Pith. sign in

REVIEW 3 major objections 4 minor 50 references

A GPU Code for Finding Microlensing Critical Curves and Caustics

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper presents a GPU code that locates microlensing critical curves and caustics for fields of thousands to ten thousand microlenses, and shows how the resulting caustic-crossing maps and distance distributions can constrain source…

desk verdict Useful GPU caustic code, but the accuracy claims need validation before the community relies on it. read the letter →

arxiv 2506.02121 v1 pith:L66AW2VA submitted 2025-06-02 astro-ph.GA astro-ph.IM

Luke Weisenbach This is my paper
classification astro-ph.GAastro-ph.IM
keywords gravitationalmicrolensingcriticalcurvescausticsGPUcomputingAberth-Ehrlichmethodfastmultipolecausticcrossingsmethods:numerical
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper presents a publicly released GPU code that solves the microlensing lens equation for the critical curves and caustics of a random field of stars, for field sizes of order $10^3$--$10^4$ microlenses. The central methodological claim is that the critical-curve problem reduces to finding the roots of a phase-parametrized family of complex equations $F(z,\phi)=0$, and that those roots can be found in parallel with the Aberth--Ehrlich method, accelerated by a fast multipole method and by growing the phase chains from the center outward. The code then exploits the orientation of the caustic curves to compute, via winding numbers, maps of the number of caustic crossings as a function of source position. A sympathetic reader should care because these caustic-crossing counts and distance-to-caustic distributions are quantities that can constrain source size, position, and velocity when interpreting light curves of lensed quasars and supernovae.

What carries the argument

The central object is the phase-parametrized critical-curve equation $F(z,\phi)=0$, a rational function of complex position $z$; for each phase $\phi\in[0,2\pi)$ its roots are image-plane points of formally infinite magnification, and following the roots as $\phi$ advances draws the critical curves, whose image under the lens equation is the caustic network. The machinery that carries the argument is the combination of the Aberth--Ehrlich simultaneous root-finding iteration (a cubically convergent method in which approximate roots repel each other and converge in parallel), implemented on a GPU; a fast multipole expansion for the deflection of distant microlenses; the replacement of the finite-field compensating mass sheet by an infinite sheet (circular fields) or a truncated Taylor polynomial (rectangular fields); and a winding-number algorithm that turns the oriented caustic polygons into a map of the number of caustic crossings per source pixel.

What would settle it

Run the code on a rectangular field with moderate $N_\star$ and compare the predicted caustic-crossing number map and caustic positions with caustic locations inferred from a direct ray-shooting magnification map for the same realization; if the winding-number map predicts crossings at positions that disagree with the ray-shooting caustics by more than a source-plane pixel, the truncated-polynomial approximation to the compensating deflection is not faithful inside the field. Equivalently, raise the Taylor order $t$ and check that the caustic positions converge; systematic shifts larger than the pixel scale would show the truncation is the limiting error.

Watch

Extended reading notes

Core claim

The central claim is that locating microlensing caustics, long treated as a sequential numerical bottleneck, is parallelizable end to end on a GPU. The critical curves are the image-plane loci where the inverse magnification vanishes; following a parametric representation, they are traced as the roots of $F(z,\phi)=0$, where $F$ is a rational function built from the macromodel convergence and shear, the microlens positions and masses, and the compensating smooth-matter deflection, and $\phi$ runs from $0$ to $2\pi$. The paper shows that all roots for one phase can be found simultaneously with a parallel Aberth--Ehrlich iteration seeded from the microlens positions, that distant microlenses can be folded into a fast multipole expansion, and that the root-following can start at $\phi=\pi$ and grow outward in both directions, or be split into independent subchains. Because each caustic is an oriented curve, a GPU winding-number calculation converts the caustic network into a map of the number of caustic crossings at every source position, and a Euclidean distance transform turns that map into distance-to-nearest-caustic distributions for both expanding and moving sources.

Load-bearing premise

The load-bearing premise is that the smooth matter spread used to balance the finite microlens field can be replaced by an infinite sheet (circular fields) or a truncated polynomial (rectangular fields) without moving the caustics inside the field; the paper bounds the truncation error but never validates caustic positions against the true lens equation.

Editorial extensions

If this is right

  • For fields of $10^3$--$10^4$ microlenses, the full critical-curve and caustic network can be computed on a single NVIDIA GPU, with runtime scaling roughly linearly with $N_\star$ at low $N_\star$ and becoming quadratic for $N_\star\gtrsim 10^4$.
  • The caustic-crossing number map splits the source plane into regions with distinct integer crossing counts, so standard image-processing algorithms can compute the distance from any source position to the nearest caustic for both expanding (supernova) and moving (quasar) sources.
  • Distance-to-caustic distributions quantify how often a source crosses caustics as a function of its velocity direction, giving a numerical factor (about 3 in the example) between motion parallel and perpendicular to the caustics without simulating light curves.
  • The same parallel root-finding machinery applies to exoplanet microlensing with an arbitrary number of lenses, and it makes feasible broader studies of the parameter that controls maximum magnification during a caustic crossing and of self-similar subdistributions of the magnification probability distribution.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural test the paper leaves implicit: comparing caustic positions and crossing counts from the truncated polynomial approximation against a direct ray-shooting magnification map for the same microlens realization would settle whether the rectangular-field approximation is faithful inside the field.
  • The crossing-number map could serve as a cheap prior in light-curve likelihood analyses, potentially skipping magnification-map evaluations in regions of the source plane with zero crossings; the paper hints at this use but does not quantify the computational saving.
  • If the observed self-similarity of magnification subdistributions across regions with different numbers of bright microimages holds at larger $N_\star$, it suggests an analytic or semi-analytic form for the magnification probability distribution, which would remove a major numerical bottleneck for microlensing modeling.
  • Runtime benchmarks are reported on one high-end GPU; on consumer hardware the crossover to quadratic scaling and the memory limits of the $(m+1)\times n_r$ root matrix may shift, so portability of the speedups should be verified on target hardware.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper presents a GPU implementation, released on GitHub, for locating microlensing critical curves and caustics in random star fields with N_star up to ~10^4. The method follows Witt's parametric representation of critical curves, solves the resulting polynomial roots with a parallel Aberth-Ehrlich iteration as in Ghidouche et al. (2017), accelerates the stellar sums with the fast multipole method, and handles finite circular and rectangular microlens fields by replacing the compensating smooth matter deflection with polynomial approximations. The paper also constructs winding-number maps of caustic crossings and uses distance transforms to estimate distances to caustics for moving and expanding sources, with applications to lensed quasars and supernovae. The central claim is that the code efficiently and accurately solves the caustic-location problem.

Significance. If the computed caustic networks are accurate, the code fills a genuine need: it combines caustic location with magnification-map analyses, uses established root-finding and FMM techniques, and is publicly available, which should help reproducibility. The runtime scaling shown in Fig. 3 is promising and the winding-number/distance-map applications are potentially useful for light-curve analyses. However, the scientific weight of the paper rests on the fidelity of the computed critical curves and caustics, and that fidelity is not demonstrated: the rectangular-field approximation is not validated against an independent solution, the convergence of the phase tracing is not tested, and the FMM accuracy parameters are never reported. The contribution is therefore significant conditional on an accuracy validation that the manuscript currently lacks.

major comments (3)
  1. [Appendix B, Eqs. (B4)-(B8), and Eq. (15)] The Taylor expansion of alpha_s(z) is the load-bearing approximation for rectangular fields, but the paper bounds the pointwise error in alpha_s itself (Eq. B5), while the critical-curve equation (Eq. 15) depends on partial alpha_s/partial z and partial alpha_s/partial zbar. A bound on alpha_s does not bound the derivatives, and replacing the exact logarithmic deflection by a polynomial changes the global structure of F(z), potentially adding or removing roots that are not roots of the true lens equation. The manuscript provides no test comparing the computed caustics in a rectangular field with an independent solution of the exact lens equation, even for a single small configuration. Such a benchmark, or a rigorous derivative error bound valid inside the rectangle, is necessary to support the central claim.
  2. [Section 3.1, Section 4.4, and Fig. 3] The only quantitative performance result is runtime scaling, but the accuracy of the critical curves is never established. The paper does not report the FMM expansion order or node size used in the calculations, nor does it study convergence of the critical-curve tracing in the phase step m; it simply states that m=200 was used for Fig. 1. A convergence test, for example comparing caustic maps obtained with several values of m and FMM settings, is needed to show that the parallel root-tracking does not miss or mix chains and that the resulting caustic positions are stable. Without this, the winding-number and distance maps of Section 5 have no demonstrated numerical reliability.
  3. [Appendix A footnote and Section 5.1] The footnote in Appendix A explicitly admits that the infinite-sheet approximation for the compensating matter can produce spurious caustics near the origin for macromodels satisfying 1-(kappa-kappa_star) +/- gamma ~ 0, because critical curves far outside the microlens field map back into the region of interest. Section 5.1 and Fig. 4 present maps of the number of caustic crossings over the source plane, and Section 5.2 uses these maps for distance and velocity constraints. The paper does not show that the simulations used in these figures avoid the problematic parameter regime, nor does it demonstrate how such spurious caustics are identified and removed. This directly affects the astrophysical applications claimed.
minor comments (4)
  1. [Appendix B, paragraph after Eq. (B7)] The line 'Using c = c * e^{-i*2*arctan(c2/c1)}' appears to contain a typo: it should state that the complex conjugate of c satisfies \bar c = c e^{-2i Arg c}. Clarifying this would help readers follow the simplification leading to Eq. (B8).
  2. [Fig. 3 caption] The caption does not define the solid versus opaque markers; the reader must infer from the text that solid markers are circular fields and opaque markers are rectangular fields. Please state this explicitly in the caption.
  3. [Section 4.2 and Appendix B] The paper states that the Taylor terms add 'a polynomial of some degree' to F(z), but it never gives the degree of that polynomial or the resulting number of roots n_r as a function of the truncation order t. Stating this explicitly would help users of the code understand the computational cost and the root-count assumptions.
  4. [Section 5.2 and Fig. 7] The statement that 'a source that has traveled 5 theta_star must be located outside the microcaustics' is not self-evident from the figure; adding a brief explanation of how the distance distribution and the caustic-crossing map imply this would improve readability.

Circularity Check

1 steps flagged · score 4.0 of 10

One load-bearing self-citation for the rectangular-field deflection; otherwise the implementation rests on independent published algorithms.

  1. self citation load bearing [Section 4.2, 'A rectangular microlens field'; relied on in Appendix B]
    "Weisenbach (in prep.) derives the form of α(z) in complex coordinates, from which one can find ∂α_s(z)/∂z and ∂α_s(z)/∂z. Discontinuities in the derivatives of α_s(z) are more prevalent in this case than for a circular field of microlenses, though as Appendices A and B show they can be circumvented and approximated with polynomials."

    The rectangular-field caustic calculation is a claimed feature of the code, and its load-bearing input is the compensating deflection α_s(z) for a rectangular field. That object is not derived in this paper: it is attributed to the author's own unpublished 'Weisenbach (in prep.)'. Appendix B then Taylor-expands that same α_s(z) and differentiates it to form ∂α_s/∂z, which enters the critical-curve equation F(z,φ) (Eq. 15). Thus the accuracy of the rectangular-field critical curves and the downstream crossing-number and distance maps rests on an author-overlapping citation whose content is not independently established in the manuscript, rather than on a derivation from stated first principles or an external benchmark.

full rationale

The core numerical method is not circular: the parametric critical-curve equation is from Witt (1990), the GPU Aberth–Ehrlich root-finding implementation follows Ghidouche et al. (2017), and the fast multipole acceleration uses Greengard & Rokhlin (1987) and Witt et al. (1993), all independent of the present author. For circular microlens fields the compensating-sheet approximation is exact inside the field, so the critical curves there are not fitted or predicted from the same data. The Section 5 crossing-number and distance maps are straightforward geometric post-processing of the computed caustics, not quantities that were used to fit the method. The one circularity-relevant step is the rectangular-field deflection α_s(z): the paper explicitly sources its complex-coordinate form to 'Weisenbach (in prep.)', an unpublished work by the same author, and Appendix B builds the critical-curve derivatives on that expression. This is a load-bearing self-citation for the rectangular-field portion of the code. The paper also openly admits in Appendix A that the infinite-sheet approximation can introduce spurious caustics for near-critical macromodels, and it does not compare rectangular-field caustic positions against an independent solution of the exact lens equation; those are validation and correctness concerns rather than circularity per se, but they reinforce that the rectangular-field claim is not yet independently established. On balance, the central implementation has substantial independent content, so the circularity score is moderate rather than high.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The computational method depends on free numerical resolution parameters (Taylor truncation order, phase steps, FMM settings) and on domain assumptions about the smooth matter sheet approximation and root continuation. No new physical entities are introduced.

free parameters (3)
  • Truncation order t for Taylor series of rectangular alpha_s = not specified
    Chosen by user; controls accuracy vs speed. The paper provides an error bound (Eq. B5) but no recommended value or sensitivity test.
  • Number of phase steps m for tracing critical curves = m=200 in Figure 1; not specified for other figures
    Resolution of the parametric curve tracing; larger m gives smoother curves but more computation.
  • FMM expansion order and node size = not specified
    FMM is used to approximate distant microlenses; accuracy depends on expansion order and node geometry, but these are not reported.
assumptions (4)
  • standard math The lens equation and Witt's parametric critical curve representation are valid for a random field of point masses.
    Standard gravitational lensing theory, cited to Saha et al. 2024 and Witt 1990.
  • domain assumption The compensating smooth matter sheet for a circular microlens field can be replaced by an infinite uniform sheet without changing critical curves inside the field.
    Appendix A: the infinite sheet matches alpha_s inside the original disk, so critical curves inside are exact; outside they are incorrect but assumed unimportant.
  • ad hoc to paper For a rectangular field, the complex logarithmic deflection can be approximated by a truncated Taylor series and the mass sheet extended to infinity, yielding a polynomial root equation.
    Appendix B: this 'approximate ruthlessly' step is introduced to make the root finding tractable; the error bound is derived but the approximation is not validated against the exact rectangular deflection.
  • domain assumption The roots for phi_j+Delta phi are close enough to the roots for phi_j that Aberth-Ehrlich with previous roots as initial guesses converges to the correct roots.
    Section 3.2: standard continuation assumption from Witt 1990; likely valid for small Delta phi but not proven for dense caustic networks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A GPU Code for Finding Microlensing Critical Curves and Caustics." pith.science (2026). https://pith.science/paper/L66AW2VA

@misc{pith2026250602121,
  author       = {Pith},
  title        = {Pith review of: A GPU Code for Finding Microlensing Critical Curves and Caustics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L66AW2VA}},
  note         = {Machine review of arXiv:2506.02121}
}
read the original abstract

Advancements in analyses of caustic crossing events in gravitationally microlensed quasars and supernovae can benefit from numerical simulations which locate the caustics in conjunction with the creation of magnification maps. We present a GPU code which efficiently solves this problem; the code is available at https://github.com/weisluke/microlensing/. We discuss how the locations of the microcaustics can be used to determine the number of caustic crossings and the distances to caustics, both of which can be used to constrain the space of nuisance parameters such as source position and velocity within magnification maps.

Figures

Figures reproduced from arXiv: 2506.02121 by the authors.

Figure 1
Figure 1. Critical curves (top) and caustics (bottom) for a field of random microlenses. The microlens positions are marked with ★ symbols. The rectangular image plane region maps back to the square source plane region; the caustics of the (approximately) single and binary lens critical curves near the origin of the image plane are visible near the origin of the source plane. Points along the critical curves and caustics are … view at source ↗
Figure 2
Figure 2. Visualization of the steps needed to find the roots of 𝐹(𝑧, 𝜙) for each value of 𝜙. 3.2 Finding roots The process of first finding the 𝑛𝑟 solutions for 𝜙 = 0, and then following each solution as 𝜙 varies, can be time consuming. This is in part due to the sequential nature of finding the roots in previous works (Witt 1990; Witt et al. 1993), which largely seem to first find one root to the desired accuracy using Newt… view at source ↗
Figure 3
Figure 3. Time taken to calculate critical curves on an NVIDIA A100 80GB GPU, vs. number of microlenses 𝑁★, for various values of the number of steps 𝑚 used to follow 𝜙 from 0 to 2𝜋. The dashed lines show the asymptotic runtime scaling 𝑡 ∝ 𝑁2 ★ of the parallelized Aberth-Ehrlich method, while the dotted lines show the approximate 𝑡 ∝ 𝑁★ of low 𝑁★. The solid markers are for a circular field of microlenses, while the opaque mar… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Visualization of the number of caustic crossings. The locations of the caustics are shown with white lines. Pixelation effects near the caustics would be seen with a sufficient zoom, but determining the number of caustic crossings at pixel locations far from the causti…
Figure 5
Figure 5. Figure 5: The distance 𝑑caustic from each (pixelated) source position to the nearest caustic for an expanding source (top) and for a moving source (bottom) traveling along the direction indicated by the black arrow. We note that edge effects must additionally be accounted for, a…
Figure 6
Figure 6. Figure 6: Sample probability distribution of the distance to the nearest caus￾tic and the source velocity angle with respect to the positive 𝑦1 axis. The distribution makes it clear that travel perpendicular to the caustics (90°) expe￾riences caustic crossings more frequently th…
Figure 7
Figure 7. Figure 7: Sample probability distribution of the distance to the nearest caustic for a source traveling along the direction 𝑦2 (i.e. a vertical cut of [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 19 canonical work pages

  1. [1]

    Aberth O., 1973, Mathematics of Computation, 27, 339–344

    work page 1973

  2. [2]

    N., Zhdanov V

    Alexandrov A. N., Zhdanov V. I., 2011, @doi [ ] 10.1111/j.1365-2966.2011.19296.x , https://ui.adsabs.harvard.edu/#abs/2011MNRAS.417..541A 417, 541

    arXiv 2011

  3. [3]

    Alpay A., 2019, https://github.com/illuhad/teralens

    2019

  4. [4]

    F., Fluke C

    Bate N. F., Fluke C. J., Barsdell B. R., Garsden H., Lewis G. F., 2010, @doi [ ] 10.1016/j.newast.2010.05.008 , https://ui.adsabs.harvard.edu/abs/2010NewA...15..726B 15, 726

    doi:10.1016/j.newast.2010.05.008 2010

  5. [5]

    Best H., Fagin J., Vernardos G., O'Dowd M., 2024, @doi [ ] 10.1093/mnras/stae1182 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.531.1095B 531, 1095

    work page doi:10.1093/mnras/stae1182 2024

  6. [6]

    R., Kantowski R., 1975, @doi [ ] 10.1086/153300 , https://ui.adsabs.harvard.edu/abs/1975ApJ...195...13B 195, 13

    Bourassa R. R., Kantowski R., 1975, @doi [ ] 10.1086/153300 , https://ui.adsabs.harvard.edu/abs/1975ApJ...195...13B 195, 13

    doi:10.1086/153300 1975

  7. [7]

    R., Kantowski R., Norton T

    Bourassa R. R., Kantowski R., Norton T. D., 1973, @doi [ ] 10.1086/152452 , https://ui.adsabs.harvard.edu/abs/1973ApJ...185..747B 185, 747

    doi:10.1086/152452 1973

  8. [8]

    S., Dai X., Poindexter S., Garmire G., 2009, @doi [ ] 10.1088/0004-637X/693/1/174 , https://ui.adsabs.harvard.edu/abs/2009ApJ...693..174C 693, 174

    Chartas G., Kochanek C. S., Dai X., Poindexter S., Garmire G., 2009, @doi [ ] 10.1088/0004-637X/693/1/174 , https://ui.adsabs.harvard.edu/abs/2009ApJ...693..174C 693, 174

    doi:10.1088/0004-637x/693/1/174 2009

Show all 50 references
  1. [9]

    Dan e k K., Heyrovsk \'y D., 2015, @doi [ ] 10.1088/0004-637X/806/1/63 , https://ui.adsabs.harvard.edu/abs/2015ApJ...806...63D 806, 63

    2015 doi

  2. [10]

    W., 1967, Communications of the ACM, 10, 107

    Ehrlich L. W., 1967, Communications of the ACM, 10, 107

    1967

  3. [11]

    Fatheddin H., Sajadian S., 2022, @doi [ ] 10.1093/mnras/stac1565 , https://ui.adsabs.harvard.edu/abs/2022MNRAS.514.4379F 514, 4379

    2022 doi

  4. [12]

    Ghidouche K., Sider A., Couturier R., Guyeux C., 2017, Journal of Computational Science, 18, 46

    2017

  5. [13]

    L., Wambsganss J., 2003, @doi [ ] 10.1086/345447 , https://ui.adsabs.harvard.edu/abs/2003ApJ...583..575G 583, 575

    Granot J., Schechter P. L., Wambsganss J., 2003, @doi [ ] 10.1086/345447 , https://ui.adsabs.harvard.edu/abs/2003ApJ...583..575G 583, 575

    2003 doi

  6. [14]

    Greengard L., Rokhlin V., 1987, @doi [Journal of Computational Physics] 10.1016/0021-9991(87)90140-9 , https://ui.adsabs.harvard.edu/abs/1987JCoPh..73..325G 73, 325

    1987 doi

  7. [15]

    Hundertmark M. P. G., 2011, PhD thesis, Georg August University of Gottingen, Germany

    2011

  8. [16]

    Jim \'e nez-Vicente J., Mediavilla E., 2022, @doi [ ] 10.3847/1538-4357/ac9e59 , https://ui.adsabs.harvard.edu/abs/2022ApJ...941...80J 941, 80

    2022 doi

  9. [17]

    Kayser R., Refsdal S., Stabell R., 1986, , https://ui.adsabs.harvard.edu/abs/1986A&A...166...36K 166, 36

    1986

  10. [18]

    S., 2004, @doi [ ] 10.1086/382180 , https://ui.adsabs.harvard.edu/abs/2004ApJ...605...58K 605, 58

    Kochanek C. S., 2004, @doi [ ] 10.1086/382180 , https://ui.adsabs.harvard.edu/abs/2004ApJ...605...58K 605, 58

    2004 doi

  11. [19]

    W., Chang K

    Lee D. W., Chang K. A., Kim S. J., 1998, Journal of Korean Astronomical Society, https://ui.adsabs.harvard.edu/#abs/1998JKAS...31...27L 31, 27

    1998

  12. [20]

    F., Miralda-Escude J., Richardson D

    Lewis G. F., Miralda-Escude J., Richardson D. C., Wambsganss J., 1993, @doi [ ] 10.1093/mnras/261.3.647 , https://ui.adsabs.harvard.edu/abs/1993MNRAS.261..647L 261, 647

    1993 doi

  13. [21]

    A., Lopez P., Mediavilla T., Abajas C., Gonzalez-Morcillo C., Gil-Merino R., 2006, @doi [ ] 10.1086/508796 , https://ui.adsabs.harvard.edu/abs/2006ApJ...653..942M 653, 942

    Mediavilla E., Mu \ n oz J. A., Lopez P., Mediavilla T., Abajas C., Gonzalez-Morcillo C., Gil-Merino R., 2006, @doi [ ] 10.1086/508796 , https://ui.adsabs.harvard.edu/abs/2006ApJ...653..942M 653, 942

    2006 doi

  14. [22]

    A., Ariza O., Lopez P., Gonzalez-Morcillo C., Jimenez-Vicente J., 2011, @doi [ ] 10.1088/0004-637X/741/1/42 , https://ui.adsabs.harvard.edu/abs/2011ApJ...741...42M 741, 42

    Mediavilla E., Mediavilla T., Mu \ n oz J. A., Ariza O., Lopez P., Gonzalez-Morcillo C., Jimenez-Vicente J., 2011, @doi [ ] 10.1088/0004-637X/741/1/42 , https://ui.adsabs.harvard.edu/abs/2011ApJ...741...42M 741, 42

    2011 doi

  15. [23]

    Natarajan P., Williams L. L. R., Brada c M., Grillo C., Ghosh A., Sharon K., Wagner J., 2024, @doi [ ] 10.1007/s11214-024-01051-8 , https://ui.adsabs.harvard.edu/abs/2024SSRv..220...19N 220, 19

    2024 doi

  16. [24]

    Neira F., Anguita T., Vernardos G., 2020, @doi [ ] 10.1093/mnras/staa1208 , https://ui.adsabs.harvard.edu/abs/2020MNRAS.495..544N 495, 544

    2020 doi

  17. [25]

    Paczynski B., 1986, @doi [ ] 10.1086/163919 , https://ui.adsabs.harvard.edu/abs/1986ApJ...301..503P 301, 503

    1986 doi

  18. [26]

    H., Bonvin V., 2022, @doi [ ] 10.1051/0004-6361/202141808 , https://ui.adsabs.harvard.edu/abs/2022A&A...659A..21P 659, A21

    Paic E., Vernardos G., Sluse D., Millon M., Courbin F., Chan J. H., Bonvin V., 2022, @doi [ ] 10.1051/0004-6361/202141808 , https://ui.adsabs.harvard.edu/abs/2022A&A...659A..21P 659, A21

    2022 doi

  19. [27]

    Ren H., Zhu W., 2025, @doi [ ] 10.3847/1538-3881/adb1b2 , https://ui.adsabs.harvard.edu/abs/2025AJ....169..170R 169, 170

    2025 doi

  20. [28]

    Saha P., Williams L. L. R., 2011, @doi [ ] 10.1111/j.1365-2966.2010.17797.x , https://ui.adsabs.harvard.edu/abs/2011MNRAS.411.1671S 411, 1671

    2011

  21. [29]

    Saha P., Sluse D., Wagner J., Williams L. L. R., 2024, @doi [ ] 10.1007/s11214-024-01041-w , https://ui.adsabs.harvard.edu/abs/2024SSRv..220...12S 220, 12

    2024 doi

  22. [30]

    Schneider P., Weiss A., 1986, , https://ui.adsabs.harvard.edu/abs/1986A&A...164..237S 164, 237

    1986

  23. [31]

    J., et al., 2022, @doi [arXiv e-prints] 10.48550/arXiv.2210.10790 , https://ui.adsabs.harvard.edu/abs/2022arXiv221010790S p

    Shajib A. J., et al., 2022, @doi [arXiv e-prints] 10.48550/arXiv.2210.10790 , https://ui.adsabs.harvard.edu/abs/2022arXiv221010790S p. arXiv:2210.10790

    2022 doi arXiv

  24. [32]

    N., Gil-Merino R., Goicoechea L

    Shalyapin V. N., Gil-Merino R., Goicoechea L. J., 2021, @doi [ ] 10.1051/0004-6361/202140527 , https://ui.adsabs.harvard.edu/abs/2021A&A...653A.121S 653, A121

    2021 doi

  25. [33]

    Sunday D., 2001, Inclusion of a Point in a Polygon, https://web.archive.org/web/20130126163405/http://geomalgorithms.com/a03-_inclusion.html

    2001

  26. [34]

    H., Goobar A., Collett T., More A., Vernardos G., 2024, @doi [ ] 10.1007/s11214-024-01044-7 , https://ui.adsabs.harvard.edu/abs/2024SSRv..220...13S 220, 13

    Suyu S. H., Goobar A., Collett T., More A., Vernardos G., 2024, @doi [ ] 10.1007/s11214-024-01044-7 , https://ui.adsabs.harvard.edu/abs/2024SSRv..220...13S 220, 13

    2024 doi

  27. [35]

    C., Fluke C

    Thompson A. C., Fluke C. J., Barnes D. G., Barsdell B. R., 2010, @doi [ ] 10.1016/j.newast.2009.05.010 , https://ui.adsabs.harvard.edu/abs/2010NewA...15...16T 15, 16

    2010 doi

  28. [36]

    Vernardos G., Tsagkatakis G., 2019, @doi [ ] 10.1093/mnras/stz868 , https://ui.adsabs.harvard.edu/abs/2019MNRAS.486.1944V 486, 1944

    2019 doi

  29. [37]

    Vernardos G., et al., 2024, @doi [ ] 10.1007/s11214-024-01043-8 , https://ui.adsabs.harvard.edu/abs/2024SSRv..220...14V 220, 14

    2024 doi

  30. [38]

    F., Weymann R

    Walsh D., Carswell R. F., Weymann R. J., 1979, @doi [Nature] 10.1038/279381a0 , https://ui.adsabs.harvard.edu/abs/1979Natur.279..381W 279, 381

    1979 doi

  31. [39]

    Wambsganss J., 1990, PhD thesis, -

    1990

  32. [40]

    J., Schneider P., 1992, , https://ui.adsabs.harvard.edu/abs/1992A&A...258..591W 258, 591

    Wambsganss J., Witt H. J., Schneider P., 1992, , https://ui.adsabs.harvard.edu/abs/1992A&A...258..591W 258, 591

    1992

  33. [41]

    Wang S., Wang L., Dong S., 2025, @doi [ ] 10.3847/1538-4365/ad9b8d , https://ui.adsabs.harvard.edu/abs/2025ApJS..276...40W 276, 40

    2025 doi

  34. [42]

    S., Krawczyk C., Vernardos G., Enzi W., Lundgren A., 2024, @doi [ ] 10.1093/mnras/stae1396 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.531.4349W 531, 4349

    Weisenbach L., Collett T., de Murieta A. S., Krawczyk C., Vernardos G., Enzi W., Lundgren A., 2024, @doi [ ] 10.1093/mnras/stae1396 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.531.4349W 531, 4349

    2024 doi

  35. [43]

    arXiv:2502.01728

    Weisenbach L., Collett T., Enzi W., Oldham L., Sainz de Murieta A., 2025, @doi [arXiv e-prints] 10.48550/arXiv.2502.01728 , https://ui.adsabs.harvard.edu/abs/2025arXiv250201728W p. arXiv:2502.01728

    2025 doi arXiv

  36. [44]

    J., 1990, , https://ui.adsabs.harvard.edu/abs/1990A&A...236..311W 236, 311

    Witt H. J., 1990, , https://ui.adsabs.harvard.edu/abs/1990A&A...236..311W 236, 311

    1990

  37. [45]

    J., 1993, @doi [ ] 10.1086/172223 , https://ui.adsabs.harvard.edu/abs/1993ApJ...403..530W 403, 530

    Witt H. J., 1993, @doi [ ] 10.1086/172223 , https://ui.adsabs.harvard.edu/abs/1993ApJ...403..530W 403, 530

    1993 doi

  38. [46]

    J., Kayser R., Refsdal S., 1993, , https://ui.adsabs.harvard.edu/abs/1993A&A...268..501W 268, 501

    Witt H. J., Kayser R., Refsdal S., 1993, , https://ui.adsabs.harvard.edu/abs/1993A&A...268..501W 268, 501

    1993

  39. [47]

    Wyithe J. S. B., Loeb A., 2002, @doi [ ] 10.1086/342121 , https://ui.adsabs.harvard.edu/abs/2002ApJ...577..615W 577, 615

    2002 doi

  40. [48]

    Wyithe J. S. B., Turner E. L., 2001, @doi [ ] 10.1046/j.1365-8711.2001.03917.x , https://ui.adsabs.harvard.edu/abs/2001MNRAS.320...21W 320, 21

    2001

  41. [49]

    Young P., 1981, @doi [ ] 10.1086/158752 , https://ui.adsabs.harvard.edu/abs/1981ApJ...244..756Y 244, 756

    1981 doi

  42. [50]

    Zheng W., Chen X., Li G., Chen H.-Z., 2022, @doi [ ] 10.3847/1538-4357/ac68ea , https://ui.adsabs.harvard.edu/abs/2022ApJ...931..114Z 931, 114

    2022 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.