Look at the test production 2007 MinBias 

I analyzed 500 files (over 2295 files available) ; the same cuts are applied than for 2007LowLuminosity test production

(lowLuminosity test)

PART I DCA resolution

Fig 1 : sigma DCA in transverse plan 

Fig 2 : sigma DCA in Z direction

These 2 plots are quite different from what it was observed for the LowLuminosity test

• sigma_DCA in transverse direction looks good up to 1/P < 1 then for the low momentum tracks, it increases and reaches the resolution given by TPC only
• sigma_DCA in Z direction is not good at all

As show with the next plots that represent the DCA vs 1/P, it seems there are a large background.

Fig. 3 : dca XY vs 1/P 

Fig. 4 : dca Z vs 1/P 

This background is also visible when we plot the DCA per momentum slices :

• dca XY : bin in momentum = 0.1 GeV/c : plot
• dca Z : bin in momentum = 0.1 GeV/c : plot

where we see the contribution of tails for the first pads.

I tried to plot the resolution of DCA as a function of p_{T} instead of P.

It seems to give the expected resolution for sigma_DCA in XY up to pT = 0.5 GeV/c (after it reaches the TPC resolution) but the sigma_DCA in Z direction is not improved.

Fig. 5 : sigma DCA in transverse direction

Fig. 6 : sigma DCA in Z direction

PART II DCA resolution : improvement

I took the projection of each bin (in Fig.3 and Fig.4) and fitted it with a double gaussian.

One gaussian is restraint to [-0.2 ; 0.2], which means a fit between [-0.2;0.2] cm for each dca distribution, the second is fit in the histo range [-1;1]

Fig 7. example for bin = 8 (eg 1/P=0.8 --> P = 1.25 MeV/c )

So I have 2 componants (2 gaussians) where :

• sigma1 (in [0.2;0.2]) = 526 um
• sigma2                    = 2.1 mm

 Fig 8. example for bin = 25 (eg 1/P=2.5 --> P = 0.4 MeV/c )

For lower momentum tracks, both sigmas are increasing : 

• sigma1 (in [0.2;0.2]) = 1.08 mm
• sigma2                    = 4.78 mm

 Fig 9. sigma1, sigma2 vs. 1/P (for dcaXY)

This figure show how the sigmas from the double-gaussian fit evolves as a function of 1/P.

I also show in green the result using FitSlicesY() simply (as it is in Fig3. and 4.) and the FitSlicesY() but for Cu+Cu (in purple) for comparison where we expect less background.

sigma1 vs 1/P seems to be linear with 1/P whereas fitting in the whole range give the increase I observed previously.

The Cu+Cu fitSlicesY() is higher than sigma1 but seems also linear as sigma1 ; it may indicate some background contribution in Cu+Cu but less dominant than in Au+Au 

Fig 10. sigma1, sigma2 vs. 1/P (for dcaZ)

The same plot but for dcaZ

Here the sigma1 is of the same magnitude than the result in Cu+Cu.

We see also that the contribution of the background is higher in Z than in XY (blue curve)