2025MNRAS.540.1608D

Metallicity (Z) estimates based on ultraviolet (UV) emission lines from the narrow-line regions of active galactic nuclei (AGNs) have been found to differ from those derived from optical lines. However, the origin of this discrepancy ($ZR$) remains poorly understood. To investigate the source of $ZR$, we compiled from the literature the fluxes of narrow near-UV [$1000 < \lambda (\rm{\mathring{\rm A}}) < 2000]$ and optical [$3000 < \lambda (\rm{\mathring{\rm A}}) < 7000]$ emission-line measurements for a sample of 11 AGNs (nine at $z<0.4$ and two at $z\sim 2.4$). Metallicity values for our sample were derived using a semi-empirical calibration based on the $C43$ = log[(C IV$\lambda$1549 + C III]$\lambda$1909)/He II$\lambda$1640] emission-line ratio and compared with those obtained via direct measurement of the electron temperature ($T_{\rm e}$-method) and via calibrations based on optical emission lines. The source of the discrepancy was investigated in terms of the ionization parameter (U), electron density ($N_{\rm e}$), and carbon abundance (C/H). We found a weak correlation between $ZR$, U, and $N_{\rm e}$. However, a moderate correlation was observed between $ZR$ and direct estimates of C/H, suggesting that the previously assumed (C/O)-Z relations in photoionization models used to derive UV carbon-line calibrations may not be valid for AGNs. By combining a large set of abundance estimates for local star-forming regions with those of our AGN sample, we derived a new (C/O)-Z relation. Comparisons between the results of photoionization models that assume this new abundance relation and the UV observational data of our sample produce Z values derived from the $C43$ index that are consistent with those obtained using the $T_{\rm e}$-method.