Image reconstruction with the JWST Interferometer

TL;DR

This work addresses non-linear detector systematics in JWST/NIRISS AMI by introducing a differentiable end-to-end forward-model (amigo) and a regularised maximum-likelihood image reconstruction framework (dorito). It implements two reconstruction pathways—image-plane and DISCO-based—within a gradient-based optimisation context to recover high-fidelity, diffraction-limited images of diverse targets (NGC 1068, Io, WR 137). The study demonstrates that forward-modelling can mitigate Brighter-Fatter Effect-related distortions and yields reconstructions that agree with literature, while highlighting the strengths and limitations of each method under varying wavefront error conditions. The open-source release of amigo and dorito, together with public AMI datasets, provides a flexible platform for advancing AMI imaging and broader interferometric applications.

Abstract

Flying on board the James Webb Space Telescope (JWST) above Earth's turbulent atmosphere, the Aperture Masking Interferometer (AMI) on the NIRISS instrument is the highest-resolution infrared interferometer ever placed in space. However, its performance was found to be limited by non-linear detector systematics, particularly charge migration - or the Brighter-Fatter Effect. Conventional interferometric Fourier observables are degraded by non-linear transformations in the image plane, with the consequence that the inner working angle and contrast limits of AMI were seriously compromised. Building on the end-to-end differentiable model & calibration code amigo, we here present a regularised maximum-likelihood image reconstruction framework dorito which can deconvolve AMI images either in the image plane or from calibrated Fourier observables, achieving high angular resolution and contrast over a wider field of view than conventional interferometric limits. This modular code by default includes regularisation by maximum entropy, and total variation defined with $l_1$ or $l_2$ metrics. We present imaging results from dorito for three benchmark imaging datasets: the volcanoes of Jupiter's moon Io, the colliding-wind binary dust nebula WR 137 and the archetypal Seyfert 2 active galactic nucleus NGC 1068. In all three cases we recover images consistent with the literature at diffraction-limited resolutions. The performance, limitations, and future opportunities enabled by amigo for AMI imaging (and beyond) are discussed.

Figures (7)

• Figure 1: Interferograms of the science targets, calibrators, and deconvolved images. Top: Images of the interferograms of science targets NGC 1068, Io, and WR 137. These are slope images, i.e. final group subtract second-to-final group. They are noticeably resolved by comparison to the middle row of psf calibrators corresponding to the above sources. Bad pixels not used in the fit are set to black. Bottom: rml image reconstructions of all three targets. The circle in the lower left of each panel indicates $\lambda/2D$. The colour tables for each image is a power stretch ($\gamma=0.4, 0.8, 0.3$ respectively for each target left to right). NGC 1068 is clipped to $25\%$ of its peak brightness in the F480M filter. Note WR 137 is displayed twice as zoomed-in as the other images due to its small angular size. NGC 1068 and WR 137 are both presented as RGB false colour images weighted by the recovered flux values in each filter. However as WR 137 was only observed in the F480M and F380M filters, the green channel was assigned the average of the other two channels. The predominant white hue of these two images is a sign that the same structure is independently recovered in all three bands, a sign of successful deconvolution of a source without strongly wavelength-dependent features.
• Figure 2: L-curve diagram used to select the optimal regularisation parameter $\lambda_{\text{TV}}$ for Io. Each point on the curve is the balance between regulariser term $\mathcal{R}$ and likelihood term $\mathcal{L}$ of a converged image reconstruction for a different regularisation hyperparameter value $\lambda_{\text{TV}}$. Shown are several reconstructed images corresponding to different points along the curve. The effects of TV regularisation on the image can be seen to strengthen with increasing $\lambda_{\text{TV}}$ (e.g. as regularisation increases, reducing ringing artefacts and forming plateaux of uniform flux with sharp edges). The optimal value for $\lambda_{\text{TV}}$ is selected from the elbow of the curve (the third image).
• Figure 3: Flow diagram depicting the Method 1 image plane-based reconstruction process. The source is modelled via a convolution with a source distribution array and is fit to the ramp data with gradient descent. The final image is obtained from the science source distribution after a specified number of gradient descent iterations are complete.
• Figure 4: Flow diagram depicting the Method 2 DISCO-based image reconstruction process, broken into three stages. In the first, the source distribution is forward-modelled with complex visibilities and then transformed to the DISCO basis in the second. This is identical to the fitting processes in Desdoigts2025, and is described in further detail there. Thirdly, the image reconstruction takes place as a source distribution array is fit to the reduced DISCOs.
• Figure 5: A comparison of the NGC 1068 images recovered in AMI with the lbti observations convolved by Isbell25, updated with recent data by private communication. The three bands of AMI data in F380M, F430M, and F480M are represented as blue, red, and green in a colour image, with the lbti 8.7 $\mu$m image flux overlaid as contours. The bright parts of the AMI colour image are close to white, indicating a consistent recovery across all three bands, and they track the bright parts of the lbti image very closely, indicating both sets of data are independently recovering the same structure.