Cornering light stops with dileptonic mT2

Supersymmetric spectra with a stop NLSP and a neutralino or gravitino LSP present a special challenge for collider searches. For stop pairs directly produced from QCD, the visible final-state particles are identical to those of top quark pair production, giving very similar kinematics but often with much smaller rates. The situation is exacerbated for compressed spectra with m((t) over tilde) similar or equal to m(t) + m(LSP), as well as for lighter stops which can suffer from low acceptance efficiencies. In this note, we explore the power of a direct stop search using dileptonic m(T2), similar to the one recently performed by ATLAS, but more optimized to cover these difficult regions of the (m((t) over tilde), m(LSP)) plane. Our study accounts for the effects of stop chirality and LSP identity, which can be significant. In particular, our estimates suggest that m((t) over tilde) = m(t) with a massless LSP is excludable for right-handed stops with bino-like (gravitino) LSP with 2012 (2011) data, but remains largely unobservable in the case of a higgsino-like singlino LSP. For each of these cases we map out the regions of parameter space that can be excluded with 2012 data, as well as currently allowed regions that would yield discovery-level significance. For the right-handed stop with bino-like LSP, we identify a narrow residual gap in coverage just above the compression line for any LSP mass, which will progressively close with more data and/or improved systematics. We also comment on the prospects of a precision mT2 shape measurement, and consider the potential of ATLAS's dileptonic (t) over tilde -> b (chi) over tilde (+)(1) searches when re-interpreted for light stops decaying directly to the LSP.