Direct High-Resolution Imaging of Earth-like Exoplanets with the Solar Gravitational Lens

Abstract: We present a scalar, aperture-averaged observability benchmark for resolved exoplanet imaging with the solar gravitational lens (SGL). A real Earth luminance map is propagated through a scanned SGL image plane using source-to-image-plane compression, a finite-aperture SGL kernel, and an Einstein-ring photon-count model. We inject photon noise, finite-exposure smear, cloud-like temporal variability, coronal and detector-calibration residuals, optical-operator mismatch, and navigation error, then reconstruct with a regularized Fourier/Wiener inverse. For an Earth-radius planet at 30 pc observed from 650 AU with a 1 m telescope, the projected image cylinder is 1.338 km across; a $128\times128$ raster has 10.46 m image-plane pitch, 99.55 km source pixels, and convolved-image SNR 43.16 for 1800 s per sample. With the stated ring-extraction, temporal, background, calibration, and navigation mitigations, the fiducial reconstruction reaches SNR 6.89, structural similarity 0.848, normalized RMS error 0.439, and a 232 km Fourier-ring-correlation resolution proxy. A higher-count case reaches SNR 10.49, structural similarity 0.927, normalized RMS error 0.287, and the 199 km two-pixel grid floor. Controlled cloud tests show that a static inverse fails for rotating cloudy data, whereas phase-registered robust coadds recover persistent surface-proxy structure in the scalar model. The dominant requirements are temporal sampling, coronal and detector calibration, ring extraction, image-plane metrology, optical-operator knowledge, and dynamic inversion, not calibrated monopole SGL blur. The benchmark demonstrates recoverable 200-230 km-class broadband spatial contrast in the stated scalar model, but does not validate propulsion, communications, coronagraphic or external-occulter propagation, physical solar multipoles, physical cloud fields, or a full dynamic retrieval pipeline.

• The paper demonstrates that SGL imaging can resolve Earth-like exoplanets using a detailed simulation pipeline incorporating photon statistics and sub-ppm calibration.
• It quantifies mission requirements by modeling instrumental noise, cloud-induced variability, navigation errors, and the convolutional 1/ρ response of the SGL.
• It shows that proper temporal sampling and dynamic inversion enable recovery of surface details at resolutions of 200–230 km under realistic conditions.

Direct Imaging of Earth-like Exoplanets with the Solar Gravitational Lens: A Technical Analysis

Introduction

This paper presents a comprehensive, scalar, aperture-averaged benchmark modeling the feasibility and limitations of direct, high-resolution imaging of Earth-like exoplanets using the Solar Gravitational Lens (SGL) (2606.14899). The study focuses on realistic end-to-end system constraints: image-plane sampling, photon statistics, systematic and random noise, instrumental and astrophysical backgrounds, temporal variability, navigation error, and numerical regularization. It establishes quantitative mission and inversion requirements and provides a basis for assessing whether SGL-enabled imaging architectures can achieve spatial resolutions and signal-to-noise ratios unattainable through conventional techniques.

SGL Imaging Framework and Forward Modeling

SGL imaging leverages the Sun's gravitational field to provide diffraction-limited angular resolution (sub-nanoarcsecond) and significant optical gain for background-limited observations. However, the SGL's extended and convolutional response—effectively a broad, aperture-averaged $1/\rho$ tail—means that unlike a conventional telescope, the source's image is spread into an Einstein ring that must be carefully sampled and algorithmically inverted.

The simulation pipeline initiates with a real, high-resolution Earth image, preprocessed to scalar luminance on a $128\times128$ grid. This image is forward-modeled through the SGL kernel, incorporating the projected cylinder geometry, photon-counting regimes for both planet and dominating coronal backgrounds, and a comprehensive suite of noise and systematics. Instrumental and physical effects injected into the modeling pipeline include rotational and cloud-induced temporal variability, finite-exposure smearing, navigation errors, calibration and coronagraphic residuals, and operator (PSF) mismatches. The resulting measurements are subject to a regularized Fourier/Wiener inversion process to test recoverability (see Figure 1 and Figure 2).

Figure 1: Target-distance and sampling scalings for SGL imaging—showing the key dependencies of image-plane coverage, sampling pitch, SNR, and dwell benchmarks.

Figure 2: Einstein-ring extraction as an optical-propagation challenge—illustrating focal-plane geometry, separation challenges, and effective aperture constraints.

Systematic Limitations and Error Budget

Contrary to the simple notions of lens-based 'super-resolution', this benchmark demonstrates that the dominant limitations on SGL imaging are not directly imposed by the solar gravitational PSF's width, but by a confluence of systematic factors:

• Temporal Sampling and Planet Variability: Rotational smearing and ephemeral cloud cover destroy the stationarity assumption of simple deconvolution, necessitating time-tagged, phase-registered coaddition and dynamic inversion.
• Solar Corona and Detector Calibration: The coronal background, although subtractable in mean, introduces shot noise orders of magnitude higher than the exoplanet signal ($Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$ for the fiducial case), making ppm-level calibration residuals essential.
• Metrology and Navigation: Meter-level knowledge of image-plane coordinates is required due to the $\sim10$ m pixel scale, strongly constraining navigation and post-facto coordinate registration.
• Operator Fidelity: Accurate knowledge and modeling of the SGL PSF (including multipole and extended-Sun effects) are prerequisite for unbiased reconstruction.

  Figure 3: SGL optical response diagnostics and the derived inverse problem's conditioning.

  

  Figure 4: Instrument-plane annular photometry—analyzing signal and coronal backgrounds in practical extraction geometry.

  

  Figure 5: Forward-model perturbations illustrating injected noise, systematics, and navigation error profiles.

Reconstruction, Regularization, and Cloud Mitigation

Reconstruction in this context is fundamentally ill-posed, with the $1/\rho$ SGL kernel leading to substantial noise amplification for high-spatial-frequency content. The scalar benchmark implements a regularized Wiener inverse on the guard-padded grid, with support and positivity constraints, guided by the discrepancy principle for regularization parameter selection. Finite photon statistics, navigation uncertainty, and calibration precision critically dictate the achievable spatial contrast and detail.

The analysis makes explicit that static one-pass rasterization is fundamentally incompatible with imaging dynamic, rotating, cloudy planets—even in scenarios of high photon SNR. This is exemplified by the C3 branch, where failure to account for temporal evolution manifests as catastrophic aliasing and loss of surface detail. Mitigation strategies (C3b, C3c) rely on repeated, registered, short-exposure sampling, robust handling of clouded pixels, and joint estimation of persistent and transient map components using phase registration and robust statistical coaddition.

Figure 6: Stepwise reconstructions across noise-free, smeared, temporally incoherent, and photon-noise-degraded scenarios—demonstrating dominant degradation modes.

Figure 7: Stochastic cloud-field stress test (C3c)—showing surface proxy, independent synthetic cloud epochs, and the recovered persistent map.

Figure 8: Structure similarity, error, contrast recovery, and spatial resolution ($FRC_{50}$) across all discrete simulation test cases.

Quantitative Numerical Results

In the fiducial $1\,{\rm m}$, $128\times128$ raster, Earth at $30\,{\rm pc}$ is compressed to a $1.338$ km image cylinder ($128\times128$0 m sampling pitch). The dominant photon-limited branch achieves post-reconstruction $128\times128$1 and SSIM$128\times128$2 (C8), with Fourier-ring-correlation resolution proxies of $128\times128$3–$128\times128$4 km, conditional on all mitigation factors. Elevating the photon statistics (C9) yields $128\times128$5 and SSIM$128\times128$6.

Cloud-mitigation demonstrations (C3b/C3c, $128\times128$7 phase-registered epochs per longitudinal bin) achieve recovery of persistent surface-like proxies with SSIMs of $128\times128$8–$128\times128$9 at comparable spatial resolution floors; they also quantify the required temporal redundancy, demonstrating that $Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$0 yields manageable $Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$1 yr single-spacecraft dwell per map, or $Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$2 d with 16 parallel spacecraft.

Calibration residuals on the coronal signal must not exceed sub-ppm levels, as $Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$3 ppm artifacts can completely compromise map fidelity ($Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$4 planet units; C5a failure mode). Proper metrology keeping residual jitters below $Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$5–$Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$6 m is essential to prevent spatial and photometric bias.

Figure 9: SNR/calibration error budget—demonstrating the translation between flat-field residuals, allowed signal artifacts, and practical calibration constraints.

Figure 10: Requirement sweeps—illustrating the scalar system's sensitivity to calibration, exposure, and photon-count budgets.

SGL vs Conventional Architectures and Mission Implications

The paper juxtaposes SGL imaging against theoretical conventional architectures (filled apertures, starshades, interferometers), showing that conventional telescopes require prohibitive apertures ($Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$7100 km) and integration times per spatial element, making them impractical for surface mapping at tens of parsecs. In contrast, the SGL's gain and angular response shift the burden to spacecraft propulsion, optical background suppression, instrumental calibration, platform metrology, and dynamic inversion.

Mission designs favor interleaved, multiple-spacecraft architectures capable of repeated, temporally resolved sampling to facilitate cloud/statistical averaging and to compress the observing schedule. The required photon time for surface map recovery is set by the product of image-plane samples and independent, cloud-mitigated visits, scaling inversely with collector area and directly with the number of epochs and phase bins.

Figure 11: SGL SNR build-up vs time and comparison with representative conventional remote-imaging scalings—highlighting the orders-of-magnitude efficiency separation.

Figure 12: Temporal sampling and architecture requirements showing trade-offs in dwell time, pixel size, and spacecraft count for mission design.

Limitations and Directions for Future Work

While the scalar, aperture-averaged model sets physically transparent requirements and demonstrates recoverability under stringent calibration and temporal control, the following major extensions are needed:

• Full wave-optical PSF incorporation (solar multipoles, plasma, extended Sun).
• Scene models including time-tagged, multispectral general-circulation radiative transfer outputs for realistic clouds and atmospheric structure.
• End-to-end detector modeling, including cosmic rays, persistence, and calibration error propagation.
• Closed-loop navigation/ephemeris estimation.
• Dynamic, Bayesian inversion of time-tagged, phase-resolved data streams jointly estimating persistent maps and nuisance states.
• Physical modeling and mitigation of host-star/exozodiacal leakage with coronagraphic/external occulter propagation.

The research pathway is thus well-specified by the operator formalism and requirements articulated in the benchmark.

Conclusion

This paper establishes, through explicit simulation and controlled degradations, that the Solar Gravitational Lens can theoretically yield resolved $Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$8–$Q_{\rm cor}/Q_{\rm exo}\approx7.7\times10^4$9 km class maps of Earth analogs at $\sim10$0 pc, conditional on stringent instrumental, calibration, and temporal requirements. Core limitations are not set by the gravitational kernel's width, but by the necessity for photon-limited calibration, dynamic inversion for temporal variability, sub-ppm background control, and precise image-plane metrology. Under the scalar model and stated mitigations, persistent surface information is statistically recoverable, but future mission planning and evaluation must incorporate multifactor dynamic models, physical scene representations, and full joint inversion methodologies.

The SGL is, therefore, a technically plausible high-resolution imaging system for exoplanetary science, provided that mission architectures can meet the quantitative requirements in calibration, temporal sampling, navigation, and dynamic inference identified by this work.