UPDATE Matt King 07/07/2015 Region Finding 2 Have been working - - PowerPoint PPT Presentation

MOEDAL SIMULATION UPDATE

Matt King

07/07/2015

1

Region Finding

 Have been working on an automated fiducial region finding

algorithm for the MMTs

 With assistance from Katarina Bendtz, who designed the

algorithm for ATLAS.

 Our situation is a little different from ATLAS

 Two distinct regions (main MMT box, box top)  No full coverage in eta  Don’t need the other analysis code in which the region finder is

embedded

 => Have been working on a new algorithm “inspired” by the

ATLAS one

07/07/2015 Matthew King, IFIC

2

The algorithm

 The basic idea of the algorithm is to find the largest

possible region, inside a mother-region, that meets some criteria.

 Algorithm *always* finds the largest possible region, not an

approximation.

 Current criteria is:

 Average efficiency in region > effmin

 More advanced selection, optimization of a function is also

possible

 Requires more study. A simple selection seems sufficient for

now.

07/07/2015 Matthew King, IFIC

3

Initial results

 Several obvious tuning parameters:

 Region size limit  Region efficiency limit  Region efficiency standard deviation limit (ignoring for now)

07/07/2015 Matthew King, IFIC

4

Efficiency*3 (because of phi coverage) Region size: theta * KEz( MeV)

“Best” regions

07/07/2015 Matthew King, IFIC

5

effmin = 0.4 effmin = 0.6 effmin = 0.5 effmin = 0.7

Geant4 steps

 Have looked further into the problem of large numbers of

Geant4 steps.

 “Remove air” solution:

 Works, but number of steps in other materials is still large for

• ther materials. Solves the immediate problem, but may run

into problems later.

 Increase “thePILfactor”:

 Experimental (undocumented) Geant4 feature that manually

multiplies the step length by some factor.

 Works, does not affect continuous energy loss but effectively

reduces effect of secondaries by the factor (in all materials).

 Observed to be a lesser effect than the continuous energy loss, but

not a rigorous solution.

07/07/2015 Matthew King, IFIC

6

Geant4 steps

 “Correct” solution.  There are two possible approaches to an energy loss

algorithm:

 PostStepDoIt (current): computes a step based on the mean

free path, does secondaries at the end of every step.

 AlongStepDoIt (target): much larger step length, secondaries

generated along the length of the step.

 Some success in getting this to work. Still needs much more work

though.

 Not trivial to make this change.

07/07/2015 Matthew King, IFIC

7