GLIMPSE JWST/NIRCam Survey: Cosmic Dawn

• The GLIMPSE survey is a lensing-assisted ultra-deep imaging program that pushes JWST/NIRCam to detect galaxies with intrinsic UV magnitudes as faint as –12.
• It refines the ultraviolet luminosity function across 6<z<16 by revealing a nearly constant faint-end slope and a shallow decline in the cosmic star formation rate density.
• Robust multi-band photometry and detailed strong-lensing models enable resolved studies of compact star-forming substructures and validate tensions with pre-JWST theoretical predictions.

Searching arXiv for the GLIMPSE JWST/NIRCam Survey papers and related context. Search query: GLIMPSE JWST NIRCam Survey Abell S1063 ultraviolet luminosity function cosmic dawn The GLIMPSE JWST/NIRCam Survey is a lensing-assisted ultra-deep imaging program centered on the galaxy cluster Abell S1063 that uses JWST/NIRCam to probe early galaxy formation, the ultraviolet luminosity function (UVLF), and the sources of cosmic reionization across approximately $6$M_{\mathrm UV}\simeq -12$, substantially below the regime accessible to comparable blank-field surveys, while also enabling resolved studies of compact star-forming structures and searches for rare highly magnified phenomena (Atek et al., 10 Nov 2025). Early GLIMPSE analyses reported two galaxy candidates at $15.7Kokorev et al., 2024). Subsequent GLIMPSE work extended the survey’s UVLF constraints across $z\sim 9$–15 using 105 galaxy candidates and found a nearly constant faint-end slope together with a shallow decline in the cosmic star formation rate density (SFRD), further sharpening the tension with many pre-JWST theoretical expectations (Chemerynska et al., 29 Sep 2025).

1. Survey definition and scientific scope

GLIMPSE was designed to push JWST’s NIRCam to its limits by combining ultra-deep imaging with strong gravitational lensing. Its stated objectives are to probe the extreme faint end of the rest-frame ultraviolet luminosity function over $6resolve star-forming substructure down to $\sim 10$ pc scales in early galaxies while also searching for highly magnified transients (Atek et al., 10 Nov 2025).

The survey targets Abell S1063 (AS1063), a Frontier-Fields cluster at $z=0.348$. The field choice was motivated by a very regular mass morphology, extensive VLT/MUSE spectroscopy with more than 70 multiple images with spectroscopic redshifts within the NIRCam module B footprint, and a large high-magnification area that can be covered by a single NIRCam module (Atek et al., 10 Nov 2025). A related early-science analysis described AS1063 as one of the highest-magnification Hubble Frontier Fields clusters, emphasizing that the combination of ultra-deep NIRCam imaging and cluster lensing makes it possible to probe intrinsically fainter galaxy populations than had previously been accessible (Kokorev et al., 2024).

This framing is central to the survey’s scientific significance. Most pre-JWST galaxy-formation models had been calibrated primarily against relatively UV-bright systems, typically $M_{\mathrm UV}<-17$, whereas theoretical uncertainty grows strongly at fainter magnitudes. GLIMPSE was explicitly constructed to measure this poorly constrained regime (Atek et al., 10 Nov 2025). A plausible implication is that the survey’s novelty lies less in extending redshift reach alone than in jointly accessing high redshift and low intrinsic luminosity.

2. Observational configuration and depth

GLIMPSE obtained ultra-deep JWST/NIRCam imaging in seven broadband filters and two medium-band filters. The broad bands are F090W, F115W, F150W, F200W, F277W, F356W, and F444W; the medium bands are F410M and F480M (Kokorev et al., 2024, Atek et al., 10 Nov 2025, Chemerynska et al., 29 Sep 2025). One overview paper describes the program as obtaining 120 h of NIRCam imaging with exposure times ranging from 20 to 40 hours per filter, yielding a $5\sigma$ limiting magnitude of 30.9 AB in a $0.2$ arcsec diameter aperture (Atek et al., 10 Nov 2025). An earlier analysis of the $M_{\mathrm UV}\simeq -12$0 candidates instead reports $M_{\mathrm UV}\simeq -12$1 h of science integration, including 19 h in F200W and 23 h in F277W, with $M_{\mathrm UV}\simeq -12$2 aperture-corrected limiting depths of 30.6 AB mag uniformly in all wide bands for $M_{\mathrm UV}\simeq -12$3 apertures (Kokorev et al., 2024). Another GLIMPSE UVLF study summarizes the point-source depths as $M_{\mathrm UV}\simeq -12$4 mag (AB) in each filter (Chemerynska et al., 29 Sep 2025). These differing numerical summaries reflect distinct presentations within the GLIMPSE literature rather than a single standardized statement.

The observational strategy employed the MEDIUM8 read pattern and a dither scheme intended both to fill detector gaps and improve sampling. One description gives a 6-point primary dither to fill SW gaps plus a 4-point subpixel dither for PSF sampling (Kokorev et al., 2024). The later overview describes a 6-point intramodule dither plus a 4-point subpixel dither, placing MODULE B on the cluster core and MODULE A on a modestly lensed blank field (Atek et al., 10 Nov 2025).

A concise summary of the filter configuration as reported in the overview paper is given below.

Filter	Exposure (h)	$M_{\mathrm UV}\simeq -12$5 depth (AB, $M_{\mathrm UV}\simeq -12$6)
F090W	39.1	30.85
F115W	39.1	30.87
F150W	22.3	30.77
F200W	19.5	30.82
F277W	22.3	30.82
F356W	19.5	30.77
F410M	16.7	30.07
F444W	39.1	30.68
F480M	22.3	29.24

In that same presentation, the limiting magnitude is computed as

$M_{\mathrm UV}\simeq -12$7

where the noise is the aperture-to-aperture background RMS and $M_{\mathrm UV}\simeq -12$8 is the filter zero point (Atek et al., 10 Nov 2025).

The role of lensing is quantitatively important. For the $M_{\mathrm UV}\simeq -12$9 search, typical magnifications of $15.7boost sensitivity by $15.7Kokorev et al., 2024). In the broader survey overview, GLIMPSE is described as probing intrinsic depths beyond 33 AB magnitudes and covering an effective source-plane area of approximately $15.7$z\sim 9$0 (Atek et al., 10 Nov 2025).

3. Lensing model and survey geometry

A core structural component of GLIMPSE is its parametric strong-lensing reconstruction for Abell S1063. The model is described as a new parametric strong-lensing model, termed “Zitrin-analytic,” comprising two cluster-scale PIEMD dark-matter halos and 303 dual-PIEMD cluster galaxies, constrained by 75 multiple images of 28 background sources, 24 of them with spectroscopic redshifts (Atek et al., 10 Nov 2025). The early $z\sim 9$1 study gives an effectively identical description in slightly different wording: two PIEMD halos corresponding to the cluster core and a northeast group, plus 303 galaxy-scale DPIEs, with $z\sim 9$2 from 75 multiple images in 28 systems, 24 with spectroscopic redshifts (Kokorev et al., 2024). The final image-plane reproduction error is reported as $z\sim 9$3 (Atek et al., 10 Nov 2025).

The magnification map for $z\sim 9$4 spans from $z\sim 9$5 in low-magnification regions to $z\sim 9$6 near the critical curve, with flow lines indicating the direction of gravitational shear (Atek et al., 10 Nov 2025). In a separate GLIMPSE summary focused on $z\sim 9$7–15 galaxies, the field-wide magnification is described as ranging from $z\sim 9$8 up to $z\sim 9$9, with typical $6Chemerynska et al., 29 Sep 2025). These statements are not contradictory: they describe different regions or uses of the magnification distribution.

The lens model enters directly into de-lensing and volume estimation. For the two $6Kokorev et al., 2024). The source-plane volume over $6Kokorev et al., 2024).

This lensing strategy defines GLIMPSE’s position relative to blank-field JWST programs. Whereas blank fields constrain the brighter end with cleaner survey geometry, GLIMPSE explicitly sacrifices simplicity in favor of intrinsic depth. A plausible implication is that its strongest leverage lies in the faint-end UVLF and in magnified substructure studies rather than in bright-end number counts.

4. Source detection, photometry, and redshift selection

GLIMPSE high-redshift candidate selection combines dropout criteria with photometric-redshift fitting. In the survey overview, candidates were selected through Lyman-break criteria together with photometric redshift estimates from EAZY, using dual detection in a short-wavelength stack (F090W–F200W) and a long-wavelength stack (F277W+F356W+F444W) to maximize sensitivity to both blue and red sources (Atek et al., 10 Nov 2025). Aperture photometry was measured in circular apertures from $\sim 10$0 to $\sim 10$1, aperture-corrected to total using the F480M PSF curve of growth, with empirical noise estimates derived from 2000 empty apertures around each source. Stringent flags were used to remove detections affected by edges, diffraction spikes, or proximity to the brightest cluster galaxy (Atek et al., 10 Nov 2025).

The final “use_phot=1” catalog contains 64,828 reliable detections, from which approximately 540 galaxy candidates at $\sim 10$2 were isolated. The redshift distribution reported in the overview is 411 sources at $\sim 10$3, 114 at $\sim 10$4, 11 at $\sim 10$5, and two remarkable candidates at $\sim 10$6 with $\sim 10$7 (Atek et al., 10 Nov 2025).

For the earliest-redshift subsample, the selection was more specific. The $\sim 10$8-dropout criteria were

$\sim 10$9

$z=0.348$0

$z=0.348$1

with detection requirements of $z=0.348$2 in F277W, F356W, and F444W, $z=0.348$3 in at least one of these bands, and non-detection at $z=0.348$4 in the blue bands F090W, F115W, and F150W (Kokorev et al., 2024). These cuts initially yielded $z=0.348$5 color-selected candidates. Photometric redshifts were then estimated with EAZY using the blue_sfhz_13 template set over $z=0.348$6–30, requiring $z=0.348$7 in at least three bands, $z=0.348$8 FWHM $z=0.348$9, and no strong low-$M_{\mathrm UV}<-17$0 peak, with a BEAGLE cross-check including nebular emission (Kokorev et al., 2024). Of 15 initial F200W dropouts with robust $M_{\mathrm UV}<-17$1–19.2 solutions, seven were removed as low-$M_{\mathrm UV}<-17$2 interlopers through a $M_{\mathrm UV}<-17$3BIC $M_{\mathrm UV}<-17$4 criterion using additional dusty templates, leaving a final sample of five sources with at least one $M_{\mathrm UV}<-17$5 detection (Kokorev et al., 2024).

For the $M_{\mathrm UV}<-17$6–15 UVLF analysis, the dropout selections were formulated differently because the target redshift range was lower. The $M_{\mathrm UV}<-17$7–11 F115W-dropout criterion required

$M_{\mathrm UV}<-17$8

and

$M_{\mathrm UV}<-17$9

while the $5\sigma$0–15 F150W-dropout criterion required

$5\sigma$1

and

$5\sigma$2

These selections further required $5\sigma$3 detections in all bands redward of the break, $5\sigma$4 in at least one redward band, no detection above $5\sigma$5 in any filter blueward of the break, visual inspection for artifacts, and morphological cuts to exclude unresolved point-like brown dwarfs (Chemerynska et al., 29 Sep 2025).

Completeness was quantified with extensive mock-source insertion. Approximately 200,000 mock galaxies were inserted in the source plane over $5\sigma$6–15 and $5\sigma$7 to $5\sigma$8, with sizes drawn from the Shibuya+15 and Yang+22 size-luminosity relations. Recovery peaks at approximately 80% for $5\sigma$9 and falls to $0.2$0 by $0.2$1 (Chemerynska et al., 29 Sep 2025).

5. Galaxy properties and the $0.2$2 candidates

The first GLIMPSE paper reported the discovery of two robust galaxy candidates at $0.2$3, identified through the combination of Lyman-break selection and photometric-redshift estimates (Kokorev et al., 2024). Their absolute magnitudes are given as $0.2$4 to $0.2$5 mag, placing them in a faint intrinsic regime relative to many earlier high-$0.2$6 JWST detections (Kokorev et al., 2024).

In that analysis, the rest-frame UV absolute magnitude at approximately 1500 Å was computed from observed F277W photometry as

$0.2$7

The UV continuum slope $0.2$8 was fit under the assumption $0.2$9 over 1260–2500 Å using F277W–F356W colors (Kokorev et al., 2024). The measured $M_{\mathrm UV}\simeq -12$00 values span approximately $M_{\mathrm UV}\simeq -12$01 to $M_{\mathrm UV}\simeq -12$02, with median $M_{\mathrm UV}\simeq -12$03, and the abstract describes the sources as having blue UV slopes around $M_{\mathrm UV}\simeq -12$04, consistent with young, dust-free stellar populations (Kokorev et al., 2024).

The same work gives the UV luminosity as

$M_{\mathrm UV}\simeq -12$05

and adopts

$M_{\mathrm UV}\simeq -12$06

following Kennicutt (2012), with negligible dust correction because $M_{\mathrm UV}\simeq -12$07. The resulting UV-based star formation rates are $M_{\mathrm UV}\simeq -12$08–1.0 $M_{\mathrm UV}\simeq -12$09 (Kokorev et al., 2024).

The paper further argues that these sources are plausible progenitors of the unusually UV-bright galaxies observed by JWST at $M_{\mathrm UV}\simeq -12$10–14. Specifically, a constant SFR of approximately $M_{\mathrm UV}\simeq -12$11 beginning near $M_{\mathrm UV}\simeq -12$12 yields growth to $M_{\mathrm UV}\simeq -12$13 by $M_{\mathrm UV}\simeq -12$14–14, while maximal $M_{\mathrm UV}\simeq -12$15CDM-limited accretion following Dekel 13 gives an exponential star-formation history that also connects $M_{\mathrm UV}\simeq -12$16 faint galaxies to UV-bright $M_{\mathrm UV}\simeq -12$17 systems (Kokorev et al., 2024). This suggests continuity between faint cosmic-dawn galaxies and later bright JWST sources without invoking exotic physics, though the paper itself frames these as results of toy or tested star-formation histories rather than direct empirical demonstration.

6. Ultraviolet luminosity function and cosmic star formation density

A central GLIMPSE product is the measurement of the UV luminosity function at very faint intrinsic luminosities. In the $M_{\mathrm UV}\simeq -12$18 candidate study, the number-density estimator is

$M_{\mathrm UV}\simeq -12$19

with $M_{\mathrm UV}\simeq -12$20 mag around $M_{\mathrm UV}\simeq -12$21 (Kokorev et al., 2024). For $M_{\mathrm UV}\simeq -12$22 at $M_{\mathrm UV}\simeq -12$23, the measured number density is

$M_{\mathrm UV}\simeq -12$24

while the abstract reports the corresponding density at $M_{\mathrm UV}\simeq -12$25 as

$M_{\mathrm UV}\simeq -12$26

The same study states that this result is in clear tension with pre-JWST theoretical predictions and extends the over-abundance of galaxies previously inferred at $M_{\mathrm UV}\simeq -12$27 out to $M_{\mathrm UV}\simeq -12$28 (Kokorev et al., 2024).

For the larger GLIMPSE sample at $M_{\mathrm UV}\simeq -12$29–12, the survey overview reports binned number densities corrected for completeness and delensed effective volume and fits a Schechter function,

$M_{\mathrm UV}\simeq -12$30

At $M_{\mathrm UV}\simeq -12$31–9 the best-fit parameters are

$M_{\mathrm UV}\simeq -12$32

and at $M_{\mathrm UV}\simeq -12$33–12

$M_{\mathrm UV}\simeq -12$34

The steep $M_{\mathrm UV}\simeq -12$35 at $M_{\mathrm UV}\simeq -12$36 is reported to imply that galaxies fainter than $M_{\mathrm UV}\simeq -12$37 can contribute 50–60% of the total UV luminosity density, whereas a shallower $M_{\mathrm UV}\simeq -12$38 would imply $M_{\mathrm UV}\simeq -12$39 (Atek et al., 10 Nov 2025).

The dedicated $M_{\mathrm UV}\simeq -12$40–15 GLIMPSE UVLF analysis instead combined GLIMPSE counts with literature bright-end data and fitted a double-power law,

$M_{\mathrm UV}\simeq -12$41

Its best-fit parameters are reported in four redshift bins:

Redshift bin	$M_{\mathrm UV}\simeq -12$42	$M_{\mathrm UV}\simeq -12$43	$M_{\mathrm UV}\simeq -12$44	$M_{\mathrm UV}\simeq -12$45
$M_{\mathrm UV}\simeq -12$46	$M_{\mathrm UV}\simeq -12$47	$M_{\mathrm UV}\simeq -12$48	$M_{\mathrm UV}\simeq -12$49	$M_{\mathrm UV}\simeq -12$50
$M_{\mathrm UV}\simeq -12$51	$M_{\mathrm UV}\simeq -12$52	$M_{\mathrm UV}\simeq -12$53	$M_{\mathrm UV}\simeq -12$54	$M_{\mathrm UV}\simeq -12$55
$M_{\mathrm UV}\simeq -12$56	$M_{\mathrm UV}\simeq -12$57	$M_{\mathrm UV}\simeq -12$58	$M_{\mathrm UV}\simeq -12$59	$M_{\mathrm UV}\simeq -12$60
$M_{\mathrm UV}\simeq -12$61	$M_{\mathrm UV}\simeq -12$62	$M_{\mathrm UV}\simeq -12$63	$M_{\mathrm UV}\simeq -12$64	$M_{\mathrm UV}\simeq -12$65

The reported evolution of the faint-end slope from $M_{\mathrm UV}\simeq -12$66 at $M_{\mathrm UV}\simeq -12$67 to $M_{\mathrm UV}\simeq -12$68 at $M_{\mathrm UV}\simeq -12$69 is described as minimal, in contrast to the stronger redshift evolution inferred at lower redshift (Chemerynska et al., 29 Sep 2025).

The same paper converts UV luminosity to SFR with the standard calibration for a Salpeter IMF,

$M_{\mathrm UV}\simeq -12$70

and no dust correction, citing UV slopes $M_{\mathrm UV}\simeq -12$71 that imply low attenuation (Chemerynska et al., 29 Sep 2025). The SFRD is written as

$M_{\mathrm UV}\simeq -12$72

Integrated to $M_{\mathrm UV}\simeq -12$73, the derived values are $M_{\mathrm UV}\simeq -12$74 at $M_{\mathrm UV}\simeq -12$75, $M_{\mathrm UV}\simeq -12$76 at $M_{\mathrm UV}\simeq -12$77, $M_{\mathrm UV}\simeq -12$78 at $M_{\mathrm UV}\simeq -12$79, and $M_{\mathrm UV}\simeq -12$80 at $M_{\mathrm UV}\simeq -12$81, with best-fit redshift evolution

$M_{\mathrm UV}\simeq -12$82

When the integration limit is extended to $M_{\mathrm UV}\simeq -12$83, galaxies with $M_{\mathrm UV}\simeq -12$84 provide more than 50% of the total cosmic SFR density out to $M_{\mathrm UV}\simeq -12$85 (Chemerynska et al., 29 Sep 2025).

7. Theoretical tension, ancillary science, and public release

GLIMPSE papers emphasize a persistent mismatch between observed galaxy abundances and many pre-JWST models. For the $M_{\mathrm UV}\simeq -12$86 detections, pre-JWST semi-analytic models such as Mason 15 and Dayal 14 and hydrodynamic models such as Behroozi 19 and Wilkins 23 underpredict the observed abundance by $M_{\mathrm UV}\simeq -12$87–$M_{\mathrm UV}\simeq -12$88, whereas post-JWST-tuned models including Muñoz 23, the FFB scenario of Li 24, and FIREbox Feldmann 24 align within $M_{\mathrm UV}\simeq -12$89–$M_{\mathrm UV}\simeq -12$90, especially near the upper envelope (Kokorev et al., 2024). That work further states that the bright end of the UVLF at $M_{\mathrm UV}\simeq -12$91 appears to decline steeply, with no detections in blank fields such as CEERS, JADES, PRIMER, and NGDEEP, while lensing reveals abundant faint galaxies near $M_{\mathrm UV}\simeq -12$92 (Kokorev et al., 2024). A toy model in which a constant 30% star-formation efficiency of halo accretion persists over 100 Myr is said to reproduce a steep UVLF that tracks the halo mass function (Kokorev et al., 2024).

The $M_{\mathrm UV}\simeq -12$93–15 UVLF paper broadens the theoretical comparison. It states that many pre-JWST models, including Mason+15 and Tacchella+18, predict a steep decline in $M_{\mathrm UV}\simeq -12$94 and a steepening to $M_{\mathrm UV}\simeq -12$95 by $M_{\mathrm UV}\simeq -12$96, thereby underestimating both bright and faint galaxies. Dust-free outflow models such as Ferrara+23 improve the bright end at $M_{\mathrm UV}\simeq -12$97–12 but still underpredict the faint population at $M_{\mathrm UV}\simeq -12$98. Stochastic bursty star-formation histories in simulations such as Sun+23 and SPHINX partially alleviate the bright-end excess but do not reproduce the measured faint-end $M_{\mathrm UV}\simeq -12$99 or the mild evolution of $15.7Feedback-free starbursts as in Dekel+23 require an unphysically rising maximum star-formation efficiency from $15.7Chemerynska et al., 29 Sep 2025). The paper concludes that none of the currently discussed scenarios—increased SFE, bursty star formation, zero dust, or an evolving IMF—fully captures the observed slowly declining UVLF normalization and nearly constant $15.7Chemerynska et al., 29 Sep 2025).

Beyond the luminosity function, GLIMPSE also supports ancillary early-science programs. By exploiting magnifications up to $15.7Atek et al., 10 Nov 2025). The same paper notes a handful of faint high-$15.7Atek et al., 10 Nov 2025).

GLIMPSE DR1, released in December 2024, includes reduced JWST and HST mosaics, photometric catalogs, and strong-lensing model files. Specifically, it provides NIRCam mosaics at 20 mas/pix in the short wavelength channel and 40 mas/pix in the long wavelength channel across all nine filters, PSF-matched and bCG/ICL-subtracted HST ACS+WFC3 mosaics, photometric catalogs containing fluxes in $15.7MCMC chain samples, and magnification maps in FITS and Lenstool formats (Atek et al., 10 Nov 2025). The data are hosted at MAST under DOI:10.17909/GLIMPSE_DR1 and mirrored at the Dawn JWST Archive; the survey documentation advises use of the provided use_phot mask and per-pixel depth and magnification maps for rigorous effective-area and volume calculations (Atek et al., 10 Nov 2025).

Taken together, these results position GLIMPSE as a survey optimized for the intrinsically faint galaxy population at cosmic dawn. The combined evidence from its $15.7Kokorev et al., 2024, Chemerynska et al., 29 Sep 2025). This suggests that future observational tests of galaxy formation and reionization at the highest redshifts will depend critically on surveys that can access the same faint intrinsic regime.