TRAPUM search for pulsars in supernova remnants and pulsar wind nebulae – II. Survey analysis and population study 2025 TRAPUM search for pulsars in supernova remnants and pulsar wind nebulae – II. Survey analysis and population study – 19 J. D. Turner, B. W. Stappers, E. Barr, M. Burgay, M. Colom i Bernadich, V. Graber, M. J. Keith, M. Kramer, L. Levin, Y. P. Men, C. Pardo-Araujo, T. Thongmeearkom, J. Tian, P. V. Padmanabh, P. Weltevrede, J. Behrend, W. Chen, E. F. Keane, A. Ridolfi Abstract We present the second and final set of TRAPUM searches for pulsars at 1284 MHz inside supernova remnants and pulsar wind nebulae with the MeerKAT telescope. No new pulsars were detected for any of the 80 targets, which include some unidentified TeV sources that could be pulsar wind nebulae. The mean upper limit on the flux density of undetected pulsars is 52 \upmu \upmu Jy, which includes the average sensitivity loss across the coherent beam tiling pattern. This survey is the largest and most sensitive multi-target campaign of its kind. We explore the selection effects that precluded discoveries by testing the parameters of the survey iteratively against many simulated populations of young pulsars in supernova remnants. For the synthetic pulsars that were undetected, we find evidence that, after beaming effects are accounted for, about 45 per cent of pulsars are too faint, 30 per cent are too smeared by scattering, and a further 25 per cent have a modelled projected location which places them outside their supernova remnant. The simulations are repeated for the S1 subband of the MeerKAT S-band receivers, resulting in a 50 − 150 50-150 per cent increase in the number of discoveries compared to L-band depending on the flux density limit achieved. Therefore, higher frequency searches that can also achieve improved flux density limits are the best hope for future targeted searches. We also report updated properties for the two previous discoveries, including a polarimetry study of PSR J1831 − - 0941 finding a rotation measure of 401 ± \pm 1 rad m -2 . 1 Introduction Pulsars are rapidly spinning neutron stars (NSs). Their high magnetic field strengths generate beams of radio waves along the magnetic field axis, which are then seen as radio pulses when the beam crosses the line of sight to an observer. NSs are born when a massive star undergoes a core collapse supernova (CCSN) at the end of its life ( Baade & Zwicky 1934 ; Heger et al. 2003 ) . CCSNe expel approximately 1 M ☉ of ejecta which moves outwards and sweeps up any circumstellar matter ( Chevalier 1982 ) , including \sim 10\,\text{M}_{☉} of matter lost by the star prior to the supernova. The expanding structure of material is visible as a supernova remnant (SNR), so the youngest pulsars are therefore typically located in the vicinity of SNRs. The discovery of more pulsar-SNR systems is important for constraining the Galactic NS birth rate ( Keane & Kramer 2008 ) and understanding their formation channels. Young pulsars are also laboratories for studying the energy budgets, X-ray and very high energy (VHE) \upgamma -ray emission of the pulsar wind nebula (PWN) and its interaction with surrounding media ( Kargaltsev & Pavlov 2008 ; Shibata et al. 2016 , e.g.) , and also the interiors of NSs ( Haskell & Melatos 2015 ; Antonopoulou et al. 2022 , e.g.) . The TRAPUM ( Stappers & Kramer 2016 , TRAnsients and PUlsars with MeerKAT;) project is a large survey for pulsars with the MeerKAT telescope. Since 2022, TRAPUM has been targeting dozens of SNRs, PWNe and also some unidentified sources from the H.E.S.S. TeV Catalogue ( Abdalla et al. 2018a ) that could be PWNe. This work follows Turner et al. 2024 , which can be deferred to for an introduction to the survey, what it might tell us about the properties of the pulsar-SNR populations, the energetics of PWNe and the Galactic birthrates of pulsars. Paper I also presented the first set of observations and two pulsars that had been discovered. PSR J1831 - 0941 is a young pulsar we discovered in candidate SNR G22.045 - 0.028 ( Dokara et al. 2021 ) , which we concluded must be powering the X-ray counterpart PWN G22.0+0.0 ( Yamauchi et al. 2016 ) . We also discovered PSR J1818 - 1502 in the foreground of G15.9+0.2, although the pulsar and SNR are probably unassociated. This new work describes all remaining searches since then. In § 2 , we provide information about how the target list was compiled, and the selection criteria used to add or reject sources from the list. In § 3 , we present the final observations of the survey, describe the most recent follow-up of the two discoveries and provide a brief interpretation and discussion of the results. The main analysis and discussion follows in § 4 , with the description of an in-depth population synthesis that aims to understand the discovery rate of the survey. New results for the two discoveries are presented and discussed in § 5 . Finally, we summarise the survey and make recommendations for future searches in § 6 . 2 Composing the target list 2.1 Target selection criteria One motivation of the TRAPUM survey was to improve the current understanding of the population of young pulsars in supernova remnants by i) analysing the upper limits and discovery rate for a large sample and ii) discovering many new pulsars and carrying out follow-up studies of their systems. This was feasible as the long total observing time allocated to the survey could be used to search a sample large enough to make inferences about the population, thus facilitating i). Moreover, ii) should follow in turn if a large number of searches yielded many new discoveries. In the early phases of the survey, a list of targets was compiled from three sources; the 309 Galactic supernova remnants in the December 2022 version of the Green catalogue ( Green 2022 , G22;) , the sources discovered and catalogued by the H.E.S.S. survey ( Abdalla et al. 2018a ) labelled as PWNe (14) or unidentified (25), as listed in TeVCat 1 1 1 http://tevcat.uchicago.edu/ v.3.400 ( Wakely & Horan 2008 ) . A further 216 candidates were identified across results from the MAGPIS ( Helfand et al. 2006 ) , second epoch Molonglo Galactic Plane Survey ( Green et al. 2014 , MGPS-2;) , THOR ( Anderson et al. 2017 ) , GLEAM ( Hurley-Walker et al. 2019b ) and GLOSTAR ( Dokara et al. 2021 ) surveys. The combined target list thus totalled 564 sources. Equal allocation of the telescope time would have allowed just under 10 minutes per target, which would not have significantly improved sensitivity over the previous best searches carried out with the Murriyang telescope at Parkes by Crawford et al. 2002 , which achieved flux density limits of 60-180 \upmu Jy at 1384 MHz for 5 SNRs. We therefore applied some criteria to reduce the list by around half and enable 20-40 minute integration times and flux density limits of 30-50 \upmu Jy. Figure 1 lists the criteria used and how they reduced the target list. Criterion A is of course necessary due to the location of MeerKAT, criterion B onwards were chosen to help maximise the number of discoveries. The final list of 281 targets was then further narrowed down by selecting candidates inferred to be filled centre or composite type, and by selecting against sources that already had competitive limits within the Max-Planck-Institut für Radioastronomie (MPIfR)–MeerKAT Galactic Plane Survey ( Padmanabh et al. 2023 , MMGPS;) or from previous targeted searches. Multiple examples in particular concerned a large number of candidates from MAGPIS, the THOR survey and the GLOSTAR survey, which generally have small diameters and are located deep in the Galactic plane. MMGPS-S is highly competitive with our searches at L-band for such targets due to the sky temperature and predicted scattering along the lines of sight. It was therefore decided that small candidates overlapped by the MMGPS-S footprint would not be observed. The MeerKAT S-band receivers were also not yet available when we were preparing the target list. We chose not to switch to S-band at a later stage, instead deciding that the uniformity of a full survey at L-band would be more useful for later population-based analyses. Initially we did not apply criterion C to targets. We chose to observe the X-ray pulsar PSR J1849 - 0001 associated with HESS J1849 - 000 to search for a radio signal at the known rotational periodicity. However, we subsequently decided to only observe known NSs if they were members of the class of central compact objects (CCOs). Conversely there were sources that did not match some of the criteria but that we still searched, because as important as observational uniformity was, there was a desire to maximise discoveries based on information available about the sources in the literature. For example, G28.8+1.5 has a very large angular diameter of \sim 100 arcmin, but we found it to be worth targeting regions of interest within the shell. Figure 1: Chart showing the sequential application of the criteria that were used on a starting list of 564 sources combined from the G22 and H.E.S.S. catalogues and SNR candidates in the literature. As can be seen, these criteria halved the number of sources, such that we could then select optimal targets on a case-by-case basis. 2.2 Data rate constraints The cap of 480 or 760 coherent beams (CBs) that the Filterbanking Beam Former User Supplied Equipment ( Barr 2018 , FBFUSE;) is capable of handling places a constraint on the size of the target that could be fully searched in a single pointing. This is the reason for criterion F in Figure 1 . A majority of the targets are circular or elliptical, so the number of beams required to fully tile them scales with the square of the target’s radius. There are several ways of increasing the size of the CBs to maximise coverage, though they all trade off sensitivity in some way: • By observing the target at a lower elevation, the projected baselines of the array in the direction pointing towards the source are smaller, thus the CBs are wider along that axis. • The inner 44 telescopes of the MeerKAT array (the core) can be used instead of the full array. This reduces the longest baseline by a factor of \sim 4, and so increases the CB coverage by a factor of about 8 after accounting for the relative weights of the baselines. • The size of the beams scales with the inverse of the observing frequency, so using the MeerKAT Ultra High Frequency (UHF-band) receivers provides a larger CB coverage than at L-band. Whilst this could have been useful for nearby, more extended targets away from the plane where the sky temperature is lower, many of those targets already have associations. Figure 2: Plot of the data ingestion rates required to tile a circular target against its angular radius. A breakdown of different observing configurations that change the size of the CBs is shown: full array versus the core, L-band versus UHF-band and whether the target is at the zenith or at an elevation of 30°. A sampling time of 153 \upmu s and 4096 channels are assumed. Thus the data limit is set by the cap of 480 on the number of CBs (see Paper I). The size of the CBs is calculated using the multibeam simulation package mosaic 2 2 2 https://github.com/wchenastro/Mosaic by WeiWei Chen for an overlap level of 0.5. footnote 2 shows the coverage supplied by using combinations of these practices. The legend is arranged approximately in order of declining sensitivity, though this strongly depends on the target’s sky position as the degradation in sensitivity at UHF-band for locations on the Galactic plane is very high due to the sky temperature contribution and also strong scattering. footnote 2 demonstrates that the best balance between CB coverage and sensitivity, for targets larger than about 5 arcmin in radius, is to use the core of the array at L-band. During the survey, we were able to push the CB coverage beyond these lines by organising blocks so that target elevations were as close as possible to MeerKAT’s horizon limit of 15°. Figure 3: The positions and types for the 124 targets searched by this survey (CCOs are excluded). A map of the targets is shown in Galactic coordinates in panel A. The footprint of the MMGPS-UHF (red upper left-lower right hatching), MMGPS-L (orange lower left-upper right hatching) and MMGPS-S (cyan crossed hatching) surveys are overlaid. Panels B and C show histograms of the longitude and latitude of the targets, respectively. Panel D shows the share of targets by type. The candidates are shown divided between the surveys in which they were first identified. 3 Second portion of the TRAPUM SNR survey 3.1 Searches since Paper I Following the targets presented in Paper I, we have searched a further 80 sources for pulsars. These observations were made under the project code SCI-20180923-MK-03. Information about the observations, properties and upper limits on the flux density and pseudo-luminosity for each target is listed in Table 1 . This brings the full list of targets 3 3 3 One of the targets of the survey was the odd radio circle J0624 - 6948 (see Paper I), which has since been confirmed to be associated with the Large Magellanic Cloud after further study ( Sasaki et al. 2025 ) . between Paper I and Table 1 to 134. Ten of these targets are CCOs so a total of 124 sources were searched for new pulsars. In Figure 3 , the positions of the targets are shown alongside their distributions in Galactic longitude, l , and latitude, b . As can be seen in panels A and B, there is a lower concentration of targets in the MMGPS survey regions, which reflects our tendency to avoid those footprints. Panel C demonstrates how concentrated our sample is on the Galactic plane (reflective of the distribution of SNRs; see Green 2025 ), and panel D shows the fractional share of each target type. Table 1: 80 targets searched for the second portion of the TRAPUM survey. Some sources have other names in the literature which are provided at the bottom of the table. As in Table 2 of Paper I, the type is defined as follows; ‘S’ for shell-type G25 SNR, ‘C’ for composite-type G25 SNR, ‘cand’ for candidate SNR, ‘U’ for unidentified TeV source. ‘?’ denotes ambiguity in the literature. Distance references are as follows: [1] Ranasinghe & Leahy 2022 , [2] Wang et al. 2007 , [3] Sofue 2023 , [4] Ranasinghe & Leahy 2018a , [5] Albert et al. 2006 , [6] Ranasinghe & Leahy 2018b , [7] Prinz & Becker 2013 , [8] Sun et al. 1999 , [9] Combi et al. 2005 , [10] Abdalla et al. 2018a , [11] Kothes & Dougherty 2007 , [12] Frail et al. 1994 , [13] Giacani et al. 2011 and [14] Karpova et al. 2016 . The sensitivity limits are for the centres of CBs as their calculation does not account for the average degradation factor which is approximately 0.65 at an of overlap level of 0.5. A handful of sources had closely spaced CBs at overlap level of 0.75. a The associated object is a CCO or candidate CCO. Source Type Associated objects Size Schedule Block ID Observation Start UTC t_{\text{obs}} N_{\text{d}} S_{\text{min}} D Dist. ref. L_{1284} limit (′) yyyy-mm-dd-hh:mm:ss (s) ( \upmu Jy) (kpc) (mJy kpc 2 ) G4.8+6.2 S G4.5+6.2 . 18 20230605 - 0014 2024 - 03 - 11 - 02:03:05 2390 44 28 G6.4 - 0.1 C . 48 20230605 - 0014 2024 - 03 - 11 - 00:42:19 2384 44 44 1.8(3) [1] 0.14(5) 6.4500–0.5583 cand G6.5 - 0.4? . 3.3 20230605 - 0014 2024 - 03 - 11 - 00:42:19 2384 44 44 G6.5 - 0.4 S 6.4500 - 0.5583? . 18 20230605 - 0014 2024 - 03 - 11 - 00:42:19 2384 44 44 3.7(2) [2] 0.60(7) G11.1 - 0.7 S . 15 20240524 - 0003 2024 - 05 - 27 - 19:25:43 2381 60 27 G11.4 - 0.1 S? . 8 20231212 - 0011 2024 - 05 - 01 - 06:33:39 2387 56 31 G16.2 - 2.7 S . 17 20241128 - 0015 2024 - 12 - 02 - 14:17:59 2388 44 28 G016.956 - 0.933 cand . 14.8 20241110 - 0005 2024 - 11 - 11 - 08:19:31 2398 44 34 G17.0 - 0.0 S . 5 20241024 - 0011 2024 - 10 - 24 - 14:56:29 2383 60 29 G17.4 - 0.1 S . 6 20240524 - 0003 2024 - 05 - 27 - 20:05:57 2395 60 28 G017.434+0.273 cand . 4.2 20241024 - 0011 2024 - 10 - 24 - 15:36:46 2386 60 28 G17.80 - 0.02 cand . 8.8 20241024 - 0011 2024 - 10 - 24 - 16:16:51 2385 60 28 G18.1 - 0.1 S . 8 20230605 - 0014 2024 - 03 - 11 - 01:22:39 2400 44 40 6.07(13) [3] 1.5(1) G018.393 - 0.816 cand . 5.2 20231212 - 0011 2024 - 05 - 01 - 04:32:13 2394 56 29 G18.8+0.3 S . 17 \times 11 20241128 - 0015 2024 - 12 - 02 - 14:58:11 2389 44 37 13.8(4) [4] 7.2(4) G18.9 - 1.1 a C? CXOU J182913.1 - 125113 33 20231219 - 0009 2023 - 12 - 19 - 15:22:33 2383 44 34 3.1(7) [2] 0.3(1) G19.75 - 0.69 cand . 26.2 20241110 - 0005 2024 - 11 - 11 - 08:59:44 2380 44 37 G19.96 - 0.33 cand . 11.8 20241128 - 0015 2024 - 12 - 02 - 15:38:23 2398 44 38 G21.0 - 0.4 S . 9 \times 7 20240524 - 0003 2024 - 05 - 27 - 20:46:13 2386 60 28 G21.6 - 0.8 S . 13 20241117 - 0009 2024 - 11 - 19 - 08:06:51 2398 44 37 G21.8 - 0.6 S . 20 20250308 - 0012 2025 - 03 - 13 - 00:43:20 2392 44 39 4.9(3) [2] 0.9(1) G021.861+0.169 cand . 5.4 20240722 - 0032 2024 - 07 - 25 - 23:21:58 2398 44 39 G022.951 - 0.311 cand . 4.8 20240722 - 0032 2024 - 07 - 25 - 22:41:26 2399 44 43 HESS J1834 - 087 U . 5.4 20241024 - 0011 2024 - 10 - 24 - 16:57:06 2397 60 32 4.1(3) [2, 5] 0.6(1) G23.85 - 0.18 cand . 5.4 20241024 - 0011 2024 - 10 - 24 - 17:37:19 2385 60 32 G024.193+0.284 cand . 4.2 20240722 - 0032 2024 - 07 - 25 - 22:01:35 2399 44 43 G24.7 - 0.6 S? . 15 20241117 - 0009 2024 - 11 - 19 - 08:47:05 2386 44 38 3.8(2) [6] 0.54(6) G26.04 - 0.42 cand . 27 20250308 - 0012 2025 - 03 - 13 - 01:23:34 2390 44 38 G26.75+0.73 cand . 10.6 20250308 - 0012 2025 - 03 - 13 - 02:43:58 2389 44 35 G27.24 - 0.14 cand . 12.2 20250308 - 0012 2025 - 03 - 13 - 02:03:48 2391 44 38 G28.21+0.02 cand . 3.4 20240524 - 0003 2024 - 05 - 27 - 21:26:31 2378 60 30 G28.22 - 0.09 cand . 6.4 20240524 - 0003 2024 - 05 - 27 - 21:26:31 2378 60 29 G28.3+0.2 cand . 14 20241110 - 0005 2024 - 11 - 11 - 10:20:11 2390 44 40 G28.33+0.06 cand . 5 20240524 - 0003 2024 - 05 - 27 - 21:26:31 2378 60 30 G028.524+0.268 cand . 6.2 20241110 - 0005 2024 - 11 - 11 - 10:20:11 2388 44 39 G28.56+0.00 cand . 1.5 20240524 - 0003 2024 - 05 - 27 - 21:26:31 2378 60 30 G28.78 - 0.44 cand . 13.2 20241128 - 0015 2024 - 12 - 02 - 16:18:37 2397 44 38 Table 2: continued Source Type Associated objects Size Schedule Block ID Observation Start UTC t_{\text{obs}} N_{\text{d}} S_{\text{min}} D Dist. ref. L_{1284} limit (′) yyyy-mm-dd-hh:mm:ss (s) ( \upmu Jy) (kpc) (mJy kpc 2 ) G028.870+0.616 cand . 4 20240722 - 0032 2024 - 07 - 25 - 21:21:55 2364 44 36 G028.877+0.241 cand . 2.6 20231212 - 0011 2024 - 05 - 01 - 05:53:17 2387 56 31 G028.929+0.254 cand . 4.4 20231212 - 0011 2024 - 05 - 01 - 05:53:17 2387 56 31 G28.92+0.26 cand . 6.4 20231212 - 0011 2024 - 05 - 01 - 05:53:17 2387 56 31 G029.329+0.280 cand . 5 20231212 - 0011 2024 - 05 - 01 - 05:53:17 2387 56 30 G29.92+0.21 cand . 4.2 20241024 - 0011 2024 - 10 - 24 - 18:17:32 2393 60 30 G261.9+5.5 S . 40 \times 30 20250212 - 0011 2025 - 02 - 14 - 18:38:29 2343 44 40 G296.6 - 0.4 cand . 14 \times 10 20250212 - 0011 2025 - 02 - 14 - 19:18:31 2385 44 29 G296.7 - 0.9 S . 15 \times 8 20241117 - 0009 2024 - 11 - 19 - 05:25:20 2383 44 29 10(1) [7] 3(1) G296.8 - 0.3 a S 2XMMi J115836.1 - 623516 20 \times 14 20250212 - 0011 2025 - 02 - 14 - 20:39:04 2396 44 29 9.6(6) [1] 2.6(3) G299.6 - 0.5 S . 13 20231219 - 0009 2023 - 12 - 19 - 12:01:18 2370 44 29 G302.3+0.7 S . 17 20250212 - 0011 2025 - 02 - 14 - 19:58:47 2382 44 29 G304.6+0.1 (i) S . 8 20231212 - 0011 2024 - 05 - 01 - 03:09:15 2371 56 27 7.9(6) [1] 1.7(3) G310.6 - 0.3 (ii) S . 8 20230605 - 0014 2024 - 03 - 10 - 22:01:17 2371 44 36 G310.8 - 0.4 (iii) S . 12 20230605 - 0014 2024 - 03 - 10 - 22:01:17 2371 44 36 G318.2+0.1 S HESS J1457 - 593? . 40 \times 35 20231219 - 0009 2023 - 12 - 19 - 12:41:16 2383 44 32 2.7(4) [1] 0.23(7) G317.3 - 0.2 S . 11 20231219 - 0009 2023 - 12 - 19 - 13:21:27 2391 44 34 HESS J1554 - 550 (iv) C G327.1 - 1.1 . 0 20240524 - 0003 2024 - 05 - 27 - 18:45:03 2392 60 25 \sim 9 [8] \sim 2 G337.3+1.0 ? . 15 \times 12 20241117 - 0009 2024 - 11 - 19 - 06:05:45 2393 44 35 G324.1+0.0 cand . 11 \times 7 20250212 - 0011 2025 - 02 - 14 - 21:19:42 2392 44 32 G326.3 - 1.8 C PWN G326.12 - 1.81 . 38 20240524 - 0003 2024 - 05 - 27 - 18:04:50 2395 60 23 3.5(6) [1] 0.3(1) HESS J1626 - 490 U G335.2+0.1 . 6 20231219 - 0009 2023 - 12 - 19 - 14:01:47 2385 44 39 G336.7 - 0.3 cand . 5 \times 3 20231212 - 0011 2024 - 05 - 01 - 03:49:33 2390 56 36 G336.7+0.5 S . 14 \times 10 20230605 - 0014 2024 - 03 - 10 - 22:41:23 2394 44 40 HESS J1634 - 472 U G337.2+0.1 . 6.6 20231219 - 0009 2023 - 12 - 19 - 14:42:01 2396 44 45 >13.5 [9, 10] >8.3 G340.6+0.3 S . 6 20250130 - 0011 2025 - 02 - 05 - 01:36:29 2370 44 37 \sim 15 [11] \sim 8 G340.4+0.4 S . 10 \times 7 20250130 - 0011 2025 - 02 - 05 - 01:36:29 2370 44 37 G341.9 - 0.3 S . 7 20241128 - 0015 2024 - 12 - 02 - 12:55:07 2374 44 38 15.8(6) [1] 9.4(7) G342.0 - 0.2 S . 12 \times 9 20241128 - 0015 2024 - 12 - 02 - 12:55:07 2376 44 38 15.8(6) [1] 9.5(7) G342.1+0.9 S . 10 \times 9 20250130 - 0011 2025 - 02 - 05 - 02:16:21 2391 44 34 \sim 6.9 [12] \sim 1.6 HESS J1702 - 420 U . 3.6 20241117 - 0009 2024 - 11 - 19 - 06:46:01 2395 44 38 G344.7 - 0.1 a C? CXOU J170357.8 - 414302 8 20240612 - 0011 2024 - 10 - 24 - 13:41:12 1577 60 34 6.3(1) [13] 1.33(4) G345.1 - 0.2 S . 6 \times 5 20240612 - 0011 2024 - 10 - 24 - 13:41:12 1577 60 34 G348.32 - 0.73 † cand G348.324 - 00.735 . 7 20240722 - 0032 2024 - 07 - 26 - 00:02:28 2389 44 37 G350.0 - 2.0 a S 1RXS J172653.4 - 382157 . 45 20230605 - 0014 2024 - 03 - 10 - 23:21:45 2388 44 31 \sim 3 [14] \sim 0.3 G351.0 - 0.6 cand . 12 20241128 - 0015 2024 - 12 - 02 - 13:35:08 2393 44 35 G351.2+0.1 C? . 7 20231212 - 0011 2024 - 05 - 01 - 05:12:43 2391 56 33 G351.4+0.2 cand . 18 \times 14 20241110 - 0005 2024 - 11 - 11 - 06:59:14 2359 44 42 G351.4+0.4 cand . 9 20241110 - 0005 2024 - 11 - 11 - 06:59:14 2364 44 43 G353.9 - 2.0 S . 13 20250308 - 0012 2025 - 03 - 13 - 00:03:13 2374 44 33 G355.4+2.7 cand . 22 20241110 - 0005 2024 - 11 - 11 - 07:39:01 2388 44 31 G355.9 - 2.5 S . 13 20241117 - 0009 2024 - 11 - 19 - 07:26:22 2393 44 30 G356.3 - 0.3 S . 11 \times 7 20230605 - 0014 2024 - 03 - 11 - 00:02:00 2398 44 38 (i) Kes 17 (ii) Kes 20B (iii) Kes 20A (iv) The Snail † The SNR candidate G348.32 - 0.73 identified by the SARAO MeerKAT Galactic Plane Survey ( Goedhart et al. 2024 ; Anderson et al. 2025 , SMGPS;) was added to the list later on, as it was noted as having a central PWN-like feature. Over half the target list consists of SNRs in the latest version of the online SNR catalogue 4 4 4 Green D. A., 2024, ‘A Catalogue of Galactic Supernova Remnants (2024 October version)’, Cavendish Laboratory, Cambridge, United Kingdom (available at https://www.mrao.cam.ac.uk/surveys/snrs/ ) ( Green 2025 , ; henceforth G25) . Eleven targets are unidentified TeV sources, which are presented in more detail in § 3.3.2 . A plurality of the candidate SNRs are from the THOR and GLOSTAR surveys. These surveys reported the largest number of new candidates. However, they also generally reported smaller shells than the more extended GLEAM or MGPS-2 discoveries. This is linked to the bias mentioned in § 2.2 that less extended targets are easier to fully tile with CBs under our observing setup. These small-diameter candidate SNRs comprise a significant portion of the survey, especially in the region 15^{\circ}0.17\times 10^{12}\,\text{G\,s}^{-2} (5) in order to be active. If it is not, the pulsar is rejected. For each pulsar, the beaming fraction, f_{b} , is calculated using the empirical relation derived by Tauris & Manchester 1998 : f_{b}=0.03+0.09\left[\text{log}_{10}\left(\frac{P}{1\,\text{s}}\right)-1\right]^{2}. (6) This is done by choosing a value, x , from a uniform distribution where 01 as the smearing threshold. In order to not interfere with the normalisation, we keep this value here. However, in § 4.1.3 , a more realistic cut of 0.75 is set when calculating TRAPUM S/N values. Having established that our model populations can replicate the PMPS and HTRU populations, we now run the full populations through a routine that mimics the TRAPUM targeted survey. Figure 6: Histograms showing the properties of 100 synthesised populations after applying an age cut of 100 kyr. On average, there are 3398 young pulsars, of which 73 are detected by PMPS and HTRU. 21 per cent of all young pulsars are beaming towards the Earth and of those, an average of 334 are too smeared to be seen by the two surveys. Each histogram has been overlaid with a normal distribution function described by the mean, \mu , and standard deviation, \sigma , of the results. 4.1.3 Replicating the TRAPUM survey Figure 7: P-\dot{P} diagram for approximately 3000 synthetic young (real age <100 kyr) pulsars produced by the evolve method under the chosen models and parameter distributions listed in Table 3 . The colours of the markers correspond to the spin-down power. The blue solid line shows the 100 kyr characteristic age contour. The input models for n , \chi and B_{0} leads to a spread in \dot{P} that for many pulsars results in characteristic ages much larger than their real ages. Figure 8: Example of a distribution of the young pulsars in Galactic coordinates ( left ) and looking down on the Galaxy in Galactocentric coordinates ( right ) after applying the sky cut that accounts for the TRAPUM survey footprint. The first step of this routine 17 17 17 The routine is performed by process_population.py , which is available in the repository github.com/jamesdturner/PsrPopPy2_trapumSNR by James Turner, a forked version of P sr P op P y 2 by Devansh Agarwal. , which is shown in the bottom strip of the flowchart in Figure 5 , is to cut all pulsars older than a SNR dissipation age of 100 kyr from the population. As was seen in Figure 6 , this leaves 3398\pm 119 pulsars. This range overlaps with the expected number of 2000-3500 Galactic SNRs extrapolated from a SN rate of 2.8\pm 0.6 ( Vink 2020 , page 36) . However, it is above an estimate of between 1600-1900 from Galactic CCSN rates over 10^{5}\,\text{yr} ( 1.63\pm 0.46 century -1 , Rozwadowska et al. 2021 ; 1.9\pm 1.1 century -1 , Diehl et al. 2006 ), though is within those propagated uncertainties. This difference could be due to the uniform age distribution or perhaps the choice of spin-down model (see § 4.3.4 ) assumed for our population synthesis producing more young pulsars than exist. This could be due to e.g., a recent drop in the Galactic star formation rate, although there is strong evidence that the rate has not changed significantly over the lifetime of massive stars (10-100 Myr) ( Soler et al. 2023 , e.g.) . Ultimately, we are satisfied with these numbers and do not attempt to infer any astrophysics from them beyond stating that our full population appears to be appropriately normalised. The young pulsars from one of the populations are shown on the P-\dot{P} diagram in Figure 7 . The spread in \dot{P} values is attributed to the spread in B_{0} and \chi , especially for pulsars with braking indices near a value of 2. We now introduce a different model of the birth velocities of the pulsars. At ages below 100 kyr, the total displacement since birth is negligible compared to the size of the Galaxy, thus the effect of the prior velocity vector is not concerning. We did not change the default velocity model for P sr P op P y 2/ evolve because it may have significantly changed the distribution of the older evolved pulsars. This could have affected the normalisation, and we wanted to be consistent with previous use of P sr P op P y . We deemed it was better to simply resample after the age cut instead. Each pulsar is reassigned a birth velocity value from the Maxwellian distribution derived by Hobbs et al. 2005 , where the 3D velocity vector has a mean of 400 km s -1 and root mean square (rms) of 265 km s -1 . The pulsars are then assigned a SNR that is offset from the position of the pulsar by an amount -\vec{v}_{\text{birth}}t . The size of the remnant is determined by the pulsar’s age, using the model described in § 4.1.4 . In the real survey, CBs were placed up to the edge of the shell, so if a pulsar has a sufficient transverse velocity component it will escape the shell and not be covered by the search area. If the tangential velocity is small, the pulsar may still be moving very quickly along the line of sight, but will still have been tiled by a coherent beam. At this stage, a flag is given if the pulsar is projected onto its SNR as seen from the Earth, which is taken to be at a position (0, 8.5, 0.006) kpc in Galactocentric coordinates. We found that approximately 9 per cent of pulsars that have left their remnants still appear to be projected onto them. The penultimate step is to select the pulsars that will be searched by TRAPUM. A 2D sky cut selects the region |b|\leq 2 °, -100 ° 95 per cent of targets were concentrated. The distribution of the pulsars after this cut is shown in Figure 8 . Then, a radial sky cut is performed by excluding remnants with an angular diameter smaller than the smallest in the real sample (the candidate G28.56+0.00 from THOR, Anderson et al. 2017 at 1.5 arcmin across). This removes very distant SNRs that we would not realistically have searched in the real survey. We searched a total of 119 shell regions 18 18 18 A figure of 119 is found by subtracting 10 CCOs and 5 TeV sources where there was no SNR from the total of 134. , so 119 pulsar-SNR pairs are sampled quasi-randomly; targets are twice as likely to be selected from l>15 °to reflect our general preference against targets in the MMGPS footprints. In the real survey, we switched between two sampling times and either the core and full array configuration were used depending on the requirements of the search block. The search parameters assigned to the synthetic pulsars therefore match those ratios. Finally, the S/N values at 1284 MHz are calculated. We obtain pulse widths using equation 8 and using the modified radiometer equation ( Dewey et al. 1985 ; also Paper I, equation 1). We include a FFT sensitivity loss of 0.7, and also the degradation factor of 0.65 which corresponds to the average coherent sensitivity across the CB tiling. Pulsars are then flagged as detected by TRAPUM if they are beaming towards Earth, they are projected onto the SNR, the duty cycle is below 75 per cent and the S/N is above the spectral threshold of 9. The results are saved, and the process is repeated 100 times. 4.1.4 Supernova remnant evolution model In this model, a supernova remnant is a spherical boundary that expands in time, t , according to a radius function R(t) . All pulsars are assigned a remnant retrospectively; the pulsar’s real age and resampled 3D velocity is used to place a SNR at the birth site. As mentioned above, we choose the dissipation time for a SNR to be 100 kyr, which is why we cut pulsars older than this from the analysis. Phase 1 (the ejecta-dominated phase) ends when the ejecta mass is approximately equal to the swept up mass. It can be shown that the radius at the end of phase 1 ( Bamba & Williams 2022 , equation 3) is R_{1}=5.8\left(\frac{M_{\text{ejecta}}}{10\,\text{M}_{\odot}}\right)^{\frac{1}{3}}\left(\frac{n_{0}}{0.5\,\text{cm}^{-3}}\right)^{-\frac{1}{3}}(\text{pc}), (9) where n_{0} is the number density of the surrounding medium. During this time, we model the growth as entirely freely expanding, such that the radius expands at a constant rate: v_{\text{exp}}=\sqrt{\frac{2E_{SN}}{M_{\text{ejecta}}}}, (10) where E_{\text{SN}} is the kinetic energy and M_{\text{ejecta}} is the mass of the ejecta released by the explosion. Conventional values are adopted for the ejecta mass of 1 M ⊙ and energy of 10 51 erg. n_{0} is set to be 1 cm -3 across the Galaxy. A pulsar’s SNR size is therefore only dependent on its age. Phase 1 is followed by the Sedov-Taylor phase (phase 2). In our modelling, the SNR continues to expand with time as t^{\frac{2}{5}} until dissipation. The chosen radius function is therefore: R(t)=\begin{cases}v_{\text{exp}}t,&\text{if}\,\,v_{\text{exp}}t\leq R_{1}.\\ 5.0\left(\frac{n_{0}}{1\,\text{cm}^{-3}}\right)^{-\frac{1}{5}}\left(\frac{E_{\text{SN}}}{10^{51}\,\text{erg}}\right)^{\frac{1}{5}}\left(\frac{t}{1\,\text{kyr}}\right)^{\frac{2}{5}}-\Delta R,&\text{otherwise}.\end{cases} (11) where the second expression is equation 7 from Bamba & Williams 2022 and \Delta R\approx 0.53 pc that accounts for the discontinuity between the two expressions. This radius function is shown in Figure 9 where it is compared to different values of E_{\text{SN}} , M_{\text{ejecta}} and n_{0} . Our choice of these values results in a sample of SNRs that represent the approximate average size. However, not including a transition to the Snowplough phase (phase 3), where the remnant expands as R(t)\propto t^{\frac{1}{5}} results in a sample of larger remnants. This is amplified by setting n_{0} to be a constant low value as Figure 9 shows, as we know the density can be much higher in the Galactic plane. The effect that this bias has on our analysis is discussed in § 4.3 . Figure 9: Plot showing the growth of a SNR during phase 1 and phase 2 for different explosion energies, CSM/ISM number densities and ejecta masses. The chosen model is the solid line, which lies approximately between the different models. 4.2 Results Figure 10: Histograms for the results of the TRAPUM SNR survey routine applied to 100 synthetic young pulsar populations. As with Figure 6 , the red solid lines are Gaussian functions plotted for the mean and standard deviation of each distribution. The results of the synthetic TRAPUM SNR survey routine iterated through 100 simulated populations are shown in Figure 10 , and also listed in Table 4 . The upper panel shows the prediction of 2\pm 2 new pulsars. Reminding ourselves that PSR J1818 - 1502 is a serendipitously discovered old foreground pulsar, the real number of our discoveries is one. This scenario represents 18 of the 100 outcomes. The lower panels break down the non-detections. Panel A shows that 96\pm 5 of the pulsars would be beaming away (consistent with the beaming fraction in Figure 6 ), panel B shows that on average 5\pm 2 pulsars are not projected onto their SNR. Then panel C shows that 12\pm 4 pulsars are smeared below the S/N threshold by propagation effects and finally panel D shows 4\pm 2 of the left over pulsars were too faint to be seen. As a quick sanity check we can sum the mean of all five panels, which returns 119 as it should. Of the smeared pulsars, we recalculated their S/N values again for zero smearing, and find that 6\pm 2 have sufficient flux densities to be detected. This demonstrates that, in these simulations, propagation effects actually explain 6 out of 17 pulsars that are beamed towards the Earth (henceforth ‘beaming pulsars’) being missed. For the real survey, we chose to keep the number of frequency channels at the maximum of 4096. As was explained in § 2.2 , this required trading off either the number of coherent beams or the sampling time, but had the benefit of ensuring that DM smearing was not significant at periods above 30 ms even at very high DMs of over 1000 pc cm -3 . This, however, does not provide any defences against the smearing due to scattering. Scattering is therefore the dominant contribution to pulse broadening for all realistic DMs under the parameters of both the real and simulated surveys. Nevertheless, starting from a position of 23 or 24 potentially detectable pulsars (based on the \sim 21 per cent beaming fraction), we find that the combined non-detections due to locations outside SNR shells, propagation effects and luminosity are able to explain our low discovery rate with a probability of 18 per cent. In order to increase the number of discoveries, the observing frequency could be increased. The mean pulsar spectral index is negative, so the integration time would also need to be increased. To investigate this further, we reran the prescription outlined above but with survey parameters of an equivalent survey conducted with the MeerKAT S-band receivers (centred on 2406.25 MHz). The CBs are smaller at S-band compared to L-band for an equivalent overlap level, so we reduced the number of frequency channels from 4096 to 1024. This would be a realistic requirement in order to decrease the data rate and allow for more CBs. The ISM, pulsar and receiver parameters that scale with frequency were converted using a spectral index of - 1.6. The result was a decrease of 5 in the number of smeared pulsars. However, the number of discoveries only increased to 3\pm 2 as the number of faint pulsars increased by 4. These results are provided in Table 4 against the results of the simulations at L-band. In order to match the flux density limits at L-band, the S-band integration times would need to increase by a factor of about 3.4. We recalculated the S/N values of the pulsars with T_{\text{obs}}\times 3.4 , which resulted in a further 2 faint pulsars becoming detectable. This is also shown in Table 4 . In the next section, we compare the properties of the simulated and real samples, discuss the short-comings of the analysis in more detail, and explore what these results could be telling us about the properties of young pulsars in the Galaxy. Table 4: The results of the synthetic TRAPUM survey showing the magnitude of the selection effects at L - band, S - band and at S - band with S/N values recalculated for a factor 3.4 increase in the observing time. Selection Effect L - band S - band S - band, T_{\rm{obs}}\times 3.4 Beaming away 96\pm 5 95\pm 5 95\pm 5 Appears outside SNR 5\pm 2 6\pm 2 6\pm 2 Too smeared 12\pm 4 7\pm 2 7\pm 2 Too faint 4\pm 2 8\pm 3 6\pm 3 Number of discoveries 2\pm 2 3\pm 2 5\pm 2 4.3 Discussion in the context of the survey The variance of the results is notable. In fact, the number of predicted discoveries is consistent with zero at 2\sigma . While 100 iterations is relatively small for a Monte Carlo approach, further iterations may not change the overall picture given that we are interested in integer numbers of pulsars. The benefit of using P sr P op P y 2/ evolve was that real pulsar ages are considered, which could then be used to model the SNRs. However, the extra physics that P sr P op P y 2 requires for a cohesively modelled and normalised population of pulsars could have injected too much complexity and inflated the final uncertainty. Future modelling of pulsar-SNR populations might use a more selective approach to how pulsars are modelled. We now consider individually some specific results and possible biases and limitations of this work. 4.3.1 Distances At the start of this section, we singled out the distance as a limitation of population studies when translating between the properties of a real and synthetic pulsar population. For the majority of real pulsars, the most accurate estimate of their distance is derived from n_{\text{e}} models, with realistic uncertainties of 30 percent, maybe even 50 per cent on the Galactic plane ( Deller et al. 2009 ; Yao et al. 2017 ) . This is a similar, if not worse, issue for SNRs as distances are arduous to measure with high precision. In Figure 11 , we show the best positions for the young pulsar and SNR populations in the Galaxy. They can both be seen to cluster along the spiral arms and in the central bulge. The vast majority of these known sources are located on the near side of the Galaxy, as is expected given that the detectability of both objects degrades significantly with distance. In P sr P op P y 2/ evolve , we populated the Galactic disk with young pulsars and SNRs uniformly up to a radius of 15 kpc. Our angular diameter cut at 1.5 arcmin filtered out \lesssim 1 per cent of supernova remnants located on the far side of the Galaxy. This resulted in an average distance of 11.5 kpc to the pulsar-SNR pairs. The real average for all known SNRs in Figure 11 is 5.2 kpc. But for those that we targeted, we find the average is 7.5 kpc for the 40 per cent that have measured distances. For known young ( \tau_{\text{c}}<100 kyr) pulsars, the average distance is 8.3 kpc assuming the ne 2001 model. Both are smaller than for what we simulated, so therein lies a bias towards apparently faint and highly scattered pulsars in the simulation. Our real sample of SNRs was chosen to be as diverse as possible, and as a result contains a large number of candidate supernova remnants which have no distance measurements. Can we infer anything about the distances of the candidate supernova remnants in our sample? It is non-trivial to infer the distance to a SNR based on its flux density or surface brightness, the latter of which is independent of the distance. Instead, the brightness can be loosely connected to the size of the remnant by using the surface brightness-diameter relation ( Clark & Caswell 1976 ) . These variables do show evidence of a correlation ( Green 2025 , e.g.) as would be expected if the radio brightness of the shell fades over time. The scatter around the correlation however makes it an extremely unreliable tool for estimating distances. With this being said, applying the basic relation to the fluxes and angular sizes of all the candidates discovered by the GLOSTAR ( Dokara et al. 2021 ) and THOR ( Anderson et al. 2017 ) surveys, we find those targets are on average located at larger distances compared to those in Figure 11 . On the other hand, the candidates from GLEAM ( Hurley-Walker et al. 2019b ) do not appear to be further away. It could be that we have sampled pulsar locations on the other side of the Galaxy by choosing to observe many GLOSTAR and THOR candidates, but it is highly unlikely that we sampled the disk up to 15 kpc as uniformly as in the simulation. In ensuing sections, we discuss how the discrepancy between the distances of the real and synthetic pulsar-SNR populations has affected other modelled properties of our pulsar-SNR populations. Figure 11: Positions of known pulsars (dots) in the Milky Way shown against the ne 2001 model. Young pulsars ( \tau_{\text{c}}<100 kyr) are highlighted as larger red dots. The positions of known supernova remnants that have distance measurements are shown as blue lines corresponding to the most up to date uncertainty or range to have been reported in the literature. The main survey region |b|\leq 2 °, -100 ° 1 TeV detected by H.E.S.S. at a resolution of 0.1°( Abdalla et al. 2018a ; available to download at https://www.mpi-hd.mpg.de/hfm/HESS/hgps/ ). Magenta circles or ellipses mark the boundaries of supernova remnants; solid lines are G25 SNRs and dashed lines are candidate SNRs from the literature. Beams may appear offset from SNR boundaries due to the centroid position precision provided by G25; beam tilings have been centred on more precise SNR positions. Red circles are the locations of pulsars from the ATNF pulsar catalogue v2.5.1 ( Manchester et al. 2005 ) . Coloured contours trace the significance levels of radio components in the field: for HESS J1457 - 593, HESS J1554 - 550 and HESS J1641 - 463 data are from the Molonglo Observatory Synthesis Telescope (MOST) Supernova Remnant Catalogue at 843 MHz ( Whiteoak 1996 ) , for HESS J1818 - 154 and HESS J1834 - 087 data are from the GaLactic and Extragalactic All-sky MWA survey ( Hurley-Walker et al. 2019a , GLEAM;) at 170-231 MHz, and finally HESS J1844 - 030 is overlaid with data from the 1.4 GHz NRAO VLA Sky Survey ( Condon et al. 1998 , NVSS;) . Appendix C RN+F2 fit with run_enterprise The corner plot produced by enterprise for the white noise, red noise and \ddot{f} parameters of PSR J1831 - 0941 is provided in Figure 19 . Figure 19: Corner plot for the fit of white and red noise parameters and F2 to the PSR J1831 - 0941 TOAs and ephemeris in Table 5 . Labels from left to right along x-axis: efac and equad for L-band data captured with the APSUSE back end, efac and equad for UHF-band data captured with the PTUSE back end, efac and equad for UWL data captured with the Medusa back end, red noise amplitude, red noise spectral index and \ddot{f} . Plot has been made using the chainconsumer package ( Hinton 2016 ) .