Performance of Jet Algorithms in CMS

The CMS Combined Software and Analysis Challenge 2007 (CSA07) is well underway and expected to produce a wealth of physics analyses to be applied to the first incoming detector data in 2008. The JetMET group of CMS supports four different jet clustering algorithms for the CSA07 Monte Carlo samples, with two different parameterizations each: \fastkt, \siscone, \midpoint, and \itcone. We present several studies comparing the performance of these algorithms using QCD dijet and \ttbar Monte Carlo samples. We specifically observe that the \siscone algorithm performs equal to or better than the \midpoint algorithm in all presented studies and propose that \siscone be adopted as the preferred cone-based jet clustering algorithm in future CMS physics analyses, as it is preferred by theorists for its infrared- and collinear-safety to all orders of perturbative QCD. We furthermore encourage the use of the \fastkt algorithm which is found to perform as good as any other algorithm under study, features dramatically reduced execution time w.r.t. previous implementations of the \kt algorithm, and is infrared- and collinear save as well.