Gravity Waves1
Is Seeing Believing? Observation in Physics
Allan Franklin
Department of Physics
University of Colorado
In their recent paper, Observation of Gravitational Waves from a
Binary Black Hole Merger, the LIGO collaboration stated, This is
the first direct detection of gravitational waves. (Abbott, Abbott
et al. 2016, p. 061102-1, emphasis added).[endnoteRef:1] This was
to distinguish their result from those of Hulse and Taylor (Hulse
and Taylor 1975) and of Weisberg and Taylor (Weisberg and Taylor
1981), in which the decrease in the period of a binary pulsar was
used to establish, with a high degree of confidence the existence
of gravitational radiation as predicted by general relativity
(Weisberg and Taylor 1981, p. 1).[endnoteRef:2] The implication by
LIGO was that the latter result was not a direct observation. [1:
In this case the LIGO collaboration is applying the 5- criterion
for an observation. In high energy physics this is a requirement
for the use of observation, which is synonymous with discovery.
This will be discussed in more detail below. For a more detailed
discussion see (Franklin 2013, Prologue).] [2: The LIGO and binary
pulsar experiments would count as successful replications. ]
This raises several interesting questions. One might ask how one
can distinguish between direct and indirect observation and whether
that distinction is seen in the practice of science. One might also
ask whether a direct observation is in some way better than an
indirect observation and, if so in what way and why. Is a direct
observation more credible? Does it have more epistemic or
evidential weight? Does it provide more support for a theory or
hypothesis?
As Harry Collins (2017) has reported, and as will be discussed
below, the use of direct was the subject of considerable discussion
within the LIGO collaboration when the discovery paper was being
written and even earlier. Philosophers of science have also weighed
in of the question of direct observation.[endnoteRef:3] These views
range from the privileging of human sense perception, certainly a
direct observation, by Bas van Fraasen (1981, pp. 13-19) to what
might certainly be considered indirect observation of an entity by
Wilfred Sellars, to have good reason for holding a theory is ipso
facto to have good reason for holding that the entities postulated
by the theory exist (1962, p. 97). Thus, in Sellars view, the
correct prediction of the energy spectrum in decay by Enrico Fermis
theory of provided grounds for belief in the existence of the
neutrino, even though the neutrino itself is not involved in the
observation. This is because the neutrino is an essential part of
Fermis theory and is needed for its predictions. The argument is
that if a group of sentences is each essential to the prediction of
an experimental result, then observation of that result supports
both the conjunction and each of the statements individually. [3:
Typically, although not always, observation is taken to apply to
entities.]
In this essay I will begin with the extensive discussion
provided by of Dudley Shapere in his essay "The Concept of
Observation in Science and Philosophy, (1982). I will also present
episodes from the history of physics to examine the roles played by
observation and their justification, in an attempt to illustrate
and clarify the distinction between direct and indirect
observation. As we shall see, the question of direct versus
indirect observation did not arise until the 20th century when both
experimental apparatuses and analysis procedures became more
complex. The earlier discussions concerned unaided human
observation as opposed to instrumental observation. In the 20th and
21st century virtually all observations in physic s require
instrumentation. As we shall see, the meaning of direct observation
has changed.
I. Shapere and Direct Observation
Shapere defined direct observation as follows. x is directly
observed (observable) if:
1. information is received (can be received) by an appropriate
receptor; and
2. that information is (can be) transmitted directly, i.e.,
without interference, to the receptor from the entity x (which is
the source of the information). (Shapere 1982, p. 492).
He emphasized that what is directly observed, what counts as
information, what is an appropriate receptor, and what is direct or
undisturbed transmission depends on the current state of physical
knowledge. Shapere specifies three elements of the observation
situation. They are the theory of the source, the theory of
transmission, and the theory of the receptor. In some cases the
theory of the phenomenon may be involved in any of these three
elements. This raises the issue of the possible theory ladenness of
the observation. Can such an observation then be considered a test
of that theory? As Shapere notes, and as illustrated below, the
fact that in a particular case the theory under test is essential
to the design of an experiment, does not guarantee that the
experimental result will be in agreement with that theory. He also
remarks that an experiment may depend, in an essential way, on
background knowledge and information. Thus, in the solar neutrino
experiment, discussed below, the calculation of the expected
neutrino flux depends on knowledge of various nuclear interaction
cross sections and decay rates. It also depends on the structure,
temperature, and composition of the Sun.[endnoteRef:4] [4: It is
also clear that ultimately the information received must be
available in a form perceivable by a human being. ]
The center of Shaperes discussion is the experiment by Ray Davis
and his collaborators (1968) that was designed to detect solar
neutrinos, those emitted by the sun.[endnoteRef:5], [endnoteRef:6]
The experiment was designed to demonstrate that nuclear processes
were occurring in the central core of the sun.[endnoteRef:7] [5:
Davis won the 2002 Nobel Prize for pioneering contributions to
astrophysics, in particular for the detection of cosmic neutrinos.]
[6: The existence of the neutrino had been established earlier in
experiments by Frederick Reines, Clyde Cowan, and their
collaborators (Cowan, Reines et al. 1956), (Reines, Cowan Jr. et
al. 1960). The directness of this observation is discussed below.]
[7: There had been earlier doubts as to whether these processes
occurred. As Arthur Eddington remarked, I am aware that many
critics consider the conditions in stars not sufficiently extreme
to bring about the transmutationthe stars are not hot enough.... we
tell them to go and find a hotter place (Eddington 1927, p.
102).]
The physics of the experiment can be schematically summarized as
follows. The major source of energy production in stars such as the
Sun proceeds by the burning of hydrogen to produce helium (41H1
4He2 + 2 e+ + 2 e), where e+ are positrons and e are electron
neutrinos.[endnoteRef:8] This process actually takes place through
many nuclear interactions. For the purpose of Daviss experiment,
the most important of these results in the production of boron
(8B5) which subsequently decayed into beryllium (8B5 8Be4 + e+ +
e). These neutrinos had sufficient energy to induce the interaction
that Davis would use to detect the neutrinos (e + 37Cl17 37A18 +
e-). The argon 37 was radioactive and would subsequently decay
(37A18 37Cl17 + e+ + e). Davis proposed to use 100,000 gallons of
perchlorethylene (C2Cl4) as a target for the solar neutrinos. The
37A produced would be collected and its decay detected. This would
provide evidence for solar neutrinos. (For more details see
(Franklin 2001, chapter 8). [8: There are three kinds of neutrinos,
those associated with the electron, the muon, and the tau lepton,
e, , and .]
Shapere quotes an unnamed philosopher of science who declared
that "There is one thing which we can be sure will never be
observed directly, and that is the central region of the sun, or,
for that matter, of any other star (cited in Shapere 1982, p.
485)." Shapere notes that this seems quite reasonable because the
Sun has a radius of 689,000 kilometers and a core temperature of 15
million degrees Centigrade. On the other hand, he quotes two
astrophysicists who take a very different view. neutrinos originate
in the very hot stellar core, in a volume less than a millionth of
the total solar volume. This core region is so well shielded by the
surrounding layers that neutrinos present the only way of directly
observing it (Weekes 1969, p. 161) and There is no way known other
than by neutrinos to see into a stellar interior (Clayton 1968, p.
388).
John Bahcall who did much of the early theoretical work on both
the solar neutrino flux and on Daviss experiment gave the following
summary.
The principal energy source for main-sequence stars like the sun
is believed to be the fusion, in the deep interior of the star, of
four protons to from an alpha particle. The fusion reactions are
thought to be initiated by the sequence 1H(p,e+v)2H(p,)3He and
terminated by the following sequences: (i) 3He( 3He,2p) 4He; (ii)
3He(,)7Be(e-v)7Li(p,)4He; and (iii) 3He(,)7Be(p,)8B() 4He. No
direct evidence for the existence of nuclear reactions in the
interior of stars has yet been obtained because the mean free path
for photons emitted in the center of a star is typically less than
10-10 of the radius of the star. Only neutrinos, with their
extremely small interaction cross sections, can enable us to see
into the interior of a star and thus verify directly the hypothesis
of nuclear energy generation in stars (Bahcall 1964, p. 300).
The last points are crucial. A photon emitted at the center of
the sun will take approximately 10 million years to reach the
surface of the Sun, but its nature will be considerably changed.
Neutrinos, because they interact only weakly with are matter, are
unchanged as they travel from the Suns core to the
Earth.[endnoteRef:9] They will also travel at the speed of light.
[9: This is not strictly true. We now know that neutrinos of one
type can transform into other types of neutrino during this
journey. ]
At the time Shapere wrote his paper the observed solar neutrino
flux was only about one third of that predicted. This led to
serious questions concerning the calculation of the solar neutrino
flux by Bahcall and others and also questions about the ability of
Daviss detector to measure that flux accurately. The solar neutrino
problem, as it came to be known, was solved when physicists found
evidence for neutrino oscillations,[endnoteRef:10] the fact that
some of the neutrinos, in their passage between the core of the sun
and the detector on Earth, transformed from one type of neutrino,
the electron neutrinos emitted in the energy production processes,
can transform into two other types of neutrino, the muon and tau
neutrinos (for more details see (Franklin 2016, chapter 14)). This
accounts for the measured, lower-than-predicted neutrino flux
because Daviss experiment was sensitive only to electron neutrinos.
By the time solar neutrinos reached the Earth some of the electron
neutrinos produced had transformed into muon or tau neutrinos.
Later experiments were able to detect all three types of neutrinos
and found that the total neutrino flux was, in fact, in agreement
with the theoretical calculation.[endnoteRef:11] [10: For more
detailed discussion see (Franklin 2001, chapters 7 and 8).] [11:
This process took more than thirty years.]
In Shaperes terms the theory of transmission, which assumed that
there was no interference with the neutrino flux in its passage
from the core of the Sun to the detector on Earth, was wrong.
Nevertheless, even the reduced flux was sufficient to demonstrate
the existence of nuclear reactions in the core of the sun, albeit
not at the predicted rate. This episode might be considered a
reasonable fit to Shaperes definition of direct observation even
though the neutrinos changed. The electron neutrinos, the products
of nuclear reactions, were transformed during their transmission to
the Daviss detector on Earth.[endnoteRef:12] Later experiments in
which the apparatuses was sensitive to all three types of
neutrinos, confirmed Bahcalls Standard Solar Model. [12: Although
it did not happen, one might speculate whether the measurement of a
lower than predicted neutrino flux might be considered as evidence
for neutrino oscillations. Oscillations had been suggested before
the advent of the solar neutrino problem. I suggest that it
shouldnt have because the theory of the source, Bahcalls solar
model, was not sufficiently well established at the time. That
measurement did, however, encourage the further investigation of
neutrino oscillations. My colleague Alysia Marino has remarked that
the SNO experiments referred to actually showed only neutrino loss
and not neutrino oscillations. Later experiments did, however, show
oscillations.]
VIII. Gravity Waves
A. The Binary Pulsar
In 1975 Russell Hulse and Joseph Taylor reported the Discovery
of a Pulsar in an Binary System (Hulse and Taylor 1975). They had
previously reported the discovery of other pulsars and described
both their equipment and their search method (Hulse and Taylor
1974). Their method used electromagnetic radiation detected by the
Arecibo Radio Telescope. The earlier paper described the discovery
of eleven new pulsars and four previously observed pulsars. Their
later paper noted that they had by now found forty pulsars of which
thirty two had not been previously observed. The detection of the
previously known pulsars gave confidence in both their equipment
and in their search method.
Hulse and Taylor inferred that one new pulsar, PSR 1913 + 16,
was part of a binary system by noticing that, periodic changes in
the pulsation rate indicate that the pulsar is a member of a binary
system (1975, p. L51). The importance of the discovery was that for
the first time it is possible to observe the gravitational
interactions of a pulsar and another massive object (p. L51). They
also remarked that the mass of the unseen companion was comparable
to the mass of the pulsar. This meant that the object was probably
a neutron star or a black hole. In addition to the obvious
potential for determining the masses of the pulsar and its
companion, this discovery makes feasible a number of studies
involving the physics of compact objects, the astrophysics of close
binary systems, and special- and general-relativistic effects (p.
L51).[endnoteRef:13] [13: The importance of this discovery is shown
by the award of the 1993 Nobel Prize in Physics to Hulse and Taylor
for the discovery of a new type of pulsar, a discovery that has
opened up new possibilities for the study of gravitation.]
Hulse and Taylor based their results and conclusions on 200
measurements of pulsar arrival times taken over five-minute
intervals. They made no explicit comments concerning the use of the
binary pulsar system to detect gravitational radiation, or gravity
waves. The did, however, set an upper limit for the change in the
orbital period of the pulsar, of dPcm/dt < 1 x 10-12, a quantity
that would be of importance in the subsequent discovery.
In 1981 Joel Weisberg and Taylor announced that, We describe an
experiment which establishes, with a high degree of confidence the
existence of gravitational radiation as predicted by general
relativity (Weisberg and Taylor 1981, p. 1). More emphatically they
stated, The test is new and different from previous tests of
relativity because it goes beyond the usual first-order corrections
to Newtonian theory. In short, the results provide the first
compelling evidence for the existence of gravitational radiation,
and the magnitude of the radiation effect is in excellent accord
with the prediction of the quadrupole formula in general
relativity. We also show that our observations are not in agreement
with the predictions of several alternative theories of gravitation
(p. 2, emphasis added).
Their data consisted of more than 1500 measurements of pulse
arrival times from the binary pulsar. These measurements allowed
them to fit a timing model of the binary system. They extracted 20
observable parameters from their data using a least-squares
procedure. Only seven of these parameters are of interest for
analyzing the orbital dynamics of the system: the projected
semimajor axis of the pulsar orbit, apsini, where i is the angle
between the plane of the orbit and the plane of the sky; the
orbital eccentricity, e, period, P, and longitude of periastron, ,
the rate of advance of periastron, dot the variable part of the
gravitational redshift and transverse Doppler shift, , and the rate
of change of the orbital period, Pdot (pp. 2-3). The measured
values of these parameters are given in Figure 37. To completely
specify the pulsars orbit, as well as the masses of the pulsar and
its companion, Weisberg and Taylor used the measured values e, ,
and apsini, and Keplers third law. The relativistic parameters, and
dot provide the two additional equations necessary to solve
explicitly for the remaining unknowns. To do so, one must assume
that the measured value of dot is entirely the result of
relativistic effects, and one must work within the framework of a
particular theory of gravity (p. 3).
After specifying the orbital parameters and the masses of the
pulsar and its companion, Weisberg and Taylor could calculate the
expected rate of orbital decay expected from gravitational
radiation. They used a calculation done by Peters and Mathews
(1963). The quoted values of P, e, mp, and mc then yield the
calculated value
Pdot = (-2.38 0.02) x 10-12, in excellent agreement with the
observed value listed in [Figure 37] (p. 4). That value was (-2.5
0.3) x 10-12.
Weisberg and Taylor also presented the values calculated from
both general relativity and from four competing theories (Figure
38). The striking result is that the Rosen, Ni, and Lightman-Lee
theories all predict an orbital period increase due to the emission
of gravitational radiation, regardless of the magnitude of the
dipole term. Thus our measurements are inconsistent with such
theories unless one introduces ad hoc effects to explain Pdot. The
Brans-Dicke theory is consistent with observation only if the
coupling constant BD is very large or if the pulsars companion has
mass and internal structure very similar to those of the pulsar so
that is small (p. 5).[endnoteRef:14] The authors also noted that
some theorists believed that Einsteins quadrupole formula used to
calculate the energy loss rate was invalid for the binary pulsar
system. Obviously the dispute about what the theory actually
predicts must be resolved, but the present experimental situation
does not by itself seem to demand any changes. The binary pulsar
system containing PSR 1913 + 16 has provided general relativity
with one of its most probing tests, and the theory has survived
unscathed (p. 5). [14: is the difference in the self-gravitational
binding energy per unit mass of the two stars.]
B. The Binary Black Hole Merger
On February 11, 2016 the LIGO-Virgo collaboration announced the
Observation of Gravitational Waves from a Binary Black Hole Merger
(Abbott, Abbott et al. 2016). The event was labelled GW 150914. The
abstract stated, On September 14, 2015 at 09:50:45 UTC the two
detectors of the Laser Interferometer Gravitational-Wave
Observatory simultaneously observed a transient gravitational-wave
signal. The signal sweeps upwards in frequency from 35 to 250 Hz
with a peak gravitational-wave strain of 1.0 1021. It matches the
waveform predicted by general relativity for the inspiral and
merger of a pair of black holes and the ringdown of the resulting
single black hole. The signal was observed with a matched-filter
signal-to-noise ratio of 24 and a false alarm rate estimated to be
less than 1 event per 203 000 years, equivalent to a significance
greater than 5.1 (p. 0611102-1). They went on to state that, This
is the first direct detection of gravitational waves and the first
observation of a binary black hole merger (p. 0611102-1, emphasis
added). This latter statement was the subject of considerable
discussion within the collaboration. What exactly should the
collaboration claim? The authors did not claim that they had
discovered gravitational radiation, but rather that they had made
the first direct observation of gravity waves. They attributed the
discovery to the work of Hulse, Taylor, and Weisberg. The discovery
of the binary pulsar system PSR B1913 + 16 by Hulse and Taylor and
subsequent observations of its energy loss by Taylor and Weisberg
demonstrated the existence of gravitational waves. This discovery,
along with emerging astrophysical understanding, led to the
recognition that direct observations of the amplitude and phase of
gravitational waves would enable studies of additional relativistic
systems and provide new tests of general relativity, especially in
the dynamic strong-field regime (p. 0611102-1).
The collaboration began their paper with a brief history of
gravitational wave theory. They noted that recent advances together
with numerical relativity breakthroughs in the past decade, have
enabled modeling of binary black hole mergers and accurate
predictions of their gravitational waveforms (p. 061102-1).
I will discuss a few of the details of the experiment because it
casts light on the question of direct and indirect observation. A
simplified diagram of the experiment is shown in Figure 39. The
LIGO detector consists of two very sophisticated and sensitive
Michelson interferometers located in Hanford, WA and Livingston,
LA, a separation of approximately10 ms for a signal; travelling at
the speed of light. A gravitational wave propagating orthogonally
to the detector plane and linearly polarized parallel to the 4-km
optical cavities will have the effect of lengthening one 4-km arm
and shortening the other during one half-cycle of the wave; these
length changes are reversed during the other half-cycle. The output
photodetector records these differential cavity length variations.
(p. 061102-4). This differential length variation alters the phase
difference between the two light fields returning to the beam
splitter, transmitting an optical signal proportional to the
gravitational-wave strain to the output photodetector (p.
061102-3). Servo controls are used to hold the arm cavities on
resonance and maintain proper alignment of the optical components
(p. 061102-4). The detector was calibrated by measuring its
response to test mass motion induced by photon pressure from a
calibrated laser.
The collaboration presented several pieces of evidence in
support of their observation of both gravity waves and of the
merger of two black holes. These included a transient signal
observed in both LIGO interferometers with a time delay of
6.9+0.5-0.4 ms between the Livingston and Hanford detectors (the
time delay was in agreement with that predicted for a signal to
travel at the speed of light between the two detectors); the fact
that the signal matched the waveform predicted by General
Relativity for the gravitational waves emitted by the merger of two
black holes; a matched filter signal to noise ratio of 24; and a
false claim rate of less than 1 in 203,000 years.
The first of these is shown in Figure 40. The left hand side
shows the signal of strain plotted against time for the Hanford
interferometer.[endnoteRef:15] The right hand side shows the
Livingston signal superimposed on the Hanford signal (inverted and
time shifted to account for the difference in location and the
difference in the relative orientation of the detectors). One can
see that the signals are extremely similar. In fact, taking into
account noise in the detector, it seems fair to say that the
signals are the same. [15: The signals were filtered to suppress
large fluctuations.]
Figure 41 shows Gravitational-wave strain projected onto each
detector in the 35350 Hz band. Solid lines show a numerical
relativity waveform for a system with parameters consistent with
those recovered from GW150914 confirmed to 99.9% by an independent
calculation. Shaded areas show 90% credible regions for two
independent waveform reconstructions. One (dark gray) models the
signal using binary black hole template waveforms. The other (light
gray) does not use an astrophysical model, but instead calculates
the strain signal as a linear combination of sine-Gaussian
wavelets. These reconstructions have a 94% overlap. P. 061102-2).
The signals from the two interferometers agreed with each other and
with the theoretical expectations.
The collaboration searched for gravity waves using two different
and independent methods. GW150914 is confidently detected by two
different types of searches. One aims to recover signals from the
coalescence of compact objects, using optimal matched filtering
with waveforms predicted by general relativity. The other search
targets a broad range of generic transient signals, with minimal
assumptions about waveforms (061102-5). For the generic waveform
search the experimenters ranked events using the detection
statistic
c = (2Ec/(1 + En/Ec)), where Ec is the dimensionless coherent
signal energy obtained by cross-correlating the two reconstructed
waveforms, and En is the dimensionless residual noise energy after
the reconstructed signal is subtracted from the data. The statistic
c thus quantifies
the SNR [signal-to-noise ratio] of the event and the consistency
of the data between the two detectors (p. 061102-6). The group
considered three mutually exclusive classes of events: events that
resembled noise transients (C1), events in which the frequency
increases with time, consistent with the expectation for a
coalescent gravity wave event (C3), and all other events (C2). The
collaboration had to estimate the number background events, which
might simulate a gravity wave event. Although the method of
estimation was slightly different for the two searches, both used a
time-shift technique in which the signal from one detector is
shifted by an offset time interval, large in comparison to the
intersite transit time. Because the comparison included the
gravitational wave signal in one detector with the noise in the
other detector, this resulted in an overestimate of the noise
background and a conservative estimate of the significance of the
candidate events. Using the background estimate the group concluded
that the false alarm rate for events that might simulate a real
signal was lower than one in 22,500 years. This gives a probability
of < 2 x 10-6 of observing one or more events with a signal as
strong as that of GW 150914. This corresponds to a statistical
significance of 4.6 .[endnoteRef:16] The results of this search are
shown ion the left side of Figure 42. The GW150914 event has a
value of c = 20.0, which is the strongest event of the entire
search. [16: A slightly different method of estimating background
gave a significance of 4.4 .]
For the binary coalescence search the collaboration compared the
observed signal with 250,000 template forms, calculated on the
basis of general relativity. The experimenters calculated (t), the
matched-filter signal-to-noise ratio for each template in each
detector. Their search maximized the value of (t) with respect to
the time of arrival of the signal. For each maximum 2r was
calculated to test whether the data were consistent in different
frequency ranges. If 2r is near one that indicates that the signal
is consistent with coalescence. If 2r was greater than one (t) was
reweighted by ^(t) = /{[1 + 2r ]3/2}1/6. The group ranked
coincident events based on the quadrature of ^c and for both
detectors. The background was computed by time shifting the SNR
maxima of detector and calculating a new set of coincident data.
The right panel of Fig. 42 shows the background for the search
class of GW150914. The GW150914 detection statistic value of c =
23.6 is larger than any background event, so only an upper bound
can be placed on its false alarm rate. Across the three search
classes this bound is 1 in
203 000 years. This translates to a false alarm probability <
2 107, corresponding to 5.1 (p. 061102-7).
Possible sources of noise such as thermal and seismic effects as
well as other environmental effects were minimized and monitored.
To monitor environmental disturbances and their influence on the
detectors, each observatory site is equipped with an array of
sensors: seismometers, accelerometers, microphones, magnetometers,
radio receivers, weather
sensors, ac-power line monitors, and a cosmic-ray detector (p.
061102-5).
The collaboration concluded, The LIGO detectors have observed
gravitational waves
from the merger of two stellar-mass black holes. The detected
waveform matches the predictions of general relativity for the
inspiral and merger of a pair of black holes and the ringdown of
the resulting single black hole. These observations demonstrate the
existence of binary stellar-mass black hole systems. This is the
first direct detection of gravitational waves and the first
observation of a binary black hole merger (p. 061102-8).
C. Discussion
In both their abstract and in their conclusion the LIGO
experimenters stated that their result was the first direct
observation of gravitational waves. This had been a topic of both
discussion and contention within the collaboration and with others
in the gravitational wave community. The discussion concerned both
the questions of directness, of whether the term black holes should
be used, and whether the group should claim Observation of or
Evidence for.[endnoteRef:17] I will concentrate on the question of
directness.[endnoteRef:18] [17: At this time the criterion for a
discovery in high-energy physics was that the signal have a
statistical significance of five standard deviations. For details
see (Franklin 2013, Prologue).] [18: In high energy physics the
inclusion of Observation of in the title of a paper requires that
the observed effect have a statistical significance of at least
five standard deviations. Evidence for indicates a significance of
leas than five standard deviations.]
1. Big Dog
On September 16, 2010 a very large signal was detected by LIGO
and nicknamed Big Dog. This generated considerable excitement
within the group because it might very well have been the detection
of the first gravity wave signal. One member of the collaboration
stated, I think this is the golden event we were all hoping for
(quoted in (Collins 2013, p. 242)). One source of worry, however,
was the possibility that the event was a blind injection, a false
simulated gravity wave signal injected into the data streams of
both interferometers. These were designed to test the ability of
the analysis procedures to detect real gravity wave signals. It was
known, at the time, that there might be zero, one, two, or three
such injections. The number was known only to the two members of
the collaboration responsible for the injections. The answer to the
question of whether Big Dog was a blind injection would be answered
at a collaboration meeting scheduled for March 2011.
The collaboration proceeded on the assumption that Big Dog was a
real signal. This included the writing of a detection paper and it
was in this process that the internal discussion of the language to
be used began. The changing answers to the question of directness
are shown in the changing titles of the proposed discovery paper.
The first title proposed, by the end of September 2010, was, The
First Detection of (or First Evidence for) Gravitational Radiation
from Black Hole Coalescence (p. 190).[endnoteRef:19] The proposer
also stated that, I agree that Direct should be used. On January
18, 2017 the title of the paper was, Direct Detection of
Gravitational Waves from Compact Binary Coalescence, but by January
19 the draft title had been changed to Observation of Gravitational
Waves from Compact Binary Coalescence. One member of the
collaboration nevertheless stated, This is a direct detection, as
opposed to the indirect detection of Hulse and Taylor. We dropped
Direct from the title and abstract to make it shorter, and we hope
it is nonetheless clear. The word direct is used in the
introduction (p. 254). [19: This discussion was far more extensive,
complex, and interesting than the brief account I have given, It
has been documented and discussed in detail in Collins (2013,
chapter 8 12).]
On February 25 a senior member of the
collaboration[endnoteRef:20] proposed, Evidence for the Direct
Detection of Gravitational Waves from a Black Hole Binary
Coalescence. His argument was, I argue for Evidence for as an
acknowledgement that a single event is far from an ideal way to
make as important a claim as this is. I dont believe that we do
ourselves any harm by adding those words of cautionI predict that
almost everyone in the community will forget those two words almost
as soon as they see our case, but it is better to let them decide
on the strength of our claim than to possibly over-sell it. We have
a very good case, about as good as I think one can do with a single
event. However, if we really want to hold ourselves to the 5-sigma
standard, we fall short (Collins 2013, p. 264).[endnoteRef:21] The
Big Dog did, however, have a significance greater than five sigma.
This was the title presented to the collaboration at the March
meeting. [20: Collins does not identify the scientists by name.]
[21: With apologies to Jacqueline Susann, once may be enough.]
In discussing the question of how much support can be provided
be a single event, one member of the collaboration mentioned the
work of Blas Cabrera (1982). Cabrera was searching for magnetic
monopoles. In 1982 he reported a single event which was not only
much larger than any other signal he had detected, but was also fit
the signal magnitude that was theoretically expected for a monopole
extremely well. Cabrera, however, made no discovery claim because
he could not eliminate all of the other possible causes or
explanations of the event. The title of Cabreras paper was "First
Results from a Superconductive Detector. Later work with larger and
more sensitive detectors failed to find a similar event. The
consensus is that Cabreras event was not a monopole. (For more
details see (Franklin 2016, Chapter 18). Another senior scientist
in the Virgo group circulated a reading list of earlier discovery
papers, including Cabreras paper. He recommended looking at the
titles of those papers in urging caution in making claims.
Some of the scientists involved in the work on the binary pulsar
had a quite different view. They were insistent that their result
was the first direct observation of gravity waves. Thibault Damour,
who performed the first complete calculation of gravity-wave
interactions argued that the binary pulsar result demonstrated that
the signal between the two bodies travelled at the speed of light
and that this constitutes direct evidence for the reality of
gravitational radiation (quoted in Collins 2013, p. 198). Joseph
Taylor agreed and offered a longer argument.
In the binary pulsar experiment, and also in a LIGO-like
experiment, one infers the presence of gravitational radiation
based on its effects it induces in a detector. If a ruler could be
used to measure the displacement of LIGOs test masses, I would
grant that detection to seem rather more direct than one based on
timing measurements of an orbiting pulsar halfway across our
Galaxy. However, LIGO cant use a ruler; instead they use
servomechanisms, very sensitive electronics, and finally long
sequences of calculations to infer that a gravitational wave has
passed by. Such a detection, like the binary pulsar timing
experiment, is arguably many stages removed from being what most
people would call direct.
There is one significant difference. The detector in the binary
pulsar experiment is the pair of orbiting neutron starsthe same
thing as the transmitter. Detection involves measuring the
back-reaction of the emitted gravitational waves on the transmitter
itself. The time and place of detection are the same as the time
and place of emission.
These things will not be true of gravitational waves detected by
LIGO. In that case, the waves will necessarily have traveled a very
large distance from transmitter to receiver.
These thoughts suggest that a better distinguishing
characteristic of the two experiments would be something like in
place and remote rather than indirect and direct. (quoted in
Collins 2013, pp. 198-99).
Members of LIGO did not agree at the time, and, as we have seen,
the 2016 LIGO observation paper stated that it was the first direct
detection of gravitational waves. I disagree that the Hulse/Taylor
pulsar was the first direct detection. It was most definitely an
indirect detection, as the waves themselves were not observed. In
fact, I think this is exactly why the paper title should use the
term direct detection. Thats kind of the whole point of this
enterprise, and we should make that clear in our first detection
paper (Collins, 2013, p. 199).The discussion became moot when on
March 14, 2012 Jay Marx, the LIGO project director announced, The
first slide will tell you the answer: the Big Dog is a blind
injection (p. 284).
2. The Event
On September 14, 2015 another very large signal was detected by
the LIGO interferometers. The led to actions by the collaboration
similar to those following Big Dog, but there were some interesting
differences. Although the possibility of a blind injection was
again raised, this was discounted by many of the scientists because
the signal was detected during an engineering run, designed to test
the apparatus before the data taking run began. It was thought that
it was unlikely that an injection would be inserted in an
engineering run. The suspense was resolved three weeks later when
it was revealed that there were no blind injections. The
possibility of a malicious injection was also discussed. This was
rejected because it would have taken a conspiracy of a large number
of the members of the collaboration to create it. We cant say that
faking GW150914 was impossible, but we can say that faking it would
have required an internal conspiracy of our most knowledgeable
people (Collins 2017, p. 101).
Once again caution was urged in stating the results of the
experiment. The Cabrera monopole was mentioned along with the more
recent claim to have observed gravity waves by the BICEP2
collaboration. That collaboration claimed to have seen a 5 effect
by looking at the polarization of the electromagnetic radiation in
the Cosmic Microwave Background (Ade, Aikin et al. 2014). That
claim was later withdrawn, or at least the size and significance of
the effect greatly reduced because the collaboration had not
adequately included the effects of cosmic dust (Mortonson and
Seljak 2014). Members of the LIGO collaboration did, however, cite
the positive result of the discovery of the - particle, discussed
earlier. In that episode a single event was sufficient to support a
discovery claim.[endnoteRef:22] [22: No mention was made of the
discovery of the positron.]
Some of the uncertainty caused by having only a single event was
removed when a second gravity wave event was detected on October
12, 2015. This was a weaker signal than The Event, it had only a 3
significance, and would certainly not have been sufficient to merit
a discovery claim, but in the presence of The Event, it did provide
additional support. Peter Saulson wrote to Harry Collins and
stated, A genuinely marginal event in the latest PyCBC box opening
that happened one hour ago. At 3 , it is going to be tougher, but
this is what we expected[endnoteRef:23] this is quite a weak event
by cWB standards, however it has a beautiful chirp shape both in
time and in TF domains (quoted in Collins 2017, pp. 118-119). [23:
The Event was a very large signal. The collaboration expected to
find more and smaller signals.]
The question of direct versus indirect observation was again
raised. Collins reports that there were 2500 e-mails concerning the
draft of the detection paper. They spanned a wide range of opinion
from caution to strong advocacy of making a strong statement about
a direct observation by LIGO. One experimenter cited ten websites
in which the binary pulsar experiment was considered to be an
indirect observation of gravity waves. In addition, when Hulse and
Taylor were awarded the 1993 Nobel Prize in Physics for the
discovery of a new type of pulsar, a discovery that has opened up
new possibilities for the study of gravitation, the press release
issued by the Nobel Prize Institute stated, The good agreement
between the observed value and the theoretically calculated value
of the orbital path can be seen as an indirect proof of the
existence of gravitational waves. We will probably have to wait
until next century for a direct demonstration of their existence
http://www.nobelprize.org/nobel_prizes/physics/laureates/1993/press.html
emphasis added.
A moderate comment was, I would also like to suggest that we
avoid the phrases "direct observation" and "direct detection", at
least in the title, abstract and introduction. While I know that we
use those terms as a way of distinguishing what we do from other
methods of GW detection I have come to understand that there are
people in our broader community who think that this terminology is
meant to diminish the importance of their work. In keeping with the
modest tone I advocate, I see no benefit in using these particular
terms. Indeed, without further definition, they do not convey any
clear meaning We have a great result, and it will not be any less
great without this characterization(Collins, 2017, pp. 147-48).
Another commentator did not think the issue of direct was very
important, but did believe that there was a significant difference
between the binary pulsar and LIGO experiments. The difference is
the Taylor crowd observed a distant GW transmitter and figured out
how it worked! We figured out how to build a sufficiently sensitive
GW receiver and since we built it, we know exactly how it works. If
anybody misses the impact of those italicized words, check out
BICEP-2's and Planck's experience to date. Those words in italics
represent a huge advance for GW physics and astronomy (Collins,
2017, p. 148).
As is seen in these quotes there was a general view that one
should avoid giving offense to others in the gravity wave
community. As Collins remarks, Taylor and Damour are very much
liked among more senior members of the gravitational wave community
and their work is hugely admired so there are many who do not want
to offend them (Collins, 2017, p. 147).[endnoteRef:24] [24: A
similar view was expressed to me by Peter Saulson (private
conversation), a former spokesperson for LIGO.]
Faced with this wide variation in opinion the writing team
conducted a poll among the LIGO group members. One question listed
eight possible titles for the discovery paper and asked for an
ordered preference.
1. Observation of Gravitational Waves from a Binary Black Hole
Merger.
2. Direct Observation of Gravitational Waves from a Binary Black
Hole Merger.
3. Detection of Gravitational Waves from a Binary Black Hole
Merger.
4. Direct Detection of Gravitational Waves from a Binary Black
Hole Merger.
5. LIGO Observation of Gravitational Waves from a Binary Black
Hole Merger.
6. LIGO Detection of Gravitational Waves from a Binary Black
Hole Merger.
7. Observation by LIGO of Gravitational Waves from a Binary
Black Hole Merger.
8. Direct Observation by LIGO of Gravitational Waves from a
Binary Black Hole Merger. (from Collins 2017, p. 154-55)
Another question asked for an opinion on whether direct
(detection and/or observation should be used in the body of the
text.
1. No problem to use direct in the paper.
2. Use direct once in the introduction and abstract only.
3. Use direct once in the conclusion only.
4. Dont use direct. (Collins 2017, p. 155)
The results were
Poll 1. No direct and no LIGO in the title. Preferences are
Observation or Detection of GW from BBH merger.
Poll 3. It is OK to use direct (detection/observation) in the
body of the paper. (Collins 2017, p. 156)
As we have seen, the initial LIGO discovery paper followed the
results of the poll.
A further argument for the difference between the binary pulsar
and LIGO experiments was presented in the Outlook section of the
LIGO paper. This was that the observation combined with the future
observations expected when other detectors came online would open a
new era in gravity wave astronomy. Efforts are under way to enhance
significantly the global gravitational-wave detector network. These
include further commissioning of the Advanced LIGO detectors to
reach design sensitivity, which will allow detection of binaries
like GW150914 with 3 times higher SNR [signal to noise ratio].
Additionally, Advanced Virgo, KAGRA, and a possible third LIGO
detector in India will extend the network and significantly improve
the position reconstruction and parameter estimation of sources
(Abbott, Abbott et al. 2016, p. 061102-8).
In terms of Shaperes discussion of direct observation we see
that the LIGO-VIRGO collaboration did make a direct observation.
The gravity waves interacted with the two interferometers. The fact
that there was complex instrumentation and analysis should not
change that conclusion. It seems fair to say that the binary pulsar
observation was indirect. The existence of gravity waves was
inferred from the decrease in period of the pulsar, which was
transmitted to the detector by electromagnetic radiation. As
discussed below, I do not believe that the direct-indirect
distinction had any epistemological significance. Recall that LIGO
attributed the discovery of gravity waves to binary pulsar work of
Hulse, Taylor, and Weisberg
Figure Captions
Figure 37. Observed orbital parameters for the binary pulsar.
From Weisberg and Taylor (1981).
Figure 38. Predictions for the rate of orbital period change of
the binary pulsar for various theories (normalized to general
relativity).
Figure 39. Simplified diagram of an Advanced LIGO detector.
Inset (a): Location and orientation of the LIGO detectors at
Hanford, WA (H1) and Livingston, LA (L1). Inset (b): The instrument
noise for each detector near the time of the signal detection. From
Abbott et al. (2016).
Figure 40. The gravitational-wave event GW150914 observed by the
LIGO Hanford and Livingston detectors. The left column: H1 strain;
right: L1 strain. GW150914 arrived first at L1 and 6.9 +0.5 0.4 ms
later at H1; for a visual comparison, the H1 data are also shown,
shifted in time by this amount and inverted (to account for the
detectors relative orientations). From Abbott et al. (2016).
Figure 41. Gravitational-wave strain projected onto each
detector in the 35350 Hz band. Solid lines show a numerical
relativity waveform for a system with parameters consistent with
those recovered from GW150914 confirmed to 99.9% by an independent
calculation. Shaded areas show 90% credible regions for two
independent waveform reconstructions. One (dark gray) models the
signal using binary black hole template waveforms. The other (light
gray) does not use an astrophysical model, but instead calculates
the strain signal as a linear combination of sine-Gaussian
wavelets. These reconstructions have a 94% overlap. From Abbott et
al. (2016).
Figure 42. Search results from the generic transient search
(left) and the binary coalescence search (right). These histograms
show the number of candidate events (orange markers) and the mean
number of background events (black lines) in the search class where
GW150914 was found as a function of the search detection statistic
and with a bin width of 0.2. The scales on the top give the
significance of an event in Gaussian standard deviations based on
the corresponding noise background. The significance of GW150914 is
greater than 5.1 and 4.6 for the binary coalescence and the generic
transient searches, respectively. From Abbott et al. (2016).
F Figure 37
FFFF Figure 38
Figure 39
Figure 40
Figure 41
Figure 42