The IPChi2 tagger is based on the impact parameter distribution of mis-tags, which can be derived directly from data and does not require the use of Monte Carlo templates as in the likelihood ratio b-tagger.
For every region of interest selected by the level 1 trigger, the level 2 b-tagging algorithm uses tracks reconstructed by the level 2 trigger, to determine the probability that the RoI originates form the primary vertex. b-jets are identified by chi2 probability cut. The performance is measured as rejection as a function of efficiency. For instance, for a b-tagging efficiency of 60%, the Level 2 b-tagging algorithm rejection is 4, which means that, on average, only one out of four light quark jets will pass the selection.
link to ATLAS page We are working in the design of different trigger chains (set of trigger signatures and operational points at Level 1, Level 2 and Event Filter) optimized for different physics analyses, in particular all hadronic top, and inclusive jet+MET and b+MET searches. In particular, we are investigating the performance of asymmetric double b-tagging, which consists of requiring one loose and one tight b-tag jets. We have seen that this type of selection improves the rejection (at a same efficiency) compared to the more standard symmetric tagging. link to ATLAS page b+MET combined triggerThe SLAC group has proposed a combined b-jet+MET trigger, to be used in physics signatures containing b-jets and missing ET. The missing ET requirement provides an additional handle to reduce the output rate, while maintaining high efficiency for signal events. We are currently studying the effect of this trigger signature in hadronic top quark final states, and b+MET Super-symmetric events. link to ATLAS page | 
