arXiv:gr-qc/0308043v3 17 Sep 2003
Detector Description and Performance
for the First Coincidence Observations
between LIGO and GEO
The LIGO Scientific Collaboration
Corresponding Author: David Shoemaker, MIT NW17-161, 175 A
lbany St., Cambridge, MA 02139
Tel: 617 253 6411 Fax: 617 253 7014 Email:
dhs@ligo.mit.edu
B. Abbott
m
, R. Abbott
p
, R. Adhikari
n
, B. Allen
am
, R. Amin
ah
,
S. B. Anderson
m
, W. G. Anderson
ac
, M. Araya
m
, H. Armandula
m
, F. Asiri
m
,
28
,
P. Aufmuth
ae
, C. Aulbert
a
, S. Babak
g
, R. Balasubramanian
g
, S. Ballmer
n
,
B. C. Barish
m
, D. Barker
o
, C. Barker-Patton
o
, M. Barnes
m
, B. Barr
ai
,
M. A. Barton
m
, K. Bayer
n
, R. Beausoleil
z
,
8
, K. Belczynski
w
, R. Bennett
ai
,
21
,
S. J. Berukoff
a
,
30
, J. Betzwieser
n
, B. Bhawal
m
, G. Billingsley
m
, E. Black
m
,
K. Blackburn
m
, B. Bland-Weaver
o
, B. Bochner
n
,
7
, L. Bogue
m
, R. Bork
m
,
S. Bose
an
, P. R. Brady
am
, J. E. Brau
ak
, D. A. Brown
am
, S. Brozek
ae
,
26
,
A. Bullington
z
, A. Buonanno
f
,
33
, R. Burgess
n
, D. Busby
m
, W. E. Butler
a
ℓ
,
R. L. Byer
z
, L. Cadonati
n
, G. Cagnoli
ai
, J. B. Camp
u
, C. A. Cantley
ai
,
L. Cardenas
m
, K. Carter
p
, M. M. Casey
ai
, J. Castiglione
ah
, A. Chandler
m
,
J. Chapsky
m
,
11
, P. Charlton
m
, S. Chatterji
n
, Y. Chen
f
, V. Chickarmane
q
,
D. Chin
aj
, N. Christensen
h
, D. Churches
g
, C. Colacino
ae
,
b
, R. Coldwell
ah
,
M. Coles
p
,
20
, D. Cook
o
, T. Corbitt
n
, D. Coyne
m
, J. D. E. Creighton
am
,
T. D. Creighton
m
, D. R. M. Crooks
ai
, P. Csatorday
n
, B. J. Cusack
c
, C. Cutler
a
,
E. D’Ambrosio
m
, K. Danzmann
ae
,
b
,
t
, R. Davies
g
, E. Daw
q
,
25
, D. DeBra
z
,
T. Delker
ah
,
1
, R. DeSalvo
m
, S. Dhurandar
ℓ
, M. D
́
iaz
ac
, H. Ding
m
,
R. W. P. Drever
d
, R. J. Dupuis
ai
, C. Ebeling
h
, J. Edlund
m
, P. Ehrens
m
,
E. J. Elliffe
ai
, T. Etzel
m
, M. Evans
m
, T. Evans
p
, C. Fallnich
ae
, D. Farnham
m
,
M. M. Fejer
z
, M. Fine
m
, L. S. Finn
ab
,
́
E. Flanagan
i
, A. Freise
b
,
4
, R. Frey
ak
,
P. Fritschel
n
, V. Frolov
p
, M. Fyffe
p
, K. S. Ganezer
e
, J. A. Giaime
q
,
A. Gillespie
m
,
10
, K. Goda
n
, G. Gonz ́alez
q
, S. Goßler
ae
, P. Grandcl ́ement
w
,
A. Grant
ai
, C. Gray
o
, A. M. Gretarsson
p
, D. Grimmett
m
, H. Grote
b
,
S. Grunewald
a
, M. Guenther
o
, E. Gustafson
z
,
15
, R. Gustafson
aj
,
W. O. Hamilton
q
, M. Hammond
p
, J. Hanson
p
, C. Hardham
z
, G. Harry
n
,
A. Hartunian
m
, J. Heefner
m
, Y. Hefetz
n
, G. Heinzel
b
, I. S. Heng
ae
,
M. Hennessy
z
, N. Hepler
ab
, A. Heptonstall
ai
, M. Heurs
ae
, M. Hewitson
ai
,
N. Hindman
o
, P. Hoang
m
, J. Hough
ai
, M. Hrynevych
m
,
12
, W. Hua
z
, R. Ingley
ag
,
M. Ito
ak
, Y. Itoh
a
, A. Ivanov
m
, O. Jennrich
ai
,
5
, W. W. Johnson
q
,
W. Johnston
ac
, L. Jones
m
, D. Jungwirth
m
,
22
, V. Kalogera
w
, E. Katsavounidis
n
,
K. Kawabe
t
,
b
, S. Kawamura
v
, W. Kells
m
, J. Kern
p
, A. Khan
p
, S. Killbourn
ai
,
C. J. Killow
ai
, C. Kim
w
, C. King
m
, P. King
m
, S. Klimenko
ah
, P. Kloevekorn
b
,
S. Koranda
am
, K. K ̈otter
ae
, J. Kovalik
p
, D. Kozak
m
, B. Krishnan
a
, M. Landry
o
,
J. Langdale
p
, B. Lantz
z
, R. Lawrence
n
, A. Lazzarini
m
, M. Lei
m
,
Preprint submitted to Elsevier Science
4 February 2008
V. Leonhardt
ae
, I. Leonor
ak
, K. Libbrecht
m
, P. Lindquist
m
, S. Liu
m
,
J. Logan
m
,
18
, M. Lormand
p
, M. Lubinski
o
, H. L ̈uck
ae
,
b
, T. T. Lyons
m
,
18
,
B. Machenschalk
a
, M. MacInnis
n
, M. Mageswaran
m
, K. Mailand
m
,
W. Majid
m
,
11
, M. Malec
ae
, F. Mann
m
, A. Marin
n
,
6
, S. M ́arka
m
, E. Maros
m
,
J. Mason
m
,
16
, K. Mason
n
, O. Matherny
o
, L. Matone
o
, N. Mavalvala
n
,
R. McCarthy
o
, D. E. McClelland
c
, M. McHugh
s
, P. McNamara
ai
,
19
,
G. Mendell
o
, S. Meshkov
m
, C. Messenger
ag
, G. Mitselmakher
ah
, R. Mittleman
n
,
O. Miyakawa
m
, S. Miyoki
m
,
34
, S. Mohanty
a
, G. Moreno
o
, K. Mossavi
b
,
B. Mours
m
,
13
, G. Mueller
ah
, S. Mukherjee
a
, J. Myers
o
, S. Nagano
b
, T. Nash
j
,
H. Naundorf
a
, R. Nayak
ℓ
, G. Newton
ai
, F. Nocera
m
, P. Nutzman
w
, T. Olson
x
,
B. O’Reilly
p
, D. J. Ottaway
n
, A. Ottewill
am
,
32
, D. Ouimette
m
,
22
, H. Overmier
p
,
B. J. Owen
ab
, M. A. Papa
a
, C. Parameswariah
p
, V. Parameswariah
o
,
M. Pedraza
m
, S. Penn
k
, M. Pitkin
ai
, M. Plissi
ai
, M. Pratt
n
, V. Quetschke
ae
,
F. Raab
o
, H. Radkins
o
, R. Rahkola
ak
, M. Rakhmanov
ah
, S. R. Rao
m
,
D. Redding
m
,
11
, M. W. Regehr
m
,
11
, T. Regimbau
n
, K. T. Reilly
m
,
K. Reithmaier
m
, D. H. Reitze
ah
, S. Richman
n
,
23
, R. Riesen
p
, K. Riles
aj
,
A. Rizzi
p
,
9
, D. I. Robertson
ai
, N. A. Robertson
ai
,
z
, L. Robison
m
, S. Roddy
p
,
J. Rollins
n
, J. D. Romano
ac
, J. Romie
m
, H. Rong
ah
,
10
, D. Rose
m
, E. Rotthoff
ab
,
S. Rowan
ai
, A. R ̈udiger
t
,
b
, P. Russell
m
, K. Ryan
o
, I. Salzman
m
, G. H. Sanders
m
,
V. Sannibale
m
, B. Sathyaprakash
g
, P. R. Saulson
aa
, R. Savage
o
, A. Sazonov
ah
,
R. Schilling
t
,
b
, K. Schlaufman
ab
, V. Schmidt
m
,
3
, R. Schofield
ak
,
M. Schrempel
ae
,
29
, B. F. Schutz
a
,
g
, P. Schwinberg
o
, S. M. Scott
c
, A. C. Searle
c
,
B. Sears
m
, S. Seel
m
, A. S. Sengupta
ℓ
, C. A. Shapiro
ab
,
24
, P. Shawhan
m
,
D. H. Shoemaker
n
, Q. Z. Shu
ah
,
14
, A. Sibley
p
, X. Siemens
am
, L. Sievers
m
,
11
,
D. Sigg
o
, A. M. Sintes
a
,
af
, K. Skeldon
ai
, J. R. Smith
b
, M. Smith
n
,
M. R. Smith
m
, P. Sneddon
ai
, R. Spero
m
,
11
, G. Stapfer
p
, K. A. Strain
ai
,
D. Strom
ak
, A. Stuver
ab
, T. Summerscales
ab
, M. C. Sumner
m
, P. J. Sutton
ab
,
J. Sylvestre
m
, A. Takamori
m
, D. B. Tanner
ah
, H. Tariq
m
, I. Taylor
g
,
R. Taylor
m
, K. S. Thorne
f
, M. Tibbits
ab
, S. Tilav
m
,
2
, M. Tinto
d
,
11
, C. Torres
ac
,
C. Torrie
m
,
ai
, S. Traeger
ae
,
31
, G. Traylor
p
, W. Tyler
m
, D. Ugolini
ad
,
M. Vallisneri
f
,
11
, M. van Putten
n
, S. Vass
m
, A. Vecchio
ag
, C. Vorvick
o
,
L. Wallace
m
, H. Walther
t
, H. Ward
ai
, B. Ware
m
,
11
, K. Watts
p
, D. Webber
m
,
A. Weidner
t
,
b
, U. Weiland
ae
, A. Weinstein
m
, R. Weiss
n
, H. Welling
ae
, L. Wen
m
,
S. Wen
q
, J. T. Whelan
s
, S. E. Whitcomb
m
, B. F. Whiting
ah
, P. A. Willems
m
,
P. R. Williams
a
,
27
, R. Williams
d
, B. Willke
ae
,
b
, A. Wilson
m
, B. J. Winjum
ab
,
30
,
W. Winkler
t
,
b
, S. Wise
ah
, A. G. Wiseman
am
, G. Woan
ai
, R. Wooley
p
,
J. Worden
o
, I. Yakushin
p
, H. Yamamoto
m
, S. Yoshida
y
, I. Zawischa
ae
,
17
,
L. Zhang
m
, N. Zotov
r
, M. Zucker
p
, J. Zweizig
m
a
Albert-Einstein-Institut, Max-Planck-Institut f ̈ur Gra
vitationsphysik, D-14476
Golm, Germany
b
Albert-Einstein-Institut, Max-Planck-Institut f ̈ur Gra
vitationsphysik, D-30167
Hannover, Germany
c
Australian National University, Canberra, 0200, Australi
a
d
California Institute of Technology, Pasadena, CA 91125, US
A
e
California State University Dominguez Hills, Carson, CA 90
747, USA
2
f
Caltech-CaRT, Pasadena, CA 91125, USA
g
Cardiff University, Cardiff, CF2 3YB, United Kingdom
h
Carleton College, Northfield, MN 55057, USA
i
Cornell University, Ithaca, NY 14853, USA
j
Fermi National Accelerator Laboratory, Batavia, IL 60510,
USA
k
Hobart and William Smith Colleges, Geneva, NY 14456, USA
ℓ
Inter-University Centre for Astronomy and Astrophysics, P
une - 411007, India
m
LIGO - California Institute of Technology, Pasadena, CA 911
25, USA
n
LIGO - Massachusetts Institute of Technology, Cambridge, M
A 02139, USA
o
LIGO Hanford Observatory, Richland, WA 99352, USA
p
LIGO Livingston Observatory, Livingston, LA 70754, USA
q
Louisiana State University, Baton Rouge, LA 70803, USA
r
Louisiana Tech University, Ruston, LA 71272, USA
s
Loyola University, New Orleans, LA 70118, USA
t
Max Planck Institut f ̈ur Quantenoptik, D-85748, Garching,
Germany
u
NASA/Goddard Space Flight Center, Greenbelt, MD 20771, USA
v
National Astronomical Observatory of Japan, Tokyo 181-858
8, Japan
w
Northwestern University, Evanston, IL 60208, USA
x
Salish Kootenai College, Pablo, MT 59855, USA
y
Southeastern Louisiana University, Hammond, LA 70402, USA
z
Stanford University, Stanford, CA 94305, USA
aa
Syracuse University, Syracuse, NY 13244, USA
ab
The Pennsylvania State University, University Park, PA 168
02, USA
ac
The University of Texas at Brownsville and Texas Southmost C
ollege,
Brownsville, TX 78520, USA
ad
Trinity University, San Antonio, TX 78212, USA
ae
Universit ̈at Hannover, D-30167 Hannover, Germany
af
Universitat de les Illes Balears, E-07071 Palma de Mallorca
, Spain
ag
University of Birmingham, Birmingham, B15 2TT, United King
dom
ah
University of Florida, Gainsville, FL 32611, USA
ai
University of Glasgow, Glasgow, G12 8QQ, United Kingdom
aj
University of Michigan, Ann Arbor, MI 48109, USA
ak
University of Oregon, Eugene, OR 97403, USA
a
ℓ
University of Rochester, Rochester, NY 14627, USA
am
University of Wisconsin-Milwaukee, Milwaukee, WI 53201, U
SA
an
Washington State University, Pullman, WA 99164, USA
3
Abstract
For 17 days in August and September 2002, the LIGO and GEO inte
rferometer
gravitational wave detectors were operated in coincidence
to produce their first
data for scientific analysis. Although the detectors were st
ill far from their design
sensitivity levels, the data can be used to place better uppe
r limits on the flux of
gravitational waves incident on the earth than previous dir
ect measurements. This
paper describes the instruments and the data in some detail,
as a companion to
analysis papers based on the first data.
Key words:
LIGO, gravitational wave, interferometer, observatory
PACS:
04.89.Nn, 07.60.Ly, 95.45.+i, 95.55.Ym
4
1 Introduction
A number of laboratories around the world [TAMA[1], VIRGO[2
], GEO[3],[4],
LIGO[5,6]] are developing detectors for gravitational wav
es using laser in-
terferometers to sense the very small strains anticipated f
rom astrophysical
sources. In a joint effort, two of these laboratories, LIGO an
d GEO 600, have
performed their first scientific observations. This note is i
ntended to provide
greater detail in the description of the detectors themselv
es as a companion
to papers describing the data analysis and astrophysical co
nclusions from this
Science Run (designated S1).
Both GEO 600 and LIGO use the principle of the Michelson inter
ferometer,
1
Currently at Ball Aerospace Corporation
2
Currently at University of Delaware
3
Currently at European Commission, DG Research, Brussels, B
elgium
4
Currently at European Gravitational Observatory
5
Currently at ESA Science and Technology Center
6
Currently at Harvard University
7
Currently at Hofstra University
8
Currently at HP Laboratories
9
Currently at Institute of Advanced Physics, Baton Rouge, LA
10
Currently at Intel Corp.
11
Currently at NASA Jet Propulsion Laboratory
12
Currently at Keck Observatory
13
Currently at Laboratoire d’Annecy-le-Vieux de Physique de
s Particules
14
Currently at LightBit Corporation
15
Currently at Lightconnect Inc.
16
Currently at Lockheed-Martin Corporation
17
Currently at Laser Zentrum Hannover
18
Currently at Mission Research Corporation
19
Currently at NASA Goddard Space Flight Center
20
Currently at National Science Foundation
21
Currently at Rutherford Appleton Laboratory
22
Currently at Raytheon Corporation
23
Currently at Research Electro-Optics Inc.
24
Currently at University of Chicago
25
Currently at University of Sheffield
26
Currently at Siemens AG
27
Currently at Shanghai Astronomical Observatory
28
Currently at Stanford Linear Accelerator Center
29
Currently at Spectra Physics Corporation
30
Currently at University of California, Los Angeles
31
Currently at Carl Zeiss GmbH
32
Permanent address: University College Dublin
33
Permanent address: GReCO, Institut d’Astrophysique de Par
is (CNRS)
34
Permanent address: University of Tokyo, Institute for Cosm
ic Ray Research
5
with its high sensitivity to differential changes ∆
L
=
L
1
−
L
2
of the lengths
L
in the two perpendicular arm lengths
L
1
and
L
2
, to detect strains of the order
of
h
= ∆
L/L
= 10
−
20
over a wide frequency range. The required sensitivity of
the interferometric readout is achieved through the use of h
igh circulating laser
power (to improve the shot-noise limited fringe resolution
) and through tech-
niques to store the light in the interferometer arms (to incr
ease the phase shift
due to a passing gravitational wave). The frequency range of
interest for these
instruments lies in the audio band (
∼
50-5000 Hz), leading to gravitational
wavelengths
λ
gw
=
c/f
gw
of hundreds of km. Because practical ground-based
detectors are short compared to the wavelength, long interf
erometer arms are
chosen to increase the sensitivity of the instrument. A vacu
um system protects
the beams from variations in the light path due to air density
fluctuations.
The test masses, which also serve as mirrors for the Michelso
n interferometer,
are suspended as pendulums and respond as free masses above t
heir
∼
1 Hz
resonant frequency. External mechanical disturbances are
suppressed through
seismic isolation systems, and the in-band intrinsic therm
al noise is controlled
via careful choice of materials and construction technique
s.
1.1 LIGO
The LIGO Observatory construction started in 1994 at the LIG
O site in Han-
ford, Washington, USA. Construction at the Livingston, Lou
isiana USA site
began a year later in June 1995. The buildings and 4 km concret
e arm-support
foundations were completed in 1998. The vacuum systems were
completed in
1999, and detector installation was substantially complet
ed in 2000. The first
operation of a LIGO interferometer took place in October 200
0. This marked
the initiation of the commissioning, consisting of periods
of intense testing and
tuning of subsystems, separated by periods where the interf
erometers were run
as complete systems. These engineering runs were primarily
intended to assess
the progress toward full detector operation. However, they
were also used to
collect data that could be used to test data handling, archiv
ing and analysis
software. Progress through the commissioning phase has bee
n steady, both in
terms of improving sensitivity and in terms of reliable oper
ation with a rea-
sonable duty cycle. By summer 2002, the improvements had bee
n sufficient
that a short duration Science Run could be expected to achiev
e limits on the
observations of gravitational waves that would be comparab
le to or better
than previous experimental limits. Consequently, a two-we
ek observation pe-
riod was scheduled, and other laboratories operating inter
ferometer detectors
were invited to join in simultaneous operation, as document
ed here. Further
progress in sensitivity has subsequently been achieved thr
ough additional com-
missioning.
6
1.2 GEO 600
The construction of GEO 600 started in 1995 as a German/Briti
sh collabo-
ration on a site near Hannover, Germany. Because the site con
strained the
length of the arms to 600m, an advanced optical layout and nov
el techniques
for the suspension systems were included in the detector des
ign. After the
buildings and the trenches were finished in 1997 the complete
vacuum system
was installed and tested. The construction phase was follow
ed by the installa-
tion of the two suspended triangular mode cleaners which hav
e been operating
reliably since 2000. To gain experience with the alignment a
nd length control
of long baseline cavities the commissioning continued with
the installation
of a 1200 m long Fabry-Perot cavity formed by one interferome
ter arm and
the power recycling mirrors. To reduce the risk of contamina
ting or damaging
the expensive main interferometer mirrors, lower grade tes
t mirrors suspended
in steel wire slings were used for the 1200 m cavity experimen
t and for the
commissioning of the power recycled Michelson interferome
ter which started
in summer 2001. A first engineering test run was conducted in J
an 2002 in
coincidence with a LIGO engineering run.
The installation of the automatic alignment system for the M
ichelson interfer-
ometer and for the power-recycling cavity was a key step towa
rds a duty cycle
of more than 98% which was achieved in the 17 day S1 Science Run
. The in-
strument ran as a power-recycled folded-arm Michelson for S
1; commissioning
of signal recycling started after S1 and is expected to bring
the GEO detector
a significant step closer to its design sensitivity.
2 Purpose of the S1 Science Run
The primary goal of the S1 run was to collect a significant body
of data to
analyze for gravitational waves. Although the sensitivity
of all the instruments
was far from the design goal and the relatively short run time
made it unlikely
that a positive detection would be made, it was expected that
upper limits
could be derived from the data that would be comparable to or b
etter than
previous gravitational wave observations. Furthermore, t
he analysis provided
the opportunity to test the methodologies with real gravita
tional wave detec-
tor data. Estimates of sensitivity for gravitational wave i
nterferometers have
almost always been based on the assumption of Gaussian noise
. While this is
a good point of departure for many of the limiting noise sourc
es (e.g., shot
noise or thermal noise), many others (e.g., seismic noise) a
re not expected to
be so well-behaved. Thus, letting the data analysis confron
t the behavior of
real noise as early as possible is crucial to developing and t
esting the analysis
techniques.
7
Other goals for S1 were aimed at improving our understanding
of the detectors
and their operation. These include:
(1) Investigating the factors that influence duty cycle for t
he interferometers.
Long periods of operation with stable conditions are import
ant for under-
standing the causes for the interferometers to lose ‘lock’ (
loss of resonance
condition for light in the interferometer cavities and cons
equent loss of
linear operation of the sensing system)
(2) Characterization of drifts in alignment and optimizati
on of the alignment
control systems
(3) Testing and optimization of on-line monitoring tools fo
r assessing perfor-
mance and maintaining high sensitivity
(4) Training and practice for instrument operators and scie
ntific monitors.
This paper provides a description of the LIGO and GEO interfe
rometers as
they were used in the S1 run.
1
It is intended as a companion to the data
analysis papers based on data from this run. Because commiss
ioning was still
underway, many parts of the detectors were not in their inten
ded final op-
erational configuration, and an important emphasis of this p
aper will be to
identify and highlight those differences.
3 The LIGO detector array
The LIGO detector array comprises three interferometers at
two sites. The
LIGO Livingston Observatory (LLO) contains three main inst
rument bays at
the vertex and ends of the L-shaped site and houses a single in
terferometer
with 4 km long arms (designated L1). The LIGO Hanford Observa
tory (LHO)
has two additional experimental halls at the midpoint in eac
h arm which
enable it to accommodate two interferometers, one with 4 km l
ong arms (des-
ignated H1) and one with 2 km arms (H2). The orientation of the
Hanford
site was chosen to be as closely aligned (modulo 90
◦
) to the Livingston site as
possible, consistent with the earth’s curvature and the nee
d for the sites to be
level; this maximizes the common response to a signal, given
the quadrupolar
form of the anticipated gravitational waves. The arms have a
n included angle
of 90
.
000
◦
. The locations and orientations of the two LIGO sites are giv
en in
Table 1.
The observatories have a support infrastructure of clean ro
oms, preparation
laboratories, maintenance shops, and computer networking
for control, data
1
A shorter period of simultaneous observations between TAMA
, GEO, and LIGO
also took place during the period of this science run. That eff
ort will be documented
elsewhere.
8
acquisition and analysis, as well as offices for site staff and m
eeting spaces for
larger gatherings. The vacuum system can be divided into two
main pieces: the
4 km beam tube arms (through which the laser beams pass betwee
n the ver-
tex and end test masses), and the vacuum chambers that house t
he suspended
optics and associated equipment. The vacuum tubing for the a
rms[13,14] is
1.2 m in diameter, fabricated from 3 mm thick 304L stainless s
teel, processed
to reduce the outgassing to very low levels (1 to 8
×
10
−
14
mbar
·
L
·
s
−
1
cm
−
2
).
Expansion bellows are placed periodically along the arms. A
n extended bake
at elevated temperature was used to remove adsorbed water. T
he tubing is
supported by a ground-level concrete slab, protected by a co
ncrete cover, and
aligned to centimeter accuracy[15]. Ion pumps at 2 km interv
als and liquid
nitrogen cooled cryogenic traps where the arms enter the bui
ldings at the ver-
tex, end, and midstations maintain the base pressure in the a
rms between 10
−
8
and 10
−
9
mbar, with the residual gas being mainly hydrogen. This pres
sure
is sufficient to put the residual gas scintillation well below
the LIGO design
sensitivity.
The seismic isolation system, test masses, and other interf
erometer optics are
housed in vacuum chambers at the vertex, mid-stations (at Ha
nford), and
end-stations. Large gate valves where the beam tubes enter t
he buildings al-
low the vacuum chambers to be isolated from the beam tubes and
brought
to atmospheric pressure for work on the suspended optics whi
le maintaining
the vacuum in the 4 km arms. The vacuum chambers have large doo
rs to
aid in the access to install and align optics. When the chambe
rs are at at-
mospheric pressure they are purged continuously with clean
(Class 10) dry
air. They have numerous viewports (for laser beams to enter a
nd exit the
vacuum system and for video camera monitoring of the interio
r components)
and electrical feedthroughs. The pumping system includes r
oughing pumps
with Roots blowers, and hydrocarbon-free turbopumps. Only
ion pumps and
cryogenic traps are used when the interferometers are opera
ting. The vacuum
chambers are fully instrumented with gauges and residual ga
s analyzers; pres-
sures range between 4
×
10
−
8
and 3
×
10
−
9
mbar. All materials used in the
vacuum chambers and for the installed detector equipment ar
e carefully pro-
cessed and screened to minimize the amount of hydrocarbons i
ntroduced into
Table 1
Location and orientation of the LIGO detectors. Note that th
e Livingston Observa-
tory is rotated by
∼
90
◦
with respect to the Hanford Observatory, when the earth’s
curvature is taken into account.
LIGO Observatory:
Hanford
Livingston
Vertex Latitude
46
◦
27’18.5” N
30
◦
33’ 46.4” N
Vertex Longitude
119
◦
24’ 27.6” W
90
◦
46’ 27.3” W
Orientation of X arm
324
.
0
◦
(NW)
252
.
3
◦
(WSW)
9
Fig. 1. Schematic layout of a LIGO interferometer.
the vacuum system as a precaution against mirror contaminat
ion[16].
The basic optical configuration of each LIGO detector is that
of a power-
recycled Michelson interferometer with resonant arm cavit
ies, shown in Fig-
ure 1. Gravitational waves produce strains in space. The arm
cavity mirrors
serve as the inertial test bodies (test masses), which move i
n response to these
strains. For example a sinusoidal wave incident on the plane
of the interfer-
ometer will cause an apparent shortening of the optical path
along one arm
and a lengthening along the other arm, and this process rever
ses half a cycle
later in the signal evolution. Laser light is incident from t
he bottom-left on
the beamsplitter, which divides it and sends it to low-loss c
avities in the arms.
The transmission of the input mirror in each cavity is much la
rger than the
losses in the cavity, and thus when the cavities are on resona
nce, almost all
of the light is returned to the beamsplitter. The beamsplitt
er is held so that
the light emerging from the antisymmetric port of the interf
erometer (right)
is at a minimum, and almost all of the light is reflected back to
ward the laser.
The power-recycling mirror forms a resonant optical cavity
with the interfer-
ometer, causing a build-up of power in the recycling cavity.
The arm cavity
mirrors serve as the inertial test bodies (test masses), mov
ing in response to
the gravitational wave.
3.1 Laser
Each interferometer is illuminated with a continuous-wave
Nd:YAG laser op-
erating in the TEM
00
Gaussian spatial mode at 1064 nm, and capable of 10
W output power[17]. A small portion of the beam is used to stab
ilize the
10
laser frequency using a reference cavity in an auxiliary vac
uum chamber (Fig-
ure 2). The beam going to the reference cavity is double-pass
ed through an
acousto-optic modulator driven by a voltage controlled osc
illator; this allows
an offset between the laser frequency and the reference cavit
y frequency to
permit the laser to follow the arm cavity length change due to
tidal strains.
This initial level of stabilization is at 0
.
1 Hz
/
Hz
1
/
2
or better in the gravita-
tional wave band. The main portion of the beam is passed throu
gh a 45 cm
path length triangular cavity to strip off non-TEM
00
light and to provide pas-
sive filtering of the laser intensity noise with a pole freque
ncy of 1.5 MHz (0.5
MHz at Livingston for the S1 run). An intensity noise control
system designed
to reduce relative intensity fluctuations below 10
−
8
Hz
−
1
/
2
was only partially
implemented during the S1 run, leaving the intensity noise a
t approximately
10
−
7
Hz
−
1
/
2
level. Electro-optic modulators impress radio-frequency
sidebands
on the light at 24.5 and 33 MHz (29.5 and 26.7 MHz for H2) for sen
sing respec-
tively the interferometer, and suspended mode cleaner[18]
degrees of freedom.
The design for the LIGO interferometers is for 8 W to be incide
nt on the
mode cleaner. However, the commissioning of the instrument
for high input
power was not completed at the time of S1, and the powers incid
ent on the
mode cleaner had been adjusted (through the use of attenuato
rs and reduced
laser power) to approximately 1 W for H1 and L1 and approximat
ely 6 W for
H2.
3.2 Input Optics
After the laser beam enters the vacuum system, it passes thro
ugh a set of
input optics to condition it before it passes to the main inte
rferometer. First,
it passes through a mode cleaner – a
∼
24 m path length triangular ring cavity
with a finesse of
∼
1350, formed from separately suspended mirrors. This cavit
y
stabilizes the beam size, position and pointing. It also blo
cks the 33 (or 26.7)
MHz sidebands, but transmits the 24.5 (or 29.5) MHz sideband
s used for the
interferometer sensing, which are at multiples of the mode c
leaner free spectral
range. In addition, it serves as an auxiliary reference for t
he laser frequency
control servo, reducing frequency noise in the transmitted
laser light to the
10
−
3
Hz
/
Hz
1
/
2
level for the S1 run parameters. After the mode cleaner, the
beam passes through a Faraday isolator, which diverts light
returning from the
interferometer onto a photodetector. This prevents the ret
urning light from
reaching the laser and causing excess noise, and allows the c
ommon-mode
motions of the test-mass mirrors to be sensed. Finally, the b
eam passes through
an off-axis telescope formed by three suspended mirrors, whi
ch expands the
beam to match the
∼
4 cm (1
/e
2
radius in power) mode of the arm cavities.
11
Fig. 2. Simplified schematic of laser stabilization. EOM: El
ectro-Optic Modulator;
AOM: Acousto-Optic Modulator; VCO: Voltage Controlled Osc
illator; PD: Photo
Diode; PMC: Pre-Mode Cleaner; IOO: Input Optics; LSC: Lengt
h Sensing/Control
system
wideband
phase
correcting
EOM
LIGO
10-W laser
reference
VCO
cavity
EOM
amp.
input (from IOO)
tidal
input (from LSC)
pre-mode
(PMC)
PMC
amp.
to input optics
(IOO)
pwr. stab. input
(from IOO)
POWER
ADJUST
pwr.
stab.
PD
cleaner
cavity
reference
pwr. stab. amp.
AOM
tidal stab. amp.
reference
cavity
PD
freq. stab. amp.
SLOW
FAST
thermal enclosure
3.3 Interferometer Optics
The main interferometer optics[19,20] are fabricated from
high-purity fused
silica, 25 cm in diameter and 10 cm thick (except the beamspli
tter which is
4 cm thick). Radii of curvature of the cavity optics are chose
n so that the
arm cavities have a stability
g
= (1
−
L/R
1
)(1
−
L/R
2
) (
L
is the cavity
length and
R
n
are the radii of curvature of the two cavity mirrors) of 0.33 (
H1
and L1) or 0.67 (H2), to minimize the excitation of higher ord
er transverse
modes by separating them in frequency from the laser frequen
cy and its RF
modulation sidebands. The surface figure accuracy of the pol
ished optic is
better than 1 nm; the coatings have a thickness uniformity th
at holds their
contribution to the apparent surface flatness negligible. T
he coatings have a
power absorption less than 1 ppm and scatter less than 70 ppm.
All optics are
wedged (typically about 2 degrees) to reduce the possibilit
y of stray reflections
interfering with the main beam and to give access to samples o
f the light inside
the interferometer. Transmission of the input mirrors to th
e arm cavities is
2.7% and the end mirrors have a transmission of approximatel
y 12 ppm, to
give an arm cavity pole frequency of 85 Hz ( 170 Hz for H2). The b
eamsplitter
reflectivity was specified as 50
±
0
.
5%. The recycling mirror transmission is also
2.7%, to give a design recycling factor (or increase in the ci
rculating power)
of
∼
50 for the optics as designed and at full power.
During the S1 run, the low light input power led to the optical
configurations
of the three interferometers operating away from their desi
gn point. At full
operating power, absorption of light in the substrate and co
ating of the input
mirrors for the arm cavities is expected to create significan
t thermal lensing.
As a result, the curvature of the recycling mirrors was figure
d to compensate
for this anticipated thermal lensing. Since the incident la
ser power in the H1
and L1 interferometers was significantly under the design le
vel, the lack of
12
thermal lensing makes the recycling cavities slightly unst
able for the modula-
tion sideband light. This has little effect on the carrier rec
ycling gain (since the
carrier spatial mode is stabilized by resonance in the arm ca
vities) but reduces
the transmission of sideband light to the antisymmetric por
t by more than a
factor of 10. This further reduces the main differential arm l
ength sensitivity
in the high frequency region where shot noise is expected to b
e dominant.
In the case of the 2 km interferometer H2, although it was rece
iving nearly
the design input laser power, an out-of-specification anti-
reflection coating on
the input mirror of one arm caused excess loss in the recyclin
g cavity and
reduced the recycling gain for the carrier by more than a fact
or of two. As a
result it also did not develop the required thermal lens and i
ts transmission
of sidebands to the dark port was also degraded by a similar fa
ctor. These
limitations contributed to the relatively high noise level
of the instruments
seen in the sensing-noise limited regime (f
>
200 Hz).
3.4 Suspensions
Each interferometer optic is suspended as a pendulum from vi
bration-isolated
platforms to attenuate external disturbances in the gravit
ational wave band;
see Figure 3 for a schematic drawing. The suspension fiber is a
steel piano
wire, loaded at approximately 40% of its yield stress, passi
ng under the optic
as a simple loop. Small, notched glass rods are glued to the si
de of the optic
a few millimeters above the center of mass to define the suspen
sion point and
minimize frictional losses. The normal modes of the test mas
s optic suspen-
sion are approximately 0.74 Hz (pendulum mode), 0.5 Hz (yaw m
ode), 0.6
Hz (pitch mode), 12 Hz (bounce mode), 18 Hz (roll mode) and mul
tiples of
345 Hz (violin mode). Thermal noise is managed in interferom
etric gravita-
tional wave detectors by placing resonances above or below t
he detection band
when possible, and by choosing materials and assembly techn
iques which yield
high resonance
Q
’s[21]. This gathers the thermal noise power into a narrow
band and lowers the values on either side of the resonance. In
the case of the
suspensions, high resonance
Q
’s (measured to be typically 2 to 4
×
10
5
) in all
suspension modes yield a negligible level of off-resonance t
hermal noise for the
S1 sensitivity.
The suspension system also provides the means for applying c
ontrol forces
and torques to align the mirrors and hold the interferometer
in resonance.
Four small Nd:Fe:B magnets are attached to the back of the mir
ror using
aluminum stand-offs and a vacuum compatible epoxy, with alte
rnating po-
larities to reduce coupling to environmental magnetic field
s. The suspension
structure supports voice coils on ceramic forms near the mag
nets to produce
control forces. Each of these assemblies also incorporates
an LED/photodiode
pair arranged so that the magnet partially obstructs the pat
h between them
13
Fig. 3. LIGO Suspension
(a “shadow sensor”). This provides a read-out of the longitu
dinal position of
the magnet with a noise level of approximately 10
−
10
m
/
Hz
1
/
2
in the gravita-
tional wave band. Similar magnets are attached to the sides o
f the optic and
a shadow sensor/voice coil assembly acts on one of these to da
mp sideways
motion.
The magnet and coil actuators are driven by several sensors,
via servo con-
trollers, allowing control of their positions and orientat
ions with respect to
both the local structures and the globally-measured length
s and angles. Local
damping of the modes of the suspension is provided by feeding
appropriately
filtered and mixed signals from the shadow sensors to the coil
s to create a
damping force near the pendulum frequencies. Signals from i
nterferometric-
based wavefront sensors and optical levers (described belo
w) are also applied
to maintain the pointing of the test mass. Lastly, the interf
erometer length sig-
nals are applied to acquire resonance and hold the operating
lengths for the
interferometer to within
∼
10
−
13
m rms. The suspension controllers, which
combine and filter these signals appropriately, were of two s
tyles during S1:
an original analog system with some digital gain and filter co
ntrols (for H2 and
L1), and a system with the signal processing performed digit
ally[22] (H1). In
all cases, a significant low-frequency noise contributor wa
s the final amplifier,
which for S1 had to deliver stronger control forces than thos
e expected for
the final configuration. This then compromised the gravitati
onal wave-band
performance.
Thermal noise internal to the mirrors is minimized by mainta
ining high
Q
’s in
all the internal modes. The fused silica internal losses are
anticipated to make
the dominant contribution to thermal noise. However, the di
electric coating
on the mirror will also contribute noticeably because of its
proximity to the
14
beam[23]. The attachments to the mirror for the suspension a
nd the magnets
can degrade the individual modal
Q
’s but, because of their distance from the
front surface of the optic, their effect on thermal noise is ne
gligible. In-situ
measurements of Q’s typically range from 2
×
10
5
to 1
.
6
×
10
7
, depending on
the mode. Calculations indicate that the thermal noise is ne
ar the design goal
and thus negligible for the S1-run sensitivity.
3.5 Seismic Isolation
The vibration isolation systems are four layer passive isol
ation stacks[24]. The
final stage in each vacuum chamber is an aluminum optical tabl
e that holds
the optic suspensions. Each optical table is supported by fo
ur legs. Each leg
consists of a series of three heavy stainless steel cylinder
s, supported by coil
springs made with phosphor bronze tubing containing inner c
onstrained layers
which are sealed from the vacuum via electron-beam welding.
The transfer
function of ground motion to table motion shows a series of br
oad peaks
between 1.5 and 12 Hz, representing the normal modes (typica
l
Q
∼
10
−
30) of the masses moving on the springs, followed by a steep fa
lloff above
the highest resonance. The total attenuation reaches a valu
e of 10
6
by about
50 Hz. The high
Q
’s of the resonances in the 1.5 to 12 Hz band presents a
particular problem at LLO, where they amplify anthropogeni
c ground noise
in this frequency range, and cause difficulties in locking dur
ing daylight hours.
A planned six degree-of-freedom external active isolation
system to cope with
this excitation was not in place during S1. The support point
s for the seismic
isolation stack penetrate the vacuum chamber through bello
ws that decouple
the seismic isolation stack from vacuum chamber vibrations
and drift. External
coarse actuators at the support points permit translations
and rotations to
minimize the control forces that are needed to align the opti
cs during the
initial installation and to compensate for any long-term se
ttling.
In addition, the systems at the ends of the arms are equipped w
ith a fine
actuator aligned with the arm that can translate the entire a
ssembly (seis-
mic isolation stack and optic suspension) by approximately
±
90
μ
m over the
frequency range from DC to
∼
10 Hz. This system is used during the inter-
ferometer operation to compensate for earth tides, using a s
imple predictive
model and a very slow feedback from the differential and commo
n mode arm
length controls. At LLO, an additional microseismic feed-f
orward system[25]
was used to reduce the length fluctuations of the arms at the mi
croseismic fre-
quency (approximately 0.16 Hz). Also, the L1 detector’s fine
actuators were
used together with seismometers in a beam-direction active
seismic isolation
system at each test mass chamber, which reduced seismic exci
tation of the
most troublesome stack modes by a factor of
∼
5.
15
3.6 Length and Angle Control
There are four longitudinal degrees of freedom that must be h
eld to allow the
interferometer to function: the two arm lengths are held at t
he Fabry-Perot
cavity resonance condition, the beamsplitter position is s
et to maintain the
light intensity minimum at the antisymmetric port and the re
cycling mirror
position is positioned to meet the resonance condition in th
e recycling cavity.
These lengths are sensed using RF phase modulation sideband
s on the incident
light in an extension[26] of the Pound-Drever-Hall techniq
ue. The modulation
frequency was chosen so that the phase modulation sidebands
are nearly an-
tiresonant in the arm cavities; the carrier light is strongl
y overcoupled so that
0.97 of the light is reflected on resonance, and it receives a
π
phase shift on
reflection. By making the recycling round trip cavity length
an odd number
of RF half-wavelengths, the recycling cavity can be simulta
neously resonant
for the carrier and sidebands. A small length asymmetry (
∼
30 cm) is intro-
duced between the beamsplitter and the two input test masses
to couple the
sideband light out the dark port.
Three interferometer output beams (Figure 4) are used to det
ermine the lon-
gitudinal degrees of freedom[27], which are best thought of
as two differen-
tial motions (arm cavities or strain readout, and the Michel
son), and two
common-mode motions (common mode ‘breathing’ of the arm cav
ities, and of
the power recycling cavity). A photodiode signal at the anti
symmetric port is
demodulated with the 90
◦
quadrature of the modulation drive to give a signal
proportional to the difference in arm cavity lengths (differe
ntial arm length).
A second photodiode monitors the light reflected from the rec
ycling mirror
(separated from the incident beam by a Faraday isolator); it
is demodulated
in phase with the modulation drive and is primarily sensitiv
e to the average
of length of the two arm cavities (common mode arm length). Th
e third pho-
todiode monitors light from inside the recycling cavity, pi
cked off from the
back (anti-reflection-coated) side of the beamsplitter wit
h the aid of the small
wedge angle in the substrate. The in-phase signal is primari
ly sensitive to the
recycling cavity length, while the quadrature phase is sens
itive to the Michel-
son path difference from the beamsplitter to the input test ma
sses. One major
deviation from the final interferometer design during S1 was
that attenuators
were placed in front of the antisymmetric port photodiodes o
n all three inter-
ferometers, reducing the effective power used in each interf
erometer to about
50 mW instead of the 6 W nominal value. These attenuators prot
ected the
photodiodes from saturation, and possible damage, during t
he commission-
ing phase before the complete mirror angular controls were i
mplemented and
when large fluctuations in the power on the photodiodes were p
resent. This
had a particularly significant impact on the performance in t
he high frequency
region (above a few hundred Hz), where the low effective light
level combined
with the reduced sideband efficiency noted above to cause the n
oise to be well
16
Fig. 4. Schematic drawing of a LIGO interferometer showing l
aser, input light mode
cleaner, and the locations of the photodiodes (S
xxx
) used to sense and control the
resonance conditions.
L
1
and
L
2
are the arm cavity lengths; a gravitational wave
produces a differential signal of the form (
L
1
−
L
2
), and (
L
1
+
L
2
) is a sensitive
measure of the laser frequency noise. The Michelson degrees
of freedom are differ-
ential (
l
1
−
l
2
) and common-mode (
l
1
+
l
2
), the latter measured with respect to
the recycling mirror. PBS: Polarizing Beam Splitter. AOM: A
cousto-Optic Mod-
ulator. PC: Pockels Cell. VCO: Voltage Controlled Oscillat
or. S
mc
: Signal, Mode
Cleaner. S
ref
: Signal, Reference Cavity. S
refl
: Signal, reflected light. FI: Faraday Iso-
lator. PRM: Power Recycling Mirror. S
prc
: Signal, Power Recycling Cavity. S
anti
:
Signal, Antisymmetric Port. BS: Beamsplitter. ITM: Input T
est Mass. ETM: End
Test Mass.
above the design level.
The signals from these three photodiodes, appropriately de
modulated and fil-
tered, are used to control the lengths and hold the interfero
meter in resonance.
The high frequency portion of the reflected photodiode signa
l S
refl
is fed back
(via an analog path at Hanford for S1, digital at Livingston)
to the mode
cleaner and laser to stabilize the input laser frequency to t
he average arm
length. The signals S
prc
and S
anti
from the other two photodiodes are used
to control the positions of the interferometer optics. The d
emodulated signals
from all three photodiodes are whitened with an analog filter
, digitized with
a 16 bit ADC operating at 16384 samples per second, and digita
lly filtered
with the inverse of the analog whitening filter to return them
to their full
dynamic range. A dedicated real-time signal processor comb
ines these error
signals via a matrix (whose coefficients are adjusted in real t
ime during the
lock acquisition process) to form appropriate control sign
als, filters them, and
sends the results to combinations of optics to control the in
terferometer. It
17
also passes the photodiode signals (error signals) and the f
eedback signals to
the data acquisition system. The flexibility of the digital c
ontrol system to re-
spond in changes to the interferometer response function du
ring the ‘locking’
process as a function of sensed light levels, and to allow spe
cialized filters to be
implemented on the fly, has been crucial to the ability to acqu
ire lock on the
interferometers[28], to aid in the commissioning, and ulti
mately to suppress
noise in the control systems.
As noted above, an ensemble of optical levers and wavefront s
ensors is de-
signed to sense and control the angular degrees of freedom of
the suspended
optics in the main part of the interferometer[29,30]. Each l
arge (25 cm) optic
is equipped with an optical lever, consisting of a fiber-coup
led diode laser and
a (position sensitive) quadrant photodiode, which is inten
ded to hold the optic
stable while the interferometer is unlocked. These compone
nts are mounted on
piers outside the vacuum system and operate through viewpor
ts at distances
between 1 and 25 meters from the optic; their long-term stabi
lity and inde-
pendence from the interferometric sensing system allows a m
anual alignment
to be maintained continuously.
The full instrument design includes a wavefront sensing con
trol system to
optimize the alignment during operation. Quadrant photodi
odes are placed at
the output ports of the interferometer, in the near field and (
via telescopes)
in the far field. The photocurrents are demodulated as for the
length control
system, and sums and differences can be formed to develop a com
plete set of
alignment information which is then used to control the mirr
or angles, using
the suspension actuators. However, at the time of the S1 run,
this system
was only partially commissioned, and only the mode cleaner a
nd two degrees
of freedom of the interferometer, the differential pitch and
yaw of the end
test masses (cavity end mirrors), were controlled by wavefr
ont sensors. As
an interim measure, the incomplete wavefront sensing was co
mplemented by
signals from the optical levers during operation. However,
the optical lever
angular sensing noise is much greater than that for the wavef
ront sensors.
Even after careful control-law shaping, the optical levers
remained one of the
principal contributors to the low-frequency noise of the in
strument for S1.
Baffles to capture stray light are placed along the 4 km beam tub
es, and at
specific places near the optics inside the vacuum chambers, t
o reduce the pos-
sibility of a scattered beam or one from an intentional wedge
in the optics from
recombining with the main beams. Some of the baffles for the fina
l installation
were not in place for the S1 run, but calculations indicate th
at this should not
have been a source of noise at our present level of sensitivit
y.
A noise model of the instruments summarizes the limits to the
performance of
the interferometer at the time of this science run. The model
for the Livingston
detector is shown in Figure 5. In general, the contributions
are evaluated by
18
Fig. 5. A frequency-domain model of the noise sources at the t
ime of the S1 run for
the Livingston (L1) detector. The noise sources, discussed
in the text, are assumed to
add in quadrature. The actual noise curve is also shown, alon
g with the performance
expected for the instrument when working at the design level
10
1
10
2
10
3
10
4
10
−22
10
−20
10
−18
10
−16
Strain (h/rtHz)
Frequency (Hz)
S1 L1 Noise spectrum, 08/29/02
Optical Lever Servo
Suspension Coil Driver
Shot and Electronic Noise
Performance Goal
measuring a source term (e.g., laser frequency noise), and m
easuring a coupling
function (e.g., the transfer function from an intentional f
requency modulation
to the response in the strain channel), and then multiplying
these two together
to make a prediction. In some cases, analytical models are us
ed (for example
the mechanical
Q
of the suspension systems is measured and then used in a
model of the thermal noise contribution). For this model, al
l the terms are
considered to be independent, and the noises are added as the
square root
of the sum of the squares. Many sources of noise have been mode
lled; this
figure only shows those that limit the present performance. T
he model explains
the overall instrument noise performance well, and subsequ
ent commissioning
efforts have shown that reductions in the leading noise terms
also leads to the
anticipated reduction in the overall instrument noise.
3.7 Simulations
In addition to subsystem dynamics and control models, two si
mulations played
a significant role in the design and commissioning of the LIGO
detectors. The
first is an FFT-based optical propagation code[31] that mode
ls the power-
19
recycled Michelson interferometer with Fabry Perot arms. T
his code was used
to develop the specifications for the interferometer optics
, and has been used
in comparisons with commissioning data to evaluate the perf
ormance of the
optics as installed. The second simulation is an end-to-end
time-domain sim-
ulation of the LIGO interferometers[32]. This model includ
es a modal-based
optical propagation, accurate modeling of the electronic f
eedback, simplified
models for the suspension systems, and typical noise inputs
. This model proved
to be invaluable in developing the lock acquisition softwar
e.
3.8 Environmental Monitoring
A system of auxiliary sensors is installed at each LIGO site t
o monitor possible
environmental disturbances. The Physics Environment Moni
tor system (PEM)
contains seismometers and tiltmeters to monitor low freque
ncy ground distur-
bances, accelerometers and microphones to monitor higher f
requency mechan-
ical disturbances, magnetometers to monitor magnetic field
s that might affect
the test masses, and monitors of the line power. Sensors are p
resent in all
buildings and near all key sensitive components. They have b
een used to e.g.,
help identify sources of acoustic and electromagnetic coup
ling, and to help de-
sign improvements to the apparatus; as the instrument sensi
tivity improves,
they will be used as veto signals in the astrophysical analys
es. Planned cosmic
ray detectors and rf monitors were not operational at the tim
e of the S1 run.
3.9 Control and Data Systems
Supervisory control of the interferometers is accomplishe
d using EPICS (Ex-
perimental Physics and Industrial Control System[33]). EP
ICS establishes a
communications protocol within a non-hierarchical comput
er network and pro-
vides an operator interface from networked workstations lo
cated in the control
room. Processors distributed in all electronics racks can m
odify amplifier gains,
offsets, filtering, on/off controls, etc., allowing either ma
nual or automated
(scripted) control of the state of the electronics. The EPIC
S processors also
interface to analog-to-digital converters to provide moni
tors for the electron-
ics inputs and outputs. Each interferometer has approximat
ely 5000 EPICS
variables (either control or monitor points). EPICS also pr
ovides tools for cap-
turing and restoring the state of the instrument to ensure th
at this complex
instrument can be reliably brought to a known configuration.
The data acquisition and control system collects signals fr
om the interferom-
eter and from the environment, and delivers signals for the l
ength and angle
controls. VME-based converters and processors are used, an
d acquisition sys-
tems are placed in the vertex building and the mid- and end-st
ations. Analog
20
signals are digitized with 16-bit resolution. Fiber optics
are used to link the
instrument racks together, and a shared memory approach all
ows data to be
collected and shared by a number of systems over the multi-km
distances. The
data are collected with 16-bit resolution. The complete dat
a are formatted into
the standard data ‘Frames’ (a format used by all of the interf
erometric grav-
itational wave community) and initially are stored on spinn
ing media for a
quick ‘look-back’ buffer of roughly two weeks. All data are ar
chived to tapes
for later analysis. Reduced data sets also in the standard Fr
ame format, con-
figured for a given science run, are produced as well; these se
rve most analysis
needs.
It is important that the data acquisition system accurately
time-stamp the
data it records. The fundamental timing for both sites is der
ived from GPS
receivers located at each building (vertex, mid and end). A 2
22
Hz (approxi-
mately 4.2 MHz) clock signal is generated from the GPS as well
as a 1 pulse-
per-second synchronization signal. Together these are use
d to synchronize the
data collected by the various processors. Ramp signals are u
sed to monitor any
timing errors and alarms are set for the operators. This moni
toring has proven
useful during S1. It showed that the timing was subject to jum
ps (typically
10’s of milliseconds, but sometimes larger) when the length
control system
processors were rebooted, with the consequence that some S1
data had to
be eliminated from some analyses because of uncertain timin
g (These timing
jumps have since been cured, and a redundant and independent
atomic clock
reference is being implemented for the future).
3.10 Diagnostics and Monitoring
Two closely related systems, the Global Diagnostic System a
nd the Data Mon-
itoring Tool, provide the instrument operators and scienti
fic monitors with
tools for evaluating interferometer performance both duri
ng commissioning
and scientific running[34]. The Global Diagnostic System (o
r GDS) can ac-
cess data from any signal collected by the data acquisition s
ystem, including
test-point signals that can be stored for post-analysis if i
ndicated. It can dis-
play the time series and the power spectrum for individual si
gnals, and the
transfer function and coherence for pairs of signals. The GD
S also has the abil-
ity to apply arbitrary-waveform excitations to various tes
t points within the
interferometer control systems. These can be used to measur
e transfer func-
tions through stimulus-response testing. The data can be fil
tered, decimated,
calibrated, stored and recalled for comparisons.
The Data Monitoring Tool (or DMT) is a package of software com
ponents run-
ning on a set of processors on a dedicated network. A high-spe
ed connection
to the data acquisition system makes the full data set availa
ble with only a
21
one-second latency. The emphasis in the DMT is on relatively
simple measures
of instrument performance applied to the full data stream in
realtime. Thus it
can give the operators and scientific monitors rapid feedbac
k about interfer-
ometer performance. These include such measures as the non-
stationarity and
burst-like behavior of various types in the interferometer
outputs, band-limited
rms amplitudes of interferometer outputs and environmenta
l monitors, mon-
itors of calibration lines, histograms to monitor the gauss
ianity of the data,
and real-time estimates of detector sensitivity to neutron
star binary inspiral
events. The DMT is also an important element of the data analy
sis process,
analyzing the auxiliary channels for veto signals in parall
el with the strain
channel analysis.
3.11 Data Analysis System
To analyze the large volume of data generated, LIGO has devel
oped the LIGO
Data Analysis System (LDAS). The LDAS provides a distribute
d software en-
vironment with scalable hardware configurations to provide
the computational
needs for both on and off site data analysis. The architectura
l design of the
system is based on the concept of multiple concurrent data an
alysis pipelines
in which data is fed into the pipeline as it is collected and pr
oceeds down the
pipeline where necessary signal analysis procedures are ap
plied depending on
the particular type of analysis that is being carried out[35
].
LDAS is complemented by the LIGO/LSC Algorithm Library, whi
ch is a set
of
C
-language routines that can run under LDAS or be used indepen
dently.
They are carefully vetted to ensure that the algorithms and r
esults are correct.
The LDAS distributed software environment is composed of ro
ughly 12 mod-
ules called LDAS Application Programming Interfaces (APIs
), each of which
is a separate process under the Unix operating system. Each m
odule is de-
signed to carry out the multitude of steps associated with ea
ch unique pipeline.
For example, one module has computational elements for read
ing and writing
LIGO channel data in the Frame format, another module has com
putational
elements for signal processing in the time or frequency doma
in, and another
module has computational elements necessary to perform par
allel analysis
across a cluster of tightly networked CPUs[36]. Upon comple
tion of the data
analysis pipeline, data products and results are stored to d
isk or inserted into
the LIGO relational database. The database has tables desig
ned to capture
results associated with detector characterization, on lin
e and off line astro-
physical searches and multi-detector analyses. The softwa
re can be scaled to
run on a wide variety of computing hardware.
During the first LIGO Science Run, the LDAS at the LIGO Observa
tories and
22
data analysis centers located at Caltech and MIT, LDAS at oth
er institutions,
and other configuration-controlled computational systems
were operated with
commissioning configurations of the hardware and software.
The software was
in the late stage of beta development, having a complete set o
f modules.
The hardware systems consisted of a complement of servers wi
th tens of ter-
abytes of disk storage for the raw data and the LIGO database,
along with
scaled down computation centers with approximately 200 meg
aflops of aggre-
gate computational performance between them. In its final co
nfiguration, the
LDAS hardware will include upgrades to the current servers,
and expanded
high performance computation clusters with over two teraflo
ps of aggregate
computational performance. In addition, new tape storage s
ystems will be put
on line which will provide adequate storage at the observato
ries for six months
of local data and storage for all of the data at Caltech; this i
s where the data
for the multiple detectors are brought together. The softwa
re is expected to
double in performance as we upgrade from beta versions to the
first completed
version later this year. In addition, the software is being a
dapted to support
Grid Computing technology and security protocols allowing
for LIGO data
analysis once the computational Grid is deployed in the near
future[37].
3.12 LIGO Data
The full data stream from each of the LIGO interferometers co
nsists of several
thousand channels, recorded at rates from 1 Hz to 16384 Hz wit
h a total data
rate of 5MB/s per interferometer. These channels include EP
ICS process vari-
ables that define the state of the interferometer, signals fr
om environmental
monitors, signals from auxiliary servos in the interferome
ter (for example, op-
tical lever signals), as well as the main gravitational wave
signal. For the servos
not operational during S1, the corresponding data channels
were recorded, but
of course they contain only zeros. The non- gravitational wa
ve data channels
can be used in a number of ways:
(1) They can be used to determine the operational “health” of
the interfer-
ometer (how well it was aligned, whether large offsets were pr
esent in any
servos, etc.).
(2) They can be used to regress noise from the main gravitatio
nal wave chan-
nel (for example, measurements of the laser frequency noise
can be used
to correct the gravitational wave channel to remove any resi
dual effects
from laser frequency noise coupling to mismatches in the arm
s).
(3) They can be used to veto non-gaussian noise in the interfe
rometer (for
example seismometer data could be used to keep noise from imp
ulsive
seismic disturbances from being misinterpreted as a gravit
ational wave).
23
At this point in the commissioning, few of these techniques h
ave been explored
and developed. In part, this is because the majority of the no
ise sources in the
interferometer at present are attributable to electronic n
oise entering through
imperfect tuning, and consequently, few of the auxiliary ch
annels are expected
to be useful. The DMT capabilities to perform this analysis w
ere exercised in
preparation and performed very well.
The main signal for the analysis to search for gravitational
wave signals is
the output of the photodiode at the antisymmetric port, demo
dulated in the
quadrature phase at 24.5 MHz (H1 and L1), or 29.5 MHz (H2). Thi
s analog
signal is amplified and digitized. An analog filter whitens th
e signals before
digitization, and a precise inverse of this filter ‘de-white
ns’ the signal in the
digital domain, to best take advantage of the dynamic range a
nd noise in
the Analog-to Digital Converter (ADC). Since it is the error
point in the
servo control system which holds the differential arm length
, its interpretation
requires correction for the loop gain of the servo. This sign
al represents the
phase difference of the light from the two arms, filtered only b
y a roll-off
at high frequencies because of the arm cavity storage time, a
nd while the
interferometer is operating, is a continuous measure of the
differential strain
between the two arms and thus potentially of gravitational w
ave signals.
4 The GEO Detector
The GEO Detector is situated at the perimeter of an agricultu
ral research sta-
tion to the south-east of Hannover, Germany; see Table 2. The
buildings are
intended to be just sufficient to accommodate the instrument a
nd its acqui-
sition and control hardware. Data recording, and much of the
operation and
on-line monitoring of the instrument, will be performed at t
he Max Planck In-
stitute in downtown Hannover, once continuous science oper
ation is underway.
A microwave link maintains a high-bandwidth dedicated conn
ection between
the two.
Table 2
Location and orientation of the GEO 600 detector. Note that t
he arms form an
angle of 94
◦
19’ 53”. This deviation from perpendicular has negligible e
ffect on the
sensitivity.
Vertex Latitude
52
◦
14’ 42.5” N
Vertex Longitude
9
◦
48’ 25.9” E
Orientation of North arm
334
.
1
◦
(NNW)
Orientation of East arm
68
.
4
◦
(ENE)
One central building (13 m
×
8 m in size) and two end buildings (6 m
×
3
24
m) accommodate the vacuum chambers (2 m tall, 1m in diameter)
in which
the optical components are suspended. In the central buildi
ng, nine vacuum
chambers form a cluster which can be subdivided into three se
ctions to allow
mirror installation without venting the whole cluster. Thi
s arrangement allows
a minimum of down-time for a change of the signal-recycling m
irror (to change
the detector bandwidth). To avoid fluctuations of the optica
l path caused by
a time-varying index of refraction, all light paths in the in
terferometer are in a
high-vacuum system. For this purpose GEO 600 uses two 600 m lo
ng vacuum
tubes of 60 cm diameter which are suspended in a trench under g
round. A
novel convoluted-tube design, allowing a wall thickness of
only 0.8 mm, was
used to reduce weight and cost of the stainless-steel vacuum
tube.
The whole vacuum system, except for the mode cleaner and sign
al recycling
section, is pumped by four magnetically levitated turbo pum
ps with a pumping
speed of 1000 l/s, each backed by a scroll pump (25 m
3
/
h). Due to the use of
stainless steel with a low outgassing rate, a 2 day air bake at
200
◦
C and a 5
day vacuum bake at 250
◦
C, a pressure of 1
×
10
−
8
mbar can be achieved in
the tubes. Large gate valves allow the beam tubes to be tempor
arily closed off
and maintained under vacuum whenever the instrument vacuum
chambers are
opened for installation work. Additional dedicated pumpin
g systems are used
for the mode cleaner section and for the signal-recycling se
ction. The pressure
in the vacuum chambers is in the mid 10
−
8
mbar range. Great care was taken to
minimize contamination of the all-metal vacuum system by hy
drocarbons. For
this reason the seismic isolation stacks, which contain sil
icone elastomer and
other materials containing hydrocarbons, are sealed by bel
lows and pumped
separately. Furthermore, the light emitting diodes (LEDs)
, the photodiodes
and the feedback coils used as ‘shadow’ sensors and actuator
s in the pendulum
collocated damping and actuation systems are encapsulated
in glass.
The buildings of GEO 600 are split into three regions with diff
erent cleanroom
classes: the so-called gallery where people can visit and st
aff can work with
normal clothes, the inner section which has a cleanroom clas
s of 1000 and a
movable cleanroom tent installed over open chambers with a c
leanroom class
100.
4.1 Suspension and Seismic Isolation
Two different types of seismic isolation and suspension syst
ems are imple-
mented in GEO 600. The first one, used to isolate the mode clean
er optics,
consists of a double pendulum suspended from a pre-isolated
top-plate. To
avoid an excitation of the pendulum mode, four collocated co
ntrol systems
measure the motion of the intermediate mass with respect to a
coil-holder
arm which is rigidly attached to the top plate, and feed back t
o the mirror via
25