The 2D clustering in the Ifr is done merging together 1D hits which have the (square of the) distance in the 2D projection (normalized to the iron gap thicknes) smaller than a given cut. This "blob" criteron does not rely on a particular shape of the track traversing the Ifr and permits a reconstruction of both muon tracks as well as K0L showers.
Varying the value of this cut may result in:
Looking at cluster multiplicities of reconstructed cosmics event with different cut values gives an indication of the optimum cut values.
All the following plots show the variation of several quantities as a function of the clustering cut (on the x axis). The way to change this cut in reconstruction is the following:
module talk IfrMake2DClusters
cutSqr set <value>
exit
The first two plots show the average number of reconstructed 2D and 3D clusters in a sample of 200 cosmics events taken with the Ifr trigger (warning: the first 17 events contain no data).
The events contain both the barrel and the forward end cap. A 2D cluster is included in the plot and may be candidate for 3D clustering if it hits at least 5 layers. For low cut values ( <~ 6 ), tracks are split into multiple clusters; a fraction of those is so small that does not satisfy the minimum number of layers requirement (and does not enter the plot). If the cut is greater than ~9 the number of clusters starts to decrease up to a pleteau, because contiguous tracks in the same sector are merged together. Those tracks are mainly due to a straight track broken into two pieces due to the inefficiency of one layer.
The following plot shows the average number of hits per 3D clusters. Since increasing the cuts generates larger clusters, the trend is monothonic.
The next two plots show the dependency of multiplicity of reconstructed 2D clusters and 3D clusters respectively as a function of the cut. In particular, the number of events with high multiplicities (# of 2D = 6, #of 3D = 3) drops as the cut goes over 9. Conversely, the number of events with only one reconstructed 3D clusters drops (because the 2D segments are too short) as the cut reaches values smaller than ~6.
The values 10 to 12 seem to give stable results and is candidate for placing default cuts. Anyway, excluding the 17 events with no Ifr data, all the events reconstruct at least one valid cluster in all the cut range ([2, 25]) studied.
Two 2D clusters reconstructed in two views of the same sector are matched to form a 3D clusters if all the hit layers match with the exception of a maximum number of mismatches specified by a cut value. In order to change the value of this cut in the tcl script, the following command lines are needed:
More studies varying the minimum # of layer and scanning events by hand will help better understanding the results.
module talk IfrMake3DClusters
cut set <value>
exit
If the cut is set to zero, all hit layers must match (i.e. there must be no layer hit in one view and not in the other); in this case, a view with inefficient FECs may cause the loss of 3D clusters. If the cut is set to 19, essentially all 2D clusters of the same view are associated together, so more 3D clusters are reconstructed, but a larger number of ambiguities may arise (those ambiguities may be reduced later using the tracking for charged clusters). The average number of reconstructed 3D clusters is monothonically increasing with the cut value, as shown in the next plot.
The distribution of multiplicity varyes with the cut value according to the curves shown in the next plot. For cut values smaller than 3 some events have no reconstructed cluster at all (aside the kown 17...), while above this value all avents have at least one reconstructed 3D cluster
Using cutSqr = 5 and minimum # of layer set to 5, running on 1000 events, the fraction of ambiguous clusters varies from 2.1% with cut on 3D clusters set to 20 (all 2D clusters of the same sector are associated) down to 0.6% if the cut is lowered to 3. As mentioned before, the number of ambiguities may be further reduced with the association with charged tracks.