WorksPhysics and Astronomy
Generalized Schwarzian thermodynamics from gravitational perturbations of near-extremal Kerr
Abstract
We derive the generalized Schwarzian thermodynamics of near-extremal Kerr directly from four-dimensional gravitational perturbations of the near-horizon geometry. We construct two families of exact linear perturbations of the near-NHEK background: a non-normalizable mode that deforms the asymptotic boundary, and a normalizable mode supported near the horizon. Both perturbations are smooth solutions of the vacuum Einstein equations reduced to the Jackiw–Teitelboim equation for the dual scalar, without conical singularities. We study their asymptotic symmetries and show that they generate an algebra of time reparametrizations; requiring the gauge parameters to depend on the Schwarzian mode, the isometry condition selects the stationary KdV hierarchy. The same hierarchy is independently enforced by the vacuum Einstein equations on the perturbation coefficient. In the Euclidean regime, we evaluate the Hamiltonian action and show that each mode independently generates a tower of odd-power corrections to the near-extremal entropy.
Introduction
The near-horizon region of extremal black holes is the natural arena where the infrared dynamics of gravity becomes tractable. For the extremal Kerr black hole, the decoupling limit yields the Near-Horizon Extreme Kerr (NHEK) geometry [1], a fibration of a circle over whose isometry group is enhanced to . This enhancement underlies the Kerr/CFT proposal [2] and its extensions to near-extremal configurations [3]. However, the rigid throat does not support finite-energy excitations: any perturbation that carries energy into the throat back-reacts and destroys its asymptotic structure [4, 5, 6]. Moving slightly away from extremality therefore requires breaking the conformal symmetry in a controlled way.
The universal mechanism for this breaking is captured by nearly- holography and Jackiw–Teitelboim (JT) gravity [7, 8]. The reparametrizations of the boundary time are spontaneously and explicitly broken down to , and their low-energy dynamics is governed by the Schwarzian action [9, 10, 11, 12]. These edge modes of the throat control the thermodynamics of near-extremal black holes at low temperatures, including the quantum corrections that resolve the apparent breakdown of the semiclassical description [13, 14, 15, 16, 17, 18]. For Kerr, the near-horizon region was perturbed by a JT sector in [19, 20], where gravitational perturbations of NHEK were organized in terms of the JT equation; recently, more arguments have appear in favour that JT gravity is the effective theory that dominates at low temperature [21]. On the other hand, in the perturbation analyses of [20, 22], the JT sector is sourced by conical singularities of the four-dimensional geometry.
The connection between KdV integrability and black hole thermodynamics was first uncovered in three-dimensional gravity. In [23], it was shown that gravity with KdV boundary conditions admits an infinite set of commuting conserved charges, and that the resulting thermodynamics organizes into a generalized Gibbs ensemble with odd-power corrections to the entropy. This structure was subsequently linked to a bigger integrable asymptotic symmetry algebra of [24, 25], and its two-dimensional avatar—the generalized Schwarzian theory with KdV symmetry-breaking—was developed in [26], where it was shown to govern the thermodynamics of nearly- black holes with an infinite tower of KdV deformations. The present work establishes that this integrable structure is not merely a feature of lower-dimensional models: it is encoded directly in the four-dimensional near-horizon geometry of near-extremal Kerr.
In this article we realize this mechanism explicitly: starting from the integrable structure of the near-NHEK symmetry algebra, we construct two families of exact gravitational perturbations and derive their thermodynamic contributions. Our starting point is the observation that the infinitesimal symmetries of any locally geometry fibred as in NHEK are generated by a scalar dual to the Killing vector, which satisfies precisely the JT equation. Rather than sourcing this scalar with a conical singularity, we promote it to a genuine dynamical field and construct two families of exact linear perturbations of the near-NHEK geometry that are smooth solutions of the vacuum Einstein equations: a non-normalizable mode that deforms the asymptotic boundary, and a normalizable mode localized near the horizon. These two profiles are related by the inversion of the isometry group, and they carry complementary information: is the source seen from the asymptotic boundary, while is the response seen from the horizon.
When the symmetry parameters are allowed to depend locally on the Schwarzian mode , the Killing equations reproduce the stationary Korteweg–de Vries (KdV) hierarchy [23, 26, 27]. In this language the near-NHEK black hole is a particular soliton of the KdV equation, and the near-horizon region is endowed with an infinite set of conserved integrals of motion. The two modes carry different conformal weights under boundary reparametrizations: transforms as a weight-one object and as a weight-two density.
In the Euclidean regime we evaluate the Hamiltonian action within the Regge–Teitelboim formalism [28, 29]. The asymptotic Killing vectors associated with generate a surface term at the outer boundary proportional to the KdV Hamiltonians, reproducing the generalized Schwarzian thermodynamics of [26] with the coupling fixed by the angular momentum; the resulting corrections to the near-extremal entropy are odd powers of the temperature. The near-horizon Killing vectors associated with generate a finite charge at the horizon with the same structure as the asymptotic charge of . Because and correspond to alternative perturbations, with their own asymptotic behaviour, they define independent theories, each with their own generalized Gibbs ensemble, yet both analysis yields an independent tower of odd-power corrections,
where are the KdV chemical potentials sourced by at , and those sourced by at .
The article is organized as follows. § 2 reviews the symmetries of the near-horizon geometry and their reformulation in terms of the dual scalar. § 3.1 constructs the two families of gravitational perturbations. § 4 computes the asymptotic and near-horizon Killing vectors, evaluates the Hamiltonian action in the Euclidean regime, and derives the thermodynamics at the saddle point. We close with our conclusions.
Symmetries of the near-horizon geometry: Gauge transformations via scalar duality
The NHEK geometry [1] is the vacuum solution
with angular functions
where is the mass of the extremal Kerr black hole from which it is obtained by a decoupling limit. The sector is an space, the coordinate parametrizes a fibre with connection , and surfaces of constant are warped geometries [30, 31]. The isometry group is , generated by
so that and is a constant. The vectors (3) close the Lie algebra
and the generators ,
satisfy for , and commute with the generator . The global can be enhanced to the full group of time reparametrizations by allowing the throat to fluctuate, as in nearly- geometries, which are responsible for driving black holes away from extremality. Following [32], we take the improper coordinate transformation1
where and primes denote derivatives with respect to the new time, . It maps (1) into
where
is the Schwarzian derivative, invariant under with . The family (7) contains NHEK for , i.e. , and the near-NHEK black hole [3] for , i.e. , with horizon at . The latter describes a closer zoom into the horizon of a slightly non-extremal Kerr black hole, where measures the deviation from extremality, with Hawking temperature in terms of the decoupling parameter [3].
Following [20], we consider the geometry
where , , are the angular functions defined in (2), is now any locally AdS metric and is a vector field. With this ansatz, the vacuum Einstein equations with vanishing cosmological constant
reduce to the conditions
Here is the curvature from the two-dimensional metric and denotes the Levi-Civita tensor with sign convention .
We look for symmetries of this general geometry (9). We start by proposing the vector field
i.e., the vector field has no azimuthal component and is independent of . From the equation , one gets the condition
The equation , reads
Now we insert (11), reducing (12) to
Since only and depend on and these are linearly independent functions, this gives the pair of equations
The remaining variation is given by , that decouples into the variation of the 2-dimensional metric and the gauge field . Considering again the linear independence of and and (14) we get,
Finally, the equation , is identically satisfied. As a summary, we have the set of constraint equations
and the set of gauge transformations
From the above set of transformations, it becomes clear that part of the symmetry transformations along can be recovered from the ones associated to the 2-dimensional metric. Moreover, they can be rewritten considering that and contracting the transformation with the antisymmetric tensor,
To obtain the equation (21), we have considered the equation of motion , which is simply a recast of the second equation in (10). Here we focus on the fact that the symmetries of the full fibration (9) can be encoded in a single scalar function. Following [20], one can show that the Killing equations for the ansatz (9), with independent of , reduce after separation of the angular dependence to the two-dimensional conditions
where is a scalar, is an integration constant, and the two-dimensional component of the Killing vector is recovered as . The first equation in (22) is precisely the Jackiw–Teitelboim equation: the Hessian of is pure trace on the background , so plays the role of the JT dilaton.
For the near-NHEK metric (7), the solution of (22) renders
with an arbitrary function of time. The component then yields the transformation law of the Schwarzian mode,
which is the action of a Virasoro coadjoint vector on a stress tensor of central charge by restoring units through with central charge . The transformation of under a second Killing parameter ,
reproduces the Lie bracket (4).
Gravitational perturbations
In this section we construct two families of exact linear perturbations of the (near)-NHEK background using the gauge formulation of constant-curvature spacetimes. The key feature is that the perturbation can be recast as the Jackiw–Teitelboim equation, so the KdV hierarchy emerges directly from the vacuum Einstein equations. The two radial profiles have complementary fall-offs: one decays faster than the background as , while the other is well-defined near .
Near-NHEK as a solution of the KdV equation and integrable perturbations
To propose new gravitational perturbations, we draw on the equations (10). First, we construct a locally spacetime using the first-order formulation of the constant-curvature equation in 2D [34, 35, 36], which has been extensively used in the literature to construct 2D asymptotic conditions [37, 38, 39]. Here, its dynamics can be recast in terms of the JT scalar. In a second step, this enables us to construct a 4D spacetime using the ansatz (9), demanding the radial dependence to decay faster than the near-NHEK background.
Locally AdS2 metric
We rewrite the constant curvature equation in terms of the one-form connection ,
being the field strength. The equation (26) encompass the first order equations
by means of the relationship with the first order fields ,
for 2, where is the antisymmetric invariant tensor. Here the relevant algebra is ,
in which the fields are spanned. Additionally, the associated invariant bilinear form is and the metric components in terms of the vielbein can be retrieved using the formula
It is convenient to work in the basis, as it is isomorphic to . Being with the generators of , the transformations among the two basis are given by
Their algebra satisfy the commutation relation , and the only non-zero components of the invariant bilinear form are given by and 3.
Using (28) that relate the components of vielbein and the spin connection with the gauge fields and the transformation (31), one can establish the relationship
Here we propose the conditions
where , and are free functions which are constrained by constant curvature equation (26). A similar gauge choice was previously considered in [26] to recover nearly AdS spacetimes. Indeed, analizing the and components of (26), one get the equations,
that relate the functions with . Employing now the relationship between the 2D metric and the gauge field components (30), leads to the following ansatz for any locally AdS geometry4,
Using (35) and (34), the constant negative curvature equation is solved by,
As one can observe, the equation is only characterized by temporal derivatives of whereas the radial dependence is left unconstrained. We will fix the radial dependence in the next subsection, depending on the features of the fall-off perturbation.
Radial redundancy and the two independent modes
The freedom in the radial profile of can be seen as reflection of a deeper symmetry of AdS. The isometry group SL(2, R) acts on the Poincaré coordinate by inversions , which interchanges the two independent power-law solutions of the radial part of the gauge connection. Concretely, adopting the Poincaré coordinate , the two linearly independent radial profiles
are mapped into each other by the inversion , confirming that the radial coordinate is redundant up to this SL(2, R) freedom. In the standard AdS/CFT language [40, 41, 42], has conformal dimension (non-normalizable mode, source) while has conformal dimension (normalizable mode, response or VEV), with the two related by the inversion symmetry.
Both profiles satisfy the same temporal equation (45) independently, so the full two-parameter family
is a solution for any pair satisfying the KdV hierarchy equation
Perturbed near-NHEK
Using (10) and (9), we aim to construct a spacetime of the form,
where is the background spacetime associated to the near-NHEK solution (7) including the Schwarzian mode, and is an axisymmetric perturbation.
We take the locally AdS spacetime found in the previous subsection (35) and exploit the two independent radial profiles (37) to construct two distinct classes of gravitational perturbation.
Asymptotic perturbation ().
We first select the non-normalizable profile , which ensures that the deviation from the NHEK geometry decays faster than the background as . Both perturbations constructed below are smooth vacuum solutions with no conical singularities at the poles [22]. This choice determines the vector via (10),
and leads to the perturbation (renaming for brevity),
All components decay as , and the perturbation vanishes when .
Near-near-horizon perturbation ()
The complementary radial mode vanishes at and grows toward the asymptotic boundary. We refer to it as the near-near-horizon perturbation because it is regular at the near-NHEK horizon, in analogy with the near-horizon character of the NHEK geometry itself relative to the full Kerr spacetime. In the AdS/CFT dictionary, is the normalizable mode in the sense that its coefficient is not freely prescribed as an asymptotic boundary condition but is instead determined dynamically as the response (VEV) of the dual operator. The two modes play complementary roles: sources the operator from the boundary at , while encodes the near-horizon response. Substituting into (34) gives
and the vector , determined to linear order in from the curvature condition (10), reads
Inserting (43) into (9) yields the near-near-horizon perturbation,
In contrast to (41), the dominant components (, , , ) grow with , while decays. The growing behavior of the metric components makes this mode incompatible with the asymptotic boundary conditions imposed for ; it is instead well-defined as a perturbation near , where the ratio , confirming that is a valid small perturbation only in the near-near-horizon region of the NHEK. Note also that receives no contribution from at linear order, a structural difference from the asymptotic perturbation. The perturbation vanishes when .
The two perturbations (41) and (44) satisfy the same equations of motion. The presented spacetimes are Ricci flat solution, where
which is the generalized stationary KdV equation when are the Gelfand–Dikii polynomials. This equation can be rewritten as
where is the nearly AdS spacetime, associated to the background metric (7). By taking the trace of the above equation, it reduces to
which is the wave equation in a nearly AdS background, which was previously obtained in [20] as the mode of an axisymmetric perturbation, but sourced by a conical singularity. Here this equation does not come from a (singular) mode expansion of the perturbation. Indeed, in the present case, the angular part is fully solved by the functional form of and .5
The perturbations give a non-trivial dynamics to the Schwarzian mode by means of the scalar , which reduces to the JT equation with no conical singularities. It is straightforward to see by fixing the coordinates in the metric (39) perturbed by (41). In doing so, the spacetime reduces to the sphere one,
For the asymptotic perturbation (41), a power expansion of the linearized Ricci components gives while the remaining components are or smaller, confirming that produces a well-defined asymptotic perturbation. For the near-near-horizon perturbation (44), the dominant components of grow as , so the linearized Ricci tensor grows as ; near , however, the relative perturbation , confirming that is a valid small perturbation only in the near-near-horizon region.
Similar simplifications occur in the remaining components.
In the Appendix A, we study the propagating content of the asymptotic perturbation , that can be characterised via the Newman–Penrose Weyl scalars and . The unperturbed near-NHEK background is of Petrov type D (), so any perturbation that makes and non-vanishing breaks the type-D structure and carries gravitational degrees of freedom with helicity . Using the null tetrad adapted to the near-NHEK geometry, one finds that and are both non-zero for generic , confirming that the perturbation carries propagating gravitational degrees of freedom; the explicit expressions, computed in the asymptotic region . The equations of motion (45) relate to the KdV scalar, so the propagating content of the asymptotic perturbation is governed by the same integrable hierarchy.
Asymptotic symmetries and Hamiltonian action
We construct the Hamiltonian action of the near-NHEK geometry within the Regge–Teitelboim Hamiltonian formalism [28]. The symmetries and thermodynamic properties of NHEK geometries have been studied extensively in the literature [43, 44, 45, 46, 47, 48, 49]. Working in the Euclidean regime, the on-shell action reduces entirely to a boundary surface term whose value at the saddle point reproduces the generalized Schwarzian thermodynamics. The technical details of the Euclidean section, the regularity condition, and the derivation of the surface term are collected in Appendix B; here we present the asymptotic symmetry algebra that leave invariant the two set of perturbations (41) and (44), the conserved charges from which the Hamiltonians are constructed, and the thermodynamic results.
Asymptotic symmetries
Symmetries of the asymptotic perturbation . The set of transformations that preserve the form of the perturbed near-NHEK metric (39) are given by the equation
provided the asymptotic Killing vector is determined by the components
Here is the generator parameter and is an arbitrary function of time related to the time reparametrization symmetry. Additionally, the symmetry transformations are
where we observe that scalar perturbation function transforms as a vector and as a spin-two density.
It is important to note that the above asymptotic Killing vectors (52) reduce to the ones obtained for the previously analized cases in Section 2, for , and when we additionally set , to the symmetries (3).
Near-near-horizon symmetries of . The near-near-horizon perturbation introduced in (37) defines a complementary set of asymptotic symmetries, now associated with the inner boundary . Using the scalar-duality construction of Section 2, one can find the exact asymptotic Killing vectors that preserve the full perturbed metric with . They read
where is an arbitrary function of time. As with , these reduce to , , , reproducing the leading-order result familiar from the unperturbed case. The exact form has the same structural character as (52) for : in both cases the denominator involves a single combination of and , namely for the non-normalizable mode and for the normalizable one, reflecting their different conformal weights.
Under (54), the near-horizon data transform as
Crucially, transforms as a weight-two density under reparametrizations, in contrast with , which transforms as a weight-one vector. This difference in conformal weight is the origin of the even versus odd power series in the temperature in the thermodynamic partition function, as we discuss in Section 4.5.
KdV symmetries and conserved charges
The asymptotic Killing vectors (52) and (54) depend on the background through , so the reparametrization parameters and may themselves be functionals of and its derivatives, and . Allowing such field-dependent parameters does not break the algebra of time reparametrizations established in § 2; it selects a particular field-dependent representative within that infinite-dimensional algebra. Following [23, 26], the condition that such a field-dependent vector generates an integrable charge — that is, that is an exact variation on the phase space of near-NHEK geometries — requires
and the same condition holds for . This is the stationary KdV equation for the Schwarzian mode : it is not an additional restriction on the dynamics but rather the integrability condition that selects, within the full reparametrization algebra, those generators whose associated charges are well-defined functionals on phase space. The solutions are the Gelfand–Dikii polynomials , generated by the bi-Hamiltonian recursion
whose first two non-trivial members are
with integration constants. Each satisfies (56) by construction: , so the integrability condition is . Applied to this gives (the rigid near-NHEK saddle), while applied to it gives the stationary KdV equation
and each higher member yields the next equation of the stationary hierarchy. The near-NHEK geometry () is the common fixed point of all flows simultaneously. Crucially, as shown in § 3.1, the vacuum Einstein equations impose precisely (56) on the perturbation coefficient via (45), so the KdV hierarchy is not merely a kinematic feature of the symmetry algebra but is dynamically enforced once is promoted to a dynamical field. The associated conserved charges
form an infinite set of commuting integrals of motion that label the near-NHEK saddle. The same set of charges is generated by both and , since both parameters satisfy the same integrability condition (56); as we will see in the next subsection, the two modes therefore contribute to the same KdV hierarchy, but from complementary boundaries, as made explicit in the thermodynamic analysis.
Euclidean action and thermodynamics
The role of the infinite set of symmetries found in the previous sections becomes transparent in the Euclidean regime. We continue and compute the on-shell value of the Hamiltonian action on the perturbed near-NHEK geometry (39). The regularity condition that fixes the thermal period , whose derivation is presented in Appendix § 8, here, we develop the key result that the on-shell action reduces entirely to the boundary term (68), which at the saddle point reproduces the generalized Schwarzian thermodynamics of [26] with a coupling fixed by the angular momentum.
The two radial modes introduced in (37) correspond to alternative quantization schemes — standard (source, ) and alternate (response, ) — and therefore define independent theories. Within each theory, the relevant mode contributes to the partition function from its own boundary: from and from . The two modes are related by the inversion of the isometry group (37), which maps the asymptotic boundary to the horizon and exchanges . As a consequence, the near-horizon boundary action of has the same structure as the asymptotic one of , and both independently generate odd-power corrections to the entropy of the same form. The asymptotic mode is treated first in Section 4.4; the near-horizon mode is treated in Section 4.5.
Surface term and integrability
In what follows, the analysis will carried out for the asymptotic mode , whose associated symmetries are the Killing vectors (52) generated by . For simplicity we take . The Euclidean action in Hamiltonian form reads6
On-shell the constraints vanish. The kinetic term evaluates to
which vanishes because is periodic in Euclidean time. The on-shell action is therefore entirely given by the surface term . Using the asymptotic symmetries (52) as deformation parameters (B.5), the variation of the charge at large radius is
The divergent term is removed by holographic renormalization: placing the boundary at and rescaling , then taking with ,
The term vanishes at the saddle but encodes the coupling between the JT scalar and the reparametrization mode relevant for fluctuations. Since the phase space is one-dimensional ( is the only variation), the boundary term is always integrable:
Parametrizing , with , independent of and ,
Choosing (the Gelfand-Dikii polynomials (57)) and using the identity
the boundary term becomes
This is the boundary action of [26], , now derived from the four-dimensional surface term with a coupling set by . The dictionary with [26] is
Saddle point and thermodynamics
Extremizing over the reparametrization mode, , and using that is antisymmetric under integration by parts,
The saddle point condition is therefore the stationary KdV equation , which coincides with the Killing condition (56). It states that is constant along the thermal cycle, not that it vanishes. The constant representative solves it for every , and it is selected by the regularity condition (B.9).
At the saddle point all Gelfand-Dikii polynomials are constants. From (57), , , , one finds
Consequently and the term coupling in (68) drops out. The perturbation function is also constant at the saddle. From , it scales as . Hence the perturbation survives at extremality, while the higher KdV deformations switch off as .
The on-shell action is given by the surface term, , supplemented by the contribution of the horizon, which in the Hamiltonian approach yields the Bekenstein-Hawking entropy of extremal Kerr, [50]. Evaluating (68) with (71), the partition function at the saddle point is
The term linear in plays the role of the rotational chemical potential and does not contribute to the entropy. The energy above extremality and the entropy are
and they satisfy the first law at fixed , , with specific heat . The comparison with [26], where , shows that our result coincides with the generalized Schwarzian thermodynamics after the identification
Thermodynamic stability, , requires , which for selects odd values of . This mirrors the condition found in [26] for the convergence of the one-loop Gaussian integrals, and it contains the Schwarzian case , where is linear in the temperature, as expected for near-extremal black holes.
KdV symmetries at the saddle: generalized Gibbs ensemble
The constant representative is a common fixed point of all the KdV flows: since every is constant, for all . Hence, nothing prevents us from switching them on simultaneously, , which turns (68) into a generalized Gibbs ensemble,
where the are the chemical potentials of the higher KdV charges. This gives a precise meaning to the integration constants of the recursion (57): a single generator with integration constants, with , is already a combination of this type with . Since the regularity condition (B.9) does not depend on the boundary action, the saddle point is still , and
At low temperatures the entropy is dominated by the linear Schwarzian term, and the higher KdV charges provide odd-power corrections. The values of the KdV charges at the saddle follow from (67) and (71),
Near-horizon mode and its thermodynamic charge
We now evaluate the Regge–Teitelboim charge for the complementary near-horizon mode introduced in (37), whose asymptotic symmetries are the near-near-horizon vectors (54) generated by .
Similarly to the previous case, we analyze the kinetic contribution to the Euclidean action which one can show that is total time-derivative in . Since satisfies the same periodic boundary conditions as , the integral over the thermal cycle vanishes identically, exactly as in (62) for the asymptotic mode.
We now evaluate the Regge–Teitelboim charge for the complementary near-horizon mode introduced in (37), whose asymptotic symmetries are the near-near-horizon vectors (54) generated by . The contribution along remains the same as in the previous case. The near-near-horizon Killing vector (54) determines the deformation vector (B.5) lapse via the ADM decomposition. Substituting (3.19) into the Regge–Teitelboim integrand (B.3), and integrating over the horizon 2-sphere, one obtains
This is the near-horizon analogue of (63).
At the saddle , the integrability condition requires . The coefficient is not a free parameter: under the near-horizon vector (54) its transformation law (4.6) is
identifying as a weight-two primary. At the saddle , requiring that be invariant under the flow generated by forces to satisfy the same stationarity condition as itself. The unique weight-two functional of that is stationary under and is polynomial in and its derivatives is the Gelfand-Dikii polynomial one level below,
for a numerical coefficient .
By the inversion symmetry discussed in § 3.1, the charge (4.29) at has the same algebraic structure as the asymptotic charge (4.14) at , with and . We therefore apply the same parametrization strategy as in § 4.4: setting and with , independent of , the boundary term takes the same form as (4.17) with independent coupling constant . Setting and simultaneously and using the identity (4.18), the near-horizon boundary term at levels evaluates at the saddle to
in direct analogy with (4.19). At the saddle, and , so the near-horizon partition function at levels scales as
giving odd powers for each total level . The double sum over with therefore contributes to the same odd-power tower as the asymptotic mode, with independent couplings .
Generalized Gibbs ensemble
Because and correspond to alternative asymptotic behaviours, they define independent theories. In the standard quantization ( sourced at ), the partition function organizes into the generalized Gibbs ensemble
reproducing the generalized Schwarzian thermodynamics of [26] with KdV chemical potentials . In the alternate quantization, sourced, the inversion symmetry guarantees that the near-horizon partition function takes the same form,
with an independent set of KdV chemical potentials . Both towers run over odd powers , , , of the temperature; the identical parity is a direct consequence of the inversion symmetry relating the two modes. The two ensembles (83)–(84) provide complementary descriptions of the near-extremal Kerr thermodynamics, one organized from the asymptotic boundary and one from the horizon.
Conclusions
We have shown that the near-horizon geometry of near-extremal Kerr supports two families of exact linear gravitational perturbations — a non-normalizable mode and a normalizable mode , both smooth vacuum solutions of the Einstein equations. The key ingredient is the promotion of the JT scalar, dual to the Killing vector of any locally fibration, to a genuine dynamical field without introducing conical singularities. When the symmetry parameters depend on the Schwarzian mode , the Killing equations reduce to the stationary KdV hierarchy, the same condition enforced independently by the vacuum Einstein equations on the perturbation coefficient. The near-NHEK geometry is thus equipped with an infinite tower of commuting conserved charges.
In the Euclidean regime, the two modes contribute to the thermodynamic partition function from complementary boundaries and correspond to alternative quantization schemes. The asymptotic mode generates a surface term that reproduces the generalized Schwarzian thermodynamics of [26] with odd-power temperature corrections to the entropy,
The normalizable mode is regular at the near-NHEK horizon and contributes a finite Regge–Teitelboim charge from the near-horizon boundary. The radial redundancy from which both perturbations were constructed, justify that generates a near-horizon partition function of the same form, with an independent set of KdV chemical potentials. Both theories organize into generalized Gibbs ensembles with odd-power temperature corrections, providing complementary descriptions of the near-extremal Kerr thermodynamics: one organized from the asymptotic boundary and one from the horizon.
Several directions remain open. First, the Euclidean path integral beyond the saddle. One example to follow would be the one proposed in [51], starting from warped geometry. The stationary KdV equation (59) admits periodic Jacobi-elliptic solutions
which are further saddle points of generalized Gibbs ensembles with two or more chemical potentials. Their Euclidean regularity requires the Hill operator to have trivial monodromy along the thermal cycle; we defer this monodromy analysis to future work.
Second, the one-loop contribution. In the present four-dimensional setting, promoting the Schwarzian mode to a phase-space variable and accounting for the coupling between the JT scalar and the reparametrization mode requires care. The U(1) mode associated with also contributes to the path-integral measure, as in the one-loop analyses of near-extremal Kerr [13, 14, 15]. Recent computations of one-loop corrections from quasinormal modes [18] and from semiclassical Virasoro blocks [17] suggest that the dominant logarithmic correction originates entirely from the near-horizon region, consistent with the throat geometry we analyze here. For the pure generalized Schwarzian theory the analogous correction is [26]; the four-dimensional counterpart is left for future work. It would also be interesting to identify the holographic dual of the full generalized Gibbs ensemble as a one-dimensional quantum mechanical system with KdV symmetry.
Acknowledgements
M.C. is grateful to I. J. Araya, R. Mancilla and R. Olea, for valuable discussions. The authors also thank C. Muñoz and H. Van de Bosch, who where part of the initial discussions that led to this work. The authors acknowledge support from the Center of Mathematical Modeling (CMM), BASAL fund FB210005 for centers of excellence from ANID-Chile. M.C. also acknowledges support from ANID through Fondecyt Regular grant No. 1231810. J.T. additionally acknowledges support from ANID through Fondecyt Postdoctoral grant No. 3250260.
Newman–Penrose analysis of the asymptotic perturbation
We analyse the propagating content of the perturbations constructed in § 3.1 using the Newman–Penrose formalism [52].
Null tetrad. The null tetrad adapted to the unperturbed near-NHEK background (7) is7
The background (7) is of Petrov type D, with and
By the Goldberg–Sachs theorem, any perturbation that renders and non-vanishing breaks the type-D structure and carries gravitational degrees of freedom with helicity . Computing the Weyl scalars for the asymptotic perturbation (41) in the region using the basis (A.1) gives
Both scalars are non-vanishing for generic , confirming that the asymptotic perturbation carries propagating gravitational degrees of freedom. The computation is performed in the asymptotic region ; a complete analysis valid also near requires extending the tetrad to the full perturbed background and lies beyond the scope of the present work.
Hamiltonian formalism and Euclidean thermodynamics
We collect here the technical details of the Regge–Teitelboim Hamiltonian formalism and the Euclidean computation whose results are used in § 4.
ADM decomposition and canonical generator
The metric is written in ADM form,
where is the temporal coordinate, describes the 3-dimensional boundary hypersurface, and , are the lapse and shift. The Hamiltonian is
where and are the Hamiltonian and momentum constraints (proportional to Einstein’s equations, hence vanishing on-shell) and is the boundary term that makes the variational problem well defined. The canonical generator is , with
with and momentum
The gauge symmetry vectors are related to the asymptotic Killing vectors by
To pass to the radial-foliation signature one uses the analytic continuation of [53]: , , .
Euclidean section and regularity
A saddle point of the Euclidean path integral must be a smooth solution. A conical singularity at the horizon would act as a delta-function source of curvature and violate the vacuum field equations. In contrast with [20], where the JT sector is supported by conical sources, we require the absence of any such defect.
Consider (2.7) with a constant Schwarzian mode . The horizon is located at the zero of ,
so that corresponds to the near-NHEK solution . The gauge field does not vanish at the horizon, , and the Killing vector that becomes null there is . Using the co-rotating angle , the fibre reads
Since vanishes quadratically, the fibre does not affect the regularity analysis. Expanding near the horizon with , the Euclidean metric becomes
which is flat in polar coordinates only if has period , i.e.
Because multiplies the whole block, the same period removes the conical singularity at every . In terms of the original angle, the identification is complex, , while is real; this is the familiar complex saddle of rotating black holes, through which enters as the potential conjugate to .
The condition (B.9) admits three equivalent readings. First, ; relative to the asymptotic Kerr time, [3]. Second, the saddle reparametrization is , satisfying . Third, demanding trivial holonomy of the Hill equation along the thermal cycle gives ; is the smooth disk and are -fold covers [26].
The perturbation (41) does not modify (B.9) at the saddle, where : the -dependent components are pure gauge, the -dependent ones vanish quadratically at , and the remaining cross-term is removed by a periodicity-preserving shift of . Hence, to linear order in the perturbation, the JT scalar does not shift the temperature.
References
- [1]J. M. Bardeen and G. T. Horowitz, Phys. Rev. D 60 (1999), 104030 [arXiv:hep-th/9905099 [hep-th]].DOI
- [2]M. Guica, T. Hartman, W. Song and A. Strominger, Phys. Rev. D 80 (2009), 124008 [arXiv:0809.4266 [hep-th]].DOI
- [3]I. Bredberg, T. Hartman, W. Song and A. Strominger, JHEP 04 (2010), 019 [arXiv:0907.3477 [hep-th]].DOI
- [4]O. J. C. Dias, H. S. Reall and J. E. Santos, JHEP 08 (2009), 101 [arXiv:0906.2380 [hep-th]].DOI
- [5]A. J. Amsel, G. T. Horowitz, D. Marolf and M. M. Roberts, JHEP 09 (2009), 044 [arXiv:0906.2376 [hep-th]].DOI
- [6]P. Breitenlohner and D. Z. Freedman, “Stability in Gauged Extended Supergravity,” Annals Phys. 144 (1982), 249–281 doi:10.1016/0003-4916(82)90116-6DOI
- [7]R. Jackiw, “Lower Dimensional Gravity,” Nucl. Phys. B 252 (1985), 343-356 doi:10.1016/0550-3213(85)90448-1.DOI
- [8]C. Teitelboim, “Gravitation and Hamiltonian Structure in Two Space-Time Dimensions,” Phys. Lett. B 126 (1983), 41-45 doi:10.1016/0370-2693(83)90012-6DOI
- [9]A. Almheiri and J. Polchinski, JHEP 11 (2015), 014 [arXiv:1402.6334 [hep-th]].arxiv.org/abs/1402.6334
- [10]K. Jensen, Phys. Rev. Lett. 117 (2016) no.11, 111601 [arXiv:1605.06098 [hep-th]].DOI
- [11]J. Maldacena, D. Stanford and Z. Yang, PTEP 2016 (2016) no.12, 12C104 [arXiv:1606.01857 [hep-th]].arxiv.org/abs/1606.01857
- [12]J. Engelsøy, T. G. Mertens and H. Verlinde, JHEP 07 (2016), 139 [arXiv:1606.03438 [hep-th]].
- [13]L. V. Iliesiu and G. J. Turiaci, JHEP 05 (2021), 145 [arXiv:2003.02860 [hep-th]].DOI
- [14]D. Kapec, A. Sheta, A. Strominger and C. Toldo, Phys. Rev. Lett. 133 (2024) no.2, 021601 [arXiv:2310.00848 [hep-th]].DOI
- [15]I. Rakic, M. Rangamani and G. J. Turiaci, JHEP 06 (2024), 011 [arXiv:2310.04532 [hep-th]].arxiv.org/abs/2310.04532
- [16]N. Banerjee and M. Saha, JHEP 07 (2023), 010 [arXiv:2303.12415 [hep-th]].DOI
- [17]P. Arnaudo, G. Bonelli and A. Tanzini, Phys. Rev. Lett. 134 (2025), 221601 [arXiv:2412.16057 [hep-th]].arxiv.org/abs/2412.16057
- [18]D. Kapec, A. Sheta, A. Strominger and C. Toldo, [arXiv:2410.08048 [hep-th]].
- [19]A. Castro and V. Godet, SciPost Phys. 8 (2020), 089 [arXiv:1906.09083 [hep-th]].DOI
- [20]A. Castro, V. Godet, J. Simón, W. Song and B. Yu, JHEP 07 (2021), 218 [arXiv:2102.08060 [hep-th]].DOI
- [21]A. Castro, R. Mancilla and I. Papadimitriou, JHEP 11 (2025), 083 doi:10.1007/JHEP11(2025)083 [arXiv:2507.01126 [hep-th]].DOI
- [22]F. Mariani and C. Toldo, JHEP 02 (2026), 052 [arXiv:2505.02674 [hep-th]].arxiv.org/abs/2505.02674
- [23]A. Pérez, D. Tempo and R. Troncoso, JHEP 06 (2016), 103 [arXiv:1605.04490 [hep-th]].DOI
- [24]M. Cárdenas, F. Correa, K. Lara and M. Pino, Phys. Rev. Lett. 127 (2021) no.16, 161601 doi:10.1103/PhysRevLett.127.161601 [arXiv:2104.09676 [hep-th]].DOI
- [25]M. Cárdenas, F. Correa and M. Pino, Phys. Rev. D 114 (2026) no.2, 026010 doi:10.1103/n2hm-lk94 [arXiv:2504.20292 [hep-th]].DOI
- [26]M. Cárdenas, JHEP 10 (2024), 052 [arXiv:2405.03128 [hep-th]].DOI
- [27]J. L. Gervais, Phys. Lett. B 160 (1985), 279–282 doi:10.1016/0370-2693(85)91327-9DOI
- [28]T. Regge and C. Teitelboim, Annals Phys. 88 (1974), 286 doi:10.1016/0003-4916(74)90404-7DOI
- [29]G. W. Gibbons and S. W. Hawking, Phys. Rev. D 15 (1977), 2752 doi:10.1103/PhysRevD.15.2752DOI
- [30]D. Anninos, W. Li, M. Padi, W. Song and A. Strominger, JHEP 03 (2009), 130 [arXiv:0807.3040 [hep-th]].DOI
- [31]G. Compère and S. Detournay, JHEP 03 (2009), 098 [arXiv:0808.1911 [hep-th]].arxiv.org/abs/0808.1911
- [32]D. Kapec and A. Lupsasca, Class. Quant. Grav. 37 (2020) no.1, 015006 [arXiv:1905.11406 [hep-th]].DOI
- [33]M. M. Roberts, JHEP 12 (2012), 027 [arXiv:1204.1982 [hep-th]].DOI
- [34]T. Fukuyama and K. Kamimura, Phys. Lett. B 160 (1985), 259–262 doi:10.1016/0370-2693(85)91322-XDOI
- [35]K. Isler and C. A. Trugenberger, Phys. Rev. Lett. 63 (1989), 834–837 doi:10.1103/PhysRevLett.63.834DOI
- [36]A. H. Chamseddine and D. Wyler, Nucl. Phys. B 340 (1990), 595–616 doi:10.1016/0550-3213(90)90460-UDOI
- [37]D. Grumiller, J. Salzer and D. Vassilevich, “AdS₂ holography is (non-)trivial for (non-)constant dilaton,” JHEP 12 (2015), 015 doi:10.1007/JHEP12(2015)015 [arXiv:1509.08486 [hep-th]].DOI
- [38]D. Grumiller, R. McNees, J. Salzer, C. Valcárcel and D. Vassilevich, “Menagerie of AdS₂ boundary conditions,” JHEP 10 (2017), 203 doi:10.1007/JHEP10(2017)203 [arXiv:1708.08471 [hep-th]].DOI
- [39]V. Godet and C. Marteau, “New boundary conditions for AdS₂,” JHEP 12 (2020), 020 doi:10.1007/JHEP12(2020)020 [arXiv:2005.08999 [hep-th]].DOI
- [40]I. R. Klebanov and E. Witten, Nucl. Phys. B 556 (1999), 89–114 [arXiv:hep-th/9905104 [hep-th]].DOI
- [41]I. Ishibashi and R. M. Wald, Class. Quant. Grav. 21 (2004), 2981–3014 [arXiv:hep-th/0402184 [hep-th]].DOI
- [42]D. Marolf and S. F. Ross, JHEP 11 (2006), 085 [arXiv:hep-th/0606113 [hep-th]].DOI
- [43]A. Castro and F. Larsen, “Near Extremal Kerr Entropy from AdS₂ Quantum Gravity,” JHEP 12 (2009) 037, doi:10.1088/1126-6708/2009/12/037, [arXiv:0908.1121 [hep-th]].DOI
- [44]K. Hajian, A. Seraj and M. M. Sheikh-Jabbari, “NHEG Mechanics: Laws of Near Horizon Extremal Geometry (Thermo)Dynamics,” JHEP 03 (2014) 014, doi:10.1007/JHEP03(2014)014, [arXiv:1310.3727 [hep-th]].DOI
- [45]G. Compère, “The Kerr/CFT correspondence and its extensions: a comprehensive review,” Living Rev. Rel. 15 (2012) 11, doi:10.12942/lrr-2012-11, [arXiv:1203.3561 [hep-th]].DOI
- [46]K. Hajian, A. Seraj and M. M. Sheikh-Jabbari, “Near Horizon Extremal Geometry Perturbations: Dynamical Field Perturbations vs. Parametric Variations,” JHEP 10 (2014) 111, doi:10.1007/JHEP10(2014)111, [arXiv:1407.1992 [hep-th]].DOI
- [47]G. Compère, K. Hajian, A. Seraj and M. M. Sheikh-Jabbari, “Wiggling Throat of Extremal Black Holes,” JHEP 10 (2015) 093, doi:10.1007/JHEP10(2015)093, [arXiv:1506.07181 [hep-th]].DOI
- [48]G. Compère, K. Hajian, A. Seraj and M. M. Sheikh-Jabbari, “Extremal Rotating Black Holes in the Near-Horizon Limit: Phase Space and Symmetry Algebra,” Phys. Lett. B 749 (2015) 443–447, doi:10.1016/j.physletb.2015.08.027, [arXiv:1503.07861 [hep-th]].DOI
- [49]K. Hajian, M. M. Sheikh-Jabbari and H. Yavartanoo, “Extreme Kerr black hole microstates with horizon fluff,” Phys. Rev. D 98 (2018) 026025, doi:10.1103/PhysRevD.98.026025, [arXiv:1708.06378 [hep-th]].DOI
- [50]M. Bañados, C. Teitelboim and J. Zanelli, Phys. Rev. Lett. 72 (1994), 957–960 [arXiv:gr-qc/9309026 [gr-qc]].DOI
- [51]E. Despontin, S. Detournay, R. Mancilla and C. Toldo, [arXiv:2607.08482 [hep-th]].arxiv.org/abs/2607.08482
- [52]E. Newman and R. Penrose, J. Math. Phys. 3 (1962), 566–578 doi:10.1063/1.1724257DOI
- [53]M. Banados, A. Schwimmer and S. Theisen, JHEP 05 (2004), 039 [arXiv:hep-th/0404245 [hep-th]].DOI