Background from PID Contamination

Summary:

The goal of this analysis is to estimate the contribution to A_LL from particles that aren't charged pions but nevertheless make it into my analysis sample. So far I have calculated A_LL using a different dE/dx window that should pick out mostly protons and kaons, and I've estimated the fraction of particles inside my dE/dx window that are not pions by using a multi-Gaussian fit in each pt bin. I've assumed that this fraction is not spin-dependent.


Points to remember:
  • my analysis cuts on -1 < nSigmaPion < 2
Multi-Gaussian fits to nSigmaPion distributions:

pi- is on the left, pi+ on the right. Each row is a pt bin corresponding to the binning of my asymmetry measurement. The red Gaussian corresponds to pions, green is protons and kaons, blue is electrons. So far I've let all nine parameters float. I tried fixing the mean and width of the pion Gaussian at 0. and 1., respectively, but that made for a worse overall fit. So far, the fit results for the first two bins seem OK.

2 < pt < 4:
4 < pt < 6:
6 < pt < 8:
8 < pt < 10:

I extracted the the integral of each curve from -1..2 and got the following fractional contributions to the total integral in this band:

pi- bin pion p/K electron
2-4 0.91 0.09 0.01
4-6 0.92 0.05 0.03
6-8 0.78 0.07 0.15
8-10 0.53 0.46 0.01

pi+ bin pion p/K electron
2-4 0.90 0.09 0.01
4-6 0.91 0.06 0.03
6-8 0.68 0.06 0.26
8-10 0.88 0.05 0.08
A_LL for protons and kaons

I repeated my A_LL analysis changing the dE/dx window to [-inf,-1] to select a good sample of protons and kaons. The A_LL I calculate for a combined MB || JP1 || JP2 trigger (ignore the theory curves) is


For comparison, the A_LL result for the pion sample looks like

I know it looks like I must have the p/K plots switched, but I rechecked my work and everything was done correctly. Anyway, since p/K is the dominant background the next step is to use this as the A_LL for the background and use the final contamination estimates from the fits to get a systematic on the pion measurement