Unveiling Extended Components of 'Little Red Dots' in Rest-Frame Optical

Abstract

Recent JWST observations have revealed a population of red, compact, high-redshift objects called 'Little Red Dots'(LRD), whose host components have remained largely unconstrained, possibly due to their extreme compactness. Current morphological studies have been limited by small samples, as well as by insufficient imaging depth, which may not allow reliable separation between point-like and extended components, leaving the existence and properties of extended components in LRD largely unconstrained. Here, we perform the image stacking analysis of 217 LRDs in four NIRCam bands, representing the largest and homogeneous sample observed from COSMOS-Web survey to date. Our results reveal an unambiguous detection of faint extended emission in the F444W band, with a typical size of ~200 parsecs and magnitude of ~27.7 AB at z~6.5. We perform four-band photometric SED fitting based on galaxy templates and derive a stellar mass of 8.91+-~0.1 logM_sun. Given this stellar mass, the host galaxy is compact, i.e., ~2.5 times smaller than star-forming populations at similar mass, and the typical black hole mass of LRDs is elevated by ~1.5 dex above the local MBH-M* relation. This work provides direct observational evidence for the existence of LRD host galaxies and offers crucial insights into the growth of the host galaxy and the co-evolution of galaxies and their black holes within the first billion years after the Big Bang.

Figures (11)

• Figure 1: Imaging stacking analysis reveals the underlying extended emission that is not clearly detectable in individual PS-only decompositions.Top left: An example of 2D decomposition for an individual LRD using a PS-only model. From left to right: original LRD image (data), model image generated from the PSF (model), residual image after subtracting the point source component (data$-$PS), and the normalized residual. Bottom left: Illustration of the stacking procedure. Residuals (data$-$PS) from 217 LRDs are aligned and averaged to produce a stacked image that enhances the visibility of low-surface-brightness extended emission in the outskirts. Right: The SNR map of the residual feature. The standard deviation image (error map) is computed by measuring pixel-by-pixel dispersion across different PSFs at the same location and corresponding background noise. The stacked data$-$PS image is used as the signal in this calculation.
• Figure 1: Distribution of magnitude and redshift for the selected LRD sample in four bands. The sample includes 217 sources satisfying the criteria F115W - F150W $< 0.8$, total magnitude $< 30$, and $\chi^2 < 2$. Red, yellow, green, and blue dots represent F444W, F277W, F150W, and F115W, respectively. The sample spans a redshift range of $z = 5$–9, with source brightness in the short-wavelength bands approaching the detection limit.
• Figure 2: Imaging modeling of stacked image across four NIRCam bands. Stacked data-PS residual image based on PS + Sérsic fitting, representing the average extended emission of LRD. Each panel corresponds to one of the four NIRCam filters (F115W, F150W, F277W, and F444W), illustrating the wavelength-dependent morphology of the mean extended component. Top: Stacked image across the four NIRCam bands (averaged). A scale bar of 0.5$\hbox{$^{\prime\prime}$}$ corresponds to $2.8$ kpc at $z\sim6.45$. Middle: Best-fit models using a single Sérsic model. Third row: Corresponding normalized residuals. Bottom: Azimuthally averaged light profiles of the stacked data (blue) compared to the PSF (orange).
• Figure 2: Distribution of individual LRD extended emission to total flux ratio and the redshift evolution.Left: Histogram of the LRD extended emission to total flux ratio. The step histograms represent the distribution of ratio within 10% intervals. The dashed lines mark the median values of individual LRD, while the triangle symbols indicate the extended emission to total flux ratio of the corresponding averaged image in each band using the stacking method. Right: The position of each scatter point represents the mean extended emission to total flux ratio in a given redshift bin, while the color of the solid lines distinguishes different filters. The scatter points are color-coded according to the number of sources in each bin, as shown by the colorbar, and their marker size is proportional to the number of sources, such that larger markers indicate bins with more sources.
• Figure 3: SED fitting of the extended emission using four-band photometry based on galaxy templates via Bagpipes. Red data points with error bars show the inferred host fluxes. The blue region indicates the 1$\sigma$ uncertainty range of the SED templates derived from Markov-Chain Monte-Carlo (MCMC) sampling, represents the stellar template, with orange diamonds marking the predicted fluxes from this template.