Do AGNs Really Change the Look of Galaxies?

Astronomers have found that active galactic nuclei (AGN) can trick us when we classify galaxy shapes. Starting from over 2,000 nearby inactive galaxies, researchers added a fake “bright point” representing an AGN, contributing 5%–75% of the total light. They then simulated how these galaxies would look fainter, as if they were farther away, up to redshift z≈2, typical of the COSMOS field. Measuring six morphological parameters—concentration, asymmetry, GINI, M20, CCON and CABR—they saw the strongest effect when the galaxy is faint (magnitude > 23) and the AGN is very bright (> 25% of the light). Spirals are the most “confused” because their arms become blurred. Among the parameters, the GINI coefficient is the most stable, followed by M20, CCON and finally CABR. To spot active galaxies at high redshift, the best combo is CABR + CCON + asymmetry, or CABR + GINI. These findings help refine galaxy catalogs and deepen our grasp of cosmic evolution.

The analysis relies on a sample of local galaxies with established visual classification, to which a simulated point-like nucleus was added. The nucleus was modeled as an unresolved source, using a PSF typical of the Hubble telescope, and its flux fraction varied from 5% to 75% of the total. To mimic higher‑redshift observations, images were degraded in resolution and noise, and magnitudes were increased to 25–26 mag. The six morphological parameters measured include: (i) GINI, quantifying light distribution; (ii) M20, the second moment of the brightest regions; (iii) CCON (Conselice‑Bershady), a concentration index based on growth curves; (iv) CABR (Abraham), concentration based on fixed apertures; (v) Asymmetry, comparing the image with its 180° rotation; (vi) Smoothness, measuring fine structures. Results show GINI is minimally affected by AGN contamination because it depends on the cumulative light distribution, whereas CABR is strongly impacted as it relies on central apertures that include the bright nucleus. Statistical tests (Kolmogorov‑Smirnov) confirm significant parameter differences for AGN > 25% and magnitude > 23. Moreover, the combined use of CABR, CCON, and asymmetry improves separation between active and inactive galaxies, achieving an ROC‑AUC > 0.85 versus < 0.70 for single parameters. These findings suggest that future deep‑field surveys must incorporate multiple parameters and account for AGN effects to avoid morphological classification bias.