A new radio transient found with the MeerKAT telescope

(1)

M

ASTER

’

S

T

HESIS

A new radio transient found with the

MeerKAT telescope

Author: Jelle Meijn 10275258 Supervisor: Dr. Antonia Rowlinson Examiner: Prof. dr. Ralph Wijers

February 15, 2021 60 ECTS

Anton Pannekoek

Institute

(2)

Abstract

A new radio transient found with the MeerKAT telescope

by Jelle Meijn

In one of the commensal searches of the ThunderKAT program for the MeerKAT telescope, we found a new radio transient. Using weekly observations at 1.28 GHz, the flux increased from∼3 mJy to∼6 mJy before dropping back down to∼2 mJy over the course of two months. In contrast to the fast rise exponential decay typi-cally seen in transients of an explosive nature, this transient has a very Gaussian-like lightcurve. After a two-month period of relatively stable flux, it increases again and fluctuates between∼3 mJy and∼4.5 mJy for around three months. After another two months, it stabilises slightly below∼2 mJy. Combining the MeerKAT data with entries in PAN-STARRS, VLASS and observations taken with SWIFT’s XRT, we con-clude that the transient is consistent with an AGN. We argue that the transient is most likely caused by intrinsic flaring of the AGN. We also argue the importance of direction dependent calibration and primary beam correction in future commensal searches.

(3)

Introduction

1.1 Radio Astronomy

Since the age of the Babylonians and the ancient Greeks, humans have always had an interest in the night sky. Until the 19th century, this has always been done by the visual part of the electromagnetic spectrum. However, with the invention of radio and gamma-ray telescopes and everything in between, the whole of the electromag-netic spectrum became available to study the sky.

This began back in the 18th century with James Clerk Maxwell. With his for-mulation of the Maxwell equations, he unified electricity, magnetism and light as part of one fundamental force of nature, electromagnetism (Maxwell, 1865). The Maxwell equations predicted that visual light was only a small part of the whole electromagnetic spectrum, which could have an infinite amount of frequencies. This also led physicists to believe electromagnetic radiation could have an astronomical source. Later in 1900, Max Planck derived his now well-known Planck’s law, an equation that relates the spectral density of a blackbody to its temperature (Planck, 2013). According to this equation, the amount of black body radiation emitted in radio is very low compared to the visual part of the spectrum. For objects like stars, which have surface temperatures of over 2500K, most of the radiation is emitted in the visual part. This would mean that to detect an astronomical radio signal emitted by blackbody radiation, the receiver had to be really sensitive.

After several failed attempts by different scientists, Karl Jansky was the first to discover an astronomical radio source. While he was working at Bell Laboratories, he built an antenna which could detect radio waves. He picked up a radio signal that seemed to repeat every 23 hours and 56 minutes, every sidereal day. Therefore, the signal had to be extraterrestrial. After comparing it with optical observations, he concluded that the signal was coming from the center of the Milky Way (Jansky, 1933). However, because of the Great Depression scientists were hesitant to follow up on the research, causing the progress in radio astronomy to be slow at the start.

The discovery of Jansky grabbed the attention of Grote Reber. After his failed attempt to land a job at Bell Laboratories, he set out to build his own radio telescope. He constructed a 9 meter parabolic radio antenna in his own backyard in 1937 and started looking for the source Jansky found at 3300 MHz. At the time, only thermal (blackbody) radiation was considered to be a natural emission mechanism of radio waves. This radiation would be stronger at higher frequencies, so Reber initially used a receiver at 3300 MHz. After not detecting any signals, Reber kept lowering the frequency until at 160 MHz he finally detected the source previously discovered by Jansky (Reber,1940). The fact that he detected this signal in such a low frequency also let him to the conclusion that it couldn’t be thermal emission. After this detec-tion, he continued working with his telescope and made the first radio sky survey.

(5)

Reber would be the only radio astronomer for almost a decade, until after WWII. The advancements in sensitivity and efficiency in RADAR made during the war led to great progress in radio astronomy. More researchers started focusing on radio astronomy and more and bigger radio telescopes would follow.

1.2 Interferometry

In the quest for better sensitivity and higher angular resolution, radio telescopes were getting bigger. One example of this is the 305 meter Arecibo radio dish in Puerto Rico and the newly build FAST in China with a diameter of 500 meters. The angular resolution of a telescope depends on both the diameter of the dish and the observing frequency and is given by

θ= λ

D (1.1)

with θ the angular resolution in radians, λ the observed wavelength and D the di-ameter of the telescope. This tells us that for longer wavelengths the dish has to be larger if we want the same angular resolution. For example, to get an angular res-olution of 1 arcsecond at an observing wavelength of 500 nm, the diameter of the telescope has to be around 10 cm. In optical this is very easy to build and the very first telescopes build by Huygens in the 17th century were approximately this size. However, when we go down the electromagnetic spectrum to radio wavelengths, it gets more complicated. When observing at 21 cm, to obtain an angular resolution of 1 arcsecond, the diameter of the dish would have to be around 40 kilometers, difficult to build in your own backyard.

To get to higher angular resolutions, radio astronomers had to find an alternative way of building telescopes. Instead of building one single huge dish, the telescope can be divided in multiple smaller dishes. All the signals can then be combined to simulate a telescope the size of the entire array of dishes (Thompson, 1999). This is called aperture synthesis. The maximum angular resolution of the telescope now depends on the ’diameter’ of the entire array and equation1.1becomes

θ = λ

B (1.2)

with B the maximum baseline, the distance between the two dishes furthest apart. It exploits the wave properties of light. Assuming light from a distant radio source arrives parallel on Earth, two different antennas, 1 and 2, will detect the same signal, which will make a stronger signal when combined. These waves will however have a geometric delay given by τg = (~B·ˆs)/c, where~B is the baseline between the two antennas and ˆs the unit vector in the direction of the source. When~B is known precisely, the signal of one of the antennas can be delayed for the correction. This can be done by physically adding cable, but is mostly done by correcting it with software later. A schematic figure of two antennas is given in figure1.1. Here the signal received by antenna 1 is delayed by τg. The signals received in both antennas is then given by:

(6)

FIGURE1.1: Schematic overview of a two antenna telescope, combin-ing the signal of a distant point source.

These signals are then multiplied and averaged over time by a correlator to get the correlator response:

Rc = hV1V2i = V 2

2 cos(ωτg) (1.4)

For an extended source with specific intensity Iν at ν = ω/(2π) we integrate

over the total solid angle: Rc=

Z

Iνcos(2πν~B·ˆs/c)dΩ (1.5)

Now the intensity of the source is linked to the response of the interferometer. However, a cosine is an even function, so the response will only sample the even part of Iν. Since the total intensity of the source can be written as the sum of the odd

and the even part, it can be detected by adding a correlator with an odd output. This is done by simply adding a 90◦ phase delay to the output of the antenna and since cos(ωτg−π/2) =sin(ωτg)we get:

Rs= Z

Iνsin(2πν~B·ˆs/c)dΩ (1.6)

To make everything easier we can define a complex visibility Vν and using

Eu-ler’s formula we get:

Vν ≡Rc−iRs=

Z Iνe

(7)

Now the source intensity is linked to the response of the correlator. Using the Van Cittert-Zernike theorem (Thompson et al.,2017), we can now take the inverse Fourier transform of the visibility to get the brightness distribution of the source.

The signal of a distant point source causes an interference pattern with notice-able sidelobes in the case of a two antenna setup. We can reduce these sidelobes by adding more dishes with baselines differing in length and direction. One example of a radio telescope with different baseline lengths but all in the same direction is the Westerbork Synthesis Radio Telescope. It consists of 14 antennas of 25 meter di-ameter, all aligned in an East-West configuration. Instead of having antennas in the North-South direction, the telescope uses the Earth’s rotation to simulate a circular dish. This technique is called Earth-rotation aperture synthesis. Especially in the early days of interferometry when computing power was still very limited, the sim-ple symmetrical configuration of the dishes made the data reduction significantly easier. However, with the advancement of computers, new and more complex an-tenna configurations became possible.

A simple configuration results in a lot of redundancy in baselines. What a lot of the newer built telescopes use is a pseudorandom configuration, which tries to minimize the amount of similar baselines. An advantage of these configurations is that it’s less dependent on the Earth’s rotation to have different baselines in all directions, so less total observation time is needed. Examples of telescopes using a pseudorandom configuration are ALMA, LOFAR and MeerKAT. The last one being the telescope used in this research.

1.3 MeerKAT

The MeerKAT telescope (Jonas,2009; Jonas et al.,2018) is built as a pathfinder to the Square Kilometer Array (SKA), to test technologies needed for what will be become the most sensitive radio telescope with a total collecting area of around a square kilometer (Carilli and Rawlings, 2004). While first light for the SKA is still years away, MeerKAT saw its first light in 2016 and was inaugurated in 2018. It consists of 64 13.5 meter wide satellite dishes, which were expanded from the 7 dishes that made up its precursor, the KAT7 telescope, located at the same site. Once the SKA is completed, it will not only have great sensitivity but will also combine this with a very high angular resolution. It accomplishes this by its very long baselines of up to 3000 km with part of its stations in Australia.

The pseudorandom configuration is already utilised in the MeerKAT telescope, which has a dense core while also having long baselines up to 8 kilometers, as can be seen in figure1.2. This feature allows the telescope to observe a large field of view of over a square degree, while having a great angular resolution of 5 arcseconds. This makes the MeerKAT telescope perfect for large surveys. During the first few years of operation, 70% of observing time is allocated to large survey projects (LSP). This research is done as part of one these LSPs, the ThunderKAT program, The Hunt for Dynamic and Explosive Radio Transients with MeerkAT. (Fender et al.,2017). 1.3.1 ThunderKAT

ThunderKAT searches for radio transients in the image plane. It targets a variety of synchrotron sources like X-ray binaries and short gamma ray bursts on a weekly basis. This makes it possible to monitor closely how these sources evolve over time. Besides the data on these sources, we will also be able to search the rest of the images

(8)

FIGURE1.2: Positions of the antennas of MeerKAT. On the left the whole array and on the right zoomed-in on the dense core (Booth et

al.,2009).

being made. Because of the large field of view of MeerKAT, a lot of new interesting science can be done. This is where the commensal search program comes into play. This program aims to analyse all the data that’s already obtained and searches in the rest of the field for new transients.

Other blind searches are already being done with for example AARTFAAC (Prasad et al.,2014; Prasad et al.,2016) using LOFAR (Haarlem et al.,2013) and VAST (Mur-phy et al.,2013) using ASKAP (Johnston et al., 2008). LOFAR, the Low Frequency Array, is a radio telescope located mainly in the Netherlands with some stations spread across the rest of Europe. It observes the sky in the frequency range of 10-240 MHz and unlike most radio telescopes, it doesn’t use large movable dishes. Instead, it consists of around 20000 small and relatively simple omnidirectional antennas, meaning they are sensitive to almost the entire sky. This very wide field of view is utilized with AARTFAAC, which focuses on finding transients in real time.

On the other hand, ASKAP is more similar to MeerKAT. It is a precursor to the SKA as well, consisting of 36 antennas, each 12 meters in diameter. Each antenna has a phased array feed, increasing the field of view to up to 30 degrees2. This is significantly higher than MeerKATs field of view. However, with a higher sensitivity, MeerKAT will take less time per observation compared to ASKAP and therefore are very comparable in their survey capabilities.

With the weekly cadence of the ThunderKAT program, we can make an estima-tion of what kind of transients we are able to find. In figure1.3a parameter space is shown for all different radio transients with on the x-axis the observed frequency multiplied by the timescale of the transient and on the y-axis the luminosity. This gives us a nice distinction between the different kind of transients. Assuming a transient timescale of between a week and a couple of months and an observing fre-quency of around 1 GHz, we can get a quick overview of the different transients we are most likely to observe within the ThunderKAT program. This is marked as the red rectangle in figure1.3. Here we can see that flare stars (RSCVn/Algols), X-ray bi-naries, supernovae, Gamma Ray Bursts (GRBs) and Active Galactic Nuclei (AGNs)

(9)

fall within the ThunderKAT timescales.

FIGURE1.3: Parameter space of radio transients with on the x-axis the the frequency multiplied by the timescale and on the y-axis the luminosity of the source (Pietka et al., 2015). Shown in red is the

region of the parameter space where ThunderKAT will operate.

This research uses data obtained as part of the X-ray binary program within ThunderKAT, specifically Maxi J1820+070. This means that if we find a new tran-sient, it will vary on the timescales of weeks. In the next section we will discuss some possible transient sources and whether we can expect to find them as part of the ThunderKAT program.

1.4 Transients

When we think about the night sky, we think about the occasional planet and a vast number of long-lasting and unchanging stars. These stars have been a constant source of light for such a long time that they were named in constellations centuries ago which are still the same to this day. However, when taking a closer look, the sky seems to be full of activity. Stars moving around in the Milky Way, varying in brightness and exploding when they are at the end of their lives. Supermassive black holes swallowing entire stars in a matter of months. We typically can’t see these phe-nomena with the naked eye, but with the advancement of technology and better and more sensitive telescopes, we can observe them across not only the electromagnetic spectrum, but through gravitational waves as well.

For this research, a couple of the following transient phenomena may be relevant. 1.4.1 Supernovae

One of the more obvious transients is the supernova. A star over 10 times the mass of the Sun exploding at the end of their lifetime. These were already observed hun-dreds of years ago. Around the year 1000 a bright ’star’ appeared in the sky, even

(10)

visible during the day. This turned out to be a supernova of which we can still see the remnants today as the Crab Nebula.

FIGURE 1.4: Radio lightcurve of supernova SN 1994I showing the typical fast rise exponential decay found in explosive transients

(Weiler et al.,2011).

While the occurrence of a supernova in our own Milky Way is estimated to only be twice per century (Li et al., 2011), we currently observe multiple extragalactic supernovae per day. These are, however, observed in optical wavelengths. Radio telescopes are only sensitive enough to observe supernova up to around 100 Mpc. Therefore, most supernovae haven’t been detected with radio telescopes. In the past 20 years less than 30 radio supernovae have been detected (Weiler et al.,2002).

Radio ligthcurves for supernovae typically show a fast rise in brightness be-fore an exponential decay. This can clearly be seen in figure 1.4, where the radio lightcurve of SN1994I is plotted at a wavelength of 20 cm (1.5 GHz).

1.4.2 Fast Radio Bursts

In 2007, Duncan Lorimer and David Narkevic discovered something peculiar when looking through archival data of the Parkes Observatory. In the data obtained in 2001 they found a short, but bright millisecond radio burst across multiple wavelengths of unknown origin (Lorimer et al., 2007). This marked the start of a new field in radio transients. The burst, which is now known as the Lorimer burst, is shown in figure1.5. It shows a ’sweep’ across the different frequencies, where the signal arrives later the lower the frequency. This is called the dispersion delay and tells us how much neutral hydrogen is present between the source and the Earth. Because of this high dispersion delay, they concluded that the source was extragalactic.

Since then multiple of these, called fast radio bursts (FRB), were found, initially only as single events. This led to cataclysmic events, like supernovae and star merg-ers, being the most likely explanation of these bursts.

However, in 2015 the first repeating FRB was found, FRB 121102, coming from a faint dwarf galaxy 3 billion light-years away (Chatterjee et al., 2017). After its

(11)

FIGURE1.5: The radio burst found by Lorimer et al. (2007). It shows the total integrated pulse on the top right corner as well as the fre-quency evolution where the frefre-quency is plotted against time. The dispersion delay can be clearly seen across the frequency band, where

the peak flux arrives later for lower frequencies.

first detection in 2012, repeating bursts were detected in 2015, confirming the burst couldn’t be caused by a cataclysmic event as previously thought. It also enabled astronomers to localise the burst more precisely compared to single burst events.

Because of the short time span in which these bursts happen, FRBs will probably only be detected in the short timescale imaging searches. While FRB detections are still expected within ThunderKAT, another MeerKAT program MeerTRAP will most likely find hundreds of new FRBs (Stappers,2016).

1.4.3 Gamma Ray Bursts

Gamma Ray Bursts (GRBs) are extremely energetic and bright explosions originating from far outside of the Milky Way. They are non-repeating and last from a couple of milliseconds to a couple of minutes. The first GRB was discovered by US Vela satellites originally meant to observe space detonations of nuclear bombs. In 1967 they observed a gamma ray flash which couldn’t be explained by a detonation of a nuclear bomb. In the next three years sixteen of these bursts were detected which all had an extragalactic origin (Klebesadel et al.,1973).

Around 25% of all observed GRBs are classified as short GRBs, with a duration of up to 2 seconds (Zhang and Meszaros,2004). For a long time the origin of short GRBs was unknown, but one leading hypothesis was that these events were caused by the merger of two neutron stars or a black hole and a neutron star. It took till 2017 to confirm this when the first afterglow of the gravitational wave event GW170817, a merger between two neutron stars, was observed (Abbott et al.,2017). The other

(12)

75% are classified as long GRBs, with a duration of more than 2 seconds. These mostly originate from star-forming regions within galaxies and are associated with core collapse supernovae.

While GRBs are most luminous in X-ray, they can also be observed in other wave-lengths. The initial burst of X-ray emission is typically followed by fading emissions in other wavelengths known as the afterglow. This afterglow is caused by the ejecta of the burst interacting with interstellar matter, which can tell us a lot about the envi-ronment in which the GRB originated. However, these afterglows are very faint and the sensitivity of available telescopes limited the amount of afterglows which could be studied. With the higher sensitivity of MeerKAT, it will be possible to observe more faint GRB afterglows in the future.

1.4.4 Flare stars

Just like our Sun exhibits solar flares, other stars can flare as well. These stars can vary in brightness over the course of a couple minutes to hours, typically a fast in-crease followed by a slower dein-crease in brightness. These flares mainly occur in K and M-type stars with strong magnetic fields at their surface. They happen when charged particles are accelerated due to magnetic reconnection and interact with the plasma medium of the star. The best known flare star is UV Ceti, an M dwarf discov-ered in 1948 (Joy and Humason,1949). UV Ceti has since then become a designation for other similar flare stars.

These stellar flares emit radiation across the whole electromagnetic spectrum, so can show up in radio as well. In 2018, the first radio transient was discovered by MeerKAT. This transient seems to be caused by a stellar flare from a K-star Driessen et al. (2020). The lightcurve is shown in figure1.6. Here an initial bright burst was detected, followed by some a variable decrease in brightness. It is argued that the variability is caused by star spots on the surface, but further investigation is still needed to determine the true nature of the transient.

FIGURE1.6: Radio lightcurve of MKT J170456.2-482100, the first radio transient discovered with MeerKAT (Driessen et al.,2020).

1.4.5 X-ray binary and AGN variability

An X-ray binary is a binary system which is shining bright in X-rays. It consists of one compact object, a stellar mass black hole or a neutron star, and a donor star. When the donor star comes within the Roche limit of the compact object, matter

(13)

FIGURE1.7: Radio/X-ray correlation for the X-ray binary GX339-4, Sgr A and a sample of AGNs (Falcke et al.,2004).

starts transferring onto the compact object (Schatz and Rehm, 2006). It forms an accretion disk around the compact objects heating up the matter the closer it gets. Around the innermost part of the accretion disk, the matter heats up so much due to friction it starts to emit X-ray. On the other hand, this accretion can also cause the formation of jets perpendicular to the accretion disk. Within these jets charged particles are accelerated to relativistic velocities causing synchrotron emission which can be observed in radio. A similar mechanism applies to AGNs. Here, matter is accreting onto a supermassive black hole in the center of a galaxy, causing it to shine bright in X-ray and radio.

This relationship between the accreting matter and the jet is evident in the radio/X-ray correlation shown in figure1.7. This is typically used to categorize and distin-guish X-ray binaries and AGNs. Here the X-ray binary GX339-4 is shown in the lower left corner. AGNs are inherently more luminous in both radio and X-ray and therefore are located in the top right corner of the plot.

1.4.6 Tidal disruption events

A tidal disruption event happens when a star approaches a supermassive black hole at such a distance that it is ripped apart and swallowed by the black hole (Rees, 1988). This will increase the amount of matter accreting onto the black hole and therefore increasing its luminosity in both X-ray and radio wavelengths. The flares caused by the relativistic jets have so far only been detected in optical and X-ray and only later have been followed up by radio observations. With the increased sensitivity of the MeerKAT and later the SKA, it’s predicted that more of these events will be discovered in surveys.

A well-studied tidal disruption event is Swift J1644+57, shown in figure1.8 in red. It was originally discovered as a GRB known as GRB 110328A. It was later found that a star getting torn apart by a 106to 107solar mass black hole caused the GRB (Bloom et al.,2011). Similar to other explosive transients, the lightcurve has a fast rise followed by exponential decay.

(14)

FIGURE 1.8: Radio lightcurves of several tidal disruption events (Alexander et al.,2017).

1.4.7 Lensing effects

Finally, lensing effects are a different kind of transient compared to ones mentioned above. Instead of being of an astrophysical origin like a supernova or binary merger, with a lensing effect we observe a constant background source where a plasma lens moves across the line of sight. Depending on the dimensions and trajectory of the lens it will bend radiation coming from the background source causing variations in brightness.

In figure1.9both a theoretical model as well as observations of this lensing effect is shown. The left shows the effect different paths of the lens have on the lightcurve, with the path color coded from red to blue. The dark red line with b/ri → 1, cor-responds to only the very edge of the lens moving in front of a background source. Dark blue with b/ri →0 corresponds to the center of the lens moving in front of the background source.

These effects are also seen in observations, of which one is shown on the right. Here the lightcurves of quasar 0954+658 in 2.25 GHz (blue) and 8.3 GHz (orange) are plotted. Especially in the lower frequencies, a significant dip surrounded by two peaks in brightness can be seen. What also stands out is how frequency dependent this effect is. At lower frequencies it follows the model closely, but at higher frequen-cies the lightcurve has a very irregular shape.

(15)

(A) (B)

FIGURE1.9: On the left a model describing the effect on the lightcurve of a background source by an axisymetrical plasma lens moving in front of it. On the right, observations of a plasma lensing effect seen

(16)

Chapter 2

Data Analysis

2.1 MeerKAT observations

As discussed in section1.3.1, the ThunderKAT program observes its targets every week. This is however on the condition that the source of interest is active and can be detected by MeerKAT. For this research we used the observations made by Bright et al.,2020, who monitored the X-ray binary MAXI J1820+070 (hereafter, J1820). This means that the observations were carried out only when J1820 was active. The first set of observations with the MeerKAT telescope of J1820 were obtained between September 28, 2018 and February 1, 2019 on a weekly basis and an integration time of 15 minutes each. After this period of activity, J1820 went into quiescence until the beginning of March. After another outburst, weekly monitoring started again on March 9, 2019, lasting until May 25, 2019. Another period of quiescence followed, until August 10, 2019, when weekly monitoring started once again. The observa-tions were taken at a central frequency of 1.28 GHz with a bandwidth of 0.86 GHz consisting of 4096 channels. The data was flagged for RFI in CASA and AOFlagger. Bandpass and complex gain calibration was also done in CASA, as well as the flux scaling. Then the data was averaged in time and frequency for imaging. To image the whole field WSClean was used with uniform weighting. More details of the observations can be found in Bright et al.,2020.

While the uniform weighting applied to the data, also known as direction in-dependent calibration, suffices for studying J1820, it can be quite problematic for commensal searches when studying objects in the rest of the image. Because of the relatively large field of view of the MeerKAT telescope, distortions caused by differences in the Earth’s atmosphere will vary across the image plane. Therefore, when the same calibration is applied to the entire image, parts of the image may be wrongly calibrated. This can increase the strength of artifacts around the outer edges of the image. To combat this, J. Bright applied direction dependent calibration to the image. Here the image is divided in multiple smaller areas to which each a different calibration is applied depending on the smaller areas itself. The algorithm used here isDDFACETand is described in detail in Tasse et al.,2018.

The effect of this is shown in figure 2.1. Here on the left, a part of the field is shown of the direction independent calibrated image. It shows a lot of sidelobes caused by bright sources in the field, mainly the two at the top left of the image. The combination of these artifacts of different bright sources alters the brightness on other parts of the image causing flux readings to differ from reality. On the right the same part of the image is shown, but now with direction dependent calibration applied to the image. It is immediately noticeable that the sidelobes are reduced significantly.

Another correction that needs to be done is a primary beam correction. The sen-sitivity of a radio telescope changes depending on what part of the field you look at.

(17)

FIGURE2.1: Part of the field centered on the transient, with on the left the direction independent calibrated image and on the right the direction dependent calibrated image. The sidelobes caused by bright source seen on the left disappear when direction dependent

calibra-tion is applied.

It will be most sensitive around the center of the field and drops towards the outer edges of the field. Therefore, the flux readings will be lower than the actual flux for sources further from the center of the image.

2.2 The LOFAR Transients Pipeline

To make the identification of transients easier, we used the LOFAR Transients Pipeline (TraP; Swinbank et al., 2015). This pipeline was developed to support the au-tomation needed when handling the massive amounts of data produced by LOFAR. While specifically build for LOFAR, the pipeline can handle radio images from other telescopes as well as long as it’s in the right format, making it a useful tool in all of radio transient research.

An overview of the pipeline is shown in figure2.2. It takes multi frequency image cubes with two spatial dimensions and an optional monitoring list to track specific target locations as input. The data goes through a series of quality control steps and images identified as poor quality are excluded from the further processing. Then a source-finding process is applied which identifies pixel ’islands’ above a certain level of noise which are fitted with an elliptical Gaussian. These found sources are then associated across the different images, either as a new source or as an update to previously found sources. For sources appearing in earlier images, but not present in the image being analysed an extra null measurement is made so an upper limit can be taken. It then stores the lightcurves along with a series of calculated properties and variability parameters of the sources in a database.

The two variability parameters η and V are very useful in identifying transients in a certain field. The parameter η is the reduced χ squared value compared to a constant source with the brightness of the weighted mean flux density over N measurements and is defined as

ην = 1 N−1 N

∑

i=1 (Iν,i−ξIν) 2 σ_ν2_,i (2.1)

with Iνthe flux density at a frequency ν, ξIν the weighted mean flux density and

(18)

FIGURE 2.2: Schematic representation of the LOFAR Transients Pipeline (Swinbank et al.,2015).

a high value for η indicates an increased variation compared to a stable source. V is the coefficient of variation and is the defined as the ratio between the mean flux density and the standard deviation:

Vν = s Iν = 1 Iν r N N−1(I 2 ν−Iν 2 ) (2.2)

with s the standard deviation. Therefore, V indicates the magnitude of the variation compared to its mean flux density.

In this research the default settings for TraP were used, except for the following parameters: Setting Value extraction_radius 3600 force_beam True beamwidths_limit 3 new_source_sigma_margin 0

(19)

2.3 The MAXI J1820 field

The distribution of values for η and V for all the identified sources in the field of MAXI J1820 is shown in figure2.3. Three sources immediately stand out; the one with the highest value for both η and V (circled in red) corresponds to the X-ray binary MAXI J1820+070, a well-studied object and the original science goal of these observations. Its outburst causing these high values for η and V is described in Bright et al.,2020. A second one to the left of J1820 (circled in purple) coincides with an unknown object which only appeared in one of the observations and four times as non-detections. Further research will be necessary to identify whether this was an actual transient. A third interesting source is identified as an outlier in both η and V and circled in green. The parameters for this source are η = 96 and V = 0.355 and it is located at 18h18m49.7 s and 06h28m45.1 s right ascension and declination respectively. Its lightcurve is shown in figure2.4and clearly shows a variable source.

FIGURE2.3: Distribution of the η and V variability parameters in the MAXI J1820 field

When we take a closer look at the transient, we see something peculiar; the source brightens and dims over the course of two months in a very Gaussian way. After returning to flux levels similar to its initial flux, it doubles in brightness again for a period of around 2 months. This time however, the Gaussian features seen in the first peak seem to be absent and the flux seems to vary around a higher flux

(20)

level. So the symmetry we see in the first peak seems to be absent in the second one. AppendixBincludes the flux measurements for the transient source as measured by TraP.

FIGURE2.4: Lightcurve of the radio transient in the MAXI J1820 field. In blue the flux measured by MeerKAT and in orange the flux after

applying the corrections described in equations2.3and2.4.

To check whether this is an actual astrophysical transient we can take a look at the lightcurves of nearby sources. In figure2.5the lightcurves of six nearby sources are plotted. These are all the sources that lie within a 0.3◦ x 0.3◦area of the transient source. Although some variation in flux can be seen in the other sources, the flux of the transient clearly stands out. These nearby sources were also used to make a small correction to the flux of the transient, trying to minimize the periodic flux variation between epochs due to systematic errors. To make this correction we calculated a scaling factor per epoch, which depends on the average flux per source:

An,ν(t) = Fn,ν(t) −Fn,ν

Fn,ν +1 (2.3)

with An,ν(t)the scaling factor of a source n at frequency ν and epoch t, Fn,ν(t) the uncorrected flux of source n at frequency ν and epoch t and Fn,ν the mean flux of source n at frequency ν taken over all epochs. Now we take the mean of the scaling factors per frequency and epoch to get the scaling factor corresponding to a frequency and epoch:

Aν(t) = An,ν(t) (2.4)

To calculate the corrected flux of the source we multiply the scaling factor by the flux. The result of this correction for the lightcurve of the transient can be seen in figure2.4marked in orange.

2.4 Comparison to other catalogs

In the Pan-STARRS catalog (Chambers et al., 2016) two objects are present at the location of the transient, one of magnitude 21 on the left and one of magnitude 19 on the right. Since the source in the radio image overlaps with both of these sources, it is unclear to which source in Pan-STARRS it corresponds. However, a radio source at the transient location is also present in the VLASS catalog (Lacy et al.,2016). The angular resolution of VLASS is around 2.5 arcseconds compared to 5 arcseconds for

(21)

FIGURE2.5: Uncorrected lightcurves of nearby sources with the tran-sient source in blue. The flux variation of the trantran-sient source is not

seen in the other nearby sources.

the MeerKAT. This allows us to constrain the position of the source better and we can conclude that the radio transient originated from the left fainter source. Figure 2.6shows a side-by-side comparison of the MeerKAT image, VLASS image and Pan-STARRS. All are centered around the transient source and clearly show that the left faint source in Pan-STARRS corresponds to the transient radio source in both the MeerKAT and VLASS images.

FIGURE2.6: Side by side comparison of the MeerKAT image (left), VLASS (middle) and Pan-STARRS (right).

ASAS-SN (Shappee et al.,2014; Kochanek et al.,2017) shows an object varying in brightness around the same coordinates as well. However, the angular resolution of the ASAS-SN observations is larger than the separation between the two objects. This combined with the fact that the source is three orders of magnitude brighter, we can conclude that this variation is coming from the source on the right.

2.5 Optical observations

After the detection of the transient with MeerKAT, we observed the area around the source with the South African Large Telescope (SALT) (Stobie et al., 2000). SALT is the largest single optical telescope in the southern hemisphere with a diameter of 11 meter and is designed for spectroscopy. Since the right source is a couple of magnitudes brighter than the other source, the SALT observation only shows the spectrum of the bright source. From this spectrum we concluded that this is a K-star at a distance of 2808+₋3047₁₅₉₈ pc (Bailer-Jones et al., 2018). Unfortunately, we weren’t able to obtain a spectrum of the transient source.

(22)

2.6 X-ray observations

The left source in the Pan-STARRS image is also found in the WISE AGN candidate catalog (Assef et al.,2018) and is classified as an AGN candidate based on mid-IR photometry. To confirm the source as an AGN, we requested observation time on Swift’s XRT (Gehrels et al., 2004; Burrows et al., 2005). Compared to other X-ray telescopes, Swift has a very high slew rate, making it perfect for short observations. As mentioned in section1.4.5 there is a relationship between the radio and X-ray radiation emitted by an XRB or AGN, with the X-ray emission coming from the inner regions of the accretion disk and the radio emission coming from the jet. We can use this relation to make an estimation of the expected X-ray flux coming from the transient source. Corbel et al.,2013found this relation with Frad the radio flux density at 9 Ghz and FXthe 3-9 KeV X-ray flux to be

F_rad ∝ a F_Xb (2.5)

with a=1.85 and b=0.62. This correlation was found using observations of the X-ray binary GX 339-4 and therefore wouldn’t immediately apply to AGNs. However, to make a rough estimation we can use the similarity between AGNs and galactic black holes shown in McHardy et al.,2006. We can estimate the expected X-ray flux to be around 8·10−12 _{erg s}-1_cm-2 _{assuming a radio spectral index for synchrotron} sources of -0.7. This corresponds to a predicted 0.4 counts/sec for the 0.2-10 KeV band of XRT assuming an X-ray spectral index of 2.

FIGURE2.7: SWIFT X-ray observation on the left and the MeerKAT observation on the right. 4 detections can be seen in X-ray that

origi-nate from around the transient source.

A total observation time of 819 seconds was obtained and was carried out on November 6, 2019. To analyse the X-ray data and to obtain a flux measurement we usedSOSTA within the X-ray image packageXIMAGE (Giommi et al., 1991). After specifying a box, the program performs a statistical analysis on that part of the im-age. It uses the background enclosed by this box to estimate the source intensity and its statistical significance. A total of 4 detections were found around the source with a probability of it being a fluctuation of the background of 8·10−4_{. This corresponds} to a source intensity of 0.0056 counts/sec. When assuming a photon index of 2, this amounts to 2.1±1.2·10−13ergs cm−2s−1in the 0.2-10 KeV band and 8.5±4.7·10−14 ergs cm−2_s−1 _{in the 3-9 KeV band. An image of the X-ray observation can be seen} in figure2.7.

(23)

Chapter 3

Discussion

As part of the ThunderKAT program for MeerKAT we found a new radio transient. The transient is quite different compared to other known radio transient sources. While a typical transient shows a fast brightening followed by an exponential de-cay, here the first peak has a very Gaussian-like lightcurve. Over the course of two months, it doubles in brightness before returning to its original brightness. Using data from the PAN-STARRS catalog, VLASS and observations taken with SWIFT, the source seems to be consistent with an AGN. We can divide the nature of the transient into two separate phenomena, intrinsic and extrinsic variability.

3.1 Extrinsic variability

First, we take a look at extrinsic variability. In this case the radiation emitted by the source is assumed to be constant, but because of some interference in its path to Earth, the source appears to be variable. One way this can be caused is by dis-turbances in the Earth’s atmosphere. Differences in air density and temperature can cause scintillation effects in radiation coming from distance sources. How-ever, this effect is only noticeable on smaller timescales (seconds) and not on the weekly timescale where ThunderKAT operates. Also, if local atmospheric distur-bances cause this variability, it should be seen in other parts of the field as well, especially sources surrounding the transient. As shown in figure 2.5, the nearby sources show no similar behaviour. Therefore, scintillation can’t explain the cause of the transient found in this research.

A different kind of extrinsic variability that can cause similar variability is a plasma lensing effect as discussed in section1.4.7. For a specific configuration, when the very edge of a plasma lens moves in front of a background source, the resulting lightcurve can have a Gaussian shape. This could explain the first peak, but as we see in figure2.4the source brightens again after a period of around two months, now with a very different shape. A typical lightcurve of a lensing event shows two sepa-rate peaks when it’s not the very edge of the lens moving in front of the background source. This is shown in figure1.9afor b/ri . 0.9. However, these lightcurves are expected to be very symmetrical, with a significant dip in brightness between the two peaks, especially for b/ri . 0.7. This is confirmed in multiple observations of these lensing events, one of which is shown in figure1.9b.

When looking at the lightcurve of the transient in the J1820 field, we see two sep-arate peaks. However, the dip between these two peaks has a similar flux level as before the first peak. Furthermore, the second peak has a very different shape com-pared to the first one. Both the symmetry as well as the significant dip in brightness seem to be a characteristic property of these lensing effects, as shown in Fey et al., 1996and Fiedler et al., 1994. Therefore, we can rule out a plasma lensing effect as the possible source of this transient.

(24)

3.2 Intrinsic AGN variability

Another possible explanation for this transient is inherent AGN flaring. In other words, the source itself causes the variation in brightness, not the medium through which the emitted radiation travels. This variability is most likely caused by differ-ences in the amount of matter accreting onto the black hole. In figure3.1the radio emission of X-ray binary Cyg X-3 is shown over the period of 100 days. While flares I-IV are shorter, flare V has a duration of around 10 days, comparable to the dura-tion of the MeerKAT transient. Also the shape of the peak is similar. If we consider an AGN as a scaled-up version of an X-ray binary, the transient can be caused by intrinsic flaring of the AGN similar to the flare seen in Cyg X-3.

FIGURE3.1: Radio flaring seen in Cyg X-3 (Fender et al.,1997). While the duration of flares I-IV are shorter, flare V has a similar Gaussian shape and comparable duration to the MeerKAT transient, especially

at the 13.3 cm wavelength.

An example of an AGN flaring is shown in figure3.2. It shows the variability of the blazar J1043+2408 over the period of 10 years. We can see peaks in brightness in similar shapes but on larger, multiple month timescales compared to the MeerKAT transient. Also, this blazar shows a strong periodicity with constantly varying flux levels. In the MeerKAT observations we currently have, the transient shows a period of stable flux at the end of 2019. Therefore, it will be interesting to see if more recent observations show whether there is any actual periodicity. On the other hand, Ven-turi et al.,2001shows a population of blazars where very little periodicity is found. One of these blazars is shown in figure 3.3. It shows a blazar with initially what seems to be a stable flux period, followed by a single year-long peak in brightness. While the timescale here is higher compared to the month-long timescales of the MeerKAT transient, the shape of the lightcurve is very similar. Both these examples show that the Gaussian-like variability isn’t completely uncommon in AGNs.

So now that we’ve seen that the radio emission is consistent with other AGN flaring events, we can check if the X-ray observations are also consistent with X-ray emissions from other AGNs. With SWIFT, on November 6, 2019 we observed X-ray emission of 8.5±4.7·10−13 ergs cm−2 s−1 in the 3-9 keV band assuming a photon index of 2. The probability of it being a fluctuation of the background is 8·10−4_,

(25)

FIGURE3.2: Lightcurve of the blazar J1043 over the period of 10 years at 15 GHz(Bhatta,2018), showing similar shaped peaks compared to

the MeerKAT transient.

FIGURE3.3: Lightcurve of J0238+166 at 8 GHz showing a single year-long peak (Venturi et al.,2001)

corresponding to a 3.4σ detection. Combining this with the closest MeerKAT, obser-vation we can see how the radio emission correlates with the X-ray emission. Us-ing these quasi simultaneous observations, we can compare the source to the other AGNs shown in figure1.7. Since we don’t know the distance to the source, we can’t calculate the actual luminosity, which we need if we want to make the comparison. We can however estimate the luminosities for a range of distances, typical for AGNs. To calculate the luminosity L at distance d we use

L=4πd2F (3.1)

with F the measured flux. For the estimation we assumed a distance to the source between 1 Mpc and 10 Gpc. The observations used in Falcke et al.,2004were at 5 GHz. Therefore, we have to extrapolate the MeerKAT flux values, which were taken at 1.28 GHz, to 5 GHz. We do this using the formula

F5 GHz= F1.28 GHz· 5 1.28 α (3.2) with α the spectral index. Currently, we don’t have an accurate estimate of the radio spectral index of the source, so for calculating the correlation we assume −1.5 < α< 1. Because of the low luminosity of the source in X-ray and the limited photon counts with SWIFT’s XRT, we also don’t have an accurate measurement of the X-ray photon indexΓ. Therefore, we assume 1 < Γ < 2. Appendix Ashows

(26)

FIGURE3.4: Similar to figure1.7, but with the possible correlation for the transient source added in gray. Here we assume a radio spectral index between -1.5 and 1, an X-ray photon index between 1 and 2 and

a distance to the source between 1 Mpc and 10 Gpc.

the code implementation of this estimation. The result of the estimation is shown in figure3.4. Here the same AGNs as in figure 1.7 are plotted and overlapped by the possible correlations for the transient source, shown in grey. While the spread of the correlation for the transient source is large, it clearly shows it lies within the area expected for AGNs. Therefore, the X-ray emission we observed is consistent with the source being an AGN.

3.3 Direction dependent calibration

A different point we want to argue is the importance of direction dependent cali-bration and primary beam correction. As shown in section2.1 and figure 2.1, di-rection dependent calibration highly suppresses the noise caused by bright sources in a field. For the initial identification of new transients this offers no real limita-tions, since these can still be identified in the direction independent calibrated data. However, for further investigation it will be necessary to get the correct flux mea-surements. Therefore, in addition to the direction dependent calibration, a primary beam correction will also be necessary.

(27)

Chapter 4

Conclusion

We report the discovery of a new radio transient from MeerKAT observations be-tween September 18, 2018 and October 19, 2019. In contrast to the typical fast rise exponential decay of explosive transients, the transient source shows an initial two month-long 6 mJy peak in brightness with a Gaussian-like shape. After another two months, a second increase in brightness follows. Now with a different pattern where the flux fluctuates between 3 mJy and 4.5 mJy. At the end of 2019, the flux level sta-bilizes below 2 mJy.

Besides the MeerKAT data, we observed the transient source with SWIFT’s XRT. Although faint, a 3.4σ detection of X-ray emission is observed at the location of the source. Combining this with data from PAN-STARRS and VLASS and comparing the radio/X-ray correlation with other AGNs, we conclude that the transient source is consistent with an AGN. We argue that the cause of the transient is intrinsic to the AGN itself and not a result of extrinsic variability like lensing and scintillation effects.

Additionally, we argue that the use of direction dependent calibration and pri-mary beam correction is necessary for future commensal searches within the Thun-derKAT program. Not for the initial discovery of new transient sources, but for further analysis of these sources.

A lot about the transient source found in this research is still unknown. Not yet all MeerKAT observations available of the field were analysed in this research. As mentioned before, the more recent observations will probably give us more insight in the periodicity of the source or confirm the transient as a onetime event. Besides the radio observations, we are also planning to observe the source in optical. At the moment the radio/X-ray correlation lies within the range of other AGNs but is still unconstrained. This is mostly because we don’t know the distance to the source. To make an estimation of the distance, we are planning to observe the source with the VLT in optical wavelengths. This will enable us to obtain a redshift and thus constrain where the source lies on the radio/X-ray correlation even further.

(28)

Acknowledgements

First, I want to thank Antonia Rowlinson for her guidance during the project. I really appreciate the advice she gave me, both in and outside of doing research. I also want to thank everyone from the API Radio Transients Group for all their ideas, interesting discussions and helpful feedback over the course of this project. Also a big thank you to the ThunderKAT group, especially Joe Bright, who did a lot of the data reduction needed for this research.

(29)

Appendix A

Code

A.1 Radio/X-ray correlation estimation

import numpy as np

import matplotlib.pyplot as plt

def calculate_flux(flux, goal_freq, observed_freq, index):

"""

Extrapolates the flux at the observed frequency to a different frequency according to a specified spectral index.

"""

return flux * (goal_freq / observed_freq)**index

def calculate_luminosity(distance, flux):

"""

Calculate the luminosity by multiplying the observed flux with the surface area of a sphere with the distance as radius. """

return 4 * np.pi * distance**2 * flux

def parsec_to_centimeter(parsec):

return 3.1E18 * parsec

# Radio flux observed with MeerKAT at 1.28 GHz

mkt_rad = 4.4E-26

# X-ray fluxes for photon indices between 1 and 2

swift_xrays = [2.57E-13, 2.34E-13, 2.12E-13, 1.92E-13, 1.73E-13, 1.55E-13,

1.39E-13, 1.24E-13, 1.097E-13, 9.7E-14, 8.55E-14]

# Range of values for the radio spectral index between -1.5 and 1

rad_indices = np.linspace(-1.5, 1, num=101)

# Range of possible distances to the source between 1 Mpc and 10 Gpc

distances = np.geomspace(parsec_to_centimeter(1.0E6),

(30)

# Initiate lists for luminosities

radio_lums = [] xray_lums = []

# Start looping over all values for the distance and the X-ray and # radio luminosities

for swift_xray in swift_xrays:

for distance in distances:

# Multiply by 10E7 to correct for AGN mass

xray_lum = calculate_luminosity(distance, swift_xray) * 1.0E7

for index in rad_indices:

rad_flux = calculate_flux(mkt_rad, 5, 1.28, index)

radio_lum = calculate_luminosity(distance, rad_flux) * 5.0E9

radio_lums.append(radio_lum) xray_lums.append(xray_lum)

(31)

Appendix B

Tables

B.1 MeerKAT measurements

Observation Frequency Flux Flux error

date (GHz (mJy) (mJy)

2018-09-28 1.284 0.002594 0.000051 2018-10-05 1.284 0.002350 0.000065 2018-10-12 1.284 0.002811 0.000054 2018-10-19 1.284 0.002606 0.000091 2018-10-27 1.284 0.003754 0.000067 2018-11-03 1.284 0.005234 0.000057 2018-11-10 1.284 0.005515 0.000071 2018-11-17 1.284 0.006003 0.000084 2018-11-24 1.284 0.005724 0.000078 2018-12-02 1.284 0.005090 0.000098 2018-12-08 1.284 0.005027 0.000073 2019-01-05 1.284 0.002392 0.000054 2019-01-12 1.284 0.001809 0.000079 2019-01-19 1.284 0.002010 0.000068 2019-01-26 1.284 0.002023 0.000071 2019-02-01 1.284 0.001919 0.000079 2019-03-09 1.284 0.003303 0.000073 2019-03-18 1.284 0.004042 0.000055 2019-03-25 1.284 0.003802 0.000063 2019-04-01 1.284 0.004121 0.000058 2019-04-09 1.284 0.004596 0.000073 2019-04-15 1.284 0.003800 0.000065 2019-04-20 1.284 0.003254 0.000140 2019-04-29 1.284 0.003545 0.000050 2019-05-04 1.284 0.003202 0.000071 2019-05-11 1.284 0.003331 0.000077 2019-05-18 1.284 0.003651 0.000103 2019-05-25 1.284 0.003614 0.000070 2019-08-10 1.284 0.001600 0.000104 2019-08-16 1.284 0.001433 0.000095 2019-08-23 1.284 0.001769 0.000067 2019-08-31 1.284 0.001620 0.000070 2019-09-14 1.284 0.001563 0.000052 2019-09-29 1.284 0.001458 0.000096 2019-10-19 1.284 0.001714 0.000101

(32)

Bibliography

Abbott, Benjamin P et al. (2017). “GW170817: observation of gravitational waves from a binary neutron star inspiral”. In: Physical Review Letters 119.16, p. 161101. Alexander, KD et al. (2017). “Radio Observations of the Tidal Disruption Event XMMSL1

J0740–85”. In: The Astrophysical Journal 837.2, p. 153.

Assef, RJ et al. (2018). “The WISE AGN Catalog”. In: The Astrophysical Journal Supple-ment Series 234.2, p. 23.

Bailer-Jones, CAL et al. (2018). “Estimating distance from parallaxes. IV. Distances to 1.33 billion stars in Gaia data release 2”. In: The Astronomical Journal 156.2, p. 58. Bhatta, Gopal (2018). “Detection of periodic radio signal from the blazar J1043+

2408”. In: Galaxies 6.4, p. 136.

Bloom, Joshua S et al. (2011). “A possible relativistic jetted outburst from a massive black hole fed by a tidally disrupted star”. In: Science 333.6039, pp. 203–206. Booth, RS et al. (2009). “MeerKAT key project science, specifications, and proposals”.

In: arXiv preprint arXiv:0910.2935.

Bright, JS et al. (2020). “An extremely powerful long-lived superluminal ejection from the black hole MAXI J1820+ 070”. In: Nature Astronomy, pp. 1–7.

Burrows, David N et al. (2005). “The Swift X-ray telescope”. In: Space science reviews 120.3-4, pp. 165–195.

Carilli, Christopher and Steve Rawlings (2004). “Science with the Square Kilometer Array: Motivation, key science projects, standards and assumptions”. In: arXiv preprint astro-ph/0409274.

Chambers, KC et al. (2016). “The pan-starrs1 surveys”. In: arXiv preprint arXiv:1612.05560. Chatterjee, S et al. (2017). “A direct localization of a fast radio burst and its host”. In:

Nature 541.7635, pp. 58–61.

Corbel, S et al. (2013). “The ‘universal’radio/X-ray flux correlation: the case study of the black hole GX 339- 4”. In: Monthly Notices of the Royal Astronomical Society 428.3, pp. 2500–2515.

Dong, Lingyi, Maria Petropoulou, and Dimitrios Giannios (2018). “Extreme scatter-ing events from axisymmetric plasma lenses”. In: Monthly Notices of the Royal Astronomical Society 481.2, pp. 2685–2693.

Driessen, LN et al. (2020). “MKT J170456. 2–482100: the first transient discovered by MeerKAT”. In: Monthly Notices of the Royal Astronomical Society 491.1, pp. 560– 575.

Falcke, Heino, Elmar Körding, and Sera Markoff (2004). “A scheme to unify low-power accreting black holes-Jet-dominated accretion flows and the radio/X-ray correlation”. In: Astronomy & Astrophysics 414.3, pp. 895–903.

Fender, Rob et al. (2017). “ThunderKAT: the MeerKAT large survey project for image-plane radio transients”. In: arXiv preprint arXiv:1711.04132.

Fender, Robert P et al. (1997). “Cygnus X-3 in outburst: quenched radio emission, radiation losses and variable local opacity”. In: Monthly Notices of the Royal Astro-nomical Society 288.4, pp. 849–858.

(33)

Fey, Alan L, Andrew W Clegg, and Ralph L Fiedler (1996). “VLBI Observations of Eight Extreme Scattering Event Sources: Milliarcsecond-Scale Structure”. In: The Astrophysical Journal 468, p. 543.

Fiedler, R et al. (1994). “A summary of extreme scattering events and a descriptive model”. In: The Astrophysical Journal 430, pp. 581–594.

Gehrels, Neil et al. (2004). “The Swift gamma-ray burst mission”. In: The Astrophysical Journal 611.2, p. 1005.

Giommi, P, L Angelini, and N White (1991). The XIMAGE User’s Guide.

Haarlem, Michael P van et al. (2013). “LOFAR: The low-frequency array”. In: Astron-omy & astrophysics 556, A2.

Jansky, Karl G (1933). “Electrical disturbances apparently of extraterrestrial origin”. In: Proceedings of the Institute of Radio Engineers 21.10, pp. 1387–1398.

Johnston, Simon et al. (2008). “Science with ASKAP”. In: Experimental astronomy 22.3, pp. 151–273.

Jonas, Justin et al. (2018). “The MeerKAT radio telescope”. In: MeerKAT Science: On the Pathway to the SKA. Vol. 277. SISSA Medialab, p. 001.

Jonas, Justin L (2009). “MeerKAT—The South African array with composite dishes and wide-band single pixel feeds”. In: Proceedings of the IEEE 97.8, pp. 1522–1530. Joy, AH and ML Humason (1949). “Observations of the faint dwarf star L 726-8”. In:

Publications of the Astronomical Society of the Pacific 61, pp. 133–134.

Klebesadel, Ray W, Ian B Strong, and Roy A Olson (1973). “Observations of gamma-ray bursts of cosmic origin”. In: The Astrophysical Journal 182, p. L85.

Kochanek, CS et al. (2017). “The all-sky automated survey for supernovae (ASAS-SN) light curve server v1. 0”. In: Publications of the Astronomical Society of the Pacific 129.980, p. 104502.

Lacy, Mark et al. (2016). “The VLA Sky Survey (VLASS): Description and Science Goals”. In: American Astronomical Society Meeting Abstracts# 227. Vol. 227, pp. 324– 09.

Li, Weidong et al. (2011). “Nearby supernova rates from the Lick Observatory Super-nova Search–III. The rate–size relation, and the rates as a function of galaxy Hub-ble type and colour”. In: Monthly Notices of the Royal Astronomical Society 412.3, pp. 1473–1507.

Lorimer, DR et al. (2007). “A bright millisecond radio burst of extragalactic origin”. In: Science 318.5851, pp. 777–780.

Maxwell, James Clerk (1865). “VIII. A dynamical theory of the electromagnetic field”. In: Philosophical transactions of the Royal Society of London 155, pp. 459–512.

McHardy, IM et al. (2006). “Active galactic nuclei as scaled-up Galactic black holes”. In: Nature 444.7120, pp. 730–732.

Murphy, Tara et al. (2013). “VAST: an ASKAP survey for variables and slow tran-sients”. In: Publications of the Astronomical Society of Australia 30.

Pietka, M, RP Fender, and EF Keane (2015). “The variability time-scales and bright-ness temperatures of radio flares from stars to supermassive black holes”. In: Monthly Notices of the Royal Astronomical Society 446.4, pp. 3687–3696.

Planck, Max (2013). The theory of heat radiation. Courier Corporation.

Prasad, P et al. (2014). “Real-time calibration of the AARTFAAC array for transient detection”. In: Astronomy & Astrophysics 568, A48.

Prasad, Peeyush et al. (2016). “The AARTFAAC all-sky monitor: System design and implementation”. In: Journal of Astronomical Instrumentation 5.04, p. 1641008. Reber, Grote (1940). “Cosmic static”. In: Proceedings of the IRE 28.2, pp. 68–70.

Rees, Martin J (1988). “Tidal disruption of stars by black holes of 10 6–10 8 solar masses in nearby galaxies”. In: Nature 333.6173, pp. 523–528.

(34)

Schatz, Hendrik and KE Rehm (2006). “X-ray binaries”. In: Nuclear physics A 777, pp. 601–622.

Shappee, Benjamin J et al. (2014). “The man behind the curtain: X-rays drive the UV through NIR variability in the 2013 active galactic nucleus outburst in NGC 2617”. In: The Astrophysical Journal 788.1, p. 48.

Stappers, BW (2016). “MeerTRAP: Real time commensal searching for transients and pulsars with MeerKAT”. In: Proc. Sci.(MeerKAT2016) 10.

Stobie, Robert, Jacobus G Meiring, and David AH Buckley (2000). “Design of the Southern African Large Telescope (SALT)”. In: Optical Design, Materials, Fabrica-tion, and Maintenance. Vol. 4003. International Society for Optics and Photonics, pp. 355–362.

Swinbank, John D et al. (2015). “The LOFAR transients pipeline”. In: Astronomy and Computing 11, pp. 25–48.

Tasse, C et al. (2018). “Faceting for direction-dependent spectral deconvolution”. In: Astronomy & Astrophysics 611, A87.

Thompson, A Richard (1999). “Fundamentals of radio interferometry”. In: Synthesis Imaging in Radio Astronomy II. Vol. 180, p. 11.

Thompson, A. Richard, James M. Moran, and George W. Swenson (2017). “Van Cittert– Zernike Theorem, Spatial Coherence, and Scattering”. In: Interferometry and Syn-thesis in Radio Astronomy. Cham: Springer International Publishing, pp. 767–786.

ISBN: 978-3-319-44431-4.

Venturi, T et al. (2001). “Radio monitoring of a sample of X-and-ray loud blazars”. In: Astronomy & Astrophysics 379.3, pp. 755–766.

Weiler, Kurt W et al. (2002). “Radio emission from supernovae and gamma-ray bursters”. In: Annual Review of Astronomy and Astrophysics 40.1, pp. 387–438.

Weiler, Kurt W et al. (2011). “Radio emission from SN 1994I IN NGC 5194 (M 51): The best-studied type Ib/c radio supernova”. In: The Astrophysical Journal 740.2, p. 79.

Zhang, Bing and Peter Meszaros (2004). “Gamma-ray bursts: progress, problems & prospects”. In: International Journal of Modern Physics A 19.15, pp. 2385–2472.

A new radio transient found with the MeerKAT telescope

M

’

T

A new radio transient found with the

MeerKAT telescope

Anton Pannekoek

Institute

Abstract

Contents

Chapter 1

Introduction

1.1

Radio Astronomy

1.2

Interferometry

1.3

MeerKAT

1.4

Transients

Chapter 2

Data Analysis

2.1

MeerKAT observations

2.2

The LOFAR Transients Pipeline

∑

2.3

The MAXI J1820 field

2.4

Comparison to other catalogs

2.5

Optical observations

2.6

X-ray observations

Chapter 3

Discussion

3.1

Extrinsic variability

3.2

Intrinsic AGN variability

3.3

Direction dependent calibration