Evidence for dark matter in MG 1654+134

The radio ring lens MG 1654+134 strongly constrains the structure and mass of the 19 mag lens galaxy at its center. The best-fit effective radius for a de Vaucouleurs model of the lens galaxy is R_e_ = 2.0" (1.5" < R_e_ < 3.0") marginally consistent with fits to the observed lens galaxy. The B band mass-to-light ratio interior to the ring that the galaxy would have today is (M/L)_B_ = (20.4 +/- 2.8)(f_e_/1.4)h, where a passively evolving early type galaxy fades by a factor of f_e_~ 1.4 between the lens redshift and the present. This (M/L)_B_ is inconsistent with stellar dynamical models of nearby ellipticals, but it is consistent with estimates from the statistics of gravitational lens numbers and separations. Isothermal models fit the ring with a halo velocity dispersion of σ_DM_ = (216+/-2) km s^-1^, consistent with the fundamental plane, the stellar dynamics of ellipticals in dark matter halos [predicted dispersion (212+/-12)(f_e_/14)^0.29^ km s^-1^], and the statistics of gravitational lenses. This strongly supports the conclusion from lens statistics that early type galaxies do not have constant mass- to-light ratios even on scales comparable to the effective radius. Density distributions with ρ is proportional to (r^2^ +s^2^)^(α - 3)/2^ fit the data only in the restricted range 0.9 <~ α <~ 1.1 where α = 1 is an isothermal model. Centrally concentrated models such as the Plummer or modified Hubble models require large, finite core radii (800 h^-1^ and 600 h^-1^ pc, respectively) but fail to fit the structure of the ring. Isothermal models only have upper limits of order 100 h^-1^ pc on the core radius. The major axis of the mass distribution is 7^deg^ + 10^deg^ from the major axis of the luminosity distribution. The models clearly distinguish between different angular structures for the lens model, and a quadrupole expansion of an ellipsoidal density distribution fits the data better than a similar monopole in an external shear.