Can we constrain cosmology with compound gravitational lenses?
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.
Vera: I'm Vera.
Jocelyn: And I'm Jocelyn, here with Subrahmanyan.
Vera: We're looking at the paper: "Can we constrain cosmology with compound gravitational lenses?".
Jocelyn: Here's the gist of the paper. Compound lensing in double-source-plane lenses (DSPLs) introduces multiple mass-sheet transformations that bias the inferred distance ratios used to constrain cosmology<ref:2611.15714#pg4>.
Vera: The recovered distance ratio from lens models of simulated compound lenses showed a strong correlation between the stellar-to-halo mass and the recovered beta<ref:2611.15714#pg10>.
Jocelyn: The study found that while cosmology was biased when using the full mock Euclid sample, it was reduced substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2sigma and ∼ 1sigma of the ground-truth cosmology<ref:2611.15714#pg10>.
Vera: Therefore, for extreme compound lensing, the model complexity must be sufficient to reproduce the lensing observations<ref:2611.15714#pg4>.
Paper summary: Vera: So, to recap this paper on "Can we constrain cosmology with compound gravitational lenses?", they argue that multiple mass-sheet transformations in double-source-plane lenses introduce biases into the distance ratios used for cosmological constraints <ref:2611.15714#pg4>.
Jocelyn: The core claim is that these biases are significant, especially since we rely on those distance ratios to pin down parameters like w zero and omega m in a flat wCDM cosmology <ref:2611.15714#pg3> <ref:2611.15714#pg1>.
Subrahmanyan: It matters because strong lenses are becoming a promising independent probe for cosmology, and they found that if you don't account for the mass distribution complexities, you get skewed results <ref:2611.15714#pg2>.
Vera: They explored this by looking at how deflection by the first lens can introduce a second mass-sheet transformation within the first source plane which is often ignored <ref:2611.15714#pg3>.
Jocelyn: Essentially, they’re showing that for these systems to be useful cosmological probes, we need to model them with enough complexity to capture all these transformations <ref:2611.15714#pg4>.
Subrahmanyan: The paper suggests that while cosmology can be constrained, it depends heavily on the accuracy of the lens modeling used, particularly regarding the mass distribution of the deflector galaxies <ref:2611.15714#pg10>.
Vera: They looked at simulated observations based on a specific SDSS source, J0946+one thousand six and varied how they modeled that source's matter distribution to see if that systematic bias propagated to a more varied population <ref:2611.15714#pg4> <ref:2611.15714#pg1,to a more varied population>.
Jocelyn: So they weren't just looking at one case; they were testing whether this issue was a fluke or a general problem across the whole sample of DSPLs <ref:2611.15714#pg4>.
Subrahmanyan: The implication here is that future surveys like Euclid need to be aware of these specific modeling challenges when trying to extract cosmological information from strong lenses <ref:2611.15714#pg10>.
Conclusion: Vera: So, thinking about the title "Can we constrain cosmology with compound gravitational lenses?", what it really means is that these complex lensing systems are a potential tool for measuring dark energy, but only if we handle their internal mass structure correctly <ref:2611.15714#pg4>.
Jocelyn: It’s not just about finding more lenses; it’s about understanding the systematic uncertainties inherent in using them as standard cosmological probes <ref:2611.15714#pg3>.
Subrahmanyan: The paper points to the need for sophisticated lens modeling, specifically that if you want to constrain dark energy with DSPLs, you must ensure your model complexity is sufficient to reproduce the observed lensing signals <ref:2611.15714#pg4>.
Vera: They showed that even with careful filtering of a gold sub-sample, they still have a tension with our true cosmological values in some areas, so we can't just assume these lenses will give us perfect results <ref:2611.15714#pg10>.
Jocelyn: The authors are suggesting that the next step is to build better models for these complex systems so we can reduce those biases and get tighter constraints on parameters like w zero and omega m <ref:2611.15714#pg10> <ref:2611.15714#pg1>.
Subrahmanyan: Ultimately, this work shows that DSPLs are a promising avenue, but the success of that probe hinges entirely on our ability to accurately model the mass distributions involved <ref:2611.15714#pg4>.
Department of Physics, University of Oxford · Institute of Cosmology and Gravitation, University of Portsmouth
astro-ph.CO
Submitted: 2026-11-11
Updated: 2026-11-11
Comments: 16 pages + Appendix, 20 figures. Submitted for publication in the Monthly Notices of the Royal Astronomical Society (MNRAS)
Code: https://github.com/tcollett/LensPop
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 88/100
The gist: Compound lensing in double-source-plane lenses (DSPLs) introduces multiple mass-sheet transformations that bias the inferred distance ratios used to constrain cosmology<ref:2611.15714#pg4>.
Key concepts
- Compound Lensing
- This occurs in strong lenses where two or more deflector galaxies are located along the line of sight at different redshifts. It involves multiple source planes, meaning the light from a single source passes through several gravitational potentials, creating complex lensing effects that are hard to model accurately.
- Mass-Sheet Transformation (MST)
- This is a systematic uncertainty in strong lensing analysis where an incorrect assumption about the mass distribution of the lens galaxy causes errors. Compound lensing introduces an additional MST within the first source plane that must be accounted for, often ignored in simpler models.
- Distance Ratio ($eta$)
- This parameter relates angular positions of a source across different redshift planes. It is used to constrain cosmological parameters like dark energy's equation of state ($w_0$) and matter density ($\Omega_m$). The study shows $\beta$ is sensitive to the assumed mass model, especially for extreme compound lensing systems.
Terminology
Summary
Compound lensing in double-source-plane lenses (DSPLs) introduces multiple mass-sheet transformations that bias the inferred distance ratios used to constrain cosmology<ref:2611. p. 0 Collett2
and Philip Holloway2
1Department of Physics, University of Oxford, Keble Road, Oxford OX1 3RH, UK
2
Institute of Cosmology and Gravitation, University of Portsmouth, Burnaby Road, Portsmouth PO1 3FX, UK
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Compound lensing in double-source-plane lenses (DSPLs) introduces multiple mass-sheet transformations that bias the inferred
distance ratios 𝛽 used to constrain cosmology. We examined simulated
observations based on SDSS J0946+1006, in which we varied the matter distribution of source 1. We found a strong correlation
between the stellar-to-halo mass and the recovered 𝛽. We examined simulated
observations based on SDSS J0946+1006, in which we varied the matter distribution of source 1. We found a strong correlation
between the stellar-to-halo mass and the recovered 𝛽. We inferred 𝑤0 and Ωm for a flat
𝑤CDM cosmology from our 𝛽 posteriors for our combined sample. We found the cosmology biased when using the full mock
Euclid sample, but reduced the bias substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2𝜎
and ∼ 1𝜎 of the ground-truth cosmology for simulated Euclid and Hubble Space Telescope imaging, respectively. We inferred 𝑤0 and Ωm for a flat
𝑤CDM cosmology from our 𝛽 posteriors for our combined sample. We found the cosmology biased when using the full mock
Euclid sample, but reduced the bias substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2𝜎
and ∼ 1𝜎 of the ground-truth cosmology for simulated Euclid and Hubble Space Telescope imaging, respectively. We found the cosmology biased when using the full mock
Euclid sample, but reduced the bias substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2𝜎
and ∼ 1𝜎 of the ground-truth cosmology for simulated Euclid and Hubble Space Telescope imaging, respectively. Therefore, we
must eliminate systems with unreliable lensing information, and for extreme compound lensing, the model complexity must be
sufficient to reproduce the lensing observations. We found the cosmology biased when using the full mock
Euclid sample, but reduced the bias substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2𝜎
and ∼ 1𝜎 of the ground-truth cosmology for simulated Euclid and Hubble Space Telescope imaging, respectively. Therefore, we
must eliminate systems with unreliable lensing information, and for extreme compound lensing, the model complexity must be
sufficient to reproduce the lensing observations.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Modeling Compound Lensing Theory
Compound gravitational lensing occurs in multi-source-plane strong lenses where two or more deflector galaxies are located along the LOS in different redshift planes<ref:2611.15714#pg3>. The unlensed angular position of a source in the jth redshift plane theta j can be expressed as a function of the observed angular position in the i-th redshift plane theta i by equation (1)<ref:2611. p. 1 The lens equation, which describes the deflection of
the true source position to the observed position, can be generalised
to compound lenses with any number of deflector planes (Schneider
2019). The unlensed angular position of a source in the 𝑗
𝑡 ℎ redshift
plane 𝜃 𝑗 can be expressed as a function of the observed angular
position in the 𝑖
𝑡 ℎ redshift plane 𝜃𝑖 by:
𝜃 𝑗 = 𝜃1 −
∑︁
𝑗−1
𝑖=1
𝛽𝑖 𝑗𝛼
′
𝑖
(𝜃𝑖), (1)
where 𝜃1 is the observed angular position of the source in the image
plane (1
𝑠𝑡 deflector plane), and 𝛼
′
𝑖
is the scaled deflection angle of
the 𝑖
𝑡 ℎ deflector plane, a factor of 𝐷𝑖𝑠
𝐷𝑠
times the physical deflection
angle 𝛼ˆ𝑖
. In this work, we will be considering the simplest compound lensing
case where there are two source planes. Therefore, we can simplify
Equation 1 to write the double-source-plane lens equation:
𝜃3 = 𝜃1 − 𝛼
′
1
(𝜃1) − 𝛼
′
2
(𝜃2), (3)
which describes the unlensed angular position of S2 𝜃3 as a function
of its observed position 𝜃1, the scaled deflection angle of the lens 𝛼
′
1
,
the scaled deflection angle of S1 𝛼
′
2
and the angular position in the
first source plane 𝜃2 given by:
𝜃2 = 𝜃1 − 𝛽12𝛼
′
1
(𝜃1), (4)
where 𝛽12 is the distance ratio;
𝛽12 =
𝐷12𝐷3
𝐷2𝐷13
, (5)
henceforth referred to as 𝛽. The details of this lens modelling are given in Section 4. 1 Cosmological Inference
Within a flat 𝑤CDM cosmology, the angular diameter distance between redshifts 𝑧𝑖 and 𝑧𝑗 along the LOS is given by:
𝐷𝑖 𝑗 =
𝑐
𝐻0 (1 + 𝑧𝑗)
∫ 𝑧 𝑗
𝑧𝑖
𝑑𝑧
𝐸(𝑧)
, (6)
where 𝑐 is the speed of light, 𝐻0 is the Hubble constant and the
normalized Hubble parameter 𝐸(𝑧) ≡ 𝐻(𝑧)/𝐻0 can be expressed as:
𝐸(𝑧) =
√︁
Ωm(1 + 𝑧)
3 + (1 − Ωm) (1 + 𝑧)
3(1+𝑤)
, (7)
where Ωm is the matter density parameter and 𝑤0 is the dark energy
equation of state parameter. 1 Cosmological Inference
Within a flat 𝑤CDM cosmology, the angular diameter distance between redshifts 𝑧𝑖 and 𝑧𝑗 along the LOS is given by:
𝐷𝑖 𝑗 =
𝑐
𝐻0 (1 + 𝑧𝑗)
∫ 𝑧 𝑗
𝑧𝑖
𝑑𝑧
𝐸(𝑧)
, (6)
where 𝑐 is the speed of light, 𝐻0 is the Hubble constant and the
normalized Hubble parameter 𝐸(𝑧) ≡ 𝐻(𝑧)/𝐻0 can be expressed as:
𝐸(𝑧) =
√︁
Ωm(1 + 𝑧)
3 + (1 − Ωm) (1 + 𝑧)
3(1+𝑤)
, (7)
where Ωm is the matter density parameter and 𝑤0 is the dark energy
equation of state parameter. Consequently, 𝛽 is a function of only
redshift, 𝑤0 and Ωm due to the factor of 𝐻0 cancelling.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Systematic Bias Analysis
The mass-sheet transformation (MST) is a major source of systematic uncertainty in strong lensing analyses<ref:2611.15714#pg3>. Compound lensing introduces a second MST within the first source plane, which is often ignored and has not been studied in detail<ref:2611. p. 1 2020; Fleury
et al. Compound lensing introduces a second MST within
the first source plane (Schneider 2014), which is often ignored and
has not been studied in detail. However, the differences between a true galaxy and commonly used
parameterisations leave room for the MST. In this work, we explored the bias that deflection by S1 can introduce when the ground-truth mass distribution is more complex
than our model. In this work, we explored the bias that deflection by S1 can introduce when the ground-truth mass distribution is more complex
than our model. The difference between the true and model profiles
allows for an internal MST that modifies the Einstein radius and density slope of a mass profile when not explicitly parameterised.Where the paper says it
Where the paper says it
Where the paper says it
Data Population and Constraints
The study produced a simulated catalogue of DSPLs detectable by the Euclid telescope to see if the same systematic propagates to a more varied population<ref:2611.15714#pg4>. They selected a gold sample of 20 compound lenses based on criteria that reduce the spread in inferred values of beta<ref:2611. p. 9 2 Euclid population
From our LenPop catalogue, we found 165 detectable compound
lenses (𝜃E,s1 > 0. 1) out of 2236 detectable DSPLs within 4200 deg2
of the sky (30% of the EWS) based on our criteria from Section
3. The details of this lens modelling are given in Section 4. 1 Cosmological Inference
Within a flat 𝑤CDM cosmology, the angular diameter distance between redshifts 𝑧𝑖 and 𝑧𝑗 along the LOS is given by:
𝐷𝑖 𝑗 =
𝑐
𝐻0 (1 + 𝑧𝑗)
∫ 𝑧 𝑗
𝑧𝑖
𝑑𝑧
𝐸(𝑧)
, (6)
where 𝑐 is the speed of light, 𝐻0 is the Hubble constant and the
normalized Hubble parameter 𝐸(𝑧) ≡ 𝐻(𝑧)/𝐻0 can be expressed as:
𝐸(𝑧) =
√︁
Ωm(1 + 𝑧)
3 + (1 − Ωm) (1 + 𝑧)
3(1+𝑤)
, (7)
where Ωm is the matter density parameter and 𝑤0 is the dark energy
equation of state parameter.Where the paper says it
Where the paper says it
Cosmological Results
Using all 74 of their successfully modelled Euclid simulations failed to produce physically plausible cosmological constraints due to significant bias in some models<ref:2611.15714#pg10>. By filtering out systems with tiny Einstein radii, overlapping sources, fewer images and undetectable counters, they produced a gold sample of 19 simulated compound lenses which de-biased their inferred cosmology, leaving a ∼ 2sigma tension with the truth but with a large credible region within the w0−Ωm parameter space<ref:2611. p. 0 We inferred 𝑤0 and Ωm for a flat
𝑤CDM cosmology from our 𝛽 posteriors for our combined sample. We found the cosmology biased when using the full mock
Euclid sample, but reduced the bias substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2𝜎
and ∼ 1𝜎 of the ground-truth cosmology for simulated Euclid and Hubble Space Telescope imaging, respectively.Where the paper says it
Conclusion and Outlook
DSPLs are promising potential probes of dark energy that can leverage distance ratios to constrain cosmological parameters<ref:2611.15714#pg4>. The findings suggest that S1 must be treated with care if one wishes to use DSPLs as a future probe of dark energy<ref:2611. p. 1 We discuss the implications of our findings in
Section 6, provide recommendations for future cosmological DSPL
studies and finally summarise these findings in Section 7. Throughout
the simulations presented in this work, we use a consistent cosmological model of a flat ΛCDM cosmology with a matter density fraction
of Ωm = 0.Where the paper says it
Pruning redundant reasoning tokens
Compound lensing in double-source-plane lenses (DSPLs) introduces multiple mass-sheet transformations that bias the inferred distance ratios used to constrain cosmology<ref:2611.15714#pg4>. The recovered distance ratio from lens models of simulated compound lenses showed a strong correlation between the stellar-to-halo mass and the recovered beta<ref:2611.15714#pg10>. The study found that while cosmology was biased when using the full mock Euclid sample, it was reduced substantially using a limited, well-modelled “gold” sub-sample, bringing it to within ∼ 2sigma and ∼ 1sigma of the ground-truth cosmology<ref:2611.15714#pg10>. Therefore, for extreme compound lensing, the model complexity must be sufficient to reproduce the lensing observations<ref:2611.15714#pg4>.
How the training signal is built
The study produced a simulated catalogue of DSPLs detectable by the Euclid telescope to see if the same systematic propagates to a more varied population<ref:2611.15714#pg4>. They examined simulated observations based on SDSS J0946+1006, in which they varied the matter distribution of source 1<ref:2611. p. 0 We present the recovered 𝛽 from lens models of simulated compound lenses
and explore the systematic biases that arise from a mass-sheet transformation in the first source plane. We examined simulated
observations based on SDSS J0946+1006, in which we varied the matter distribution of source 1. We found a strong correlation
between the stellar-to-halo mass and the recovered 𝛽. We produced a simulated catalogue of DSPLs detectable by the Euclid
telescope to see if the same systematic propagates to a more varied population.Where the paper says it
Where the paper says it
Where the gains come from
The study found that for systems where M∗/Mh ∈ [0.01, 0.03], beta is more accurately recovered as 0. p. 8 04, but with less significance within the range of values tested. By splitting 𝑀∗/𝑀h into three regimes, we can identify
the range over which 𝛽 is recovered accurately. 04, but with less significance within the range of values tested. By splitting 𝑀∗/𝑀h into three regimes, we can identify
the range over which 𝛽 is recovered accurately. 04, but with less significance within the range of values tested. By splitting 𝑀∗/𝑀h into three regimes, we can identify
the range over which 𝛽 is recovered accurately. The improvement for DSPL is especially significant since we will
transition from a handful of known systems to > 103
candidates from
Euclid (Collett 2015; Euclid Collaboration 2025b), giving us the advantage of large-number statistics previously unavailable. Spectroscopic surveys will provide the information required to confidently
© 2026 The Authors
arXiv:2611. The improvement for DSPL is especially significant since we will
transition from a handful of known systems to > 103
candidates from
Euclid (Collett 2015; Euclid Collaboration 2025b), giving us the advantage of large-number statistics previously unavailable. Spectroscopic surveys will provide the information required to confidently
© 2026 The Authors
arXiv:2611. The improvement for DSPL is especially significant since we will
transition from a handful of known systems to > 103
candidates from
Euclid (Collett 2015; Euclid Collaboration 2025b), giving us the advantage of large-number statistics previously unavailable. Spectroscopic surveys will provide the information required to confidently
© 2026 The Authors
arXiv:2611. The improvement for DSPL is especially significant since we will
transition from a handful of known systems to > 103
candidates from
Euclid (Collett 2015; Euclid Collaboration 2025b), giving us the advantage of large-number statistics previously unavailable. Spectroscopic surveys will provide the information required to confidently
© 2026 The Authors
arXiv:2611. The improvement for DSPL is especially significant since we will
transition from a handful of known systems to > 103
candidates from
Euclid (Collett 2015; Euclid Collaboration 2025b), giving us the advantage of large-number statistics previously unavailable. Spectroscopic surveys will provide the information required to confidently
© 2026 The Authors
arXiv:2611. 03], 𝛽 is more accurately recovered as 0. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03], 𝛽 is more accurately recovered as 0. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03], 𝛽 is more accurately recovered as 0. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03], 𝛽 is more accurately recovered as 0. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03], 𝛽 is more accurately recovered as 0. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03] but less so outside this range. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03] but less so outside this range.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The study found that for systems where M∗/Mh ∈ [0.01, 0.03], beta is more accurately recovered as 0.726±0.007<ref:2611.15714#pg10>. This indicates that their mass model profile is a suitable approximation to the NFW + Sérsic ground truth for M∗/Mh ∈ [0.01, 0.03] but less so outside this range<ref:2611.15714#pg10>.
The study found that for systems where M∗/Mh ∈ [0.01, 0.03], beta is more accurately recovered as 0.726±0.007<ref:2611.15714#pg10>. This indicates that their mass model profile is a suitable approximation to the NFW + Sérsic ground truth for M∗/Mh ∈ [0.01, 3] but less so outside this range<ref:2611. p. 8 This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03] but less so outside this range. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03] but less so outside this range. This indicates that our mass model
profile is a suitable approximation to the NFW + Sérsic ground truth
for 𝑀∗/𝑀h ∈ [0. 03] but less so outside this range.Where the paper says it
Where the paper says it
Where the paper says it
The study found that for systems where M∗/Mh ∈ [0.01, 3], beta is more accurately recovered as 0. p. 8 04, but with less significance within the range of values tested. By splitting 𝑀∗/𝑀h into three regimes, we can identify
the range over which 𝛽 is recovered accurately. 04, but with less significance within the range of values tested. By splitting 𝑀∗/𝑀h into three regimes, we can identify
the range over which 𝛽 is recovered accurately.Where the paper says it
Where the paper says it
The study found that for systems where M∗/Mh ∈ [0.01, 3], beta is more accurately recovered as 0.726±0.007<ref:2611.15714#pg10>. This indicates that their mass model profile is a suitable approximation to the NFW + Sérsic ground truth for M∗/Mh ∈ [0.01, 3] but less so outside this range<ref:2611.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
System can recover cosmological parameters w0 and Omega m from lens model inference of distance ratio by using Equations 5-7, which are derived from a flat wCDM cosmology. This allows for constraints on dark energy equation of state parameter w0 and matter density parameter Omega m through the posterior distributions of the inferred distance ratio parameter b.
-
AI system can identify and mitigate systematic biases arising from mass-sheet transformations (MST) by explicitly marginalizing them, as the paper suggests that
With the addition of stellar kinematic measurements and galaxy number counts, several studies have shown that the MST can be constrained and explicitly marginalised.
-
The improved system can perform end-to-end simulations of compound lenses to explore the bias associated with the first source plane in a Bayesian framework, specifically by examining how
the difference between the true and model profiles allows for an internal MST that modifies the Einstein radius and density slope of a mass profile when not explicitly parameterised.
-
The system can be trained to distinguish between different regimes of galaxy mass-to-halo ratios (Mstar/Mh) to identify where the inferred distance ratio b is recovered accurately, noting that
for values of Mstar/Mh in [0.01, 0.03] b is more accurately recovered as 0.726±0.007.
-
The system can be adapted to use high-resolution follow-up observations to improve constraints on cosmological parameters, as demonstrated by the fact that
Using simulated HST observations of our gold sample, we improved the precision of our b posteriors, reducing the mean standard deviation by 37%.
Abstract
Compound lensing in double-source-plane lenses (DSPLs) introduces multiple mass-sheet transformations that bias the inferred distance ratios β used to constrain cosmology. We present the recovered β from lens models of simulated compound lenses and explore the systematic biases that arise from a mass-sheet transformation in the first source plane. We examined simulated observations based on SDSS J0946+1006, in which we varied the matter distribution of source 1. We found a strong correlation between the stellar-to-halo mass and the recovered β. We produced a simulated catalogue of DSPLs detectable by the Euclid telescope to see if the same systematic propagates to a more varied population. But at Euclid's resolution, the properties of source 1 were poorly constrained, and any correlated bias was lost in the statistical uncertainty. We inferred w 0 and Ω m for a flat w CDM cosmology from our β posteriors for our combined sample. We found the cosmology biased when using the full mock Euclid sample, but reduced the bias substantially using a limited, well-modelled ''gold'' sub-sample, bringing it to within about2σ and about1σ of the ground-truth cosmology for simulated Euclid and Hubble Space Telescope imaging, respectively. Therefore, we must eliminate systems with unreliable lensing information, and for extreme compound lensing, the model complexity must be sufficient to reproduce the lensing observations. DSPLs remain a promising cosmological probe that could provide independent and complementary constraints to existing probes.
Sources
- TDCOSMO XXXI: New techniques in line-of-sight studies of time delay lenses
- A Catalog of Mass Models for Gravitational Lensing
- Spatially Resolved Kinematics of SLACS Lens Galaxies. II: Breaking Degeneracies with Lensing and Dynamical Models
- Disentangling the dark and stellar mass through precise lens modelling of the JWST observation of lensed quasar WFI2033--4723
- MAGNUS III: Mild evolution of the total density slope in massive early-type galaxies since z$\sim$1 from dynamical modeling of MUSE integral-field stellar kinematics
- TDCOSMO XXV: Measuring H0 to 6.5% precision with quasar strong lensing and maximally flexible mass-sheet dynamics
- TDCOSMO 2026: A 3% measurement of the Hubble constant from strong lensing and stellar kinematics
- Automated Variational Inference in Probabilistic Programming
- AdaBelief Optimizer: Adapting Stepsizes by the Belief in Observed Gradients
Related papers
- 21cmEMUv3: a hybrid diffusion-LSTM emulator of 21cmFAST summary observables
- Connecting field-level and n-point function measurements of galaxy bias
- The Impact of Cosmic Variance and Satellites on JWST Clustering Measurements at Redshift around 6
- Universality of Primordial Anisotropies in Gravitational Wave Background
- KiDS-1000 cosmic shear reanalysis using MetaCalibration
- 2Flow2SN: Dusty Supernova Cosmology with Photometric Redshifts and Contamination using Neural Simulation-Based Inference