One new HLT1 line for downstream photon conversions
New HLT1 line(Hlt1DownstreamGammaToEE) dedicated to reconstruct downstream photons.
Channels
There are a list of channels that can take profit of this implementation:
- \tau \rightarrow \mu \gamma
- \Lambda_b \rightarrow \Lambda \gamma
- B \rightarrow \mu \nu \gamma
- D^0 \rightarrow \phi \gamma
- B^+ \rightarrow K^{*+} ( \rightarrow K^* \pi^+) \gamma
The MC merge request for this line is here
Selection
The tentative selection is (everyone interested is welcome to contribute):
- m(\gamma_{ee})<100 MeV
- dA_is_electron & dB_is_electron
- armenteros_y < 60
- p_T(\gamma_{ee})>2.5 GeV
Rate
The rate is tested with:
- evt_max: 100k
- sequence: hlt1_pp_forward_then_matching_and_downstream_1200KHz
- input: exp_24_minbias_Sim10c_magdown
- input-rate: 30000 kHz
The rate is 36.000 kHz:
Hlt1DownstreamGammaToEEDecision Incl: 36.000 +/- 10.386 kHz, Excl: 27.000 +/- 8.996 kHzFor more check: rate_100k.log
Bandwidth Tunning:
- Parameters to tune the rate: minPt (default minPt=1500 MeV)
- MC to tune against:
- Existing MC:
11102202(B^0 \rightarrow K^{*0}\gamma) - Ideal MC: the one produced in the MC merge request here
- Private MDFs were produced locally using mdf_split_for_standalone_Allen.py and are available if needed.
- Existing MC:
Single hidden layer fully connected neural network (FCNN)
The preselection results in a very pure sample of downstream converted photons.
We also investigated if a NN could help to separate between prompt and displaced composites.
The implemented NN is inspired on the other downstream lines. The features used are:
- fd, eta_simple, dira_angle, log(ip), log(pt), log(min_ip), log(max_ip)
The predictions for the different samples are:
The results show that the pt performance is slightly better than the NN, then just a pt>1.5 GeV is applied.
Armentero Podolanski Plot
For the bandwidth division
| Line name | Tuning parameter | Simulation sample |
|---|---|---|
Hlt1DownstreamGammaToEE |
minPt | 11102202 |
Note: for the sim10f release we would like to request 12203303 and one D→hhɣ mode if possible.