ) m/R = 1/(16\dot{j}_{\Psi}^{2}) m / R = 1/ ( 16 j ˙ Ψ 2 ) , a fixed j ˙ Φ > 0 \dot{j}_{\Phi} > 0 j ˙ Φ > 0 implies
j ˙ Ψ ≤ 1 8 j ˙ Φ , 2 m R = 1 8 j ˙ Ψ 2 ≥ 8 j ˙ Φ 2 . (282) \dot{j}_{\Psi} \le\frac{1}{8\dot{j}_{\Phi}}, \qquad\frac{2m}{R} = \frac{1}{8\dot{j}_{\Psi}^{2}} \ge8\dot{j}_{\Phi}^{2}. \tag*{(282)} j ˙ Ψ ≤ 8 j ˙ Φ 1 , R 2 m = 8 j ˙ Ψ 2 1 ≥ 8 j ˙ Φ 2 . (282)
A nonzero fixed transverse spin therefore prevents this isolated ring from becoming arbitrarily thin. Its leading extremal endpoint is j ˙ Ψ = 1 / ( 8 j ˙ Φ ) \dot{j}_{\Psi} = 1/(8\dot{j}_{\Phi}) j ˙ Ψ = 1/ ( 8 j ˙ Φ ) , whereas exact PS gives j ˙ Ψ = 1 / ( 8 j ˙ Φ ) + j ˙ Φ \dot{j}_{\Psi} = 1/(8\dot{j}_{\Phi}) + \dot{j}_{\Phi} j ˙ Ψ = 1/ ( 8 j ˙ Φ ) + j ˙ Φ , with limiting area 2 2 j ˙ Φ 2\sqrt{2}\dot{j}_{\Phi} 2 2 j ˙ Φ . At j ˙ Φ = 0.10 \dot{j}_{\Phi} = 0.10 j ˙ Φ = 0.10 the endpoints are 1.25 and 1.35, respectively. This illustrates sensitivity to corrections near transverse extremality. The isolated-ring thickness bound does not apply directly to Saturn, whose reduced spins include the central contribution.
Accuracy of the thermal blackfold comparison. At j ˙ Ψ = 1.2 \dot{j}_{\Psi} = 1.2 j ˙ Ψ = 1.2 , linear interpolation along each plotted continuation at fixed j ˙ Φ \dot{j}_{\Phi} j ˙ Φ gives the following rounded values. The shortfall is 100 ( a m e t r i c − a b f ) / a m e t r i c 100(a_{\rm metric} - a_{\rm bf})/a_{\rm metric} 100 ( a metric − a bf ) / a metric . The last two columns refer to the Saturn blackfold at the same two total reduced spins, with α \alpha α as in (C.22).
Table C.1.
These values diagnose the plotted approximation. On the sampled j ˙ Φ = 0.10 \dot{j}_{\Phi} = 0.10 j ˙ Φ = 0.10 , j ˙ Ψ ≥ 1 \dot{j}_{\Psi} \ge1 j ˙ Ψ ≥ 1 blackfold segment, the smallest 2 m / R 2m/R 2 m / R is approximately 0.0758. At j ˙ Ψ = 1.2 \dot{j}_{\Psi} = 1.2 j ˙ Ψ = 1.2 , its separation proxy is 2 m / ( R − r + , H ) ≃ 0.0794 2m/(R-r_{+,{\rm H}}) \simeq0.0794 2 m / ( R − r + , H ) ≃ 0.0794 , while α ≃ 0.965 \alpha\simeq0.965 α ≃ 0.965 is close to the transverse Kerr bound. The corresponding minimum thickness ratios on the sampled j ˙ Φ = 0.01 \dot{j}_{\Phi} = 0.01 j ˙ Φ = 0.01 and 0.05 segments are approximately 0.000801 and 0.0198.
The j ˙ Φ = 0.10 \dot{j}_{\Phi} = 0.10 j ˙ Φ = 0.10 Saturn slice is only moderately thin and is close to the transverse Kerr bound. Its roughly 21% area mismatch at j ˙ Ψ = 1.2 \dot{j}_{\Psi} = 1.2 j ˙ Ψ = 1.2 is consistent with omitted finite-thickness and interaction effects, but the leading calculation does not predict the magnitude or sign of the correction. In five dimensions, self-gravity enters at the same order as spin-orbit coupling so adding the latter alone is not a complete next-order result. The exact solution therefore offers a way to test the spin–orbit contribution derived in [14 section 3.2.5] , provided that self-gravity effects at the same order are accounted for. The thick-regime blackfold cusps are therefore extrapolations.
Obstruction to doubly extremal mutual equilibrium
Zero temperature at both horizons does not imply mutual equilibrium: the two angular velocities must also agree in a common physical frame. We first prove that this is impossible for a regular member of the finite double-extremal family in § 4 . The proof is uniform in its remaining parameters and sign sectors, before imposing balance. We then describe exact boundary exclusions and the numerical searches with independently retuned BZ data. The latter strengthen the evidence that such thermodynamic equilibrium is not possible to achieve for doubly spinning extremal black Saturns.
Conventions and the two exact matching equations
We use the rational version of the coefficient construction in section 4, with
λ = τ 2 / 2 , h = H 2 / 4 , C 3 = v τ , q b o o s t = τ − 1 . (283) \lambda= \tau^2/2,\qquad h = H^2/4,\qquad C_3 = v\tau,\qquad q_{\mathrm{boost}} = \tau^{-1}. \tag*{(283)} λ = τ 2 /2 , h = H 2 /4 , C 3 = v τ , q boost = τ − 1 . (283)
Here H = H ∗ H = H_\ast H = H ∗ and v = v ∗ v = v_\ast v = v ∗ in the notation of section 4, in particular v v v multiplies τ \tau τ , rather than λ \sqrt{\lambda} λ . The constants ( Z , c , d , H , v , w , r 0 , β ) (Z,c,d,H,v,w,r_0,\beta) ( Z , c , d , H , v , w , r 0 , β ) remain free before balance, and the horizon points are z R = 0 z_{\mathrm{R}} = 0 z R = 0 and z H = Z z_{\mathrm{H}} = Z z H = Z . This overparameterized description contains the finite doubly extremal boundary of the six-parameter balanced family. We do not fix Z Z Z , r 0 r_0 r 0 , β \beta β , or the value of w w w used in earlier examples. Throughout this calculation
Z > 0 , r 0 > 0 , c H v β ≠ 0 , ∣ c ∣ , ∣ H ∣ , ∣ v ∣ , ∣ β ∣ < ∞ . (284) Z > 0,\qquad r_0 > 0,\qquad cHv\beta\ne0,\qquad|c|,|H|,|v|,|\beta| < \infty. \tag*{(284)} Z > 0 , r 0 > 0 , cH v β = 0 , ∣ c ∣ , ∣ H ∣ , ∣ v ∣ , ∣ β ∣ < ∞. (284)
All signs of c c c , H H H , v v v , β \beta β are retained. Write the intermediate coefficient metric as
d s 2 = G a b d x a d x b + f ( d ρ 2 + d z 2 ) , x a = ( t , ψ , ϕ ) , det G = − ρ 2 . (285) \mathrm{d}s^2 = G_{ab}\mathrm{d}x^a\mathrm{d}x^b + f(\mathrm{d}\rho^2+\mathrm{d}z^2),\qquad x^a=(t,\psi,\phi),\qquad\det G=-\rho^2. \tag*{(285)} d s 2 = G ab d x a d x b + f ( d ρ 2 + d z 2 ) , x a = ( t , ψ , ϕ ) , det G = − ρ 2 . (285)
The symbols in this section refer to this raw coefficient metric. The physical asymptotic frame is obtained by one constant, nonsingular change of Killing basis and a finite positive normalization of f f f . Equality of the two horizon generator rays is invariant under that common basis change. Therefore, when the raw time components are nonzero, equality of both physical angular velocities is equivalent to
ω R ψ = ω H ψ , ω R ϕ = ω H ϕ . (286) \omega_{\mathrm{R}\psi}=\omega_{\mathrm{H}\psi},\qquad\omega_{\mathrm{R}\phi}=\omega_{\mathrm{H}\phi}. \tag*{(286)} ω R ψ = ω H ψ , ω R ϕ = ω H ϕ . (286)
Individual raw angular velocities need not equal the physically normalized ones. Exact extraction from the horizon kernels gives
ω R ψ = 1 β − c / Z + r 0 2 4 β , ω R ϕ = 1 β + c / Z + r 0 2 4 β , (287) \omega_{\mathrm{R}\psi}=\frac{1}{\beta}-\frac{c/Z+r_0^2}{4\beta},\qquad
\omega_{\mathrm{R}\phi}=\frac{1}{\beta}+\frac{c/Z+r_0^2}{4\beta}, \tag*{(287)} ω R ψ = β 1 − 4 β c / Z + r 0 2 , ω R ϕ = β 1 + 4 β c / Z + r 0 2 , (287)
ω H ψ = − H c v Z D H + N H 8 H D H , ω H ϕ = − H c v Z D H − N H 8 H D H , (288) \omega_{\mathrm{H}\psi}=-\frac{Hcv}{ZD_{\mathrm{H}}}+\frac{N_{\mathrm{H}}}{8HD_{\mathrm{H}}},\qquad
\omega_{\mathrm{H}\phi}=-\frac{Hcv}{ZD_{\mathrm{H}}}-\frac{N_{\mathrm{H}}}{8HD_{\mathrm{H}}}, \tag*{(288)} ω H ψ = − Z D H H c v + 8 H D H N H , ω H ϕ = − Z D H H c v − 8 H D H N H , (288)
where
D H = 2 c v − H d , (289) D_{\mathrm{H}}=2cv-Hd, \tag*{(289)} D H = 2 c v − H d , (289)
N H = 2 ( Z + 1 ) Z H 2 c 2 v − H 2 d v Z + 2 w H 2 v + 4 H c v 2 Z + 4 H d − 4 c v . (290) N_{\mathrm{H}}=\frac{2(Z+1)}{Z}H^2c^2v-\frac{H^2dv}{Z}+2wH^2v+\frac{4Hcv^2}{Z}+4Hd-4cv. \tag*{(290)} N H = Z 2 ( Z + 1 ) H 2 c 2 v − Z H 2 d v + 2 w H 2 v + Z 4 H c v 2 + 4 H d − 4 c v . (290)
The sum of equations (D.4) fixes d d d , and their difference fixes w w w :
d = c v ( 2 H + β Z ) , (291) d=cv\left(\frac{2}{H}+\frac{\beta}{Z}\right), \tag*{(291)} d = c v ( H 2 + Z β ) , (291)
w = c 2 Z 2 + c Z r 0 2 − ( Z + 1 ) c 2 Z − c v H Z + β c v 2 Z 2 − 2 c H 2 − 2 β c H Z . (292) w=\frac{c^2}{Z^2}+\frac{c}{Zr_0^2}-\frac{(Z+1)c^2}{Z}-\frac{cv}{HZ}+\frac{\beta cv}{2Z^2}-\frac{2c}{H^2}-\frac{2\beta c}{HZ}. \tag*{(292)} w = Z 2 c 2 + Z r 0 2 c − Z ( Z + 1 ) c 2 − H Z c v + 2 Z 2 β c v − H 2 2 c − H Z 2 β c . (292)
Substitution makes both angular-velocity differences vanish identically. In particular,
D H = − β H c v Z ≠ 0 (293) D_{\mathrm{H}}=-\frac{\beta Hcv}{Z}\ne0 \tag*{(293)} D H = − Z β H c v = 0 (293)
on the matching locus. The locus D H = 0 D_{\mathrm{H}}=0 D H = 0 is not an omitted matching branch: for finite nonzero H c v H_{\mathrm{cv}} H cv the unnormalized hole generator has zero time component and a nonzero angular sum, whereas the ring generator in (D.5) has nonzero time component. Their rays cannot agree. This also explains why clearing a vanishing normalization denominator must not create spurious solutions.
A uniform obstruction on the middle axis
We can establish the obstruction by examining the exterior axis at z = Z / 2 z=Z/2 z = Z /2 . A convenient exact rational approach is
z = Z / 2 , ρ 2 = Z 2 s ( 1 − s ) 2 , s ↓ 0. (294) z = Z/2,\qquad\rho^2 = \frac{Z^2s}{(1-s)^2},\qquad s\downarrow0. \tag*{(294)} z = Z /2 , ρ 2 = ( 1 − s ) 2 Z 2 s , s ↓ 0. (294)
Insert (D.9)–(D.10) directly into the full Laurent-coefficient metric, before expanding in s s s . Exact rational arithmetic gives
f ( s , Z / 2 ) = s P ( s ∣ c , H , v , Z , r 0 , β ) 8 H 4 Z 6 r 0 4 β 4 ( 1 + s ) 7 , (295) f(s,Z/2)=\frac{sP(s\mid c,H,v,Z,r_0,\beta)}{8H^4Z^6r_0^4\beta^4(1+s)^7}, \tag*{(295)} f ( s , Z /2 ) = 8 H 4 Z 6 r 0 4 β 4 ( 1 + s ) 7 s P ( s ∣ c , H , v , Z , r 0 , β ) , (295)
where P P P is a polynomial. The degree is at most six. Its coefficients are obtained by the finite scalar construction in subsection B.2. We supply the exact coefficient calculation with the ancillary material. Crucially, the denominator at s = 0 s=0 s = 0 is nonzero throughout (D.2). Hence
lim ρ → 0 f ( ρ , Z / 2 ) = 0. (296) \lim_{\rho\to0}f(\rho,Z/2)=0. \tag*{(296)} ρ → 0 lim f ( ρ , Z /2 ) = 0. (296)
No sign assumption, numerical minimization, balance equation, or division by a potentially vanishing leading numerator is used here. For a shorter description of the generic behavior, define
k = 2 Z H , C = 2 c + 2 Z r 0 2 , (297) k=\frac{2Z}{H},\qquad C=2c+\frac{2Z}{r_0^2}, \tag*{(297)} k = H 2 Z , C = 2 c + r 0 2 2 Z , (297)
E = C + 1 − v ( β + k ) − 2 β k − k 2 , (298) \mathcal{E}=C+1-v(\beta+k)-2\beta k-k^2, \tag*{(298)} E = C + 1 − v ( β + k ) − 2 β k − k 2 , (298)
Δ e q = ( β v + 1 ) 2 − 4 β 4 − 2 β 2 k v − 4 β 3 k + 2 β k − 3 β 2 k 2 . (299) \Delta_{\mathrm{eq}}=(\beta v+1)^2-4\beta^4-2\beta^2kv-4\beta^3k+2\beta k-3\beta^2k^2. \tag*{(299)} Δ eq = ( β v + 1 ) 2 − 4 β 4 − 2 β 2 k v − 4 β 3 k + 2 β k − 3 β 2 k 2 . (299)
This Δ e q \Delta_{\mathrm{eq}} Δ eq is unrelated to the earlier Killing-frame determinant. The first coefficient of (D.13) is
f ( s , Z / 2 ) = c 2 Δ e q E 2 2 Z 4 k 2 β 4 s + O ( s 2 ) . (300) f(s,Z/2)=\frac{c^2\Delta_{\mathrm{eq}}\mathcal{E}^2}{2Z^4k^2\beta^4}s+O(s^2). \tag*{(300)} f ( s , Z /2 ) = 2 Z 4 k 2 β 4 c 2 Δ eq E 2 s + O ( s 2 ) . (300)
On E = 0 \mathcal{E}=0 E = 0 or Δ e q = 0 \Delta_{\mathrm{eq}}=0 Δ eq = 0 this leading coefficient vanishes. Equation (D.13), whose full denominator contains neither factor, still gives (D.14). Thus these exceptional loci cannot evade the obstruction.
Proposition D.1 (No regular mutual equilibrium in the finite chart). Within (D.2), the direct correlated doubly extremal metric cannot simultaneously have two distinct regular horizons, a regular spacelike axis between them, a finite nonsingular common asymptotic Killing frame, and equality of both horizon angular velocities. The conclusion includes every real sign sector of the finite chart and therefore every balanced subfamily contained in it.
Proof. At an interior point of a smooth spacelike axis, let q q q be proper radial distance. The vanishing rotational Killing field has squared norm proportional to q 2 q^2 q 2 , with a finite positive proportionality constant. The complementary two-dimensional Killing block has a finite nonzero Lorentzian determinant. The identity det G = − ρ 2 \det G=-\rho^2 det G = − ρ 2 then implies ρ = a q + O ( q 3 ) \rho=aq+O(q^3) ρ = a q + O ( q 3 ) with a ≠ 0 a\ne0 a = 0 . Consequently the radial base coefficient has a finite strictly positive axis limit, f → a − 2 > 0 f\to a^{-2}>0 f → a − 2 > 0 . The same necessary property holds with a finite conical deficit. A divergent or degenerate complementary block would itself violate the assumed smooth spacelike axis. Equations (D.4) force (D.9)–(D.10), and the exact identity (D.13) instead gives f → 0 f\to0 f → 0 at the interior axis point z = Z / 2 z=Z/2 z = Z /2 . A finite positive conformal normalization and a nonsingular Killing-basis change cannot remove that zero. This contradicts the necessary axis behavior, independently of balance.
There is also a generic invariant symptom of the failure. For the fixed raw Killing vector ξ = ∂ ψ − ∂ ϕ \xi=\partial_\psi-\partial_\phi ξ = ∂ ψ − ∂ ϕ , the exact metric gives, when E Δ e q ≠ 0 \mathcal{E}\Delta_{\rm eq}\ne0 E Δ eq = 0 ,
g ( ξ , ξ ) = 16 Z 2 β 2 Δ e q s + O ( 1 ) = 16 Z 4 β 2 Δ e q ρ 2 + O ( 1 ) . (301) g(\xi,\xi)=\frac{16Z^2\beta^2}{\Delta_{\rm eq}s}+O(1)=\frac{16Z^4\beta^2}{\Delta_{\rm eq}\rho^2}+O(1). \tag*{(301)} g ( ξ , ξ ) = Δ eq s 16 Z 2 β 2 + O ( 1 ) = Δ eq ρ 2 16 Z 4 β 2 + O ( 1 ) . (301)
The norm of a finite fixed linear combination of smooth Killing fields cannot diverge at a regular axis point. We do not use this generic residue on its exceptional loci, and do not assume that ξ \xi ξ is itself a periodic physical rotation. The uniform conformal identity already excludes those loci without either assumption.
Checks and boundary limits
We checked the matching and axis identities with 13 exact algebraic tests, including the exceptional loci. Independent Laurent calculations at two points in each of the 16 sign sectors of ( c , H , v , β ) (c,H,v,\beta) ( c , H , v , β ) reproduce the horizon generators and axis coefficients with normalised errors below 4 × 10 − 14 4\times10^{-14} 4 × 1 0 − 14 . These checks support the symbolic identities. They do not give interval bounds on a wider parameter region. For example, Z = r 0 = c = H = β = 1 Z=r_0=c=H=\beta=1 Z = r 0 = c = H = β = 1 , v = 6 v=6 v = 6 , d = 18 d=18 d = 18 and w = − 7 w=-7 w = − 7 give equal raw angular velocities ( 1 / 2 , 3 / 2 ) (1/2,3/2) ( 1/2 , 3/2 ) , but f = 275.625 ρ 2 + O ( ρ 4 ) f=275.625\rho^2+O(\rho^4) f = 275.625 ρ 2 + O ( ρ 4 ) at z = 1 / 2 z=1/2 z = 1/2 . The middle axis is singular.
Limits that remain matched. We can extend the obstruction analytically to convergent limits that satisfy angular matching at every nonzero value of the limiting parameter. After rescaling the separation to keep the horizon points distinct, the middle-slice base coefficient belongs to
V = span { s ( 1 + s ) 7 , s 2 ( 1 + s ) 7 , … , s 7 ( 1 + s ) 7 } . (302) \mathcal{V}=\operatorname{span}\left\{\frac{s}{(1+s)^7},\frac{s^2}{(1+s)^7},\ldots,\frac{s^7}{(1+s)^7}\right\}. \tag*{(302)} V = span { ( 1 + s ) 7 s , ( 1 + s ) 7 s 2 , … , ( 1 + s ) 7 s 7 } . (302)
Constant parameter-dependent normalisations preserve this space. Finite limits at seven distinct off-axis values of s s s force the seven coefficients to converge, since the evaluation matrix is invertible. The limit therefore still vanishes at s = 0 s=0 s = 0 and fails the regular-axis condition. This includes correlated zero or infinite parameters without exchanging parameter and axis limits. Paths that become matched only at their endpoint require separate tests.
We examined 53 exact rational boundary paths near Z = c = H = r 0 = β = 1 Z=c=H=r_0=\beta=1 Z = c = H = r 0 = β = 1 , v = 6 v=6 v = 6 . Seventeen remain matched and 36 perturb d , w d,w d , w independently. Forty-four retain the axis zero. The other nine fail the primitive angular identification required for a Saturn. In a further 192 simultaneous limits extracted directly from the six-pole metric, we scaled one of ( c , H , v , β , r 0 , Z ) (c,H,v,\beta,r_0,Z) ( c , H , v , β , r 0 , Z ) as τ ± 1 \tau^{\pm1} τ ± 1 in all 16 sign sectors, with d , w d,w d , w fixed by the matching relations. Ninety-six retain a zero middle-axis coefficient, 80 have a negative base coefficient and 16 lose the central extremal double pole. None retains two regular extremal horizons.
Independent retuning of the BZ coefficients. We also allowed the physical BZ parameters ( α , b 0 , C 3 , η ) (\alpha,b_0,C_3,\eta) ( α , b 0 , C 3 , η ) to vary independently and wrote χ = − α b 0 \chi=-\alpha b_0 χ = − α b 0 . Two direct boundary calculations yield exact obstructions. For S = 1 / r 0 2 S=1/r_0^2 S = 1/ r 0 2 , α = ( − 1 + S L ) / β \alpha=(-1+SL)/\beta α = ( − 1 + S L ) / β , b 0 = β b_0=\beta b 0 = β and L = τ 2 / 2 L=\tau^2/2 L = τ 2 /2 , the matching equations are (D.9)–(D.10) with r 0 − 2 r_0^{-2} r 0 − 2 replaced by S S S . We obtain
f m a t c h = s P 6 ( s ) 8 H 4 Z 6 β 4 ( 1 + s ) 7 . (303) f_{\mathrm{match}}=\frac{sP_6(s)}{8H^4Z^6\beta^4(1+s)^7}. \tag*{(303)} f match = 8 H 4 Z 6 β 4 ( 1 + s ) 7 s P 6 ( s ) . (303)
This calculation includes S = 0 S=0 S = 0 imposed before eliminating the Gram matrix. For the simultaneous scaling H = H 0 / τ H=H_0/\tau H = H 0 / τ , matching instead requires
d = c α β Z , w = c 2 Z 2 + c S Z − ( Z + 1 ) c 2 Z + β c v 2 Z 2 , (304) d=\frac{c\alpha\beta}{Z},\qquad w=\frac{c^2}{Z^2}+\frac{cS}{Z}-\frac{(Z+1)c^2}{Z}+\frac{\beta c v}{2Z^2}, \tag*{(304)} d = Z c α β , w = Z 2 c 2 + Z c S − Z ( Z + 1 ) c 2 + 2 Z 2 β c v , (304)
and the rescaled base coefficient is
f ~ m a t c h = lim τ → 0 τ 2 f m a t c h = s P ~ 6 ( s ) 8 Z 6 β 4 ( 1 + s ) 7 . (305) \widetilde{f}_{\mathrm{match}}=\lim_{\tau\to0}\tau^2 f_{\mathrm{match}}=\frac{s\widetilde{P}_6(s)}{8Z^6\beta^4(1+s)^7}. \tag*{(305)} f match = τ → 0 lim τ 2 f match = 8 Z 6 β 4 ( 1 + s ) 7 s P 6 ( s ) . (305)
Both polynomials have degree at most six. Thus the regular-axis obstruction persists in these two boundary classes. This conclusion assumes the leading Gram rank used in the extraction. Further rank degenerations require a separate calculation.
We followed one exceptional cancellation through two more orders, allowing independent subleading BZ coefficients. The ring then has at most a simple radial pole, while the central double pole survives. This excludes a finite-area extremal ring on that branch. A separate finite-auxiliary-pole ansatz also fails its necessary leading-order cancellation condition. These are exclusions of specific mechanisms. The ancillary material retains the ansatz and exact coefficient calculations.
The wider numerical searches give the following results. We extract the limiting metric coefficients before testing the horizons.
Table Search.
We checked representative exclusions by increasing the precision and truncation order and comparing exact rational calculations with the direct six-pole metric. Full parameter lists are in the ancillary material.
No searched branch retains two regular extremal horizons with equal angular velocities. The exact no-go result covers the finite family of § 4 and the boundary classes specified above. These wider searches provide numerical evidence for the nonexistence of doubly spinning extremal Saturns in thermodynamic equilibrium.
References