Cluster ID – a specific implementation
We have described the Cluster ID concept in general terms above and we now turn to describing the specific implementation we have developed and tested. For the cluster builder we use the contiguous hits method described above. We will plot specific results derived using a detector with the following description… We will also show results that illustrate the difference between gaps and nogaps. For the particles we simulate we will use 500 GeV ttbar events to determine both single particle distributions and illustrate efficacy in a real event…
For the discriminating properties we have selected a set of 15 discriminators, each one has been chosen to contribute to solving particular problems of discrimination. We now described the discriminators in some detail.
The general question of describing the shape and size and orientation of a cluster is facilitated by defining a (symmetric) energy tensor for each cluster and calculating its three eigenvalues and the three principal axes. The tensor is defined exactly as one would define an inertia tensor except that the mass of each element in the object is replaced here with the energy of each hit. The location of each hit is taken to be the center of the hit cell. We label the eigenvalues for a cluster E1, E2 and E3 such that E1 < E2 < E3. It turns out to be advantageous to normalize the eigenvalues with the energy of the cluster calling the resulting normalized energy eigenvalues NE1, NE2 and NE3. This way clusters created by the same particle type but with varying energies have similar numerical NEn values.
Plots in figure 1 show the distribution of NE values for single photons, pions, and K0Ls [charged hadrons (pions, kaons,…) and neutral hadrons (K0L, K0S, neutron, lambda…)] that interact in the calorimeter. NE1 is the smallest normalized eigenvalue and corresponds to the…TO BE CONTINUED
The LCD code system will have a default net which will be automatically invoked if a user does not override this with their own custom version. The defaults will be specific a detector type but will apply to all event types simulated in that detector type. These standard default nets can be used as a standard for comparison with test versions.