We study the use of red sequence selected galaxy spectroscopy for unbiased estimation of galaxy cluster masses .
We use the publicly available galaxy catalog produced using the semi-analytic model of on the Millenium Simulation ( ) .
We make mock observations to mimic the selection of the galaxy sample , the interloper rejection and the dispersion measurements for large numbers of simulated clusters spanning a wide range in mass and redshift .
We explore the impacts on selection using galaxy color , projected separation from the cluster center , and galaxy luminosity .
We probe for biases and characterize sources of scatter in the relationship between cluster virial mass and velocity dispersion .
We identify and characterize the following sources of bias and scatter : intrinsic properties of halos in the form of halo triaxiality , dynamical friction of red luminous galaxies and interlopers .
We show that due to halo triaxiality the intrinsic scatter of estimated line-of-sight dynamical mass is about three times larger ( 30 - 40 \% ) than the one estimated using the 3D velocity dispersion ( \sim 12 \% ) and a small bias ( \lesssim 1 \% ) is induced .
Furthermore we find evidence of increasing scatter as a function of redshift and provide a fitting formula to account for it .
We characterize the amount of bias and scatter introduced by dynamical friction when using subsamples of red-luminous galaxies to estimate the velocity dispersion .
We study the presence of interlopers in spectroscopic samples and their effect on the estimated cluster dynamical mass .
Our results show that while cluster velocity dispersions extracted from a few dozen red sequence selected galaxies do not provide precise masses on a single cluster basis , an ensemble of cluster velocity dispersions can be combined to produce a precise calibration of a cluster survey mass–observable relation .
Currently , disagreements in the literature on simulated subhalo velocity dispersion- mass relations place a systematic floor on velocity dispersion mass calibration at the 15 % level in mass .
We show that the selection related uncertainties are small by comparison , providing hope that with further improvements to numerical studies this systematic floor can be substantially reduced .