First Demonstration of Kernel Phase Interferometry on JWST/MIRI: Prospects for Future Planet Searches Around Post Main Sequence Stars

Kernel phase interferometry (KPI) is a post-processing technique that treats a conventional telescope as an interferometer by accurately modeling a telescope pupil as an array of virtual subapertures. KPI provides angular resolution within the diffraction limit by eliminating instrumental phase errors to first order. It has been successfully demonstrated to boost angular resolution on both space- and ground-based observatories, and is especially useful for enhancing space telescopes, as their diameters are smaller than the largest ground-based facilities. Here we present the first demonstration of KPI on JWST/MIRI data at 7.7 microns, 10 microns, and 15 microns. We generate contrast curves for 16 white dwarfs from the MIRI Exoplanets Orbiting White dwarfs (MEOW) Survey, finding significantly deeper contrast at small angular separations compared to traditional imaging with JWST/MIRI, down to within $λ$/D. Additionally, we use our KPI setup to successfully recover four known companions orbiting white dwarfs and brown dwarfs. This analysis shows that at these wavelengths KPI can uniquely access the orbital parameter space where inward-migrating post-main-sequence giant exoplanets are now thought to exist. We discuss the prospects for applying KPI to a larger sample of white dwarfs observed with JWST, increasing the volume of directly imaged close-in post-main-sequence exoplanets.

💡 Research Summary

This paper presents the first application of kernel phase interferometry (KPI) to data from the Mid‑Infrared Instrument (MIRI) on the James Webb Space Telescope (JWST). KPI treats a conventional telescope pupil as a dense interferometric array of virtual sub‑apertures, allowing the extraction of “kernel phases” that are immune to first‑order wavefront errors. By combining JWST’s high Strehl ratio (≥0.85) with the long wavelengths accessible to MIRI (7.7 µm, 10 µm, and 15 µm), the authors demonstrate that KPI can achieve diffraction‑limited angular resolution and significantly improved contrast at separations well inside the classical λ/D limit.

The authors built a realistic MIRI pupil model using the STPSF (formerly webbpsf) package and the XARA Python toolkit. A “grey” transmission model was adopted with a transmission cutoff of 0.70, and the sub‑aperture pitch was set to 0.35 m after visual inspection of the (u,v) sampling density. This configuration yields an effective field of view that scales with λ/s, guiding the choice of image cropping for each wavelength band (≈4.4″ for 7.7 µm, ≈5.5″ for 10 µm, and up to ≈7″ for 15 µm).

The data set comprises publicly available MIRI observations from three JWST programs: the MEOW survey (PID 4403) covering 17 white dwarfs (16 of which are analyzed in the F770W filter), two brown dwarfs (W0146 and W1711, PID 2124), and two white dwarfs with known wide companions (WD 1202‑232 and WD 2105‑82, PID 1911). Standard reduction steps include background subtraction with photutils, bad‑pixel correction, and application of a super‑Gaussian window (order 4, HWHM≈0.5 λ/s) to suppress noise far from the PSF core. After Fourier transformation, the sampled complex visibilities are projected onto the kernel phase space using the pre‑computed K matrix.

Because the observations lack dedicated calibrator stars or multiple roll angles, the kernel phases are uncalibrated; errors are estimated from the standard deviation across all kernel phases and treated as uniform. Companion detection proceeds by generating a three‑dimensional grid of binary models (separation, position angle, contrast), computing χ² between model and data kernel phases, and identifying the minimum. Detection limits are defined at Δχ² = 25 (≈5σ). Contrast curves for the 16 MEOW white dwarfs reach 5–6 mag at separations down to λ/D (≈250 mas, ≈3 AU at the median distance of 12 pc), a factor of ~4 improvement over conventional MIRI imaging, which typically attains similar contrast only beyond ~580 mas.

Four known companions are recovered: two brown dwarf companions (≈150 mas, ≈0.07 mag contrast; ≈701 mas, ≈0.07 mag contrast) and two white dwarf companions (≈1.1″, ≈4.45 mag; ≈2.1″, ≈3.83 mag). For three of the four objects, KPI-derived separations, position angles, and contrasts agree with direct‑imaging (DI) results within 1σ. The closest brown dwarf companion (W0146) shows larger discrepancies (≈40 mas, ≈37°, ≈1 mag), likely due to systematic errors in the uncalibrated kernel phases and higher‑order noise. Correlation plots of measured versus model kernel phases confirm strong agreement for the well‑behaved cases and highlight the need for proper kernel‑phase calibration.

The discussion emphasizes that KPI’s strength lies in probing the inner working angle (IWA) down to λ/D, where post‑main‑sequence (post‑MS) planets are predicted to reside after stellar mass loss and orbital migration. At MIRI’s long wavelengths, even cool (few hundred Kelvin) giant planets emit sufficient thermal radiation to be detectable, opening a new parameter space inaccessible to ground‑based facilities. The authors outline future improvements: employing angular differential imaging (ADI) via multiple roll angles, reference‑star differential imaging (RDI) with contemporaneous calibrators, refining the pupil pitch and transmission threshold, and incorporating higher‑order wavefront error models. Scaling KPI to the full MEOW sample (and beyond) could dramatically increase the census of close‑in, low‑mass companions to white dwarfs, providing critical constraints on planetary survival, dynamical evolution, and the fate of planetary systems after the host star leaves the main sequence.