2.1.1. Compact configuration

The compact configuration of Phase II consists of 56 tiles from the original Phase I core (including a few tiles an intermediate distance from the centre) and adds two hexagonal (‘hex’) cores of 36 tiles each. The size of the hex cores was chosen to be comparable to the original core, and the spacing was chosen to yield relatively smooth uv coverage to $\sim\!~100\, \text{m}$ . The spacing is also approximately that of the hydrogen epoch of reionisation array (HERA; DeBoer et al. Reference DeBoer2017), which will allow cross checks for EoR science. The layout is shown in Figure 1.

Figure 1.

The compact and extended configurations of the Phase II MWA. The blue and orange squares show the tiles which are correlated in the compact and extended configurations, respectively. Note the linear radial scale within 200 m to show the dense pseudo-random/redundant hybrid core, and logarithmic radial scale beyond to capture the nearly 6 km diameter.

The primary motivation for the compact configuration is to increase surface brightness sensitivity relative to the Phase I array. The net effect is a significant increase in sensitivity to diffuse signals like the 21 cm signal from the EoR. Figure 2 demonstrates the thermal noise levels achievable in a power spectrum measurement of the 21 cm signal for Phase I (red), Phase II (blue), and a hypothetical Phase II array with all 256 tiles correlated (black). Phase II achieves lower noise levels by a factor of $\sim\! 2$ –5 (in $\text{mK}^2$ ) compared with Phase I over a wide range of scales. Comparisons with other facilities such as HERA, the Low-Frequency Array (LOFAR), and the Precision Array for Probing the Epoch of Reionisation (PAPER) can be found in DeBoer et al. (Reference DeBoer2017) (their Figure 4) and the appendix of Pober et al. (Reference Pober2014).

Figure 2.

Typical EoR power spectrum model at 150 MHz with associated noise levels available to the Phase I and Phase II arrays with 1000 h observation. ‘Phase II 256’ represents the result from a future MWA upgrade where all 256 tiles are used simultaneously. (From Wayth et al. Reference Wayth2018).

Beyond the increased surface brightness sensitivity, the substantial number of redundant baselines serves a dual-purpose. First, they enable coherent addition of redundant visibilities without requiring gridding or imaging. This redundancy can ultimately serve to increase the sensitivity of a class of non-imaging EoR power spectrum estimators like the delay spectrum approach pioneered by PAPER (Parsons et al. Reference Parsons, Pober, McQuinn, Jacobs and Aguirre2012a,b). For a pedagogical description of imaging and non-imaging EoR power spectrum estimators, see Morales et al. (Reference Morales, Beardsley, Pober, Barry, Hazelton, Jacobs and Sullivan2019). Second, redundancy enables a new class of interferometric calibration techniques that derive antenna-based gains by minimising deviations between visibilities from redundant baselines (Wieringa Reference Wieringa1992; Liu et al. Reference Liu, Tegmark, Morrison, Lutomirski and Zaldarriaga2010; Zheng et al. Reference Zheng2014). The Phase II compact configuration is somewhat unique, however, in retaining a large number of non-redundant baselines to preserve a high-fidelity point spread function (PSF) for imaging and image-based calibration.

2.1.2. Extended configuration

The extended configuration of Phase II consists of 72 tiles from the original Phase I array and 56 new long baseline tiles (Figure 1). The extended layout nearly doubles the longest baselines (from 3 to 5.3 km), resulting in a resolution of $\sim\!1.3\,\text{arcmin}$ at 154 MHz. Franzen et al. (Reference Franzen2016) estimated the Phase I classical confusion limit to be $\sim1.7~\text{mJy}$ at 154 MHz, and the improved resolution is expected to reduce this by a factor of 5–10.

In addition, the naturally weighted PSF of the extended configuration is significantly smoother than that of the Phase I array, owing to the lack of a dense core and instead a more uniform sampling of the uv plane (Wayth et al. Reference Wayth2018). Therefore, we expect lower sidelobe confusion noise, and improved point source sensitivity.

2.2.1. Correlator data rate

The data rate of the archive restricted the correlator during Phase I to a time–frequency resolution set by $$\tau \Delta f \mathbin{\lower.3ex\hbox{$\buildrel>\over {\smash{\scriptstyle\sim}\vphantom{_x}}$}} 2{\mkern 1mu} 000$$ , where $\tau$ is the visibility integration time and $\Delta f$ is the bandwidth of a single frequency channel. However, the longer baselines of the extended configuration of Phase II necessitated finer time–frequency resolution. This was made possible by implementing a lossless in-situ compression on the visibilities (Kitaeff Reference Kitaeff, Taylor and Rosolowsky2015) and an expansion of the onsite archive system. The result is a factor of 4 increase in possible output data rate, for example, typical observations in the extended configuration have frequency and time resolution of 10 kHz and 0.5 s, respectively.

2.2.2. Voltage capture buffer

The MWA voltage capture system (VCS; Tremblay et al. Reference Tremblay2015) facilitates recording of raw antenna voltage data (before correlation and tied-array beamforming) at $${\rm{100\mu s}}$$ and 10 kHz resolution. This capability has been exploited for a variety of pulsar science applications such as investigating pulsar emission physics and studies of millisecond pulsars (e.g. Bhat et al. Reference Bhat, Ord, Tremblay, McSweeney and Tingay2016; McSweeney et al. Reference McSweeney, Bhat, Tremblay, Deshpande and Ord2017; Meyers et al. Reference Meyers2017; Bhat et al. Reference Bhat2018; Meyers et al. Reference Meyers2018). A new feature has enabled buffering voltage data up to 150 s, which can be recorded after receiving a trigger (Subsection 2.3). In this mode, instead of recording directly to disk, the voltages are stored in a ring buffer in the on-board memory of the VCS servers (Hancock et al. Reference Hancock2019). The voltages are kept in the rolling memory buffer for as long as possible (depending on the available memory) on a first-in-first-out basis.

2.2.3. Breakthrough listen

Although replacement of the correlator hardware was not included in the Phase II upgrades, work is currently underway to enable much higher frequency resolution and more flexible beamforming. Additionally, a fibre link from the MRO to Curtin University means that raw voltage data can be accessed offsite, enabling easier deployment of commensal instruments that can produce beams and spectra within the primary FOV, independent of the science user who is controlling the primary beam pointing. The breakthrough listen (BL; Worden et al. Reference Worden2017) team has deployed hardware to Curtin as part of a pilot program to perform experiments to SETI (Section 7.2). The BL hardware will also serve as a general purpose instrument for targets such as fast transients and pulsars—in essence an enhanced version of the existing VCS.

4.1.1. Cosmic ray tomography

Below $\sim\!150$ MHz, most H II regions become optically thick and therefore appear in absorption relative to the background Galactic synchrotron. When the distance to the H II region is known, the difference between its surface brightness and a nearby (source-free) region gives a direct measurement of the integrated cosmic ray electron emissivity between the H II region and the edge of the Galactic plane. Phase I of the MWA was successfully used by Hindson et al. (Reference Hindson2016) to detect and measure 306 H II regions. Su et al. (Reference Su2017); Su et al. (Reference Su2018) went on to use the distinct spectral signature of these regions to perform cosmic ray tomography of the Galactic plane. Foreground measurements, between the H II region and Earth, are more difficult, as an absolute measurement of the emissivity must be calculated, which is impossible for an interferometer which does not measure total power. However, Su et al. (Reference Su2018) spectrally scaled the 408 MHz total power image of Haslam et al. (Reference Haslam, Klein, Salter, Stoffel, Wilson, Cleary, Cooke and Thomasson1981); Haslam et al. (Reference Haslam, Salter, Stoffel and Wilson1982) to fit the foreground emissivity, albeit with large errors. Their results showed an increase in emissivity towards the Galactic Centre and a decrease with galactocentric radius, consistent with other results in the literature.

The limitations of this work are due to two main factors: the low resolution of Phase I, which reduces both the number of detectable H II regions due to confusion noise, and the separability of H II regions in complex areas; and the lack of distance estimates towards H II regions, which are necessary in order to perform the tomography measurement. The improved resolution of Phase II MWA will considerably improve the detectability and separability of H II regions, revealing $\sim\!2\times$ more, which, with distance estimates, could be used for tomography. Upcoming radio recombination line surveys such as the H II Region Discovery Survey (HRDS; Anderson et al. Reference Anderson, Bania, Balser and Rood2011; Bania et al. Reference Bania, Anderson and Balser2012; Anderson et al. Reference Anderson, Armentrout, Johnstone, Bania, Balser, Wenger and Cunningham2015) and its Southern counterpart, SHRDS (Brown et al. Reference Brown2017a), aim to find distances to the hundreds of H II regions catalogued by the WideField Infrared Survey Explorer (Anderson et al. Reference Anderson, Bania, Balser, Cunningham, Wenger, Johnstone and Armentrout2014). The combination of improved resolution and distance estimates also gives the ability to measure more distant (smaller apparent size) H II regions, which yields more 3D sampling of the Galactic plane, considerably improving the leverage of the data over the models of cosmic ray electron distribution (Strong et al. Reference Strong, Orlando and Jaffe2011) and magnetic field distribution (Han Reference Han2017).

4.1.2. Planetary nebulae

More than 300 planetary nebulae (PNe), visible to MWA, have angular sizes larger than 1 arcmin (Parker et al. Reference Parker2006; Filipović et al. Reference Filipović2009; Miszalski et al. Reference Miszalski, Parker, Acker, Birkby, Frew and Kovacevic2008; Parker et al. Reference Parker, Bojičić and Frew2016), which can be resolved by Phase II of the MWA. Nearly all PNe are optically thick and will only show self-absorption features at the MWA frequencies, thus can be used to perform cosmic ray tomography, similarly with that of the H II region absorption. Furthermore, excellent MWA low-frequency coverage, in combination with high-frequency measurements, will be extremely useful for PNe spectral energy distribution (SED) construction and examination of potential radial density gradients and the emission measure to angular diameter relationship (Leverenz et al. Reference Leverenz, Filipović, Vukotić, Urošević and Grieve2017).

4.1.3. Supernova remnants and pulsar wind nebulae

The currently detected population of supernova remnants (SNRs) is considerably lower than the total number of SNRs expected in our Galaxy (Modjaz et al. Reference Modjaz, Gutiérrez and Arcavi2019). In particular, examination of pulsar creation rates, heavy element abundances, OB star counts and stellar life-cycle models, as well as detection rates of SNRs in other galaxies, all suggest that there should be well over 1 000 SNRs detectable in the plane of the Galaxy (Li et al. Reference Li, Wheeler, Bash and Jefferys1991; Tammann et al. Reference Tammann, Loeffler and Schroeder1994; Brogan et al. Reference Brogan, Gelfand, Gaensler, Kassim and Lazio2006). To date, only a fraction of these (<30%) have been detected despite the ever-increasing number of observational surveys of the Galactic plane; over 90% of those detections have been made at radio frequencies.

Older SNRs are expected to have various angular scales and low-surface brightnesses (e.g. G $108.2-0.6$ , a low-surface brightness SNR detected in the Canadian Galactic Plane Survey; Tian et al. Reference Tian, Leahy and Foster2007). It is detecting this latter category of SNRs to which the MWA is well-suited, with its superb diffuse source sensitivity. As expected (Bowman et al. Reference Bowman2013), the Phase I MWA proved itself to be a powerful machine for the detection of new SNRs (Hurley-Walker et al. Reference Hurley-Walker2019a; Maxted et al. Reference Maxted2019; Onić et al. Reference Onić2019), including several that had been previously misclassified as Hii regions (Hindson et al. Reference Hindson2016). These reclassifications were possible, thanks to the low-frequency coverage and the high-spectral resolution of the MWA: the spectra of H II regions turn over and go into absorption at the lowest part of the MWA band, and are thus readily distinguished from SNRs, which display power law spectra.

However, the angular resolution of the Phase I array confined detections to SNRs that had angular sizes greater than $0.1^\circ$ , leaving potential populations of smaller SNRs waiting for discovery with the Phase II array. Combining Phase I and Phase II data will potentially reveal low surface brightness emission below previous confusion limits, both within and outside of the Galactic Plane, as well as having the higher angular resolution of Phase II with the large-angular scale sensitivity of Phase I (see Figure 6 for an example of the combined imaging potential). The complete population of SNRs in the nearby Magellanic Clouds will also be detectible (Bozzetto et al. Reference Bozzetto2017; For et al. Reference For2018).

Figure 6.

A region of sky centred on $\text{RA}\, 17^{\rm h}22^{\rm m}$ , $\text{Dec} -36^\circ$ at 139–170 MHz is shown using three different data sets and imaging techniques. Left: A single 2-min snapshot from the original Phase I configuration, imaged with multiscale WSCLean and a Briggs weighting of minus;1 (Hurley-Walker et al. Reference Hurley-Walker2019 c); Middle: A single 2-min snapshot from the extended array, imaged with multiscale WSClean and a Briggs weighting of 0; Right: The two observations imaged together using image-domain-gridding (van der Tol et al. Reference van der Tol, Veenboer and Offringa2018) in WSClean and a Briggs weighting of 0. Known SNRs are shown with solid lines, while SNRs detected by Hurley-Walker et al. (Reference Hurley-Walker2019b) are shown with dotted lines. The increased resolution and imaging quality of Phase II MWA make it possible to discern these SNRs in 4 min, instead of $\approx\!30$ min.

4.4.1. The synchrotron cosmic web

The large-scale structure of the Universe requires the presence of intergalactic shocks, which are in turn expected to accelerate electrons and amplify intergalactic magnetic fields (Keshet et al. Reference Keshet, Waxman and Loeb2004b; Brüggen et al. Reference Brüggen, Ruszkowski, Simionescu, Hoeft and Dalla Vecchia2005; Hoeft & Brüggen Reference Hoeft and Brüggen2007; Battaglia et al. Reference Battaglia, Pfrommer, Sievers, Bond and Enßlin2009; Araya-Melo et al. Reference Araya-Melo, Aragón-Calvo, Brüggen and Hoeft2012). These shocks should thus produce faint synchrotron emission, which can act as a tracer of large-scale structure, cosmic filaments, and primordial magnetic fields (Keshet et al. Reference Keshet, Waxman and Loeb2004a; Wilcots Reference Wilcots2004; Donnert et al. Reference Donnert, Dolag, Lesch and Müller2009; Vazza et al. Reference Vazza, Ferrari, Bonafede, Brüggen, Gheller, Braun and Brown2015a,b; Brown et al. Reference Brown2017b). Detection of this ‘synchrotron cosmic web’ can provide a direct image of the large-scale structure of the Universe, act as a laboratory for studying particle acceleration in low-density shocks, lead to a measurement of the magnetic field strength of the intergalactic medium, and provide a direct discriminant on competing models for the origin of cosmic magnetism. It is predicted that the signal from the synchrotron cosmic web should dominate other radio signals on scales of $\sim10'$ to $\sim1^\circ$ at frequencies around 100 MHz (Keshet et al. Reference Keshet, Waxman and Loeb2004b; Vazza et al. Reference Vazza, Ferrari, Bonafede, Brüggen, Gheller, Braun and Brown2015a), making the MWA a well-suited facility to search for these structures.

Vernstrom et al. (Reference Vernstrom, Gaensler, Brown, Lenc and Norris2017) have carried out a search for the synchrotron cosmic web with the Phase I MWA, in which they cross-correlated diffuse radio emission imaged at 180 MHz with large-scale structure traced by infrared galaxy surveys. They were able to place upper limits on the surface brightness of the synchrotron cosmic web of $0.01\,-\,0.3\text{mJy\,arcmin}^{-2}$ , which translates to upper limits on the magnetic field strength of 0.03 – 1.98 $${\rm{\mu }}$$ G, assuming equipartition. While these constraints are not yet deep enough to differentiate between different cosmic magnetism models, the limits are comparable to other limits from cluster observations (e.g. Feretti et al. Reference Feretti, Perley, Giovannini and Andernach1999; Brunetti et al. Reference Brunetti, Setti, Feretti and Giovannini2001; Brown et al. Reference Brown, Farnsworth and Rudnick2010) or predictions from MHD simulations (e.g. Donnert et al. Reference Donnert, Dolag, Lesch and Müller2009; Vazza et al. Reference Vazza, Ferrari, Brüggen, Bonafede, Gheller and Wang2015b).

The depth of the search reported by Vernstrom et al. (Reference Vernstrom, Gaensler, Brown, Lenc and Norris2017) was limited by confusion, in that large numbers of unresolved extragalactic radio sources have not been subtracted from the data, and may mimic diffuse radio emission that traces large-scale structure. As for many other continuum science programs, the improved angular resolution and reduced confusion levels offered by MWA Phase II will enable much deeper searches for the synchrotron cosmic web, whether by direct imaging (e.g. Kronberg et al. Reference Kronberg, Kothes, Salter and Perillat2007), statistical cross-correlations (Vernstrom et al. Reference Vernstrom, Gaensler, Brown, Lenc and Norris2017; Brown et al. Reference Brown2017b), or also in polarimetry (Rudnick & Brown Reference Rudnick and Brown2009; Brown et al. Reference Brown, Rudnick, Farnsworth, Strassmeier, Kosovichev and Beckman2009).

4.4.2. Diffuse emission in galaxy clusters

In addition to the potential detection of the large-scale, diffuse emission predicted by the cosmic web, the Phase I MWA has proven itself to be a powerful instrument for the detection and study of diffuse, low-surface brightness emission in galaxy clusters.

Diffuse emission in clusters manifests in a variety of forms including central radio halos, which are believed to be the result of turbulence in the cluster core; mini-halos powered by central AGN; peripheral radio relics which result from strong shocks generated in cluster mergers; and radio phoenices which trace the passage of these shocks across the lobes of AGN (see Kempner et al. (Reference Kempner, Blanton, Clarke, Enßlin, Johnston-Hollitt, Rudnick, Reiprich, Kempner and Soker2004) for the taxonomy of cluster sources and Brunetti & Jones (Reference Brunetti and Jones2014) for a review of the physics of emission in galaxy clusters). Recently, Govoni et al. (Reference Govoni2019) observed diffuse emission at 140 MHz connecting the clusters Abell 0399 and Abell 0401 with LOFAR.

Although on different physical scales from hundreds of kpc (mini-halos and phoenices) to Mpc-scale (halos and relics), one defining characteristic of these sources is their steep radio spectral indices. In the case of halos, the average spectral index is $-1.3$ , with some examples with spectral indices steeper than $-2$ being detected in MWA data (Duchesne et al. Reference Duchesne, Johnston-Hollitt, Offringa, Pratt, Zheng and Dehghan2017). Such emission characteristics make the detection of diffuse cluster emission considerably easier at low frequencies and the large physical scales make detection easier with instruments sensitive to angular scales of the order of arcminutes. The MWA is thus ideal, and a large number of new and existing sources of diffuse emission in clusters and groups have already been detected and studied with the Phase I array (Hindson et al. Reference Hindson2014; Schellenberger et al. Reference Schellenberger, Vrtilek, David, O’Sullivan, Giacintucci, Johnston-Hollitt, Duchesne and Raychaudhury2017; Duchesne et al. Reference Duchesne, Johnston-Hollitt, Offringa, Pratt, Zheng and Dehghan2017).

A particular strength of the MWA is the ability to perform detailed spectral studies of these sources across the MWA band. Figure 7 shows the in-band spectral measurements and resultant fit for the source complex extended cluster source NGC 741 taken from the MWA’s GLEAM survey (Wayth et al. Reference Wayth2015). This plot demonstrates the extremely fine spectral resolution of the MWA as compared to existing arrays (for a full analysis of these data, see Schellenberger et al. Reference Schellenberger, Vrtilek, David, O’Sullivan, Giacintucci, Johnston-Hollitt, Duchesne and Raychaudhury2017). Having such high-fidelity spectral information has been invaluable in constraining the physical processes at work to generate the emission, allowing separation of sources that would otherwise be morphologically indistinct into classes based on the underlying acceleration mechanisms and age of the electron population (Schellenberger et al. Reference Schellenberger, Vrtilek, David, O’Sullivan, Giacintucci, Johnston-Hollitt, Duchesne and Raychaudhury2017).

Figure 7.

MWA data for the complex cluster source NGC 741, showing the unprecedented ability of the MWA to determine spectral indices across the MWA band. Further information and analysis of these data are available in Schellenberger et al. (Reference Schellenberger, Vrtilek, David, O’Sullivan, Giacintucci, Johnston-Hollitt, Duchesne and Raychaudhury2017).

4.5.1. AGN evolution

The long baselines have allowed us to conduct new broad-frequency continuum surveys: the eXtended GLEAM survey (GLEAM-X) and the MWA Interestingly Deep Astrophysical (MIDAS) survey. GLEAM-X is a repeat of the all-sky GLEAM, but with longer total integration in addition to the higher resolution (reaching more than $\sim\! 5\times$ the depth of GLEAM). MIDAS is going at least twice as deep as GLEAM-X, targeting six well-studied extragalactic fields: the five Galaxy And Mass Assembly (GAMA) survey fields (Driver et al. Reference Driver2009) and the Spitzer South Pole Telescope Deep Field (Ashby et al. Reference Ashby2014). The GAMA survey fields cover $250 \text{deg}^{2}$ with exquisite photometry from UV to the far-IR. These fields also include spectroscopic redshifts, group catalogues, and derived data products such as stellar mass and star formation rate. The fusion of these data sets will allow a plethora of science including determining the luminosity function, and the nature and evolution of the low-frequency selected radio population.

In concert with deep ongoing surveys from ASKAP (Norris et al. Reference Norris2011), uGMRT (Swarup et al. Reference Swarup, Ananthakrishnan, Kapahi, Rao, Subrahmanya and Kulkarni1991), and the Australia Telescope Compact Array (ATCA; GAMA Legacy ATCA Southern Survey minus;GLASS) MWA observations will yield broadband SEDs for tens of thousands of sources in the GAMA fields. A large fraction of these will have spectroscopic redshifts and derived host galaxies properties. This unique data set will provide estimates of jet powers for these radio-loud AGN which can be compared to other processes such as accretion rate and star formation (e.g. Gürkan et al. Reference Gürkan2018).

The improvement in sensitivity is also beneficial to the study of Giant Radio Galaxies (GRGs), which are AGN with emission that extends over $\sim$ 700 kpc. Fine details of this extended emission can now be resolved with the addition of the long baselines, allowing for studies of spectral properties such as the spectral ageing of the lobes, enabling detailed modelling of source environments. As an example, Figure 8 compares images of the giant radio galaxy ESO 422 minus;G028 using Phase I and Phase II data. The Phase II data were imaged using robust $=+1.0$ to improve sensitivity to ultra-diffuse large-scale emission from this GRG. The improvement in sensitivity and resolution offered by Phase II of the MWA is illustrated by the ultra-faint diffuse emission from the tails of ESO 422 recovered at much greater significance compared to Phase I. With the factor $\sim\!2-3$ improvement in resolution offered by Phase II, such data can be used in conjunction with higher-frequency data from other cutting edge radio telescopes (e.g. uGMRT, ASKAP, and the ATCA) to probe the spectral properties of extended radio galaxies at much improved angular resolution compared to Phase I.

Figure 8.

Images of ESO 422 minus;G028 from Phase I (left, from GLEAM; Hurley-Walker et al. (Reference Hurley-Walker2017)) and Phase II (right, from GLEAM-X). Colour-scale and contours both denote the surface brightness at 200 MHz. Colour scale ranges from $-3$ to $100\sigma$ on an arcsinh stretch to emphasise faint diffuse emission. Contours start at $3\sigma$ and scale by a factor $\sqrt{2}$ , where $\sigma=14.9 \text{mJy beam}^{-1}$ ( $2.0 \text{mJy beam}^{-1}$ ) in the Phase I (Phase II) image. The beam size is shown by the hatched ellipse in the lower-left corner.

4.5.2. Powerful AGN

The GLEAM 4-Jy Sample (White et al. Reference White2018), selected at 151 MHz, contains 1 863 bright radio sources and is dominated by AGN. This includes 77 radio galaxies that are so extended that they are resolved into multiple GLEAM components by the MWA Phase I beam. However, approximately one-fifth of the sample suffers from confusion which will be alleviated by higher resolution allowing them to be more easily classified and cross-matched. The 4-Jy sample will be a benchmark for the bright radio galaxy population, like 3CRR (Laing et al. Reference Laing, Riley and Longair2003), but is an order of magnitude larger.

GLEAM has also allowed us to search for powerful very high-redshift sources through a number of techniques including ultra-steep spectrum, curved SED, and compactness (from IPS, see Section 6.2.1). This ongoing work to obtain their redshifts is proving fruitful with follow-up on 8 m class telescopes and with the Atacama Large Millimetre Array (e.g. Seymour Reference Seymour2019).

4.6.1. Compact steep spectrum and gigahertz-peaked spectrum radio sources

Compact steep spectrum (CSS) and gigahertz-peaked spectrum (GPS) radio sources are a class of compact radio-loud AGN that are thought to be the young precursors to large-scale radio galaxies (O’Dea Reference O’Dea1998). The lower radio luminosity ( $< 10^{25}\text{ W/Hz}$ at 5 GHz) sample of GPS and CSS sources has been suggested to be dominated by the objects strongly affected during their evolution by an interaction with the ISM or instability in the accretion disk (Czerny et al. Reference Czerny, Siemiginowska, Janiuk, Nikiel-Wroczyński and Stawarz2009; Bicknell et al. Reference Bicknell, Mukherjee, Wagner, Sutherland and Nesvadba2018). Such sources could be the short-lived precursors needed to account for the overabundance of GPS and CSS sources relative to the large-scale radio galaxies (Kunert-Bajraszewska et al. Reference Kunert-Bajraszewska, Gawroński, Labiano and Siemiginowska2010). However, none of the low-luminosity peaked-spectrum samples to date have been large enough, or devoid of selection biases, to justify this conclusion. The increased sensitivity of MWA Phase II will be able to probe a significantly fainter population of peaked-spectrum sources than previously possible (Callingham et al. Reference Callingham2017). While other telescopes such as LOFAR have the required sensitivity to access the low-luminosity peaked-spectrum population, LOFAR currently lacks the wide fractional bandwidth needed to identify peaked-spectrum candidates nontrivially when solely using LOFAR data. Therefore, the large fractional bandwidth of the MWA, combined with the high sensitivity of Phase II, represents a significant advantage for this science.

4.6.2. Dying radio galaxies

As with the diffuse emission in galaxy clusters, the MWA has a niche capability in the detection of diffuse, low surface brightness emission from the lobes of old, dead, or dying radio galaxies. One such early example was the detection of previously unknown giant radio lobes associated with the lenticular galaxy NGC 1534 (Hurley-Walker et al. Reference Hurley-Walker2015; Duchesne & Johnston-Hollitt Reference Duchesne and Johnston-Hollitt2018). Work in this field across a range of low-frequency telescopes has seen the detection of several more such sources, suggesting a significant and previously unknown population of sources (e.g. de Gasperin et al. Reference de Gasperin, Intema, Williams, Brüggen, Murgia, Beck and Bonafede2014; Brienza et al. Reference Brienza2016; Mahatma et al. Reference Mahatma2018). However, as with the diffuse emission in galaxy clusters, the volume over which remnant or dying radio galaxies can be detected was limited with the Phase I array to the very local Universe. If we assume a typical size for such sources of 700 kpc, in order to resolve the lobes of such galaxies, the Phase I MWA would have been limited to sources with redshifts less than 0.17. Even detecting such sources in the nearby Universe, the Phase I MWA had insufficient resolution to perform spectral imaging across the lobes of the radio sources, which is vital to understanding the physics of such sources. As a result while the Phase I MWA was able to detect some of this class of object, no detailed imaging could be undertaken. The detection limits with the Phase II array can be pushed out to redshifts of z $\sim$ 0.42, which considerably expands the volume in which such sources can be found. Furthermore, there will be sufficient resolution to perform detailed statistical spectral studies over the radio lobes of nearby sources, which is vital to understand the physics of such systems.

5.2.1. Gravitational waves

In 2016, the LIGO/Virgo Consortium (LVC) reported the first two detections of GWs (Abbott et al. Reference Abbott2016a,b). The LVC sent private alerts to the EM follow-up community (Abbott et al. Reference Abbott2016 d; Abbott et al. Reference Abbott2016 c) to identify coincident EM transients. While identifying an EM counterpart would greatly enhance the utility of the GW signal (e.g. Phinney Reference Phinney2009; Metzger & Berger Reference Metzger and Berger2012; Singer et al. Reference Singer2014; Chu et al. Reference Chu, Howell, Rowlinson, Gao, Zhang, Tingay, Boër and Wen2016; Branchina & De Domenico Reference Branchina and De Domenico2016), it is not a simple task owing to very large uncertainty regions for the GW events, which can be hundreds of degrees (e.g. Kasliwal & Nissanke Reference Kasliwal and Nissanke2014; Singer et al. Reference Singer2014).

As mentioned in Section 5.1, this hard work bore amazing fruit in 2017, when LVC detected a neutron star merger GW 170817 (Abbott et al. Reference Abbott2017a) coincident with a short (underluminous) gamma-ray burst, GRB 170817A (Abbott et al. Reference Abbott2017 c) which was subsequently detected across the EM spectrum (where MWA participated Abbott et al. Reference Abbott2017b). While this one event has generated a wealth of information, there are still fundamental questions as to the nature of the multi-wavelength emission, the energetics of the explosion, the properties of the environment, and the degree of beaming (e.g. Kasliwal et al. Reference Kasliwal2017; Mooley et al. Reference Mooley2018b; Mooley et al. Reference Mooley2018a; Gottlieb et al. Reference Gottlieb, Nakar, Piran and Hotokezaka2018; Troja et al. Reference Troja2018; Lazzati et al. Reference Lazzati, Perna, Morsony, Lopez-Camara, Cantiello, Ciolfi, Giacomazzo and Workman2018; Alexander et al. Reference Alexander2018; Dobie et al. Reference Dobie2018). Many of these can only be answered by looking at a much larger sample of objects, and it must be determined what strategies and types of signals will give the best return in the radio domain.

A number of models for neutron star-neutron star mergers predict a prompt flash of coherent radio emission that should accompany (slightly before or after) the GW signal and be detectable by the MWA (see Section 5.1 for specific examples). In fact, low-frequency telescopes have a number of advantages over optical/infrared searches: they have fields-of-view of hundreds to thousands of square degrees; the radio sky is relatively quiet at these frequencies (Karastergiou et al. Reference Karastergiou2015; Tingay et al. Reference Tingay2015; Stewart et al. Reference Stewart2016; Rowlinson et al. Reference Rowlinson2016; Polisensky et al. Reference Polisensky2016) with very few transients unassociated with the GW event (e.g. Hotokezaka et al. Reference Hotokezaka, Nissanke, Hallinan, Lazio, Nakar and Piran2016); and owing to the new rapid-response mode, the MWA can respond within seconds to an external trigger. A detection of a prompt coherent flash would give immediate localisation of the GW signal with a very low false-positive rate, and would help probe the distance to the source through constraints on the line-of-sight electron density if dispersion can be measured. Unfortunately, GW 170817 was not visible to the MWA when the trigger was announced, and observations from other sites were not sufficiently sensitive (Callister et al. Reference Callister2017, using the LWA was a factor of $\sim\!100$ less sensitive than typical MWA observations discussed below).

Along with the VOEvent triggering mechanism discussed in Section 2.3, an optimal strategy for MWA follow-up of prompt emission from GW transients has been implemented (Kaplan et al. Reference Kaplan, Murphy, Rowlinson, Croft, Wayth and Trott2016). This is complementary to approaches taken with other facilities (e.g. Yancey et al. Reference Yancey2015) where the wide FoV, rapid response, and Southern location of the MWA give it an advantage (Chu et al. Reference Chu, Howell, Rowlinson, Gao, Zhang, Tingay, Boër and Wen2016; Howell et al. Reference Howell2015). Kaplan et al. (Reference Kaplan, Murphy, Rowlinson, Croft, Wayth and Trott2016) used simulated GW events from Singer et al. (Reference Singer2014) to compute the expected fraction of events that the MWA would observe and the sensitivity to them, given the optimised pointing strategy. For optimum conditions, the limiting (10 $\sigma$ ) flux density is $\sim 0.1$ Jy. However, given the influence of Galactic synchrotron emission and the limited collecting area away from zenith, only 5% of simulated events are close to that limit; a more typical sensitivity is 1 Jy. For typical distances ( $\sim 200\,$ Mpc), it corresponds to luminosity limits of $10^{38-39}\,{\rm erg/s}$ . These luminosities are squarely in the range of various predictions for prompt radio emission (Usov & Katz Reference Usov and Katz2000; Pshirkov & Postnov Reference Pshirkov and Postnov2010; Totani Reference Totani2013; Zhang Reference Zhang2014; Falcke & Rezzolla Reference Falcke and Rezzolla2014) and should seriously constrain the underlying physics (Kaplan et al. Reference Kaplan2015 c) even with a non-detection. Note that the MWA’s rapid (6–14 s) repointing is crucial to capturing prompt emission, because dispersive delays of only tens of seconds are expected at the MWA frequencies. During the LIGO/Virgo O3 run (started 2019 April) there is an expectation of low-latency alerts with reduced information, which will be ideal for triggering MWA follow-up. Note that if prompt gamma-ray emission is detected again the system will trigger on the GRB itself, since that notice is sent within seconds of the event.

Recently, James et al. (Reference James, Anderson, Wen, Bosveld, Chu, Kovalam, Slaven-Blair and Williams2019) proposed triggering the MWA on negative latency GW alerts (generated from the GWs emitted by the inspiral of the binary components prior to merger). These alerts have relatively poor positional constraints, and so this observing mode would disable 15 of the 16 dipoles on each tile, resulting in a FoV covering about a quarter of the sky. This mode is expected to improve the MWA’s response time by several seconds, significantly increasing the potential to detect FRB-like prompt emission from mergers.

5.2.2. Neutrinos

Multi-messenger astrophysics at MWA also involves searching for counterparts to astrophysical neutrino sources. In addition to archival searches for serendipitous observations of neutrino triggers (Section 7.3), MWA is undertaking pointed observations to follow up triggers from neutrino observatories including IceCube (IceCube Collaboration et al. 2006) and ANTARES (Ageron et al. Reference Ageron2011).

Recent multi-wavelength follow-up observations of an IceCube neutrino alert identified the blazar TXS 0506+056 as the first detected high-energy ( $\textit{E}\,>\,10^{12}\text{eV}$ ) astrophysical source (IceCube Collaboration et al. 2018). However, the mechanism behind the neutrino emission is unknown, with the most optimistic models predicting only 1% of the observed signal (Keivani et al. Reference Keivani2018). Furthermore, no more than 27% of IceCube’s high-energy astrophysical flux can be explained by blazar emission (Aartsen et al. Reference Aartsen2017).

EM follow-up of neutrino events to identify the remaining neutrino sources is therefore a priority. Candidate astrophysical sources of the remaining flux of high-energy neutrinos include GRBs, core-collapse SNe, microquasars, or AGN. The recent observation by the ANITA Antarctic balloon experiment of two $\sim$ 1 EeV neutrino events in apparent contradiction of standard model physics also serves as a reminder that neutrino observations could point the way to as-yet unknown phenomena (Gorham et al. Reference Gorham2016).

Similar to GWs, the wide FoV, increased resolution and sensitivity, and new rapid-response mode of the MWA are well suited to react to neutrino alerts. IceCube and ANTARES are expected to generate a total of 6 – 8 alerts per annum with a position visible to the MWA. Triggering on these events and performing the subsequent observations will allow for the strongest limits to date on prompt radio emission from neutrino transients and may aid in localisation of these new astrophysical probes.

5.2.3. Cosmic rays

The origin of cosmic rays is unknown. In the region of the energy spectrum between the ‘knee’ at $10^{15}$ – $10^{16}$ eV and the ‘ankle’ at $10^{18}$ – $10^{19}$ eV, cosmic rays are thought to transition from a predominantly Galactic to an extragalactic origin. In this region, experiments aim to measure the spectrum of each primary particle species (p, He, CNO, Fe, etc.) (Apel et al. Reference Apel2013), with spectral features identifying Galactic accelerators via the magnetic rigidity of the primaries (Peters Reference Peters1961). Using the low-band antennas from its inner stations, LOFAR has established that a dense array of low-frequency radio antennas can accurately reconstruct key cosmic ray properties.

When a cosmic ray impacts the upper atmosphere, it produces an extensive air shower of secondary particles, which in turn emit a sub-microsecond burst of radio waves (Huege Reference Huege2016). The frequency structure and ground pattern of these bursts depend on the properties of the particle cascade and, hence, on the primary cosmic ray itself.

By studying the ground pattern of cosmic ray cascades (Schellart et al. Reference Schellart2013), radio data from LOFAR have allowed the height of shower maximum, $X_{\rm max}$ , to be reconstructed with unprecedented accuracy (Buitink et al. Reference Buitink2014). The importance of $X_{\rm max}$ is that it can be statistically related to the cosmic ray composition.

LOFAR’s detection threshold of approximately $10^{16}$ eV lies in the Galactic–extragalactic transition region, and its measurements have indicated a new light-mass component to the flux in the $10^{17}$ – $10^{17.5}$ eV range (Buitink et al. Reference Buitink2016).

The MWA’s bandwidth of $30.72$ MHz and flatter bandpass will provide a comparable sensitivity to the 30–80 MHz range used by LOFAR’s Cosmic Ray Key Science Project. The rate of cosmic ray detections is also expected to be comparable to that of LOFAR due to the concentrated sensitivity of the compact configuration of Phase II. The core region, including hexes, spans an area similar to that of the LOFAR superterp. Beamforming $N_{\rm dip}$ (here, 16) dipoles per tile increases signal-to-noise by $N_{\rm dip}^{0.5}$ and reduces the cosmic ray energy detection threshold $E_{\rm thresh}$ by $N_{\rm dip}^{-0.5}$ . Since the cosmic ray rate $R_{\rm CR}$ falls with energy $E_{\rm CR}$ approximately as $\frac{dR_{\rm CR}}{dE_{\rm CR}} \sim E_{\rm CR}^{-3}$ in this range, the integrated rate varies as $E_{\rm thresh}^{-2}$ , that is, with $N_{\rm dip}$ . This exactly compensates the $N_{\rm dip}$ -fold loss of solid angle from beamforming, so that the rate is constant. This will therefore allow the MWA to probe lower-energy regions of the cosmic ray spectrum with similar statistics to LOFAR, and test predictions of a rigidity-dependent spectrum.

Identification of cosmic ray signals at the MWA site could be performed using either onsite particle detectors, or triggered on radio-only data. Particle detectors are being developed for SKA1-Low and tested with the MWA (Figure 9). An array of eight detectors is planned to trigger MWA radio observations of high-energy cosmic ray events. While radio-only triggering has proven infeasible for most modern experiments, it has recently been tested successfully at the Owens Valley Radio Observatory Long Wavelength Array (Monroe Reference Monroe2018). The feasibility of both techniques is currently under investigation, in particular the resynthesis of time-domain data from 24 coarse channels provided by the VCS.

Figure 9.

The prototype SKAPA cosmic ray detector (white box) deployed near the MWA core at the MRO. It is raised off the ground with palettes to avoid surface water. Power and data return are enclosed via the black cable housing. The detector has been tested to ensure compliance with MRO RFI requirements and has been detecting cosmic ray muons since its deployment in October 2018. Photo credit: Justin Bray.

While cosmic ray emission below 80 MHz is well-understood, the frequency range above 80 MHz is relatively unexplored, with the only observations being from experiments measuring the radiation pattern at essentially a single point (e.g. the Antarctic Impulse Transient Antenna, ANITA Hoover et al. Reference Hoover2010; Cosmic-Ray Observation via Microwave Emission, CROME Šmída et al. 2014). Simulations predict that 110–190 MHz is the most sensitive frequency range at which to observe cosmic rays (Balagopal et al. Reference Balagopal, Haungs, Huege and Schröder2018). This opens the possibility of observing radio emission from the flux of PeV ( $10^{15}$ eV) gamma-rays from the Galactic Centre source detected by the High Energy Stereoscopic System, H.E.S.S. (HESS Collaboration et al. Reference Collaboration2016). Furthermore, Corsika-based Radio Emission from Air Showers (CoREAS) simulations suggest that the ground pattern in Stokes V (circular polarisation) changes shape at higher frequencies (Huege et al. Reference Huege, Ludwig, James, Lahmann, Eberl, Graf, James, Huege, Karg and Nahnhauer2013). This arises from interference along the cascade axis and points to a method of resolving air shower structure—and hence composition—in more detail than can be provided by $X_{\rm max}$ alone.

5.6.1. All-sky pulsar surveys

In the compact configuration, with the two hexes and the core providing a collecting area equivalent to that of the full Phase I array, the processing cost for beamforming across the FoV is considerably reduced. Specifically, the number of tied-array beams required to fill the FoV (at a gain level down to half power point) is reduced from $2.7 \times 10^5$ (Phase I) to $3.9 \times 10^3$ (Phase II compact). This corresponds to a reduction of more than two orders of magnitude in the computational cost, thereby making large-scale, high-sensitivity pulsar searches more tractable (and affordable) with the MWA.

Large pulsar surveys have established histories of delivering high-impact science returns in the long run. A noteworthy example is the Parkes multibeam survey (Manchester et al. Reference Manchester2001) that discovered the double pulsar system J0737-3039A/B and the eccentric neutron star-white dwarf binary J1141-6545, both of which are proven laboratories for testing general relativity and alternate theories of gravity (Kramer et al. Reference Kramer2006b; Bhat et al. Reference Bhat, Bailes and Verbiest2008). Its successor survey (Keith et al. Reference Keith2010) found $>$ 200 pulsars, besides the very first population of FRBs (Thornton et al. Reference Thornton2013). As such, conducting a full Galactic census of pulsars is a key science driver for the SKA and its pathfinders (Keane et al. Reference Keane2015). While the $\sim\!1 - 2$ GHz band proved efficient for probing deeper into the Galactic plane, low frequencies offer the advantages of higher survey speeds and benefit from the increased brightness of pulsars.

The above developments have led to the conception of the Southern-sky MW A Rapid Two-metre (SMART) survey—an all-sky survey for pulsars with the Phase II compact MWA, where the goal is to systematically survey the entire sky visible to the MWA (i.e. declination south of $+30^{\circ}$ ) at a frequency band of $140 - 170$ MHz. A dwell time of $\sim\!1.5$ h per pointing (maximum duration of uninterrupted recording feasible with the VCS) translates to a 10- $\sigma$ detection limit of $S_{150} \sim 3$ mJy (the mean flux density at a frequency of 150 MHz), assuming a 30.72 MHz bandwidth. This is more than a factor of 2 improvement ^{Footnote d} over the Parkes southern pulsar survey from 1990s (at a frequency of 430 MHz) which found 101 pulsars including 17 millisecond pulsars (MSPs) (Lyne et al. Reference Lyne1998; Manchester et al. Reference Manchester1996). Searching at low frequencies will necessitate a large number of trial DMs and will be limited by a ‘pulse broadening horizon’, that is, the distance beyond which temporal smearing from multipath propagation ( $\tau _d$ ) limits (or degrade) the pulsar detectability. At 150 MHz, this is $\sim\! 5$ kpc ( $\tau _d \,\sim\!100$ ms) for directions away from the Galactic disk (cf. Bhat et al. Reference Bhat, Cordes, Camilo, Nice and Lorimer2004). Pulsars with periods $$\mathop > \limits_ \sim 100$$ ms, DMs $\mathop < \limits_ \sim 300$ ${\rm pc \ cm ^ {-3}}$ , and $S_{150} \mathop > \limits_ \sim 3$ mJy will be detectable with the MWA survey.

Accurate modelling and forecasting of the expected yield from such a survey is however difficult, due to a number of unknowns, such as a possible spectral turnover at low frequencies, and temporal broadening from multipath scattering. It is further complicated due to the loss of sensitivity from projection (at large zenith angles) and the resulting non-uniformity in sensitivity (e.g. when observations are made in the drift-scan modes). Even then, a conservative forecast based on simulation studies performed using the PSRPOPPy software (Bates et al. Reference Bates, Lorimer, Rane and Swiggum2014), after calibrating with recent MWA pulsar census work (Xue et al. Reference Xue2017), predicts $\sim$ 230 new pulsar discoveries (see Figure 10) including $\sim$ 20–40 MSPs. Aside from this scientific promise, this will produce a digital archive of the full MWA-visible sky and will serve as an important demonstrator survey for the low-band SKA.

Figure 10.

Simulated pulsar detections (red dots) from an all-sky high time resolution survey with the MWA. The grey region represents the MWA’s visible sky, and the dark-grey region the sky that is exclusively accessible by the MWA at frequencies below 300 MHz (declination $<-55^{\circ}$ ). The blue dots are pulsars from the ATNF pulsar catalogue v1.59.

5.6.2. High-time resolution studies of pulsar emission

Pulsar emission phenomena occur on timescales spanning several orders of magnitude, with temporal structures lasting mere nanoseconds (e.g. Hankins et al. Reference Hankins, Kern, Weatherall and Eilek2003) to microseconds (e.g. Hankins Reference Hankins1971; Cordes Reference Cordes, Sieber and Wielebinski1981; De et al. Reference De, Gupta and Sharma2016) and even longer (e.g. Deshpande & Rankin Reference Deshpande and Rankin2001; McSweeney et al. Reference McSweeney, Bhat, Tremblay, Deshpande and Ord2017). In general, smaller temporal structures map directly to smaller physical structures in the pulsar magnetospheres. High-time resolution studies of individual pulses from pulsars that exhibit both regular pulse-to-pulse variations such as sub-pulse drifting and microstructure (i.e. the finer components that make up individual subpulses) can thus provide important insights into the relationship between physical magnetospheric structures of different length scales. Such studies (which are still in their infancy) are now ripe to be explored with the new capabilities of the MWA-VCS’s pulsar processing pipeline, where the beamformer pipeline (Ord et al. Reference Ord, Tremblay, McSweeney, Bhat, Sobey, Mitchell, Hancock and Kirsten2019) has been enhanced with an implementation that allows re-synthesising high-time resolution voltage time series by undoing the polyphase filter bank operation prior to the correlation stage. This allows attaining a time resolution down to $\sim$ 1 $${\rm{\mu }}$$ s and is highly promising for studying short timescale pulsar emission features at the low frequencies of the MWA (Figure 11) where pulsars generally tend to be brighter.

Figure 11.

A sequence of 210 individual pulses from PSR J0034 minus;0721 that shows the phenomenon of sub-pulse drifting. The observations were made with Phase II over the 140–170 MHz band and processed using the newly enhanced tied-array beam-former pipeline that resynthesises the VCS-recorded data to produce high-time resolution ( $\sim$ 1 $${\rm{\mu }}$$ s) time series. The pulsar switches between three distinct drift bands (designated as Modes A, B, and C) and exhibits the phenomenon of nulling (i.e. cessation of pulses) and is a promising target for studying emission mechanisms.

This new implementation has been successfully tested using observations of MSPs (Kaur et al. 2019). The ability to perform coherent de-dispersion on high-time resolution voltage time series allows the complete removal of the deleterious effects of temporal smearing caused by interstellar dispersion (Hankins & Rickett Reference Hankins, Rickett, Alder, Fernbach and Rotenberg1975). Together with the large frequency lever arm provided by the MWA, this allows the measurement of pulsar DMs with unprecedented levels of precision. The high quality pulse profiles that capture fine temporal structures by virtue of being able to perform phase coherent de-dispersion also significantly improve the precision with which the pulse arrival times can be measured. These new high-time resolution capabilities have wide-ranging applications from studies of millisecond pulsars to the exploration of a new parameter space in the quest to understand the pulsar emission mechanism.

5.6.3. Sporadic and intermittent emission

While the emission from most pulsars is seen as regular sequences of pulses (e.g. McSweeney et al. Reference McSweeney, Bhat, Tremblay, Deshpande and Ord2017), some pulsars exhibit irregular emission, in the form of sporadic emission (e.g. giant pulses), or switching emission states (e.g. Kramer et al. Reference Kramer, Lyne, O’Brien, Jordan and Lorimer2006a; Lorimer et al. Reference Lorimer, Lyne, McLaughlin, Kramer, Pavlov and Chang2012; Young et al., Reference Young, Weltevrede, Stappers, Lyne and Kramer2014; Meyers et al. Reference Meyers2017; Meyers et al. Reference Meyers2018; Meyers et al. Reference Meyers2019). Such emission occurs on timescales of seconds to months, and the pulsars which exhibit such types of emission pose a major challenge to our understanding of the underlying physics that govern the radio emission mechanism. While there are several models that attempt to describe such intermittency (e.g. Cheng Reference Cheng1985; Cordes & Shannon Reference Cordes and Shannon2008; Jones Reference Jones2012; Li et al. Reference Li, Spitkovsky and Tchekhovskoy2012; Seymour & Lorimer Reference Seymour and Lorimer2013; Melrose & Yuen Reference Melrose and Yuen2014), none satisfy the enormous diversity of emission phenomenology observed. Furthermore, the sporadic and intermittent emission behaviour appears to be extremely broadband, based on simultaneous high-energy and wideband radio observations (e.g. Abdo et al. Reference Abdo2010; Hermsen et al. Reference Hermsen2013; Kaplan et al. Reference Kaplan2015b; Meyers et al. Reference Meyers2017; Hermsen et al. Reference Hermsen2018). In general, low-frequency coverage is lacking for the vast majority of southern hemisphere pulsars, and in particular for pulsars with sporadic or intermittent emission. Observations at low frequencies with the MWA are thus potentially promising to reveal emission characteristics that are substantially different to those observed at higher frequencies.

However, these sporadically emitting pulsars are inherently difficult to observe without regular, unusually long dwell times, which is generally difficult, and in particular not currently feasible using the VCS-like functionality. The newly developed voltage buffer mode can be suitably exploited to address this shortcoming. This mode allows the MWA to circumvent the inherent limitation in studying intermittently emitting pulsars to some extent, for example, through suitable coordination with the observing programs that regularly observe such intermittent targets, such as UTMOST (Bailes et al. Reference Bailes2017) and the Parkes radio telescope (e.g. Kerr et al. Reference Kerr, Hobbs, Shannon, Kiczynski, Hollow and Johnston2014). The recent detection of low-frequency emission from the intermittent pulsar J1107 minus;5907 (Meyers et al. Reference Meyers2018) provides an excellent demonstration of the MWA’s unique advantage for conducting such coordinated broadband observations. As it is, broadband simultaneous observations involving multiple telescopes ostensibly provide valuable insights into the pulsar emission mechanism (e.g. Bhat et al. Reference Bhat, Gupta, Kramer, Karastergiou, Lyne and Johnston2007; Hermsen et al. Reference Hermsen2013; McSweeney et al. Reference McSweeney, Bhat, Wright, Tremblay and Kudale2019). With the prospects of the VCS buffer mode and BL back-end becoming available for routine observations in the coming years, the Phase II MWA will be promising for studying these sub-populations of pulsars.

6.1.1. Calibration and imaging challenges

At MWA frequencies, even the quiet Sun is orders of magnitude brighter than any Southern-hemisphere compact source, with radio bursts reaching flux densities many orders of magnitude brighter still. This means that there is sufficient signal to noise to image the Sun easily, even when using a very short snapshot with extremely narrow bandwidth. However, it also means that astrophysical sources are not typically visible when viewing the Sun unless extremely high dynamic range can be attained. Consequently, a technique was developed for robust absolute flux density calibration using a sky model of diffuse Galactic emission (Oberoi et al. Reference Oberoi, Sharma and Rogers2017). This technique exploits the MWA’s wide FOV and numerous short baselines to which the Sun appears as an unresolved source. Only a few baselines are required (and no imaging is necessary) making the technique computationally lean. A prescription to transfer this flux calibration to solar images has also been developed (Mohan & Oberoi Reference Mohan and Oberoi2017).

In order to follow the spatial, spectral, and temporal evolution of solar emission, it is desirable to image the Sun with the maximum time and frequency resolution possible (typically 0.5 s and 40 kHz time and frequency resolution, respectively). However, this can mean $\sim1\!\times10^5$ images per minute of observation. Therefore, in order to be able to analyse significant amounts of MWA data, it is imperative to use automated pipelines for calibration and imaging. This has motivated the development of an Fermi Automated Imaging Routine for Compact Arrays for Radio Sun (AIRCARS; Mondal et al. Reference Mondal, Mohan, Oberoi, Morgan, Benkevitch, Lonsdale, Crowley and Cairns2019b). This end-to-end interferometric calibration and imaging pipeline also corrects for bandwidth decorrelation due to differing cable lengths, tunes the analysis to the needs of solar imaging, and is able to improve imaging quality well beyond what had been achieved by other fully automated approaches. AIRCARS has successfully been used in a completely hands-off manner for imaging data spanning a large range of solar conditions (peak brightness temperatures ranging from $10^6$ to $10^9$ K) and can routinely produce images with dynamic ranges between $10^3$ and $10^5$ .

6.1.2. Radio bursts

Type III bursts are caused by semi-relativistic electron beams that stream through and perturb the background plasma, generating Langmuir waves that ultimately produce intense radio emission. The electrons are accelerated by magnetic reconnection during solar flares, which was directly evidenced by observations from the 32-tile MWA prototype (Cairns et al. Reference Cairns2018), and the beams then propagate away from the Sun along magnetic field lines connected to the flare site. In particular, this analysis showed definitive evidence for reconnection occurring in association with type III bursts and with X-ray emissions. Additionally, since the associations are not 1:1 in either direction, special conditions are required to produce observable radio emissions and X-rays.

McCauley et al. (Reference McCauley, Cairns, Morgan, Gibson, Harding, Lonsdale and Oberoi2017) reported on Phase I MWA observations of type III source regions that repetitively split apart in a direction roughly perpendicular to the inferred electron beam trajectory, and they concluded that this behaviour resulted from a distinctive magnetic field configuration implied by contemporaneous extreme-ultraviolet observations. These results highlight the MWA’s capability to probe electron beam trajectories, and thereby local magnetic field structures, in new and unexpected ways. Type III bursts can also be used to probe the coronal density, which was explored by McCauley et al. (Reference McCauley, Cairns and Morgan2018), who found much larger densities than expected from standard models. This replicated decades-old findings from previous instruments, and MWA observations of the quiescent corona were used to estimate the extent to which the type III source heights were altered by radio propagation effects, which were found to be important but not entirely sufficient to explain the apparent density enhancements. Comparisons between observations of the scattered source (which includes any angular broadening and ducting) and simulations which typically ignore these effects offer one way to determine the impact of these processes. Since some observations (Mercier et al. Reference Mercier, Subramanian, Kerdraon, Pick, Ananthakrishnan and Janardhan2006) and theory suggest that sometimes scattering is very weak and sometimes strong (Subramanian & Cairns Reference Subramanian and Cairns2011), and the amount of scattering depends on a line integral involving the spectrum of density irregularities, these theory-data comparisons might remotely constrain the spectrum and spatial variations of density turbulence in the corona.

Mohan et al. (Reference Mohan, Mondal, Oberoi and Lonsdale2019b) have succeeded in unambiguously disentangling the effects of scattering from intrinsic variations in emission morphology of a weak type III burst. This has led to the discovery of second-scale quasi-periodic oscillations in the angular size and orientation of the burst source with simultaneous oscillations in the intensity of the burst. These observations cannot be explained in terms of the conventional Alfvènic oscillations. In the same data set, Mohan et al. (Reference Mohan, McCauley, Oberoi and Mastrano2019a) also find an unrelated weak flare seen in the EUV and soft X-rays. Their analysis shows that this flare was responsible for some local coronal heating. A detailed study of the radio emissions associated with the magnetic loop undergoing heating has revealed a strong tendency for the observed short-lived narrow-band nonthermal radio emissions to cluster on a timescale of 30 s. This clustering is very well formed and prominent when the heating event is in progress, but is seen to exist even prior to the start of the heating phase, suggesting that this timescale is independent of the heating event itself. This discovery was made possible by the high dynamic range spectroscopic snapshot imaging capability of the MWA, which enabled a reliable detection of the weak nonthermal emissions originating from the source of interest, even in the presence of other stronger activity. The improved PSF of the MWA Phase II and its higher angular resolution will enable detection of even finer features and further increase the imaging dynamic range of MWA solar images.

Type II bursts are interpreted in terms of shocks accelerating electrons that generate Langmuir waves and then radiation at the electron plasma frequency and its harmonic (Nelson & Melrose Reference Nelson and Melrose1985; Cairns Reference Cairns2011). Observational analyses and simulations predict that sometimes multiple regions of the shock produce the radio emission simultaneously or sequentially, depending on the event (Schmidt & Cairns Reference Schmidt and Cairns2012; Schmidt et al. Reference Schmidt, Cairns and Lobzin2014; Kozarev et al. Reference Kozarev, Raymond, Lobzin and Hammer2015). MWA Phase I observations of the 7 September 2014 type II burst show a stably located, weakly/un-resolved source despite predictions for multiple type II subsources a factor of $\sim 4 - 10$ smaller than the beam size at 140 MHz. The increased resolution of MWA Phase II will provide a stronger test of the multiple-source hypothesis.

Another window on these processes is provided by polarimetry. Studies of the polarised time- and spatially varying source regions of type II and III bursts will also produce new science and constraints on emission mechanisms and scattering processes.

One long-standing issue is that the standard theory for type I, II, and III bursts involves generation of radiation near the electron plasma frequency, that is, 100% circularly polarised in the sense of the ‘ordinary’ (or ‘o’) mode, yet this is not observed for all three: specifically, type Is are almost invariably 100% circularly polarised, type IIs (except for herringbone fine structures) less than $\sim\!20\%$ , and type IIIs are occasionally up to 60% but typically $\mathop < \limits_ \sim 20\%$ (Nelson & Melrose Reference Nelson and Melrose1985; Suzuki & Dulk Reference Suzuki and Dulk1985). One interpretation is that the standard theories are wrong but another is that scattering depolarises the radiation (Melrose Reference Melrose2006). MWA Phase II can address these issues by explicitly measuring the time- and spatially varying sources in all four Stokes parameters, searching for evidence of strongly polarised substructures that vary with time and space (e.g. depolarisation fronts moving away from the cores of subsources versus random flickering polarisation regions).

In conclusion, studies of solar radio bursts are expected to benefit substantially from MWA-Phase II’s improvements in linear spatial resolution (a factor close to 2) and better Fermi uv plane coverage (better imaging fidelity). These should result in much better characterisation of the complex, sub-structured sources of type II and III bursts in position, time, and polarisation, thereby enabling more quantitative tests of theoretical ideas and predictions.

6.1.3. Radio counterparts of coronal mass ejections

The bulk of solar dynamics is dictated by solar magnetic fields, but beyond the photosphere their measurements have remained difficult. Radio observations form the only known techniques to measure these fields. The first successful attempt to model the spectrum of radio emission from the Coronal Mass Ejection (CME) plasma as gyrosynchrotron emission and hence provide an estimate of the CME magnetic field was by Bastian et al. (Reference Bastian, Pick, Kerdraon, Maia and Vourlidas2001). In spite of their well-recognised merits, and considerable effort expended towards them, there are only a handful of successful detections available in the literature (Maia et al. Reference Maia, Gama, Mercier, Pick, Kerdraon and Karlicky2007; Tun & Vourlidas Reference Tun and Vourlidas2013; Bain et al. Reference Bain, Krucker, Saint-Hilaire and Raftery2014; Carley et al. Reference Carley, Vilmer, Simões and Ó Fearraigh2017), some of which are different attempts to model the same CME. The key reason for this lack of success is that the intrinsic $T_B$ of gyrosynchrotron emission is at least a few orders of magnitude lower than that of the other nonthermal emissions usually present during the times when a CME has recently lifted off, and the imaging dynamic range provided by instrumentation available until recently fell short of the requirement to be able to image this weak emission. Recent work with the MWA has shown that with its improved imaging dynamic range, and ability to simultaneously sample its entire observing band, it should now be possible to routinely image the radio counterparts of CMEs and model their spectra in a more robust manner than possible before (Mondal et al. Reference Mondal, Oberoi and Vourlidas2019a). The improved resolution of the MWA Phase II will reveal the structures in CMEs in greater detail and its better imaging dynamic range will permit tracking of these emissions further out into the heliosphere, making these measurements relevant from a space weather perspective.

6.1.4. Weak non-thermal emission

Early observations with the 32-element MWA prototype revealed a previously unappreciated abundance of weak, short-lived (few s), narrow-band (few MHz) impulsive nonthermal emission features carpeting the frequency–time plane (Oberoi et al. Reference Oberoi2011). Detailed non-imaging investigations of weak impulsive nonthermal emissions have been carried out using the MWA Phase I with the aim of providing robust statistics on these emissions. Using an automated continuous wavelet transforms-based approach, Suresh et al. (Reference Suresh2017) found that these impulsive emissions take place at the rate of many thousand per hour (as measured across 30.72 MHz bandwidth). Individual emission features were found to last for 1–2 s and span 4–5 MHz in bandwidth and $\sim\! 1-100$ SFU ^{Footnote e} in peak flux densities. Their characteristics suggest that these features might represent the weaker end of the distribution of solar type I bursts. Using a Gaussian mixtures-based statistical analysis technique, Sharma et al. (Reference Sharma, Oberoi and Arjunwadkar2018) modelled the solar radio emission as a superposition of a slowly varying emission component of thermal origin and an impulsive, and hence nonthermal, component. They estimated the flux density distribution as well as the prevalence of impulsive nonthermal emission in the frequency–time plane. They found the fractional occupancy of nonthermal impulsive emission to lie in the range 17%–45% even during a period of medium solar activity, and that the flux density radiated in the impulsive component is very similar in strength to that radiated in the thermal emission-dominated slowly varying component.

These studies were motivated by the desire to explore the use of weak nonthermal emissions in looking for signatures of weak magnetic reconnection events, the presence of which was originally suggested by Parker (Reference Parker1988). Referred to as the nanoflare hypothesis, the simultaneous presence of a large number of weak reconnection events carpeting the solar surface, but too weak to be detected individually, was proposed as a possible resolution of the coronal heating problem and remains one of the likely possibilities. At flux densities of $\sim 0.2$ SFU, these are the weakest nonthermal emissions reported at low radio frequencies using non-imaging techniques. Ongoing imaging studies using MWA Phase I suggest RMS variability during quiet Sun conditions at the levels of $10^{-3}$ SFU (i.e. 10 Jy), by far the weakest to be reported ever. The high dynamic range techniques described in Sec. 6.1.1 along with the higher angular resolution of MWA Phase II will enable even deeper studies to come. Although a lot remains to be done before their role in coronal heating can be elucidated, these studies have firmly established the presence and abundance of nonthermal emissions significantly weaker than were known before.

6.1.5. Quiescent sun

It is not only bursts that are of interest but also sources on the quiescent Sun, for instance, associated with coronal holes, active regions, and the quiet Sun itself. Work with MWA Phase I intensity data shows new features of coronal holes and active regions (McCauley et al. Reference McCauley, Cairns, Morgan, Gibson, Harding, Lonsdale and Oberoi2017). MWA Phase II’s better angular resolution and imaging fidelity will allow better characterisation of these sources and better comparisons with the SDO, SOHO, and RHESSI data from the UV to X-rays.

In addition, recent work using Phase I data has produced the first low-frequency images of the quiescent solar corona in circular polarisation (McCauley et al. Reference McCauley, Cairns, White, Mondal, Lenc, Morgan and Oberoi2019). The Stokes Fermi V structure at the lowest MWA frequencies is found to be generally well-correlated with the structure of the line-of-sight magnetic field component in a widely used global magnetic field model at a height roughly corresponding to that of the radio limb. Coronal magnetic field models are typically extrapolated from photospheric observations, and there are very few observations that can be used to constrain the field at the heights and scales probed by MWA observations. This is a powerful and unique capability of the MWA that will be greatly enhanced by the improved spatial resolution of the Phase II extended mode, as Phase I could not resolve many of the features that would discriminate between competing models.

6.2.1. Interplanetary scintillation

Interplanetary Scintillation (IPS) is the rapid ( $\sim$ 1 s) variability in brightness of compact ( $\mathop < \limits_ \sim $ 1 arcsecond) sources due to scattering of the propagating wave by the solar wind: a supersonic outflow of turbulent plasma which fills interplanetary space. A detection of IPS with the MWA was first made serendipitously in night-time astrophysical data by Kaplan et al. (Reference Kaplan2015a). Following this, a pilot study was made using observations designed for the purpose, at solar elongations most suitable for detecting IPS (Morgan et al. Reference Morgan2018). This pilot study demonstrated that the MWA has a unique capability for detecting IPS of several hundred sources simultaneously in a single 5-min observation. From a solar science perspective, this means that the heliosphere can be mapped in unprecedented detail in a very short amount of time.

These findings motivated a 6-month long-observing campaign, described in detail by Morgan et al. (Reference Morgan, Macquart, Chhetri, Ekers, Tingay and Sadler2019). The analysis of these data is ongoing; however, much of the work so far has focused on determining the astrophysical nature of the sources detected. They differ somewhat from compact source populations at higher frequencies, most notably they are dominated by peaked-spectrum sources (Chhetri et al. Reference Chhetri, Morgan, Ekers, Macquart, Sadler, Giroletti, Callingham and Tingay2018a). The precise counts of the various sub-populations are described in Chhetri et al. (Reference Chhetri, Ekers, Morgan, Macquart and Franzen2018b). Sadler et al. (Reference Sadler, Chhetri, Morgan, Mahony, Jarrett and Tingay2019) have shown that the most compact IPS sources are high-redshift sources, 1/3 of which have a redshift $>\!2$ .

A further key finding (Morgan et al. Reference Morgan2018) was the necessity to maximise the sensitivity and imaging fidelity on short timescales ( $\sim$ 1 s). In contrast to most astrophysical applications, where MWA sensitivity is limited by confusion noise (see, e.g., Wayth et al. Reference Wayth2015), IPS sources are more rare and therefore the confusion limit is at a considerably lower flux density. However, since IPS detections are made by measuring an increased variance in flux density, the detection limit in an observation only reduces with the fourth root of observing time. This makes maximising the instantaneous sensitivity of the array critical.

This can be maximised in synthesis images by choosing a natural weighting scheme (where all baselines are weighted equally). However, this comes with two disadvantages: firstly, the resolution is poor compared to a weighting scheme which increases the weight of longer baselines (i.e. a uniform weighting scheme); secondly, the PSF of the array in a naturally weighted image typically has higher sidelobes. With the Phase I MWA, this latter effect is very strong due to the very high concentration of tiles in the core. This is in contrast to uniformly weighted images which have an exceptionally good PSF due to the large number of interferometer elements in the MWA and their pseudo-random arrangement. For IPS observations (where the Sun is typically in the sidelobes), uniform weighting has the added advantage that it downweights the extended quiet Sun significantly.

As shown in Wayth et al. (Reference Wayth2018), by utilising the extended Phase II MWA and using natural weighting, we combine the resolution and exceptionally low sidelobes of uniformly weighted Phase I images with the sensitivity of naturally weighted Phase I images. This should lead to a factor of 2 improvement in sensitivity (Wayth et al. Reference Wayth2015).

6.2.2 Faraday rotation measurements

Kinetic and magnetic energy from a CME can create major, potentially damaging disturbances in the near-earth environment, and there is high interest in being able to predict both the arrival time and the effects of CME impacts with sufficient precision and lead-time. Since the geo-effectiveness of CME impacts is strongly dependent on the orientation of the magnetic field in the CME, a long-standing goal of low-frequency observations in the context of space weather is to measure the orientation of the magnetic fields in propagating CMEs. The only known means of remote detection of this orientation is via the measurement of Faraday rotation (FR) in the plasma of the CME (Oberoi & Lonsdale Reference Oberoi and Lonsdale2012).

Because of the radial dependence of coronal/solar wind density and magnetic field strength, attempts to detect FR from heliospheric plasma, and in particular from CME events, have been confined to regions close to the sun. Some measurements have used linearly polarised signals from spacecraft (e.g. the Helios spacecraft at distances of 2–20 solar radii, using the 2.295 GHz carrier or the MESSENGER spacecraft using the 8 GHz carrier, probing the corona at a few solar radii Hollweg et al. Reference Hollweg, Bird, Volland, Edenhofer, Stelzried and Seidel1982; Jensen et al. Reference Jensen, Nolan, Bisi, Chashei and Vilas2013; Efimov et al. Reference Efimov, Lukanina, Rogashkova, Samoznaev, Chashei, Bird and Paetzold2015). Others have used pulsars (Ord et al. Reference Ord, Johnston and Sarkissian2007; You et al. Reference You, Coles, Hobbs and Manchester2012), again at small solar elongations where the coronal plasma and field strengths are high. Observations of polarised extragalactic sources have also been conducted (e.g. Le Chat et al. Reference Le Chat, Kasper, Cohen and Spangler2014), most recently with considerable success using the upgraded Karl G. Jansky VLA (Kooi et al. Reference Kooi, Fischer, Buffo and Spangler2017), also at small solar elongations. The pulsar and extragalactic source observations have typically been done at low GHz frequencies.

In order to probe heliospheric and CME FR at large fractions of an au, much lower frequencies must be used so that the amount of rotation is detectable. This requires that a source of background linearly polarised emission must be present, but discrete polarised sources tend to be few and weak at MWA frequencies. Furthermore, to be of practical utility, any FR measurements must target the inner heliosphere for CMEs which may be on a collision course, which generally means performing the FR observations during the daytime. Due to the strength of solar radio emission, this places extraordinary demands on the imaging dynamic range of these observations.

The MWA Phase I system has allowed major advances in these studies for two key reasons. First, as has been demonstrated by MWA observations (Lenc et al. Reference Lenc2016), the polarised galactic synchrotron background has much more power on large angular scales than on small scales, a discovery made possible by the excellent short-baseline Fermi uv coverage and imaging capability of the MWA. Thus, by employing low angular resolution imaging, a strong polarised signal against which to measure FR is readily available in essentially all directions. This is a dramatically more favourable scenario than trying to use a sparse grid of weak polarised discrete sources. Second, the excellent instantaneous Fermi uv coverage of the MWA combined with thousands of independent closure quantities has allowed the development of very high dynamic range solar imaging, which in turn creates a realistic possibility of detecting much weaker emission (the polarised Galactic background) during the daytime, and making FR measurements of CME plasma in the inner heliosphere. Work continues to further improve dynamic range via the development of direction-dependent ionospheric calibration on small linear scales.

For FR work, the Phase II system differs from Phase I in one important respect, namely that for either high or low angular resolution work, essentially all 128 tiles are available, instead of a subset. For low-resolution FR studies, the compact configuration thus yields not only greater sensitivity but also denser Fermi uv coverage with extensive redundancy (by design) to constrain calibration solutions. Improved calibration, the inclusion of direction-dependent, small linear scale ionospheric calibration, and the much simpler nature of the angular structure of the solar emission on scales of $0.5^{\circ}$ and larger combine to offer the prospect of meaningful daytime FR measurements. Such measurements are potentially possible using existing Phase I data, but compact configuration Phase II data will be inherently superior.

The first detection and characterisation of spatially resolved FR associated with an interplanetary CME would be a major result, offering a densely sampled large-volume view of a quantity directly related to the magnetic field geometry. This would transform studies of CME dynamics and would point the way to a future network of low-frequency imaging arrays, modelled after the MWA, that could form a crucial part of a space weather prediction network.