Optimal jet finder. D.Y. Grigoriev, E. Jankowski, F.V. Tkachov.

PROGRAM SUMMARY
Title of program: OJF-014
Catalogue identifier: ADSB
Ref. in CPC: 155(2003)42
Distribution format: tar gzip file
Operating system: Linux/Windows98
High speed store required: 1MK words
Number of lines in distributed program, including test data, etc: 27472
Keywords: Hadronic jets, jet finding algorithms, Particle Physics, Elementary, Event reconstruction.
Programming language used: Fortran 77
Computer: Any with a F77 compiler .

Nature of physical problem:
Analysis of hadronic final states in high energy particle collision experiments often involves identification of hadronic jets. A large number of hadrons detected in the calorimeter is reduced to a few jets by means of a jet finding algorithm. The jets are used in further analysis which would be difficult or impossible when applied directly to the hadrons. Grigoriev et al [hep-ph/0301185] provide a brief introduction to the subject of jet finding algorithms and a general review of the physics of jets can be found in [Rep. Prog. Phys. 36 (1993) 1067].

Method of solution:
The software we provide is an implementation of the so-called optimal jet definition (OJD). The theory of OJD was developed by Tkachov [Phys. Rev Lett 73 (1994) 2405; 74 (1995) 2618; Int. J. Mod. Phys. A 12 (1997 5411; 17 (2002) 2783]. The desired jet configuration is obtained as the one that maximises ΩR, a certain function of the input particles and jet configuration.

Restrictions:
The size of the largest data structure the program uses is (maximal number of particles in the input) x (maximal number of jets in the output) x 8 bytes. (For the standard settings <1 MB). Therefore, there is no memory restriction for any conceivable application for which the program was designed.

Typical running time:
The running time depends strongly on the physical process being analyzed and the parameters used. For the benchmark process we studied e+e- -> W+W- -> 4 jets with the average number of ~80 particles in the input, the running time was <10-2 s on a modest PC per event with ntries=1. (We took ntries=10 for all events in the benchmark process, however for the majority of events ntries~3 would suffice. For small values of ntries the global minimum of ΩR may be missed in some fraction of events resulting in the deterioration of the precision of measurements based on the jet algorithm.) For a fixed number of jets the complexity of the algorithm grows linearly with the number of particles (cells) in the input, in contrast with other known jet finding algorithms for which this dependence is cubic. The reader is referred to Grigoriev et al [hep-ph/0301185] for a more detailed discussion of this issue.