2011A&A...528A..44M

The classical method to infer microturbulent velocity in stellar spectra requires that the abundances of the iron lines are not correlated with the observed equivalent widths. An alternative method, requiring the use of the expected line strength, is often used to bypass the risk of spurious slopes caused by the correlation between the errors in abundance and equivalent width. Our aim is to compare the two methods, describe their respective advantages and disadvantages, and show how they apply to the typical practical cases. We performed a test with a grid of synthetic spectra, including instrumental broadening and Poissonian noise. For these spectra, we derived the microturbulent velocity by using the two approaches and compared the results with the original value with which the synthetic spectra were generated. The two methods provide similar results for spectra with SNR≥70, while for lower signal-to-noise (SNR) both approaches underestimate the true microturbulent velocity, depending on the SNR and the possible selection of the lines based on the equivalent width errors. Basically, the values inferred by using the observed equivalent widths better agree with those of the synthetic spectra. Indeed, the method based on the expected line strength is not totally free from a bias that can heavily affect the determination of microturbulent velocity. Finally, we recommend to use the classical approach (based on the observed equivalent widths) to infer this parameter. For a full spectroscopical determination of the atmospherical parameters, the difference between the two approaches is reduced, which leads to an absolute difference in the derived iron abundances of less than 0.1dex.