2.1 Intergalactic magnetism

We do not know the origin of magnetism in the Universe. Broadly speaking there are two possible scenarios (Widrow, Reference Widrow2002; Rees, Reference Rees2006): a “top-down” model in which magnetism was generated everywhere in the Universe through plasma instabilities or phase transitions at very early times, or a “bottom-up” model in which magnetism was originally generated in stellar interiors and accretion disks, and then spread by various astrophysical processes. The key to distinguishing between these models are the magnetic properties of intergalactic gas since the initial conditions are expected to be quickly washed out by turbulent amplification processes in galaxy clusters (Stasyszyn et al., Reference Stasyszyn, Nuza, Dolag, Beck and Donnert2010; Hackstein et al., Reference Hackstein, Brüggen, Vazza, Gaensler and Heesen2019).

POSSUM offers significant advancements in measuring these properties. With its dense, wide-area RM grid, POSSUM is sensitive to magnetic structures on scales of megaparsecs, allowing us to probe magnetic field strengths down to the nanogauss level (Akahori et al., Reference Akahori, Gaensler and Ryu2014). Just as fast radio bursts (FRBs) have begun to reveal the gas distribution of the intergalactic medium (IGM) through dispersion measures (DMs; Macquart et al., Reference Macquart2020), POSSUM’s RM grid can be used to reveal the magnetic field of the IGM.

Specifically, RM $\propto \int n_e B_\parallel dl$ (where $B_\parallel$ is the line-of-sight component of the magnetic field $\mathbf{B}$ ) can be combinedFootnote ^c with FRB DMs, emission measures, and other tracers to directly estimate $B_\parallel$ , which can then be used to infer the total magnetic field strength under certain assumptions. Additionally, RMs are exceptionally sensitive probes of $n_e$ when other tracers are unavailable, allowing magnetic fields to “illuminate” the ionised gas via Faraday rotation. The effectiveness of this approach was demonstrated in POSSUM’s first Early Science paper, which mapped the gaseous extent of the Fornax cluster purely through an RM grid (Anderson et al., Reference Anderson2021).

2.1.2 Galaxy clusters and groups

Magnetic fields play a crucial yet still poorly understood role in the evolution of galaxy clusters and groups, particularly within their intracluster medium (ICM) and intragroup medium (IGrM). These fields are essential for accelerating and guiding cosmic rays (e.g. Ruszkowski & Pfrommer, Reference Ruszkowski and Pfrommer2023) and for regulating mixing and thermal conduction (e.g. Barnes et al., Reference Barnes2019). In rich clusters, the ICM typically hosts fields of $0.1$ – $10\,\mu$ G (Donnert et al., Reference Donnert, Vazza and Brüggen2018), likely amplified from primordial seeds whose exact origin remains unclear.

The work of Anderson et al. (Reference Anderson2021) demonstrated POSSUM’s transformational capability for studying the magnetised ICM in detail. As shown in Figure 1, ASKAP commissioning data on the Fornax cluster achieved an RM grid density of 27 RMs deg $^{-2}$ , revealing multiple cluster components and tracing the ICM out to the virial radius–well beyond its apparent X-ray extent. Building on this, the full POSSUM survey will map several other nearby clusters (e.g. Coma, Shapley, Virgo), each with hundreds of RM measurements inside the projected virial radius, compared to the $\lesssim10$ previously available (e.g. Bonafede et al., Reference Bonafede, Feretti, Murgia, Govoni, Giovannini, Dallacasa, Dolag and Taylor2010, Reference Bonafede, Vazza, Brüggen, Murgia, Govoni, Feretti, Giovannini and Ogrean2013; Stuardi et al., Reference Stuardi, Bonafede, Lovisari, Domnguez-Fernández, Vazza, Brüggen, van Weeren and de Gasperin2021). At least 100 additional high-mass, low-z clusters will have $\gtrsim 50$ RM sight lines, filling a major observational gap. Such a substantial increase in RM statistics will facilitate Faraday rotation analyses of large-scale ICM fields and depolarisation studies of smaller-scale features (Bonafede et al., Reference Bonafede, Feretti, Murgia, Govoni, Giovannini, Dallacasa, Dolag and Taylor2010), including how the field strength may scale with gas density (e.g. Osinga et al., Reference Osinga2025). Meanwhile, thousands of more distant or smaller clusters will still each have on the order of 1–10 POSSUM RM measurements, making RM stacking experiments feasible in bins of redshift, cluster richness, environment, and other key formation-related parameters (Krause et al., Reference Krause, Alexander, Bolton, Geisbüsch, Green and Riley2009).

Figure 1.

Comparison of NVSS (left; Taylor et al., Reference Taylor, Stil and Sunstrum2009) and POSSUM (right) Band 1 RMs within the Fornax cluster’s virial radius (grey dashed circle), showcasing POSSUM’s transformative capability to probe magnetised gas in clusters, groups, and many other $\sim$ degree-scale extragalactic objects. The background is a Digitized Sky Survey optical image (greyscale). Solid circles indicate the position, magnitude, and sign of RMs measured by each survey, with diameters proportional to (RM) $^2$ and colours indicating RM sign (red for positive, blue for negative). While only a few clusters or groups contain $\gtrsim 10$ NVSS RMs at sub-virial radii, POSSUM’s RM grid will sample thousands of these systems, including over a dozen with 100 or more RMs.

In addition, the IGrM of galaxy groups – the most common environment for galaxy evolution – are even less thoroughly characterised. Observations suggest the IGrM is shaped by shallower gravitational potentials, mergers, outflows, and galactic motions (Lovisari et al., Reference Lovisari, Ettori, Gaspari and Giles2021; Oppenheimer et al., Reference Oppenheimer, Babul, Bahé, Butsky and McCarthy2021), yet direct RM data for these lower-mass halos have been scarce. POSSUM’s order-of-magnitude increase in RM density now extends RM-grid techniques to hundreds of group-scale systems, revealing magnetised gas that can extend well beyond group boundaries into the cosmic web (Anderson et al., Reference Anderson2024a).

A key strength of POSSUM’s wide sky coverage is that it also samples the outskirts of clusters/groups and the Galactic ISM foreground, regions typically undersampled by targeted RM campaigns. This comprehensive approach supports both single-object studies and ensemble statistical methods. Furthermore, clusters and groups host polarised radio galaxies and relics that trace field configurations via 2D maps of RM, intrinsic polarisation, or depolarisation (e.g. van Weeren et al., Reference van Weeren, de Gasperin, Akamatsu, Brüggen, Feretti, Kang, Stroe and Zandanel2019; Stuardi et al., Reference Stuardi2019, Reference Stuardi, Bonafede, Rajpurohit, Brüggen, de Gasperin, Hoang, van Weeren and Vazza2022). Capturing such low surface brightness ( $\lesssim1\,\mu\textrm{Jy arcsec}^{-2}$ ) emission on $\lesssim10'$ scales requires a sensitive survey at sub- $1\,\textrm{GHz}$ frequencies, such as POSSUM, which can detect RM variations at the $\sim10\,\textrm{rad m}^{-2}$ level from shocks or cluster interactions (e.g. Bonafede et al., Reference Bonafede, Vazza, Brüggen, Murgia, Govoni, Feretti, Giovannini and Ogrean2013).

Finally, these observations will feed back into numerical simulations of the ICM/IGrM, whose hot ( $10^{7}$ – $10^{8}$ K), dilute ( $10^{-2}$ – $10^{-3}$ cm $^{-3}$ ), and weakly collisional nature (Kunz et al., Reference Kunz, Jones and Zhuravleva2022) results in complex, poorly understood coupling across a wide range of scales, posing significant challenges. While simulations have become increasingly powerful (Achikanath Chirakkara et al., Reference Achikanath Chirakkara, Seta, Federrath and Kunz2024), they still lack comprehensive observational constraints. By providing a dense RM grid across a wide range of cluster/group masses and redshifts, POSSUM will supply the necessary observational data to anchor these models. This will not only illuminate how magnetic fields evolve in weakly collisional plasmas over cosmic time but also clarify how these fields differ from those in collisional plasmas, where MHD is typically applicable.

2.2 The magnetism of galaxies

Magnetic fields are crucial to the dynamics and evolution of galaxies, influencing the energy balance of the ISM, regulating star formation, and controlling gas flows into and out of galactic disks. While detailed observations have revealed typical magnetic field strengths and structures (Beck & Wielebinski, Reference Beck, Wielebinski, Oswalt and Gilmore2013, updated 2023), it remains unclear whether these fields are shaped by standard dynamo processes or faster, more complex mechanisms. By simultaneously providing polarisation spectra of background radio sources and ISM emission, POSSUM will probe magnetic field structures across a wide range of scales, providing insights into the complex 3D magneto-ionic structures of the Milky Way and other galaxies (e.g. Basu et al., Reference Basu, Mao, Fletcher, Kanekar, Shukurov, Schnitzeler, Vacca and Junklewitz2018; Shah & Seta, Reference Shah and Seta2021).

2.3 The interplay between galactic and intergalactic magnetism

Magnetism plays a crucial role in both influencing and being influenced by the gaseous and energetic interactions between galaxies and the circumgalactic medium (CGM). This includes processes such as star formation (Steinwandel et al., Reference Steinwandel, Dolag, Lesch and Burkert2022), AGN-driven outflows (Vazza et al., Reference Vazza, Wittor, Brunetti and Brüggen2021), and gas accretion (Grønnow et al., Reference Grønnow, Tepper-Garca and Bland-Hawthorn2018). Understanding these processes requires answers to questions such as: How far do galactic magnetic fields extend into the CGM? How strong is the magnetism in halos and the CGM? How does this depend on environment, and how has this evolved over cosmic time? POSSUM will address these issues for the Milky Way and other star-forming galaxies, for interacting galaxies, and for AGN-driven outflows.

2.4 Other science topics

POSSUM will enable a broad range of scientific investigations beyond those described in detail above. Some specific examples include:

• • Using POSSUM’s RM grid in conjunction with FRBs, both to correct for Galactic foregrounds of FRB RMs, and to combine with an FRB “dispersion measure grid” to study the Milky Way’s halo (Pandhi et al., Reference Pandhi, Hutschenreuter, West, Gaensler and Stock2022);

• • Constraining properties of unusual objects, for instance Odd Radio Circles (Norris et al., Reference Norris2021a), via their spectra and polarisation, in combination with EMU data;

• • Searches for violations of Lorentz invariance, quintessence, and axion-like particles via cosmic birefringence (see Basu et al., Reference Basu, Goswami, Schwarz and Urakawa2021, and references therein); and

• • Studying both linear and circular polarisation properties of unusual transients, in conjunction with VAST data.

3.1 The Australian SKA Pathfinder (ASKAP)

ASKAP is a radio interferometric array located at Inyarrimanha Ilgari Bundara, the CSIRO Murchison Radio-astronomy Observatory in outback Western Australia (WA; Hotan et al., Reference Hotan2021). Situated within the Australian Radio-Quiet Zone WA, the array benefits from exceptional protection against terrestrial radio frequency interference (RFI). This is particularly advantageous for the critical 800–900 MHz range in POSSUM’s main survey band, which is generally unusable at other synthesis arrays due to its allocation for mobile phone networks. At ASKAP, all such transmitters are far beyond the horizon, with interference only occurring during rare atmospheric ducting events. When these events do happen, they affect bands of about 10 MHz wide for just a few hours, and the data is automatically flagged, resulting in a negligible impact on POSSUM’s data quality.

The ASKAP array comprises 36 fully steerable 12-meter Alt-Az mounted dishes equipped with an innovative third roll axis (described below), distributed over baselines ranging from 22 meters to 6 kilometres. The telescope can observe in the 700 to 1800 MHz frequency range. Each dish is equipped with a phased array feed (PAF), providing an instantaneous field of view of approximately 30 square degrees at 800 MHz, using 36 electronically formed dual linearly polarised beams. These beams are arranged to optimise both survey speed and uniformity of sensitivity across the field of view.

For each formed beam, the ASKAP correlator generates full polarisation visibility spectra (XX, XY, YX, and YY) with 15,552 fine spectral channels across an instantaneous bandwidth of up to 288 MHz. These data are organised into CASA measurement sets by the ASKAP ingest system (Voronkov, Reference Voronkov2020) and are recorded on disk at the Pawsey Supercomputing Research Centre for reduction by the ASKAP Pipeline.

Each beam requires independent calibration. During the beamforming process, an on-dish calibration system at the vertex of each antenna radiates noise that is used to correct any relative phase differences between the X and Y formed beams of the antenna (Chippendale & Anderson, Reference Chippendale and Anderson2019), ensuring high intrinsic Stokes-V purity. While the noise sources are switched off during science observations to avoid residual pickup by nearby antennas, they are periodically reactivated to correct for slow drifts in the gains of the PAF elements, which could otherwise degrade beam quality.

The instrumental bandpass and on-axis instrumental polarisation calibration for each beam are typically derived shortly before or after the science observations, usually once every 24 hours, using the primary calibrator PKS B1934–638. This calibrator is observed for a few minutes at the centre of each formed beam, and the resulting calibration is then applied to the corresponding science data during processing (see Section 4 for details).

Models of the ASKAP primary beams are derived from holographic measurements using the unresolved, unpolarised calibrator PKS B0408–65. These observations allow for the accurate modelling and correction of the off-axis instrumental polarisation response. A comprehensive characterisation of ASKAP’s off-axis polarisation leakage performance for various use cases – such as band-averaged RM synthesis or frequency-dependent depolarisation modelling – will be provided with the first data release. Early results indicate that ASKAP’s leakage from Stokes I into Q, U, and V is generally around 0.2% across most of the observing footprint, averaged across the band (with a caveat discussed in Section 5.1).

This exceptional wide-field performance is made possible by ASKAP’s axisymmetric antenna design and third roll axis, which maintain a constant sky orientation, akin to that of an equatorially mounted telescope. This capability is fully realised through routine automated observations and data processing, making ASKAP the only GHz-band radio telescope in the world capable of delivering high-purity polarisation data over wide fields almost immediately after observation. Additionally, ASKAP’s distinctive observing mode – where each sky area is typically covered by multiple independent beams, and each formed beam observes multiple independent regions of the sky, with all data readily accessible in the online archive (Section 4.2) – provides unique and powerful capabilities to identify and correct residual leakage or other biases during post-processing, using the statistics accumulated from field sources present in the main survey observations.

3.2 Survey design

To achieve the science goals outlined in Section 2 and ensure POSSUM’s broad serendipitous discovery potential and legacy value, we identified four key survey characteristics determined by the observational setup (as opposed to the intrinsic imaging capabilities of the array) that define the capability of POSSUM:

• • The average sky density of detected linearly polarised background sources;

• • The RM precision of the detected polarised sources;

• • The ability to detect and characterise ‘Faraday complexity’, that is, variation of RM within or across the source; and

• • The total sky coverage, and the overlaps with complementary radio surveys (see Section 7) and with ancillary multi-wavelength data.

These survey characteristics are controlled by the $\lambda^2$ coverage through frequency band choice, target sensitivity through per-field integration time choice, and the selection of sky area covered.

The choice of wavelength coverage involved significant trade-offs. Longer wavelengths improve RM precision due to the proportional relationship between polarisation angle and $\lambda^2$ , and increase source fluxes, following $\lambda^{0.75}$ , which can potentially yield more detected sources. However, these benefits come at the cost of reduced angular resolution (which impacts the study of intrinsic emission from resolved objects) and increased differential Faraday rotation both within sources and at scales below the observing resolution (leading to depolarisation; e.g. see Burn (Reference Burn1966); Tribble (Reference Tribble1991); Sokoloff et al. (Reference Sokoloff, Bykov, Shukurov, Berkhuijsen, Beck and Poezd1998)). This depolarisation becomes particularly pronounced at wavelengths longer than 30 cm (frequencies below 1 GHz), where most bright sources are substantially affected (e.g. Conway et al., Reference Conway, Birch, Davis, Jones, Kerr and Stannard1983). Balancing these trade-offs required careful consideration of POSSUM’s scope and science priorities, alongside empirical measurements of polarised source density and array performance across the available observing bands.

To inform this decision, a pilot survey was conducted, confirming that a greater density of polarised sources is detected in ASKAP’s band 1 (0.7–1.2 GHz) compared to band 2 (1.15–1.44 GHz). This difference was particularly pronounced due to the negative impact of RFI from global navigation satellites in the 1.15–1.3 GHz range, which significantly reduced overall sensitivity. The low prevalence of RFI in band 1 also allowed for precise frequency tuning to avoid RFI almost entirely and mitigate system temperature roll-up at the lower end of the band. As a result, the decision was made to conduct the main survey commensally with EMU (Norris et al., Reference Norris2011, Hopkins et al., 2025, in press) in ASKAP’s band 1, tuned to 800–1088 MHz.

While the choice of frequency tuning was flexible, our imaging strategy faced significant constraints. Modern spectropolarimetric analysis ideally balances finer image cube channelisation, which enhances the recovery of polarised emission with large |RM| values, against coarser channelisation, which reduces data processing and storage demands while improving image deconvolution. However, due to the current limitations of ASKAPsoft, POSSUM’s band must be imaged and deconvolved at its native resolution, with each 1 MHz channel processed individually across the frequency range. This approach, though currently necessary, is suboptimal and constrains image quality in certain respects, particularly near bright extended objects. Nevertheless, the archiving of POSSUM visibilities offers the potential to apply alternative custom calibration and imaging pipelines (e.g. Thomson et al., Reference Thomson2023) that could mitigate these issues. Additionally, ongoing developments in ASKAPsoft may eventually address these limitations in future data products, which we plan to highlight in subsequent data releases should this occur.

POSSUM’s band 1 observations use the closepack36 (Hotan et al., Reference Hotan2021) beam footprint with a beam spacing (pitch) of $0.^{\!\circ}9$ to balance between the telescope’s field of view ( $\approx 30\,\textrm{deg}^2$ ), the spatial uniformity of sensitivity, and the wide-field instrumental polarisation response. The planned sky coverage is shown in Figure 2, including a total area of 20,630 deg $^2$ (50.0% of the sky) that is presently constrained by the planned survey load for ASKAP over a 5-year period. The entire sky south of declination $\delta \approx -11^\circ 40^{\prime}$ will be covered, with additional sky coverage extending northward to $\delta \approx -7^\circ 0^\prime$ within a J2000 right ascension (RA) range of $03^\textrm{h} 05^\textrm{m}$ to $08^\textrm{h} 05^\textrm{m}$ , to $\delta \approx +2^\circ 20^{\prime}$ within a RA range of $08^\textrm{h} 05^\textrm{m}$ to $08^\textrm{h} 25^\textrm{m}$ , and to $\delta \approx +7^\circ 0^{\prime}$ within both RA ranges of $08^\textrm{h} 25^\textrm{m}$ to $15^\textrm{h} 15^\textrm{m}$ as well as $20^\textrm{h} 55^\textrm{m}$ to $03^\textrm{h} 05^\textrm{m}$ (through $00^\textrm{h}$ ). The irregular declination limit maximises overlap with extragalactic multi-wavelength survey projects.Footnote ^d The Galactic plane is covered at longitudes $\ell \lt 19.^{\!\circ} 8$ and $\ell \gt 226.^{\!\circ} 1$ . The survey area has been divided into 853 fields following the standard all-sky tiling scheme of ASKAP in equatorial coordinates (see Section 9.11 of Hotan et al., Reference Hotan2021). The integration time per tile is set to 10 hr, yielding an expected full-band sensitivity of $\sim 18\,\mu\textrm{Jy}/\textrm{beam}$ at 943 MHz. For the 692 southern fields at $\delta \lt -11^\circ 40^\prime$ , each tile will be observed in a single 10 hr session. The remaining 161 northern fields will each be observed in two 5-hr sessions, due to the limited time per day that these fields will be above the ASKAP horizon. The entire survey in band 1 will therefore consist of 1014 observations, or scheduling blocks (SBs).

The band 1 survey will be complemented by ancillary observations in band 2 (1296–1440 MHz), obtained through a full polarisation processing of the continuum data from the WALLABY survey (Koribalski et al., Reference Koribalski2020). The WALLABY survey will cover a sky area of 15,470 deg $^2$ (37.5% of the sky) using the square_6x6 PAF footprint with a pitch of $0.^{\!\circ}9$ at two interleaved positions, totalling 552 fields with 1104 pointings. As outlined by the green-shaded region in Figure 2, the band 2 sky coverage will approximately include the entire area within $\delta = -13^\circ 50^\prime$ to $-56^\circ30^\prime$ , as well as a southern extension to $\delta = -62^\circ 50^\prime$ within RA range of $00^{\textrm{h}} 10^\textrm{m}$ – $18^{\textrm{h}} 05^\textrm{m}$ , a northern extension to $\delta = +7^\circ 45^\prime$ within RA range of $01^{\textrm{h}} 55^\textrm{m}$ – $02^{\textrm{h}} 45^\textrm{m}$ , another northern extension to $\delta = +13^\circ 10^\prime$ within RA range of $11^{\textrm{h}} 35^\textrm{ m}$ – $13^{\textrm{h}} 30^\textrm{m}$ , and two equatorial bands within $\delta = -2^\circ 50^\prime$ to $+7^\circ 45^\prime$ within RA ranges of $08^{\textrm{h}} 20^\textrm{m}$ to $15\textrm{h} 20^\textrm{m}$ as well as $22^{\textrm{h}} 00^\textrm{m}$ to $02^{\textrm{h}} 45^\textrm{m}$ (through $00^{\textrm{h}}$ ). The integration time per field (with the two interleaves combined) will be $16\,\textrm{hr}$ , resulting in a sensitivity of $\sim 19\,\mu\textrm{Jy}$ per $10^{\prime\prime}$ beam.

Figure 2.

Planned sky coverage of POSSUM. The coverage of the band 1 (800–1088 MHz) component, commensal with the EMU survey, is shaded purple, and the band 2 (1296–1440 MHz) component, commensal with the WALLABY survey, is shaded green. (Regions covered by both components appear dark green.) The location of the Galactic Centre is marked by the cyan cross. The background greyscale map in the top panel is the reprocessed Haslam et al. (Reference Haslam, Salter, Stoffel and Wilson1982) 408 MHz all-sky synchrotron emission map (Remazeilles et al., Reference Remazeilles, Dickinson, Banday, Bigot-Sazy and Ghosh2015) in linear colour scale, while the background colour map in the bottom panel represents the Milky Way contributions to the Faraday rotation experienced by the linearly polarised emission from extragalactic sources derived by Hutschenreuter et al. (Reference Hutschenreuter2022). Both maps are in equatorial coordinates centred at (J2000) $\alpha = 0^{\textrm{h}}$ , $\delta = 0^\circ$ and in Mollweide projection.

We note that the frequency and sky coverage of bands 1 and 2 are also well-suited for complementary multi-GHz broadband follow-up using the Australia Telescope Compact Array (ATCA), whose lowest frequency is $1.4\,\textrm{GHz}$ (Wilson et al., Reference Wilson2011) when accounting for RFI, and for complementary polarisation surveys currently being conducted with other instruments, as discussed in Section 7.2.

3.3 Expected survey yield

The POSSUM survey is designed to produce three key outputs: (1) a dense grid of polarised extragalactic radio sources, (2) intrinsic polarisation measurements of individual resolved Galactic and extragalactic objects, and (3) mapping of Galactic diffuse synchrotron emission. The scientific objectives for resolved and diffuse polarisation products are detailed in Sections 2 and 7.3, respectively. In this section we focus on the expected characteristics of POSSUM’s RM grid.

Previous polarisation studies suggested that POSSUM’s band 1 survey would yield 30–50 RMs deg $^{-2}$ (e.g. Rudnick & Owen, Reference Rudnick and Owen2014b; Loi et al., Reference Loi, Murgia, Govoni, Vacca, Prandoni, Bonafede and Feretti2019) based on a conservative $8\sigma$ detection threshold (Macquart et al., Reference Macquart, Ekers, Feain and Johnston-Hollitt2012) of 144 $\mu$ Jy/beam. This projection is now supported by results from POSSUM’s pilot survey (Vanderwoude et al., Reference Vanderwoude2024), which achieved an RM density of 42 deg $^{-2}$ in an extragalactic field and 20 deg $^{-2}$ in a Galactic-plane field, with densities slightly decreasing to 35 and 14 RMs deg $^{-2}$ , respectively, for Faraday-simple sources (i.e. those showing no evidence of multiple polarised emission components; see e.g. Anderson et al., Reference Anderson, Gaensler, Feain and Franzen2015). Consequently, across the survey region, we expect to measure $(6.2-10.3) \times 10^5$ RMs for robustly detected polarised radio sources. We emphasise, however, that POSSUM will catalogue polarisation data for all sources catalogued by the EMU survey in Stokes I, irrespective of their measured polarised signal-to-noise ratio (see Section 5.1), allowing users to set their own detection thresholds and balance the trade-off between RM grid density and reliability.

With band 1’s frequency coverage, the full width at half maximum (FWHM) of the rotation measure spread function (RMSF) in RM synthesis is $58\ \textrm{rad m}^{-2}$ . An $8\sigma$ polarised signal-to-noise cutoff allows for an RM precision of approximately 3 rad m $^{-2}$ at the detection limit, with mean and median RM uncertainties of 2.1 rad m $^{-2}$ and 1.6 rad m $^{-2}$ , respectively, across the detected source population (Vanderwoude et al., Reference Vanderwoude2024). These uncertainties can be compared to the intrinsic RM scatter of extragalactic point sources at this resolution and frequency, which is approximately 5 rad m $^{-2}$ (e.g. Schnitzeler, Reference Schnitzeler2010; Taylor et al., Reference Taylor2024).

For sky areas with both band-1 and band-2 coverage, the combined sensitivity of $\sim 12\,\mu\textrm{Jy/beam}$ is expected to yield a polarised source density of 45– $60\,\textrm{deg}^{-2}$ at $8\sigma$ threshold, with a measured value of 48 $\textrm{deg}^{-2}$ (37 deg $^{-2}$ for Faraday-simple sources) delivered by the Pilot Survey (Vanderwoude et al., Reference Vanderwoude2024). The RMSF FWHM of $36\,\textrm{rad m}^{-2}$ will yield a median RM precision of $\approx 1\,\textrm{rad m}^{-2}$ across all detected polarised sources.

From the Vanderwoude et al. (Reference Vanderwoude2024) study, we anticipate approximately 16% of band-1 polarised components in extragalactic regions will be complex. Near the Galactic plane, this number can be as high as 33%. Using the catalogues published in that same study, we predict that around 4-5% of all EMU Stokes I components will be polarised with a signal-to-noise ratio $\gt8$ . For Stokes I components brighter than 10 mJy, these catalogues show that up to 85% of sources are polarised. Of all the polarised sources, around 90% have polarised fraction $\gt$ 1% and around 50% have polarised fractions $\gt$ 5%. These values are summarised in Figure 3. The Vanderwoude et al. (Reference Vanderwoude2024) study also compares the POSSUM RMs to existing values in the literature. Comparisons are challenging and limited due to the low source density of the existing catalogues, but they find generally good agreement where comparisons can be made.

Figure 3.

The expected percentage of polarised sources for Stokes I fluxes above various thresholds (top) and percentage of polarised sources for various polarised fractions (bottom) computed using the Vanderwoude et al. (Reference Vanderwoude2024) prototype POSSUM catalogues.

3.4 Survey timeline

POSSUM began trial survey observations in November 2022, and commenced the main survey in May 2023. The survey is expected to run for five years, concluding around May 2028. To indicate the typical pace of observations and data availability, by 16 July 2024, 200 of the 1014 planned band 1 scheduling blocks (SBs) had been validated and released, with an additional 19 observed and awaiting processing. For band 2, 24 out of 1104 SBs had been validated, with 2 more observed and pending processing. The survey status is regularly updated online.Footnote ^e

3.5 Performance relative to other polarisation surveys

We now compare POSSUM’s design (Section 3.2) and performance (Section 3.3) to other past and present radio polarisation surveys, to underscore POSSUM’s unique potential for advancing studies of cosmic magnetic fields. Key parameters for these surveys are listed in Table 1 and summarised in Figure 4.

Table 1.

POSSUM compared to other major modern interferometric polarisation surveys.

^a Lowest and highest usable frequencies. Bands containing substantial gaps (e.g. due to RFI) are marked with $\dagger$ .

^b Noise in Stokes Q and U.

^c Largest Angular Scale of recoverable emission, estimated as LAS $= 0.6\lambda_{\text{c}}/B_{\text{min}}$ , where $\lambda_{\text{c}}$ is the wavelength at the centre of the band, and $B_{\text{min}}$ is the minimum baseline length.

^d FWHM of the sky beam used for polarisation analysis; $\ge$ resolution of lowest frequency channel, so usually lower than associated Stokes I images.

^e $\delta\phi = 3.8/\Delta\lambda^2$ : FHWM of RM spread function; $\phi_\textrm{max} = \sqrt{3}/\delta\lambda^2$ : maximum measurable RM; $\phi_\textrm{max-scale} = \pi/\lambda_\textrm{min}^2$ : maximum detectable scale of extended structure in Faraday space.

^f Apertif re-tuned to 1220-1520 MHz (1292-1520 MHz usable) for the final year, to reduce RFI contamination.

^g A later LoTSS data release will have 6” angular resolution.

^h NVSS did not employ Faraday synthesis as only two frequency channels were available. For comparison, the median value of the tabulated NVSS RM uncertainties is ten times larger than for the POSSUM band 1 survey.

ⁱ SPICE-RACS spatial resolution is roughly declination-dependent.

^j VLASS $\phi_\textrm{max}$ is for the initial data release; final 16-MHz channels will give $\sim$ 16,000 rad m $^{-2}$ .

References: [1] This paper; [2] Adebahr et al. (Reference Adebahr2022), Adams et al. (Reference Adams2022); [3] Shimwell et al. (Reference Shimwell2022); O’Sullivan et al. (Reference O’Sullivan2023); [4] Taylor et al. (Reference Taylor2024); [5] Padmanabh et al. (Reference Padmanabh2023); [11] Taylor et al. (Reference Taylor, Stil and Sunstrum2009); [6] Riseley et al. (Reference Riseley2018); Riseley et al. (Reference Riseley2020); [7] Zhang et al. (in prep); [8] Thomson et al. (Reference Thomson2023); [9] Beuther et al. (Reference Beuther2016), Shanahan et al. (Reference Shanahan2022); [10] Lacy et al. (Reference Lacy2020).

As highlighted in Section 2, the polarisation re-analysis of the NVSS by Taylor et al. (Reference Taylor, Stil and Sunstrum2009) ushered in a revolution in RM grid science. However, the NVSS RM grid is relatively sparse by modern standards, with a density of approximately 1 RM deg $^{-2}$ . Additionally, its median RM uncertainty of around 10 rad m $^{-2}$ is comparable to the signal strengths of the very targets we now aim to study (see Section 2). Finally, it suffers from issues of precision and reliability due to the limited number of spectral channels used (e.g. Ma et al., Reference Ma, Mao, Stil, Basu, West, Heiles, Hill and Betti2019a, Reference Ma, Mao, Stil, Basu, West, Heiles, Hill and Bettib).

Subsequent surveys have improved on the NVSS in key respects, often dramatically so, but these advancements typically involve trade-offs in other critical areas, such as coverage area, resolution, or sensitivity. For example, Rudnick & Owen (Reference Rudnick and Owen2014a) conducted a 1.5-GHz polarisation survey with a detection threshold of 14.5 $\mu$ Jy beam $^{-1}$ , yielding a grid density of approximately 100 RMs deg $^{-2}$ at 10” resolution, but only covering an area of 0.27 square degrees (0.00065% of the sky).Footnote ^f The MeerKAT International GHz Tiered Extragalactic Exploration (MIGHTEE) survey (Taylor et al., Reference Taylor2024) aims to provide polarisation data with a sensitivity of a few $\mu$ Jy/beam and a resolution of 18”, though its coverage is limited to 20 deg $^2$ (0.048% of the sky). The Very Large Array Sky Survey (VLASS) (VLASS; Lacy et al., Reference Lacy2020) has covered the entire sky north of declination $-40^\circ$ (33,827 deg $^2$ , or $\sim$ 82% of sky) at 2–4 GHz with 5” FWHM in polarisation, significantly improving upon NVSS in resolution and enabling the mapping of RM structure across radio galaxies, albeit with substantially greater RM uncertainties than NVSS, and a largest angular scale of recoverable emission dramatically lower than most other large-area $\sim$ GHz-frequency surveys. At lower frequencies, surveys conducted by the Low-Frequency Array (LOFAR) and Murchison Widefield Array (MWA) together cover the entire sky at meter wavelengths (see Section 7.2.3), achieving high RM precision, though depolarisation effects result in a much lower RM grid density compared to higher frequency surveys.

POSSUM will vastly outperform the NVSS RM catalogue and polarimetric imaging in several key areas, including RM sky density and RM uncertainties, improving them by factors of approximately 35 and 10, respectively. This will provide an RM grid containing hundreds of times more information. Additionally, POSSUM offers enhanced spatial resolution at 20”, and its fully-sampled band will enable the detailed identification and characterisation of Faraday complexity – i.e. multiple polarised emission components with distinct RMs or components that span a range of RMs. The exceptional image quality, surface brightness sensitivity, the approximately 30 arcminute largest angular scale of recoverable emission provided by ASKAP, and the excellent wide-field polarisation purity achieved in routine end-to-end data processing (see Section 3.1) provides peerless new capabilities for mapping intrinsic polarisation in large samples of resolved objects.

Figure 4.

Comparison of POSSUM’s performance with other surveys listed in Table 1 across several key metrics. The RM precision is depicted by the colour of each circle, according to the logarithmic scale on the right. The size of each circle corresponds linearly to the angular resolution of the respective survey. While NVSS is included for reference, its RM precision is not displayed, as it is more suitable for characterizing broadband surveys. Low-frequency surveys like LoTSS, POGS, and POGS-X lie off the bottom of the plot due to their considerably lower RM Grid density compared to mid-frequency surveys, despite offering much better RM precision.

The unique combination of deep, large-area observations processed in an automated, end-to-end manner, covering the previously under-sampled southern sky – now a focal point for some of the world’s most advanced telescopes across all wavelengths – will make POSSUM’s survey data an invaluable legacy resource. This resource will remain unsurpassed until science-ready data products from all-sky SKA surveys become available.

4.1 Observatory pipelines

The ASKAP observatory maintains and operates data processing pipelines to produce the basic survey data products. The observatory’s POSSUM-specific pipeline handles all aspects of interferometric calibration and imaging, producing images, data cubes, and related products. This pipeline is intended to operate in a near-real-time mode, with calibration and imaging being performed as quickly as possible after an observation is complete. As a result, each observation is processed through this pipeline fully independently, even where multiple observations cover the same sky area (i.e. interleaved observations).

The bandpass is derived from observations of the unpolarised calibrator source PKS B1934–638. For observations with visibilities recorded at higher spectral resolution (i.e. in observations commensal with WALLABY), the calibrated visibilities are averaged to 1-MHz coarse resolution channels after bandpass calibration. Flagging is applied to the bandpass observations prior to computing bandpass solutions, and to the target data after applying bandpass solutions.

Post primary calibration, the on-axis leakages for ASKAP are usually less than 1% (see Section 3.1). Data on the bandpass calibrator source are used to derive the on-axis leakage spectra per antenna per formed ASKAP beam, which are applied to the target science data after the time-dependent gains are refined using self calibration. This calibration corrects the on-axis instrumental leakages in ASKAP data to better than 0.1%.

Intermittent observation of secondary calibrators to refine time-dependent gain variations is not feasible for multiple formed beams. For ASKAP, gain variations for each beam are derived using the iterative technique of self-calibration. For POSSUM data, a single amplitude and phase self-calibration iteration is used. This process leads to a small but non-zero astrometric shift with respect to reference catalogues, an issue which is described in detail in the EMU overview paper (Hopkins et al., 2025, in press).

Images are generated and CLEANed independently per beam. Stokes IQUV spectral cubes are produced, with each channel and Stokes parameter imaged and CLEANed independently. Meanwhile. Stokes I and V multi-frequency synthesis (MFS) images, using the full observation bandwidth, are made for EMU and WALLABY (see Section 7.1).

The synthesised beams or point spread functions (PSFs) corresponding to the different formed ASKAP beams are not necessarily identical, because each beam samples slightly different spatial frequencies (due to having different target directions) and because there is different flagging for each beam. The smallest common PSF size for all 36 formed beams in a given observation is calculated on a per-channel basis and all beams are convolved to have this PSF. Spectral variations in the PSF size are preserved, so higher frequency channels will typically have a smaller PSF.

The images for each formed beam are linearly mosaicked together to form a single set of image and cube products, each covering the full area of a single observation. The mosaicking uses full-Stokes models of the primary beam derived from holographic observations of the point source PKS B0407–638 to apply corrections for both the sensitivity pattern of the primary beam and the off-axis leakage of Stokes I into Q, U, and V.

After mosaicking, the source finding software Selavy (Whiting & Humphreys, Reference Whiting and Humphreys2012) is run on the Stokes I MFS image, producing catalogues of total intensity source components and emission islands.Footnote ^g These Stokes I catalogues will report fitted and deconvolved sizes along with fitted peak and total intensities. The 1D POSSUM pipeline will be run on each Stokes I component and thus the polarised and total intensity values can be linked between the catalogues. Several additional tools, such as those that generate the data quality reports used for validation, are also run on the mosaicked images and data cubes. The mosaicked images, cubes, and other ancillary data are then deposited in the ASKAP data archive.

4.2 Archiving, quality control, and validation

The observatory pipelines deposit data into the CSIRO ASKAP Science Data Archive (CASDA; Huynh et al., Reference Huynh, Dempsey, Whiting, Ophel, Ballester, Ibsen, Solar and Shortridge2020) and assign an “Unreleased” state that makes them visible only to members of the relevant survey teams.

Once the data have been uploaded to CASDA, the survey team makes an initial assessment to evaluate whether the data are of sufficient quality to enable science; if so, the data are released according to the ASKAP data release policy.Footnote ^h

Since the output data cubes are very large (up to 180 GB), manual download and inspection of each observation is difficult. The quality assessment is therefore performed using an HTML-format report that is created within the ASKAP pipeline. This report extracts a variety of numerical statistics, diagnostic plots, polarisation spectra of bright sources, JPEG versions of the images, and GIF movies that cycle through the cube channels, all of which are used to perform an initial manual inspection of the data.

4.3 Science-ready processing pipeline

The observatory pipeline model of near-real-time processing of individual observations has a few limitations that prevent it from performing all the processing needed to produce the final survey data products: it cannot perform processing that requires multiple observations or that requires additional data that are not yet available immediately after an observation is completed. POSSUM has developed an additional science-ready processing pipeline that performs steps that cannot be accomplished with the observatory pipelines, to get the data to a final science-ready state (i.e. requiring no additional processing steps prior to beginning scientific analysis).

The science-ready processing pipeline has been built and operated as a partnership between POSSUM team and the Australian SKA Regional Centre (AusSRC), as part of the AusSRC’s interest in developing expertise with SKA-like large-data pipelines. The pipeline performs four primary steps: convolution to survey resolution, correction for ionospheric Faraday rotation, tiling and reprojection, and mosaicking of overlapping/adjacent observations.

Convolution to survey resolution is required for three reasons: to enforce uniformity across the survey area, which simplifies some types of analysis that can be affected by resolution; to allow adjacent observations to have a common PSF to allow them to be mosaicked together; and to ensure that all frequency channels have a common PSF prior to spectral analysis. The convolution to survey resolution could be performed in the observatory pipeline, but it was decided that there is benefit to having an archived version of each observation at the best possible resolution, since this would allow users to use the better resolution data (at the cost of additional processing by the user) if needed for their specific scientific needs.

The survey resolution was chosen to be 20”, as this was found to be a reasonable compromise between preserving the best resolution possible (approximately 18”, based on the typical PSF size at the lowest frequency of 800 MHz) and keeping channels or observations for which data flagging has resulted in slightly worse PSFs. In individual observations, any channels so affected by flagged visibilities that the resolution is worse than 20” are flagged out in the data cubes; if more than 10% of channels are affected, an observation will be rejected during validation and will require re-observation.

Ionospheric correction is performed to subtract the Faraday rotation caused by the Earth’s ionosphere, which would otherwise contaminate the astrophysical Faraday rotation and introduce systematic errors. This correction would ideally be applied to the visibilities prior to imaging using external measurements of ionospheric electron content, which are published on the internet with latency of hours to days. However, the ASKAP observatory’s goal of near-real-time processing does not allow ionospheric correction prior to imaging, preventing a full correction of the time-dependent ionospheric Faraday rotation. Imaging without correcting the time-dependent effects introduces some amount of depolarisation, but at POSSUM frequencies this effect is typically much smaller than a 1% loss of polarised intensity, so it is primarily the systematic shift in RMs and polarisation angles that requires correction. We have developed a software toolFootnote ⁱ to calculate and apply a time-averaged correction to the Stokes Q and U cubes. Typical corrections are of order $1\ \textrm{rad}\,m^{-2}$ .

Tiling the data (i.e. breaking it into multiple files each covering smaller sky area) is necessary, as the large (>150 GB) FITS cubes produced by the observatory are difficult to work with outside of a high-performance computing facility. We have developed a tiling scheme based on the HEALPix (Górski et al., Reference Górski, Hivon, Banday, Wandelt, Hansen, Reinecke and Bartelmann2005) pixel grid: each tile is defined as the sky area covered by a single pixel of the HEALPix grid with $N_{\textrm{side}} = 32$ ; each tile thus has a sky area of approximately 3.36 deg $^2$ (or approximately 110’ per side). Each tile contains $2048\times 2048$ pixels with an approximate linear size of 3 $''\!\!.\,$ 22 (equivalent in size and position to HEALPix pixels of $N_{\textrm{side}}$ = 65 536) in the HPX projection.Footnote ^j The principle is similar to the HiPS format (Fernique et al., Reference Fernique2017), but without the hierarchical scheme of lower resolution versions present in HiPS. Our tile size is approximately 4.2 GB, small enough to be easily manipulated even on reasonably modern laptops. The survey area spans 6319 tiles in band 1 and 4985 tiles in band 2.

Finally, tiles from overlapping or adjacent observations are linearly mosaicked together, which mitigates the noisy edges around each observation and provides a more uniform sensitivity across the boundaries of observations. These final tiles are downloaded and ingested to the POSSUM collection at the Canadian Astronomical Data Centre archive.Footnote ^k

The POSSUM analysis pipeline comprises two parallel components. The first focuses on polarisation catalogues and polarisation spectra of source components. This is called the 1D pipeline, since it reduces each source component to a single dimension (frequency). The second component focuses on resolved and extended sources and on diffuse polarised emission. This is the 3D pipeline, using three dimensions: right ascension, declination, and frequency or RM.Footnote ^l

4.4 1D polarimetry pipeline

The 1D component of POSSUM’s science-ready processing pipeline was built around the idea of providing polarisation information for every single Stokes I source component present in the EMU catalogues, including non-detections. Rather than performing an untargeted search for polarised sources, each EMU source component is assigned a counterpart in the POSSUM catalogue with the same name/ID, allowing the catalogues to be easily joined and giving users direct access to the full power of both surveys together.

For each EMU source component, the Stokes IQU Footnote ^m spectra (with corresponding measurement uncertainties) are extracted from the tiles produced as described previously. An estimate of the diffuse foreground/background signal present around each source is also derived, and is subtracted from the on-source spectra; this estimate is stored along with the on-source spectra.

Using these spectra, RM synthesis is performed using the RM-Tools package (Purcell et al., Reference Purcell, Van Eck, West, Sun and Gaensler2020)Footnote ⁿ to transform the observed complex fractional polarised signal as a function of wavelength-squared, $\mathbf{P}(\lambda^2)$ , into the “Faraday dispersion function” (FDF; Burn, Reference Burn1966; Brentjens & de Bruyn, Reference Brentjens and de Bruyn2005). The FDF encodes properties such as the polarised fraction and RM of the strongest polarised emission peak, and can be used to estimate the degree of Faraday complexity.

The polarisation information is then merged with that from the Stokes I source-finding, along with relevant information from the data quality assessment/validation and observation metadata, to create the POSSUM catalogue. At this stage, various flags for properties such as significant polarised intensity or suitability for RM grid analyses are assigned; these will be discussed in detail in the forthcoming data release and pipeline description papers.

4.5 3D polarimetry pipeline

The 3D component of POSSUM’s science-ready processing pipeline focuses on the production of products to enable the analysis of spatially-resolved polarised emission, which can include resolved sources as well as diffuse Galactic emission. This pipeline is very straightforward: it performs RM synthesis on every pixel location in the Stokes IQU datacubes, producing FDF cubes matching the area of each frequency cube tile. An FDF characterisation routine is also run producing a series of 2D maps of polarisation properties as a function of sky position, such as peak Faraday depth, polarised intensity and fraction, and more, along with uncertainty estimates on these quantities.

4.6 Future processing possibilities

Several additional pipelines and products are currently being considered, and will be implemented depending on the level of team/community interest and on available development and computational resources. For example, a project is being developed to provide storage and computing resources from the China SKA Regional Centre Prototype (CSRC-P) to process calibrated and leakage-corrected measurement sets to produce image cubes optimised for diffuse polarised emission. Single-dish data from the PEGASUS project (see Section 7.3), when available, can then be merged with these cubes to produce images sensitive to all spatial scales, which will be crucial to study diffuse emission. Pipelines to perform alternative or additional polarisation analysis of individual sources, such as searches for multiple significant peaks in FDFs, QU-fitting to sources with detected polarisation using selected models, or characterisation of partially resolved sources, plus pipelines to combine band-1 and band-2 data, have been discussed and may be developed in the future.

5.1 Observatory products

The primary POSSUM products from the observatory pipeline are continuum channelised cubes for all Stokes parameters, I, Q, U, and V, at 1-MHz channel resolution. These are released as large FITS files, covering the full area of a single observation (approximately 6 $^\circ$ by 6 $^\circ$ ).The channels are not convolved to a common beam size, but are instead kept at the best resolution possible during the beam mosaicking process (on a per channel basis); the PSF for each channel is stored as a FITS extension in the cube files. The resulting data cubes are typically 192 GB per Stokes parameter for band-1 observations and 81 GB for band 2.

Each observation also has corresponding archival products from EMU or WALLABY. For both surveys, this includes Stokes I and V MFS images, Stokes I component and island catalogues, visibilities (averaged to 1-MHz channels), and validation reports; WALLABY observations also include a Stokes I spectral line cube and additional visibilities.

Despite ASKAP’s excellent polarimetric performance as highlighted in Section 3.1, we must raise a caveat: SBs validated and accepted prior to October 5, 2023, had the on-axis leakage correction applied twice in error. This resulted in an expected leakage of Stokes I into Q and U of approximately 1%. This issue affects 98 SBs, or about 10% of the total survey area. We will endeavour to reprocess and correct these affected SBs as soon as possible.

5.2 Enhanced data products

The pipeline described in Section 4.3 produces Stokes-frequency data cubes for Stokes I, Q, and U (Stokes V is not processed), as well as the EMU Stokes I MFS images convolved to POSSUM resolution, to match the cubes. These files follow the POSSUM HPX-projection tiling scheme (see Section 4.3), resulting in cubes of approximately 4.5 GB for band 1 and 2.3 GB for band 2.

We intend to produce a tool to enable easy creation and download of image/cube cutouts (including stitching/mosaicking across adjacent tiles). However, this has not been developed at the time of writing, and is anticipated for the first data release.

5.3 1D pipeline products

The primary product of the POSSUM 1D pipeline described in Section 4.4 is a catalogue of the polarisation properties for every source component that appears in the EMU component catalogue. This catalogue is constructed as a one-to-one match to the EMU catalogue and contains the EMU component names, which allows trivial matching of source components between these two catalogues and give users easy access to the full information provided by both catalogues. The POSSUM catalogue contains information on the basic polarisation properties (e.g. fractional polarisation, RM, polarisation angle), sourcefinder properties (e.g. deconvolved size, peak intensity, integrated flux density), observation metadata (e.g. observation IDs and dates, data quality metrics), and source assessment flags (e.g. polarisation detection/non-detection, Faraday simplicity/complexity, suitability for use in an RM grid). Full descriptions of the pipeline and the resulting catalogue will be presented in the first data release and pipeline overview papers.

The specifics of the archiving and public access to the catalogue are not finalised at the time of writing, but are expected to include a live service allowing queries through a Table Access Protocol (TAP) interface.Footnote ^q

In addition to the catalogue, the 1D pipeline also outputs the Stokes I, Q, and U spectra extracted for each source component, and the resulting FDFs computed from those spectra. The spectra may be useful for re-analysis with different techniques, or for combining with observations at other frequencies. The spectra are archived per-tile in the PolSpectraFootnote ^r FITS binary table format (Van Eck et al., Reference Van Eck2023) and will be made available through a TAP-accessible database; the FDFs will similarly be stored in both FITS binary tables and a TAP interface.

5.4 3D pipeline products

The primary outputs of the 3D pipeline (see Section 4.5) are FDF cubes, containing the FDF for each pixel of sky area. These cubes share the same tiling scheme and pixel grid as the Stokes-frequency cubes from which they are derived. The FDF-characterisation part of the 3D pipeline produces a series of maps of various polarisation properties, such as the polarised intensity, peak rotation measure, and polarisation angle of the strongest peak in the FDF, along with maps of the (statistical) measurement uncertainties in these quantities. The complex RMSF is also stored; all of the required information needed to deconvolve the FSF using the RMSF (e.g. using RMCLEAN; Heald et al., Reference Heald, Braun and Edmonds2009) is retained, to allow users to perform this type of analysis if desired.

6.1 The RM grid

The dramatic improvement in RM grid capability brought about by POSSUM is vividly illustrated in Figures 5, 6, and 1 (see also Vanderwoude et al., Reference Vanderwoude2024). Figure 5 highlights a region centred on a Galactic latitude of $+29^{\circ}$ , where the overlapping NVSS and (incomplete) POSSUM RM grids are compared. The POSSUM grid reveals coherent and subtle RM structures not visible in the NVSS data, along with distinct outlier RMs that markedly deviate in magnitude and sign from the surrounding smoother structures. A wider view is provided in Figure 6, showing a wide range of coherent structures on various scales and an overall variation in RM magnitude and grid density with Galactic latitude.

Figure 5.

An illustration of the dramatic improvement in RM grid density and precision between the NVSS (top; Taylor et al., Reference Taylor, Stil and Sunstrum2009) and POSSUM (bottom) band 1 surveys, focusing on mid-northern Galactic latitudes where the surveys currently overlap. POSSUM provides approximately 35 times higher RM grid density and 10 times better RM uncertainties, significantly enhancing the visibility and detail of coherent structures therein. The RM grids are shown using nearest-neighbour interpolation, where each cell’s colour represents the RM of an individual linearly polarised source, normalised using Matplotlib’s SymLogNorm. The NVSS coverage has been masked to align with POSSUM’s survey progress, including only NVSS sources within 1.5 degrees of a POSSUM source. Gray regions indicate areas not yet observed by POSSUM. The square tile boundaries represent ASKAP footprints, each covering approximately $5 \times 6$ degrees with 36 formed beams. The spiked polygons at the periphery are artifacts from the interpolation method used.

Figure 6.

Similar to the bottom panel of Figure 5 but focusing on a larger region extending south from the Galactic plane, this figure illustrates the rich variety of structures revealed by the POSSUM RM grid. These structures include small-scale filaments and “interfaces” where the RMs suddenly change value or sign, with the latter indicating reversals in the electron-density-weighted mean line-of-sight magnetic field direction. Closer to the Galactic plane, the absolute RM values tend to increase, while the RM grid density decreases. This reduction in grid density is due to higher noise levels in the maps of these regions, generally caused by bright diffuse Galactic ISM emission and artifacts resulting from the current absence of single-dish (zero-uv-spacing) data.

In Figure 1, we compare RMs measured by NVSS and POSSUM within the virial radius of the Fornax cluster. This comparison highlights the significant improvements in RM densities and uncertainties offered by POSSUM. Notably, POSSUM’s ability to probe extragalactic degree-scale objects is greatly enhanced compared to previous surveys, with a mean RM separation of only $\sim10$ arcminutes. These refined measurements allowed Anderson et al. (Reference Anderson2021) to identify a massive, previously undetected reservoir of warm plasma in the Fornax cluster.

6.2 Extragalactic radio sources

POSSUM’s combination of a broad range between maximum recoverable angular scale and observing resolution, excellent polarimetric purity, and high RM precision, all within a frequency band where typical sources remain highly polarised, makes it an exceptionally powerful tool for producing detailed polarisation maps across a wide range of radio galaxies and other resolved objects. This capability is illustrated in Figure 7, which presents polarisation maps of several radio galaxies spanning a range of brightness and angular scales in POSSUM’s band 1.

One of the objects shown is Fornax A, which exceeds ASKAP’s maximum recoverable scale in Stokes I, leading to a negative bowl effect. However, the smaller-scale structure in Stokes Q and U allows for accurate recovery of polarised intensity, RM, and angle maps. Remarkably, the quality of these maps is comparable to those produced by Anderson et al. (Reference Anderson, Gaensler, Heald, O’Sullivan, Kaczmarek and Feain2018b) using over 100 hours of targeted Australia Telescope Compact Array observations, whereas POSSUM achieves this with just 10 hours of integration in routine survey mode with automatic calibration and imaging. Intricate RM and polarisation structures are visible in Fornax A and the other galaxies, even at sub-beam scales, due to the vector nature of polarisation. Such structures can be used to infer complex internal dynamics within radio lobes (e.g. Anderson et al., Reference Anderson, Gaensler, Heald, O’Sullivan, Kaczmarek and Feain2018b) and interactions with their environments (e.g. Guidetti et al., Reference Guidetti, Laing, Murgia, Govoni, Gregorini and Parma2010; Guidetti et al., Reference Guidetti, Laing, Bridle, Parma and Gregorini2011; Guidetti et al., Reference Guidetti, Laing, Croston, Bridle and Parma2012).

In addition to nearby, highly resolved radio galaxies, POSSUM will image many fainter and less extended sources. Based on detections from EMU (Gupta et al., Reference Gupta2024) and LoTSS (Williams et al., Reference Williams2019), and assuming an average polarisation fraction of 25% for resolved radio lobes with a spectral index of -0.8, we expect POSSUM to map polarisation across more than 100, 1000, and 10,000 objects similar in scale and brightness to those shown in rows 2, 3, and 4, respectively, of Figure 7. Though less prominent than the brighter 3C sources, these galaxies are still sufficiently bright and well-resolved to provide detailed polarisation and RM maps. This will significantly enhance our ability to conduct population studies of radio galaxies and other astrophysical objects.

6.3 Galactic objects and ISM-dominated fields

POSSUM provides excellent surface brightness sensitivity to discrete Galactic objects and resolved Galactic ISM emission. The polarised emission from the diffuse Galactic ISM serves as a key probe of magnetic fields and thermal gas. POSSUM can detect diffuse polarised emission over much of the sky (see Figures 8 and 9), enabling new studies of the Galactic ISM, helping to disentangle Galactic emission from extragalactic polarisation and RM studies.

Figure 7.

POSSUM band 1 images of selected radio galaxies, illustrating POSSUM’s remarkable combination of resolution and sensitivity to large-scale emission. Each row features a different object, decreasing in angular size from top to bottom. POSSUM is expected to map the polarisation across the lobes of more than 100, 1,000, and 10,000 radio galaxies (respectively) of similar sizes to those shown in rows 2, 3, and 4, significantly expanding the number of galaxies we can map in detail. The left, middle, and right columns display Stokes I, polarised intensity, and RM maps, respectively, at POSSUM’s native 20" resolution. Magenta markers on the polarised intensity maps indicate the orientation of the sky-projected magnetic field, derived from polarisation angle and RM. The synthesised beam is shown in the lower left of the Stokes I plots, overlaid on a $1\times1$ arcminute box for clarity and scale. Notably, while Fornax A exceeds ASKAP’s maximum recoverable scale in Stokes I (resulting in a negative bowl effect), Stokes Q and U reveal smaller-scale structures, enabling accurate recovery of polarised intensity, RM, and angle maps (cf. Anderson et al., Reference Anderson, Gaensler, Heald, O’Sullivan, Kaczmarek and Feain2018b). With its broad scale sensitivity, excellent polarimetric precision, and minimal depolarisation compared to lower-frequency surveys, POSSUM is ideal for RM and polarisation mapping, and complements surveys at other wavelengths.

Figure 8.

Part of a field centred at $(\ell,b)=(49.75^\circ,-29.5^\circ)$ , demonstrating POSSUM’s excellent sensitivity to Galactic ISM emission structures. The top left panel shows the all-sky Wilkinson Microwave Anisotropy Probe (WMAP; Bennett et al., Reference Bennett2003) linearly polarised intensity image for context, with an inset highlighting a region of substantial polarised emission. A cyan box indicates the area shown in the other panels. The top right panel provides a zoomed-in view of this area, displaying the ASKAP Stokes I map from the EMU survey. The bottom left panel shows the ASKAP linearly polarised intensity measured by the POSSUM survey, while the bottom right panel presents the peak RM derived from RM synthesis using POSSUM data. These observations reveal significant Galactic diffuse emission, sufficiently bright to reveal notable and coherent RM structures in this region. Unlike Figures 5 and 6, where coherent RM structures are revealed by extragalactic polarised background sources, here the RM structure is both generated and illuminated by the Galactic ISM.

Figure 9.

Comparison of NVSS (Condon et al., Reference Condon, Cotton, Greisen, Yin, Perley, Taylor and Broderick1998) and ASKAP EMU/POSSUM maps in a typical extragalactic field. The top left panel shows the NVSS Stokes I map, while the top right panel presents the corresponding NVSS linearly polarised intensity. The bottom left panel displays the ASKAP Stokes I map from the EMU survey, and the bottom right panel shows the linearly polarised intensity derived from RM synthesis using POSSUM data. Notably, POSSUM reveals approximately 27 polarised detections in this region, whereas the NVSS data shows no catalogued polarised sources. The POSSUM polarisation map also reveals faint ISM emission structure, even in a largely empty field, a feature frequently observed in POSSUM data but not seen in NVSS.

7.1 Other ASKAP surveys

7.2 Other interferometric polarisation surveys

Table 1 lists the parameters of current large-area polarisation surveys, most of which will be complementary with POSSUM. We summarise these surveys and their relevance in the following subsections.

7.3 Single-dish polarisation surveys

Modern single-dish polarisation surveys are motivated both by the need for foreground correction of cosmic microwave background (CMB) polarisation (e.g. Carretti et al., Reference Carretti2019; Jones et al., Reference Jones2018; Rubiño-Mart n et al., 2023), and by the exploration of the Faraday structure of diffuse Galactic synchrotron emission. This effort is dominated by the Galactic Magneto-Ionic Medium Survey (GMIMS; Wolleben et al., Reference Wolleben2019, Reference Wolleben2021), which aims to observe the entire sky over 300–1800 MHz by combining six component surveys.

Thanks to ASKAP’s excellent short-spacing coverage, POSSUM maps are sensitive to structures up to $0.^{\!\circ} 5$ across, although these scales are significantly undersampled. Such structures occur along the Galactic plane and in large extragalactic objects like Centaurus A (e.g. Anderson et al., Reference Anderson2018a) and the Magellanic Clouds, creating substantial artifacts in the images, including large negative ‘bowls’ around extended sources. Similarly, the diffuse polarised emission in POSSUM images is often undersampled, especially at the short-wavelength end of the band where the fragmentation of diffuse polarisation described at the end of Section 7.2.3 is less effective. The long-term plan is therefore to remove such artifacts by combining ASKAP continuum survey data (both POSSUM and EMU) with single-dish surveys. There is excellent overlap in the Fourier plane between long-track ASKAP observations and those from the 64-m Murriyang Radio Telescope at Parkes, for example, as required for accurate combination.

POSSUM will be combined with different single dish surveys for each of its band-1 and band-2 components. The frequency range of POSSUM band 2 is almost entirely covered by the Southern Twenty-centimetre All-sky Polarisation Survey (STAPS), conducted with Murriyang (Raycheva et al., Reference Raycheva2025; Sun et al., Reference Sun2025). STAPS is the high-band South component of GMIMS: it covers $-90^\circ \lt \delta \lt0^\circ$ over 1328–1768 MHz, with a frequency resolution of 1 MHz. The missing coverage at $\delta \gt 0^\circ$ could be supplied by GMIMS high-band North (Wolleben et al., Reference Wolleben2021), although there is little diffuse Galactic emission in this part of POSSUM. STAPS was observed commensally with the 2.3 GHz S-PASS survey (Carretti et al., Reference Carretti2019), using an off-axis feed. To cover band 1, the POSSUM-EMU-GMIMS All Stokes UWL Survey (PEGASUS, Carretti et al., Parkes project ID P1123) is currently underway on Murriyang. PEGASUS will cover $-90^\circ \lt \delta \lt +20^\circ$ over 700–1440 MHz, with a final frequency resolution of 0.5 MHz, with a scan strategy similar to S-PASS and STAPS. The frequency overlap with STAPS will allow cross-calibration, although the new PEGASUS data are superior due to improved RFI rejection and an on-axis feed.

At the low angular resolution of the single-dish maps (20’–60’), the surface brightness sensitivity is significantly higher than in the POSSUM images, which will not detect faint Galactic emission at high latitude. Although it may seem excessive to map the entire hemisphere when coverage is primarily required over some 7000 deg $^2$ along the Galactic plane and in a few other specific regions, (a) it is not practical to scan restricted regions of the sky while preserving the large-scale structure, and (b) PEGASUS also functions as the mid-band South component of the broader GMIMS survey project.

PEGASUS observations are expected to be complete in 2025, three years ahead of POSSUM. Unlike POSSUM, the PEGASUS data cannot be fully reduced until observations are completed, since adjacent scans are observed in different observing runs and the final reduction relies on scan crossings to solve for residual baseline offsets (destriping). The first images of the diffuse polarisation from combined POSSUM and PEGASUS data will be created as enhanced products once PEGASUS reduction is complete.

7.4 POSSUM exploitation projects: QUOCKA and QUOLL

Like other large surveys, POSSUM will often serve as a finder survey for objects deserving more detailed follow-up.

An early example is the QU Observations at Cm Wavelength with Km baselines using ATCA project (QUOCKA; Heald et al., in prep.), aimed at clarifying the magnetoionic structure of AGN jets and lobes and their surroundings. QUOCKA targets 537 radio galaxies with reliably detected ASKAP polarisation and brightness suitable for detection with the Australia Telescope Compact Array, selected from 13 ASKAP Early Science fields.

QUOCKA observations cover 1.1–8.5,GHz, spanning a factor of 60 in $\lambda^2$ , thereby dramatically improving sensitivity to Faraday complexity generated by complex, strongly magnetised environments such as radio galaxies. QUOCKA also provides a sensitive probe of inner jet structure, measuring circular polarisation with minimal residual leakage. Each source was observed for about 30 min in each band, split into six 5-min snapshots over 12 hours to improve uv coverage. After convolving all frequency planes to a common resolution, QUOCKA data typically have $10''-15''$ resolution and a broadband sensitivity of $\approx50$ $\mu$ Jy beam $^{-1}$ . QUOCKA’s broadband spectropolarimetry is expected to be invaluable as a polarisation reference, and was proven so during the commissioning of ASKAP’s polarimetric performance, helping to identify spurious features in the ASKAP data and guiding the necessary corrections to achieve the data quality required by POSSUM.

Since many QUOCKA sources are dominated by structures unresolved by ATCA, the Australian Long Baseline Array (LBA) has been used to map a few of the brighter QUOCKA sources with the QU Observations Leveraging the LBA project (QUOLL; Kaczmarek et al., in prep.) project. The milliarcsecond resolution at both 1.4 and 10 GHz will be sufficient to resolve sub-kpc scale structures, mapping the detailed polarised morphology similar to the MOJAVE survey (e.g. Hovatta et al., Reference Hovatta, Lister, Aller, Aller, Homan, Kovalev, Pushkarev and Savolainen2012; Lister & Homan, Reference Lister and Homan2005; Lister et al., Reference Lister, Aller, Aller, Hodge, Homan, Kovalev, Pushkarev and Savolainen2018) but at frequencies suitable for tracing depolarisation effects, and linking directly with the QU spectra from ATCA.