Earliest Galaxy Evolution in the CANUCS+Technicolor fields: Galaxy Properties at $z\sim10-16$ seen with the Full NIRCam Medium and Broad Band Filters

TL;DR

Using a rich set of JWST/NIRCam MB+BB observations from CANUCS and TEC, this work identifies eight $z>9.5$ galaxy candidates across three independent sightlines, including a highly robust $z\sim15.4$ source. It derives a redshift-evolving UV luminosity function with $\eta=-0.21^{+0.11}_{-0.12}$ and a normalization $\log_{10}\phi_{\star,0}=-4.89^{+0.17}_{-0.20}$ (fixing $\alpha=-2.1$, $M_{\rm UV}^*=-21.0$), revealing a moderate decline in abundance from $z\sim11$ to $z\sim15$ and substantial field variance. The study demonstrates that MB filters are essential to suppress low-$z$ interlopers and to recover faint, red high-$z$ galaxies, showing that BB-only selections can overestimate the UV LF normalization by up to $\sim0.6$ dex and bias other statistics like the size–$M_{\rm UV}$ relation and $\beta_{\rm UV}$. It also documents how cosmic variance and interloper contamination complicate early galaxy evolution inferences, and argues for wide-area, MB+BB JWST surveys (e.g., MINERVA) to robustly chart galaxy formation at $z>10$.

Abstract

We present a sample of $z_{\rm phot}\sim10-16$ galaxies by exploiting one of the richest JWST NIRCam imaging data, taken in the CANUCS survey in Cycle 1 and the Technicolor (TEC) survey in Cycle 2. The combination of the CANUCS+TEC provides multi-epoch, deep NIRCam images in all medium bands (MBs) and broad bands (BBs) onboard NIRCam (22 filters in total), over $\sim23\ {\rm arcmin}^2$ in three independent lines of sight. We select high-$z$ galaxy candidates based on photometric redshifts, and obtain eight candidates at $z\sim10-16$, including a very robust candidate at $z\sim15.4$. The ultraviolet (UV) luminosity function (LF) from our sample is consistent with previous JWST studies showing a scatter of $\sim0.6$ dex across the literature, marking the significance of the field-to-field variance in interpreting galaxy abundance measurements at $z>10$. We find that the UV LF moderately evolves at $z>10$, and the LF normalization and the luminosity density decline by a factor of $\sim7$ from $z\sim11$ to $z\sim15$, indicating less steep evolution than $z<11$. We highlight the importance of MB filters, not only to minimize the contamination by low-$z$ interlopers but also to maximize the completeness. In particular, faint and less blue galaxies could be missed when the sample is built solely on BB data. The contamination and incompleteness of BB-only selected samples can bias our views of earliest galaxy evolution at $z>10$, including the UV LF by $\sim0.6$ dex, the size-magnitude relation by $\sim0.6$ dex, and the UV slope-magnitude relation by $Δβ_{\rm UV}\sim-0.3$.

Figures (13)

• Figure 1: Illustration showing the NIRCam filter transmission curves and the Lyman-$\alpha$ break feature in $z>10$ galaxies. The MB filters are shown by filled curves.
• Figure 2: The highest redshift candidate of CANUCS-3214552 at $z_{\rm phot}\sim15.4$. Top: cutouts of the source in all NIRCam filter images, except for the two narrow band filters. Cutouts are $1.\!\!^{\prime\prime}6$ on the side. Image scalings are homogeneous. Bottom: the SED and the PDZ of CANUCS-3214552. The SED shows a clear dropout between F210M and F200W, and the sharp drop between this narrow wavelength range strongly supports the high-$z$ solution template (black solid curve in the bottom left panel) and results in negligible low-$z$ probability (blue curve in the bottom right panel). There is no variability between Cycle 1 and Cycle 2 observation with 1 year separation (blue and red squares in bottom left), which rules out the possibility of $z\sim4$ SN contamination. If MB data were not available, the SED could also be fit by a low-$z$ solution (gray curve in bottom left) and the PDZ would have a non-negligible secondary peak at $z\sim4$ (gray dashed line in bottom right). Only multiple-epoch rich MB+BB observations can secure this galaxy as a very robust $z\sim15$ galaxy candidate.
• Figure 3: High-$z$ galaxy selection completeness in the absolute UV magnitude vs redshift plane. Yellow points present our main $z>9.5$ galaxy sample selected by criteria (\ref{['eqn:selection']}). Our selection is 50 % complete down to $M_{\rm UV}\sim-18.7$ mag over the full redshift range of $10\lesssim z\lesssim16$. Though, only one galaxy is found at $z>12$ whereas seven are found at $z<12$, which indicates redshift evolution of the UV LF across this redshift range.
• Figure 4: The UV LFs at $z\sim11$ (left) and $z\sim14$ (right) from CANUCS+TEC data. Black solid curve is the best estimation of the redshift-evolving UV LF obtained in Sec. \ref{['subsec:LFparams']} at $z=10.5$ (left) and $z=14$ (right), with the red-shaded area showing 16th- to 84th-percentile. Red filled circles with black error bars present the binned UV LFs in the $9.5<z<12$ bin (left) and $12<z<16$ bin (right) measured in Sec. \ref{['subsec:binLF']}. UV LF measurements in literature at similar redshifts are also shown for comparison Perez_Gonzalez2023Adams2024ApJDonnan2024Robertson2024ApJWillott2024ApJPerez_Gonzalez2025Castellano2025arXiv.
• Figure 5: The redshift evolution of cosmic UV luminosity density. Black solid line is the median estimation from CANUCS+TEC data in this work, based on the redshift-evolving UV LF measurement in Sec. \ref{['subsec:LFparams']}, and red shaded area shows the 16th- to 84th-percentile range. Black dashed line is the $\rho_{\rm UV}$ measurement from JOF observation Robertson2024ApJ, which uses similarly rich NIRCam MB+BB filters in high-$z$ galaxy selection. For comparison, literature $\rho_{\rm UV}$ measurements from other JWST observations are shown filled small symbols Perez_Gonzalez2023Adams2024ApJDonnan2024Willott2024ApJHarikane2025ApJ. Gray dash-dot curve marks the prediction of $\rho_{\rm UV}$ evolution assuming a constant star-formation efficiency at $z>6$ by Harikane2022. Dotted curves present predictions by several theoretical simulations (THESAN-ZOOM, Kannan2025; Universe Machine, Behroozi2020; FLARES, Vijayan2021Wilkins2023; SC SAM, Yung2024). The $\rho_{\rm UV}$ evolution slope in our work at $z\gtrsim10$ (black solid line) is somewhat less steep than previous measurements at $z\sim8$ and in a great agreement with that from JOF data, and starts to deviate from the constant SFE prediction above $z\sim11$.