1. Introduction
Large pulsar surveys have played a pivotal role in advancing pulsar astronomy. They have collectively led to a substantial increase in the population of known pulsars (e.g., Manchester et al. Reference Manchester1978, Reference Manchester2001; Han et al. Reference Han2021), with the discoveries of exotic objects (e.g., the double pulsar, the transitional millisecond pulsar) leading to transformational science (e.g., Lyne et al. Reference Lyne2004; Kramer et al. Reference Kramer2021; Archibald et al. Reference Archibald2009). As such, conducting a full Galactic census and exploiting the specialised targets for high-profile science, such as testing strong-field gravity and detecting nanohertz-frequency gravitational waves, is a headline science theme for the Square Kilometre Array (SKA) and many of its precursor and pathfinder facilities (Keane et al. Reference Keane2015; Shao et al. Reference Shao2015; Janssen et al. Reference Janssen2015; Bailes et al. Reference Bailes2020).
Figure 1. Sky plot summarising the observing strategy adopted for the SMART pulsar survey and the progress made to date: the full visible sky (i.e., declination
$< +30^{\circ}$
) is covered using 70 pointings that overlap
$10^{\circ}$
in LST and
$15^{\circ}$
in declination. The coloured contours represent the half-power points of the main lobe of the tile (primary) beam (at 155 MHz). The blue, red, and green pointing sets, as well as 10 of the 13 purple ones, have already been observed through four dedicated observing campaigns undertaken in the 2018B, 2019B, 2020A, 2021A semesters. The red filled circles are the known pulsar detections from the survey data (120 so far), and the black stars are the three new pulsar discoveries, including PSR J0026
$-$
1955, an independent discovery from the SMART. The grey star is PSR J1357
$-$
2530, a re-discovery of a previously incorrectly characterised pulsar.
The radically new designs of next-generation facilities bring significant changes (as well as challenges) to the pulsar survey landscape. For instance, all-sky surveys, which have been mainly a single-dish arena for long, will now have to be conducted using distributed interferometric arrays or aperture arrays that make up these new generation facilities such as the Low Frequency Array (LOFAR; van Haarlem et al. Reference van Haarlem2013) and the Murchison Widefield Array (MWA; Tingay et al. Reference Tingay2013; Wayth et al. Reference Wayth2018). They present a substantial complexity given the inherent beamforming cost and large data rates, but also provide the benefits of significantly larger survey speeds, and new avenues for candidate confirmation and follow-up. At frequencies below 300 MHz, the ongoing LOFAR Tied-Array All-Sky (LOTAAS; Sanidas et al. Reference Sanidas2019) survey is the first such major step, while the South African MeerKAT telescope is proving to be highly successful in sensitive targeted searches (e.g., Ridolfi et al. Reference Ridolfi2022). In the first paper of this series (hereafter referred to as Paper I), we described the Southern-sky MWA Rapid two-metre (SMART) survey, a large project to search for pulsars and fast transients with the MWA’s 140-170 MHz band, in the entire sky south of +30° in declination.
While there has been substantial progress in terms of data collection effort, the SMART survey is currently in its early stages in terms of data processing and analysis. The adoption of voltage recording for survey data collection requires full-scale tied-array beamforming as the front-end of the processing chain, but it offers a number of unique benefits for candidate follow-up and confirmation. Most notably, the SMART survey has minimal reliance on new observations for immediate follow-up and/or confirmation of pulsar candidates. Access to the voltage data even allows accurate positional determination through a dense-grid localisation, using either survey data alone, or in combination with archival data, as vividly demonstrated in Paper I (see also Swainston et al. Reference Swainston2021) through examples from early discoveries.
Nevertheless, scientific merits of most new pulsar discoveries are typically revealed in due course of time, often through targeted follow-ups using more sensitive telescopes, or long-term timing observations. Two important pre-requisites are: (1) an accurate sky position, and (2) a full coherent timing solution. The latter typically relies on a series of regular observations carried out over the course of time. Again, with the advent of sensitive high-resolution imaging instruments such as the upgraded Giant Metre-wave Radio Telescope (uGMRT; Gupta et al. Reference Gupta2017), and large survey data releases from sensitive imaging facilities such as the Australian SKA Pathfinder (ASKAP; McConnell et al. Reference McConnell2020), source identification and localisation can be expedited for arc-second level positional determination early on, thereby accelerating convergence to an initial timing solution. The design of the SMART also offers additional benefits, by virtue of its overlapping pointings, as detailed in Paper I. Moreover, when archival data are available, they can be utilised for an investigation of source variability, or for extending timing analysis, as demonstrated in Swainston et al. (Reference Swainston2021).
In Paper I, we described the survey design, science goals, current processing pipelines and follow-up strategies. In this sequel to Paper I, we present the survey status and further details on the initial discoveries, demonstrate and verify some of the detailed follow-up strategies, and provide a brief census of pulsar re-detections. In Section 2, we present the survey status, and in Section 3 details on confirmation and follow-up. Initial pulsar discoveries are summarised in Section 4, where we also present re-detections of a large sample of previously known pulsars from our SMART data processing. A summary is presented in Section 5.
2. Survey status
Even though the use of the voltage capture system (VCS) and the MWA’s large field-of-view (FoV) means the data collection part of the survey can be completed in
$<$
100 h of telescope time, practical considerations such as the data rate (42 TB per observation) and the availability of the compact configuration necessitate that the survey be carried out in multiple observing campaigns. By taking advantage of such windows of opportunity in the past few years, we have advanced the data collection effort to 75%. Specifically, through four dedicated campaigns in the 2018B, 2019B, 2020A and 2021A semesters, we have covered much of the sky within the right ascension (RA) range
$<$
16 h and
$>$
21 h and declination south of
$+30^{\circ}$
, as shown in Figure 1. The fourth campaign to cover the 13–17 h RA range was only partially complete, as the system was no longer available for observations for an extended period since 2021 May. Further details of the pointings and observing campaigns are summarised in Table 1. The remaining 25% of the survey will be undertaken once the array returns to the compact configuration (by around mid-2023). Substantial progress has already been made with the science commissioning of the new corelator/voltage capture system (MWAX VCS), with the successful demonstration of the ability to attain full coherent beam sensitivity.
Table 1. Summary of the SMART survey observing campaigns to date.
aThe number in parentheses are common redetections with the previous semester; e.g., 3 pulsars were detected in both 2018B and 2019B observations.
In the ongoing first-pass search, 10 min of data from each observation are processed in 2358 trial dispersion measures (DMs) out to 250
${\textrm{pc cm}}^{-3}$
, thus effectively performing a shallow survey of a large part of the southern sky. The limiting sensitivity achieved is comparable, or slightly better than, that of the Parkes 70 cm survey from the 1990s (Manchester et al. Reference Manchester1996), at least for pulsars with spectral index
$\alpha \lesssim -1.6$
. At the time of writing, processing is completed for
${\sim}$
10% of the data collected thus far, covering
${\sim}$
55% of the sky searched in the shallow survey phase, resulting in
${\sim}$
1 million candidates, after filtering through the current machine-learning (ML) tools. Efforts are under way to substantially improve the efficacy of candidate filtering schemes, using semi-supervised generative adversarial network (SGAN) based classifiers (e.g., Balakrishnan et al. Reference Balakrishnan2021), and will be integrated into the processing pipelines in the near future. Initial results are encouraging, and details including current implementation and the efficacy achieved, will be described in a future publication.
The current efficacy of ML tools means that a large volume of candidates require human scrutiny; however, only a small fraction (
$\lesssim$
3%) of these have been examined for promising candidates to follow-up, resulting in four pulsar discoveries, including an independent discovery and a re-discovery of a previously incorrectly characterised pulsar, as well as re-detections of 120 previously known pulsars. All four discoveries have been followed up with Parkes and uGMRT, and also using archival/new data from the MWA, for early science extraction, and with the intent to develop and demonstrate effective follow-up strategies. Detection plots of PSRs J1002
$-$
2044 and J1357
$-$
2530 are shown in Figure 2. The parameters of the three new pulsars are summarised in Table 2, and some of their detailed characteristics are described in the sections below.
Table 2. Parameter summaries of SMART pulsar discoveries.
aIndependent rediscovery. The source was first detected as a GBNCC candidate (cf. McSweeney et al. Reference McSweeney2022). A full timing solution from MWA data is presented in Table 4.
bIndependent rediscovery of a previously incorrectly characterised pulsar. See text (Section 4.1.4) for details.
Figure 2. Detection plots of PSRs J1002
$-$
2044 and J1357
$-$
2530 from SMART search processing. Panels (a) and (c) are diagnostic plots (from PRESTO) of the original candidate detections (PSR J1002
$-$
2044 and PSR J1357
$-$
2530, respectively), while panels (b) and (d) show the improved detections from follow-up processing (confirmation) using the same inital detection observation, but by forming a tied-array beam for the full 80-min observing duration.
3. Follow-up of pulsar discoveries
As described in Paper I, the SMART survey’s unique design offers a range of flexible reprocessing avenues for candidate confirmation and improved positional determination, without having to rely on making immediate new observations. Furthermore, the archival data from past VCS projects can be leveraged for further detailed follow-up and characterisation, as demonstrated in Swainston et al. (Reference Swainston2021) and McSweeney et al. (Reference McSweeney2022); e.g., extending an initial timing solution derived from regular observations to determine the spin period (P) and its first time-derivative (
${\dot P}$
). An improved sky position is vital for follow-ups with more sensitive telescopes operating at higher frequencies, but a convergence toward this can be expedited via high-resolution imaging using facilities such as the uGMRT.
In the following sections we outline some additional strategies that are natural extensions to those already described in Paper I. These include: (1) an interferometric localisation via high-resolution imaging, (2) the processing of archival VCS observations for an extended timing analysis, and (3) high-frequency follow-up using sensitive telescopes such as the uGMRT and Parkes. As we have demonstrated in earlier publications from the SMART survey, some of these are proving useful for rapid convergence to a working pulsar timing solution (e.g., Swainston et al. Reference Swainston2021), and for ascertaining the detailed characteristics in terms of scientific significance (e.g., McSweeney et al. Reference McSweeney2022).
3.1. Imaging for improved localisation
3.1.1. High resolution imaging with the uGMRT
Beyond the tied-array beam localisation as detailed in Paper I, the position of SMART-discovered pulsars can be further improved via high-resolution continuum imaging with interferometric arrays, for which the high sensitivity and angular resolution of the upgraded GMRT (uGMRT) is highly appealing, provided the source is north of
$-55^{\circ}$
in declination, and the spectrum is not ultra-steep (spectral index,
$ \alpha \gtrsim -3 $
). For PSR J0036
$-$
1033, the first discovered MWA pulsar, this was convincingly demonstrated in Swainston et al. (Reference Swainston2021), for which the localisation uncertainty was significantly improved from an initial
${\sim}$
$10^{\prime}$
(i.e., discovery observation) to
${\sim}$
$10^{\prime\prime}$
through a dense-grid tied-array beam (TAB) localisation using the three different array configurations, and then eventually to
$ \lesssim$
$1^{\prime\prime}$
via high-resolution imaging of the pulsar using uGMRT Band 3 and 4 (300–750 MHz) observations.
For PSR J0026
$-$
1955, even though uGMRT observations were made in similar bands using the identical setup (i.e., concurrent visibility recording and phased-array beams), and analysed using similar procedures, imaging of Band 3 data (300–500 MHz) proved difficult due to the presence of a bright source (
$S_{400} \sim 4\,\textrm{Jy})$
in the vicinity (
${\sim}$
$10.5^{\prime}$
from the putative pulsar position). Fortunately, our earlier observations that were made as part of the DDTC185 project (and before the realisation that the original detection was in fact through one of the grating lobes), provided significant attenuation of the source owing to the initial
${\sim}$
$42 ^{\prime}$
offset in (and hence incorrect) position. Analysis of these observations resulted in a
${\sim}$
$15\sigma$
detection, as shown in the top left panel of Figure 3, notwithstanding the fact that the pulsar was at more than half-power beam width offset from the phase centre. The resultant rms of
$ 276\,{{\unicode{x03BC} {\rm Jy \, beam} ^ {-1}}} $
is nearly a factor of two larger than that was achieved with the imaging analysis of PSR J0036
$-$
1033.
Figure 3.
Upper panels: uGMRT images of PSR J0026
$-$
1955 in Band 3 (300–500 MHz; left) and Band 4 (550–750 MHz; right), obtained from integration times of 136 and 56 min, respectively. The rms sensitivity is
$\approx$
276
$\unicode{x03BC} {\rm{Jy}}\, \rm{beam}^{-1}$
in Band 3, and
$\approx$
210
$\unicode{x03BC} {\rm{Jy}}\, \rm{beam}^{-1}$
in Band 4, respectively, yielding
${\sim} 15\sigma$
and
${\sim} 10\sigma$
detection of the pulsar. The presence of a bright nearby source (4 Jy at 400 MHz, at
${\sim} 10.5^{\prime}$
from the pulsar position) impacts the quality of calibration achievable, as seen in Band 4 observations. Band 3 observations were made at an offset position from the pulsar, and the effect of the bright source was subdued owing to attenuation of the primary beam. Lower panels: uGMRT images of PSR J1357
$-$
2530 in Band 3 (300–500 MHz; left) and Band 4 (550–750 MHz; right), obtained from integration times of 90 and 71 min, respectively. The rms sensitivity is
$\approx$
153
$\unicode{x03BC} {\rm{Jy}}\, \rm{beam}^{-1}$
in Band 3, and
$\approx$
53
$\unicode{x03BC} {\rm{Jy}}\, \rm{beam}^{-1}$
in Band 4, respectively, yielding
${\sim}$
$6\sigma$
and
${\sim}$
$11\sigma$
detection of the pulsar. Band 3 imaging of the initial (poorly localised) pulsar position revealed multiple compact sources, one of which turned out to be the pulsar (see Section 4.1.4 for details).
The imaging of Band 4 data (550–750 MHz) benefited from a much narrower primary beam (and hence a substantially higher attenuation toward the bright, nearby source), besides a reduced flux density due to spectral index, yielding a
${\sim}$
$10\sigma$
detection of the pulsar, as shown in the right panel of Figure 3. The rms attained in this case is
$\approx$
$210\,{{\unicode{x03BC} {\rm{Jy \, beam}} ^ {-1}}} $
, almost an order of magnitude larger than that is typically attainable for deep imaging of 1 h in uGMRT Band 4. This illustrates the complexity in interferometric localisation, when the imaging analysis is hampered by bright sources in the vicinity.
Table 3. Localisation summary of PSR J0026
$-$
1955.
Figure 4. Detection summary of PSR J0026
$-$
1955, over a
${\sim}$
3.5-yr time span; the PRESTO detection significance (
$\sigma$
) is plotted against the MJD of observation. The detections are predominantly from observations over a 30.72 MHz bandwidth centred at 154.24 MHz, with time integration varying from 10 min to 1 h. Note that the
$\sigma$
reported for the discovery observation is for the full 80-min data that was processed for follow-up and not for the 10-min observation for initial (first-pass) search, in which the pulsar was discovered. The 3-
$\sigma$
upper limits indicate non-detection in a number of 10–20 min observations.
A summary of uGMRT-derived positions are given in Table 3, along with the initial MWA-derived position. There is
$\approx$
$32^{\prime\prime}$
discrepancy between the uGMRT- and MWA-derived positions, which can be possibly attributed to a residual ionospheric refraction as discussed in Swainston et al. (Reference Swainston2022a). It is possible that the
$\approx$
$3^{\prime\prime} $
offset between the two uGMRT bands may also be caused by the ionosphere; even so, the final position, which we adopt as the mean of those from the two uGMRT bands, is still over an order of magnitude more precise than that reported by McSweeney et al. (Reference McSweeney2022). Further resolution to this will be possible once an improved (and fully coherent) timing solution becomes available.
3.1.2. Imaging with the MWA
For sources outside the uGMRT’s sky (
$\delta < -55^{\circ} $
), in principle, the recorded VCS data (if available) from long-baseline configurations of the array (Phase I or Phase II extended) can be correlated offline and imaged to verify the position obtained via the TAB localisation. However, as we demonstrated in Swainston et al. (Reference Swainston2021), even with co-addition of multiple observations we were unable to reach a sensitivity better than
${\sim}$
6–8
${\rm{mJy}}\, \rm{beam}^{-1}$
(for Stokes I), and hence an imaging detection may be possible only for pulsars that are sufficiently bright (
${S_{150}}\,\gtrsim\,30$
mJy). For long-period pulsars, such as PSR J0026
$-$
1955, ‘on’ and ‘off’ gated imaging may also be attempted for improved detection in imaging, though this methodology is currently under development.
3.2. Archival observations
Apart from the observations taken as part of the SMART campaign, the MWA archive includes many VCS observations dating back to 2014 when early pulsar and high-time resolution science observations commenced. These observations were typically parts of projects focusing on individual pulsars, so the distribution of the pointing directions across the sky is non-uniform, and some areas of the sky are not covered at all. Nevertheless, if a promising SMART candidate happens to be in an historically observed part of the sky, then the archival VCS data can be used for further confirmation of the candidate without requiring extra observations be taken.
Again we use PSR J0026
$-$
1955 as an example, whose location had fortuitously been covered by multiple previous observations. Six of these observations are included in the analysis described in McSweeney et al. (Reference McSweeney2022), but since then, more archival observations have been identified and processed, as shown in Figure 4. Depending on the intervals between detections, their duration, and the quality of the detections themselves, such archival observations can be used to constrain otherwise hard-to-measure properties of the candidate pulsars which require long follow up campaigns, such as the flux variability due to scintillation, and their timing properties.
3.2.1. Leveraging archival data to achieve timing coherence
As an example of using the archival MWA data and improved positions from imaging follow-up (see Section 3.1), we are able to extract times-of-arrival (TOAs) from MWA observations alone and, in principle, phase-connect a timing solution. Here, we demonstrate that for PSR J0026
$-$
1955 across a
${\sim}$
1.7-yr time span.
Examining the MWA VCS observation database revealed observations containing the nominal pulsar position dating back to 2017 April, including dedicated followup observations for PSR J0036
$-$
1033 in late-2020 into 2021. These serendipitous observations are well suited to generating a phase-connected timing solution as there was a dense observing campaign where the same field was observed 8 times in three days (2020 October 26–28), surrounded by regular observations spaced every 3–4 d for approximately 2 weeks before (2020 October 10–22) and roughly weekly cadence for one month after (2020 October 30 to 2020 November 27). Monthly cadence data were then collected until the system became unavailable for science observing (see Section 2). These observing time scales are adequate to accurately determine the rotation frequency, detect spin-down and measure the DM.
The initial rotational characteristics presented by McSweeney et al. (Reference McSweeney2022) are sufficient to fold PSR J0026
$-$
1955 for a single observation, but quickly lose phase connection after
${\sim}$
1 d. Using the uGMRT imaging position (Figure 3) and the initial estimates of the spin period and DM, we processed some of the more recent archival observations (two observations in 2019 April) and the majority of the PSR J0036
$-$
1033 timing followup observations (17 observations spanning 2020 October–November), creating 10-s subintegration archives with DSPSR. These archives were then subsequently processed using PSRCHIVE routines to remove radio frequency interference (RFI), specifically using the paz median-filtered bandpass algorithm. Since PSR J0026
$-$
1955 is a prolific nuller (cf. McSweeney et al. Reference McSweeney2022), we manually inspected each observation with pazi and masked the null subintegrations to construct a high-quality average (integrated) profile. Out of the total 21 observations, we processed for this work, 10 were excluded either due to non-detections (unsurprising for a pulsar with a
$>$
70% nulling fraction) or very weak detections (signal-to-noise ratio, S/N < 15 for the fully time- and frequency-averaged profile). The five observations with the highest S/N were then combined and artificially phase aligned using psradd and a basic frequency-averaged template was created using paas, where two Gaussian components modelled the profile.
Each observation was then pre-processed such that there were four frequency subbands per epoch, from which TOAs were extracted using the pat utility within PSRCHIVE. The initial ephemeris used to fold the observations and the TOAs were passed to both the Tempo2 (Hobbs, Edwards, & Manchester Reference Hobbs, Edwards and Manchester2006; Edwards, Hobbs, & Manchester Reference Edwards, Hobbs and Manchester2006) and PINT (Luo et al. Reference Luo2021) pulsar timing packages. With each timing package, we were able to phase-connect all available observations to the archival data points from nearly two years earlier by updating the spin frequency, adding a spin frequency time-derivative (
$28\sigma$
significance), and small adjustment to the DM. Both packages produced consistent results and uncertainties. Given the sparse sampling on longer time scales we did not fit for the pulsar sky position, however, as we accrue and/or process additional observations (with the MWA or other telescopes) we will be able to constrain the position to better than the current
${\sim}$
$3^{\prime\prime}$
uncertainty obtained from the uGMRT imaging. The current best timing solution (from Tempo2) is presented in Table 4.
Table 4. Timing solution for PSR J0026
$-$
1955 using Tempo2.
aThe position of the pulsar was fixed at the uGMRT imaging position, as given in Section 3.1.
3.3. High-frequency follow-ups
Our ability to attain a positional accuracy of
${\sim}$
1–2
$^{\prime}$
from discovery observations, or even down to
${\sim}$
10–20
$^{\prime\prime}$
when archival VCS data are also available from the long-baseline configurations of the MWA (see Paper I), facilitates high-frequency follow-ups with sensitive telescopes such as the uGMRT and Parkes that share substantial common skies with the MWA. The combination of these telescopes and their wide-band instrumentation allow nearly seamless coverage in frequency from
${\sim}$
100 MHz to 4 GHz, thereby enabling a detailed characterisation of spectral and propagation properties of new pulsar discoveries. For instance, in the case of PSR J0036
$-$
1033, such follow-up observations helped establish the steep-spectrum (
$ S _{\nu} \propto \nu^{-2.0 \pm 0.2} $
, where
$ S _{\nu} $
is the flux density at frequency
$ \nu $
) low-luminosity nature of the object (with a pseduo-luminosity
$ L_{1400} \sim 0.1 $
${\rm{mJy}}\, \rm{kpc}^{2}$
at 1400 MHz). Observational studies of this kind are an integral part of our follow-up strategies, and are important for an improved understanding of the Galactic pulsar population in general.
Example detections from such high-frequency follow-ups are shown in Figure 5 for PSR J0026
$-$
1955. These were obtained using data collection and analyses procedures similar to those described in Swainston et al. (Reference Swainston2021), i.e., concurrent imaging and phased-array observations using the uGMRT Band 3 (300–500 MHz) and Band 4 (550–750 MHz) receivers, and observations using the Parkes (Murriyang) telescope. Observations at higher frequencies were made in the search mode of the Parkes ultra-wideband low-frequency (UWL; Hobbs et al. Reference Hobbs2020) receiver that spans the frequency range from 704 to 4032 MHz, and followed analysis procedures as described in Dai et al. (Reference Dai2019).
Figure 5. Detection plots of PSR J0026
$-$
1955 with the MWA (left panel), uGMRT (two central panels), and Parkes (right panel) telescopes, spanning a frequency range from 140 to 1400 MHz: the top panel is the integrated pulse profile and the waterfall plot below shows the pulse strength vs. pulse phase and frequency. MWA observations were made with the Phase II compact configuration of the array, whereas those with the uGMRT made use of the 200 MHz mode of the phased-array beamformer comprised of 11–13 antennas located within the central square. Parkes observations were made using the Ultra-Wideband Low-frequency receiver (704–4032 MHz); however, the pulsar was detected only in the band below 1.4 GHz. The profiles have been downsampled to 256 bins across pulse phase and 64 frequency channels for plotting clarity.
The best detection was made in the uGMRT Band 3 frequency range, where S/N
${\sim}$
300 was achieved in one of the observing sessions in
${\sim}$
30-min integration. Despite the severe RFI contamination of the Parkes data, the pulsar was clearly detected in the low band range 750–1300 MHz, as shown in Figure 5. Estimated flux densities are
${\sim}$
4.2 and 2
${\rm{mJy}}$
, respectively at 400 and 650 MHz (from uGMRT imaging, see Figure 3). The pulsar was also detected in the TIFR GMRT Sky Survey (Intema et al. Reference Intema, Jagannathan, Mooley and Frail2017), with a flux density of
${\sim}$
19
${\rm{mJy}}$
(at 150 MHz) and also in the ASKAP-low and -high bands, with flux densities of
${\sim}$
1.3
${\rm{mJy}}$
(at 888 MHz) and
${\sim}$
0.3
${\rm{mJy}}$
(at 1.5 GHz), respectively (E. Lenc, private communication). The combination of these measurements suggest a spectral index
$\alpha \sim -1.6$
for this pulsar, assuming no turnover down to
${\sim}$
150 MHz. However, as reported in McSweeney et al. (Reference McSweeney2022), the pulsar exhibits a large nulling fraction of
${\sim}$
75%, and complex sub-pulse drifting behaviour, including rapid changes in the drift rate and consistent evolution within its two main drift modes. Further analysis using these data are deferred to a future publication (Janagal et al., in preparation).
4. Pulsar discoveries and census
As described in Section 2, a first-pass search processing is currently in progress, to realise a shallow survey for long-period pulsars. About
${\sim}$
10% of the data collected so far have been processed in this mode, though it covers
${\sim}$
55% of the planned sky coverage. Only a small fraction (
$\lesssim$
3%) of the candidates have been examined for promising candidates to follow up, resulting in four pulsar discoveries, one of which is an independent discovery of a Green Bank Northern Celestial Cap (GBNCC) pulsar candidate, and another a re-discovery of a pulsar that was originally found in the Parkes 70 cm survey (Manchester et al. Reference Manchester1996) but reported with an incorrect name and DM. The processing has also led to a re-detection of 120 previously known pulsars, which brings the number of MWA pulsar detections to 180, a significant improvement over the initial census (at 185 MHz) reported in Xue et al. (Reference Xue2017). All four pulsars are being followed up with Parkes and the uGMRT, and also using archival and new data from the MWA, for detailed characterisation including their timing, variability, and early science, and also with the intent to develop and demonstrate effective follow-up strategies. Basic parameters of these pulsars are summarised in Table 2, and we elaborate on some specifics in the sections below.
4.1. Individual pulsar details
4.1.1. PSR J0036
$-$
1033
PSR J0036
$-$
1033 is the first new pulsar discovery from the SMART survey, the details and initial follow-up of which have been reported in Swainston et al. (Reference Swainston2021). This non-recycled pulsar with a period
$P=0.9$
s and a
${\rm{DM}}=23.123\,{\rm{pc cm}}^{-3}$
has a fairly typical magnetic field strength (
$ B \sim 4.4 \times 10^{11}$
G) and characteristic age (
$\tau _c \sim 67$
Myr) but has a spectral index
$\alpha = -2.0 \pm 0.2 $
. An estimated luminosity at 1400 MHz,
$L_{1400}$
${\sim}$
0.1
${\rm{mJy}}\, \rm{kpc}^{2}$
, places this pulsar in the lowermost two percentile of the currently known population of long-period pulsars in luminosity. Its DM and a location at a high Galactic latitude (
$b \approx -72.9^{\circ}$
) make it an excellent target for scintillation and variability studies for probing the nature of plasma turbulence in a region where it is poorly constrained. The related observations are currently under way at the uGMRT and the Five-hundred-metre Aperture Spherical Telescope (FAST) and will be reported in a future publication.
4.1.2. PSR J0026
$-$
1955
The ‘discovery’ of PSR J0026
$-$
1955 revealed some remarkable features of the SMART survey; most notably, the complexity of the tied-array beam pattern, resulting from a large fraction of the tiles configured as hexagonal layouts (see Paper I; Figures 7 and 8). This can result in multiple detections of pulsars through the near and far side (grating) lobes, especially when the pulsar is sufficiently bright. For instance, PSR J0026
$-$
1955, which was initially detected as a 11
$\sigma$
candidate (in a 10-min observation) was eventually re-detected in a different pointing, almost
${\sim}$
45
$^{\prime}$
offset from the initial position, and with a much higher significance of
$\rm{S/N} = 60$
, leading to the revelation that the initial detection was in fact through one of the grating lobes.
The pulsar was also particularly noted for its sub-pulse drifting nature which became the focus of immediate investigation. Incidentally, this turned out to be an independent discovery, as the pulsar was first reported as a candidate in the GBNCC survey data; nevertheless, it was blindly discovered in our SMART processing. Further details of the pulsar, including analysis of its intriguing sub-pulse drifting and long-duration nulls, are reported in McSweeney et al. (Reference McSweeney2022).
4.1.3. PSR J1002
$-$
2044
The discovery of PSR J1002
$-$
2044 followed the new (revised) scheme for candidate selection and scrutiny, as described in Section 3.3 of Paper I. It is a long-period pulsar with
$P=1.67$
s and a moderate
$\rm{DM} = 43.172$
${\textrm{pc cm}}^{-3}$
. The NE2001 model of electron density (Cordes & Lazio Reference Cordes and Lazio2002) places it at a distance of
${\sim}$
1.4 kpc. Its non-detection in the Parkes 70 cm survey implies a spectral index
$\alpha \lesssim -1.8 $
, and non-detections in the TGSS (TIFR GMRT Sky Survey) and NVSS (NRAO VLA Sky Survey) continuum imaging sky surveys, which reached rms sensitivity limits of
${\sim}$
5
${\rm{mJy}}\, \rm{beam}^{-1}$
and
${\sim}$
0.5
${\rm{mJy}}\, \rm{beam}^{-1}$
(in the pulsar vicinity) at their respective frequencies (i.e. 150 and 1400 MHz), implying a flux density at 150 MHz,
$ S_{150} \lesssim 10 $
${\rm{mJy}}$
. This hints at the possibility of the pulsar being another low-luminosity object, with an implied
$ L_{1400} \lesssim 0.3 $
${\rm{mJy}}\, \rm{kpc}^{2}$
, which is confirmed by our recent Parkes detection, suggesting
$ S_{1400} \sim 0.1$
mJy, and hence akin to PSR J0036
$-$
1033. The pulsar is currently being followed up for timing and localisation (via imaging) using Parkes and the uGMRT.
4.1.4. PSR J1357
$-$
2530
Of the 120 redetections from SMART data processing (see Section 4.2), PSR J1357
$-$
2530 warrants some additional contextual discussion. It was first ‘discovered’ in the untargetted SMART search workflow as a 6
$\sigma$
candidate in a 10-min observation, with a period (P) of 0.912 s and DM of 16.046
${\textrm{pc cm}}^{-3}$
(see Figure 2). Follow-up processing of the full 80-min observation resulted in a significantly improved detection, with a four-fold increase in S/N, as shown in the figure. This pulsar has a small duty cycle of
${\sim}$
2–3%, and the DM implies a distance of
${\sim}$
0.8 kpc as per the NE2001 model of Cordes & Lazio (Reference Cordes and Lazio2002). However, a closer examination of the published literature including those from low-frequency surveys with Parkes and the Green Bank Telescope (GBT), has essentially confirmed that this is a rediscovery. The Parkes 70 cm survey appears to have been reported this as PSR J1358
$-$
2533 with a similar period but with a significantly discrepant DM =
$28\pm3$
${\textrm{pc cm}}^{-3}$
and at
${\sim}$
20
$^{\prime}$
offset from our MWA position (Manchester et al. Reference Manchester1996). The DM was subsequently revised upward by D’Amico et al. (Reference D’Amico1998) to
$31.3 \pm 1$
${\textrm{pc cm}}^{-3}$
. It appears that GBNCC has re-found the pulsar but with a
$\rm{DM} = 16.0 \pm 0.2\,{\rm{pc cm}}^{-3}$
(McEwen et al. Reference McEwen2020). Even so, the position (and hence the pulsar’s exact name) remained uncertain and a timing solution is currently not available. The pulsar also appears to be another low-luminosity object, even if we are to assume a spectral index (
$\alpha$
) of
$-1.6$
.
In order to resolve this ambiguity, we recently undertook uGMRT follow-up observations of this pulsar (GTAC Cycle 43), where we used an observing set-up that is similar to those described in Section 3.1 (i.e. concurrent imaging and phased-array beams). Following an initial non-detection, which is attributable to a poorly localised position, imaging of the recorded visibility data revealed multiple (five) compact sources in the field within
${\sim}$
$5 ^{\prime}$
from the phase centre; one of these turned out to be the pulsar, as shown in the lower panels of Figure 3. This was confirmed in subsequent observations, where we cycled through the prospective sources by making separate phased-array beam observations. Table 5 gives a summary of the pulsar positions from our uGMRT imaging along with the initial MWA position (from the TAB localisation). The pulsar position, with R.A. 13
$^{\rm h}$
57
$^{\rm m}$
24
$^{\rm s}$
, and decl.
$-25^{\circ}30^{\prime}39^{\prime\prime}$
, while precisely determined down to
$\lesssim 0.3^{\prime\prime}$
, has a systematic offset of
${\sim}$
3
$^{\prime\prime}$
, possibly due to ionospheric refraction (cf. Swainston et al. Reference Swainston, Lee, McSweeney and Bhat2022a). The estimated flux densities are
${\sim}$
1 and
${\sim}$
0.5 mJy at 400 and 650 MHz, respectively. A summary of the pulsar parameters are included in Table 2. The new position warrants renaming the pulsar to PSR J1357
$-$
2530.
Table 5. Localisation summary of PSR J1357
$-$
2530.
Figure 6. All-sky distribution of known pulsars in the ATNF pulsar catalogue (grey filled circles). The 120 pulsars detected in the SMART processing are shown as red filled circles, along with 57 pulsars detected from the processing of non-SMART MWA observations (shown as filled circles in dark blue, respectively, for incoherent and coherent beam detections). The black star symbols are the three new pulsar discoveries from the SMART, and the grey star is re-discovery of a previously incorrectly characterised pulsar. The declination limit of the SMART survey (
$\delta < +30^{\circ}$
) is shown as the solid red line, whereas for surveys with other (northern) telescopes, they are shown as dashed lines at the respective lower limits in declination (i.e.,
$\delta > 0^{\circ}$
for LOTAAS,
$\delta > -40^{\circ}$
for GBNCC and
$ -40^{\circ} > \delta > -55^{\circ}$
for GHRSS).
Figure 7. Integrated pulse profiles for 120 re-detected pulsars in the SMART data processing. The period, DM, and the number of phase bins are shown in each panel. All detections were made in the SMART survey band 140–170 MHz.
4.2. Redetections of known pulsars
As described earlier, a secondary goal of the SMART survey is to map out the southern sky for low-frequency detections of known pulsars. This is important for two main reasons. Firstly, all past large pulsar surveys in the southern hemisphere have been at frequencies
$\gtrsim$
400 MHz (e.g., Manchester et al. Reference Manchester1978, Reference Manchester1996, Reference Manchester2001; Keith et al. Reference Keith2010). With the majority of the pulsars in the southern sky originally found and followed up at such high frequencies, low-frequency information (
$\lesssim $
300 MHz) is lacking for a substantial number of known pulsars, especially for those at declination,
$\delta \lesssim -30^{\circ} $
. Secondly, despite the MWA opening a new low-frequency window for pulsar astronomy in the south, the limitations of legacy VCS (e.g., a maximum recording capacity of 1.5 h)Footnote a and the logistics of handling large data rates prevented undertaking a complete pulsar census in an efficient manner. In an effort to partly address this, and as a preparatory work toward SKA-Low science, Xue et al. (Reference Xue2017) carried out the first pulsar census with the MWA at 185 MHz, using the then available resources and data sets (covering
${\sim}$
55% of the sky) and software and processing capabilities (i.e., incoherent detection, reaching
$\lesssim$
10% of the full coherent beam sensitivity). This led to the low-frequency detections of 50 pulsars (including 6 millisecond pulsars), 10 of which were detected for the first time at low frequencies. This modest sample was also used for simulation studies and forecasting the detectable population of pulsars with SKA-Low, as summarised by Xue et al. (Reference Xue2017).
A summary of re-detections from the SMART processing thus far is shown in Figure 6, where we have also shown those from non-SMART observations (i.e., archival data from other VCS-based projects). In total, there have been 120 re-detections from the SMART project; integrated profiles of these (at 155 MHz) are shown in Figure 7. Table 6 provides a summary of these detections, along with the pulsar parameters P and DM, and the signal-to-noise ratio of detection
$\rm (S/N)_{det}$
, listed as S/N in column 5 in Table 6, and column 6 in Table 7. Given the strong direction and elevation dependencies of the MWA’s sensitivity, we have accounted for the expected degradation in
$\rm (S/N)_{det}$
due to the pulsar’s offset within the primary beam (relative to the beam centre), while estimating the mean flux density at the observing frequency (listed in the column 6 of Table 6). As a result of the large extent of the MWA’s primary beam, these offsets can be as large as
${\sim}$
10–20° relative to the centre of the beam; for example, PSR J0034
$-$
0721, a bright long-period pulsar (and a well-known sub-pulse drifter; McSweeney et al. Reference McSweeney, Bhat and Tremblay2017), which would be detected with
$\rm S/N\sim 400$
in
${\sim}$
1 h observation, will likely be detected with
$\rm (S/N)_{det} \lesssim 100$
when observations are made at an elevation of
${\sim}$
30
$^ \circ $
and at
${\sim}$
10
$^ \circ $
offset from the beam centre.
Including 57 pulsars from non-SMART observations, 18 of which are from targeted observations toward PSR J1822-2256 (from the G0071 project; by Janagal et al.), this brings the current tally to 180 pulsars detected by the MWA. This is more than triple the number of pulsar detections with the MWA since the publication of Xue et al. (Reference Xue2017), but with the important distinction that many of these are full coherent beam detections, and the data can be reprocessed for full polarimetric information and/or higher time resolution (down to
$\lesssim$
1
$\unicode{x03BC} {\rm{s}}$
) depending on the scientific case (e.g., millisecond pulsars). Of the 177 pulsars shown in Figure 6 (and summarised in Figures 7 and 8, and Tables 6 and 7), the detection of 96 was facilitated by the SMART survey data, i.e. doubling of the number of pulsars detected with the MWA. For 34 pulsars, these are also the first low-frequency detections (
$\lesssim$
300 MHz); 30 of our pulsars have also been detected by the LOFAR (Bilous et al. Reference Bilous2016). Collectively, these will help extend the low-frequency coverage of pulsars, providing a useful sample for further detailed studies of pulsar emission and/or population analysis.
For many pulsars detected in SMART observations (Figure 7), there have been multiple detections, by virtue of our sky tessellation strategy (Figure 1), where the adjacent SMART pointings overlap
${\sim}$
10
$^ \circ $
in R.A. and
${\sim}$
12
$^ \circ $
in declination. In most such cases, we have selected the best detection. The number of phase bins
$N_{\rm{bin}}$
is chosen to optimise
$\rm (S/N)_{\rm det}$
while retaining the ability to discern main profile features, and varies from 64 to 1024. For MSPs with short periods (
$P \lesssim 5$
ms),
$ N_{\rm{bin}} \approx P_{\rm{ms}}/0.1 $
, where
$P_{\rm{ms}}$
is the period in ms.
4.2.1. Detection sensitivity and flux densities
The re-detections from SMART are used to ascertain the survey sensitivity at 150 MHz by comparing the measured pulsar flux densities
$S_{150}^{\rm{meas}}$
with their expected values
$S_{150}^{\rm{exp}}$
. To estimate
$S_{150}^{\rm{meas}}$
, folded profile data were flux density calibrated using the method outlined in Meyers et al. (Reference Meyers2017). In brief, this makes use of the array factor formalism whereby the tied-array beam patterns are simulated by modelling them as the product of the tile beam pattern (primary beam response) and the array factor (in this case for the compact configuration of the Phase II array). This analysis yields estimates of the system temperature (
${T_{\rm{sys}}}$
) and gain (G), which are then used to estimate the system equivalent flux density,
${\rm{SEFD}} = {T_{\rm{sys}}} / G$
. The simulations also naturally take into account the strong directional dependencies of these parameters. The variability in
${T_{\rm{sys}}}$
and G over the duration of observation is accounted for as a source of error in the flux density estimate.
The expected flux densities
$S_{150}^{\rm{exp}}$
are essentially based on the extrapolation from the spectral forms deduced using the published flux density measurements of pulsars. This involved the use of a large body of flux density measurements made available by Jankowski et al. (Reference Jankowski2018) as well as a number of other past publications. The model fits to the spectra are based on the algorithms described in the work of Jankowski et al. that employs the Akaike information criterion (AIC) to determine the best spectral model. Further details on the database repository and software description are presented in Swainston et al. (Reference Swainston, Lee, McSweeney and Bhat2022b). The uncertainties in
$S_{150}^{\rm{exp}}$
thus largely reflect the quality and uncertainty of the model fit to the spectra (e.g., the error on the spectral index, or other spectral characteristics). An important caveat is that the bulk of the flux density measurements are from observations at frequencies
$\gtrsim$
400 MHz and as a result may likely over-predict the expected values of
$S_{150}^{\rm{exp}}$
for pulsars that exhibit turnover, or a spectral break, or flattening at low frequencies.
A comparison of the measured flux densities of detected pulsars and the expected values as described above is shown in Figure 9. This includes 100 pulsars for which reasonable estimates of
$S_{150}^{\rm{exp}}$
were possible (i.e., four or more measurements available to allow meaningful estimates of the spectral form and the index
$\alpha$
), which limited the analysis to
${\sim}$
80% of the detected pulsars.
As shown in the figure, within the caveats of the uncertainties associated with
$S_{150}^{\rm{meas}}$
and
$S_{150}^{\rm{exp}}$
(e.g., from beam simulations, and the spectral model fits) and the assumptions involved (e.g., spectral extrapolation to 150 MHz), we find there is reasonable agreement, with the measured flux densities within a factor of two of their expected values. Evidently there are a small number of outliers where the analysis tends to over-predict
$S_{150}^{\rm{exp}}$
, and this may be to do with a spectral extrapolation error. In the case of some high DM pulsars, it is also possible there may be an underestimation of
$S_{150}^{\rm{meas}}$
, especially in the cases where the scatter broadening is substantial. The ratio between the measured and expected values is
$S_{150}^{\rm{meas}}$
/
$S_{150}^{\rm{exp}}$
=
$1.05 ^{+0.42} _{-0.31}$
(after excluding a small number of outlier points where the ratio
$>$
10 or
$<$
0.1). Given this, and with the caveat that in general large flux density variabilities expected at low frequencies (as much as by a factor of
${\sim}$
2–3 for moderate to high-DM pulsars, and even larger for low-DM ones), this provides a good demonstration that our SMART data processing is attaining a detection sensitivity in line with the expectations.
The measured flux densities of detected pulsars also validate the minimum expected sensitivity (
${S_{\rm{min}}}$
) for the shallow pass of our survey, where
${S _{\rm{min}}} \sim 10$
mJy for a 10
$\sigma$
detection of long-period pulsars (see Figure 9). As seen from the figure, the faintest pulsar detections are
${{S _{\rm 150} ^{\rm{meas}}}} \sim$
7–8
${\rm{mJy}}$
. In general, the measurements of low to moderate-DM pulsars are in better agreement compared to high-DM ones (i.e., DM
$\gtrsim$
50
${\textrm{pc cm}}^{-3}$
). Further details on the data used for the analysis, and the software description, are available in Swainston et al. (Reference Swainston, Lee, McSweeney and Bhat2022b).
4.2.2. Data release
Considering the inherent difficulties in accessing and processing VCS recorded data for pulsar detections and searching, we will make beamformed data of re-detected pulsars publicly accessible via a dedicated database repository (to be served by Data Central). We will adopt the preferred data format by other large pulsar databases (e.g., Hobbs et al. Reference Hobbs2011), i.e., folded archives in 10-s sub-integrations, with all 3072 channels and in full polarisation (i.e. four Stokes). For millisecond pulsars, we will provide coherently-dedispersed profiles in 1024 or more phase bins, in 24 coarse channels (1.28 MHz). Our intent is a continual growth of this database as further detections accrue from the ongoing and future data processing. Given the paucity of low-frequency pulsar detections in the southern sky, we hope this will become a useful resource, and potentially a reference catalogue of low-frequency pulsar detections in the SKA-Low era. Additionally, SMART survey raw data from 2018-2019 will be made available via a data retention project supported by Australian Research Data Commmons.Footnote b
Table 6. Known pulsars detected in SMART survey observations
aTarget offset from observation centre.
bLow-frequency (<400 MHz) information available in the literature: ‘M’ for MWA-incoh/MWA-image (Bell et al. Reference Bell2016; Murphy et al. Reference Murphy2017; Xue et al. Reference Xue2017; Kaur et al. Reference Kaur2019), ‘L’ for LOFAR (Kondratiev et al. Reference Kondratiev2016; Bilous et al. Reference Bilous2016; Sanidas et al. Reference Sanidas2019; Bondonneau et al. Reference Bondonneau2020), ‘W’ for LWA (Stovall et al. Reference Stovall2015), ‘G’ for GMRT (Bhattacharyya et al. Reference Bhattacharyya2016; Frail et al. Reference Frail, Jagannathan, Mooley and Intema2016), ‘g’ for GBT-GBNCC (McEwen et al. Reference McEwen2020), ‘E’ for Effelsberg (Sieber Reference Sieber1973), ‘P’ for Pushchino (Izvekova et al. Reference Izvekova, Kuzmin, Malofeev and Shitov1981; Malofeev, Malov, & Shchegoleva Reference Malofeev, Malov and Shchegoleva2000; Kuzmin & Losovsky Reference Kuzmin and Losovsky2001; Tyul’bashev et al. Reference Tyul’bashev, Tyul’bashev, Oreshko and Logvinenko2016), ‘U’ for Ukrainian T-shaped Radio telescope-UTR2 (Zakharenko et al. Reference Zakharenko2013).
5. Summary and conclusions
The SMART survey, enabled by the advent of the Phase II upgrade of the MWA, is an ambitious all-sky project to search for pulsars and fast transients at low frequencies. It exploits the large FoV and voltage capture functionality of the MWA, the combination of which provides a survey speed of
${\sim}$
450
${\rm deg ^ {2} \, h^{-1}}$
, but at the expense of large data rates of 28
${\rm{TB}}\, {\rm{h}^{-1}}$
, and consequently, substantial computational costs in beamforming and search processing. The execution of the survey relies on the availability of the compact configuration of the Phase II array. Through a series of four dedicated campaigns over the past few years, the data collection part of the survey is now
${\sim}$
75% completed.
In the first-pass processing as described in Paper I, where 10 min of data from each observation (out of 80 min) are processed (in 2358 trial DMs, out to a maximum DM of 250
${\textrm{pc cm}}^{-3}$
), effectively a shallow survey is performed, reaching a sensitivity about one-third of that will eventually be attainable with the planned deep-pass processing. Already three new pulsars have been found from the initial processing, where
${\sim}$
3% of the candidates have been scrutinised for promising ones. This initial processing has also led to re-detection of 120 previously known pulsars, thus bringing the total tally of MWA-detected pulsars to 180, i.e. more than a tripling of the number of pulsars detected with the MWA before the advent of the SMART survey.
The voltage recording strategy employed for the survey enables a multitude of avenues for follow-ups and confirmations, including improved detection and localisation to sub-arcminute position, thereby facilitating prompt follow-ups using sensitive telescopes such as Parkes and uGMRT that operate at
$\gtrsim$
300 MHz. Not only that this strategy is enabling an accelerated convergence to initial pulsar timing solution; e.g., by taking advantage of high-resolution (
${\sim}$
arcsecond) imaging with the uGMRT, archival VCS observations (where available) can also be potentially exploited to obtain a full coherent timing solution, as demonstrated in this paper.
The early discoveries and science from the SMART survey provide an excellent demonstration of the potential benefits of using a next-generation survey instrument such as the MWA, in tandem with current-generation instruments (e.g., Parkes, uGMRT) for fruitful scientific exploitation. For example, PSR J0036
$-$
1033, the first MWA-discovered pulsar turned out to be a low-luminosity object (Swainston et al. Reference Swainston2021), and a promising target for probing the high-Galactic latitude plasma turbulence. PSR J0026
$-$
1955, the second pulsar discovered, exhibits intriguing sub-pulse drifting patterns and has a large nulling fraction of
${\sim}$
77% (McSweeney et al. Reference McSweeney2022). Follow-up studies of these pulsars are in progress with the uGMRT and FAST. The third pulsar, PSR J1002
$-$
2044, another potential low-luminosity object, is currently being followed up using the uGMRT and Parkes, as well as with the MWA. Finally, for PSR J1357
$-$
2530, a pulsar originally reported in the literature with an incorrect name and DM, we have obtained a precise position via uGMRT follow-up to facilitate a high-resolution imaging and source identification.
The initial discoveries and results presented in this paper hint at the promising future for the SMART. As efforts around data processing and candidate scrutiny ramp up in the coming years, we anticipate the discovery rate to increase. The SMART survey is well positioned to emerge as a low-frequency version of the High Time Resolution Universe survey with Parkes, and as an important reference for even more ambitious surveys planned with the SKA-Low.
Figure 8. Integrated pulse profiles for 57 pulsars detected in observations of various non-SMART projects. The period, DM, and the number of phase bins are shown in each panel. Th majority of these detections were made in the 170–200 MHz band, while a few in the SMART survey band (140–170 MHz).
Table 7. Known pulsars detected with MWA observations not associated with SMART.
aTarget offset from observation centre.
bThe mode of beamforming used. I: incoherent, where the detected powers of each tile are summed. C: coherent/tied-array, where the voltages of each tile are summed (after appropriate phase rotations to the pointing-direction) and then detected. For incoherent detections, flux density estimates are using the method outlined in Xue et al. (Reference Xue2017), whereas for coherent beam detections, they are estimated using the method described in Section 4.2.
Figure 9. Measured flux densities
$S_{\rm{meas}}$
against expected flux densities
$S_{\rm{exp}}$
for 100 re-detections from the SMART survey processing; the measured values are corrected for the pulsar’s offset relative to the centre of the primary beam, and this can be as much as a factor of three. The expected values are extrapolations based on spectral fitting on the published flux density measurements of the pulsars (see text for details). The red line indicates where the two flux densities are equal whereas the pair of orange lines indicates a factor of two difference, relative to the red line.
Acknowledgement
We thank an anonymous referee for several useful comments that helped to improve the content and presentation of this paper. The scientific work made use of Inyarrimanha Ilgari Bundara, the CSIRO Murchison Radio-astronomy Observatory. We acknowledge the Wajarri Yamaji people as the traditional owners of the Observatory site. This work was supported by resources provided by the Pawsey Supercomputing Centre with funding from the Australian Government and the Government of Western Australia. This work was supported by resources awarded under Astronomy Australia Ltd’s ASTAC merit allocation scheme on the OzSTAR national facility at the Swinburne University of Technology. The OzSTAR program receives funding in part from the Astronomy National Collaborative Research Infrastructure Strategy (NCRIS) allocation provided by the Australian Government. The GMRT is run by the National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research, India. The Parkes radio telescope is part of the Australia Telescope National Facility which is funded by the Australian Government for operation as a National Facility managed by CSIRO.
Software: We acknowledge the use of the following software/packages for this work: CASA (McMullin et al. Reference McMullin, Waters, Schiebel, Young, Golap, Shaw, Hill and Bell2007), DSPSR (van Straten & Bailes Reference van Straten and Bailes2010, Reference van Straten and Bailes2011), PRESTO (Ransom Reference Ransom2001, Reference Ransom2011), PSRCHIVE (Hotan, van Straten, & Manchester Reference Hotan, van Straten and Manchester2004; van Straten et al. Reference van Straten2011), Tempo2 (Hobbs et al. Reference Hobbs, Edwards and Manchester2006; Hobbs & Edwards Reference Hobbs and Edwards2012), PINT (Luo et al. Reference Luo2019, Reference Luo2021), PsrPopPy (Bates et al. Reference Bates, Lorimer, Rane and Swiggum2014), Nextflow (Di Tommaso et al. Reference Di Tommaso2017).
Data Availability
It would be appropriate to include the Data Central link https://apps.datacentral.org.au/smart where we intend to publish all useful data products, and the ARDC link (see footnote Footnote b of this manuscript) https://doi.org./10.25917/qd25-pg50 through which raw data (from 2018 and 2019 observations) will be made available to the wider community.
Open access
