Galaxy-scale lens search in the PEARLS NEP TDF and CEERS JWST fields

We present four galaxy scale lenses discovered in two JWST blank-fields: the ~ 54 arcmin^2 of the PEARLS North-Ecliptic-Pole Time-Domain Field (NEP TDF) and in the ~ 90 arcmin^2 of CEERS. We perform the search by visual inspection of NIRCam photometric data, obtaining an initial list of 16 lens candidates. We down-select this list to 5 high-confidence lens candidates, based on lens modelling of the image configuration and photometric redshift measurements for both the source and the deflector. We compare our results to samples of lenses obtained in ground-based and space-based lens searches and theoretical expectations. We expect that JWST observations of field galaxies will yield approximately 1 galaxy scale lens every three to four NIRCam pointings of comparable depth to these observations (~ 9 arcmin^2 each). This shows that JWST, compared to other lens searches, can yield an extremely high number of secure lenses per unit area, with redshift and size distributions complementary to lens samples obtained from ground-based and wide-area surveys. We estimate that a single JWST pure-parallel survey of comparable depth could yield $\sim 80$ galaxy scale lenses, with a third of them having z_lens>1 and z_source>3.

💡 Research Summary

This paper presents the discovery and characterization of galaxy‑scale strong gravitational lenses in two JWST blank fields: the PEARLS North‑Ecliptic‑Pole Time‑Domain Field (NEP TDF, ~54 arcmin²) and the CEERS field (~90 arcmin²). By visually inspecting the NIRCam imaging (8 filters in NEP TDF, 7 in CEERS) combined with HST/ACS optical data (F606W, F814W), the authors identified an initial set of 16 lens candidates. Photometric redshifts for both foreground deflectors and background sources were obtained through SED fitting using EAZY and BAGPIPES, while structural parameters of the deflectors were measured with GALFIT.

To confirm the lensing nature, each candidate was modeled with a parametric mass model: a Singular Isothermal Ellipsoid (SIE) plus external shear for the deflector, and an elliptical Sérsic profile for the source. Lens modeling was performed with the lenstronomy package, employing particle‑swarm optimization to find initial parameters and emcee‑based MCMC to sample the posterior. Four candidates were successfully reproduced with this approach, showing Einstein radii between 0.3″ and 0.8″, deflector redshifts z_lens≈0.9–2.1, and source redshifts z_source≈2.5–5.6.

Recognizing that a central mask (≈0.3″) can hide images close to the deflector core and that a single Sérsic profile may be insufficient for complex lens light, the authors applied an alternative method based on Multi‑Gaussian Expansion (MGE) for both lens‑light subtraction and source reconstruction, using pyautolens. This more flexible modeling recovered an additional high‑confidence lens (PEARLS J172339.6+654936) and provided a qualitatively good fit for a CEERS system that was otherwise excluded.

Overall, five high‑confidence galaxy‑scale lenses are reported (four from the SIE/lenstronomy pipeline plus one from the MGE approach). The lenses are characterized by relatively small Einstein radii, higher redshifts, and modest angular sizes compared with samples from ground‑based surveys, reflecting JWST’s superior spatial resolution (0.03″/pixel) and depth (28.5–29.3 AB mag in 0.32″ apertures).

Statistically, the authors find roughly one lens per three to four NIRCam pointings (~9 arcmin² each), corresponding to a surface density of ~0.25 lenses per pointing. This detection efficiency exceeds that of wide‑area optical surveys (Euclid, LSST, Roman) by an order of magnitude, owing to JWST’s ability to resolve faint, high‑redshift sources behind relatively low‑mass deflectors. Extrapolating to a pure‑parallel JWST program covering ~300 NIRCam pointings at comparable depth, they predict ~80 galaxy‑scale lenses, with about one‑third at z_lens > 1 and z_source > 3. Such high‑z lenses are valuable for probing the inner mass profiles of early‑type galaxies, studying the evolution of the velocity‑dispersion function, and measuring cosmological parameters via time‑delay analyses.

The paper also discusses the role of visual inspection versus automated machine‑learning searches. While CNNs, SVMs, and self‑attention models have shown promise, the authors argue that human visual inspection still yields higher purity and completeness, especially when combined with sophisticated modeling. They suggest that future JWST parallel surveys, coupled with machine‑learning pre‑selection, could dramatically increase the lens sample size while retaining the high confidence achieved here.

In conclusion, JWST’s deep, high‑resolution NIR imaging enables the discovery of a new regime of galaxy‑scale strong lenses—smaller angular scales and higher redshifts—complementary to existing ground‑based and space‑based samples. The five lenses presented here demonstrate JWST’s potential to build statistically robust lens catalogs that will advance our understanding of galaxy evolution and cosmology.