Chandra Rules Out Super-Eddington Accretion Models For Little Red Dots

Andrea Sacchi, Akos Bogdan

First listed 2025-05-14 | Last updated 2025-07-29

Abstract

One of the most puzzling discoveries by JWST is the population of high-redshift, red, and compact galaxies dubbed little red dots (LRDs). Based on broad-line diagnostics, these galaxies have been argued to host accreting $10^7-10^8$ M$_\odot$ supermassive black holes (SMBHs), a claim with crucial consequences for our understanding of how the first black holes form and grow over cosmic time. A key feature of LRDs is their extreme X-ray weakness: analyses of individual and stacked sources have yielded non-detections or only tentative, inconclusive X-ray signals, except for a handful of individual cases. Although high obscuration is the most straightforward way to explain the X-ray weakness of LRDs, JWST rest-frame optical/UV spectra initially argued against the presence of Compton-thick gas clouds. Instead, several authors have proposed that LRDs are intrinsically X-ray weak due to super-Eddington accretion rates. In this work, we observationally test these tailored models by stacking X-ray data for 55 LRDs in the Chandra Deep Field South, accumulating a total exposure time of nearly 400 Ms. Despite reaching unprecedented X-ray depths, our stack still yields a non-detection. The corresponding upper limits are deep enough to rule out current super-Eddington accretion models, and are compatible only with extremely high levels of obscuration ($N_{\rm H}\gtrsim10^{25}$ cm$^{-2}$). To explain the X-ray weakness of LRDs, we therefore speculate that the SMBHs in these systems are neither as massive nor as luminous as currently believed.

Short digest

Stacks Chandra data for 55 JADES/NGDEEP-selected little red dots in the 7 Ms CDF-S, reaching ~390 Ms effective exposure, yet finds no soft- or hard-band signal. The resulting flux limits translate to bolometric luminosities more than an order of magnitude below those inferred from JWST, directly ruling out current super-Eddington SED models tailored for LRDs. The non-detections are instead consistent only with extreme obscuration, requiring NH ≳ 10^25 cm^-2. The authors therefore argue the SMBHs in LRDs are likely less massive and/or less luminous than currently claimed.

Key figures to inspect

• Figure 1: Use the exposure-corrected CDF-S mosaic with 55 LRD positions to assess field coverage, crowding, and off-axis distribution that sets the stacking PSF/EEF choices and background control.
• Figure 2: Compare the stacked soft/hard-band upper limits against super-Eddington model swaths (Pacucci & Narayan; Inayoshi et al.; agnslim) to see these predictions sit above the limits across viewing angles and Eddington ratios.
• Figure 2 (red curves): Inspect the Compton-thick borus02 tracks; only NH ≳ 10^25 cm^-2 falls below the limits, illustrating how extreme columns are required even in the hard band.