I repeated the exercise of matching tracks with new and old errors, as done in my previous blog post, but this time I used primary tracks. For matching, I required:

• Same event
• |Δη| < 0.1
• |Δz_vtx| < 4 cm
• |Δψ_vtx| < 10 °

...where ψ_vtx is the azimuthal direction of the momentum vector at the primary vertex. I also apply quality cuts that tracks have flag>0 (seems to be unnecessary, as all primary tracks appear to have a flag of either 301 or 311) and N_hit>22 (which removes the flag=311 special forward-going tracks anyhow).

Here is the distribution of matched primary tracks log₁₀(pT):

And here are the ratios of unmatched to matched tracks vs. log₁₀(pT), where red is found only with old errors, and blue is found only with new errors:

Over most pT, the distributions are the same, which probably means I'm not perfectly matching tracks. At the highest pT bin of unmatched tracks there is a hint of a possible difference, but this is countered by the fact there are 124 matched tracks with log₁₀(pT) greater than 1.5, and no unmatched tracks there. If it were an effect similar to RT ticket 2098 for high pT track loss, the trend of the ratio might be expected to continue to rise with growing pT, but such a trend is pretty well excluded by the high pT results.

In the [1.3,1.5] bin, there are 21 unmatched old error tracks, and 474 matched, or 495 total. Using sqrt(N), the expected statistical error on any measurement in that bin is ±22.2 tracks. This puts 21 unmatched tracks within error of 0. It is very likely then that the high value in this bin is just a statistical fluctuation.

For the bins above 1.5 (pT = 31.6 GeV/c), the result of 0 unmatched in 124 sets a 95% Poisson confidence level that the losses are below 2.42%.

_______________

Looking at matching vs. N_hit (removing the aforementioned N_hit cut), we can see that matching is (not surprisingly) the worst at low N_hit. Plots are the same, just linear and log scales. In general there is little difference is the two unmatched distributions. If anything, the new errors give slightly more unmatched low N_hit tracks.

________________

Looking at matching vs. η, we see that many of the primary tracks found only in the old errors sample are high positive η. I suspect that these are the specially tracked forward-going (towards the EEMC) tracks which have N_hit large enough not to be specially flagged.

I remade the pT distribution cuts using |η|<1.2 but I will not bother posting them because they are nearly identical to the first plots on this page. High η tracks are a very small fraction of the track population, and there are no unmatched tracks with |η|>1.2 and pT>16 GeV/c [log₁₀(pT) = 1.2], while there is just one matched track right on the edge of those cuts: η=1.208, pT = 45.9 GeV/c, N_hit=23.

________________

My conclusion is that there is no significant loss of primary tracks at high pT.

-Gene