Leveraging the null stream to detect strongly lensed gravitational waves

Jef Heynen; Justin Janquart; Soumen Roy

REVIEW 3 major objections 6 minor 2 cited by

Leveraging the null stream to detect strongly lensed gravitational waves

T0 review · 3 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read Strongly lensed gravitational-wave images can be found by a joint network null stream: correcting antenna responses for magnification, time shift, and Morse phase cancels lensed pairs, while unrelated pairs do not.

desk verdict A genuinely new null-stream construction for lensed GW pairs, but Eq. (8) omits the relative time delay and the demo still leans on PE waveforms; worth refereeing as a methods note. read the letter →

arxiv 2509.06745 v1 pith:S6GMR463 submitted 2025-09-08 gr-qc astro-ph.HE

Jef Heynen , Soumen Roy , Justin Janquart This is my paper

classification gr-qcastro-ph.HE MSC 83C35 PACS 04.30.-w

keywords gravitationalwavesstronglensingnullstreamdetectornetworkMorsephasedetectionstatisticnestedsampling

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

Strong gravitational lensing produces several images of one gravitational-wave signal that differ only by a magnification, an arrival-time shift, and a Morse phase. The paper argues that these three parameters let the null stream of the detector network be generalized into a joint null stream for an entire set of candidate images: at the correct sky position and lensing parameters, the projector that cancels signals cancels all images at once. Genuinely lensed pairs then show joint null energy near zero, while unrelated events cannot be mapped onto each other and leave a large residual. The authors demonstrate the split with an injection study of 100 lensed and 1000 unlensed pairs and report that evaluating one pair takes a few minutes on a single processing unit, in contrast to full joint parameter estimation. If it holds up, the statistic gives lensing searches a cheap classical route toward real-time analysis, and a fully waveform-independent version is described as the next step.

What carries the argument

The joint antenna response matrix F of Eq. (7): the responses of every detector to every candidate image, stacked, with the rows of image j rescaled by the relative lensing factor √μ_{j1} e^{iπ n_{j1} sign(f)} (the square root of the relative magnification times the Morse phase). Absorbing the lensing into the detector geometry rather than the waveform makes a single two-dimensional polarization subspace describe all images at once. The null projector P_null = 1 − F(F†F)^{−1}F† built from this matrix then removes every lensed image from the stacked data, and the summed residual power — the joint null energy — is the detection statistic. Nested sampling over the sky and lensing parameters min

What would settle it

Inject strongly lensed pairs at O4 sensitivity whose true waveform belongs to a family the reconstruction omits (e.g., precession or tides), and check whether the lensed joint-null-energy distribution stays far from the unlensed distribution; if it moves to overlap, the discrimination is carried by reconstruction fidelity, not by the null-stream formalism.

Watch

Extended reading notes

Core claim

In the geometric-optics regime, every lensed image j of one source shares a single waveform up to a relative magnification μ_{j1}, an arrival-time shift t_{j1}, and a Morse phase n_{j1}. The central move is to fold these factors into the antenna response matrix: stacking all images' data across all detectors, the joint matrix F of Eq. (7) rescales image j's rows by √μ_{j1} e^{iπ n_{j1} sign(f)}. At the true sky position and lensing parameters, F spans exactly the two polarization directions over the whole network, so the null projector P_null = 1 − F(F†F)^{−1}F† removes every image at once and only noise remains. Lensed pairs therefore give near-zero joint null energy; unrelated pairs cannot

Load-bearing premise

The lensed-versus-unlensed separation rests on the maximum-likelihood reconstructed waveform being an accurate stand-in for the true signal — the paper's own injections show that imperfect reconstruction is what pushes lensed pairs away from zero joint null energy and creates the overlap with unlensed pairs.

Editorial extensions

If this is right

A lensing search can bypass joint parameter estimation of the binary: only the sky and lensing parameters are sampled, and a pair costs a few minutes on one core, pointing toward real-time screening of the growing detection catalog.
The posterior on sky position and lensing parameters, obtained directly from the null-energy sampling, gives an electromagnetic follow-up pointing without a separate joint PE analysis.
The formalism stacks any number M of images, so a triplet or quadruplet of lensed images is treated by the same joint null energy as a pair.
Because the waveform can come from wavelet-based reconstructions such as cWB rather than PE, the statistic extends to strongly lensed burst searches, though with reduced sensitivity for compact binary coalescences.
The stated next step—working directly on noisy data using the chi-squared statistics of the null energy, with a time-frequency filter to suppress noise—would remove the dependence on parameter estimation altogether.

Reading between the lines

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

If the separation holds at scale, the joint null energy could serve as a cheap pre-filter that triages candidate pairs for expensive joint parameter estimation in existing lensing searches—an ordering the authors leave implicit.
The overlap between the histograms is attributed to waveform reconstruction error, which makes the method's discrimination a direct function of reconstruction fidelity; a controlled study injecting waveforms with known mismatch (for example, missing precession or tidal effects) would map exactly where the statistic loses power.
The same null-energy logic applied to a single event is already a known glitch veto; the joint version extends that idea to a coherence test between events, so a pair that refuses to form a null stream at any parameters could flag a non-astrophysical member.
A working waveform-independent time-frequency version would make lensed-image identification a streaming operation on the detection pipeline, a regime where the computational saving over PE-based searches grows with the third-generation detector event rate.

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.

Desk Editor's Note

read the letter

First, the useful thing to know: this short note has a real idea. The authors construct a joint null stream for two or more lensed GW images by modifying each image's antenna response matrix by the magnification and Morse phase, then stacking. As far as I can tell from the cited literature, that exact construction hasn't been proposed before. It's a natural extension, but a good one, and the toy injection study (100 lensed vs 1000 unlensed pairs) shows the expected separation. Credit where due: the core formalism in Eqs (1)-(8) is straightforward and the authors are explicit about the limitations—they call the results preliminary and note the overlap is due to imperfect waveform reconstruction.

Now the soft spots. The stress-test concern is real: Eq. (8) does not include the relative time delay δt_{j1}. The lensed waveform (1) has the factor e^{-2πif δt_j}, and without it (or without an explicit time-alignment step before stacking) the joint null stream will not cancel for a genuinely lensed pair. The text mentions 't_j the observed time of coalescence' but doesn't say how the data are aligned across images. If the authors align each data segment to a common geocentric time, they need to show that. If they don't, the statistic in Section 4 doesn't follow from the equations. This is a real mathematical gap, not a presentation nit. The fix is easy—include the phase factor in F or state the alignment—but it has to be done.

The second soft spot is the reliance on PE-derived maxL waveforms. The abstract says the method removes the need for PE, but the demonstration uses PE to get the waveforms. The authors acknowledge this and point to cWB and future waveform-independent work, so it's an honest limitation, but it means the computational advantage claimed in the intro is not yet realized. The paper is a proof of principle, not a search-ready pipeline.

The injection study itself is minimal: one histogram, no significance statement, no comparison to existing searches (e.g., GstLAL or ML methods). For a methods note that's acceptable, but the reader shouldn't mistake it for a full validation.

Who is this for? The strong lensing search community—especially people working on rapid follow-up pipelines. It deserves a serious referee: the idea is novel and the missing time delay is easily fixable. I'd send it to a referee with a flag to check the derivation and ask for a more complete description of the alignment. Reading group? Maybe—it's short enough and the flaw is instructive. Cite? Probably not in my own work until the derivation is corrected. Serious thinker: yes, the paper is internally coherent in its goals and honest about limitations; the gap is an omission, not a deep confusion.

Referee Report

3 major / 6 minor

Summary. The paper proposes a method to search for strongly lensed gravitational-wave events using the null stream of a detector network. For a set of M candidate images, the method stacks the frequency-domain data from all detectors and constructs a joint antenna-response matrix (Eqs. 7-8) that includes relative magnification and Morse phase; a null projector then defines a joint null energy. The claim is that at the correct sky position and lensing parameters the joint null energy vanishes for lensed pairs but not for unrelated events. The authors demonstrate the idea with injections (100 lensed pairs, 1000 unlensed pairs), using maximum-likelihood waveforms from parameter estimation in place of the data, and obtain a histogram with the bulk of lensed pairs near zero. The paper frames this as a basis for a computationally lighter, ultimately waveform-independent search.

Significance. If the formalism is correct, this is an attractive and simple addition to the strong-lensing search toolbox: the null stream is a well-established concept, and the joint construction avoids the need for a fully joint parameter-estimation of the two events and may eventually be applied directly to data or to non-template reconstructions. The authors are transparent about the preliminary nature of the results and identify the main limitation (waveform-reconstruction fidelity). The paper provides a falsifiable prediction: lensed pairs can be separated from unlensed pairs by the near-zero joint null energy. However, the current evidence is minimal and the time-delay treatment in the formalism needs clarification before the central claim can be accepted.

major comments (3)

[Section 4, Eq. (8)] Equation (8) defines the joint response for image j with relative magnification and Morse phase but omits the relative time delay e^{-2π i f δt_j1} that follows from Eq. (1). Without including this factor, or without explicitly aligning the stacked data vectors to a common reference time, the joint null stream will not cancel a genuinely lensed pair at the correct parameters. The text introduces t_j and t_j1 but does not state how the time shift is absorbed. This is a load-bearing issue: if the factor is truly unnecessary because the data are aligned, that must be stated; if not, Eq. (8) must include the phase.
[Section 5] The injection study consists of one histogram (Fig. 1) with no quantitative statistical characterization. No detection efficiency, false-alarm probability, ROC curve, or threshold is computed, and the overlap region is not quantified. The authors attribute the overlap to imperfect maxL waveform reconstruction, which means the discriminating power of the statistic is dependent on reconstruction fidelity rather than the formalism alone. For a proposed detection statistic, this validation is insufficient. Please provide a statistical analysis of the separation (e.g., distributions, significance, or at least the fraction of lensed pairs falling below a given false-alarm threshold).
[Abstract and Section 5] The abstract motivates the method by removing the need for parameter estimation, but the implemented and tested pipeline uses PE maxL waveforms (Section 5). The cWB-based non-PE reconstruction is described only as a preliminary negative result. Thus the advertised computational advantage is not yet demonstrated. Either present a non-PE (or raw-data) validation, or clearly frame the current contribution as a proof of principle that still relies on PE reconstructions.

minor comments (6)

[Section 2, Eq. (1)] Define δt_j explicitly with respect to a reference time (e.g., image 1) to support later relative definitions.
[Section 4] The symbol t_j1 appears without definition; likely δt_j1 is intended.
[Fig. 1] The x-axis label is ambiguous; state that the plotted quantity is the difference E_joint - (E_1 + E_2) or clarify if it is a sum.
[Section 3, Eq. (6)] Specify whether N is the number of positive-frequency bins and note that the chi-squared distribution assumes a known sky position and whitened data.
[Section 5] Typo: 'supertreshold' should be 'suprathreshold'.
[References] A few references are incomplete (e.g., [3] is a preprint, [14] is a software link); consistent formatting is recommended.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the joint null-stream statistic is a self-consistency test built from standard lensing relations; it is not a fitted-input prediction and does not rely on load-bearing self-citation.

full rationale

The paper constructs a joint null stream by stacking lensed-image data and defining a joint antenna response matrix (Eqs. 7–8) that incorporates relative magnification and Morse phase. For a perfect waveform and correct lensing/sky parameters, the lensed signals lie in the subspace spanned by that matrix, so the null energy vanishes by construction. This is a mathematical identity conditional on the standard geometric-optics lensing model (Eq. 1, cited to [3]) and the standard null-stream formalism (Eqs. 3–6, cited to [8,9]); it is not circular because the detection statistic's practical discriminatory power is an empirical question that the paper tests with injections. The PE-derived maxL waveforms are used as a stand-in for the true signal, and Section 5 explicitly attributes the histogram overlap to imperfect waveform reconstruction — a stated limitation, not a circular step. No parameter is fitted to the null-energy outcome and then presented as an independent prediction. The only self-citation is [6] (Janquart et al.), used to support the claim that no compelling lensing evidence exists so far; this is not load-bearing for the derivation. One non-circular correctness concern: Eq. (8) omits the relative lensing time delay e^{-2πi f δt_{j1}} from the joint response matrix, so as written the cancellation claim does not follow unless that delay is included elsewhere or the data are aligned; this is a derivation gap, not circularity.

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

No new physical constants or entities are introduced. The free_parameters list is empty because the lensing parameters (magnification, Morse phase, time shift) are the targets of estimation, not ad hoc fit parameters. The axioms are standard assumptions in GW data analysis, stated or implicitly used in the paper.

assumptions (4)

domain assumption Geometric optics applies to strongly lensed GW images in the relevant band
Stated in Section 2, citing refs [2,3]; the lensed waveform is a magnification, time shift, and Morse phase applied to the unlensed waveform.
domain assumption Detector noise is Gaussian, stationary, and independent across detectors and time segments
Stated in Section 3 (Eq. 2 and following); underlies the chi-squared distribution of the null energy and the stacking of data from different epochs.
domain assumption The detector network has more working detectors than polarizations (D>P=2)
Implicit in the null stream construction in Eq. (4); with D=2 the null projector is null and the statistic collapses.
domain assumption The maximum-likelihood waveform from PE is an unbiased proxy for the true signal
Used throughout Section 5 where the data stream is replaced by the maxL waveform; the paper acknowledges reconstruction residuals.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Leveraging the null stream to detect strongly lensed gravitational waves." pith.science (2026). https://pith.science/paper/S6GMR463

@misc{pith2026250906745,
  author       = {Pith},
  title        = {Pith review of: Leveraging the null stream to detect strongly lensed gravitational waves},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S6GMR463}},
  note         = {Machine review of arXiv:2509.06745}
}

read the original abstract

Gravitational lensing of gravitational waves is expected to be observed in current and future detectors. In view of the growing number of detections, computationally light pipelines are needed. Detection pipelines used in past LIGO-Virgo-KAGRA searches for strong lensing require parameter estimation to be performed on the gravitational wave signal or are machine learning based. Removing the need for parameter estimation in classical methods would alleviate the ever growing demand of computational resources in strong lensing searches and would make real-time analysis possible. We present a novel way of identifying strongly lensed gravitational wave signals, based on the null stream of a detector network. We lay out the basis for this detection method and show preliminary results confirming the validity of the formalism. We also discuss the next development steps, including how to make it independent of parameter estimation.

Figures

Figures reproduced from arXiv: 2509.06745 by the authors.

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Search for strong lensing of gravitational waves in the binary black hole events from O1-O4a
gr-qc 2026-07 accept novelty 6.0 of 10

Posterior Overlap 2.0 finds no lensed BBH pairs in O1–O4a (p_L < 0.6% for all pairs) and sets a 90% upper bound of 1.4% on the strong-lensing fraction.
Identifying lensed gravitational waves with physics-informed posterior learning
gr-qc 2026-07 conditional novelty 6.0 of 10

Fusing a simulation-trained common-source mass posterior with waveform features raises lensed-event detection efficiency from 20.8% to 35.2% at 1% false-positive rate and lowers the SNR for 50% efficiency from 45.3 to 33.5.

Reference graph

Works this paper leans on

14 extracted references · 8 canonical work pages · cited by 2 Pith papers

[1]

Schneider P, Ehlers J and Falco E E 1992Gravitational Lenses

work page
[2]

J.5951039–1051 (Preprintastro-ph/0305055)

Takahashi R and Nakamura T 2003Astrophys. J.5951039–1051 (Preprintastro-ph/0305055)

work page arXiv
[3]

Dai L and Venumadhav T 2017 (Preprint1702.04724)

arXiv 2017
[4]

Hannuksela O A, Haris K, Ng K K Y, Kumar S, Mehta A K, Keitel D, Li T G F and Ajith P 2019 Astrophys. J. Lett.874L2 (Preprint1901.02674)

arXiv 2019
[5]

J.970191 (Preprint2304

Abbott Ret al.(LIGO Scientific, KAGRA, VIRGO) 2024Astrophys. J.970191 (Preprint2304. 08393)

work page
[6]

Janquart Jet al.2023Mon. Not. Roy. Astron. Soc.5263832–3860 (Preprint2306.03827)

arXiv
[7]

Wright Met al.2025 (Preprint2507.20256)

work page Pith review arXiv 2025
[8]

Chatterji S, Lazzarini A, Stein L, Sutton P J, Searle A and Tinto M 2006Phys. Rev. D74082005 (Preprintgr-qc/0605002)

work page arXiv

Show all 14 references

[9]

Phys.12053034 (Preprint0908.3665)

Sutton P Jet al.2010New J. Phys.12053034 (Preprint0908.3665)

arXiv
[10]

Skilling J 2006Bayesian Analysis1833–859
[11]

Acernese Fet al.(VIRGO) 2015Class. Quant. Grav.32024001 (Preprint1408.3978)

arXiv
[12]

Capote Eet al.2025Phys. Rev. D111(6) 062002 URLhttps://link.aps.org/doi/10.1103/ PhysRevD.111.062002
[13]

C ¸ alı¸ skan M, Ezquiaga J M, Hannuksela O A and Holz D E 2023Phys. Rev. D107063023 (Preprint 2201.04619)

arXiv
[14]

Klimenko S, Vedovato G, Necula V, Salemi F, Drago M, Chassande-Mottin E, Tiwari V, Lazzaro C, O’Brian B, Szczepanczyk M, Tiwari S and Gayathri V 2021 cwb pipeline library: 6.4.0 URL https://doi.org/10.5281/zenodo.4419902

2021 doi

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

Schneider P, Ehlers J and Falco E E 1992Gravitational Lenses

work page

[2] [2]

J.5951039–1051 (Preprintastro-ph/0305055)

Takahashi R and Nakamura T 2003Astrophys. J.5951039–1051 (Preprintastro-ph/0305055)

work page arXiv

[3] [3]

Dai L and Venumadhav T 2017 (Preprint1702.04724)

arXiv 2017

[4] [4]

Hannuksela O A, Haris K, Ng K K Y, Kumar S, Mehta A K, Keitel D, Li T G F and Ajith P 2019 Astrophys. J. Lett.874L2 (Preprint1901.02674)

arXiv 2019

[5] [5]

J.970191 (Preprint2304

Abbott Ret al.(LIGO Scientific, KAGRA, VIRGO) 2024Astrophys. J.970191 (Preprint2304. 08393)

work page

[6] [6]

Janquart Jet al.2023Mon. Not. Roy. Astron. Soc.5263832–3860 (Preprint2306.03827)

arXiv

[7] [7]

Wright Met al.2025 (Preprint2507.20256)

work page Pith review arXiv 2025

[8] [8]

Chatterji S, Lazzarini A, Stein L, Sutton P J, Searle A and Tinto M 2006Phys. Rev. D74082005 (Preprintgr-qc/0605002)

work page arXiv