Amplification of metric perturbations by extremal horizons

A famous prediction of general relativity is that an observer entering a macroscopic black hole is oblivious to his impending doom, feeling nothing special at the moment of horizon crossing. An exception to this rule is the Aretakis instability of extremal black holes [8, 9, 10, 11, 13, 12, 67, 70, 14, 24, 48, 94, 49, 50, 46, 53, 3, 47, 4, 6, 38, 77, 27, 75, 5, 41], which implies that observers momentarily experience large curvature as they cross the horizon; however, the tidal deformation remains small. In this paper we will discuss a related phenomenon in which infalling observers do experience noticeable tidal deformation. We show that nearly co-rotating perturbations of extremal black holes are effectively amplified near the horizon, such that infalling observers experience a sudden increase in tidal deformation as they cross. At precise co-rotation the tidal deformation is formally infinite, heralding a breakdown of perturbation theory in a regime where it would naively be expected to be valid.

The construction is inspired by recent work showing that the smoothness of stationary, axisymmetric extremal horizons in general relativity is in some sense accidental [57, 54, 55, 58, 56]. While that body of work found singularities by incorporating effective field theory corrections or axion fields, we will find analogous behavior within ordinary Einstein-Maxwell theory by going beyond stationarity and axisymmetry. The key point is that horizon co-rotating perturbations (sometimes called “synchronous” [78]) enjoy near-horizon self-similarity in the same way as stationary, axisymmetric perturbations, while allowing a much richer behavior, including fractional and complex scaling exponents.

The potential “danger” of co-rotating perturbations can be seen immediately from the Teukolsky equation [83]. After enforcing the co-rotating condition $\omega=m\Omega_{H}$ (with $\Omega_{H}$ the horizon frequency and $m$ the azimuthal mode number), the usual wave behavior at the extremal horizon $r_{H}=M$ disappears and is instead replaced by power laws in horizon deviation $x=(r-M)/M$ (see [82, 78] or Eq. (29) below),

$\displaystyle{}_{s}\hat{R}_{\ell m(m\Omega_{H})}\sim ax^{h-1+s+im}+bx^{-h+s+im}.$

(1)

Here ${}_{s}\hat{R}_{\ell m(\omega)}$ is the radial dependence of a term in the mode sum for a Weyl scalar $\psi_{s}$ done in horizon-regular coordinates and a horizon-regular tetrad (precise definitions given in the main body), and $h$ is a real or complex number, the “conformal weight” of the mode.

In the presence of power-law end behavior, one normally eliminates a singular power by imposing regularity. However, the Kerr spacetime has “principal modes” where $h=1/2+i\delta$ for real $\delta$. In this case both terms scale as $x^{-1/2+s}$; they are indistinguishable based on regularity, and both are singular for $s\leq 0$. Choosing $s=-2$ corresponds to a component of spacetime curvature seen by an infalling observer, meaning that an observer would see curvature diverging as $x^{-5/2}$. The net tidal deformation $\Delta\xi$ scales like two integrals, so $\Delta\xi\sim x^{-1/2}\to\infty$. There are two logical possibilities: either these modes are somehow impossible to excite, or linearized perturbations can produce large tidal deformation.

To elucidate the situation, we consider nearly co-rotating modes, where the standard ingoing-wave conditions can be imposed. In this case the power laws (1) arise in the regime $k\ll x\ll 1$, where $k=2(\omega-m\Omega_{H})/\Omega_{H}$ is the deviation from precise co-rotation (with a conventional factor of $2$). We show that the modes of the retarded Green function become arbitrarily large as $k\to 0$, with the same scaling $k^{-1/2}$, confirming the presence of fields that produce large tidal deformation. The analysis is done analytically for the Kerr spacetime and numerically for gravitoelectromagnetic (GEM) perturbations of the KN spacetime (mass $M>0$ and charge $Q\geq 0$). Since the region $k\ll x\ll 1$ of large field approaches the horizon as $k\to 0$, we say as a shorthand that the horizon is singular.

These singular modes can be excited physically using a co-rotating source, or at least a source with discrete Fourier support at the co-rotating modes $\omega=m\Omega_{H}$. In the parameter range $Q/M>\sqrt{3}/2$ there are timelike co-rotating equatorial orbits, so a single orbiting particle suffices. For lower black hole charge, one can instead consider a non-circular orbit tuned to have a non-zero co-rotating component. We expect that such a construction is always possible, and provide a definite example in the Kerr spacetime. We also show how to build a precisely co-rotating source from a swarm of such particles. For purely electromagnetic perturbations of Kerr, the source need not execute a geodesic motion, so one may choose a point charge executing a small circular motion about the symmetry axis.

These results can be understood geometrically in terms of the invariant weight of a tensor field under near-horizon scaling [44, 49]. In Ref. [44] this concept was used to understand the black hole Meissner effect [18, 63, 25, 61, 60, 21, 80, 73, 74, 17, 64, 44, 51, 66, 52, 43], showing that the positive integer weight of stationary, axisymmetric perturbations gives rise to the characteristic screening of EM fields.¹¹ 1 Refs. [44, 49] use an opposite sign convention for the definition of weight; see below (69) for details. Later, the notion of weight was applied to the Aretakis instability, where the weight was identified with the late-time power-law decay or growth, with the positive real part indicating that scalar invariants decay [49]. This paper shows that co-rotating modes have self-similar weight with negative real part ($w=-1/2+i\alpha$ for $\alpha\in\mathbb{R}$), indicating singular near-horizon behavior. We utilize the interrelationships between these phenomena (Tab. 1) to confirm the presence of a gravitational Meissner effect and make predictions for the precise decay and growth rate of GEM perturbations of the KN spacetime.

Our analysis shows that there exist linearized solutions of the sourced Einstein-Maxwell equations displaying near-horizon amplification, with observable consequences for infalling observers. However, the construction requires sources that contain exactly co-rotating harmonics (e.g., a particle on a particular bound orbit for all time), which neglects radiation reaction and can only be regarded as a local-in-time approximation to an evolving spacetime. Furthermore, any evolution toward the co-rotating state will presumably evolve the black hole away from an initially extremal state. Further work is required to estimate the degree of amplification that is realizable with smooth evolution from initial data in the full nonlinear theory.

This paper is organized as follows. In Sec. II we review the perturbation framework and introduce notation. In Sec. III we consider co-rotating perturbations of the Kerr spacetime, and in Sec. IV we consider co-rotating perturbations of the KN spacetime. We then review the notion of tensor weight in Sec. V and use it to connect with the Meissner and Aretakis effects in Secs. VI and VII, respectively. Our metric signature is $(-+++)$ and we use units with $G=c=4\pi\epsilon_{0}=1$.

The Kerr-Newman (KN) metric [62, 72] is reviewed in [1]. We consider only the extremal case,

$\displaystyle a=\sqrt{M^{2}-Q^{2}},$

(2)

where $M>0$ is the mass, $Q\geq 0$ is the charge, and $a\geq 0$ is the spin parameter (angular momentum $aM$). The horizon radius and angular frequency are

$\displaystyle r_{H}=M,\qquad\Omega_{H}=\frac{a}{M^{2}+a^{2}}.$

(3)

We will also find it useful to introduce

$\displaystyle\Delta=(r-M)^{2},\qquad\rho=r+ia\cos\theta.$

(4)

We will be using the Newman-Penrose (NP) approach to black hole perturbations [71, 83] throughout this paper. We form the standard NP quantities suitable for perturbations of the KN spacetime [26, 35, 68],

	$\displaystyle\psi_{-1}$	$\displaystyle=\frac{3(\rho^{*})^{2}(M\rho-Q^{2})}{\sqrt{2}\,Q\rho}\delta\phi_{2}$		(5)
		$\displaystyle=\frac{3(\rho^{*})^{2}(M\rho-Q^{2})}{\sqrt{2}\,Q\rho}\delta F_{\mu\nu}m^{*}{}^{\mu}n^{\nu}$		(6)
	$\displaystyle\psi_{-2}$	$\displaystyle=\rho^{*}{}^{4}\Psi_{4}=\rho^{*}{}^{4}\delta C_{\mu\nu\rho\sigma}n^{\mu}m^{*}{}^{\nu}n^{\rho}m^{*}{}^{\sigma},$		(7)

presented here in the gauge with $\Psi_{3}=0$. Here $\delta F_{\mu\nu}$ and $\delta C_{\mu\nu\rho\sigma}$ are the perturbations to the field strength and Weyl tensor, respectively.

For calculations we use the Kinnersley tetrad in Boyer-Lindquist (BL) coordinates ($t,r,\theta,\phi)$,

	$\displaystyle\ell$	$\displaystyle=\frac{r^{2}+a^{2}}{\Delta}\partial_{t}+\partial_{r}+\frac{a}{\Delta}\partial_{\phi},$		(8)
	$\displaystyle n$	$\displaystyle=\frac{r^{2}+a^{2}}{2\rho\rho^{*}}\partial_{t}-\frac{\Delta}{2\rho\rho^{*}}\partial_{r}+\frac{a}{2\rho\rho^{*}}\partial_{\phi},$		(9)
	$\displaystyle m$	$\displaystyle=\frac{1}{\sqrt{2}\rho}\left(ia\sin\theta\partial_{t}+\partial_{\theta}+\frac{i}{\sin\theta}\partial_{\phi}\right).$		(10)

where a star denotes complex conjugation.

The Kinnersley tetrad and BL coordinates are not regular on the event horizon. For near-horizon analysis we will instead use ingoing coordinates $(v,r,\theta,\varphi)$ and the Hartle-Hawking tetrad $\{\hat{\ell},\hat{n},\hat{m},\hat{m}^{*}\}$. The ingoing coordinates are defined by

$\displaystyle v=t+r_{*},\qquad\varphi=\phi+r_{\sharp},$

(11)

with $r_{*}$ and $r_{\sharp}$ satisfying

$\displaystyle\frac{dr_{*}}{dr}=\frac{r^{2}+a^{2}}{\Delta},\qquad\frac{dr_{\sharp}}{dr}=\frac{a}{\Delta}.$

(12)

The Hartle-Hawking tetrad is defined by

$\displaystyle\hat{\ell}=B\ell,\ \ \ \hat{n}=B^{-1}n,\ \ \ \hat{m}=m$

(13)

with boost

$$B=\frac{\Delta}{2(r^{2}+a^{2})}.$$

(14)

We use a hat to denote a quantity constructed in the regular tetrad and coordinates. The hatted Weyl scalars satisfy

$\displaystyle\hat{\psi}_{s}=B^{s}\psi_{s}.$

(15)

When considering Fourier mode solutions of the form

$\displaystyle\psi_{s}=f(r,\theta)e^{im\phi}e^{-i\omega t},$

(16)

we will express the hatted version as

$\displaystyle\hat{\psi}_{s}=B^{s}\psi_{s}=\hat{f}(r,\theta)e^{im\varphi}e^{-i\omega v},$

(17)

so that

$\displaystyle\hat{f}=B^{s}e^{-imr_{\sharp}}e^{i\omega r_{*}}f.$

(18)

We will commonly work with co-rotating perturbations with $\omega=m\Omega_{H}$. In this case the relationship is

$\displaystyle\hat{f}=B^{s}e^{-im(r_{\sharp}-\Omega_{H}r_{*})}f.$

(19)

This special combination is logarithmically divergent near the horizon,

$\displaystyle r_{\sharp}-\Omega_{H}r_{*}\sim-2M\Omega_{H}\log x,\qquad x\to 0,$

(20)

where $\sim$ denotes asymptotic equality and

$\displaystyle x=\frac{r-M}{M}.$

(21)

Thus we have

$\displaystyle\hat{f}\sim x^{2s+2iM\Omega_{H}m}f,$

(22)

which for the Kerr spacetime $(a=M)$ becomes

$\displaystyle\hat{f}\sim x^{2s+im}f.$

(23)

We will start with the special case of a Kerr black hole ($Q=0$), where the mathematics simplifies significantly. In this limit the gravitational and electromagnetic perturbations decouple,²² 2 In our convention for $\psi_{-1}$, a factor of $1/Q$ must be stripped off from (6) in order to obtain a finite Kerr limit. controlled respectively by $\psi_{\pm 2}$ and $\psi_{\pm 1}$, which separately satisfy the spin-$s$ Teukolsky equation [83].

A general perturbation may be Fourier transformed and expanded with respect to the spin-weighted spheroidal harmonics ${}_{s}S_{\ell m\omega}$. Each individual “term” then has the form

$\displaystyle{}_{s}\psi_{\ell m\omega}={}_{s}R_{\ell m\omega}(r){}_{s}S_{\ell m\omega}(\theta)e^{im\phi}e^{-i\omega t},$

(24)

leading to an ODE for ${}_{s}R_{\ell m\omega}$,

$\displaystyle{}_{s}\mathcal{E}_{\ell m\omega}[{}_{s}R_{\ell m\omega}(r)]=0,$

(25)

with the operator ${}_{s}\mathcal{E}_{\ell m\omega}$ given in Eq. (28) of Ref. [49]. A co-rotating perturbation depends on $\phi$ and $t$ through the combination $\phi-\Omega_{H}t$ and hence has frequency $\omega=m\Omega_{H}$. Such modes have been called “synchronous” in Ref. [78].

It is easy to check that the Teukolsky equation at $\omega=m\Omega_{H}$ predicts near-horizon power-law behavior,

$\displaystyle{}_{s}R_{\ell m(m\Omega_{H})}\sim ax^{h-1-s}+bx^{-h-s},\quad x\to 0,$

(26)

where $a$ and $b$ are constant coefficients, $x=(r-M)/M$ was defined in (21), and the “conformal weight” $h$ is

$\displaystyle h=\frac{1}{2}+\sqrt{\frac{1}{4}+{}_{s}K_{lm}-2m^{2}},$

(27)

where ${}_{s}K_{lm}>0$ is the spheroidal harmonic eigenvalue evaluated at co-rotation $\omega=m\Omega_{H}$. (See e.g. Eq. (24) of Ref. [49].) This eigenvalue is defined with the convention that $K=\ell(\ell+1)$ and hence $h=\ell+1$ in the axisymmetric case ($m=0$), which is also stationary on account of the co-rotating assumption. The numerically-observed properties of ${}_{s}K_{lm}$ indicate that $h$ subdivides into three cases (see e.g. (67) of Ref. [49]),

$\displaystyle h\textrm{ is}\begin{cases}1/2+i\delta,&|m|\gtrsim 0.74l\quad\quad\ \!\text{(principal)},\\ >1,&0<|m|\lesssim 0.74l\ \text{(supplementary)},\\ l+1,&m=0\quad\quad\quad\quad\ \text{(axisymmetric)},\end{cases}$

(28)

where $\delta>0$. The transition from supplementary to principal at $m=0.74\ell$ is exact in the large-$\ell$ limit [91].

In the horizon-regular Hartle-Hawking tetrad, the two power laws are instead

$\displaystyle{}_{s}\hat{R}_{\ell m(m\Omega_{H})}\sim ax^{h-1+s+im}+bx^{-h+s+im}$

(29)

according to (23). In the axisymmetric case $m=0$, the conformal weight $h=\ell+1$ is always a positive integer larger than $|s|$, meaning that the $a$ term is always regular, while the $b$ term is always singular. When considering stationary, axisymmetric sources, one therefore imposes $b=0$ as a boundary condition.

The same approach does not work for a truly co-rotating perturbation $m\neq 0$. For the supplementary modes $h>1$, the $a$ term is still more regular than the $b$ term, but both terms fail to be smooth because of the non-integer power law. The situation is even worse for the principal modes $h=1/2+i\delta$, where both terms behave as $x^{-1/2+s}$ up to a phase. The terms are indistinguishable based on regularity, and both diverge for $s\leq 0$.

Horizon regularity is therefore not a reasonable boundary condition for the co-rotating modes. To understand the situation we will need to make two changes. First, we will consider modes that are nearly co-rotating,

$\displaystyle k\equiv 2\frac{\omega-m\Omega_{H}}{\Omega_{H}}\ll 1,$

(30)

defined with a conventional factor of 2 [49]. Second, we will explicitly include a source,

$\displaystyle{}_{s}\mathcal{E}_{\ell m\omega}[{}_{s}g_{\ell m\omega}(r,r^{\prime})]=\Delta^{-s}\delta(r-r^{\prime}),$

(31)

again with a conventional prefactor $\Delta^{-s}$ as in Eq. (32) of Ref. [49]. With these two changes, we may impose the standard boundary conditions of ingoing-wave behavior at the horizon, and outgoing-wave behavior at infinity. In this context, ${}_{s}g_{\ell m\omega}(r)$ are interpreted as the modes of the retarded Green function for the Teukolsky equation.

The solution for ${}_{s}g_{\ell m\omega}$ may be constructed explicitly using the method of matched asymptotic expansions (e.g., [49]), considering separately the “near” approximation $x\ll 1$ and “far” approximation $x\gg k$ before matching in the region of overlap $k\ll x\ll 1$. The near approximation is used to impose ingoing-wave behavior at the event horizon, and the far approximation is used to impose outgoing-wave behavior at infinity. The power laws (26) appear in the overlap region, where near and far are matched together.

In order to understand the excitation of the overlap-region power laws by external sources, we choose the source point to be in the far region ($x^{\prime}\gg k$) and the field point to be in the overlap region ($k\ll x\ll 1$). The non-axisymmetric Green function modes are then given by expanding Eq. (54) of Ref. [49] in the overlap region,

s

$\displaystyle g_{lm\omega}(x,x^{\prime})\approx\frac{R(x^{\prime})}{1-2h}\frac{x^{h-1-s}+\eta(-ik)^{2h-1}x^{-h-s}}{1-\mathcal{R}\eta(-ik)^{2h-1}},$

(32)

where $R(x^{\prime})$ is a far-zone solution (called $R^{\rm far,up}$ in Ref. [49]), and

	$\displaystyle\eta$	$\displaystyle=-\,\frac{\Gamma(2-2h)\,\Gamma(h-im-s)}{\Gamma(2h)\,\Gamma(1-h-im-s)},$		(33)
	$\displaystyle\mathcal{R}$	$\displaystyle=-\,\frac{\Gamma(2-2h)\,\Gamma(h-im+s)}{\Gamma(1-h-im+s)\,\Gamma(2h)}(-im)^{2h-1}.$		(34)

In the limit $k\to 0$, we may compare with the precisely co-rotating behavior (26). For a supplementary (or axisymmetric) mode, we have $2h-1>0$ so that $k^{2h-1}\to 0$ as $k\to 0$, and we recover just the $a$ term of (26), showing that $b=0$ is indeed the correct boundary condition for these modes. However, for a principal mode, $2h-1=2i\delta$ is purely imaginary, and $k^{2h-1}$ is an infinitely oscillating phase as $k\to 0$.

For these modes we cannot set $k=0$ precisely and instead must regard the factors of $k^{2h-1}$ as a “random phase” for small but non-zero values of $k$. In other words, both terms of (26) contribute as $k\to 0$ for nearly co-rotating sources, with coefficients that oscillate with $k$ (but remain bounded) according to (32). Therefore, the perturbation is of order $x^{-1/2-s}$ when $k\ll x\ll 1$.

Using Eq. (23), we see that the regular-tetrad Green function modes $\hat{g}$ are of order $x^{-1/2+s}$ in the overlap region $k\ll x\ll 1$. This indicates arbitrarily large $s\leq 0$ perturbations as $k\to 0$. Evaluating at $x\sim k^{p}$ for $0<p<1$, the perturbation has size

$\displaystyle\hat{g}(x\sim k^{p})\sim k^{p(-1/2+s)},\qquad k\to 0.$

(35)

To understand the physical effects, we use the estimate $x^{-1/2+s}$ for the magnitude of the perturbation with a cutoff at $x\sim k$ where the overlap region ends. For $s=-1$ we have a component of electric field of order $x^{-3/2}$ experienced by an infalling observer. Since $m^{\mu}$ represents angular directions, from (6) we see that the leading singular electric (and magnetic) field is transverse to the horizon. The energy $\Delta E$ imparted to the observer scales as one integral of the behavior, i.e. $x^{-1/2}\sim k^{-1/2}$, and hence is large:

$\displaystyle\Delta E$

$\displaystyle\sim k^{-1/2}\to\infty\qquad(\textrm{electromagnetic}).$

(36)

However, the contribution to the integrated displacement goes as $x^{1/2}\sim k^{1/2}$ and is small. In other words, the body experiences a large kick transverse to the horizon, while the angular displacement remains finite.

For $s=-2$ the perturbation represents a component of perturbed spacetime curvature of order $x^{-5/2}$ experienced by an infalling observer. Again from (7) we see that the leading tidal field is transverse. The tidal deformation $\Delta\xi$ scales as two integrals of this behavior, giving $x^{-1/2}\sim k^{-1/2}$,

$\displaystyle\Delta\xi$

$\displaystyle\sim k^{-1/2}\to\infty\qquad(\textrm{gravitational}).$

(37)

In other words, the infalling body experiences large tidal deformation transverse to the horizon. It also experiences a large rotation transverse to the horizon (though only logarithmically so), stemming from the imaginary part of the principal exponent $x^{-1/2+i\alpha}$.

We emphasize that claims of arbitrarily large deformation apply only within linearized theory. In fact, it seems clear that the metric perturbation itself should scale in the same way as the tidal deformation, such that arbitrarily large deformation just means that linearized theory is breaking down. Within linearized theory, the proper interpretation of “large” tidal deformation is larger than expected. As an observer falls toward a perturbed Kerr black hole under normal circumstances, he experiences a gradual increase in tidal force set by the typical scales of the background spacetime, with no special sign of horizon-crossing. However, if the perturbation is nearly co-rotating, the observer instead feels a sudden increase in tidal force right near the horizon. More precisely, if $\epsilon$ represents the typical strength of the perturbation far from the black hole, the observer experiences deformation of order $\epsilon k^{-1/2}$ over the last moments of his journey passing through the region $k\ll x\ll 1$.

We now discuss how co-rotating modes can be excited by a persistent source. (Although we will say “co-rotating” for short, we always have in mind a limiting process involving a nearly co-rotating mode.) The retarded solution for $\psi_{s}$ resulting from a physical source (conserved stress-energy or charge-current) is constructed by applying certain differential operators [83] to the stress-energy or charge-current and then integrating against the Teukolsky equation Green function that we study. These operators do not mix Fourier modes, so we may focus on the co-rotating modes of the source. We restrict consideration to sources having a discrete set of non-zero Fourier modes that include co-rotating frequencies. The simplest example is a co-rotating particle, i.e. a particle following an orbit of $\partial_{t}+\Omega_{H}\partial_{\phi}$ in a region where that vector is timelike. For electromagnetic perturbations, the particle need not follow a geodesic, and we may simply place it sufficiently close to the symmetry axis at any given “height”. For gravitational perturbations, however, the source must follow a geodesic, and the extremal Kerr spacetime has no co-rotating timelike geodesics. In particular, $\partial_{t}+\Omega_{H}\partial_{\phi}$ is everywhere spacelike on the equatorial plane outside the horizon.

However, we may instead consider eccentric and/or inclined orbits that are tuned to have co-rotating harmonics. The key point is that bound orbits are characterized by three frequencies $\Omega_{r},\Omega_{\theta},\Omega_{\phi}$ associated respectively with radial, polar, and azimuthal oscillations [23, 79]. As a consequence, any scalar covariantly constructed from the orbit along with other stationary, axisymmetric tensors will have time dependence given by a sum over frequencies

$\displaystyle\omega_{mnq}=m\Omega_{\phi}+n\Omega_{r}+q\Omega_{\theta},$

(38)

where $m,n,q$ are integers. A harmonic is co-rotating when $\omega=m\Omega_{H}$, or

$\displaystyle n\Omega_{r}+q\Omega_{\theta}+m(\Omega_{\phi}-\Omega_{H})=0.$

(39)

These resonances can be obtained by tuning a single orbital parameter. For example, we parameterize an equatorial orbit by a dimensionless radial parameter $p_{\mathrm{orb}}$ and eccentricity $e$, defined by the radial turning points $r_{\min}/M=p_{\mathrm{orb}}/(1+e)$ and $r_{\max}/M=p_{\mathrm{orb}}/(1-e)$. Setting $e=0.1$, we find that the resonance condition

$$\Omega_{\phi}+3\Omega_{r}=\Omega_{H}$$

(40)

is satisfied at $p_{\mathrm{orb}}=1.68854$, where $M\Omega_{r}=0.0619771$ and $M\Omega_{\phi}=0.314069$. This orbit has specific energy $\mathcal{E}=E/\mu=0.741171$ and specific angular momentum $\mathcal{L}=L_{z}/\mu=1.55742M$, and oscillates between $r_{\min}=1.53504M$ and $r_{\max}=1.87616M$.

The $q=0,n=3m$ harmonics of this orbit are co-rotating (39). In particular, the $m=2$ harmonic

$$\omega_{260}=2\Omega_{\phi}+6\Omega_{r}=2\Omega_{H}$$

(41)

contains the $\ell=m=2$ dominant principal mode and associated singular horizon behavior. In Appendix A we show how one can manipulate orbital phases of a large number of particles in order to cancel off the non co-rotating harmonics and create a precisely co-rotating source.

We now consider the general case of non-zero black hole charge, $Q\neq 0$. The coupled electromagnetic and gravitational perturbations are encoded in $\psi_{-1}$ and $\psi_{-2}$ [26, 35, 68]. Each mode $\psi_{i}(r,\theta)e^{im\phi}e^{-i\omega t}$ satisfies the coupled equations

	$\displaystyle\left(F_{-2}+Q^{2}G_{-2}\right)\psi_{-2}+Q^{2}H_{-2}\psi_{-1}$	$\displaystyle=0,$		(42)
	$\displaystyle\left(F_{-1}+Q^{2}G_{-1}\right)\psi_{-1}+Q^{2}H_{-1}\psi_{-2}$	$\displaystyle=0,$		(43)

where $F_{i},G_{i},H_{i}$ are second-order differential operators in $r$ and $\theta$. (We use the notation of Ref. [35], where explicit expressions may be found.) In this section, $\psi_{s}(r,\theta)$ refers to the radial and polar dependence of a mode; we suppress dependence on $m$ and $\omega$.

Although some aspects of these equations are amenable to analytic analysis [33], in general they must be solved numerically. We will proceed in two steps, mirroring our analytical analysis of the Kerr case. First we will assume precise co-rotation ($k=0$) and seek special solutions with power-law dependence in $x$ as $x\to 0$. At each value of black hole charge $Q$ and azimuthal number $m$, we find a discrete, apparently infinite, set of solutions that generalize the mode-by-mode power laws of Kerr (26), to which they reduce in the zero-charge limit. We then confirm these exponents arise in physical solutions by numerically constructing the retarded Green function at small-but-finite values of $k$.

In the previous analysis we found the emergence of power-law behavior near the horizon for specific NP quantities. This behavior can be understood more geometrically through the notion of self-similarity in the near-horizon limit [44, 49]. We can assign an invariant weight to the self-similar tensor and use this notion to connect with other near-horizon phenomena, including the Meissner effect and the Aretakis instability.

The near-horizon limit may be constructed by a coordinate transformation from BL coordinates $(t,r,\theta,\phi)$ to “barred coordinates” $x^{\bar{\mu}}=(\bar{t},\bar{r},\theta,\bar{\phi})$ defined by [92, 65, 2]

$\displaystyle\bar{t}=\lambda\Omega_{H}t,\qquad\bar{r}=\frac{r-M}{\lambda M},\qquad\bar{\phi}=\phi-\Omega_{H}t.$

(57)

(In terms of the near-horizon deviation $x$, we have $\bar{r}=x/\lambda$.)

We denote barred coordinate components by placing a bar on the index, and place a bar on any tensor that is defined by a limiting process $\lambda\to 0$ in barred coordinates. The KN near horizon geometry and Maxwell field are defined as

$\displaystyle\bar{g}_{\bar{\mu}\bar{\nu}}=\lim_{\lambda\to 0}g_{\bar{\mu}\bar{\nu}},\quad\bar{F}_{\bar{\mu}\bar{\nu}}=\lim_{\lambda\to 0}F_{\bar{\mu}\bar{\nu}}.$

(58)

The details of the limiting values are not important; the crucial property is that the limiting metric remains non-degenerate (Lorentzian). Since further rescalings do not change the limit, the resulting configuration is guaranteed to have an enhanced scaling symmetry $\bar{t}\to\bar{t}/C$ and $\bar{r}\to C\bar{r}$.⁴⁴ 4 The near-horizon geometry also picks up an additional enhanced symmetry; this symmetry is “accidental” from the perspective of the scaling limit and not important for our considerations. This demonstrates the emergence of scaling symmetry near the KN horizon.

Perturbing fields will not, in general, possess the full scaling symmetry, but in many cases they have an emergent scaling self-similarity, meaning that their barred coordinate components behave as a power law in $\lambda$ as $\lambda\to 0$. We say that a tensor field $T_{\mu_{1}\dots\mu_{n}}$ has weight $w$ if⁵⁵ 5 This is the opposite convention from our previous work [44, 49], in which such a field would have weight $-w$.

$\displaystyle T_{\bar{\mu}_{1}\dots\bar{\mu}_{n}}\sim\lambda^{w}\bar{T}_{\bar{\mu}_{1}\dots\bar{\mu}_{n}},\qquad\lambda\to 0,$

(59)

where $\sim$ indicates asymptotic equality. It follows from (59) that the ‘‘coefficient’’⁶⁶ 6 In general, when we speak of the weight of a tensor, we allow the possibility of a “zero coefficient” $\bar{T}_{\bar{\mu}_{1}\dots\bar{\mu}_{n}}=0$. For example, if we say that $T$ has weight one, its barred coordinates could in fact scale to zero faster than $\lambda$. If the exact weight is known we will emphasize that fact. $\bar{T}_{\bar{\mu}_{1}\dots\bar{\mu}_{n}}$ is self-similar under further rescalings,

$\displaystyle\bar{T}\to C^{w}\bar{T}\qquad\textrm{under}\quad\bar{t}\to\bar{t}/C,\quad\bar{r}\to C\bar{r}.$

(60)

The weight of a tensor is additive under tensor product. In particular, since the KN metric has weight 0 according to (58), the weight of a tensor is unaffected by raising/lowering indices. In general, covariant operations involving the metric do not affect the weight of a tensor. For example, the covariant derivative of a tensor has the same weight as the original tensor.

In our analysis of co-rotating perturbations, we found that Kinnersley-tetrad $\psi_{s}$ of a co-rotating mode behaves as $x^{w-s}$ (for some $w$) near the horizon,

$\displaystyle\psi_{s}\sim x^{w-s}S_{s}(\theta)e^{im(\phi-\Omega_{H}t)},\qquad x\to 0.$

(61)

Changing to barred coordinates shows that

$\displaystyle\psi_{s}\sim\lambda^{w-s}\bar{\psi}_{s},\quad\lambda\to 0,$

(62)

with

$\displaystyle\bar{\psi}_{s}=\bar{r}^{w-s}S_{s}(\theta)e^{im\bar{\phi}}.$

(63)

Thus $\psi_{s}$ has weight $w-s$. The co-rotating angular dependence is crucial for the existence of a well-defined weight; otherwise a $\lambda$ would appear in the exponential.

It is easy to check that the Kinnersley tetrad vectors have weights

$\displaystyle\ell\sim\lambda^{-1}\bar{\ell},\quad n\sim\lambda\bar{n},\quad m\sim\lambda^{0}\bar{m}.$

(64)

Going back to the definition (6)–(7) and counting the appearances of $\ell$ and $n$ in each formula, one begins to suspect that the underlying tensors $\delta F_{\mu\nu}$ and $\delta C_{\mu\nu\rho\sigma}$ must have weight $w$ (at least ignoring trivial changes in mass, spin, and charge), such that the weights of the tetrad vectors account for the shifted weight $w-s$ of $\psi_{s}$.

We can show this to be the case in the Kerr spacetime ($Q=0$) using the metric/field reconstruction procedure [29, 28, 86, 87]. Schematically, the reconstructed metric may be written $h_{\mu\nu}=\mathcal{O}_{\mu\nu}[\psi_{-2}]$, where $\mathcal{O}_{\mu\nu}$ is a set of operations involving four angular integrals and then two derivatives. (For details, we recommend the excellent presentation of Ref. [16].) Using the Geroch-Held-Penrose [42, 59] version of the latter step (e.g., Eq. (19) of Ref. [49]), it is straightforward to establish that the reconstructed metric is invariant under tetrad boosts $\ell\to B\ell,n\to B^{-1}n$. By using the boost $B=x$, we make the tetrad have weight $0$. In this tetrad, $\tilde{\psi}_{-s}$ has weight $w$ and the metric reconstruction procedure manifestly preserves the weight (since it is built from the background metric and the weight-0 tetrad vectors), so that the resulting perturbation $h_{\mu\nu}$ also has weight $w$. Wald’s theorem [86] guarantees that the reconstruction is unique up to gauge and changes in mass and spin.

Similar arguments establish that the Maxwell field $F_{\mu\nu}$ reconstructed from $\psi_{-1}$ in the Kerr spacetime has weight $w$. We fully expect the same to be true of reconstructed metric and field perturbations in the KN spacetime, but we are unable to make an analogous argument at this time.

It is helpful to also introduce a version of this story better adapted to the regular coordinates and tetrad. We instead consider “tilde coordinates” $x^{\tilde{\mu}}=(\tilde{v},\tilde{r},\theta,\tilde{\varphi})$

$\displaystyle\tilde{v}=\lambda\Omega_{H}v,\qquad\tilde{r}=\frac{r-M}{\lambda M},\qquad\tilde{\varphi}=\varphi-\Omega_{H}v.$

(65)

Near-horizon weights are defined in the same way,

$\displaystyle T_{\tilde{\mu}_{1}\dots\tilde{\mu}_{n}}\sim\lambda^{\tilde{w}}\tilde{T}_{\tilde{\mu}_{1}\dots\tilde{\mu}_{n}},\qquad\lambda\to 0,$

(66)

from which it follows that

$\displaystyle\tilde{T}\to C^{\tilde{w}}\tilde{T}\qquad\textrm{under}\quad\tilde{v}\to\tilde{v}/C,\quad\tilde{r}\to C\tilde{r}.$

(67)

Recalling (22), a single co-rotating power-law may be expressed in either coordinate system as

$\displaystyle x^{w}e^{im(\phi-\Omega_{H}t)}=x^{w+2iM\Omega_{H}m}e^{im(\varphi-\Omega_{H}v)}.$

(68)

The weights are thus related by

$\displaystyle\tilde{w}=w+2imM\Omega_{H}.$

(69)

The weights are the same in the axisymmetric case $m=0$, and are otherwise related by a purely imaginary shift. The weights $p$ defined in Ref. [49] are related to these weights by $p=-\tilde{w}$.

For future reference we note that the Hartle-Hawking tetrad vectors have weights

$\displaystyle\hat{\ell}\sim\lambda\hat{\bar{\ell}},\quad\hat{n}\sim\lambda^{-1}\hat{\bar{n}},\quad\hat{m}\sim\lambda^{0}\hat{\bar{m}}.$

(70)

Since the tetrad is axisymmetric, there is no distinction between $w$ and $\tilde{w}$.

In the axisymmetric case $m=0$, our co-rotating perturbations are stationary. That is, $m=0$ is the simpler case of stationary, axisymmetric perturbations. For EM perturbations of the Kerr spacetime, the EM scaling weights are integers and give rise to the Meissner effect [44]. We now review this result and generalize it to GEM perturbations of the extremal KN spacetime.

For an extremal Kerr black hole with stationary, axisymmetric perturbations, the general solution regular at the poles is

$\displaystyle\psi_{s}=\sum_{\ell=|s|}^{\infty}\left(a_{\ell}x^{\ell-s}+b_{\ell}x^{-(\ell+1)-s}\right)P^{(s)}_{\ell}(\cos\theta),$

(75)

where $P^{(s)}_{\ell}$ are associated Legendre Polynomials. The $b_{\ell}$ family is singular on the horizon and discarded. From the $a_{\ell}$ family we immediately conclude that for regular perturbations, each $\ell$-mode of $\psi_{s}$ has weight $\ell-s$. Below Eq. (64) we established that the reconstructed perturbation (maxwell field or metric perturbation) has weight $\ell$, up to gauge and trivial changes in mass, charge and spin.

Restricting to stationary, axisymmetric electromagnetic perturbations of extremal Kerr that do not change the charge of the spacetime, the general solution is thus

$\displaystyle F_{\mu\nu}=\sum_{\ell=1}^{\infty}F^{(\ell)}_{\mu\nu}(r,\theta),\quad F^{(\ell)}_{\tilde{\mu}\tilde{\nu}}\sim\lambda^{\ell}.$

(76)

where $F^{(\ell)}_{\mu\nu}$ has weight $\ell$ as indicated by $F^{(\ell)}_{\tilde{\mu}\tilde{\nu}}\sim\lambda^{\ell}$. Here we use the ingoing version of the near-horizon limit since we will be evaluating tensors on the future horizon. The weight $\tilde{w}=w=\ell$ is the same as the BL version for stationary, axisymmetric perturbations.

To connect to the traditional Meissner screening, consider expressing this field in co-rotating regular coordinates $(v,r,\theta,\tilde{\varphi})$. Changing to the tilde coordinates (65) entails $v\to\tilde{v}/(\Omega_{H}\lambda)$ and $r\to M(1+\lambda\tilde{r})$, so that expanding a $v$-independent (stationary) tensor component in $\lambda$ is just a Taylor expansion in radius about the horizon, together with an overall weighting depending on how many times $v$ and $r$ appear in the component. For example,

	$\displaystyle F_{\tilde{v}\tilde{\theta}}=\frac{1}{\Omega_{H}\lambda}\!\Big($	$\displaystyle F_{v\theta}+\lambda M\partial_{r}F_{v\theta}\tilde{r}$
		$\displaystyle+\frac{1}{2}M^{2}\lambda^{2}\partial_{r}\partial_{r}F_{v\theta}\tilde{r}^{2}+...\Big)_{\!\!r=M}\!.$		(77)

The general form (76) requires that this perturbation vanish until order $\lambda^{1}$ (the lowest weight-term being $\ell=1$), which means that $F_{v\theta}$ and its first derivative must vanish on the horizon, i.e. $F_{v\theta}=O((r-M)^{2})$. Making a similar argument for the other components, the complete collection behaves as

	$\displaystyle F_{v\theta}\sim F_{v\tilde{\varphi}}\sim(r-M)^{2}$			(78)
	$\displaystyle F_{vr}\sim F_{\theta\tilde{\varphi}}\sim(r-M)^{1}$			(79)
	$\displaystyle F_{r\theta}\sim F_{r\tilde{\varphi}}\sim(r-M)^{0}$	$\displaystyle.\$		(80)

In particular, the horizon pullback of $F$ and its dual are vanishing, which is the invariant statement of the electromagnetic black hole Meissner effect.