PHYSICAL REVIEW D 90, 124011 (2014)

Quasinormal modes of charged fields around a Reissner-Nordström

black hole
Maurício Richartz* and Davi Giugno†
Centro de Matemática, Computação e Cognição, Universidade Federal do ABC (UFABC),
09210-170 Santo André, São Paulo, Brazil
(Received 25 September 2014; published 3 December 2014)
The quasinormal spectrum of a charged scalar field around a nonextremal Reissner-Nordström black
hole, especially in the limit of large electromagnetic interaction, has been comprehensively studied only
very recently. In this paper, we extend the analysis to Dirac fields using the continued fraction method
and compare the results with the scalar case. In particular, we study the behavior of the fundamental
quasinormal mode as a function of the black hole’s charge and of the electromagnetic interaction
parameter. We derive an analytical formula for the quasinormal frequencies in the limit of large
electromagnetic interaction. As the extremal limit of black hole charge is approached, we show that,
unlike the case of neutral fields, the imaginary part of the quasinormal frequencies approaches zero for
charged fields.

DOI: 10.1103/PhysRevD.90.124011 PACS numbers: [Link], [Link]

I. INTRODUCTION According to the uniqueness theorems, once a black hole

The study of black hole perturbations was pioneered by has reached a stable configuration, it will be characterized
Regge and Wheeler in the 1950s while attempting to only by three parameters: its mass M, its charge Q, and its
establish the stability of Schwarzschild black holes [1]. angular momentum J. The QNM ωn of a black hole, on the
Later, Vishveshwara [2] identified a special type of per- other hand, will in general be described by an interacting
turbation characterized by purely outgoing waves at spatial field which is characterized by a mass parameter μ, a charge
infinity and purely ingoing waves in the vicinity of the parameter q, an orbital number l, an azimuthal number m
event horizon. Such perturbations, dubbed quasinormal and a spin parameter s (depending on the type of pertur-
modes (QNMs) by Press [3], are only permitted for a bation, i.e., scalar, Dirac, electromagnetic or gravitational).
discrete set of complex frequencies ωn , the so-called There exists in the literature an extensive amount of
quasinormal frequencies. Assuming a time dependence research on QNMs, for almost any kind of black hole
of expð−iωtÞ, if their imaginary part is negative and any kind of perturbation—see e.g., [16–19] for recent
[Imðωn Þ < 0], the perturbation is damped; similarly, if reviews on the subject. There is one situation, however,
Imðωn Þ > 0, the mode is dynamically unstable and its which only recently has drawn the attention of the physics
amplitude grows in time until the linear approximation community. The QNMs of charged fields in a charged black
ceases to be valid. hole background have only been studied in detail in the
QNMs have been studied in many different contexts, regime μM ≪ 1 and qQ ≪ 1 [20–25], the exception being
including (but not limited to) the scattering of particles by Refs. [26–28], where the scalar case was analyzed. In view
black holes [4–6], the formation of black holes by the of that, our main objective in this paper is to use the
collapse of a star [7,8], the stability of black holes [9,10], continued fraction method to present a detailed study of the
the quantization of black holes [11–13], the AdS/CFT quasinormal modes of a massless charged Dirac field for
correspondence [14], and the perturbation of analogue arbitrary values of qQ, comparing the obtained results with
black holes [15]. Perhaps the most famous manifestation the scalar case. We also discuss in detail the behavior of
of QNMs occurs in astrophysics, when gravitational waves the QNM frequencies in the limit Q → M of extremal
are emitted by a perturbed black hole. Such a process can black holes.
be divided into three stages. First, there is an outburst of
radiation which is highly dependent on the initial pertur- II. FIELD DYNAMICS
bation; second, there is a long period of quasinormal
oscillation, and finally, at later times, the QNMs damp Let us start by describing the dynamics of massless
out and become dominated by a power law tail. charged fields around a Reissner-Nordström (RN) black
hole of mass M and charge Q whose metric is given by

* Δ 2 r2 2
[Link]@[Link] ds2 ¼ − dt þ dr þ r2 ðdθ2 þ sin2 θdφ2 Þ; ð2:1Þ
†
dgiugno@[Link] r2 Δ

1550-7998=2014=90(12)=124011(9) 124011-1 © 2014 American Physical Society

MAURÍCIO RICHARTZ AND DAVI GIUGNO PHYSICAL REVIEW D 90, 124011 (2014)
2
where Δ ¼ r2 − 2Mr þ Q2 . The locations of the event where K ¼ ωr − qQr and λs ¼ ðj − sÞðj þ s þ 1Þ is a
horizon and of the ffiCauchy horizon p are, respectively, separation constant. The equation above is analogous to the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
rþ ¼ M þ M − Q and r− ¼ M − M − Q2 . The
2 2 2 radial Teukolsky equation for the Kerr metric [32].
RN metric can be obtained from the Einstein-Maxwell If we now make the transformation f s ¼ Δs=2 rRs and
equations by assuming stationarity and spherical symmetry. adopt the tortoise coordinate r (defined by dr =dr ¼
The electromagnetic field is obtained from the electromag- r2 =Δ), the master equation (2.9) becomes
netic potential Aμ , whose only nonvanishing component
d2 f s
is A0 ¼ −Q=r. þ W s ðω; r Þf s ¼ 0; ð2:10Þ
A massless charged scalar field ψ propagating on a RN dr2
background obeys the Klein-Gordon equation,
where the complex function W s is given by
ð∇μ − ieAμ Þð∇μ − ieAμ Þψ ¼ 0: ð2:2Þ 
Δ ðK − isðr − MÞÞ2
W s ðω; r Þ ¼ 4 þ 4isωr
r Δ
A massless charged Dirac field, on the other hand, can be 
described by a pair of spinors PA and Q̄A which satisfy the M Q2 2
−2isqQ − jðj þ 1Þ þ s − 2 þ 2 2 :
Dirac equations, r r
ð2:11Þ
σ μ AB0 ð∇μ − iqAμ ÞPA ¼ 0; ð2:3Þ
The equation above can be solved analytically in the
σ μ AB0 ð∇μ þ iqAμ ÞQA ¼ 0; ð2:4Þ asymptotic limits of the tortoise coordinate space (note that
r → −∞ and r → ∞ correspond, respectively, to the
where σ μ AB0 are generalizations of the Pauli matrices [29]. event horizon r → rþ and the spatial infinity r → ∞). By
Note that, by using the differential operator ∇μ − iqAμ, we further imposing the boundary conditions naturally asso-
have assumed the minimal coupling between the scalar/ ciated with quasinormal oscillation, i.e., purely outgoing
Dirac field and the electromagnetic field of the black hole. waves far away from the black hole and purely ingoing
All the equations above can be separated if one uses waves near the black hole’s event horizon, one is able to
appropriate ansatzes. In fact, let us rewrite the scalar field as obtain the asymptotic form of the quasinormal modes. For a
ψ ¼ R0 ðrÞY 0jm ðθ; ϕÞe−iωt and the fermionic spinors as nonextremal black hole (0 ≤ Q < M), the result is

−1
( −s−iqQ þiωr
R−12 ðrÞY jm2 ðθÞ Zout
s r e ; r → ∞
0
P ¼ e−iωt imφ
e ; ð2:5Þ fs → r −r
−2sð þ 2 − Þr −iðω−qQ
ð2:12Þ
rþ Þr
r Ztrs e r
þ ; r → −∞;
þ1 where Zout tr
P1 ¼ Rþ12 ðrÞY jm2 ðθÞe−iωt eimφ ; ð2:6Þ s and Z s are constants.
Equation (2.10), together with the boundary conditions
above, becomes an eigenvalue problem for ω so that only a
0 −1
Q̄1 ¼ Rþ12 ðrÞY jm2 ðθÞe−iωt eimφ ; ð2:7Þ discrete set of frequencies (the quasinormal frequencies ωn )
is allowed. In the next section we describe the continued
þ1 fraction method, a commonly used method to determine the
R−12 ðrÞY jm2 ðθÞ quasinormal frequencies, and discuss its implementation
00
Q̄ ¼ − e−iωt eimφ ; ð2:8Þ
r for charged fields around a RN black hole.
where ω is the frequency of the field, j ≥ jsj and −j ≤
III. CONTINUED FRACTIONS
m ≤ j are integers (scalar case) or half-integers (fermionic
case) and Y sjm ðθÞ are the corresponding spin-weighted In 1985, inspired by a technique due to Jaffe to calculate
spherical harmonics [30] (note that for scalar fields, l is the energy eigenvalues of the H þ 2 ion, Leaver devised a
usually used instead of j). After some algebra, one can numerical method to compute QNMs of Schwarzschild and
show that the radial components of the fields, Rs ðrÞ, satisfy Kerr black holes using continued fractions [33,34]. His
the following master equation [24,31]: approach consists in expressing the solution to the wave
equation, with the appropriate boundary conditions for
   2
d
−s sþ1 dRs K − 2isðr − MÞK QNMs, as a power series which is everywhere convergent,
Δ Δ þ except possibly at the asymptotic limit r → ∞. By requir-
dr dr Δ
 ing convergence also at infinity, one obtains a continued
þ 4isωr − 2isqQ − λs Rs ¼ 0; ð2:9Þ fraction relation which must be satisfied by the expansion
coefficients of the power series.

124011-2
QUASINORMAL MODES OF CHARGED FIELDS AROUND A … PHYSICAL REVIEW D 90, 124011 (2014)
Leaver’s original method was later improved by Nollert singularity at r ¼ þ∞ and three regular singularities at
[35] and has since been used to calculate QNMs in a variety r ¼ 0, r ¼ r− and r ¼ rþ . The solution which satisfies the
of situations [19]. Recently, it has been used in Ref. [28] to boundary conditions (2.12) can be expanded in a powers
obtain the QNMs of a charged scalar field around a Kerr- series around r ¼ rþ ,
Newman black hole. In this paper, besides the scalar case,
X
∞  
we also apply Leaver’s technique to determine the QNMs r − rþ nþδ
of a massless Dirac field around a RN black hole without RðrÞ ¼ eiωr ðr − r− Þϵ an ; ð3:1Þ
n¼0
r − r−
any restrictions on the parameters involved. To the best of
our knowledge, it is the first time that these QNMs are
where ϵ ¼ −iqQ þ iωðrþ þ r− Þ − 2s − 1 and δ¼
calculated for Dirac fields without the assumption of small
ir2þ ðω−qQ
rþ Þ
qQ. For small qQ, the QNMs of charged Dirac fields −s − rþ −r− .
The coefficients an must satisfy the three-
around a RN black hole have been calculated in Refs. [23] term recurrence relations,
and [25] using, respectively, the WKB and the Pöschl-
Teller methods. We also note that the continued fraction
α0 a1 þ β0 a0 ¼ 0;
method has been applied to the more general context of a
Kerr-Newman–de Sitter background in Ref. [22]. αn anþ1 þ βn an þ γ n an−1 ¼ 0; n ≥ 1; ð3:2Þ
To implement Leaver’s method, we start by noting
that, in the nonextremal case, Eq. (2.9) has an irregular where αn , βn and γ n are given by

αn ¼ −ðn þ 1Þ½r− ðn − s þ 1Þ þ rþ ð−n − 2iqQ þ 2irþ ω þ s − 1Þ; ð3:3Þ

βn ¼ −rþ ½λs þ 2n2 − 4irþ ωð2n þ 3iqQ þ 1Þ þ 6inqQ þ 2n − 4ðqQÞ2 þ 3iqQ − 8r2þ ω2 þ s þ 1
þ r− ½λs þ 2nðn þ iqQ þ 1Þ þ iqQ þ s þ 1 − 2ið2n þ 1Þr− rþ ω; ð3:4Þ

γ n ¼ −fn þ 2i½qQ − ωðrþ þ r− Þg½nðr− − rþ Þ þ irþ ð−2qQ þ 2rþ ω þ isÞ þ r− s: ð3:5Þ

When q ¼ Q ¼ s ¼ 0, these coefficients reduce to the αn−1 γ n αn−2 γ n−1 αγ α γ α γ

βn −    0 1 ¼ n nþ1 nþ1 nþ2    ;
original ones obtained by Leaver for the Schwarzschild βn−1 − βn−2 − β0 βnþ1 − βnþ2 −
case [33]. Additionally, for uncharged fields (qQ ¼ 0), ð3:7Þ
these coefficients are compatible with the ones obtained in
Ref. [36]; for scalar fields (s ¼ 0), they are compatible with whose most stable root is exactly the nth QNM frequency.
the ones obtained in Ref. [12]. As discovered by Nollert [35] for Schwarzschild
Since rþ is a regular singular point, convergence of black holes, the convergence of Leaver’s method can be
the series is automatically guaranteed for rþ ≤ r < ∞. improved if one estimates the “error” associated with the
P also at r ¼ ∞ implies convergence of
Convergence truncation of the continued fraction. To implement the
the sum n an and, therefore, the coefficients an
technique in our problem, we first note that RN ¼
must satisfy the following infinite continued fraction −aNþ1 =aN satisfies the recursive equation
equation [37]: γ Nþ1
RN ¼ ; ð3:8Þ
βNþ1 − αNþ1 RNþ1
α0 γ 1 α1 γ 2
0 ¼ β0 − : ð3:6Þ and, therefore, corresponds exactly to the “rest” of the
β1 − β2 − continued fraction (3.6) truncated at order N. We then
expand RN in a power series of N 1=2 ,
Any QNM can be found by solving the equation above X
∞
if the infinite continued fraction is truncated at some RN ¼ Ck N −k=2 ; ð3:9Þ
sufficiently large index N. However, in practice, the k¼0
equation above is only used to calculate the fundamental
frequency (its most stable root). In order to find the nth and substitute it in Eq. (3.8) to determine the coefficients
quasinormal frequency, it is more convenient to invert Ck . The first coefficients obtained this way are given by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the equation above n times, resulting in the following C0 ¼ −1, C1 ¼ 2iωðr− − rþ Þ and C2 ¼ 34 − s − iqQþ
equation: 2iωrþ .

124011-3
MAURÍCIO RICHARTZ AND DAVI GIUGNO PHYSICAL REVIEW D 90, 124011 (2014)
TABLE I. QNM frequencies of the fundamental mode obtained with the continued fraction method for different values of Q=M and
qQ. For each value of qQ, the first line corresponds to the scalar case s ¼ j ¼ 0 and the second line to the fermionic case s ¼ j ¼ 1=2.

Q=M ¼ 0.01 Q=M ¼ 0.1 Q=M ¼ 0.5 Q=M ¼ 0.99 Q=M ¼ 0.9999
jqQj jReðωÞj ImðωÞ jReðωÞj ImðωÞ jReðωÞj ImðωÞ jReðωÞj ImðωÞ jReðωÞj ImðωÞ
0.110457 −0.104896 0.110649 −0.104938 0.115764 −0.105751 0.133570 −0.095641 0.133459 −0.095844
0
0.182966 −0.096983 0.183295 −0.097033 0.192120 −0.098106 0.236845 −0.088925 0.238169 −0.087697
0.106575 −0.104213 0.106760 −0.104254 0.111673 −0.105068 0.138611 −0.095570 0.010010 −0.014155
0.01
0.179503 −0.096446 0.179824 −0.096496 0.188446 −0.097572 0.231900 −0.088850 0.010008 −0.021229
0.072255 −0.096978 0.072377 −0.097016 0.075611 −0.097778 0.186342 −0.091844 0.100110 −0.014005
0.1
0.149158 −0.091157 0.149416 −0.091206 0.156303 −0.092289 0.189365 −0.087148 0.100081 −0.021159
0.547816 −0.125258 0.549058 −0.125264 0.583393 −0.124817 0.895177 −0.054159 0.987906 −0.006805
1
0.590192 −0.120632 0.591511 −0.120655 0.627858 −0.120744 0.933299 −0.058255 0.996092 −0.008213
5.006341 −0.125074 5.018771 −0.125073 5.364954 −0.124410 8.765415 −0.054161 9.860731 −0.006875
10
5.012548 −0.124996 5.024987 −0.124996 5.371354 −0.124353 8.769147 −0.054172 9.861243 −0.006875
50.00188 −0.125001 50.12625 −0.125000 53.59044 −0.124356 87.63741 −0.054172 98.60556 −0.006875
100
50.00250 −0.125000 50.12688 −0.124999 53.59108 −0.124356 87.63779 −0.054172 98.60561 −0.006875
500.0126 −0.125000 501.2564 −0.124999 535.8984 −0.124356 876.3725 −0.054172 986.0554 −0.006875
1000
500.0126 −0.125000 501.2564 −0.124999 535.8985 −0.124356 876.3725 −0.054172 986.0554 −0.006875

IV. NUMERICAL RESULTS also calculated the fundamental QNMs for the scalar case
s ¼ j ¼ 0. The results for different sets of parameters are
We shall now use a root-finding algorithm to determine
presented in Table I.
the QNMs associated with Eqs. (3.6) and (3.7). It is
In order to get a complete description of the dependence
convenient to rescale the equations so that the parameters
of the quasinormal modes on the electromagnetic coupling
and variables become dimensionless. Indeed, if we rescale
r → r=M and t → Mt, it is straightforward to see that the between the field and the black hole, we plot the real and
nth quasinormal frequency Mωn will depend only on the imaginary parts of the fundamental frequency as a function
following parameters: Q=M, qQ, s and j. We focus our of qQ. The results are presented in Fig. 1 for the case
analysis in the case s ¼ j ¼ 1=2 and investigate the Q=M ¼ 0.5. In fact, in order to highlight the symmetry of
behavior of the fundamental (n ¼ 0) QNMs as Q=M the quasinormal modes with respect to the transformation
and qQ are varied. As explained in Refs. [23,25], the (ω→−ω , qQ → −qQ), we have plotted both the ReðωÞ>0
potentials for s ¼ 1=2 produce the same QNM spectrum and the ReðωÞ < 0 quasinormal branches in Fig. 1. For
and, therefore, there is no loss of generality in considering sufficiently small values of jqQj, when qQ is positive, the
only the s ¼ þ1=2 case. As mentioned before, we pay mode with ReðωÞ < 0 is, in general, more stable than the
special attention to the regime qQ ≫ 1 and also discuss in mode with ReðωÞ > 0 (because of the symmetry, this
detail what happens when the limit Q → M of extremal behavior is reversed for negative qQ). As jqQj increases,
charge is approached. For comparison reasons, we have the real part of the fundamental frequency approaches zero

FIG. 1 (color online). Real (left plot) and imaginary (right plot) parts of the most stable QNMs of the Dirac field s ¼ j ¼ 1=2 as a
function of qQ for Q=M ¼ 0.5. Note the symmetry with respect to the simultaneous transformations ω → −ω and qQ → −qQ. Note
also the critical value of jqQj above which one of the branches disappears from the spectrum.

124011-4
QUASINORMAL MODES OF CHARGED FIELDS AROUND A … PHYSICAL REVIEW D 90, 124011 (2014)

FIG. 2 (color online). Real (left plot) and imaginary (right plot) parts of the fundamental quasinormal mode as a function of qQ for the
fermionic case s ¼ j ¼ 1=2 [only the ReðωÞ > 0 branch is represented here]. From bottom to top: Q=M ¼ 0.01 (blue), Q=M ¼ 0.5
(red), Q=M ¼ 0.9 (green), Q=M ¼ 0.95 (brown), Q=M ¼ 0.97 (orange), Q=M ¼ 0.99 (magenta). Note that the imaginary parts for
Q=M ¼ 0.01 and Q=M ¼ 0.5 are almost indistinguishable in this plot.

until its branch ceases to exist and disappears from the frequency approaches a constant value, while the real part
spectrum at some critical value of the electromagnetic seems to grow linearly. Besides that, the quasinormal
coupling. Such behavior, which is not uncommon, was frequencies for the scalar and fermionic cases approach
observed in Ref. [28] for scalar fields. We have now verified the same value when jqQj → ∞, as one can observe in
that it is also manifest for fermionic fields. As this critical Fig. 3, where the scalar and Dirac spectra for Q=M ¼ 0.5
value is approached, the continued fraction method seems to have been superimposed. These observations are also
converge slower. Nonetheless, we were able to verify that evident in Table I if one looks at it from top to bottom
this critical value is almost unchanged as Q=M varies: for as jqQj is increased. Analytically, they can be confirmed by
the scalar case (s ¼ j ¼ 0) it is jqQj ≈ 0.3 and for the Dirac expanding the continued fraction equation (3.7) in powers
case (s ¼ j ¼ 1=2) it is jqQj ≈ 0.7. of qQ and solving for ω, producing
In Fig. 2, we have similar plots of the QNM frequencies
as a function of qQ for Q=M ¼ 0.01, 0.5, 0.9, 0.95, 0.97,
qQ ð1 þ 2nÞðrþ − r− Þ
0.99. However, due to the symmetry discussed above, we ωn ¼ −i
restrict the analysis to ReðωÞ > 0 without loss of generality. rþ 4r2þ
Note the change in the behavior of ImðωÞ as the extremal ðrþ − r− Þð2rþ ð1 þ 2s þ 2λs Þ þ r− ð4s2 − 1ÞÞ
charge limit is approached, at some point between Q=M ¼ þ
16r3þ qQ
0.9 and Q=M ¼ 0.95 (compare the top curves with the
ð1 þ 2nÞð3r− − rþ Þð4s2 − 1Þðrþ − r− Þ2
bottom ones). In all cases, as the electromagnetic inter- −i ; ð4:1Þ
action increases, the imaginary part of the quasinormal 64r4þ q2 Q2

FIG. 3 (color online). Comparison between scalar (s ¼ j ¼ 0, blue, dashed) and Dirac (s ¼ j ¼ 1=2, red, solid) quasinormal modes.
Real (left plot) and imaginary (right plot) parts are shown as functions of qQ for Q=M ¼ 0.5. Note the critical values of the
electromagnetic interaction around jqQj ≈ 0.3 for s ¼ j ¼ 0 and around jqQj ≈ 0.7 for s ¼ j ¼ 1=2.

124011-5
MAURÍCIO RICHARTZ AND DAVI GIUGNO PHYSICAL REVIEW D 90, 124011 (2014)
−3
plus terms of order ðqQÞ . A similar result [up to order for ω, we write ω ¼ qQ=rþ þ Kðrþ − r− Þ and solve the
ðqQÞ−1 ] was obtained previously in Refs. [27] and [28], but corresponding equation for K. An interesting consequence
only for the scalar case. We would like to note that our is that the coefficients αn , βn and γ n become proportional to
result also holds in the fermionic case, and that deviations ðrþ − r− Þ. Eliminating this common factor, the coefficients
from the scalar case only show up in terms of order ðqQÞ−1 can be redefined as
and higher, as shown explicitly in the equation above.
Another important result which we have observed in our αn ¼ −ð1 þ nÞð1 þ n − 2iKr2þ − sÞ; ð4:2Þ
simulations (see the last column of Table I) is that the
imaginary part of the fundamental frequency becomes βn ¼ 1 − iqQ þ 2iKrþ r− þ 2ðn − 2iKr2þ Þ
smaller as the extremal limit Q ¼ M is approached, while
the real part approaches qQ=rþ ≈ q. However, these modes × ½n þ 1 − iqQ − 2iKrþ ðrþ − r− Þ þ s þ λs ; ð4:3Þ
have a strange behavior since they correspond to solutions of
i
the continued fraction equation which do not exist in the γn ¼ ð2qQr− þ inrþ − 2Kr2− rþ þ 2Kr3þ Þ
neutral limit qQ ¼ 0. More precisely, as pointed out in rþ
Ref. [28] for the scalar case, if we fix Q=M and take the limit × ðn − 2iKr2þ þ sÞ; ð4:4Þ
qQ → 0, there will be some critical value of qQ below which
these modes disappear from the spectrum. Similarly to the and then Eq. (3.6) can be normally solved for K. With the
scalar case, these modes exist for all values of Q=M, but far original coefficients, Eq. (3.6) is always proportional to
from the extremal limit they correspond to higher overtones ðrþ − r− Þ and therefore one needs more precision in the
and therefore cannot be the fundamental mode (see Fig. 4 for root-finding algorithm to correctly determine the QNMs
a plot of these “strange” modes showing the exact value of when the extremal limit is approached.
Q=M above which they become fundamental modes). The observations above suggest that, for exactly
Analytically, an indication of the existence of such extremal black holes, the fundamental quasinormal
modes can be obtained if, instead of solving Eq. (3.6) frequencies for charged fields (scalar and Dirac) have

FIG. 4 (color online). Real (left plots) and imaginary (right plots) parts of the Dirac quasinormal modes (s ¼ j ¼ 1=2) near the
extremal limit for qQ ¼ 0.01 (top panels) and qQ ¼ 0.1 (bottom panels). Note that near the extremal limit (above Q=M ≈ 0.9985), the
most stable quasinormal mode corresponds to the blue-solid branch, with a real part given approximately by ReðωÞ ≈ qQ=rþ ≈ q. Far
from the extremal limit, the red-dashed branch corresponds to the fundamental frequency. The dots, which correspond to solutions of
Eq. (A8), are in excellent agreement with the results of the continued fraction method.

124011-6
QUASINORMAL MODES OF CHARGED FIELDS AROUND A … PHYSICAL REVIEW D 90, 124011 (2014)
vanishing imaginary parts, with real parts given by the The case of nearly extremal black holes is particularly
field’s charge q. An analogous result was derived by interesting due to the appearance of strange modes which do
Detweiler [38,39] for nearly extreme Kerr black holes not exist for uncharged fields. The imaginary component
(through the calculation of the poles of the reflection ImðωÞ of such modes approaches zero as Q → M, while
coefficient), leading him to suggest that extremal Kerr ReðωÞ approaches qQ=rþ ≈ q. This is exactly the upper
black holes are marginally unstable. It was later argued that limit of frequencies for which superradiant scattering occurs
this marginal instability does not really occur [40–42]. In [47,48]. The existence of these modes indicates that QNMs
our case, the existence of these modes with arbitrarily small of exactly extremal black holes have vanishing imaginary
imaginary parts could be an artifact of the complicated parts, similarly to what occurs for Kerr black holes.
continued fraction equation (3.6), as suggested in Ref. [28]. It is important to remark, however, that Leaver’s original
Fortunately, in the near extremal limit, Eq. (2.9) is method fails to converge for extremal black holes. More
amenable to analytic methods (see the Appendix) and precisely, when M ¼ Q, the regular singularities at r ¼ r−
one is able to verify the existence of these modes by an and at r ¼ rþ merge at r ¼ M, becoming an irregular
alternative calculation. In fact, the QNMs obtained through singularity. Therefore it is hopeless that a power series
the algebraic equation (A8) are in excellent agreement with expansion around r ¼ rþ [like (3.1)] will have a nonzero
the ones obtained through the continued fraction method radius of convergence. Onozawa et al. [49] have proposed a
in the limit Q → M, as shown in Fig. 4. We have also verified modification of Leaver’s method to deal with such type of
that the last two columns of Table I are compatible with the equations and have successfully applied it to uncharged
solutions of Eq. (A8). In order to complement our analysis, it fields around an extremal RN black hole. The obtained
would be very interesting to check if these modes lead to a results are in very good agreement with the ones obtained
marginal instability of the extremal RN black hole or not. for nearly extremal black holes using Leaver’s original
Additionally, we would like to point out that a similar method. As discussed by Leaver in Ref. [34] for uncharged
behavior of purely real frequencies is also observed for fields, this is not a surprise. Even though the convergence
massive fields. As reported in [43,44], arbitrarily long of the original method becomes worse as the extremal limit
living modes (the so-called quasiresonant modes) can exist is approached, it can still be used with great accuracy quite
if the field mass has special values. If one changes the close to the extremal limit. In our case, we have confirmed
boundary condition at infinity to consider quasibound this fact by checking the results of the continued fraction
states instead of quasinormal modes, similar results can method with an alternative calculation. Nonetheless, it
be observed. In fact, it has been shown that the frequencies would be interesting to implement the ideas of Ref. [49]
of quasibound states of massive scalar [45] and Proca [46] to calculate the QNM spectrum of charged fields around
fields have zero imaginary part in the limit Q → M. The exactly extremal black holes.
results of Ref. [46] even indicate that this behavior may Finally, we would like to point out another interesting
occur not only for extremal black holes, but also in more fact concerning our simulations: we have not observed any
generic situations. QNM with positive imaginary part, indicating that non-
extremal RN black holes are stable under massless charged
V. FINAL REMARKS perturbations.
In this paper, we have implemented Leaver’s continued ACKNOWLEDGMENTS
fraction method, together with Nollert’s improvement, to
determine the charged Dirac quasinormal modes of a RN The authors are grateful to Alberto Saa for enlightening
black hole. We were able to compute the quasinormal discussions. M. R. was funded by the São Paulo Research
frequencies for arbitrary values of the interaction parameter Foundation (FAPESP), Grant No. 2013/09357-9.
qQ. For qQ ¼ 0, our results exactly agree with the ones
APPENDIX
obtained in Ref. [36]. For small values of jqQj our results
agree with the ones obtained in Refs. [23,25] through In the limit of extremal charge Q → M, together with the
approximation methods (WKB and Pöschl-Teller poten- assumption ω → qQ=rþ , it is possible to find the QNMs
tial). As we increase the electromagnetic interaction, our of charged fields around a RN black hole by calculating
simulations indicate the existence of a critical value of jqQj the poles of the reflection/transmission coefficients of the
for which the fundamental quasinormal modes have arbi- associated scattering problem. Such analysis has been
trarily small real oscillation frequencies. In the limit of performed in Ref. [26] for scalar fields. In what follows,
large jqQj, on the other hand, the QNMs where obtained we generalize the analysis of Ref. [26] to include not only
both numerically and analytically for the first time, and the the scalar case s ¼ 0 but also the Dirac case s ¼ 1=2. It is
results where compared with the scalar case. Furthermore, important to remark that such a type of calculation is based
our description unifies the treatment of charged QNMs for on the seminal work of Press and Teukolsky [9] and has
scalar and Dirac fields through the master equation (2.9) been used for the first time by Detweiler to determine the
and the continued fraction coefficients (3.3) and (3.4). QNMs of a nearly extreme Kerr black hole [38].

124011-7
MAURÍCIO RICHARTZ AND DAVI GIUGNO PHYSICAL REVIEW D 90, 124011 (2014)

Inspired by the notation of Ref. [26], we use new dimensionless variables, namely,
r − rþ rþ − r−
x¼ ; τ¼ ; ω̂ ¼ ωrþ ; k ¼ 2ω̂ − qQ; ϖ ¼ ω̂ − qQ; ðA1Þ
rþ rþ

in order to rewrite Eq. (2.9) as

 2 
2 2 00 K 0 K
x ðx þ τÞ R þ ðs þ 1Þð2x þ τÞxðx þ τÞR þ 2 − isð2x þ τÞ þ xðx þ τÞð4isω̂x þ 2isk − λs Þ R ¼ 0; ðA2Þ
rþ rþ

where K=rþ ¼ ω̂x2 þ kx þ ϖ, R ¼ RðxÞ is the radial function, and 0 denotes the derivative with respect to x. We are
interested in solving the equation above in the double limit τ → 0, ϖ → 0. To do that, we first note that in the far region
x ≫ maxðτ; ϖÞ, Eq. (A2) can be approximated by

x2 R00 þ 2ðs þ 1ÞxR0 þ ½ðω̂x þ kÞ2 − 2isðω̂x þ kÞ þ ð4isω̂x þ 2isk − λs ÞR ¼ 0: ðA3Þ

Its most general solution can be written in terms of confluent hypergeometric functions,
 
1 1
R ¼ C1 e−iω̂x x−2−2þiδ 1 F1 − s þ ik þ iδ; 1 þ 2iδ; 2iω̂x þ C2 ðδ → −δÞ; ðA4Þ
2

where C1 and C2 are constants, δ2 ¼ k2 − ðj þ 12Þ2 , and the notation δ → −δ means that the preceding term should be
repeated with −δ replacing δ.
Near the horizon (x ≪ 1), on the other hand, the radial equation can be approximated by

x2 ðx þ τÞ2 R00 þ ðs þ 1Þð2x þ τÞxðx þ τÞR0 þ ½ðkx þ ϖÞ2 − isð2x þ τÞðkx þ ϖÞ þ xðx þ τÞð2isk − λs ÞR ¼ 0: ðA5Þ

The only solution of the equation above which is compatible with the boundary condition of no outgoing waves near the
event horizon is
 
−s−iϖτ −s−ikþiϖτ 1 1 ϖ x
R¼x ðx þ τÞ 2 F1 − s − ik þ iδ; − s − ik − iδ; 1 − s − 2i ; − : ðA6Þ
2 2 τ τ

By matching the solutions (A4) and (A6) above in the overlap region maxðτ; ϖÞ ≪ x ≪ 1, the coefficients C1 and C2 can
be determined:
1 1
Γð1 − s − 2i ϖτÞΓð2iδÞτ2−s−ik−iδ Γð1 − s − 2i ϖτÞΓð−2iδÞτ2−s−ikþiδ
C1 ¼ 1 ; C2 ¼ 1 : ðA7Þ
Γð2 − s − ik þ iδÞΓð12 þ ik þ iδ − 2i ϖτÞ Γð2 − s − ik − iδÞΓð12 þ ik − iδ − 2i ϖτÞ

Finally, the requirement that Eq. (A4) be compatible with (2.12) in the limit x → ∞ (i.e., no ingoing waves) produces an
algebraic equation for the QNM frequencies:

Γð2iδÞΓð1 þ 2iδÞΓð12 − s − ik − iδÞΓð12 þ s − ik − iδÞΓð12 þ ik − iδ − 2i ϖτÞ

þ ð−2iω̂τÞ2iδ ¼ 0: ðA8Þ
Γð−2iδÞΓð1 − 2iδÞΓð12 − s − ik þ iδÞΓð12 þ s − ik þ iδÞΓð12 þ ik þ iδ − 2i ϖτÞ

As expected, the equation above reduces to the However, the results of Ref. [24] do not seem to reduce
expression obtained in Ref. [26] when s ¼ 0. We also to the expression above in the limit of zero rotation (and
note that a similar calculation was performed in Ref. [24] consequently, they are also incompatible with the results
for charged perturbations of a Kerr-Newman black hole. of Ref. [26]).

124011-8
QUASINORMAL MODES OF CHARGED FIELDS AROUND A … PHYSICAL REVIEW D 90, 124011 (2014)
[1] T. Regge and J. A. Wheeler, Phys. Rev. 108, 1063 (1957). [25] Y.-J. Wu and Z. Zhao, Phys. Rev. D 69, 084015 (2004).
[2] C. Vishveshwara, Nature (London) 227, 936 (1970). [26] S. Hod, Phys. Lett. A 374, 2901 (2010).
[3] W. H. Press, Astrophys. J. 170, L105 (1971). [27] S. Hod, Phys. Lett. B 710, 349 (2012).
[4] S. L. Detweiler, Proc. R. Soc. A 352, 381 (1977). [28] R. Konoplya and A. Zhidenko, Phys. Rev. D 88, 024054
[5] S. L. Detweiler and E. Szedenits, Astrophys. J. 231, 211 (2013).
(1979). [29] J. Stewart, Advanced General Relativity (Cambridge
[6] M. Davis, R. Ruffini, W. Press, and R. Price, Phys. Rev. University Press, Cambridge, 1991).
Lett. 27, 1466 (1971). [30] J. N. Goldberg, A. J. Macfarlane, E. T. Newman, F.
[7] C. Cunningham, R. Price, and V. Moncrief, Astrophys. J. Rohrlich, and E. C. G. Sudarshan, J. Math. Phys. (N.Y.)
224, 643 (1978). 8, 2155 (1967).
[8] C. Cunningham, R. Price, and V. Moncrief, Astrophys. J. [31] M. Richartz and A. Saa, Phys. Rev. D 84, 104021 (2011).
230, 870 (1979). [32] S. A. Teukolsky, Phys. Rev. Lett. 29, 1114 (1972).
[9] W. H. Press and S. A. Teukolsky, Astrophys. J. 185, 649 [33] E. Leaver, Proc. R. Soc. A 402, 285 (1985).
(1973). [34] E. W. Leaver, Phys. Rev. D 41, 2986 (1990).
[10] B. F. Whiting, J. Math. Phys. (N.Y.) 30, 1301 (1989). [35] H.-P. Nollert, Phys. Rev. D 47, 5253 (1993).
[11] S. Hod, Phys. Rev. Lett. 81, 4293 (1998). [36] J.-l. Jing, J. High Energy Phys. 12 (2005) 005.
[12] S. Hod, Classical Quantum Gravity 23, L23 (2006). [37] W. Gautschi, SIAM Rev. 9, 24 (1967).
[13] C. Corda, Int. J. Mod. Phys. D 21, 1242023 (2012). [38] S. Detweiler, Astrophys. J. 239, 292 (1980).
[14] D. Birmingham, I. Sachs, and S. N. Solodukhin, Phys. Rev. [39] V. Cardoso, Phys. Rev. D 70, 127502 (2004).
Lett. 88, 151301 (2002). [40] V. Ferrari and B. Mashhoon, Phys. Rev. D 30, 295 (1984).
[15] E. Berti, V. Cardoso, and J. P. S. Lemos, Phys. Rev. D 70, [41] K. Glampedakis and N. Andersson, Phys. Rev. D 64,
124006 (2004). 104021 (2001).
[16] K. D. Kokkotas and B. G. Schmidt, Living Rev. Relativity 2, [42] E. Berti, Conf. Proc. C0405132, 145 (2004).
2 (1999). [43] A. Ohashi and M. Sakagami, Classical Quantum Gravity 21,
[17] H.-P. Nollert, Classical Quantum Gravity 16, R159 (1999). 3973 (2004).
[18] E. Berti, V. Cardoso, and A. O. Starinets, Classical Quantum [44] R. Konoplya and A. Zhidenko, Phys. Lett. B 609, 377
Gravity 26, 163001 (2009). (2005).
[19] R. A. Konoplya and A. Zhidenko, Rev. Mod. Phys. 83, 793 [45] J. C. Degollado and C. A. R. Herdeiro, Gen. Relativ. Gravit.
(2011). 45, 2483 (2013).
[20] R. Konoplya, Phys. Lett. B 550, 117 (2002). [46] M. O. P. Sampaio, C. Herdeiro, and M. Wang, Phys. Rev. D
[21] R. Konoplya, Phys. Rev. D 66, 084007 (2002). 90, 064004 (2014).
[22] R. A. Konoplya and A. Zhidenko, Phys. Rev. D 76, 084018 [47] J. D. Bekenstein, Phys. Rev. D 7, 949 (1973).
(2007). [48] M. Richartz, S. Weinfurtner, A. J. Penner, and W. G. Unruh,
[23] J. Chang and Y. Shen, Int. J. Theor. Phys. 46, 1570 (2007). Phys. Rev. D 80, 124016 (2009).
[24] J. Jing, Q.-Y. Pan, and X. He, Int. J. Mod. Phys. D 16, 81 [49] H. Onozawa, T. Mishima, T. Okamura, and H. Ishihara,
(2007). Phys. Rev. D 53, 7033 (1996).

124011-9