Black Hole

Quasinormal Modes
Hing Tong Cho
(Tamkang University)

IPAS, Jan 2018

Gravitational wave signal of a perturbed black hole
or binary black holes merger ringdown

initial wave pulse

quasinormal “ringing”

late time power-law tail

 Content
• What is a quasinormal mode?
• Black hole quasinormal modes
• Quasinormal modes and tails
• Computations of quasinormal mode
frequencies
• Extending to rotating black holes
• Discussions
What is a quasinormal mode?
1 2 2
Harmonic oscillator potential V ( x ) =  x
4
Asymptotic solutions

 x2 / 4  x2 / 4
e e

− x 2 / 4 − x 2 / 4
e e
discrete bound state spectrum
  x2 / 4
e
B

A C
− x 2 / 4 − x 2 / 4
e e
2nd order differential eq.
− x 2 / 4  x2 / 4 − x 2 / 4
Ae Be + Ce

B = 0  discrete energy eigenvalues

 Scattering states: incident wave from the left

 Be − ikx

Ae−ikx  → Ceikx

− ikx 2nd order differential eq.

Ae Be−ikx + Ceikx

Reflection probability | C |2 / | B |2
Transmission probability | A | / | B |
2 2

No restrictions on the energy: continuous spectrum

 Discrete spectra from scattering states (I)

− ikx
Ae   Be − ikx

− ikx 2nd order differential eq.

Ae Be−ikx + Ceikx

C = 0  discrete energy (real) eigenvalues

This is called the total transmission spectrum

 For example: square barrier potential

V0
 e − ikx

Te − ikx
 → R eikx
0 a

n 2 2
R = 0  En = V0 + 2 , n = 1,2,3,
a
Discrete total transmission spectrum ( En  V0 )
 Discrete spectra from scattering states (II)

− ikx
Ae  → Ce ikx

− ikx 2nd order differential eq.

Ae Be−ikx + Ceikx

B = 0  discrete energy (complex) eigenvalues

This is called the quasinormal spectrum representing

decay modes
 For example: Poschl-Teller potential

V ( x ) = sech2 x

e − ikx
 → e ikx
The quasinormal spectrum can be solved exactly
2
 3  1 
En =  − i  n +  , n = 0,1,2,
 2  2 
-800 -600 -400 -200 n=0
Re ( En )
-10

-20

-30

-40

-50 Im ( En )
n = 30
Black hole quasinormal modes

Black hole perturbation: From the

Einstein equation

R = 0

where the metric is given by

g  = g  + h
 The background metric is the
Schwarzschild black hole metric

−1
 2M  2  2M 
ds = −1 −
2
 dt + 1 −  dr 2
+ r 2
d 2

 r   r 

To first order in h

 R = 0
   h +    h −    h  −   h = 0
Since the spacetime is spherically symmetric,
one can expand the fields in terms of the
tensor harmonics on the 2-sphere.

(V )
a
(1)
lm
= Ylm ;a ; (V )
a
( 2)
lm
=  a Ylm ;b
b

(T )
(1)
ab lm = Ylm ;ab ; (T )
( 2)
ab lm = Ylm  ab

(T )
( 3)
ab lm
1 c

=  a Ylm ;cb +  b Ylm ;ca
2
c

where Ylm are the spherical harmonics, and
 ab and  a b are the metric and anti-symmetric
tensor on the 2-sphere.
For example, the Regge-Wheeler
perturbation

 
0 0 h0 (r , t )Va( 2) ( ,  ) 
 
h = 0 0 h1 (r , t )Va( 2) ( ,  ) 
* * 
 h2 (r , t )Tab(3) ( ,  ) 
* * 
Note that h is symmetric.
Gauge symmetric: general coordinate
transformation

h' = h +   +  

where   is the gauge vector.

The gauge vector can be chosen such

that
h2 (r , t ) = 0
Putting the Regge-Wheeler ansatz into
the perturbation equations, one obtain
the Regge-Wheeler equation

 2  2
− 2 + VRW  = 0
t 2
x
where

1  2M 
 = 1 −  h1 (r , t )
r r 
Note that we have used the tortoise
coordinate

 r 
x = r + 2 M ln  − 1
 2M 

where outside the horizon

2M  r    −  x  
Regge-Wheeler potential

 2M   l (l + 1) 6 M 
VRW = 1 −  2 − 3 
 r  r r 

VRW

x (M )
Since the spacetime is static, VRW is
independent of time. One can take
− i t
~e
Then the Regge-Wheeler equation

 2
− 2 + VRW  =  2
x
is Schroedinger equation-like and would
have the corresponding quasinormal
modes.
Quasinormal modes and tails

initial wave pulse

quasinormal “ringing”

late time power-law tail

Initial value problem with

 ( x,0) = u( x ) ;  t ( x, t ) |t =0 = v ( x )

Then using the retarded Green’s function

 2G  2 G
− 2 + VRW G =  (t ) ( x − x ' )
t 2
x

with G( x, x' ; t ) = 0 , t  0
The solution can be expressed as

 ( x, t )
=  dx' u( x ' ) t G ( x, x ' ; t ) + v ( x ' )G ( x, x' ; t )

Our main task now is to analyze the

Green’s function for the corresponding
Regge-Wheeler potential
Using the Fourier transform

 ~
G ( x, x ' ; t ) =  d e −i t
G ( x, x ' ;  )
−

~
where G ( x , x ' ;  ) is analytic in the
upper half  -plane because
G ( x, x ' ; t ) = 0 , t  0
~
G ( x, x ' ;  ) satisfies
2~
2~ G ~
−  G − 2 + VRW G =  ( x − x ' )
x
For x  x' we have
 2
f
−  f − 2 + VRW f = 0
2

x
f ( x,  ) → e − i x , x → −
That is, wave goes into the horizon
For x  x' we have

 2
g
−  g − 2 + VRW g = 0
2

x

g ( x,  ) → ei x , x → 

Wave goes out to infinity

Then the Green’s function is given by

 f ( x,  ) g ( x ' ,  )
 x  x'
~  W ( )
G ( x, x ' ;  ) = 
f ( x ' ,  ) g ( x,  )
 x  x'

 W ( )

where
W () = g ( x, ) f ' ( x, ) − g ' ( x, ) f ( x, )
is the Wronskian
For t  0,

we need to close the contour in the

lower half plane to obtain G( x, x' ; t )
Examine the singularities in the lower
half plane
 ~
Poles of G ( x, x ' ;  ) occur at the
zeros of the Wronskian

f ' g'
W = gf '− g ' f = 0  =
f g
 f g

A solution with outgoing boundary

conditions at x =  : Quasinormal mode
Suppose
i) Quasinormal modes are the only
singularities
ii) The contribution from the semi-circle
at infinity vanishes

This would imply the completeness of

the quasinormal modes.
However, there are other singularities for
the Regge-Wheeler potential

In particular, there is a cut on the negative

imaginary axis which is relevant to the late-
time power-law tail of the perturbation

This singularity comes from the non-

analytic behavior of g ( x,  )
Consider
 2
g
−  g − 2 + VRW g = 0
2

x
and as x → 
l (l + 1) l (l + 1)4 M  x 
VRW = + ln   +
 2M 
2 3
x x

The cut singularity is related to this

asymptotic behavior of the potential
Treating the inverse-square part
exactly, that is,

 2 ( 0)
g l (l + 1) ( 0 )
− g −
2 ( 0)
+ g =0
x 2
x 2

i x
g ( x,  ) → e
( 0)
, x→

We have
g ( 0 ) ( x,  ) = e i ( l +1) / 2 ( x)hl(1) ( x)
Then g ( x,  ) satisfies the integral
equation
g ( x,  )

= g ( x,  ) +  dx' M ( x, x' ;  )V ( x' ) g ( x' ,  )
( 0)
x

where
l (l + 1)
V ( x) = V ( x) − 2
x
is the subtracted potential
 and

M ( x, x ' ;  )
i

= −  x x' hl(1) ( x' )hl( 2 ) ( x) − hl(1) ( x)hl( 2 ) ( x' )
2

is the zeroth order Green’s function
First we consider a model power-law
potential

K  x0 
V ( x) = 2  
x0  x 

Then to the first Born approximation

g ( x,  )

= g ( x,  ) +  dx' M ( x, x' ;  )V ( x' ) g ( x' ,  )
( 0) ( 0)
x
and after some manipulation one has

g ( x,  )  C (l ,  )( x0 ) − 2

Thus there is a branch cut which can

be put on the negative imaginary axis

The contours on both sides of the cut

give a nonzero contribution

G ( x, x ' ; t )  C (l ,  ) t − ( 2 l + ) , t → 
For the Regge-Wheeler potential
l (l + 1)4 M  x 
V = ln  
 2M 
3
x

one can obtain the result for this potential

by differentiating with respect to 

G ( x, x ' ; t ) 



C (l ,  ) t −( 2 l + ) 
 =3

C (l ,  ) − ( 2 l + 3)
 t
  =3
because we have

C (l ,3) = 0

One obtain the famous power-law tail for

the Schwarzschild black hole background

 ( x, t )  G ( x, x ' ; t )
−( 2 l + 3)
t
for t → 
 late time power-law tail

intermediate time quasinormal

early time initial pulse modes
Computation of QNM frequencies
Quasinormal modes are solutions to the
Regge-Wheeler equation

 2
−   − 2 + VRW = 0
2

x
 ( x,  ) → e  i x , x → 

The corresponding frequencies are in

general complex
Direct numerical evaluation (Chandrasekhar
and Detweiler)
Problems with direct numerical evaluations:

 = R + iI
− i t I t
I  0 for stability as e ~ e →0 , t →

However,

 ( x,  ) → e  i x ~ e | | x , x → 
I

which is the dominant solution. It will easily be

contaminated by the sub-dominant solution. This
will lead to instabilities in numerical schemes.
Solvable potential (Ferrari and Mashhoon)

Approximate the black hole potential by

a solvable one, for example, the Poschl-
Teller potential

V0
VPT ( x) =
cosh2 ( x − x0 )
The corresponding quasinormal frequencies are

1/ 2
 2
  1
EPT = V0 −  − i  n + 
 4   2

n = 0,1,2,
Note that for the Poschl-Teller potential

− x
VPT ( x)  4V0e as x → 

while for the Schwarzschild potential

l (l + 1)
V ( x)  2
as x → 
x
− x / 2M
e as x → −
Matching the potentials near the maxima,

Height: V0 = V ( xm ax)

1  d 2
V ( x) 
Curvature:  =−
2
 2

2V0  dx  x = x
m ax
Schwarzschild quasinormal frequencies
(Ferrari and Mashhoon 1984)
Continued fraction method (Leaver)

The solution to the Regge-Wheeler equation

 r −1 
 m

(r ) = (r − 1) −i
r i 2 i ( r −1)
e 
m =0
am 
 r 


which satisfies the boundary condition at the

event horizon

r = 2M → 1
Substituting this ansatz into the Regge-Wheeler
equation

 0 a1 +  0 a0 = 0
 m am +1 +  m am +  m am −1 = 0, m = 1,2,

where

 m = m 2 − (2i − 2)m − 2i + 1

 m = −2m 2 + (8i − 2)m − 8 2 − 4i
+ l (l + 1) − 3
 m = m 2 − 4i m − 4 2 − 4
To satisfy the boundary condition at infinity


 Convergence of the sum a
m =0
m

 Continued fraction relation

am +1 −  m +1
=
am  m+1 m + 2
 m +1 −
 m + 2 m + 3
 m+ 2 −
 m+3 − 
Quasinormal frequency condition is given by

a1 0 − 1
=− =
a0 0  − 1 2
1
 2 3
2 −
3 − 
 0 1
 0 − =0
1 2
1 −
 2 3
2 −
3 − 
Schwarzschild quasinormal frequencies with
(Leaver 1985)
 WKB approximation (Schutz and Will)

 e −iEr
 T e − iEr III II I
→ R e iEr

r2 r0 r1 r
Regions I and III:
Standard WKB wavefunctions
r
r 1
iEr dr '[ E 2 −V − E ]
A+ e i
I  e
(E 2
−V )
1/ 4

r
−iEr
A− e −i r1 dr '[ E 2 −V − E ]
+ e
(E 2
−V )
1/ 4

where

r 1
−i dr '[ E 2 −V − E ]
A+ = R E e

r 1
i dr '[ E 2 −V − E ]
A− = E e
 r
− iEr
Be −i r 2 dr '[ E 2 −V − E ]
III  e
(E 2
−V )1/ 4

where

r2

B = T E e −
−i dr '[ E 2 −V − E ]
Region II:
Parabolic approximation for the middle part
1
V (r )  V (r0 ) + V ' ' (r0 )(r* − r0 ) 2
2
The Schrodinger equation becomes
d 2  2
( ) 2
+  E − V0 − V0 ' ' (r − r0 )   = 0
1
 
2
dr 2
d 2
(
 2 + z2 +  2  = 0
dz
)
where
z = 1/ 4 (r − r0 );  2 = (E 2 −V0 ) /  ;  = −V0 ' ' / 2
The wavefunctions in region II can be
expressed in terms of parabolic cylinder
functions D (z )

II   D 1 i
− −
2 ( 2 e i / 4 z )
2 2

+D 1 i 2
− −
(− 2 e i / 4 z )
2 2
Asymptotic matchings the wavefunctions in
regions I, II, and III give the transmission and
reflection probabilities

| T |2 = 1 − | R |2
1
=  2
(1 + e )
Quasinormal mode condition:
out-going waves only

| T |2 , | R |2 → 

   2 = −i (2n + 1) , n = 0,1,2,...

  ( E − V0 ) /  = −i (2n + 1)
2

 1
(
 E = V0 − i n +  − 2V0''
2
)
1/ 2

 2
Quasinormal mode condition to 3rd order in
the WKB approximation (Schutz, Will, and
Iyer)
 1
E = V0 − i n +  − 2V0
2
(
'' 1 / 2
)
 2

('' 1 / 2
)
 1
+ − 2V0  − i n +  − 2V0'' ( )1/ 2

 2
where

    
2

1 V0  1 2
( )

( 4) '''
1 1 V0
= 
  ''   +   −  ''  7 + 60 2 
(− 2V )
0
'' 1/ 2
  0 
 8 V 4  288  V0  

  5  V '' '  4  ' '' 2 ( 4 )  
  0
 ''  (
77 + 188 2
− )
1  V 0 V0
384  V0''3 
(
 51 + 100 2 )

 6912  V0  
 
 1  V0'''V0(5)  
2
1  V0 
( ) ( )
( 4)
1
=  +   +  2
+ +  2

( )
− 2V0''  2304  V0'' 
67 68
288  V0'' 2 
19 28

 
− 1  V0  5 + 4 2
( )
(6)

 288  V0''  
 

1 d nV
 =n+ ; V0( n ) = n
2 dr* r* = r* ( r0 )
Schwarzschild quasinormal frequencies
(Iyer 1987)
The WKB approximation is accurate for low-
lying modes, error of the order of a few percent.

The approximation is systematic. The order of

approximation has been given by Zonoplya
to the 6th order recently.
Extending to rotating black holes
Astrophysical black holes usually possess
angular momentum: Kerr black holes
 2
ds 2 = − (dt − a sin 2 d ) 2 + dr 2 +  2 d 2
2 
sin 2
+ [(r 2 + a 2 ) d − a dt ]2
2

 = r 2 − 2M r + a 2 ;  2 = r 2 + a 2 cos 2
Two Killing vector fields:
Time translation related to energy
conservation
Rotation symmetry related to angular
momentum conservation

Hidden symmetry:
Killing tensor gives another conserved
quantity
 Teukolsky equation:

 (r 2 + a 2 ) 2   2
 4 aM r  2

− − a sin   2 −
2 2

   t  t 
 M (r 2 − a 2 )  
+ 4 r + ia cos − 
   t
    1       1 a 2
  2

 sin  
−1
+ 2
 + + 2 −  2
r  r  sin      sin    
 a (r − M ) i cos  
− 4 +  − ( 4 cot2
 + 2) = 0
  sin   
2
The Teukolsky equation is separable

 = ei t eim R(r ) S ( )

d  −1 dR   K 2 + 4i(r − M ) K 
 2
 + − 8i r −   R = 0
dr  dr    

1 d  dS 
 sin  
sin d  d 
 2 2 2 m2 4mcos 
+ a  cos  − + 4a cos + − 4cot  + E + 4 S = 0
2

 sin 
2
sin 
2


K = (r 2 + a 2 ) − am ;  = E − 2 + a 2 2 − 2am
 Discussions
1. The quasinormal mode frequencies can be
evaluated quite accurately using semi-numerical
methods, like the continued fraction and WKB.
2. The study of black hole quasinormal modes has
been termed the “Black Hole Spectroscopy”.
3. The observations of gravitational waves provide
the opportunity to examine the theory of general
relativity in detail through the study of black hole
spectroscopy.
4. The Kerr power or log tails have not been
examined fully. More work needs to be done.
However, this is less relevant to the current
observations of gravitational waves
5. The hidden symmetry study can be extended
to higher dimensional rotating Myers-Perry
black holes