3.2.1 Calibration

The method used to calibrate MWA-VCS flux densities is described in Meyers et al. (Reference Meyers2017). To summarise, the system temperature ( $T_{\mathrm{sys}}$ ) is a combination of the antenna temperature ( $T_{\mathrm{ant}}$ ), the receiver temperature ( $T_{\mathrm{rec}}$ ), and the ambient temperature ( $T_{\mathrm{amb}}$ ):

(1) \begin{equation} T_{\mathrm{sys}} = \eta T_{\mathrm{ant}} + (1 - \eta) T_{\mathrm{amb}} + T_{\mathrm{rec}},\end{equation}

where $\eta\approx 1$ is the radiation efficiency of the array. To compute the antenna temperature, we perform an integral of the sky temperature weighted by the tied-array beam power over the sky. The sky temperature was determined by scaling the 408-MHz Haslam map (2014 reprocessed version; Haslam et al. Reference Haslam, Salter, Stoffel and Wilson1982; Remazeilles et al. Reference Remazeilles, Dickinson, Banday, Bigot-Sazy and Ghosh2015) to our observing frequencies assuming a spectral index of $-2.55$ (Guzmán et al. Reference Guzmán, May, Alvarez and Maeda2011). The tied-array beam power can be expressed as the product of the primary beam power and the array factorFootnote ^g power pattern. Given that the product being integrated depends on the primary beam power, we performed the integral for only the sky regions where the primary beam power exceeded 0.1% of the power at zenith.

The tied-array gain is defined as $G=A_{\mathrm{e}}/2k_{\mathrm{B}}$ , where $A_{\mathrm{e}}$ is the tied-array effective area and $k_{\mathrm{B}}$ is the Boltzmann constant. The effective area depends on the square of the observing wavelength ( $\lambda^2$ ) and the beam solid angle, which is an integral of the array factor power pattern over the whole sky.

The system equivalent flux density (SEFD) of the array was calculated as

(2) \begin{equation} \mathrm{SEFD} = f_{\mathrm{c}} \frac{T_{\mathrm{sys}}}{G},\end{equation}

where $f_{\mathrm{c}}$ is the coherency factor, which quantifies the coherent beamforming efficiency. The efficiency of the beam formation is affected by the accuracy of the antenna phasing, differences in ionospheric refraction between the calibration observation and the target observation, and many other factors (Ord et al. Reference Ord, Tremblay, McSweeney, Bhat, Sobey, Mitchell, Hancock and Kirsten2019). Rather than measuring $f_{\mathrm{c}}$ for each observation, we assigned a value of 0.7, and factored this assumption into the uncertainties. We computed the SEFD at four uniformly spaced time and frequency steps over the total observing time and frequency band (for a total of 16 samples per observation) and used the mean SEFD for calibrating the flux scale.

Figure 4. Distribution of census targets in the period-DM parameter space. The detected targets are shown as blue filled circles and the non-detected targets are shown as red pluses. As the initial census was incoherently dedispersed, we indicate where the intrachannel dispersive smearing ( $\textit{t}_{\text{DM}}$ ) is equal to the period and half of the period with dashed and dotted lines, respectively.

Figure 5. Comparison of MWA flux densities with LOFAR (left panel; from Kondratiev et al. Reference Kondratiev2016) and MeerKAT (right panel; from Spiewak et al. Reference Spiewak2022). Pulsars are labelled by right ascension. For the LOFAR comparison, we indicate the line of equal fluxes and the factor-of-two deviation band. For the MeerKAT comparison, we indicate the implied spectral index of a power-law model between the MWA and MeerKAT centre frequencies.

The data were then processed as follows. Each pulsar observation was first averaged over all sub-integrations and frequency channels, and the total intensity (Stokes I) pulse profile was extracted. The profiles were then normalised by the standard deviation of the off-pulse noise to convert them to $\mathrm{S}/\mathrm{N}$ units. We identified the off-pulse region by searching for the phase window in which the integrated flux density is minimal. For pulsars without a clear off-pulse region due to scattering, we measured the noise from a profile dedispersed to a DM of $500\,\mathrm{cm}^{-3}\,\mathrm{pc}$ , which is sufficient to fully disperse the signal over the pulse phase. We then measured the mean $\mathrm{S}/\mathrm{N}$ over the profile, $\langle \mathrm{S}/\mathrm{N} \rangle$ , and calculated the mean flux density ( $S_{\mathrm{mean}}$ ) using the radiometer equation:

(3) \begin{equation} S_{\mathrm{mean}} = \langle \mathrm{S}/\mathrm{N} \rangle \times \frac{\mathrm{SEFD}}{\sqrt{n_{\mathrm{p}}\Delta\nu \Delta t_{\mathrm{obs}} N_{\mathrm{b}}^{-1}}},\end{equation}

where $n_{\mathrm{p}}$ is the number of polarisations summed ( $n_{\mathrm{p}}=2$ for our data), $\Delta\nu$ is the bandwidth, $\Delta t_{\mathrm{obs}}$ is the integration time, and $N_{\mathrm{b}}$ is the number of phase binsFootnote ^h .

We assign conservative uncertainties to account for the following factors that influence the measured flux densities:

Beam coherence: Based on past measurements, we estimate that the coherency factor of will vary by no more than ${\sim}$ 20–30% from the nominal value of 0.7.

Galactic emission spectrum: The spectral index of the Galactic continuum emission is typically estimated to be between $-2.5$ and $-2.6$ away from the Galactic plane, and flattens due to thermal absorption to between $-2.2$ and $-2.5$ on the Galactic plane (Guzmán et al. Reference Guzmán, May, Alvarez and Maeda2011; Eastwood et al. Reference Eastwood2018). Spectral indices derived between the 159-MHz Engineering Development Array 2Footnote ⁱ (EDA2) sky map and the reprocessed 408-MHz Haslam sky map are generally consistent with these results (Kriele et al. Reference Kriele, Wayth, Bentum, Juswardy and Trott2022). Given our assumed spectral index of $-2.55$ , we expect an error of up to ${\sim}$ 5% of our extrapolated $T_{\mathrm{sky}}$ at high Galactic latitudes ( $|b|\gt 10^{\circ}$ ), and up to ${\sim}$ 30% on the Galactic plane ( $|b|\lt 10^{\circ}$ ). Given that the primary beam is much wider than the Galactic plane, the total error in the sky integral should always be lower than this.

Beam jitter: Ionospheric refraction can cause beam ‘jitter’ (i.e. dephasing) on timescales of ${\sim}$ hours when the angular scale of the source position offsets (between the calibrator and the target) are comparable to the half width at half maximum (HWHM) of the tied-array beam. Observations from 2013–2016 (around Solar cycle maximum) show that ionospheric offsets are typically $\mathbin{\sim}0.1$ – $0.2^\prime$ at the MRO (Waszewski, Morgan, & Jordan Reference Waszewski, Morgan and Jordan2022). On rare occasions, extreme ionospheric activity (e.g. travelling ionospheric disturbances) can cause ionospheric shifts of up to $\mathbin{\sim}1^\prime$ (Loi et al. Reference Loi2015; Jordan et al. Reference Jordan2017). The SMART observations were collected in 2018–2023 (at a comparable or lower Solar activity level to 2013–2016Footnote ^j ) in the compact configuration of the MWA, with an effective tied-array beam HWHM of $\mathbin{\sim}10$ – $15^\prime$ at $154\,\mathrm{MHz}$ . Therefore, even in extreme circumstances, we expect the beam power attenuation to be negligible ( $\lt1\%$ ). Furthermore, most ionospheric activity evolves on time scales longer than the duration of the SMART observations, so the calibration solutions are likely an accurate representation of the ionospheric state, even during active periods.

To account for the cumulative error of the above factors, we assign a flux density uncertainty of 30% to pulsars at $|b|\gt10^{\circ}$ , and 40% to pulsars at $|b|\lt10^{\circ}$ . Several other factors remain unaccounted because they are difficult or impossible to quantify. These are:

Beam simulation: The array factor calculation assumes an array of identical elements, which is not necessarily true for the MWA due to some tiles having one or more failing dipoles. Whilst we do not account for this, it is not likely to be a dominant error.

Interstellar scintillation: Low-DM pulsars ( ${\mathrm{DM}}\lesssim 10\,{\textrm{cm}}^{-3}\,\mathrm{pc}$ ) are susceptible to diffractive interstellar scintillation (DISS) with characteristic timescales of $\mathbin{\sim}10\,\mathrm{min}$ and bandwidths of $\mathbin{\sim} 1\,\textrm{MHz}$ (e.g. PSR J0437 $-$ 4715; Bhat et al. Reference Bhat, Ord, Tremblay, McSweeney and Tingay2016). DISS has been observed to cause flux density variations of ${\sim}$ 5–6 for J0437 $-$ 4715. Additionally, refractive interstellar scintillation (RISS) can cause variations on timescales of weeks to months. Quantifying DISS and RISS for each pulsar requires long-term monitoring on timescales much larger than the characteristic RISS timescale, which is beyond the scope of this census.

Interstellar scattering: For significantly scattered pulsars (where the scattering timescale is comparable to or exceeds the pulse period), our measurements may underestimate the pulsar flux density owing to the pulsar signal contaminating our estimate of the off-pulse noise. We note which flux densities are likely to be underestimated in Table 1.

3.2.2 Comparison with LOFAR and MeerKAT

We present mean flux densities for all of the detected MSPs in Table 1. Since our flux density calibration is based on a priori assumptions about the telescope and sky temperature, it is important to validate our measurements against independent measurements from other telescopes. In the left panel of Figure 5, we compare flux densities for the MSP detections in common with LOFAR. We see good agreement, with nearly all of the MWA flux densities comfortably falling within a factor of two of the LOFAR flux densities. There are three notable outliers: PSRs J0737 $-$ 3039A, B1937+21, and J2145 $-$ 0750. Both J0737 $-$ 3039A and B1937+21 experience significant broadening due to interstellar scattering below $200\,\mathrm{MHz}$ . Therefore, the measurements will have mismatched systematic errors due to the differing receiver bandwidths and contamination of the off-pulse noise estimate with the pulsar signal. Radio continuum images of B1937+21 indicate a flux density of $\mathbin{\sim}1$ – $2\,\mathrm{Jy}$ in the MWA frequency band (Erickson & Mahoney Reference Erickson and Mahoney1985; Kuniyoshi et al. Reference Kuniyoshi, Verbiest, Lee, Adebahr, Kramer and Noutsos2015), however these may be overestimated due to source confusion. Our flux density for J2145 $-$ 0750 is consistent with measurements from the Engineering Development Array (EDA) in the same frequency band, which are averaged over 9 observations spanning 100 days (Bhat et al. Reference Bhat2018). The discrepancy with LOFAR is most likely due to refractive interstellar scintillation.

In the right panel of Figure 5, we plot the MWA measurements against the MPTA census flux densities. We indicate on the plot the spectral index required to account for the difference in flux density between the centre frequencies of the two telescopes, assuming a simple power-law model. Due to the unaccounted uncertainties in our measurements (especially the intrinsic variability due to scintillation), the implied spectral indices are only an indicative value. We also caution that although MSP spectra are typically described adequately by a simple power-law, spectral breaks (with flattening at lower frequencies) have been well constrained in some cases (e.g. PSRs J0437 $-$ 4715, B1937+21, and J2145 $-$ 0750; Kuniyoshi et al. Reference Kuniyoshi, Verbiest, Lee, Adebahr, Kramer and Noutsos2015; Lee et al. Reference Lee, Bhat, Sokolowski, Swainston, Ung, Magro and Chiello2022). In these cases, the spectral index implied by our comparison may be too steep at lower frequencies and too shallow at higher frequencies. Regardless, the distribution of spectral indices we observe is consistent with previous MSP population studies, which find that the mean spectral index is around $-1.6$ to $-1.9$ (e.g. Toscano et al. Reference Toscano, Bailes, Manchester and Sandhu1998; Kramer et al. Reference Kramer, Lange, Lorimer, Backer, Xilouris, Jessner and Wielebinski1999; Swainston Reference Swainston2023; Karastergiou et al. Reference Karastergiou, Johnston, Posselt, Oswald, Kramer and Weltevrede2024).

The shallowest spectral index in the compared sample is for PSR J0437 $-$ 4715, being $-0.52^{+0.17}_{-0.13}$ . This is shallower than the value of $-0.83$ expected for this frequency range from spectral modelling of published flux densities (Lee et al. Reference Lee, Bhat, Sokolowski, Swainston, Ung, Magro and Chiello2022). The discrepancy can be accounted for by the fact that J0437 $-$ 4715 is known to exhibit strong diffractive scintillation at MWA frequencies (Bhat et al. Reference Bhat2014; Bhat et al. Reference Bhat2018), and the observation appears to have been taken when the pulsar was scintillating down. The steepest spectral index is for PSR J1536 $-$ 4948, being $-3.61^{+0.28}_{-0.21}$ . This is consistent with the steep in-band spectral index of $-3.2\pm 0.6$ from MeerKAT (Spiewak et al. Reference Spiewak2022). A more detailed spectral analysis using a larger set of flux densities compiled from the literature is deferred to a future publication.

3.3.1 RM synthesis

We can describe the complex linear polarisation in terms of the Stokes parameters I, Q, and U as

(4) \begin{equation} \boldsymbol{P} = \boldsymbol{p}I = Q + \mathrm{i}U,\end{equation}

where $\boldsymbol{p} \equiv \boldsymbol{P}/I$ . Written as a complex exponential, this is

(5) \begin{equation} \boldsymbol{P} = \lVert \boldsymbol{p} \rVert I{\mathrm{e}}^{2\mathrm{i}\psi},\end{equation}

where $\lVert \boldsymbol{p} \rVert$ is the degree of linear polarisation and $\psi$ is the position angle of the polarisation vector. The linearly polarised intensity, L, is obtained by adding Q and U in quadrature:

(6) \begin{equation} L \equiv \lVert \boldsymbol{p} \rVert I = \sqrt{Q^2 +U^2}.\end{equation}

Polarised signals propagating through magnetised plasma experience a rotation of the polarisation vector due to the Faraday effect. At an observing wavelength $\lambda$ , the measured polarisation angle is

(7) \begin{equation} \psi = \psi_0 + {\mathrm{RM}}\lambda^2,\end{equation}

where $\psi_0$ is the intrinsic polarisation angle and the rotation measure (RM) is the rate of change of the polarisation angle as a function of $\lambda^2$ , usually in units of $\mathrm{rad\,m}^{-2}$ . Physically, the RM is a measure of the integrated magnetic field projected along the line of sight, weighted by the number density of free electrons. Equation (7) is valid assuming that the linear polarisation originates from a single source with no internal Faraday rotation. It is typically assumed that Faraday rotation within the pulsar magnetosphere is negligible, which is theoretically supported under specific magnetospheric conditions (Wang, Han, & Lai Reference Wang, Han and Lai2011). Therefore, unambiguous evidence for magnetospheric Faraday rotation (e.g. RM variations as a function of pulse phase or deviations from linearity in $\lambda^2$ ) can provide valuable constraints on magnetospheric models. As such, it is important to consider these effects when measuring the RM from pulsar observations.

The simplest way to measure the RM is to fit Equation (7) to the position angle as a function of $\lambda^2$ . However, we instead chose to use the technique of RM synthesis developed by Burn (Reference Burn1966) and Brentjens & de Bruyn (Reference Brentjens and de Bruyn2005), which has been widely adopted due to its superior sensitivity when analysing low- $\mathrm{S}/\mathrm{N}$ data and its ability to separate instrumental polarisation from astrophysical signals. Following Burn (Reference Burn1966), we consider a generalisation of the RM, known as the Faraday depth ( $\phi$ ), which is defined at every point along the line of sight to the source. Assuming $\psi_0$ is the same at all Faraday depths, we can define the Faraday dispersion function (FDF), which represents the intrinsic linear polarisation at each Faraday depth integrated over all $\lambda^2$ :

(8) \begin{equation} \boldsymbol{F}(\phi) = \int_{-\infty}^{+\infty} \boldsymbol{P}(\lambda^2)\, {\mathrm{e}}^{-2\mathrm{i}\phi\lambda^2}\,\mathrm{d}\lambda^2.\end{equation}

If the polarised signal is negligibly Faraday rotated by the pulsar magnetosphere, then it will appear as a compact source in the FDF, with the amplitude peaking at the RM defined by Equation (7). Hence, any excess signal in the FDF other than the RM peak implies some deviation from the $\lambda^2$ relation (which may be astrophysical or instrumental in nature). In particular, weakly chromatic or achromatic instrumental effects can appear as a ‘zero peak’ in the FDF (i.e. a peak at $\phi \sim 0\,\textrm{rad m}^{-2}$ ). If the RM peak can be resolved from the zero peak, then the RM can be measured without contamination from the instrumental signal.

As we can only observe $\boldsymbol{P}(\lambda^2)$ at discrete and positive values of $\lambda^2$ , the practical implementation is a discrete Fourier transform,

(9) \begin{equation} \tilde{\boldsymbol{F}}(\phi) \approx \frac{1}{N_{\mathrm{c}}} \sum_{c=1}^{N_{\mathrm{c}}} \boldsymbol{P} (\lambda^2_c)\, {\mathrm{e}}^{-2\mathrm{i}\phi (\lambda_c^2-\lambda_0^2)},\end{equation}

where c is the frequency channel index, $N_{\mathrm{c}}$ is the number of channels, $\lambda_c$ is the centre wavelength of channel c, and $\lambda_0$ is a reference wavelength to which all polarisation vectors are de-Faraday rotated (Brentjens & de Bruyn Reference Brentjens and de Bruyn2005). For this work, we are only concerned with the amplitude of the FDF, so the choice of $\lambda_0$ is arbitrary.

As shown by Brentjens & de Bruyn (Reference Brentjens and de Bruyn2005), the reconstructed FDF can be described as the convolution of the true FDF with the RM spread function (RMSF),

(10) \begin{equation} \boldsymbol{R}(\phi) \approx \frac{1}{N_{\mathrm{c}}} \sum_{c=1}^{N_{\mathrm{c}}} {\mathrm{e}}^{-2\mathrm{i}\phi (\lambda_c^2-\lambda_0^2)}.\end{equation}

The RMSF can be deconvolved from the reconstructed FDF using the rm-clean algorithm (Heald, Braun & Edmonds Reference Heald, Braun and Edmonds2009). This reduces sidelobe confusion, which can help to distinguish the zero peak from the RM peak. The rm-clean model components can also help to identify sources of polarisation in the FDF. The effective resolution of the FDF is essentially limited by the width of the main peak of the RMSF, which we quantify with the full width at half maximum (FWHM):

(11) \begin{equation} \delta\phi\ \left[\mathrm{rad\,m}^{-2}\right] \approx \frac{3.8}{\Delta\lambda^2},\end{equation}

where $\Delta\lambda^2$ is the total span of the data in $\lambda^2$ space in $\mathrm{m}^{-2}$ (Schnitzeler, Katgert, & de Bruyn Reference Schnitzeler, Katgert and de Bruyn2009). For the SMART frequency band, $\delta\phi \sim 2.50\,\textrm{rad m}^{-2}$ .

3.3.2 RM estimation and uncertainties

As described by Brentjens & de Bruyn (Reference Brentjens and de Bruyn2005), the uncertainty in the RM can be estimated analytically. However, this method relies on assumptions about the shape of the channel bandpass that often do not hold for real data (Schnitzeler & Lee Reference Schnitzeler and Lee2015). It can also be difficult to estimate a reliable analytic uncertainty for noisy data. We therefore adopt a bootstrapping approach, similar to the method used by Ilie, Johnston, & Weltevrede (Reference Ilie, Johnston and Weltevrede2019), as follows. White noise was added to Stokes Q and U with a standard deviation determined by taking the median (over all off-pulse phase bins) of the phase-resolved standard deviations estimated over all frequency channels. RM synthesis was then performed on the simulated data and the RM was measured by fitting a parabola to the peak of the FDFFootnote ^k and interpolating to find the maximum. This was repeated for a large number of iterations ( $\geq 10^4$ ) to create a Gaussian distribution of simulated RM measurements. The RM measurement and statistical uncertainty were taken as the mean and standard deviation of the bootstrap distribution, respectively. For some low- $\mathrm{S}/\mathrm{N}$ detections, the bootstrap distribution was multimodal due to the noise in the FDF having a comparable amplitude to the RM peak. In these cases, if the majority of samples converged on a single RM, then we discarded the outlier samples before measuring the distribution.

3.3.3 Ionospheric RM correction

The observed RM is the sum of contributions from the IISM and the terrestrial ionosphere:

(12) \begin{equation} {\mathrm{RM}}_{\mathrm{obs}} = {\mathrm{RM}}_{\mathrm{IISM}} + {\mathrm{RM}}_{{\mathrm{iono}}}.\end{equation}

To estimate the ionospheric RM, we used spinifex (Mevius, Thomson, & Dijkema Reference Mevius, Thomson and Dijkema2025). The total electron content (TEC) was estimated using a single-layer ionospheric model with global ionosphere maps provided by the Jet Propulsion Laboratory (JPL; Martire et al. Reference Martire2024). The JPL maps are created using measurements from the International GNSS Service ground station network,Footnote ^l including one station at the MRO. spinifex performs spatial and temporal interpolation of the TEC, and calculates uncertainties using the root-mean-square ionosphere maps. Since spinifex is essentially a re-write of rmextract (Mevius Reference Mevius2018), this method is practically equivalent to the one validated by Lee et al. (Reference Lee, Bhat, Sokolowski, Meyers and Magro2024) using pulsar observations at the MRO. To estimate ${\mathrm{RM}}_{\mathrm{iono}}$ , we calculated the mean of 10 time steps spanning the duration of each pulsar observation. An indicative statistical uncertainty was estimated by bootstrapping ${\mathrm{RM}}_{\mathrm{iono}}$ and calculating the standard deviation; this was then doubled to bring the mean ionospheric uncertainty of our dataset to ${\sim}0.1\,\textrm{rad m}^{-2}$ , which is more reflective of the accuracy observed by Lee et al. (Reference Lee, Bhat, Sokolowski, Meyers and Magro2024). We then subtracted ${\mathrm{RM}}_{\mathrm{iono}}$ from the observed RM to obtain the estimated contribution from the IISM.

3.3.4 Profile-averaged RM

Following Ilie et al. (Reference Ilie, Johnston and Weltevrede2019), we determined a profile-averaged RM by summing together the amplitudes of all the on-pulse FDFs – i.e.

(13) \begin{equation} \tilde{F}_{\mathrm{total}}(\phi) = \sum_{\theta\in\Theta} \lVert \tilde{\boldsymbol{F}}_{\theta}(\phi) \rVert,\end{equation}

where $\Theta$ is the set of all on-pulse phase bins – and measuring the RM from the peak of $\tilde{F}_{\mathrm{total}}(\phi)$ . The on-pulse region was identified by searching for the phase window in which the integrated flux density was maximal. The size of the phase window was chosen by eye to match the pulse width due to the large variation in pulse shapes and $\mathrm{S}/\mathrm{N}$ . The uncertainty was estimated by bootstrapping the profile-averaged RM following the method described in Section 3.3.2.

We obtained significant profile-averaged RM measurements for 25 MSPs, which are listed in Table 2 along with the estimated contributions from the ionosphere and IISM. For comparison, we list the RM measurements from MeerKAT (Spiewak et al. Reference Spiewak2022) and Murriyang/Parkes (Dai et al. Reference Dai2015), as well as the z-score of the reference measurements (from the literature) relative to the MWA measurements; i.e.

(14) \begin{equation} z_{\mathrm{RM}} = \frac{{\mathrm{RM}}_{\mathrm{ref}} - {\mathrm{RM}}_{\mathrm{MWA}}}{\sqrt{\sigma_{\mathrm{ref}}^2 + \sigma_{\mathrm{MWA}}^2}},\end{equation}

where ${\mathrm{RM}}_{\mathrm{ref}} \pm \sigma_{\mathrm{ref}}$ are the reference measurements, and ${\mathrm{RM}}_{\mathrm{MWA}} \pm \sigma_{\mathrm{MWA}}$ are measurements from this work. Figure 6 shows a visual comparison of our measurements with the literature. In general, our measurements are in good agreement with both MeerKAT and Parkes. Of the 23 MSPs with both MWA and MeerKAT measurements, 17 are consistent to within $3\sigma$ and only three exceed $5\sigma$ (PSRs J0125 $-$ 2327, J0437 $-$ 4715, and J2129 $-$ 5721). Of the 10 MSPs with both MWA and Parkes measurements, 5 are consistent to within $3\sigma$ and two exceed $5\sigma$ (PSRs J1730 $-$ 2304 and J2129 $-$ 5721). The noted discrepancies with MeerKAT and Parkes may be a result of different measurement methods. For example, Dai et al. (Reference Dai2015) measured the RM for selected pulse phase regions in which the position angles were stable across the Parkes receiver bands. This may also be the reason for the discrepancies between MeerKAT and Parkes; for J1730 $-$ 2304, we observe a $7.48\sigma$ discrepancy with Parkes, but only a $0.54\sigma$ discrepancy with MeerKAT. Overall, we find that the precision of our measurements is consistently comparable (and in some cases better) than the MeerKAT measurements (which are averages over $\gt6$ observations) and the Parkes measurements (which are fits over the full frequency range of the Parkes dual-band receiver). Notably, PSRs J1038+0032 and J1400 $-$ 1431 are depolarised at higher frequencies, and as a result the MWA measurements are more than an order of magnitude more precise than MeerKAT.

Table 2. RM measurements for 25 MSPs detected in the census. For comparison, we list RM measurements from MeerKAT (Spiewak et al. Reference Spiewak2022, labelled ‘sbm22’) and Murriyang/Parkes (Dai et al. Reference Dai2015, labelled ‘dhm15’), as well as the $\textit{z}$ -score of the reference measurements relative to the MWA measurements, $\textit{z}_{\text{RM}}$ (see Equation 14). Observed RMs which may be contaminated by instrumental polarisation (i.e. $\left|{\text{RM}}_{\text{obs}}^{\text{MWA}}\right| \lt \delta\phi/\text{2}$ ), are indicated with an asterisk.

Figure 6. Comparison of the RM measurements from this work ( ${\text{RM}}_{\text{MWA}}$ ) and published RM measurements from the literature ( ${\text{RM}}_{\text{ref}}$ ). The red squares are Parkes measurements from Dai et al. (Reference Dai2015) and the black circles are MeerKAT measurements from Spiewak et al. (Reference Spiewak2022). The shaded grey band shows the FWHM of the RMSF centred at $\text{0}\,\text{rad}\,\text{m}^{-\text{2}}$ ; MWA measurements within this range could potentially be contaminated with instrumental polarisation. For visual clarity, we show the measurement for PSR J0737 $-$ 3039A in an inset. In the top panel, we show the absolute value of $\textit{z}_{\text{RM}}$ (see Equation 14).

Two MSPs do not have RM measurements from MeerKAT or Parkes: PSRs J2051 $-$ 0827 and J2256 $-$ 1024. For J2051 $-$ 0827, the pulsar catalogue reports a value of $-46(10)\,\mathrm{rad\,m}^{2}$ from Han et al. (Reference Han, Manchester, van Straten and Demorest2018). Our measurement of $-31.74(15)\,\mathrm{rad\,m}^{2}$ is consistent within $1.5\sigma$ of the catalogued measurement and improves the precision by two orders of magnitude. For J2256 $-$ 1024, Crowter et al. (Reference Crowter2020) report observed RM measurements (i.e. without ionospheric corrections) for three frequency bands: 350, 820, and $1\,500\,\mathrm{MHz}$ . Our measurement of $-14.00(12)\,\mathrm{rad\,m}^{2}$ is consistent to within $1.2\sigma$ and $0.3\sigma$ of the Crowter et al. measurements at 820 and $1\,500\,\mathrm{MHz}$ , respectively. The 350-MHz measurement is higher than ours by $8\sigma$ ; however, this could reasonably be attributed to the uncorrected ionospheric contribution.

3.3.5 Phase-resolved RM

The apparent RM can vary significantly with pulse phase, which has been attributed to interstellar scattering, profile evolution with frequency, and magnetospheric effects (e.g. Noutsos et al. Reference Noutsos, Karastergiou, Kramer, Johnston and Stappers2009; Noutsos et al. Reference Noutsos2015; Dai et al. Reference Dai2015; Ilie et al. Reference Ilie, Johnston and Weltevrede2019; Xu et al. Reference Xu2025). We therefore measured the RM for the individual phase bins within the on-pulse region of each MSP. We considered an RM measurement significant if the $1\sigma$ confidence interval (calculated via bootstrapping) was smaller than the FWHM of the RMSF, and the linear polarisation exceeded $5\sigma_I$ (where $\sigma_I$ is the standard deviation of the off-pulse noise). To quantify the variation in RM, we calculated the maximum statistical difference between any two RM measurements over the profile. The phase-resolved RM measurements and pulse profiles for the four MSPs with the most significant variations are shown in Figure 7.

Following the analysis by Noutsos et al. (Reference Noutsos, Karastergiou, Kramer, Johnston and Stappers2009), we also measured the change in fractional circular polarisation over the observing band, $\Delta(V/I)$ , by fitting a linear model $V/I\propto\nu$ to the spectral data for each phase bin. Correlations between $\Delta(V/I)$ and V as a function of pulse phase can be an indication of interstellar scattering (see also Ilie et al. Reference Ilie, Johnston and Weltevrede2019). Additionally, propagation through relativistic plasma in the pulsar magnetosphere can cause a frequency-dependent conversion from linear into circular polarisation, known as Faraday conversion (Kennett & Melrose Reference Kennett and Melrose1998). The effect of Faraday conversion on the position angle depends on the physical environment, and is therefore commonly modelled using a generalised RM (GRM); i.e.

(15) \begin{equation} \psi = \psi_0 + \mathrm{GRM}\lambda^\alpha,\end{equation}

where $\alpha$ is the wavelength dependence. Lower et al. (Reference Lower2024) demonstrated that incorrectly modelled Faraday conversion (e.g. assuming $\psi\propto\lambda^2$ ) can produce apparent phase-dependent RM variations. A detailed search for Faraday conversion (as in Lower et al. Reference Lower2024) is beyond the scope of this work. However, we note that a correlation between $\Delta(V/I)$ and the RM could be an indication of Faraday conversion.

We observe the largest apparent RM variations for PSR J2145 $-$ 0750 (see Figure 7). The apparent RM is stable at ${\sim}{-}2.2\,\mathrm{rad\,m}^{-2}$ for most of the profile, but jumps to ${\sim}0.5\,\mathrm{rad\,m}^{-2}$ near the centre of the trailing profile component. The phase-resolved FDF (see Figure 8, right) shows that the polarisation is distributed between peaks at Faraday depths of ${\sim}0.5$ and ${\sim}{-}2.2\,\mathrm{rad\,m}^{-2}$ . It is particularly notable that between pulse longitudes of $30^{\circ}$ and $45^{\circ}$ , the polarisation is simultaneously present at both ${\sim}0.5$ and ${\sim}{-}2.2\,\mathrm{rad\,m}^{-2}$ (as indicated by the rm-clean model components in Figure 8). This strongly suggests that the peak closer to ${\sim}0\,\mathrm{rad\,m}^{-2}$ is instrumental in origin. To verify this hypothesis, we performed a jackknife cross-validation test by dividing the observation into four separate 15 min segments and re-performing the analysis on each segment. As expected, the polarisation near ${\sim}0.5\,\mathrm{rad\,m}^{-2}$ varies significantly in intensity over time, while the peak near ${\sim}{-}2.2\,\mathrm{rad\,m}^{-2}$ remains stable. We therefore conclude that the apparent variation is instrumental, and the true pulsar RM does not vary significantly with phase.

PSR J0437 $-$ 4715 shows RM variations of ${\sim}1\,\mathrm{rad\,m}^{-2}$ around the steepest part of the position angle curve (see Figure 7). In this case, the peak in the FDF appears to shift with phase (see Figure 8, left). Additionally, the rm-clean model components indicate that – unlike for J2145 $-$ 0750 – the FDF is accurately described by a single point source convolved with the RMSF at all phases. This suggests that there is no significant achromatic instrumental polarisation. As above, we performed a jackknife test by dividing the observation into two 15 min segments; in this case, we do not see any significant variations over time. This is again consistent with there being no significant instrumental contamination in the signal. Since this pulsar has a very low DM ( $2.64\,\textrm{cm}^{-3}\,\mathrm{pc}$ ), the estimated scattering timescale is much smaller than the bin width, which effectively rules out scattering as the cause of the variations. Apparent RM variations can also be induced when the amplitudes of overlapping emission components evolve differently with frequency. We investigated the profile evolution by averaging the frequency channels into three subbands and comparing the profiles; no significant profile shape variations were observed in the phase range of interest. We therefore cannot attribute the variations to any known systematic effect. We also do not measure any notable correlation between $\Delta(V/I)$ and the change in RM, so our analysis does not show any evidence of Faraday conversion. Therefore, the origin of the variations is as-yet unclear. Observing at lower frequencies or over a larger frequency range would help to rule out weakly-chromatic instrumental effects by reducing potential confusion in the FDF. A larger fractional bandwidth would also improve the sensitivity to changes in $\Delta(V/I)$ . These requirements could be met by using non-contiguous frequency configurations (e.g. Kaur et al. Reference Kaur, Ramesh Bhat, Dai, McSweeney, Shannon, Kudale and van Straten2022).

Considering only the pulsars with unambiguous RM measurements (i.e. ${\mathrm{RM}}\gg \delta\phi/2$ ), we see notable variations for PSRs J2241 $-$ 5236 and J2256 $-$ 1024. For both pulsars, the phase-dependent RM variations are correlated with the rate of change of the position angle. For J2241 $-$ 5236, there is also a clear correlation between $\Delta(V/I)$ and the rate of change of circular polarisation. These results are consistent with interstellar scattering. The estimated scattering timescales for these pulsars are ${\sim}10$ – $20\,\unicode{x03BC}\textrm{s}$ at $154\,\mathrm{MHz}$ (given the DMs of 11.4 and $13.8\,\mathrm{cm^{-3}}\,\mathrm{pc}$ ; Bhat et al. Reference Bhat, Cordes, Camilo, Nice and Lorimer2004), compared to a bin width of $3.125\,\unicode{x03BC}\textrm{s}$ (i.e. ${\sim}$ 3–6 bins). This is consistent with the temporal scale of the RM variations observed in Figure 7.

It is interesting to note that PSRs J0437 $-$ 4715, J2145 $-$ 0750, and J2241 $-$ 5236 all show larger RM variations in the Parkes data from Dai et al. (Reference Dai2015) than in the MWA data. Noutsos et al. (Reference Noutsos2015) showed that smaller apparent RM variations at low frequencies are a natural result of interstellar scattering. However, since both J0437 $-$ 4715 and J2145 $-$ 0750 have scattering timescales smaller than the bin width, this cannot explain the inconsistency with Parkes for these pulsars.

3.4.1 Polarimetry

For each of the detections, we generated full-polarimetric integrated pulse profiles. We followed the methods described by Spiewak et al. (Reference Spiewak2022), based on the methods from Everett & Weisberg (Reference Everett and Weisberg2001), to de-bias the linear polarisation measurements and to compute the uncertainties in the position angle measurements. For detections with measured RMs, we removed the Faraday rotation using the pam utility in psrchive.

Figure 7. Polarimetric integrated pulse profiles for MSPs showing the most significant apparent phase-dependent RM variations. The bottom panels show the total intensity ( $\textit{I}$ ; black solid line), linear polarisation ( $\textit{L}$ ; red dashed line), and circular polarisation ( $\textit{V}$ ; blue dotted line). The top panels show the change in circular polarisation over the observing band for bins with $\textit{I}\gt\text{10}\sigma_{\textit{I}}$ , and the RM and position angle (P.A.) for bins with $\textit{L}\gt\text{5}\sigma_{\textit{I}}$ . The red dashed lines and shaded RM ranges indicate the profile-averaged RMs and 1- $\sigma$ uncertainties as described in Section 3.3.4. The black dashed lines and shaded regions indicate the location and FWHM of the RMSF centred at $\text{0}\,\text{rad}\,\text{m}^{-\text{2}}$ (i.e. the zero peak).

Figure 8. Phase-resolved and profile-averaged FDF amplitudes for PSRs J0437 $-$ 4715 (left) and J2145 $-$ 0750 (right). The top panels show the polarimetric integrated pulse profiles (as in Figure 7). The bottom left panels show the amplitude of the FDFs as a function of Faraday depth ( $\phi$ ) and pulse longitude. The right panels show the average amplitudes of the on-pulse FDFs: the red dashed lines and shaded ranges indicate the RMs measured from the profile-averaged FDF amplitudes and their 1- $\sigma$ uncertainties; and the black dashed lines and shaded ranges indicate the location and FWHM of the RMSF at $\text{0}\,\text{rad}\,\text{m}^{-\text{2}}$ (i.e. the zero peak). rm-clean was applied to each phase-resolved FDF using a S/N threshold of 3; the rm-clean model components are shown as red dots.

In Figure 9, we show a comparison between the MWA and LOFAR polarimetric pulse profiles (from Noutsos et al. Reference Noutsos2015) for PSRs J1022+1001 and B1257+12. The profiles were aligned by minimising the standard deviation of the total intensity profile residuals. As neither data were calibrated for absolute polarisation, the mean residual position angle is arbitrary. For J1022+1001, there are negligible differences in the total and polarised intensities, but some systematic variations in the position angle residuals where the position angle is changing most rapidly with phase. This may be due to frequency evolution of the position angle over the much larger LOFAR observing band. For B1257+12, the position angles and polarised intensities are consistent, but there is a difference in the total intensity profiles at the leading edge which is likely due to frequency evolution of the pulse shape. This is the first published verification of MWA-VCS polarimetry using an independent instrument at the same frequency.

Figure 9. Comparison between MWA and LOFAR polarimetry for PSRs J1022+1001 (left) and B1257+12 (right). The LOFAR data are from Noutsos et al. (Reference Noutsos2015) and span a larger bandwidth, but have similar centre frequencies to the MWA data. The top two panels show the position angles (P.A.) and the residuals ( $\Delta$ P.A.) with the mean subtracted (the mean residual is arbitrary). The bottom three panels show the polarimetric integrated pulse profiles (the colours are the same as in Figure 7) and the residuals in intensity.

3.4.2 Frequency evolution

Characterising the frequency evolution of MSP pulse profiles is important for both pulsar timing and understanding the pulsar emission mechanism. In this section, we provide a qualitative comparison of the MWA pulse profiles to archival profiles at other frequencies. We compiled profiles from the MPTA census data release (Spiewak et al. Reference Spiewak2022) and the European Pulsar Network Database of Pulsar Profiles.Footnote ^m For PSR J2256 $-$ 1024, we used the coherently-dedispersed GBT profiles from Crowter et al. (Reference Crowter2020). Since timing information for the archival profiles was not available, the profiles were aligned by eye based on our own interpretation of the profile features, particularly using components common across multiple frequencies. Figure 10 shows pulse stacks for 20 selected MSPs; we particularly focus on those with interesting frequency evolutions and/or high-quality MWA profiles. In the following, we describe the salient features and frequency evolution of the profiles for each of the selected MSPs.

Figure 10. Frequency evolution of integrated pulse profiles for selected MSPs. MWA profiles are from this work. Profiles from other telescopes are from the following publications: MeerKAT – Spiewak et al. (Reference Spiewak2022); Murriyang/Parkes – Manchester & Johnston (Reference Manchester and Johnston1995), Bailes et al. (Reference Bailes1997), and Dai et al. (Reference Dai2015); Effelsberg – Kijak et al. (Reference Kijak, Kramer, Wielebinski and Jessner1997) and Kramer et al. (Reference Kramer, Xilouris, Lorimer, Doroshenko, Jessner, Wielebinski, Wolszczan and Camilo1998), Lovell – Stairs, Thorsett, & Camilo (Reference Stairs, Thorsett and Camilo1999); LPA – Kuz’min & Losovskii (Reference Kuz’min and Losovskii1999); GBT – Crowter et al. (Reference Crowter2020); Arecibo – Camilo, Nice, & Taylor (Reference Camilo, Nice and Taylor1993) and Wahl et al. (Reference Wahl, Rankin, Venkataraman and Olszanski2023).

PSR J0030+0451. The MWA profile suggests that both the leading and trailing components of the main pulse in the MeerKAT profile weaken at low frequencies. This leads to a simpler, single-peaked main pulse. The interpulse is also much weaker (though still detectable) in the MWA profile.

PSR J0125 $-$ 2327. Although the $\mathrm{S}/\mathrm{N}$ is low in the MWA profile, it is clear that the trailing component is significantly weaker than in the MeerKAT profile. The components also appear to be more separated in the MWA profile, leading to a larger profile width. The MWA profile also shows a higher degree of linear polarisation.

PSR J0407+1607. The MWA profile shows a new leading component that is not seen in the MeerKAT profile. Since profile component separations are known to change very slowly with frequency for MSPs, the two components in the MWA profile are not likely to be a bifurcation of the MeerKAT profile (as one may expect for a non-recycled pulsar). Rather, this appears to be an entirely new component which has been enhanced at low frequencies.

PSR J0437 $-$ 4715. The dramatic frequency evolution of this pulsar has previously been described by Bhat et al. (Reference Bhat2018). The spacing and relative amplitude of the outer profile components evolve slowly with frequency, whilst the central peak appears to weaken and shift towards the trailing edge of the profile. Interestingly, the inversion in circular polarisation near the centre of the profile, which is observed at higher frequencies, is not seen by the MWA.

PSR J0737 $-$ 3039A. This is the MSP from the famous double pulsar system, which is a high-priority target for relativistic binary timing programs (e.g. Kramer et al. Reference Kramer2021b). This pulsar is significantly scattered at MWA frequencies, and as such a detailed characterisation of the scattering properties will be presented in a separate publication. We note here that the relative amplitude of the main pulse and interpulse changes significantly between the MeerKAT and MWA profiles, implying different spectral indices.

PSR J1022+1001. This pulsar has a particularly interesting profile evolution in which the trailing component is prominent between $\mathbin{\sim}400\,\mathrm{MHz}$ – $2\,\mathrm{GHz}$ and weakens at higher and lower frequencies. We aligned the MWA profile with the leading component, consistent with the interpretation of Noutsos et al. (Reference Noutsos2015). This alignment implies that the trailing component is completely missing in the MWA frequency band.

PSR J1024 $-$ 0719. The frequency evolution for this pulsar is perplexing. There is a clear enhancement of the trailing component in the Parkes profile at $436\,\mathrm{MHz}$ compared to higher frequency profiles. However, the MWA (and LOFAR) profile shows only a single component. The LPA profile at $103\,\mathrm{MHz}$ then hints at two components (although it should be cautioned that this is an incoherently dedispersed profile which may be broadened due to intrachannel smearing). There is therefore an ambiguity as to whether the component visible in the MWA profile corresponds to the leading or trailing component of the Parkes and LPA profiles. The linear polarisation provides a clue: the MWA profile is linearly polarised towards the leading edge, similarly to the higher-frequency profiles. This implies that the MWA profile should be aligned with the leading edge of the other profiles, contrary to what was assumed by Kondratiev et al. (Reference Kondratiev2016). This alignment is more consistent with the LPA profile, as should be expected given the relatively small difference in frequency. The overall trend is therefore similar to PSR J1022+1001: the trailing component is enhanced between $\mathbin{\sim}400\,\mathrm{MHz}$ – $2\,\mathrm{GHz}$ , and weakens at higher and lower frequencies. The difference is that for J1024 $-$ 0719, the trailing component appears to be enhanced again at even lower frequencies.

PSR J1400 $-$ 1431. The MeerKAT profile shows two distinct peaks, whereas the MWA profile shows only one. It is unclear whether the MWA profile should be aligned with the leading or trailing peak of the MeerKAT profile, but we favour the leading peak due to the asymmetry in the MWA profile. This alignment implies that the trailing component is not visible in the MWA profile. The MWA profile is also strongly linearly polarised, whereas the MeerKAT profile is completely depolarised.

PSR J1455 $-$ 3330. The trailing side of the profile (in particular the most prominent component) is enhanced in the MWA profile. However, the pulse width remains similar to the MeerKAT profile.

PSR J1536 $-$ 4948. The profile for this pulsar has three pulses which together span nearly the entire rotation phase. The MWA profile is noticeably scattered compared to the MeerKAT profile, and the trailing pulse is enhanced compared to the two other pulses.

PSR J1730 $-$ 2304.The profile has three main components, with the central component having a steeper spectral index than the two outer components. The leading component is significantly weakened in the MWA profile, and the trailing component is enhanced compared with the Parkes 50-cm profile.

PSR B1957+20. This is the first black widow MSP to be discovered. The MWA profile is scattered, but it is clear that the main pulse and interpulse have different spectral indices, with the amplitudes being more similar in the MWA profile.

PSR J2051 $-$ 0827. The MWA profile shape is similar to the Lovell profile at $410\,\mathrm{MHz}$ . Notably, this pulsar exhibits rapid depolarisation with frequency: the MWA profile shows significant linear polarisation, whereas the profile is depolarised at $410\,\mathrm{MHz}$ .

PSR J2129 $-$ 5721. The profile for this pulsar comprises a pulse with two components which is present across all frequency bands, as well as a post-cursor pulse which is only visible at high frequencies. The MWA profile is slightly broader and more linearly polarised than the Parkes 50-cm profile, but is otherwise consistent.

PSR J2145 $-$ 0750. The frequency evolution of this pulsar has been studied by several authors (e.g. Kramer et al. Reference Kramer, Lange, Lorimer, Backer, Xilouris, Jessner and Wielebinski1999; Kondratiev et al. Reference Kondratiev2016; Bhat et al. Reference Bhat2018; Kaur Reference Kaur2021). At low frequencies, the trailing component is significantly enhanced and shifts towards the central component, and the precursor component disappears. Both the linear and circular polarisation in the MWA profile is consistent with the Parkes and Lovell profiles. The position angle curve (see Figure 7) shows an orthogonal polarisation mode jump (i.e. a jump of $90^{\circ}$ ), but is otherwise relatively linear.

PSR J2150 $-$ 0326. The relative amplitude between the main pulse and interpulse changes considerably between the MeerKAT and MWA profiles, with the amplitudes being comparable in the MWA profile. Again we observe rapid depolarisation, with both pulses in the MWA profile being completely linearly polarised, whilst the MeerKAT profile is only marginally polarised.

PSR J2222 $-$ 0137. This is a mildly-recycled relativistic binary pulsar with a very low DM of $3.27\,\textrm{cm}^{-3}\,\mathrm{pc}$ (Boyles et al. Reference Boyles2013). The MWA profile is broader than the MeerKAT profile, which is partly due to an enhancement of the post-cursor components. The main component also appears wider in the MWA profile.

PSR J2241 $-$ 5236. This pulsar has a very narrow profile which shows little variation with frequency. The profile is weakly polarised at high frequencies, but is significantly polarised (both linearly and circularly) at MWA frequencies. As first noted by Kaur et al. (Reference Kaur2019), the precursor components visible in the Parkes profiles with very low amplitude are quite prominent in the MWA profile and are offset earlier in phase. Also notable is the trailing component visible as a shoulder in the MeerKAT and Parkes profiles, which cannot be seen in the MWA profile.

PSR J2256 $-$ 1024. There is a clear enhancement in the central profile component at lower frequencies, to the extent that the leading and trailing components cannot be seen in the MWA profile. We also see an enhancement in a post-cursor component in the MWA profile. The post-cursor can be seen with a much lower amplitude in the 352-MHz GBT profile.

PSR J2317+1439. The leading profile component that is prominent at higher frequencies is significantly weaker in the MWA profile. This can be seen more clearly in the LOFAR profile (see Kondratiev et al. Reference Kondratiev2016).

In general, the profile component separations and widths appear to vary very little with frequency, whereas the relative amplitudes vary considerably. This is different from non-recycled pulsars, which often show a separation of pulse components at low frequencies. Our results are consistent with Kramer et al. (Reference Kramer, Lange, Lorimer, Backer, Xilouris, Jessner and Wielebinski1999) and Kondratiev et al. (Reference Kondratiev2016), who attribute the slow frequency evolution of MSP profiles to the emission originating from a narrow altitude range due to the more compact magnetospheres.

A common trend observed for MSPs is asymmetric evolution of the profile components. Dyks, Wright, & Demorest (Reference Dyks, Wright and Demorest2010) suggested that enhancement of the trailing side of the profile at low frequencies can be attributed to aberration and retardation, whilst weakening of the trailing side is due to asymmetry in curvature radiation with respect to the dipole axis. Therefore, the observed frequency evolution can indicate the dominant of these effects. Kondratiev et al. (Reference Kondratiev2016) identified a clear enhancement of the trailing side for PSRs J1022+1001, B1257+12, J2145 $-$ 0750, and J2317+1439. Our observations support these results; however, we also observed that the trailing side of PSR J1022+1001 weakens again at low frequencies (i.e. the component amplitude turns over). In addition, we identify an enhancement of the trailing side of the profile for PSRs J1455 $-$ 3330, J2222 $-$ 0137, and J2256 $-$ 1024. We also identify weakening of the trailing side for PSRs J0030+0451, J0125 $-$ 2327, J1400 $-$ 1431, J2129 $-$ 5721, and J2241 $-$ 5236. Overall, there are a comparable number of MSPs that exhibit each of the two trends.

We consistently observe that the MWA profiles are comparably or more polarised than higher-frequency profiles. The depolarisation of pulsar emission with frequency is a well-known trend (e.g. Manchester et al. Reference Manchester, Taylor and Huguenin1973; Kramer et al. Reference Kramer, Lange, Lorimer, Backer, Xilouris, Jessner and Wielebinski1999), and is commonly attributed to the superposition of orthogonal polarisation modes (von Hoensbroech, Lesch, & Kunzl Reference von Hoensbroech, Lesch and Kunzl1998; McKinnon Reference McKinnon1997). However, the specifics of the depolarisation mechanism are not agreed on, and it is likely that a combination of magnetospheric, co-rotational, and physical effects determine the polarisation evolution with frequency (Noutsos et al. Reference Noutsos2015). Regardless of the physical explanation, the observation of this trend provides an indirect verification of MWA-VCS polarimetry.