LIGO: PROGRESS AND PROSPECTS
Barry C. Barish
California Institute of Technology, Pasadena, CA 90405, USA
ABSTRACT
The status, progress and plans for the Laser Interferometer Gravitational-wave Observatory (LIGO) are
presented. Ground based gravitational wave detectors are complementary to the initiatives in space. The
ground-based detectors will cover higher frequencies (10 to 10 4 Hz) and will therefore be sensitive to different
sources or different regimes for the same astrophysical source.
INTRODUCTION
The Laser Interferometer Gravitational-wave Observatory (LIGO) is a joint Caltech-MIT project supported by
the National Science Foundation (Barish and Weiss, 1999). The LIGO observations will be carried out with long
baseline suspended mass interferometers that are located at Hanford, Washington and Livingston, Louisiana. To
unambiguously make detections of these rare events a time coincidence within 10ms (transit time at the speed of
light) between detectors separated by 3030 km will be sought. The construction of LIGO has been completed and
the two facilities can be seen in the aerial photograph in Figure 1.
Fig 1. Aerial photographs of the 4km long suspended mass interferometers installed a Livingston, Louisiana on the
left and Hanford, Washington on the right are shown.
2
LIGO is now entering the commissioning phase for this complex instrument. Following a planned two year
commissioning program, we expect the first dedicated sensitive broadband searches for astrophysical gravitational
waves at an amplitude (strain) of h ~ 10-21 to begin during 2002. The initial search with LIGO will be the first
attempt to detect gravitational waves with a detector having sensitivity that intersects plausible estimates for
known astrophysical source strengths. The initial detector constitutes a 100 to 1000 fold improvement in both
sensitivity and bandwidth over previous searches.
The facilities developed to support the initial interferometers will allow for the evolution of the detectors to
probe the field of gravitational wave astrophysics with more and more sensitivity over the next two decades.
Sensitivity improvements and special purpose detectors will be needed either to enable detection if strong enough
sources are not found with the initial interferometer, or following detection, to increase the rate in order to enable
the detections to begin to become a new tool for astrophysical research. The other detectors in the world offer
potential simultaneous detection of astrophysical sources and there are plans underway to correlate signals from all
operating detectors in a worldwide network.
THE LIGO SUSPENDED MASS INTERFEROMETERS
The optical configuration in LIGO is a power recycled Fabry-Perot Michelson interferometer. It is this type
of configuration that will be used for the initial interferometers. The layout and parameters are shown in Figure 2.
There are many other possible types of interferometer configurations that have been considered and might be used
in future detectors. Considerable research has been done at Stanford on Sagnac interferometers, as the power of
the lasers increase an all reflective geometry might be used, and there is considerable work toward adding a signal
recycling mirror on the output side of the interferometer, allowing an ability to tune the interferometer for
increased sensitivity at a particular frequency in a narrow band mode.
The initial LIGO interferometer
parameters have been chosen such that the sensitivity will be consistent both with estimates needed for possible
detection of known sources. Although the rate for such sources have large uncertainty, improvements in sensitivity
linearly improves the distance searched for detectable sources, which increases the rate by the cube of the
improvement. Our plans are to upgrade the detector in about 2006 with improved mirrors, suspensions, seismic
isolation and laser giving about a factor of 20 sensitivity improvement, and to also add the signal recycling feature
for narrow band searches.
Fig 2. The optical layout of the LIGO suspended
mass Michelson interferometer with Fabry-Perot arm
cavities
Fig 3. The limiting and other noise sources
for LIGO are shown. The shaded area indicates the
detector sensitivity.
4
Noise Sources and Sensitivity
The success of the detector ultimately will depend on how well we are able to control the noise in the
measurement of very small strains. Noise is broadly but also usefully categorized in terms of those phenomena
which limit the ability to sense and register the small motions (sensing noise limits) and those that perturb the
masses by causing small motions (random force noise). Eventually one reaches a practical but not fundamental
limiting noise, the quantum limit, which combines the sensing noise with a random force limit. This orderly and
intellectually satisfying categorization presumes that one is careful enough as experimenters in the execution of the
experiment that one has not produced less fundamental, albeit, real noise sources that are caused by faulty design
or poor implementation. We have dubbed these as technical noise sources and in real life these have often been the
impediments to progress. The primary noise sources for the initial LIGO detector are shown in Figure 3.
In order to control the technical noise sources, extensive use has been made of two concepts. The first is the
technique of modulating the signal to be detected at frequencies far above the 1/f noise due to the drift and gain
instabilities experienced in all instruments. For example, the optical phase measurement to determine the motion of
the fringe is carried out at radio frequency rather than near DC. Thereby, the low frequency amplitude noise in the
laser light does not directly perturb the measurement of the fringe position. (The low frequency noise still causes
radiation pressure fluctuations on the mirrors through the asymmetries in the interferometer arms.) A second
concept is to apply feedback to physical variables in
the experiment to control the large excursions at low
frequencies and to provide damping. The variable is
measured through the control signal required to hold
it stationary. Here a good example is the position of
the interferometer mirrors at low frequency. The
interferometer fringe is maintained at a fixed phase
by holding the mirrors at fixed positions at low
frequencies. Feedback forces to the mirrors
effectively hold the mirrors “rigidly”. In the initial
LIGO interferometers the forces are provided by
permanent magnet/coil combinations. The mirror
motion that would have occurred is then read in the
control signal required to hold the mirror.
Great care has been taken in LIGO to control
these technical noise sources. In order to test and
understand our sensitivity and the noise limitations,
we have performed extensive tests with a 40-meter
prototype interferometer on the Caltech campus.
This interferometer has essentially all the pieces and
Fig 4. The displacement noise measured in the
the same optical configuration as is used in LIGO, so
LIGO 40m suspended mass interferometer prototype
it represents a good place to demonstrate our
on the Caltech campus. The general shape and level
understanding of these issues before confronting
are well simulated by our understanding of the
them in LIGO. The 40m prototype has achieved a
limiting noise sources: seismic noise at the lowest
displacement sensitivity of h ~ 10-19 m, which is
frequencies; suspension thermal noise at the
comparable to the displacement sensitivity required
intermediate frequencies; and shot noise or
in the 4 km LIGO interferometers. Figure 4 shows
photostatistics at the highest frequencies. Also, the
the measured noise curve in the 40m prototype and
primary line features are understood as various
our understanding of the contributions from various
resonances in the suspension system.
noise sources.
ASTROPHYSICAL SOURCES OF GRAVITATIONAL WAVES
There are a many known astrophysical processes in the Universe that produce gravitational radiation (Thorne
1987). Terrestrial interferometers, like LIGO, will search for signals from in the 10Hz - 10KHz frequency band.
Characteristic signals from astrophysical sources will be sought over background noise from continuously recorded
time-frequency series of the strain. Examples of sources and their characteristic signals include the following:
Chirp Signals
The inspiral of compact objects such as a pair of
neutron stars or black holes gives radiation that
characteristically increases in both amplitude and
frequency as the system evolves toward its the final
coalescence. This characteristic ‘chirp signal’ can be
calculated in detail, and the final waveform is
sensitive to the masses, separation, ellipticity of the
orbits, etc. A variety of search techniques have been
developed, including a direct comparison with an
array of templates that cover the range of possible
parameters. The waveform for the inspiral phase is
well understood and has been calculated in sufficient
detail for neutron star-neutron star inspiral. This
inspiral phase is well matched to the LIGO
sensitivity band for neutron star binary systems. For
heavier systems, like a system of two black holes, the
final coalescence and even the ring down phases are
in the LIGO frequency band. This is potentially very
interesting from a scientific point of view, but the
lack of detailed knowledge of the waveforms make
the searches more challenging.
The expected rate of coalescing binary neutron
star systems (with large uncertainties) is expected to
be a few per year within about 200 Mpc.
Coalescence of neutron star/black hole or black
hole/black hole airs may provide stronger signals but
their rate of occurrence (as well as the required
detection algorithms) are more uncertain. Recently,
enhanced mechanisms for ~10M blackholeblackhole mergers have been proposed, making these
systems of particular interest. The expected signal
for different compact sources is indicated in Figure
5.
Fig 5. The sensitivity curves for detection of
compact binary inspirals for the initial and the
improved LIGO interferometers are shown and
compared with the expected signal from the neutron
star – neutron star binary inspiral benchmark events.
Note that the sensitivity of the initial detector has
been chosen as a balance of the arguments above
making detection plausible and the use of
demonstrated technologies. Also shown is the
sensitivity for 20 M blackhole mergers, showing the
sensitivity to merger and ringdown phase.
Periodic Signals
Radiation from rotating non-axisymmetric neutron stars will produce periodic signals in the detectors. The
emitted gravitational wave frequency is twice the rotation frequency. For many known pulsars, the frequency falls
within the LIGO sensitivity band. Searches for signals from spinning neutron stars will involve tracking the system
for many cycles, taking into account the doppler shift for the motion of the Earth around the Sun, and including the
effects of spin-down of the pulsar. Both targeted searches for known pulsars and general sky searches are
anticipated.
Stochastic Signals
Signals from gravitational waves emitted in the first instants of the early universe, as far back as the Planck
epoch at 10-43 sec, can be detected through correlation of the background signals from two or more detectors.
Gravitational waves can probe earlier in the history of the Universe than any other radiation due to the very weak
6
interaction. Some models of the early Universe can result in detectable signals. Observations of this early
Universe gravitational radiation would provide an exciting new cosmological probe.
Burst Signals
The gravitational collapse of stars (e.g. supernovae) will lead to emission of gravitational radiation. Type I
supernovae involve white dwarf stars and are not expected to yield substantial emission. However, Type II
collapses can lead to strong radiation if the core collapse is sufficiently non-axisymmetric. The rate of Type II
supernovae is roughly once every 30 years in our own Galaxy. This is actually a lower bound on the rate of stellar
core collapses, since that rate estimate is determined from electromagnetic observations and some stellar core
collapses could give only a small electromagnetic signal. The ejected mantle dominates the electromagnetic signal,
while the gravitational wave signal is dominated by the dynamics of the collapsing core itself.
Numerical modeling of the dynamics of core collapse and bounce has been used to make estimates of the
strength and characteristics of the signal emitted in gravitational radiation. This is a very complicated problem and
the predictions are quite model dependent, depending both on detailed hydrodynamic processes and the initial
rotation rate of the degenerate stellar core before collapse. Estimating the event detection rate is consequently
difficult and the rate may be as large as many per year with initial LIGO interferometers, or less than one per year
with advanced LIGO interferometers. Probably a reasonable guess is that the initial detectors will not see far
beyond our own galaxy, while an advanced detector should see out to the Virgo cluster, where the rate is a
few/year.
THE STATUS OF THE LIGO
The construction of LIGO infrastructure in both
Hanford, Washington and in Livingston, Louisiana
began in 1996 and was completed on schedule at the
end of last year. . The long beam pipes are kept
under high vacuum at all times and can be isolated
from the large chambers containing the mirror-test
masses and associated optics and detectors by the
means of large gate valves that allow opening the
chambers without disturbing the vacuum in the long
beam pipes (see Figure 6)
The infrastructure for LIGO consists of
prepartion of both sites, civil construction for both
the laboratory buildings and the enclosures for the
vacuum pipes, as well as the large volume high
vacuum system to house the interferometers. The
Fig. 6. A photograph of the large vacuum chambers
large vacuum system was the most challenging part
containing the various LIGO detector components is
of the infrastructure for LIGO, involving 16 km or
shown. These chambers are isolated from the long
1.2 m diameter high vacuum pipe. That system
vacuum pipes by gate valves to access the equipment.
achieved 10-6 torr vacuum pumping only from the
ends with vacuum and turbo is in place and pumps.
The long vacuum pipes were then ‘baked’ to
accelerate the outgassing by wrapping the pipes in insulation and running a current of 2000 amps down the pipes,
which raised the temperature to ~ 1600 C. We baked at this temperature for about one week. Following
cooldown, the pipes have no leaks and achieved a vacuum of better than 10-9 torr. All 16 km of beam pipe are now
under high vacuum and the level of vacuum achieved is such that noise from scattering off residual molecules
should not be a problem for either initial LIGO or envisioned upgrades. The long beam pipes are kept under high
vacuum at all times and can be isolated from the large chambers that contain the mirror-test masses and associated
optics and detectors by the means of large gate valves that allow opening the chambers without disturbing the
vacuum in the pipes. .
Having completed the infrastructure, the installation and commissioning of the detector subsystems began this
past year. The laser we have developed for LIGO is a 10W Nd:YAG laser at 1.064 m in the TEM00 mode. The
laser has been developed for production through Lightwave Electronics, using their 700-mwatt NPRO laser as the
input to a diode pumped power amplifier. This commercialized laser is now sold by Lightwave as a catalog item.
8
We have been running one laser continuously for about a year with good reliability. We are optimistic that this
laser will make a reliable input light source for the LIGO interferometers.
For the LIGO application, the laser must be further stabilized in frequency, power and pointing. We have
developed a laser prestabalization subsystem, which is performing near our design requirements. We require for
40 Hz < f < 10 KHz,
Frequency noise:
Intensity noise:
dn(f) < 10-2 Hz/Hz1/2
-6
1/2
dI(f)/I < 10 /Hz
This low noise highly stabilized laser system has been tested and characterized in detail and is performing near
LIGO design specifications. Figure 7 shows some performance measurements of the prestabilized laser system.
Detailed characterization and improvement of noise sources continue
.
Fig 7. Performance of the LIGO prestabilized laser in frequency (left) and power (right). The lines indicate the
noise requirements for the interferometers.
The prestabilized laser beam is further conditioned by a 12 m triangular mode cleaner, which is also
operational. The University of Florida group in LIGO has developed the input optics containing this mode cleaner.
The beam has been successfully transported through that system and sent down the first 2 km arm.
LIGO at Hanford has both a half-length (2km) and a full-length (4km) interferometer installed in the same
vacuum chamber. The extra constraint of requiring a ½ size signal in the shorter interferometer will be used to
eliminate common noise and lower the singles rate in the coincidence between the sites. Recently, the long 2 km
arm cavities have been locked one at a time for typical few hour times, at which point lock is lost due to tidal
effects. The electronics to compensate for tidal effects are not yet installed. This first success at locking a long
cavity is an important accomplishment and gives us confidence we will soon lock the entire LIGO interferometer.
Various monitoring signals for a 15 minute locked period are shown in Figure 8. The next and final step for the
first interferometer, which we are using as a pathfinder to find any big problems, is to lock both arms at the same
time, at which point we will have achieved the full LIGO suspended mass Michelson Interferometer with FabryPerot arms.
Following the locking of the long arms early this year, The first step for us this summer was to turn on the
power-recycled Michelson part of the interferometer, formed by the input mirrors, as well as the beam splitter and
the recycling mirror. To make LIGO as sensitive as possible, we want to have as much light as possible returning
to the beam splitter. This occurs when the recycling mirror is placed at the correct distance from the beam splitter,
‘trapping’ the reflected light in the Michelson interferometer. This causes the laser light in the interferometer to
build up to a high level. For the full LIGO-I system, recycling will cause the light to build up by about a factor of
thirty. When this build up occurs, the power-recycled Michelson interferometer is resonating. That resonant state
was achieved within the last few weeks for this short arm system.
We now plan to add one of the 2-km-long arms resonating, which will be the final step in preparation for
turning on the full interferometer. We have now installed all of the electronics for the entire interferometer, and
we expect to achieve ‘first lock’ of LIGO this fall.
Fig 8. A locked stretch of a 2km arm of LIGO showing the transmitted light and various control and error
signals is shown. This marks an important milestone in making LIGO operational.
LIGO DATA ANALYSIS PREPARATIONS AND ISSUES
We have been preparing for analysis of LIGO interferometer data. These preparartions have involved
designing and implementing a computer system capable of LIGO data analysis, development and implementing
algorithms for the different searches. In addition, we have been testing our ability to extract signals from noise
expected in real interferometer data. I will only briefly discuss the last item here.
Fig 9. An illustration is shown of the types of instrumental signals that must be dealt with in the LIGO
interferometers. These are from the 40m prototype data.
1
0
In order to test the ability to pull signals out of real interferometer data, a test end-to-end analysis was
performed using 40m data (noise curve is shown in Figure 4). The data taken in 1994 consists of a stretch of 44.8
hours of reasonably stable running. The data has been analyzed for neutron star binary inspirals using the
technique of matched templates (a total of 687 filters). The sensitivity of the interferometer was h ~ 3.5 10-19
mHz-1/2, which corresponds to a sensitivity capable of making a detection within our galaxy, but not beyond. It
should be noted that the expected rate in our galaxy is only ~10 -6/yr, so this data has a negligibly small chance for
making a real detection. The real value is that it allows us to test how well the data analysis techniques work, in
advance of obtaining real LIGO interferometer data.
The data consists of 39.9 hrs of running where the interferometer was locked. This locked data was analyzed
using good data segments within that data, which amounted to a reduction to 25.0 hrs of data (after removing
obvious problems, sections too near the time lock was acquired, etc). For the LIGO interferometers, we hope to
improve the fraction of good data available during a running period to at least 90%, and we have required such
performance reliability in our interferometer design. It is important to note that the 40m interferometer data
contains lots of features that that complicate the data analysis. Using such data provides a good test of our
analysis. Some of the types of peculiar instrumental signals observed in the 40m data are shown in Figure 9.
The data was first cleaned up using various procedures, for example removing sinusoidal artifacts by multi-taper
methods. The data contains both significant non-stationary and non-gaussian noise, which we will need to deal
with in LIGO data. In order to test how well one can extract a binary inspiral chirp signals from such noise,
simulated events were superposed on the actual 40m noise and then the analysis code was used to extract the
signal. The simulated signals covered the range of parameters for a binary system. A comparison then could be
made of the efficiency of extracting these signals and how well the reconstructed parameters compared with the
injected parameters (e.g. distance of the source).
Figure 10 shows one such comparison by comparing the distance injected and reconstructed as a function
of the distance. Perfect reconstruction would lie on
the diagonal, so the deviation from that line is an
indication of the error in reconstruction due to the
presence of noise. It is clear that the events cluster
quite well along the diagonal as desired.
An analysis of the errors in distance
measurements is consistent with the signal to noise
ratio (SNR) fluctuations of our expectations. This
demonstrates that the reconstruction of distance in
the presence of this non-gaussian and non-stationary
noise was not significantly degraded. From this data,
an upper limit on event rate was determined using a
technique based on the SNR of the ‘loudest’ event.
The loudest event is the one that fit the template
analysis best. From this analysis the limit on rate of
binary inspirals in our galaxy was determined to be R
< 0.5/hour with 90% CL. The overall detection
efficiency  = 0.33. It is worth noting that an ideal
detector for this amount of good data would set a
limit of R < 0.16/hour. From this study, we conclude
that analysis for signals for binary inspirals works on
Fig 10. Analysis of simulated events inserted into
this within a factor of 3 degradation of what might be
real 40m noise comparing the distance of the inserted
expected in an ideal analysis. A challenge for us in
event with the reconstructed distance from the data
the development of the data analysis for the LIGO
analysis
interferometers is to improve this efficiency to be as
close to the ideal case as possible.
CONCLUSIONS AND PLANS
We are optimistic that we will achieve the major milestone this year of locking the first full interferometer as
a suspended mass Michelson interferometer with Fabry Perot arms. We will then characterize the performance and
begin to concentrate our efforts on noise and reliability issues. We also will be commissioning the second and
third interferometers, using lessons learned on the first interferometer. We expect to schedule some short
engineering test runs during this long turn period. We already had one such engineering data run last March
recording data for a single locked arm.
We expect to have a similar short run once we lock the entire
interferometer this fall and similarly another after we achieve coincidence running between the sites, probably by
next summer. Our long-term plan is evolve into a science data-taking mode, where we suspend interferometer
studies, and try to collect a significant amount of data some time during 2002. Our eventual goal to collect at least
1 year of integrated coincidence data between the two sites with sensitivity near 10 -21. Depending on how well we
do at making LIGO robust and how quickly we obtain design noise levels, we estimate that goal should be
reachable by sometime in 2005. As soon as possible following that achievement, we want to be prepared to
undertake improvements that will yield a significant improvement in sensitivity. .
As described above, the initial LIGO detector is a compromise between performance and technical risk. The
design incorporates some educated guesses concerning the directions to take to achieve a reasonable probability for
detection. It is a broadband system with modest optical power in the interferometer arms and a low risk vibration
isolation system. The suspensions and other systems have a direct heritage to the demonstration interferometer
prototypes we have tested over the last decade. As ambitious as the initial LIGO detectors seem, there are clear
technical improvements we expect to make, following the initial search. The initial detector performance and
results will guide the specific directions and priorities to implement from early data runs.
We anticipate making both reductions of noise from stochastic sources and in the sensing noise. These
improvements will include improvements in the suspension system to improve the thermal noise, the seismic
isolation and improvements to the sensing noise through the use of higher power lasers in conjunction with
improved optical materials for the test masses/mirrors to handle this higher power. We believe it is quite realistic
to improve the sensitivity at 100 Hz by at least a factor of 10, and to broaden the sensitive bandwidth by about a
factor without any radically new technologies or very large changes. This will improve the rate (or volume of the
universe searched) at a fixed sensitivity by a factor of 1000. If the physics arguments favor an even greater
sensitivity in a narrower bandwidth, it will be possible to change the optical configuration and make a narrow band
device. In conclusion, I might note that over the next decade ground-based detectors offer good prospects for
detection of gravitational waves. They are complementary in approach and in scientific coverage to space based
detectors. Hopefully within about a decade we will be detecting signals from gravitational waves, both on the earh
and in space.
References
1. Barish, B.C. and R. Weiss, Physics Today, October 1999 and LIGO Web site http://www.ligo.caltech.edu/
2. Thorne, K.S. 300 Years of Gravitation Edited by S.W. Hawking and W. Israel Cambridge University Press,
Cambridge, England 1987 Chapter 9 “Gravitational radiation”