Professional Documents
Culture Documents
and
S. W. Hawking
Institute of Theoretical Astronomy, University of Cambridge, England
(Received 30 November 1970)
It is argued that the short bursts of gravitational radiation which Weber reports most prob-
ably arise from the gravitational collapse of a body of stellar mass or the capture of one col-
lapsed object by another. In both cases the bulk of the energy would be emitted in a burst
lasting about a millisecond, during which the Riemann tensor would change sign from one to
three times. The signal-to-noise problem for the detection of such bursts is discussed, and
it is shown that by observing fluctuations in the phase or amplitude of the Brownian oscilla-
tions of a quadrupole antenna one can detect bursts which impart to the system an energy of
a small fraction of kT. Applied to Weber's antenna, this method could improve the sensitivi-
ty for reliable detection by a factor of about 12. However, by using an antenna of the same
physical dimensions but with a much tighter electromechanical coupling, one could obtain an
improvement by a factor of up to 250. The tighter coupling would also enable one to deter-
mine the time of arrival of the bursts to within a millisecond. Such time resolution would
make it possible to verify that the radiation was propagating with the velocity of light and to
determine the direction of the source.
2191
2192 G. W. GIBBONS AND S. W. HAWKING
means that gravitational-wave detectors are very frequencies which he does not observe. The only
inefficient: A flow of 2 x 10~ erg/cm' sec for a reasonable way in which neutron stars might be
duration of 10 sec in a bandwidth of 0.016 Hz at responsible for the bursts that Weber observes is
1660 Hz is needed to excite Weber's detectors to if we happen to be untypically near a neutron star
an amplitude of the same order as that of the that was undergoing violent eruptions: A neutron
Brownian motion of the apparatus. Weber observes star at 1 pc would have enough energy to produce
the amplitude of his detectors rising above thresh- a burst of 4 x 10' erg/cm' once a day for about
old within 0. 5 sec of one another. If the duration 150000 years.
of the bursts were much longer than 0.5 sec, the Gravitational collapse, or the capture of one col-
amplitudes of the two detectors would rise slowly lapsed object by another, would have the advantage
and would probably not cross the threshold within over neutron stars as a source of gravitational ra-
0.5 sec of each other as the initial Brownian mo- diation that a much greater fraction of the rest-
tion of the two detectors would be different. This mass energy might be released. This makes the
would be inconsistent with Weber's report that energy problem easier but the amount is still em-
shortening the coincidence time did not significantly barrassingly high. '-' A source at the galactic cen-
affect the rate of coincidences. A burst of duration ter needs an energy of about 1.4 x 10" erg (VOTO}
less than 0. 5 sec must have a bandwidth of at least to produce a burst of 1.2 x 10" erg/cm' at the
0.3 Hz and so each burst must carry an energy of earth.
at least 4 x 10' erg/cm'. If one allowed for an In a collapse or a capture most of the energy
equal energy in the other polarization which Weber would not be released until matter has fallen to
does not observe, the total energy would be at near the Schwarzchild radius or one collapsed ob-
least 8 x 106 erg/cm'. It is not reasonable to ject was near to the Schwarzchild radius of the
suppose that there is gravitational radiation only other one. Thus one would expect most of the en-
in a very narrow bandwidth centered on the fre- ergy of the gravitational radiation to be in a period
quency of Weber's detector. Therefore we pre- 7 of the order of the dynamical time at this stage,
sume that either each burst has a much larger i.e. , ~-GMc '-10 'NM. ' sec.
bandwidth, say of the order of 1000 Hz, and so The energy flux in the gravitational wave at the
carries an energy of the order of 1.2 x 10" observer is
erg/cm' per polarization, or that each burst has t
a narrow bandwidth but that there are something = c'(4vG) '
F(t) — R„„(u)du
like a thousand bursts at other frequencies for
every one that Weber observes. In either case 2
the energy flux is at least 10'0 erg/cm' per day. + ff, ~,(u) du erg/cm' sec,
A source emitting at 1660 Hz, the frequency at
which Weber observes, must presumably be under-
going collective dynamical motions at that frequen-
where the wave is taken to be traveling in the
cy. This would seem to imply that its size was
less than 100 km, which is the distance light would three directions and the Riemann tensor compo-
travel in half a cycle. Possible sources of this nents Pypyp and pyplpp represent the two states of po-
size might be a neutron star, a star undergoing larization of the wave. From this formula it can
gravitational collapse, or the capture of one col- be seen that for a burst of finite energy the time
lapsed object by another. A neutron star would integrals of the Riemann tensor components over
have an available nuclear and gravitational energy
the duration of the burst must be zero. This means
that the sign of the Riemann tensor components
equal to about 1% of its rest mass, i.e. , about 2
& 10~ erg. Therefore to produce a gravitational-
must reverse during the burst.
wave burst of 1.2 x 10'o erg/cm' it would have to In the linearized theory one has the relation
be within a distance of 3 x 10" cm, i.e. , 100 par- G 1 d
secs (pc). A distance of 100 pc includes about one 3c' (2)
r dt'
part in 10' of the galactic disk. If, therefore,
sources are distributed uniformly throughout the where D ~ is the quadrupole moment of the source
galactic disk, ten thousand of them are using up and r isthe distance from the source.
all their available energy every day, a number From this formula it follows that if the quadru-
which seems unreasonably high. This number is pole is initially and finally time-independent (as
not changed if one supposes that the neutron stars one would expect for a gravitational collapse}, then
emit in a narrower bandwidth; for although the not only is the time integral of the components of
sources that Weber observes can be further away, the Riemann tensor zero but the second and third
there will be a corresponding number at other integrals as well, i.e. , '
THEORY OF THE DETECTION OF SHORT BURSTS. . . 2193
dt dt 8 ~0~0 t dt
0 - 0 0
R io io
This implies that the components of the Riemann
tensor must change sign at least three times dur-
ing the burst. For gravitational capture, on the
other hand, the quadrupole moment will initially
be a quadratic function of the time and so only the
first time integral of the components of the Rie-
mann tensor will be zero, and the components of (a)
the Riemann tensor need change sign only once
during the burst.
In the nonlinear theory one does not have the sim-
ple relation (2) between the source and the radia-
tion field. Thus Eq. (3) will not in general be sat- R io lo
IV. THE SIGNAL/NOISE RATIO cillations of the detector one could detect a gravi-
Assuming that extraneous electrical and mechan- tational burst that imparted to the detector in that
ical disturbances can be eliminated, the main time an energy greater than 2mnkTQ '.' Obviously,
sources of noise are the Brownian movements of using a smaller value of n gives a better time res-
the detector masses, the Johnson noise associated olution.
with the electrical losses in the transducer, and The most practical way to observe the motions
the noise produced in the first stage of the ampli- of the detector seems to be to use a piezoelectric
fier. We shall discuss first the Brownian move- transducer' whose output is fed into a high-imped-
ments and show that their effect can be minimized ance amplifier. A piezoelectric transducer can be
by using a detector with a fairly high Q and observ- represented electrically as a charge generator in
ing its motion with a good time resolution. parallel with a capacitance and a resistance in
The two masses and the spring of the detector series (Fig. 2). Here C, is the electrical capaci-
form a simple harmonic oscillator. By the law of tance of the transducer, n is its voltage output per
equipartition of energy the oscillation will have an unit change of separation x of the two masses of
average energy of oscillation of kT provided that the detector, and &, = (&uoC, )-'tan5 represents the
it is in equilibrium with its surroundings. The am- electrical losses in the transducer, where tan5 is
plitude x~ of the thermal oscillations will be given the "dissipation factor. "
by The Johnson noise of the resistor produces a
mean-square voltage in a bandwidth ~(d of
m(u ~x '=4kT . (9)
Vr' = 2~ 'kTR, d&u = 2v 'kT tan5 6&v(~,C, )-'.
For re-10' g, T-300'E, and ~, -10', this gives (10)
"
-4 x 10 cm. One might think that one could de-
tect only bursts that induced an amplitude greater This may be compared to the mean-square voltage
than this or equivalently that imparted to the de- produced by the Brownian movements of the detec-
tector an energy larger than kT. However, it is in tor masses
fact possible to detect much smaller bursts. VB '=n'x B'
For although the average energy of the thermal
oscillation is kT in equilibrium, the noise poorer, where xs is given by (9). Thus,
i.e. , the rate at which energy enters and leaves the V~' =4n'kT(m&u, ') '. (12)
thermal oscillations of the detector, is very small
if Q is large (we are grateful to P. Aplin for point- By observing fluctuations over n cycles, one can
ing this out). This can be seen as follows: One detect against the Brownian noise a burst that im-
can regard the thermal oscillations as arising parts to the detector an energy greater than
from a large number of small impulses applied to 2mnkTQ '. For such a burst the signal output volt-
the detector with random phases and amplitudes. age V~ satisfies V~'~ 2wnV~'Q '. In order to de-
The oscillations induced by each small impulse tect a signal against both the transducer noise and
will die away in the damping time Qcu, -'. Thus at the Brownian noise, one needs
any time the energy of the detector arises effec- V~ ~ V~ +2nnQ 'VB'.
tively from the random impulses applied in the last
Q(2v) ' cycles. Since the phases of the oscillations The greatest sensitivity, therefore, is obtained
induced by the small impulses are uncorrelated, by choosing the value of n for which
it follows that the mean value of the energy of the Vz' =2pnV~ Q
oscillation induced by the small impulses in one
cycle is kT2vQ-'. Thus by observing the change This gives
over n cycles of the phase or amplitude of the os- n' = Q tan6 (8w 'p) ',
THEORY OF THE DETECTION OF SHORT BURSTS. . . 2195
C,
Z3
ZI Za
L,
~
J. Weber, General Relativity and Gravitational Waves These values can be deduced from the equivalent cir-
(Interscience, New York, 1961). cuit given in J. Weber, Lett. Nuovo Cimento 4, 653 (1970).
2J. Weber, Phys. Rev. Letters 22, 1320 (1969). ~~The effective value of l is about 27(. ~ times the total
3J. Weber, Phys. Rev. Letters 24, 276 (1970). length of the detector.
4J. Weber, Phys. Rev. Letters 25, 180 (1970). J. Weber, Lett. Nuovo Cimento 4, 653 (1970).
5D. W. Sciama, G. B. Field, and M. J. Rees, Phys. 3The resonant frequency of this circuit is uo when the
Rev. Letters 23, 1514 (1969). output of the transducer is short circuited. When it is
~D. W. Sciama, G. B. Field, and M. J. Rees, Comments not short circuited the resonant frequency is (1+p) p.
Astrophys. Space Phys. 1, 187 (1969). Taking ~0& cL)g improves the signal/(Brownian noise)
7D. W. Sciama, Nature 224, 1263 (1969). ratio but makes the signal/(transducer noise) ratio worse.
This equation also holds in the linearized Brans-Dicke One cannot make cup & (dg without reducing the length of
theory. the detector.
9This point has also been made by V. Braginski7. and ~~S. Deser and B. E. Laurent, Ann. Phys. (N. Y.) 50, 76
V. N. Rudenko, Usp. Fiz. Nauk 100, 395 (1970) [Soviet (1968) .
Phys. Usp. 13, 165 (1970)] .