PHYSICAL REVIEW D 76, 062002 (2007)

Local readout enhancement for detuned signal-recycling interferometers

Henning Rehbein,1 Helge Müller-Ebhardt,1 Kentaro Somiya,2 Chao Li,3 Roman Schnabel,1
Karsten Danzmann,1 and Yanbei Chen2
1
Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut), Institut für Gravitationsphysik, Leibniz Universität Hannover,
Callinstr. 38, 30167 Hannover, Germany
2
Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut), Am Mühlenberg 1, 14476 Potsdam, Germany
3
California Institute of Technology, M/C 130-33, Pasadena, California 91125, USA
(Received 21 May 2007; published 21 September 2007)
High power detuned signal-recycling interferometers currently planned for second-generation inter-
ferometric gravitational-wave detectors (for example Advanced LIGO) are characterized by two reso-
nances in the detection band, an optical resonance and an optomechanical resonance which is upshifted
from the suspension pendulum frequency due to the so-called optical-spring effect. The detector’s
sensitivity is enhanced around these two resonances. However, at frequencies below the optomechanical
resonance frequency, the sensitivity of such interferometers is significantly lower than non-optical-spring
configurations with comparable circulating power; such a drawback can also compromise high-frequency
sensitivity, when an optimization is performed on the overall sensitivity of the interferometer to a class of
sources. In this paper, we clarify the reason for such a low sensitivity, and propose a way to fix this
problem. Motivated by the optical-bar scheme of Braginsky, Gorodetsky, and Khalili, we propose to add a
local readout scheme which measures the motion of the arm-cavity front mirror, which at low frequencies
moves together with the arm-cavity end mirror, under the influence of gravitational waves. This scheme
improves the low-frequency quantum-noise-limited sensitivity of optical-spring interferometers signifi-
cantly and can be considered as an incorporation of the optical-bar scheme into currently planned second-
generation interferometers. On the other hand it can be regarded as an extension of the optical-bar scheme.
Taking compact binary inspiral signals as an example, we illustrate how this scheme can be used to
improve the sensitivity of the planned Advanced LIGO interferometer, in various scenarios, using a
realistic classical-noise budget. We also discuss how this scheme can be implemented in Advanced LIGO
with relative ease.

DOI: 10.1103/PhysRevD.76.062002 PACS numbers: 04.80.Nn, 03.65.Ta, 42.50.Dv, 42.50.Lc

configurations, the optical resonance of the interferometer

I. INTRODUCTION
is away from the carrier frequency, creating a peak sensi-
First-generation laser interferometric gravitational-wave tivity to GWs away from DC.
(GW) detectors (LIGO [1], VIRGO [2], GEO [3], and As demonstrated theoretically by Buonanno and Chen
TAMA [4]) are reaching design sensitivities. These inter- [7–9] and experimentally by Somiya et al. [10] and
ferometers are usually Michelson interferometers with Miyakawa et al. [11], detuned signal-recycling also makes
Fabry-Perot cavities in the arms, with power-recycling the power inside the interferometer depend on the motion
(PR) at the laser input port (with the exception of GEO, of the mirrors, creating an optical spring, and can shift the
which uses dual-recycling [5]), and operating close to the eigenfrequency of the test masses from the pendulum
dark-port condition. frequency (
1 Hz) up to the detection band. The optical
In order to have a flexible sensitivity to specific astro- spring helps to improve the interferometer’s response to
physical sources, and for other technical reasons such as GWs around the optomechanical resonant frequency, even
lowering power at the beam splitter (BS), second- allowing the interferometer to surpass the free-mass stan-
generation interferometers, such as Advanced LIGO [6], dard quantum limit (SQL). However, the quantum-noise-
plan to use the so-called signal-recycling (SR) configura- limited sensitivity of optical-spring interferometers at fre-
tion, in which an additional mirror is placed at the dark port quencies below the optomechanical resonant frequency is
of a Fabry-Perot Michelson interferometer, modifying the dramatically lower than the one of non-optical-spring in-
optical resonant structure of the interferometer. The adjust- terferometers. Such a limitation in sensitivity is caused by
ment of the location and reflectivity of the signal-recycling the optical spring, which rigidly connects the front and the
mirror varies the optical resonance frequency and band- end mirror of the arm cavities at frequencies below the
width, respectively. Near the optical resonance, sensitivity optomechanical resonance. The general principle under-
to GWs is improved. When the signal-recycling cavity, the lying this effect has already been explained in the works of
cavity formed by the input test-mass mirrors and the Braginsky, Gorodetsky, and Khalili, namely, in their pro-
signal-recycling mirror is neither resonant nor antiresonant posal of the optical-bar detection scheme [12]. In order to
with respect to the carrier frequency, the optical configu- understand this more conveniently, we need to use the local
ration is called detuned signal-recycling. In these detuned inertial frame of the BS, in which the effect of GWs can be

1550-7998= 2007=76(6)=062002(10) 062002-1 © 2007 The American Physical Society

HENNING REHBEIN et al. PHYSICAL REVIEW D 76, 062002 (2007)
described completely as a tidal force field, which induces ETM
forces only on the end test-mass mirrors (ETMs), but not
on the input test-mass mirrors (ITMs). We make the ap-
proximation that the ITMs and the BS are colocated. In this
frame, the propagation of the light is unaffected by GWs.
Remember that the optical spring connects the ITM and the
ETM. At frequencies substantially below the optomechan-
ical resonance, the optical spring behaves like a rigid ITM
optical bar, connecting the ITM and the ETM of each P RM
arm rigidly. It is then easy to understand that the carrier BS
laser 1
light, which senses the change in arm-cavity length, or the
difference in ITM and ETM motion, cannot be used to laser 2
measure GWs efficiently at these frequencies. On the other
hand, since the ITM and the ETM are rigidly connected, ITM ETM
they both move, in the local inertial frame of the BS, by
1=2 the amount the ETM would have moved if there were S RM

no optical spring present (assuming ITM and ETM to have

equal masses). To illustrate this situation, assume that a
low-frequency GW with amplitude h is incident from right FIG. 1 (color online). Schematic plot of a power- and signal-
above our detector (with arm length L), then in the local recycled Michelson interferometer with arm cavities and double-
inertial frame of the BS, the motion of the ETM of a non- readout. The added local readout sensing the ITM is realized by
optical-spring interferometer would be Lh, the motion of a secondary laser which does not resonate in the arm cavities.
ITM and ETM of an optical-spring interferometer below
resonance will both be
Lh=2. For this reason, if one also
detect more efficiently the population of compact binary
measures the local motion of the ITM using an additional
objects with a broad range of masses (and hence signal
local readout scheme, one can recover low-frequency sen-
frequency band).
sitivity dramatically. Note that as viewed by the local
In order to construct the local meter, we consider a
meter, the ITM has an effective mass that is equal to the
scheme where a second carrier is injected into the bright
total mass of the ITM and the ETM. If one applies a local
port, which does not enter the arm cavities, but instead
readout scheme to the ETM, the same sensitivity recovery
senses the location of the ITMs, as shown in Fig. 1. An
is possible, since the ETM also moves with respect to a free alternative strategy would be attaching auxiliary interfer-
colocated mirror by Lh=2. Braginsky, Gorodetsky, and ometers at the ETMs. These two strategies are quite
Khalili proposed an optical-bar detection scheme, in which equivalent in the ideal situation, but differ from each other
only the local motion of the ITM is measured [12]. In this in terms of difficulty in implementation, in terms of quan-
sense, what we are proposing can be considered as directly tum noise, and in terms of technical noise sources such as
incorporating the optical-bar scheme into currently laser noise as we will discuss in some more detail.
planned second-generation interferometers. This paper is organized as follows. In Sec. II, we study
Local readout schemes have also been proposed for the dynamics, sensing, and control of our double-readout
interferometers without optical spring, with a different interferometer. In Sec. II A, we write down and solve the
motivation. In those interferometers, the motion of mirrors joint Heisenberg equations of motion of test masses, beam
with respect to their local inertial frames are caused by splitter, and optical fields; in Sec. II B, we evaluate the
radiation-pressure noise (if we only consider signal and optimal combined GW sensitivity of the two readout chan-
quantum noise sources); results of local readout schemes nels; in Sec. II C, we prove that the use of control schemes
can thus be used to cancel radiation-pressure noise and does not affect this sensitivity. In Sec. IV we show the
improve low-frequency sensitivity [13,14]. Furthermore, benefit which the local readout scheme will provide for the
such schemes are able to cancel parts of the classical noise. detection of intermediate-mass black holes, using a realis-
Our treatment here can also be viewed as a generalization tic Advanced LIGO noise budget. In Sec. IV we consider
of these schemes because by setting detuning in our treat- practical issues for a possible implementation in Advanced
ment to zero, we will recover their results. LIGO. In Sec. V we summarize our main conclusions.
From an astrophysical point of view, the addition of the
local readout scheme, which broadens the detection band, II. DYNAMICS, SENSING AND CONTROL
will allow the interferometer to search for multiple sources
simultaneously, as well as to examine a wider frequency A. Equations of motion
range of the same source. As an example, we will explore Let us consider a configuration where the ITM motion of
how the increase in detection bandwidth can allow us to a signal- and power-recycled Michelson interferometer

062002-2
LOCAL READOUT ENHANCEMENT FOR DETUNED SIGNAL- . . . PHYSICAL REVIEW D 76, 062002 (2007)
with arm cavities is locally sensed by a small interferome- (the small interferometer, formed by the BS and the ITMs),
ter which has the ITMs as its end mirrors (cf. Fig. 1). This respectively. The operators F^ i and F^ i BP describe the
is realized by injecting a second carrier into the bright port, radiation-pressure forces which would act on fixed mirrors
which does not resonate in the arms (preferably antireso- caused by the incoming vacuum fields at the dark port and
nant). Because the frequency (and the polarization) of the the laser light fluctuations from the bright port, respec-
second carrier is (are) different from that of the first, we tively. The operators Y^ ji account for the shot noise in the
effectively obtain a two interferometers in one scheme
case of fixed mirrors. Each optical component is subject to
where parameters such as detuning and mirror reflectivities
for each interferometer can be chosen independently; input classical noise generated by the corresponding operator
^
vacuum fluctuations associated with the two lasers are also and has its own mechanical susceptibility Rxx . The suscep-
i
independent. tibilities RFF describe the optical springs [7] and RYii F the
Throughout this paper, we will assume the GW with transformation of the mirror motion into the two outputs.
amplitude h as incident from right above the interferome- In the following we will present all these quantities in more
ter, with a polarization that maximizes the response of our detail while all appearing parameters are summarized in
L-shaped Michelson interferometers. In the following we Table I.
will list the Heisenberg equations of motion in the fre- The free radiation-pressure force and the free shot noise
quency domain [7–9,15,16] for the differential mode of in each of the two interferometers are given by [9]
motion (i.e., opposite in the two arms) of the input mirrors s




















x^ ITM and the end mirrors x^ ETM , respectively, as well as for i i m@ i
 i a^ i i ^ i
1  a
F^ i  2
;
the BS motion normal to its reflective surface x^ BS and for 2 
 i  ii 
 i  ii 
the two measurement outputs y^ i :
i 2  i 2 
2 a^ i i i ^ i
1  2  a
Y^ 1i  2
;
x^ ITM  Rxx 
F 
 ^ 1 R1
FF 
x x^ ITM 
^ ETM 
 i  ii 
 i  ii 
p



 F^ 2 
  R2 ^ ITM  2x^ BS  
^ ITM ;
FF 
x
2i i a^ i i 2 i 2
1       
a
2 ^ i
Y^ 2i  2
;
(1) 
 i  ii 
 i  ii 
8Pi !i
where i  mLi c0 has units of frequency cubed. Note that
^ 1
x^ ETM  Rxx 
F 
  R1
FF 
x
^ ETM  x^ ITM 
Pi refers to the circulating power in each arm, respec-
 Lh 
^ ETM ; (2) tively. Here a^ i ^ i
1 and a 2 are the amplitude and phase
p


quadrature operators of the incoming vacuum fields at
^ 2 2
x^ BS  RBS
xx 
F 
  RFF 
x
^ ITM  2x^ BS  the dark port [15], associated with the first and second

 F^ 1 ^ 2 ^
BP 
  FBP 
 
BS ; (3)
TABLE I. Technical data and parameter values for large-scale
interferometer and local meter used throughout the calculations.
y^ 1  Y^ 1
1 
 sin
1
 Y^ 1
2 
 cos
1 Note that we have defined SR with respect to the first carrier
while the local meter requires a different reflectivity in order to
 R1
Y1 F 
 sin
1  R1 
 cos 1 
Y2 F achieve the bandwidth 2 .
x^ ETM  x^ ITM ; (4) Symbol physical meaning value
m single mirror mass 40 kg
y^ 2  Y^ 2
1 
 sin
2
 Y^ 2
2 
 cos
2
mBS beam splitter mass 40 kg
c=!1 laser wavelength of 1st carrier 1064 nm
 R2Y1 F 
 sin
2
 R2 2
Y2 F 
 cos  P1
0
circulating power of 1st carrier 0:1 . . . 0:8 MW
p



x^ ITM  2x^ BS : (5) L1 large-scale interferometer arm length 4









p km
PR power-recycling mirror reflectivity 0:94
Note that x^ ITM and x^ ETM account for the differential motion  detuning phase for 1st carrier 0p.









..
between two mirrors while x^ BS describes the motion of a SR signal-recycling mirror reflectivity 0:93
single mirror o cavity half bandwidth for 1st carrier 215 Hz
p


with an angle of 45 deg. This explains the  1 detection angle for 1st carrier 0...
factor of 2 in front of the BS motion. The out-going fields
at the dark port belonging to the two different carriers are c=!20 laser wavelength of 2nd carrier 1064 nm
each sensed by homodyne detection such that the measure- P2 circulating power of 2nd carrier 0 . . . 16 kW
L2 local meter arm length 15 m
ment outputs are a certain combination of amplitude and
2 detuning for 2nd carrier 0 Hz
phase quadratures (described by the phases  1;2 ). Note 2 cavity half bandwidth for 2nd carrier 24 kHz
that we have labeled all quantities with superscripts (1) and  2 detection angle for 2nd carrier 0
(2) for the large-scale interferometer and the local meter

062002-3
HENNING REHBEIN et al. PHYSICAL REVIEW D 76, 062002 (2007)
carrier field, respectively. The susceptibilities are given by The operators
^ ITM ,
^ ETM , and
^ BS model the classical
[9] noise at ITM, ETM, and BS, respectively. We assume that
p


they are uncorrelated but all have the same spectrum,
BS 2 namely, one fourth of the classical-noise spectrum gener-
Rxx   ;
mBS
2 ally expected for the differential mode of motion. By using
2 the following only nonvanishing correlation functions,
Rxx   ;
m
2
i m i ha^ i ^ j
k 
a
y 0 0
l  
isym  



ij
kl ;
Ri
FF  ;
4 
 i  ii 
 i  ii  hb^i ^i y 0 0 i
k 
bl  
isym  



kl Sl 
;
s

















i i m i h
^ ITM 

^ ITM y 
0 isym  2


0 Scl 
;
i
RY1 F  ; (6)
2@ 
 i  ii 
 i  ii  h
^ ETM 

^ ETM y 
0 isym  2


0 Scl 
;
s

















i i i m i  i
h
^ BS 

^ BS y 
0 isym  


0 Scl 
;
RY2 F   ;
2@ 
 i  ii 
 i  ii 
we obtain the single-sided noise spectral densities. Here
where the (free) optical resonant frequency of the large- Si
l 
 is the spectrum of technical input laser noise while
scale interferometer at
 1  i1 is determined by Scl 
 characterizes the spectrum of all the other classical-
2SR sin2 noise sources. In further calculations we will assume the
1  o ; laser noise to be white and 10 dB above the vacuum noise
1  2SR  2SR cos2 level. For other classical-noise sources, we use the current
1  2SR noise budget of Advanced LIGO, as given in Bench [18];
1  o : contributions such as suspension thermal noise, seismic
1 2SR  2SR cos2
noise, and thermal fluctuations in the coating and gravity
As already mentioned, the second carrier does not reso- gradient noise are presented in Fig. 4.
nate in the arm cavities and therefore the local meter is just Note that we can obtain two input-output relations from
equivalent to an interferometer configuration without cav- the equations of motion in Eq. (1)–(5) and write them in
ities in the arms. Thus, in Eq. (3) we only take into account the following compact form:
the forces on the BS due to field fluctuations around the
second carrier, in the same way as in Ref. [17]: the first two y^ 1  n~ T1 ~  s1 h; y^ 2  n~ T2 ~  s2 h; (7)
terms in the bracket on the right-hand side of Eq. (3) are
due to dark-port fluctuations around the second carrier, where ~ T  a^ 1
1 ;a^ 1
2 ;a^ 2
1 ;a^ 2 ^1 ^2 ^ ^ ^
2 ; b1 ; b1 ;
ITM ;
ETM ;
BS 
while the third and fourth term, given by and T denotes transposed. Here the two vectors n~ 1;2 are
q






































the linear transfer functions from the noise channels ~ into
1
L1 1 m@1  2PR  0 the two output channels, while the two functions s1;2 are
F^ BP  0 p


b1
1 ;
2c 0 1  PR   i1  PR 
 the linear transfer functions from the signal, i.e., the GW
s










































2 mL2 @1  PR  2
strain h, into the output channels.
2
F^ BP  b1 ;
2c1  PR  B. Combined sensitivity
are forces due to bright-port fluctuations, where b1i
are the Now we seek for a linear combination of the two output
amplitude quadrature of fluctuations around the first and channels, y^ 1 and y^ 2 ,
second carrier, at the input port. Forces due to fluctuations
around the first carrier are usually negligible, because the y^  K1 
y^ 1  K2 
y^ 2 ; (8)
intensity of the first carrier at the beam splitter is lower than
that of the second carrier; in addition, fluctuations associ- which has optimal sensitivity to gravitational waves. In this
ated with the first carrier also do not build up as much as optimization, we only consider the signal-referred noise
those associated with the second carrier, both in common spectral density of y,
^
and in differential mode. 

In Eq. (4), we make the approximation that the first K
 K1 K2 N 1

carrier only senses the cavity length, x^ ETM  x^ ITM , ignor- K

Sh 
  
2  (9)
ing the slight difference between its sensitivities to ITM K1
 K1 K2 S

and ETM, as well as motion of the BS. In Eq. (5), the K2

second carrier only senses the ITM and BS motions, since
it does not enter the arm cavities. with

062002-4
LOCAL READOUT ENHANCEMENT FOR DETUNED SIGNAL- . . . PHYSICAL REVIEW D 76, 062002 (2007)

1st carrier output 101 1st carrier output

24 nd
10 2 carrier output 2nd carrier output
Signal transfer function

Noise transfer function

1023

100
1022

100 101 102 103 100 101 102 103

f Hz f Hz

FIG. 2 (color online). Example for the signal (left panel) and noise (right panel) transfer functions in a signal-recycled Michelson
interferometer with two carriers and double-readout, for a configuration with the parameters as given in Table I but  1  0,  
=2  0:014, and P1  800 kW.

2 3
14 M  N1 S; (12)
 T 6 6 S1 7
7
n~ 1 66
l 7
7
 with the corresponding eigenvector providing the optimal
N  S2 n~ 1 n~
2
n~ T2 6
6 l 7
7 filters K1 ; K2 .
4 2Scl 12 5
We now illustrate the local readout scheme using the
Scl
following configuration: the parameters are given in Table I
(10) as well as phase quadrature readout  1  0, signal-
and recycling cavity detuning phase   =2  0:014, and
power P1  800 kW of the first carrier are used. In Fig. 2
 
s s
s1 s
2 we plot the individual signal- and noise-transfer functions
S  1 1
; (11) of the first and second carriers, for the configuration with
s2 s1 s2 s
2
P2  4 kW. As we can see from these plots, the first
where 1k stands for a k-dimensional identity matrix. One carrier mainly senses frequencies above the optical-spring
way of obtaining the minimum noise is to impose the resonance with signal-transfer function suppressed at
constraint that the value of the denominator always re- lower frequencies by the optical spring; the second carrier
mains unity, and minimize the numerator under this con- offers complementary sensitivity for frequencies below the
straint. Note that an overall rescaling of the vector K1 ; K2  optical-spring resonance, when the ITM is dragged to-
does not affect Sh . The resulting minimum noise is one gether with the ETM by the optical spring. As a conse-
over the bigger eigenvalue of the 2-by-2 matrix quence, as we see in the left panel of Fig. 3, at frequencies

21
10
1.0
Hz

2 P2 0 kW
K1
2 SQ P2 1 kW
Noise Spectral Density 1

K2 L
10 22 P2 4 kW
Filter function

P2 16 kW

0.5
23
10

24
10
0.0
100 101 102 103 100 101 102 103
f Hz f Hz

FIG. 3 (color online). Left panel: filter functions K1 and K2 , for the same configuration as in Fig. 2. Here each filter function is
rescaled such that it gives the percentage of how much GW strain it feeds into the combined output. Right panel: quantum noise curves
for our proposed scheme with different powers of the 2nd carrier. Again phase quadrature readout  1  0, signal-recycling cavity
detuning phase   =2  0:014, and power P1  800 kW of the 1st carrier are used.

062002-5
HENNING REHBEIN et al. PHYSICAL REVIEW D 76, 062002 (2007)
above the optical-spring resonance, the optimal combina- signal-referred noise spectrum of K1 ; K2 y,
~ is then clearly
tion depends mostly on the first readout, while at frequen- invariant with respect to changes in D.
cies below the optical-spring resonance, the optimal
combination depends mostly on the second readout.
III. IMPROVEMENTS IN ADVANCED LIGO
Noise curves with optimal filters are plotted for different
SENSITIVITY
powers of the second carrier (0 kW, 1 kW, 4 kW, and
16 kW) in the right panel of Fig. 3 where only quantum A. Matched-filtering signal-to-noise ratio
noise is taken into account. This plot illustrates that the To quantify the astrophysical merit of various configu-
local readout scheme can directly improve the sensitivity rations, we will calculate the improvement in the matched-
only below the optomechanical resonance frequency. It filtering signal-to-noise ratio (SNR) or the detectable dis-
turns out that 4 kW in each arm of the local meter already tance for a given threshold SNR, respectively, for inspiral
gives a remarkable increase in sensitivity. In the following waves from compact binary systems. For a known wave-
studies we fix P2  4 kW. form (in the frequency domain) hf, the optimal SNR
One could imagine that the combination of a signal- achievable by correlating the data with a known template is
recycled Michelson interferometer with a local readout s





























may indirectly help improve the sensitivity at high fre- Z 1 jhfj2
quencies or increase the detection bandwidth, once an 2 df (16)
0 Sh f
overall optimization to a broadband source is performed.
The underlying effect is that the sensitivity of the large- where Sh f is the single-sided noise spectral density. For
scale interferometer can be shifted to higher frequencies by compact binary objects, the lowest post-Newtonian ap-
choosing its detection angle to be closer to the phase proximation gives (see, e.g., [19])
quadrature while the local meter helps to maintain sensi-
G5=6 1=2 M1=3
tivity at low frequencies. This will be studied more care- jhfj  p





2=3 3=2 f7=6 fmax  f (17)
fully in Sec. III. 30 c D
with
C. Control
M1 M2
As it has been shown in Refs. [7–9] the optical spring M  M1  M2  and  ; (18)
M1  M2
introduces an instability, which must be stabilized using a
feedback control system. In single-readout systems, it is where , M, M1 , and M2 are the reduced, total, and single
easy to show that such a control system does not give rise to masses of the binary and D is the distance from the source
any fundamental change in our GW sensitivity [7–9], to the detector. Here the amplitude is the one where the rms
intuitively because signal and noise are fed back with the average over all directions is already taken into account.
same proportion onto the test masses. Our double-readout There is an upper cutoff frequency, fmax , in Eq. (17) be-
system is more complex, but the same intuition still ap- yond which the system undergoes a transition from adia-
plies. If we denote x~  x^ ITM ; x^ ETM ; x^ BS T and y~  batic inspiral into nonadiabatic merger, and Eq. (17) is no
y^ 1 ; y^ 2 T , the Eqs. (1)–(5) can be written schematically longer a valid approximation. This frequency is usually
as taken to be the GW frequency at the last stable circular
orbit given, for a test mass in a Schwarzschild space-time
~
x~  A
x~  B
~  C
h  D
y;
~ (13) with mass M,
 
M
y~  F
x~  G
:
~ (14) fmax  4400 Hz : (19)
M
Here matrix A describes mirror dynamics, matrix B de- A lower cutoff frequency fmin should also be applied to the
scribes how the noise sources in ~ are applied as forces integration in Eq. (16), below which it is no longer possible
onto the mirrors, vector C~ describes how GW signal h to treat the system as stationary. We take fmin  7 Hz.
directly influences the mirrors, F describes how the output Considering binaries of averaged orientation the observ-
channels y~ sense the various motions x,
~ G describes sens- able distance for a given SNR 0 reaches
ing noise in y, ~ and finally D describes the feedback. s





v
u































Solving Eqs. (13) and (14) jointly, we obtain 2 G5=6 1=2 M1=3 uZ fmax f7=3
D t df : (20)
~
y~  12  HD1 HB  G~  HCh; (15) 15 2=3 c3=2 0 fmin Sh f

where we have defined H  F12  A1 . In Eq. (15) the In this paper, we assume event rate to be proportional to the
only dependence of y~ on the control system is through D, cube of detectable distance, i.e.,
which only appears in the first factor on the right-hand side. R / D3 : (21)
The optimal sensitivity, which is obtained by maximizing

062002-6
LOCAL READOUT ENHANCEMENT FOR DETUNED SIGNAL- . . . PHYSICAL REVIEW D 76, 062002 (2007)
B. Improvement in the event rate ture). Such a prescription is justified, because a local meter
The tools reviewed in the previous subsection enable us with such a short arm length, low power, and finesse (as we
to optimize a specific interferometer configuration for have chosen) is mostly dominated simply by shot noise.
given binary inspirals by maximizing its SNR with respect If we compare the two schemes with and without an
to certain interferometer parameters. Note that we now also added local meter at the binary mass they are optimized for
take classical noise into account as is indicated by the gray we find moderate improvement in event rates (cf. last
lines in Fig. 4. In this paper we assume that Advanced column in Table II). The improvement increases for higher
LIGO refers to a signal-recycled interferometer without binary masses since our scheme helps to enhance sensitiv-
local readout and optimized for neutron-star–neutron-star ity mainly at low frequencies. Such a moderate improve-
(NS-NS) binary systems, i.e., binary systems with M  ment has been limited mainly due to low-frequency
1:4  1:4M . We then vary the optical power P1 , detun- classical noise.
The advantage of the local readout scheme can be ap-
ing 1 , and detection angle  1 in such a way that the
preciated better when we realize that there are different
SNR of the signal-recycled interferometer without local
populations of likely sources (e.g., binary total mass M can
readout is maximized for the total mass of a given binary
reside in a range, M), whose signals extend to different
system in Table II. When we optimize our scheme, we
frequency bands. We need to investigate how well a con-
maximize the SNR varying the same set of parameters of
figuration optimized for a particular system with total mass
the large-scale interferometer (cf. Table II) but with im-
M would perform for other possible masses in M. In this
posing a fixed power for the second carrier (P2  4 kW), paper, we consider M  M ; 630M  with maximum
requiring the second carrier to be resonant in the signal- mass determined by the condition fmax  fmin . In Fig. 5,
recycling cavity (2  0), and fixing a detection quad- we show the improvements in event rates (with respect to
rature phase of  2  0 (i.e., detecting the phase quadra- Advanced LIGO baseline, optimized for NS-NS binaries)

21 21
10 10
Hz

Hz

w local meter w local meter

w o local meter w o local meter

su

su
Noise Spectral Density 1

Noise Spectral Density 1

sp

sp
en

en

22 22
10 10
sio

sio
n

SQ SQ
L L

coati coati
23 ng 23 ng
10 10
grad

grad
seism

seism
ient

ient
ic

ic

24 24
10 10
101 102 103 101 102 103
f Hz f Hz
21
10
Hz

w local meter

w o local meter
su
Noise Spectral Density 1

sp
en

22
10
sio
n

SQ
L

coati
23 ng
10
grad
seism

ient
ic

24
10
101 102 103
f Hz

FIG. 4 (color online). Noise curves for the scheme with local readout (power of the 2nd carrier fixed to P2  4 kW) and without
local readout both optimized for binary systems with total mass M  2:8M (upper left), M  40M (upper right), and M  120M
(lower). Special parameters used for optimizations are given in Table II and all others in Table I. Here classical noise ( gray lines) is
included. Single contributions of the classical- noise are labeled according to their appearance: suspension thermal noise results from
the fluctuations in the suspension system; seismic noise is due to motion of the ground; thermal fluctuations in the coating dominates
the one in the substrate; gravity gradient noise accounts for time-changing Newtonian gravitational forces.

062002-7
HENNING REHBEIN et al. PHYSICAL REVIEW D 76, 062002 (2007)
TABLE II. Parameters used when optimizing our proposed double-readout scheme and the usual Advanced LIGO–like configu-
ration each for different binary systems. The last column gives the improvement in the event rate for our proposed scheme compared to
the usual scheme, both optimized for the given equally distributed total binary mass. Reasonable errors in  and  1 may decrease the
event rate —but not more than 1%.

M=M optimization parameters w/ local meter optimization parameters w/o local meter improvement in event rate
1 1 1 1
P in kW  in radian  in radian P in kW  in radian  in radian
2.8 800 0:48 0:7 800 0:48 0:49 29%
20 450 0:47 0:58 500 0:48 0:48 28%
30 250 0:46 0:46 200 0:46 0:49 30%
40 150 0:45 0:43 150 0:45 0:46 33%
80 100 0:45 0:38 100 0:45 0:46 44%
120 100 0:46 0:32 100 0:47 0:41 42%
160 110 0:47 0:25 100 0:47 0:30 45%
200 110 0:48 0:25 100 0:48 0:27 48%

obtainable by Advanced LIGO configurations (solid lines) cies. Although Fig. 5 (solid curve) does not show a
and double-readout configurations (dashed lines) for bi- significant increase in binary event rates, this configuration
naries with M 2 M, when the configurations are opti- is potentially interesting for detecting other sources above
mized specifically for M  2:8M (black), 40M (dark 300 Hz, for example, pulsars, and low-mass x-ray binaries.
gray) and 120M (light gray). In Fig. 4, we show the Detector for intermediate-mass black-hole binaries. The
corresponding noise spectral densities of these configura- double-readout configuration optimized for 40M systems
tions, together with classical noise. Figures 4 and 5 pro- (dark gray curve in Fig. 5) has the same sensitivity to
vides us with at least two possible applications of the low-mass binary systems as Advanced LIGO baseline
double-readout scheme. (up to M  10M ), while improving event rates for
Detector with broader frequency band. The sensitivity 60M –300M by factors of 2–4:5. This allows us to build
of the double-readout configuration optimized for 2:8M a detector sensitive to the more speculative (yet in some
systems (solid curve on the upper left panel of Fig. 4) is sense astrophysically more interesting) intermediate-mass
broader in band and globally better than the baseline design black-hole binaries, without sacrificing sensitivity at low-
of Advanced LIGO (dashed curve in the same figure), mass systems which are more certain to exist. As we see
particularly at higher frequencies; this demonstrates that from dashed curves in Fig. 5, such broad improvement
when an overall optimization is performed, the local read- simultaneously for systems with different total masses is
out can indirectly improve sensitivity at higher frequen- not achievable by single-readout Advanced LIGO–like
configurations. It is also interesting to note that this con-
figuration only requires a circulating power of 150 kW in
10 the arms.
8 Optimized for: The improvement in event rate increases significantly
Improvement in Event Rate

6 M 2.8 Msolar for higher binary masses (cf. gray curve in Fig. 5 optimized
M 40 Msolar for M  120M ) since the local meter helps to enhance
4 M 120 Msolar sensitivity mainly at low frequencies. But if we optimize
for such high masses the sensitivity for lower masses
2 cannot keep up with Advanced LIGO.
It turns out that our scheme even improves sensitivity in
the low-frequency regime when sensitivity is dominated by
1
0.8
classical noise, as can be seen in Fig. 5, since for high
0.6
binary masses the dashed curves meet at a factor of
4=33=2 below the solid curves. We explain this factor in
1 2 3 45 10 20 40 100 200 400 the appendix.
M Msolar

FIG. 5. Improvement in the event rate compared to Advanced IV. IMPLEMENTATION ISSUES
LIGO versus total binary mass with fixed optimization parame-
ters for each curve. Signal-recycled interferometer with (solid In this section we discuss the possibility of implement-
lines) and without (dashed lines) local readout are optimized for ing this technique explicitly in the Advanced LIGO detec-
three different binary masses. Power of the 2nd carrier is fixed to tor. In fact, the so-called central Michelson degree of
P2  4 kW. freedom in the detector, already to be measured to keep

062002-8
LOCAL READOUT ENHANCEMENT FOR DETUNED SIGNAL- . . . PHYSICAL REVIEW D 76, 062002 (2007)
the signal-extraction port of the interferometer in dark other. In addition, it is easy to combine the two beams
fringe, is exactly what our local readout scheme proposes before injection into the interferometer without losing the
to measure. However, sensitivity of the current Michelson power. An alternative way to the orthogonal polarizations
control signal must be improved dramatically in order to be is to use a cavity that can separate the beams at different
turned into our regime. We note that more precise mea- frequencies, where one resonates in the cavity while the
surement of this Michelson degree of freedom also helps to other does not. The cavity, a so-called output-mode-
decrease control-loop noise, which is shot noise imposed cleaner, is already planned to be used at the detection
on the control signal coupling to the main signal due to port in Advanced LIGO. In the same way an input mode
unavoidable imbalances [20]. cleaner cavity can be used to combine two beams before
Optical Power. In the baseline design, a pair of radio the injection into the interferometer.
frequency (RF) sidebands created around the main carrier Alternative configuration. One may also place the local
frequency will be injected to probe the motion as is already meters around the ETMs. In this case, a single laser beam,
done in current detectors. However, the power level of which can be different in frequency from the carrier light,
current RF sidebands is not high enough for our local should be split and brought to each end of the arms so that
readout. In the baseline design, the input power is 125 W, laser noise can be canceled out after taking a subtraction of
which is amplified to
1:0 kW at each ITM, due to power the two ETMs’ motion measurements. A cavity can be
recycling. Only about 1% of the power at the input port is implemented as well as it is proposed for a radiation-
pumped into the RF sidebands that resonate in the power- pressure-noise reduction method in [13,14]. In this way
recycling cavity but not in the arms. Taking into account the secondary laser for the local readout does not need such
the fact that the RF sidebands do not enter the arm cavities high power and there is no concern of a heat problem at the
and thus suffer from less optical losses, the power of the BS and the ITMs. However, in this case much more addi-
Michelson-control sidebands at the ITM is currently tional optical components are required to realize this
planned to be
34 W. Thus, one needs to raise the current configuration.
power by
120 times in order to achieve P2  4 kW.
Another more realistic way of realization is to use a phase- V. CONCLUSION
locked secondary laser with its frequency shifted by an odd
Motivated by the optical-bar schemes [12] and quantum-
number of half free-spectral ranges from the primary laser
locking schemes [13,14], we have proposed injecting
to satisfy the off-resonant condition in the arms.
a second laser beam into detuned signal-recycled
Furthermore, this subcarrier should almost be in dark
Michelson interferometers, sensing the differential motion
fringe at the signal-extraction port and should be resonant
of the input mirrors, and improving low-frequency sensi-
in both recycling cavities. To achieve a circulating power
tivities of these interferometers, currently at low frequen-
of P2  4 kW for the subcarrier we even need a little cies being limited by the rigidity of the optical spring. We
more input power than for the primary laser. But we can derived the optimal combined sensitivity of this double-
hope to use the higher-power laser for the subcarrier while readout scheme, and demonstrated that this optimal sensi-
the parametric instability [21,22] in the arm cavity may tivity is invariant with respect to the application of a feed-
limit the power of the primary laser. Indeed, a circulating back control scheme.
power of P2  4 kW is only a few times more than the Taking into account the current classical-noise budget of
carrier power of the current GEO detector which has a Advanced LIGO, as well as constraints on optical power,
similar topology compared to the local meter. we performed an optimization of our double-readout
Detection. Each signal at the dark port should be ex- schemes toward the detection of compact-binary inspirals.
tracted with some reference field, which will be another set This scheme is shown either to be able to broaden the
of RF sidebands in the RF readout scheme, or DC offset detection band and (indirectly) significantly improve
light in the DC readout scheme. The former one leaks high-frequency sensitivities, or to allow the detection of
through the dark port via macroscopic asymmetry in the intermediate-mass black-hole binaries with a broad fre-
central Michelson interferometer, and the latter one leaks quency range without sacrificing sensitivity to neutron-
through the dark port via microscopic asymmetry between star binaries and stellar-mass black-hole binaries.
the two arm cavities. Either way, the reference fields for the We also discussed briefly how the sensing of the
carrier and the subcarrier should be isolated before the Michelson degree of freedom in the current plan of
photo-detection, otherwise the reference field which is Advanced LIGO can be made dramatically more sensitive
not used for the signal extraction will just impose extra and turned into our local readout scheme.
shot noise. One way to solve the problem is to make use of Finally, we would like to point out that this scheme
orthogonal polarizations. Before the photo-detection, the should be further investigated as a candidate design for
carrier and the subcarrier accompanied with the reference third-generation detectors, possibly in conjunction with the
fields can be separated by a polarized beam splitter, which injection of squeezed vacuum states [23,24] into the inter-
is all-reflective to one polarization and transmissive to the ferometer’s dark port [25,26].

062002-9
HENNING REHBEIN et al. PHYSICAL REVIEW D 76, 062002 (2007)
ACKNOWLEDGMENTS proportional to
We thank Stan Whitcomb for very useful discussions. p



y^ 2 /
^ ETM 
^ ITM  2 2
^ BS  Lh; (A2)
Research of H. M.-E., K. S., and Y. C. is supported by the
Alexander von Humboldt Foundation’s Sofja Kovalevskaja where
BS is the classical noise acting on the BS. Suppose
Programme. Research of H. R. and R. S. is supported by the again that
ITM ,
ETM , and
BS have independent noise at
Deutsche Forschungsgemeinschaft through the EGC pro- the same level for ITM and ETM but half as high for the BS
gram and the SFB No. 407, respectively. Research of C. L. [cf. Eq. (6)]. We obtain that the optimal filter uses 3=4 of
is supported in part by NSF grants No. PHY-0099568 and the output of the large-scale interferometer and 1=4 of the
No. PHY-0601459. small interferometer in the units as above. This is in con-
trast to the optimal filter functions when only quantum
APPENDIX A: DOUBLE-READOUT SCHEME noise is taken into account as in the left panel of Fig. 3.
DOMINATED BY CLASSICAL NOISE Then the combined output is given by
Suppose at low frequencies, sensing noise is negligible, 1 1
and noise is dominated by the classical force noise acting y^ /
ETM 
ITM  p



BS  Lh: (A3)
2 2
on the mirrors. Then, the first carrier offers the following:
Then the large-scale interferometer’s noise spectral density
y^ 1 /
^ ETM 
^ ITM  Lh; (A1)
versus the optimal noise spectral density reads 2= 32 which
where
ETM and
ITM are classical noise on the ETM and gives the factor in Fig. 5. In this way the double-readout is
the ITM, respectively. The output of the second carrier is able to cancel some fraction of the classical noise.

[1] D. Shoemaker (LIGO Scientific Collaboration), Nucl. [14] J. M. Courty, A. Heidmann, and M. Pinard, Phys. Rev.
Instrum. Methods Phys. Res., Sect. A 517, 154 (2004). Lett. 90, 083601 (2003).
[2] L. D. Fiore (VIRGO collaboration), Classical Quantum [15] H. J. Kimble, Y. Levin, A. B. Matsko, K. S. Thorne, and
Gravity 19, 1421 (2002). S. P. Vyatchanin, Phys. Rev. D 65, 022002 (2001).
[3] B. Willke et al., Classical Quantum Gravity 19, 1377 [16] T. Corbitt, Y. Chen, and N. Mavalvala, Phys. Rev. A 72,
(2002). 013818 (2005).
[4] M. Ando (TAMA collaboration), Phys. Rev. Lett. 86, 3950 [17] J. Harms, R. Schnabel, and K. Danzmann, Phys. Rev. D
(2001). 70, 102001 (2004).
[5] G. Heinzel, A. Freise, H. Grote, K. Strain, and [18] http://www.ligo.mit.edu/bench/bench.html.
K. Danzmann, Classical Quantum Gravity 19, 1547 [19] T. Damour, B. R. Iyer, and B. S. Sathyaprakash, Phys. Rev.
(2002). D 62, 084036 (2000).
[6] Advanced LIGO Reference Design, LIGO M060056, [20] K. Somiya, O. Miyakawa, P. Fritschel, and R. Adhikali,
http://www.ligo.caltech.edu/docs/M/M060056-07/ Report No. LIGO-T060272-00-I, 2006.
M060056-07.pdf. [21] V. B. Braginsky, S. E. Strigin, and S. P. Vyatchanin, Phys.
[7] A. Buonanno and Y. Chen, Phys. Rev. D 64, 042006 Lett. A 287, 331 (2001).
(2001). [22] V. B. Braginsky, S. E. Strigin, and S. P. Vyatchanin, Phys.
[8] A. Buonanno and Y. Chen, Phys. Rev. D 65, 042001 Lett. A 305, 111 (2002).
(2002). [23] H. Vahlbruch, S. Chelkowski, B. Hage, A. Franzen, K.
[9] A. Buonanno and Y. Chen, Phys. Rev. D 67, 062002 Danzmann, and R. Schnabel, Phys. Rev. Lett. 97, 011101
(2003). (2006).
[10] K. Somiya, P. Beyersdorf, K. Arai, S. Sato, S. Kawamura, [24] K. McKenzie, N. Grosse, W. P. Bowen, S. E. Whitcomb,
O. Miyakawa, F. Kawazoe, S. Sakata, A. Sekido, and N. M. B. Gray, D. E. McClelland, and P. K. Lam, Phys. Rev.
Mio, Appl. Opt. 44, 3179 (2005). Lett. 93, 161105 (2004).
[11] O. Miyakawa et al., Phys. Rev. D 74, 022001 (2006). [25] J. Harms, Y. Chen, S. Chelkowski, A. Franzen, H.
[12] V. B. Braginsky, M. L. Gorodetsky, and F. Y. Khalili, Phys. Vahlbruch, K. Danzmann, and R. Schnabel, Phys. Rev.
Lett. A 232, 340 (1997). D 68, 042001 (2003).
[13] S. Kawamura, O. Miyakawa, and K. Somiya, ‘‘Reduction [26] A. Buonanno and Y. Chen, Phys. Rev. D 69, 102004
of Radiation Pressure Noise Using an Auxiliary (2004).
Interferometer in a Laser Interferometric Gravitational
Wave Antenna’’ (unpublished).

062002-10