HST STIS spectroscopy of the triple nucleus of M31:: Two nested disks in keplerian rotation around a supermassive black hole

We present Hubble Space Telescope (HST) spectroscopy of the nucleus of M31 obtained with the Space Telescope Imaging Spectrograph (STIS). Spectra that include the Ca II infrared triplet (lambda similar or equal to 8500 angstrom) see only the red giant stars in the double brightness peaks P1 and P2. In contrast, spectra taken at lambda similar or equal to 3600 - 5100 angstrom are sensitive to the tiny blue nucleus embedded in P2, the lower surface brightness nucleus of the galaxy. P2 has a K-type spectrum, but we find that the blue nucleus has an A- type spectrum: it shows strong Balmer absorption lines. Hence, the blue nucleus is blue not because of AGN light but rather because it is dominated by hot stars. We show that the spectrum is well described by A0 giant stars, A0 dwarf stars, or a 200 Myr old, single-burst stellar population. White dwarfs, in contrast, cannot fit the blue nucleus spectrum. Given the small likelihood for stellar collisions, recent star formation appears to be the most plausible origin of the blue nucleus. In stellar population, size, and velocity dispersion, the blue nucleus is so different from P1 and P2 that we call it P3 and refer to the nucleus of M31 as triple. Because P2 and P3 have very different spectra, we can make a clean decomposition of the red and blue stars and hence measure the light distribution and kinematics of each uncontaminated by the other. The line-of-sight velocity distributions of the red stars near P2 strengthen the support for Tremaine's eccentric disk model. Their wings indicate the presence of stars with velocities of up to 1000 km s(-1) on the anti-P1 side of P2. The kinematics of P3 are consistent with a circular stellar disk in Keplerian rotation around a supermassive black hole. If the P3 disk is perfectly thin, then the inclination angle i similar or equal to 55 degrees is identical within the errors to the inclination of the eccentric disk models for P1+ P2 by Peiris & Tremaine and by Salow & Statler. Both disks rotate in the same sense and are almost coplanar. The observed velocity dispersion of P3 is largely caused by blurred rotation and has a maximum value of sigma 1183 +/- 201 km s(-1) . This is much larger than the dispersion sigma similar or equal to 250 km s(-1) of the red stars along the same line of sight and is the largest integrated velocity dispersion observed in any galaxy. The rotation curve of P3 is symmetric around its center. It reaches an observed velocity of V 618 +/- 81 km s(-1) at radius 0''.05 0.19 pc, where the observed velocity dispersion is sigma 674 +/- 95 km s(-1). The corresponding circular rotation velocity at this radius is similar to 1700 km s(-1). We therefore confirm earlier suggestions that the central dark object interpreted as a supermassive black hole is located in P3. Thin-disk and Schwarzschild models with intrinsic axial ratios b/a less than or similar to 0.26 corresponding to inclinations between 55 degrees and 58 degrees match the P3 observations very well. Among these models, the best fit and the lowest black hole mass are obtained for a thin- disk model with M center dot = 1.4 x 10(8) M circle dot. Allowing P3 to have some intrinsic thickness and considering possible systematic errors, the 1 sigma confidence range becomes (1.1- 2.3) x 10(8) M circle dot. The black hole mass determined from P3 is independent of but consistent with Peiris & Tremaine's mass estimate based on the eccentric disk model for P1+ P2. It is similar to 2 times larger than the prediction by the correlation between M. and bulge velocity dispersion sigma bulge. Taken together with other reliable black hole mass determinations in nearby galaxies, notably the Milky Way and M32, this strengthens the evidence that the M center dot-sigma bulge relation has significant intrinsic scatter, at least at low black hole masses. We show that any dark star cluster alternative to a black hole must have a half- mass radius less than or similar to 0''.03 = 0.11 pc in order to match the observations. Based on this, M31 becomes the third galaxy (after NGC 4258 and our Galaxy) in which clusters of brown dwarf stars or dead stars can be excluded on astrophysical grounds.