Phase Retrieval using Nonlinear Curvature Sensing within Convergent Beams

Justin R. Crepp, Caleb G. Abbott, James Smous, Matthew Engstrom, Brian Sands · 2026

Read Phase Retrieval using Nonlinear Curvature Sensing within Convergent Beams by Justin R. Crepp, Caleb G. Abbott, James Smous, Matthew Engstrom, Brian Sands (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-11832.txt.

Introduction

Crepp and co-authors study wavefront sensing in a convergent beam for adaptive optics. Path-length diversity methods read intensity on several planes to recover phase and amplitude. Free-space versions, like the nonlinear curvature sensor, are sharp but bulky. Recording in a convergent beam shrinks the instrument and packs light onto smaller detector areas. The team builds a physical optics model and checks it in simulations and lab tests. Their Fourier recipe lets standard Gerchberg-Saxton pipelines use the convergent-beam data, cutting size, weight, and cost.

Catalog record

CREATORJustin R. Crepp, Caleb G. Abbott, James Smous, Matthew Engstrom, Brian Sands
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37792
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.11832

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Phase Retrieval using Nonlinear Curvature Sensing within Convergent Beams Justin R. Crepp, Caleb G. Abbott, James Smous, Matthew Engstrom, Brian Sands Abstract Path-length diversity methods may be used for adaptive optics (AO) systems to retrieve phase and amplitude information by measuring intensity across multiple planes. Observations that rely on free-space propagation, such as the nonlinear curvature wavefront sensor (WFS), have been shown to offer excellent sensitivity and robustness to scintillation. However, the default design results in a large opto-mechanical footprint due to unavoidable geometric-optics and wave-optics effects. Measurements recorded in a convergent beam would improve instrument compactness, while concentrating light into smaller detector regions of interest, improving signal-to-noise ratio and possibly wavefront reconstruction speed. In this paper, we study path-length diversity wavefront sensing using four planes of contemporaneous intensity measurements made in a convergent beam. We develop a physical optics propagation model and validate the model by performing wavefront reconstructions in both simulations and lab experiments. The manuscript’s core contribution is a practical, intensity-domain, Fourier-transform-based recipe to use a conventional multi-plane Gerchberg-Saxton (or comparable) reconstruction pipeline with convergent-beam measurements, enabling a compact optical layout. We find that this approach offers practical benefits over an equivalent free-space wavefront sensor, in particular reducing size, weight, complexity and cost. a]University of Notre Dame, Department of Physics and Astronomy, Notre Dame, IN, USA keywords imaging, adaptive optics, wavefront sensing, beam control † † authorinfo: Send correspondence to [email protected] 1 Introduction Diffraction and interference of spatially coherent radiation may be used to estimate the electric field of light waves that become aberrated by inhomogeneous media [ 22 ] . In the case of atmospheric turbulence, phase diversity or path-length diversity techniques may be employed to retrieve phase and amplitude information [ 8 , 9 , 28 ] . Applications include astronomy, space domain awareness, remote sensing, power beaming, and related areas [ 13 , 20 , 6 , 2 , 17 ] . Deciding where to place intensity measurement planes that detect and record diffraction and interference patterns has important implications for practical implementation [ 18 ] . A variety of multi-image phase diversity methods have been developed for near-focal-plane and/or near-pupil-plane measurements using various reconstruction algorithms [ 12 , 29 , 24 , 25 , 7 , 3 ] . The family of curvature sensors, including both linear and nonlinear signals, often rely on multi-plane, passive intensity measurements with no moving parts [ 13 ] . The principal optical design challenge typically involves creating an optical relay that generates a small diameter beam, long back focal distance, and small opto-mechanical footprint. A small beam diameter is needed to induce diffraction and record multiple images onto a single detector; a long back focal distance is needed to fit beam-splitting components into the instrument; and a small opto-mechanical footprint is desired for practical implementation. These requirements often conflict with one another. As a consequence, recording intensity measurements in semi-collimated space to introduce path-length diversity for coherent sensors creates unavoidably large opto-mechanical footprints (tens of centimeters on a side). In practice, the optical bench real-estate needed for pure free-space propagation measurements may not exist. Otherwise, multiple high-speed cameras may be required [ 30 ] , further increasing size, weight, power, complexity and cost. Operating in a convergent beam would address these practical concerns while also concentrating light into a fewer number of pixels to help combat detector read noise. Figure 1: Cartoon level optical design concepts for capturing the sensing channels of a multi-plane wavefront sensor. Beam-splitting components introduce optical path-length differences between channels (individual derivative beams are not shown). Operating in collimated space requires a large opto-mechanical foot-print (top three designs). Operating in a converging beam improves compactness (bottom design). As an example, it is instructive to compare the spatial frequency response of the conventional curvature sensor to the non-linear curvature sensor. As a convention, we use \nu=\sqrt{\nu_{x}^{2}+\nu_{y}^{2}} for spatial frequencies (cycles / m) so that in two-dimensions \hat{\phi}(\nu_{x},\nu_{y})=\int\int\phi(x,y)e^{-i2\pi(\nu_{x}x+\nu_{y}y)}dxdy. (1) The (linear) curvature sensor estimates the Laplacian of the phase, \phi(x,y) , using intensity differences recorded between two planes located near the pupil, \Delta I(x,y)=C\nabla^{2}\phi(x,y), (2) where C is a proportionality constant that depends on the optical set-up. The power spectrum of curvature phase estimates is given by \Phi_{\rm Curvature}(\nu_{x},\nu_{y})=\;<|\mathrm{FT}(\nabla^{2}\phi(x,y))|^{2}>, (3) where <> is an expectation operator for a statistically isotropic phase field on the pupil. Since |\mathrm{FT}(\nabla^{2}\phi(x,y))|=4\pi^{2}(\nu_{x}^{2}+\nu_{y}^{2})\hat{\phi}(\nu_{x},\nu_{y})=4\pi^{2}\nu^{2}\hat{\phi}(\nu) , and \Phi_{\rm atm}(\nu)=<|\hat{\phi}(\nu)|^{2}>\sim\nu^{-11/3} is the power spectrum of phase aberrations caused by atmospheric turbulence, the sensitivity of the curvature sensor scales with spatial frequency according to \Phi_{\rm Curvature}(\nu)\propto\nu^{4}\nu^{-11/3}\propto\nu^{1/3} (4) assuming Kolmogorov statistics. Differential intensity measurements recorded near the pupil thus emphasize high spatial frequencies [ 24 ] . While curvature sensing enhances small-scale turbulence features, operating further from the pupil allows diffraction to smooth high spatial frequency content. Including measurement planes at greater distances (lower Fresnel number), such as with the non-linear curvature sensor, has been shown to significantly improve the sensitivity of curvature sensing [ 13 , 6 , 18 ] . In the limit of operating in the far field, atmospheric phase distortions begin to imprint their statistics onto the intensity pattern, \Phi_{\rm Fraunhofer}(\nu)\approx\Phi_{\rm atm}(\nu)\propto\nu^{-11/3}, (5) emphasizing lower spatial frequency content which contain more power. As in the case of a four-plane non-linear curvature sensor, the combination of inner-planes and outer-planes leads to a broader, more balanced spatial frequency response. In practice however, the far field generally remains inaccessible using free-space propagation due to unrealistically large z -distances. Operating a non-linear curvature WFS closer to focus, e.g. lower Fresnel number than what a free-space propagation system nominally provides access to, would permit measurements at larger effective propagation distances while also improving signal-to-noise ratio in the presence of read noise. Concentrating the light beam can probe smaller spatial scales, such as in microscopy [ 16 ] , and may reduce computational latency by illuminating a smaller number of pixels. While lab testbeds and field experiments warrant the use of a focusing element to conveniently access larger effective propagation distances, the surface figure of focusing elements creates large phase variations (hundreds of waves of curvature) that preclude common wavefront reconstruction algorithms from converging due to spatial sampling limitations. It is therefore desirable to develop a technique whereby a lens or mirror is used in hardware while not requiring explicit modeling of the optical surface in software. Curvature sensing and phase diversity near focus have a long history, starting with Roddier 1988 [ 24 ] and later nonlinear curvature sensing. [ 13 ] Several nlCWFS implementations have been demonstrated in the lab or using ground-based telescopes, typically in an effort to compare performance with conventional curvature sensing and Shack-Hartmann devices. Mateen et al. built a nlCWFS that used dichroics with fold mirrors to record four intensity measurement planes contemporaneously [ 21 , 20 ] . Crass et al. 2012 and Velasco et al. 2015 developed the AOLI instrument, which uses dichroics with a pyramid mirror and two high-speed CCD cameras, to explore combining lucky imaging with AO at the William Herschel Telescope. [ 5 , 30 ] Crepp et al. 2020 and Letchev et al. 2024 conducted lab experiments using a nlCWFS that split the incoming aberrated beam into four measurement planes based on polarization state. [ 6

Read the complete text