Gravitational wave energy spectrum of a parabolic encounter

Christopher P. L. Berry and Jonathan R. Gair

Institute of Astronomy, Madingley Road, Cambridge, CB3 0HA, United Kingdom
(Dated: November 19, 2010)
We derive an analytic expression for the energy spectrum of gravitational waves from a parabolic
Keplerian binary by taking the limit of the Peters and Mathews spectrum for eccentric orbits. This
demonstrates that the location of the peak of the energy spectrum depends primarily on the orbital
periapse rather than the eccentricity. We compare this weak-field result to strong-field calculations
and find it is reasonably accurate ( 10%) provided that the azimuthal and radial orbital frequencies
do not differ by more than 10%. For equatorial orbits in the Kerr spacetime, this corresponds
to periapse radii of rp & 20M . These results can be used to model radiation bursts from compact
objects on highly eccentric orbits about massive black holes in the local Universe, which could be
arXiv:1010.3865v2 [gr-qc] 18 Nov 2010

detected by the Laser Interferometer Space Antenna (LISA).

PACS numbers: 04.30.w, 04.25.Nx, 97.60.Lf, 98.35.Jk

I. INTRODUCTION apse rp & 20M . The parabolic spectrum takes a neat

analytic form; deriving it from the bound spectrum will
An important source of gravitational waves for the allow corrections for high-eccentricity bound orbits to be
proposed space-based gravitational wave detector, the found in the future. We hope this note will be a useful
Laser Interferometer Space Antenna (LISA) [1, 2], are resource for future work on gravitational radiation from
the inspirals of stellar-mass compact objects into mas- high-eccentricity orbits.
sive black holes in the centres of galaxies. During the
last few years of inspiral these systems generate continu-
ous gravitational waves in the LISA band, which will al- II. PARABOLIC LIMIT
low the detection of as many as several hundred systems
out to redshift z 1 [3]. However, prior to this phase, A. Energy Spectrum
the inspiraling object spends many years on a highly ec-
centric orbit, generating bursts of gravitational radiation For an orbit of eccentricity e with periapse radius rp ,
at each periapse passage. LISA could resolve individual Peters and Mathews [6] give the power radiated into the
bursts from sources in the nearby Universe. Initial es- nth harmonic of the orbital angular frequency as
timates [4] suggested a LISA event rate of 18 yr1 ,
including 15 yr1 from the centre of the Milky Way, 32 G4 M12 M22 (M1 + M2 )(1 e)5
P (n) = g(n, e), (1)
although this was subsequently revised downwards to 5 c5 rp5
1 yr1 with negligible contribution from extragalac-
tic sources [5]. If even a single burst from the Galactic where the function g(n, e) is defined in terms of Bessel
centre is detected during the LISA mission, this will pro- functions of the first kind
vide an unparalleled probe of the structure of spacetime
n4

there. g(n, e) = Jn2 (ne) 2eJn1 (ne)
The spectrum of radiation from these bursts will be 32
2
well approximated by the spectrum of a parabolic orbit.1 2
In this note we derive an analytic approximation to this + Jn (ne) + 2eJn+1 (ne) Jn+2 (ne)
n
spectrum by taking the limit of the Peters and Math- 2
+ 1 e2 [Jn2 (ne) 2Jn (ne) + Jn+2 (ne)]


ews [6, 7] (PM) energy spectrum for eccentric Keplerian 
binaries. We show that the peak of the spectrum de- 4 2
+ 2 [Jn (ne)] . (2)
pends primarily on the orbital periapse and only weakly 3n
on the eccentricity. We also estimate the range of va-
lidity of the approximation (in Sec. III) by comparing The Keplerian orbital frequency is
to numerical Teukolsky data, finding that it is a good
approximation for equatorial orbits in Kerr with peri- G(M1 + M2 )(1 e)3
12 = = (1 e)3 c2 , (3)
rp3

where c is defined as the angular frequency of a circular

cplb2@ast.cam.ac.uk orbit of radius rp . The energy radiated per orbit into the
jgair@ast.cam.ac.uk nth harmonic, that is, at frequency n = n1 , is
1 We use parabolic to refer to marginally bound orbits.
Marginally bound Keplerian orbits are parabola; in curved space- 2
times they do not retain such a simple shape. E(n) = P (n); (4)
1
 2

as e 1 for a parabolic orbit, 1 0 as the orbital where K (z) is a modified Bessel function of the second
period becomes infinite. The energy radiated per orbit is kind. Using this to evaluate the limit gives
then the total energy radiated. The spacing of harmonics r r !
is = 1 , giving energy spectrum 1 2( z) 23/2 ( z)3
J (z) K1/3 . (16)
 3z 3 z
dE 
1 = E(n). (5)
d n For our case,
!
23/2 f
r
Changing to linear frequency 2f = , 1 2
Jn (ne) (1 e)1/2 K1/3 , (17)
3 3
128 2 G3 M12 M22

dE 
= (1 e)2 g(n, e) (6)
df fn 5 c5 rp2 and the first limiting function is well defined,
2
4 3
G M12 M22
= (n, e), (7)
!
3/2
r
5 c5 rp2 1 2 2 f
A(f) = K1/3 . (18)
3 3
where the function (n, e) is defined in the last line. For
a parabolic orbit, we must take the limit of (n, e) as To find the derivative we combine (9) and (16), and
e 1. expand to lowest order yielding
We simplify (n, e) using the recurrence formulae " !
23/2 f
r
(Watson [8] 2.12) 1 2 3/2
Jn (ne) (1 e) 2 K1/3
2 3 3
2
J1 (z) + J+1 (z) = J (z) (8) !#
z 1 23/2 f
+ K1/3 . (19)
J1 (z) J+1 (z) = 2J (z), (9) f 3
and eliminate n using We may re express the derivative using the recurrence
n formula (Watson [8] 3.71)
n= = (1 e)3/2 f, (10)
1 K1 (z) + K+1 (z) = 2K (z) (20)

where f = n /c = fn /fc is a dimensionless frequency. to give

To find the limit we define two new functions " ! !
1 e 23/2 f 23/2 f

Jn (ne)
 
Jn (ne)
 Jn (ne) K2/3 + K4/3

A(f ) = lim
; B(f ) = lim . 3 3 3
e1 (1 e)1/2 e1 1e !#
(11) 1 23/2 f
To give a well-defined energy spectrum, both of these K1/3 . (21)
2f 3
must be finite.
The Bessel function has an integral representation And so finally we obtain the well-defined
" ! !
1 1 23/2 f 23/2 f
Z
J (z) = cos( z sin ) d; (12)
B(f ) = K2/3 + K4/3
0 3 3 3
!#
we want the limit of this for , z , with z . 1 23/2 f
K1/3 . (22)
Using the stationary phase approximation, the dominant 2f 3
contribution to the integral comes from the regime in
which the argument of the cosine is approximately zero Having obtained expressions for A(f) and B(f) in
(Watson [8] 8.2, 8.43), for small : terms of standard functions, we can now calculate the
energy spectrum for a parabolic orbit. From (7)
1 z 
Z 
J (z) cos z + 3 d (13)
0 6 4 2 G3 M12 M22
 
dE f
Z = , (23)
1  z  df 5 c5 rp2 fc
cos z + 3 d; (14)
0 6
where we have used the limit
this last expression is an Airy integral and has a standard h i2
form (Watson [8] 6.4) (f) = 8f2 B(f) 2fA(f)

!
Z
x
 3/2 
2x 4 4f2 h i2
3
cos(t + xt) dt = K1/3 , (15) + 128f + A(f ) . (24)
3 33/2 3
0
 3

20
Integrating the energy spectrum (23) gives
18 Z
2 G3 M12 M22
16 Eint = c (f) df. (29)
5 c5 rp2 0
14

The integral can be evaluated numerically as

12
Z
425
(n, e)

10
(f) df = 12.5216858 . . . = 7/2 . (30)
0 2 3
8

6
The two total energies are consistent, Eint = Esum .
4

2 III. APPLICABILITY
0
0 1 2 3 4 5 6
A. Limit of Approximation
f /fc

FIG. 1. The relative energy (per orbit) spectrum (n, e) for

The PM approach assumes Keplerian orbits in flat
e = 0.2 (heavy line), e = 0.5 (medium line), e = 0.7 (light spacetime. This should be a valid approximation in
line), and the limiting result for e = 1 (dashed line) versus the weak-field regime far from a massive body. To find
frequency. the limit of this approximation, we can compare the
PM results with those from more accurate techniques.
Energy spectra for parabolic orbits do not seem to be
This agrees with the e = 1 form of Turners result, which available in the literature yet, so we will make do with
was computed by direct integration along unbound or- the total energy fluxes calculated by Martel [10], who
bits [9]. Figure 1 shows how (n, e) changes with ec- uses time-domain black hole perturbation theory for a
centricity including our result for a parabolic encounter Schwarzschild black hole of mass M . Figure 2 shows the
(cf. Figure 3 of Peters and Mathews [6]). Although ratio of the two energies as a function of periapse dis-
more power is radiated into higher harmonics, the peak tance. As expected the PM result is more accurate for
of the spectrum does not move much: it is always be- larger periapses. The agreement worsens as the periapsis
tween f = fc and f = 2fc , with f = 2fc for e = 0 and decreases. At rp = 4M , corresponding to the radius of
f 1.637fc for e = 1. the innermost stable circular orbit (ISCO), the energy
flux calculated by Martel diverges, so the ratio tends to
zero. This divergence is because in Schwarzschild (or
B. Total Energy Kerr) spacetime a parabolic orbit may have a zoom-whirl
structure where it undergoes a number of near circular
To check the validity of this limit we can calculate the rotations (whirls) about the black hole. As the radius of
total energy radiated by integrating (23) over all frequen- the ISCO is approached, the number of whirls tends to
cies, or by summing the energy radiated into each har- infinity (in the absence of radiation reaction), so an in-
monic. These must yield the same result. Summing: finite amount of energy is radiated. Figure 2 also shows
how the ratio of energies follows the number of rotations,
64 G3 M12 M22 X defined as N = /2, where is the total change in
Esum = 5 2
c (1 e)7/2 g(n, e), (25) the azimuthal angle over one orbit. As N increases, the
5 c rp n PM approximation worsens because the Keplerian orbit
does not include this extra rotation. The accuracy of
where we have used equations (1), (3) and (4). Peters
the PM result deteriorates rapidly once the orbit transi-
and Mathews [6] provide the result
tions to a zoom-whirl trajectory and is therefore far from
parabolic in shape.
X 1 + (73/24)e2 + (37/96)e4
g(n, e) = . (26) The PM result is accurate to 10% for orbits with
n
(1 e2 )7/2 N . 1.1. We will adopt this as a cutoff point. For an
equatorial orbit in Kerr spacetime,
Using this,
1 d
Z
N= dr
64 G3 M12 M22 1 + (73/24)e2 + (37/96)e4 rp dr
Esum = c ,
5 c5 rp2 (1 + e)7/2 Lz
Z 2
r 2M (1 a/Lz )r
(27) = dr, (31)
2M rp (r2 2M r + a2 )w
which is perfectly well behaved as e 1,
where
85 G3 M 2 M 2 L2z 2
Esum = 5/2 5 1 2 2 c . (28) w2 = r3 r + (Lz a)2 r; (32)
2 3 c rp 2M
 4

1.0 22

0.9 21

0.8
20

0.7
E PM /E Martel ; 1/N

19

0.6
18

r p /M
0.5
17
0.4

16
0.3

15
0.2

0.1 14

0.0 13
0 5 10 15 20 25 1.0 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0

r p /M a/M

FIG. 2. Ratio of the total energy radiated as calculated using FIG. 3. Periapse radius corresponding to N = 1.1 as a func-
the Peters and Mathews [6] approach to that calculated by tion of spin parameter a (solid line). The curve may be ap-
Martel [10] using black hole perturbation theory (solid line) proximated by a straight line rp = 3.91a + 17.36M (dashed
versus periapse radius rp . The latter approach should give line).
more accurate results. Also shown is the reciprocal of the
number of rotations 1/N (dashed line). The Keplerian limit
corresponds to N = 1. R /2 p
where K(m) = 0 d/ 1 m sin2 is the complete
elliptic integral of the first kind. Figure 3 shows the peri-
Lz is the specific angular momentum about the z-axis; a apsis for which N = 1.1 for a range of spins. Equatorial
is the spin parameter, and we have adopted units with orbits with larger periapses should be reasonably approx-
G = c = 1. We will find it useful to define imated by the PM result.
p Nonequatorial orbits are more complicated because of
r = M M 2 a2 , (33) the additional precession of the orbital plane. This extra
rotation will mean that the PM approach is less accurate;
and the two nonzero roots of the cubic w2 however, this should be subdominant to the perihelion
r precession effect and so the cutoff periapsis should not
L2z L4z be much larger than for the equatorial case.
rp, 1 = (Lz a)2 ; (34)
4M 16M 2
the periapsis is the larger root rp > r1 . This equation
implicitly gives Lz as a function of rp . The integral may B. Astrophysical Implications
be rewritten as
Lz
Z 
1 +
 Considering bursts from the Galactic centre, orbits
N= 1+ + dr, (35) with periapses of rp . 120M could generate bursts that
2M rp w r r+ r r would be detectable with LISA [4, 5]. It is therefore
where likely that any such burst that was detected would be
in the regime of validity of the Peters and Mathews ap-
2M ar a2 Lz proach, rp & 20M for equatorial orbits. The results de-
= . (36) scribed in this note will therefore have application in that
2Lz M 2 a2
context, and it should be possible to explore the major-
This may be evaluated using elliptic integrals (Grad- ity of parameter space using this approximation. The
shteyn and Ryzhik [11] 3.131.8, 3.137.8) most interesting orbits, those which come deep within
s       the strong-field region of the black holes spacetime, will
be beyond the range of validity of this approximation,

Lz 2 + r+  r1 r  r1
N= + , but these represent a small subset of all plausible events.
rp M r+ rp  rp r rp  rp
(37) This result may also be applicable for studying
R /2 p parabolic encounters between stellar mass black holes;
where (n|m) = 0 d/(1 n sin2 ) 1 m sin2 is these may occur in densely populated environments such
the complete elliptic integral of the third kind. In the as globular clusters [13] or the Galactic centre [14].
limit of a 0 we recover the Schwarzschild result [12] Bursts from these encounters should be detectable with
s   near-future ground-based detectors, such as the Ad-
Lz 2 r1 vanced Laser Interferometric Gravitational-Wave Obser-
N= K , (38)
rp M rp vatory [13, 14].

[1] P. Bender, A. Brillet, I. Ciufolini, A. M. Cruise,
C. Cutler, K. Danzmann, F. Fidecaro, W. M. Folkner,
J. Hough, P. McNamara, M. Peterseim, D. Robertson,
M. Rodrigues, A. R udiger, M. Sandford, G. Sch afer,
R. Schilling, B. Schutz, C. Speake, R. T. Stebbins,
T. Sumner, P. Touboul, J. Vinet, S. Vitale, H. Ward, and
W. Winkler, LISA Pre-Phase A Report, Tech. Rep. (Max-
Planck-Institut f
ur Quantenoptik, Garching, 1998).
[2] K. Danzmann and A. R udiger, Classical Quantum Grav-
ity, 20, S1 (2003).
[3] J. R. Gair, Classical Quantum Gravity, 26, 094034
(2009).
[4] L. J. Rubbo, K. Holley-Bockelmann, and L. S. Finn,
Astrophys. J., 649, L25 (2006).
[5] C. Hopman, M. Freitag, and S. L. Larson, Mon. Not. R.
Astron. Soc., 378, 129 (2007).
5
ACKNOWLEDGMENTS
CPLB is supported by STFC. JRG is supported by the
Royal Society.
[6] P. C. Peters and J. Mathews, Phys. Rev., 131, 435
(1963).
[7] P. C. Peters, Phys. Rev., 136, B1224 (1964).
[8] G. N. Watson, A Treatise on the Theory of Bessel Func-
tions, 2nd ed., Cambridge Mathematical Library (Cam-
bridge University Press, Cambridge, 1995).
[9] M. Turner, Astrophys. J., 216, 610 (1977).
[10] K. Martel, Phys. Rev. D, 69, 044025 (2004).
[11] I. Gradshteyn and I. Ryzhik, Table of Integrals, Series,
and Products, sixth ed., edited by A. Jeffrey (Academic
Press, London, 2000).
[12] C. Cutler, D. Kennefick, and E. Poisson, Phys. Rev. D,
50, 3816 (1994).
[13] B. Kocsis, M. E. Gaspar, and S. Marka, Astrophys. J.,
648, 411 (2006).
[14] R. M. OLeary, B. Kocsis, and A. Loeb, Mon. Not. R.
Astron. Soc., 395, 2127 (2009).