ARC review NO.1

Cesar

[*] ========= AN - v4 ===========

==>2 Data and MC samples

--General: I understood you use track reconstruction from HI Reco for D meson reco and pp Reco for

inclusive charged particles. Do you expect any influence in your comparisons due to this?

They are both above 2 GeV, where tracking between PbPb and pp are not very different.

--l99: I think is better you use a subsection "2.2" instead of "2.1.1"

Done

--l105: SubSec2.2 --> SubSec.2.3 (if apply change above)

Done

--Tables 3 and 4: could you please refer more explicitly these tables in

your text?

Done

--l113: "PYTHIA TuneZ2 " . I only see Pythia8 samples in the tables.

Changed to Pythia8

==>3 Triggering and Event Selections

--l139-140: Not clear to me which ratio gives unit. Do you normalize the

ratio of the two flat distributions?

Yes. Or it is a ratio of the two normalized distribution. The point of the figure is to show a flat distribution, so there is no bias, the absolute scale is not important here.

--l164: "Overlap between ...removed." . Can you please detail more on

this? Or add a reference if done elsewhere.

The events that fire both "MB trigger" and "centrality trigger" are removed, since we have MB and centrality trigger PDs.

==>4 Track selections and performance

--l177: the number of pixel layers with measurements...

Done

--l181: I understood that for your analysis you are using pT>2GeV and

not 0.3<pT<3GeV

Indeed, done.

==>5 D meson reconstruction

--l234: Why was needed a re-optimisation of the D meson selection?

Which figure of merit do you use for this optimisation?

"Re-optimization" means only an optimization with respect to HIN-16-007, got rid of "re"

==>6 D meson signal extraction

--l241: Section??

Done.

--l246-248: Have you tested a convolution of Gaussian and Breit-Wigner

(Voigtian)?

This was looked into from previous D meson analysis I believe, and it is concluded that we all use double Gaussian.

--Fig.16: These contributions you plot for "Dzero + Dzero^bar signal" and "k-pi swap" are obtained

from the fit or directly from simulation?

If is from the fit, how it compares with simulation?

I would put it a "semi" data driven approach. We used the same reconstruction in the MC, where we know exactly the k-pi swapped contribution is, and use that fraction as a fixed parameter, to do the invariant mass fitting in the data, so it is indeed from the fit but also constrained by the MC. This approach has been adopted by many previous analysis.

--Fig.18: The Chi2/Ndof is not good. What is the impact of this in your analysis?

I see you discuss this in the systematic uncertainty section, then it would be good you mention and refer here.

Done. We have studied it as a systematics.

==>7 Analysis Technique

--l266-267: A, B??

Typo. Done.

--l269-271: "Because of the symmetric...three sub-event methods[27]."

I could not understand what you mean with this sentence.

The asymmetric system would not be able to use the 2-sub events as was done in this analysis, for example pPb. Therefore, that's why we use three-sub event methods there. Here, as for v1 analysis, it is more trivial to use 2-sub events.

--l335: Equation 13 and 17 --> Eq. 10 and 11.

Done

--l336-337: Why linear function? Okay, I see you discuss this in the systematics section.

Please, refer to this here?

Done.

--l338-340: Why you say that the "k-pi swapped" are from "real Dzero or Dzero^bar". I am not sure the word

"real" is correct to be used here.

Rephrased, "signal Dzero or Dzero^bar"

--Figs19-22:

1) What determines if the v1 will have a "peak" or a "valley" in the Dzero/Dzero^bar mass?

From my understanding it should be the sign of the value. If it is positive, it should be a peak. Otherwise it should be valley. However, since the value is very close to zero, it is not so obvious to see that for some bins.

2) Why the peak/valley is more pronounced in the Dzero/Dzero^bar case than in the "Dzero+Dzero^bar"?

I think it is essentially the statistical flucutations are larger in the separate case, making the struture more obivous. The statistical uncertainty should be considered, which is one of the parameter from the fit that you cannot see from the figure.

--l360-361 and l363-364: What you present in the figure24 is a MC

simulation or a fit?

It is a MC template fit on the data. Exactly the same approach as HIN-16-007 and HIN-17-003

--l364: "pT range 3-30GeV" . In the plot I see 2-30GeV.

Done. 2-30 GeV

--Eq19: I could not understand how you obtain this equation. Could please describe how you do it or add a ref

if this is the case.

This can be also found in HIN-16-007 analysis note. But if you spend 5 minutes, it can be derived.

==>8 Results

--l389-390: "As goes to forward...starts to decrease."

Where can I see this? the plot goes until |y|=1.5. Add ref if needed.

Changed.

--l394: "..field []"??

Done

--l409: Indicate here that you discuss this systematic uncertainty in

Section9.

Done.

==>9 Systematic uncertainties and cross checks

--l436-439: Not clear to me how you got the systematic in the slope. Do you consider the

difference between the slope with a fit only considering the statistical uncertainties and a

fit considering all the systematics in the points?

I use the systematic uncertainty on each point, and make a set of 64 (there are 6 points) combinations with each point at its upper or lower bound of their systematic uncertainty while keeping other 5 points the same. Linear fits to all of them and use the maximum deviation on the slope for the systematic uncertainty. When the points are moved according to their systematic uncertainty, their statistical uncertainty is not changed.

--Fig.29: Why the magnitude of the uncertainties in the right plot is

much smaller than the previous cases?

This is due to the fit. The uncorrelated error for correlated dataset is sqrt(sigma_a^2 - sigma_b^2), and when the errors are close to each other, their uncorrelated error is very small. The Pol2 has a more similar statistical error with respect to Pol1, comparing to Pol0 fit.

========= PAS - v1 =================

--l19: the ALICE Collaboration

Done

--l42: "...conductivity, of the ...". remove comma.

Done

--l64: one --> ones

Done

--l68: Minimum bias -> Minimum-bias

Done

--l72: please, define primary vertex.

I believe this is ok to leave it like this. Previous publications did not need to define primary vertex.

--l105: units for DCA?

Done. cm

--l115: A,B?

Removed.

--l125: "\Pi" -> \pi

Done

--Figure1:

1) the notation in y-axis looks not consistent with the one used in the text/eqs. Can you use the same notation?

2) Can you try to increase the size of the font in the plots? Overall it looks too small for me.

Done.

--l155: Eq.5???

Done

--l174-175: please, mention here that these corrections are taken from

MC simulations.

Done.

--l197-198: "As goes to forward...decrease." . Do you have a ref for

this?

Done.

--l202: "filed []" .Ref???

Done.

-->References

--l232: "Fiz.74,408(1978)" ..Please add space

--l243: "Nucl. Phys. A" (A not boldface).

Use just the first page "102".

--l247: same as in "l243"

--l342: "Phys.Rev.C". Please add space

--l351: "Eur. Phys. J C". "C" not boldface.

Done

Helena

=> section 7.1: Which tracking efficiency did you use for calculating

v1odd, pp or HI?

For charged particles, since it is from a pp reconstruction dataset, the corresponding efficiency from pp reconstructed MC was used. The default result of D meson is not efficiency corrected, but the systematic check of efficiency correction is done on the HI reconstructed MC sample.

=> L. 238: D^0 reconstruction: in AN referenced in the note (HIN-15-005),

the track selection is for |eta| < 1.1 or |eta| < 1.5 (HIN-16-001). In this

analysis it is |eta| < 2.4. What was the motivation for that? What happens

to the eta region that wasn't covered previously?

Previous analysis was looking into the pT dependence with mid-rapidity. Therefore, higher rapidities were not considered. That's why the "re-optimization" of the signal significance is needed for this analysis, mainly for the forward rapidity. This v1odd is studied as a function of rapidity usually, and it is our advantage to have a large coverage to perform the study.

=> Figures 16 -- 18: Is the bin really 5 MeV as indicated?

No. Typo. It is 1 MeV instead.

=> L. 317: Is there a plot of the confirmation that the v1,bkg is the same

for the D0 and \bar{D}^{0}?

Yes. Added in the analysis note.

=> section 9.2 D: it would be nice to have a clearer explanation on how

this was done. and how was the systematic uncertainty quote chosen.

This is explained in detail in AN, and rephrased.

=> Fig. 28, right (but it also applies to others): if I understood it

correctly, you plot the difference of the constant and pol.2 assumptions

for the background wrt the default v1 bkg used in the analysis. then you

fit a constant to this difference and use it to determine the systematics.

For the positive rapidity Pol2 difference, what is the chi2/ndf of this

fit? And how from those fits do you determine the systematics?

Unfortunately, the Chi2 of these fits cannot really tell much about how to determine the systematics. Also, recommended by the Statstics Committee, the analyzers need to judge the situation case by case about how to quote the systematic uncertainty. For this analysis, the quoted systematic uncertainty covers the constant fit to the "diff".

Fuqiang Wang

My major comment at the moment concerns systematic uncertainties. It's

good to have the breakdowns of the syst. uncertainties in Table 7 of the

AN. You quote an average syst. uncertainty for all rapidity bins since

no obvious dependence is observed. I agree this is a sensible way to

quote the uncertainty. You have v1 of D+Dbar, D, Dbar listed separately,

which is good. I want to see the syst. uncertainties to be also listed

separately for v1 of D-Dbar. I'm not sure we can simply take them to be

correlated between D and Dbar. I know it's tough with limited

statistics, but we should look at v1_D - v1_Dbar for each set of cuts

and see the variations in the results. I think we should also have the

syst. uncertainty studied for the slope, i.e. get the slope for each set

of the cuts, and then get the variations in the slope.

The problem of doing what you suggested is that it is very hard to quote the systematic uncertainty everywhere consistently. Using the difference from the slope directly from each systematic source, it is then not consistent with what we quote from the D-Dbar difference directly, or vice versa. In addition, as pointed out once, the difference on D-Dbar, or the slope of each systematic source, are much more sensitive to the statistical fluctuations. This will make our systematic uncertainty mostly dominated by statistical uncertainty.

Can you zoom into the y-axis for those v1 vs mass plots?

Done.

-- ZhoudunmingTu - 2018-03-19