Stellar Kinematics of the Double Nucleus of M31

Abstract

We report observations of the double nucleus of M31 with the f/48 long-slit spectrograph of the Hubble Space Telescope Faint Object Camera. We obtain a total exposure of 19,000 s over seven orbits, with the 0.063"-wide slit along the line between the two brightness peaks (P.A. 42 deg). Careful correction of the raw data for detector backgrounds and geometric distortion is essential. A spectrum of Jupiter obtained with the same instrument is used as a spectral template to extract rotation and velocity dispersion profiles by cross-correlation. The rotation curve is resolved and reaches a maximum amplitude of ~250 km s^-1 roughly 0.3" to either side of a rotation center lying between peaks P1 and P2, 0.16"+/-0.05" from the optically fainter P2. We find the velocity dispersion to be <~250 km s^-1 everywhere except for a narrow ``dispersion spike'' centered 0.06"+/-0.03" on the anti-P1 side of P2, in which sigma peaks at 440+/-70 km s^-1. At much lower confidence, we see local disturbances to the rotation curve at P1 and P2 and an elevation in sigma at P1. At very low significance we detect a weak asymmetry in the line-of-sight velocity distribution opposite to the sense usually encountered. Convolving our V and sigma profiles to Canada-France-Hawaii Telescope resolution, we find good agreement with the forthcoming results of Kormendy & Bender, although there is a 20% discrepancy in the dispersion that cannot be attributed to the dispersion spike. Our results are not consistent with the location of the maximum dispersion found by Bacon et al. in 1994. Comparing with published models, we find that the more recent sinking star cluster model of Emsellem & Combes does not reproduce either the rotation curve or the dispersion profile. The eccentric-disk model proposed by Tremaine fares better and can be improved somewhat by adjusting the original parameters. However, detailed modeling will require dynamical models of significantly greater realism.