Sample used is : 

• 5 D⁺, D⁰ , L_c mixed with hijing Au+Au central event
• flat pt
• geometry : UPGR15
• 10k events
• it corresponds to the 3rd sample of Yifei's list

GOAL : We have 2 available type of file : the mixed and hijing background. Therefore, by running the BFC chain and after our analysis code, we can have histograms for the signal + background (S + B) and background only (B).

Then one can build the signifcance as a function of a cut range and find the appropriate value to use

The first step is to find single tracks cut (using the mixed file contains both D⁺ candidates and daughter candidate but there not enough entries and we're not sure to use ONLY daughters from D⁺)

So I :

• used the hijing file and run a macro to fill only the tracks properties
• generated 50k D⁺ flat and real pt spectrum as a signal (1particle/event)

1. Flat pt : FZ file

Some histo to check the D⁺ are correctly generated with Starsim

Fig. 1: phase space of D⁺ :flat in pt and uniform in -1 < y < 1

Fig. 2 : life time of D⁺ : the parameter of the exponential fit shoud be equal to 1/ctau

The fit here gives 1/ctau = 31.67 --> ctau = 315.75 microns

2. Flat pt : file after reco+analysis code

A useful cut we looked at [to reduce the background] is the dca of track to Primary vertex. A fixed cut (for e.g, |dca|>40 microns) will reduce some backgound but may not be appropriated for p_t dependence. The significance of the track S_t = dca /error is better.

Results: the analysis code is run with 'open cuts" , then  the reco . invariant mass (triplet of tracks identified as Kaon, pion and pion) will have large background due to wrong sign association ( = pid misentification)

Requiring K<0, and the 2 pions >0 gives the correct invariant mass (Fig 3)

Fig. 3 : (Kpipi) invariant mass for all associations (black line) and correct sign for the daughters (yellow fill)

The next plot is the track significance distribution for tracks identified as kaon, for the 3 samples (background hijing, single D⁺ with flat and real pt)

I also used 3 sets of cuts :

1. |ndEdx| < 2 for kaon
2. |ndEdx| < 2 and (pixel hits =2,Ist hit=1, Ssd hit =1)
3. |ndEdx| < 2 and (pixel hits =2,Ist hit=1, Ssd hit =1) and |dca|<.2 cm to remove tails in the DCA distribution 

Fig. 4 : Track significance distributions [kaon] for different quality cuts and for the 3 samples

Clearly we see that a cut at |S_t|>2 should discriminate the background (blue line) from the signal (green and red lines)

The above plots are for all P_t : as I used flat and real P_t D⁺, the pt dependence of the dca distribution has to be check.

I used the DCA resolution (standard deviation of the DCA resolution vs 1/P for this

Fig. 5 : DCA resolution vs 1/P  [kaon] for different quality cuts and for the 3 samples

We see that the DCA resolution ( = the broadening of the DCA distribution) is higher for Kaon tracks from D⁺ compared to background, over a large momentum range

As a check, the P_t distribution for the 3 samples is below :

Fig. 6 : P_t distribution [kaon] for the 3 samples

The real pt D⁺ sample seems for appropriate for comparison with hijing

Fig. 7 : Track significance distributions [pion] for different quality cuts and for the 3 samples

_________________________________

The second step is to look at the tracks association. As a result of the first step, I used the cut |S_t|>2 for all tracks candidate (ie for kaon and pion)

The signal is the real P_t D⁺ and the background are some hijing files

I then compared dca1V = distance of the track to the secondary vertex point reconstructed

• cut_set_1 = |ndEdx| < 2 for all tracks, correct sign and invariant mass of D⁺ candidate in [1.82 ; 1.92]
• cut_set_2 = |ndEdx| < 2 for all tracks, correct sign and invariant mass of D⁺ candidate in [1.82 ; 1.92] and (pixel hits =2,Ist hit=1, Ssd hit =1) and |dca|<.2 cm

Fig. 8 : dca1V with cut_set1

Fig. 9 : dca1V with cut_set2

Fig 10 : zoom of Fig. 8

• TEST  with 1k events

Cuts applied directly in the analysis code :

1. |S_t| >2 for all daughters tracks
2. dca1v < 0.01 cm ( = 100 microns) , dca2v < 0.01 cm, dca3v < 0.01 cm

Fig 11 : invariant mass of (Kpipi) : no D⁺ peak is seen

Fig 12 : invariant mass of (Kpipi) with tighter cut on dca of daughters to the secondary vertex.

The cut used is 50 microns, which is actually (see Fig. 10) right the cross between signal and background.

Adding another cut on the decay length to be positive ( = real decay, no fake) enhanced a little bit the inv. mass

 

** July 7 : update **

The cuts used in the analysis code are [slightly different from those I quote earlier]

• |S_t|>2.5
• |DCA_max| < 0.1
• probability of fit > 0.01

item 2 is motivated by the following plot 

Fig 13 : DCA_XY distribution of tracks to PV : from middle and right panel, we can (almost) see that for the signal, the DCA_XY looks broader than the one for the background.

From the middle panel, a cut at |DCA_max|<.2 could be used

--> to check : rerun with :

• |S_t|>2 
• |DCA_max| < 0.2
• probability of fit > 0.001
•  

 ** UPDATE : July 8th : **
 cuts used :

• |S_t|>2.5 
• |DCA_max| < 0.1
• probability of fit > 0.001

Fig. 14 : invariant mass of D⁺ candidate for several cuts combinations [of dca1V and decay length significance] 

 

** UPDATE : July 10 **

Look at the dca1V dependence vs P_t of D⁺ candidate

Fig. 15 : P_t of D+ 

Fig. 16 : dca1V distribution for different Pt D⁺ : clearly for high Pt, the dca1V could be tightened

Fig. 17: P_t distribution of (Kππ) association for background [from hijing files used for mixing]

Compared to the signal, the distribution looks the same, allowing a comparison of dca1V for different P_t bins

Fig. 18 : 0 < P_t D⁺ < 0.5 GeV/c

 

Fig. 19 : 0.5 < P_t D⁺ < 1 GeV/c

 

Fig. 20 : 1 < P_t D⁺ < 2 GeV/c

 

 Fig. 21 : 2 < P_t D⁺ 

➲possible cut :

• 0 < P_t < 1 ➟ dca1V< 50 μm
• 0 < P_t > 1 ➟ dca1V< 40 μm

** UPDATE : July 11th : **

From the previous normalized histograms, I estimated the signal/noise and significance for the 4 bins

Fig. 22 : 0 < P_t D⁺ < 0.5

Fig. 23 : 0.5 < P_t D⁺  < 1

Fig. 24 : 1 < P_t D⁺  < 2

Fig. 25 : 2 < P_t D⁺ 

 

** update for the D+ reco ** :

I tried to use the unlike-like sign for background subtraction in the inv. mass plots.

• unlike sign is defined by (K^-π⁺π⁺)
• Like sign is defined by (K⁺π⁺π⁺)