Mengzhen - Gamma conversion in Au+Au 14.5 GeV run 14
Vz cut - not necessary, but study for SSD, so gamma should come from center of SSD
cut on DCA of V0 to PV - reduces combinatorial BG since most gammas come from pi0
BG - usually we see an increase of the BG (in this case it is rotational BG, 180 degree rotation) with mass, not decrease. This needs clarification. Suggestion - plot Foreground and BG together and plot Signal = Foreground - Background
- slide 7 - PXL detector was installed during run 14, vertical structure for IST looks strange - maybe these are cooling pipes (shown this way due to worst resolution?)
- should we see the beam pile? The radiation length is 1/3 of PXL (~0.2%Xo), so if PXL not visible, then beam pipe won't be visible either.

Suggestions:
- try ToF PID to suppress BG
- try to subtract combinatorial BG (like-sign)
- combinatorial BG should go to 0 when mass goes to 0, try like-sign BG
- remove/play with dca1 to PV, dca2 to PV cuts and see if this helps to see PXL

Robert - E/p efficiency in U+U 193 GeV
- slide 6 - firs two bins discarded, used photonic results instead (slide 7)
- slide 7 - background = side-bands BG

- why energy from simulations and data is so different? Not clear
- Mustafa: this could be related to StTriggerSimMaker - some of tables are not available (and then it uses the latests table) - not clear if this has any impact on the BEMC simulations or not (hopefully not)
- small impact of this update on the final results, the largest on R_AA for peripheral collisions
- comments to paper proposal: Fig 2 - remove red points (this is intermediate step), remove lower panels of fig. 2 (technical details of simulation tuning, not necessary)

Jonathan - D0 with TMVA -
- pileup could be important (due to ghosting), also pT dependence of variables could be useful
- for background - unlike sign pairs may be not appropriate: for instance: K+pi- in D0, this pairs could come from D0-bar, which have similar kinematics to D0
- using opening angle and momentum could be dangerous since TMVA could create a fake signal in the mass (which depends on momentum and opening angle). Jonathan indeed sees a fake signal in the like-sign mass distribution. Solution - remove either momentum or opening angle from training - probably that one, which looks similar for Signal and Background.
- suggestion: try use narrower/wider mass range for background
- using sign of daughter in the training phase could help to avoid fake signals for like-sign pairs, but may hide problem for unlike-sign pairs
- which parameters have the largest discrimination power? Jonathan will try to find those variables

Long - D0 correlations with Pythia

- slide 2 - Pythia parameters tuned by Mustafa, Xin and Shusu to STAR and LHC data,would be good to see this study. Mustafa said, that this study it will be in the paper submitted soon to the journal and publicly available
- multiplicity distribution - all simulated events, all tracks in the event included (no rapidity cut)
- slide 4 - method 2 and 3 are equivalent, but these are meant as a check if selection of D-mesons using hadronic decays gives any bias compared to the overall D-meson sample
- slide 6 - compare simulations with STAR data
- slide 7 -  not visible correlations, although there should be one. Request: make a projection plot for a fixed, high pT of D0 (for example pT = GeV/c)
- slide 10 - wrong-sign method fails if S/B is low
- slide 12 - slide-band works better, but there are some issues at low -pT, where background is larger and an assumption about the background (average of side bands) may not be valid. Possible solution  -find a function that describe the shape of the background, then calculate the BG under the peak taking into account that shape
- slide 15 - what is the S/B in the data? S/B = 1/20 for pT trigger > 2 GeV
large chi2/ndf at low S/B could be due to high statistics in the simulations, could be better in the data.

Michael - D+/- with HFT
- peak width - smaller than for D0, could be due to smaller decay energy (difference between D-meson mass an a sum of decay products)?
- DCA cut - right now cut on DCA between different pairs, but this cut should also include cut between all three tracks

Linag - KF vertex refitting
- how many D0 candidates are removed? Need to check
- slide 4 - number of D0 candidates decreases with centrality, that's why there is a sharp peak around 0 at low multiplicity
- for difference distribution - fit slices and quantify the widths of this distribution
- suggestion - use global momentum for mass calculation
- question - does the track cuts impact KF vertex re-fitting? Mustafa will produce picoDST without track quality cuts and then we could check if track quality cuts in picDST make any difference
- related comment - DCA (vector) of D0 vertex not stored in D0 trees. If we use all tracks for vertex-refitting, then D0 trees have to be reproduced. There is no such issue for HF trees developed by Jochen
- instead of removing D0 candidates using mass cut, we could remove tacks forming D0 candidate. Comment from Wei: This could (possibly) generate a bias in low-multiplicity events. Will be good to have a histogram with number removed tracks as a function of event multiplicity to check is this this is indeed an issue.