New Class of Gravitational Wave Templates for Inspiralling Compact Binaries

Achamveedu Gopakumar1, ∗
1
Theoretisch-Physikalisches Institut, Friedrich-Schiller-Universität Jena, Max-Wien-Platz 1, 07743 Jena, Germany
(Dated: February 2, 2008)
Compact binaries inspiralling along quasi-circular orbits are the most plausible gravitational wave
(GW) sources for the operational, planned and proposed laser interferometers. We provide new class
of restricted post-Newtonian accurate GW templates for non-spinning compact binaries inspiralling
along PN accurate quasi-circular orbits. Arguments based on data analysis, theoretical and astro-
physical considerations are invoked to show why these time-domain Taylor approximants should be
interesting to various GW data analysis communities. [? ]

PACS numbers: 04.30.Db, 04.25.Nx

arXiv:0712.3236v1 [gr-qc] 19 Dec 2007

Introduction.— Inspiralling comparable mass com- various types of search templates and let us take a closer
pact binaries are the most plausible sources of gravita- look at the so-called TaylorT1 and TaylorT2 waveforms
tional radiation for the operational, planned and pro- implemented in the LSC Algorithms Library (LAL) [2].
posed laser interferometric GW interferometers. GW These two template families employ the following expres-
data analysts, analyzing noisy data from the interferom- sion for the so-called restricted PN waveform
eters, require accurate and efficient temporally evolving  2/3
GW polarizations, h+ (t) and h× (t), the so-called GW G m ω(t)
h(t) ∝ cos 2 φ(t) , (1)
search templates. It is expected that weak GW signals, c3
buried in the noisy interferometric data, should be ex-
tracted by employing the technique of ‘matched filter- where the proportionality constant may be set to unity
ing’. This is an optimal technique if and only if one for non-spinning compact binaries. At a given PN order,
can construct search templates that accurately model ex- the above mentioned two families provide two slightly
pected GW signals from astrophysical sources, especially different ways to compute ω(t) and φ(t). The TaylorT1
in their phase evolution. Till the late stages of binary family numerically solves the following two differential
inspiral, GW signals may be accurately modeled using equations:
the post-Newtonian (PN) approximation to general rel- 
dφ(t) d ω(t) dE
ativity. The PN approximation to the dynamics of in- = ω(t) ; = −L(ω) , (2)
dt dt dω
spiralling compact binaries, usually modeled to consist
of point masses, provides, for example, the equations of where, for example, L(ω) and E are respectively the
motion as corrections to the Newtonian one in terms of 3.5PN accurate GW energy luminosity and the 3PN ac-
(v/c)2 ∼ Gm/c2 r, where v, m, and r are the character- curate orbital energy for TaylorT1 3.5PN waveforms. In
istic orbital velocity, the total mass, and the typical or- other words, for a given PN member of the TaylorT1 fam-
bital separation, respectively. In PN computations, it is ily, ω(t) and φ(t) are computed by numerically solving the
customary to treat a non-spinning inspiralling compact related approximants in Eq. (2). To construct a mem-
binary to consist of two point masses moving in quasi- ber of TaylorT2 family, say TaylorT2 3.5PN, we require
circular orbits. These PN computations, to date provided 3.5PN (Taylor expanded) accurate version of d ω(t)/dt,
four quantities that are required to do astrophysics with appearing in Eq. (2). The differential equations that de-
GW interferometers. For inspiralling compact binaries, fine ω(t) and φ(t) for TaylorT2 3.5PN waveforms can be
the relevant four quantities are the 3PN accurate dy- symbolically displayed as
namical (orbital) energy E(x), expressed as a PN series

2/3 5/3
in terms of x = G m ω3PN /c3 , ω3PN (t) being the

GMω

dφ(t) d ω(t) 96 2
3PN accurate orbital angular frequency, the 3.5PN ac- = ω(t) ; = ω 1
dt dt 5 c3
curate expression for GW energy luminosity L(x) and
the 2.5PN amplitude corrected expressions for h+ (t) and + O(ν) + O(ν 3/2 ) + O(ν 2 ) + O(ν 5/2 )

h× (t), written in terms of the orbital phase φ and x [1].
+ O(ν 3 ) + O(ν 7/2 ) , (3)
GW data analysts employ these inputs to construct
where ν = 1/c2 is a PN ordering parameter and the ex-
plicit expressions for these PN contributions may be ex-
11 This tracted from Refs. [1]. In the above equation, the chirp
is roughly the version submiited to Phys. Rev. Lett. on
18/08/2007. A substantially modified version that also addresses mass M ≡ m η 3/5 , where η is the usual symmetric mass
objections of the referee is under preparation. ratio and m being the total mass of the binary.
 2

In this paper, we provide prescriptions to compute convenient to define these quantities in the PN accu-
three new types of time-domain Taylor approximants that rate Keplerian type parametric solution to the conserva-
should be, in our opinion, interesting to various GW data tive PN accurate compact binary dynamics, available in
analysis communities. Let us first list the salient fea- Refs. [8]. When eccentricity parameter, say time eccen-
tures of these new templates that also employ an expres- tricity et , associated with the PN accurate Keplerian type
sion similar to Eq. (1) to generate waveforms. The three parametrization approaches zero, one can define PN ac-
important features of our Taylor approximants are the curate orbital angular frequency ω ≡ dφ/dt = n (1 + k).
following. The first point is that, in comparison with This implies that for PN accurate circular orbits, the
TaylorT1 and Taylor T2 waveforms, for a given GW fre- angular part of the orbital motion is simply given by
quency window and at a given PN order, our prescrip- φ − φ0 = n × (1 + k) × (t − t0 ). To 3PN order, using
tions will provide more accumulated GW cycles. Further, Ref. [9], in the limit et → 0, we have
our approaches to compute h(t) are numerically as cheap    
(expensive) as TaylorT1 and Taylor T2 waveforms. Let 2/3 39 4/3 315
ω3PN =n 1 + 3 ξ + −7η ξ +
us consider the second point. It is desirable to construct 2 2
GW templates using the mathematical formulation em-
   
817 123
ployed to construct the (heavily employed) PN accurate + 7 η2 + − + π2 η ξ 2 , (4)
4 32
relativistic Damour-Deruelle timing formula for binary
pulsars [3]. This is because formally GW phasing for in- where ξ = G m n/c3 . Let us now compute employing
spiralling compact binaries and timing of relativistic bi- PN accurate expressions for E(x) and L(x), available in
nary pulsars are quite similar. Our construction of these Refs. [1], the following 3PN accurate expression for the
new Taylor approximants are indeed influenced by the orbital energy E and 3.5PN accurate GW energy lumi-
GW phasing formalism, available in Ref. [4], that pro- nosity L, in terms of ξ, as
vided a method to construct GW templates for compact    
binaries of arbitrary mass ratio moving in inspiralling ec- 2/3 5 η 2/3 45 21
Ẽ(ξ) = ξ 1+ − ξ + − η
centric orbits. We recall that the techniques adapted in 4 12 8 8
Ref. [4] were influenced by the mathematical formulation,
 
1 2 4/3 7975 35 3 1031 2
developed in Ref. [5], to compute the Damour-Deruelle − η ξ + − η + η
24 192 5184 288
timing formula. Finally, a recent preliminary investi- 
30403 41 2
  
gation indicates that our new Taylor approximants, at + − + π η ξ2 , (5a)
576 96
the dominant radiation reaction order, should be very   
efficient in capturing GWs from compact binaries inspi- 32 2 10/3 35 2113 2/3
L(ξ) = η ξ 1+ − η+ ξ
ralling along PN accurate and mildly eccentric orbits [6]. 5 12 336
This is, in our opinion, a very attractive feature for GW
 
458461 20129 65 2 4/3
data analysts as GWs from inspiralling (astrophysical) +4πξ + − η+ η ξ
9072 504 18
compact binaries should have some tiny eccentricities, 
583 26753
 
16
when their GWs enter the bandwidth of laser interfer- + − η+ π ξ 5/3 +
24 672 3
ometers. 
41 13106635373 6881951
Let us describe how we construct these new types + η π2 + − η
of PN accurate time-domain Taylor approximant GW 3 23284800 7776
 
search templates. GWs from inspiralling (astrophysi- 375997 2 775 3 1712
+ η − η − γ + log(4 ξ 1/3 ) ξ 2
cal) compact binaries will have some tiny eccentricities 3024 324 105
around orbital frequencies of 20Hz. For example, using 
771833 624559 193385 2
 
Ref. [4], it is not that difficult to show that the orbital + − η+ η π ξ 7/3 ,
2016 1728 3024
eccentricity of the Hulse-Taylor binary pulsar when its (5b)
orbital frequency reaches around 20 Hz will be ∼ 10−6 .
Therefore, let us take a closer look at how one can de- where Ẽ = −2 E, E being the dimensionless non-
scribe, in a PN accurate manner, eccentric orbits, mo- relativistic energy per unit reduced mass [4] and γ be-
tivated by the fact that GW phasing requires accurate ing the Euler’s gamma. We are now in a position to
orbital description. Couple of decades ago, it was demon- construct, in our terminology, TaylorK1 3.5PN and Tay-
strated that associated with a PN accurate non-circular lorK2 3.5PN restricted PN waveforms. In our approach,
orbit, there exists two gauge invariant quantities, if ex- the form of the restricted PN waveform, Eq. (1), becomes
pressed in terms the conserved orbital energy and an-  2/3
gular momentum of the binary [7]. These are the PN h(t) ∝ G mc3n(t) cos 2 φ(t) , This is allowed because
accurate mean motion n and k that measures the ad- at Newtonian order ω = n and the amplitude is indeed
vance of periastron in the time interval T , T being the Newtonian accurate in Eq. (1). For TaylorK1 3.5PN ac-
radial orbital period such that n = 2 π/T . It is quite curate waveform, n(t) and φ(t) are numerically obtained
 3

using the following two differential equations NGW with what is expected from TaylorT1 and Tay-
lorT2 waveforms at these four different PN orders . The
dφ numbers, relevant for initial LIGO, are listed in Table I
= ω3PN , (6a)
dt  where we compare NGW resulting from TaylorK2 and
dn dE TaylorT2 prescriptions [results are similar while compar-
= −L(ξ) (6b)
dt dn ing TaylorK1 and TaylorT1]

To construct our TaylorK2 3.5PN waveforms, as ex-

pected, we Taylor expand, in terms of ξ, the RHS of
Eq. (6b) and this leads to TABLE I: Accumulated number of GW cycles, relevant for
initial LIGO, for three types of canonical binaries at four dif-
dφ ferent PN orders using TaylorK2 and TaylorT2 waveforms.
= ω3PN , (7a)
dt The values of NGW arising from TaylorT2 waveforms are
given in parentheses. We note that TaylorK2 waveforms pro-
  
dn 96 1273 11
= η n2 ξ 5/3 1 + − η ξ 2/3 + 4 π ξ vide more NGW compared to TaylorT2 waveforms. For high
dt 5 336 4 mass binaries, the convergence of NGW is not that pronounced
  
438887 59 2 49507 4/3 20033 for TaylorK2 waveforms compared to TaylorT2 waveforms.
+ + η − η ξ +
18144 18 2016 672 m1 /M⊙ : m2 /M⊙ 1.4 : 1.4 1.4 : 10 10 : 10
189
 
16 287
 2PN 1616.4 (1613.5) 345.6 (333.8) 57 (52.6)
− η π ξ 5/3 + + η π2 2.5PN 1613.5 (1605.8) 333.8 (333.1) 53.8 ( 52.6)
8 3 24 3PN 1623.4 (1616) 347.3 ( 330.9) 57.6 (52.9)
5605 3 617285 2 16554367 3.5PN 1620.6 (1615.4) 342.4 (330.5) 56.2 (52.5)
− η + η − η
2592 8064 31104
38047038863 1712  
1/3
+ − γ + log 4 ξ ξ2
139708800 105 We are aware that LAL also provides routines to create
  
91495 2 1608185 971011 7/3 TaylorT3 waveforms. In this prescription, both φ(t) and
+ η − η+ πξ ,
1512 6048 4032 ω(t), appearing in Eq. (1), are given as explicit PN accu-
(7b) rate functions of time. These explicit time dependencies
are usually expressed in terms of the so-called ‘adimen-
Let us now specify, for example, the limits of integration 3
sional’ time variable θ = 5cG ηm (tc − t), where tc is the PN
for n to construct TaylorK1 3.5PN and TaylorK2 3.5PN accurate coalescence time. It is indeed possible for us to
waveforms. For initial LIGO, it is customary to use ωi compute n(t), using Eqs. (7), as a PN series in terms of
and ωf , the initial and final final values of ω, to be 40 π θ. However, we are reluctant to repeat what is done in
Hz and (6(3/2) m)−1 Hz, where ωf is twice the conven- TaylorT3 waveforms to get φ(t) with the help of Eqs. (7).
tional orbital angular frequency of the innermost stable Observe that radiation reaction and hence temporal evo-
circular orbit for a test particle around a Schwarzschild lution of n first appears at 2.5PN order and therefore, in
black hole. With these inputs, the initial and final values our opinion, it is better to keep dφ/dt to at least 2PN or-
of n, denoted by ni and nf , are numerically computed us- der in Eqs. (7) to be consistent in a PN way[see Refs. [4, 6]
ing Eq. (4). This is justified because of the observation where similar approaches are employed].
in Ref. [10] that the quadrupolar GW frequency from a
compact binary, having PN accurate orbital motion, ap- It is important to note, while constructing these time-
pears at (1 + k) n/π. At 3PN order, for a compact binary domain Taylor waveforms, that we employed the follow-
having m = 11.4M⊙ and η ∼ 0.108 we have ni ∼ 111.32 ing two arguments. The first one is the standard argu-
Hz and nf ∼ 679.3Hz. In our approaches to construct, ment that equates the rate of decrease of the conserved
for example, TaylorK1 2PN and TaylorK2 2PN wave- orbital energy of a compact binary to the opposite of GW
forms, we use only the 2PN accurate relation connecting luminosity. However, for constructing TaylorT1, Tay-
ω and n. lorT2, TaylorK1 and TaylorK2 waveforms, one requires
Let us now compute in the time domain the accumu- additional PN accurate relations relating ω (or n as the
lated number of GW cycles, NGW , in a given GW fre- case may be) to the conserved orbital energy. Further,
quency window, by numerically integrating Eqs. (6) and we speculate that the two different ways of computing
(7) representing temporal evolutions for TaylorK1 and dω/dt, enforced in TaylorT1 and TaylorT2 waveforms,
TaylorK2 waveforms at four different PN orders, namely may be based on the fact that observationally dω/dt (or
2PN, 2.5PN, 3PN and 3.5PN orders, for three canonical the above mentioned standard argument) is only tested
compact binaries usually considered in the GW litera- to the Newtonian radiation reaction order by the accurate
ture [we restrict these orbital evolutions such that emit- timing of binary pulsars. Therefore, it is natural to ask if
ted GWs are in the GW frequency window defined by we can construct h(t) employing only the energy balance
40 Hz and (63/2 π m)−1 Hz]. Let us also compare these argument. This is indeed possible as demonstrated below
 4

cumulated GW cycles in a given GW frequency window

and may be useful in detecting GWs from inspiraling
h(t̂) ∝ Ẽ(t̂) cos 2 φ(t̂) , (8a) compact binaries that should have ‘teeny-weeny’ orbital
  
dφ 1 891 201 11 2 2 eccentricities. Further, our approaches are influenced by
= ζ 3/2 1 + [9 + η] ζ + − η+ η ζ
dt̂ 8 128 64 128 the way PN accurate Damour-Deruelle timing formula

41445

309715 205 2

1215 2 was constructed. Therefore, we feel that our TaylorK1,
+ + − + π η+ η TaylorK2 and TaylorEt waveforms should be of certain
1024 3072 64 1024
  interest to the practitioners of LAL. Further, we feel that
45 3 3
+ η ζ , (8b) our restricted PN waveforms should be useful for the the
1024 recently initiated mock LISA data challenge task force.
  
dζ 64 5 13 5 The data analysis implications of these templates, rele-
= ηζ 1+ − η ζ + 4 π ζ 3/2
dt̂ 5 336 2 vant for both ground and space based GW detectors, are
under active investigations in collaborations with Stas
  
117857 12017 5 4913
+ − η + η2 ζ 2 + Babak, Sukanta Bose, Christian Röver and Manuel Tess-
18144 2016 2 672
   mer. The GW phase evolution under our prescription is
177 369 16
− η π ζ 5/2 + η+ π2 also being compared with its counterpart in numerical
8 32 3
relativity based binary black inspiral.
37999588601 24861497 488849 2
+ − η+ η I am indebted to Gerhard Schäfer for illuminating dis-
279417600  72576 16128
 cussions and persistent encouragements. Lively discus-
85 3 1712  p 
sions with Manuel Tessmer are warmly acknowledged.
− η − ln 4 ζ + γ ζ 3
64 105 This work is supported in part by the DFG (Deutsche
  
613373 2 129817 3207739 Forschungsgemeinschaft) through SFB/TR7 “Gravita-
+ η + − η π ζ 7/2 ,
12096 2304 48384 tionswellenastronomie” and the DLR (Deutsches Zen-
(8c) trum für Luft- und Raumfahrt) through “LISA Ger-
many”.
where t̂ = t c3 /G m and ζ = Ẽ. We call the resulting h(t̂)
as TaylorEt waveforms. The values of ζ corresponding
to ωi and ωf can numerically evaluated using the RHS
of Eq. (8b) for dφ/dt̂ = ω̂.
∗
We evaluated NGW associated with TaylorEt 3.5PN Electronic address: [email protected]
waveforms for the three canonical compact binaries and [1] T. Damour, P. Jaranowski, and G. Schäfer, Phys. Lett.
B 513, 147 (2001); L. Blanchet, G. Faye, B. R. Iyer,
the numbers are the following. For neutron star binaries, and B. Joguet, Phys. Rev. D 65, 061501(R) (2002); 71,
m = 2.8M⊙ and η = 0.25, NGW = 1617.4 and for the 129903(E) (2005); L. Blanchet, T. Damour, G. Esposito-
usual black hole-neutron star binaries, m = 11.4M⊙ and Farèse, and B. R. Iyer, Phys. Rev. Lett. 93, 091101
η = 0.108, we have NGW = 335.4. For typical stellar (2004); K. G. Arun, L. Blanchet, B. R. Iyer, and
mass black hole binaries, m = 20M⊙ and η = 0.25, one M. S. S. Qusailah, Class. Quant. Grav. 21, 3771 (2004);
gets NGW = 54.0. It is interesting to note that we get 22, 3115(E) (2005) and references therein.
[2] Lsc Algorithm Library (LAL),
larger NGW , compared to TaylorT 3.5PN waveforms and
http://www.lsc-group.phys.uwm.edu/lal.
lower values compared to TaylorK 3.5PN waveforms. It [3] T. Damour and N. Deruelle, Ann. Inst. Henri Poincare
should be related to the fact that it takes more time Phys. Theor. 43, 107 (1985); T. Damour and J. Taylor,
for TaylorEt prescription to reach ωf form ωi compared Phys. Rev. D 45, 1840 (1992).
to TaylorT1 (or TaylorT2) approach and the opposite [4] T. Damour, A. Gopakumar, and B. R. Iyer, Phys. Rev.
is true for the cases of TaylorK1 (or TaylorK2). The D 70, 064028 (2004).
observation that TaylorEt waveforms also provide more [5] T. Damour, in Proceedings of Journées Relativistes 1983,
number of GW cycles in a give GW frequency window, in edited by S. Benenti, M. Ferraris, and M. Francaviglia
(Pitagora Editrice, Bologna, 1985), pp. 89–110.
our opinion, makes it our third prescription to compute [6] M. Tessmer and A. Gopakumar, submitted to Phys. Rev.
h(t). D (2007), arXiv:0712.3199[gr-qc].
Conclusions.— We provided new ways of construct- [7] T. Damour and G. Schäfer, Nuovo Cimento Soc. Ital.
ing restricted time-domain PN accurate waveforms for Fis., B 101, 127 (1988).
non-spinning compact binaries inspiralling along PN ac- [8] T. Damour and N. Deruelle, Ann. Inst. Henri Poincaré
curate quasi-circular orbits. Our prescriptions employed Phys. Théor. 44, 263 (1986); T. Damour and G. Schäfer,
Nuovo Cimento Soc. Ital. Fis., B 101, 127 (1988);
PN accurate expressions for the conserved orbital energy
G. Schäfer and N. Wex, Phys. Lett. 174 A, 196, (1993);
and GW luminosity, available in Refs. [1], in a democratic erratum 177, 461; R.-M. Memmesheimer, A. Gopaku-
manner and heavily depended on certain PN accurate mar, and G. Schäfer, Phys. Rev. D 70, 104011 (2004).
gauge invariant quantities, first introduced in Ref. [7]. [9] C. Königsdörffer and A. Gopakumar, Phys. Rev. D 73,
These template waveforms provide more number of ac- 124012 (2006).
 5

[10] M. Tessmer and A. Gopakumar, Mon. Not. Roy. Astron.

Soc. 374, 721 (2007) [arXiv:gr-qc/0610139].