Introduction
Among all physics associated with black holes, Hawking radiation offers one of the clearest links between gravitation and quantum physics. In 1974, Hawking showed that black holes formed by gravitational collapse emit radiation at a temperature determined by their surface gravity [1 , 2 ] . His approach treats quantum fields on a classical collapsing spacetime, showing that a state with no incoming particles gives rise to an outgoing flux at late times. The spectrum reaching infinity is modulated by greybody factors, which account for scattering by the exterior geometry. Beyond predicting radiation from an object classically unable to emit it, this result gave physical content to the analogy between black hole mechanics and thermodynamics [3 ] . It also fixed the normalization of Bekenstein’s proposal that a black hole possesses an intrinsic entropy directly proportional to its horizon area, an idea motivated by the apparent loss of entropy when matter falls into a black hole [4 , 5 ] . The outgoing radiation, in turn, carries energy away, so the geometry cannot remain unchanged as the black hole loses mass and evaporates. An isolated Schwarzschild black hole offers a simple illustration of this feedback; as its mass decreases, its temperature rises and the emission accelerates. Even so, the estimated evaporation times for stellar masses far exceed the present age of the Universe [6 ] .
The low Hawking temperature of stellar-mass black holes also makes their radiation exceedingly difficult to distinguish from the astrophysical background [2 ] . While Hawking radiation was originally a purely theoretical prediction, Unruh opened a way to study its analogue in the laboratory by considering sound in a transonic flow [7 ] . In this setting, sound perturbations propagate on an effective spacetime with a sonic horizon, and where their quanta, the phonons, play the role of the particles emitted in the gravitational case. The idea extends beyond fluids. In a nonlinear optical fiber, for instance, a light pulse can produce a moving perturbation of the refractive index that acts as a horizon for probe light, an arrangement used to observe stimulated analogue Hawking emission [8 , 9 ] . What makes these analogues especially useful is that they bring the conversion of modes near a horizon within experimental reach, allowing a key part of the emission process to be studied under controlled conditions.
In the gravitational setting, Hawking radiation also raises a question about what becomes of the initial quantum state. If a black hole formed from a pure state evaporates completely, leaving behind radiation in a mixed state, the evolution would not be unitary [10 ] . A major advance in describing the emission process came with a series of seminal works [11 , 12 , 13 , 14 ] that laid the foundations of the Kraus–Parikh–Wilczek (KPW) approach, a semiclassical framework that describes Hawking radiation as quantum tunneling across the event horizon. By enforcing energy conservation, the approach accounts for the energy carried away by the outgoing shell and the accompanying shift of the horizon. It is this backreaction that gives rise to corrections to the leading thermal tunneling factor. These departures from thermality have been discussed in connection with the information paradox [15 ] , although the emission probabilities alone do not establish whether the full evolution is unitary.
This tunneling picture can be extended to anti-de Sitter spacetimes, with an early application to Schwarzschild–AdS black holes [16 ] . Closer to our problem is Medved’s analysis of neutral emission from the electrically charged Ba\u{n}ados–Teitelboim–Zanelli (BTZ) black hole, which also considered the approach to extremality [17 ] . For rotating BTZ, Wu and Jiang accounted for the angular momentum carried away by the emitted shell [18 ] . These examples bring changes in the black hole parameters into the calculation of the emission probability. Here we turn the question around and ask what emission does to the horizon structure itself. The nonlinearly charged BTZ family of Ref. [19 ] is particularly well suited to this question, since it admits configurations in which simple and degenerate horizons coexist. The loss of mass and charge can then affect not only the positions of the horizons but also their number and degeneracy.
A black hole can lose charge and still move closer to extremality, as Hiscock and Weems found by following the coupled evolution of mass and charge in sufficiently massive Reissner–Nordström black holes early in their evaporation [20 ] . In the BTZ family considered here, the charge carried away for a given loss of energy selects a path through the stationary configurations. Following that path lets us see how horizon degeneracy changes along the way and whether the black hole moves toward extremality. The rate of energy loss then sets the pace of the evolution. As extremality draws near, the effect of a single emission becomes especially important. Kraus and Wilczek showed that self-interaction corrections can substantially modify the spectrum in this regime [21 ] . The background may evolve slowly enough on the timescale of the outgoing modes for an adiabatic description [22 ] , yet a single quantum can carry away an appreciable fraction of the energy above extremality, making the canonical approximation unreliable [23 ] . More recently, Brown et al. found that quantum fluctuations of the near-extremal geometry can substantially alter the emission spectrum of four-dimensional charged black holes [24 ] .
Organization of this work. We begin in Sec. 2 by studying Hawking emission at finite energy within the KPW description and developing a local inversion of the horizon-defining equation, arriving at formal series expressions for the tunneling probabilities and finite horizon displacements. This construction is extended to degenerate horizons in Sec. 3, where we examine their splitting and possible subsequent coalescence as mass and charge are extracted. We then turn to the nonlinearly charged BTZ family in Sec. 4, following the horizon histories selected by different emission profiles. The approach to outer coalescence motivates the analysis of Sec. 5, which examines the limiting tunneling exponent alongside the duration of the process and the conditions for an adiabatic and thermal description. A complementary application to the sector reproducing the quantum-backreacted BTZ geometry is presented in Sec. 6. We close with a discussion of the results and further directions in Sec. 7, leaving the detailed coefficient formulas and complementary calculations to the appendices.
KPW probability for finite-energy emission
The KPW prescription describes a finite emission between members of a stationary black hole family, with the mass lost by the black hole carried by the outgoing shell. Consider a static family in D ≥ 3 D \ge3 D ≥ 3 spacetime dimensions whose line element can be written in Schwarzschild-like gauge,
d s 2 = − f ( r ; M , λ A ) d t 2 + d r 2 f ( r ; M , λ A ) + r 2 d Σ D − 2 , k 2 . (1) \mathrm{d}s^2=-f(r;\mathcal{M},\lambda^A)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{f(r;\mathcal{M},\lambda^A)}+r^2\mathrm{d}\Sigma^2_{D-2,k}. \tag*{(1)} d s 2 = − f ( r ; M , λ A ) d t 2 + f ( r ; M , λ A ) d r 2 + r 2 d Σ D − 2 , k 2 . (1)
Here, M \mathcal{M} M denotes the black hole mass, while λ A \lambda^A λ A (A = 1 , … , n p A=1,\ldots,n_p A = 1 , … , n p ) collectively denotes the integration constants, couplings, and other parameters held fixed during neutral, radial emission. The metric d Σ D − 2 , k 2 \mathrm{d}\Sigma^2_{D-2,k} d Σ D − 2 , k 2 describes the transverse section of the horizon, with normalized curvature k k k . For a noncompact section, extensive thermodynamic quantities are understood as densities with respect to its volume.
At leading semiclassical order, the probability is given by Γ ≃ exp ( − 2 Im I ) \Gamma\simeq\exp(-2\operatorname{Im} I) Γ ≃ exp ( − 2 Im I ) , where I I I denotes the semiclassical tunneling action and where we use ℏ = 1 \hbar=1 ℏ = 1 throughout. Its imaginary contribution arises from the radial pole crossed by the outgoing trajectory. In the gauge of Eq. (1) , the relevant pole occurs at the event horizon, which marks the causal barrier forbidding classical escape from the interior. The same function therefore determines the event horizon, its kinematics as the mass decreases due to emission, and the radial pole contributing to the tunneling probability.
This class of spacetimes (1) includes asymptotically flat, de Sitter (dS), and anti-de Sitter (AdS) black holes with admissible transverse sections. Since the corresponding residue is determined locally near the horizon, the asymptotic boundary conditions governing the subsequent propagation of radiation do not enter its evaluation. The interpretation of a sequence of such emissions as a complete evaporation process, however, requires boundary conditions that allow the outgoing flux to escape the system. This condition is particularly relevant for asymptotically AdS geometries, where reflecting boundary conditions would return the emitted radiation from infinity. Evaporation with absorbing boundary conditions at AdS infinity has been analyzed, for instance, in Ref. [25 ] .
Concretely, let r = r + r=r_{+} r = r + denote the position of the initial event horizon, which we first assume to be a simple outer zero of the blackening function,
f ( r + ; M , λ A ) = 0 , ∂ r f ( r + ; M , λ A ) > 0 , (2) f(r_{+};\mathcal{M},\lambda^{A})=0,\qquad\partial_{r}f(r_{+};\mathcal{M},\lambda^{A})>0, \tag*{(2)} f ( r + ; M , λ A ) = 0 , ∂ r f ( r + ; M , λ A ) > 0 , (2)
where for our notations ∂ x f : = ∂ f / ∂ x \partial_{x}f:=\partial f/\partial x ∂ x f := ∂ f / ∂ x throughout. In this case, the near-horizon zero is linear, and both the outgoing trajectory and the horizon displacement admit the ordinary local expansions developed below. Degenerate roots appear later as limiting configurations in which these expansions change character; they play a central role in what follows.
Throughout the emission, we follow the outer branch continuously connected to this initial horizon. The stationary Killing field ∂ t \partial_{t} ∂ t is assumed to preserve its normalization across the family, and the energy and temperature below use this normalization throughout the sequence. Here, horizon motion refers to stationary Killing horizons parametrized by the emitted energy.
Neutral emission along the tunneling path
A Gullstrand–Painlevé-type time coordinate [26 ] makes the metric (1) regular near the horizon,
d τ = d t + 1 − f f d r . (3) \mathrm{d}\tau=\mathrm{d}t+\frac{\sqrt{1-f}}{f}\mathrm{d}r. \tag*{(3)} d τ = d t + f 1 − f d r . (3)
In this chart, the outgoing radial null geodesic obeys
r ˙ = 1 − 1 − f = 1 2 ∂ r f ( r + ) ( r − r + ) + O ( ( r − r + ) 2 ) , (4) \dot{r}=1-\sqrt{1-f}=\frac{1}{2}\partial_{r}f(r_{+})(r-r_{+})+O\left((r-r_{+})^{2}\right), \tag*{(4)} r ˙ = 1 − 1 − f = 2 1 ∂ r f ( r + ) ( r − r + ) + O ( ( r − r + ) 2 ) , (4)
where the dot denotes differentiation with respect to τ \tau τ and the expansion is about the initial horizon. The pole in 1 / r ˙ 1/\dot{r} 1/ r ˙ gives the usual Boltzmann factor when the geometry is kept fixed. The KPW prescription follows the displacement of this pole as energy is extracted from the black hole. Let E > 0 \mathcal{E}>0 E > 0 denote the total emitted energy and E ′ ∈ [ 0 , E ] \mathcal{E}'\in[0,\mathcal{E}] E ′ ∈ [ 0 , E ] its running value. The instantaneous black hole mass and event horizon are
M ′ = M − E ′ , R + ( E ′ ) = r + ( M − E ′ , λ A ) , (5) \mathcal{M}'=\mathcal{M}-\mathcal{E}',\qquad R_{+}(\mathcal{E}')=r_{+}(\mathcal{M}-\mathcal{E}',\lambda^{A}), \tag*{(5)} M ′ = M − E ′ , R + ( E ′ ) = r + ( M − E ′ , λ A ) , (5)
with R + ( 0 ) = r + R_{+}(0)=r_{+} R + ( 0 ) = r + and satisfying the horizon condition
f [ R + ( E ′ ) ; M − E ′ , λ A ] = 0. (6) f\left[R_{+}(\mathcal{E}');\mathcal{M}-\mathcal{E}',\lambda^{A}\right]=0. \tag*{(6)} f [ R + ( E ′ ) ; M − E ′ , λ A ] = 0. (6)
at each stage of the emission process. Throughout the text, we retain this convention for other functions of the emitted energy and similar quantities, where a prime denotes an intermediate running value.
For this neutral contour calculation, we assume that the outer branch shrinks as energy is emitted, equivalently ∂ M f < 0 \partial_{\mathcal{M}} f < 0 ∂ M f < 0 at the running horizon. Hamilton’s equation replaces the radial momentum by the black hole Hamiltonian H = M − E ′ \mathcal{H} = \mathcal{M} - \mathcal{E}' H = M − E ′ . The resulting imaginary part of the action is
Im I = Im ∫ r i n r o u t ∫ M M − E d H r ˙ ( r ; H , λ A ) d r . (7) \operatorname{Im} I = \operatorname{Im} \int_{r_{\mathrm{in}}}^{r_{\mathrm{out}}} \int_{\mathcal{M}}^{\mathcal{M}-\mathcal{E}} \frac{d\mathcal{H}}{\dot{r}(r;\mathcal{H},\lambda^{A})}\,dr. \tag*{(7)} Im I = Im ∫ r in r out ∫ M M − E r ˙ ( r ; H , λ A ) d H d r . (7)
The shell starts just inside the initial horizon and emerges just outside the final one. The radial endpoints therefore approach r i n = r + r_{\mathrm{in}} = r_{+} r in = r + and r o u t = R + ( E ) r_{\mathrm{out}} = R_{+}(\mathcal{E}) r out = R + ( E ) , with r o u t < r i n r_{\mathrm{out}} < r_{\mathrm{in}} r out < r in for the shrinking branch, despite the outgoing character of the shell [14 ] . At every intermediate energy, the radial contour crosses r = R + ( E ′ ) r = R_{+}(\mathcal{E}') r = R + ( E ′ ) . The instantaneous Hawking temperature is
T H ( E ′ ) = 1 4 π ∂ r f [ R + ( E ′ ) ; M − E ′ , λ A ] . (8) T_{\mathrm{H}}(\mathcal{E}') = \frac{1}{4\pi}\partial_{r}f\left[R_{+}(\mathcal{E}');\mathcal{M}-\mathcal{E}',\lambda^{A}\right]. \tag*{(8)} T H ( E ′ ) = 4 π 1 ∂ r f [ R + ( E ′ ) ; M − E ′ , λ A ] . (8)
At the running pole, this orientation and the positive-energy contour prescription give
Im ∫ d r r ˙ = − 1 2 T H ( E ′ ) . \operatorname{Im}\int\frac{dr}{\dot{r}} = -\frac{1}{2T_{\mathrm{H}}(\mathcal{E}')}. Im ∫ r ˙ d r = − 2 T H ( E ′ ) 1 .
Combining this with d H = − d E ′ d\mathcal{H} = -d\mathcal{E}' d H = − d E ′ therefore gives
Im I = 1 2 ∫ 0 E d E ′ T H ( E ′ ) . (9) \operatorname{Im} I = \frac{1}{2}\int_{0}^{\mathcal{E}}\frac{d\mathcal{E}'}{T_{\mathrm{H}}(\mathcal{E}')}. \tag*{(9)} Im I = 2 1 ∫ 0 E T H ( E ′ ) d E ′ . (9)
At each E ′ \mathcal{E}' E ′ , the black hole enclosed by the shell is identified with a stationary member of the same family, to which an instantaneous entropy and the stationary first law can be assigned. The entropy S S S is the gravitational entropy entering that first law and need not coincide with the area entropy in a general theory [27 , 28 ] .
Let Q a = Q a ( λ A ) Q^{a} = Q^{a}(\lambda^{A}) Q a = Q a ( λ A ) , with a = 1 , … , n q a = 1,\ldots,n_{q} a = 1 , … , n q , denote the independent global charges determined by the parameters λ A \lambda^{A} λ A , and let Φ a \Phi_{a} Φ a be their conjugate potentials. Keeping the couplings fixed, the first law at an intermediate stage takes the form
d M ′ = T H ( E ′ ) d S + ∑ a = 1 n q Φ a ( E ′ ) d Q a . (10) d\mathcal{M}' = T_{\mathrm{H}}(\mathcal{E}')dS + \sum_{a=1}^{n_{q}}\Phi_{a}(\mathcal{E}')dQ^{a}. \tag*{(10)} d M ′ = T H ( E ′ ) d S + a = 1 ∑ n q Φ a ( E ′ ) d Q a . (10)
For the neutral radial channel, we assume that only the mass varies among the independent solution parameters. Thus λ A \lambda^{A} λ A remains fixed and d Q a = 0 dQ^{a} = 0 d Q a = 0 , so d M ′ = T H ( E ′ ) d S d\mathcal{M}' = T_{\mathrm{H}}(\mathcal{E}')dS d M ′ = T H ( E ′ ) d S throughout the tunneling path. Since d M ′ = − d E ′ d\mathcal{M}' = -d\mathcal{E}' d M ′ = − d E ′ , the entropy difference is
Δ S = S ( M − E , λ A ) − S ( M , λ A ) = − ∫ 0 E d E ′ T H ( E ′ ) . (11) \Delta S = S(\mathcal{M}-\mathcal{E},\lambda^{A}) - S(\mathcal{M},\lambda^{A}) = -\int_{0}^{\mathcal{E}}\frac{d\mathcal{E}'}{T_{\mathrm{H}}(\mathcal{E}')}. \tag*{(11)} Δ S = S ( M − E , λ A ) − S ( M , λ A ) = − ∫ 0 E T H ( E ′ ) d E ′ . (11)
Combining this expression with Eq. (9) yields the usual entropy form of the backreacting tunneling probability [29 , 30 ] ,
Im I = − 1 2 Δ S , Γ K P W ≃ exp ( Δ S ) . (12) \operatorname{Im} I = -\frac{1}{2}\Delta S,\qquad\Gamma_{\mathrm{KPW}} \simeq\exp(\Delta S). \tag*{(12)} Im I = − 2 1 Δ S , Γ KPW ≃ exp ( Δ S ) . (12)
without assuming an area law for S S S , provided the stationary first law holds along the tunneling path.
Inversion of the horizon condition
The finite-emission dependence of the KPW exponent can be recovered from the horizon displacement once the entropy is parametrized along the stationary family. This amounts to locally inverting the horizon equation.
At fixed λ A \lambda^{A} λ A , provided ∂ M f ( r + ; M , λ A ) ≠ 0 \partial_{\mathcal{M}}f(r_{+};\mathcal{M},\lambda^{A}) \ne0 ∂ M f ( r + ; M , λ A ) = 0 , the implicit function theorem allows the horizon equation to be solved locally in terms of a mass function
M = m ( r ) . (13) \mathcal{M}=\mathrm{m}(r). \tag*{(13)} M = m ( r ) . (13)
Thus M = m ( r + ) \mathcal{M}=\mathrm{m}(r_{+}) M = m ( r + ) . Writing Δ r + ( E ) : = R + ( E ) − r + \Delta r_{+}(\mathcal{E}) := R_{+}(\mathcal{E})-r_{+} Δ r + ( E ) := R + ( E ) − r + for the finite displacement, the final horizon position obeys M − E = m ( r + + Δ r + ) \mathcal{M}-\mathcal{E}=\mathrm{m}(r_{+}+\Delta r_{+}) M − E = m ( r + + Δ r + ) . We reserve δ \delta δ for infinitesimal variations and perturbative corrections. Defining
h ( y ) : = m ( r + ) − m ( r + + y ) , (14) h(y):=\mathrm{m}(r_{+})-\mathrm{m}(r_{+}+y), \tag*{(14)} h ( y ) := m ( r + ) − m ( r + + y ) , (14)
the displacement is determined by E = h ( Δ r + ) \mathcal{E}=h(\Delta r_{+}) E = h ( Δ r + ) . Its local inverse exists because
∂ y h ( 0 ) = − ∂ r m ( r + ) = ∂ r f ( r + ; M , λ A ) ∂ M f ( r + ; M , λ A ) ≠ 0. (15) \partial_{y}h(0)=-\partial_{r}\mathrm{m}(r_{+})=\frac{\partial_{r}f(r_{+};\mathcal{M},\lambda^{A})}{\partial_{\mathcal{M}}f(r_{+};\mathcal{M},\lambda^{A})}\ne0. \tag*{(15)} ∂ y h ( 0 ) = − ∂ r m ( r + ) = ∂ M f ( r + ; M , λ A ) ∂ r f ( r + ; M , λ A ) = 0. (15)
If m ( r ) \mathrm{m}(r) m ( r ) is analytic near the initial simple horizon, the Lagrange–Bürmann (LB) formula [31 ] gives Δ r + = ∑ n ≥ 1 a n E n \Delta r_{+}=\sum_{n\ge1}a_{n}\mathcal{E}^{n} Δ r + = ∑ n ≥ 1 a n E n , with
a n = 1 n ! [ ∂ y n − 1 { ( y h ( y ) ) n } ] y = 0 . (16) a_{n}=\frac{1}{n!}\left[\partial_{y}^{n-1}\left\{\left(\frac{y}{h(y)}\right)^{n}\right\}\right]_{y=0}. \tag*{(16)} a n = n ! 1 [ ∂ y n − 1 { ( h ( y ) y ) n } ] y = 0 . (16)
Since ∂ y h ( 0 ) ≠ 0 \partial_{y}h(0)\ne0 ∂ y h ( 0 ) = 0 , the ratio y / h ( y ) y/h(y) y / h ( y ) has a regular extension at y = 0 y=0 y = 0 .
Writing the entropy restricted to the stationary family as S = S ( r , λ A ) S=S(r,\lambda^{A}) S = S ( r , λ A ) , the composed LB formula, assuming this entropy is analytic near r + r_{+} r + , similarly gives Δ S = ∑ n ≥ 1 b n E n \Delta S=\sum_{n\ge1}b_{n}\mathcal{E}^{n} Δ S = ∑ n ≥ 1 b n E n , where
b n = 1 n ! [ ∂ y n − 1 { ∂ y S ( r + + y , λ A ) ( y h ( y ) ) n } ] y = 0 . (17) b_{n}=\frac{1}{n!}\left[\partial_{y}^{n-1}\left\{\partial_{y}S(r_{+}+y,\lambda^{A})\left(\frac{y}{h(y)}\right)^{n}\right\}\right]_{y=0}. \tag*{(17)} b n = n ! 1 [ ∂ y n − 1 { ∂ y S ( r + + y , λ A ) ( h ( y ) y ) n } ] y = 0 . (17)
These coefficients reconstruct the horizon displacement and entropy exponent from the local mass–radius relation and gravitational entropy, without requiring a closed-form horizon radius. The derivations and explicit coefficient extractions are given in Appendix § 9 . The inverse branch satisfying R + ( 0 ) = r + R_{+}(0)=r_{+} R + ( 0 ) = r + admits local continuation while the horizon remains simple, although its Maclaurin series in E \mathcal{E} E about E = 0 \mathcal{E}=0 E = 0 need not converge throughout the emission interval. Independently of convergence, the construction also holds for formal power series [32 ] .
The construction separates the inversion from the specific form of the entropy. In Einstein gravity with minimally coupled matter,
S ( r + ) = Vol ( Σ D − 2 , k ) 4 G D r + D − 2 , (18) S(r_{+})=\frac{\operatorname{Vol}(\Sigma_{D-2,k})}{4G_{D}}r_{+}^{D-2}, \tag*{(18)} S ( r + ) = 4 G D Vol ( Σ D − 2 , k ) r + D − 2 , (18)
with the volume factor omitted for entropy densities. For D = 3 D=3 D = 3 , the entropy is linear in r + r_{+} r + and the exponent is directly proportional to the horizon displacement.
A particularly enlightening specialization of the above result occurs when the blackening function is affine in the black hole mass with a constant coefficient,
f ( r ; M , λ A ) = c [ m ( r ) − M ] , c > 0. (19) f(r;\mathcal{M},\lambda^{A})=c\left[\mathrm{m}(r)-\mathcal{M}\right],\qquad c>0. \tag*{(19)} f ( r ; M , λ A ) = c [ m ( r ) − M ] , c > 0. (19)
The coefficient c c c may depend on the fixed parameters but is independent of r r r and M \mathcal{M} M . The geometric temperature and the first law at fixed λ A \lambda^{A} λ A then require d S / d r + = 4 π / c dS/dr_{+}=4\pi/c d S / d r + = 4 π / c . This is a consistency condition on the chosen family and its gravitational entropy, including a Wald entropy. It therefore applies directly to the three-dimensional Einstein family considered in § 4 . Writing f 0 ( r ) : = f ( r ; M , λ A ) f_{0}(r):=f(r;\mathcal{M},\lambda^{A}) f 0 ( r ) := f ( r ; M , λ A ) for the blackening function of the initial state, neutral emission produces the constant shift
f ( r ; M − E , λ A ) = f 0 ( r ) + c E . (20) f(r;\mathcal{M}-\mathcal{E},\lambda^{A})=f_{0}(r)+c\mathcal{E}. \tag*{(20)} f ( r ; M − E , λ A ) = f 0 ( r ) + c E . (20)
The first inverse terms are
Δ r + = − c E ∂ r f 0 ( r + ) − c 2 ∂ r 2 f 0 ( r + ) 2 [ ∂ r f 0 ( r + ) ] 3 E 2 + O ( E 3 ) . (21) \Delta r_{+}=-\frac{c\mathcal{E}}{\partial_{r}f_{0}(r_{+})}-\frac{c^{2}\partial_{r}^{2}f_{0}(r_{+})}{2[\partial_{r}f_{0}(r_{+})]^{3}}\mathcal{E}^{2}+O(\mathcal{E}^{3}). \tag*{(21)} Δ r + = − ∂ r f 0 ( r + ) c E − 2 [ ∂ r f 0 ( r + ) ] 3 c 2 ∂ r 2 f 0 ( r + ) E 2 + O ( E 3 ) . (21)
Here h ( y ) = − f 0 ( r + + y ) / c h(y)=-f_{0}(r_{+}+y)/c h ( y ) = − f 0 ( r + + y ) / c , so every higher order follows from application of Eq. (16) .
Thermal limit and nonthermal corrections
The entropy dependence in Eq. (12) also admits a statistical-mechanical interpretation. For an isolated composite system in equilibrium, the relative weight of a specified radiation state is determined by the number of compatible states available to the emitting body, giving the factor exp ( Δ S ) \exp(\Delta S) exp ( Δ S ) [33 ] . The usual Boltzmann factor follows upon linearizing the entropy change in the transferred energy, while the quadratic correction accounts for the change in the body’s temperature due to its finite heat capacity.
Accordingly, expanding the entropy difference about the initial configuration results in
Δ S = − E T H − E 2 2 C Q T H 2 + O ( E 3 ) , (22) \Delta S=-\frac{\mathcal{E}}{T_{\mathrm{H}}}-\frac{\mathcal{E}^{2}}{2C_{Q}T_{\mathrm{H}}^{2}}+O(\mathcal{E}^{3}), \tag*{(22)} Δ S = − T H E − 2 C Q T H 2 E 2 + O ( E 3 ) , (22)
where T H : = T H ( 0 ) T_{\mathrm{H}}:=T_{\mathrm{H}}(0) T H := T H ( 0 ) and
C Q : = ( ∂ M ∂ T H ) Q a , ∀ Q a , (23) C_{Q}:=\left(\frac{\partial\mathcal{M}}{\partial T_{\mathrm{H}}}\right)_{\mathcal{Q}^{a}},\qquad\forall\ \mathcal{Q}^{a}, \tag*{(23)} C Q := ( ∂ T H ∂ M ) Q a , ∀ Q a , (23)
are evaluated at the initial state. Any remaining parameters in λ A \lambda^{A} λ A are also held fixed in the derivative defining C Q C_{Q} C Q . The first nonthermal correction therefore suppresses emission relative to the initial temperature Boltzmann factor for C Q > 0 C_{Q}>0 C Q > 0 and enhances it for C Q < 0 C_{Q}<0 C Q < 0 , coinciding with the results in [33 ] . For a finite neutral emission, the suppression extends beyond this expansion if C Q > 0 C_{Q}>0 C Q > 0 throughout the simple outer branch. The running temperature then decreases as energy is extracted, and Eq. (11) gives Δ S ≤ − E / T H \Delta S\leq-\mathcal{E}/T_{\mathrm{H}} Δ S ≤ − E / T H .
Charged emission
For a shell carrying one conserved U ( 1 ) U(1) U ( 1 ) charge, the blackening function depends explicitly on both the mass and electric charge,
f = f ( r ; M , Q , λ A ) . f=f(r;\mathcal{M},\mathcal{Q},\lambda^{A}). f = f ( r ; M , Q , λ A ) .
The remaining parameters λ A \lambda^{A} λ A , including charges associated with other global symmetries and the couplings of the theory, are held fixed. Let Q e m \mathcal{Q}_{\mathrm{em}} Q em be the signed charge transported by the emitted shell. Emission of charge with the same sign as Q \mathcal{Q} Q reduces the magnitude of the black hole charge, whereas oppositely charged matter increases it. We first consider charge extraction within a fixed sign sector. Opposite-sign emission is discussed in § 5.1 .
An emission channel connects ( M , Q ) (\mathcal{M},\mathcal{Q}) ( M , Q ) to ( M − E , Q − Q e m ) (\mathcal{M}-\mathcal{E},\mathcal{Q}-\mathcal{Q}_{\rm em}) ( M − E , Q − Q em ) through a quasistationary path C C C in the space of black hole states. When the extracted energy is strictly monotonic, this path can be parametrized by E ′ \mathcal{E}' E ′ as
C : [ 0 , E ] ⟶ R 2 , E ′ ⟼ ( M − E ′ , Q − Q e m ( E ′ ) ) . (24) C:\ [0,\mathcal{E}]\longrightarrow\mathbb{R}^{2},\qquad\mathcal{E}'\longmapsto\left(\mathcal{M}-\mathcal{E}',\mathcal{Q}-\mathcal{Q}_{\rm em}(\mathcal{E}')\right). \tag*{(24)} C : [ 0 , E ] ⟶ R 2 , E ′ ⟼ ( M − E ′ , Q − Q em ( E ′ ) ) . (24)
Here, Q e m ( E ′ ) \mathcal{Q}_{\rm em}(\mathcal{E}') Q em ( E ′ ) is a continuously differentiable charge profile satisfying Q e m ( 0 ) = 0 \mathcal{Q}_{\rm em}(0)=0 Q em ( 0 ) = 0 and Q e m ( E ) = Q e m \mathcal{Q}_{\rm em}(\mathcal{E})=\mathcal{Q}_{\rm em} Q em ( E ) = Q em . The coordinates on C C C are denoted by ( M ′ , Q ′ ) (\mathcal{M}',\mathcal{Q}') ( M ′ , Q ′ ) , and its image is restricted to the admissible stationary branch at fixed λ A \lambda^{A} λ A . At the endpoint, M ′ ( E ) = M − E \mathcal{M}'(\mathcal{E})=\mathcal{M}-\mathcal{E} M ′ ( E ) = M − E and Q ′ ( E ) = Q − Q e m ( E ) \mathcal{Q}'(\mathcal{E})=\mathcal{Q}-\mathcal{Q}_{\rm em}(\mathcal{E}) Q ′ ( E ) = Q − Q em ( E ) , with M \mathcal{M} M and Q \mathcal{Q} Q referring to the initial state of the emission. The running horizon now satisfies
f [ R + ( E ′ ) ; M ′ , Q ′ , λ A ] = 0 , R + ( 0 ) = r + . (25) f\left[R_{+}(\mathcal{E}');\mathcal{M}',\mathcal{Q}',\lambda^{A}\right]=0,\qquad R_{+}(0)=r_{+}. \tag*{(25)} f [ R + ( E ′ ) ; M ′ , Q ′ , λ A ] = 0 , R + ( 0 ) = r + . (25)
Conservation fixes the endpoints. It does not, by itself, determine the charge profile Q e m ( E ′ ) \mathcal{Q}_{\rm em}(\mathcal{E}') Q em ( E ′ ) .
The gauge potential associated with this charge is A = A μ d x μ A=A_{\mu}dx^{\mu} A = A μ d x μ . In the static, purely electric sector considered here, it can be written as A = A t ( r ) d t A=A_{t}(r)dt A = A t ( r ) d t . The potential Φ ( r ) : = − A t ( r ) \Phi(r):=-A_{t}(r) Φ ( r ) := − A t ( r ) and the conserved energy use the same reference convention as the black hole Hamiltonian. With this convention, the instantaneous horizon value Φ + : = Φ ( R + ) \Phi_{+}:=\Phi(R_{+}) Φ + := Φ ( R + ) is the thermodynamic potential conjugate to Q \mathcal{Q} Q along the radiation process. Both the potential and the horizon position vary along C C C .
For minimally coupled charged matter, the near-horizon radial phase depends on the energy after the gauge work has been subtracted. We adopt the quasistationary charged KPW prescription, which subtracts the electrostatic work from the mass variation entering the radial action [34 , 35 ] . Taking the residue at the instantaneous outer horizon gives
Im I = − 1 2 ∫ C d M ′ − Φ + d Q ′ T H . (26) \operatorname{Im} I=-\frac{1}{2}\int_{C}\frac{d\mathcal{M}'-\Phi_{+}d\mathcal{Q}'}{T_{\mathrm{H}}}. \tag*{(26)} Im I = − 2 1 ∫ C T H d M ′ − Φ + d Q ′ . (26)
All quantities in this integral are evaluated at that horizon. Other zeros of f f f , if present, are not swept by the contour for emission across the exterior horizon. The charged horizon residue and its integration along C C C are derived in Appendix § 10 , using a regular gauge at the future horizon and the quasistationary KPW prescription.
By the stationary first law d M ′ = T H d S + Φ + d Q ′ d\mathcal{M}'=T_{\mathrm{H}}dS+\Phi_{+}d\mathcal{Q}' d M ′ = T H d S + Φ + d Q ′ at fixed λ A \lambda^{A} λ A , the one-form in Eq. (26) is d S dS d S . Thus Eq. (12) holds with
Δ S = S ( M − E , Q − Q e m , λ A ) − S ( M , Q , λ A ) , (27) \Delta S=S(\mathcal{M}-\mathcal{E},\mathcal{Q}-\mathcal{Q}_{\rm em},\lambda^{A})-S(\mathcal{M},\mathcal{Q},\lambda^{A}), \tag*{(27)} Δ S = S ( M − E , Q − Q em , λ A ) − S ( M , Q , λ A ) , (27)
The leading horizon exponent depends only on the endpoints for paths on the same admissible stationary branch.
In the parametrization of Eq. (24) , the running-horizon charged integral (26) becomes
Im I = 1 2 ∫ 0 E 1 − Φ + ( E ′ ) ( d Q e m ( E ′ ) / d E ′ ) T H ( E ′ ) d E ′ . (28) \operatorname{Im} I=\frac{1}{2}\int_{0}^{\mathcal{E}}\frac{1-\Phi_{+}(\mathcal{E}')\left(d\mathcal{Q}_{\rm em}(\mathcal{E}')/d\mathcal{E}'\right)}{T_{\mathrm{H}}(\mathcal{E}')}d\mathcal{E}'. \tag*{(28)} Im I = 2 1 ∫ 0 E T H ( E ′ ) 1 − Φ + ( E ′ ) ( d Q em ( E ′ ) / d E ′ ) d E ′ . (28)
We restrict attention to ordinary positive horizon-energy channels, whose extracted increments obey d E ′ − Φ + d Q e m > 0 d\mathcal{E}'-\Phi_{+}d\mathcal{Q}_{\rm em}>0 d E ′ − Φ + d Q em > 0 . They give a positive imaginary action and a decreasing black hole entropy. For bosonic modes, the opposite sign is associated with the superradiant sector, where the exponential enters an emission-to-absorption ratio rather than a standalone emission probability, see [36 ] .
The local mass function now becomes M = m ( r , Q ) \mathcal{M} = \mathfrak{m}(r,\mathcal{Q}) M = m ( r , Q ) , with the remaining λ A \lambda^{A} λ A suppressed. The final horizon satisfies
M − E = m ( r + + Δ r + , Q − Q e m ) . (29) \mathcal{M} - \mathcal{E} = \mathfrak{m}(r_{+} + \Delta r_{+},\mathcal{Q} - \mathcal{Q}_{\mathrm{em}}). \tag*{(29)} M − E = m ( r + + Δ r + , Q − Q em ) . (29)
The radial displacement follows by inverting the mass–radius relation at the final charge Q − Q e m \mathcal{Q} - \mathcal{Q}_{\mathrm{em}} Q − Q em . This gives a family of inverses parametrized by the charge transfer, whose evaluation along the chosen profile Q e m ( E ) \mathcal{Q}_{\mathrm{em}}(\mathcal{E}) Q em ( E ) determines the evolving horizon. To retain a function vanishing at y = 0 y = 0 y = 0 , define, analogously,
h ( y ; Q e m ) : = m ( r + , Q − Q e m ) − m ( r + + y , Q − Q e m ) . (30) h(y;\mathcal{Q}_{\mathrm{em}}) := \mathfrak{m}(r_{+},\mathcal{Q} - \mathcal{Q}_{\mathrm{em}}) - \mathfrak{m}(r_{+} + y,\mathcal{Q} - \mathcal{Q}_{\mathrm{em}}). \tag*{(30)} h ( y ; Q em ) := m ( r + , Q − Q em ) − m ( r + + y , Q − Q em ) . (30)
The final horizon equation is then equivalent to
h ( Δ r + ; Q e m ) = E + m ( r + , Q − Q e m ) − M . (31) h(\Delta r_{+};\mathcal{Q}_{\mathrm{em}}) = \mathcal{E} + \mathfrak{m}(r_{+},\mathcal{Q} - \mathcal{Q}_{\mathrm{em}}) - \mathcal{M}. \tag*{(31)} h ( Δ r + ; Q em ) = E + m ( r + , Q − Q em ) − M . (31)
This choice of reference mass makes the neutral inversion formula applicable without assuming an explicit relation between emitted energy and charge. For a jointly analytic mass function with ∂ r m ≠ 0 \partial_{r}\mathfrak{m} \ne0 ∂ r m = 0 , the inverse coefficients depend analytically on the charge transfer Q e m \mathcal{Q}_{\mathrm{em}} Q em near Q e m = 0 \mathcal{Q}_{\mathrm{em}} = 0 Q em = 0 . The resulting horizon and entropy coefficients are worked out in Appendix A. If S ( r , Q ) S(r,\mathcal{Q}) S ( r , Q ) has explicit charge dependence at fixed radius, the entropy change also contains that variation. Of course, such a contribution vanishes for the standard Einstein area entropy.
At first order, differentiation of the horizon condition gives
Δ r + = ( ∂ M f ) E + ( ∂ Q f ) Q e m ∂ r f + ⋯ , (32) \Delta r_{+} = \frac{(\partial_{\mathcal{M}}f)\mathcal{E} + (\partial_{\mathcal{Q}}f)\mathcal{Q}_{\mathrm{em}}}{\partial_{r}f} + \cdots, \tag*{(32)} Δ r + = ∂ r f ( ∂ M f ) E + ( ∂ Q f ) Q em + ⋯ , (32)
with derivatives at the initial horizon and omitted terms of total degree at least two. When S = S ( r ) S = S(r) S = S ( r ) and ∂ r S ( r ) > 0 \partial_{r}S(r) > 0 ∂ r S ( r ) > 0 , the first law gives T H ( ∂ r S ) ( R + ) d R + = − ( d E ′ − Φ + d Q e m ) < 0 T_{\mathrm{H}}(\partial_{r}S)(R_{+})\,dR_{+} = -(d\mathcal{E}^{\prime} - \Phi_{+}d\mathcal{Q}_{\mathrm{em}}) < 0 T H ( ∂ r S ) ( R + ) d R + = − ( d E ′ − Φ + d Q em ) < 0 for an ordinary channel. Thus the outer horizon shrinks due to the emission, although the charge contribution may oppose the mass contribution. With explicit charge dependence in the entropy, its decrease alone does not determine the sign of the radial displacement.
A joint Taylor expansion about the initial state expresses the quadratic correction in terms of the entropy Hessian [37 ] ,
Δ S = − E − Φ + Q e m T H + 1 2 [ ( ∂ M 2 S ) E 2 + 2 ( ∂ M ∂ Q S ) E Q e m + ( ∂ Q 2 S ) Q e m 2 ] + ⋯ . (33) \Delta S = -\frac{\mathcal{E} - \Phi_{+}\mathcal{Q}_{\mathrm{em}}}{T_{\mathrm{H}}} + \frac{1}{2}\left[(\partial_{\mathcal{M}}^{2}S)\mathcal{E}^{2} + 2(\partial_{\mathcal{M}}\partial_{\mathcal{Q}}S)\mathcal{E}\mathcal{Q}_{\mathrm{em}} + (\partial_{\mathcal{Q}}^{2}S)\mathcal{Q}_{\mathrm{em}}^{2}\right] + \cdots. \tag*{(33)} Δ S = − T H E − Φ + Q em + 2 1 [ ( ∂ M 2 S ) E 2 + 2 ( ∂ M ∂ Q S ) E Q em + ( ∂ Q 2 S ) Q em 2 ] + ⋯ . (33)
All derivatives, T H T_{\mathrm{H}} T H , and Φ + \Phi_{+} Φ + here refer to the initial state. The omitted terms have total degree at least three. The linear term is the thermal exponent with its chemical potential arising from the electromagnetic sector.
For the case of a linear charge profile Q e m ( E ) = η E \mathcal{Q}_{\mathrm{em}}(\mathcal{E}) = \eta\mathcal{E} Q em ( E ) = η E , substitution in Eq. (33) gives the correction along the emission direction. As η → 0 \eta\to0 η → 0 , this reduces to Eq. (22) , since ∂ M 2 S = − 1 / ( C Q T H 2 ) \partial_{\mathcal{M}}^{2}S = -1/(C_{\mathcal{Q}}T_{\mathrm{H}}^{2}) ∂ M 2 S = − 1/ ( C Q T H 2 ) . The mixed and charge derivatives also enter for charged emission, so C Q C_{\mathcal{Q}} C Q alone does not determine the sign.
A smooth nonlinear profile with the same initial slope d Q e m / d E ∣ 0 = η \left.\mathrm{d}Q_{\rm em}/\mathrm{d}\mathcal{E}\right|_0=\eta d Q em / d E ∣ 0 = η adds the following quadratic contribution to the linear-profile expansion:
Φ + 2 T H ( ∂ E 2 Q e m ) ( 0 ) E 2 . \frac{\Phi_+}{2T_H}\left(\partial_{\mathcal{E}}^2Q_{\rm em}\right)(0)\mathcal{E}^2. 2 T H Φ + ( ∂ E 2 Q em ) ( 0 ) E 2 .
This contribution is already included when the thermal exponent of the initial state − ( E − Φ + Q e m ( E ) ) / T H -\left(\mathcal{E}-\Phi_+Q_{\rm em}(\mathcal{E})\right)/T_H − ( E − Φ + Q em ( E ) ) / T H retains the full charge profile.
The neutral and charged prescriptions above apply while the outer horizon remains simple. If this branch approaches a degenerate configuration, the entropy difference may still have a finite limit. The change in the local inverse at degeneracy, and its consequences for the other horizon branches, are examined next.
Horizon splitting and coalescence under emission
At a degenerate horizon, the linear term in the radial displacement vanishes, so the ordinary inverse of § 2 must be replaced by a Puiseux-type expansion. We use it to determine the nearby real roots, when they exist, and how their separation depends on the emission channel. A sequence of quasistationary emissions can change both the positions and the number of horizons. More precisely, an initially degenerate inner horizon may split into two simple zeros, one of which can later merge with the event horizon. Figure 1 illustrates one such sequence, allowing the profile of f ( r ) f(r) f ( r ) to deform as energy and charge are lost.
Figure 1. Schematic horizon motion and rearrangement along an illustrative charged emission channel. (a) An internal double zero at r d r_d r d coexists with a simple event horizon at r + r_+ r + . (b) The inner pair separates immediately after emission, giving three simple roots r − < r 0 < R + r_- < r_0 < R_+ r − < r 0 < R + . (c) The middle and event horizons merge at r e r_e r e , while r − r_- r − remains simple. Red, violet, and blue markers follow the innermost, middle, and event horizons, respectively, with two-color markers indicating double zeros. In panels (b) and (c), the dashed curve is the preceding profile and the shaded region lies between the two curves. The event-horizon slope is positive in panels (a,b) and vanishes at the merger in panel (c). The asymptotic form shown is illustrative; the horizon sequence depends on the stationary family, initial state, and chosen emission channel.
Neutral splitting under a uniform mass shift
A good starting point is the affine mass dependence of Eq. (19) , for which neutral emission produces the uniform shift f ( r ; M − E , λ A ) = f 0 ( r ) + c E f(r;\mathcal{M}-\mathcal{E},\lambda^A)=f_0(r)+c\mathcal{E} f ( r ; M − E , λ A ) = f 0 ( r ) + c E , with all parameters λ A \lambda^A λ A held fixed.
Let r = r d r=r_{\mathrm d} r = r d be a double zero of the initial blackening function,
f 0 ( r d ) = ∂ r f 0 ( r d ) = 0 , ∂ r 2 f 0 ( r d ) ≠ 0. (34) f_0(r_{\mathrm d})=\partial_r f_0(r_{\mathrm d})=0,\qquad\partial_r^2 f_0(r_{\mathrm d})\ne0. \tag*{(34)} f 0 ( r d ) = ∂ r f 0 ( r d ) = 0 , ∂ r 2 f 0 ( r d ) = 0. (34)
The local analysis applies to any such zero. The causal character of the associated horizons depends on the global geometry. In particular, an internal double horizon may coexist with a simple event horizon through which the tunneling process takes place.
Writing the displaced root at the final stage as r d , f : = r d + Δ r d r_{\mathrm d,f}:=r_{\mathrm d}+\Delta r_{\mathrm d} r d , f := r d + Δ r d , its horizon condition becomes
0 = c E + ∂ r 2 f 0 ( r d ) 2 ( Δ r d ) 2 + ∂ r 3 f 0 ( r d ) 6 ( Δ r d ) 3 + O ( ( Δ r d ) 4 ) . (35) 0=c\mathcal{E}+\frac{\partial_r^2 f_0(r_{\mathrm d})}{2}(\Delta r_{\mathrm d})^2+\frac{\partial_r^3 f_0(r_{\mathrm d})}{6}(\Delta r_{\mathrm d})^3+O((\Delta r_{\mathrm d})^4). \tag*{(35)} 0 = c E + 2 ∂ r 2 f 0 ( r d ) ( Δ r d ) 2 + 6 ∂ r 3 f 0 ( r d ) ( Δ r d ) 3 + O (( Δ r d ) 4 ) . (35)
The absence of a linear term prevents a Maclaurin expansion of the inverse about E = 0 \mathcal{E}=0 E = 0 . The leading quadratic equation has discriminant − 2 c E ∂ r 2 f 0 ( r d ) -2c\mathcal{E}\,\partial_r^2f_0(r_{\mathrm d}) − 2 c E ∂ r 2 f 0 ( r d ) . For sufficiently small E > 0 \mathcal{E}>0 E > 0 , a positive discriminant gives two distinct real roots, whereas a negative one gives none. Thus, under a decrease of the physical mass due to matter tunneling out of the black hole, a double zero at a local maximum of f 0 f_0 f 0 will unfold into two real horizons, whereas a double zero at a local minimum has no nearby real continuation in the same direction. A vanishing second derivative requires the higher-multiplicity analysis developed in Appendix § 11 .
In the splitting sector,
∂ r 2 f 0 ( r d ) < 0. (36) \partial_r^2 f_0(r_{\mathrm d})<0. \tag*{(36)} ∂ r 2 f 0 ( r d ) < 0. (36)
Assuming that f 0 f_0 f 0 is analytic near r d r_{\mathrm d} r d , each branch admits a locally convergent Puiseux expansion [32 , 38 ] of the form
Δ r d ( ϵ ) ( E ) = ∑ n = 1 ∞ a n ( ϵ ) E n / 2 , ϵ = ± 1. (37) \Delta r_{\mathrm d}^{(\epsilon)}(\mathcal{E})=\sum_{n=1}^{\infty}a_n^{(\epsilon)}\mathcal{E}^{n/2},\qquad\epsilon=\pm1. \tag*{(37)} Δ r d ( ϵ ) ( E ) = n = 1 ∑ ∞ a n ( ϵ ) E n /2 , ϵ = ± 1. (37)
The half-integer powers follow from the multiplicity of the initial zero. Applying the LB construction to the ordinary inverse in the signed variable ϵ E \epsilon\sqrt{\mathcal{E}} ϵ E gives the coefficients, with a n ( − ) = ( − 1 ) n a n ( + ) a_n^{(-)}=(-1)^n a_n^{(+)} a n ( − ) = ( − 1 ) n a n ( + ) . The integer powers of E \mathcal{E} E therefore produce a common displacement, while the half-integer powers separate the branches. The coefficient formulas and their extension to arbitrary multiplicity are developed in Appendix § 11 .
The first few terms are
Δ r d ( ϵ ) ( E ) = ϵ 2 c E ∂ r 2 f 0 ( r d ) + c ∂ r 3 f 0 ( r d ) 3 [ ∂ r 2 f 0 ( r d ) ] 2 E + O ( E 3 / 2 ) . (38) \Delta r_{\mathrm d}^{(\epsilon)}(\mathcal{E})=\epsilon\sqrt{\frac{2c\mathcal{E}}{\partial_r^2f_0(r_{\mathrm d})}}+\frac{c\,\partial_r^3f_0(r_{\mathrm d})}{3[\partial_r^2f_0(r_{\mathrm d})]^2}\mathcal{E}+O(\mathcal{E}^{3/2}). \tag*{(38)} Δ r d ( ϵ ) ( E ) = ϵ ∂ r 2 f 0 ( r d ) 2 c E + 3 [ ∂ r 2 f 0 ( r d ) ] 2 c ∂ r 3 f 0 ( r d ) E + O ( E 3/2 ) . (38)
The two signs identify simultaneous zeros of the same final geometry, continuous in radius but nonanalytic in emitted energy. The common term linear in E \mathcal{E} E cancels from their separation,
r d , f ( + ) − r d , f ( − ) = 2 2 c E ∂ r 2 f 0 ( r d ) + O ( E 3 / 2 ) . (39) r_{\mathrm d,f}^{(+)}-r_{\mathrm d,f}^{(-)}=2\sqrt{\frac{2c\mathcal{E}}{\partial_r^2f_0(r_{\mathrm d})}}+O(\mathcal{E}^{3/2}). \tag*{(39)} r d , f ( + ) − r d , f ( − ) = 2 ∂ r 2 f 0 ( r d ) 2 c E + O ( E 3/2 ) . (39)
At these roots, the radial slopes of the final blackening function behave as
∂ r f ( r d , f ( ϵ ) ; M − E , λ A ) = − ϵ − 2 c E ∂ r 2 f 0 ( r d ) + O ( E ) . (40) \partial_r f\left(r_{\mathrm d,f}^{(\epsilon)};\mathcal{M}-\mathcal{E},\lambda^{A}\right)=-\epsilon\sqrt{-2c\mathcal{E}\,\partial_r^2f_0(r_{\mathrm d})}+O(\mathcal{E}). \tag*{(40)} ∂ r f ( r d , f ( ϵ ) ; M − E , λ A ) = − ϵ − 2 c E ∂ r 2 f 0 ( r d ) + O ( E ) . (40)
Therefore, immediately after a sufficiently small emission of energy, the originally degenerate horizon splits into two simple horizons, with positive slope at the smaller radius and negative slope at the larger one, while their surface gravities have the same leading magnitude, proportional to E \sqrt{\mathcal{E}} E .
Charged emission and the critical mass curve
Charge variation changes not only the level of f f f but also its shape. Along the chosen emission path C \mathcal{C} C in Eq. (24) , the mass and charge vary together, with the cumulative emitted charge Q e m ( E ) Q_{\mathrm{em}}(\mathcal{E}) Q em ( E ) specified by the emission channel. To follow a double zero under this variation, suppose that f ( r ; M , Q , λ A ) f(r;\mathcal{M},Q,\lambda^{A}) f ( r ; M , Q , λ A ) is jointly analytic near the initial degenerate horizon and that ∂ M f ≠ 0 \partial_{\mathcal{M}}f \ne0 ∂ M f = 0 there. The local mass function M = m ( r , Q ) \mathcal{M}=m(r,Q) M = m ( r , Q ) introduced around Eq. (29) is then well defined. At the double zero, ∂ r m = 0 \partial_{r}m=0 ∂ r m = 0 and ∂ r 2 m = − ∂ r 2 f / ∂ M f ≠ 0 \partial_{r}^{2}m=-\partial_{r}^{2}f/\partial_{\mathcal{M}}f \ne0 ∂ r 2 m = − ∂ r 2 f / ∂ M f = 0 . The remaining λ A \lambda^{A} λ A are held fixed and suppressed below.
For each running value of the charge Q ′ Q' Q ′ , we first determine the mass and radius at which a double zero continuously connected to the initial one occurs. We denote this radius by r d ( Q ′ ) r_{\mathrm{d}}(Q') r d ( Q ′ ) and the corresponding critical mass by M d ( Q ′ ) \mathcal{M}_{\mathrm{d}}(Q') M d ( Q ′ ) , defined by
M d ( Q ′ ) : = m [ r d ( Q ′ ) , Q ′ ] , ∂ r m [ r d ( Q ′ ) , Q ′ ] = 0. (41) \begin{aligned}
\mathcal{M}_{\mathrm{d}}(Q') &:= m[r_{\mathrm{d}}(Q'),Q'],\\
\partial_{r}m[r_{\mathrm{d}}(Q'),Q'] &= 0. \tag*{(41)}
\end{aligned} M d ( Q ′ ) ∂ r m [ r d ( Q ′ ) , Q ′ ] := m [ r d ( Q ′ ) , Q ′ ] , = 0. (41)
Since ∂ r 2 m ≠ 0 \partial_{r}^{2}m \ne0 ∂ r 2 m = 0 at the initial double zero, these functions are analytic near Q Q Q , with r d ( Q ) = r d r_{\mathrm{d}}(Q)=r_{\mathrm{d}} r d ( Q ) = r d and M d ( Q ) = M \mathcal{M}_{\mathrm{d}}(Q)=\mathcal{M} M d ( Q ) = M . The emission path generally leaves this critical curve. At the emission endpoint, where Q ′ ( E ) = Q − Q e m Q'(\mathcal{E})=Q-Q_{\mathrm{em}} Q ′ ( E ) = Q − Q em , define the signed mass difference
Δ M d : = M d ( Q − Q e m ) − ( M − E ) . (42) \Delta\mathcal{M}_{\mathrm{d}}:=\mathcal{M}_{\mathrm{d}}(Q-Q_{\mathrm{em}})-(\mathcal{M}-\mathcal{E}). \tag*{(42)} Δ M d := M d ( Q − Q em ) − ( M − E ) . (42)
It vanishes on the critical curve and is positive when the actual mass lies below it. Writing r = r d [ Q ′ ( E ) ] + Δ r d r=r_{\mathrm{d}}[Q'(\mathcal{E})]+\Delta r_{\mathrm{d}} r = r d [ Q ′ ( E )] + Δ r d , with Δ r d \Delta r_{\mathrm{d}} Δ r d now measured from the critical radius at the final charge, the horizon condition M − E = m [ r , Q ′ ( E ) ] \mathcal{M}-\mathcal{E}=m[r,Q'(\mathcal{E})] M − E = m [ r , Q ′ ( E )] becomes
0 = Δ M d + ∂ r 2 m 2 ( Δ r d ) 2 + ∂ r 3 m 6 ( Δ r d ) 3 + O ( ( Δ r d ) 4 ) . (43) 0=\Delta\mathcal{M}_{\mathrm{d}}+\frac{\partial_{r}^{2}m}{2}(\Delta r_{\mathrm{d}})^{2}+\frac{\partial_{r}^{3}m}{6}(\Delta r_{\mathrm{d}})^{3}+O((\Delta r_{\mathrm{d}})^{4}). \tag*{(43)} 0 = Δ M d + 2 ∂ r 2 m ( Δ r d ) 2 + 6 ∂ r 3 m ( Δ r d ) 3 + O (( Δ r d ) 4 ) . (43)
All radial derivatives here and in the following local expansion are evaluated at
[ r d [ Q ′ ( E ) ] , Q ′ ( E ) ] . [r_{\mathrm{d}}[Q'(\mathcal{E})],Q'(\mathcal{E})]. [ r d [ Q ′ ( E )] , Q ′ ( E )] .
The leading quadratic discriminant is therefore
− 2 ( ∂ r 2 m ) Δ M d . (44) -2(\partial_{r}^{2}m)\Delta\mathcal{M}_{\mathrm{d}}. \tag*{(44)} − 2 ( ∂ r 2 m ) Δ M d . (44)
For sufficiently small nonzero Δ M d \Delta\mathcal{M}_{\mathrm{d}} Δ M d , a positive discriminant gives two nearby real roots, whereas a negative one gives none. At Δ M d = 0 \Delta\mathcal{M}_{\mathrm{d}}=0 Δ M d = 0 , the double zero persists at r d [ Q ′ ( E ) ] r_{\mathrm{d}}[Q'(\mathcal{E})] r d [ Q ′ ( E )] . On the real-root side, the first terms are
r d , f ( ϵ ) = r d [ Q ′ ( E ) ] + ϵ − 2 Δ M d ∂ r 2 m + ∂ r 3 m 3 ( ∂ r 2 m ) 2 Δ M d + O ( ∣ Δ M d ∣ 3 / 2 ) , (45) r_{\mathrm{d},f}^{(\epsilon)}=r_{\mathrm{d}}[Q'(\mathcal{E})]+\epsilon\sqrt{-\frac{2\Delta\mathcal{M}_{\mathrm{d}}}{\partial_{r}^{2}m}}+\frac{\partial_{r}^{3}m}{3(\partial_{r}^{2}m)^{2}}\Delta\mathcal{M}_{\mathrm{d}}+O(|\Delta\mathcal{M}_{\mathrm{d}}|^{3/2}), \tag*{(45)} r d , f ( ϵ ) = r d [ Q ′ ( E )] + ϵ − ∂ r 2 m 2Δ M d + 3 ( ∂ r 2 m ) 2 ∂ r 3 m Δ M d + O ( ∣Δ M d ∣ 3/2 ) , (45)
with ϵ = ± 1 \epsilon=\pm1 ϵ = ± 1 . Charge variation thus moves the critical radius about which the two branches separate and changes their inverse coefficients. Their separation begins at 2 − 2 Δ M d / ∂ r 2 m 2\sqrt{-2\Delta\mathcal{M}_{\mathrm{d}}/\partial_{r}^{2}m} 2 − 2Δ M d / ∂ r 2 m . At a zero, ∂ r f = − ( ∂ M f ) ∂ r m \partial_{r}f=-(\partial_{\mathcal{M}}f)\partial_{r}m ∂ r f = − ( ∂ M f ) ∂ r m , hence the two slopes have opposite signs and the same leading magnitude, proportional to ∣ Δ M d ∣ 1 / 2 |\Delta\mathcal{M}_{\mathrm{d}}|^{1/2} ∣Δ M d ∣ 1/2 . For the affine family with c > 0 c>0 c > 0 and ∂ r 2 m < 0 \partial_{r}^{2}m<0 ∂ r 2 m < 0 , the smaller radius branch has positive slope, just as in the neutral unfolding. For the internal pair in Figure 1 (b), the signs ϵ = − 1 \epsilon=-1 ϵ = − 1 and ϵ = + 1 \epsilon=+1 ϵ = + 1 identify r − r_{-} r − and r 0 r_{0} r 0 , respectively.
At each fixed charge, the inversion therefore follows the neutral construction, with the displacement measured from r d [ Q ′ ( E ) ] r_{\mathrm d}[Q'(\mathcal{E})] r d [ Q ′ ( E )] . Its coefficients and the derivatives of the critical curve are collected in Appendix § 11 .
To express this unfolding along C \mathcal{C} C , suppose that its charge profile is analytic near E ′ = 0 \mathcal{E}'=0 E ′ = 0 . The mass difference along the path has the expansion
Δ M d ( E ) = ( 1 − η d M d d Q ) E + O ( E 2 ) , (46) \Delta\mathcal{M}_{\mathrm d}(\mathcal{E})=\left(1-\eta\frac{d\mathcal{M}_{\mathrm d}}{dQ}\right)\mathcal{E}+\mathcal{O}(\mathcal{E}^{2}), \tag*{(46)} Δ M d ( E ) = ( 1 − η d Q d M d ) E + O ( E 2 ) , (46)
Here η = ( d Q e m / d E ) ∣ 0 \eta=(dQ_{\mathrm{em}}/d\mathcal{E})|_0 η = ( d Q em / d E ) ∣ 0 , and the critical derivatives are evaluated at the initial state. Three local cases can be distinguished:
If ( 1 − η d M d / d Q ) ∂ r 2 m < 0 (1-\eta\,d\mathcal{M}_{\mathrm d}/dQ)\partial_r^2m<0 ( 1 − η d M d / d Q ) ∂ r 2 m < 0 , the path enters the side with two real roots. Their separation is proportional to E 1 / 2 \mathcal{E}^{1/2} E 1/2 .
If 1 − η d M d / d Q = 0 1-\eta\,d\mathcal{M}_{\mathrm d}/dQ=0 1 − η d M d / d Q = 0 , the path is tangent to the critical curve. The first nonzero higher-order mass difference decides whether a real pair is present. If it occurs at order p p p and has the sign required by Eq. (44) , the separation is proportional to E p / 2 \mathcal{E}^{p/2} E p /2 ; a quadratic departure gives a separation of order E \mathcal{E} E .
If ( 1 − η d M d / d Q ) ∂ r 2 m > 0 (1-\eta\,d\mathcal{M}_{\mathrm d}/dQ)\partial_r^2m>0 ( 1 − η d M d / d Q ) ∂ r 2 m > 0 , the nearby real pair is absent immediately after emission.
Initial tangency alone does not ensure that the double horizon persists. The degeneracy is preserved only if M ′ = M d ( Q ′ ) \mathcal{M}'=\mathcal{M}_{\mathrm d}(Q') M ′ = M d ( Q ′ ) throughout the path. The higher coefficients are given in Appendix § 11 .
When the charge derivative of the mass function satisfies ∂ Q m ( r , Q ) = Φ ( r ) \partial_Qm(r,Q)=\Phi(r) ∂ Q m ( r , Q ) = Φ ( r ) in the potential convention of § 2.4 , the critical slope equals the electric potential at the degenerate horizon,
d M d d Q = Φ d , Φ d : = Φ [ r d ( Q ) ] . (47) \frac{d\mathcal{M}_{\mathrm d}}{dQ}=\Phi_{\mathrm d},\qquad\Phi_{\mathrm d}:=\Phi[r_{\mathrm d}(Q)]. \tag*{(47)} d Q d M d = Φ d , Φ d := Φ [ r d ( Q )] . (47)
Then the initial departure from the critical curve is governed by 1 − η Φ d 1-\eta\Phi_{\mathrm d} 1 − η Φ d , whereas ordinary emission through the simple event horizon requires 1 − η Φ + > 0 1-\eta\Phi_+>0 1 − η Φ + > 0 . Since the two potentials are evaluated at different radii, this condition alone does not determine whether the internal pair splits or disappears.
From splitting to outer coalescence
After the internal double horizon splits, the larger of the two resulting roots can merge with the event horizon, leaving the innermost horizon simple, as illustrated in Figure 1 (b)-(Figure 1 c). The splitting and outer merger occur at different extrema of the stationary mass–radius relation.
For the neutral affine family, suppose that m ( r ) m(r) m ( r ) has an internal local maximum at r d r_{\mathrm d} r d and an adjacent outer local minimum at r e > r d r_{\mathrm e}>r_{\mathrm d} r e > r d . Denote the two critical masses by M i \mathcal{M}_{\mathrm i} M i and M e \mathcal{M}_{\mathrm e} M e , and their positive difference by the critical energy E c \mathcal{E}_{\mathrm c} E c
M i : = m ( r d ) , M e : = m ( r e ) , E c = M i − M e > 0. (48) \mathcal{M}_{\mathrm i}:=m(r_{\mathrm d}),\qquad\mathcal{M}_{\mathrm e}:=m(r_{\mathrm e}),\qquad\mathcal{E}_{\mathrm c}=\mathcal{M}_{\mathrm i}-\mathcal{M}_{\mathrm e}>0. \tag*{(48)} M i := m ( r d ) , M e := m ( r e ) , E c = M i − M e > 0. (48)
Taking M = M i \mathcal{M}=\mathcal{M}_{\mathrm i} M = M i as the initial mass, assume that the double internal horizon coexists with a separate simple event horizon at r + > r e r_+>r_{\mathrm e} r + > r e , and that these are the only positive zeros. Neutral emission lowers the mass to M ′ = M i − E ′ \mathcal{M}'=\mathcal{M}_{\mathrm i}-\mathcal{E}' M ′ = M i − E ′ , initially splitting the internal double zero into two simple horizons. Provided these three branches persist without other horizon transitions up to the critical energy, the larger of the internal pair approaches the event horizon until they merge at E ′ = E c \mathcal{E}' = \mathcal{E}_c E ′ = E c , while the innermost horizon remains simple.
Let f c ( r ) : = f 0 ( r ) + c E c f_c(r) := f_0(r) + c\mathcal{E}_c f c ( r ) := f 0 ( r ) + c E c denote the limiting blackening function. The outer coalescence occurs at
f c ( r e ) = ∂ r f c ( r e ) = 0 , ∂ r 2 f c ( r e ) > 0. (49) f_c(r_e) = \partial_r f_c(r_e) = 0,\qquad\partial_r^2 f_c(r_e) > 0. \tag*{(49)} f c ( r e ) = ∂ r f c ( r e ) = 0 , ∂ r 2 f c ( r e ) > 0. (49)
For E < E c \mathcal{E} < \mathcal{E}_c E < E c , the horizon equation is f c ( r ) − c ( E c − E ) = 0 f_c(r) - c(\mathcal{E}_c - \mathcal{E}) = 0 f c ( r ) − c ( E c − E ) = 0 . The outer member of the pair obeys
R + ( E ) = r e + 2 c ∂ r 2 f c ( r e ) ( E c − E ) 1 / 2 + O ( E c − E ) . (50) R_+(\mathcal{E}) = r_e + \sqrt{\frac{2c}{\partial_r^2 f_c(r_e)}}(\mathcal{E}_c - \mathcal{E})^{1/2} + \mathcal{O}(\mathcal{E}_c - \mathcal{E}). \tag*{(50)} R + ( E ) = r e + ∂ r 2 f c ( r e ) 2 c ( E c − E ) 1/2 + O ( E c − E ) . (50)
The other member has the opposite leading radial displacement. Their separation vanishes as ( E c − E ) 1 / 2 (\mathcal{E}_c - \mathcal{E})^{1/2} ( E c − E ) 1/2 , and the instantaneous Hawking temperature behaves as
T H ( E ) = 2 c ∂ r 2 f c ( r e ) 4 π ( E c − E ) 1 / 2 + O ( E c − E ) . (51) T_{\mathrm{H}}(\mathcal{E}) = \frac{\sqrt{2c}\,\partial_r^2 f_c(r_e)}{4\pi}(\mathcal{E}_c - \mathcal{E})^{1/2} + \mathcal{O}(\mathcal{E}_c - \mathcal{E}). \tag*{(51)} T H ( E ) = 4 π 2 c ∂ r 2 f c ( r e ) ( E c − E ) 1/2 + O ( E c − E ) . (51)
The inverse temperature is integrable at this endpoint. Using again d S / d r + = 4 π / c dS/dr_+ = 4\pi/c d S / d r + = 4 π / c for the affine family gives
Δ S ( E ) = Δ S ( E c ) + 4 π 2 c ∂ r 2 f c ( r e ) ( E c − E ) 1 / 2 + O ( E c − E ) . (52) \Delta S(\mathcal{E}) = \Delta S(\mathcal{E}_c) + 4\pi\sqrt{\frac{2}{c\,\partial_r^2 f_c(r_e)}}(\mathcal{E}_c - \mathcal{E})^{1/2} + \mathcal{O}(\mathcal{E}_c - \mathcal{E}). \tag*{(52)} Δ S ( E ) = Δ S ( E c ) + 4 π c ∂ r 2 f c ( r e ) 2 ( E c − E ) 1/2 + O ( E c − E ) . (52)
Here Δ S ( E c ) \Delta S(\mathcal{E}_c) Δ S ( E c ) denotes the limit from simple outer horizons. The next radial correction is given in Appendix § 11 .§ 4 .
For charged emission, the outer critical level becomes a curve M e ( Q ′ ) = m [ r e ( Q ′ ) , Q ′ ] \mathcal{M}_e(\mathcal{Q}') = \mathfrak{m}[r_e(\mathcal{Q}'),\mathcal{Q}'] M e ( Q ′ ) = m [ r e ( Q ′ ) , Q ′ ] , defined by the same stationary point construction, now at the mass minimum. In the affine charged family with c > 0 c > 0 c > 0 , the outer pair exists on the side M ′ > M e ( Q ′ ) \mathcal{M}' > \mathcal{M}_e(\mathcal{Q}') M ′ > M e ( Q ′ ) . Along C \mathcal{C} C , the mass difference appropriate to this branch is
Δ M e ( E ′ ) : = M e [ Q − Q e m ( E ′ ) ] − M + E ′ . (53) \Delta\mathcal{M}_e(\mathcal{E}') := \mathcal{M}_e[\mathcal{Q}-\mathcal{Q}_{\mathrm{em}}(\mathcal{E}')] - \mathcal{M}+\mathcal{E}'. \tag*{(53)} Δ M e ( E ′ ) := M e [ Q − Q em ( E ′ )] − M + E ′ . (53)
It is negative before coalescence and vanishes when the path reaches the outer critical curve. Applying Eq. (45) at the minimum gives a horizon separation of order ∣ Δ M e ∣ 1 / 2 |\Delta\mathcal{M}_e|^{1/2} ∣Δ M e ∣ 1/2 . The temperature has the same leading order at a regular double outer zero. The relation between this mass difference and the remaining emitted energy depends on the channel: a transverse approach gives the square-root dependence found above, whereas a tangent approach can change that power.
The charged limiting action is obtained from the full residue in Eq. (28) , retaining the gauge electric work factor 1 − Φ + ( d Q e m / d E ′ ) 1-\Phi_+(d\mathcal{Q}_{\mathrm{em}}/d\mathcal{E}') 1 − Φ + ( d Q em / d E ′ ) . For either channel, a finite limiting entropy gives a finite leading tunneling exponent. This limit is taken through simple outer horizons. Radiation from an initially degenerate geometry requires a separate treatment [39 ] .
The nonlinearly charged BTZ family examined in § 4 provides an explicit realization of these critical levels and the associated horizon rearrangement.
Finite-energy emission from a nonlinearly charged BTZ black hole
The exact solution presented in Ref. [19 ] is an extension of the BTZ geometry [40 ] through nonlinear electrodynamics. Finite-energy tunneling and the near-extremal regime were studied for the standard charged BTZ black hole in Ref. [17 ] . Here the nonlinear matter permits a degenerate inner horizon to coexist with a simple event horizon, providing a setting for the distinct transitions described above. This broader classical family contains, as a particular member, the quantum-backreacted BTZ geometry of New Massive Gravity (NMG) obtained to leading order in backreaction in Ref. [41 ] , whose horizon emission is examined in § 6 . An explicit blackening function allows the horizons to be followed along neutral and charged emission paths, including the intervals between encounters with the critical curves.
Tunneling probability from a two-horizon configuration
The solution arises in three-dimensional Einstein gravity with a negative cosmological constant Λ = − 1 / ℓ 2 \Lambda=-1/\ell^2 Λ = − 1/ ℓ 2 , coupled to nonlinear electrodynamics. We describe the supporting matter in Plebanski’s H ( P ) \mathcal{H}(P) H ( P ) formulation [42 ] , with a gauge potential A μ A_\mu A μ and an independent 2-form with components P μ ν P^{\mu\nu} P μν . The structural function of the electrodynamics depends on P = 1 4 P μ ν P μ ν P=\frac{1}{4}P_{\mu\nu}P^{\mu\nu} P = 4 1 P μν P μν , and the field strength is F μ ν = 2 ∂ [ μ A ν ] F_{\mu\nu}=2\partial_{[\mu}A_{\nu]} F μν = 2 ∂ [ μ A ν ] . The supporting structural function, first-order action, and field equations are collected in Appendix § 12 .§ 1 . This description provides the matter realization throughout r > 0 r>0 r > 0 . The resulting spacetime is described by the static metric
d s 2 = − f ( r ) d t 2 + d r 2 f ( r ) + r 2 d φ 2 , φ ∼ φ + 2 π , (54) ds^2=-f(r)dt^2+\frac{dr^2}{f(r)}+r^2d\varphi^2,\qquad\varphi\sim\varphi+2\pi, \tag*{(54)} d s 2 = − f ( r ) d t 2 + f ( r ) d r 2 + r 2 d φ 2 , φ ∼ φ + 2 π , (54)
with
f ( r ) = r 2 ℓ 2 − M − 2 2 κ q r γ ln ( β γ r 2 q ) , (55) f(r)=\frac{r^2}{\ell^2}-M-\frac{2\sqrt{2}\kappa q r}{\gamma}\ln\left(\frac{\beta\gamma r}{\sqrt{2q}}\right), \tag*{(55)} f ( r ) = ℓ 2 r 2 − M − γ 2 2 κ q r ln ( 2 q β γ r ) , (55)
A = 2 r γ [ ln ( β γ r 2 q ) − 1 ] d t . (56) A=\frac{\sqrt{2}r}{\gamma}\left[\ln\left(\frac{\beta\gamma r}{\sqrt{2q}}\right)-1\right]dt. \tag*{(56)} A = γ 2 r [ ln ( 2 q β γ r ) − 1 ] d t . (56)
The electric contribution in Eq. (55) contains an additional radial factor multiplying the logarithm compared with the Maxwell-charged BTZ solution [43 ] . Here, κ = 8 π G 3 \kappa=8\pi G_3 κ = 8 π G 3 , M M M and q q q are integration constants, whereas β \beta β and γ \gamma γ are couplings of the nonlinear electromagnetic theory. The real configurations considered below require
β γ q > 0. (57) \frac{\beta\gamma}{q}>0. \tag*{(57)} q β γ > 0. (57)
The multihorizon configurations used here have q / γ > 0 q/\gamma>0 q / γ > 0 and β > 0 \beta>0 β > 0 , where the mass–radius relation can develop both a local maximum and a local minimum. Horizon-adapted representations and their extremal limits are derived in Appendix § 12 .
We shall focus on the critical two-horizon configuration identified in Appendix § 12 , in which a degenerate inner Killing horizon at r = r − r=r_- r = r − coexists with a simple event horizon at r = r + r=r_+ r = r + , with
0 < r − < r + . 0<r_-<r_+. 0 < r − < r + .
At r = r − r=r_- r = r − , the initial blackening function has a local maximum that touches zero, whereas r = r + r=r_+ r = r + is a simple outer zero with positive slope. The corresponding horizon-adapted representation is given by Eq. (D.9), supplemented by the degeneracy condition ∂ r f 0 ( r − ) = 0 \partial_r f_0(r_-)=0 ∂ r f 0 ( r − ) = 0 . During emission we vary M M M and, for charged shells, q q q in the original blackening function, keeping ( κ , β , γ , ℓ ) (\kappa,\beta,\gamma,\ell) ( κ , β , γ , ℓ ) fixed.
We begin with neutral emission from this state and then allow the charge to vary. Throughout the analysis, r + r_+ r + denotes the initial event-horizon radius.
The physical mass, electric charge, and entropy are
M = π M κ , Q = 2 π q , S = 4 π 2 κ r + . (58) \mathcal{M} = \frac{\pi M}{\kappa}, \qquad\mathcal{Q} = 2\pi q, \qquad S = \frac{4\pi^2}{\kappa}r_{+}. \tag*{(58)} M = κ π M , Q = 2 π q , S = κ 4 π 2 r + . (58)
The Hawking temperature and electric potential in the gauge (56) are
T H = ∂ r f ( r + ) 4 π = r + 2 π ℓ 2 − 2 κ 2 π γ [ ln ( β γ r + 2 q ) + 1 ] , (59) T_{\mathrm{H}} = \frac{\partial_r f(r_{+})}{4\pi} = \frac{r_{+}}{2\pi\ell^2} - \frac{\sqrt{2\kappa}}{2\pi\gamma}\left[\ln\left(\frac{\beta\gamma r_{+}}{\sqrt{2q}}\right)+1\right], \tag*{(59)} T H = 4 π ∂ r f ( r + ) = 2 π ℓ 2 r + − 2 π γ 2 κ [ ln ( 2 q β γ r + ) + 1 ] , (59)
Φ e = − A t ( r + ) = 2 r + γ [ 1 − ln ( β γ r + 2 q ) ] . (60) \Phi_{\mathrm{e}} = -A_t(r_{+}) = \frac{\sqrt{2}r_{+}}{\gamma}\left[1-\ln\left(\frac{\beta\gamma r_{+}}{\sqrt{2q}}\right)\right]. \tag*{(60)} Φ e = − A t ( r + ) = γ 2 r + [ 1 − ln ( 2 q β γ r + ) ] . (60)
The instantaneous value of Φ e \Phi_{\mathrm{e}} Φ e is the horizon potential Φ + \Phi_{+} Φ + used in § 2.4 .
The thermodynamic description provided in Appendix § 12 .§ 5 follows the standard approach with cosmological constant and nonlinear electromagnetic couplings fixed. The Hamiltonian variation then identifies M \mathcal{M} M as the internal energy exchanged with the shell. This is the fixed-coupling restriction of the extended phase-space treatment in Ref. [19 ] , where the mass has an enthalpy interpretation. The first law reads
d M = T H d S + Φ e d Q , (61) \mathrm{d}\mathcal{M} = T_{\mathrm{H}}\mathrm{d}S + \Phi_{\mathrm{e}}\mathrm{d}\mathcal{Q}, \tag*{(61)} d M = T H d S + Φ e d Q , (61)
together with the corresponding Smarr relation
M = 1 2 ( T H S + Φ e Q ) . (62) \mathcal{M} = \frac{1}{2}\left(T_{\mathrm{H}}S+\Phi_{\mathrm{e}}\mathcal{Q}\right). \tag*{(62)} M = 2 1 ( T H S + Φ e Q ) . (62)
The horizon equation takes the mass–radius form introduced in § 2.2 and § 2.4 , with the charge argument written as q = Q / ( 2 π ) q=\mathcal{Q}/(2\pi) q = Q / ( 2 π ) and the identification
m ( r , q ) : = π κ [ r 2 ℓ 2 − 2 2 κ q r γ ln ( β γ r 2 q ) ] . (63) m(r,q) := \frac{\pi}{\kappa}\left[\frac{r^2}{\ell^2}-\frac{2\sqrt{2}\kappa q r}{\gamma}\ln\left(\frac{\beta\gamma r}{\sqrt{2q}}\right)\right]. \tag*{(63)} m ( r , q ) := κ π [ ℓ 2 r 2 − γ 2 2 κ q r ln ( 2 q β γ r ) ] . (63)
Equivalently, the blackening function takes the affine form
f ( r ; M , q ) = κ π [ m ( r , q ) − M ] . (64) f(r;\mathcal{M},q) = \frac{\kappa}{\pi}\left[m(r,q)-\mathcal{M}\right]. \tag*{(64)} f ( r ; M , q ) = π κ [ m ( r , q ) − M ] . (64)
For the emission of a neutral shell, ( q , β , γ , ℓ ) (q,\beta,\gamma,\ell) ( q , β , γ , ℓ ) remain fixed and the mass parameters change according to
M ⟼ M − E , M ⟼ M − κ π E . (65) \mathcal{M}\longmapsto\mathcal{M}-\mathcal{E}, \qquad M\longmapsto M-\frac{\kappa}{\pi}\mathcal{E}. \tag*{(65)} M ⟼ M − E , M ⟼ M − π κ E . (65)
The final horizon equation therefore becomes
f ( r ; M − E , q ) = f 0 ( r ) + κ π E = 0. (66) f(r;\mathcal{M}-\mathcal{E},q)=f_0(r)+\frac{\kappa}{\pi}\mathcal{E}=0. \tag*{(66)} f ( r ; M − E , q ) = f 0 ( r ) + π κ E = 0. (66)
For a charged shell, the mass changes as in Eq. (65) , while the integration constant q q q varies along the channel C \mathcal{C} C of § 2.4 . Its running value and the corresponding event horizon satisfy
q ( E ′ ) = q − Q e m ( E ′ ) 2 π , m [ R + ( E ′ ) , q ( E ′ ) ] = M − E ′ , R + ( 0 ) = r + . (67) q(\mathcal{E}')=q-\frac{Q_{\mathrm{em}}(\mathcal{E}')}{2\pi}, \qquad m\left[R_{+}(\mathcal{E}'),q(\mathcal{E}')\right]=\mathcal{M}-\mathcal{E}', \qquad R_{+}(0)=r_{+}. \tag*{(67)} q ( E ′ ) = q − 2 π Q em ( E ′ ) , m [ R + ( E ′ ) , q ( E ′ ) ] = M − E ′ , R + ( 0 ) = r + . (67)
The charge profile is specified by the emission channel, with the running states restricted by Eq. (57) . Charge variation therefore changes the radial profile of f f f in addition to the mass shift in Eq. (66) .
Since the entropy is linear in the event horizon radius and has no explicit charge dependence, its displacement determines the tunneling exponent for either channel while the instantaneous event horizon remains simple,
ln Γ K P W ≃ Δ S = 4 π 2 κ Δ r + ( E ) . (68) \ln\Gamma_{\rm KPW} \simeq\Delta S = \frac{4\pi^{2}}{\kappa}\Delta r_{+}(\mathcal{E}). \tag*{(68)} ln Γ KPW ≃ Δ S = κ 4 π 2 Δ r + ( E ) . (68)
For neutral emission, q ( E ′ ) = q q(\mathcal{E}')=q q ( E ′ ) = q and the local series about the initial simple event horizon gives
ln Γ K P W ≃ 4 π 2 κ ∑ n = 1 ∞ a n E n . (69) \ln\Gamma_{\rm KPW} \simeq\frac{4\pi^{2}}{\kappa}\sum_{n=1}^{\infty}a_n\mathcal{E}^{n}. \tag*{(69)} ln Γ KPW ≃ κ 4 π 2 n = 1 ∑ ∞ a n E n . (69)
Within the convergence domain of the inverse, this series reconstructs the finite-emission exponent. For the initial two-horizon configuration, its coefficients acquire the compact form
a n = ( κ ℓ 2 π ) n r + 1 − 2 n A n ( r + r − ) , (70) a_n=\left(\frac{\kappa\ell^{2}}{\pi}\right)^{n}r_{+}^{1-2n}A_n\left(\frac{r_{+}}{r_{-}}\right), \tag*{(70)} a n = ( π κ ℓ 2 ) n r + 1 − 2 n A n ( r − r + ) , (70)
where the dimensionless function A n A_n A n encodes the degeneracy condition at the internal root. An explicit closed expression for A n A_n A n , together with the first coefficients and their derivation from the LB formula, is presented in Appendix § 13 . For charged emission, the corresponding inverse is obtained from the coefficients in Appendix § 9 , using Eq. (63) at the final charge. The prescribed profile Q e m ( E ) Q_{\rm em}(\mathcal{E}) Q em ( E ) then gives the displacement along C \mathcal{C} C .
The dependence on r + / r − r_{+}/r_{-} r + / r − in Eq. (70) comes from the initial degeneracy condition, which constrains the geometry at the event horizon. The internal roots are not crossed by the outgoing contour. Their rearrangement nevertheless follows from the same change of mass and charge, as we now describe.
From inner horizon fine-structure to extremality
For fixed ( q , β , γ , ℓ ) (q,\beta,\gamma,\ell) ( q , β , γ , ℓ ) , every positive horizon solves the level set equation m ( r , q ) = M \mathfrak{m}(r,q)=\mathcal{M} m ( r , q ) = M , corresponding to an intersection of the mass–radius curve with the horizontal line at mass M \mathcal{M} M . This picture places the initial inner degeneracy and the later outer merger at the two stationary points of the same curve. Its first derivatives are
∂ r m = π κ { 2 r ℓ 2 − 2 2 κ q γ [ ln ( β γ r 2 q ) + 1 ] } , ∂ r 2 m = π κ ( 2 ℓ 2 − 2 2 κ q γ r ) , ∂ r 3 m = 2 2 π q γ r 2 . (71) \begin{aligned}
\partial_r\mathfrak{m} &= \frac{\pi}{\kappa}\left\{\frac{2r}{\ell^{2}}-\frac{2\sqrt{2}\kappa q}{\gamma}\left[\ln\left(\frac{\beta\gamma r}{\sqrt{2}q}\right)+1\right]\right\},\\
\partial_r^{2}\mathfrak{m} &= \frac{\pi}{\kappa}\left(\frac{2}{\ell^{2}}-\frac{2\sqrt{2}\kappa q}{\gamma r}\right),\\
\partial_r^{3}\mathfrak{m} &= \frac{2\sqrt{2}\pi q}{\gamma r^{2}}. \tag*{(71)}
\end{aligned} ∂ r m ∂ r 2 m ∂ r 3 m = κ π { ℓ 2 2 r − γ 2 2 κ q [ ln ( 2 q β γ r ) + 1 ] } , = κ π ( ℓ 2 2 − γ r 2 2 κ q ) , = γ r 2 2 2 π q . (71)
When q / γ < 0 q/\gamma<0 q / γ < 0 , the curve is strictly convex and admits only one positive intersection for M > 0 \mathcal{M}>0 M > 0 . A configuration with several positive horizons instead requires
q γ > 0. \frac{q}{\gamma}>0. γ q > 0.
Together with Eq. (57) , this condition also implies β > 0 \beta> 0 β > 0 .
For κ β ℓ 2 > 1 \kappa\beta\ell^{2} > 1 κ β ℓ 2 > 1 , the stationary condition ∂ r m = 0 \partial_{r}m = 0 ∂ r m = 0 admits two positive solutions. These are expressed through the two real branches of Lambert’s W W W function, defined implicitly by
W ( x ) exp [ W ( x ) ] = x . W(x)\exp[W(x)] = x. W ( x ) exp [ W ( x )] = x .
On the interval − 1 / e < x < 0 -1/e < x < 0 − 1/ e < x < 0 , they correspond to the principal branch W 0 ( x ) W_{0}(x) W 0 ( x ) and the lower branch W − 1 ( x ) W_{-1}(x) W − 1 ( x ) [44 ] . Since their common argument appears repeatedly, we introduce the shorthand
w k : = W k ( − 1 e κ β ℓ 2 ) , k = 0 , − 1. (72) w_{k} := W_{k}\left(-\frac{1}{e\kappa\beta\ell^{2}}\right), \qquad k = 0,-1. \tag*{(72)} w k := W k ( − e κ β ℓ 2 1 ) , k = 0 , − 1. (72)
The two stationary radii can then be written as
r max = − 2 κ q ℓ 2 γ w 0 , (73) r_{\max} = -\frac{\sqrt{2}\kappa q\ell^{2}}{\gamma}w_{0}, \tag*{(73)} r m a x = − γ 2 κ q ℓ 2 w 0 , (73)
r min = − 2 κ q ℓ 2 γ w − 1 . (74) r_{\min} = -\frac{\sqrt{2}\kappa q\ell^{2}}{\gamma}w_{-1}. \tag*{(74)} r m i n = − γ 2 κ q ℓ 2 w − 1 . (74)
The ordering of the two real Lambert branches,
− 1 < w 0 < 0 , w − 1 < − 1 , -1 < w_{0} < 0, \qquad w_{-1} < -1, − 1 < w 0 < 0 , w − 1 < − 1 ,
implies the corresponding ordering of the radii
0 < r max < 2 κ q ℓ 2 γ < r min . 0 < r_{\max} < \frac{\sqrt{2}\kappa q\ell^{2}}{\gamma} < r_{\min}. 0 < r m a x < γ 2 κ q ℓ 2 < r m i n .
These are the local maximum and minimum of m ( r , q ) m(r,q) m ( r , q ) , respectively, as shown in Figure 2 . For the neutral sequence of § 3.3 , they give r d = r max r_{\mathrm{d}} = r_{\max} r d = r m a x and r e = r min r_{\mathrm{e}} = r_{\min} r e = r m i n . The critical masses defined in Eq. (48) therefore evaluate to
Figure 2. Structure of the local mass function for sampling constant values κ = ℓ = γ = q = 1 \kappa= \ell= \gamma= q = 1 κ = ℓ = γ = q = 1 and β = 6 / 5 \beta= 6/5 β = 6/5 . At the upper critical mass M i \mathcal{M}_{\mathrm{i}} M i , the two inner horizons coincide at r − = r 0 = r m a x r_{-} = r_{0} = r_{\mathrm{max}} r − = r 0 = r max , while a separate simple event horizon remains. Intermediate mass levels intersect the curve at three simple horizons. At the lower critical mass M e \mathcal{M}_{\mathrm{e}} M e , the middle and event horizons merge at r 0 = r + = r m i n r_{0} = r_{+} = r_{\mathrm{min}} r 0 = r + = r min , while the innermost root survives and remains simple. The plotted r + r_{+} r + denotes R + ( E ) R_{+}(\mathcal{E}) R + ( E ) along the neutral sequence, with r m i n = r e r_{\mathrm{min}} = r_{\mathrm{e}} r min = r e in the notation of Figure 1 . Neutral emission connects these two critical levels at fixed charge, with E c = M i − M e ≃ 1.856 \mathcal{E}_{\mathrm{c}} = \mathcal{M}_{\mathrm{i}} - \mathcal{M}_{\mathrm{e}} \simeq1.856 E c = M i − M e ≃ 1.856 .
M i = m ( r max , q ) = − 2 π κ q 2 ℓ γ 2 w 0 ( w 0 + 2 ) , (75) \mathcal{M}_{\mathrm{i}} = m(r_{\max},q) = -\frac{2\pi\kappa q^{2}\ell}{\gamma^{2}}w_{0}(w_{0}+2), \tag*{(75)} M i = m ( r m a x , q ) = − γ 2 2 π κ q 2 ℓ w 0 ( w 0 + 2 ) , (75)
M e = m ( r min , q ) = − 2 π κ q 2 ℓ γ 2 w − 1 ( w − 1 + 2 ) . (76) \mathcal{M}_{\mathrm{e}} = m(r_{\min},q) = -\frac{2\pi\kappa q^{2}\ell}{\gamma^{2}}w_{-1}(w_{-1}+2). \tag*{(76)} M e = m ( r m i n , q ) = − γ 2 2 π κ q 2 ℓ w − 1 ( w − 1 + 2 ) . (76)
These values are intrinsic thresholds of the black hole family. At fixed q q q , the number and multiplicity of the positive roots change when the mass crosses either of them. The subscripts indicate the causal character of the corresponding degeneracy: M i \mathcal{M}_{\mathrm{i}} M i marks the merger of the two inner roots, whereas M e \mathcal{M}_{\mathrm{e}} M e marks the coalescence of the middle and event horizons.
The upper critical mass is positive for every κ β ℓ 2 > 1 \kappa\beta\ell^{2} > 1 κ β ℓ 2 > 1 , while the lower one is positive only within
1 < κ β ℓ 2 < e 2 . (77) 1 < \kappa\beta\ell^{2} < \frac{e}{2}. \tag*{(77)} 1 < κ β ℓ 2 < 2 e . (77)
For the multihorizon emission paths considered below, we restrict to this interval, for which
0 < M e < M i . 0 < \mathcal{M}_{\mathrm{e}} < \mathcal{M}_{\mathrm{i}}. 0 < M e < M i .
This is the sector in which the outer coalescence has positive mass and leaves a separate positive innermost horizon. Since m ( 0 + , q ) = 0 m(0^{+},q) = 0 m ( 0 + , q ) = 0 and m ( r , q ) → + ∞ m(r,q) \to+\infty m ( r , q ) → + ∞ at large radius, the three monotonic portions separated by the maximum and minimum intersect every mass level between M e \mathcal{M}_{\mathrm{e}} M e and M i \mathcal{M}_{\mathrm{i}} M i exactly once.
Starting from the upper critical configuration, M = M i \mathcal{M} = \mathcal{M}_{\mathrm{i}} M = M i , neutral emission at fixed q q q lowers the mass to M ′ ( E ) = M i − E \mathcal{M}'(\mathcal{E}) = \mathcal{M}_{\mathrm{i}} - \mathcal{E} M ′ ( E ) = M i − E . It reaches the lower critical level at the neutral critical energy of Eq. (48) ,
E c = M i − M e . (78) \mathcal{E}_{\mathrm{c}} = \mathcal{M}_{\mathrm{i}} - \mathcal{M}_{\mathrm{e}}. \tag*{(78)} E c = M i − M e . (78)
The intersections in Figure 2 show the resulting sequence from inner splitting to outer coalescence. Above the initial mass level M i \mathcal{M}_{\mathrm{i}} M i , the mass curve admits only one positive intersection, corresponding to a simple event horizon. The two inner branches have the common initial limit r − ( 0 ) = r 0 ( 0 ) = r m a x r_{-}(0) = r_{0}(0) = r_{\mathrm{max}} r − ( 0 ) = r 0 ( 0 ) = r max , while r − ( E ) r_{-}(\mathcal{E}) r − ( E ) denotes the smaller branch after splitting. The initial configuration satisfies
f ( r m a x ) = ∂ r f ( r m a x ) = 0 , ∂ r 2 f ( r m a x ) < 0. (79) f(r_{\mathrm{max}}) = \partial_{r} f(r_{\mathrm{max}}) = 0,\qquad\partial_{r}^{2} f(r_{\mathrm{max}}) < 0. \tag*{(79)} f ( r max ) = ∂ r f ( r max ) = 0 , ∂ r 2 f ( r max ) < 0. (79)
A separate simple event horizon r + > r m i n r_{+} > r_{\mathrm{min}} r + > r min remains, with finite positive temperature. The initial double zero consequently lies in the splitting sector (36) .
Any positive neutral emission lowers the mass below M i \mathcal{M}_{\mathrm{i}} M i and immediately unfolds the double zero into two simple roots. Throughout 0 < E < E c 0 < \mathcal{E} < \mathcal{E}_{\mathrm{c}} 0 < E < E c , the geometry therefore possesses three distinct positive horizons,
0 < r − ( E ) < r 0 ( E ) < R + ( E ) . 0 < r_{-}(\mathcal{E}) < r_{0}(\mathcal{E}) < R_{+}(\mathcal{E}). 0 < r − ( E ) < r 0 ( E ) < R + ( E ) .
Writing r a = r a ( E ) r_{a} = r_{a}(\mathcal{E}) r a = r a ( E ) for a ∈ { − , 0 , + } a \in\{-, 0, +\} a ∈ { − , 0 , + } , with r + ( E ) = R + ( E ) r_{+}(\mathcal{E}) = R_{+}(\mathcal{E}) r + ( E ) = R + ( E ) , the metric function at mass M − E \mathcal{M} - \mathcal{E} M − E and charge q q q is negative for 0 < r < r − 0 < r < r_{-} 0 < r < r − and changes sign across each of the three simple horizons. The outermost root R + ( E ) R_{+}(\mathcal{E}) R + ( E ) corresponds to the event horizon, while the intermediate root r 0 ( E ) r_{0}(\mathcal{E}) r 0 ( E ) defines the Cauchy horizon associated with the exterior region. The smallest root r − ( E ) r_{-}(\mathcal{E}) r − ( E ) corresponds to an additional inner Killing horizon. At fixed charge, their motion under emission obeys
d r a d E = − κ π ∂ r f ( r a ) . (80) \frac{d r_a}{d\mathcal{E}}=-\frac{\kappa}{\pi\partial_r f(r_a)}. \tag*{(80)} d E d r a = − π ∂ r f ( r a ) κ . (80)
The positive slopes at the innermost and event horizons make both contract, while the negative slope at the middle horizon makes it expand as energy is radiated away. The middle and event horizons consequently approach one another until they reach the minimum of the mass–radius curve.
At the critical energy, the middle and event horizons merge at the local minimum,
r 0 ( E c ) = R + ( E c ) = r min , r_0(\mathcal{E}_c)=R_{+}(\mathcal{E}_c)=r_{\min}, r 0 ( E c ) = R + ( E c ) = r m i n ,
with
f ( r min ) = ∂ r f ( r min ) = 0 , ∂ r 2 f ( r min ) > 0. (81) f(r_{\min})=\partial_r f(r_{\min})=0,\qquad\partial_r^2 f(r_{\min})>0. \tag*{(81)} f ( r m i n ) = ∂ r f ( r m i n ) = 0 , ∂ r 2 f ( r m i n ) > 0. (81)
The resulting event horizon is degenerate and has zero Hawking temperature, while the innermost simple root survives. If the family is formally continued to positive masses below M e \mathcal{M}_e M e , the outer pair is absent and the continuation of the smallest root becomes the sole event horizon.
The same structure can be represented globally in the charge–mass state space. At fixed ( ℓ , β , γ ) (\ell,\beta,\gamma) ( ℓ , β , γ ) , the local mass function can be checked to be homogeneous of second degree according to
m ( λ r , λ q ) = λ 2 m ( r , q ) , λ > 0. (82) m(\lambda r,\lambda q)=\lambda^2m(r,q),\qquad\lambda>0. \tag*{(82)} m ( λ r , λ q ) = λ 2 m ( r , q ) , λ > 0. (82)
In particular, for q ′ = Q ′ / ( 2 π ) q'=Q'/(2\pi) q ′ = Q ′ / ( 2 π ) , any positive horizon satisfies
r a ( M ′ , q ′ ) = q ′ q r a ( M ′ q 2 q ′ 2 , q ) , q ′ q > 0. (83) r_a(\mathcal{M}',q')=\frac{q'}{q}r_a\left(\mathcal{M}'\frac{q^2}{q'^2},q\right),\qquad\frac{q'}{q}>0. \tag*{(83)} r a ( M ′ , q ′ ) = q q ′ r a ( M ′ q ′2 q 2 , q ) , q q ′ > 0. (83)
Charge variation thus changes the radial scale, while the ratio M ′ / Q ′ 2 \mathcal{M}'/Q'^2 M ′ / Q ′2 determines the positions of the scaled horizons. This scaling also relates the tunneling exponents of corresponding channels. After scaling the initial state as ( M , Q ) ↦ ( λ 2 M , λ Q ) (\mathcal{M},Q)\mapsto(\lambda^2\mathcal{M},\lambda Q) ( M , Q ) ↦ ( λ 2 M , λ Q ) and choosing the corresponding charge profile to satisfy Q e m , λ ( λ 2 E ) = λ Q e m ( E ) Q_{\mathrm{em},\lambda}(\lambda^2\mathcal{E})=\lambda Q_{\mathrm{em}}(\mathcal{E}) Q em , λ ( λ 2 E ) = λ Q em ( E ) , Eq. (68) gives
Δ S λ ( λ 2 E ) = λ Δ S ( E ) . (84) \Delta S_\lambda(\lambda^2\mathcal{E})=\lambda\Delta S(\mathcal{E}). \tag*{(84)} Δ S λ ( λ 2 E ) = λ Δ S ( E ) . (84)
Hence Δ S / S ( 0 ) \Delta S/S(0) Δ S / S ( 0 ) has the same dependence on E / M \mathcal{E}/\mathcal{M} E / M for these charge-scaled states, while their absolute exponents, and therefore their tunneling factors, differ. Since Q = 2 π q Q=2\pi q Q = 2 π q , Eqs. (75) and (76) extend to the two critical curves
M i ( Q ) = − κ ℓ 2 Q 2 2 π γ 2 w 0 ( w 0 + 2 ) , (85) \mathcal{M}_i(Q)=-\frac{\kappa\ell^2Q^2}{2\pi\gamma^2}w_0(w_0+2), \tag*{(85)} M i ( Q ) = − 2 π γ 2 κ ℓ 2 Q 2 w 0 ( w 0 + 2 ) , (85)
M e ( Q ) = − κ ℓ 2 Q 2 2 π γ 2 w − 1 ( w − 1 + 2 ) . (86) \mathcal{M}_e(Q)=-\frac{\kappa\ell^2Q^2}{2\pi\gamma^2}w_{-1}(w_{-1}+2). \tag*{(86)} M e ( Q ) = − 2 π γ 2 κ ℓ 2 Q 2 w − 1 ( w − 1 + 2 ) . (86)
In the sign sector Q / γ > 0 Q/\gamma>0 Q / γ > 0 , they delimit the strip
M e ( Q ) < M < M i ( Q ) , (87) \mathcal{M}_e(Q)<\mathcal{M}<\mathcal{M}_i(Q), \tag*{(87)} M e ( Q ) < M < M i ( Q ) , (87)
occupied by the three-horizon geometries. Its upper boundary contains a degenerate inner horizon together with a simple event horizon, whereas its lower boundary contains a simple innermost root and a degenerate event horizon.
The two-boundary structure itself collapses at the limiting coupling κ β ℓ 2 = 1 \kappa\beta\ell^2=1 κ β ℓ 2 = 1 . In this particular choice, both negative branches of Lambert’s function coincide w 0 = w − 1 = − 1 w_0=w_{-1}=-1 w 0 = w − 1 = − 1 , and
r max = r min = r t : = 2 κ q ℓ 2 γ , M i = M e = M t : = 2 π κ q 2 ℓ γ 2 . (88) r_{\max}=r_{\min}=r_{\mathrm{t}}:=\frac{\sqrt{2}\kappa q\ell^2}{\gamma},
\qquad
\mathcal{M}_{\mathrm{i}}=\mathcal{M}_{\mathrm{e}}=\mathcal{M}_{\mathrm{t}}:=\frac{2\pi\kappa q^2\ell}{\gamma^2}. \tag*{(88)} r m a x = r m i n = r t := γ 2 κ q ℓ 2 , M i = M e = M t := γ 2 2 π κ q 2 ℓ . (88)
The local maximum and minimum have then merged into a stationary inflection point, with f ( r t ) = ∂ r f ( r t ) = ∂ r 2 f ( r t ) = 0 f(r_{\mathrm{t}})=\partial_r f(r_{\mathrm{t}})=\partial_r^2 f(r_{\mathrm{t}})=0 f ( r t ) = ∂ r f ( r t ) = ∂ r 2 f ( r t ) = 0 and ∂ r 3 f ( r t ) > 0 \partial_r^3 f(r_{\mathrm{t}})>0 ∂ r 3 f ( r t ) > 0 , producing a horizon of multiplicity three. Its local response belongs to the higher-multiplicity hierarchy developed in Appendix § 11 . At fixed couplings in Eq. (77) , varying the charge scales both radii by the same positive factor and cannot bring them together at nonzero charge. The triple zero is therefore a coupling-space limit, distinct from the emission paths considered here.
Each charge profile defines a horizon history , namely the sequence of stationary horizon configurations as E \mathcal{E} E increases along C \mathcal{C} C . To follow it, we evaluate the critical masses at the running charge. With M = M i ( Q ) \mathcal{M}=\mathcal{M}_{\mathrm{i}}(Q) M = M i ( Q ) as the initial mass, their signed mass differences are explicitly
Δ M l ( E ) = M l ( Q ) [ 1 − Q e m ( E ) Q ] 2 − M i ( Q ) + E , l ∈ { i , e } . (89) \Delta\mathcal{M}_{l}(\mathcal{E})=\mathcal{M}_{l}(Q)\left[1-\frac{Q_{\mathrm{em}}(\mathcal{E})}{Q}\right]^2-\mathcal{M}_{\mathrm{i}}(Q)+\mathcal{E},
\qquad l\in\{\mathrm{i},\mathrm{e}\}. \tag*{(89)} Δ M l ( E ) = M l ( Q ) [ 1 − Q Q em ( E ) ] 2 − M i ( Q ) + E , l ∈ { i , e } . (89)
Here the charge keeps the sign of the chosen sector, Q ′ ( E ) = Q − Q e m ( E ) Q'(\mathcal{E})=Q-Q_{\mathrm{em}}(\mathcal{E}) Q ′ ( E ) = Q − Q em ( E ) with Q ′ ( E ) / γ > 0 Q'(\mathcal{E})/\gamma>0 Q ′ ( E ) / γ > 0 . The three-horizon geometry occurs precisely when Δ M i > 0 \Delta\mathcal{M}_{\mathrm{i}}>0 Δ M i > 0 and Δ M e < 0 \Delta\mathcal{M}_{\mathrm{e}}<0 Δ M e < 0 . The first inequality places the state below the inner critical curve, resolving the inner double zero. The second keeps it above the outer critical curve, retaining the middle and event horizons. At either equality the corresponding pair coincides. Their local separation is given by the charged inverse of § 3.2 , evaluated about the critical radius at the running charge.
Differentiating the horizon condition along any differentiable charge profile, with ∂ q m = 2 π Φ \partial_q m=2\pi\Phi ∂ q m = 2 π Φ , gives
d r a d E = − 1 − Φ ( r a ) d Q e m / d E ∂ r m [ r a , q ( E ) ] . (90) \frac{d r_a}{d\mathcal{E}}=-\frac{1-\Phi(r_a)\,dQ_{\mathrm{em}}/d\mathcal{E}}{\partial_r m[r_a,q(\mathcal{E})]}. \tag*{(90)} d E d r a = − ∂ r m [ r a , q ( E )] 1 − Φ ( r a ) d Q em / d E . (90)
with Φ ( r ) = − A t ( r ) \Phi(r)=-A_t(r) Φ ( r ) = − A t ( r ) evaluated at the running charge and the formula applied to simple roots. For linear charge extraction, take Q e m ( E ) = η E Q_{\mathrm{em}}(\mathcal{E})=\eta\mathcal{E} Q em ( E ) = η E , with constant η \eta η . The inner mass difference is then exactly
Δ M i ( E ) = ( 1 − 2 η M i Q ) E + η 2 M i Q 2 E 2 , (91) \Delta\mathcal{M}_{\mathrm{i}}(\mathcal{E})=\left(1-\frac{2\eta\mathcal{M}_{\mathrm{i}}}{Q}\right)\mathcal{E}+\frac{\eta^2\mathcal{M}_{\mathrm{i}}}{Q^2}\mathcal{E}^2, \tag*{(91)} Δ M i ( E ) = ( 1 − Q 2 η M i ) E + Q 2 η 2 M i E 2 , (91)
Since the inner critical point is a maximum, 1 − 2 η M i / Q > 0 1-2\eta\mathcal{M}_{\mathrm{i}}/Q>0 1 − 2 η M i / Q > 0 realizes case (i), whereas a negative value realizes case (iii). At η = Q / ( 2 M i ) \eta=Q/(2\mathcal{M}_{\mathrm{i}}) η = Q / ( 2 M i ) , the positive quadratic term resolves the tangent case (ii) in favor of a real pair, with separation proportional to E \mathcal{E} E .
Each of these cases can occur under ordinary emission through the event horizon. At the tangent value, homogeneity gives
1 − η Φ + = r + ∂ r m ( r + , q ) 2 M i > 0 , 1-\eta\Phi_+=\frac{r_+\partial_r m(r_+,q)}{2\mathcal{M}_{\mathrm{i}}}>0, 1 − η Φ + = 2 M i r + ∂ r m ( r + , q ) > 0 ,
at the initial state, so a neighborhood on either side of this value also obeys the condition ensuring a positive horizon energy. For constant η \eta η , an initially positive factor remains positive along the simple outer horizon1 .
For 1 / 2 < η M i / Q < 1 1/2<\eta\mathcal{M}_i/Q<1 1/2 < η M i / Q < 1 , the inner pair initially absent in case (iii) reappears at the positive emitted energy value
E s : = Q 2 η 2 M i ( 2 η M i Q − 1 ) , (92) \mathcal{E}_{\mathrm{s}}:=\frac{Q^2}{\eta^2\mathcal{M}_i}\left(\frac{2\eta\mathcal{M}_i}{Q}-1\right), \tag*{(92)} E s := η 2 M i Q 2 ( Q 2 η M i − 1 ) , (92)
where the path meets the upper critical curve again. The charge and mass are still nonzero there. A double inner zero is recovered at E s \mathcal{E}_{\mathrm{s}} E s and splits into two simple horizons with separation proportional to ( E − E s ) 1 / 2 (\mathcal{E}-\mathcal{E}_{\mathrm{s}})^{1/2} ( E − E s ) 1/2 , before the middle horizon later meets the event horizon. The charged history can therefore contain an interval with a single event horizon between two encounters with the inner critical curve. For a constant-η \eta η emission channel with 0 < η M i / Q < 1 0<\eta\mathcal{M}_i/Q<1 0 < η M i / Q < 1 , the outer merger is reached at positive mass and nonzero charge of the initial sign, at the unique positive zero of Δ M e \Delta\mathcal{M}_{\mathrm{e}} Δ M e . If η M i / Q ≥ 1 \eta\mathcal{M}_i/Q\ge1 η M i / Q ≥ 1 , that curve is not reached before the charge vanishes, where the nonzero-charge construction ends.
The nonlinear paths in Figure 3 allow the charge extracted per unit energy to vary during emission. We take Q > 0 Q>0 Q > 0 and retain M ′ = M i − E \mathcal{M}'=\mathcal{M}_i-\mathcal{E} M ′ = M i − E and Q ′ ( E ) = Q − Q e m ( E ) Q'(\mathcal{E})=Q-Q_{\mathrm{em}}(\mathcal{E}) Q ′ ( E ) = Q − Q em ( E ) , where unprimed M i \mathcal{M}_i M i and M e \mathcal{M}_{\mathrm{e}} M e denote the critical masses evaluated at the initial charge. For each path, E c \mathcal{E}_{\mathrm{c}} E c will denote its own outer coalescence energy. Concretely, we consider the following two profiles.
Figure 3. Critical charge–mass structure and emission paths of the logarithmically deformed BTZ family for κ = ℓ = 1 \kappa= \ell= 1 κ = ℓ = 1 , β = 27 / 20 \beta= 27/20 β = 27/20 , and q = γ = 4 / π q = \gamma= 4/\pi q = γ = 4/ π , so the initial charge is Q i = Q = 8 \mathcal{Q}_{i} = \mathcal{Q} = 8 Q i = Q = 8 . The critical curves M i ( Q ) \mathcal{M}_{i}(\mathcal{Q}) M i ( Q ) and M e ( Q ) \mathcal{M}_{e}(\mathcal{Q}) M e ( Q ) delimit the shaded three-horizon strip. The pale regions contain a single event horizon. At nonzero charge, the upper curve contains a double internal horizon and a simple event horizon, whereas the lower curve contains a simple innermost horizon and a double event horizon. All five paths start at ( Q , M i ) (\mathcal{Q},\mathcal{M}_{i}) ( Q , M i ) and are followed in the direction of increasing emitted energy. The solid black path is neutral. The solid blue and violet paths have Q e m = η E \mathcal{Q}_{\rm em} = \eta\mathcal{E} Q em = η E , with η M i / Q = 1 / 2 \eta\mathcal{M}_{i}/\mathcal{Q} = 1/2 η M i / Q = 1/2 and 3 / 4 3/4 3/4 , respectively: the former enters tangentially, while the latter leaves the strip and returns through the upper critical curve at E s > 0 \mathcal{E}_{s} > 0 E s > 0 . The dashed indigo path is the exponential profile (93) , with a = 3 a = 3 a = 3 and Q f / Q = 0.86 \mathcal{Q}_{f}/\mathcal{Q} = 0.86 Q f / Q = 0.86 . The dash-dotted green path is the fractional profile (94) , with p ≃ 0.1063 p \simeq0.1063 p ≃ 0.1063 and Q f / Q = 0.65 \mathcal{Q}_{f}/\mathcal{Q} = 0.65 Q f / Q = 0.65 . Lower-boundary markers identify the merger of the middle and event horizons. The additional upper-boundary marker on the 3 / 4 3/4 3/4 path identifies the delayed recovery of the inner double zero.
Exponential charge depletion. Suppose that charge extraction becomes less efficient as the remaining charge approaches a residual value 0 ≤ Q r e s < Q 0\le Q_{\mathrm{res}}<Q 0 ≤ Q res < Q . A relaxation law analogous to Newton’s cooling law, here parametrized by emitted energy, is
d Q ′ d E = − a E c ( Q ′ − Q r e s ) , a > 0. \frac{dQ'}{d\mathcal{E}}=-\frac{a}{\mathcal{E}_{\mathrm{c}}}(Q'-Q_{\mathrm{res}}), \qquad a>0. d E d Q ′ = − E c a ( Q ′ − Q res ) , a > 0.
Integrating with Q ′ ( 0 ) = Q Q'(0)=Q Q ′ ( 0 ) = Q gives the exponential charge-depletion profile
Q e m ( E ) = ( Q − Q r e s ) [ 1 − e − a E / E c ] . (93) Q_{\mathrm{em}}(\mathcal{E})=(Q-Q_{\mathrm{res}})\left[1-e^{-a\mathcal{E}/\mathcal{E}_{\mathrm{c}}}\right]. \tag*{(93)} Q em ( E ) = ( Q − Q res ) [ 1 − e − a E / E c ] . (93)
Here a a a is dimensionless and the charge extracted per unit energy decreases exponentially from η 0 = a ( Q − Q r e s ) / E c \eta_0=a(Q-Q_{\mathrm{res}})/\mathcal{E}_{\mathrm{c}} η 0 = a ( Q − Q res ) / E c .
Proportional fractional loss. A different prescription takes the fractional charge loss to be proportional to the fractional mass loss,
d Q ′ Q ′ = p d M ′ M ′ , p > 0. \frac{dQ'}{Q'}=p\frac{d\mathcal{M}'}{\mathcal{M}'}, \qquad p>0. Q ′ d Q ′ = p M ′ d M ′ , p > 0.
Integrating with M ′ = M i − E \mathcal{M}'=\mathcal{M}_i-\mathcal{E} M ′ = M i − E and Q ′ ( 0 ) = Q Q'(0)=Q Q ′ ( 0 ) = Q gives
Q e m ( E ) = Q [ 1 − ( 1 − E M i ) p ] , 0 ≤ E < M i . (94) Q_{\mathrm{em}}(\mathcal{E})=Q\left[1-\left(1-\frac{\mathcal{E}}{\mathcal{M}_i}\right)^p\right], \qquad0\le\mathcal{E}<\mathcal{M}_i. \tag*{(94)} Q em ( E ) = Q [ 1 − ( 1 − M i E ) p ] , 0 ≤ E < M i . (94)
Here η 0 = p Q / M i \eta_0 = pQ/\mathcal{M}_i η 0 = pQ / M i , and d Q e m / d E = η 0 ( 1 − E / M i ) p − 1 \mathrm{d}Q_{\mathrm{em}}/\mathrm{d}\mathcal{E} = \eta_0(1-\mathcal{E}/\mathcal{M}_i)^{p-1} d Q em / d E = η 0 ( 1 − E / M i ) p − 1 . For 0 < p < 1 / 2 0 < p < 1/2 0 < p < 1/2 , charge extraction accelerates as the mass decreases, while the path remains below the upper critical curve. Indeed, Eq. (89) gives
Δ M i = M i [ ( 1 − E M i ) 2 p − ( 1 − E M i ) ] > 0 \Delta\mathcal{M}_i = \mathcal{M}_i \left[\left(1-\frac{\mathcal{E}}{\mathcal{M}_i}\right)^{2p}-\left(1-\frac{\mathcal{E}}{\mathcal{M}_i}\right)\right] > 0 Δ M i = M i [ ( 1 − M i E ) 2 p − ( 1 − M i E ) ] > 0
for 0 < E < M i 0 < \mathcal{E} < \mathcal{M}_i 0 < E < M i . The profile is analytic for E < M i \mathcal{E} < \mathcal{M}_i E < M i for any real p > 0 p > 0 p > 0 .
For either of these profiles, the endpoint can be selected by prescribing the remaining charge fraction s f : = Q f / Q s_f := Q_f/Q s f := Q f / Q , where Q f = Q ′ ( E c ) Q_f = Q'(\mathcal{E}_c) Q f = Q ′ ( E c ) . The lower critical curve then fixes
E c = M i − M e s f 2 . (95) \mathcal{E}_c = \mathcal{M}_i-\mathcal{M}_e s_f^2. \tag*{(95)} E c = M i − M e s f 2 . (95)
The endpoint condition determines the residual charge in Eq. (93) and the exponent in Eq. (94) , respectively:
Q r e s Q = s f − e − a 1 − e − a , p = ln s f ln [ ( M e / M i ) s f 2 ] . (96) \begin{aligned}
\frac{Q_{\mathrm{res}}}{Q} &= \frac{s_f-e^{-a}}{1-e^{-a}},\\
p &= \frac{\ln s_f}{\ln\left[(\mathcal{M}_e/\mathcal{M}_i)s_f^2\right]}. \tag*{(96)}
\end{aligned} Q Q res p = 1 − e − a s f − e − a , = ln [ ( M e / M i ) s f 2 ] ln s f . (96)
A nonnegative residual requires s f ≥ e − a s_f \ge e^{-a} s f ≥ e − a . For 0 < s f < 1 0 < s_f < 1 0 < s f < 1 and 0 < M e < M i 0 < \mathcal{M}_e < \mathcal{M}_i 0 < M e < M i , the fractional prescription automatically gives 0 < p < 1 / 2 0 < p < 1/2 0 < p < 1/2 .
For the nonlinear examples in Figure 3 and Figure 4 , we choose the parameters so that both initial slopes lie below the tangent value, η 0 M i / Q < 1 / 2 \eta_0\mathcal{M}_i/Q < 1/2 η 0 M i / Q < 1/2 . The inner pair therefore splits immediately with square-root separation in both cases, even though the charge extracted per unit energy decreases in one profile and increases in the other. The subsequent horizon positions depend on the full charge profile. In particular, the remaining charge at coalescence fixes the merger radius through r min ∝ Q ′ r_{\min} \propto Q' r m i n ∝ Q ′ and the total emitted energy through Eq. (95) . The fractional-loss example reaches the lower critical curve with less remaining charge than the exponential one. Its merger therefore occurs at a smaller radius and a larger emitted energy, as seen in Figure 4 (d) and Figure 4 (e). For the parameters shown, both paths remain inside the three-horizon strip until this merger and satisfy 1 − Φ + d Q e m / d E > 0 1-\Phi_+ \mathrm{d}Q_{\mathrm{em}}/\mathrm{d}\mathcal{E} > 0 1 − Φ + d Q em / d E > 0 throughout the simple outer branch. Both profiles prescribe charge loss as a function of energy. Their realization by a particular radiating field depends on its energy and charge fluxes.
Figure 4. Running horizons along the five paths in Figure 3 , with the same parameters and initial inner-degenerate state. Red, dashed violet, and blue denote the innermost, middle, and event horizons, respectively. Radii are normalized by the common initial event-horizon radius r + ( 0 ) r_{+}(0) r + ( 0 ) , and the emitted energy by the coalescence energy E c \mathcal{E}_{c} E c of each path. (a) Neutral emission resolves the internal double zero immediately, with separation proportional to E 1 / 2 \mathcal{E}^{1/2} E 1/2 . (b) The linear η M i / Q i = 1 / 2 \eta M_{i}/Q_{i}=1/2 η M i / Q i = 1/2 path enters tangentially and yields a linear separation. (c) The linear 3 / 4 3/4 3/4 path initially retains only the simple event horizon. The isolated internal point at E = 0 \mathcal{E}=0 E = 0 records the initial double zero, recovered later at E s ≃ 3.6477 \mathcal{E}_{s}\simeq3.6477 E s ≃ 3.6477 before splitting as ( E − E s ) 1 / 2 (\mathcal{E}-\mathcal{E}_{s})^{1/2} ( E − E s ) 1/2 . (d) Exponential depletion and (e) fractional loss both produce immediate square-root splitting, with different subsequent horizon histories. The first five panels cover the full emission interval and have the same radial range. Their coalescence energies are E c ≃ 3.9347 \mathcal{E}_{c}\simeq3.9347 E c ≃ 3.9347 , 4.0606, 4.0930, 3.9787, 4.0323, respectively. In each case the middle and event horizons merge while a simple innermost horizon survives. (f) Zoom to the final interval of panel (c), with a different radial range. Dashed and dotted guides mark delayed inner splitting and outer coalescence, respectively.
The horizon histories in Figure 4 follow from the full horizon condition, evaluated using Eq. (83) .
At the internal transitions, the outer denominator in Eq. (90) stays positive and nonzero. For the analytic profiles considered here, R + R_+ R + and Δ S \Delta S Δ S are therefore analytic through these transitions, including the delayed splitting at E s \mathcal{E}_s E s . The simple-root description of the outer horizon ends at the later coalescence. The internal rearrangement therefore produces no nonanalytic feature in the leading horizon emission exponent. Smooth differences between charge profiles remain possible, and the full radiation spectrum also depends on propagation outside the horizon.
Further emission channels
The interplay between charge and energy emission can also produce profiles that preserve the inner degeneracy throughout the whole process. The limiting choice p = 1 / 2 p = 1/2 p = 1/2 in Eq. (94) gives such a channel,
Q e m ( E ) = Q [ 1 − 1 − E M i ] , 0 ≤ E < M i , (97) \mathcal{Q}_{\rm em}(\mathcal{E}) = \mathcal{Q}\left[1-\sqrt{1-\frac{\mathcal{E}}{\mathcal{M}_{i}}}\right], \qquad0 \leq\mathcal{E} < \mathcal{M}_{i}, \tag*{(97)} Q em ( E ) = Q [ 1 − 1 − M i E ] , 0 ≤ E < M i , (97)
for which M − E = M i [ Q ′ ( E ) ] \mathcal{M}-\mathcal{E}=\mathcal{M}_{i}[\mathcal{Q}'(\mathcal{E})] M − E = M i [ Q ′ ( E )] at every stage. All horizon radii scale by 1 − E / M i \sqrt{1-\mathcal{E}/\mathcal{M}_{i}} 1 − E / M i , so the double inner horizon and the simple event horizon persist without an intervening three-horizon geometry. Its outer horizon-energy factor is the positive constant
r + ∂ r m ( r + , q ) 2 M i . \frac{r_{+}\partial_{r}m(r_{+},q)}{2\mathcal{M}_{i}}. 2 M i r + ∂ r m ( r + , q ) .
The limit Q ′ → 0 Q' \to0 Q ′ → 0 lies outside the nonzero-charge sector used here. Another possibility is that a curved channel instead unfolds the inner double horizon at the beginning of the process and later brings the pair together again at a lower point on the same critical curve. An explicit profile of such a channel is
Q e m ( E ) = Q [ 1 − 1 − E 2 M i E r ] , (98) Q_{\mathrm{em}}(\mathcal{E}) = Q\left[1-\sqrt{1-\frac{\mathcal{E}^{2}}{\mathcal{M}_{i}\mathcal{E}_{r}}}\right], \tag*{(98)} Q em ( E ) = Q 1 − 1 − M i E r E 2 , (98)
where E r \mathcal{E}_{r} E r is the return energy, chosen within 0 < E r < M i − M e 0 < \mathcal{E}_{r} < \mathcal{M}_{i} - \mathcal{M}_{e} 0 < E r < M i − M e , and 0 ≤ E ≤ E r 0 \le\mathcal{E} \le\mathcal{E}_{r} 0 ≤ E ≤ E r . Charge extraction remains monotonic, with ( ∂ E Q e m ) ( 0 ) = 0 (\partial_{\mathcal{E}}Q_{\mathrm{em}})(0)=0 ( ∂ E Q em ) ( 0 ) = 0 . It starts at order E 2 \mathcal{E}^{2} E 2 , so the leading response is the neutral square-root splitting. The inner mass difference is exactly
Δ M i ( E ) = E ( 1 − E E r ) . (99) \Delta\mathcal{M}_{i}(\mathcal{E})=\mathcal{E}\left(1-\frac{\mathcal{E}}{\mathcal{E}_{r}}\right). \tag*{(99)} Δ M i ( E ) = E ( 1 − E r E ) . (99)
Together with the bound on E r \mathcal{E}_{r} E r , this places the path inside the three-horizon strip for 0 < E < E r 0 < \mathcal{E} < \mathcal{E}_{r} 0 < E < E r . The two histories are compared in Figure 5 . Along the returning path, the inner pair separates as E 1 / 2 \mathcal{E}^{1/2} E 1/2 and merges again as ( E r − E ) 1 / 2 (\mathcal{E}_{r}-\mathcal{E})^{1/2} ( E r − E ) 1/2 . The event horizon remains simple and at nonzero temperature throughout this return to inner degeneracy. For the parameters shown, the positive horizon energy condition 1 − Φ + d Q e m / d E > 0 1-\Phi_{+}\,\mathrm{d}Q_{\mathrm{em}}/\mathrm{d}\mathcal{E}>0 1 − Φ + d Q em / d E > 0 also holds throughout.
Figure 5. Further charged emission channels for κ = ℓ = γ = q = 1 \kappa= \ell= \gamma= q = 1 κ = ℓ = γ = q = 1 , β = 6 / 5 \beta= 6/5 β = 6/5 , and E r = ( M i − M e ) / 2 \mathcal{E}_{r} = (\mathcal{M}_{i} - \mathcal{M}_{e})/2 E r = ( M i − M e ) /2 . (a) The dash-dotted profile (97) strictly follows the M i ( Q ) \mathcal{M}_{i}(Q) M i ( Q ) curve, whereas the dashed profile (98) enters the shaded three-horizon region and returns to this curve. Both paths are shown up to E r \mathcal{E}_{r} E r and have the same initial and final states. (b) Full horizon radii along the returning channel. Red, violet, and blue denote the innermost, middle, and event horizons. Both endpoints contain an internal double horizon and a simple event horizon.
At E r \mathcal{E}_{r} E r , the returning channel and the profile preserving the inner degeneracy reach the same black hole state. Equation (27) therefore gives their common leading tunneling exponent,
Δ S ( E r ) = S ( 0 ) [ 1 − E r M i − 1 ] , (100) \Delta S(\mathcal{E}_{r})=S(0)\left[\sqrt{1-\frac{\mathcal{E}_{r}}{\mathcal{M}_{i}}}-1\right], \tag*{(100)} Δ S ( E r ) = S ( 0 ) [ 1 − M i E r − 1 ] , (100)
The same leading exponent thus accompanies two different internal horizon histories.
Approach to outer coalescence
The histories presented in Figure 4 reach the lower critical curve at points determined by their charge profiles. There the event horizon merges with the middle root and its temperature vanishes. The limiting tunneling exponent can still be obtained from the stationary family, while the duration of this approach requires the energy and charge fluxes. The finite-energy exponent also determines when a single emission can be treated thermally.
Limiting tunneling exponent
For an approach from the three-horizon sector at nonzero charge, the coalescence energy satisfies
M − E c = M e [ Q − Q e m ( E c ) ] (101) \mathcal{M}-\mathcal{E}_{c}=\mathcal{M}_{e}\left[Q-Q_{\mathrm{em}}(\mathcal{E}_{c})\right] \tag*{(101)} M − E c = M e [ Q − Q em ( E c ) ] (101)
and therefore depends on the charge profile. Since the lower critical mass scales as Q ′ 2 Q'^{2} Q ′ 2 , removing charge of the same sign as the black hole, within the chosen fixed sign sector, lowers this mass and requires more emitted energy than the neutral value in Eq. (78) . Emission of oppositely charged matter instead increases ∣ Q ′ ∣ \lvert Q'\rvert ∣ Q ′ ∣ and raises the critical mass, allowing coalescence at a smaller emitted energy. For either sign, the channel must satisfy 1 − Φ + d Q e m / d E > 0 1-\Phi_{+}\,\mathrm{d}Q_{\mathrm{em}}/\mathrm{d}\mathcal{E}>0 1 − Φ + d Q em / d E > 0 while the event horizon remains simple.
At fixed couplings, charge variation moves r min r_{\min} r m i n without changing the radial curvature of m m m at this minimum. On the three-horizon side Δ M e < 0 \Delta\mathcal{M}_{e}<0 Δ M e < 0 , the event horizon and its temperature consequently behave as
R + ( E ) = r min [ q ( E ) ] + ℓ κ w − 1 π ( 1 + w − 1 ) − Δ M e + O ( ∣ Δ M e ∣ ) , R_{+}(\mathcal{E})=r_{\min}[q(\mathcal{E})]+\ell\sqrt{\frac{\kappa w_{-1}}{\pi(1+w_{-1})}}\sqrt{-\Delta\mathcal{M}_{e}}+O(\lvert\Delta\mathcal{M}_{e}\rvert), R + ( E ) = r m i n [ q ( E )] + ℓ π ( 1 + w − 1 ) κ w − 1 − Δ M e + O (∣ Δ M e ∣) ,
T H ( E ) = 1 2 π ℓ κ ( 1 + w − 1 ) π w − 1 − Δ M e + O ( ∣ Δ M e ∣ ) . (102) T_{\mathrm{H}}(\mathcal{E})=\frac{1}{2\pi\ell}\sqrt{\frac{\kappa(1+w_{-1})}{\pi w_{-1}}}\sqrt{-\Delta\mathcal{M}_{e}}+O(\lvert\Delta\mathcal{M}_{e}\rvert). \tag*{(102)} T H ( E ) = 2 π ℓ 1 π w − 1 κ ( 1 + w − 1 ) − Δ M e + O (∣ Δ M e ∣) . (102)
The middle horizon approaches the same running minimum with the opposite leading displacement. For neutral emission, − Δ M e = E c − E -\Delta\mathcal{M}_{e}=\mathcal{E}_{c}-\mathcal{E} − Δ M e = E c − E , recovering the square-root merger and cooling of § 3.3 . For charged emission, it is the mass difference from the running critical curve that enters this square-root law. If − Δ M e -\Delta\mathcal{M}_{e} − Δ M e is proportional to ( E c − E ) n (\mathcal{E}_{c}-\mathcal{E})^{n} ( E c − E ) n , the horizon separation and temperature vanish as ( E c − E ) n / 2 (\mathcal{E}_{c}-\mathcal{E})^{n/2} ( E c − E ) n /2 . For a path analytic at the endpoint, n ≥ 1 n\geq1 n ≥ 1 is the first nonzero integer order. The scaling analysis below also allows a leading power n > 0 n>0 n > 0 when the approach is not analytic.
Throughout the neutral trajectory, R + ≥ r min R_{+}\geq r_{\min} R + ≥ r m i n and ∂ r 2 f ( R + ) > 0 \partial_{r}^{2}f(R_{+})>0 ∂ r 2 f ( R + ) > 0 . Together with Eq. (80) , this implies d T H / d E < 0 \mathrm{d}T_{\mathrm{H}}/\mathrm{d}\mathcal{E}<0 d T H / d E < 0 along the simple outer branch and hence
d 2 Δ S d E 2 = 1 T H 2 d T H d E < 0. (103) \frac{\mathrm{d}^{2}\Delta S}{\mathrm{d}\mathcal{E}^{2}}=\frac{1}{T_{\mathrm{H}}^{2}}\frac{\mathrm{d}T_{\mathrm{H}}}{\mathrm{d}\mathcal{E}}<0. \tag*{(103)} d E 2 d 2 Δ S = T H 2 1 d E d T H < 0. (103)
The finite-emission exponent lies strictly below its initial thermal tangent throughout the trajectory:
Δ S ( E ) < − E T H ( 0 ) , 0 < E ≤ E c . (104) \Delta S(\mathcal{E})<-\frac{\mathcal{E}}{T_{\mathrm{H}}(0)},\qquad0<\mathcal{E}\leq\mathcal{E}_{c}. \tag*{(104)} Δ S ( E ) < − T H ( 0 ) E , 0 < E ≤ E c . (104)
The endpoint is understood as a limit through simple outer horizons. Figure 6 compares the full entropy difference with the initial thermal exponent and the quadratic and quartic inverse truncations. The higher-order truncation improves the local approximation, but the square-root endpoint cannot be reproduced by a finite polynomial in E \mathcal{E} E . The entropy difference remains finite while its derivative d Δ S / d E = − 1 / T H \mathrm{d}\Delta S/\mathrm{d}\mathcal{E} = -1/T_{\mathrm{H}} d Δ S / d E = − 1/ T H diverges.
Figure 6. Finite-energy entropy exponent for neutral emission, for κ = ℓ = γ = q = 1 \kappa= \ell= \gamma= q = 1 κ = ℓ = γ = q = 1 and β = 6 / 5 \beta= 6/5 β = 6/5 , as in Figure 2 . The solid curve is the full Δ S \Delta S Δ S , evaluated on the outer root giving ln Γ K P W \ln\Gamma_{\mathrm{KPW}} ln Γ KPW at leading semiclassical order. The dashed line is the thermal exponent − E / T H ( 0 ) -\mathcal{E}/T_{\mathrm{H}}(0) − E / T H ( 0 ) . The dash-dotted and dotted curves truncate the inverse expansion in Appendix § 13 at second and fourth order, respectively. The horizontal axis is normalized by the neutral E c = M i − M e \mathcal{E}_{c} = \mathcal{M}_{i} - \mathcal{M}_{e} E c = M i − M e . The exponent obeys the global bound (104) and approaches Δ S ( E c ) ≃ − 43.209 \Delta S(\mathcal{E}_{c}) \simeq-43.209 Δ S ( E c ) ≃ − 43.209 with a divergent negative slope. The dot denotes this limit from simple outer horizons. At the endpoint, the quadratic and quartic exponents have relative errors of approximately 40% and 29%, respectively. These finite-order expansions resolve the initial emission, but lose accuracy as the outer root approaches its double-zero limit.
The limiting exponent is finite for both neutral and charged paths approaching the lower critical curve at nonzero final charge Q f = Q ′ ( E c ) Q_{f} = Q'(\mathcal{E}_{c}) Q f = Q ′ ( E c ) ,
lim E → E c − ln Γ K P W ≃ 4 π 2 κ [ r min ( Q f 2 π ) − r + ] . (105) \lim_{\mathcal{E}\to\mathcal{E}_{c}^{-}} \ln\Gamma_{\mathrm{KPW}} \simeq\frac{4\pi^{2}}{\kappa}\left[r_{\min}\left(\frac{Q_{f}}{2\pi}\right)-r_{+}\right]. \tag*{(105)} E → E c − lim ln Γ KPW ≃ κ 4 π 2 [ r m i n ( 2 π Q f ) − r + ] . (105)
For neutral emission, Q f = Q Q_{f} = Q Q f = Q . For a charged approach, the full electric work factor in Eq. (28) must be retained, as shown in Appendix § 11 .§ 4 . For a tangent approach, it is this full integrand, rather than 1 / T H 1/T_{\mathrm{H}} 1/ T H alone, that determines the limiting action. The resulting positive-mass, zero-temperature configuration contains the double event horizon and a surviving simple innermost horizon. For 0 < M ′ < M e ( Q ′ ) 0 < M' < \mathcal{M}_{e}(Q') 0 < M ′ < M e ( Q ′ ) , the surviving innermost root is the only positive horizon. Thus the lower critical mass terminates the tracked outer branch, even though smaller-mass black holes remain in the stationary family. A transition to this smaller horizon would require a dynamical description.
Time dependence and adiabaticity
The horizon histories above are parametrized by emitted energy. The KPW integral compares states along such a path, but does not determine the time separating successive emissions. To discuss this duration, we now use the stationary family as an instantaneous description of an averaged evaporation process. The cumulative energy becomes E ( t ) E(t) E ( t ) , where t t t is the time at the AdS boundary with the Killing normalization used for the mass and temperature. The evolution of the parameters is supplied by energy and charge balance.
Let L r a d = − d M ′ / d t > 0 L_{\rm rad}=-d\mathcal{M}'/dt>0 L rad = − d M ′ / d t > 0 denote the net luminosity, measured as the black hole energy lost per unit boundary time. Similarly, J r a d = − d Q ′ / d t J_{\rm rad}=-dQ'/dt J rad = − d Q ′ / d t is the rate of loss of signed electric charge. Conservation gives
M ′ ( t ) = M − E ( t ) , Q ′ ( t ) = Q − Q e m [ E ( t ) ] . \mathcal{M}'(t)=\mathcal{M}-E(t), \qquad Q'(t)=Q-Q_{\rm em}[E(t)]. M ′ ( t ) = M − E ( t ) , Q ′ ( t ) = Q − Q em [ E ( t )] .
In particular, an evolution following the prescribed path must satisfy
d E d t = L r a d , J r a d L r a d = d Q e m d E . (106) \frac{dE}{dt}=L_{\rm rad}, \qquad\frac{J_{\rm rad}}{L_{\rm rad}}=\frac{dQ_{\rm em}}{dE}. \tag*{(106)} d t d E = L rad , L rad J rad = d E d Q em . (106)
The charge profile fixes this ratio, while the luminosity sets the rate at which the path is traversed. Its determination requires the radiation spectrum and its transmission through the exterior, together with boundary conditions allowing a net loss of energy, as discussed in § 2 . The stationary first law also gives the instantaneous entropy rate,
d S d t = − L r a d − Φ + J r a d T H . \frac{dS}{dt}=-\frac{L_{\rm rad}-\Phi_+J_{\rm rad}}{T_{\rm H}}. d t d S = − T H L rad − Φ + J rad .
The positive horizon-energy condition used above therefore becomes L r a d − Φ + J r a d > 0 L_{\rm rad}-\Phi_+J_{\rm rad}>0 L rad − Φ + J rad > 0 .
We measure the approach to the selected outer coalescence by the remaining energy
Δ E ( t ) : = E c − E ( t ) > 0 , d Δ E d t = − L r a d . (107) \Delta E(t):=E_{\rm c}-E(t)>0, \qquad\frac{d\Delta E}{dt}=-L_{\rm rad}. \tag*{(107)} Δ E ( t ) := E c − E ( t ) > 0 , d t d Δ E = − L rad . (107)
For neutral emission this is the mass above the lower critical value. For charged emission the critical mass itself varies, and the mass difference governing the double zero instead has the form
− Δ M e ∼ K ( Δ E ) n , K > 0 , n > 0. -\Delta M_{\rm e}\sim K(\Delta E)^n, \qquad K>0,\quad n>0. − Δ M e ∼ K ( Δ E ) n , K > 0 , n > 0.
Here K K K and n n n are fixed by the chosen path. The analytic case is derived in Appendix § 11 .§ 4 . For a more general power-law approach, we assume that the leading expansion can also be differentiated along the path. The two gaps coincide for neutral emission, with K = 1 K=1 K = 1 and n = 1 n=1 n = 1 . Equation (102) then determines the temperature near the endpoint,
T H ∼ A T ( Δ E ) n / 2 , T_{\rm H}\sim A_T(\Delta E)^{n/2}, T H ∼ A T ( Δ E ) n /2 ,
A T : = 1 2 π ℓ κ K ( 1 + w − 1 ) π w − 1 > 0. (108) A_T:=\frac{1}{2\pi\ell}\sqrt{\frac{\kappa K(1+w_{-1})}{\pi w_{-1}}}>0. \tag*{(108)} A T := 2 π ℓ 1 π w − 1 κ K ( 1 + w − 1 ) > 0. (108)
Choose t 0 t_0 t 0 sufficiently near coalescence that these expansions apply, and write Δ E 0 : = Δ E ( t 0 ) > 0 \Delta E_0:=\Delta E(t_0)>0 Δ E 0 := Δ E ( t 0 ) > 0 . Expressing the luminosity along the path as L r a d ( Δ E ) L_{\rm rad}(\Delta E) L rad ( Δ E ) , integration of Eq. (107) gives
t − t 0 = ∫ Δ E ( t ) Δ E 0 d ξ L r a d ( ξ ) . (109) t-t_0=\int_{\Delta E(t)}^{\Delta E_0}\frac{d\xi}{L_{\rm rad}(\xi)}. \tag*{(109)} t − t 0 = ∫ Δ E ( t ) Δ E 0 L rad ( ξ ) d ξ . (109)
The duration of the final approach is the limit as the lower integration bound tends to zero. For a luminosity whose leading behavior is L r a d ∼ L 0 ( Δ E ) ν L_{\mathrm{rad}} \sim L_0(\Delta\mathcal{E})^\nu L rad ∼ L 0 ( Δ E ) ν , with L 0 > 0 L_0 > 0 L 0 > 0 , retaining this leading term yields
Δ E ( t ) ≃ { [ ( Δ E 0 ) 1 − ν − ( 1 − ν ) L 0 ( t − t 0 ) ] 1 1 − ν , ν ≠ 1 , Δ E 0 e − L 0 ( t − t 0 ) , ν = 1. (110) \Delta\mathcal{E}(t) \simeq\begin{cases} \left[(\Delta\mathcal{E}_0)^{1-\nu}-(1-\nu)L_0(t-t_0)\right]^{\frac{1}{1-\nu}}, & \nu\ne1, \\ \Delta\mathcal{E}_0 e^{-L_0(t-t_0)}, & \nu=1. \end{cases} \tag*{(110)} Δ E ( t ) ≃ { [ ( Δ E 0 ) 1 − ν − ( 1 − ν ) L 0 ( t − t 0 ) ] 1 − ν 1 , Δ E 0 e − L 0 ( t − t 0 ) , ν = 1 , ν = 1. (110)
The endpoint is reached in finite time only for ν < 1 \nu< 1 ν < 1 , with t c − t 0 ≃ ( Δ E 0 ) 1 − ν / [ ( 1 − ν ) L 0 ] t_c-t_0 \simeq(\Delta\mathcal{E}_0)^{1-\nu}/[(1-\nu)L_0] t c − t 0 ≃ ( Δ E 0 ) 1 − ν / [( 1 − ν ) L 0 ] . Otherwise the approach is exponential for ν = 1 \nu=1 ν = 1 and algebraic for ν > 1 \nu>1 ν > 1 . The convergence criterion follows directly from Eq. (109) , independently of the subleading terms in the luminosity.
To approximate the radiation temperature by the stationary T H T_{\mathrm{H}} T H , the inverse surface gravity timescale ( 2 π T H ) − 1 (2\pi T_{\mathrm{H}})^{-1} ( 2 π T H ) − 1 must be short compared with the temperature-variation timescale T H / ∣ d T H / d t ∣ T_{\mathrm{H}}/|dT_{\mathrm{H}}/dt| T H /∣ d T H / d t ∣ . Their ratio gives the adiabaticity condition
ϵ a d : = ∣ d T H / d t ∣ 2 π T H 2 ≪ 1. (111) \epsilon_{\mathrm{ad}} := \frac{|dT_{\mathrm{H}}/dt|}{2\pi T_{\mathrm{H}}^2} \ll1. \tag*{(111)} ϵ ad := 2 π T H 2 ∣ d T H / d t ∣ ≪ 1. (111)
This implements the slow variation criterion for a Hawking-like spectrum [22 ] , assuming that the instantaneous surface gravity approximates the peeling rate of the outgoing rays. Using the differentiable endpoint expansion and d Δ E / d t = − L r a d d\Delta\mathcal{E}/dt=-L_{\mathrm{rad}} d Δ E / d t = − L rad gives
ϵ a d ∼ n L 0 4 π A T ( Δ E ) ν − 1 − n / 2 . (112) \epsilon_{\mathrm{ad}} \sim\frac{nL_0}{4\pi A T}(\Delta\mathcal{E})^{\nu-1-n/2}. \tag*{(112)} ϵ ad ∼ 4 π A T n L 0 ( Δ E ) ν − 1 − n /2 . (112)
Consequently, ϵ a d \epsilon_{\mathrm{ad}} ϵ ad decreases towards coalescence when ν > 1 + n / 2 \nu>1+n/2 ν > 1 + n /2 . At ν = 1 + n / 2 \nu=1+n/2 ν = 1 + n /2 , it remains small if n L 0 / ( 4 π A T ) ≪ 1 nL_0/(4\pi A T)\ll1 n L 0 / ( 4 π A T ) ≪ 1 . For a transverse approach the marginal power is ν = 3 / 2 \nu=3/2 ν = 3/2 , while quadratic contact raises it to ν = 2 \nu=2 ν = 2 . For ν < 1 + n / 2 \nu<1+n/2 ν < 1 + n /2 , the adiabaticity parameter grows without bound. In particular, an infinite approach time does not suffice: the interval 1 ≤ ν < 1 + n / 2 1\le\nu<1+n/2 1 ≤ ν < 1 + n /2 has an infinite duration but still loses adiabatic control.
The same calculation locates this loss of control. For a fixed small tolerance ϵ ∗ > 0 \epsilon_* > 0 ϵ ∗ > 0 , a power law with ν < 1 + n / 2 \nu<1+n/2 ν < 1 + n /2 reaches ϵ a d = ϵ ∗ \epsilon_{\mathrm{ad}}=\epsilon_* ϵ ad = ϵ ∗ at
Δ E a d ≃ ( n L 0 4 π A T ϵ ∗ ) 1 1 + n / 2 − ν . (113) \Delta\mathcal{E}_{\mathrm{ad}} \simeq\left(\frac{nL_0}{4\pi A T\epsilon_*}\right)^{\frac{1}{1+n/2-\nu}}. \tag*{(113)} Δ E ad ≃ ( 4 π A T ϵ ∗ n L 0 ) 1 + n /2 − ν 1 . (113)
This estimate applies if the evolution was initially adiabatic and the resulting gap lies within the endpoint expansion. Without specifying a luminosity law, a uniform bound ϵ a d ≤ ϵ ∗ \epsilon_{\mathrm{ad}}\le\epsilon_* ϵ ad ≤ ϵ ∗ implies ∣ d ( 1 / T H ) / d t ∣ ≤ 2 π ϵ ∗ |d(1/T_{\mathrm{H}})/dt|\le2\pi\epsilon_* ∣ d ( 1/ T H ) / d t ∣ ≤ 2 π ϵ ∗ . Integrating gives
T H ( t ) ≥ T H ( t 0 ) 1 + 2 π ϵ ∗ T H ( t 0 ) ( t − t 0 ) . (114) T_{\mathrm{H}}(t) \ge\frac{T_{\mathrm{H}}(t_0)}{1+2\pi\epsilon_*T_{\mathrm{H}}(t_0)(t-t_0)}. \tag*{(114)} T H ( t ) ≥ 1 + 2 π ϵ ∗ T H ( t 0 ) ( t − t 0 ) T H ( t 0 ) . (114)
A zero temperature is therefore excluded at finite boundary time while this adiabatic bound remains satisfied. The incompatibility between finite evaporation time and continued adiabaticity was also found for regular four-dimensional black holes with a Stefan–Boltzmann loss law [45 ] .
For neutral emission into massless bulk modes, an ideal blackbody estimate in 2 + 1 2+1 2 + 1 dimensions takes L r a d ∝ ( 2 π R + ) T H 3 L_{\mathrm{rad}}\propto(2\pi R_+)T_{\mathrm{H}}^3 L rad ∝ ( 2 π R + ) T H 3 [25 ] . Since R + → r min > 0 R_+\to r_{\min}>0 R + → r m i n > 0 , this yields ν = 3 / 2 \nu=3/2 ν = 3/2 , the marginal adiabatic case for n = 1 n=1 n = 1 . The actual net luminosity depends on greybody transmission and returning fluxes, and this estimate requires the thermal regime examined next.
Finite-energy emission and the thermal regime
Slow variation in time does not ensure that one emission produces only a small change in the black hole state. A canonical thermal treatment also requires this latter condition [23 ] . We examine it on the neutral branch, where M ′ = M e + Δ E \mathcal{M}' = \mathcal{M}_{\mathrm{e}} + \Delta\mathcal{E} M ′ = M e + Δ E and T H = A T Δ E + O ( Δ E ) T_{\mathrm{H}} = A_{\mathrm{T}}\sqrt{\Delta\mathcal{E}} + O(\Delta\mathcal{E}) T H = A T Δ E + O ( Δ E ) , with K = 1 K = 1 K = 1 and n = 1 n = 1 n = 1 in Eq. (108) . Writing S e S_{\mathrm{e}} S e for the limiting entropy, the first law and the fixed-charge heat capacity give
S − S e = 2 Δ E A T + O ( Δ E ) , S-S_{\mathrm{e}}=\frac{2\sqrt{\Delta\mathcal{E}}}{A_{\mathrm{T}}}+O(\Delta\mathcal{E}), S − S e = A T 2 Δ E + O ( Δ E ) ,
C Q = 2 Δ E A T + O ( Δ E ) . (115) C_{Q}=\frac{2\sqrt{\Delta\mathcal{E}}}{A_{\mathrm{T}}}+O(\Delta\mathcal{E}). \tag*{(115)} C Q = A T 2 Δ E + O ( Δ E ) . (115)
Consider one additional neutral emission of Killing energy 0 < ω ≤ Δ E 0<\omega\leq\Delta\mathcal{E} 0 < ω ≤ Δ E , leaving a gap Δ E − ω \Delta\mathcal{E}-\omega Δ E − ω on the tracked outer branch. The same ω \omega ω denotes the angular frequency conjugate to the normalized time t t t . The leading KPW exponent is ln Γ K P W ( ω ) ≃ Δ S ω \ln\Gamma_{\mathrm{KPW}}(\omega)\simeq\Delta S_{\omega} ln Γ KPW ( ω ) ≃ Δ S ω , where
Δ S ω : = S ( M ′ − ω , Q ) − S ( M ′ , Q ) = 2 A T [ Δ E − ω − Δ E ] + O ( ω ) . (116) \begin{aligned}
\Delta S_{\omega}&:=S(\mathcal{M}'-\omega,Q)-S(\mathcal{M}',Q)\\
&=\frac{2}{A_{\mathrm{T}}}\left[\sqrt{\Delta\mathcal{E}-\omega}-\sqrt{\Delta\mathcal{E}}\right]+O(\omega). \tag*{(116)}
\end{aligned} Δ S ω := S ( M ′ − ω , Q ) − S ( M ′ , Q ) = A T 2 [ Δ E − ω − Δ E ] + O ( ω ) . (116)
The remainder follows from the regular next term in the entropy expansion.
For this neutral NLED branch, the same square-root expression also gives a bound on the full entropy exponent. Equation (71) shows that second and third derivatives of the mass function are positive for r ≥ r min r\geq r_{\min} r ≥ r m i n . Since ∂ r m ( r min ) = 0 \partial_{r}m(r_{\min})=0 ∂ r m ( r m i n ) = 0 , integration of ∂ r [ ( ∂ r m ) 2 ] = 2 ( ∂ r m ) ( ∂ r 2 m ) \partial_{r}[(\partial_{r}m)^{2}]=2(\partial_{r}m)(\partial_{r}^{2}m) ∂ r [( ∂ r m ) 2 ] = 2 ( ∂ r m ) ( ∂ r 2 m ) gives
[ ∂ r m ( r ) ] 2 < 2 [ m ( r ) − M e ] ∂ r 2 m ( r ) , r > r min . [\partial_{r}m(r)]^{2}<2[m(r)-\mathcal{M}_{\mathrm{e}}]\partial_{r}^{2}m(r),\qquad r>r_{\min}. [ ∂ r m ( r ) ] 2 < 2 [ m ( r ) − M e ] ∂ r 2 m ( r ) , r > r m i n .
Together with T H = κ ∂ r m / ( 4 π 2 ) T_{\mathrm{H}}=\kappa\partial_{r}m/(4\pi^{2}) T H = κ ∂ r m / ( 4 π 2 ) , this means that T H / Δ E T_{\mathrm{H}}/\sqrt{\Delta\mathcal{E}} T H / Δ E increases with the mass above coalescence. Integrating the fixed-charge first law from M ′ − ω \mathcal{M}'-\omega M ′ − ω to M ′ \mathcal{M}' M ′ therefore yields
− Δ S ω > ω T H 2 1 + 1 − ω / Δ E , 0 < ω ≤ Δ E . (117) -\Delta S_{\omega}>\frac{\omega}{T_{\mathrm{H}}}\frac{2}{1+\sqrt{1-\omega/\Delta\mathcal{E}}},\qquad0<\omega\leq\Delta\mathcal{E}. \tag*{(117)} − Δ S ω > T H ω 1 + 1 − ω /Δ E 2 , 0 < ω ≤ Δ E . (117)
Here T H T_{\mathrm{H}} T H is evaluated before the emission, and ω = Δ E \omega=\Delta\mathcal{E} ω = Δ E is understood as a limit through simple outer horizons. The bound is saturated at leading order in the endpoint expansion. At that order, the ratio of the suppression exponent to its thermal estimate increases from one to two as the fraction ω / Δ E \omega/\Delta\mathcal{E} ω /Δ E increases from zero to one.
For ω ≪ Δ E \omega\ll\Delta\mathcal{E} ω ≪ Δ E , the thermal expansion (22) , evaluated at the current state, has linear term − ω / T H -\omega/T_{\mathrm{H}} − ω / T H and quadratic term − ω 2 / ( 2 C Q T H 2 ) -\omega^{2}/(2C_{Q}T_{\mathrm{H}}^{2}) − ω 2 / ( 2 C Q T H 2 ) . The relative size of the quadratic correction is
ω 2 C Q T H = ω 4 Δ E [ 1 + O ( Δ E ) ] . (118) \frac{\omega}{2C_{Q}T_{\mathrm{H}}}=\frac{\omega}{4\Delta\mathcal{E}}\left[1+O(\sqrt{\Delta\mathcal{E}})\right]. \tag*{(118)} 2 C Q T H ω = 4Δ E ω [ 1 + O ( Δ E ) ] . (118)
For radiation with characteristic energies ω ∼ T H \omega\sim T_{\mathrm{H}} ω ∼ T H , the ratio ω / Δ E \omega/\Delta\mathcal{E} ω /Δ E measures the fraction of the remaining energy removed by one quantum. The thermal approximation therefore requires
T H Δ E ≪ 1 , or equivalently Δ E ≫ A T 2 . (119) \frac{T_{\mathrm{H}}}{\Delta\mathcal{E}}\ll1,\qquad\text{or equivalently}\qquad\Delta\mathcal{E}\gg A_{\mathrm{T}}^{2}. \tag*{(119)} Δ E T H ≪ 1 , or equivalently Δ E ≫ A T 2 . (119)
For the neutral solution the scale on the right is explicit,
A T 2 = κ ( 1 + w − 1 ) 4 π 3 ℓ 2 w − 1 . (120) A_T^2 = \frac{\kappa(1+w_{-1})}{4\pi^3\ell^2w_{-1}}. \tag*{(120)} A T 2 = 4 π 3 ℓ 2 w − 1 κ ( 1 + w − 1 ) . (120)
At fixed couplings, w − 1 w_{-1} w − 1 is independent of charge by Eq. (72) . The leading energy scale A T 2 A_T^2 A T 2 is therefore the same at different charges, even though the critical mass and entropy scale quadratically and linearly with the charge magnitude, respectively. This follows from the charge-independent radial curvature at the nondegenerate minimum.
Due to Eq. (115) , the thermal window is also the regime C Q ∼ S − S e ≫ 1 C_Q \sim S-S_{\mathrm e}\gg1 C Q ∼ S − S e ≫ 1 . The entropy excess, rather than the possibly large limiting entropy, controls the change caused by a typical quantum. To use the endpoint expansion in this regime, the higher radial terms in the mass–radius relation must remain small as well. Equation (71) gives the ratio of its cubic to quadratic term as ( S − S e ) / [ 3 S e ∣ 1 + w − 1 ∣ ] (S-S_{\mathrm e})/[3S_{\mathrm e}|1+w_{-1}|] ( S − S e ) / [ 3 S e ∣1 + w − 1 ∣ ] , where S − S e = 4 π 2 ( R + − r min ) / κ S-S_{\mathrm e}=4\pi^2(R_+-r_{\min})/\kappa S − S e = 4 π 2 ( R + − r m i n ) / κ is exact. The two requirements consequently give
1 ≪ S − S e ≪ S e ∣ 1 + w − 1 ∣ . (121) 1\ll S-S_{\mathrm e}\ll S_{\mathrm e}|1+w_{-1}|. \tag*{(121)} 1 ≪ S − S e ≪ S e ∣1 + w − 1 ∣. (121)
This competition can also be expressed in energy. Define the scale associated with the quadratic endpoint expansion by
E g e o m : = A T 2 4 [ S e ∣ 1 + w − 1 ∣ ] 2 = 2 π κ q 2 ℓ 2 γ 2 ∣ w − 1 ∣ ∣ 1 + w − 1 ∣ 3 . (122) \mathcal{E}_{\mathrm{geom}}:=\frac{A_T^2}{4}\left[S_{\mathrm e}|1+w_{-1}|\right]^2=\frac{2\pi\kappa q^2\ell^2}{\gamma^2}|w_{-1}||1+w_{-1}|^3. \tag*{(122)} E geom := 4 A T 2 [ S e ∣1 + w − 1 ∣ ] 2 = γ 2 2 π κ q 2 ℓ 2 ∣ w − 1 ∣∣1 + w − 1 ∣ 3 . (122)
It is a characteristic scale, not a sharp boundary. Within the endpoint expansion, Eq. (121) is equivalent parametrically to
A T 2 ≪ Δ E ≪ E g e o m . (123) A_T^2\ll\Delta\mathcal{E}\ll\mathcal{E}_{\mathrm{geom}}. \tag*{(123)} A T 2 ≪ Δ E ≪ E geom . (123)
This interval requires S e ∣ 1 + w − 1 ∣ ≫ 1 S_{\mathrm e}|1+w_{-1}|\gg1 S e ∣1 + w − 1 ∣ ≫ 1 . Increasing the charge at fixed couplings raises its upper scale as q 2 q^2 q 2 , while leaving its lower scale unchanged.
The approach to triple degeneracy has the opposite effect. Writing d : = ∣ 1 + w − 1 ∣ ∈ ( 0 , 1 ) d:=|1+w_{-1}|\in(0,1) d := ∣1 + w − 1 ∣ ∈ ( 0 , 1 ) in the positive-mass sector (77) , the Lambert relation gives κ β ℓ 2 = e d / ( 1 + d ) \kappa\beta\ell^2=e^d/(1+d) κ β ℓ 2 = e d / ( 1 + d ) , so d ∼ 2 ( κ β ℓ 2 − 1 ) d\sim\sqrt{2(\kappa\beta\ell^2-1)} d ∼ 2 ( κ β ℓ 2 − 1 ) as κ β ℓ 2 → 1 + \kappa\beta\ell^2\to1^+ κ β ℓ 2 → 1 + . The two scales then behave as
A T 2 ∼ κ 4 π 3 ℓ 2 d , E g e o m ∼ 2 π κ q 2 ℓ 2 γ 2 d 3 . (124) \begin{aligned}
A_T^2&\sim\frac{\kappa}{4\pi^3\ell^2}d,\\
\mathcal{E}_{\mathrm{geom}}&\sim\frac{2\pi\kappa q^2\ell^2}{\gamma^2}d^3. \tag*{(124)}
\end{aligned} A T 2 E geom ∼ 4 π 3 ℓ 2 κ d , ∼ γ 2 2 π κ q 2 ℓ 2 d 3 . (124)
At fixed ( q , κ , ℓ , γ ) (q,\kappa,\ell,\gamma) ( q , κ , ℓ , γ ) , the upper scale therefore contracts as d 3 d^3 d 3 , faster than the thermal scale, which decreases as d d d . A large limiting entropy alone does not ensure that the two approximations overlap near triple degeneracy. The conclusion concerns the quadratic double-root description. In the triple-zero regime the cubic term must be retained.
A luminosity computed from a thermal spectrum must be reconsidered when typical quanta carry a substantial fraction of the remaining energy. Here Δ E \Delta\mathcal{E} Δ E bounds emissions that stay on the tracked outer branch. Smaller masses still admit a different event horizon. The adiabaticity condition and the finite-energy condition thus test the variation in time and the change of state during one emission, respectively.
The distinction can be seen in the neutral marginal case, with n = 1 n = 1 n = 1 and ν = 3 / 2 \nu= 3/2 ν = 3/2 . Taking L r a d ≃ L 0 ( Δ E ) 3 / 2 L_{\mathrm{rad}} \simeq L_0(\Delta\mathcal{E})^{3/2} L rad ≃ L 0 ( Δ E ) 3/2 as an averaged loss law, the leading solution reads
Δ E ( t ) ≃ [ ( Δ E 0 ) − 1 / 2 + L 0 2 ( t − t 0 ) ] − 2 , T H ( t ) ≃ A T ( Δ E 0 ) − 1 / 2 + L 0 ( t − t 0 ) / 2 . (125) \begin{aligned}
\Delta\mathcal{E}(t) &\simeq\left[(\Delta\mathcal{E}_0)^{-1/2} + \frac{L_0}{2}(t-t_0)\right]^{-2},\\
T_{\mathrm{H}}(t) &\simeq\frac{A_T}{(\Delta\mathcal{E}_0)^{-1/2}+L_0(t-t_0)/2}. \tag*{(125)}
\end{aligned} Δ E ( t ) T H ( t ) ≃ [ ( Δ E 0 ) − 1/2 + 2 L 0 ( t − t 0 ) ] − 2 , ≃ ( Δ E 0 ) − 1/2 + L 0 ( t − t 0 ) /2 A T . (125)
At this order, the adiabaticity parameter is constant, ϵ a d ≃ L 0 / ( 4 π A T ) \epsilon_{\mathrm{ad}} \simeq L_0/(4\pi A_T) ϵ ad ≃ L 0 / ( 4 π A T ) , and the slow-variation condition requires this ratio to be small. The horizon separation and entropy excess both decrease as t − 1 t^{-1} t − 1 , so S e S_{\mathrm{e}} S e is approached only at infinite time along this formal history.
For this marginal history to begin in a controlled thermal endpoint regime, its initial gap must satisfy Eq. (123) . As Δ E \Delta\mathcal{E} Δ E decreases, the endpoint expansion improves, but the thermal approximation eventually fails even if ϵ a d \epsilon_{\mathrm{ad}} ϵ ad stays small. Continuing a luminosity derived from a thermal spectrum beyond that range requires a treatment of finite-energy emission. The entropy exponent, the time to coalescence, and the validity of a thermal spectrum thus require separate information. The stationary family fixes the first and constrains the other two once a radiation law is supplied. The complementary one-horizon sector considered next allows us to examine the local temperature and the finite-energy exponent away from these horizon mergers.
Horizon emission from the quantum-backreacted BTZ geometry
Finally, a complementary application of particular interest is provided by exploring the one-horizon sector under a suitable choice of parameters. This subfamily incorporates the leading quantum-backreacted BTZ geometry coupled to NMG. Accordingly, we provide a comparison of the neutral probe horizon factor before following finite-energy emission in the Einstein–NLED realization. Corrections to the BTZ horizon and temperature from the backreaction of conformally coupled scalars were studied in Einstein gravity by Martínez and Zanelli [46 ] , where the blackening function acquires a 1 / r 1/r 1/ r term. In NMG, the solution obtained in Ref. [41] at leading order in the backreaction of N ≫ 1 N \gg1 N ≫ 1 conformal scalar fields has
f s c ( r ; M ) = r 2 ℓ 2 − M + α s c ( M ) r ln ( r e ℓ ) , (126) f_{\mathrm{sc}}(r;M)=\frac{r^2}{\ell^2}-M+\alpha_{\mathrm{sc}}(M)r\ln\left(\frac{r}{e\ell}\right), \tag*{(126)} f sc ( r ; M ) = ℓ 2 r 2 − M + α sc ( M ) r ln ( e ℓ r ) , (126)
α s c ( M ) : = 2 N ℓ p F s c ( M ) M ℓ 2 > 0. (127) \alpha_{\mathrm{sc}}(M):=\frac{2N\ell_{\mathrm{p}}F_{\mathrm{sc}}(M)}{M\ell^2}>0. \tag*{(127)} α sc ( M ) := M ℓ 2 2 N ℓ p F sc ( M ) > 0. (127)
Here M > 0 M>0 M > 0 is the seed BTZ mass parameter and ℓ p = 8 π G 3 \ell_{\mathrm{p}}=8\pi G_3 ℓ p = 8 π G 3 . The positive, dimensionless function F s c ( M ) F_{\mathrm{sc}}(M) F sc ( M ) enters the scalar expectation value ⟨ T ν μ ⟩ \langle T^\mu_{\ \nu}\rangle ⟨ T ν μ ⟩ with transparent boundary conditions at infinity. Matching Eq. (55) at a given M M M requires [19 ]
q γ = − α s c ( M ) 2 2 κ , β = − α s c ( M ) 2 e κ ℓ . (128) \frac{q}{\gamma}=-\frac{\alpha_{\mathrm{sc}}(M)}{2\sqrt{2}\kappa},\qquad\beta=-\frac{\alpha_{\mathrm{sc}}(M)}{2e\kappa\ell}. \tag*{(128)} γ q = − 2 2 κ α sc ( M ) , β = − 2 e κ ℓ α sc ( M ) . (128)
For a neutral minimally coupled test scalar, the same normalization of ∂ t \partial_t ∂ t in the two realizations gives the same leading WKB pole. It can be evaluated on the corrected mean geometry without using a gravitational entropy formula. Denote the corrected event horizon radius by r h r_{\mathrm{h}} r h , and introduce
r B : = ℓ M , ε q ( M ) : = α s c ( M ) ℓ 2 M = N ℓ p F s c ( M ) ℓ M 3 / 2 , r_{\mathrm{B}}:=\ell\sqrt{M},\qquad\varepsilon_q(M):=\frac{\alpha_{\mathrm{sc}}(M)\ell}{2\sqrt{M}}=\frac{N\ell_{\mathrm{p}}F_{\mathrm{sc}}(M)}{\ell M^{3/2}}, r B := ℓ M , ε q ( M ) := 2 M α sc ( M ) ℓ = ℓ M 3/2 N ℓ p F sc ( M ) ,
where r B r_{\mathrm{B}} r B is the uncorrected BTZ radius. Expanding the horizon condition about this simple root recovers the displacement obtained in Ref. [41 ] ,
r h r B = 1 + ε q ( 1 − ln M ) + O ( ε q 2 ) . (129) \frac{r_{\mathrm{h}}}{r_{\mathrm{B}}}=1+\varepsilon_{\mathrm{q}}(1-\ln\sqrt{M})+\mathcal{O}(\varepsilon_{\mathrm{q}}^{2}). \tag*{(129)} r B r h = 1 + ε q ( 1 − ln M ) + O ( ε q 2 ) . (129)
The remainder is an expansion at fixed M M M and ℓ \ell ℓ . Small relative corrections to the horizon radius and temperature require
ε q ∣ 1 − ln M ∣ ≪ 1 , ε q ≪ 1. \varepsilon_{\mathrm{q}}\left|1-\ln\sqrt{M}\right|\ll1,\qquad\varepsilon_{\mathrm{q}}\ll1. ε q 1 − ln M ≪ 1 , ε q ≪ 1.
The temperature must be evaluated at the displaced horizon. The radial derivative of Eq. (126) is 2 r / ℓ 2 + α s c ln ( r / ℓ ) 2r/\ell^{2}+\alpha_{\mathrm{sc}}\ln(r/\ell) 2 r / ℓ 2 + α sc ln ( r / ℓ ) , so the first-order change in its horizon value is
4 π δ T H = 2 ( r h − r B ) ℓ 2 + α s c ln ( r B ℓ ) = α s c . 4\pi\delta T_{\mathrm{H}}=\frac{2(r_{\mathrm{h}}-r_{\mathrm{B}})}{\ell^{2}}+\alpha_{\mathrm{sc}}\ln\left(\frac{r_{\mathrm{B}}}{\ell}\right)=\alpha_{\mathrm{sc}}. 4 π δ T H = ℓ 2 2 ( r h − r B ) + α sc ln ( ℓ r B ) = α sc .
The logarithms cancel between the displacement and the explicit change in the radial slope. Writing T s c T_{\mathrm{sc}} T sc for the corrected Hawking temperature, we obtain
T s c T B = 1 + ε q + O ( ε q 2 ) , T B : = M 2 π ℓ . (130) \frac{T_{\mathrm{sc}}}{T_{\mathrm{B}}}=1+\varepsilon_{\mathrm{q}}+\mathcal{O}(\varepsilon_{\mathrm{q}}^{2}),\qquad T_{\mathrm{B}}:=\frac{\sqrt{M}}{2\pi\ell}. \tag*{(130)} T B T sc = 1 + ε q + O ( ε q 2 ) , T B := 2 π ℓ M . (130)
The correction to the temperature is positive at fixed seed mass and Killing-time normalization. For M > e 2 M>e^{2} M > e 2 the horizon correction in Eq. (129) is negative, so the quantum-backreacted horizon is smaller than the BTZ horizon and nevertheless hotter.
For a neutral excitation of Killing energy ω > 0 \omega>0 ω > 0 , treated here without its own backreaction, the leading horizon emission-to-absorption ratio is R H ( ω ) ≃ exp ( − ω / T H ) \mathcal{R}_{\mathrm{H}}(\omega)\simeq\exp(-\omega/T_{\mathrm{H}}) R H ( ω ) ≃ exp ( − ω / T H ) [36 ] . Equations (127) and (130) then give
ln ( R H s c ( ω ) R H B T Z ( ω ) ) = 2 π N ℓ p F s c ( M ) M 2 ω + O ( ω T B ε q 2 ) . (131) \ln\left(\frac{\mathcal{R}_{\mathrm{H}}^{\mathrm{sc}}(\omega)}{\mathcal{R}_{\mathrm{H}}^{\mathrm{BTZ}}(\omega)}\right)=\frac{2\pi N\ell_{\mathrm{p}}F_{\mathrm{sc}}(M)}{M^{2}}\omega+\mathcal{O}\left(\frac{\omega}{T_{\mathrm{B}}}\varepsilon_{\mathrm{q}}^{2}\right). \tag*{(131)} ln ( R H BTZ ( ω ) R H sc ( ω ) ) = M 2 2 π N ℓ p F sc ( M ) ω + O ( T B ω ε q 2 ) . (131)
The correction enhances the horizon emission factor relative to BTZ at the same M M M and ω \omega ω in both matched geometries. This comparison retains the leading WKB action on the background corrected to first order in N ℓ p N\ell_{\mathrm{p}} N ℓ p .
Emission at finite energy additionally probes the path followed through the family. In the semiclassical NMG solution, the source fixes α s c \alpha_{\mathrm{sc}} α sc as a function of the seed mass. By Eq. (128) , this variation differs from neutral emission at fixed NLED couplings, since fixed β \beta β fixes α s c \alpha_{\mathrm{sc}} α sc . The finite-energy NMG calculation also requires its own Hamiltonian energy and entropy, including the quantum source 2 . For the Einstein–NLED realization, the first law at fixed couplings derived in Appendix § 12 .§ 5 allows an explicit finite-energy comparison.
Match the NLED solution at an initial seed mass M 0 M_{0} M 0 and write α 0 : = α s c ( M 0 ) \alpha_{0}:=\alpha_{\mathrm{sc}}(M_{0}) α 0 := α sc ( M 0 ) . Neutral emission keeps α 0 \alpha_{0} α 0 and the logarithmic reference scale fixed. The final mass parameter is
M f = M 0 − κ E π > 0 , M_{\mathrm{f}}=M_{0}-\frac{\kappa\mathcal{E}}{\pi}>0, M f = M 0 − π κ E > 0 ,
and the final blackening function is r 2 / ℓ 2 − M f + α 0 r ln [ r / ( e ℓ ) ] r^2/\ell^2-M_f+\alpha_0 r\ln[r/(e\ell)] r 2 / ℓ 2 − M f + α 0 r ln [ r / ( e ℓ )] . Applying the horizon expansion at M 0 M_0 M 0 and M f M_f M f and using the Einstein entropy in Eq. (68) gives ln Γ N L E D ≃ Δ S N L E D \ln\Gamma_{\mathrm{NLED}}\simeq\Delta S_{\mathrm{NLED}} ln Γ NLED ≃ Δ S NLED , with
Δ S N L E D = 4 π 2 ℓ M 0 κ [ 1 − κ E π M 0 − 1 ] − 2 π 2 N ℓ p F s c ( M 0 ) κ M 0 ln ( 1 − κ E π M 0 ) + O [ ε q ( M 0 ) 2 ] . (132) \Delta S_{\mathrm{NLED}}=\frac{4\pi^2\ell\sqrt{M_0}}{\kappa}\left[\sqrt{1-\frac{\kappa\mathcal{E}}{\pi M_0}}-1\right]-\frac{2\pi^2N\ell_{\mathrm{p}}F_{\mathrm{sc}}(M_0)}{\kappa M_0}\ln\left(1-\frac{\kappa\mathcal{E}}{\pi M_0}\right)+\mathcal{O}\left[\varepsilon_{\mathrm{q}}(M_0)^2\right]. \tag*{(132)} Δ S NLED = κ 4 π 2 ℓ M 0 [ 1 − π M 0 κ E − 1 ] − κ M 0 2 π 2 N ℓ p F sc ( M 0 ) ln ( 1 − π M 0 κ E ) + O [ ε q ( M 0 ) 2 ] . (132)
This expression retains the full dependence on emitted energy at first order in the initial background correction. The first term is the finite-energy BTZ exponent. The second is positive for 0 < E < π M 0 / κ 0<\mathcal{E}<\pi M_0/\kappa 0 < E < π M 0 / κ and expresses the correction in the quantum parameters of the initial matched geometry. Its small E \mathcal{E} E limit reproduces − E / T s c ( M 0 ) -\mathcal{E}/T_{\mathrm{sc}}(M_0) − E / T sc ( M 0 ) to the same perturbative order. The remainder is taken at fixed M 0 M_0 M 0 and E \mathcal{E} E .
Along the fixed-α 0 \alpha_0 α 0 continuation, ∂ r f = 2 / ℓ 2 + α 0 / r > 0 \partial_r f=2/\ell^2+\alpha_0/r>0 ∂ r f = 2/ ℓ 2 + α 0 / r > 0 , so the temperature decreases as the event horizon shrinks. The exact NLED exponent consequently obeys Δ S N L E D < − E / T H ( 0 ) \Delta S_{\mathrm{NLED}}<-\mathcal{E}/T_{\mathrm{H}}(0) Δ S NLED < − E / T H ( 0 ) for a nonzero emission while this horizon remains simple. The initial temperature T H ( 0 ) T_{\mathrm{H}}(0) T H ( 0 ) agrees with T s c ( M 0 ) T_{\mathrm{sc}}(M_0) T sc ( M 0 ) at the retained order. The positive first-order background correction thus raises the exponent relative to the BTZ emission result, while the exact NLED result remains below its own initial thermal tangent.
The expansion requires small relative corrections to both the temperature and the horizon radius,
α 0 ℓ 2 M ′ ≪ 1 , α 0 ℓ 2 M ′ ∣ 1 − ln M ′ ∣ ≪ 1 , \frac{\alpha_0\ell}{2\sqrt{M'}}\ll1,\qquad\frac{\alpha_0\ell}{2\sqrt{M'}}\left|1-\ln\sqrt{M'}\right|\ll1, 2 M ′ α 0 ℓ ≪ 1 , 2 M ′ α 0 ℓ 1 − ln M ′ ≪ 1 ,
throughout M f ≤ M ′ ≤ M 0 M_f\le M'\le M_0 M f ≤ M ′ ≤ M 0 . The logarithmic correction in Eq. (132) cannot be continued to M f = 0 M_f=0 M f = 0 using this expansion.
Within its domain of validity, the comparison separates the part of horizon emission fixed by the shared geometry from the additional input needed for a finite change of state. The corrected temperature determines the neutral-probe horizon factor in both realizations. The finite-energy exponent computed here follows the Einstein–NLED family at fixed couplings and its related first law. The matching at the initial geometry does not make it the evaporation exponent of the semiclassical NMG family.
Discussion and concluding remarks
The KPW approach to Hawking radiation has found extensive application in black hole physics by incorporating energy conservation into the semiclassical description of emission. Revisiting the horizon condition within this framework reveals how the accompanying loss of mass and charge can affect the horizon locations, their number and causal nature. For the analytic static spacetimes of the form (1) , wherever the local mass inversion is well defined, the Lagrange–Bürmann formula and its extension through Puiseux series provide local expressions for the motion of simple and degenerate roots without requiring an explicit solution for the horizon radii. Furthermore, composing this inverse with the appropriate gravitational entropy determines the leading KPW exponent, provided the stationary first law holds along the emission path. At the same time, the coefficient formulas developed in § 2 and Appendices § 9 and § 11 make the corrections to the thermal exponent accessible to arbitrary order.
A natural extension of this construction arises when a horizon becomes degenerate, since an emission can then change the number of nearby real roots. The corresponding critical curve consists of the mass and charge values for which the double horizon exists. The sign of the mass difference from this curve, together with the radial curvature of the mass function, determines whether two simple horizons emerge or the nearby real pair is absent. The charge emission profile is therefore an essential part of the description. Once the pair exists, its separation is proportional to the square root of the magnitude of this mass difference, giving a square-root dependence on the emitted energy for a transverse departure and a linear dependence for a quadratic one. The double horizon itself is preserved along paths that remain on the critical curve throughout.
We found a particularly well-suited case study in the recently presented logarithmic deformation of the BTZ black hole [19 ] , which realizes these possibilities within a single matter theory. Starting with a double inner horizon and a separate simple event horizon, we show that neutral emission necessarily splits the former into two simple horizons, revealing an internal horizon fine structure. The behavior is richer for charged channels, which allow for immediate splitting, delayed recovery of an inner pair, or persistent inner degeneracy, while the event horizon remains simple. Here the analyticity of the charge emission profile has a useful consequence: the outer radius and entropy exponent remain locally analytic through these internal transitions, so the creation or loss of an inner pair produces no nonanalytic marker in the leading horizon emission factor. An instructive comparison is given in Figure 5 , where we present one channel that preserves the double inner horizon, while another separates it into two simple horizons and later brings them together again. Both connect the same initial and final states and consequently have the same leading KPW exponent, despite their different intermediate horizon configurations.
Through a formal continuation of the process, we find that the larger of the newly separated inner horizons can coalesce with the event horizon, driving the sequence towards an extremal black hole. Its temperature then tends to zero but, surprisingly, the entropy difference and hence the leading tunneling exponent have a finite limit. For charged approaches tangent to the outer critical curve, retaining the electric work in the residue is essential to recover this behavior. Smaller positive masses below the coalescence value still admit the continuation of the innermost root as a simple event horizon. This critical mass therefore ends the tracked outer branch, with a transition to the smaller horizon requiring a dynamical description.
This approach towards extremality has an interesting counterpart in Ref. [47 ] . The fully nonlinear evolutions of the four-dimensional Einstein–Maxwell–charged-scalar system produce near-extremal black holes from regular, horizon-free initial data, with this character independently identified through radiation at future null infinity. There, the near-extremal regime emerges during formation, whereas in our case Hawking emission can drive an already formed black hole towards a degenerate event horizon. A common feature of the two settings is the role of the relative loss of mass and charge. In their evolutions these fluxes are determined dynamically, while the emission profiles considered here make their influence on the horizon motion explicit.
The prospect of reaching extremality naturally raises the question of how long this approach takes, since the entropy exponent does not specify the interval between successive emissions. To that end, the energy and charge balance examined in § 5.2 relates the charge emission profile to the ratio of charge and energy fluxes, while the luminosity determines how quickly the path is traversed. A sufficiently rapid decrease of the luminosity makes the approach asymptotic in time. Adiabaticity imposes a stronger restriction, however, because the inverse surface gravity also diverges and the geometry must evolve slowly on this increasing time scale. More generally, the cooling bound in Eq. (114) excludes reaching zero temperature in finite boundary time while a uniform bound on adiabaticity remains valid, independently of the particular luminosity law.
Interestingly, even such a slow approach need not remain thermal. A quantum with energy of order T H T_{\mathrm H} T H can eventually carry away an appreciable fraction of the energy remaining before coalescence, so that the temperature can change substantially during a single emission. For neutral emission, this effect is captured by the exact bound (117) , which quantifies the suppression relative to the Boltzmann estimate without truncating the entropy change at quadratic order. Near extremality, the available thermal reservoir is measured by the entropy above the limiting configuration, rather than by the total entropy [23 ] . Requiring this entropy excess to be large while remaining within the quadratic radial expansion gives the overlap interval in Eq. (123) . This interval widens with increasing charge magnitude at fixed couplings, but contracts towards triple degeneracy. In the latter case, the geometric upper scale decreases faster than the thermal lower scale. To examine the thermal regime near a triple horizon, the mass–radius relation must therefore be expanded through cubic order in the displacement from the critical radius.
Our final application concerns the one-horizon sector, which reproduces the quantum-backreacted BTZ geometry of NMG [41 ] . It offers an opportunity to examine Hawking emission from a black hole whose geometry already incorporates a semiclassical effect. At the matched background and with the same normalization of time, both realizations give the same temperature and leading horizon emission factor for a neutral minimally coupled probe. At fixed seed mass, the leading semiclassical correction raises the temperature, even when the corrected horizon is smaller than the BTZ horizon. Following a finite emission in the Einstein–NLED realization at fixed couplings gives, in addition, a positive logarithmic correction to the BTZ entropy exponent. The resulting exponent recovers the corrected temperature in the limit of small emitted energy. This enhancement relative to BTZ coexists with suppression relative to the initial thermal estimate of the corrected black hole, since the NLED configuration cools as energy is extracted. These results apply within the perturbative regime specified in § 6 . Following the family obtained through semiclassical backreaction in NMG [41 ] would instead change the quantum source with the mass and require the energy and entropy of that theory for the finite-emission calculation.
An interesting question is what happens if the initial state is already an extremal black hole. In the family studied here, neutral mass loss from the outer critical curve removes the double outer horizon, leaving the previously innermost root as a simple event horizon. A charged path along that curve, on the other hand, preserves extremality within the stationary family as the mass and charge magnitude decrease. To recover two distinct outer horizons, the remaining mass must lie above the critical mass at the new charge. A transverse departure into this region requires negative horizon energy at leading order, outside the ordinary channels considered here, while tangential departures must be decided at higher orders. The role of this energy–charge balance recalls classical absorption tests of cosmic censorship [48 , 49 ] , with the distinction that the smaller horizon survives here and the issue is a transition between horizon configurations. Determining a physical emission rate for any of these paths calls for a calculation of tunneling directly from the degenerate state [39 ] .
One of the main remaining questions is which of these stationary histories can be realized during evaporation. This requires the radiation spectrum and an analysis of the exterior’s linear stability. Backscattering in the exterior, or reflection at the AdS boundary, can return energy and, for charged radiation, charge to the hole, changing the net fluxes that select the emission path. Alongside this scattering problem, the coupled gravitational and NLED perturbations would determine whether growing modes can disrupt the slow evolution assumed here. Charged-scalar studies of other three-dimensional AdS black holes illustrate the dependence of superradiance and mode stability on the matter model and boundary conditions [50 , 51 ] .
The fate of the internal horizon fine structure also depends on how these perturbations evolve inside the black hole. The mass inflation mechanism of Poisson and Israel [52 ] and the analysis of strong cosmic censorship for linearly charged BTZ [53 ] make this a necessary part of the dynamical problem. Semiclassical effects can also destabilize an inner horizon with vanishing surface gravity, as shown for stationary regular black holes in Ref. [54 ] . The role of time dependence is illustrated by the finite renormalized stress tensor obtained for trapped regions of finite lifetime in Ref. [55 ] , within the s -wave Polyakov approximation. Bringing these effects into the present family would help determine which features of its horizon fine structure persist during evaporation.
Finally, a natural extension of this study is to account for non-static spacetimes. Recent calculations of near-extremal Kerr–Newman evaporation at fixed electric charge show how the competition between s -wave and superradiant emission, with quantum corrections included, can slow the late-time evolution [56 ] . Closer to our setting, generalized Chern–Simons interactions offer a way to support charged rotating geometries by supplying force-free electromagnetic currents without contributing directly to the stress tensor [57 , 58 ] . This approach accommodates the original charged rotating BTZ solution and extends to nonlinear electrodynamics, with further applications to rotating AdS and Lifshitz black holes [59 , 60 ] . These constructions suggest a concrete route to rotating counterparts of the family studied here through an enlargement of its matter sector, allowing angular-momentum extraction to join the loss of mass and charge in selecting paths through the critical configurations. A separate extension to genuinely time-dependent geometries can build on the tunneling description of evolving trapping horizons in spherical symmetry [36 ] . In both directions, an appealing question is whether the radiation itself favors histories that preserve horizon degeneracy or lead towards extremality.
Acknowledgments
The authors are thankful to Eloy Ayón-Beato and Borja Diez for their feedback on this work, and to Radouane Gannouji for useful discussions on recent results complementary to those presented here. MAAC was partially supported by the Pro-Mejoras program of Universidad Veracruzana. MB is supported by Proyecto Interno UCM, Línea Regular UCM-IN-25202 and FONDECYT 1262452. ER acknowledges encouragement from PRODEP-México, CA-UV-320: Algebra, Geometría y Gravitación. JAMZ was partially funded by the Estancias Posdoctorales por México program of SECIHTI. ER and JAMZ wer partially supported by the Sistema Nacional de Investigadoras e Investigadores (SNII), also under SECIHTI.
Lagrange–Bürmann formulae
The horizon and entropy coefficients in Eqs. (16) and (17) follow from the same inversion problem set by the horizon condition. We give its derivation in the notation of § 2 , keeping all parameters λ A \lambda^{A} λ A fixed. The analytic and formal versions of the LB construction are extensively discussed in Refs. [31 , 32 ] .
The inverse horizon map
Consider the function h h h defined in Eq. (14) . Its Maclaurin expansion about y = 0 y=0 y = 0 is
h ( y ) = − ∑ j = 1 ∞ ∂ r j m ( r + ) j ! y j , ∂ y h ( 0 ) = − ∂ r m ( r + ) ≠ 0. (133) h(y)=-\sum_{j=1}^{\infty}\frac{\partial_r^j m(r_+)}{j!}y^j,\qquad\partial_y h(0)=-\partial_r m(r_+)\ne0. \tag*{(133)} h ( y ) = − j = 1 ∑ ∞ j ! ∂ r j m ( r + ) y j , ∂ y h ( 0 ) = − ∂ r m ( r + ) = 0. (133)
If m m m is analytic near r + r_+ r + , the inverse function theorem gives a unique analytic solution y = Δ r + ( E ) y=\Delta r_+(\mathcal{E}) y = Δ r + ( E ) of E = h ( y ) \mathcal{E}=h(y) E = h ( y ) with Δ r + ( 0 ) = 0 \Delta r_+(0)=0 Δ r + ( 0 ) = 0 . The ratio y / h ( y ) y/h(y) y / h ( y ) has a removable singularity at the origin, with value − 1 / [ ∂ r m ( r + ) ] -1/[\partial_r m(r_+)] − 1/ [ ∂ r m ( r + )] .
Let g ( y ) g(y) g ( y ) be analytic in the same neighborhood, and denote by [ E n ] [\mathcal{E}^n] [ E n ] the coefficient of E n \mathcal{E}^n E n . For n ≥ 1 n\ge1 n ≥ 1 , a change of variable in the residue gives
[ E n ] g [ Δ r + ( E ) ] = R e s E = 0 g [ Δ r + ( E ) ] E n + 1 = R e s y = 0 g ( y ) ∂ y h ( y ) h ( y ) n + 1 = 1 n R e s y = 0 ∂ y g ( y ) h ( y ) n . (134) [\mathcal{E}^n]g[\Delta r_+(\mathcal{E})]=\mathop{\rm Res}_{\mathcal{E}=0}\frac{g[\Delta r_+(\mathcal{E})]}{\mathcal{E}^{n+1}}=\mathop{\rm Res}_{y=0}\frac{g(y)\partial_y h(y)}{h(y)^{n+1}}
=\frac{1}{n}\mathop{\rm Res}_{y=0}\frac{\partial_y g(y)}{h(y)^n}. \tag*{(134)} [ E n ] g [ Δ r + ( E )] = Res E = 0 E n + 1 g [ Δ r + ( E )] = Res y = 0 h ( y ) n + 1 g ( y ) ∂ y h ( y ) = n 1 Res y = 0 h ( y ) n ∂ y g ( y ) . (134)
The last equality follows because the residue of d [ g ( y ) / h ( y ) n ] / d y d[g(y)/h(y)^n]/dy d [ g ( y ) / h ( y ) n ] / d y vanishes. Extracting the coefficient of y − 1 y^{-1} y − 1 yields
g [ Δ r + ( E ) ] = g ( 0 ) + ∑ n = 1 ∞ E n n ! [ ∂ y n − 1 { ∂ y g ( y ) ( y h ( y ) ) n } ] y = 0 . (135) g[\Delta r_+(\mathcal{E})]=g(0)+\sum_{n=1}^{\infty}\frac{\mathcal{E}^n}{n!}\left[\partial_y^{n-1}\left\{\partial_y g(y)\left(\frac{y}{h(y)}\right)^n\right\}\right]_{y=0}. \tag*{(135)} g [ Δ r + ( E )] = g ( 0 ) + n = 1 ∑ ∞ n ! E n [ ∂ y n − 1 { ∂ y g ( y ) ( h ( y ) y ) n } ] y = 0 . (135)
Taking g ( y ) = y g(y)=y g ( y ) = y reproduces Eq. (16) . Taking instead g ( y ) = S ( r + + y , λ A ) g(y)=S(r_++y,\lambda^A) g ( y ) = S ( r + + y , λ A ) gives Eq. (17) , including the subtraction of the initial entropy through the term g ( 0 ) g(0) g ( 0 ) .
For a computation at a prescribed order, only finitely many radial derivatives of the mass function are needed. Direct coefficient extraction gives
a n = ( − 1 ) n n [ ∂ r m ( r + ) ] n [ y n − 1 ] [ 1 + ∑ j = 1 n − 1 ∂ r j + 1 m ( r + ) ( j + 1 ) ! ∂ r m ( r + ) y j ] − n , (136) a_n=\frac{(-1)^n}{n[\partial_r m(r_+)]^n}[y^{n-1}]\left[1+\sum_{j=1}^{n-1}\frac{\partial_r^{j+1}m(r_+)}{(j+1)!\partial_r m(r_+)}y^j\right]^{-n}, \tag*{(136)} a n = n [ ∂ r m ( r + ) ] n ( − 1 ) n [ y n − 1 ] [ 1 + j = 1 ∑ n − 1 ( j + 1 )! ∂ r m ( r + ) ∂ r j + 1 m ( r + ) y j ] − n , (136)
where the sum is understood to not contribute for n = 1 n=1 n = 1 . In particular,
a 1 = − 1 ∂ r m , a 2 = − ∂ r 2 m 2 ( ∂ r m ) 3 , a 3 = ( ∂ r m ) ( ∂ r 3 m ) − 3 ( ∂ r 2 m ) 2 6 ( ∂ r m ) 5 , (137) a_1=-\frac{1}{\partial_r m},\qquad a_2=-\frac{\partial_r^2m}{2(\partial_r m)^3},\qquad a_3=\frac{(\partial_r m)(\partial_r^3m)-3(\partial_r^2m)^2}{6(\partial_r m)^5}, \tag*{(137)} a 1 = − ∂ r m 1 , a 2 = − 2 ( ∂ r m ) 3 ∂ r 2 m , a 3 = 6 ( ∂ r m ) 5 ( ∂ r m ) ( ∂ r 3 m ) − 3 ( ∂ r 2 m ) 2 , (137)
a 4 = 10 ( ∂ r m ) ( ∂ r 2 m ) ( ∂ r 3 m ) − ( ∂ r m ) 2 ∂ r 4 m − 15 ( ∂ r 2 m ) 3 24 ( ∂ r m ) 7 . a_4=\frac{10(\partial_r m)(\partial_r^2m)(\partial_r^3m)-(\partial_r m)^2\partial_r^4m-15(\partial_r^2m)^3}{24(\partial_r m)^7}. a 4 = 24 ( ∂ r m ) 7 10 ( ∂ r m ) ( ∂ r 2 m ) ( ∂ r 3 m ) − ( ∂ r m ) 2 ∂ r 4 m − 15 ( ∂ r 2 m ) 3 .
with all derivatives evaluated at r + r_+ r + .
Entropy coefficients and specialization for affine mass dependence
Expanding the composed entropy to third order gives
b 1 = − ∂ r S ∂ r m , b_1=-\frac{\partial_r S}{\partial_r m}, b 1 = − ∂ r m ∂ r S ,
b 2 = ( ∂ r m ) ( ∂ r 2 S ) − ( ∂ r 2 m ) ( ∂ r S ) 2 ( ∂ r m ) 3 , b_2=\frac{(\partial_r m)(\partial_r^2S)-(\partial_r^2m)(\partial_rS)}{2(\partial_r m)^3}, b 2 = 2 ( ∂ r m ) 3 ( ∂ r m ) ( ∂ r 2 S ) − ( ∂ r 2 m ) ( ∂ r S ) ,
b 3 = ( ∂ r m ) ( ∂ r 3 m ) − 3 ( ∂ r 2 m ) 2 6 ( ∂ r m ) 5 ∂ r S + ( ∂ r 2 m ) ( ∂ r 2 S ) 2 ( ∂ r m ) 4 − ∂ r 3 S 6 ( ∂ r m ) 3 . (138) b_3=\frac{(\partial_r m)(\partial_r^3m)-3(\partial_r^2m)^2}{6(\partial_r m)^5}\partial_rS+\frac{(\partial_r^2m)(\partial_r^2S)}{2(\partial_r m)^4}-\frac{\partial_r^3S}{6(\partial_r m)^3}. \tag*{(138)} b 3 = 6 ( ∂ r m ) 5 ( ∂ r m ) ( ∂ r 3 m ) − 3 ( ∂ r 2 m ) 2 ∂ r S + 2 ( ∂ r m ) 4 ( ∂ r 2 m ) ( ∂ r 2 S ) − 6 ( ∂ r m ) 3 ∂ r 3 S . (138)
All radial derivatives here are evaluated at the initial r + r_+ r + along the stationary family at fixed λ A \lambda^A λ A . Using the neutral first law, ∂ r S = ∂ r m / T H \partial_r S=\partial_r m/T_{\mathrm{H}} ∂ r S = ∂ r m / T H and d T H / d r + = ∂ r m / C Q \mathrm{d}T_{\mathrm{H}}/\mathrm{d}r_+=\partial_r m/C_Q d T H / d r + = ∂ r m / C Q , the first two coefficients recover the thermal expansion in Eq. (22) .
There is a useful consistency condition between the mass dependence of the metric and the entropy. Differentiating the horizon equation along the stationary family yields
∂ r m ( r + ) = − ∂ r f ∂ M f , ∂ r S ( r + ) = − 4 π ∂ M f (139) \partial_r m(r_+)=-\frac{\partial_r f}{\partial_M f},\qquad\partial_r S(r_+)=-\frac{4\pi}{\partial_M f} \tag*{(139)} ∂ r m ( r + ) = − ∂ M f ∂ r f , ∂ r S ( r + ) = − ∂ M f 4 π (139)
where the partial derivatives are evaluated at ( r + , m ( r + ) , λ A ) (r_+,m(r_+),\lambda^A) ( r + , m ( r + ) , λ A ) and the second equality uses the fixed-parameter first law and T H = ∂ r f / ( 4 π ) T_{\mathrm{H}}=\partial_r f/(4\pi) T H = ∂ r f / ( 4 π ) at a simple horizon. Thus, for the radially constant affine coefficient in Eq. (19) ,
∂ r S ( r + ) = 4 π c , b n = 4 π c a n . (140) \partial_r S(r_+)=\frac{4\pi}{c},\qquad b_n=\frac{4\pi}{c}a_n. \tag*{(140)} ∂ r S ( r + ) = c 4 π , b n = c 4 π a n . (140)
The entropy is consequently linear in the horizon radius along each connected portion of this fixed-parameter family. For the charged BTZ solution, c = κ / π c=\kappa/\pi c = κ / π , and Eq. (A.8) exactly matches Eq. (68) .
In the same specialization, h ( y ) = − f 0 ( r + + y ) / c h(y)=-f_0(r_++y)/c h ( y ) = − f 0 ( r + + y ) / c and ∂ r j m ( r + ) = ∂ r j f 0 ( r + ) / c \partial_r^j m(r_+)=\partial_r^j f_0(r_+)/c ∂ r j m ( r + ) = ∂ r j f 0 ( r + ) / c for j ≥ 1 j\ge1 j ≥ 1 . These substitutions in Eq. (A.5) give Eq. (21) , while Eq. (A.4) determines every higher order.
The residue identities also hold for formal power series. In the analytic case, the convergence radii of the horizon and entropy series are set by the nearest complex singularities of the chosen inverse and of the composed entropy, respectively. A multiple zero on another branch does not, by itself, obstruct the event-horizon expansion.
Charged horizon and entropy coefficients
For charged emission, the remaining parameters λ A \lambda^A λ A are held fixed, while the final charge is the endpoint value Q ′ ( E ) = Q − Q e m Q'(\mathcal{E})=Q-Q_{\mathrm{em}} Q ′ ( E ) = Q − Q em . The function h ( y ; Q e m ) h(y;Q_{\mathrm{em}}) h ( y ; Q em ) in Eq. (30) satisfies h ( 0 ; Q e m ) = 0 h(0;Q_{\mathrm{em}})=0 h ( 0 ; Q em ) = 0 . Its inverse argument is now
E ~ : = E + m ( r + , Q ′ ( E ) ) − M , E ~ = h ( Δ r + ; Q e m ) . (141) \widetilde{\mathcal{E}}:=\mathcal{E}+m(r_+,Q'(\mathcal{E}))-\mathcal{M},\qquad\widetilde{\mathcal{E}}=h(\Delta r_+;Q_{\mathrm{em}}). \tag*{(141)} E := E + m ( r + , Q ′ ( E )) − M , E = h ( Δ r + ; Q em ) . (141)
This shift accounts for the change of the reference mass at fixed radius. For sufficiently small Q e m Q_{\mathrm{em}} Q em , the initial simple zero ensures ∂ y h ( 0 ; Q e m ) = − ∂ r m ( r + , Q ′ ( E ) ) ≠ 0 \partial_y h(0;Q_{\mathrm{em}})=-\partial_r m(r_+,Q'(\mathcal{E}))\ne0 ∂ y h ( 0 ; Q em ) = − ∂ r m ( r + , Q ′ ( E )) = 0 . Holding this endpoint charge fixed in the radial residue calculation gives
Δ r + = ∑ n = 1 ∞ a n ( Q e m ) E ~ n . (142) \Delta r_+=\sum_{n=1}^{\infty}a_n(Q_{\mathrm{em}})\widetilde{\mathcal{E}}^n. \tag*{(142)} Δ r + = n = 1 ∑ ∞ a n ( Q em ) E n . (142)
where the coefficients are given by Eq. (16) with h ( y ) h(y) h ( y ) replaced by h ( y ; Q e m ) h(y;Q_{\mathrm{em}}) h ( y ; Q em ) . Equations (A.4) and (A.5) remain valid with all radial derivatives evaluated at ( r + , Q ′ ( E ) ) (r_+,Q'(\mathcal{E})) ( r + , Q ′ ( E )) .
Writing the entropy on the stationary family as S ( r , Q ) S(r,Q) S ( r , Q ) and taking g ( y ) = S ( r + + y , Q ′ ( E ) ) g(y)=S(r_++y,Q'(\mathcal{E})) g ( y ) = S ( r + + y , Q ′ ( E )) at fixed Q ′ ( E ) Q'(\mathcal{E}) Q ′ ( E ) , the same composed inverse gives
Δ S = S ( r + , Q ′ ( E ) ) − S ( r + , Q ) + ∑ n = 1 ∞ b n ( Q e m ) E ~ n , (143) \Delta S=S(r_+,Q'(\mathcal{E}))-S(r_+,Q)+\sum_{n=1}^{\infty}b_n(Q_{\mathrm{em}})\widetilde{\mathcal{E}}^n, \tag*{(143)} Δ S = S ( r + , Q ′ ( E )) − S ( r + , Q ) + n = 1 ∑ ∞ b n ( Q em ) E n , (143)
with b n ( Q e m ) b_n(\mathcal{Q}_{\rm em}) b n ( Q em ) given by Eq. (17) at fixed Q ′ ( E ) \mathcal{Q}'(\mathcal{E}) Q ′ ( E ) , using the same replacement of h h h . The fixed-radius entropy difference vanishes for the standard Einstein area entropy, but need not do so for a general gravitational entropy.
The charged first law, restricted to the stationary family, reads
∂ r S = ∂ r m T H , ∂ Q S = ∂ Q m − Φ + T H . (144) \partial_r S = \frac{\partial_r m}{T_{\rm H}}, \qquad\partial_{\mathcal{Q}} S = \frac{\partial_{\mathcal{Q}}m-\Phi_+}{T_{\rm H}}. \tag*{(144)} ∂ r S = T H ∂ r m , ∂ Q S = T H ∂ Q m − Φ + . (144)
Consequently, the joint expansion of Eq. (A.11) begins with − ( E − Φ + Q e m ) / T H -(\mathcal{E}-\Phi_+\mathcal{Q}_{\rm em})/T_{\rm H} − ( E − Φ + Q em ) / T H and reproduces the entropy Hessian in Eq. (33) at quadratic order. For the affine family, Eq. (A.8) also applies to the coefficients at each fixed Q ′ ( E ) \mathcal{Q}'(\mathcal{E}) Q ′ ( E ) ; the fixed-radius charge contribution must still be retained when present.
Along the chosen channel C \mathcal{C} C , one finally substitutes Q e m = Q e m ( E ) \mathcal{Q}_{\rm em}=\mathcal{Q}_{\rm em}(\mathcal{E}) Q em = Q em ( E ) to obtain the horizon and its entropy at the emission endpoint.
Charged horizon residue and the emission path
The radial contour calculation of § 2.1 extends to charged emission once the electrostatic work is included. We use a regular gauge and follow the outgoing pole along the path C \mathcal{C} C of Eq. (24) . The finite-emission construction uses the quasistationary charged KPW prescription [34 , 35 , 36 ] .
At each fixed intermediate state, the time transformation in Eq. (3) sends A = A t d t A=A_tdt A = A t d t to a potential with radial component − A t 1 − f / f -A_t\sqrt{1-f}/f − A t 1 − f / f . A gauge transformation removes this component:
A r e g = A + d Λ = A t d τ , ∂ r Λ = A t 1 − f f , I r e g = I + e Λ . (145) A^{\rm reg}=A+\mathrm{d}\Lambda=A_t\mathrm{d}\tau,
\qquad
\partial_r\Lambda=\frac{A_t\sqrt{1-f}}{f},
\qquad
I^{\rm reg}=I+e\Lambda. \tag*{(145)} A reg = A + d Λ = A t d τ , ∂ r Λ = f A t 1 − f , I reg = I + e Λ. (145)
The potential is regular at the future horizon, although Λ \Lambda Λ may be logarithmically singular on the overlap with the original exterior chart. The kinetic momentum ∂ μ I − e A μ \partial_\mu I-eA_\mu ∂ μ I − e A μ is unchanged. We use this representative below, with A τ = − Φ A_\tau=-\Phi A τ = − Φ and A r = 0 A_r=0 A r = 0 , retaining the energy reference of § 2.4 .
For a radial excitation of charge e e e , let ε = − ∂ τ I \varepsilon=-\partial_\tau I ε = − ∂ τ I and p r = ∂ r I p_r=\partial_r I p r = ∂ r I . In the Painlevé metric, the leading massless Hamilton–Jacobi equation becomes
f p r 2 − 2 1 − f ( ε − e Φ ) p r − ( ε − e Φ ) 2 = 0. (146) fp_r^2-2\sqrt{1-f}(\varepsilon-e\Phi)p_r-(\varepsilon-e\Phi)^2=0. \tag*{(146)} f p r 2 − 2 1 − f ( ε − e Φ ) p r − ( ε − e Φ ) 2 = 0. (146)
Its outgoing solution is
p r o u t = ε − e Φ 1 − 1 − f = ε − e Φ r ˙ , (147) p_r^{\rm out}=\frac{\varepsilon-e\Phi}{1-\sqrt{1-f}}=\frac{\varepsilon-e\Phi}{\dot r}, \tag*{(147)} p r out = 1 − 1 − f ε − e Φ = r ˙ ε − e Φ , (147)
where r ˙ \dot r r ˙ is given by Eq. (4) . Mass and transverse kinetic terms do not change the leading simple pole when ε − e Φ + > 0 \varepsilon-e\Phi_+>0 ε − e Φ + > 0 .
In the reduced KPW action, Hamilton’s equation converts the radial momentum integration into an integration over this energy transfer. Conservation along C \mathcal{C} C gives d M ′ = − d E ′ \mathrm{d}\mathcal{M}'=-\mathrm{d}\mathcal{E}' d M ′ = − d E ′ and d Q ′ = − d Q e m \mathrm{d}\mathcal{Q}'=-\mathrm{d}\mathcal{Q}_{\rm em} d Q ′ = − d Q em , so the singular increment is
d p r ∣ p o l e = d E ′ − Φ ( r ; M ′ , Q ′ ) d Q e m r ˙ ( r ; M ′ , Q ′ ) ∣ p o l e . (148) \left.\mathrm{d}p_r\right|_{\rm pole}
=
\left.
\frac{\mathrm{d}\mathcal{E}'-\Phi(r;\mathcal{M}',\mathcal{Q}')\mathrm{d}\mathcal{Q}_{\rm em}}
{\dot r(r;\mathcal{M}',\mathcal{Q}')}
\right|_{\rm pole}. \tag*{(148)} d p r ∣ pole = r ˙ ( r ; M ′ , Q ′ ) d E ′ − Φ ( r ; M ′ , Q ′ ) d Q em pole . (148)
Here d p r d p_r d p r denotes the momentum increment of the reduced action after subtraction of the electromagnetic canonical term, not the total variation of Eq. (B.3) through changing backgrounds. Hamilton’s equation is applied at fixed r r r in each instantaneous state. The profile Q e m ( E ′ ) \mathcal{Q}_{\rm em}(\mathcal{E}') Q em ( E ′ ) fixes the charge transferred between successive black hole states.
Inserting p r = ∫ d p r p_r=\int d p_r p r = ∫ d p r into the reduced radial action I = ∫ p r d r I=\int p_r d r I = ∫ p r d r and parametrizing the transfers by E ′ \mathcal{E}' E ′ gives
I p o l e = ∫ 0 E d E ′ ∫ γ ( E ′ ) d r 1 − Φ ( r ; M ′ , Q ′ ) r ˙ ( r ; M ′ , Q ′ ) d Q e m d E ′ ∣ p o l e . (149) I_{\mathrm{pole}}=\int_{0}^{\mathcal{E}}d\mathcal{E}'\int_{\gamma(\mathcal{E}')}d r\left.\frac{1-\Phi(r;\mathcal{M}',Q')}{\dot{r}(r;\mathcal{M}',Q')}\frac{d\mathcal{Q}_{\rm em}}{d\mathcal{E}'}\right|_{\mathrm{pole}} . \tag*{(149)} I pole = ∫ 0 E d E ′ ∫ γ ( E ′ ) d r r ˙ ( r ; M ′ , Q ′ ) 1 − Φ ( r ; M ′ , Q ′ ) d E ′ d Q em pole . (149)
Here γ ( E ′ ) \gamma(\mathcal{E}') γ ( E ′ ) crosses the instantaneous outer zero R + ( E ′ ) R_{+}(\mathcal{E}') R + ( E ′ ) and follows its displacement along C \mathcal{C} C . The positive-energy contour prescription is imposed before exchanging the integrations. Terms regular at the horizon do not contribute to this imaginary part.
To evaluate the running residue, center each radial contour at its pole by writing r = R + ( E ′ ) + y r=R_{+}(\mathcal{E}')+y r = R + ( E ′ ) + y . Each radial integral is taken at fixed E ′ \mathcal{E}' E ′ , so d r = d y d r=d y d r = d y . The local expansions are
r ˙ = 2 π T H ( E ′ ) y + O ( y 2 ) , Φ ( r ) = Φ + ( E ′ ) + O ( y ) . (150) \dot{r}=2\pi T_{\mathrm{H}}(\mathcal{E}')y+O(y^{2}),\qquad\Phi(r)=\Phi_{+}(\mathcal{E}')+O(y). \tag*{(150)} r ˙ = 2 π T H ( E ′ ) y + O ( y 2 ) , Φ ( r ) = Φ + ( E ′ ) + O ( y ) . (150)
The radial variation of Φ \Phi Φ is regular after division by r ˙ \dot{r} r ˙ and leaves the residue unchanged. Here the positive shell transfer is paired with the local outgoing contour, oriented from inside to outside. With this orientation, the prescription y − i 0 y-i0 y − i 0 gives Im ∫ d y / ( y − i 0 ) = π \operatorname{Im}\int d y/(y-i0)=\pi Im ∫ d y / ( y − i 0 ) = π . Thus
Im I = 1 2 ∫ 0 E 1 − Φ + ( E ′ ) T H ( E ′ ) d Q e m d E ′ d E ′ = − 1 2 ∫ C d M ′ − Φ + d Q ′ T H . (151) \operatorname{Im} I=\frac{1}{2}\int_{0}^{\mathcal{E}}\frac{1-\Phi_{+}(\mathcal{E}')}{T_{\mathrm{H}}(\mathcal{E}')}\frac{d\mathcal{Q}_{\rm em}}{d\mathcal{E}'}d\mathcal{E}'=-\frac{1}{2}\int_{\mathcal{C}}\frac{d\mathcal{M}'-\Phi_{+}dQ'}{T_{\mathrm{H}}}. \tag*{(151)} Im I = 2 1 ∫ 0 E T H ( E ′ ) 1 − Φ + ( E ′ ) d E ′ d Q em d E ′ = − 2 1 ∫ C T H d M ′ − Φ + d Q ′ . (151)
This reproduces Eq. (26) . The stationary first law identifies the last integrand with d S dS d S , giving the finite entropy difference.
A change of gauge reference must be accompanied by the corresponding change of canonical energy, preserving ϵ − e Φ \epsilon-e\Phi ϵ − e Φ . The conserved mass, charge and potential for the first-order H ( P ) \mathcal{H}(\mathcal{P}) H ( P ) realization are identified by the fixed-coupling Hamiltonian derivation in Appendix § 12 .§ 5 .
Puiseux–Lagrange series for degenerate horizons
At a multiple horizon, the inverse in Appendix § 9 is naturally expanded in fractional powers of the emitted energy for neutral emission, or of the mass difference from the critical curve for charged emission. We derive the coefficients by reducing the problem to an ordinary inverse with a nonzero linear term. This gives the particular Puiseux construction needed here. More general treatments may be found in Refs. [32 , 38 ] .
A horizon of arbitrary multiplicity
Let r d r_d r d be a zero of multiplicity m ≥ 2 m\ge2 m ≥ 2 at the reference mass M \mathcal{M} M , so that
∂ r j f ( r d ; M , λ A ) = 0 , ( 0 ≤ j < m ) , \partial_r^{j}f(r_d;\mathcal{M},\lambda^{A})=0,\qquad(0\le j<m), ∂ r j f ( r d ; M , λ A ) = 0 , ( 0 ≤ j < m ) ,
∂ r m f ( r d ; M , λ A ) ≠ 0. (152) \partial_r^{m}f(r_d;\mathcal{M},\lambda^{A})\ne0. \tag*{(152)} ∂ r m f ( r d ; M , λ A ) = 0. (152)
We assume analyticity and ∂ M f ( r d ; M , λ A ) ≠ 0 \partial_{\mathcal{M}}f(r_d;\mathcal{M},\lambda^A) \ne0 ∂ M f ( r d ; M , λ A ) = 0 , so the local mass function in Eq. (13) remains well defined. Repeated differentiation of f [ r , m ( r ) , λ A ] = 0 f[r,m(r),\lambda^A]=0 f [ r , m ( r ) , λ A ] = 0 then gives
∂ r j m ( r d ) = 0 , ( 1 ≤ j < m ) , \partial_r^j m(r_d)=0,\qquad(1\le j<m), ∂ r j m ( r d ) = 0 , ( 1 ≤ j < m ) ,
∂ r m m ( r d ) = − ∂ r m f ∂ M f . (153) \partial_r^m m(r_d)=-\frac{\partial_r^m f}{\partial_{\mathcal{M}}f}. \tag*{(153)} ∂ r m m ( r d ) = − ∂ M f ∂ r m f . (153)
with the partial derivatives evaluated at the degenerate horizon. Taking r d r_d r d as the reference radius in Eq. (14) , we write
h ( y ) : = m ( r d ) − m ( r d + y ) , E = h ( y ) . (154) h(y):=m(r_d)-m(r_d+y),\qquad\mathcal{E}=h(y). \tag*{(154)} h ( y ) := m ( r d ) − m ( r d + y ) , E = h ( y ) . (154)
where y y y is the displaced radius relative to r d r_d r d . The ratio h ( y ) / y m h(y)/y^m h ( y ) / y m is analytic and nonzero at the origin. Choose one of its analytic m m m -th roots and introduce the local variable
u = y ( h ( y ) y m ) 1 / m , u m = E . (155) u=y\left(\frac{h(y)}{y^m}\right)^{1/m},\qquad u^m=\mathcal{E}. \tag*{(155)} u = y ( y m h ( y ) ) 1/ m , u m = E . (155)
Since d u / d y ≠ 0 du/dy\ne0 d u / d y = 0 at y = 0 y=0 y = 0 , the inversion formula in Appendix § 9 applies to y ( u ) y(u) y ( u ) . The m m m values u = ζ j E 1 / m u=\zeta_j\mathcal{E}^{1/m} u = ζ j E 1/ m , with ζ j = exp ( 2 π i j / m ) \zeta_j=\exp(2\pi i j/m) ζ j = exp ( 2 π ij / m ) and j = 0 , … , m − 1 j=0,\ldots,m-1 j = 0 , … , m − 1 , give
Δ r j ( E ) = ∑ n = 1 ∞ a n ( j ) E n / m , \Delta r_j(\mathcal{E})=\sum_{n=1}^{\infty}a_n^{(j)}\mathcal{E}^{n/m}, Δ r j ( E ) = n = 1 ∑ ∞ a n ( j ) E n / m ,
a n ( j ) = ζ j n n ! [ ∂ y n − 1 ( y m h ( y ) ) n / m ] y = 0 . (156) a_n^{(j)}=\frac{\zeta_j^n}{n!}\left[\partial_y^{n-1}\left(\frac{y^m}{h(y)}\right)^{n/m}\right]_{y=0}. \tag*{(156)} a n ( j ) = n ! ζ j n [ ∂ y n − 1 ( h ( y ) y m ) n / m ] y = 0 . (156)
The fractional power in this expression is the n n n th power of the reciprocal of the chosen analytic m m m -th root of h ( y ) / y m h(y)/y^m h ( y ) / y m . Changing that choice only permutes the m m m solutions. Their series converge for sufficiently small ∣ E ∣ 1 / m |\mathcal{E}|^{1/m} ∣ E ∣ 1/ m , or can be read as formal Puiseux series. For the affine mass dependence of Eq. (19) , one substitutes h ( y ) = − f 0 ( r d + y ) / c h(y)=-f_0(r_d+y)/c h ( y ) = − f 0 ( r d + y ) / c in Eq. (C.5). In particular,
( a 1 ( j ) ) m = − c m ! ∂ r m f 0 ( r d ) . (157) \left(a_1^{(j)}\right)^m=-\frac{c\,m!}{\partial_r^m f_0(r_d)}. \tag*{(157)} ( a 1 ( j ) ) m = − ∂ r m f 0 ( r d ) c m ! . (157)
For real analytic f 0 f_0 f 0 and sufficiently small E > 0 \mathcal{E}>0 E > 0 , a zero of even multiplicity therefore produces two nearby real zeros when ∂ r 2 f 0 ( r d ) < 0 \partial_r^2 f_0(r_d)<0 ∂ r 2 f 0 ( r d ) < 0 , and none when ∂ r 2 f 0 ( r d ) > 0 \partial_r^2 f_0(r_d)>0 ∂ r 2 f 0 ( r d ) > 0 . An odd multiplicity produces one real zero for either sign. The remaining zeros are complex. This is a local count: horizons away from r d r_d r d are unaffected by this classification.
In this specialization, the slopes at the displaced zeros satisfy
∂ r f [ r d + Δ r j ( E ) ; M − E , λ A ] = − m c a 1 ( j ) E ( m − 1 ) / m + O ( E ) . (158) \partial_r f[r_d+\Delta r_j(\mathcal{E});\mathcal{M}-\mathcal{E},\lambda^A]=-\frac{mc}{a_1^{(j)}}\mathcal{E}^{(m-1)/m}+O(\mathcal{E}). \tag*{(158)} ∂ r f [ r d + Δ r j ( E ) ; M − E , λ A ] = − a 1 ( j ) m c E ( m − 1 ) / m + O ( E ) . (158)
Thus each of these zeros is simple for sufficiently small nonzero energy. When a real zero represents an event horizon with positive slope, its temperature inherits the power E ( m − 1 ) / m \mathcal{E}^{(m-1)/m} E ( m − 1 ) / m .
Application to double and triple zeros
For m = 2 m=2 m = 2 and ∂ r 2 f 0 ( r d ) < 0 \partial_r^2 f_0(r_d)<0 ∂ r 2 f 0 ( r d ) < 0 , choose the positive square root at y = 0 y=0 y = 0 and label the two real solutions by ϵ = ± 1 \epsilon=\pm1 ϵ = ± 1 . The coefficients through order E \mathcal{E} E are those in Eq. (38) ; the next one is
a 3 ( ϵ ) = c a 1 ( ϵ ) 36 [ ∂ r 2 f 0 ( r d ) ] 3 { 3 ∂ r 2 f 0 ( r d ) ∂ r 4 f 0 ( r d ) − 5 [ ∂ r 3 f 0 ( r d ) ] 2 } . (159) a_3^{(\epsilon)}=\frac{c\,a_1^{(\epsilon)}}{36[\partial_r^2 f_0(r_d)]^3}\left\{3\partial_r^2 f_0(r_d)\partial_r^4 f_0(r_d)-5[\partial_r^3 f_0(r_d)]^2\right\}. \tag*{(159)} a 3 ( ϵ ) = 36 [ ∂ r 2 f 0 ( r d ) ] 3 c a 1 ( ϵ ) { 3 ∂ r 2 f 0 ( r d ) ∂ r 4 f 0 ( r d ) − 5 [ ∂ r 3 f 0 ( r d ) ] 2 } . (159)
The identity a n ( − ) = ( − 1 ) n a n ( + ) a_n^{(-)} = (-1)^n a_n^{(+)} a n ( − ) = ( − 1 ) n a n ( + ) holds at every order. For m = 3 m = 3 m = 3 , choose
a 1 ( 0 ) = − 6 c ∂ r 3 f 0 ( r d ) 3 , (160) a_1^{(0)} = \sqrt[3]{-\frac{6c}{\partial_r^3 f_0(r_d)}}, \tag*{(160)} a 1 ( 0 ) = 3 − ∂ r 3 f 0 ( r d ) 6 c , (160)
where the cube root is real. With ζ j = exp ( 2 π i j / 3 ) \zeta_j = \exp(2\pi i j/3) ζ j = exp ( 2 π ij /3 ) , the first coefficients are
a 1 ( j ) = ζ j a 1 ( 0 ) , a_1^{(j)} = \zeta_j a_1^{(0)}, a 1 ( j ) = ζ j a 1 ( 0 ) ,
a 2 ( j ) = − ζ j 2 [ a 1 ( 0 ) ] 2 ∂ r 4 f 0 ( r d ) 12 ∂ r 3 f 0 ( r d ) , a_2^{(j)} = -\frac{\zeta_j^2 [a_1^{(0)}]^2 \partial_r^4 f_0(r_d)}{12\partial_r^3 f_0(r_d)}, a 2 ( j ) = − 12 ∂ r 3 f 0 ( r d ) ζ j 2 [ a 1 ( 0 ) ] 2 ∂ r 4 f 0 ( r d ) ,
a 3 ( j ) = [ a 1 ( 0 ) ] 3 { [ ∂ r 4 f 0 ( r d ) ] 2 48 [ ∂ r 3 f 0 ( r d ) ] 2 − ∂ r 5 f 0 ( r d ) 60 ∂ r 3 f 0 ( r d ) } . (161) a_3^{(j)} = [a_1^{(0)}]^3 \left\{\frac{[\partial_r^4 f_0(r_d)]^2}{48[\partial_r^3 f_0(r_d)]^2} - \frac{\partial_r^5 f_0(r_d)}{60\partial_r^3 f_0(r_d)}\right\}. \tag*{(161)} a 3 ( j ) = [ a 1 ( 0 ) ] 3 { 48 [ ∂ r 3 f 0 ( r d ) ] 2 [ ∂ r 4 f 0 ( r d ) ] 2 − 60 ∂ r 3 f 0 ( r d ) ∂ r 5 f 0 ( r d ) } . (161)
The solution j = 0 j = 0 j = 0 is real, while j = 1 , 2 j = 1, 2 j = 1 , 2 form a complex-conjugate pair. Thus a stationary inflection does not split into three real horizons under the uniform mass shift. The real zero moves toward smaller radii if ∂ r 3 f 0 ( r d ) > 0 \partial_r^3 f_0(r_d) > 0 ∂ r 3 f 0 ( r d ) > 0 , and toward larger radii if ∂ r 3 f 0 ( r d ) < 0 \partial_r^3 f_0(r_d) < 0 ∂ r 3 f 0 ( r d ) < 0 . Its slope vanishes as E 2 / 3 \mathcal{E}^{2/3} E 2/3 .
Charged emission about the critical curve
For a double zero, the charged inverse is centered at r d ( Q ′ ) r_d(Q') r d ( Q ′ ) , as in § 3.2 . Since ∂ r 2 m ≠ 0 \partial_r^2 m \ne0 ∂ r 2 m = 0 , the equation ∂ r m [ r d ( Q ′ ) , Q ′ ] = 0 \partial_r m[r_d(Q'), Q'] = 0 ∂ r m [ r d ( Q ′ ) , Q ′ ] = 0 defines this radius analytically for nearby charges. Differentiating this equation and the critical mass M d ( Q ′ ) = m [ r d ( Q ′ ) , Q ′ ] \mathcal{M}_d(Q') = m[r_d(Q'), Q'] M d ( Q ′ ) = m [ r d ( Q ′ ) , Q ′ ] gives
d r d d Q ′ = − ∂ r ∂ Q m ∂ r 2 m , d M d d Q ′ = ∂ Q m , d 2 M d d Q ′ 2 = ∂ Q 2 m − ( ∂ r ∂ Q m ) 2 ∂ r 2 m . (162) \begin{aligned}
\frac{d r_d}{d Q'} &= -\frac{\partial_r \partial_Q m}{\partial_r^2 m},\\
\frac{d\mathcal{M}_d}{d Q'} &= \partial_Q m,\\
\frac{d^2\mathcal{M}_d}{d Q'^2} &= \partial_Q^2 m - \frac{(\partial_r\partial_Q m)^2}{\partial_r^2 m}. \tag*{(162)}
\end{aligned} d Q ′ d r d d Q ′ d M d d Q ′2 d 2 M d = − ∂ r 2 m ∂ r ∂ Q m , = ∂ Q m , = ∂ Q 2 m − ∂ r 2 m ( ∂ r ∂ Q m ) 2 . (162)
All derivatives of m m m are evaluated at ( r d ( Q ′ ) , Q ′ ) (r_d(Q'), Q') ( r d ( Q ′ ) , Q ′ ) . At the endpoint charge Q ′ ( E ) Q'(\mathcal{E}) Q ′ ( E ) , define
h d ( y ; Q ′ ( E ) ) : = M d ( Q ′ ( E ) ) − m [ r d ( Q ′ ( E ) ) + y , Q ′ ( E ) ] , Δ M d = h d ( y ; Q ′ ( E ) ) . (163) \begin{aligned}
h_d(y; Q'(\mathcal{E})) &:= \mathcal{M}_d(Q'(\mathcal{E})) - m[r_d(Q'(\mathcal{E})) + y, Q'(\mathcal{E})],\\
\Delta\mathcal{M}_d &= h_d(y; Q'(\mathcal{E})). \tag*{(163)}
\end{aligned} h d ( y ; Q ′ ( E )) Δ M d := M d ( Q ′ ( E )) − m [ r d ( Q ′ ( E )) + y , Q ′ ( E )] , = h d ( y ; Q ′ ( E )) . (163)
Here y y y is the dummy radial variable of the inverse construction; at a root it equals the displacement Δ r d \Delta r_d Δ r d measured from the critical radius at the final charge. The second equality is then the final horizon condition, with Δ M d \Delta\mathcal{M}_d Δ M d defined in Eq. (42) . Since h d ( y ; Q ′ ( E ) ) = − ( ∂ r 2 m ) y 2 / 2 + O ( y 3 ) h_d(y; Q'(\mathcal{E})) = -(\partial_r^2m)y^2/2 + O(y^3) h d ( y ; Q ′ ( E )) = − ( ∂ r 2 m ) y 2 /2 + O ( y 3 ) , the construction in Eq. (C.4) applies with m = 2 m = 2 m = 2 and Δ M d \Delta\mathcal{M}_d Δ M d in place of E \mathcal{E} E .
For ∂ r 2 m < 0 \partial_r^2m < 0 ∂ r 2 m < 0 and Δ M d > 0 \Delta\mathcal{M}_d > 0 Δ M d > 0 , choose the positive square root at y = 0 y = 0 y = 0 . Accordingly, the two real branches are
r d , f ( E ) = r d ( Q ′ ( E ) ) + ∑ n = 1 ∞ a n ( E ) ( Q ′ ( E ) ) ( Δ M d ) n / 2 , (164) r_{d,f}^{(\mathcal{E})} = r_d(Q'(\mathcal{E})) + \sum_{n=1}^{\infty} a_n^{(\mathcal{E})}(Q'(\mathcal{E}))(\Delta\mathcal{M}_d)^{n/2}, \tag*{(164)} r d , f ( E ) = r d ( Q ′ ( E )) + n = 1 ∑ ∞ a n ( E ) ( Q ′ ( E )) ( Δ M d ) n /2 , (164)
where ϵ = ± 1 \epsilon=\pm1 ϵ = ± 1 and
a n ( ϵ ) ( Q ′ ( E ) ) = ϵ n n ! [ ∂ y n − 1 ( y 2 h d ( y ; Q ′ ( E ) ) ) n / 2 ] y = 0 . (165) a_n^{(\epsilon)}\bigl(\mathcal{Q}'(\mathcal{E})\bigr)=\frac{\epsilon^n}{n!}\left[\partial_y^{n-1}\left(\frac{y^2}{h_d\bigl(y;\mathcal{Q}'(\mathcal{E})\bigr)}\right)^{n/2}\right]_{y=0}. \tag*{(165)} a n ( ϵ ) ( Q ′ ( E ) ) = n ! ϵ n ∂ y n − 1 ( h d ( y ; Q ′ ( E ) ) y 2 ) n /2 y = 0 . (165)
The first coefficients follow from Eqs. (38) and (C.8) with the replacement
∂ r j f 0 ( r d ) c ⟶ ∂ r j m [ r d ( Q ′ ( E ) ) , Q ′ ( E ) ] . \frac{\partial_r^j f_0(r_d)}{c}\longrightarrow\partial_r^j m\bigl[r_d\bigl(\mathcal{Q}'(\mathcal{E})\bigr),\mathcal{Q}'(\mathcal{E})\bigr]. c ∂ r j f 0 ( r d ) ⟶ ∂ r j m [ r d ( Q ′ ( E ) ) , Q ′ ( E ) ] .
At a local minimum, ∂ r 2 m > 0 \partial_r^2m>0 ∂ r 2 m > 0 , the real pair instead requires Δ M d < 0 \Delta\mathcal{M}_d<0 Δ M d < 0 . Replacing h d h_d h d by − h d -h_d − h d and Δ M d \Delta\mathcal{M}_d Δ M d by − Δ M d -\Delta\mathcal{M}_d − Δ M d in the inverse construction gives its real coefficients. In either case, the first terms take the form of Eq. (45) . The neutral formulas are recovered when Q ′ ( E ) = Q \mathcal{Q}'(\mathcal{E})=\mathcal{Q} Q ′ ( E ) = Q and Δ M d = E \Delta\mathcal{M}_d=\mathcal{E} Δ M d = E ; in the affine specialization, ∂ r j m = ∂ r j f 0 / c \partial_r^jm=\partial_r^jf_0/c ∂ r j m = ∂ r j f 0 / c for j j j radial derivatives. No affine mass dependence is needed for the charged construction itself.
The expansion along C C C follows by inserting its charge profile into the critical mass. For an analytic profile near the initial state,
Δ M d ( E ) = E + ∑ k = 1 ∞ ( − 1 ) k k ! d k M d d Q k [ Q e m ( E ) ] k , (166) \Delta\mathcal{M}_d(\mathcal{E})=\mathcal{E}+\sum_{k=1}^{\infty}\frac{(-1)^k}{k!}\frac{d^k\mathcal{M}_d}{d\mathcal{Q}^k}\left[\mathcal{Q}_{\rm em}(\mathcal{E})\right]^k, \tag*{(166)} Δ M d ( E ) = E + k = 1 ∑ ∞ k ! ( − 1 ) k d Q k d k M d [ Q em ( E ) ] k , (166)
where the critical derivatives are evaluated at the initial charge. With η = ( ∂ E Q e m ) ( 0 ) \eta=(\partial_{\mathcal{E}}\mathcal{Q}_{\rm em})(0) η = ( ∂ E Q em ) ( 0 ) , this gives
Δ M d ( E ) = ( 1 − η d M d d Q ) E + 1 2 [ η 2 d 2 M d d Q 2 − d M d d Q ( ∂ E 2 Q e m ) ( 0 ) ] E 2 + O ( E 3 ) . (167) \Delta\mathcal{M}_d(\mathcal{E})=\left(1-\eta\frac{d\mathcal{M}_d}{d\mathcal{Q}}\right)\mathcal{E}+\frac{1}{2}\left[\eta^2\frac{d^2\mathcal{M}_d}{d\mathcal{Q}^2}-\frac{d\mathcal{M}_d}{d\mathcal{Q}}\left(\partial_{\mathcal{E}}^2\mathcal{Q}_{\rm em}\right)(0)\right]\mathcal{E}^2+O(\mathcal{E}^3). \tag*{(167)} Δ M d ( E ) = ( 1 − η d Q d M d ) E + 2 1 [ η 2 d Q 2 d 2 M d − d Q d M d ( ∂ E 2 Q em ) ( 0 ) ] E 2 + O ( E 3 ) . (167)
Thus the quadratic term in a tangent channel depends both on the curvature of the critical curve and on that of the charge profile. Together with the common displacement r d [ Q − Q e m ( E ) ] r_d[\mathcal{Q}-\mathcal{Q}_{\rm em}(\mathcal{E})] r d [ Q − Q em ( E )] , this gives the path-dependent splitting described in § 3.2 . These formulas apply while the critical zero remains double. If ∂ r 2 m \partial_r^2m ∂ r 2 m also vanishes, its continuation requires a higher-multiplicity analysis, since charge variation need not preserve that multiplicity.
Approach to the outer coalescence
The next radial term in Eq. (50) follows from the double-zero inverse at the limiting minimum r e r_e r e ,
R + ( E ) = r e + 2 c ∂ r 2 f c ( r e ) ( E c − E ) 1 / 2 − c ∂ r 3 f c ( r e ) 3 [ ∂ r 2 f c ( r e ) ] 2 ( E c − E ) + O ( ( E c − E ) 3 / 2 ) . (168) R_+(\mathcal{E})=r_e+\sqrt{\frac{2c}{\partial_r^2f_c(r_e)}}(\mathcal{E}_c-\mathcal{E})^{1/2}-\frac{c\,\partial_r^3f_c(r_e)}{3[\partial_r^2f_c(r_e)]^2}(\mathcal{E}_c-\mathcal{E})+O\bigl((\mathcal{E}_c-\mathcal{E})^{3/2}\bigr). \tag*{(168)} R + ( E ) = r e + ∂ r 2 f c ( r e ) 2 c ( E c − E ) 1/2 − 3 [ ∂ r 2 f c ( r e ) ] 2 c ∂ r 3 f c ( r e ) ( E c − E ) + O ( ( E c − E ) 3/2 ) . (168)
For charged emission, the expansion in Eq. (45) is evaluated at the outer minimum r e ( Q ′ ) r_e(\mathcal{Q}') r e ( Q ′ ) , with ∂ r 2 m > 0 \partial_r^2m>0 ∂ r 2 m > 0 also at the endpoint and the mass difference Δ M e < 0 \Delta\mathcal{M}_e<0 Δ M e < 0 of Eq. (53) . In the affine family with a regular positive coefficient c c c , T H = c ∂ r m / ( 4 π ) T_{\rm H}=c\,\partial_rm/(4\pi) T H = c ∂ r m / ( 4 π ) at the simple event horizon, and hence
T H ( E ) = c 4 π − 2 ∂ r 2 m Δ M e + O ( ∣ Δ M e ∣ ) . (169) T_{\rm H}(\mathcal{E})=\frac{c}{4\pi}\sqrt{-2\partial_r^2m\Delta\mathcal{M}_e}+O\bigl(|\Delta\mathcal{M}_e|\bigr). \tag*{(169)} T H ( E ) = 4 π c − 2 ∂ r 2 m Δ M e + O ( ∣Δ M e ∣ ) . (169)
Here Q ′ Q' Q ′ denotes Q ′ ( E ) = Q − Q e m ( E ) Q'(\mathcal{E})=Q-Q_{\mathrm{em}}(\mathcal{E}) Q ′ ( E ) = Q − Q em ( E ) , and ∂ r 2 m \partial_r^2m ∂ r 2 m is evaluated at ( r e ( Q ′ ) , Q ′ ) (r_e(Q'),Q') ( r e ( Q ′ ) , Q ′ ) . If c c c depends on the charge, it is evaluated at the same Q ′ Q' Q ′ .
For an analytic channel with Δ M e ∼ − K ( E c − E ) n \Delta M_e\sim-K(\mathcal{E}_c-\mathcal{E})^n Δ M e ∼ − K ( E c − E ) n , where K > 0 K>0 K > 0 and n ≥ 1 n\geq1 n ≥ 1 is an integer, the radial correction is of order ( E c − E ) n / 2 (\mathcal{E}_c-\mathcal{E})^{n/2} ( E c − E ) n /2 , while r e [ Q ′ ( E ) ] r_e[Q'(\mathcal{E})] r e [ Q ′ ( E )] varies analytically near the endpoint. If the entropy is analytic in the radius and charge there, its derivative therefore receives a regular contribution from the motion along the critical curve and a term of order ( E c − E ) n / 2 − 1 (\mathcal{E}_c-\mathcal{E})^{n/2-1} ( E c − E ) n /2 − 1 . The first law gives
1 − Φ + ( E ) ( ∂ E Q e m ) ( E ) T H ( E ) = − d d E S [ R + ( E ) , Q ′ ( E ) ] = O ( 1 ) + O ( ( E c − E ) n / 2 − 1 ) . (170) \frac{1-\Phi_+(\mathcal{E})(\partial_{\mathcal{E}}Q_{\mathrm{em}})(\mathcal{E})}{T_H(\mathcal{E})}
=-\frac{d}{d\mathcal{E}}\mathcal{S}[R_+(\mathcal{E}),Q'(\mathcal{E})]
=\mathcal{O}(1)+\mathcal{O}\left((\mathcal{E}_c-\mathcal{E})^{n/2-1}\right). \tag*{(170)} T H ( E ) 1 − Φ + ( E ) ( ∂ E Q em ) ( E ) = − d E d S [ R + ( E ) , Q ′ ( E )] = O ( 1 ) + O ( ( E c − E ) n /2 − 1 ) . (170)
Although 1 / T H 1/T_H 1/ T H alone is not integrable for n ≥ 2 n\geq2 n ≥ 2 , the full charged integrand remains integrable for every n ≥ 1 n\geq1 n ≥ 1 . Tangency can thus change the endpoint power without making the tunneling action divergent.
Details on the nonlinearly charged BTZ black hole
This Appendix collects the supporting matter, horizon-adapted parametrizations, and fixed-coupling thermodynamics of the static nonlinearly charged BTZ family used in § 4 . The parametrizations extend the stationary-point analysis of Ref. [19 ] by resolving the one, two, and three-horizon configurations explicitly.
Plebanski formulation of the supporting matter
Consider Einstein gravity coupled to nonlinear electrodynamics in Plebanski’s H ( P ) \mathcal{H}(\mathcal{P}) H ( P ) formulation [42 ] . The electromagnetic fields are the gauge potential A μ A_\mu A μ and an independent antisymmetric tensor P μ ν P^{\mu\nu} P μν , with invariant P : = P μ ν P μ ν / 4 \mathcal{P}:=P_{\mu\nu}P^{\mu\nu}/4 P := P μν P μν /4 . The first-order action is
I [ g μ ν , A μ , P μ ν ] = ∫ d 3 x − g [ 1 2 κ ( R + 2 ℓ 2 ) + 1 2 P μ ν F μ ν − H ( P ) ] . (171) I[g_{\mu\nu},A_\mu,P^{\mu\nu}]=\int d^3x\sqrt{-g}\left[\frac{1}{2\kappa}\left(R+\frac{2}{\ell^2}\right)+\frac{1}{2}P^{\mu\nu}F_{\mu\nu}-\mathcal{H}(\mathcal{P})\right]. \tag*{(171)} I [ g μν , A μ , P μν ] = ∫ d 3 x − g [ 2 κ 1 ( R + ℓ 2 2 ) + 2 1 P μν F μν − H ( P ) ] . (171)
with F μ ν = 2 ∂ [ μ A ν ] F_{\mu\nu}=2\partial_{[\mu}A_{\nu]} F μν = 2 ∂ [ μ A ν ] . It is worth pointing out that this ( P , H ) (\mathcal{P},\mathcal{H}) ( P , H ) formulation is especially suitable for constructing non-rotating and rotating black hole configurations (see Refs. [61 , 62 , 63 , 64 , 65 , 66 ] ). The theory supporting the black hole family of Ref. [19 ] is specified by
H ( P ) = β − 4 P ∣ β γ ∣ [ 1 − ln ( − 4 P ∣ β γ ∣ ) ] , P < 0. (172) \mathcal{H}(\mathcal{P})=\frac{\beta\sqrt{-4\mathcal{P}}}{|\beta\gamma|}\left[1-\ln\left(\frac{\sqrt{-4\mathcal{P}}}{|\beta\gamma|}\right)\right],\qquad\mathcal{P}<0. \tag*{(172)} H ( P ) = ∣ β γ ∣ β − 4 P [ 1 − ln ( ∣ β γ ∣ − 4 P ) ] , P < 0. (172)
The couplings ( κ , ℓ , β , γ ) (\kappa,\ell,\beta,\gamma) ( κ , ℓ , β , γ ) are held fixed, with β γ ≠ 0 \beta\gamma\neq0 β γ = 0 . The additive constant in H \mathcal{H} H is set to zero, keeping the cosmological scale ℓ \ell ℓ fixed as in the metric.
Variation with respect to A μ A_\mu A μ and P μ ν P^{\mu\nu} P μν gives, respectively,
∇ μ P μ ν = 0 , F μ ν = H P P μ ν , (173) \nabla_\mu P^{\mu\nu}=0,\qquad F_{\mu\nu}=\mathcal{H}_{\mathcal{P}}P_{\mu\nu}, \tag*{(173)} ∇ μ P μν = 0 , F μν = H P P μν , (173)
where H P : = ∂ H / ∂ P \mathcal{H}_{\mathcal{P}}:=\partial\mathcal{H}/\partial\mathcal{P} H P := ∂ H / ∂ P . The metric variation yields standard Einstein equations G μ ν − ℓ − 2 g μ ν = κ T μ ν G_{\mu\nu}-\ell^{-2}g_{\mu\nu}=\kappa T_{\mu\nu} G μν − ℓ − 2 g μν = κ T μν , with
T μ ν = − H P P μ α P ν α + ( 2 P H P − H ) g μ ν . (174) T_{\mu\nu}=-\mathcal{H}_{\mathcal{P}}P_{\mu\alpha}P_\nu{}^\alpha+(2\mathcal{P}\mathcal{H}_{\mathcal{P}}-\mathcal{H})g_{\mu\nu}. \tag*{(174)} T μν = − H P P μα P ν α + ( 2 P H P − H ) g μν . (174)
Here the constitutive relation has been used after varying the action. For the static metric Eq. (54) and a purely electric potential A = A t ( r ) d t A=A_t(r)\mathrm{d}t A = A t ( r ) d t , the first equation in Eq. (D.3) integrates to
r P r t = q , P r t = − q r , P ( r ) = − q 2 2 r 2 . (175) r\mathcal{P}^{rt}=q,\qquad\mathcal{P}_{rt}=-\frac{q}{r},\qquad\mathcal{P}(r)=-\frac{q^2}{2r^2}. \tag*{(175)} r P r t = q , P r t = − r q , P ( r ) = − 2 r 2 q 2 . (175)
In the sector β γ / q > 0 \beta\gamma/q>0 β γ / q > 0 , the constitutive relation gives
F r t = ∂ r A t = 2 γ ln ( β γ r 2 q ) . (176) F_{rt}=\partial_r A_t=\frac{\sqrt{2}}{\gamma}\ln\left(\frac{\beta\gamma r}{\sqrt{2}q}\right). \tag*{(176)} F r t = ∂ r A t = γ 2 ln ( 2 q β γ r ) . (176)
Integrating this expression recovers the potential in Eq. (56) , with its additive gauge constant kept as specified there.
In this sector, the stress tensor obeys T t t = T r r = − H T^t_t=T^r_r=-\mathcal{H} T t t = T r r = − H and T φ φ = 2 P H P − H T^\varphi_\varphi=2\mathcal{P}\mathcal{H}_{\mathcal{P}}-\mathcal{H} T φ φ = 2 P H P − H . The radial Einstein equation therefore becomes
∂ r f ( r ) 2 r − 1 ℓ 2 = − κ H [ P ( r ) ] = − 2 κ q γ r [ 1 + ln ( β γ r 2 q ) ] . (177) \frac{\partial_r f(r)}{2r}-\frac{1}{\ell^2}=-\kappa\mathcal{H}[\mathcal{P}(r)]=-\frac{\sqrt{2}\kappa q}{\gamma r}\left[1+\ln\left(\frac{\beta\gamma r}{\sqrt{2}q}\right)\right]. \tag*{(177)} 2 r ∂ r f ( r ) − ℓ 2 1 = − κ H [ P ( r )] = − γ r 2 κ q [ 1 + ln ( 2 q β γ r ) ] . (177)
Its integration gives Eq. (55) , with M M M as the remaining integration constant. The angular Einstein equation is satisfied by the same solution.
It is worth noticing that the structural function and constitutive relation are single-valued and smooth on the electric domain P < 0 \mathcal{P}<0 P < 0 , including P = − β 2 γ 2 / 4 \mathcal{P}=-\beta^2\gamma^2/4 P = − β 2 γ 2 /4 , where H P = 0 \mathcal{H}_{\mathcal{P}}=0 H P = 0 and the electric field vanishes. Since P ( r ) \mathcal{P}(r) P ( r ) is monotonic for nonzero q q q , this formulation provides the supporting matter throughout r > 0 r>0 r > 0 . Moreover, neither M M M nor q q q enters the structural function: both integration constants may vary within the same theory while ( κ , ℓ , β , γ ) (\kappa,\ell,\beta,\gamma) ( κ , ℓ , β , γ ) remain fixed.
Horizon-adapted parametrizations
For fixed ( q , β , γ , ℓ ) (q,\beta,\gamma,\ell) ( q , β , γ , ℓ ) , the positive roots of Eq. (55) determine the Killing horizons of the geometry. It is nevertheless useful to eliminate combinations of the original parameters in favor of the horizon radii themselves. This produces representations adapted to a prescribed root configuration and makes the corresponding degeneracy conditions manifest. For fixed couplings, the horizon radii remain subject to the parameter relations inherited from Eq. (55) .
Consider first a geometry with a positive horizon at r = r + r=r_+ r = r + . Using f ( r + ) = 0 f(r_+)=0 f ( r + ) = 0 to eliminate the constant logarithmic contribution, the blackening function can be written as
f ( r ) = r 2 ℓ 2 − M − 2 2 κ q r γ r + ln ( r r + ) − r r + ( r + 2 ℓ 2 − M ) . (178) f(r)=\frac{r^2}{\ell^2}-M-\frac{2\sqrt{2}\kappa q r}{\gamma r_+}\ln\left(\frac{r}{r_+}\right)-\frac{r}{r_+}\left(\frac{r_+^2}{\ell^2}-M\right). \tag*{(178)} f ( r ) = ℓ 2 r 2 − M − γ r + 2 2 κ q r ln ( r + r ) − r + r ( ℓ 2 r + 2 − M ) . (178)
This form satisfies f ( r + ) = 0 f(r_+)=0 f ( r + ) = 0 identically. In the sector q / γ < 0 q/\gamma<0 q / γ < 0 , Eq. (71) shows that the mass–radius curve is strictly convex. Together with f ( 0 + ) = − M < 0 f(0^+)=-M<0 f ( 0 + ) = − M < 0 and f ( r ) → + ∞ f(r)\to+\infty f ( r ) → + ∞ as r → ∞ r\to\infty r → ∞ , this ensures that a positive mass geometry has exactly one simple positive horizon.
A form adapted to two prescribed positive roots 0 < r − < r + 0<r_-<r_+ 0 < r − < r + follows by imposing f ( r − ) = f ( r + ) = 0 f(r_-)=f(r_+)=0 f ( r − ) = f ( r + ) = 0 . Eliminating the logarithmic coefficient gives
f ( r ) = − M + r 2 ℓ 2 − r r + ( r + 2 ℓ 2 − M ) − r r − [ r 2 ℓ 2 − M − r r + ( r + 2 ℓ 2 + M ) ] ln ( r / r + ) ln ( r − / r + ) . (179) \begin{aligned}
f(r)={}&-M+\frac{r^2}{\ell^2}-\frac{r}{r_+}\left(\frac{r_+^2}{\ell^2}-M\right)\\
&-\frac{r}{r_-}\left[\frac{r^2}{\ell^2}-M-\frac{r}{r_+}\left(\frac{r_+^2}{\ell^2}+M\right)\right]\frac{\ln(r/r_+)}{\ln(r_-/r_+)}. \tag*{(179)}
\end{aligned} f ( r ) = − M + ℓ 2 r 2 − r + r ( ℓ 2 r + 2 − M ) − r − r [ ℓ 2 r 2 − M − r + r ( ℓ 2 r + 2 + M ) ] ln ( r − / r + ) ln ( r / r + ) . (179)
The conditions f ( r − ) = f ( r + ) = 0 f(r_{-})=f(r_{+})=0 f ( r − ) = f ( r + ) = 0 alone do not exclude an additional positive root between them. The two-horizon configuration selected in § 4 is obtained when the inner zero is degenerate,
f ( r − ) = ∂ r f ( r − ) = 0 , ∂ r 2 f ( r − ) < 0 , (180) f(r_{-})=\partial_{r}f(r_{-})=0,\qquad\partial_{r}^{2}f(r_{-})<0, \tag*{(180)} f ( r − ) = ∂ r f ( r − ) = 0 , ∂ r 2 f ( r − ) < 0 , (180)
while r + r_{+} r + remains a simple event horizon.
To express this restriction more efficiently, we define
α : = r − r + , μ : = M ℓ 2 r + 2 , 0 < α < 1. (181) \alpha:=\frac{r_{-}}{r_{+}},\qquad\mu:=\frac{M\ell^{2}}{r_{+}^{2}},\qquad0<\alpha<1. \tag*{(181)} α := r + r − , μ := r + 2 M ℓ 2 , 0 < α < 1. (181)
The condition ∂ r f ( r − ) = 0 \partial_{r}f(r_{-})=0 ∂ r f ( r − ) = 0 fixes the dimensionless mass,
μ ( α ) = α ( 1 − α + α ln α ) α − 1 − ln α , (182) \mu(\alpha)=\frac{\alpha(1-\alpha+\alpha\ln\alpha)}{\alpha-1-\ln\alpha}, \tag*{(182)} μ ( α ) = α − 1 − ln α α ( 1 − α + α ln α ) , (182)
with M = r + 2 μ ( α ) / ℓ 2 M=r_{+}^{2}\mu(\alpha)/\ell^{2} M = r + 2 μ ( α ) / ℓ 2 . Thus, the initial degenerate configuration may be parametrized by ( r + , α ) (r_{+},\alpha) ( r + , α ) , with r − = α r + r_{-}=\alpha r_{+} r − = α r + . This is the origin of the dependence on r + / r − r_{+}/r_{-} r + / r − displayed by the inverse coefficients in Eq. (70) .
Equations (D.9) and (D.12) parametrize the initial degenerate member of the family. The subsequent horizon positions follow from the original blackening function at the running mass and charge, as in Eq. (66) for neutral emission.
The three-horizon parametrization
Before either pair of roots coalesces, the geometry admits three distinct positive horizons,
0 < r − < r 0 < r + . (183) 0<r_{-}<r_{0}<r_{+}. \tag*{(183)} 0 < r − < r 0 < r + . (183)
Starting from Eq. (D.9), the remaining condition f ( r 0 ) = 0 f(r_{0})=0 f ( r 0 ) = 0 fixes the mass in terms of the three radii. Introducing
r ~ − : = r − r + , r ~ 0 : = r 0 r + , 0 < r ~ − < r ~ 0 < 1 , (184) \widetilde r_{-}:=\frac{r_{-}}{r_{+}},\qquad\widetilde r_{0}:=\frac{r_{0}}{r_{+}},\qquad0<\widetilde r_{-}<\widetilde r_{0}<1, \tag*{(184)} r − := r + r − , r 0 := r + r 0 , 0 < r − < r 0 < 1 , (184)
one obtains
M = M ~ : = r + 2 r ~ 0 r ~ − ℓ 2 ( 1 − r ~ − ) ln r ~ 0 − ( 1 − r ~ 0 ) ln r ~ − r ~ − ( 1 − r ~ 0 ) ln r ~ − − r ~ 0 ( 1 − r ~ − ) ln r ~ 0 . (185) M=\widetilde M:=\frac{r_{+}^{2}\widetilde r_{0}\widetilde r_{-}}{\ell^{2}}\frac{(1-\widetilde r_{-})\ln\widetilde r_{0}-(1-\widetilde r_{0})\ln\widetilde r_{-}}{\widetilde r_{-}(1-\widetilde r_{0})\ln\widetilde r_{-}-\widetilde r_{0}(1-\widetilde r_{-})\ln\widetilde r_{0}}. \tag*{(185)} M = M := ℓ 2 r + 2 r 0 r − r − ( 1 − r 0 ) ln r − − r 0 ( 1 − r − ) ln r 0 ( 1 − r − ) ln r 0 − ( 1 − r 0 ) ln r − . (185)
This expression is positive throughout the ordered domain (D.14). Dividing both its numerator and denominator by ( 1 − r ~ − ) ( 1 − r ~ 0 ) > 0 (1-\widetilde r_{-})(1-\widetilde r_{0})>0 ( 1 − r − ) ( 1 − r 0 ) > 0 leaves differences of ln x / ( 1 − x ) \ln x/(1-x) ln x / ( 1 − x ) and x ln x / ( 1 − x ) x\ln x/(1-x) x ln x / ( 1 − x ) , respectively. For 0 < x < 1 0<x<1 0 < x < 1 , the first function is strictly increasing and the second strictly decreasing; the ordering r ~ − < r ~ 0 \widetilde r_{-}<\widetilde r_{0} r − < r 0 therefore makes both differences positive.
The ordering (D.13), together with f ( 0 + ) = − M < 0 f(0^{+})=-M<0 f ( 0 + ) = − M < 0 and f ( r ) → + ∞ f(r)\to+\infty f ( r ) → + ∞ at large radius, yields the alternating sign structure stated in § 4.2 . Accordingly, r + r_{+} r + is the event horizon, r 0 r_{0} r 0 is the Cauchy horizon associated with the exterior region, and r − r_{-} r − is the innermost positive Killing horizon.
Extremal limits of the three-horizon configuration
The two boundaries of the three-horizon sector follow directly from Eq. (D.15). When the middle root approaches the innermost one, r ~ 0 → r ~ − = α \tilde r_0 \to\tilde r_- = \alpha r ~ 0 → r ~ − = α ,
lim r ~ 0 → r ~ − M ~ = r + 2 ℓ 2 μ ( α ) > 0. (186) \lim_{\tilde r_0\to\tilde r_-}\widetilde{M} = \frac{r_+^2}{\ell^2}\mu(\alpha)>0. \tag*{(186)} r ~ 0 → r ~ − lim M = ℓ 2 r + 2 μ ( α ) > 0. (186)
This reproduces Eq. (D.12) and corresponds to the upper critical configuration (79) , where the two inner branches merge while the event horizon remains simple.
The opposite boundary is reached when the middle root approaches the event horizon, r ~ 0 → 1 \tilde r_0 \to1 r ~ 0 → 1 . Equation (D.15) then gives
lim r ~ 0 → 1 M ~ = r + 2 r ~ − ℓ 2 r ~ − − 1 − ln r ~ − 1 − r ~ − + r ~ − ln r ~ − > 0. (187) \lim_{\tilde r_0\to1}\widetilde{M} = \frac{r_+^2\tilde r_-}{\ell^2}\frac{\tilde r_- - 1-\ln\tilde r_-}{1-\tilde r_-+\tilde r_-\ln\tilde r_-}>0. \tag*{(187)} r ~ 0 → 1 lim M = ℓ 2 r + 2 r ~ − 1 − r ~ − + r ~ − ln r ~ − r ~ − − 1 − ln r ~ − > 0. (187)
This limit describes the lower critical configuration (81) . The middle and event horizons merge into a double outer root, while the innermost horizon remains simple. Unlike the inner coalescence, this boundary has ∂ r f ( r + ) = 0 \partial_r f(r_+)=0 ∂ r f ( r + ) = 0 and hence vanishing Hawking temperature. It is the configuration approached by the neutral emission trajectory in § 4 . The limiting behavior through simple outer horizons is described in § 3.3 .
Revisiting the black hole thermodynamics
We derive the first law by computing the Hamiltonian charge associated with the Killing vector ξ = ∂ t \xi=\partial_t ξ = ∂ t by means of the covariant phase space formalism [27 , 28 ] . Variations are taken within the nonlinearly charged BTZ family at fixed ( κ , ℓ , β , γ ) (\kappa,\ell,\beta,\gamma) ( κ , ℓ , β , γ ) , so only the integration constants ( M , q ) (M,q) ( M , q ) vary. The variation of the first-order action (D.1) gives the surface term
J μ = − g [ 2 P μ ( α β ) γ ∇ γ δ g α β + P μ ν δ A ν ] , (188) \mathcal{J}^{\mu}=\sqrt{-g}\left[2\mathcal{P}^{\mu(\alpha\beta)\gamma}\nabla_\gamma\delta g_{\alpha\beta}+\mathcal{P}^{\mu\nu}\delta A_\nu\right], \tag*{(188)} J μ = − g [ 2 P μ ( α β ) γ ∇ γ δ g α β + P μν δ A ν ] , (188)
with
P μ ν ρ σ : = 1 2 κ ( ∂ R ∂ R μ ν ρ σ ) = 1 4 κ ( g μ ρ g ν σ − g μ σ g ν ρ ) . \mathcal{P}^{\mu\nu\rho\sigma}:=\frac{1}{2\kappa}\left(\frac{\partial R}{\partial R_{\mu\nu\rho\sigma}}\right)=\frac{1}{4\kappa}\left(g^{\mu\rho}g^{\nu\sigma}-g^{\mu\sigma}g^{\nu\rho}\right). P μν ρ σ := 2 κ 1 ( ∂ R μν ρ σ ∂ R ) = 4 κ 1 ( g μ ρ g ν σ − g μ σ g ν ρ ) .
Since P μ ν ρ σ \mathcal{P}^{\mu\nu\rho\sigma} P μν ρ σ is covariantly constant, its derivatives do not contribute to the surface term. The corresponding symplectic potential is the two-form
Θ ( 2 ) = 1 2 ϵ μ ν ρ J μ − g d x ν ∧ d x ρ . \Theta_{(2)}=\frac{1}{2}\epsilon_{\mu\nu\rho}\frac{\mathcal{J}^{\mu}}{\sqrt{-g}}\,dx^\nu\wedge dx^\rho. Θ ( 2 ) = 2 1 ϵ μν ρ − g J μ d x ν ∧ d x ρ .
On shell, the Noether current associated with ξ \xi ξ is locally exact. Using the orientation ϵ t r ϕ = r \epsilon_{tr\phi}=r ϵ t r ϕ = r , the charge one-form is
Q ( 1 ) = 1 2 ϵ α μ ν Q μ ν d x α , (189) Q_{(1)}=\frac{1}{2}\epsilon_{\alpha\mu\nu}Q^{\mu\nu}dx^\alpha, \tag*{(189)} Q ( 1 ) = 2 1 ϵ α μν Q μν d x α , (189)
with
Q μ ν = − 2 P μ ν ρ σ ∇ ρ ξ σ + P μ ν ξ ρ A ρ . (190) Q^{\mu\nu}=-2\mathcal{P}^{\mu\nu\rho\sigma}\nabla_\rho\xi_\sigma+\mathcal{P}^{\mu\nu}\xi^\rho A_\rho. \tag*{(190)} Q μν = − 2 P μν ρ σ ∇ ρ ξ σ + P μν ξ ρ A ρ . (190)
For variations between stationary solutions, the one-form δ Q ( 1 ) − ı ξ Θ ( 2 ) \delta Q_{(1)}-\imath_\xi\Theta_{(2)} δ Q ( 1 ) − ξ Θ ( 2 ) is closed, where ı ξ \imath_\xi ξ denotes contraction with ξ \xi ξ . The Hamiltonian variation
δ H = ∫ Σ ( 1 ) ( δ Q ( 1 ) − ı ξ Θ ( 2 ) ) (191) \delta H=\int_{\Sigma_{(1)}}\left(\delta Q_{(1)}-\imath_\xi\Theta_{(2)}\right) \tag*{(191)} δ H = ∫ Σ ( 1 ) ( δ Q ( 1 ) − ξ Θ ( 2 ) ) (191)
is therefore independent of the circle Σ ( 1 ) \Sigma^{(1)} Σ ( 1 ) on which it is evaluated in the exterior.
For the static ansatz, the electromagnetic first integral reads r P r t = q rP^{rt}=q r P r t = q . The pullbacks of the charge and symplectic potential to a circle of constant t t t and r r r are then
Q ( 1 ) = ( r ∂ r f 2 κ − q A t ) d φ , Q_{(1)}=\left(\frac{r\partial_r f}{2\kappa}-qA_t\right)d\varphi, Q ( 1 ) = ( 2 κ r ∂ r f − q A t ) d φ ,
i ξ Θ ( 2 ) = [ r ∂ r δ f + δ f 2 κ − q δ A t ] d φ . i_\xi\Theta_{(2)}=\left[\frac{r\partial_r\delta f+\delta f}{2\kappa}-q\,\delta A_t\right]d\varphi. i ξ Θ ( 2 ) = [ 2 κ r ∂ r δ f + δ f − q δ A t ] d φ .
Here, variations act on ( M , q ) (M,q) ( M , q ) at fixed r r r . Integrating over φ ∼ φ + 2 π \varphi\sim\varphi+2\pi φ ∼ φ + 2 π gives
δ H = − π κ δ f ( r ) − 2 π A t ( r ) δ q . \delta\mathcal{H}=-\frac{\pi}{\kappa}\delta f(r)-2\pi A_t(r)\delta q. δ H = − κ π δ f ( r ) − 2 π A t ( r ) δ q .
From Eqs. (55) –(56) , the metric variation satisfies
δ f ( r ) = − δ M − 2 κ A t ( r ) δ q . \delta f(r)=-\delta M-2\kappa A_t(r)\delta q. δ f ( r ) = − δ M − 2 κ A t ( r ) δ q .
Thus the growing terms in the metric and gauge sectors cancel, leaving the finite, integrable variation δ H = π δ M / κ \delta\mathcal{H}=\pi\delta M/\kappa δ H = π δ M / κ . With the mass normalization adopted in Eq. (58) ,
δ H ∞ = π κ δ M = δ M . (192) \delta\mathcal{H}_\infty=\frac{\pi}{\kappa}\delta M=\delta\mathcal{M}. \tag*{(192)} δ H ∞ = κ π δ M = δ M . (192)
The gauge is kept as specified in Eq. (56) throughout this construction. At the event horizon, varying f ( r + ) = 0 f(r_+)=0 f ( r + ) = 0 along the stationary family gives
δ f ( r + ) = − ∂ r f ( r + ) δ r + . \delta f(r_+)=-\partial_r f(r_+)\delta r_+. δ f ( r + ) = − ∂ r f ( r + ) δ r + .
Evaluating the same Hamiltonian variation on the limiting horizon circle therefore yields
δ H + = π κ ∂ r f ( r + ) δ r + − 2 π A t ( r + ) δ q = T H δ S + Φ e δ Q , (193) \begin{aligned}
\delta\mathcal{H}_+&=\frac{\pi}{\kappa}\partial_r f(r_+)\delta r_+-2\pi A_t(r_+)\delta q\\
&=T_{\mathrm{H}}\delta S+\Phi_{\mathrm{e}}\delta Q, \tag*{(193)}
\end{aligned} δ H + = κ π ∂ r f ( r + ) δ r + − 2 π A t ( r + ) δ q = T H δ S + Φ e δ Q , (193)
where the entropy, charge, temperature, and potential are those in Eqs. (58) –(60) . The equality δ H ∞ = δ H + \delta\mathcal{H}_\infty=\delta\mathcal{H}_+ δ H ∞ = δ H + establishes the first law (61) .
Finally, the fixed-coupling homogeneity of the mass function, Eq. (82) , gives through Euler’s theorem
2 M = r + ∂ m ( r + , q ) ∂ r + + q ∂ m ( r + , q ) ∂ q = T H S + Φ e Q . (194) \begin{aligned}
2\mathcal{M}&=r_+\frac{\partial m(r_+,q)}{\partial r_+}+q\frac{\partial m(r_+,q)}{\partial q}\\
&=T_{\mathrm{H}}S+\Phi_{\mathrm{e}}Q. \tag*{(194)}
\end{aligned} 2 M = r + ∂ r + ∂ m ( r + , q ) + q ∂ q ∂ m ( r + , q ) = T H S + Φ e Q . (194)
The second equality follows from the first law and the linear dependence of S S S and Q Q Q on r + r_+ r + and q q q , respectively. This recovers the fixed-coupling Smarr relation stated in Eq. (62) . A related derivation from Euler homogeneity for Maxwell-charged AdS3 _3 3 black holes, with boundary conditions preserving the scaling symmetry, is given in Ref. [67 ] . Here the identity follows directly from the NLED mass function and the Hamiltonian normalization above.
Tunneling coefficients for the two-horizon nonlinearly charged BTZ black hole
In this appendix we derive the functions A n A_n A n entering Eq. (70) . The initial inner degeneracy constrains the derivatives of f f f at the event horizon through Eq. (D.9), allowing the coefficients to be written in terms of r + / r − r_+/r_- r + / r − .
Writing ρ : = r + / r − > 1 \rho:= r_+/r_- > 1 ρ := r + / r − > 1 , the initial conditions f 0 ( r − ) = ∂ r f 0 ( r − ) = f 0 ( r + ) = 0 f_0(r_-) = \partial_r f_0(r_-) = f_0(r_+) = 0 f 0 ( r − ) = ∂ r f 0 ( r − ) = f 0 ( r + ) = 0 allow the parameters entering the logarithmic term to be expressed in terms of ( r + , ρ ) (r_+,\rho) ( r + , ρ ) . In particular, the two dimensionless combinations relevant to the expansion at the event horizon are
ν ( ρ ) : = 2 2 κ q ℓ 2 γ r + = ( ρ − 1 ) 2 ρ ( 1 − ρ + ρ ln ρ ) (195) \nu(\rho) := \frac{2\sqrt{2}\kappa q\ell^2}{\gamma r_+} = \frac{(\rho-1)^2}{\rho(1-\rho+\rho\ln\rho)} \tag*{(195)} ν ( ρ ) := γ r + 2 2 κ q ℓ 2 = ρ ( 1 − ρ + ρ ln ρ ) ( ρ − 1 ) 2 (195)
σ ( ρ ) : = ℓ 2 r + ∂ r f 0 ( r + ) = ( ρ − 1 ) [ ( ρ + 1 ) ln ρ − 2 ( ρ − 1 ) ] ρ ( 1 − ρ + ρ ln ρ ) (196) \sigma(\rho) := \frac{\ell^2}{r_+}\partial_r f_0(r_+) = \frac{(\rho-1)[(\rho+1)\ln\rho-2(\rho-1)]}{\rho(1-\rho+\rho\ln\rho)} \tag*{(196)} σ ( ρ ) := r + ℓ 2 ∂ r f 0 ( r + ) = ρ ( 1 − ρ + ρ ln ρ ) ( ρ − 1 ) [( ρ + 1 ) ln ρ − 2 ( ρ − 1 )] (196)
Thus, σ \sigma σ is the dimensionless slope of the initial event horizon, whereas ν \nu ν measures the strength of the logarithmic contribution. For the two-horizon configuration considered here, σ > 0 \sigma> 0 σ > 0 .
A considerable simplification follows from the specific radial dependence of the blackening function (55) . Beyond second order, only the term proportional to r ln r r\ln r r ln r contributes, and its derivatives satisfy
∂ r n f 0 ( r ) = ( − 1 ) n + 1 ( n − 2 ) ! 2 2 κ q γ r n − 1 , n ≥ 3. (197) \partial_r^n f_0(r) = (-1)^{n+1}(n-2)!\frac{2\sqrt{2}\kappa q}{\gamma r^{n-1}}, \qquad n \ge3. \tag*{(197)} ∂ r n f 0 ( r ) = ( − 1 ) n + 1 ( n − 2 )! γ r n − 1 2 2 κ q , n ≥ 3. (197)
At the initial event horizon, ∂ r 2 f 0 ( r + ) = ( 2 − ν ) / ℓ 2 \partial_r^2 f_0(r_+) = (2-\nu)/\ell^2 ∂ r 2 f 0 ( r + ) = ( 2 − ν ) / ℓ 2 . With the dimensionless displacement x : = Δ r + / r + x := \Delta r_+/r_+ x := Δ r + / r + and emitted energy u : = κ ℓ 2 E / ( π r + 2 ) u := \kappa\ell^2\mathcal{E}/(\pi r_+^2) u := κ ℓ 2 E / ( π r + 2 ) , Eq. (66) becomes
0 = u + σ x + x 2 + ν [ x − ( 1 + x ) ln ( 1 + x ) ] . (198) 0 = u + \sigma x + x^2 + \nu\left[x-(1+x)\ln(1+x)\right]. \tag*{(198)} 0 = u + σ x + x 2 + ν [ x − ( 1 + x ) ln ( 1 + x ) ] . (198)
This exact equation retains the full radial dependence of the blackening function and resums the derivative tower above.
Since σ ≠ 0 \sigma\ne0 σ = 0 , the solution connected to the initial event horizon admits a Maclaurin expansion about u = 0 u=0 u = 0 ,
x ( u ) = ∑ n = 1 ∞ A n ( ρ ) u n . (199) x(u) = \sum_{n=1}^{\infty} A_n(\rho)u^n. \tag*{(199)} x ( u ) = n = 1 ∑ ∞ A n ( ρ ) u n . (199)
Applying the Lagrange–Bürmann formula directly to Eq. (E.4) gives the closed expression
A n ( ρ ) = ( − 1 ) n n ! [ d n − 1 d x n − 1 { σ + x + ν [ 1 − 1 + x x ln ( 1 + x ) ] } − n ] x = 0 . (200) A_n(\rho) = \frac{(-1)^n}{n!}\left[\frac{d^{n-1}}{dx^{n-1}}\left\{\sigma+x+\nu\left[1-\frac{1+x}{x}\ln(1+x)\right]\right\}^{-n}\right]_{x=0}. \tag*{(200)} A n ( ρ ) = n ! ( − 1 ) n [ d x n − 1 d n − 1 { σ + x + ν [ 1 − x 1 + x ln ( 1 + x ) ] } − n ] x = 0 . (200)
The expression in braces is understood by its regular extension at x = 0 x=0 x = 0 , where its value is σ \sigma σ . Equation (E.6) determines every order solely in terms of the initial ratio ρ = r + / r − \rho= r_+/r_- ρ = r + / r − .
The first coefficients are
A 1 = − 1 σ , A 2 = − 1 − ν / 2 σ 3 , A 3 = σ ν / 6 − 2 ( 1 − ν / 2 ) 2 σ 5 , A 4 = 1 σ 7 [ − 5 ( 1 − ν 2 ) 3 + 5 6 σ ν ( 1 − ν 2 ) + 1 12 σ 2 ν ] . (201) \begin{aligned}
A_1 &= -\frac{1}{\sigma},\\
A_2 &= -\frac{1-\nu/2}{\sigma^3},\\
A_3 &= \frac{\sigma\nu/6-2(1-\nu/2)^2}{\sigma^5},\\
A_4 &= \frac{1}{\sigma^7}\left[-5\left(1-\frac{\nu}{2}\right)^3+\frac{5}{6}\sigma\nu\left(1-\frac{\nu}{2}\right)+\frac{1}{12}\sigma^2\nu\right]. \tag*{(201)}
\end{aligned} A 1 A 2 A 3 A 4 = − σ 1 , = − σ 3 1 − ν /2 , = σ 5 σ ν /6 − 2 ( 1 − ν /2 ) 2 , = σ 7 1 [ − 5 ( 1 − 2 ν ) 3 + 6 5 σ ν ( 1 − 2 ν ) + 12 1 σ 2 ν ] . (201)
with σ = σ ( ρ ) \sigma=\sigma(\rho) σ = σ ( ρ ) and ν = ν ( ρ ) \nu=\nu(\rho) ν = ν ( ρ ) as given in Eqs. (E.2) and (E.1). Restoring dimensions in Eq. (E.5) reproduces Eq. (70) . In particular,
a 1 = − κ π ∂ r f 0 ( r + ) = − κ 4 π 2 T H a_1=-\frac{\kappa}{\pi\partial_r f_0(r_+)}=-\frac{\kappa}{4\pi^2 T_{\mathrm{H}}} a 1 = − π ∂ r f 0 ( r + ) κ = − 4 π 2 T H κ
so the leading contribution to the tunneling exponent is − E / T H -\mathcal{E}/T_{\mathrm{H}} − E / T H , as required.
The inverse can be continued locally while this outer zero remains simple. The Maclaurin series in u u u about u = 0 u=0 u = 0 , however, reconstructs that branch only within its own convergence domain. At the later merger of the middle and event horizons, the outer slope vanishes and the local inverse takes the Puiseux form developed in Appendix § 11 .