STAR HardProbes PWG

[wy157543 AT mail.ustc.edu.cn]

Hi Nihar and Isaac,

Thank you for your kind comments. The paper and note have been updated and 
can be accessed at the following links:
https://drupal.star.bnl.gov/STAR/system/files/psi2s_paper_pwg_v2.pdf
https://drupal.star.bnl.gov/STAR/system/files/psi2s_note_pwg_v2.pdf

Below are the detailed replies.

For paper:

Title: “First observation of charmonium sequential suppression in 
heavy-ion collisions at  Relativistic Heavy Ion Collider” -> “ 
Measurement of charmonium sequential suppression in heavy-ion collisions 
at  Relativistic Heavy Ion Collider”  [ “First …” in title sounds 
awkward and a reader can judge it. ]

I removed the "First" and kept the observation
--------

Line7: "We report on measurements of charmonium …” -> “We report the 
measurement of charmonium …”

Done.
--------

L48-51: “…has  been measured in Pb+Pb collisions at …” Better to 
mention what we observed at the LHC experiments and what more 
information we can get at RHIC measurement and why it is important to 
measure at RHIC energy.

Done.  The importence for psi2s measurements is described on lines 60-76
--------

L90-92: This sentence sounds redundant and you already mentioned in 
intro. Please drop it.

Done.
--------

L138: “… from data for different BDT cuts, “ it is not clear from the 
text what are BDT cuts. Please elaborate and avoid words like “cuts”. 
Use condition or criteria, etc. [Similarly for other places]

BDT cut->BDT threshold,
Added "by requiring the BDT response to be above a threshold." on line133-134 
to introduce "BDT threshold".  
--------

_ Many places “200 GeV” is used, use “ \sqrt s_NN = 200 GeV”

Done.
--------

_ It would be important to point out that what is advantage of using 
BDT/ supervised ML technique over traditional invariant Mass 
combinatorial subtraction method. A account will enhance this paper. 
Please indicate if such comparisons are discussed in AN.

The comparison was added in AN (319-322): The jpsi significence is improved 
from 133 (traditional method) to 170  (ML method) the ML technique did 
significantly improve signal salience.
--------

_ L280-289: “While all three measurements are consistent within 
uncertainties, the centrality dependence in 200 GeV Ru+Ru and Zr+Zr 
collisions seem to more closely resemble the measurements at 17.3 GeV…” 
This is not true within uncertainty . “…seem to more closely resemble….” 
Please paraphrase.

The original content is modified to：
While all three measurements are consistent within uncertainties  at 
comparable $\langle N_{\rm part} \rangle$ values, the centrality dependence 
trend observed  in $\sqrt{s_{_{\rm NN}}}$ = 200 GeV Ru+Ru and Zr+Zr 
collisions seems to more closely resembles 
the measurements at 17.3 GeV than those at 5.02 TeV. 

"at comparable $\langle N_{\rm part} \rangle$" and "trend" were added to make 
the expression clearer.
--------

I think it would be better to make Fig,3 as final money plot for this 
paper. Whereas Fig.4 and it discussion should come before double ratio 
vs Npart plot.
I find Ratio vs pT plot is less significant, due to uncertainty, than 
current Fig.3 to convey the message.

Fig,3 does convey the more important message. However, if the positions of 
Fig,3, Fig,4 are exchanged, it is a little strange to discuss ratio instead 
of double ratio after calculating pp, pA reference. If Fig,4 is placed before 
calculating pp, pA reference, the discussion of physical results is 
separated. Another reason is that Fig,3 contains calculations from both 
models, and therefore they can be introduced together. Personally, I think it 
would be better to keep the current order. it is not necessary that the money.
--------

Fig.3:
_ Please add in the legend what is that gray band on unity line. It is 
unclear.
_ Can we have p/d+Au 200 GeV CNM effect calculation from Tsinghua group? 
That would even strengthen the claim.

_ Is there any discussion or information on QGP Temperature from 
Tsinghua's group calculation?


Added " The band at unity represents the relative uncertainty in interpolated 
$\psi$(2S) to J/$\psi$ 
ratio in $p$+$p$ collisions." to the captain. 
Regarding the calculation of "CNM effect" and "QGP Temperature",
we may need to further consult Tsinghua group.
--------

---
----------------------------------------------------------
---
For note:

142. I see that after applying the procedure, there is still a residual eta 
dependence although the phi dependence is nicely removed. Is this caused by 
the imperfect agreement of the fit with the data especially in the negative 
rapidity region? Is this non-zero nsigma_e close enough to zero that it 
produces a minimal effect in the analysis?


Yes, the slight \eta dependence that still exists after recalibration does 
result from slight differences between the fit results and the data, but 
these differences have little effect on the analysis results. Because the 
recalibration is only convenient for the selection of nsigmaE cut and the 
calculation of nsigmaE cut efficiency. In addition, the small deviation is 
taken into account in efficiency correction since the correction is estimated 
with a data-driven method.

---

175. Just to make sure I understand, this cut on the invariant mass implies 
that only Dalitz decays of pi0s are included in the sample, while photon 
conversion typically falls just outside the requirement, right?

Electrons from Dalitz decay are equally good for estimating efficiencies 
compared to photon conversion. What matters is the
electron purity, not its source. The selection of the pairMass cut takes into 
account both the statistics and the electron purity.
For example, in 40–80% centrality, photon conversions are also included to 
increase statistics. (New 20-40% and 40-80% centrality graphs added)

---

204. This isn't 100% true is it? We see in the Background plot in Fig. 11 
that the track pT or p has a correlation of ~ 0.25 for example. And this 
makes sense given the relation between invariant mass and momentum. Did you 
check whether the results were modified significantly if the observables 
which have some degree of correlation with the pairMass were removed from 
consideration as training features?

Fig. 11 includes more information than the features used in ML. Only the 
features shown in Fig. 17 have been used for identification, 
all of which have low correlation with pairMass. More detail describes for 
the features used in ML is added in AN (190-200)
---

Eq. 4. I understand the procedure you're following to change the mean and 
sigma of the embedding nsigma distributions, but the subscripts 1,2,3 were a 
little confusing to me for a second. Could you either explicitly explain them 
in the text (e.g. "the numerical subscripts refer to a value in the embedding 
at the corresponding step in the correction procedure" or similar), or use a 
less opaque notation?

Thanks for your reminding, I have added some content to better describe these 
steps. (222-227)

---
219. Sorry for this possibly silly question, but isn't there a concern that 
if you use embedding for signal and data for background that the model may 
distinguish the two based on some subtle difference not between signal and 
background but between embedding and data (that is left over after the 
feature alignment procedure)? Was the training tried with embedding for 
background as well, and if so was it approximately the same result?


We do not see the need to use embedding for background since utilizing data 
is always better. In fact, we do not have embedding for background since our 
simulation only embeds psi(2S) signal. Then the question is whether there is 
any residual difference between data and embedding after the alignment 
procedure, and if so how would that affect the results. We do not have a 
clear answer to that since if we know there is a difference we would have 
corrected it. What we can do is to check the dependence of corrected Jpsi and 
psi(2S) counts, with raw counts from data and efficiency from machine 
learning, on the BDT response. A flat distribution is seen in Fig. 19, which 
shows that embedding agrees with data quite well after alignment. 
Furthermore, variations in the results from using different BDT cuts are 
included in the uncertainty, which is another way to gauge the difference 
between data and embedding
---

Fig. 15. Just curious, what is the y-axis here? Is it bin-width and 
integral-normalized count? Or something else?

Yes, it's normalized count

---

Fig. 20. I can't tell if my eye is fooling me or not, but it looks like the 
relative statistical uncertainties on the 0-80% range are slightly bigger 
than the 40-80% range. Is this true?

Yes, this is true for ψ(2S). Because more significant suppression in central 
collisions for psi(2S) results in worse signal to background ratio. 
(At BDT cut = 0.7, the signal significance is only 1.7σ.)  

---

Fig. 25. What should we make of the fact that the Crystal Ball fits are quite 
poor?

Indeed, the templates obtained by ToyMC cannot be perfectly described by the 
Crystal Ball function. This is an interesting feature and we will continue to 
investigate that. However, what we take from the fitting are the ratios of 
various Crystal-ball parameters between Jpsi and psi(2S), which should still 
be good even if the templates do not fit ver well. In data analysis, Jpsi 
peak is fit with all Crystal-ball function parameters being free (the fit 
works well as seen in Figs. 26 and 27), while psi(2S) peak fit is constrained 
by Jpsi fit parameters and the parameter ratios from ToyMC.
---

Fig. 28. What is causing the bump ~ 500 MeV? Comparing to e.g. 
https://drupal.star.bnl.gov/STAR/system/files/JianZhou_hf_20220120_v3.pdf s. 
14 [from run 16], this bump does appear but a) to a seemingly lesser extent 
(for the same set of cuts), and b) with more of a centrality dependence. Do 
you have an insight into the difference?

In terms of the bump, I always think that comes from interplay of two 
effects: 
i) efficiency increases with pt as electron's dE/dx increases with pt and 
thus more likely to generate a TPC hit, and also the track length increases 
with pt; 
ii) on the other hand, electron's efficiency decreases with pt since at low 
pt electron tracks can bend through TPC boundaries and still get 
reconstructed while at high pt electron tracks are straight and they are more 
likely to be lost in the boundaries.

The major differences between Jian's analysis and this analysis are high 
luminosity in run16 which could affects TPC performance in terms of the bump. 
The variation in TPC occupancy from central to peripheral events is much 
larger for Au+Au than for Isobar, which is why Jian seems a larger centrality
dependence


Best,

Yan and PAs



> -----原始邮件-----
> 发件人: "Mooney, Isaac" <isaac.mooney AT yale.edu>
> 发送时间:2025-01-05 09:06:27 (星期日)
> 收件人: "star-hp-l AT lists.bnl.gov" <star-hp-l AT lists.bnl.gov>
> 主题: Re: [[Star-hp-l] ] Measurement of ψ(2S) production in Ru+Ru and Zr+Zr 
> collisions at 200 GeV is Ready for PWG Review
> 
> Hi Yan and PAs,
> 
> Due to my delay in reviewing the materials, I focused on the analysis note 
> as the paper draft can be tuned up during the GPC if necessary. 
> Congratulations to you on an extremely well-written note. I have a handful 
> of comments below, and I find that the analysis is ready to move to GPC 
> formation, pending any input from Qian. I may send comments on the paper 
> draft at some point, but that doesn’t need to delay the formation request.
> 
> Thanks,
> Isaac
> 
> 142. I see that after applying the procedure, there is still a residual eta 
> dependence although the phi dependence is nicely removed. Is this caused by 
> the imperfect agreement of the fit with the data especially in the negative 
> rapidity region? Is this non-zero nsigma_e close enough to zero that it 
> produces a minimal effect in the analysis?
> 
> 175. Just to make sure I understand, this cut on the invariant mass implies 
> that only Dalitz decays of pi0s are included in the sample, while photon 
> conversion typically falls just outside the requirement, right?
> 
> 204. This isn't 100% true is it? We see in the Background plot in Fig. 11 
> that the track pT or p has a correlation of ~ 0.25 for example. And this 
> makes sense given the relation between invariant mass and momentum. Did you 
> check whether the results were modified significantly if the observables 
> which have some degree of correlation with the pairMass were removed from 
> consideration as training features?
> 
> Eq. 4. I understand the procedure you're following to change the mean and 
> sigma of the embedding nsigma distributions, but the subscripts 1,2,3 were 
> a little confusing to me for a second. Could you either explicitly explain 
> them in the text (e.g. "the numerical subscripts refer to a value in the 
> embedding at the corresponding step in the correction procedure" or 
> similar), or use a less opaque notation?
> 
> 219. Sorry for this possibly silly question, but isn't there a concern that 
> if you use embedding for signal and data for background that the model may 
> distinguish the two based on some subtle difference not between signal and 
> background but between embedding and data (that is left over after the 
> feature alignment procedure)? Was the training tried with embedding for 
> background as well, and if so was it approximately the same result?
> 
> Fig. 15. Just curious, what is the y-axis here? Is it bin-width and 
> integral-normalized count? Or something else?
> 
> Fig. 20. I can't tell if my eye is fooling me or not, but it looks like the 
> relative statistical uncertainties on the 0-80% range are slightly bigger 
> than the 40-80% range. Is this true?
> 
> Fig. 25. What should we make of the fact that the Crystal Ball fits are 
> quite poor?
> 
> Fig. 28. What is causing the bump ~ 500 MeV? Comparing to e.g. 
> https://drupal.star.bnl.gov/STAR/system/files/JianZhou_hf_20220120_v3.pdf 
> s. 14 [from run 16], this bump does appear but a) to a seemingly lesser 
> extent (for the same set of cuts), and b) with more of a centrality 
> dependence. Do you have an insight into the difference?
> 
> > On Dec 24, 2024, at 02:04, Nihar Sahoo <nihar AT rcf.rhic.bnl.gov> wrote:
> >
> > Dear PAs,
> >
> > I have gone through the AN and paper draft. I find both the materials in 
> > good shape, besides my following comments.
> > Please take a look at my comments below.
> > If Isaac and Qian will sign off with all these comments, then we can 
> > request for the GPC formation around 1st week of January.
> >
> >
> > Title: “First observation of charmonium sequential suppression in 
> > heavy-ion collisions at  Relativistic Heavy Ion Collider” -> “ 
> > Measurement of charmonium sequential suppression in heavy-ion collisions 
> > at  Relativistic Heavy Ion Collider”  [ “First …” in title sounds awkward 
> > and a reader can judge it. ]
> >
> > Line7: "We report on measurements of charmonium …” -> “We report the 
> > measurement of charmonium …”
> >
> > L48-51: “…has  been measured in Pb+Pb collisions at …” Better to mention 
> > what we observed at the LHC experiments and what more information we can 
> > get at RHIC measurement and why it is important to measure at RHIC energy.
> >
> > L90-92: This sentence sounds redundant and you already mentioned in 
> > intro. Please drop it.
> >
> > L138: “… from data for different BDT cuts, “ it is not clear from the 
> > text what are BDT cuts. Please elaborate and avoid words like “cuts”. Use 
> > condition or criteria, etc. [Similarly for other places]
> >
> > _ Many places “200 GeV” is used, use “ \sqrt s_NN = 200 GeV”
> >
> > _ It would be important to point out that what is advantage of using BDT/ 
> > supervised ML technique over traditional invariant Mass combinatorial 
> > subtraction method. A account will enhance this paper. Please indicate if 
> > such comparisons are discussed in AN.
> >
> > _ L280-289: “While all three measurements are consistent within 
> > uncertainties, the centrality dependence in 200 GeV Ru+Ru and Zr+Zr 
> > collisions seem to more closely resemble the measurements at 17.3 GeV…” 
> > This is not true within uncertainty . “…seem to more closely resemble….” 
> > Please paraphrase .
> >
> > I think it would be better to make Fig,3 as final money plot for this 
> > paper. Whereas Fig.4 and it discussion should come before double ratio vs 
> > Npart plot.
> > I find Ratio vs pT plot is less significant, due to uncertainty, than 
> > current Fig.3 to convey the message.
> >
> > Fig.3:
> > _ Please add in the legend what is that gray band on unity line. It is 
> > unclear.
> > _ Can we have p/d+Au 200 GeV CNM effect calculation from Tsinghua group? 
> > That would even strengthen the claim.
> >
> > _ Is there any discussion or information on QGP Temperature from 
> > Tsinghua's group calculation?
> >
> > Thank you
> > Nihar
> >
> >
> > On 2024-11-02 16:21, wy157543 wrote:
> >> Dear HPs,
> >> The paper draft and analysis note for “First observation of
> >> charmonium sequential suppression in heavy-ion collisions at
> >> Relativistic Heavy Ion Collider” are ready for pwg review. The
> >> associated documents can be found as follows:
> >> Webpage:
> >> https://drupal.star.bnl.gov/STAR/blog/wy157543/Measurement-psi2S-production-RuRu-and-ZrZr-collisions-200-GeV-0
> >> Paper
> >> draft:https://drupal.star.bnl.gov/STAR/system/files/psi2s_paper_pwg.pdf
> >> Analysis note:
> >> https://drupal.star.bnl.gov/STAR/system/files/psi2s_note_pwg.pdf
> >> We would appreciate it if you could review the documents and provide
> >> us with your valuable comments and suggestions!
> >> Best regards,
> >> Yan for PAs
>