Data Scouting for Low-Mass Dijet Resonances

We present the results of further investigation into a novel data analysis technique called "data scouting". This technique allows us probe below the lower dijet mass limit, which is imposed by bandwidth limitations on the high-level trigger (HLT) at large luminosities. By introducing a new trigger at HLT which runs a reduced version of the offline analysis and saves only a very small amount of data per event, we obtain a trigger rate approximately $10^3$ times larger than the corresponding rate for full event storage. After appropriate selection cuts, identifying each jet from this trigger with a corresponding jet reconstructed offline allows a direct comparison between jets. We find that the transverse momentum $p_t$ of the HLT jets is dependent on both the jet $p_t$ and on the pseudorapidity $\eta$, and quantify this relationship by fitting a crystal ball function to the fractional difference for various combinations of $p_t$ and $\eta$. This function can now be incorporated into the Monte Carlo simulation of RECO jets in order to simulate HLT jets instead, allowing HLT jets to be used for resonance searches at low masses ($<1~\textrm{TeV}$) with much more data than previously ($20~\textrm{fb}^{-1}$). FInally, we consider additional work that needs to be done to make this technique feasible for a rigorous analysis, and make a start on this by conducting an F-test to assess the suitability of the background model for the dijet spectrum.