2010ApJ...721.1725D

We present a systematic study of the physical properties of late-M dwarfs based on high-quality dynamical mass measurements and near-infrared (NIR) spectroscopy. We use astrometry from Keck natural and laser guide star adaptive optics imaging to determine orbits for the late-M binaries LP 349-25AB (M7.5+M8), LHS 1901AB (M6.5+M6.5), and Gl 569Bab (M8.5+M9). We find that LP 349-25AB (M_tot= 0.120^+0.008_–0.007M_☉) is a pair of young brown dwarfs for which Lyon and Tucson evolutionary models jointly predict an age of 140±30 Myr, consistent with the age of the Pleiades. However, at least the primary component seems to defy the empirical Pleiades lithium depletion boundary, implying that the system is in fact older (if the parallax is correct) and that evolutionary models under-predict the component luminosities for this magnetically active binary. We find that LHS 1901AB is a pair of very low-mass stars (M_tot= 0.194^+0.025_–0.021M_☉) with evolutionary model-derived ages consistent with the old age (>6 Gyr) implied by its lack of activity. Our improved orbit for Gl 569Bab results in a higher mass for this binary (M_tot= 0.140^+0.009_–0.008M_☉) compared to previous work (0.125±0.007 M_☉). We use these mass measurements along with our published results for 2MASS J2206-2047AB (M8+M8) to test four sets of ultracool model atmospheres currently in use. Fitting these models to our NIR integrated-light spectra provides temperature estimates warmer by 250 K than those derived independently from Dusty evolutionary models given the measured masses and luminosities. We propose that model atmospheres are more likely to be the source of this discrepancy, as it would be difficult to explain a uniform temperature offset over such a wide range of masses, ages, and activity levels in the context of evolutionary models. This contrasts with the conclusion of Konopacky et al. that model-predicted masses (given input T_eff and L_bol) are at fault for differences between theory and observations. In addition, we find an opposite (and smaller) mass discrepancy from what they report when we adopt their model-testing approach: masses are too high rather than too low because our T_eff estimates derived from fitting NIR spectra are 650 K higher than their values from fitting broadband photometry alone.