Seidr update: photonic 'black magic' for high-contrast interferometry using kernel-nulling and photonic lanterns

Adam K Taras; Akira Rodziewicz-Ryan; Barnaby Norris; Christopher H Betters; Daniel S Dahl; Frantz Martinache; Jin Wei; Julia J Bryant; Jyotirmay Paul; Marc-Antoine Martinod

Hybrid photonic lanterns plus a transformer network raise the fraction of starlight injected into a kernel-nuller by 28 percent in simulation, tightening the path to high-contrast H-band interferometry.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 03:34 UTC pith:KJYIQLCE

load-bearing objection Solid Asgard instrument update: open-loop 28% lantern injection gain is cleanly simulated, but remains idealized and not yet closed-loop or lab-validated. the 2 major comments →

arxiv 2607.03280 v2 pith:KJYIQLCE submitted 2026-07-03 physics.optics astro-ph.IM

Seidr update: photonic 'black magic' for high-contrast interferometry using kernel-nulling and photonic lanterns

Nathan K Long , Daniel S Dahl , Christopher H Betters , Julia J Bryant , Nick Cvetojevic , Michael J Ireland , Stefan Kraus , Sergio Leon-Saval

show 8 more authors

Frantz Martinache Marc-Antoine Martinod Barnaby Norris Jyotirmay Paul Akira Rodziewicz-Ryan Adam K Taras Jin Wei Peter G Tuthill

This is my paper

classification physics.optics astro-ph.IM

keywords kernel-nullingphotonic lanternhybrid mode-selectivewavefront sensingneural networkVLTIhigh-contrast interferometryH-band

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Seidr is an H-band kernel-nulling beam combiner for the VLTI that sits inside the Asgard suite. Its key design choice is to feed each telescope beam into a hybrid mode-selective photonic lantern before the nulling chip: the lantern funnels the fundamental mode into a single science core while the remaining cores measure residual wavefront errors. A transformer neural network maps those core powers into a wavefront correction that is sent to an upstream deformable mirror. Under realistic atmospheric turbulence the open-loop correction lifts the mean fraction of light in the science core from 0.55 to 0.71. Because kernel observables are already robust to second-order phase errors, the extra light and intensity stability translate into a lower noise floor and better contrast for young giant planets, circumstellar dust, and, eventually, warm exomoons.

Core claim

Open-loop wavefront estimation that uses the intensity outputs of hybrid mode-selective photonic lanterns can increase the mean power ratio in the mode-selective core by 28 percent (0.55 to 0.71) under Von Kármán seeing, thereby improving light injection into a kernel-nulling photonic chip and reducing residual intensity fluctuations that set the null depth.

What carries the argument

The hybrid mode-selective photonic lantern (HMS-PL) transfer matrix that maps pupil-plane phase into six single-mode core powers, together with a transformer network that inverts those powers into a deformable-mirror command maximizing the fundamental-mode core.

Load-bearing premise

That open-loop, noise-free simulations of lantern powers already capture the residual non-common-path aberrations, detector noise and closed-loop dynamics that will exist between the fringe tracker, the deformable mirror and Seidr on the sky.

What would settle it

On-sky or closed-loop laboratory measurements that show the mode-selective-core power ratio remaining near 0.55 after the transformer correction is applied, or that show no narrowing of the kernel-output distributions when intensity fluctuations are reduced.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

Solid Asgard instrument update: open-loop 28% lantern injection gain is cleanly simulated, but remains idealized and not yet closed-loop or lab-validated.

read the letter

This is a straightforward instrument-update paper for Seidr, the H-band kernel-nuller sitting inside the Asgard suite at VLTI. The one new quantitative result is the open-loop transformer-NN correction of 6-core hybrid mode-selective photonic lantern powers, which lifts the mean mode-selective-core power ratio from 0.55 to 0.71 under standard Von Kármán parameters (r0 = 0.4 m, L0 = 10 m, 10 m/s wind). That 28 % figure is clearly stated, backed by a BPM-derived transfer matrix and a clean PDF comparison, and is then used to motivate narrower kernel-output distributions at reduced intensity noise.

What the paper does well is the integration story. It shows how four HMS-PLs feed both the kernel-nulling chip and Baldr’s DM, re-uses the established mixing and kernel operators from Martinache & Ireland, and places Seidr cleanly relative to Heimdallr, Baldr, Bifrost and Nott. The lantern design (central MS core twice the radius of the five WFS cores) and the intensity/phase transfer matrices are documented carefully enough that someone could reproduce the simulation chain. Citations to the prior lantern and kernel-nulling literature are appropriate; the self-cites are just the Asgard suite papers one would expect.

The soft spot is exactly what the authors flag: everything is open-loop and noise-free. No detector noise, no residual NCPA between Heimdallr/Baldr and Seidr, no closed-loop dynamics, and no accounting for the SNR drop that occurs once light is concentrated into the mode-selective core. The subsequent Monte-Carlo kernel-width plots therefore illustrate a potential benefit rather than a demonstrated on-sky improvement. That is a real limitation for an instrument paper, but it is not hidden and does not break the internal logic of the simulations.

This is useful reading for anyone working on photonic injection, nulling interferometry, or VLTI instrumentation. It is not a discovery paper and does not reorganize the field, but it is a clear progress report with a reproducible numerical claim. I would send it to referees; they will ask for the closed-loop and lab steps that are already planned, but the present material is solid enough to deserve that scrutiny. Worth engaging if you care about the Asgard path or lantern-based sensing.

Referee Report

2 major / 5 minor

Summary. This instrument-update paper describes Seidr, an H-band kernel-nulling beam combiner for the Asgard suite at the VLTI. It combines four hybrid mode-selective photonic lanterns (HMS-PLs) for simultaneous science injection and wavefront sensing with a photonic kernel-nulling chip. The authors present a finite-difference BPM transfer matrix for a 6-core HMS-PL, open-loop transformer-NN wavefront estimation that raises the mean mode-selective-core power ratio from 0.55 to 0.71 under Von Kármán seeing (r0 = 0.4 m), and simulations of kernel-output distributions under residual phase and intensity noise. The work positions Seidr as complementary to Nott and reliant on Heimdallr/Baldr infrastructure.

Significance. If the HMS-PL + NN injection scheme survives closed-loop and on-sky conditions, Seidr would provide a practical route to deeper H-band nulls at the VLTI by reducing non-common-path intensity fluctuations that limit kernel-nulling contrast. The paper supplies a concrete transfer-matrix characterization, a quantified open-loop injection gain (Fig. 6), and a clear link between intensity stabilization and kernel-output narrowing (Fig. 8). These are useful design milestones for the Asgard community even though they remain simulation-only. The architecture itself (mode-selective lanterns feeding a kernel-nuller with NN feedback to Baldr) is a coherent and novel instrument concept.

major comments (2)

Section 4 and Fig. 6: the central numerical claim (mean mode-selective-core power ratio rising from 0.55 to 0.71) is obtained from open-loop, noise-free transfer-matrix simulations. The authors themselves state that the ratio “does not represent a closed-loop AO system” and does not account for overall light injected into the chip. Residual NCPA between Heimdallr/Baldr and Seidr, photon/read noise on the five WFS cores, and the SNR degradation that accompanies concentration of light into the mode-selective core are therefore untested. Without at least a closed-loop residual or a noise-injection study, the 28 % figure cannot be taken as a reliable predictor of on-sky null depth.
Section 5 and Fig. 8: the claimed narrowing of kernel-output distributions for σ_I = 0.05 and 0.02 is presented as a potential improvement offered by the HMS-PLs, yet no quantitative mapping from the open-loop lantern correction of §4 to these σ_I values is given. The link between the demonstrated injection gain and the assumed intensity-noise reductions remains illustrative rather than demonstrated; a short calculation or Monte-Carlo chain connecting the two would make the contrast argument load-bearing.

minor comments (5)

Equation (1): the mixing operator M is written with a free phase offset θ that is never specified numerically; a brief statement of the value used (or that results are independent of θ) would aid reproducibility.
Table 2 and Fig. 3: the lantern is characterized only at λ = 1550 nm; a short remark on chromatic behaviour across the H-band would strengthen the design claim.
Figure 4 caption and surrounding text: the example phase screen is shown without stating the corresponding r0 or residual RMS; adding these numbers would make the illustration quantitative.
Section 2: the interaction with Bifrost’s C-RED One camera and the routing of the bright nuller output into the same NN are mentioned only briefly; a one-sentence clarification of the optical path would help readers unfamiliar with Asgard.
References: several Asgard companion papers are cited as SPIE proceedings; ensuring the most recent arXiv or journal versions are listed would improve accessibility.

Circularity Check

0 steps flagged

No load-bearing circularity; open-loop NN injection gain and kernel narrowing are measured outcomes of independent transfer-matrix + Von Kármán simulations, not tautologies or forced fits.

full rationale

The paper is an instrument-design update whose central numerical claims (mean mode-selective-core power ratio rising from 0.55 to 0.71 under stated Von Kármán parameters; narrowing of kernel-output distributions under reduced σ_I) are generated by forward simulation: a finite-difference BPM transfer matrix A for the 6-core HMS-PL, injection of known phase screens, training of a transformer NN on the resulting (P_out, wavefront) pairs, open-loop application of the estimated correction, and re-measurement of the power ratio (Fig. 6) and kernel statistics (Fig. 8). These steps do not reduce by construction to their inputs; the improvement is an empirical outcome of the trained model on held-out turbulence realizations. Kernel-nulling operators M and K are taken from the independent literature (Martinache & Ireland 2018; Chingaipe et al. 2023) and are not redefined here. Self-citations to the Asgard suite papers and to the authors’ companion SPIE abstract [14] supply context and defer detailed NN architecture, but are not used to justify uniqueness or to forbid alternatives; the numerical results stand on the simulations presented. No fitted parameter is re-labeled a prediction, no uniqueness theorem is imported, and no ansatz is smuggled. Minor self-citation of the instrument suite is normal and non-load-bearing, yielding a score of 1.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 1 invented entities

The performance claims rest on a standard atmospheric model, a linear lantern transfer matrix obtained by BPM, and the established algebraic properties of kernel-nulling. Free parameters are the turbulence and residual-error numbers chosen for the Monte-Carlo runs; no new physical entities are postulated.

free parameters (6)

Fried parameter r0 = 0.4 m
Set to 0.4 m for the Von Kármán seeing case that produces the 28 % injection gain; chosen by hand as a representative H-band value.
outer scale L0 = 10 m
Set to 10 m; standard but free choice that affects the low-order content of the phase screens.
transverse wind speed = 10 m/s
10 m/s under Taylor frozen-flow hypothesis; free parameter controlling temporal sequence statistics.
residual phase error sigma_phi = 50 nm / 25 nm
50 nm (Heimdallr baseline) and 25 nm (optimistic on-chip fringe tracking) used for kernel Monte-Carlo; free performance assumptions.
relative intensity fluctuation sigma_I = 0.10, 0.05, 0.02
0.10 / 0.05 / 0.02 levels chosen to illustrate HMS-PL benefit; free parameters not measured on sky.
kernel phase offset theta
Pre-defined phase offset appearing in the mixing matrix M; free design parameter of the photonic chip.

axioms (4)

domain assumption Von Kármán turbulence spectrum plus Taylor frozen-flow hypothesis adequately models residual wavefronts after GPAO/NAOMI+Baldr.
Used throughout §4 to generate the training and test phase screens; standard but not verified for the specific non-common-path residuals of Seidr.
domain assumption Light propagation through the HMS-PL is a linear complex transfer matrix A obtained by finite-difference BPM.
Stated in §3; underpins all power-to-wavefront mappings.
domain assumption Kernel operator K annihilates second-order instrumental phase errors (K·Θ = 0).
Taken from Martinache & Ireland 2018 and used in §5; algebraic property assumed exact for the chosen mixing matrix.
ad hoc to paper Open-loop correction of the estimated wavefront fully translates into the reported mode-selective power gain.
Implicit in Fig. 6; the paper itself flags that the metric is not closed-loop.

invented entities (1)

Seidr instrument (HMS-PL + kernel-nuller + TNN architecture) no independent evidence
purpose: Concrete realization of H-band kernel-nulling inside Asgard/Bifrost.
The instrument itself is the new system being proposed; its components are drawn from prior literature but the integrated design is the paper’s contribution.

pith-pipeline@v1.1.0-grok45 · 13799 in / 3248 out tokens · 26769 ms · 2026-07-12T03:34:00.976563+00:00 · methodology

0 comments

read the original abstract

Seidr is a new interferometric beam combiner within the Asgard Suite, utilizing infrastructure common to the BIFROST instrument at the Very Large Telescope Interferometer. Seidr combines hybrid mode-selective photonic lantern injection modules with a kernel-nulling photonic chip backend to enable deep H-band nulling for high-contrast studies of exoplanets, exomoons, and circumstellar dust. This instrument update summarizes Seidr's current design maturity and recent simulations of the point source - to - lantern outputs. We also outline progress on our neural network-based wavefront estimation scheme, which uses the photonic lantern outputs to sense phase fluctuations, designed to feed back to Baldr's deformable mirror, and improve nuller light injection.

Figures

Figures reproduced from arXiv: 2607.03280 by Adam K Taras, Akira Rodziewicz-Ryan, Barnaby Norris, Christopher H Betters, Daniel S Dahl, Frantz Martinache, Jin Wei, Julia J Bryant, Jyotirmay Paul, Marc-Antoine Martinod, Michael J Ireland, Nathan K Long, Nick Cvetojevic, Peter G Tuthill, Sergio Leon-Saval, Stefan Kraus.