The evolution of the bar fraction and bar lengths in the last 12 billion years

TL;DR

This study extends bar fraction and bar-length investigations to 1 ≤ z ≤ 4 using JWST NIRCam CEERS data, comparing short (F200W) and long (F356W+F444W) rest-frame wavelengths to capture bars traced by different stellar populations. Through a two-step morphological approach—Sérsic-based structural fitting plus expert visual classification—the authors quantify f_bar and L_bar in a mass-complete disc sample, finding f_bar declines with redshift (≈0.16 at z~1–2 to ≈0.07 at z~3–4) while long-wavelength bars show little evolution in mean length (~3.6 kpc) and short-wavelength bars average ~2.9 kpc, with a small growth toward z≈1. Bars and discs grow in tandem; when bar length is normalised by disc size, L_bar/R_90 remains roughly constant (~0.5) from z=4 to z=0, implying coevolution of bar and disc growth. The results reveal that bars at z>1 are already long and robust, some reaching local-bar strengths, and that bar formation initiates earlier in the Universe than previously inferred, particularly in the most massive galaxies (downsizing). They also highlight wavelength-dependent biases and resolution effects that influence bar measurements, underscoring the value of multi-band JWST imaging for disentangling stellar mass structure from star formation and dust.

Abstract

We investigate the evolution of the bar fraction and length using an extended JWST NIRCam imaging dataset of galaxies at $1 \leq z \leq 4$. We assess the wavelength dependence of the bar fraction and bar length evolution by selecting a nearly mass-complete CEERS disc sample and performing visual classifications on the short (F200W) and long (F356W+F444W) wavelength channels. A similar bar fraction is observed for both samples, and combined, we find a declining bar fraction from $0.16^{+0.03}_{-0.03}$ to $0.07^{+0.03}_{-0.01}$ over the redshift range. No evolution in the F356W+F444W bar length is measured, with a mean of 3.6 kpc. A slight increase of $\sim 1$ kpc towards $z = 1$ is measured in the F200W sample, with a mean of 2.9 kpc. We find that the correlation between bar length and galaxy mass, for massive galaxies at $z < 1$, is unseen at $z > 1$. By incorporating barred galaxies at $z<1$, we show that there is a modest increase in the bar length ($\approx 2$ kpc) towards $z=0$, but bars longer than $\approx8$ kpc are only found at $z<1$. We show that bars and discs grow in tandem, for the bar length normalised by disc size does not evolve. Not only is a significant population of bars forming beyond $z = 1$, but our results also show that some of these bars are as long and strong as the average bar at $z\approx0$.

Figures (12)

• Figure 1: Stellar mass-redshift distribution of CEERS galaxies in the redshift range $1 \leq z \leq 4$. We adopt the 95% empirical completeness of Duncan_2019, originating from CANDELS data. Our sample is selected to be those with stellar masses greater than the redshift-dependent stellar mass limit (purple).
• Figure 2: Masking of neighbouring sources with SExtractor performed on the example galaxy EGS$\_1246$. Top left - JWST NIRCam F444W image with fitted elliptical isophotes. Top right - SExtractor segmentation map. Bottom left - Resulting image with identified neighbouring sources masked out. Bottom right - Masked image with newly fitted elliptical isophotes. This figure shows the influence of nearby sources on the orientation and elongation of isophotal ellipse fits.
• Figure 3: Distribution of the best-fit structural parameters from a single 2D Sérsic function fitting using Imfit, performed on the optimised samples for galaxies between $1 \leq z \leq 4$. Models fit to F356W and F444W NIRCam images are filled distributions, and those as a result of fits to F200W NIRCam images are unfilled distributions with black edges. Left: single Sérsic best-fit ellipticity, $e$. Middle: single Sérsic best-fit Sérsic index, $n$. Right: single Sérsic best-fit effective radius, $r_{e}$.
• Figure 4: Evolution of the bar fraction in disc galaxies. At high-$z$ the bar fraction is found in three redshift bins, $1 \leq z < 2$, $2 \leq z < 3$ and $3 \leq z \leq 4$, using the visual classification of the images from the JWST NIRCam filters F356W+F444W (red stars) and F200W (blue stars) and combining these results to find a total bar fraction (yellow stars). Error bars in $f_{bar}$ are the $1\sigma$ bimodal interval, and the shaded area is the upper and lower bounds of the bar fraction (see text for details). Dashed horizontal error bars show the full range in $z$ of the identified bars, while thick horizontal solid lines show the corresponding 25%-75% inter-quartile range. The insert shows the bar fraction for the breakdown of strongly (right-pointing triangle) and weakly (left-pointing triangle) barred galaxies. The results of Paper 1 are black unfilled squares (HST) and stars (JWST), and in grey are the JWST bar fraction of Guo_2025 (diamond), Salcedo_2025 (circle) and Géron_2025 (square).
• Figure 5: Stellar mass - redshift distribution of barred galaxies found by visual classification. For the independent studies using NIRCam filters F200W (blue) and F356W+F444W (red), galaxies are defined as strongly barred (right-pointing triangle, dotted) and weakly barred (left-pointing triangle, dashed). For the strongly and weakly barred galaxies, we show the mean stellar mass across the redshift range as dotted and dashed lines, respectively. The 95% empirical completeness adopted from Duncan_2019 (black step function) sets the lower limit to the stellar mass distribution.