Checking the reliability of equivalent width R23 for estimating the metallicities of galaxies

Aims. We verify whether the O/H abundances of galaxies can be derived from the equivalent width (EW) R-23 instead of the extinctioncorrected flux R-23, and eventually propose a method of improving the reliability of this empirical method, which is often used for the non- flux calibrated spectra of galaxies. Methods. We select 37 173 star-forming galaxies from the Second Data Release of the Sloan Digital Sky Survey (SDSS-DR2), which offers a wide range of properties to test the EW method. Results. The EW-R-23 method brings with it a significant bias: for the bulk of SDSS galaxies, it may a. ect the determination of log (O/H) by factors ranging from - 0.2 to 0.1 dex and for some galaxies by factors ranging from - 0.5 to 0.2 dex. We characterize this discrepancy (or bias) by alpha = (I-[OII]/I-H beta)/(EW[OII]/EWH beta), which is virtually independent of dust extinction, while tightly correlating with D-n(4000), although at a lower significance, with (g - r) colors. Conclusions. The EW-R-23 method cannot be used as a proxy for the extinction-corrected flux R-23 method. From analytical third-order polynomial fits of a versus (g - r) colors, we have been able to correct the EW-R-23 method. With this additional and easy correction, the EW-R-23 method provides O/H abundance values similar to those derived from the extinction-corrected flux R-23 method with an accuracy of 0.1 dex for > 92% of the SDSS galaxies.