From fdiakono@phys.uoa.gr Wed Aug 29 21:49:22 2012 Date: Wed, 29 Aug 2012 22:49:13 +0300 From: Fotios Diakonos To: Peter Seyboth , Nikos Antoniou Subject: [Fwd: Your_manuscript LV13012 Anticic] ----------------------- Αρχικό Μήνυμα ------------------------ Θέμα: Your_manuscript LV13012 Anticic Από: prl@aps.org Ημερομηνία: Τετ, Αύγουστος 29, 2012 22:08 Προς: fdiakono@phys.uoa.gr -------------------------------------------------------------------------- Re: LV13012 Critical fluctuations of the proton density in A+A collisions at 158A GeV by T. Anticic, B. Baatar, D. Barna, et al. Dear Dr. Diakonos, The above manuscript has been reviewed by our referees. The resulting reports include a critique which is sufficiently adverse that we cannot accept your paper on the basis of material now at hand. We append pertinent comments. If you feel that you can overcome or refute the criticism, you may resubmit to Physical Review Letters. With any resubmittal, please include a summary of changes made and a brief response to all recommendations and criticisms. Yours sincerely, Abhishek Agarwal Assistant Editor Physical Review Letters The premier APS journal for current research Email: prl@ridge.aps.org Fax: 631-591-4141 http://prl.aps.org/ ---------------------------------------------------------------------- Report of Referee A -- LV13012/Anticic ---------------------------------------------------------------------- Let me start off by saying that the determination of a critical point in the QCD phase diagram is a rather controversial issue, theoretically and experimentally, as I am sure the authors are aware. Overall, I believe that although the paper addresses a novel measurement and potentially a dramatic and positive result (discovery of a critical end point), the paper in its present form is not suited for PRL. First and foremost, I have problems with the experimental evidence in light of other measurements in the same energy regime, that do not seem to corroborate the presented result. I also believe that the chosen method, namely intermittency, requires a much more extensive discussion of the potentially irreducible trivial effects that could cause a positive result. Furthermore the authors need to confront the fact that not all theoretical papers suggest the existence of a critical point in the QCD phase diagram. In the following I will go through my objections in detail and offer suggestions for a revised manuscript. Generally the paper and the result deserve an airing in PRL, but I am worried that a proper paper which addresses all my objections will be too detailed for PRL submission. Thus the authors should seriously consider a longer but more valuable article in another journal. Detailed objections/suggestions: - In terms of viable theoretical calculations my concerns are two-fold: a.) I would suggest that the authors modify their introductory sentence and take into account much more recent calculations, in particular by the Fodor group based on lattice QCD, which pretty clearly indicated that there is little to no evidence on the lattice for a critical point (see for example JHEP1104, 001 (2011) by the some of the same authors than Refs. 1 and 2). b.) Most of the theoretical predictions of the critical exponent in the power law behavior are drawn from the authors own theory paper (Ref.6). It would be good to have independent confirmation of the calculated critical limit (phi**2=0.83) by another theory group, for example the Stephanov group. It is certainly unusual for a theory group to team up with an experimental collaboration to confirm their own predictions. - In terms of experimental evidence I have five main objections/suggestions to the paper is: a.) The paper ignores measurements by the same collaboration on K/pi and pt-fluctuations, which seem to indicate enhanced fluctuations at lower beam energies rather than the highest SPS energies. The paper also ignores measurements from the RHIC beam energy scan which shows no enhanced fluctuations at any SPS energies or above. b.) More importantly the paper does not address the STAR measurements of the higher factorial moments (skewness, kurtosis) of the net-proton density at all energies, including the SPS energies used here. which again show no evidence of criticality. If intermittency is indeed a more sensitive measure to the baryon density fluctuations than the higher moments, then the paper needs to explain why this would be the case. c.) The paper does not specify any trivial background which could lead to correlations at high M**2, and it does not sufficiently address the effect of very fine momentum binning (what is the smallest momentum bin chosen for the largest M**2 ?). Intermittency measurements in relativistic heavy ion collisions, away from the critical point, had historically quite a bit of systematics due to 'trivial effects' such as binning and acceptance effects. Therefore the effect of fine binning needs to be better addressed by showing the evolution of the systematic error as a function of M**2. The systematic error in Fig.2 is solely based on comparing different fitting procedures rather than adding a systematic uncertainty from the measurement itself, except for a dependence on the proton purity. d.) The paper does not indicate the relevance of the system size dependence (is it really conceivable that the SiSi system is closer to criticality than the PbPb system?). I presume the error bars in SiSi are only larger due to statistics (around 200 K in SiSi and 1.5 Million events in PbPb. 200 K is a rather small data sample. Is there a chance for more statistics in the smaller system? e.) A key assumption in the paper is that the net-proton density is perfectly flat in the rapidity range chosen for the analysis (around mid-rapidity). Since the bin is rather wide, the authors should consider showing the actual measurement, in particular for the smaller systems in order to corroborate their assumption. ---------------------------------------------------------------------- Report of Referee B -- LV13012/Anticic ---------------------------------------------------------------------- This paper presents an interesting new result from the NA49 collaboration on fluctuations of the density in momentum space of emitted protons in A+A collisions at the SPS. The result is potentially extremely interesting since it would constitute possible observation of one of the proposed signatures that the system produced in A+A collisions has passed in the vicinity of the critical point in the QCD phase diagram. A convincing demonstration that this is the most likely explanation of the observation would certainly be worthy of publication in PRL. However, spectacular claims require a very thorough justification. I do not find the results, as presented, sufficiently convincing to warrant the conclusion presented. The paper must demonstrate not just that there are non-Poissonian fluctuations, which it does, but it must also demonstrate that the fluctuations cannot be explained by any other mundane explanation more plausible than the proposed critical fluctuations due to the QCD critical point. The paper does not sufficiently address this point. The authors further claim, not only that the behavior is that expected for a system near the QCD critical point, but that the value of the power is in quantitative agreement with the QCD expectation for the Si+Si system, as if this observation is in itself evidence that this interpretation must be correct. However, the authors do not convincingly demonstrate that all relevant corrections have been applied, necessary to claim that the quantitative result can be fully attributed to the critical fluctuations. The authors have used the method of investigation of the variation of the scaled factorial moments of the proton multiplicity with bin size, using bins in the two-dimensional transverse momentum space. This is a form of the so-called "intermittency analysis" method. The virtue of the scaled factorial moments is that they are extremely sensitive to non-Possionian fluctuations, since the moments will not depend on the bin size if the underlying distribution (particle production mechanism) is Poissonian. In that case, the power of the scaled factorial moment versus the number of bins is identically zero. Any deviation of the power from zero is an indication that the underlying distribution is non-Poissonian. A value greater than zero indicates an enhanced probability (as for a negative-binomial distribution) while a value less than zero is indicative of a suppressed probability (as for a binomial distribution). If the nature of the probability distribution (production mechanism) acts on the smallest scale (bin size) then the variation should be described well with a constant power over all bin sizes. Rather than simply analyze the scaled factorial moments as extracted from the event data, the authors go further and extract the power from \Delta F, the difference between the measured scaled factorial moments, and the same result obtained for mixed events. The mixed event procedure is not described in detail, but if the mixed events are created by a Poissonian selection process, as they should, then one would expect that the scaled factorial moments for the mixed events must not depend on the bin size. Therefore, there should be no reason to subtract such properly constructed mixed event results. On the other hand, one can construct mixed events with unintended correlations, for example, by imposing a binomial selection which would give rise to a decrease with decreasing bin size. If the mixed events have exactly the same multiplicity distribution as the real events, then they must have the same moments. it is notable and highly suspicious, that in the Si+Si case, where the most spectacular claim is made, the mixed event results do depend on the bin size, decreasing with decreasing bin size, which then will give a larger power law dependence to \Detla F for this case, apparently bringing it up to a value "consistent" with the QCD expectation. This raises the question of why there is a bin size dependence of the mixed events at all, and why apparently only for this case? If the mixed events are not truly mixed events, but do include some other correlations, such as two-track merging effects, this needs to be clarified in the text, together with an explanation of why the Si+Si case seems to decrease and is so different from the other cases. Furthermore, contrary to statements made in the paper, the theoretical expectation (as shown in Ref. 6) is that the power law dependence should extend over the full range of bin sizes, with the same power. This doesn't appear to be the case for the Si+Si results after the mixed event subtraction. The low M^2 regions of the data have been suppressed in the figures after Fig.1, and the result isn't presented as a log-log plot as in the last two figures, which prevents easy judgment on this point. From the text, it's suggested that the fit results and uncertainties have only been extracted from fits to the limited region shown in Fig. 2. It should be clarified why this is justified. The absolute value of the scaled factorial moments depend on the overall multiplicity. Subtracting the results for mixed events with a power of zero just subtracts a constant value from F, and so just changes the M=1 intercept. Mathematically it's expected that the F should be proportional to M^q where M is the number of bins and q is the power (see Ref. 6), so one could write F= F1 * M^q. The intercept for a single bin M=1 is then F1. Presuming, that the mixed events are properly normalized with power of zero, one expects F_m = F1 * M^0 = F1. So subtracting the mixed events gives \Delta F = F1 *M^q - F1. The paper states that the power is extracted from the fit of \Delta F to the form \Delta F = DF0 * M^q', which obviously does not give the desired exponent q. If the mixed events don't have a power of zero, as appears to be the case for the Si+Si result, then the effective power extracted from the subtracted result will be even more distorted. Also, a relatively few high multiplicity "abnormal" events in the tail of the distribution can strongly effect the scaled moments. Therefore the multiplicity distributions, including those for mixed events should be presented. There are many effects which can give rise to non-Poissonian fluctuations which are not discussed in this letter. Physics effects include radial or elliptic flow, proton-proton final state interactions, or decays, while detector effects might include track splitting or particle ID effects. Normally, one uses an event generator like AAMPT with a GEANT implementation of the detector and analyzes the simulation results in the same manner as the detector to confirm whether or not there might be uninteresting effects to explain the observations. This should be done and shown for this analysis. Effects like flow and p-p final state interactions are not included in most event generators, but can be investigated by imposing the expected correlations on the generated MC events, or on mixed events. This should also be done and compared to the observed results. The systematic error discussions is inadequate. Only two systematic errors are discussed: 1) the fit systematic error for the Si+Si case, in which case the quoted systematic error is based on the deviation of three results which each have a fit error twice as large as the quoted systematic error, and 2) the purity of the Pb+Pb case, where the quoted systematic error was extracted from the deviation of three results each with a fit error equal to the quoted error. Quoting systematic errors based on a results that themselves have errors larger than the quoted error seems dubious. Also, shouldn't the fit systematic error be similar for the Pb+Pb case? Also, there must be other systematic errors from detector effects and other physics correlations mentioned above that are not discussed at all. Any of these effects which give rise to a dependence of the factorial moments on bin size certainly must be give rise to a systematic error on the power to be attributed to critical fluctuations. Other comments and questions: * The analysis is based on the "scaled factorial moment F_2" given in Eq. 1, which is not the same as the "factorial moment". The text should refer to F_2 properly throughout. * Why don't figures 2, 4, and 5 show the data over the full range of measured M^2 shown in Fig.1? And since the interest is to confirm a power-law dependence all of these figures, including Fig.1 should be shown as log-log plots. Also, Figs. 1 and 2 could be shown in much expanded forms in order to view the possible M^2 dependence of the data more clearly. * Were all of the measurements obtained under the same experimental conditions? For example, was the lower energy result obtained with the same B field conditions as the higher energy results? * Were any of the measurements repeated with different experimental conditions (i.e. different B-field) but with the same kinematic analysis, i.e. same transverse momentum range and bins, to confirm that the results did not depend on the experimental conditions? * Since the authors seem to suggest that the central Si+Si system size is optimal for coming into the vicinity of the critical point, has the analysis been done for intermediate Pb+Pb centralities with a similar number of participant nucleons as central Si+Si? * Similarly, critical fluctuations presumably should vary slowly above the "critical" system size, while other effects like flow and occupancy will change more dramatically. Therefore it would be important to confirm how much variation with centrality is seen for the Pb+Pb case. Is this information available? In light of the seriousness of the above concerns I don't add further comment on details of the text. In summary, although the measured results as presented may be correct, the authors do not prevent a convincing demonstration that their interpretation of the results as due to critical fluctuations at the QCD critical point is the only interpretation, or the most likely interpretation. Therefore further work will be necessary for the paper to warrant publication in PRL.