TL;DR
This paper introduces a novel wavefront-based method to precisely estimate the Point Spread Function (PSF) across an entire telescope's field of view using only observations of field stars. This approach bypasses the need for risky in-flight telescope maneuvers, offering a more robust and efficient way to characterize telescope aberrations for missions like Euclid and WFIRST.
What Problem Does This Paper Address?
Future astronomical missions, particularly those focused on weak gravitational lensing like Euclid and WFIRST, require extremely precise measurements of galaxy shapes. The accuracy of these measurements is heavily dependent on understanding the telescope's Point Spread Function (PSF), which describes how light from a point source is spread out. A significant challenge is the "anisoplanatism" of wide-field telescopes, meaning the PSF varies across the field of view. Traditionally, estimating the PSF at every point has been difficult. While field stars can provide PSF measurements, they are often at random positions, and interpolating to the exact positions of galaxies is problematic. Furthermore, in-flight calibration of telescope aberrations, which directly influence the PSF, is often risky and complex. This paper tackles the problem of simultaneously estimating these varying aberrations and thus the PSF across the entire field of view using readily available star observations.
Methods
The core of the proposed method is a wavefront-based approach that models the telescope's optical system as a series of interfaces (mirrors). For each interface, aberrations are described using Zernike polynomials, and these aberration parameters are the unknowns to be estimated. The forward model simulates the propagation of starlight (modeled as a plane wave) through the telescope, accounting for mirror shapes and aberrations, and predicts the intensity pattern at the detector. This model is non-linear with respect to the aberration parameters.
The estimation problem is framed as a constrained optimization problem. The authors employ an Augmented Lagrangian approach, breaking down the problem into two main steps:
- Phase Retrieval Problem: An inner minimization step that estimates intermediate values related to the complex amplitude of the light at the detector plane. This is solved using a proximity operator.
- Tomography Problem: An outer minimization step that uses these intermediate values to refine the estimates of the telescope's aberration parameters. This is solved using a continuous optimization routine (VMLMB) with derivatives computed via recursive back-propagation.
This "back-propagating" of errors allows for the simultaneous estimation of aberrations across all mirrors and for all stars, effectively solving for the entire wavefront.
Datasets
The authors tested their algorithm using simulated data. They simulated a Ritchey-Chrétien telescope similar to the Hubble Space Telescope, with specific parameters for mirror diameters, curvature radii, and conic constants. Aberrations were introduced by drawing random coefficients for Zernike polynomials for the first and second mirrors. The dataset consisted of observations of 50 stars, randomly distributed across the field of view, with their fluxes adjusted to simulate realistic photon counts. The simulation also included background noise and Poisson distribution to mimic detector behavior.
Key Insights and Ideas
- Unified Aberration Estimation: The method treats telescope aberrations as a global property that can be estimated from multiple star observations across the field of view, rather than trying to estimate the PSF independently at each location.
- Leveraging Field Stars: By using field stars, the method avoids the need for dedicated calibration time or risky maneuvers like defocusing the telescope.
- Back-propagation for Optimization: The use of back-propagation to compute gradients for the optimization algorithm is crucial for efficiently updating the estimated aberration parameters.
- Joint Problem Solving: The approach elegantly combines PSF estimation and aberration characterization into a single, coherent framework.
Key Results
The paper presents qualitative results demonstrating the effectiveness of the algorithm. Figure 3 shows that the estimated PSFs (left column) closely match the "truth" PSFs (middle column) when starting from an aberration-free model. The "No Aberration" column shows what the PSF would look like without any optical errors, highlighting the impact of the simulated aberrations. The method successfully converges to a PSF that is very similar to the ground truth, even with simulated noise and aberrations. The authors are currently working on quantitative results related to PSF ellipticity and size.
Limitations and Future Work
The authors acknowledge that this is a "proof of concept" paper and that significant work remains before applying the method to real data. Key limitations and future directions include:
- Quantitative Results: Developing robust quantitative metrics for PSF estimation accuracy.
- Real Data Processing: Adapting the algorithm to handle real-world data, which may include complexities not fully captured in simulations.
- Undersampled and Broadband PSFs: Extending the method to accurately estimate PSFs for undersampled images and across a broader range of wavelengths.
Summary updated