Multimessenger observations¶

Multimessenger astronomy: A very brief history¶

While the discovery of GW170817 marked the beginning of with , it did not mark the beginning of as a field. The earliest observations can be traced to astroparticle observations of the solar flares in the 1940s :cite:`2015paas.book.....S`. The observation of SN1987A, a type-II supernova in the Large Magellanic Cloud, was made first in visible light at Las Campanas, but also later in ultraviolet. Around three hours prior to the event’s detection in , a number of observatories detected a burst of antineutrinos (with around 25 particles observed across three observatories). The first detection of gravitational waves was made in 2015, with the observation of a coalescence, and just under two years later the first observation of from a coalescence was made, which coincided with a observed by the Fermi satellite. An extensive observing campaign was launched as a result of these observations, with observatories covering the entire spectrum making observations of the event. These will be discussed in detail in the section discussing GW170817.

Multimessenger astrophysical sources¶

The anticipated sources of gravitational wave signals fall into two broad categories: transient sources, which produce a burst of over a short period of time; and sources, which produce continuously. A number of transient sources are considered as promising targets for astronomy, and these are often referred to as bright.

Preparing GW alerts¶

While detections can be interesting in their own right, the development of relies on rapid communication between the detectors and observatories. This is challenging, as not all events are likely to produce emission, and the location of the event in the sky must be determined. Once these quantities are determined events are reported using the and on GraceDB (see https://gracedb.ligo.org/superevents/public/O3/).

Localising GW signals on the sky¶

If a network of at least two geographically separated detectors observes a signal it is possible to ascertain the location in the sky, \(\hat{\vec{\Omega}}\), from the difference in arrival times between the two sites. For a detector at a position, \(\vec{r}_{D}\), and an arbitrary reference location, \(\vec{r}_{0}\), this time delay, \(\delta t\), will be

\[\label{eq:intro:detectors:timedelay} \delta t (\hat{\vec{\Omega}}) = \frac{1}{c} (\vec{r}_{0} - \vec{r}_{D}) \cdot \hat{\vec{\Omega}}.\]

This allows the location of the signal to be confined to a ring on the sky corresponding to constant \(\Delta t\). Examples of these rings for a source are plotted in figure [fig:det:advanced-timing]. Timing uncertainty in the signal, which arises both from clock uncertainties and uncertainties in defining a reference point in the received signal increase the area of this region. As more detectors are added to the network it is possible to reduce this area, as increasing the number of detector pairs works to reduce the sky area compatible with the observed delay times.

Isochrones for the three detector pairs in the advanced network. For a single detector pair the localisation is a ring; with three detectors there are three pairs of detectors, and so three rings, and we can reduce the plausible locations the signal could have come from to the two places where all of the rings overlap. Isochrones for the three detector pairs in the advanced network. For a single detector pair the localisation is a ring; with three detectors there are three pairs of detectors, and so three rings, and we can reduce the plausible locations the signal could have come from to the two places where all of the rings overlap.¶

Additional localisation information can be attained from the observed amplitude of the signal in each detector. The signal will be convolved with the antenna pattern (see figure [fig:det:aligo-antenna]); as each detector is insensitive to some regions of the sky, the total plausible localisation of the signal is reduced.

For a approaching the detector from an azimuth (relative to one of the arms) and altitude (relative to the plane of the detector), \((\alpha, \delta)\) on the sky these patterns for the \(+\)- and \(\times\)-polarisations, \(F_{+}\) and \(F_{\times}\), will be

\[\begin{split}\begin{aligned} \label{eq:detectors:antennapattern:plus} F_{+} &= \frac{1}{2} (1 + \sin^{2}\delta) \cos 2\alpha \cos 2\psi - \sin\delta\sin 2 \alpha \sin 2 \psi \\ F_{\times} &= \frac{1}{2} (1 + \sin^{2}\delta) \cos 2\alpha \sin 2\psi - \sin\delta\sin 2 \alpha \cos 2 \psi.\end{aligned}\end{split}\]

where \(\phi\) is the polarisation angle of the .

Transient astronomy¶

Challenges for GW event follow-up¶

While preparing alerts based on observations is challenging, attempting to make observations to follow these up is not without problems. The localisation of most events is poor, meaning that the event could originate anywhere within a large patch (or large patches) of the sky. The majority of observatories can perform observations over only a small field of view, however, and the emission related to a event may be short-lived. As a result an observatory must be able to rapidly survey a large area of sky with high sensitivity.

The sky localisations which are published by detectors are divided into observing “tiles” by each follow-up observatory . The size of each tile will vary depending on the sensitivity and field-of-view of the telescope. Each tile is then prioritised using probability information from the analysis :cite:`2017ApJ...834...84C,2019MNRAS.489.5775C`, and taking into account difficulties in moving the telescope and the period of local night.

GW170817: A case-study¶

[sec:gw170817]

|image|

On 17 August 2017, during the second observing run of advanced LIGO, and a few days after advanced Virgo had started making observations a signal, GW170817, was detected by both LIGO detectors and the Virgo detector. In contrast to previous detections which had all been signals, GW170817 was identified as being produced by a system.

Independently of the detection the Fermi and INTEGRAL satellites detected a slightly less than two seconds after the time the was detected in . GCN alerts were issued rapidly for both the Fermi detection (within 14 seconds) and the LIGO/Virgo detection (within 40 minutes).

The (recently-expanded) three detector network initially localised the signal to within 31 square degrees in the southern celestial hemisphere, however later analysis allowed this to be reduced to a 28 square degree patch of sky. The localisation areas from the various detections are shown in figure [fig:gw170817-localisation] for the detections in green and the detections in blue.

The three-detector localisation was calculated by around 17:54 UTC, which allowed telescopes in South America to search the localisation area for an optical transient 5. The Swope supernova survey was the first collaboration to observe the transient :cite:`2017ApJ...848L..12A,2017Sci...358.1556C` (although six observatories would independently discover the optical counterpart :cite:`2017ApJ...848L..12A`). The optical counterpart was observed in NGC 4993.

The highly-precise localisation which was produced by imaging the optical counterpart allowed observations to be made across the entire spectrum.

Ultraviolet emission was detected 15.3 hours after the event by Swift, and 9 days later X-ray emission was detected by the Chandra X-ray Observatory. 16 days after the was observed radio emission was observed by the VLA in New Mexico.

observations continued until 2019, with the Hubble Space Telescope unable to detect any optical afterglow after 584 days :cite:`2019ApJ...883L...1F`. Superluminal radio emission was also reported :cite:`2018Natur.561..355M` between 75 and 230 days after the merger.

Cosmology from multimessenger astronomy¶

The observation of an counterpart to GW170817 allowed the galaxy it originated in to be identified. In turn this allowed the recession velocity of the to be determined with high precision from its redshift. The detection allows the distance to the source to be measured directly (although with a fairly large uncertainty, thanks to a degeneracy between the distance to the source and the angle at which it is inclined relative to the observer.

Since the distance, \(d\), and recession velocity, \(v\), are related by Hubble’s Law,

if we know both \(v\) and \(d\) we can infer \(H_{0}\).

The distance to the source of GW170817 inferred from the is \(d = \SI[parse-numbers=false]{48.8^{+2.9}_{-6.9}}{\mega\parsec}\), and the measured recession velocity is \(v = \SI{3017\pm166}{\kilo\meter\ \second^{-1}}\).

This allowed \(H_{0}\) to be inferred to be \(\SI[parse-numbers=false]{70.0^{+12.0}_{-8.0}}{\kilo\meter\ \second^{-1}\ \mega\parsec^{-1}}\)

:cite:`2017Natur.551...85A`.

While we get the greatest amount of information from events which can be localised by observations, it is also possible to infer the Hubble constant using only observations. This means that events can be used, which are much more frequently observed than events.

In order to make inferences without knowing which galaxy the event occurred in we need accurate three-dimensional galaxy catalogues. By identifying a list of galaxies which lie within the localised volume (through the sky localisation and distance estimate of the ) we can use a Bayesian analysis to combine the inferences from each plausible galaxy to give an overall estimate :cite:`2019arXiv190806050G,2019arXiv190806060T`.

From the first two observation runs’ detections it is possible to update the GW170817-only estimate of \(H_{0}\) to \(\SI[parse-numbers=false]{68.0^{+14.0}_{-7.0}}{\kilo\meter\ \second^{-1}\ \mega\parsec^{-1}}\)

:cite:`2019arXiv190806060T`.

GW follow-up of EM events¶

In addition to attempts to identify electromagnetic counterparts to signals, there are ongoing efforts to identify signals produced by events observed by observatories. Thanks to the near-continuous, all-sky, broadband observations made by a network of detectors, it is possible to conduct searches for counterparts in high-latency in recorded data (whereas an observatory may need to be pointed to the appropriate area of sky, for example).

There have been targeted searches for from , motivated by observations. The sky localisation provided by the observation simplifies the process of searching for the signal :cite:`2019arXiv190803584T`.

Pulsars are the most promising source of continuous , and since these are observed by radio telescopes, which can determine their rotation frequency we can target searches for from pulsars both by sky location and frequency (the frequency is twice the rotation frequency, since are emitted from the quadrupole mode). To date we’ve not been successful in detecting from pulsars, but the non-detection allows us to place limits on the physical properties of known pulsars :cite:`2019PhRvD..99l2002A`. Pulsars are also observed to glitch when observed in radio: a glitch is a sudden change in the rotational frequency of the pulsar; the mechanism which causes these is poorly understood, but may produce . The time at which these glitches occur is well known from observations, so searches for these can be carried out over a short stretch of data :cite:`2019PhRvD.100f4058K`.

Observations are made of frequently, and events are known to be a progenitor source for these events. These events are very well localised in time, however gamma ray detectors are not normally able to give a very precise sky localisation for an event, so a search can be made over a short span of detector data, but a large sky area :cite:`2019arXiv190701443T`.

The future: multi-band multimessenger astronomy¶

The current generation of detectors are designed to operate in a frequency range where the merger and ringdown components of a or low-mass system will produce a detectable signal. However, space-based detectors, such as , will be able to make observations at much lower frequencies. As a result the inspiral of these events will be observable for a much longer period of time than is currently possible.

For an inspiralling event the frequency of the inspiral signal can be used to predict the time at which the two systems will merge :cite:`1994PhRvD..50.7111S`. This means if the lowest frequency a detector can measure an inspiral signal at is \(f_{\text{low}}\) then the time, \(t\), between observing the start of the inspiral and the merger is approximately

\(\SI{1.25}{\solMass}\).

1

Within the solar system, and more broadly, the heliosphere, it’s possible to argue that additional messengers exist, for example, through sample return missions, or magnetometer measurements, however, these are not available for the vast majority of the universe, so I’ll not give them any further consideration here.

2

Though it’s generally accepted that this was a coincidence, as no event following this one has been coincident with an event, and the poor localisation of the GW150914 signal provides little evidence that the two events were spatially coincident.

3

If the Earth was equivalently spherical the highest mountains would be around \(\SI{25}{\meter}\) high.

4

For a fuller discussion of the effects of spin and the orbit on the determination of the orbital compactness see section 4 of :cite:`2017AnP...52900209A`.

5

The search was complicated by the proximity of the search region to the sun, which meant observations were only possible shortly after the onset of twilight for optical telescopes.