17th Cambridge Workshop on Cool Stars, Stellar Systems and the Sun

Type: Oral

Topic: Fundamental Parameters of Cool Stars and Brown Dwarfs

Abstract
We present a new technique, and improved calibrations of existing techniques, to determine M dwarf metallicities from optical and/or NIR spectra. Deriving metallicities of M dwarfs from stellar spectra is difficult due to the presence of complex molecules in their atmospheres. Recently, several methods have been developed using either NIR (Rojas-Ayala et al. 2011; Terrien et al. 2012) or visible wavelength (Woolf et al. 2009) spectra. These techniques use atomic (for NIR) or molecular (for visible wavelengths) lines to produce models that approximate M-dwarf metallicities. The models are calibrated using well-separated M dwarf companions to FGK stars with precise metallicities. Unfortunately, most of these studies have made use of a relatively small number of calibrator pairs ($\lesssim 30$). As a result, they rely heavily on stars in a narrow range of metallicities ($-0.4 \lesssim $[Fe/H]$ \lesssim +0.4$) and spectral types ($\simeq$ M0 to M4). We improve on this by identifying $>$120 FGK-M suitable stellar pairs, more than 70 of which have primaries with published metallicities. For the remaining primary stars, we derive metallicities from high-resolution spectra taken with ESPaDOnS on CFHT. We obtain spectra for each M dwarf companion with SNIFS on the UH2.2m and SpeX on the IRTF in the visible and NIR respectively. Our resulting sample spans a wide range of metallicities ($-1.1 < $[Fe/H]$ < +0.5$) and spectral types (K7 to M7). We then use this sample to test and recalibrate known metallicity estimators as well as to identify new methods of determining M dwarf metallicities. We find that metallicities derived from NIR spectra are precise (RMSE $\simeq 0.12$~dex) near solar metallicity, but are unreliable at extremely low metallicities ([Fe/H]$<-0.5$). We find the most accurate metallicities can be determined using a joint fit of visible wavelength and NIR spectroscopic features Although our results are preliminary, this new method already shows promising results (RMSE $\simeq 0.07$) over the full range of metallicities in our sample.