Dear authors, Congratulations with this interesting analysis and well-written paper. Please find below the comments from the Nikhef group. We think that the paper has a relatively high level of detail but do actually appreciate it. Cheers, Wouter ------------------------------------------ Concerning the analysis we have the following questions/issues. 1. Since the phi(1020) is the dominant contribution to the KK final state, it seems very important that the phi lineshape in the fit is correctly described: small problems in the phi lineshape may have a big impact on the extracted contributions of the other resonances. Also the relative contribution of phi and S-wave (fig 18,19) must be very sensitive to the exact description of the phi. Unfortunately, there are no plots shown in the paper from which one can conclude that the lineshape is fine. The only plot that is there, fit 22 top-left, looks very poor in the phi region. The analysis note reports a chi value of 73/20. The conclusion in the text that the fit described the data well (line 554) does not seem justified. Also, the overall chi2 of the fit (table 3) is very poor, with a 'probability' smaller than 10^-7. Such problems cannot be ignored and must either be solved or explained in the paper. 2. Concerning the phi line shape we are particularly worried about the method that is used to account for the finite KK mass resolution. For the phi(1050) one cannot really ignore the resolution since it is a good fraction of the natural width. In the paper there is nothing written about this. In the appendix of the analysis note you write that you artifically increase the natural width of the phi to account for the resolution. However, while the resolution mostly affects the 'peak', changing the natural width affects the shape of the phi also in its tails. I have attached a plot thay illustrates this: in green is the relativistic breit wigner (with a phase space factor taking the lower KK threshold into account, but not the upper threshold). In blue is the same pdf convolved with a Gaussian (using RooFFTConv). In red is the relativistic breit wigner with the width decribed in the analysis note. You see that the effects on both peak and tail are different: the left side of the peak will ook different; the contribution to the tail from the 'fudged' breit wigner is too large. We understand that it is not trivial to take the resolution into account: It is a strong function of the KK invariant mass for instance and this dependence needs to be modeled. Furthermore, you'll need to rely on numeric integration which makes the fit notably slower. However, we think that given the level of detail with which you report for instance the phi/S-wave fraction, the resolution is such an important issue that it needs to be dealt with properly. 3. We are a bit worried about your use of sidebands for background model, in particular because of the B mass constraint. For signal events the mass constraint will improve the mKK and mJpsiK resolution. However, for background it will make it worse and the effect if bigger the further away you go from the B signal region. This potentially affects a real phi contribution in the background. Such a contribution was observed in the 0.3/fb time-dependent analysis note, see ANA 2011-036, figure 6. Your projection of the mKK sideband does not show evidence of a sharp phi peak in the combinatorial background. It is unclear to us if this is because of binning, because there none (due to the tight selection), or because it is smeared out as a result of the mass constraint. If it is the latter, then the background parameterization extracted from the sideband is not representative for the signal region. 4. You explain that you use toy simulations to compute the uncertainties on the fit parameters because it is hard to propagate uncertainties. We are really puzzled by this: if you can compute "f(x)", then propagating uncertainties is as easy as computing f(x+sigma(x)). If you need correlations, you can use the same method to compute the Jacobian and then transport the covariance matrix. Why did you use toys? The fact that the chi2 of your fit comes out so poorly makes us also worried about the statistical uncertaintiesin general. Do the toy simulations also have such a poor chi2/dof? 5. You report the ratio of the lambda=0 and lamba=1 components of the phi(1020) and claim that it is compatible with the time-dependent analyis. However, we see a few problems with this comparison. First, at face value the reported numbers are incompatible: the uncertainty for the time-dependent analysis is dominated by the systematic error on the angular acceptance. Whether that error be right or not, there is no reason to assume that it is any different for this analysis, so they are fully correlated. The statistical errors should also be correlated. So, the numbers are not close at all. Second, as you know, we are comparing apples to pears here: the time-dependent analysis reports amplitudes at t=0, while these are integrated amplitudes. For dG!=0 these are not supposed to be identical. Finally, the number that you report is not corrected for the finite decay time acceptance. We do not know what the effect of that is, since you don't discuss it anywhere, but it may well account for part of the difference. In short, in our view this ratio should not be reported in this paper: without the acceptance correction it is not useful to anybody. The comparison the our other result is off. Instead, it suffices to say something like "The value of the ratio of the two phi(1020) contributions is consistent with that extracted from the time-dependent analysis." We also think that the effect on the branching fractions of the time acceptance in case dG!=0 needs to be discussed somewhere, for example when discussing systematic uncertainties. 6. In your discussion of the systematic uncertainties you discuss one test where you increase the IP chi2 of the kaons. What systematic effect if this supposed to probe? You explain that it changes the chi2/dof of the fit marginally, but isn't it more relevant how it affects the fitted parameters? (You could have a perfect fit with an incompatible result.) We have also a small number of editorial comments: - line 310: it looks odd to define the helicity angle as an angle between vectors in two different lorentz frames. Instead, define it as "the angle between the muplus and minus the direction of the B in the J/psi rest frame." (You can take mu-minus if you don't like the 'minus' for the B direction.) - line 314: Only in case you have actually used DecayTreeFitter: I (WH) agree that we reference the DTF paper far too often in LHCb, but since these Dalitz analyses actually really profit from it, I propose to add a reference to DOI: 10.1016/j.nima.2005.06.078. - equation (15): where does this expression come from? Please explain or add a reference. - line 373: the reported chi2 (198/303) is extremely unlikely. Is this a typo, or is something wrong with estimated errors? If the latter, fix it, or don't report it. - figure 8: why is the scale arbitrary? cannot you define it properly? - line 417: "The situation ... is .. confused" Replace 'confused' with 'confusing' (or 'uncertain' or something equivalent). - line 441: The parameterization of the acceptance doesn't look trivial. Does this really allow you to do a 'analytical' integration over the helicity angle? - figure 15b, and fig 17 are identical appart from the curves. Please choose one. - figure 17: there is either a real problem with the fit, or the projection of the blue curve (total pdf) is wrong: it is absent for the smallest bin. - figure 18 and 19: From the text it is unclear what it in these figures. At first we had not realized that this is not the result of a binned fit to the data, but rather just an integral of your fitted PDF. We propose to correct the text. Also, it should be explained that the error bars in the figrue are highly correlated. Finally, we see little reason not to make the bin size in figures much smaller, or better still, just draw continuous bands. Is there a good motivation to show both figures? - line 553-555: as we commented above, this conclusion does not seem justified by the plots. See e.g. figure 22, top left and fit 23, bottom left. - line 599-602: from the description we do not understand what you have done, so it is hard to imagine that somebody from outside the collaboration could. Please, rephrase. - line 619: remove 'long' in "long tracks"