AGN--Host Galaxy Image Decomposition with JWST

Callum Dewsnap, Pauline Barmby, Sarah C. Gallagher

First listed 2025-10-31 | Last updated 2025-10-31

Abstract

The ability to disentangle the light of an AGN from its host galaxy is strongly dependent on the spatial resolution and depth of the imaging. As the capabilities of imaging systems improve with time, confirming that our standard techniques adequately model the increasingly complex structures unveiled is essential. With JWST providing unprecedented image quality, we can test how measurements of galaxy morphology vary with the choice of point-spread function (PSF) and fitting software. We perform two-component Sérsic+PSF fits of the surface brightness profiles of 87 X-ray AGNs $(0.1 < z < 4)$ from the CEERS survey. We create model PSFs for NIRCam F115W imaging using both photutils and PSFEx. We find that PSFEx models consistently fail to match the radial profile of typical point sources within our sample. We then perform AGN--host decompositions on each source by creating Sérsic+PSF models using both Galfit and AstroPhot. We find that Galfit and AstroPhot converge to different regions of the parameter space, providing consistently differing host galaxy properties. While we can measure the AGN and host magnitudes accurately, we find that the host galaxy morphological parameters are not well-determined -- the Sérsic index and effective radius are strongly covariant. Significant changes in the host galaxy parameters do not correspond to changes in the statistical quality of fit, nor to significant changes in the model's radial profile. These results indicate that the Sérsic profile does not uniquely well-represent typical AGN host galaxies in extragalactic survey fields. We also provide recommendations for studies of AGN hosts comparable to ours.

Short digest

Tests JWST/CEERS F115W AGN–host decompositions for 87 X-ray AGNs (0.1<z<4) using Sérsic+PSF models with PSFs from photutils vs PSFEx and fits from Galfit vs AstroPhot. PSFEx PSFs systematically mis-reproduce point-source radial profiles, while Galfit and AstroPhot converge to different regions of parameter space, yielding inconsistent host properties. AGN and host magnitudes are robust, but host morphology is poorly constrained: Sérsic n and Re are strongly covariant and large parameter shifts barely affect χ² or radial profiles. Conclusion: single-Sérsic models do not uniquely represent typical AGN hosts in JWST extragalactic fields, and the authors offer practical recommendations.

Key figures to inspect

• Fig. 3 — Compare PSF FWHM distributions for photutils (red) vs PSFEx (blue) across JWST and HST; note systematic offsets and where PSFEx broadens/narrows relative to the point-source ensemble and external benchmarks (green stars).
• Fig. 4 — Inspect the 2D PSF stamps and 1D radial profiles: quantify where PSFEx departs from the empirical point-source profile (core/wings) and how photutils better tracks the gray point-source stack.
• Fig. 1 — Check the Lx–z plane and flux-limit line to see the dynamic range of rest-frame 2–10 keV luminosities and redshifts feeding the 87-object sample; verify spectroscopic vs photometric-z coverage.
• Fig. 2 — Evaluate point-source vetting: artifacts in the HST/ACS F814W core that contaminate PSF training, motivating strict selection for reliable JWST/HST PSFs.
• (If available in the main text) Decomposition residual/radial-profile comparison — verify that markedly different n and Re solutions produce nearly indistinguishable residuals/profiles, illustrating the covariance-driven non-uniqueness.