Upsilon pp, dAu, AuAu Paper Documents

This page is for collecting the following documents related to the Upsilon pp (2009), dAu (2008) and AuAu (2010) paper:

• Paper Proposal (Most Recent: Version 3)
• New in v3: Now says we're going for PLB and has the E772 and MC plots included. Also has |y|<1.0 results
• Technical Note (Most Recent: Version 6)
• New in v3: AuAu consistency analysis and expanded summary table
• New in v4: Added JPsi study of linewidth and 1S numbers
•         New in v6 : Final version for paper as resubmitted to PLB
• Paper Draft (Latest: Version 25)
• New in v26: Updated with changes made in PLB proof
• v25-resub: Version as re-submitted to PLB (no line numbers).
• New in v25: Updated acknowledgements.
• New in v24: Minor changes to discussion or TPC misalignment
• New in v23.1: Added systematics to Fig. 3
• New in v23: Updated with comments from Lanny and Thomas. Changes are in red.
• New in v22: Made changed based on GPC responses to our responses to the referees. Also, all Tables are now correct. Changes are in blue.
• New in v21: Changed in response to PLB referee comments. Changed results to likelihood fits. Added binding energy plot. Tabs. II and III are NOT correct.
• New in v20: ???
• New in v19: Updated with minor comments from Thomas on Nov 25.
• New in v18: Incorporated lost changes from v16. Added 3 UC Davis undergrads to the author list.
• New in v17: A few more changes from GPC comments and addition of AuAu cross sections
• New in v16: Changes from GPC comments after collaboration review
• New in v15: Collaboration review changes
• New in v14: English QA changes
• New in v13: Mostly minor edits suggested by Lanny and Thomas
• New in v12: Updated the MC section to addredd |y|<0.5. Also did some other, minor graphwork on fig 3b
• New in v11: Updated with latest comments from the GPC. Official version before the first GPC meeting
• New in v10: Updated from PWG discussion. Cleaned and enchanced plots
• New in v9: Cleaned up v8
• New in v8: Added analysis of 1S state and discussion of E772
• New in v7: Made many changes based on first round of GPC e-mails. Summaries of changes and responses can be found on the responses sub-page.
• New in v6: Cleaned up most plots. Reworded end of intro. Cleaned up triggering threshold discussion. Added labels for subfigures.
• New in v5.1: Made stylistic clarifications and fixed a few typos. Updated dAu mass spectrum legend to explain grey curve.
• New in v5: PLB formatting and some plot clean-up
• New in v4: E772 results and |y|<1.0 and |y|<0.5 both included for AuAu

Thanks to all the people who submitted comments. These have helped to improve the draft.  Please find the responses to the comments below.

1) Page 3, line 40, „The cross section for bottomonium production is smaller than that of charmonium [8-10]...“, check it, is there any cross section for bottomonium production in these papers?

Answer: Both papers report a quarkonium result.  The PHENIX papers quote a J/psi cross section of ~178 nb. Our paper from the 2006 data quotes the Upsilon cross section at 114 pb.

2) Page 3, lines 51-52, „compared to s_ccbar approx 550 - 1400 mb [13, 14]). ...“. It should be checked, in [13] s is about 0.30 mb and in [14], Tab.VII, s is about 0.551 – 0.633 mb.“.

Answer: In Ref. 13, the 0.3 mb is for dsigma/dy, not for sigma_cc.  To obtain sigma_cc, one has to use a multiplicative factor of 4.7 that was obtained from simulations (Pythia), as stated in that reference.  This gives a cross section of ~ 1.4 mb, which is the upper value we quote (1400 \mu b). In reference 14, in Table VIII the low value of 550 \mu b is the lower value we use in the paper.  So both numbers we quote are consistent with the numbers from those two references.

3) Page 3, line 78, „...2009 (p+p)...“ and line 80 „20.0 pb-1... “, In Ref. [10] the pp data taken during 2006 were used, 7.9 pb-1, it seems that this data sample was not included in the present analysis. Am I true? If yes – please explain, why? If the data from 2006 are included in the present draft, then add such information in the text, please.

Answer: That is correct: the data from 2006 was not included in the present analysis.  There were two major differences.  The first difference is the amount of inner material. In 2006 (and 2007), the SVT was still in STAR. In 2008, 2009, and 2010, which are the runs we are discussing in this paper, there was no SVT. This makes the inner material different in 2006 compared to 2009, but it is kept the same in the entire paper.  This is the major difference. The inner material has a huge effect on electrons because of bremsstrahlung, and this distorts the line shape of the Upsilons.  The second difference is that the trigger in 2006 was different than in 2009. This difference in triggers is not insurmountable, but given the difference in the amount of inner material, it was not worth to try to join the two datasets. We have added a comment to the text about this:

"All three datasets were taken with the same detector configuration.
Note that the data from our previous pp result was not added to this analysis because the
amount of material in the detector was different during 2006 than in all the three datasets discussed here, preventing a uniform data analysis."

4) Page 4, Fig.1, numbers on the y-axe should be checked, because in [10], Fig.10, are practically the same acounts, but the statistic is 3 times smaller;

Answer: The number that matters is the counts in the Upsilon signal.  In Fig. 10 of Ref. 10, there is a lot more combinatorial background (because of the aforementioned issue with the inner material), so when looking at the total counts one sees a similar number than in the present paper. However, in the case of the 2006 data, most of the counts are from background.  The actual signal counts in the highest bin of the 2006 data are ~55-30 = 25, whereas the signal counts in the present paper are ~ 50 - 5 = 45 in the highest bin. When you also notice that the 2006 plot had bins that were 0.5 GeV/c^2 wide, compared to the narrower bins of 0.2 GeV/c^2 we are using in Figure 1 (a), it should now be clear that the 2009 data has indeed more statistics.

5) Page 5, line 31, „114 ± 38+23-14 pb [10]“, value 14 should 24;

Answer: Correct. We have fixed this typo. Thank you.

6) Page 5, Fig.2, yee and yY should be identical;

Answer: We will fix the figures to use one symbol for the rapidity of the upsilons throughout the paper.

7) Page 5, Fig.2 – description, „Results obtained by PHENIX are shown as filled tri-angles.“ à diamond;

8) Page 6, Fig.3a, here should be hollow star for STAR 1S (dAu) as it is in Fig.3b;

9) Page 8, line 7, „we find RAA(1S) = 0.44 ± ...“ à should be 0.54;

10) Page 8, lines 9-12, „The ratio of RAA(1S) to RAA(1S+2S+3S) is consistent with an RAA(2S+3S) approximately equal to zero, as can be seen by examining the mass range 10-11 GeV/c2 in Fig. 4.“, it is not clear, check this phrase, please;

Answer: We have modified this phrase to the following: "If 2S+3S states were completely dissociated in Au+Au collisions, then R_AA(1S+2S+3S) would be approximately equal to $R_AA(1S) \times 0.69$.  This is consistent with our observed R_AA values, and can also be inferred

by examining the mass range 10--11 GeV/c^2 in Fig. 4,
where no significant 2S or 3S signals are seen."

11) Page 8, line 26, „CNM“, it means Cold Nuclear Matter suppression? – should be explained in text;

Answer: The explanation of the CNM acronym is now done in the Introduction.

12) Page 9, line 30-31, „The cross section in d+Au collisions is found to be = 22 ± 3(stat. + fit)+4- 3(syst.) nb.“, but there is no such results in the draft before;

Answer: This result is now given in the same paragraph where the corresponding pp cross section is first
stated, right after the description of Figure 1.

13) Page 9, line 34, „0.08(p+p syst.).“ à „0.07(p+p syst.).“, see p.7;

14) Page 10, Ref [22], should be added: Eur. Phys. J C73, 2427 (2013);

15) Page 10,, Ref [33] is missing in the draft.

Answer: We have now removed it. It was left over from a previous version of the draft which included text that has since been deleted.

1) Replace 'Upsilon' in the title and text with the Greek symbol.

2) use the hyphen consistently across the whole paper, for example, sometimes you use 'cold-nulcear matter', and at another place 'cold-nuclear-matter'. Another example is 'mid-rapidity', 'mid rapidity', 'midrapidity'...

Answer: On the hyphenation, if the words are used as an adjectivial phrase, then those need to be hyphenated.  In the phrase "the cold-nuclear-matter effects were observed", the words "cold-nuclear-matter" are modifying
the word "effects", so they are hyphenated. However, from a previous comment we decided to use the acronym "CNM" for "cold-nuclear matter", which avoids the hyphenation.  We now use "mid-rapidity" throughout the paper.

3) For all references, remove the 'arxiv' information if the paper has been published.

Answer: We saw that published papers in PLB do include the arxiv information in their list of references. For the moment, we prefer to keep it there since not all papers are freely available online, but arxiv manuscripts are. We will leave the final word to the journal, if the editors ask us to remove it, then we will do so.

4) Ref. [33] is not cited in the text. For CMS, the latest paper could be added, PRL 109, 222301 (2012).

Answer: Ref [33] was removed. Added the Ref. to the latest CMS paper on Upsilon suppression.

5) For the model comparisons, you may also compare with another QGP suppression model, Y. Liu, et al., PLB 697, 32-36 (2011)

Answer: This model is now included in the draft too, and plotted for comparison to our data in Fig. 5c.

6) page 3, line 15, you may add a reference to lattice calculations for Tc ~ 175 MeV.

7) Fig 1a, \sqrt{s_{NN}} -> \sqrt{s}. In the caption, |y| -> |y_{ee}|

8) For the dAu rapidity, the directions of Au-going and p-going should be explicitly defined.

Answer: We also realized that this was important to do. This is now done by adding the sentence: "Throughout this paper, the positive rapidity region is the deuteron-going direction, and the
negative rapidity region is the Au-going direction. "

9) Fig.2a, the label of x-axis, 'y_{ee}' -> 'y_{\Upsilon}'. In the caption for Fig. 2a, Ref. [21] should be cited after 'EPS09 nPDF'.

Answer: We moved the citation to the first part of the caption.

10) page 5, around line 28-29, please mention explicitly this result is for p+p 200 GeV.

Answer: Done. The text now reads "we calculate a production cross section in p+p collisions..."

11) page 7, line 33, add space after N_{part}

12) page 7, line 36, Fig. 5c -> Figure 5c

13) page 7, line 46, remove 'bin from'

14) page 7, line 55, 'the latter' -> 'the former' ?

Answer: Split the sentence into two, and explicitly stated "The level of suppression we observe for
|y|<0.5 stays approximately constant from dAu up to central AuAu collisions. " to make it clear.

15) Fig. 4 a, b, and c, '30%-60%' -> 30-60%, '10%-30%' -> '10-30%', '0%-10%' -> '0-10%' In the caption, |y| -> |y_{ee}|

16) Fig. 5, the label of the y axis better to be the same style as Fig. 2

17) Page 9, line 33, line 45, when quoting the RdAu and RAA, why omit the p+p stat. errors? Also the p+p syst. err. in line 34 is not the same as that in page 7, line 41, please check.

Answer: The p+p stat. errors are combined together with the Au+Au stat. errors because it is straightforward to combine stat. errors, and we just quote the combined stat. error. Syst errors are fixed.

1. In the legends of Fig 1 and Fig 4, the line color for the like-sign and unlike-sign should be blue and red, instead of black.

2. On page 5, line 29, it is not specified whether this is for p+p or dAu.

Answer: Done. The text now reads "we calculate a production cross section in p+p collisions..."

3. The directions of the d and Au beams were not defined: which goes forward and which backward in y? It will be good to specifiy the direction, and briefly discuss the different physics we expect from the forward and backward regions.

Answer: We also realized that this was important to do. This is now done by adding the sentence: "Throughout this paper, the positive rapidity region is the deuteron-going direction, and the
negative rapidity region is the Au-going direction. "

4. Page 7, line 50, "Pb+Pb" should be upright.

5. Page 7, line 55-56, "the latter" should be the model, which doesn't look constant. It seems you are talking about the measurements. Then it should be "the former".

Answer: Split the sentence into two, and explicitly stated "The level of suppression we observe for
|y|<0.5 stays approximately constant from dAu up to central AuAu collisions. " to make it clear.

6. Page 8, line 13-14, "in d+Au to be $2\sigma$ from unity and consistent with unity in peripheral" -> "to be $2\sigma$ from unity in d+Au and consistent with unity in peripheral"

7. Page 8, line 22, "modeling"

Answer: There are two aims: to incorporate... and to model ... Since we use the infinitive form in the description of the first aim ("to incorporate") we also use the infinitive form ("to model") in the second aim.

8. Page 3, line 82, "pQCD" -> "perturbative QCD (pQCD)"
9. Page 5, line 6, "perturbative QCD" -> "pQCD"

10. Page 5, Fig 2, the caption says "Results obtained by PHENIX are shown as filled triangles", but they are "diamonds", not triangles in figure.

11. Pg 4 Line 1 : Barrel Electro-Magnetic Calorimeter (EMC) - Barrel Electro-Magnetic Calorimeter (BEMC) and replace EMC with BEMC throughout.

12. Pg 4 Line 65 : |y_{\upsilon}| - |y|. In the following Figure 1, its |y_{ee}| < 0.5 in figure panels and |y| < 0.5 in caption. Inconsistency, if all of them are same.

13. Pg 5 Line 1 : The data are fit .. - The data are fitted ..

Answer: Both forms are grammatically correct. The past participle can be either "fit" or "fitted".
http://en.wiktionary.org/wiki/fit#Verb
We kept the text as is.

14. Pg 5 Line 6 : via a perturbative (pQCD) next to leading order (NLO) - via a next to leading order (NLO) pQCD

15. Pg 5 Line 41 : ... with respect to ... - ... with respect to the ...

Answer: It is correct as written, usage: with respect to (something). One could also use "with respect to the" but then we would need to add another noun, for example as in, "with respect to the binary-collision-scaling expectation". We felt the original form was ok.

16. Pg 5 Line 46 : ... yield ... - ... yields ...

17. Pg 6 Line 25 : The present data ... - The present data in which figure ?

Answer: It is now made clear in the text that this refers to Figure 2b.

18. Pg 6 Caption for Fig. 3 : Use a) and b) instead of top and bottom

19. Pg 6 Caption for Fig. 3 : x_{F} in caption and X_{F} in figure

20. Pg 8 Line 26 : when CNM first appears, it needs to be spelled out.

Answer: Done, it is now given in the Introduction.

21. Pg 9 Line 28 & 31 : The term B_{ee} \times is missing in front of d\sigma/dy

Page 3, Line 71. Why only p+p and d+Au? Why is the Au+Au cross-section not extracted?

Answer: We typically extract the yield per event in AA. This can be transformed into a cross section if we use the integrated luminosity. To get from a total number of minimum-bias events to an integrated luminosity all that is needed is the hadronic cross section for AuAu collisions, which is typically obtained using a Glauber model.  We typically don't quote it mainly because what the community wants to know is R_AA itself.  That is the quantity that the theorists typically calculate, and so we had received guidance to not include a cross section. (It was actually included in earlier versions of the draft.) Given this call for including it, we have now brought it back to the draft.

Figure 2. It might be more appropriate to include the description of the symbols in the figure caption rather than in the text. The legend might be reformatted so the description of symbols has the same structure for STAR, PHENIX, and Ramona Vogt. Why not use a consistent label for what we understand to be the same quantity expressed on the horizontal axis? (Figure 2a uses the rapidity for e+e- while Figure 2b uses rapidity for the upsilon.)

Answer: The caption now describes the symbols too. We left the description in the text also, to help the reader.

Page 5, Line 4. The wording in the text makes it sound like the red line in Figure 2 could refer exclusively to the upsilon production.

Answer: We have reworded this part to:
"The data are fit with a parameterization consisting of the sum of various contributions to the

electron-pair invariant-mass spectrum. The lines in
Fig. 1 show the yield from the combinatorial background (dashed blue line),
the physics background from Drell-Yan and \bbbar\ pairs
(dot-dashed green line), and finally the inclusion of the \upsi\ contribution
(solid red line)."

Page 6, Line 18. It might be more appropriate to discuss here why the mid-rapidity point is lower than the prediction (rather than later in the text).

Answer: In a sense, the next paragraphs and figures are meant to discuss this point being low. We use R_dAu to have more discussion of the model predictions (and show their uncertainties). We next compare our result to previous measurements, which show a similar suppression.  We added the sentence "To study this observation for \dAu\ further, we make a closer comparison to models and to previous measurements of \upsi\ production in p+A collisions. " to highlight this.

Page 7, Line 11/Figure 3b. It is unclear how the plot in terms of Feynman-x improves the comparison of rapidity coverage.

Answer: We added the x_F plot because the E772 data were given in x_F. We can massage our data to get x_F from rapidity making some estimates about the pT, which we can do because we have all the information on the Upsilon 4-momenta for our data, but we do not have this information for E772. So in order to compare to their result, it was best to not touch their data and massage ours, with intimate knowledge of ours, than to keep everything in y_Upsilon but having to massage their data without knowledge of their pT distribution so that we would only be guessing as to the correct y_Upsilon that would correspond to a particular x_F range.

Page 9, Line 30. This result in the conclusions does not seem to have been presented in the body of the paper.

Answer: This result is now given in the same paragraph where the corresponding pp cross section is first
stated, right after the description of Figure 1.

1. legend of Fig. 1b
--------------------
I would rather put R_{dA}=1 (not R_{AA}) to be consistent with the figure caption and the main text

2. Fig 2a and discussion in the text
of the results for pp at positive and negative rapidities.
----------------------------------------------------------
I found it a bit awkward that we are presenting results just after folding in data at positive and negative y.
Of course the physics for pp is symmetric wrt y=0,
but it would be better to present separately results
for -1 < y < -0.5 and 0.5 < y < 1.0 to show that indeed the results are consistent.
(Also as a cross check of correctness of including all experimental corrections, and nothing to hide)

Answer: We did check that the results were consistent for pp, but we wanted to maximize the statistical power of the data, given that we are still somewhat statistics limited.  Note that the acceptance and efficiency is lower for the 0.5 < |y|< 1.0 region, so that is why we wanted to add the two in pp, thanks to the symmetry, to show our best results.  For the d+Au case, as we say in the paper, we did leave the analysis in distinct rapidity regions because the system is not symmetric.

3. legend of Fig. 2a
--------------------
For STAR and PHENIX points it would be more transparent,
if the legend would have similar layout as for NLO pQCD CEM.
I.e. 'STAR' in a single line followed by two lines
'pp' and 'dAu/1000' and analogously for
PHENIX/

4. line 2 on page 7
------------------
"their deuterium result" => "their pd result"
would be more straightward statement
(I assume E772 had a liquid deuterium target to study pd collisions)

Answer: Done. And yes, we say in the text that they had a liquid deuterium target.

5. Fig. 4
---------
The curves for combinatorial background should be made smooth
like for all other curves, not going in steps.

page 4, line 1 and in further occurences: shouldn't it be BEMC instead of EMC ?
--------------------

page 5, line 1: shouldn't it be "The data are fitted"
---------------------------------------
Answer: Both forms are grammatically correct. The past participle can be either "fit" or "fitted".
http://en.wiktionary.org/wiki/fit#Verb
We kept the text as is.

Overall it is a very well written paper and important results.

1. Acronyms in the introduction should be defined there (RHIC, LHC, pQCD or even QCD)
--------------------

p. 3, l. 60: you use "cold-nuclear-matter effects" without defining what "cold" and "hot" nuclear matter is. It would be good to introduce these terms when you talk about QGP and then other possible sources of suppression (line 52-63)
--------------------

p.8 l.26 - CNM should be defined
--------------------

Answer: It is now defined in the Introduction.

p.8 l.44-48 - it is not clear from the text how exactly CNM and QGP effects were combined for the scenario 4.
--------------------

Answer: We now state "For scenario 4), the expected suppression is simply taken to be the product of the suppression from scenario 2) and scenario 3)."

p.9 l.29 "with NLO" -> with "pQCD NLO"
--------------------

Answer:  In the rest of the paper, we have used NLO pQCD, so at this point, it should be clear that when we are talking about
a Next-to-Leading Order calculation, we are implicitly talking about a perturbative QCD calculation (the fact that we are talking
about "orders" in a calculation implies
that we are talking about perturbation theory,
and this entire paper deals with QCD), so it should be clear from the context.

Figures: Caption of Fig 2: " from EPS09 with shadowing" - "EPS09" is nPDF which includes shadowing already, maybe write "due to shadowing using EPS09"?
--------------------

prediction uses the EPS09 nPDF which includes shadowing"

Fig 2 and Fig 6 - the contrast of the figures could be improved - for instance lines for models in Fig. 2 are barely visible when printed in black and white

Reader 4: Fig. 1 and Fig 4 - The information on pT range,
in which the signal is presented, can be added.
-------------------------------------

Answer: We added a sentence at the end of the "Experimental Methods" section to state: "For all results we quote, the Upsilon data are integrated over all transverse momenta."

The new p+p result is significant, why is it not in title?

Answer: We already have one paper that is all about the pp cross section. Our result in this paper is an improvement, but the new results on suppression are the highlight of the paper, and we felt they deserved to be emphasized in the title.  If we change the title to something like "Upsilon production in pp, d+Au, and Au+Au collisions at sqrt(s_NN) = 200 GeV" would include the pp result in the title, but it will not mention suppression.  We prefer to emphasize the suppression, as that is the new, important result. Since we are attempting to publish in Physics Letters B, we felt it was more appropriate.

The paper is not clear in many places, and would be helped from a re-­write keeping the audience in mind, i.e. not nesc. an expert in HI.  It was commented that in particular the introduction on page 2 line 56 to 66 has much expert knowledge assumed, but does cover the field. Some examples are given below in the individual comments.

Answer: We tried to make the introduction section a short review of the field so that a non-expert could follow.  We don't understand which expert knowledge is assumed in the introduction in lines 56-66.  We certainly have strived to make the paper clear, and we will look for the specific comments and suggestions below.

The different RAA values appears multiple places in text. We think it is important to present these in tabular form, particular since so many numbers are presented RAA |y| <0.5, 1 centrality and collision system. Noted by several readers. Page 7,30-­‐50 Page 8, 4-­‐20

Answer: A table with all the values has now been added to the paper.

The definition of RAA seems a bit colloquial, normally this is defined vs. e.g pt, but in the case of the Upsilon it is our understanding this is an integral of the cross section over all (or some) pt-­‐range divided by the pp . The paper should define this clearly.

Answer: We specifically wrote in the paper the equation used for R_AA.  This is as clear a definition as we can make.  We also now specify that our measurements are integrated over all pt.

The abstract should reflect the conclusion of the paper, this does not at present.

Answer: The abstract includes the most central R_AA and the R_dAu values, which are some of the most important results of the paper.  We also state three of the most important conclusions we draw from the data:

Our results are consistent with complete suppression of excited-state Υ mesons in Au+Au collisions. The additional suppression in Au+Au is consistent with the level expected in model calculations that include the presence of a hot, deconfined Quark-Gluon Plasma. However, understanding the effects seen in d+Au is still needed before fully interpreting the Au+Au results.

The most important observation, which is the unexpected observation that R_dAu is the same as R_AA for central events in the |y|<0.5 region, is the reason why we wrote the last sentence in the abstract.

The paper needs clarification in regard to the material budgets for the 3 running periods. The text alludes to differences, e.g. how its included in the fits. Why not summarize the rad lengths for pp, dAu and AuAu to be precise. If not, it is very hard to follow the different figures, and clearly different response functions for the Upsilon peaks.
Answer: This is now fixed. The sentence in question alluded to differences in the material budget, but for the three years there were no differences in the material budget. Only the differences in the detector occupancy and the detector calibrations affect the width.  In the new version, we also mention explicitly that the 3 running periods have the same material budget.
Page 6 Please define XF, how you used XF. It was not found in the analysis note, and we have problems to understand how we can reach XF~0.4 when measuring at mid-­‐ rapidity Xf= pz/pzmax normally, so are we seeing Upsilons with Pz=40 Gev in y<1? In any case its not defined.

Answer: Good catch! We made a mistake in the calculation for STAR, we accidentaly used the E772 value for the beam momentum. We were originally thinking of transforming their values of x_F to rapidity, but then when we decided not to move their data and change ours to x_F, we did not use our value for the beam momentum.  The figure is updated.  But the most important point which is at y=0 remains at x_F=0, so the comparison to the level of suppression seen by STAR and E772 at x_F=0 stays the same.

On page 8 line 7 it say RAA = 0.44+-­‐… where as figure 5 c clear as R > 0.5. Please clarify.

Answer: It should be 0.54, it was a typo, and is now fixed.

The discussion between the |y|<1 and |y|<0.5 is not clear cut, particular for the AuAu; It is surprising to have such difference. Is it possible this reflects un-­accounted systematic error or is it all statistical? It does take away from the final conclusions since for |y|<0.5 there is no suppression relative to dAu where as there is for |y|<1. This clearly translates into the interpretation of the interesting model comparisons presented in fig 6. Conclusions in the text are iffy. The data in fig 5 as given do NOT indicate any (significant) centrality dependence vs. Npart , only for RAA(1S). Is that the message that should come across?

Answer: We have discussed the differences in the |y|<1 and |y|<0.5 in the PWG, precisely to try to make sure that the results we are observing in |y|<1 and in |y|<0.5 are statistically consistent.  One of the results is a subset of the other, so one must be careful to take into account the correlations.  This study is in the technical note, in section 6A (page 33).  We concluded that the results are self consistent.  As to whether the result is statistical fluctuation, this is a possibility, but that is the case for any result, and the only way to remedy that situation is to run more dAu.  As to whether it could be a systematic effect, we have done the analysis in |y|<0.5, in |y|<1, and in 0.5<|y|<1 where for each we use the same methods for extracting the signal, for applying efficiency and acceptance corrections, for estimating the backgrounds, etc.  So if there is a systematic effect, it would affect the |y|<0.5 and the 0.5<|y|<1 region in the same way, and therefore it would not lead to differences between these two regions.  We do not think that this "takes away" from the final conclusions, because it is an observation that is not expected if there should be binary scaling in dAu, and it makes the result more interesting.  The reason why we included the E772 data was precisely because we observed such a striking suppression in dAu. So indeed, the fact that the data in Fig. 5b do not show a significant centrality dependence vs. Npart is one of the most important observations of the paper. And with the E772 data, we can point to a previous result that shows a similar level of suppression in pA.  Therefore, this paper will serve to exhort the community to take a closer look at Upsilon suppression in pA or dA.  We do not understand the comment about conclusions being "iffy". If there is a specific conclusion that does not seem to be supported by the data, then we can address that.

The last sentence in conclusion seem exaggerated, and not documented from text just remove.

Answer: One of the main points of the paper is that in Fig. 5b, as we explain in the previous answer, there is no evidence for a significant centrality dependence of Upsilon suppression in dAu.  The models predict the level of suppression we see in AuAu, but one of the key results of the paper is the suppression seen in dAu. The GPC strongly advocated to include a sentence in the conclusions of the paper that cautions readers that one must understand the dAu suppression before any strong claims can be made. The last sentence was rephrased slighly to better reflect this.

In abstract suggest the remove the sentence “Our measurements p+p…” and add to the text where relevant in the introduction. Not really relevant.

Page 3 line 34: it is not at all obvious how the 2 statements (deconfinement and high temperature phase of lattice QCD where color is an active degree of freedom) in this sentence are scientifically connected.

Answer: The connection is that color Debye screening, which is the original effect proposed by Matsui and Satz,
requires a quark-gluon plasma where the color charges of the high-temperature plasma screen the heavy-quark potential that binds the bottomonium (or charmonium) states.  This is one of the key ideas in QGP physics.

Page 3 line 59 for a non HI guru this argumentation is basically impossible to follow. Also ccbar and bbar pairs are produced the same way through gg fusion so why should there be a difference.

Answer: It seems that the question arises because the inquirer did not follow that the arguments presented are about final state effects, since the comment about ccbar and b-bbar pairs being produced through gluon fusion is about their production in the inital state, not about the possible ways that they can be broken up in the final state.  The comment about
the interaction cross section of the Upsilon with hadrons applies to the final state, once the hadrons are produced.  The size of the upsilon meson is much smaller than the J/psi meson, and the corresponding cross section of an Upsilon to interact with a final state pion (and then break up into a pair of B mesons) is much smaller than the cross section for a J/psi to be broken up by a pion into a pair of D mesons.  We will add a comment that the effects discussed in this section are final state effects.

Page 3 line 46. There is no reference to statistical recombination.

Page 3 line 78 there is no issues using the 2008 dAu data even so other analysis claim they cannot publish because of the non perfect tpc alignment?

Answer: We put a lot of work to take into account the effect of the TPC misalignment.  This is discussed in the Technical Note in Section V.F, page 29. In particular, the 2009 pp data was originally processed with the same misalignment that the 2008 dAu data and the 2010 AuAu data both have.  The 2009 pp data was subsequently reprocessed with fixed calibrations, and we studied the effect that the distortions had on our invariant mass reconstruction on an event-by-event basis, i.e. comparing the mass obtained in the production with the misalignment and then with the misalignment fixed on the exact same event.  This allowed us to characterize the effect of the misalignment and to take it into account in embedding for the line-shapes and then in the extraction of the Upsilon yield via the fits using those line-shapes.  This was studied extensively in the PWG in large part because we wanted to make sure that any issues regarding the misalignment would be dealt with appropriately.  We cannot comment on other analyses, but if they can also study the differences in the two pp 2009 productions, that could help them to account for the TPC misalignment in their own analyses.

Fig 1 caption – comment to fit: the chi^2 of the pp fit must be horrible, any reason why the fit does not describe the data better.

Answer: The chi^2/NDF is 1.37 in the pp fit.  This is not something we would characterize as "horrible".  Given the statistics, there is not a strong reason to change the fit from using components we expect to have, namely the Upsilon states, the Drell-Yan and b-bbar continuum, and the combinatorial background.

2nd question: was the setup of STAR, especially the material budget, the same? If not, which I assume, how different are they?

Answer: The material budget was the same.  The TPC misalignment in dAu, and AuAu increases the width compared to pp. The higher occupancy in AuAu also contributes to a broadening compared to pp.  As noted above, we now explicitly state
in the paper that the material budget in all three datasets is the same.

page 5 line 6 (fig caption) ‘band’ -> box/square

Answer: The NLO calculations are shown as a band, and that is what is mentioned in the caption.

page 6 line 48: the effect at mid rapidity taking the systematic uncertainty into account is 2 sigma max. I think this is a number which needs to be stated.

Answer: We state the value of R_dAu with statistic and systematic uncertainties. We will also provide a table with all the R_AA and R_dAu values. The sentence we use in page 6 line 48 says that the suppresion is "indicative" of effects beyond shadowing, initial-state parton energy loss, or absorption by spectator nucleons.
Using "indicative" is usually warranted for effects that are of ~2 to 3 sigma significance, we certainly not claim a "discovery" (5sigma).

Itʼs a bit hard to follow the various R_AA and R_dAu quoted in the paper. A table listing the R_{AB} for the various combinations might be more useful than scattering the values through the text.

Answer: A table is now provided.

Abstract: I realize that in the abstract you donʼt want to get too technical, but omitting the rapidity range and whether it is 1S or 1S+2S+3S makes the numbers not useful.

Answer: We added a short clarification in the abstract as to the result quoted being 1S+2S+3S, and in the rapidity range |y|<1.

p. 4: Lines 55-57: the tracking and electron identification efficiencies would be the same across the three datasets, but in the previous paragraph there was discussion about differences in efficiency. Needs to be made clearer.

Answer: The text is now clear that the main thing that was chosen to be the same was the electron identification efficiency.

Fig1 The N_{--} is unclear the – runs together with the N

Fig. 2: Vogt band does not print well.

Answer: Increased the line weights and changed the colors to darken them so that they print better.

fig 2a needs ""Phenix"" in dAu/1000 (open diamonds)

Fig 2: “are shown as triangles There are no triangles,

Answer: Done. It should be diamonds.

c) Fig 3a The label A^0.96 is not the actual black curve which is (A/2)^{-0.04) according" "to the text in pg 7. Maybe writting the A^{alpha} scaling of cross section in the figure may help.

in Fig 4 where the CB in all three panels is not a smooth curve nor a histogram; it has an unusual "mexican pyramid" shape

Answer: What's wrong with Mexican pyramids? :-) The plot will now be a smooth curve.

The A to the 0.96 does not match the text in line 5 page 7

Answer: As noted above, the Figure will now display A^0.96 scaling to make clear that the line shown is not A^0.9, but
rather derived from a cross section that scales as A^0.96.

Fig. 5: Are the shaded boxes systematics in the AB system? If so, needs to be in the caption.

Fig. 6: "Our data is shown as a red vertical line with systematics shown by the pink box. There are two systematics (pp and AB). What was done with these? The pp is common to d+Au and Au+Au, so not clear, actually, what should be done.

Answer: The two systematics were added in quadrature for Fig. 6, we now state that in the paper. (Agree that it is debatable how to best combine them, but we should state what was done.)

.p. 6, lines 43-44: Do you mean y<~-1.2, rather than 1.2? Otherwise the argument doesnʼt make sense. And, where is the 1.2 from (citation)?

Answer: Correct, it should be -1.2.  We do give the reference (23) for this statement in the previous sentence.

p. 8, line 11: consistent with an RAA(2S+3S) approximately equal to zero. Would be better to quantify this as an upper limit.

Answer: This section was reworded based on suggestions from another reader. The argument now starts with the hypothesis of an approximately zero yield of the 2S+3S, states what that would imply for the R_AA(1S) and R_AA(1S+2S+3S) values, and
then notes that this is consistent with our data.

p. 9, line 1: at how many sigma was the exclusion? At 4.2 sigma, as quoted later?

Answer: The exclusion the "no-suppression" scenario had a p-value of less than 1 in 10^7 (better than 5 sigma) for all R_AA cases in AuAu. The R_dAu had a different p-value of 1.8 * 10^-5 (~4.2 sigma).

Line 18: result rather than effect reads better.

How were systematics taken into account in the quoting of “sigma”?"

Answer: The only time we quote "sigma" are for the exclusion of the "no-suppression" scenarios.  For R_AA, they would still
be excluded at better than ~5 sigma even including systematics.  For the dAu case, if the p-value is calculated with the systematic uncertainty shift we get 1.5 x 10^-3, which is about 3sigma.

a) The style of the paper is too colloquial for my taste, but I'm told that journals have relaxed their style requirements.

Answer: This is a style issue, we are certainly willing to discuss this with the editors of the journal if need be.

d) Reference [10] explains that the Combinatorial background is obtained by fitting the same charge sign pair distribution and that appears to be the case in this paper except in Fig 4 where the CB in all three panels is not a smooth curve nor a histogram; it has an unusual ""mexican pyramid"" shape.

Answer: The plots will all have a smooth curve.

Page 7 top (line)9 From the figs its not obvious there is 4.2 deviation, more like 3, can you cross check.

Clearly the difference between y 0.5 and 1.0 make the conclusion a bit waffly.

Answer: For dAu, in both scenarios we are excluding the no-suppression scenario. Both datasets are supporting this conclusion. Furthermore, the comment we make in about a 4.2sigma exclusion of the no-suppression scenario comes
from the |y|<1 measurement, which is the weaker exclusion of the two. The |y|<0.5 only serves to make this conclusion stronger.

The notation and fonts for RAA and Upsilon(1S+2S+3s) not not consistent across paper.
Answer: plots are now consistent (For Anthony).

Page 8 line 48 “ assumed a flat prior..” This reference to statistics may or may not come across well to the general reader,
Possible expand on this.

Answer: We have followed other papers in the Physics Letters B which use these same statistical techniques, and this usage was accepted.

One minor comment:" "In Fig.6, “CMN effects” should be “CNM effects”

Page 4 line 28&57, the three datasets” clarify to indicate “between the datasets from the three collisions systems”"

"line 57" "‘be the same” Really, should it not be “approximately the same".

Responses to GPC, April 2014

Thomas:

1. General: with the new text (in red) there's no a wild mix
of Au in roman and italic in normal text and in super/sub-scripts. Since Au is a chemical symbol I would put it all in roman
consistently.

I fixed the remaining instances of italicized "Au"s in the text.
2. Page 3, line 30, Sentence starting "Additionally ...".

This sentence doesn't say a lot and as I already mentioned that
I do not think the feed-down pattern is any more complex
than that in the charmomium sector. I attached a schematic
diagram. Replace Y with Psi and chi_b with chi_c and h_b with h_c
and that's it.

Why not simply saying here that the amount of feed-down into
the Y(states) is not measured at RHIC energies and then give
numbers of the next closest energy (which is Tevatron I guess).

We've changed the discussion of feed down in the introduction. We added a reference to the Tevatron results. We also discussed the direct fraction and its implication for the interpretation of the 1S results.

3. Fig 1,: I already mentioned that I suggest to turn this
into a table. The plot doesn't really provide any new insight.

Done.

4. Page 9, line 13.
"*" -> "\times" or just leave it out

Changed to \times. It helps distinguish the (1S) as an argument and the next set of parens as a mathematical expression.

5. Page 11, Line 18.

I wonder if one should add one sentence mentioning the Y suppression
in high multiplicity pp events seen by CMS. Fits in the context.

I would argue it's a little ambiguous whether we should do this. If we were citing LHC results, I think this would be prudent. However, we have yet to see evidence of Upsilon suppression (or J/Psi suppression) in pp collisions at RHIC energies.

6. Page 11, line 50.

Delete "However".

Done.

7. Fig. 6. The font size of the legend is a bit on the small side.

There's enough room to make it a tic bigger.

I've tried to squeeze a larger font in there. Thoughts?

8. Table II.
Can we really say that d-Au is 0-100%? That would be zero bias.
Wasn't there a min-bias mixed with the Y trigger. To my knowledge
we never quoted anything above 80/90%. What about simply saying

Done.

Lanny:

P3 L30 -- remove "complex" (it is an unnecessary adjective here)

OK

P4 (new) Fig.1 and red text lines 50-51, 65-69:  The efficiencies are
about the same for the 30-60, 10-30 and 0-60 at each rapidity bin.
This information probably should be in the text since HF reco. eff.
are useful to know by others in the business. I recommend putting
this information in the text in place of the above Figure 1 and lines, e.g.

"The $\Upsilon$ acceptance $\times$ efficiency for three centrality
bins (30-60%, 10-30%, 0-60%) are XX, XX and XX for respective
rapidity bins |y|<0.5, |y|<1.0 and 0.5<|y|<1.0. For the 0-10% centrality
the corresponding total efficiencies are reduced by approximately XX%."

Please check that the various uses of "total efficiency", "reconstruction
efficiency", "acceptance times efficiency" etc are used consistently and
avoid extra such terms if possible.

P5 Fig2b -- The legend "p+p x <Ncoll>" is misleading and may be what ref.2
is asking about. The grey band in 2b is not simply the red curve in 2a
multiplied by a constant (Ncoll).  There are resolution effects as discussed
on P6. The caption should say, "The grey band shows the expected yield if
RDAu = 1 including resolution effects (see text)."

P5 L8 -- Are b-bbar pair backgrounds NPE from open HF meson
decays (B-mesons)?  Just curious.

You got it.

P5 Tabel I -- I assume momentum resolution effects are included in
the line shape entries.  Ref.1 is concerned about p-resolution and in
addition to the response, this table caption should note that p-resol.
is included in the line shape errors if that is true.

It is included and is now noted in the caption.

P5 L17-28 -- I did not find any discussion in the paper about the
use of max likelihood fitting. This turned out to be a big deal and
will be discussed in the response. This parag. would be the place
to say, briefly how the fits were done.

Good idea. The following was added: "The fit is permormed simultaneously to the like-sign and unlike-sign spectra using a maximum-likelihood method."

P6 L6 -- "miscalibration" sounds scary. Can this issue be explained
in the text, and more so in the responses, so that neither referee nor
the readers are put-off by the statement and dismiss the paper's results?

We now refer to it as a misalignment as well as quantifying the effect it had on the line shape.

P6 L26 -- I recommend against arguing with the referee over simple
wording changes that have equivalent meanings.  Is there a subtlety
here that I don't recognize?

We changed the wording and Manuel played diplomat. I never intended this to be the real response; it was more for the GPC. It's now fixed.

P6 Fig.3a -- The referee is asking that the legend "Upsilon -> e+e-"
say explicitly "Upsilon(1S+2S+3S) -> e+e-".  But also change to l+l-.
She/he wants the states listed explicitly.

Done.

P8 Fig 5c caption - same issue as above with the grey band. The
last sentence in the caption should read: "The grey band ... number
of binary collisions including resolution effects (see text)."

Done.

P9 L8 -- Referring to Fig. 6c, the 10-30% RAA is consistent with unity
also. This sentence should say, "..consistent with unity in peripheral
to more-central Au+Au collisions..."  BTW, "events" is jargon which we
all use, but I think it is better to say "collisions" here and throughout
the paper unless we are specifically discussing a triggered event in
DAQ.

Fair enough. I've updated the text to reflect that the RAA in 10-30% is consistent with unity as well.

Also, I changed event to collision where approriate in the text. Those changes are unhighlighted.

P10 L8 -- "With two possibilities.." implies that CNM and QGP are

the only possibilities for reducing yields.  There is at least the
possibility of modified fragmentation of HF quarks in a
dense system.  I recommend saying "Considering two possible
sources..."  which more accurately reflects what was done; we
considered these two effects and not others.

Good point! We fixed it. Thanks.

P10 L37-39 -- Isn't the "QGP only" preferred in Fig. 8b? Why
mention the other as "consistent" and not also mention the

one that fits best?

One thing to note here is that the "QGP only" model also includes the "no suppression" model in dAu. Seeing as no suppression is disfavored by the dAu study, we can argue that "QGP only" is not really favored. We made this more clear in the text.

Thorsten:

- p3, l31-l32: I don't like the formulation too much, maybe "...there exists a feed-down pattern in the bottomonium sector, and thus melting of the higher states affects also the measured yield of the lower states."

We've changed this section. See responses to Thomas' comments for further info.

- fig 1 take a lot of space for basically not much information, maybe a table would be sufficient?
Done

- p6, l6: TPC miscalibration sounds scary, maybe non-perfect TPC calibration?

We now refer to it as a misalignment as well as quantifying the effect it had on the line shape.

- p11, l5: I'm not too happy with the A^alpha discussion: after all it is a just a fit to the data. Have you used for this statement the alpha value from our own measurement, e.g. fig 4 bottom or the integrated one from fig 4 top? The integrated one is significantly above the midrapidity one, also for E772
Fair point. We discussed what we need to in the previous paragraph and we've removed this sentence.

Responses to PLB Referees

We would like to thank the referees for the insightful and constructive comments. We discuss below our detailed replies to your questions and the corresponding explanation of changes to the manuscript. But before we go into the replies to the comments, we want to make the referees aware of changes to the results that were prompted via our studies of the systematic effects on the yield extraction.  Since this paper deals with cross sections and with nuclear-modification factors, both of which involve obtaining the yields of the Upsilon states, this change affects all the results in the paper. We therefore wanted to discuss this change first. Please note that the magnitude of these effects do not change the overall message of the paper.

We wanted to alert the referees up-front about this important change before we proceeded into the detailed responses.  This study was indirectly prompted by one of the questions from Referee 2 regarding systematic effects from yield extraction.

In the process of investigating the systematic difference between extracting the upsilon yield through simultaneous fitting compared to background subtraction as requested by the referee, we also studied the effects of chi^2 fits (specifically of Modified Least-Squares fits) compared to maximum-likelihood fits. We used chi^2 fits in our original submission. We were aware that extracting yields using a chi^2 fit introduces a bias (e.g see Glen Cowan's "Statistical Data Analysis", Sec. 7.4). The size of the bias is proportional to the value of the chi^2 of the fit.  In the case of the Modified Least-squares fit, when fitting a histogram including the total yield as a fit parameter, the yield will on average be lower than the true yield by an amount equal to chi^2.  The relative bias, i.e. the size of the bias divided by the extracted yield, goes to zero in the large yield limit, which is why for cases with large statistics this effect can be negligible.  We had attempted to mitigate the effects of this bias by using the integral of the data, since this removes the bias completely in the signal-only case.  But a bias remains in the case where there are both signal and background present. For our case, the yield extracted from the fit for the background is also biased toward lower values, and since we used this background estimate to subtract from the integral of the data in the extraction of the Upsilon yields, these biased the Upsilon yields towards higher values.   Through simulation studies, where we include signal and 3 background components as in our analysis, we were able to quantify these effects. Given that in some cases the biases could be of order 10-20%, the fits needed to be redone in order to remove the bias. The solution is straightforward since the extraction of yields using a maximum-likelihood fit is unbiased.  We have studied the difference of a modified-least squares fit and a maximum-likelihood fit and confirmed that the yield extraction in the latter method is essentially unbiased. We therefore have redone all the fits to extract the Upsilon yields via maximum-likelihood fits. The revised results are now quoted in the paper. The overall message of the paper is not affected by these changes.

We proceed next to answer the specific points raised by the reviewers.

Reviewer #1: This paper reports results on Y production in pp, dAu, and AuAu
collisions at top RHIC energy. It contains original and important
results and clearly qualifies for publication in PLB. However, there
are many aspects of the paper which need attention and/or improvement
prior to publication. They are detailed below:

1. the introduction is carelessly written. For example, the value
quoted for the pseudo-critical temperature near mu = 0 of 173 MeV is
taken from an old publication in 2003. Recent lattice results from the
Wuppertal-Budapest group (PoS LATTICE2013 (2013) 155) and the Hot QCD
Collaboration (Phys.Rev. D81 (2010) 054504) imply much lower T values
near 150 MeV and are far superior in terms of lattice sizes and spacing.

There are certainly newer results, which we now cite in the paper. However, we note that the
the results from the Hot QCD collaboration (Phys.Rev. D81 (2010) 054504) do not imply
much lower T values.  In that paper, in section IV "Deconfinement and Chiral aspects of the QCD transition", when discussing the deconfinement transition temperature range the authors write:
"...we have seen that the energy density shows a rapid rise in the temperature interval T = 170200. MeV. This is usually interpreted to be due to deconfinement, i.e., liberation of many new degrees of freedom".
Therefore, this does not indicate T values near 150 MeV. In addition, they also mention this range when discussing their results for the renormalized Polyakov loop, which
is the parameter most closely related to the deconfiment transition, being that it is the exact order parameter in the pure
gauge case:
"The renormalized Polyakov loop rises in the temperature interval T = 170200 MeV where we also see the rapid increase of the energy density."
Therefore, the results from the Hot QCD collaboration do not imply T values near 150 MeV.

In addition, in reference 9 of the Wuppertal-Budapest group (JHEP 1009 (2010) 073 arXv:1005.3508), which is a paper comparing the various results for Tc between the Wuppertal-Budapest and HotQCD groups, again the results for the renormalized Polyakov loop (figure 7, right) indicate a broad transition region in the region T=160-200 MeV.  They do have a table discussing values of Tc of about 147 MeV, but that is for the chiral transition, which is not the most relevant one for quarkonium suppression.
When they look at the trace anomaly (e-3p)/T^4, they see 154 MeV for the Tc value.  They in addtion make the point that the transition is a broad crossover, which is something we also say in our paper.  The fact that the transition is a broad crossover leads to differences in the estimates of the pseudo-critical temperatures depending on which observable is used.  As an example, in the caption of Table 2, where they give the values of Tc for many observables, they mention that the Bielefeld-Brookhaven-Columbia-Riken Collaboration obtained Tc=192. They also note "It is more informative to look at the complete T dependence of observables, than
just at the definition-dependent characteristic poins of them." So given the above, we will modify the paper to give a range of temperatures, 150-190, and cite the papers from the
Wuppertal-Budapest and HotQCD collaborations.

also the discussion on whether charmonium or bottomonium 'is a cleaner
probe..' does not get to any of the real issues, such as the complex
feeding pattern in the Y sector and the crucial question of whether Y
mesons reach equilibrium in the hot fireball as required to interpret
the apparent sequential melting pattern in terms of 'break-up'
temperatures.

The issues we discuss, in our opinion, are real issues.  We discuss co-mover absorption and the interplay between suppression and statistical recombination of uncorrelated charm pairs. These have been a topic of intense interest in the charmonium case for over a decade.  We certainly agree that these are not the only issues, but in this paper we aim to present the result of our measurement, so we cannot give a detailed review of all issues. However, the aim was to point out that for the bottomonium case, the expected contributions to either suppression or enhancement from both of these mechanisms are much smaller, and hence studying Upsilons is cleaner. The reviewer brings up the importaint issue of feed-down that affects the bottomonium as it does the charmonium sector.  We have added a few sentences regarding feed-down.  Regarding bottomonium, the feed-down contributions to the Upsilon states are not measured at RHIC energies yet. It is therefore assumed in the models used by Strickland, Rapp, etc. that the fraction of directly-produced Upsilon(1S) is ~51%, as measured in pp collisions at high pT at Tevatron energy. The original paper motivating the quarkonium sequential suppression by Digal, Petreczky, and Satz discussed feed-down as part of the impetus for looking for suppression of the Upsilon(1S). Given the ~51% direct Upsilon fraction, an R_AA of the Upsilon(1S) as low as ~0.51 would not necessarily imply suppression of the direct 1S, but could be due solely to suppression of the excited states. We have added text about this point in the paper, in discussing the R_AA(1S) result.

The reviewer also mentions that there is a crucial question as to
whether the Upsilon mesons reach equilibrium with the fireball as a requirement to interpret the sequential melting pattern. We respectfully disagree with the referee in this matter. The Upsilon is by definition not in equilibrium. The only requirement of course is that the medium is deconfined. In lattice QCD studies only the medium is thermalized; the potential between the heavy quarks is screened independent of whether the Upsilon is in equilibrium or not. We discussed this issue with lattice expert Peter Petreczky who confirmed our view.

furthermore, statistical recombination is not a 'complication' but a
direct measure of deconfinement. And the smallness of off-diagonal
terms in the recombination matrix does not imply absence of
recombination as the diagonal terms can be substantial.

We agree that statistical recombination is an indication of deconfinement, but from the experimental side, it has made the interpretation of the results more complicated, because one needs to take into account the interplay of suppression and recombination.  Because this effect is negligible for the bottomonium states even at LHC energies, the quantitative interpretation of the experimental results is less complicated. It is in this sense that the word "complication" is meant.

Also the newest results on p-Pb collisions from the LHC are entirely
ignored, see, e.g., arXiv:1308.6726.

We are aware of the quarkonium pPb results from LHC, however there is not a way to make a direct comparison to LHC results, because there are no pPb results on the nuclear modification factor of Upsilons.  The results from ALICE in the reference given above are for the J/psi meson, and they are also for the forward-backward kinematic range.  There are also results from CMS (arXiv:1312.6300) for Upsilon mesons at midrapidity, but these are in the form of ratios of the yield of the excited states to the ground states in a given system (pp, pPb, PbPb), and of double ratios, i.e. excited-to-ground-state ratios in pPb divided by excited-to-ground-state ratios in pp.  These give us relative suppression of the excited states, whereas our results are for absolute suppression, and are therefore not directly comparable. The only quantitative comparison to the CMS data we can make is to estimate a double ratio for the excited states. The double ratio we find is consistent with the result from pPb from CMS, but it is also consistent with 1, i.e. no suppression of the excited states relative to the ground state in pPb compared to pp (We find the double ratio = 0.72 +/- 0.37). We will make a comment about this in the paper, and cite the CMS pPb result.  But the advantage of the results we are presenting is that we have fully corrected nuclear modification factors, which convey more information than relative suppression.

2. section on experimental methods

no detail is given concerning the crucial momentum resolution but it
is stated at the end of this section that cuts were adjusted for
different systems such that 'tracking and electron id would be the
same across the 3 data sets'. On the other hand, already in Fig. 1 we
see a strong dependence of the mass resolution on the system even for
low multiplicities as in pp and p-Pb. The effect must be much stronger
in Pb-Pb as is indeed visible in Fig. 4. Especially in view of the
importance of resolution for the separation of excited Y states this
referee has to be convinced that the systematic errors are under
control for momentum and pid measurements as a function of
multiplicity. Also how the systematic errors for the separation of Y'
and Y'' from Y are determined as a function of multiplicity needs to
be demonstrated explicitely.

We agree that the mass resolution is very important for the results of the paper, and need to be discussed.  We added text to discuss the Upsilon mass resolution, and how it was studied as a function of TPC multiplicity (we focus on mass resolution, but this is directly related to the momentum resolution of the electron tracks used to reconstruct the Upsilon).  We studied the mass resolution using both simulations and data-driven methods.  Regarding the momentum resolution and the difference of the mass width seen in the pp compared to the dAu and AuAu plots, the majority of the difference between the pp lineshape and the dAu/AuAu lineshapes comes from a miscalibration in the TPC which was corrected in the pp dataset via a reproduction of the raw data, but due to time constraints was not corrected in the dAu and AuAu datasets. With the distortion correction, the pp mass resolution is found to be 1.3%.  If there were no distortions in Au+Au, we find in peripheral events a mass resolution for the Upsilon(1S) of 1.7%, which widens to 2.0% for central events, based on simulations. In order to ensure that we had this resolution effect under quantitative control, in addition to studying it via embedding simulated tracks into real collision events, we also studied the difference in the reconstructed mass of every reconstructed di-electron pair between the corrected and uncorrected pp datasets on an event-by-event basis. We were able to determine the additional mass smearing introduced by this TPC distortion, and this data-driven knowledge was used in the determination of the line shapes in dAu and in Au+Au.  The additional smearing introduced by the TPC distortion resulted in a mass resolution for the Upsilon(1S) of 2.7%, widening to 2.9% for central events.  For d+Au, the mass resolution is also found to be 2.7%, consistent with the peripheral events in AuAu. Finally, given the importance of the resolution for the separation of the states, we also used one additional data-driven method to check the resolution. We used the data from Au+Au and performed a chi^2 scan varying the mass-width parameter of the Upsilon line shape to see if this would give the same results as those found from the pp event-by-event mass-smearing data-driven study.  The results were consistent with each other, giving us confidence that the mass resolution is under control.  We used the shape of the chi^2 vs. resolution-parameter derived from the data to assign an uncertainty to our mass-resolution parameter knowledge, and used this to estimate a systematic uncertainty on the yields.
We have added a few sentences to the text towards the end of the Experimental  Methods section to give the relevant mass resolution numbers, and to make it clear that the mass resolution is different for pp compared to both dAu and AuAu due to the TPC distortion.  We also added some sentences to discuss this in the description of the invariant mass figures.
The systematic uncertainties due to our knowledge of the line-shapes, which are directly related to the separation between the Upsilon(1S) and the excited states, are also listed in the systematic uncertainty table that we added to the paper. The rows listing the uncertainty due to the line shape, including these mass resolution effects and the uncertainty in the knowledge of these TPC distortions, are given in the table.  Finally, we also added a systematic uncertainty to the extraction of the Upsilon(1S) yield based on the purity of our mass cut.  This is affected not only by the knowledge of the line shapes, as discussed above, but also by the possible suppression of the excited states.  We estimated this uncertainty by comparing the case of no suppression to that of complete suppression of the excited states, and recalculated the Upsilon(1S) purity for each.  The systematic uncertainty table has a row giveing the value of this uncertainty on the Upsilon(1S) yield, and we also added text to clarify it in the paper.

3.  Fig. 2b

even at y = 0 the difference between data and models is less than 2
sigma, taking uncertainties due to the pp reference into account and I
and initial state parton energy loss in these data.

With the new fit results, only the point at y=0 is different from the models, and while the difference is now of order ~3sigma, the two other points do not show any deviation from the models. The text has been changed removing the sentence mentioning effects beyond shadowing.  Also, thanks to this comment, we realized that we had plotted the full pp cross section systematic uncertainty on the figure, but for the purposes of R_dAu, some of these systematic uncertainties cancel.  Therefore, the band illustrating the systematics due to the pp reference should have been smaller, and this has now been corrected.

4. in Fig. 3 the size of the systematic errors should be indicated.

The problem is that none of the data points from E772 had systematic uncertainties, so we cannot include them.  We have indicated the size of our systematics in the plots.

5.  in Fig. 5 it is demonstrated that the observed suppression near
midrapidity is independent of system size (N_part). This could imply
that the higher Y states are already disappeared in dAu
collisions. This is mentioned briefly in the conclusion, but could be
stressed more.

The statement we made in the paper is complementary to the one suggested by the referee. We state at the end of the paper that we cannot claim that the suppression in AuAu is unambiguously from color deconfinement in AuAu given the suppression in dAu.  The reason we stated it like this is that the expectation was for only a minimal amount of suppression in dAu, but our results are a call for caution and for considering other hypotheses.  The referee's comment is a call to consider a specific hypothesis: that the higher Upsilon states already dissappear in dAu collisions. This is a hypothesis that is not accounted for yet in any model. While we are not advocating any particular hypothesis for dAu suppression, we can add a sentence phrased as suggested here, just to stress that our data elicit new thinking about Upsilon suppression in dAu.

6.  At LHC energy, the anisotropic model of Strickland reproduces well

the centrality dependence of R_AA but not the rapidity dependence,
see, e.g. the final session of the recent hard probes meeting in South
Africa.

The predictions from Strickland shown in Fig. 4 are rapidity-dependent. Our data agree fairly well in both rapidity ranges. The rapidity range examined by the slides shown in Strickland's Hard Probes talk (|y|<4.0) is much wider than the ranges we examine (|y|<1.0). Looking at the models, we do not expect to see much variation at all in the range we examine which is consistent with our observation. To see this variation, we would need to examine a much wider rapidity range which is not the focus of this paper.  (To constrain models at larger rapidities, it will be interesting to see the PHENIX results near |y|~2, which should be submitted for publication soon.)

7.  The presentation in Fig. 6 on the quantitative evaluation of
different model assumptions compared to data depends again strongly on
the size of the systematic errors, see the comment in section 2.

We've added a table and a new plot summarizing efficiencies and systematics.

----------------------------------------------------------------------------------------------------------------------------------------------------------------------

Reviewer #2: I have read the manuscript PLB-D-13-01645 submitted to me for review.
The authors present a detail analysis on the suppression of Y production in d+Au and Au+Au collisions at sqrt(s_NN)=200 GeV using the STAR detector at RHIC. The article is very well written and deserves publication. However, I would like to suggest considering the following remarks to improve the understandability of the article:

1. Page 1, column 1, paragraph 1: The now accepted value for the critical temperature (chiral transition) is Tc = 150 - 160 MeV (depending on the exact definition of the observables). Reference 3 is outdated and should be replaced by more recent publications, i.e. arXiv:1005.3508 [hep-lat]

See response to first reviewer's comment #1. References have been updated. Furthermore, the Tc noted here (the chiral transition) is not the relevant phase transition for quarkonium suppression. The more relevant one is the deconfinement transition (which is somewhat broad as currently noted in the text).

2. Page 2, column 2, paragraph 1: Please quantify the corrections due to the trigger bias w.r.t. the event centrality. Same for the tracking efficiency as a function of N_part. How does acceptance times efficiency for detecting Y as a function of rapidity and N_part looks like?

Added a new figure summarizing the total efficiency as a function of N_part and rapidity.

3. Page 3, column 1, paragraph 1: statement "some information will be lost" is too general! What are the systematic uncertainties arising from the different methods (same-event like-sign CB, fit to the CB) of the combinatorial background subtraction? What is the signal significance, in particular in the d+Au measurement? How does the signal looks like after CB and physical background subtraction? Systematic errors should be clearly mentioned.

Many thanks to the referee for mentioning this. In the process of investigating the systematic difference between extracting the upsilon yield through simultaneous fitting compared to background subtraction, we also investigated the effects of chi^2 fits compared to likelihood fits. We used chi^2 fits in our original submission. We were aware that extracting yields using a chi^2 fit introduces a bias. We attempted to minimize this bias by extracting the yields using the integral of the data. However, we still needed to subtract the backgrounds and the bakground yields came from the fits. Through investigation, we have demonstrated that chi^2 fits systematically underestimate the background yield, leading to an overestimate of the upsilon signals.

We have studied these effects through various MC simulations in order to extract the biases. The likelihood fits have negligible biases. Furthermore, and to get back to the original question posed by the referee, in these simulations we obtained the variance of our results when doing simultaneous fits when compared to background-subtracted fits. We found a reduction in the variance when using simultaneous fits which was our original impetus. We also found no systematic effect in the expectation values of the yields obtained by the two different fitting methods. However, given the reduction in the variance of the extracted yield (i.e. in their error) in the simultaneous fit, we favor this method since it introduces a smaller uncertainty. We have redone all of our fits using the likelihood method and we corrected for any extraction biases seen through simulation.

Regarding signal significance, in all cases we see significant signals in d+Au. This can be infered by examining Fig. 3a and comparing the size of the statistical+fit error bars to the measured value of the cross section. This ratio is a good indicator of the statistical significance of our signal. For example, the dAu signal at |y|<0.5 has a significance of 11.7/3.2 = 3.7 sigma.

4. Fig 1: It would be easier for reader if the range of the y axis would be the same in Fig 1a and Fig 1b. Why is the explanation of the grey curves in the figure discussed in this complicated way, to my understanding the gray band simply shows the pp yield scaled by the number of binary collisions? If so, the label could read simply pp*<N_coll>.

The axes in Figs. 1a and 1b now match. We've relabeled the gray band.

5. Fig 1a: From where the line shape for pp comes from? It seems NOT to fit experimental data, i.e. all data points around 9 GeV/c^2 and below. Is it then evident to take as a cross section the integral of the data points?

The line shape in pp comes from simulations embedded in real data. Below 9 GeV, the lineshape threads between high and low datapoints. It cannot fit exactly to all of them without introducing wiggles in the function.

6. Page 3, column 2, paragraph 1: How was the measured Y(1S+2S+3S) yield transformed to cross section?

The cross section was calculated by correcting for EID efficiencies, triggering efficiencies, and acceptance to get a corrected yield. We then divided by the integrated luminosity to get a cross section.

7. Page 3, column 2, paragraph 3 (wording):  "Hence, averaging between forward and reverse rapidities is not warranted as it is in

p+p." -->  "Hence, averaging between forward and backward rapidities is not justified as it is in p+p." sounds more understandable.

Since the two words are very closely related (Merriam-Webster includes "justification" as one of the definitions of "warrant"), this is more a matter of style. The authors prefer the word "warranted."

8. Page 4: Try to arrange the placement of Figs such that there will not be a single line of the text within one column.

Done

9. Fig 2: also here Fig a and Fig b could be presented with the same range on the Y axis, e.g. from -3 to 3.

Done.

10. Fig 2a : what is shown here is Y(1S+2S+3S), moreover PHENIX results on Y -> mu+mu- are shown in the same plot, that is why the figure label should be changed, i.e. Y->e+e- should be replaced by Y(1S+2S+3S)

We changed the label to Y->l+l- to represent leptons.

11. Page 4, column 2, paragraph 2: <N_coll> (not <N_bin>) is commonly used as notation for the number of binary collisions. Sigma_AA is sigma^tot_AA (same for pp). It is important to indicate in the text the values for the total inelastic cross sections in pp, dA and AA and <N_coll> used to calculate R_AA.

Now using <N_coll>. Inelastic cross sections are provided inline.

12. Page 4, column 2, paragraph 3: In view of the discussion would it be helpful to also show R_AuAu vs. Rapidity?

We have addede new plots to the paper, and given the length considerations, and that this plot doesn't really add any new information beyond the existing tables and figures, for the sake of space, we would prefer to leave this plot out.

13. Page 6, column 1, paragraph 1: Which function has been used to fit the CB - exponential? Again, what are the systematic uncertainty arising from the different methods (same-event like-sign CB, fit to the CB) of the combinatorial background subtraction. See also comment 4. concerning the label.

The function used to model the CB is now discussed in the text. Systematics from the fit methods are summarized in Tab. I.

14. Page 6, column 1, paragraph 2: The statement "Similar suppression is found by CMS in PbPb collisions (37)" should be moved to the paragraph 4 where the authors discuss Y(1S) suppression. Actually, for the same value of N_part=325 R_AuAu=0.54+-0.09 as for R_PbPb=0.45

Done.

15. Page 6, column 1, paragraph 4: How did the authors derived: R_AA(1S+2S+3S) = R_AA(1S)*0.69?

We calculate this number by relating the two nuclear modification factors. For the (1S+2S+3S) case, this needs the ratio of the yields of (1S+2S+3S) in AA to the same yield in pp.  Since the R_AA(1S) is the ratio of the yield of the 1S state in AA to that in pp, one can take this out as a common factor in the R_AA(1S+2S+3S), obtaining the relation R_AA(1S+2S+3S)=R_AA(1S) * (1+ N_AA(2S+3S)/N_AA(1S))/(1+N_pp(2S+3S)/N_pp(1S)), where N_AA refers to the yield obtained in AA collisions and N_pp refers to the yield obtained in pp collisions.  This equation makes no assumptions.  When one takes the additional hypothesis that the yield in AA of the excited states is zero, the factor becomes 1/(1+N_pp(2S+3S)/N_pp(1S)). So with the ratio of excited states to ground state in pp collisions, one can find the multiplicative factor.  We calculated this ratio in two manners: first, by using the PDG branching ratios together with NLO pQCD calculations for the upsilon production cross sections (from Ref 21 by Frawley, Ullrich, and Vogt), and, second, by using the measured 2S/1S and 3S/1S ratios. For example, these ratios have been measured at sqrt{s}=38.8 GeV and also at sqrt(s)=2.76 TeV from CMS, and are relatively independent of sqrt(s), or even whether the collision system is pp or pA. In the first case where we used the pQCD cross section and PDF branching ratios, we get 0.69 for the multiplicative factor. In the second case where we only used measured ratios, we get 0.72 +/- 0.02.  The difference between using the low-sqrt(s)-pA data or the CMS pp data at 2.76 TeV is 0.01, which is smaller than the statistical error of the CMS data.
We had aimed to keep the text brief, since we were mindful of the space constraints, but given this question, we have added a few more sentences and references to clarify the R_AA(1S+2S+3S)=R_AA(1S)*0.7 statement, and also reduced our significant figures, quoting only a factor of 0.7.

16. Page 6, column 2, paragraph 2: What are the uncertainties on Drell-Yan and bbbar cross sections and how does it influence the significance of the signal.

Various normalizations are used in the fit. This is accounted for in the correllation

17. http://arxiv.org/pdf/1109.3891.pdf reports on the first measurement of the Y nuclear modification factor with STAR. It is probably worth to mention this work in the ms.

We are certainly aware of the proceeding mentioned here, which showed preliminary results for these analyses. The author of the proceedings was a member of the institute where the primary analsys shown in this paper was done. The reason we omit the citation to this and to other proceedings where the preliminary results have been shown is that it is a policy of the STAR collaboration to not cite our own proceedings showing preliminary data. This is partly with the goal to make it clear that the final results presented in a given paper, which have gone through the full collaboration review and through the external peer-review process, are the ones that should be referenced once they are available.

18. The R_AA of J/psi (p_T > 5 GeV), Y(1S) and an upper limit on the R_AA (2S+3S) was obtained in STAR. I would like to suggest to add a plot showing R_AA as a function of binding energy as a summary figure (also as a key figure to the long discussion on the extraction of the upper limit on R_AA(2S+3S)).

Good idea. We added this figure towards the end of the paper.

In summary, this ms. contains very interesting results and I propose publication in Phys. Letter B after the authors have taken care of the remarks above.

We thank the referee for her/his comments and remarks, which have helped improve the paper.  We hope that we have addressed the issues raised, and adequately answered the questions posed, and look forward to the publication of the paper.

Upsilon pp, dAu, AuAu GPC E-mail Responses

This is a page to house long e-mail responses.