pith. sign in

arxiv: 2605.26347 · v1 · pith:ABAWGAUTnew · submitted 2026-05-25 · 🌌 astro-ph.IM

A distributed resource-adaptive implementation of the widefield radio-interferometric measurement model for scalable image formation

Arwa Dabbech , Yves Wiaux This is my paper

Pith reviewed 2026-06-29 20:09 UTC · model grok-4.3

classification 🌌 astro-ph.IM
keywords radio interferometrywidefield imagingw-stackingw-projectiondistributed computingmeasurement modelimage formationNUFFT
0
0 comments X

The pith

A hybrid w-stacking/w-projection approach with automated bin selection enables distributed widefield radio interferometric imaging under memory constraints.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a distributed and resource-adaptive implementation of the widefield measurement model in radio interferometry. It relies on a hybrid decomposition that combines w-stacking and w-projection, with the number of w-bins chosen automatically to reduce computational cost while respecting system memory limits. The model is broken into low-dimensional per-bin operators, residual w-offsets are absorbed into Fourier kernels, and an optional holographic matrix can reduce data dimensionality when needed. Further parallelization comes from block decomposition of the sparse matrices under memory controls. The approach supports repeated application of the measurement operator and its adjoint for image formation on large wideband datasets.

Core claim

The widefield measurement model can be realized as a distributed resource-adaptive operator through a hybrid w-stacking/w-projection decomposition in which the number of w-bins is selected automatically to minimize computational cost subject to memory constraints, with residual offsets handled by measurement-specific Fourier kernels augmenting the NUFFT de-gridding matrix and optional holographic encoding of the composite operator.

What carries the argument

The hybrid w-stacking/w-projection decomposition with automated w-bin selection that partitions the measurement model into low-dimensional per-bin operators.

If this is right

  • The measurement model decomposes into independent low-dimensional operators per w-bin that can run in parallel across distributed compute nodes.
  • Optional holographic matrices jointly encode the forward and adjoint operations to cut memory use when data volumes are large.
  • Sparse de-gridding matrices can be further split into Fourier-partitioned blocks for additional parallel scaling.
  • The same framework applies to both monochromatic and wideband imaging without case-by-case redesign.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could be combined with existing iterative solvers that rely on repeated Phi dagger Phi applications to reach larger fields of view on fixed hardware.
  • Automated bin selection might generalize to other position-dependent operators in interferometry or tomography where memory is the binding constraint.
  • Block decomposition of the matrices opens a route to real-time or streaming imaging pipelines if the per-block cost stays below latency thresholds.

Load-bearing premise

The hybrid decomposition with automated bin selection preserves enough accuracy in the position-dependent PSF for intended imaging uses without manual tuning or extra error controls.

What would settle it

A side-by-side comparison on widefield MeerKAT or ASKAP data in which the automated hybrid model produces image artifacts or flux errors exceeding those of a manually tuned reference implementation.

Figures

Figures reproduced from arXiv: 2605.26347 by Arwa Dabbech, Yves Wiaux.

Figure 1
Figure 1. Figure 1: Overview of the approach to obtain the band-lim￾ited Fourier w-kernels. From left to right: (a) real part of a low-resolution phase modulation, evaluated directly in the image domain over twice the FoV of interest, and tapered via a Hamming window, (b) magnitude of its FFT; (c) magni￾tude of the w-kernel after compression via hard-thresholding. The gist of w-stacking/w-projection approach is to dis￾cretise… view at source ↗
Figure 2
Figure 2. Figure 2 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Φ †Φ implementation workflow. Resource planning is conducted to determine the operator’s computation strategy and associated number of w-layers P (stage 1). If Φ †Φ is precomputed, memory-controlled Fourier partitioning is performed (stage 2A), followed by precomputation of the underpinning sparse (de-gridding or holographic) matrix blocks (stage 3). The resulting operator is then applied repeatedly within… view at source ↗
Figure 4
Figure 4. Figure 4: Application of the operator Φ †Φ to an input image x hosted by the master worker. Two levels of decomposition are considered for full parallelisation. The first level arises from the w-stacking, whereby P FFT workers–corresponding to the w-layers–perform Fourier transforms (specifically, the operators FZp). The second level, enabled by a memory-controlled Fourier partitioning, distributes de-gridding/gridd… view at source ↗
Figure 5
Figure 5. Figure 5: Compute resource planning: Estimated computational complexity of Φ †Φ (c[Φ†Φ] ; right y-axis) and the memory required to store the underpinning sparse matrix G (mG; left y-axis) as a function of the number of w-layers (P). Dashed horizontal lines indicate the system’s memory budget (m⋆ ) for varying number of CPU nodes ranging from 1 to 10. Vertical lines (in blue and green) indicate the corresponding opti… view at source ↗
Figure 6
Figure 6. Figure 6: Compute resource planning as a function of CPU node budget, equivalently expressed in CPU core budget (1 CPU node comprises 36 cores). The selected number of w-layers P in each setting is shown. The coloured regions in￾dicate the strategy for computing Φ †Φ (see [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Computational cost of building Φ †Φ, specifically its underpinning sparse matrix (including Fourier partition￾ing and precomputation), and its application within an it￾erative imaging algorithm, in CPU hour. Results are shown for two parameter choices for the implementation of the pre￾computed Φ †Φ: (i) using G with P = 23, the closest value to the minimiser of the computational complexity, and (ii) using … view at source ↗
Figure 8
Figure 8. Figure 8: Breakdown of the computational and commu￾nication time associated with the application of Φ †Φ to an image estimate x as a function of the number of w-layers P, in seconds (sec). Results are shown for three implementa￾tions of the precomputed Φ †Φ: (i) using G with P = 23, the closest value to the minimiser of the computational complex￾ity, (ii) using G with P = P WSClean = 126, representing an upper bound… view at source ↗
Figure 9
Figure 9. Figure 9: The dirty image (a) of the field SB9442-35 at the frequency 1069 MHz and reconstructed images obtained using WSClean (b) and the four imaging algorithms supported by our proposed framework: HyperAIRI (c), Hyper-uSARA (d), AIRI (e), uSARA (f). Each panel is overlaid with zooms on selected regions: zoom (i) is centred on the “dancing ghosts”; zooms (ii) and (iv) show regions with both extended and compact ra… view at source ↗
read the original abstract

Modern image formation algorithms in radio interferometry rely on repeated applications of the operator {\Phi} modelling the measurement process and its adjoint {Phi^\dagger} to enforce consistency with the acquired data, specifically via their composite mapping {Phi^\dagger\Phi} encoding the array's point spread function (PSF). The large data volumes produced during wideband observations yield significant computational challenges for image formation. Moreover, for widefield imaging, the baseline components along the line of sight w complicate severely the measurement model beyond the conventional 2-dimensional non-uniform Fourier transform (NUFFT), making the PSF highly position-dependent. We propose a distributed resource-adaptive implementation of the widefield measurement model, enabled by a hybrid w-stacking/w-projection approach, whereby the number of w-bins is set in a fully automated manner to minimise the computational cost under the compute system's memory constraints. The resulting measurement model is naturally decomposed and distributed into low-dimensional operators specific to w-bins. Residual w-offsets are integrated as measurement-specific Fourier kernels augmenting the sparse de-gridding matrix of the basic NUFFT model. An optional data dimensionality reduction is also introduced, jointly encoding the sequential Fourier de-gridding/gridding operations in {Phi^\dagger\Phi} into a holographic matrix when required by memory constraints. For further parallelisation, the sparse de-gridding or holographic matrices are decomposed into blocks via memory-controlled Fourier partitioning. The approach has been validated in prior works through real data case studies for both monochromatic and wideband imaging of MeerKAT and ASKAP data. We provide herein a thorough analysis of its computational efficiency using simulated MeerKAT data. A MATLAB implementation is available in BASPLib.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper proposes a distributed resource-adaptive implementation of the widefield radio-interferometric measurement model Φ and its adjoint, using a hybrid w-stacking/w-projection decomposition. The number of w-bins is chosen automatically to minimize cost subject to memory limits; residual w-offsets are absorbed into measurement-specific Fourier kernels augmenting the NUFFT de-gridding matrix. An optional holographic matrix encodes Φ†Φ for dimensionality reduction, and sparse matrices are further block-decomposed via memory-controlled Fourier partitioning. Computational efficiency is demonstrated on simulated MeerKAT data; accuracy is asserted via prior real-data MeerKAT/ASKAP case studies. A MATLAB implementation (BASPLib) is supplied.

Significance. If the automated bin-selection heuristic preserves position-dependent PSF fidelity, the method supplies a practical, hardware-aware route to scalable widefield imaging that reduces manual tuning and supports distribution across memory-constrained nodes. The hybrid decomposition, residual-kernel augmentation, and optional holographic reduction are concrete engineering contributions for next-generation arrays.

major comments (1)
  1. [Validation and results sections] The central claim that automated w-bin selection (under memory constraints) preserves sufficient position-dependent PSF accuracy without additional error controls or manual tuning is load-bearing, yet the manuscript supplies no new quantitative PSF error maps, sidelobe level comparisons, or sensitivity tests of the heuristic itself. Accuracy is referenced only to prior real-data studies; the simulated MeerKAT efficiency analysis does not address this gap.
minor comments (1)
  1. [Abstract] The abstract and introduction would benefit from a concise statement of the precise memory and accuracy trade-off metric used to set the automated bin count.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive review and for recognizing the practical engineering contributions of the hybrid decomposition and resource-adaptive implementation. We respond to the single major comment below.

read point-by-point responses

  1. Referee: [Validation and results sections] The central claim that automated w-bin selection (under memory constraints) preserves sufficient position-dependent PSF accuracy without additional error controls or manual tuning is load-bearing, yet the manuscript supplies no new quantitative PSF error maps, sidelobe level comparisons, or sensitivity tests of the heuristic itself. Accuracy is referenced only to prior real-data studies; the simulated MeerKAT efficiency analysis does not address this gap.

    Authors: We agree that the manuscript would be strengthened by direct quantitative assessment of the automated w-bin selection heuristic on the simulated MeerKAT data. The accuracy of the hybrid w-stacking/w-projection model with residual-kernel augmentation is supported by the cited prior real-data validations on MeerKAT and ASKAP, which employed the same operator construction. The present work centres on computational efficiency and distribution under memory constraints. In the revised version we will add PSF error maps, sidelobe comparisons, and sensitivity tests of the bin-selection procedure applied to the simulated dataset to close this gap. revision: yes

Circularity Check

0 steps flagged

Minor self-citation for prior accuracy validation; no reduction of claims to fitted inputs or self-referential definitions

full rationale

The manuscript presents a hybrid w-stacking/w-projection implementation with automated bin selection and optional holographic matrices, validated for computational efficiency on simulated MeerKAT data. Accuracy preservation is referenced to prior real-data case studies rather than derived or fitted within this work. No equations, parameters, or predictions reduce by construction to the inputs (no self-definitional loops or fitted-input-called-prediction patterns). The self-citation is not load-bearing for the efficiency analysis, which stands on independent simulated benchmarks. This qualifies as a normal minor self-citation (score 2) with the derivation remaining self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the contribution is described as an algorithmic implementation rather than a derivation introducing new constants or entities.

pith-pipeline@v0.9.1-grok · 5843 in / 1137 out tokens · 28833 ms · 2026-06-29T20:09:41.981622+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 42 canonical work pages · 1 internal anchor

  1. [1]

    2025, ApJS, 280, 63, doi: 10.3847/1538-4365/adfbed

    Wiaux, Y. 2025, ApJS, 280, 63, doi: 10.3847/1538-4365/adfbed

    work page doi:10.3847/1538-4365/adfbed 2025

  2. [2]

    2024, ApJS, 273, 3, doi: 10.3847/1538-4365/ad46f5

    Aghabiglou, A., San Chu, C., Dabbech, A., & Wiaux, Y. 2024, ApJS, 273, 3, doi: 10.3847/1538-4365/ad46f5

    work page doi:10.3847/1538-4365/ad46f5 2024

  3. [3]

    Arras, P., Reinecke, M., Westermann, R., & Enßlin, T. A. 2021, A&A, 646, A58, doi: 10.1051/0004-6361/202039723

    work page doi:10.1051/0004-6361/202039723 2021

  4. [4]

    2013, ApJ, 770, 91

    Bhatnagar, S., Rau, U., & Golap, K. 2013, ApJ, 770, 91

    2013

  5. [5]

    2019, MNRAS, 492, 3509, doi: 10.1093/mnras/stz3555

    Birdi, J., Repetti, A., & Wiaux, Y. 2019, MNRAS, 492, 3509, doi: 10.1093/mnras/stz3555

    work page doi:10.1093/mnras/stz3555 2019

  6. [6]

    2020, MNRAS, 500, 3821, doi: 10.1093/mnras/staa3023

    Bonaldi, A., An, T., Br¨ uggen, M., et al. 2020, MNRAS, 500, 3821, doi: 10.1093/mnras/staa3023

    work page doi:10.1093/mnras/staa3023 2020

  7. [7]

    Briggs, D. S. 1995, in AAS Meeting Abstracts, Vol. 187, AAS Meeting Abstracts

    1995

  8. [8]

    P., Habergham, S

    Carrillo, R. E., McEwen, J. D., & Wiaux, Y. 2012, MNRAS, 426, 1223, doi: 10.1111/j.1365-2966.2012.21605.x

    work page doi:10.1111/j.1365-2966.2012.21605.x 2012

  9. [9]

    Cornwell, T. J. 2008, IEEE J-STSP, 2, 793, doi: 10.1109/JSTSP.2008.2006388

    work page doi:10.1109/jstsp.2008.2006388 2008

  10. [10]

    J., Golap, K., & Bhatnagar, S

    Cornwell, T. J., Golap, K., & Bhatnagar, S. 2008, IEEE J-STSP, 2, 647, doi: 10.1109/JSTSP.2008.2005290

    work page doi:10.1109/jstsp.2008.2005290 2008

  11. [11]

    J., & Perley, R

    Cornwell, T. J., & Perley, R. A. 1992, A&A, 261, 353

    1992

  12. [12]

    2015, A&A, 576, A7, doi: 10.1051/0004-6361/201424602

    Dabbech, A., Ferrari, C., Mary, D., et al. 2015, A&A, 576, A7, doi: 10.1051/0004-6361/201424602

    work page doi:10.1051/0004-6361/201424602 2015

  13. [13]

    2022, ApJL, 939, L4, doi: 10.3847/2041-8213/ac98af

    Dabbech, A., Terris, M., Jackson, A., et al. 2022, ApJL, 939, L4, doi: 10.3847/2041-8213/ac98af

    work page doi:10.3847/2041-8213/ac98af 2022

  14. [14]

    2017, MNRAS, 471, 4300, doi: 10.1093/mnras/stx1775

    Wiaux, Y. 2017, MNRAS, 471, 4300, doi: 10.1093/mnras/stx1775

    work page doi:10.1093/mnras/stx1775 2017

  15. [15]

    2003, IEEE TSP, 51, 560, doi: 10.1109/TSP.2002.807005

    Fessler, J., & Sutton, B. 2003, IEEE TSP, 51, 560, doi: 10.1109/TSP.2002.807005

    work page doi:10.1109/tsp.2002.807005 2003

  16. [16]

    N., Starck, J

    Garsden, H., Girard, J. N., Starck, J. L., et al. 2015, A&A, 575, A90, doi: 10.1051/0004-6361/201424504

    work page doi:10.1051/0004-6361/201424504 2015

  17. [17]

    2023, RASTI, 2, 91, doi: 10.1093/rasti/rzad002 H¨ ogbom, J

    Gheller, C., Taffoni, G., & Goz, D. 2023, RASTI, 2, 91, doi: 10.1093/rasti/rzad002 H¨ ogbom, J. A. 1974, A&A, 15, 417

    work page doi:10.1093/rasti/rzad002 2023

  18. [18]

    2021, PASA, 38, e009, doi: 10.1017/pasa.2021.1

    Hotan, A., Bunton, J., Chippendale, A., et al. 2021, PASA, 38, e009, doi: 10.1017/pasa.2021.1

    work page doi:10.1017/pasa.2021.1 2021

  19. [19]

    2011, SKA Memo, 132

    Humphreys, B., & Cornwell, T. 2011, SKA Memo, 132

    2011

  20. [20]

    2016, in MeerKAT Science: On the Pathway to the SKA, 1, doi: 10.22323/1.277.0001

    Jonas, J., & MeerKAT Team. 2016, in MeerKAT Science: On the Pathway to the SKA, 1, doi: 10.22323/1.277.0001

    work page doi:10.22323/1.277.0001 2016

  21. [21]

    R., & Enßlin, T

    Junklewitz, H., Bell, M. R., & Enßlin, T. 2015, A&A, 581, A59, doi: 10.1051/0004-6361/201423465

    work page doi:10.1051/0004-6361/201423465 2015

  22. [22]

    G., Waterson, M., Alachkar, B., et al

    Labate, M. G., Waterson, M., Alachkar, B., et al. 2022, JATIS, 8, 011024

    2022

  23. [23]

    J., & de Hoog, F

    Li, F., Cornwell, T. J., & de Hoog, F. 2011, A&A, 528, A31, doi: 10.1051/0004-6361/201015045

    work page doi:10.1051/0004-6361/201015045 2011

  24. [24]

    M., & McEwen, J

    Mars, M., Betcke, M. M., & McEwen, J. D. 2025, RASTI, 4, rzaf025, doi: 10.1093/rasti/rzaf025

    work page doi:10.1093/rasti/rzaf025 2025

  25. [25]

    The MeqTrees software system and its use for third-generation calibration of radio interferometers

    Noordam, J. E., & Smirnov, O. M. 2010, A&A, 524, A61. https://arxiv.org/abs/1101.1745

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  26. [26]

    R., & Smirnov, O

    Offringa, A. R., & Smirnov, O. 2017, MNRAS, 471, 301, doi: 10.1093/mnras/stx1547

    work page doi:10.1093/mnras/stx1547 2017

  27. [27]

    R., McKinley, B., Hurley-Walker, N., et al

    Offringa, A. R., McKinley, B., Hurley-Walker, N., et al. 2014, MNRAS, 444, 606, doi: 10.1093/mnras/stu1368

    work page doi:10.1093/mnras/stu1368 2014

  28. [28]

    2017, MNRAS, 469, 938, doi: 10.1093/mnras/stx755

    Onose, A., Dabbech, A., & Wiaux, Y. 2017, MNRAS, 469, 938, doi: 10.1093/mnras/stx755

    work page doi:10.1093/mnras/stx755 2017

  29. [29]

    Perley, R. A. 1999, in ASP Conf. Ser., Vol. 180, Synthesis Imaging in Radio Astronomy II, ed. G. B. Taylor, C. L. Carilli, & R. A. Perley, 383

    1999

  30. [30]

    Pratley, L., Johnston-Hollitt, M., & McEwen, J. D. 2019, ApJ, 874, 174, doi: 10.3847/1538-4357/ab0a05

    work page doi:10.3847/1538-4357/ab0a05 2019

  31. [31]

    2020, A&A, 636, L1, doi: 10.1051/0004-6361/202037800

    Ramatsoku, M., Murgia, M., Vacca, V., et al. 2020, A&A, 636, L1, doi: 10.1051/0004-6361/202037800

    work page doi:10.1051/0004-6361/202037800 2020

  32. [32]

    2017, MNRAS, 470, 3981, doi: 10.1093/mnras/stx1267

    Repetti, A., Birdi, J., Dabbech, A., & Wiaux, Y. 2017, MNRAS, 470, 3981, doi: 10.1093/mnras/stx1267

    work page doi:10.1093/mnras/stx1267 2017

  33. [33]

    2020, in ICASSP 2020, 1434–1438, doi: 10.1109/ICASSP40776.2020.9053284

    Repetti, A., & Wiaux, Y. 2020, in ICASSP 2020, 1434–1438, doi: 10.1109/ICASSP40776.2020.9053284

    work page doi:10.1109/icassp40776.2020.9053284 2020

  34. [34]

    L., et al

    Roth, J., Frank, P., Bester, H. L., et al. 2024, A&A, 690, A387, doi: 10.1051/0004-6361/202451107 A distributed resource-adaptive RI measurement model17

    work page doi:10.1051/0004-6361/202451107 2024

  35. [35]

    2020, Royal Society of London

    Scaife, A. 2020, Royal Society of London. Proc. Math. Phys. Eng. Sci, 378

    2020

  36. [36]

    Smirnov, O. M. 2011, A&A, 527, A107, doi: 10.1051/0004-6361/201116434

    work page doi:10.1051/0004-6361/201116434 2011

  37. [37]

    P., Dewdney, P

    Swart, G. P., Dewdney, P. E., & Cremonini, A. 2022, JATIS, 8, 011021

    2022

  38. [38]

    2025, MNRAS, 542, 426, doi: 10.1093/mnras/staf1082

    Tajja, A., Aghabiglou, A., Tolley, E., et al. 2025, MNRAS, 542, 426, doi: 10.1093/mnras/staf1082

    work page doi:10.1093/mnras/staf1082 2025

  39. [39]

    2026, ApJS, 283, 9, doi: 10.3847/1538-4365/ae3149

    Tang, C., Dabbech, A., Jackson, A., & Wiaux, Y. 2026, ApJS, 283, 9, doi: 10.3847/1538-4365/ae3149

    work page doi:10.3847/1538-4365/ae3149 2026

  40. [40]

    2022, MNRAS, 518, 604–622, doi: 10.1093/mnras/stac2672

    Terris, M., Dabbech, A., Tang, C., & Wiaux, Y. 2022, MNRAS, 518, 604–622, doi: 10.1093/mnras/stac2672

    work page doi:10.1093/mnras/stac2672 2022

  41. [41]

    2025, MNRAS, 537, 1608, doi: 10.1093/mnras/staf022

    Terris, M., Tang, C., Jackson, A., & Wiaux, Y. 2025, MNRAS, 537, 1608, doi: 10.1093/mnras/staf022

    work page doi:10.1093/mnras/staf022 2025

  42. [42]

    R., Moran, J

    Thompson, A. R., Moran, J. M., & Swenson, G. W. 2007, Interferometry and Synthesis in Radio Astronomy (Wiley-VCH), doi: 10.1007/978-3-319-44431-4

    work page doi:10.1007/978-3-319-44431-4 2007

  43. [43]

    2023, MNRAS, 521, 1, doi: 10.1093/mnras/stac1521

    Thouvenin, P.-A., Abdulaziz, A., Dabbech, A., Repetti, A., & Wiaux, Y. 2023, MNRAS, 521, 1, doi: 10.1093/mnras/stac1521

    work page doi:10.1093/mnras/stac1521 2023

  44. [44]

    J., Goeke, R., Bowman, J

    Tingay, S. J., Goeke, R., Bowman, J. D., et al. 2013, PASA, 30, e007, doi: 10.1017/pasa.2012.007 van Haarlem, M. P., Wise, M. W., Gunst, A., et al. 2013, A&A, 556, A2, doi: 10.1051/0004-6361/201220873

    work page doi:10.1017/pasa.2012.007 2013

  45. [45]

    Percival, et al

    Wiaux, Y., Puy, G., Boursier, Y., & Vandergheynst, P. 2009, MNRAS, 400, 1029, doi: 10.1111/j.1365-2966.2009.15519.x

    work page doi:10.1111/j.1365-2966.2009.15519.x 2009

  46. [46]

    G., Dabbech, A., Jackson, A., & Wiaux, Y

    Wilber, A. G., Dabbech, A., Jackson, A., & Wiaux, Y. 2023a, MNRAS, 522, 5558, doi: 10.1093/mnras/stad1351

    work page doi:10.1093/mnras/stad1351

  47. [47]

    2023b, MNRAS, 522, 5576, doi: 10.1093/mnras/stad1353

    Wiaux, Y. 2023b, MNRAS, 522, 5576, doi: 10.1093/mnras/stad1353

    work page doi:10.1093/mnras/stad1353

  48. [48]

    2013, MNRAS, 436, 1993, doi: 10.1093/mnras/stt1707

    Wiaux, Y. 2013, MNRAS, 436, 1993, doi: 10.1093/mnras/stt1707

    work page doi:10.1093/mnras/stt1707 2013

  49. [49]

    2022, MNRAS, 515, 1973, doi: 10.1093/mnras/stac1843

    Xie, Y.-F., Wang, F., Deng, H., et al. 2022, MNRAS, 515, 1973, doi: 10.1093/mnras/stac1843

    work page doi:10.1093/mnras/stac1843 2022

  50. [50]

    F., Tan, S

    Ye, H., Gull, S. F., Tan, S. M., & Nikolic, B. 2022, MNRAS, 510, 4110, doi: 10.1093/mnras/stab3548 Zdenˇ ek, P. 2012, PhD thesis, Brno university of technology

    work page doi:10.1093/mnras/stab3548 2022