2003ApJ...587..407C

Use of the highly sensitive Hokupa'a/Gemini curvature wave front sensor has allowed direct adaptive optics (AO) guiding on very low mass (VLM) stars with SpT = M8.0-L0.5. A survey of 39 such objects detected nine VLM binaries (seven of which were discovered for the first time to be binaries). Most of these systems are tight (separation <5 AU) and have similar masses (ΔKs<0.8 mag; 0.85<q<1.0). However, two systems (LHS 2397a and 2M 2331016-040618) have large ΔKs>2.4 mag and consist of a VLM star orbited by a much cooler L7-L8 brown dwarf companion. On the basis of this flux-limited (Ks<12 mag) survey of 39 M8.0-L0.5 stars (mainly from the 2MASS sample of Gizis et al.), we find a sensitivity-corrected binary fraction in the range 15%±7% for M8.0-L0.5 stars with separations greater than 2.6 AU. This is less than the 32%±9% measured for more massive M0-M4 dwarfs over the same separation range. It appears M8.0-L0.5 binaries (as well as L and T dwarf binaries) have a much smaller semimajor axis distribution peak (∼4 AU) compared to more massive M and G dwarfs, which have a broad peak at larger ∼30 AU separations. We also find no VLM binary systems (defined here as systems with M_tot<0.185M_☉) with separations greater than 15 AU. We briefly explore possible reasons why VLM binaries are slightly less common, nearly equal in mass, and much more tightly bound compared to more massive binaries. We investigate the hypothesis that the lack of wide (a>20 AU) VLM/brown dwarf binaries may be explained if the binary components were given a significant differential velocity kick. Such a velocity kick is predicted by current ``ejection'' theories, where brown dwarfs are formed because they are ejected from their embryonic minicluster and therefore starved of accretion material. We find that a kick from a close triple or quadruple encounter (imparting a differential kick of ∼3 km/s between the members of an escaping binary) could reproduce the observed cutoff in the semimajor axis distribution at ∼20 AU. However, the estimated binarity (≲5%) produced by such ejection scenarios is below the 15%±7% observed. Similarly, VLM binaries could be the final hardened binaries produced when a minicluster decays. However, the models of Sterzik & Durisen and Durisen, Sterzik, & Pickett also could not produce a VLM binary fraction of 15% and a G star binary fraction of 57%. The observed VLM binary frequency could possibly be produced by cloud core fragmentation. However, our estimate of a fragmentation-produced VLM binary semimajor axis distribution contains a significant fraction of ``wide'' VLM binaries with a>20 AU in contrast to observation. In summary, more detailed theoretical work will be needed to explain these interesting results that show VLM binaries to be a significantly different population from more massive M & G dwarf binaries.