Roadmap for Exoplanet High-Contrast Imaging: Nulling Interferometry, Coronagraph, and Extreme Adaptive Optics

Abstract

The detection and characterization of exoplanets are central topics in astronomy, and high-contrast imaging techniques such nulling interferometry, coronagraphs, and extreme adaptive optics (ExAO) are key tools for the direct detection of exoplanets. This review synthesizes the pivotal role of these techniques in astronomical research and critically analyzes their role as key drivers of progress in the field. Nulling interferometry suppresses stellar light through the phase control of multiple telescopes, thereby enhancing the detection of faint planetary signals. This technology has evolved from the initial Bracewell concept to the LIFE (Large Interferometer For Exoplanets) technique, which will achieve a contrast ratio of 10⁻⁷ in the mid-infrared wavelength range in the future. Coronagraphs block starlight to create a “dark region” for direct observation of exoplanets. By leveraging innovative mask designs, theoretical contrast ratios of up to 4 × 10⁻⁹ can be achieved. ExAO systems achieve precise wavefront correction to optimize the high-contrast imaging performance and mitigate atmospheric disturbances. By leveraging wavefront sensing, thousand-element deformable mirrors, and real-time control algorithms, these systems suppress the turbulence correction residuals to 80 nm RMS, enabling ground-based telescopes to achieve a Strehl ratio exceeding 0.9. This work provides a comprehensive analysis of the underlying principles, prevailing challenges, and future application prospects of these technologies in astronomy.

1. Introduction

The detection and characterization of exoplanets stand at the forefront of astronomical research [1], driven by the profound quest to find potentially habitable worlds beyond our solar system [2]. Among these, the direct detection and spectroscopic characterization of Earth-like exoplanets orbiting within the habitable zones of Sun-like stars represent a paramount goal of modern astronomy, holding the key to answering fundamental questions about the uniqueness of our planet and the potential for life elsewhere. However, the direct observation of exoplanets, particularly Earth-like planets orbiting within the habitable zones of Sun-like stars, poses a formidable challenge due to the immense brightness contrast between a star and its orbiting planets (typically ranging from 10⁻⁶ for gas giants to 10⁻¹⁰ for terrestrial planets) and their small angular separations (often below 0.1 arcs), compounded by the distorting effects of Earth’s atmosphere for ground-based observations [3]. No single technology can overcome these barriers alone; progress is instead driven by the co-evolution and integration of advanced methods in interferometry, coronagraphy, and wavefront control. Consequently, there is a pressing need for a synthesized resource that not only details the principles and milestones of each discipline but, crucially, elucidates their interdependencies and combined potential.

Therefore, this article serves as a review article that aims to clarify the interrelationships between key technologies by integrating the scattered historical developments of high-contrast imaging techniques and to present a clear development trajectory. This article charts the technological roadmap for exoplanet HCI, focusing on its three foundational pillars: nulling interferometry, coronagraphy, and extreme adaptive optics (ExAO). Each of these three topics is covered in a separate unit, starting with the origins of the technology module, progressing to the most advanced technologies available today, and concluding with future development directions. This allows readers to understand the historical context and make vertical comparisons. At the same time, horizontal comparisons between the three units reveal the relationships between them.

When discussing each new technology, the paper follows the logic of “what problem did it solve, who invented what technology in what year, what problems remained, and how did subsequent researchers solve these problems.” Ultimately, all technologies are connected through a temporal framework, forming a comprehensive and coherent development trajectory.

We hope this article will lower the learning barrier for newcomers while providing experts with a structured reference framework to gain a clearer historical context and more intuitive horizontal and vertical comparisons. Therefore, we structure this review around the problem-solving logic that has driven instrumental innovation:

• Suppressing Stellar Light via Coherent Cancelation (Nulling Interferometry): The initial challenge was to selectively suppress the intense starlight while preserving the faint planetary signal. This motivated the concept of nulling interferometry, first proposed by Bracewell in 1978 [4]. By precisely controlling the phase difference (π-shift) between light beams collected by separated telescopes, destructive interference nullifies the on-axis starlight, creating a “dark fringe” where off-axis planetary light can constructively interfere. Early space mission concepts like DARWIN (ESA) and TPF-I (NASA) aimed to leverage this principle but faced technological hurdles [5]. Ground-based efforts, such as the OHANA project demonstrating fiber-linked interferometry on Mauna Kea and the Keck Interferometer Nuller, validated the concept but grappled with atmospheric turbulence and sensitivity limits [6]. The quest for higher stability, sensitivity, and resolution led to multi-telescope beam combiners like MIRC/MIRC-X (CHARA array) and breakthroughs in the mid-infrared with the LBTI (Large Binocular Telescope Interferometer) and its NIC (Nulling and Imaging Camera) [7]. The recent revolution in integrated photonics, exemplified by instruments like Dragonfly and GLINT, promises miniaturized, robust nulling interferometers [8]. The ultimate vision for space-based nulling is embodied in the LIFE (Large Interferometer For Exoplanets) mission concept, employing advanced “kernel-nulling” for unprecedented contrast in the mid-IR in the future [9].
• Creating Localized Dark Regions via Diffraction Control (Coronagraphy): Complementing interferometry, coronagraphy tackles the contrast challenge by locally blocking or diffracting starlight within the telescope’s focal plane [10]. Early concepts involved external occulters, but their impractical scale for exoplanets (requiring massive structures tens of thousands of kilometers from the telescope) shifted focus to internal coronagraphs. The foundational Lyot coronagraph (1930s) used a focal plane mask and pupil stop. Subsequent innovations focused on mask design to improve performance: Band-Limited Masks (BLMC) optimized diffraction suppression [11]; phase masks (e.g., Four-Quadrant Phase Mask—4QPMC, Eight-Octant Phase Mask—EOPM, Optical Vortex Coronagraph—OVPMC) utilized destructive interference with higher theoretical contrast and smaller inner working angles (IWAs), demonstrated on instruments like Subaru’s HiCIAO/SCExAO [12]; and apodized pupil concepts (e.g., Classical Pupil Apodization—CPA, Apodized Pupil Lyot Coronagraph—APLC) modulated the pupil amplitude. A significant leap came with Phase-Induced Amplitude Apodization (PIAA), using aspheric optics to reshape the beam without significant light loss, later combined with complex masks as PIAACMC, a technology baselined for missions like WFIRST-CGI and crucial for ground-based systems like VLT-SPHERE (Spectro-Polarimetric High-contrast Exoplanet REsearch) and Subaru-SCExAO [13].
• Enabling High-Fidelity Wavefront Control (Extreme Adaptive Optics—ExAO): Both nulling interferometry and coronagraphy are critically dependent on achieving near-perfect wavefront quality and stability. Atmospheric turbulence and static optical aberrations degrade performance. This necessity birthed ExAO, an advanced form of adaptive optics pushing correction fidelity to the extreme. Early conceptual work laid the groundwork. Key developments included: high-density deformable mirrors (DMs) with thousands of actuators (e.g., MEMS DMs for XAOPI concept); advanced wavefront sensors like the pyramid wavefront sensor (PWFS) and Asymmetric Pupil Fourier Wavefront Sensor (APF-WFS) [14]; sophisticated real-time control algorithms (e.g., Fourier transform methods, Multigrid Conjugate Gradient—MGCG) [15]; and techniques for Non-Common Path Aberration (NCPA) correction. Instruments like Subaru-SCExAO and MagAO-X integrate these technologies, providing the stable, high-Strehl-ratio (>0.9) point spread function (PSF) essential for modern coronagraphs and coherent beam combination in interferometry [16]. Innovations like Microwave Kinetic Inductance Detectors (MKIDs) integrated into MKID-PWFS promise further gains in sensitivity and multi-band wavefront sensing [17].

Crucially, these three pillars are not isolated but increasingly interdependent. ExAO provides the stable wavefront foundation without which advanced coronagraphs and sensitive nulling cannot function. Coronagraphs create the localized dark zones needed for planet detection after wavefront correction. Nulling interferometry offers an alternative path to extreme contrast, particularly in specific wavelength regimes. The synergy between them is paramount for pushing performance towards the 10⁻¹⁰ contrast required for Earth-like exoplanets.

Looking ahead, this review also explores the transformative impact of emerging disruptive technologies: Deep learning is revolutionizing wavefront prediction, speckle noise suppression [18], and observation optimization. Integrated photonics and optical neural networks promise paradigm shifts towards miniaturized, ultra-fast, and intelligent HCI systems capable of light-speed processing.

The structure of this review reflects this roadmap:

Section 2 (Existing High-Contrast Imaging Techniques) delves into the principles, historical development, key instruments, achievements, and limitations of each core technology: Section 2.1 Nulling Interferometry (Bracewell to LIFE). Section 2.2 Coronagraphs (External Occulters to PIAACMC). Section 2.3 Extreme Adaptive Optics (Concepts to SCExAO/MagAO-X/MKID-PWFS)

Section 3 (Development Trends) analyzes the critical convergence of these core technologies and examines the revolutionary potential of: Section 3.1 The Mutual Enhancement Between ExAO, Coronagraphs, and Nulling Interferometry. Section 3.2 Applications of Deep Learning Across the HCI Workflow. Section 3.3 Integrated Photonics and Optical Neural Networks for Next-Generation Systems. Section 3.4 Key Challenges in the Short and Long Term.

Section 4 (Summary) synthesizes the technological journey, underscores the link between current capabilities and future innovations, and reiterates the path towards the ultimate goal of characterizing Earth-like exoplanets.

By tracing this instrumental and technological evolution—from foundational concepts to cutting-edge implementations and emerging paradigms—this review aims to provide a comprehensive roadmap for researchers navigating the rapidly advancing field of high-contrast exoplanet imaging.

2. Existing High-Contrast Imaging Techniques

As shown in Figure 1, telescopes for high-contrast imaging are directly divided into nulling interferometry technology and coronagraph technology, which are the first two technologies to be introduced in this paper. The third technology, ExAO, is an enabling technology for nulling interferometry and coronagraph and is also an independent module that determines the final performance of the system. Therefore, the three together form the high-contrast imaging technology triangle.

2.1. Nulling Interferometry

While stellar interferometers can achieve extreme angular resolution—unmatched by even the next generation of single-aperture telescopes—through arrays composed of small or medium-sized telescopes, successfully constructing baselines exceeding 300 m has enabled sub-milliarcsecond resolution in the visible and near-infrared bands. However, the fundamental physics of optics imposes stringent requirements on the precision of beam combination and optical path matching between two or more telescopes, making the technical cost and challenges of building interferometric arrays immense. This section reviews the basic theory and overall design considerations of interferometry, focusing on the practical technological solutions developed to overcome these challenges. These include mitigating atmospheric turbulence, solving the problem of sub-micron optical path difference (OPD) matching, constraints on the number and layout of telescopes, and managing the throughput losses inherent in the multiple reflections and transmissions necessary to combine beams from multiple telescopes in a central beam-combining laboratory. The discussion will also cover other critical aspects of interferometric measurement. The process, from site selection and telescope array configuration to the final precise measurement of interferometric visibility and closure phase, involves complex processing through multiple subsystems. This includes active wavefront sensing for tilt error correction, and even full adaptive optics systems; telescope design for long-distance collimated beam transport; control of diffraction losses; polarization matching; OPD insertion and active compensation; correction for atmospheric refraction and dispersion differences in glass and air media; beam splitting into visible and near-infrared channels; precise alignment of long optical paths; high-accuracy metrology for the 3D geometry of the interferometric array; and various beam combination schemes, from simple two-way combiners to image-plane combination for imaging with multiple telescopes. Over the past few years, a wealth of invaluable experience has been accumulated through a series of robust and successful interferometric instruments. Today, interferometry is progressively becoming a mainstream observational method in astrophysical research [19].

2.1.1. Early Concepts and Space Mission Proposals: Bracewell, DARWIN and TPF-I

Owing to the extremely high contrast between stars and planets, direct imaging is extremely challenging. In 1978, Bracewell first introduced the concept of nulling interferometry [20]. This concept is based on precision phase manipulation achieved by introducing a precise π-phase shift into one of the arms of a dual (or multiple) telescope interferometer. A point source located at the center of the optical axis of the interferometer (which corresponds to the position of the star) is set at equal optical distances from the two arms. With the introduction of the π-phase shift, when two beams of light with equal amplitude are superimposed, complete phase-cancelation interference occurs. The wave peaks meet the troughs and cancel each other out, and the intensity is significantly suppressed (ideally to zero). Point sources that are simultaneously off-axis (corresponding to the planetary position) have a light-range difference of

$$OPD=B\cdot \sin \theta$$

(1)

(B is the baseline length and θ is the angular distance) between the two arms. This natural phase difference interacts with the artificially introduced π-phase shift to achieve phase-length interference or partial phase-length interference at a specific angle θ, allowing the planetary signal to be preserved or enhanced.

The basic interference intensity formula (double aperture) describing the intensity distribution of an arbitrary point source in an interferometer is

$$I_{total}=I_1+I_2+2\sqrt{I_1{I}_2}\cos (\mathrm{\Delta }\varphi )$$

(2)

I₁, I₂: light intensity of the two interfering arms; Δϕ: phase difference between the two beams (with artificial phase shift and optical range difference contributions).

This technique creates a high-contrast “dark zone” at the position of the target star, and weak planetary signals can be detected at very small angular distances from neighboring stars (corresponding to the habitable zone).

However, Bracewell’s dual-telescope method results in increased in light-signal leakage from stars with increasing baselines, making it difficult to detect planets near stars. In 1997, Angel and Woolf proposed using a linear array of multiple telescopes to flatten the angular transfer function of the interferometer [4], thereby achieving a wider central null region. They designed an array of four telescopes using hierarchical combinations to produce wider dark fringes that suppressed the light signal from the star more effectively.

The European Space Agency (ESA) and the National Aeronautics and Space Administration (NASA) had planned to use nulling interferometry to detect Earth-like planets. To transform Bracewell’s initial proposal into a feasible space mission, ESA proposed the DARWIN project in 1996. This project planned to use a constellation of multiple mid-infrared telescopes and employed nulling interferometry as the measurement principle to directly detect Earth-like planets [21]. Similarly, NASA proposed a concept called the Terrestrial Planet Finder Interferometer (TPF-I) as part of its Navigator program, aiming to characterize Earth-like exoplanets [22]. Unfortunately, due to a lack of clarity on the scientific yields of the missions and low technological immaturity, both missions were abandoned by the end of the 2010s [9]. While no planets were directly detected by these concepts, they transformed nulling interferometry from a theoretical proposition into an engineered approach for exoplanet characterization, providing the technical blueprint for future implementations.

2.1.2. Ground-Based Fiber Interferometric Array: OHANA Project

To meet the increasing resolution and sensitivity demands in astronomical observations, Mariotti et al. first proposed the concept of connecting large ground-based telescopes via interferometric technology in 1996 to enhance the sensitivity and resolution of optical/infrared interferometers, aiming to improve resolution by increasing the baseline length [23]. Technical challenges related to beam injection into fibers had to be addressed to achieve effective transmission of telescope beams through single-mode optical fibers. In 2000, Perrin et al. detailed the current version of the OHANA project and proposed the use of single-mode optical fibers to connect large telescopes on the summit of Mauna Kea to construct a large near-infrared interferometer [24]. In the first phase of the OHANA project, researchers focused on developing and optimizing the instruments required to inject light into a single-mode fiber. In 2002, Woillez et al. conducted the first tests at CFHT to evaluate the efficiency of injecting adaptive optics-corrected light into a single-mode fiber to assess the performance of the interferometer and observational protocols [25]. The injection efficiency is a key indicator of interferometer performance and directly affects its sensitivity and observational capabilities.

$${\eta }_{inj}={\displaystyle \frac{P_{sm}}{P_{mm}}}$$

(3)

The injection efficiency is defined as the ratio of the injection rate in a single-mode fiber to the injection rate in a multimode fiber. Here, P_sm is the optical power in a single-mode fiber and P_mm is the optical power in a multimode fiber.

The fiber loss is defined as

$$P_{out}=P_{in}\times {10}^{{\displaystyle \frac{-aL}{10}}},$$

(4)

where P out is the output optical power, P in is the input optical power, α is the loss factor (in dB/km), and L is the fiber length (in km).

Following advances in fiber-optic injection technology, the OHANA project entered a second phase (Phase II) to construct a prototype and explore the “shortest” baseline [26]. This phase focused on connecting the “shortest” baselines between the Mauna Kea hilltop telescopes to demonstrate the OHANA concept in the J, H, and K bands, as the shown of the Figure 2.

Optical path differences (OPD) in long-baseline interferometers vary due to Earth’s rotation (a phenomenon known as fringe tracking or streak drift). This angular drift causes the interference pattern to “streak” across the detector, degrading signal integration. To maintain stable interference fringes during observations: a specialized delay line system was designed for the layout of the Mauna Kea telescope. As shown in the figure, most of the delay was compensated by moving the central trolley, and the Following advances in fiber-optic injection technology was tracked by a fast-moving mirror. This innovative design effectively solved the optical-range-difference compensation problem, improved the stability and accuracy of the interferometer, and operated in the near-infrared covering the J, H, and K spectral bands [27]. The maximum length of the optical fibers used in Phase II was approximately 500 m, with a propagation loss of up to 1 dB for a baseline of up to 450 m and a conservative estimate of transmission efficiency of 50%. The delay lines, with a total travel distance of 50 m [28], used a three-phase system comprising coarse delay, fast delay, and an OPD (Optical Path Difference) error simulator, thus ensuring precise interferometric operation.

The sensitivity of the OHANA Phase II prototype was estimated based on parameters such as the system transmission and fiber injection stability. The limiting magnitude was expected to be K = 13 ± 1 when using an 8 m telescope, and the limiting magnitude was increased by approximately five orders of magnitude under streak-tracking of off-axis sources. Simultaneously, leveraging the high resolution and sensitivity of this system, astronomical observation programs, for young stellar objects (YSOs) and extragalactic sources were developed [29]. These programs make full use of OHANA’s high resolution and sensitivity of the OHANA. These observation programs fully leveraged the technology in OHANA Phase II to demonstrated its potential in high-contrast imaging.

Following progress in fiber injection technology and the “shortest” baseline interferometry operations, the OHANA project entered its third phase (Phase III) [30], which involved expanding and formally launching routine scientific operations. The focus of this phase was to integrate more Mauna Kea summit telescopes into a unified interferometric array to form an 800 m baseline and to enable routine scientific operations. A key objective of this phase was to connect all participating telescopes (such as CFHT, Gemini, Keck, Subaru, etc. [31]) via fiber optics, enabling OHANA to provide astronomical observations with higher resolution and sensitivity. In Phase III, the fiber optic system was further refined by introducing longer fibers (up to 800 m) and an upgraded delay-line system to accommodate the requirements of long baselines. In particular, the improvements to the delay lines not only expanded the light path difference compensation capability significantly, but also ensured efficient operation and light-path stability under long baseline conditions through a compact multichannel delay-line design.

OHANA Phase III used the near-infrared wavelength bands (J, H, and K bands [30]) for observations and planned to improve the transmission efficiency to achieve detailed observations of faint celestial objects. The longer fiber transmission system had a maximum transmission loss of 1 dB, which enhanced the transmission efficiency to K = 15 ± 1, suitable for a wider range of astronomical research projects. Based on the data accumulated in the OHANA Phase II, Phase III enabled high-precision interferometric imaging at longer baselines and with longer optical path differences [30].

2.1.3. Traditional Bulk-Optics Interferometers: Keck Interferometer Nuller and Nulling Depth Formula

Nulling interferometers based on traditional optical components were constructed in the early 2000s, marking an important breakthrough [25]; for example, the successful development and application of the Keck Interferometer Nuller validated the discovery of the nulling-depth formula by fitting actual observational data.

$$N\equiv {\displaystyle \frac{I_{res}}{I_{star}}}={\epsilon }_A^2+{({\displaystyle \frac{\mathrm{\Delta }{\varphi }_{error}}{2}})}^2$$

(5)

Here, N is the zeroing depth, $I_{\mathrm{res}}$ is the residual stellar light intensity after zeroing, $I_{\mathrm{star}}$ is the original stellar light intensity without zeroing, $\mathrm{\Delta }{\varphi }_{error}$ is the phase control error (radians), and ${\epsilon }_A$ is the amplitude imbalance.

2.1.4. Multi-Telescope Infrared Interferometry: MIRC and MIRC-X

Long-baseline interferometers can achieve high angular resolution, but are limited by the number of telescopes and instrument sensitivity, making it difficult to achieve model-independent imaging. To enhance the sensitivity and imaging capabilities of interferometers, Monnier et al. developed the Michigan InfraRed Combiner (MIRC) in 2006 [32], the principle is shown in Figure 3. This was the first six-telescope beam combiner operating in the near-infrared wavelength range, integrating the six telescopes of the CHARA array with a baseline extended to 330 m (the longest optical baseline for ground-based interferometers at the time), with an angular resolution of 0.6 mas (equivalent to resolving a 1 m object on the Moon), covering the J/H band (1.1–1.7 μm), and filling the gap in long-baseline infrared interferometry [33]. Single-mode fibers were used to transmit optical signals, enabling multi-beam phase-stable synthesis. This achievement led to several milestones in stellar astrophysics, including imaging of the expansion phase of a nova explosion [34], observing the occultation process of a binary star system [35], and studying the surface and spots of other stars. This addressed the limitations of traditional null interferometers that had insufficient angular resolution and imaging capabilities and supported only two to four telescopes, which made them unable to achieve model-independent imaging (such as stellar surface structures).

Although MIRC achieved significant results, higher sensitivity and accuracy were required for key scientific applications, such as the astrometric detection of faint YSOs and exoplanets. To address these issues, the MIRC-X project was launched to significantly improve the sensitivity and accuracy. In June 2017, the first phase of the MIRC-X project was completed, with the installation of the C-RED ONE camera based on the SAPHIRA detector significantly reducing the readout noise and improving the frame rate [36]. In September 2018, the second phase was completed with the installation of a new anti-crosstalk beam combiner, optimization of the pixel scale, and redesign of the high-throughput photometric channel optical components, including new J + H band fibers and custom non-polarizing beam splitters [37].

MIRC-X reduced the noise-equivalent power to 10% of that of MIRC using SAPHIRA detectors. Moreover, sensitivity was improved by two orders of magnitude (detection limit reaching K = 7), extending the target detection limit from bright extended sources (e.g., red giants) to faint YSOs within planetary disks (K > 7) [36]. The noise-equivalent power model for the sensitivity leap is

$$NE{P}_{MIRC-X}={\displaystyle \frac{NETD\cdot \sqrt{A_{\det }\cdot \mathrm{\Delta }f}}{R_x}}$$

(6)

measured at $NE{P}_{MIRC-X}=0.1\cdot NE{P}_{MIRC}$, where NEP is the noise-equivalent power, NETD is the noise-equivalent temperature difference, $A_{\det }$ is the effective detector area, $\mathrm{\Delta }f$ is the signal bandwidth, and Rx is the detector response rate.

MIRC-X operates in the J and H bands, covering the near-infrared wavelength range, and has made numerous observations since June 2017, covering a wide variety of research objects such as YSOs, stellar speckle imaging, and astrometric planetary surveys. The customized unpolarized beamsplitter in MIRC-X enables the system Mueller matrix to satisfy

$${‖M_{\mathrm{s}ys}-I‖}_F<0.003$$

(7)

Starlight exhibits partial polarization due to scattering in planetary atmospheres or circumstellar disks. Accurately measuring polarization states requires characterizing instrumental polarization effects introduced by optics (e.g., mirrors, beamsplitters). The Jones calculus describes fully polarized light using complex electric field components ([E_x, E_y]), while the Mueller calculus operates on measurable intensity-based Stokes parameters (S = [I, Q, U, V]). The instrument’s polarization response is represented by a Mueller matrix (M_sys), which transforms the true astrophysical Stokes vector (S_ast) into the observed vector (S_obs):

$$S_{ast}=M_{sys}^{-1}{S}_{obs}$$

(8)

$S_{ast}$: astro-true Stokes vector (including polarization state), $S_{obs}$: detector measured Stokes vector, $M_{sys}$: system Mueller matrix.

MIRC-X approximates the unitary matrix I by the crosstalk-resistant design, achieving visibility precision exceeding 1%. A closure phase precision exceeding 1° guarantees the phase extraction accuracy of weak planetary signals in nulling interferometry [38]. Closure phase—a robust interferometric observable immune to telescope-specific phase errors—measures the phase sum around a triangle of baselines. Non-zero closure phases (e.g., >1°) indicate asymmetric brightness distributions, enabling model-independent detection of off-axis companions despite atmospheric turbulence.

Optical-polarization-state transformation formulas are widely used in MIRC-X interferometers to analyze and design polarization measurement paths. As light waves pass through various optical elements in an interferometer, the polarization state of the output light must be determined to accurately measure the polarization characteristics of the incident light. Instrumental polarization effects in observational data are corrected to extract the true polarization state of a target object, which generally involves a detailed calibration of the instrumental polarization effects and the use of the calibration data to correct the observed polarization signal. The state-of-light polarization transformation equation [38] is given by

$$\left(\begin{array}{l}{P}_{f,1}\\ P_{f,2}\end{array}\right)=f\cdot e^{i\varphi }{M}_{MX}WP(\zeta ,\theta )\left(\begin{array}{cc}1& 0\\ 0& {\tilde{M}}_{8-19}\end{array}\right)R(A)\left(\begin{array}{cc}1& 0\\ 0& {\tilde{M}}_{4-7}\end{array}\right)R(a)\left(\begin{array}{cc}1& 0\\ 0& {\tilde{M}}_{1-3}\end{array}\right)R(q)\left(\begin{array}{l}{P}_{i,\alpha }\\ P_{i,\delta }\end{array}\right)$$

(9)

$P_{\mathrm{i}}$: Initial Polarization state defined by the Jones vector, $P_f$: final polarization state, $M_X$: polarization matrix of the MIRC-X instrument, and $WP\left(\zeta ,\theta \right)$: the contribution of the half-wave slice as a function of its orientation angle θ and delay ζ. $R_{\mathrm{i}}$: the rotation matrix that takes into account the azimuth, elevation, and parallax angles in the optical path, $\mathrm{f}$: net flux across the optical path, and $P^{\prime }$: polarization state of the detector.

The formula was derived based on the theory of optical polarization and the specific optical design of the CHARA array and MIRC-X instrument. It considers all polarization transformations that light undergoes as it passes through an optical system. A half-wave plate was used to control the polarization of the measured light. By rotating the half-wavelength, the polarization direction to which the prism is sensitive can be changed, thereby measuring the properties of the target object in different polarization directions. The half-wave slice matrix equation is

$$WP(\zeta ,\theta )=\mathrm{R}(\theta )\left(\begin{array}{cc}1& 0\\ 0& {\mathrm{e}}^{i\zeta }\end{array}\right)R(-\theta ),$$

(10)

where R(θ) and R(−θ) denote the rotation matrices that rotate the coordinate system by angles θ and −θ, respectively; and $\left(\begin{array}{cc}1& 0\\ 0& e^{i\zeta }\end{array}\right)$ denotes the delay effect of a half-wave slice, where $e^{i\zeta }$ is a complex exponential representation of the delay.

2.1.5. Breakthrough in Mid-Infrared: LBTI and NIC

The successful validation of MIRC in 2006 demonstrated the potential of multi-telescope interferometry to achieve high angular resolution and model-independent imaging in the near-infrared band. However, it also revealed the limitations of traditional nulling techniques in more complex scenarios; for example, when the contrast of the target celestial object spans multiple orders of magnitude (e.g., exoplanets and their host stars), photon noise and atmospheric turbulence in the near-infrared band significantly degrade detection capabilities. To overcome this bottleneck, interferometry techniques have been extended to longer wavelengths (mid-infrared to thermal infrared) to leverage the rapid decay of stellar radiation at longer wavelengths, thereby naturally improving the signal-to-noise ratio for weak targets (such as planets or dust disks).

In 2008, Hinz et al. proposed the large binocular telescope interferometer (LBTI) project [39], the principle of this project is shown in Figure 4. LBTI operates in the mid-to-long infrared (3–13 μm) region as its primary detection window, resolving the fundamental problem of low planet-to-star contrast. LBTI performs nulling interferometry in the thermal infrared band (>2.5 μm) to suppress light from the parent star, thereby enhancing the detection of exoplanets [40]. To achieve the scientific objectives of LBTI, a specialized camera was designed to perform both nulling interferometry and imaging, called the nulling and imaging camera (NIC). The NIC comprises two scientific channels: a nulling-optimized mid-infrared camera with a wavelength range of 8–13 μm and sensitivity reaching 10 μm, enabling direct imaging of habitable-zone giant planets, and an imaging camera with a wavelength range of 3–5 μm for concurrent acquisition of near-infrared structural information, enabling multi-band joint inversion (e.g., dust disk temperature error < 10 K). Additionally, it includes a K-band channel for phase-change sensing and closed-loop correction, controlling piezoelectric ceramics at a closed-loop frequency > 500 Hz to compensate for optical path differences at the sub-nanometer level (ΔL < λ/1000), thereby suppressing atmospheric turbulence-induced wavefront aberrations [39]. In the NIC, the three-channel data streams form a “sensing-cancellation-imaging” closed-loop system, stabilizing the cancelation depth at δ~10⁻³ (1000 times better than traditional systems), enhancing the detection capability of weak signals from exoplanets by two orders of magnitude. The NIC is particularly suitable for detecting circumstellar dust disks and giant planets around nearby stars. Using nulling interferometry, it exhibits unique sensitivity in the 3–25 μm spectral range during ground-based observations, with a demonstrated capability to resolve dust disk structures at 0.4″ from the star (equivalent to the scale of the asteroid belt in the solar system) at a 20 m baseline.

2.1.6. Integrated Photonics Revolution: Dragonfly and GLINT

LBTI has achieved breakthrough observations of exoplanets and dust disks using nulling interferometry in the thermal infrared wavelength band. However, its complex multichannel closed-loop correction system faces inherent challenges, such as optical component redundancy, phase-stability dependence on mechanical adjustment, and mechanical stability of discrete components (e.g., sub-nanometer accuracy requirements for piezoelectric ceramic phase correction). To simplify the nulling interferometer architecture fundamentally and improve system robustness, Jovanovic et al. developed the Dragonfly interferometer based on integrated photonics in 2012 [41], achieving beam splitting, routing, and recombination on a single chip for the first time. The integrated photonics nulling interferometer technology used in the Dragonfly project achieves nulling through phase control built into waveguide structures, completely distinguishing itself from traditional polarization nulling methods. Polarization cancelation relies on birefringent crystals/wave plates to modulate polarization states, and has two major drawbacks: polarization dependency and system complexity. It requires incident light with specific polarization states, while astronomical light sources are predominantly unpolarized, necessitating multistage polarization beam splitters and delay waveplates, introducing additional losses. The comparison shows that the integrated photonic nulling interferometer has advantages.

Dragonfly is an integrated photonic nulling interferometer that achieves nulling interferometry through a single-chip silicon nitride waveguide chip, enabling precise control of the light field through built-in phase shifts within the waveguide. First, the incident light is split, phase modulated, and then directed through a superluminal wave coupler to be evenly distributed across multiple waveguides. Half of the interference arms integrate thermo-optic phase shifters

$$\mathrm{\Delta }\varphi ={\displaystyle \frac{2\pi }{\lambda }}\cdot {\displaystyle \frac{\partial \mathrm{n}}{\partial T}}\cdot \mathrm{\Delta }T\cdot L$$

(11)

where ∂n/∂T is the thermo-optic coefficient (2.45 × 10⁻⁵ K⁻¹) of Si₃N₄, ΔT is the temperature control accuracy ±0.1 K (corresponding to a phase error < 0.02 rad), T is the temperature, and $\lambda$ is the wavelength. Temperature control is used to introduce a precise π-phase shift into the phase variation $\mathrm{\Delta }\varphi$. Coherent nulling synthesis is then performed, and the phase-shifted optical field is superimposed in a multimode interference coupler. Stability is guaranteed by suppressing the path error with laser direct-write waveguide length matching [42].

The innovations include the following. In terms of flux, the swift wave coupler improves the optical transmission efficiency by a factor of 20. The flux enhancement formula is

$${\eta }_{Dragonfly}={\displaystyle \frac{P_{out}}{P_{in}}}\approx e^{-\alpha L}\cdot {\left|\kappa \right|}^2$$

(12)

Among them, ${\eta }_{Dragonfly}=20\times {\eta }_{mask}$ where $\alpha$ is the waveguide transmission loss, L is the waveguide length, $\kappa$ is the swift wave coupling coefficient (design value > 0.98), P is the power, ${\eta }_{mask}$ is the optical flux efficiency of the traditional sparse aperture mask (Sparse Aperture Mask, SAM). The laser direct-write waveguide control length difference is ≤36 μm, and $\mathrm{\Delta }\varphi ={\displaystyle \frac{2\pi \mathrm{\Delta }L}{\lambda }}<0.07\lambda$ in subwavelength path matching.

In terms of imaging stability, the closed-phase fluctuation with RMS = 0.4° supports complete Fourier space sampling. This architecture provided a technical paradigm for subsequent photon extinction interferometers such as GLINT. Subsequently, the concept of an integrated photon-nullifying interferometer based on ultrafast laser direct-writing technology was proposed and gradually developed [43].

In 2015, Errmann et al. fabricated and tested a four-channel zero-ablation interferometric assembly using integrated optics [44]. They used ultrafast laser engraving to fabricate photonic chips containing three 50/50 directional couplers that combined optical signals from four telescopes in a two-stage cascade. By simulating the motions of stars across the sky, they measured the expected angular transfer function of the four-telescope nulling coupler and verified that the transmission properties agreed with theoretical calculations, thus simplifying the optical design of the multi-telescope nulling interferometer.

The Dragonfly interferometer directly responds to the demand for high flux and low noise in nulling interferometry through optical flux enhancement and phase stabilization. Leveraging traditional optical interferometry (e.g., LBTI), Dragonfly initiated photonic-chip nulling interferometry (e.g., GLINT).

Since exoplanet research has shifted to fine characterization, direct imaging must has been developed simultaneously with ultra-high contrast (${10}^4–{10}^8$) and sub-diffraction-limited angular resolutions (close to or below the diffraction limit). To test the performance of the integrated photon-nulling interferometer in a real astronomical observation, the GLINT Pathfinder has been developed. It is the first photon chip-integrated nulling interferometer in a ground-based telescope (Subaru Telescope) capable of detecting exoplanets under real atmospheric turbulence.

In 2016, the GLINT path detector was successfully tested on the Subaru Telescope for the first time. The instrument naturally filters out higher-order turbulent aberrations through the turbulence suppression principle, i.e., “single-mode waveguide filtering of turbulent higher-order modes”, by splitting the telescope pupil into sub-apertures and injecting a single-mode silicon nitride waveguide that transmits only the phase-consistent Erie spot center field (mode field diameter~λ), reducing the wavefront error to less than λ/50. At the nulling port, the directional coupler introduces a π-phase shift that suppresses the stellar light to δ~10⁻⁴ (measured value), and at the inverse nulling port, the stellar reference light is retained for real-time calibration of the instrument polarization and phase drift. The numerical self-calibration (NSC) algorithm is used to fit the two-port signals and decouple the astrophysical nulling depth using the following formula

$${\delta }_{ast}={\displaystyle \frac{V_{null}-k\cdot V_{antinull}}{V_{antinull}}}$$

(13)

k: system transmission ratio; $V_{null}$ and $V_{antinull}$ represent the voltage values of the nulling and anti-nulling signals, respectively.

The method increases the accuracy of the stellar angular diameter measurements to 0.1 mas. It also successfully resolves a binary star system at 1.6 λ/D (angular separation of 15 mas), and the information recovery capability exceeds the telescope diffraction limit (λ/D = 40 mas) [45].

The GLINT instrument operates at a wavelength of 1.6 microns (1600 nm) with a bandwidth of 50 nm. The GLINT instrument achieves an extinction depth accuracy of approximately 10⁻⁴. For large self-luminous planets, the required near-infrared contrast range is approximately 10⁻⁴. For Earth-like planets that reflect the light of their host stars, the required contrast range is approximately 10⁻⁸. The GLINT instrument can detect stars with angular diameters as small as the mas resolution. For example, measurements of α Bootis show an angular diameter of 18.9 mas, demonstrating the very high angular resolution of GLINT. Measurements of other stars show similarly high-precision results, e.g., α Herculis with an angular diameter of 31.2 mas, χ Cygni with an angular diameter of 20.5 mas, etc. [46]. The GLINT project serves as a pathfinder that integrates the concept of nulling interferometry into a compact and rugged platform of a photonic chip. The chip contains complex optical circuits such as waveguides, beamsplitters, and directional couplers, and is capable of measuring both “zeroed-out” signals (planetary rays) and “anti-zeroed-out” signals (redirected stellar rays). The astrophysical nulling depth is estimated using the NSC and is in agreement with known interferometric values. The compact chip (5 × 5 mm²) with turbulence suppression capability will play an important role in exoplanet imaging for future 30 m class telescopes [43].

2.1.7. Future Directions: Kernel-Nulling and the LIFE Mission

Integrated photonic nulling (e.g., Dragonfly, GLINT) has demonstrated on-chip beam combination and turbulence suppression. However, ground-based phase stability remains limited by atmospheric OPD (optical path difference) fluctuations. To achieve the sub-nanometer stability required for terrestrial planet characterization, space-based interferometry is essential. The Large Interferometer for Exoplanets (LIFE) mission leverages photonic advancements and introduces kernel-nulling—a robust beam-combination formalism that is insensitive to OPD errors—enabling high-contrast spectroscopy in the mid-infrared (4–18 μm). Following earlier work in the field, Martinache and Ireland later introduced the concept of kernel-nulling, which is a method for constructing a zero-space vector (kernel) for the interferometric equations, and allows multiple telescope beam combinations to be robust up to against OPD errors [47]. The LIFE mission uses a five-telescope architecture that can rigorously generate an optimal kernel K satisfying $K^{T}V=0$ (irrespective of the perturbation of the OPDs), where V = (V1, V2, …, V10)^T is the baseline complex visibility vector and K^T is the transpose of K.

LIFE’s astrometric precision builds upon heritage from the Space Interferometry Mission (SIM Lite) [48]. Though canceled, SIM pioneered space-based microarcsecond astrometry and kernel-nulling techniques critical for exoplanet detection. Its technology demonstrators validated picometer-level pathlength control and “kernel-phase” signal extraction algorithms in simulated exoplanet systems. SIM’s legacy proves that disentangling planetary signals from stellar leakage requires not only deep nulling but also ultra-precise relative astrometry of the host star—a principle central to LIFE’s ability to detect Earth-mass planets and measure orbital parameters for atmospheric retrieval.

Conventional ablation relies on a single π-phase shift and requires absolute optical range matching (ΔL < λ/100). Nuclear ablation constructs the mathematical kernel K, which is the zero-space basis of the interference equations, with the planetary signals located in their orthogonal complementary space [49]. To achieve this, the LIFE project was initiated to identify the effects of systematic uncertainties by developing a mathematical framework around a nulling-beam combiner and using a spatial-interference simulator. The transmission matrix M of the beam combiner is realized by a series of unequal beam splitters and phase shifted plates, each described by a mixing angle θ and a phase shift ϕ. To achieve the desired extinction depth (null depth), the amplitude (intensity) mismatch between the two beams must be less than a specific value, with amplitude and phase mismatch requirements:

$$\begin{array}{l}{\displaystyle \frac{\mathrm{\Delta }I}{I}}<4\sqrt{N}\\ \mathrm{\Delta }\varphi <2\sqrt{N}\end{array}$$

(14)

where N is the nulling depth defined as the dark output intensity divided by the bright output intensity, ΔI is the difference in amplitude (intensity) between the two beams, I is the average amplitude (intensity) of single light beam, and $\mathrm{\Delta }\varphi$ is the phase difference between two beams of light.

The system phase error requirement

$$\mathrm{\Delta }\varphi <{\displaystyle \frac{6K{\lambda }^2}{B^2{\theta }_{UD}^2}}$$

(15)

is used to calculate the conditions that the systematic phase error must satisfy in order to meet the calibrated nuclear zeroing depth requirements. Here, K is the constant associated with the defuzzification architecture, λ is the wavelength, B is the defuzzification baseline length, and ${\theta }_{UD}$ is the angular diameter of the star.

The condition that must be satisfied for a systematic mismatch between magnitude and phase satisfy the nulling requirement is described with a magnitude-phase mismatch term:

$$\mathrm{\Delta }\varphi \mathrm{\Delta }I<1.4K$$

(16)

The LIFE 5-satellite constellation mission concept (baseline 39–500 m) is designed to utilize integrated photonic nulling interferometry. Instrument performance simulations project that LIFE will achieve a stellar suppression ratio (null depth) of ~10⁻⁷ (corresponding to a contrast ratio of 10⁻⁷ for point sources) at 10 μm wavelength under optimal conditions, along with an angular resolution of 0.3 mas across the mid-infrared band (4–18 μm) [5,9]. Its core technologies are: 1. sub-nanometer constellation control to ensure interferometric coherence; 2. thermo-optic modulation nulling, where the nulling depth varies with the square of the angular position α in the second-order nulling (second-order null), and with the fourth power of the angular position α in the fourth-order nulling (fourth-order null). The specific nulling depth values are influenced by optical component errors, such as the reflectance error of the beam splitter and phase shift errors. When the errors are |ΔR| = 5% and Δϕ = 3°, the fourth-order null maintains high null performance in the 4–19 μm wavelength band. 3. Biomarker extraction: the planetary flux is inverted based on zero-signal visibility and combined with the biological confidence index to determine life signals. This design directly addresses the atmospheric characterization requirements for super-Earths such as Kepler-725c. By synthesizing the spectra, LIFE can detect CH₄/O₃ biomarker combination (SNR > 5) of Earth-like planets 30 light-years away, with an annual detection capacity exceeding 300 habitable zone planets. LIFE’s scientific objectives are clearly defined as dedicated to searching for biosignatures of exoplanets, detecting important biochemical substances such as methane, ozone, and carbon monoxide, analyzing whether life exists by collecting spectral lines of these chemicals, and successfully detecting life signs on Earth by testing Earth as an exoplanet. As shown in

Table 1. Advantages and upgrades of LIFE technology compared to previous technologies.

Predecessor Technology	Innovations Absorbed by LIFE	Upgrade Point
Dragonfly	Photonic chip thermo-optic phase shift	Space radiation-resistant version (single-event rollover rate < 10−7)
GLINT	Single-mode waveguide turbulence filtering	No filtering required (no turbulence in space)
LBTI	Mid-infrared band advantage	Extends to 18 μm (covers more molecules)

, LIFE technology has obvious advantages over previous technologies.

As shown in Figure 5, LIFE uses an array of multiple telescopes to collect starlight, which are arranged in a specific geometric configuration. The starlight collected from each telescope is guided through a zero-difference interferometric optical path to a central beam combiner. The light from the target star (i.e., the central star) is mutually canceled out in the interferometer, while the light from the surrounding planets is retained due to differences in phase.

As the ultimate form of nulling interferometry, LIFE integrates GLINT’s photonic chip technology, formation flying precision, and a new biological confidence model, establishing a benchmark for the “Sound Seeking Program” in the 2040s. Additionally, the LIFE project has been proposed as one of the major missions under the European Space Agency’s Voyage 2050 program.

2.1.8. Dynamic Maintenance

For the dynamic maintenance of the nulling state, the nulling interferometry must combat atmospheric turbulence, mechanical vibration, and thermal drift perturbation during the exposure period [50]. Currently, the mainstream scheme adopts three-level closed-loop control: 1. OPD stabilization layer, where the outlier interferometer solves the turbulence phase in real time ${\varphi }_{turb}(t)$ and the piezoelectric ceramic compensates the optical range with 0.1 nm accuracy (e.g., the LBTI stabilizes to the σOPD < λ/5000 in 1 h); the outlier interferometer obtains the nanosecond phase change by demodulation

$${\varphi }_{\mathrm{turb}}(t):\mathrm{\Delta }\varphi (t)=2\pi \cdot \mathrm{\Delta }f\cdot t+{\varphi }_{\mathrm{turb}}(t)$$

(17)

where Δϕ(t) denotes the phase change at time t, Δf denotes the change in frequency, t denotes time, ϕ turb(t) denotes the change in frequency due to the turbulence; 2. Polarization control layer: the reference light calibrates the Mueller matrix change ΔM, and the liquid crystal modulator dynamically corrects the fast-axis angle α and delay δ (residuals < 0.1%); 3. Thermodynamic layer: zero-expansion material and PID temperature control suppresses the drift (up to ±0.001 K for spatial systems). In ExAO systems (e.g., MagAO-X), the above process is compressed to the ms level: the anamorphic mirror corrects the wavefront at 2 kHz, and the fiber optic phase controller locks the OPD at 5 kHz, allowing the 10⁻⁵ ablation depth to be stabilized and maintained for 10 min at a 0.8″ optic.

2.1.9. Categorization

Nulling interferometry can be primarily divided into four types: spectral dispersion, chromatic aberration, polarization control and Integrated Photonic Nulling Interferometry.

Spectral Dispersion

Principle: Nulling interferometry technology is typically combined with spectral dispersion technology, which utilizes the dispersion properties of light to divide a broad spectral range into multiple narrow spectral bands (or channels) for processing. By introducing gratings or prisms into the optical system, stellar spectra are dispersed to different spatial positions, thereby canceling out light intensity within specific wavelength bands to enhance the accuracy and resolution of interferometric measurements. The advantage of this approach is that it allows independent optimization of the interferometric performance at different wavelengths while enhancing contrast. This facilitates separate measurement of interferometric signals at different wavelengths, enabling a more precise reconstruction of the image of the target source. By analyzing interferometric patterns at different wavelengths, it is possible to distinguish between interference signals generated by the target source itself and noise signals caused by the instrument or atmosphere, thereby reducing or eliminating the influence of noise signals. For example, nulling interferometry in the L and M bands (3–5 microns) can be used for high-resolution spectral observations of specific molecules (such as CH₄ and CO) in planetary atmospheres. This method has significant advantages in wideband imaging, because it effectively suppresses the continuous spectrum of stars while preserving the characteristic spectral information of faint celestial objects. The change in zero-elimination efficiency due to the change in path length at different wavelengths is given as:

$${\eta }_{\mathrm{null}}(\lambda )=1-\sin c^2({\displaystyle \frac{\pi \mathrm{\Delta }L}{\lambda }})$$

(18)

Here, η_null(λ) denotes the nulling efficiency, i.e., the efficiency at a specific wavelength λ; ΔL denotes the change in the path length, i.e., the difference in path lengths as light propagates through different media; c is determined by the core design parameters of the interferometer, mainly including the interferometer baseline length, center design wavelength, beam combiner optical parameters, etc.

For example, in the LIFE mission the interferometric signals within each channel can be processed separately by splitting the mid-infrared spectral range (4–19 µm) into multiple spectral channels. This spectral segmentation helps optimize the phase and amplitude corrections of the interferograms, particularly when using the kernel-nulling technique, which can more accurately remove stellar leakage and improve the sensitivity of exoplanet detection.

Color Difference Correction Type (Chromatic Aberration Control)

Principle: Chromatic aberration refers to a phenomenon in which light of different wavelengths refracts at different angles in an optical system, causing shifts in the image position. Chromatic aberration correction utilizes the inherent chromatic aberration characteristics of an optical system by precisely designing the parameters of the optical elements to ensure that light of different wavelengths converges at different focal points on the focal plane. By selecting an appropriate wavelength range and optical configuration, it is possible to cancel the light intensity at specific locations, thereby enhancing the image contrast. For example, in the implementation of FMOS (Fiber Multi-Object Spectrograph), the following steps are implemented:

• Prime focus corrector design: The FMOS prime focus corrector is designed to optimize the image quality within the 0.9–1.8 µm wavelength range [51]. The corrector employs a three-element lens design, with all elements made from BSM51Y glass material to minimize chromatic and other optical aberrations.
• Plan-focal field: The corrector is designed with a plan-focal field to simplify the accurate focusing of the fiber tips and module installation. The plan-focal field also ensures that the main light rays are aligned with the fiber axis across the entire field of view, minimizing the efficiency losses caused by a mismatch between the beam and the fiber acceptance cone. The advantages of chromatic aberration correction are its relatively simple optical design and high adaptability.

Correction methods:

• Using dispersion compensation elements such as prisms or gratings to guide light of different wavelengths to the correct paths.
• Optimizing optical design using achromatic lens groups to reduce the effects of chromatic aberration.
• Real-time wavefront correction using adaptive optical systems to monitor and correct wavefront distortions in real time, including chromatic aberration.

The phase difference due to color difference can be expressed as:

$$\mathrm{\Delta }\varphi ={\displaystyle \frac{2\pi \mathrm{\Delta }L(\lambda )}{\lambda }}$$

(19)

where ΔL(λ) is the wavelength-dependent path difference and λ is the wavelength of light. The range of ΔL(λ) can be reduced by sub-spectral chopping to decrease $\mathrm{\Delta }\varphi$. The phase delay and amplitude ratio of the interferometer can be adjusted at different wavelengths to optimize the nulling effect. It is mainly used for probing planetary atmospheric composition and temperature structure.

Polarization Control

A polarimeter is used to control and analyze the polarization state of light. Based on the polarization characteristics of light, polarizers and waveplates are introduced into optical systems to modulate the polarization state of starlight. Because starlight typically has a low polarization degree and the surrounding planets or dust disks may exhibit higher polarization, polarimeters can effectively distinguish and suppress starlight while enhancing the signals from faint celestial objects. In null interference, the consistency of polarization states is critical to ensure interference effects. Light with different polarization states produces additional phase differences during interference, which affects the visibility of the interference fringes. In interferometers, polarization control can be used to eliminate or reduce the effects of polarization on interference signals, thereby improving the measurement accuracy. For example, MIRC-X uses rotating birefringent plates (such as LiNbO₃ plates) to correct the polarization state of light [52]. These birefringent plates are rotated to adjust the polarization direction of the light, ensuring that the polarization states of the two interfering light beams are consistent. In addition, MIRC-X is equipped with half-wave plates and Wollaston prisms to measure and correct the full Stokes parameters of light, further optimizing polarization control performance.

Application methods:

• Polarization control: Ensures that the light waves entering the interferometer have consistent polarization states to minimize polarization-induced changes in interference signals.
• Polarization separation: In certain cases, it may be necessary to separate light waves with different polarization states for processing to eliminate the effects of polarization on specific measurement tasks.
• Polarization calibration: During calibration process of an interferometer, a polarimeter is used to measure and adjust the polarization state of the optical path to ensure the accuracy and stability of the interferogram.

Polarization states are typically described using Stokes parameters, including S0 (total light intensity), S1 (difference between horizontal and vertical polarization components), S2 (difference between +45° and −45° polarization components), and S3 (difference between right-handed and left-handed circular polarization components). In an interferometer, these parameters can be measured to assess and control polarization effects.

The total light intensity is

$$S_0=I$$

(20)

The difference between horizontally and vertically polarized light intensities is

$$S_1=I_H-I_V$$

(21)

The difference between diagonally and anti-diagonally polarized light intensities is

$$S_2=I_D-I_A$$

(22)

The difference between right- and left- circularly polarized light intensities is

$$S_3=I_R-I_L$$

(23)

Here, $I,I_H,I_V,I_D,I_A,I_R,I_L$ denote the total light intensity, horizontally polarized light intensity, vertically polarized light intensity, diagonally polarized light intensity, anti-diagonally polarized light intensity, right circularly polarized light intensity, and left circularly polarized light intensity, respectively.

The degree of polarization ($P$) can be calculated using the following formula:

$$P={\displaystyle \frac{\sqrt{S_1^2+S_2^2+S_3^2}}{S_0}}$$

(24)

Integrated Photonic Nulling Interferometry

Integrated Photonic Nulling represents a paradigm shift in nulling interferometry, replacing traditional bulk optics with photonic integrated circuits (PICs) fabricated on substrates like silicon nitride (Si₃N₄) or silica [53]. This approach miniaturizes the entire beam combination and phase control process onto a single chip, achieving starlight suppression through precise guided-wave phase manipulation rather than free-space optics. Light collected by telescopes is injected into single-mode waveguides, which act as spatial filters by rejecting high-order atmospheric turbulence modes (e.g., >λ/50 wavefront error). Within the PIC, directional couplers split the beams, while integrated thermo-optic phase shifters—modulated via localized microheaters—induce the critical π-phase shift for destructive interference. Figure 6 shows a typical integrated photonic nulling interferometry represented by GLINT.

Path-length differences are minimized to subwavelength tolerances (≤36 μm) via lithographic precision, ensuring coherence across baselines. Recombination occurs in multimode interference (MMI) couplers, with null depths stabilized by closed-loop control using “null” and “anti-null” output ports. Crucially, this architecture operates independently of polarization states, eliminating the need for complex birefringent optics required in traditional polarization-based nullers. Implementations like GLINT (Ground-based Lunar Imaging Nulling Telescope) and Dragonfly have demonstrated on-sky null depths of ~10⁻⁴ at 1.6 μm, leveraging turbulence filtering and real-time calibration algorithms (e.g., Numerical Self-Calibration) to isolate planetary signals.

2.1.10. Distinction Between Space-Borne and Ground-Based Implementations in Nulling Interferometry

Having discussed the core principles, historical evolution, and enabling technologies of nulling interferometry, it is crucial to recognize that its implementation and scientific potential are profoundly shaped by the observational platform—ground or space.

The development and application of nulling interferometry show a clear distinction between ground-based and space-borne platforms, driven by their respective environmental constraints and scientific goals. Ground-based projects like the OHANA phases (I, II, III) and MIRC/MIRC-X on the CHARA array demonstrated the feasibility of fiber-linked interferometry and high-sensitivity, multi-telescope beam combining under atmospheric turbulence. These efforts focused on technological validation, achieving high angular resolution in the near-infrared for stellar astrophysics. However, they are fundamentally limited by atmospheric turbulence, demanding sophisticated real-time phase control (e.g., via ExAO integration in the Keck Interferometer) and restricting achievable null depths (typically~10⁻⁴) and sensitivity for faint exoplanet signals. The GLINT pathfinder represents a significant step forward for ground-based telescopes, utilizing photonic chip integration and single-mode filtering to achieve deeper nulling (~10⁻⁴) and turbulence suppression.

In contrast, space-borne nulling interferometry, free from atmospheric limitations, targets the ultimate goal of detecting and characterizing Earth-like exoplanets via their thermal infrared spectra. Concepts like DARWIN and TPF-I were designed as pioneering space mission proposals for this purpose, though they were not realized. The LIFE mission concept is the current forefront of space-based nulling, aiming for unprecedented contrast (~10⁻⁷ at 10 μm) and sensitivity in the mid-infrared (4–18 μm) using a formation-flying telescope array. Its kernel-nulling approach and thermo-optic modulation are designed for the stability required in space. While ground-based projects like GENIE (a proposed ground-based nuller for the VLTI) aim to bridge the technology gap for future space missions [54], the inherent advantages of space for deep mid-infrared exoplanet characterization remain compelling, driving the focus of advanced concepts like LIFE towards the space environment. As illustrated in

Table 2. A comparison of the primary techniques used to eliminate nulling interferometry.

Technical Name	Nulling Interferometry	Spectral Band	Resolution (of a Photo)	Special Advantages (Other Data)
OHANA Phase II	~10−4	Near-infrared bands (J, H, K) vary at different	Varies at different	Low-loss transmission using fiber-optic connection technology, up to 500 m fiber length, maximum propagation loss of 1 dB, transmission efficiency conservatively estimated at 50%.
MIRC-X	Better than 1 degree (closed phase accuracy)	Near-infrared (J and H bands, future plans include K band)	Equivalent to the angular resolution of up to 330 m diameter baseline telescopes (~0.6 mas)	Sensitivity improved by about two magnitudes compared to its predecessor MIRC, C-RED ONE camera adopted, readout noise below 1 electron/pixel
LBTI, NIC (NOMIC/LMIRcam)	NOMIC: Relative tilt of less than approx. 3 mas is required for a zero suppression of 10−4 (depth of zero suppression not directly given)	NOMIC: 7–25 microns; LMIRcam: 3–5 microns	NOMIC: 0.1–0.35 arcsec (7–25 microns); LMIRcam: 0.04–0.07 arcsec (3–5 microns)	NOMIC: N = 0.1 mJy in 1 h, spectral resolution 100; LMIRcam: L’ = 20, M = 17 in one hour, spectral resolution 350
GLINT	~10−4 (reached during testing in the sky)	1.6 micron (1600 nm) with 50 nm bandwidth	~25–60 mas	Single photonic chip design for high stability and compactness, with future plans to increase the number of baselines to improve sensitivity
LIFE Cluster	Contrast ratio will achieve 10−7 (10 μm, sun-to-Earth analogy)	4 to 19 microns (mid-infrared)	Improved detection sensitivity and contrast through five-telescope kernel nulling beam synthesizer design	The five-telescope kernel nulling beam synthesizer is designed with redundancy so that even if one or both telescopes fail, the system will continue to produce robust observable data

, the performance metrics and primary advantages of various nulling interferometers reveal a clear progression in capability from ground-based validation to the future space-based LIFE concept.

2.2. Coronagraph

After systematically elucidating the physical mechanism by which nulling interferometry suppresses stellar speckles through dynamic phase modulation, we discuss another high-contrast imaging technique—the coronagraph. Unlike nulling interferometry, which relies on a precise interferometric system with multi-path phase modulation, the coronagraph employs a closed-loop control system to precisely compensate for wavefront aberrations, transforming the phase singularities of the stellar point spread function into adjustable dark-field regions. The coronagraph introduces light-blocking structures or phase-modulating elements into the optical path of the telescope to directly perform spatial filtering and energy redistribution of stellar diffraction light. This technical approach overcomes the strict constraints on baseline length and wavelength imposed by nulling interferometry, demonstrating superior engineering adaptability in the visible to near-infrared wavelength range [55]. However, these two technologies are not simple substitutes of each other: nulling interferometry has a unique advantage in terms of dynamic range in deep-space observations in the near-infrared wavelength band, while coronagraphs achieve breakthroughs in direct imaging of Jupiter-like planets on ground-based telescopes through innovative designs (such as vector fork-shaped light-blocking structures and phase masks). The complementary nature of these two technologies provides important insights for the future development of hybrid high-contrast imaging systems that integrate multiple techniques.

Coronagraphs eliminate the effects of the photosphere by blocking, absorbing, or diffracting light. The core objective is to create a “dark zone” on the image plane to eliminate diffraction-induced light from the primary star. Coronagraph designs balance three key parameters: Inner Working Angle (IWA: smallest detectable separation), Contrast (planet/star flux ratio), and Throughput (planet light transmission) [56]. The evolution of these designs has followed two main paths: External Coronagraphs (e.g., starshades): Achieve small IWA through geometric occlusion but require formation flying; Internal Coronagraphs: Integrate optics within the telescope. Subtypes address diffraction in different ways: 1. Interferometric Coronagraphs (e.g., 4QPM): Phase manipulation; 2. Lyot-type Coronagraphs: Focal-plane masking + pupil filtering; 3. Pupil Apodization Coronagraphs (e.g., PIAA): Amplitude/phase modulation at the pupil plane.

As shown in Figure 7, because of the intense light from stars, it is necessary to block or suppress the diffraction of light due to the telescope aperture. By altering the field strength of the light waves at the pupil or modulating the point spread function (PSF), the final light intensity distribution of the system can be modified, thereby achieving high-contrast imaging in specific regions.

2.2.1. External Occultation Coronagraph

The working principle of the external occultation coronagraph is similar to that of the external occultation solar coronagraph. By placing an external occulting device in front of the telescope, the light from the primary star is geometrically blocked, creating an artificial “solar eclipse” effect [57]. Compared with the internal occultation scheme, the advantages are as follows:

• Simplified optical system: No special telescope or adaptive optics systems are required; only a conventional telescope is required.
• Breakthrough in the diffraction-limited inner working angle: The size depends on the dimensions and position of the external shelter, and is not constrained by the optical system itself. The minimum theoretically detectable angular distance can reach 0.01″ (better than the diffraction limit λ/D ≈ 0.03″). However, unlike the external coronagraph, the stellar angular size is significantly smaller than that of the Sun (16 arcmin), leading to significantly increased technical challenges. For a 4 m aperture telescope, the diameter of the outer occulter must reach 50 m, and the distance from the telescope must be 80,000 km. The internal working angle of the coronagraph (generally within 0.5 arcs in the visible light band) is also significantly smaller than that of the solar coronagraph. The outer occulter must be carried by a spacecraft and operated in a very high orbit, and must remain synchronized with the telescope at all times, making implementation extremely challenging. Due to the above limitations, the inner-occultation coronagraph has emerged as the mainstream approach [58].

2.2.2. Internal Occultation Coronagraphs

An internal occultation coronagraph is a type of coronagraph that suppresses the primary star’s light within the optical system to enable planet detection. Since the late 1990s, growing exoplanet detection demands and technological advances have spurred the proposal of numerous internal occultation coronagraph designs. When classified according to their extinction principles, however, these fall into three categories: interferometric coronagraphs, improved Lyot-type coronagraphs, and Pupil Apodization Coronagraphs. The following sections will summarize these three coronagraph types.

The Interferometric Coronagraph

The interferometric coronagraph was the first occultation coronagraph proposed, with Bracewell first suggesting the use of interferometric principles to separate planetary light from the primary star in 1978 [59]. Over the following two decades, interferometric coronagraphs remained the primary technical approach, with the most representative example being Guyon’s 2006 [60] proposal of the Pupil swapping coronagraph. However, interferometric coronagraphs are limited by phase stability and bandwidth constraints. As a result, they have gradually been replaced by the more robust Lyot and pupiled coronagraphs.

Lyot-Type Coronagraphs

Invented in 1930 by the French astronomer Bernard Lyot, the core idea of the Lyot coronagraph is to simulate the effect of a total eclipse of the sun. An opaque circular baffle (mask) is placed in the focal plane of the telescope to block direct light from the star, and a Lyot diaphragm is added in the subsequent pupil plane to block the diffracted light, thus observing faint solar corona or exoplanets [61]. Imaging contrast is limited by the diffraction paraflap, which can be mathematically expressed as the energy distribution of the PSF:

$$I(\theta )\propto {\left|F\left[P(r)\cdot M(r)\right]\right|}^2$$

(25)

Here, I(θ) denotes the intensity distribution at angle θ, P(r) is the telescope pupil function, M(r) is the focal plane mask transmittance function, and F denotes the Fourier transform. The angular resolution of the conventional Lyot coronagraph is limited by the mask size, and side-valve noise makes it difficult for the contrast to exceed 10⁻⁵ orders of magnitude. According to the nature of the modulation plate, improved Lyot coronagraphs can be categorized into amplitude mask and phase-type coronagraphs.

• Band-Limited Mask Coronagraph (BLMC)

The band-limited plate coronagraph (BLMC), first proposed by Kuchner in 2002, is the main type of amplitude mask coronagraph [62]. In 2011, energy distribution of this coronagraph was optimized by amplitude modulation of the optical pupil surface. The mask uses a one- or two-dimensional sinc function-type transmittance with the following mathematical expression:

$$M(x,y)=\sin c^2({\displaystyle \frac{\pi D}{\lambda }}{\alpha }^{\prime }x)\mathrm{(one-dimensional)}$$

(26)

where M(x,y) denotes the transmittance of the mask at coordinates (x,y), D is the telescope aperture, λ is the wavelength, and α’ is the design constant. This design restricts diffracted light to a specific angular range, facilitating subsequent aperture blocking. BLMC offers advantages such as a simple optical path and a wide operational spectral range, but it also has drawbacks such as low light transmission efficiency and a large internal working angle. Moreover, it requires higher precision in micro-nano manufacturing. The basic principle is illustrated in Figure 8.

Practical application: The Nanjing Institute of Astronomy and Optics of the Chinese Academy of Sciences proposed the “finite band transmittance modulation” technique, discretizing the continuous sinc function into 5–7 ring bands. An experimental verification in 2019 achieved a contrast ratio of 10⁻⁷ at 4 λ/D (visible light band). The CPI-C module of the space telescope (2023) was combined with a liquid crystal space light modulator to dynamically optimize the BLMC mask, achieving a measured contrast ratio of 10⁻⁶ in the 4–12 λ/D range. Additionally, the 42 m aperture ground-based telescope E-ELT, currently in the planning and design phase, also plans to adopt BLMC as the primary solution for its coronagraph, aiming to achieve a contrast ratio of up to 3 × 10⁻¹⁰ within two circular fields of view in the 3 λ/D to 15 λ/D range, with a contrast ratio of 2 × 10⁻⁹ at 20% spectral bandwidth.

b.

Phase-type Coronagraphs

Unlike Lyot-type coronagraphs (which block starlight using opaque mask plates), phase-type coronagraphs introduce phase delay masks at the focal plane to redistribute the PSF energy through destructive interference of light waves. Its general form is

$$\mathrm{\Delta }\varphi (x,y)=\arg \left[e^{i\mathsf{\Psi }(\mathrm{x},y)}\right]$$

(27)

where Δϕ(x,y) denotes the amount of phase change at coordinates x,y, and arg denotes the radial angle, i.e., the phase angle, taken as a complex number. $e^{i\mathsf{\Psi }(\mathrm{x},y)}$ is a complex exponential function, where i is an imaginary unit, and $\mathsf{\Psi }(\mathrm{x},y)$ is the designed phase distribution function. By optimizing $\mathsf{\Psi }$ the stellar light phases out interference in the imaging region while preserving the planetary signal. Phase-plate coronagraphs have the advantages of a significantly high theoretical contrast, significantly small inner working angle, and high throughput efficiency. A variety of phase-plate coronagraph structures have been proposed, such as the four-quadrature phase-plate coronagraph (4QPMC), eight-quadrature phase-plate coronagraph, vortex phase-plate coronagraph, and sinusoidal phase plate. Among these, 4QPMC has been proposed as the best choice for imaging stars and planets [63]. 4QPMC uses a scalar phase mask technique so that the four quadrant phases alternate as 0/π with the formula

$$\mathrm{\Delta }\varphi (x,y)=\pi \cdot \mathrm{sgn}(x)\cdot sgn(y)$$

(28)

Phase modulation and working principle of 4QPM as the shown of Figure 9.

OVPMC (Optical Vortex Coronagraph) uses a Vector Vortex Coronagraph (VVC), which leverages the optic-axis-oriented design of the birefringent material to produce opposite helical phases for left- and right-circularly polarized light, with the phase varying continuously with azimuth:

$$\mathrm{\Delta }\varphi (\theta )=\mathcal{l}\theta$$

(29)

($\mathcal{l}$ is the topological charge number) [64]. The topological charge (ℓ) defines the number of 2π phase cycles around the optical axis. Unlike quadrant-based phase masks (e.g., 4QPM), which suffer from discontinuities at axes, the Vector Vortex Coronagraph (VVC) applies a continuous azimuthal phase ramp via subwavelength gratings. This enables broadband operation and eliminates diffractive artifacts, making VVCs highly suitable for high-throughput exoplanet imaging. For vortex coronagraphs, ℓ = 2 provides a wider dark zone than ℓ = 4 but requires stricter alignment. Higher ℓ values (e.g., ℓ = 6) improve broadband performance at the cost of throughput. A team from Princeton University team (2014) used an OVPMC with $\mathcal{l}=4$ to achieve 10⁻⁶ contrast (K-band) on the Keck telescope.

The sinusoidal phase-plate stellar coronagraph (SPM) proposed in 2012 achieved a good broadband performance by introducing sinusoidal phase modulation on the phase plate. The six-platform phase-plate stellar coronagraph (SLPM) proposed in 2014 further improved broadband performance, but introduced phase jump regions, requiring high manufacturing precision. With the growing need for coronagraphs capable of operating over wider wavelength bands, the introduction of chromatic aberration affecting imaging contrast has become a concerning issue. To address this problem, Xiegang et al. proposed a flat-top sine-phase-plate coronagraph (FSPM) in 2017 [65]. The FSPM employs a flat-top sinusoidal phase plate with an angular double-period structure that satisfies the requirements for broadband operation while maintaining a good achromatic performance.

The FSPM design combines the advantages of the SPM and SLPM, achieving a balance between broadband operation and high-contrast imaging through optimized phase distribution. Compared to SPM and SLPM, FSPM offers better internal working angles and achromatic performance. In the 490–620 nm wavelength range, FSPM achieves extinction of starlight at 10⁻³, with a broader operational wavelength range (539–561 nm) than that of the four-quarter plate phase star coronagraph, demonstrating superior achromatic performance.

For the FSPM, the complex amplitude distribution of light waves at the focal plane after passing through the coronagraph can be derived based on the principles of Fourier optics and coronagraphs. Given the incident light intensity, system parameters (such as the aperture stop radius, focal length, and central wavelength), and the phase plate transmission coefficient, the complex amplitude distribution at the focal plane can be calculated using this formula:

$$U_J({\mathrm{x}}^{\prime },{\mathrm{y}}^{\prime })=A_{1}t(\theta )\exp (i{\displaystyle \frac{{k{r}^{\prime }}^2}{2f}}){\displaystyle \frac{J_1(\beta r^{\prime })}{\beta r^{\prime }}}$$

(30)

where $A_1={\displaystyle \frac{2\pi A_0{R}_{AS}^2}{i{\lambda }_{0}f}}$, A is the amplitude of the incident light (constant), R is the radius of the aperture diaphragm, ${\lambda }_0$ is the center wavelength of the working bandwidth, $\beta$ is the focal length of the lens, and i is an imaginary unit. U_J (x′, y′) is the complex amplitude distribution at the focal plane. $R_{AS}$ is the radius of the aperture diaphragm, ${\lambda }_0$ is the center wavelength of the working bandwidth, f is the focal length of the lens, t(θ) is the transmission coefficient of the phase plate, θ is the angular coordinate at the focal plane, k is the wave number, r′ is the radial coordinate at the focal plane, $\beta$ is the parameter related to the radius of the aperture diaphragm and the focal length, and $J_1$ is the first-order Bessel function.

The Lyot-diaphragm pre-complex amplitude distribution is

$$U(x^{″},y^{″})=A_2{\displaystyle {\sum }_{-\infty }^{\infty }{(-i)}^n{C}_n{H}_n\left\{\frac{J_1(\beta r^{\prime })}{\beta r^{\prime }}\right\}}\exp (in\phi )$$

(31)

where $C_n$ is the coefficient, $H_n$ is the nth Hankel function, and ϕ is the angular coordinate on the Li-Au optical aperture. This distribution is used to analyze the effect of the phase plates on starlight and to design phase plates to achieve extinction effects.

During the design phase, the complex amplitude distribution guided the determination of the size and position of the aperture stop and the Li-de-Rauw stop, ensuring that stellar light is effectively blocked, and optimized the parameter selection of the flat-top sine phase plate to enhance the detection sensitivity for exoplanets. In terms of performance analysis, the double-amplitude distribution is used to evaluate the imaging contrast of the coronagraph by quantifying the degree of suppression of stellar light before the Li-Au aperture, thereby assessing the coronagraph’s ability to detect exoplanets. During system optimization, further improvements to the coronagraph’s achromatic performance and imaging quality can be achieved by adjusting the phase plate parameters and optical component layout. In astronomical observations, the analysis of the double-amplitude distribution ensures that the coronagraph can effectively suppress starlight, directly capture images of exoplanets, and provide accurate spectral information, thereby completing the detection and study of exoplanets.

Pupil Apodization Coronagraphs

Next, we introduce the Pupil Apodization Coronagraphs. The Pupil Apodization Coronagraphs achieves the desired effect by altering the complex amplitude distribution of incident light at the pupil, creating an artificial “solar eclipse” effect at the final imaging plane. This principle can be explained using Fourier optical imaging theory, where pupillary truncation for high-contrast imaging has been used for over fifty years [66]. To distinguish it from the phase-induced amplitude truncation coronagraph proposed later, this coronagraph is typically referred to as the Classical Pupil Apodization (CPA) [60].

Although external-occultation coronagraphs, Lyot coronagraphs, and CPA all suppress stellar light by creating an artificial “solar eclipse” effect to observe faint coronas or exoplanets, they differ significantly in their working principles, system structures, and technical performance. As summarized in

Table 3. A comparative analysis of external-occultation coronagraphs, Lyot coronagraphs, and CPA from three aspects: working position, core objectives, and typical application scenarios.

Typology	Working Position	Core Objective	Typical Application Scenarios
External-occultation Coronagraph	Positioned thousands to tens of thousands of kilometers in front of the space telescope.	Suppression of direct sunlight and observation of the low-altitude corona	Space solar observation
Lyot Coronagraph	Fully integrated within the telescope’s internal optical system.	Blocking stellar light and suppressing diffracted stray light	Ground-based planetary observations (e.g., Io plasma rings)
CPA Coronagraph	Primary modulation is applied at the telescope’s pupil plane (typically the exit pupil).	Modulation of optical pupil amplitude distribution for broadband imaging	Space broadband astronomical observations

, the fundamental distinction between external-occultation, Lyot, and CPA coronagraphs lies in their working position, core objective, and resulting application domain.

External occultation achieves near-sun observations through geometric occlusion and multi-stage diffraction suppression, making it suitable for space-based solar physics missions. The Lyot coronagraph employs a focal plane mask and conjugate aperture to suppress diffraction, balancing efficiency and resolution, and is suitable for ground-based planetary observations. CPA simplifies the structure through pupil-amplitude modulation, but sacrifices efficiency, making it suitable for wide-band space observations. CPA is essentially a pupil plane modulation technique, while external-masking and Lyot coronagraphs belong to optical path structure design. These two are classified in different dimensions. CPA can be combined with Lyot, and external masking can introduce the Lyot aperture to enhance suppression capability. Combining Phase-Induced Amplitude Apodization (PIAA) with the Lyot mask forms PIAACMC (Phase-Induced Amplitude Apodization Complex Mask Coronagraph) [13], addressing complex pupil-obstruction issues. The integration of CPA and PIAA enhances the efficiency, in addition to other improvements. The core objectives are lowering the internal working angle and achieving higher contrast and broader bandwidth adaptability [67].

The CPA coronagraph modulates the complex amplitude distribution of the optical field at the pupil by establishing a transmittance or phase plate on the pupil surface (generally set at the exit pupil) of the system. Kasdin in 2003 explained the principle of CPA [68] and proposed different pupil shapes (toe-cutting function) as

$$A(\mathrm{r})={\left[1-{({\displaystyle \frac{r}{R}})}^2\right]}^{\mu } (\mu \ge 1)$$

(32)

Here, A(r) is the amplitude modulation function that controls the edge transmittance gradient of the pupil by adjusting the exponent $\mu$ (used to attenuate the pupil edge); r is the radial distance from the center of the pupil to the edge and R is the radius of the pupil.

Figure 10 shows the structural diagram of a common CPA—Apodized Pupil Lyot Coronagraph (APLC).

CPA has many advantages such as its simple structure, reliability, wide operating spectral range and error sensitivity. For example, for the SPICA-CPA mission, the Japan Space Infrared Telescope adopted a circular cut toe with a luminous flux efficiency of only 20%, but with a working bandwidth of 25%, which was suitable for wide-band detection. The Nanjing Institute of Sky and Light, Chinese Academy of Sciences, developed discrete cut-toe panels to simplify continuous transmittance in five steps and increase the through-light efficiency to 35% (2020 experiment). However, in principle, the CPA-type coronagraph has two obvious shortcomings: one is due to the introduction of transmittance plates in the exit pupil; most of the incident light is blocked, resulting in a significant reduction in the through-light efficiency. For instance, the, through-light efficiency of SPICA-CPA is only ~20%, which increases the exposure time exponentially, which is extremely unfavorable. The second is due to the blocking of the larger system-equivalent aperture Due to the large obstruction, the equivalent aperture of the system is reduced significantly. For instance, the equivalent aperture of SPICA-CPA is only 60% of the actual value, and the inner working angle is >4 λ/D, which is not conducive to the detection of exoplanets with close angular distances.

Phase-Induced Amplitude Apodization Coronagraph

To improve the throughput efficiency and inner working angle of the coronagraph and realize higher contrast exoplanet imaging, in 2003, Guyon from the Subaru Telescope research team at the National Astronomical Observatory of Japan drew inspiration from the non-spherical mirror array method commonly used in laser beam transformation and proposed a novel pupilling method, which he applied to a coronagraph. This method achieves pupilling by using two non-spherical mirrors to alter the light field distribution of the pupil. This coronagraph also belongs to the category of pupil-apodizing coronagraphs and is referred to as the Phase-induced Amplitude Apodization Coronagraph (PIAA-type coronagraph) [69].

The surface type function of two mirrors in PIAA is satisfied by the two mirrors:

$${\displaystyle \frac{dz}{dr}}=-{\displaystyle \frac{\lambda }{2\pi }}{\displaystyle \frac{d\varphi }{dr}}$$

(33)

where λ denotes the wavelength, ϕ(r) is the target phase distribution, and dr denotes the amount of variation in the radial distance. The optical range difference is controlled by the mirror gradient dz/dr. This design has the advantages of a wide working spectral band, high throughput (theoretically, close to 100% throughput can be realized), and a small inner working angle (2 λ/D, as verified by the Subaru Telescope in 2005) [70]. While the advantages of the CPA coronagraph lead to a wide working spectral band, PIAA leverages the advantages of CPA corona limiters, such as a wide working spectrum, and overcomes the main disadvantages of CPA corona limiters, such as low throughput and large inner working angle.

A typical PIAA-type coronagraph is of the form shown in Figure 11. The central field of view of the coronagraph is aligned with the position of the primary star. The image formed on the focal plane of the telescope’s primary optical system is collimated into parallel light by a collimating mirror, then enters the PIAA system composed of two aspherical mirrors to achieve pupillary amplitude cutoffs. The emerging light is non-uniformly distributed parallel light, which is imaged by the imaging mirror I1 onto the first image plane F1. A stop plate is placed at the center of the first image plane to block the main star image at that position. The remaining light is collimated again to form parallel light and passes through a PIAA system with parameters identical to the preceding PIAA system but configured in reverse, restoring the light to uniformly distributed light. The light is then imaged by imaging mirror I2 at the second image plane F2 to obtain the planetary image. For simplicity, the diagram uses a transmissive system. However, considering chromatic aberration issues, most existing PIAA systems are designed as reflective systems. Therefore, the two cut-off mirrors are labeled as M1 and M2 in the diagram.

In 2010, Guyon proposed combining the PIAA and Lyot coronagraphs to form the PIAACMC [71]. PIAACMC improves PIAA in that it is suitable for complex pupil shapes. Subsequently, institutions such as the Ames Research Center conducted experimental studies on the PIAACMC for obstructed telescopes, targeting its application in the U.S. next-generation space astronomy project WFIRST. In February 2018, active optical control using a single deformable mirror was employed, and the incident light intensity distribution was simulated according to the WFIRST pupil obstruction configuration. At 650 nm monochromatic light incidence, the contrast ratio achieved was 2.6 × 10⁻⁸, and for 10% bandwidth broadband light, the contrast ratio reached 1.8 × 10⁻⁷ [72]. As chronicled in

Table 4. The developmental history of PIAA.

Timing	Milestone	Teams	Key Parameters
2003	Development of the PIAA concept	Subaru Telescope, National Astronomical Observatory of Japan	Theoretical throughput > 95% (A conceptual breakthrough)
2005	First experimental system validation	Subaru Observatory	Contrast ratio 10−6 (Visible light; initial on-sky validation)
2009	Reflective PIAA + Anamorphic Mirror Correction	NASA Ames	Contrast ratio 5 × 10−9 (650 nm monochromatic light; approaching fundamental limits in the lab)
2013	Vacuum environment polarization optimization	JPL	Contrast ratio 5 × 10−10 (Monochromatic light; represents the ultimate performance under ideal, narrowband conditions)
2018	PIAACMC adapted to WFIRST blocking pupil	NASA Goddard	Contrast ratio 1.8 × 10−7 (Achieved under 10% broadband light and a complex, obstructed pupil, demonstrating robustness for a real space mission)
2024	Liquid Crystal SLM Dynamic PIAA (lab phase)	Nanjing Institute of Astronomical Optics & Technology, National Astronomical Observatories, CAS	Contrast ratio 10−6 (4-12 λ/D under stitched mirrors; a new, active approach in development)

, the development of the PIAA coronagraph demonstrates a consistent trajectory of experimental validation and performance enhancement across multiple institutions.

The Fourier transform form of the optical pupil function is given by

$$PSF(x,y)=\left|{\displaystyle \iint P(x^{\prime },y^{\prime })}{e}^{2\pi i(x^{\prime }x+y^{\prime }y)}dxdy\right|$$

(34)

where PSF(x,y) denotes the PSF, which is the Fourier transform of the pupil function P(x′,y′) used to describe the distribution of light in the imaging plane. P(x′,y′) denotes the pupil function, i.e., the distribution of the light field in the plane of the pupil. $e^{2\pi i(x^{\prime }x+y^{\prime }y)}$ is the complex exponential function, which denotes the phase factor where i is an imaginary unit. dxdy denotes the integrals in the x- and y-directions, respectively.

Generally, a telescope uses a circular optical pupil whose PSF can be described as

$$PSF(r)=I_0\cdot \left[{\displaystyle \frac{2J_1(\frac{kDr}{2f})}{kDr/2f}}\right]$$

(35)

where PSF(r) denotes the PSF of a circular optical pupil. $I_0$ denotes the maximum central intensity. $J_1$ is a first-order Bessel function (Bessel function of the first kind). k is the wave number, generally defined as k = 2π/λ. where D denotes the telescope aperture. and f denotes the focal length of the lens. Moreover, r is the radial distance from the optical axis to a point on the image plane.

As compared in

Table 5. Comparison of core characteristics of three types of Apodized Pupil Lyot Coronagraph.

Norm	CPA	PIAA	PIAACMC
Throughput	15–30%	>90%	70–85%
Inner working angle (λ/D)	4–6	2–3	1.5–2.5
Operating bandwidth	Wide (>20%)	Medium (10–15%)	Wide (>20%)
processing difficulty	Low (coated mask)	Media (aspheric)	High (hybrid devices)
Typical tasks	SPICA	Subaru SCExAO	WFIRST/CGI

, the three apodized pupil Lyot coronagraph architectures exhibit distinct trade-offs between throughput, inner working angle, and operational bandwidth, which dictate their suitability for different observational tasks.

The latest development in coronagraphs is the Roman Space Telescope’s Coronagraph Instrument (CGI), demonstrating the maturity of coronagraph technology for space deployment. Integration and testing have yielded significant results, particularly regarding environmental stability—a critical factor for achieving and maintaining high contrast [73]. During thermal vacuum (TVAC) testing, the CGI optical bench demonstrated exceptional thermal stability, with drifts < 10 mK over 10 h, crucial for detecting planets in habitable zones. Dynamics testing revealed resilience to unexpected 11 g vertical accelerations encountered during transport scenarios. Finite-element modeling confirmed the design withstands 3× flight loads, with structural amplification factors kept below 12.8× at resonances. Innovative approaches, such as using spliced PRT sensors with copper Neptape shielding, reduced integration time by 30%. Multi-layer insulation (MLI) blanketing achieved a low heat leak of <0.15 W/m² in TVAC, essential for maintaining the cryogenic operating environment.

2.2.3. Distinction Between Space-Borne and Ground-Based Implementations in Coronagraph

Building upon the comparative analysis of various coronagraph architectures and their performance metrics, the choice between ground-based and space-based deployment critically influences their design philosophy, achievable contrast, and scientific objectives.

Coronagraph technology manifests distinctly in space-borne and ground-based environments, primarily due to the presence or absence of atmospheric turbulence. Space-borne coronagraphs operate under pristine conditions, enabling them to theoretically approach their fundamental photon noise limits and achieve the deepest contrasts (10⁻¹⁰ to 10⁻¹²) necessary for imaging Earth-like planets. They can utilize techniques that are impractical on the ground, such as large external occultors (starshades), which require precise formation flying over thousands of kilometers but offer inner working angles significantly inside the diffraction limit. Internal coronagraph concepts like the Phase-Induced Amplitude Apodization Complex Mask Coronagraph (PIAACMC), extensively tested for missions like WFIRST (now Roman Space Telescope) and envisioned for HabEx or LUVOIR (Large UV Optical Infrared telescope) [74], are optimized for the stable space environment and complex obscured apertures of space telescopes. The primary challenge in space is achieving and maintaining the extreme wavefront stability (ΔϕNCPA ≪ λ/100 RMS) required over long durations.

Ground-based coronagraphs, while benefiting from larger aperture telescopes, must contend with atmospheric turbulence. They are thus intrinsically coupled with Extreme Adaptive Optics (ExAO) systems (Section 2.3) to correct the wavefront before the coronagraph. Ground-based instruments primarily employ internal coronagraphs like the Vortex (e.g., on Keck, VLT-SPHERE), Four-Quadrant Phase Mask (4QPM) (e.g., on Subaru/SCExAO), and optimized Lyot-type coronagraphs (e.g., APLC on VLT-SPHERE). These are designed to be robust against residual turbulence and optical imperfections. While achieving impressive contrasts (10⁻⁶ to 10⁻⁷) at small angular separations (<0.5″), they are ultimately limited by atmospheric coherence time, residual speckle noise, and the need for very high-order wavefront correction. Ground-based facilities like Subaru/SCExAO and VLT-SPHERE serve as critical testbeds for developing and validating coronagraph technologies that will later be intended for space missions [75]. Concepts like the ground-based GENIE nulling coronagraph for the VLTI represented a specific approach combining interferometry and coronagraphy on a ground-based platform.

2.3. ExAO

ExAO (extreme adaptive optics) is an advanced adaptive optics technology centered on sub-nanometer wavefront control. In high-contrast imaging technology, ExAO is not only an enabling technology for coronagraphs and nulling interferometry, but also an independent module that determines the system’s ultimate performance; it resolves the core contradiction that the former two cannot address: dynamic disturbance noise (atmospheric turbulence) and static system errors (alignment residuals, thermal deformation). Although star coronagraphs and null interference techniques reach theoretical contrast ratios to the 10⁻¹⁰ through physical optical path design, the actual imaging performance remains constrained below 10⁻⁶ due to wavefront errors. This contradiction reveals the third dimension of high-contrast imaging: dynamic wavefront control. ExAO achieves this by establishing a real-time closed-loop system that integrates “sensing-computing-execution,” enabling real-time correction of atmospheric turbulence (with frequencies up to thousands of Hz) and system static aberrations (such as non-common-path errors NCPA). This compresses the residual wavefront errors to the sub-nanometer level, independently assuming the critical role of overcoming the measured performance limits of the system. Therefore, ExAO together with optical path design (coronagraph) and coherent control (nulling interferometry) forms a technical triangle [76,77], as shown in Figure 12.

As shown in Figure 13, light first enters the deformable mirror (DM) of an extreme adaptive optics system, where it undergoes a preliminary correction of low-order wavefront aberrations (such as tilt and defocus) caused by atmospheric turbulence. The light is then split into two paths by a beam splitter: one path (scientific light path) enters the subsequent optical system for scientific observation, and the other beam (wavefront sensing path) is projected onto a microlens array where the focusing offset of the light spot on the array is measured to quantify the slope information of the wavefront aberrations. The wavefront aberration data are transmitted in real time to a wavefront sensor for wavefront reconstruction. For example, in the pyramid sensor, the beam passes through a pyramid prism to form a four-quadrant diffraction pattern. By measuring the deformation of the pattern, the complete wavefront phase distribution, including higher-order aberrations (such as coma and spherical aberration), can be reconstructed with high precision. The reconstructed wavefront information is sent to the real-time control system (RTC), which combines control algorithms (such as the minimum mean square error method) to generate correction commands. Dynamic closed-loop correction is then performed, with the DM adjusting its surface morphology in real time via thousands of microactuators according to the RTC commands to compensate for wavefront aberrations. This process is iterated at frequencies ranging from hundreds of Hz to kilohertz, forming a closed-loop control system that enhances the Strehl Ratio (a measure of imaging quality) to levels approaching the diffraction limit.

To address diffraction effects caused by the primary mirror support structure, the ExAO system employs support frame diffraction suppression technology. By precisely controlling the tilt angle of each reflector plate, the four parts of the light pupil are displaced along specific directions, thereby “filling” the areas obstructed by the support frame and significantly reducing the impact of diffraction on the image quality. This preprocessing step eliminates the need for additional diffraction correction in the subsequent coronagraph, allowing it to focus on suppressing the intense light from stars and improving the contrast for planetary detection. The corrected light path enters the EMCCD camera, which captures rapidly changing celestial phenomena through high-speed continuous imaging (frame rate > 1 kHz). Combined with the wavefront correction capability of the ExAO system, high-resolution imaging approaching the diffraction limit is achieved in the visible light band, with contrast reaching the 10⁻⁷ level, providing critical technical support for cutting-edge research such as direct exoplanet imaging [78].

In ground-based astronomical observations, traditional adaptive optics is primarily a technical application that corrects the effects of atmospheric turbulence. The ExAO system, however, is a scientific observation application that leverages high frame rates and high resolution adaptive optics to achieve high-contrast imaging. The light intensity contrast between exoplanets and their host stars is extremely low, typically ranging from 10⁻⁶ to 10⁻¹⁰ [79,80]. Therefore, compared to conventional AO systems, ExAO systems require higher performance in multiple dimensions as well as additional capabilities that are not typically available in conventional AO systems. The specific differences are summarized in the

Table 6. Comparative analysis of ExAO and AO technologies.

Comparison Term	AO Technology	ExAO Technology
Wavefront Correction Accuracy	Achievement of a certain Strehl ratio (usually higher than 0.3–0.5)	Higher wavefront correction accuracy is required, typically to achieve Strehl ratios close to 1 (>0.9)
Calibration Speed	Usually between a few hundred Hz and a thousand Hz.	Higher calibration speeds are required (typically over a thousand Hz)
Correction Order	Typically lower order DMs are used, such as tens to hundreds of actuators	Requires higher-order morphing mirrors, often containing thousands of actuators
Contrast Performance	Limited contrast performance, often difficult to achieve contrast ratios of 10−6 or higher	Requires higher contrast performance, typically 10−7 to 10−10 contrast ratio is required
Wavefront Sensor	Possible use of Shack-Hartmann wavefront sensors, etc., but limited order and accuracy	Need for higher order wavefront sensors such as pyramidal wavefront sensors
Wavefront Control Algorithm	Using basic wavefront correction algorithms such as integral controllers	Need for more advanced wavefront control algorithms such as model predictive control or optimal control algorithms
Environmental Stability	Typically operated on ground-based telescopes, which need to cope with atmospheric turbulence, but the requirements for environmental stability are less stringent than for the ExAO system	Requires extreme thermal and mechanical stability control of the optical system
Photonic Noise Suppression	Although photon noise is also considered, it is not a major limiting factor in high contrast imaging	Special attention needs to be paid to the suppression of photonic noise
Coronagraphs and Post-Processing Techniques	Possible use of basic coronagraphs to suppress stellar light, but limited post-processing techniques	The need for more advanced coronagraph designs, such as vector vortex coronagraphs, as well as sophisticated post-processing techniques, such as differential imaging and angular differential imaging

below:

2.3.1. Concept Proposal and Early Exploration: Angel, LLNL and UCSC

The concept of the ExAO technique was first proposed by Angel (1994) [81], who recognized that direct imaging of exoplanets using ground-based telescopes is possible when the following conditions are met:

$$\mathrm{\Delta }\theta \le 1.22{\displaystyle \frac{\lambda }{D}}\cdot k_{atm}, C\ge {10}^{-9}$$

(36)

where Δθ is the angular resolution, indicating the smallest angle that the telescope can resolve; λ denotes the wavelength of light; D denotes the aperture of the telescope, i.e., the diameter of the telescope; katm is the atmospheric coherence length scaling factor; and C is the imaging contrast, which denotes the ratio of the brightness of the target to the background in the image, and is required to be at least 10⁻⁹ to meet the requirements for detecting Jovian planets.

In the early stages of the ExAO technology, researchers primarily focused on theoretical research and conceptual designs. They explored methods to achieve higher wavefront correction capabilities by increasing the number of DM drivers and improving the precision of the control algorithms. Subsequently, the key metrics of ExAO technology were identified, such as high-driver-count deformable mirrors, advanced wavefront sensors, and high-speed control algorithms. These metrics were primarily focused on in subsequent ExAO system design and optimization.

In the early 2000s, the Lawrence Livermore National Laboratory (LLNL) and University of California, Santa Cruz focused on theoretical research, conceptual design, and development of key components for ExAO technology. In the early stages of ExAO technology research, LLNL leveraged legacy equipment from its extreme ultraviolet lithography work, such as a phase-shift diffraction interferometer with 100-pm absolute precision, enabling the ExAO test platform to achieve sub-nanometer resolution measurement capabilities. This high-precision measurement instrument allowed researchers to accurately measure the surfaces of MEMS (microelectromechanical system) devices [82]. Subsequently, LLNL constructed a sodium guide star adaptive optics system prototype at the Lick Observatory of the University of California, successfully demonstrating sodium guide star adaptive optics correction. The system was installed on a 3 m Shane telescope at the Lick Observatory, based on a 127-driver continuous surface deformation mirror, Hartmann wavefront sensor equipped with a fast imaging low-noise CCD camera, and pulsed solid-state pumped dye laser tuned to the atomic sodium resonance line (589 nm). This enables the telescope to achieve K-band Streiner > 0.6 in the presence of atmospheric turbulence, with contrast enhanced to 10⁻⁶. The formula for wavefront reconstruction using a Hartmann wavefront sensor is

$$({\nabla }^2\varphi )(x)=\nabla \cdot s(x)$$

(37)

where ϕ is the wavefront phase, s(x) is the wavefront slope measured using the Hartmann sensor, x is the position of the aperture, ∇ denotes the gradient operator used to calculate the spatial gradient of the wavefront phase, and ∇² denotes the Laplace operator used to calculate the second-order spatial derivative of the wavefront phase, i.e., the wavefront curvature. This equation is solved in a wavefront reconstruction computer, and the results are then sent to the drive electronics of the DM.

Additionally, LLNL utilized an existing laser developed for the U.S. Department of Energy’s Atomic Vapor Laser Isotope Separation program to conduct feasibility experiments on sodium layer laser-guide stars. These experiments demonstrated that sodium layer laser-guided stars have the potential to enable adaptive optical systems to perform high-order wavefront corrections for atmospheric turbulence [83].

2.3.2. Algorithmic Breakthrough: Fast Wavefront Reconstruction

Following the validation of the engineering feasibility of extreme adaptive optics (ExAO) by the LLNL and UCSC research groups using a high-precision phase-shift interferometer and sodium-guide adaptive-optics system, achieving the rapid resolution of large-scale phase errors while maintaining correction accuracy has become a key challenge limiting the further development of adaptive-optics technology. In 2002, Poyneer, Gavel, and Brase proposed a fast wavefront-reconstruction algorithm using the Fourier transform, providing efficient wavefront-correction capabilities for large AO systems [84]. When discussing the frequency response of deformable mirrors (DMs), they derived a transfer-function formula for an ideal low-pass filter:

• Impulse response:

$$h_I(x,y)={\displaystyle \frac{\sin (\frac{\pi x}{d})}{\frac{\pi x}{d}}}{\displaystyle \frac{\sin (\frac{\pi y}{d})}{\frac{\pi y}{d}}}$$

(38)

• Transfer function (math.):

$$H_I(f_{x,}{f}_y)=\left\{\begin{array}{l}{d}^2 \left|f_x\right|<{(2d)}^{-1}, \left|f_y\right|<{(2d)}^{-1}\\ 0 else\end{array}\right.$$

(39)

An ideal low-pass filter is an important concept in signal processing that allows signals below a cut-off frequency to pass while blocking signals above that frequency. When the absolute value of the spatial frequency (f_x, f_y) is less than ${(2d)}^{-1}$, the value of the transfer function is $d^2$. Therefore, the frequency component can pass completely; otherwise, the value of the transfer function is zero; thus, the frequency component is completely blocked. The impulse response function $h_I(x,y)$ is a representation of the transfer function in the spatial domain that characterizes the response of the filter to the input signal.

2.3.3. Planet Imager Concept Design: XAOPI

The Fourier-transform wavefront-reconstruction algorithm proposed by Poyneer et al. provides an efficient computational framework for large AO systems [84]. After addressing the real-time bottleneck in correcting atmospheric turbulence-induced phase errors, astronomical research shifted its technical focus toward achieving higher-contrast direct imaging of exoplanets. Direct observation of young, self-luminous giant planets is considered a key method for studying planetary-formation processes. However, the variability of stars interferes with observations, and imaging young exoplanets using indirect methods is challenging. Therefore, astronomers require a technique that can directly image exoplanets to observe their features and orbits more clearly. To address this, the Macintosh team proposed the concept design of the eXtreme Adaptive Optics Planet Imager (XAOPI) in 2003, aiming to achieve near-diffraction-limited imaging on 8–10 m-class telescopes and directly detect and characterize Jupiter-like planets orbiting nearby young stars [85]. Its three major innovations include the following. (1) A 4096-element microelectromechanical system (MEMS) DM was developed to support the correction of over 200 Zernike modes, with a surface accuracy of λ/200 root mean square (RMS) (λ = 1 μm). (2) A spatial filter Shack–Hartmann wavefront sensor was developed, in which a spatial filter with a width of Δf = λ/D is placed in front of the sensor (Δf denotes the bandwidth of the spatial filter, i.e., the frequency range allowed to pass through the filter, λ denotes the wavelength, and d is the subaperture size). The size of the spatial filter was set to allow only spatial frequencies that were Nyquist-sampled by the subapertures of the wavefront sensor to pass through, thereby suppressing sampling aliasing errors. This achieved raw contrast ratios of 10⁻⁶–10⁻⁷ at angular separations of 0.2–0.8 arcs. (3) High-contrast point-spread function (PSF) control was developed. By optimizing the DM to generate clearly defined “dark aperture” regions, the signal-to-noise ratio (SNR) for planetary detection was enhanced. Dark aperture engineering deliberately shapes the point spread function (PSF) using deformable mirrors to create a low-intensity region (contrast < 10⁻⁷) at separations where planets may reside. This requires suppressing both diffraction (via coronagraphs) and speckles (via ExAO) within a predefined angular sector. The PSF intensity is calculated as follows:

$$I(\theta )={\displaystyle {\sum }_{i=1}^n I_i(\theta ){\sigma }_i^2 }$$

(40)

where $I_{\mathrm{i}}(\theta )={\displaystyle \frac{〈{\Phi }_i(\theta )〉}{{\lambda }^2}}$. This formula is used to calculate the total intensity of the PSF. $I_i\xi \theta )$ is the contribution of the ith wavefront error source to the PSF intensity, ${\sigma }_i$ is the magnitude of the phase error of the ith error source in radians, $〈{\Phi }_{\mathrm{i}}(\theta )〉$ is the spatial power spectrum of the phase error, and λ is the optical wavelength. This equation shows that the total PSF intensity is a superposition of individual error sources [86].

Technology Radiation at XAOPI:

The Gemini Planet Imager (GPI) directly realized XAOPI’s vision as the first dedicated ExAO coronagraphic instrument on an 8 m telescope. Deployed at Gemini-South in 2014, GPI combined a 4096-actuator MEMS DM, an advanced integral field spectrograph, and an apodized-pupil Lyot coronagraph [87]. It achieved groundbreaking H-band contrasts of 10⁻⁷ at 0.2 arcs, enabling the first direct spectral characterization of young, self-luminous exoplanets like 51 Eri b. GPI validated XAOPI’s core innovations: spatial-filtered wavefront sensing reduced aliasing errors, while its ‘dark hole’ PSF control demonstrated model-independent planet detection. This success spurred analogous instruments (VLT-SPHERE, Subaru-SCExAO), proving the feasibility of extreme AO for exoplanet characterization [88]. VLT-SPHERE employs a comparable spatial-filtering technique with an IF error reduced to 3 nm RMS. For the PIAAC system (2020) at the Nanjing Institute of Astrophotography, the spatial filter was replaced with a digital micromirror array (DMD) programmable filter to make the bandwidth Δf dynamically adjustable. A liquid-crystal spatial light modulator was used to replace the MEMS DM to realize an equivalent 1000-cell correction.

XAOPI realizes near diffraction-limited imaging on 8–10 m class telescopes through high-order Zernike-mode correction and “dark hole” PSF modulation. However, when the number of DM units exceeds a thousand, the complexity of traditional wavefront-reconstruction algorithms reaches O(N3), and the demanding real-time and accuracy requirements of the ExAO system become difficult to satisfy. Its reconstruction error $\epsilon$ satisfies $\epsilon \propto \kappa (D)\cdot {\sigma }_{noise}$, where κ(D) is the condition number of the reconstruction matrix D (usually >10³) and σ_noise is the detection noise [85,86].

2.3.4. Efficient Large-Scale Correction: MGCG Algorithm (Gilles)

To improve the efficiency and robustness of wavefront-reconstruction algorithms, in 2003, Gilles proposed conjugate gradient (MGCG) algorithm based on sparse approximation and multigrid acceleration for the wavefront reconstruction of ExAO systems [15]. The core of the algorithm consists of (1) a sparse approximation to reconstruct the matrix, preserve the main diagonal and neighboring nonzero elements of D, and reduce the storage requirement from O(N2) to O(N). (2) A pre-conditioned conjugate gradient iteration is included, in which the wavefront phase $\overrightarrow{\varphi }$ is solved by a two-step iteration:

$${\overrightarrow{\varphi }}_{k+1}=\overrightarrow{{\varphi }_k}+{\alpha }_k\overrightarrow{p_k}, \overrightarrow{p_k}=M^{-1}\overrightarrow{r_k}$$

(41)

where ${\varphi }_{k+1}$ and ${\varphi }_k$ denote the wavefront-phase vectors at the k + 1st and kth steps of the iteration, respectively. ${\alpha }_k$ is the step factor of the kth step, which is used to adjust the update amplitude for each iteration. $p_k$ is the search-direction vector for the kth step. m is the multigrid preconditioner used to accelerate the convergence of the conjugate gradient method, and $\overrightarrow{r_k}$ is the residual vector. (3) Multigrid low-frequency error suppression is applied with the corrected solution on the coarse grid

$$\delta \overrightarrow{\varphi }:D_c\delta \overrightarrow{\varphi }=R\overrightarrow{r_f}$$

(42)

where D_c is the coarse-grid operator, $\delta \varphi$ denotes the correction quantity for the phase of the wavefront, r is the restriction operator, and r_f is the residual vector that denotes the difference between the current and desired solutions.

2.3.5. Non-Common Path Error Correction: Differential Wavefront Sensing

The continuous improvement of the ExAO system performance has pushed wavefront-correction accuracy to new heights (Strehl ratio > 0.95 in the H band). However, it has also highlighted the bottleneck of the noncommon-path error (NCPA) between scientific cameras and wavefront sensors $\mathrm{\Delta }{\varphi }_{NCPA}$, which can cause a contrast loss of 1–2 orders of magnitude in coronagraphs. To meet the calibration requirement of $\mathrm{\Delta }{\varphi }_{NCPA}$ < λ/100 (RMS), Wallace et al. proposed a differential dual-wavefront sensing architecture in 2004, suggesting the use of a precision wavefront sensor (PWFS) to calibrate and maintain the wavefront of the scientific camera [89]. They explored various calibration architectures, including placing the PWFS before and after the coronagraph, and proposed a hybrid technique (combining pre-coronagraph and post-coronagraph methods) to measure and correct wavefront errors, i.e., the pre-coronagraph PWFS measures the incident wavefront ${\varphi }_{in}$, and the post-coronagraph PWFS monitors the output wavefront ${\varphi }_{out}$. The two are related by the systematic transfer matrix T: $\mathrm{\Delta }{\varphi }_{NCPA}=T^{-1}({\varphi }_{in}-{\varphi }_{out})$. The model guides the DM to compensate $\mathrm{\Delta }{\varphi }_{NCPA}$ in real time to achieve a closed-loop correction, improving the Strehl ratio of the scientific camera from 0.3–0.5 to >0.9.

2.3.6. System Integration Benchmark: Subaru SCExAO

Wavefront distortion caused by atmospheric turbulence severely affects the imaging quality of 8–10 m telescopes, becoming a key factor that limits the improvement of astronomical-observation resolution. To address this challenge, the Subaru Coronagraphic Extreme Adaptive Optics (SCExAO) project was launched with the aim of developing an ExAO system suitable for 8 m telescopes [90]. The SCExAO system is located between the Subaru telescope AO188 (a curvature adaptive-optics system with 188 drivers) and HiCIAO scientific camera, which consists of a phase-induced amplitude aperture (PIAA) coronagraph or an eight-quadrant phase mask (EOPM) coronagraph, an advanced wavefront-control system using a 1024-driver MEMS DM, and a coronagraph low-order wavefront sensor (CLOWFS) [91,92].

In the EOPM coronagraph, reflective masks are used instead of traditional opaque Lyot baffles, enabling diffracted starlight outside the aperture to be utilized by the CLOWFS for precise measurement and the correction of tilt errors. When starlight passes through the EOPM and diffracts outside the aperture, the CLOWFS captures these light rays and analyzes their changes to infer the tilt of the wavefront. The system then adjusts the DM based on this information to correct wavefront aberrations, thereby improving imaging quality. By designing a spider removal plate and an improved PIAA (MPIAA) optical system [92], the original blocked pupil with spider support structures is converted into a continuous circular pupil, eliminating high-frequency diffraction noise. The combination of the EOPM and MPIAA reduces the working angle inside the coronagraph to 0.1 λ/D, improving contrast by one order of magnitude. The CLOWFS ensures that residual low-order aberrations are below λ/200 RMS, significantly reducing the dynamic-range requirements for high-contrast imaging systems. The MPIAA prism angle $\theta$ is calculated as follows:

$$\theta ={\sin }^{-1}\left({\displaystyle \frac{n_0\sin \gamma }{\sqrt{n_{opt}^2+n_0^2-2n_{opt}{n}_0\cos \gamma }}}\right)$$

(43)

where $n_0$ is the refractive index of air, $n_{opt}$ is the refractive index of the MPIAA plate material at the optimized wavelength, and $\gamma$ is the angle defined by the geometry of the MPIAA plate, calculated as γ = arctan(aR⁻¹s), where s is the center-to-center spacing of the MPIAA plate, and aR is the ratio of the radius of the secondary mirror shading to the radius of the primary mirror [92].

2.3.7. Wavefront Sensing Innovation: APF-WFS (Martinache)

The SCExAO project developed an innovative extreme AO system by integrating a coronagraph, high-order MEMS DMs, and CLOWFSs, achieving multilevel correction of wavefront aberrations on the Subaru telescope platform. Building on this foundation, emerging demands, such as the direct detection of exoplanets, have imposed stricter requirements on the Strehl ratio, NCPA suppression, and dispersion control. In 2013, Martinache proposed the application of asymmetric pupil Fourier wavefront sensor (APF-WFS) technology [93], which directly senses pupil phase errors by analyzing AO-corrected images, providing a powerful tool for correcting NCPAs in ExAO systems. This effectively reduces system errors and improves imaging quality. By optimizing optical design and incorporating dispersion-compensation elements, the impact of dispersion on imaging quality was reduced, further enhancing system performance. The implementation of these technical solutions has significantly improved the performance of ExAO systems. After APF-WFS correction, the Strehl ratio of the PSF increased by approximately 5%, improving the capability for the direct imaging of exoplanets [90,93]. The APF-WFS method relies on the analysis of a PSF that has been corrected by an asymmetric hard-stop mask, and the AO method is used to correct the PSF. The APF-WFS method relies on analyzing the Fourier properties of an asymmetric hard-stop-masked AO-corrected image to directly perceive the pupil phase aberration. For small aberrations, the Fourier phase ($\Phi$) is linearly correlated with the pupil phase ($\phi$) using a unique operator T, which depends on the aperture geometry and illumination:

$$\Phi ={\Phi }_0+T\times \phi$$

(44)

where ${\Phi }_0$ denotes the Fourier phase pointing toward the target during observation. For high-contrast imaging, the target, that is, a bright star surrounded by high-contrast structures, can be considered as a point source; thus, ${\Phi }_0$ can be considered equal to 0.

This Fourier phase relationship can be reversed by introducing an asymmetry in the pupil. Thus, a direct focal-plane image containing only a small amount of additional diffraction from the pupil asymmetry can be used as a wavefront sensor:

$$\phi =T^{-1}\times \Phi$$

(45)

where T⁻¹ denotes the inverse matrix of the operator T for converting the Fourier phase Φ back to the pupil phase $\phi$.

2.3.8. Advanced Detectors and Multi-Band Capability: MKID-PWFS

During the advanced detector and multiband-detection phases (the mid-to-late 2010s to 2020s), advancements in detector technology, particularly the emergence of microkinetic energy inductive detectors (MKIDs), have brought new developments to ExAO systems. However, traditional detectors are limited in terms of sensitivity, temporal resolution, and wavelength coverage; thus, the high-precision, wide-band detection requirements of ExAO systems cannot be met. Therefore, the application of MKIDs and multiband wavefront reconstruction emerged. MKIDs are applied to pyramid wavefront sensors, leveraging their high sensitivity and high time resolution to achieve multiband detection and enhance the ability of the system to correct for light of different wavelengths [94]. ExAO leverages the energy resolution of MKIDs to achieve multiband wavefront reconstruction, further improving the correction accuracy and imaging quality of the system. The implementation of these technical solutions has driven the development of ExAO systems with higher sensitivities and broader wavelength coverages, providing a technological foundation for future ExAO systems in large-aperture telescopes.

As evidenced in

Table 7. Direct comparison between traditional charge-coupled devices (CCD) and MKIDs.

Parameters	Conventional CCD	MKID
time resolution	typically ≤500 Hz	1–10 kHz (10–100 times better turbulence-tracking capability)
wavelength resolution	Δλ/λ ≈ 0.05–0.1 (filter dependent)	Δλ/λ ≤ 0.005 (10–20 times higher dispersion-correction accuracy)
Sensitivity (noise)	Readout noise 3–5 e−/pix	Theory is zero readout noise (20–40 dB improvement in low-light signal-to-noise ratio)

, Microwave Kinetic Inductance Detectors (MKIDs) offer transformative advantages over conventional CCDs in temporal resolution, wavelength resolution, and sensitivity, which are critical for next-generation wavefront sensors.

Pyramid wavefront sensors are susceptible to stray light at high spatial frequencies when observing weak targets or in the visible-light wavelength band, which can contaminate the measurement signal and reduce the SNR, thereby affecting the quality of the wavefront correction. Pyramid wavefront sensors use prismatic optical elements to divide the incident wavefront into four quadrants, forming interference patterns on the focal plane. These interference patterns contain the phase information of the wavefront. By measuring the intensity distribution of these patterns, the phase distribution of the wavefront can be inferred, thereby improving the accuracy of wavefront detection. The complex amplitude distribution of the incident wavefront entering the pyramid wavefront sensor is as follows:

$${\Psi }_{aper}(x,y)=M(x,y)\cdot \exp \left[-i\cdot \Phi (x,y)\right], (x,y)\in R^2$$

(46)

where ${\Psi }_{aper}(x,y)$ denotes the complex amplitude at position (x,y), which contains the amplitude and phase information of the optical wavefront. $M(x,y)$ is the telescope-aperture mask, which defines the effective region of the optical wavefront. In this study, $M(x,y)$ denotes the telescope aperture with a circular center mask, and $M(x,y)$ = 1 when (x,y) is inside the aperture; otherwise, $M(x,y)$ = 0. Additionally, $\Phi (x,y)$ is the phase distribution of the optical wavefront entering the sensor in radians, which reflects the aberration of the optical wavefront due to atmospheric turbulence and other factors. $\exp \left[-i\cdot \Phi (x,y)\right]$ denotes the phase factor that describes the phase variation in the optical wavefront. In the complex representation, the phase factor is used to represent the spatial phase distribution of the wavefront.

The complex amplitude distribution of the optical wavefront reaching the detector plane is as follows:

$${\Psi }_{\det }(x,y)={\displaystyle \frac{1}{2\pi }}({\psi }_{aper}\ast PS{F}_{pyr})(x,y)$$

(47)

where ${\Psi }_{\det }(x,y)$ denotes the complex amplitude at position (x,y) in the detector plane, which reflects the state of the optical wavefront after it has been modulated by the pyramidal prism. ${\psi }_{aper}$ is the complex amplitude of the incident optical wavefront, which enters the pyramidal prism as an input signal. Additionally, $PS{F}_{pyr}$ is the PSF of the pyramidal prism, which describes the spatial-distribution changes. The PSF contains information about the modulation of the optical wavefront by the pyramid prism and is key to understanding the performance of the sensor. The symbol ∗ denotes the convolution operation, which describes the interaction of the incident optical wavefront with the PSF of the pyramidal prism [95].

A spatial filtering technique was applied to minimize the interference of higher-order spatial frequencies on the wavefront measurements by selectively blocking the light in front of a pyramidal prism. The transmittance function

$$T(k)=rect({\displaystyle \frac{k}{k_c}})$$

(48)

suppresses the high-frequency noise (cut-off frequency $k_c={\displaystyle \frac{D}{2\lambda f_{pyrm}}}$, where k_c denotes the cut-off frequency used to suppress the high-frequency noise, D denotes the aperture of the telescope, λ denotes the wavelength, which is an important parameter of a light wave and determines the frequency and color of the light, and f_max denotes the maximum frequency). The VLT-PWFS was designed to reduce the VLT-PWFS IF error from 12 to 4 nm RMS. In response to the ELT (Extremely Large Telescope) segmented mirror challenge (primary mirrors up to 25–40 m in diameter with segmented design), conventional reconstruction algorithms fail due to pupil breakage. Innovative solutions to address this included the following [14,96]:

• Pyramidal Wavefront Sensor via the Split Approach

This method splits the wavefront reconstruction into two parts, segmented piston mode ${\delta }_j$ reconstruction and continuous phase ${\varphi }_{cont}$ reconstruction, and uses existing non-segmented pupil wavefront-reconstruction algorithms to process the segmented data.

b.

Direct Segmented Piston Reconstructor (DSPR) Approach

This method introduces two methods to directly reconstruct the segmented piston by constructing the segmented orthogonal basis Bj and solving the piston offset independently by projection, which improves the computational efficiency five-fold. The two synergistically suppress the E-ELT wavefront residuals to λ/80 RMS, which meets the diffraction-limit demand of 30 m-class telescopes.

2.3.9. Performance Verification and Scientific Output

With the maturity of key technologies, the ExAO system has entered the stage of system integration and practical application; it is currently limited by the cooperative control of multiple subsystems and practical observation verification. Each subsystem must be integrated, and its performance must be verified through practical observation to meet the actual needs of astronomical observation. The two major challenges of wavefront-coronagraph-detector timing synchronization (Δt < 0.5 ms) and noncommon optical-path aberration suppression (ΔϕNCPA < λ/100) must also be overcome.

In 2015, Jovanovic et al. (Subaru Telescope, Macquarie University et al.) demonstrated the single-mode fiber injection efficiency of the SCExAO system during on-orbit operation [96], where a coupling efficiency of up to 74% was achieved by optimizing the single-mode fiber coupling as follows: $\eta \propto SR\cdot {\mathrm{e}}^{-{({\displaystyle \frac{\mathrm{\Delta }\theta }{{\theta }_0}})}^2}$, where η denotes the coupling efficiency, i.e., the ratio of the received optical power to the incident optical power of the optical fiber; SR denotes the saturation efficiency, which is a constant indicating the maximum coupling efficiency when the fiber receiving angle is zero; e is the base of the natural logarithm, which is approximately 2.71828; Δθ denotes the change in the fiber receiving angle; θ₀ is the fiber receiving angle, i.e., the maximum angle at which the fiber is able to receive. Magniez et al. (Durham University, SRON et al.) proposed the concept of utilizing MKID technology to enhance the performance of pyramidal wavefront sensors (PWFS) [97]. The MKID-PWFS architecture is divided into sensing, detection, and control ends. For sensing, the pyramidal prism splits the incident beam into four beams, creating four suboptical pupil images that are synchronized and captured by the MKID. For detection, MKID realizes single-photon time-stamping (1–10 μs) and photon-energy resolution based on the superconducting resonance-frequency shift and outputs the intensity and phase-gradient information of the four subpupils synchronously. Using the superconducting detector single-photon energy resolution (Δλ/λ = 0.01) and 5 kHz time resolution, multiband wavefront reconstruction (ϕcorr(λ) = ϕDM was realized λ0/λ, Φcorr(λ) denotes the corrected phase at a specific wavelength λ. ΦDM denotes the phase in dynamic mode. λ0 denotes a reference wavelength, typically a specific wavelength value, used for calibration and phase-gradient information, and λ is the current wavelength. This approach improves the dispersion-correction accuracy by 40%. For the control, the DM is driven to correct the aberration through the real-time reconstruction of the wavefront phase, forming a closed-loop system.

MKID outputs the arrival timestamp, spatial coordinates, and energy information (wavelength) of each photon, forming a spatio-temporal-energetic four-dimensional dataset of the tetrad pupil of light. The MKID-PWFS slope formula (four-quadrant light intensity difference) is as follows:

$$\begin{array}{l}{S}_x(x,y)={\displaystyle \frac{\left[I_1(x,y)+I_2(x,y)\right]-\left[I_3(x,y)+I_4(x,y)\right]}{{\displaystyle \sum I}}}\\ S_y(x,y)={\displaystyle \frac{\left[I_1(x,y)+I_3(x,y)\right]-\left[I_2(x,y)+I_4(x,y)\right]}{{\displaystyle \sum I}}}\end{array}$$

(49)

S_x(x,y) and S_y(x,y) denote the slopes of the x- and y-axes at the pixel position (x, y), respectively. These slope values reflect the degree of aberration in the wavefront at that position. I₁ (x,y), I₂ (x,y), I₃ (x,y), and I₄ (x,y) denote the light intensities on the four subpupil images of the PWFS at pixel position (x,y). These four subpupil images are formed by a pyramidal prism that splits the incident beam into four parts and is projected onto the detector using a reimaging lens. The light intensity of each subpupil image reflects the distortion of the wavefront in the subpupil region. ∑I denotes the sum of all relevant pixel light intensities for the normalized slope calculation. Normalization was performed to ensure that the slope values were within reasonable limits for subsequent processing and analysis.

In 2018, Currie et al. summarized the in-orbit performance and latest scientific results of the SCExAO system [12], including the discovery of a new companion star system and detailed characterization of a planetary-formation disk. The SCExAO long-exposure verification of deep-space exploration capabilities exhibited a 5σ contrast curve for a fifth-magnitude star at a 5 h integration time, with contrast values of 10⁻⁵, 2 × 10⁻⁶, and 10⁻⁶ at 0.25″, 0.4″, and 0.8″, respectively. In the same year, Morzinski et al. performed end-to-end simulations of the MagAO-X instrument, demonstrating that the MagAO-X system achieved a closed-loop contrast of 1.14 × 10⁻⁴ (open-loop value 4.3 × 10⁻⁴)in the Hα band (656 nm), validating its practicality in the visible-light band [98]. And then, a satellite-borne spacecraft concept based on laser-guide stars was proposed to enhance wavefront sensing and control (WFSC) signals [99], aiming to achieve picometer-level stability for the primary mirror segment of future large space telescopes. In the Hα band (656 nm), the MagAO-X instrument achieved average contrasts of 1.168 × 10⁻⁴ (upper dark hole) and 1.144 × 10⁻⁴ (lower dark hole) after closed-loop wavefront control (WFSC). Without wavefront control, the open-loop contrast ratios were 4.855 × 10⁻⁴ and 4.345 × 10⁻⁴, respectively [100]. Wavefront Sensing and Control (WFSC) for future segmented space telescopes (e.g., LUVOIR) requires picometer-level stability. An optical truss uses metrology lasers to monitor segment positions, enabling closed-loop control against thermal/mechanical drift—critical for maintaining coronagraph alignment over years.

Regarding ExAO, the latest development is the GMagAO-X system [101]. Pushing the boundaries of ExAO for the upcoming generation of Extremely Large Telescopes (ELTs), the GMagAO-X system for the Giant Magellan Telescope (GMT) introduces a revolutionary parallel deformable mirror (DM) architecture. This design employs seven commercial BMC 3K MEMS DMs operating in parallel, collectively providing 21,000 actuators—a solution overcoming the current unavailability of monolithic high-actuator-count DMs required for ELTs. Each DM segment offers ±7 μm stroke at 3.5 kHz, enabling diffraction-limited performance at visible wavelengths (λ = 0.8 μm). Laboratory validation on the HCAT testbed achieved a 73% Strehl ratio at 850 nm, confirming optical coherence across the parallel DM configuration. Critical for phasing GMT’s segmented primary mirror, the system incorporates a Holographic Dispersed Fringe Sensor (HDFS) achieving <50 nm RMS segment piston error under simulated on-sky turbulence (0.9″ seeing). This sensor utilizes dispersed “barber pole” fringes to circumvent phase-wrapping ambiguities. Furthermore, GMagAO-X addresses polarization-induced performance loss through a specialized “crossed-mirror” optical design, verified via Zemax modeling and laboratory tests.

2.3.10. ExAO: A Predominantly Ground-Based Enabling Technology

While the preceding discussion has detailed the components, algorithms, and performance advancements of ExAO systems, a fundamental aspect defining their role and requirements is the distinction between their primary domain—ground-based astronomy—and potential applications in space.

Extreme Adaptive Optics (ExAO) is fundamentally a technology developed for and primarily deployed on ground-based telescopes. Its core purpose is to overcome the severe image degradation caused by atmospheric turbulence, which is absent in space. Systems like SCExAO (Subaru), SPHERE (VLT), MagAO-X (Magellan), and KPIC/KPIC-II (Keck) exemplify state-of-the-art ground-based ExAO. They integrate high-order deformable mirrors (1000 s of actuators), advanced wavefront sensors (Pyramid WFS, spatially filtered Shack-Hartmann), high-speed real-time control, and sophisticated coronagraphs to achieve Strehl ratios > 0.9 in the near-infrared and contrasts of 10⁻⁶ to 10⁻⁷ at small angular separations. The key challenges driving ExAO development on the ground are the millisecond-scale turbulence evolution, requiring kHz-level correction bandwidths, and the suppression of both atmospheric and static (Non-Common Path Aberrations—NCPA) wavefront errors down to nanometers RMS.

While the primary domain of ExAO is ground-based astronomy, its principles and components find application in specific space-borne contexts, though the requirements differ significantly. In space, the dominant challenge is not dynamic turbulence but achieving and maintaining exquisite static wavefront stability at the picometer level over long durations and large apertures, crucial for future missions aiming for 10⁻¹⁰ contrast (e.g., HabEx, LUVOIR). Technologies developed for ExAO, such as high-actuator-count DMs, high-sensitivity wavefront sensing (e.g., for coarse phasing of segmented mirrors), and advanced control algorithms, are directly relevant. Furthermore, concepts like laser guide stars, while primarily a ground-based solution for atmospheric tomography, inspire analogous metrology systems in space for maintaining formation flying precision (e.g., for starshades or interferometers like LIFE) or internal wavefront sensing and control within a large telescope. However, the term “ExAO” itself, emphasizing extreme correction of dynamic disturbances, remains most closely associated with ground-based observatories pushing the boundaries of high-contrast imaging from Earth.

3. Development Trends

3.1. ExAO, Mutual Promotion Between Coronagraphs and Nulling Interferometry

The use of a single technological approach has reached its physical limits in the detection of Earth-like planets with a resolution of 0.1″ angular separation (contrast requirement of 10⁻¹⁰). Coronagraphs, while theoretically capable of achieving a blocking efficiency of 10⁻¹², are constrained by static-aberration limitations (NCPA > λ/50 results in measured contrast degradation to 10⁻⁸). The coherence-cancelation efficiency of zero-sum interferometry decreases below 90% when Δϕ > λ/200. The wavefront-correction residual of ExAO is difficult to break through the λ/100 RMS barrier with millisecond-level delays.

Therefore, a combination of the three technologies is required. ExAO solves the static-aberration problem of star coronagraphs by providing a stable phase difference for null interference. The star coronagraph and null interference expand the imaging boundaries from the light-intensity and coherence dimensions, respectively.

3.1.1. ExAO to Enhance the Performance of Coronagraphs

Static-aberration suppression is achieved through PSF-replication technology, compensating for noncommon-path errors and reducing NCPA from λ/15 to λ/100 RMS, bringing the measured contrast of coronagraphs close to the theoretical limit. For example, in VLT-SPHERE, ExAO improved the contrast of the vortex coronagraph from 10⁻⁶ to 10⁻⁸. A kilohertz-level DM performs dynamic turbulence correction to suppress the PSF jitter caused by atmospheric turbulence, ensuring the stability of the dark region of the coronagraph [102].

3.1.2. Phase Assurance for ExAO Nulling Interferometry

ExAO achieves phase synchronization at the subnanometer level through multiple beams, maintaining coherence and long-exposure stability. For example, after integrating ExAO into the Keck interferometer, the phase noise of the interferometer array is suppressed to Δϕ < λ/200, ensuring the coherence-cancelation efficiency of stellar light, thereby deepening the mid-infrared nulling depth from 10⁻⁴ to 10⁻⁶. In addition, the sodium guide star ExAO extends the halo angle, enabling hour-level interferometric integration [103]. As demonstrated in

Table 8. Typical examples of the coupled effects of ExAO, coronagraph, and nulling interferometry.

Case (Law)	ExAO Role	Coronagraph	Nonzero Interference Effect	Coupling Gain
VLT-SPHERE	Suppression of static aberration (λ/100 RMS)	Creating polarized dark zones	Central stellar residual-light offset	Contrast improved 100-fold
Keck interferometer	Phase lock (Δφ < λ/200)	Front-diffraction light filtering	Mid-infrared stellar optical-coherence cancelation	Depth of Nulling deepened 100-fold
CSST	On operate jitter suppression	Compression of internal working angle to 1.5 λ/D	Extension of valid points	Signal-to-noise ratio improved $\sqrt{10}$ times

, the synergistic coupling of ExAO with coronagraphs and nulling interferometry yields significant performance gains, such as improved contrast and deeper nulling depths.

The integration of these three technologies will be realized in the future, enabling unified control of the three systems and overcoming the “contrast-resolution-stability” bottleneck that single technologies cannot overcome in the exploration of exoplanets similar to Earth. In the short term, PSF-replication technology will be adapted to the CSST to achieve a spatial contrast ratio of 10⁻⁸. The long-term goal is to achieve photon-chip null interference combined with a 30 m-class telescope ExAO, targeting a contrast ratio of 10⁻¹⁰ at an angular separation of 0.1″, which is the ultimate goal of directly analyzing the atmospheric spectra of Earth-like planets [104].

3.2. Applications of Deep Learning

Deep learning is revolutionizing exoplanet detection in the era of intelligent sensing and driven a revolution in the high-contrast imaging of celestial bodies.

First, in the ExAO domain, deep-learning technology has been applied to turbulent evolution through spatio-temporal modeling with physically constrained LSTM networks [105]. By encoding atmospheric-layer turbulence-feature profiles, it predicts the trajectory of wavefront distortions over the next 50 ms, reducing correction delays to 0.1 ms (traditional methods > 0.5 ms). Based on observations of a small number of point light sources, a digital-twin model of telescope alignment errors was generated through a meta-learning PSF compensator. As a case study, the Nanjing University team reduced the number of iterations for VLT static-aberration correction from 1000 to 20. Moreover, the Samueli School of Engineering at the University of California, Los Angeles (UCLA) introduced a general framework for PSF engineering [106], utilizing a diffractive optical processor to synthesize arbitrary, spatially varying three-dimensional PSFs.

In the field of coronagraphs, deep learning can achieve multi-physics joint denoising through the intelligent decoupling of speckle noise and can also enhance the intensity of planetary diffraction spots in real time by integrating programmable diffraction layers (e.g., DMD + U-Net) at the coronagraph focal plane through diffraction optics–neural network collaboration. For example, the SPHERE system achieves a contrast improvement from 10⁻⁷ to 10⁻⁹ at 0.2″.

In the field of nulling interferometry, compressed-state photons carrying interference-phase information can enhance noise resistance by 30 dB. By optimizing the observation strategies through reinforcement learning, the dynamic balance between the exposure depth and target priority was optimized, thereby improving the screening efficiency of the LIFE mission by 300% [107].

In the field of the direct imaging of exoplanetary life, deep learning will lead to three breakthroughs in high-contrast imaging [18]. (1) Time dimension: nanosecond-level wavefront control (currently at the millisecond level) will be used to freeze atmospheric turbulence. (2) Sensitivity dimension: The extracting of single-photon-level signals (currently at the hundred-photon level) will be used to resolve temperature differences of 25 K between planets. (3) Cognitive dimension: AI will be used to directly interpret planetary-surface reflectance spectra to identify vegetation red-edge features (wavelength 750 nm) and urban lighting (1–10 μm). Looking ahead, by 2030, deep learning is expected to assist in discovering the first Earth-like planetary biosignatures (such as O₂-CH₄ imbalance in Proxima b); by 2035, intelligent telescopes will map the continental distribution of exoplanets (resolution at the 100 km level); and by 2040, a probability model for exoplanetary life will be established (based on a global spectral database).

3.3. Integrated Photonics and the Application of Optical Neural Networks to Astronomical High-Contrast Scenes

As the contrast requirements for the direct imaging of exoplanets using ground-based/space telescopes surpass the 10⁻¹⁰ magnitude level, traditional electronic computing architectures face fundamental limitations. (1) Real-time constraints: the timescale of atmospheric-turbulence evolution (~1 ms) far exceeds the delay of electronic wavefront correction (>0.5 ms), resulting in residual errors σ ≥ λ/50 in ExAO systems. (2) Speckle physical limits: speckle noise caused by static aberrations forms fixed patterns at the coronagraph focal plane; thus, traditional algorithms cannot distinguish them from the true planetary signal (SNR < 5 at 0.2″). (3) Interferometric phase drift: the coherence-cancelation efficiency of null interference is extremely sensitive to phase noise (when Δϕ > λ/100, ηnull <90%).

3.3.1. Integrated Photonics

Integrated photonics integrates a large number of optical components on a single photonic chip [108]. This can enhance the performance, cost-effectiveness, and scalability of optical systems that feature high-dimensional data processing and real-time control.

3.3.2. Optical Neural Networks

Optical neural networks are an important direction in the integration of photonics and artificial intelligence. They utilize optical components and optical signal-processing technology to simulate the computational process of neural networks, enabling more efficient and low-power neural-network computing [109]. Photonic neural networks leverage the unique high-speed parallel-processing capabilities of photonics to rapidly identify key features in images with high contrast, thereby enhancing real-time imaging performance and accuracy. They also offer advantages in processing large-scale data and complex computational tasks.

The integration of photonics and optical neural networks through light-speed linear computation (matrix multiplication), optoelectronic nonlinear activation (NOFU units), and multiphysical dimension-fusion processing (polarization/spectrum/phase) addresses three key challenges in high-contrast imaging: a. ExAO real-time performance: nanosecond-level wavefront correction replacing millisecond-level electronic control; b. Coronagraph speckle suppression: physical-level optical-domain noise filtering that enhances the contrast by 1–2 orders of magnitude; and c. Nulling interferometry stability: the wavelength–time interleaved architecture that achieves pm-level phase synchronization.

In the future, integrated photonics and optical neural networks will redefine the architecture of high-contrast imaging in astronomy, transitioning from a serial “acquisition-storage-processing” model to a “sensing-computing-integration” photonic intelligent terminal. By 2025–2030, ground-based telescopes will deploy optical-computing coprocessors to achieve ExAO closed-loop bandwidths > 5 kHz, and by 2030–2035, space stations will carry photon artificial-intelligence accelerators to complete in-orbit analyses of exoplanet atmospheric compositions. After 2040, photonic neuromorphic chips will fuse with quantum sensors to achieve the molecular-line imaging of cosmic protostellar systems. The ultimate goal of photonic intelligent terminals is to convert cosmic signals into scientific knowledge with a delay approaching the speed of light itself—this is the inevitable technological path for humanity’s exploration of the ultimate spacetime scale.

3.4. The Key Challenges That Are Expected to Arise in the Short and Long Term

3.4.1. Near-Term Horizon (~1–5 Years): Integration and Ground-Based Validation

Goals: Achieve robust integration and synergistic control of ExAO, coronagraphs, and/or nulling interferometry on existing 8–10 m class ground-based telescopes (e.g., VLT, Subaru, Keck); demonstrate stabilized raw contrasts of 10⁻⁸ within inner working angles (IWAs) of 2–4 λ/D for bright stars; validate key enabling technologies like photonic chip nulling (e.g., enhanced GLINT pathfinders); and deploy initial AI models for wavefront prediction and speckle suppression. Establish foundational performance for missions like the Chinese Space Station Telescope (CSST) module targeting ~10⁻⁸ contrast.

Key Challenges: Suppressing Non-Common Path Aberrations (NCPA) below λ/100 RMS; achieving robust, real-time control across complex multi-subsystem instruments (ExAO + Coronagraph/Nulling hardware + algorithms); scaling photonic chip complexity and validating on-sky performance under turbulence; acquiring sufficient high-quality, diverse datasets for effective AI training; and managing instrumental polarization effects in integrated systems.

Priority Development Directions:

• Unified Control Architectures: Develop and implement real-time control systems capable of orchestrating ExAO correction, coronagraph mask optimization, nulling phase control, and potential AI co-processing within a single, low-latency framework.
• Advanced NCPA Calibration and Suppression: Prioritize research into next-generation wavefront sensing (e.g., asymmetric pupil, modulated pyramid WFS) and techniques like PSF replication or differential optical path monitoring integrated directly into the science path for high-precision (<λ/100 RMS) static aberration control.
• Photonic Chip Prototyping and Validation: Accelerate the development, fabrication, and on-sky testing of more complex integrated photonic circuits for nulling (multi-baseline kernel-nulling chips) and beam combination, focusing on improving throughput, phase stability, and turbulence filtering capabilities.
• AI for Enhanced Operations: Deploy foundational deep learning models for predictive wavefront control (reducing latency), real-time speckle noise identification and subtraction in coronagraphic data, and optimizing observational strategies (target selection, exposure times).
• High-Fidelity System Modeling: Enhance end-to-end simulation tools to accurately model the coupled physics of turbulence, wavefront correction, coronagraphic diffraction, nulling coherence, and detector effects for design optimization and performance prediction.

3.4.2. Mid-Term Horizon (~5–15 Years): Scaling and Space Qualification

Goals: Deploy mature convergent HCI systems on next-generation 30 m-class Extremely Large Telescopes (ELTs—E-ELT, TMT, GMT), achieving stabilized contrasts of 10⁻⁹ and probing closer to the diffraction limit (IWA ~1–2 λ/D); execute space-based medium-scale missions (HabEx/LUVOIR-B scale) capable of detailed atmospheric characterization of Jupiter- and Neptune-analogs around nearby stars; integrate optical neural networks (ONNs) into ExAO real-time control loops, breaking the millisecond latency barrier; validate formation flying and photonic nulling interferometry for future large interferometers via pathfinder missions; and enable initial spectroscopic searches for bio-signatures (e.g., CH₄, O₃, CO₂) on temperate super-Earths with dedicated instruments.

Key Challenges: Scaling ExAO performance to ELT scales (~1000 s of actuators, multi-laser guide star tomography, kHz operation); designing and manufacturing ultra-stable optical systems for space coronagraphs capable of maintaining 10⁻¹⁰ contrast stability; developing efficient, low-latency, low-power photonic AI processors for real-time control; achieving picometer-level stability in formation flying for space interferometry; and mitigating the impact of exozodiacal light for faint planet detection.

Priority Development Directions:

• ELT-Scale ExAO Systems: Drive advancements in high-density deformable mirrors (e.g., MEMS, piezo), advanced wavefront sensors (e.g., MKID-PWFS, LIFT), laser guide star systems, tomographic reconstruction algorithms, and real-time computing architectures capable of handling ELT complexity and data rates.
• Space Coronagraph and Nulling Technology: Mature technologies like PIAACMC, vector vortex coronagraphs, and integrated photonic nullers for space environments. Focus on thermal/mechanical stability, radiation hardening, and in-flight calibration capabilities. Demonstrate 10⁻¹⁰ contrast stability in testbeds and sub-orbital flights.
• Photonics and ONNs for Real-Time Control: Develop and integrate photonic co-processors and ONNs capable of performing wavefront reconstruction, nonlinear control, and speckle field manipulation at nanosecond speeds, overcoming electronic latency limitations.
• Precision Formation Flying and Interferometry: Advance metrology systems, propulsion, and control algorithms for nanometer-to-picometer level maintenance of baseline distances and optical path differences in distributed spacecraft systems (e.g., LIFE precursor missions).
• Advanced Biosignature Detection and Validation: Develop sophisticated spectral retrieval pipelines, incorporate context from planetary system architecture, and establish rigorous frameworks for quantifying false positive probabilities for potential biosignatures like O₂-CH₄ disequilibrium.

3.4.3. Long-Term Horizon (>15 Years): Earth-Analog Characterization

The overarching objectives for this era are to: Launch and operate large-scale space missions: either a flagship coronagraphic telescope (LUVOIR-A scale) or a long-baseline interferometric array (LIFE mission), achieving the extreme contrast (~10⁻¹⁰) and angular resolution (<0.1″) needed to directly image and obtain high-SNR spectra of Earth-like planets in the habitable zones of nearby Sun-like stars; identify and confirm atmospheric biosignatures through multi-band spectral analysis; establish “photonics intelligent terminals” as the standard architecture for HCI, enabling light-speed processing; and begin mapping surface features (e.g., continents, clouds) on resolved exoplanets.

This path is fraught with profound challenges, including the need to maintain unprecedented wavefront stability (λ/1000 RMS) and thermal/mechanical control over large structures in space for years; achieving sufficient SNR for robust detection of weak molecular features in exo-Earth atmospheres; distinguishing true biosignatures from abiotic sources with high confidence; managing the complexity and cost of large space observatories; and developing reliable, large-scale photonic and quantum-enabled sensing systems.

Priority Development Directions:

• Extreme Stability Platforms: Pioneer revolutionary spacecraft and optical bench designs incorporating active and passive isolation, advanced materials (e.g., zero-CTE composites), and nanometer-level metrology and control systems for unprecedented dynamic stability.
• Ultra-Sensitive, Multi-Band Detection: Develop next-generation detectors (beyond MKIDs) with high quantum efficiency, extremely low noise, photon-counting capability, and intrinsic energy resolution across UV to mid-IR wavelengths critical for biosignatures.
• Advanced Atmospheric Retrieval and Biosignature Assessment: Create coupled climate-chemistry-photochemical models and AI-powered retrieval tools capable of interpreting complex, low-SNR spectra within a full planetary context to assess habitability and the probability of life.
• Photonics and Quantum Sensing Integration: Fully realize the vision of photonic intelligent terminals by deeply integrating photonic circuits for light manipulation, optical neural networks for processing, and potentially quantum sensors for ultra-precise metrology into a unified, high-efficiency system.
• Large Mission Architectures and Funding Strategies: Conduct comprehensive system studies, technology maturation programs, and international collaborations to define and secure the path towards constructing and launching these ambitious observatories.

4. Summary

This review summarizes the pivotal role of high-contrast imaging (HCI) techniques in astronomical research and the technological advancements they have spurred. By integrating advanced technologies such as ExAO systems, coronagraphs, and nulling interferometry, high-contrast imaging technology has significantly suppressed stellar glare, revealing faint planetary signals hidden behind it and opening up new avenues for the direct detection of exoplanets.

As shown in

Table 9. Summary of key technologies for high-contrast imaging of celestial bodies, as well as related parameters and their importance.

Technical Name	Parameters	Technical Significance
OHANA Phase II	Spectral band: near-infrared (NIR) band (J, H, K) Maximum length of optical fiber: approximately 500 m Propagation loss: 1 dB maximum Transmission efficiency: conservative estimate of 50% Sensitivity: K = 13 ± 1 (when using an 8 m telescope)	Low-loss transmission using fiber-optic connection technology to verify the feasibility and performance of fiber-optic connection interferometers
MIRC-X	Closure Phase Accuracy: better than 1° Spectral Band: near-infrared (J and H bands) Angular Resolution: equivalent to that of a baseline telescope up to 330 m in diameter (0.6 mas) Sensitivity Improvement: approximately two magnitudes	Sensitivity improved by approximately two orders of magnitude compared to its predecessor MIRC, enabling high-resolution interferometric imaging and precise model-independent detection of asymmetries (e.g., from exoplanets or stellar spots). The high closure phase precision (<1°) is crucial for this.
LBTI, NIC	Operating Band: Thermal infrared (>2.5 µm), utilizing NOMIC (7–25 µm) and LMIRCam (3–5 µm) cameras. Key Achievement: Demonstrated nulling interferometry from the ground, studying exozodiacal dust and giant planets.	A premier ground-based interferometer showcasing the power of traditional bulk optics for high-contrast mid-infrared astronomy. It operates as a nulling interferometer and imager.
Integrated photonic technology	Representative Instruments: GLINT: Operates at 1.6 µm, achieving null depths ~10−4 on the Subaru telescope. GRAVITY/PIONIER (VLTI): Use photonic beam combiners for astrometry and imaging. Dragonfly/Hi-5: Pathfinder instruments demonstrating on-chip beam combination. LIFE Mission Concept: Plans to use advanced photonic kernel-nulling in the mid-infrared (4–18 µm).	Enhanced optical-system performance, cost-effectiveness, and scalability. Achieved high sensitivity nulling interferometry and advanced the development of exoplanet-detection technology in the mid-infrared band.
Improved Lyot Coronagraph	Contrast ratio: up to 3 × 10−10 (within two circular fields of view ranging from 3 λ/D to 15 λ/D) Operating spectrum: broadband	Improved luminous efficacy and internal working angle for broadband operation
Flat Top Sinusoidal Phase-plate Coronagraph	Contrast ratio: 10−3 in the wavelength range of 490–620 nm for stellar-light extinction. Operating spectrum: 490–620 nm	Combined the advantages of the sinusoidal phase plate coronagraph and six-platform phase-plate coronagraph to achieve a balance of broadband operation and high-contrast imaging
Phase-induced Amplitude Toe-cut Coronagraph (PIAA)	Contrast: Theoretically capable of achieving near 100% luminous efficacy and 10−10 contrast imaging at an angle of 2 λ/D. Inner working angle: small	Overcame the traditional optical pupil cut-toe-type coronagraph low-throughput, large internal working angle, and other shortcomings
PIAA combined with Lyot Coronagraph	Contrast ratio: 2.6 × 10−8 for 650 nm monochromatic incident light, 1.8 × 10−7 for 10% bandwidth complex color light	Enabled the application of PIAA to complex optical-pupil shapes and further improved performance
SCExAO system	Contrast: 5σ contrast curves were achieved for fifth-magnitude stars at 5 h of integration time, with 10−5, 2 × 10−6, and 10−6 contrast at 0.25″, 0.4″, and 0.8″, respectively Spectral band: Visible to near-infrared band (approx. 500–900 nm)	High-contrast imaging was achieved and several exoplanets were successfully detected
Pyramid Wavefront Sensor	Accuracy: Improved accuracy of wavefront detection Spectral band: Multi-band	Utilized pyramid-shaped optics to divide the spot into four quadrants for improved wavefront-detection accuracy
Microwave Kinetic-inductance Detector (MKID) applied to PWFS	Provides higher performance than conventional CCD/CMOS detectors, especially in terms of high pixel count, high frame rate, and low readout noise	Enhanced performance of pyramidal wavefront sensors

, nulling interferometry has evolved from Bracewell’s foundational concept to sophisticated ground-based implementations (e.g., OHANA, MIRC-X, LBTI, GLINT) and ambitious space mission designs (e.g., LIFE), continuously pushing the boundaries of angular resolution and contrast in suppressing stellar light. Coronagraph technology has diversified from external occulters to advanced internal designs (e.g., band-limited masks, phase masks like 4QPMC and OVPMC, apodized pupil Lyot coronagraphs, and notably, Phase-Induced Amplitude Apodization (PIAA) and its derivatives like PIAACMC), significantly improving inner working angles, throughput, and operational bandwidth. ExAO systems have achieved remarkable wavefront correction fidelity (Strehl ratios > 0.9) through innovations in high-actuator-count deformable mirrors, advanced wavefront sensors (e.g., pyramid, APF-WFS), sophisticated control algorithms (e.g., MGCG), and novel detectors (e.g., MKID-PWFS), enabling coronagraphs and interferometers to approach their theoretical performance limits. Crucially, the synergistic interplay between these three core technologies—where ExAO provides the stable wavefront essential for coronagraphic dark zones and coherent nulling, while coronagraphs and nulling interferometry expand the achievable contrast from the intensity and coherence dimensions, respectively—has become increasingly recognized as vital for advancing high-contrast imaging.

However, despite these impressive achievements, fundamental limitations persist. Achieving the extreme contrast (~10⁻¹⁰) and angular resolution (<0.1″) required for direct imaging and spectroscopy of Earth-like planets remains elusive. Coronagraphs are ultimately constrained by static non-common path aberrations (NCPA) and wavefront stability. Nulling interferometry demands exquisite phase control (Δϕ ≪ λ/100) and long-baseline coherence. ExAO faces inherent challenges in real-time turbulence correction latency, NCPA calibration, and the computational complexity of large-scale systems.

Building directly upon the foundation laid by current ExAO, coronagraph, and nulling interferometry technologies, the future roadmap is characterized by three transformative and interconnected trends: convergence, intelligence, and photonic integration.

• Convergence: The deep integration of ExAO, coronagraphs, and nulling interferometry is transitioning from a promising concept to a necessity. Future systems will feature unified control architectures where ExAO not only corrects turbulence but also actively compensates for coronagraph-specific NCPA and maintains the sub-nanometer phase stability required for deep nulling over extended integrations. This synergistic approach addresses the core limitations of individual techniques, pushing achievable contrast closer to the theoretical limits demanded by exo-Earth characterization, as exemplified by the performance gains demonstrated in systems like VLT-SPHERE and envisioned for future missions like LIFE.
• Intelligence (Deep Learning): Deep learning represents a paradigm shift in how high-contrast imaging systems are designed and operated. Moving beyond traditional analytical models, AI techniques offer powerful solutions for critical bottlenecks: predicting atmospheric turbulence evolution for faster ExAO control, intelligently disentangling speckle noise from true planetary signals in coronagraphic data, optimizing observing strategies and fringe tracking for nulling interferometry, and even directly interpreting complex planetary spectra for biosignatures. This AI-driven approach promises orders-of-magnitude improvements in speed, sensitivity, and autonomy compared to conventional algorithms.
• Photonic Integration and Optical Neural Networks: Integrated photonics and optical neural networks offer a fundamental leap in system architecture and capability. Replacing bulk optics with photonic chips enables miniaturization, inherent stability, and novel functionalities like on-chip nulling (GLINT) or turbulence filtering. Optical neural networks, leveraging light-speed linear computation and multi-physical dimension processing, promise to overcome the fundamental real-time bottleneck of electronic systems for ExAO control and enable physical-layer noise suppression for coronagraphs at levels unattainable digitally. This convergence of photonics and AI aims to transition the entire high-contrast imaging chain from a sequential “sense-store-process” model to an integrated “sensing-computing” photonic intelligent terminal, operating at the ultimate speed limit—light itself. This represents a qualitative departure from the discrete, electronics-limited systems of today.

In essence, the roadmap for exoplanet high-contrast imaging is defined by a continuous evolution: leveraging the mature yet limited capabilities of current core technologies (nulling interferometry, coronagraphy, ExAO) and their synergistic fusion, while aggressively embracing the disruptive potential of deep learning and integrated photonics/optical neural networks. This combined path is essential to overcome the formidable challenges and finally achieve the long-sought goal of directly probing the atmospheres of Earth-like worlds around neighboring stars.

This review culminates in presenting a concrete technological roadmap for the future: The path forward is defined by the strategic integration of ExAO, coronagraphy, and nulling interferometry, empowered by the transformative potential of Artificial Intelligence (AI) and Integrated Photonics/Optical Neural Networks (ONNs). This roadmap outlines a phased approach:

• Near-Term (~1–5 years): Focus on integrating and validating convergent systems on existing large telescopes, achieving 10⁻⁸ contrast, and prioritizing unified control, advanced NCPA suppression, photonic chip prototyping, and initial AI deployment.
• Mid-Term (~5–15 years): Scale convergent technologies to ELTs (10⁻⁹ contrast) and medium space missions, prioritizing ELT-scale ExAO, space coronagraph/nullers, photonic/ONN real-time control, precision formation flying, and advanced biosignature detection.
• Long-Term (>15 years): Deploy large space observatories (flagship coronagraphs or interferometric arrays) capable of 10⁻¹⁰ contrast and <0.1″ resolution to image and obtain spectra of Earth-like exoplanets, prioritizing extreme stability platforms, ultra-sensitive detectors, sophisticated biosignature assessment, and fully realized photonic intelligent terminals.

Key development priorities highlighted throughout this roadmap include: resolving the NCPA suppression bottleneck, developing real-time control systems for complex convergent instruments, scaling ExAO to ELT dimensions, maturing ultra-stable space optics and precision formation flying, advancing photonic integration and ONNs for light-speed processing, and creating robust AI frameworks for wavefront prediction, speckle suppression, observation optimization, and biosignature interpretation. The synergy of these converging technologies and enablers provides the only viable pathway to transform the dream of directly analyzing the atmosphere of a pale blue dot orbiting another star into scientific reality. By embracing this roadmap, the field of high-contrast exoplanet imaging is poised to make the revolutionary leap from detecting gas giants to searching for signs of life on worlds beyond our solar system.