REVIEW 4 major objections 4 minor 18 references
GRAVITY+ adaptive optics (GPAO) tests in Europe
T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read GRAVITY+ adaptive optics meets all performance targets in European lab tests.
desk verdict Genuine end-to-end bench qualification of GRAVITY+ AO, but the headline Strehl curves need the bench seeing and uncertainties stated before the claimed margins mean anything. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the test bench built in Nice: it combines an illuminated source whose flux is calibrated in apparent magnitude, a rotating phase plate that reproduces Paranal-like turbulence (about 0.7 arcsecond seeing in the nominal configuration, up to 1.4 arcseconds in stress tests), and optics that emulate the UT coude focus, followed by the adaptive optics system and a camera that measures the PSF at 1.31 micrometers. Strehl is computed from that camera image using a normalized perfect-PSF comparison with a 16-percent ghost-flux correction, and the 1.31-micrometer value is converted to the 2.2-micrometer GRAVITY band with the Marechal approximation. Calibration templates generate interaction matrices, reference slopes, and non-common-path aberration corrections, and a real-time mis-registration algorithm keeps the wavefront sensor aligned to the deformable mirror during operation.
What would settle it
Measure the on-sky K-band Strehl of GRAVITY+ during first commissioning at the VLT under the same conditions as the bench specification (median seeing of 0.83 arcseconds and 9.5 m/s wind speed). If the natural-guide-star Strehl at magnitude 9 falls below 75 percent, or the laser-guide-star Strehl at magnitude 17 falls below 50 percent, after calibration, the bench predictions would be contradicted.
Extended reading notes
Core claim
The central claim is that the GRAVITY+ adaptive optics system performs as designed when tested on a full-scale simulator of the VLT Unit Telescope and Paranal atmosphere. The paper shows measured Strehl-versus-magnitude curves for all four modes: the natural-guide-star visible mode (40x40 sub-apertures), the laser-guide-star visible mode (30x30 plus a 4x4 low-order sensor), and the corresponding infrared 9x9 modes inherited from the CIAO system. After a parasitic infrared LED from a motor encoder was found to contaminate the wavefront sensor and was blocked with baffling, the NGS curve met the 75-percent Strehl requirement at magnitude 9 with margin, and the LGS curve met the 50-percent requirement at magnitude 17. The paper also reports that non-common-path aberrations were calibrated on the bench by mode-by-mode flux maximization, reaching above 75 percent Strehl at 1.31 micrometers, and that the loop remained stable under off-centered pupils, defective actuators, and turbulence up to an estimated 1.4 arcseconds seeing.
Load-bearing premise
The bench faithfully reproduces the VLT telescope focus and Paranal turbulence, so that the Strehl ratios measured in the laboratory will also be reached on the sky at Paranal; the paper itself notes that the laser-guide-star cone effect cannot be simulated in the bench.
Editorial extensions
If this is right
- The four GPAO operating modes are ready for the integration and verification phase at Paranal, with the system shipped in mid-2024.
- On-sky NGS performance around 75 percent Strehl in K band at magnitude 9 would give GRAVITY+ the high dynamic range needed to observe faint companions and exoplanets in the VLTI.
- Working LGS modes extend the interferometer to fainter targets such as active galactic nuclei, at the cost of a cone effect that cannot be tested in the lab.
- The automated calibration templates should make daytime setup and night operations faster and more repeatable than earlier AO systems.
- Robustness tests against vignetting, actuator failure, and target wandering suggest GPAO can maintain performance during real observing conditions.
Reading between the lines
- Because the bench cannot simulate the LGS cone effect, on-sky LGS Strehl at magnitude 17 could come in below the 50 percent target even though the bench shows margin; commissioning will be the real test.
- The paper cautions that the Marechal conversion is unreliable below about 20 percent Strehl at 2.2 microns, so the faint-end performance values, especially at high NGS magnitudes, are uncertain and could be lower in practice.
- If the phase plate's turbulence statistics match Paranal's median seeing, the measured margins suggest the system has headroom, but real telescope vibrations and thermal flexure were only partially represented, so final margins may shrink on sky.
- The stray-light episode shows that bench environments contain their own parasitic sources; similar effects on the telescope could degrade Strehl unless checked with an equivalent health-check template.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript reports the European laboratory test campaign of the GRAVITY+ adaptive optics system (GPAO), conducted on a bench in Nice that simulates the VLT UT coudé focus and Paranal-like turbulence. It describes the calibration templates, interaction matrices, rejection-function measurements, NCPA calibration, and Strehl-versus-magnitude performance curves for the four GPAO modes (NGS-VIS, LGS-VIS, NGS-IR, LGS-IR). The authors conclude that the bench measurements reproduce expected performances, including under non-nominal conditions, and that this justified shipping the system to Paranal for AIV. The headline quantitative targets are SR=0.75 at magR=9 for NGS and SR=0.50 at magR=17 for LGS under median seeing (0.83"), 9.5 m/s wind.
Significance. If the bench results transfer to Paranal, this is an important milestone for VLTI: the 40x40 NGS extreme AO and the LGS capability would be key enablers for high-contrast and faint-science observations. The paper's strengths are the transparency about known limitations (Maréchal breakdown, absent cone effect, parasitic light found and baffled), the detailed description of operational templates and calibration procedures, and the direct measurement of rejection functions and loop delays. The main caveats are that the headline figure-of-merit curves are not tied to the specification seeing, the LGS curve rests on a simple magnitude-shift ansatz, and the quoted Strehl values lack an uncertainty analysis.
major comments (4)
- [Section 4.1, 5.2, Fig. 8] The GPAO top-level requirements are quoted for median seeing of 0.83" and 9.5 m/s wind, but the Strehl-vs-magnitude curves in Fig. 8 were obtained with the rotating phase plate described in Sec. 4.1 as producing 0.7" seeing. The text does not indicate that the plate was reconfigured to 0.83" for these measurements, nor does it rescale the results to the specification seeing. Since Strehl is a strong function of r0, the 0.13" difference is potentially comparable to the unquantified 'some margins' claimed for the green curve; the stress tests at up to 1.4" are not presented as Strehl-vs-magnitude curves and therefore do not bracket the specification point. Please quantify the expected Strehl degradation at 0.83" (e.g., from the measured rejection functions or a simulation), report the margin at the specification point, and include error bars or an uncertainty estimate on the Fig. 8 curves.
- [Section 4.2 and 5.2 (yellow curve, Fig. 8)] The LGS magnitude calibration is obtained by shifting the NGS-VIS calibration by 5 magnitudes because the 4x4 NGS WFS has 100x fewer subapertures than the 40x40 WFS. This is an assumption, not a measurement of the LGS path: it ignores differences in detector quantum efficiency, read noise, spot sampling, and the separate constant-brightness LGS source used for high-order correction. The yellow curve should be presented as an extrapolated estimate, with an uncertainty, rather than as a measured LGS performance curve. A photon-budget comparison or an end-to-end simulation would make the 5-magnitude shift testable.
- [Section 5.2, Eq. (3) and Fig. 6] The paper correctly warns that the Maréchal conversion is not valid at low Strehl and that 1% of Strehl at 1.3 µm maps to about 20% at 2.2 µm. The LGS requirement SR=0.5 at 2.2 µm corresponds to roughly 0.14 Strehl at 1.3 µm under the same formula, i.e., precisely in the regime where the conversion is most uncertain. The yellow curve's compliance with the LGS top-level requirement is therefore not established by the 1.3 µm measurements alone. Please report the 1.3 µm values for the LGS curve, quantify the conversion error at the relevant Strehl levels, or mark the affected points as upper limits.
- [Section 5.2, Eq. (2)] The definition of the Strehl estimator is ambiguous as written. If PSF_perfect,norm and PSF_meas,norm are both normalized to unit total flux, the ratio of their sums is unity and Eq. (2) cannot produce the reported Strehl values; if 'norm' denotes peak normalization or a windowed sum, that must be stated. The omission of the telescope spider from the perfect PSF and the 16% ghost fraction are also potential biases that need to be quantified. Please specify the normalization, the summation window, and the error propagation to Fig. 8.
minor comments (4)
- [Section 5.1, Eq. (1)] The factor S0 is called an 'unknown initial strehl' but is never defined or used; the text should clarify that this equation is a relative flux estimator, not an absolute Strehl measurement.
- [Section 5.2] The expected curves in Fig. 8 (right) are said to be 'derived from the ERIS design documentation' but no citation is given; please add a public reference or make the model accessible.
- [Throughout] The spelling 'strehl'/'Strehl' and 'Maréchal'/'Marechal' is inconsistent; please standardize.
- [Section 2.1] The sentence 'GPAO#1, as well as GPAO#3 and 4 have been sent...' is unclear, since GPAO#1 was described as used in Nice and later sent back; please clarify the hardware flow.
Circularity Check
No significant circularity: bench Strehl measurements are direct and the comparison curves come from an external design specification.
full rationale
The paper's central claim is experimental: the GPAO hardware, integrated on a test bench, reached the Strehl values specified as top-level requirements. The derivation chain is not circular. The measured Strehl curves (Section 5.2, Figure 8 left) come from a bench camera flux measurement (Eq. 2), an independent PSF-perfect normalization, a Marechal wavelength conversion (Eq. 3), and a flux-calibrated bench lamp. The 'expected' comparison curves (Figure 8 right) are stated to come from ERIS design documentation, i.e., an external specification, not from the measurements themselves. The LGS magnitude calibration is obtained by a 5-magnitude shift from the NGS calibration, but this only sets the abscissa; the LGS Strehl values themselves are measured. The NCPA calibration uses relative flux maximization, which is a standard optimization procedure, not a fit of the final performance curve. Self-citations (e.g., [2] for the test bench, [10] for the mis-registration algorithm) describe hardware and algorithms whose behavior is directly exercised in the present tests; they do not supply the Strehl-versus-magnitude result. The known limitations (cone effect not simulated, Marechal approximation unreliable near 20% K-band Strehl, seeing set to 0.7" rather than the 0.83" specification) are external-validity caveats, not circular reductions. No equation in the paper defines its output in terms of its input, and no fitted parameter is renamed as a prediction.
Assumptions & free parameters
free parameters (1)
- AO loop tuning parameters (gain, leaks, number of modes, flux thresholds, weighting maps) =
Optimized per input magnitude
assumptions (4)
- domain assumption The rotating phase plate produces turbulence representative of Paranal median seeing.
- domain assumption Strehl measured at 1.31 microns can be converted to 2.2 microns using the Marechal approximation (equation 3).
- ad hoc to paper LGS performance can be extrapolated from NGS performance by a 5-magnitude shift based on the 4x4 versus 40x40 aperture ratio.
- domain assumption The bench source magnitude calibration based on known WFS characteristics is accurate.
Cite this review
Pith. "Pith review of GRAVITY+ adaptive optics (GPAO) tests in Europe." pith.science (2026). https://pith.science/paper/5IN5VFVI
@misc{pith2026250603721,
author = {Pith},
title = {Pith review of: GRAVITY+ adaptive optics (GPAO) tests in Europe},
year = {2026},
howpublished = {\url{https://pith.science/paper/5IN5VFVI}},
note = {Machine review of arXiv:2506.03721}
}
read the original abstract
We present in this proceeding the results of the test phase of the GRAVITY+ adaptive optics. This extreme AO will enable both high-dynamic range observations of faint companions (including exoplanets) thanks to a 40x40 sub-apertures wavefront control, and sensitive observations (including AGNs) thanks to the addition of a laser guide star to each UT of the VLT. This leap forward is made thanks to a mostly automated setup of the AO, including calibration of the NCPAs, that we tested in Europe on the UT+atmosphere simulator we built in Nice. We managed to reproduce in laboratory the expected performances of all the modes of the AO, including under non-optimal atmospheric or telescope alignment conditions, giving us the green light to proceed with the Assembly, Integration and Verification phase in Paranal.
Figures
Figures from the paper (6 more)
Reference graph
Works this paper leans on
-
[1]
GRA VITY+ Wavefront Sensors: fully AO assisted interferometry on 8m class telescope at VLTI ,
Bourdarot, G., Le Bouquin, J. B., Hoenig, S., and al., “GRA VITY+ Wavefront Sensors: fully AO assisted interferometry on 8m class telescope at VLTI ,” SPIE Conf. Series13095, 13095–22 (Aug. 2024)
work page 2024
-
[2]
Millour, F., B´ erio, P., Lagarde, S., and al., “Building a ...,”SPIE Conf. Series12183, 121831X (Aug. 2022)
work page 2022
-
[3]
Gravity+ Collaboration, Abuter, R., Alarcon, P., and al., “The GRA VITY+ Project: Towards All-sky, Faint-Science, High-Contrast Near-Infrared Interferometry ...,” The Messenger 189, 17–22 (Dec. 2022)
work page 2022
-
[4]
GRA VITY+: Towards faint science,
Eisenhauer, F., “GRA VITY+: Towards faint science,” in [ The VLT in 2030], 30 (July 2019)
work page 2019
-
[5]
GRA VITY+ Collaboration, Abuter, R., Allouche, F., and al., “First light for GRA VITY Wide. Large separation fringe tracking for the Very Large Telescope Interferometer,” A&A 665, A75 (Sept. 2022)
work page 2022
-
[6]
GRA VITY Collaboration, Abuter, R., Accardo, M., Amorim, A., and al., “First light for GRA VITY: Phase referencing optical interferometry for the Very Large Telescope Interferometer,”A&A 602, A94 (June 2017)
work page 2017
-
[7]
Nowak, M., Lacour, S., Abuter, R., and al., “Upgrading the GRA VITY...,” A&A 684, A184 (Apr. 2024)
work page 2024
-
[8]
Measuring and compensating vibrations at the VLTI: MANHATTAN-II self-intrinsic noise ...,
Bigioli, A., Courtney-Barrer, B., Abuter, R., and al., “Measuring and compensating vibrations at the VLTI: MANHATTAN-II self-intrinsic noise ...,” SPIE Conf. Series12183, 121831Z (Aug. 2022)
work page 2022
Show all 18 references
-
[9]
Recent improvements of high density magnetic deformable mirrors: faster, larger and stronger,
Charton, J., Bitenc, U., Curis, J.-F., and al., “Recent improvements of high density magnetic deformable mirrors: faster, larger and stronger,” SPIE Conf. Series9148, 914825 (Aug. 2014)
2014
-
[10]
Estimation of the lateral mis-registrations of the GRA VITY + adaptive optics system,
Berdeu, A., Bonnet, H., Le Bouquin, J. B., and al., “Estimation of the lateral mis-registrations of the GRA VITY + adaptive optics system,” A&A 687, A157 (July 2024)
2024
-
[11]
Characterization of OCam and CCD220: the fastest and most sensitive camera to date for AO wavefront sensing,
Feautrier, P., Gach, J.-L., Balard, P., and al., “Characterization of OCam and CCD220: the fastest and most sensitive camera to date for AO wavefront sensing,” SPIE Conf. Series7736, 77360Z (July 2010)
2010
-
[12]
A dynamical measure of the black hole mass in a quasar 11 billion years ago,
Abuter, R., Allouche, F., Amorim, A., and al., “A dynamical measure of the black hole mass in a quasar 11 billion years ago,” Nature 627, 281–285 (Mar. 2024)
2024
-
[13]
Chromatically modeling the parsec-scale dusty structure in the center of NGC 1068,
Leftley, J. H., Petrov, R., Moszczynski, N., and al., “Chromatically modeling the parsec-scale dusty structure in the center of NGC 1068,” A&A 686, A204 (June 2024)
2024
-
[14]
High contrast at short separation with VLTI/GRA VITY: Bringing Gaia companions to light,
Pourr´ e, N., Winterhalder, T. O., Le Bouquin, J. B., and al., “High contrast at short separation with VLTI/GRA VITY: Bringing Gaia companions to light,” A&A 686, A258 (June 2024)
2024
-
[15]
Dimensioning...,
Patru, F., Millour, F., Lai, O., and al., “Dimensioning...,” SPIE Conf. Series11448, 1144871 (Dec. 2020)
2020
-
[16]
Exoplanets in reflected starlight with dual-field inter- ferometry: A case for shorter wavelengths and a fifth ...,
Lacour, S., Carri´ on-Gonz´ alez,´O., and Nowak, M., “Exoplanets in reflected starlight with dual-field inter- ferometry: A case for shorter wavelengths and a fifth ...,” arXiv e-prints , arXiv:2406.07030 (June 2024)
2024 arXiv
-
[17]
Heterodyne IR interferometry and technologies for kilometric baseline interferome- try,
Bourdarot, G. and al., “Heterodyne IR interferometry and technologies for kilometric baseline interferome- try,” SPIE Conf. Series13095, 13095–145 (Aug. 2024)
2024
-
[18]
Kilometer-baseline interferometer: technology paths and sensitivity requirements,
Bourdarot, G. and al., “Kilometer-baseline interferometer: technology paths and sensitivity requirements,” SF2A Conf. Series(June 2024)
2024
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.