Hi Urs,
thanks for taking care of this.
the first comment is that the paper is now 30 column-lines too long, I
have some suggestions to reduce text in
http://babar.roma1.infn.it/~faccini/tmp/prl.pdf
take them as suggestions ...
I have also a few comments on the replies that are marked by "RF>"
>
----------------------------------------------------------------------------
> Report of Referee A -- LG9660/Aubert
>
----------------------------------------------------------------------------
>
> This paper reports an interesting measurement of V_ub using inclusive
> charmless B decays. The very high statistics data sample collected by
> BaBar allows a much higher purity sample to be obtained, limiting the
> systematic errors from background that plague the earlier
> measurements. It is therefore very welcome, and deserves publication
> in PRL. The paper is generally well written, but some parts are very
> dense and hence unclear, lacking in necessary detail. I have a number
> of questions and comments, together with some suggestions for minor
> improvements in the text. If these can be addressed satisfactorily, I
> will be happy to recommend the paper for publication as soon as
> possible.
>
> 1) Neither the introduction or conclusion makes any reference to the
> other main method of measuring V_ub, namely the measurement of
> exclusive branching ratios B->pi l nu, B->rho l nu etc. Since the
> two methods are complementary and the systematics rather different,
> I think at least some mention of this should be made in the
> introduction, and the e.g. the precision achieved (including both
> experimental and theoretical uncertainties) compared with that from
> exclusive techniques in the conclusion. A reference to the most
> recent publications on exclusive branching ratios (e.g. the Babar
> paper PLB 90, 181801 (2003) and others) would be useful.
We have added a comparison to exclusive measurements of |Vub| in the
conclusions.
RF> The citation takes 6 lines ... isn't it possible to find a review to
cite?
> 2) The inclusive method used here requires the assumption of
> parton-hadron duality, which is not universally accepted in this
> context. This is alluded to by the mechanism of reference [5] cited
> in the introduction, but I think this should be stated in the main
> text - it is an important assumption.
We have moved the statement into the main text in the conclusions.
> 3) Introduction line 11: `..and by measuring the fraction of charmless
> semileptonic decays...' This phrase in context sounds like this
> technique is new, whereas it is the same as used by previous
> experiments. The sample purity has been improved, but the basic
> technique is the same, so this should be rephrased.
This is just a statement that by measuring the ratio our experimental
error is suppressed. The distinction to the previous analyses is made
in the first half of the sentence and is separated by a reference and
a ','.
> 4) Figure 1(b): There seems to be a big dip in efficiency for m_X
> values just above 1 GeV - the dip in the spectrum is much more
> pronounced for the `all requirements' histogram than the `p*>1 GeV'
> histogram. What is the cause of this ? More generally, there is no
> comment on how the efficiency varies as a function of m_X.
**
> 5) Table 1: The first three variations (changing the m_X cut) have
> highly correlated data samples. Are the statistical errors on the
> R_u values calculated from all the data, or only the part which is
> independent in the two samples ? If it is all the data, the
> variations look a bit large to be coming from pure statistical
> fluctuations as is suggested in the text. E.g. for m_X<1.55 GeV the
> result is 2.06+-0.25, then for m_X<1.70 it is 2.35+-0.28.
> Taking the quadrature difference of the statistical errors as an
> (admittedly crude) estimate of the uncorrelated part, the
> difference is 0.29+-0.13.
RF> With the correct calculation of the uncorrelated errors they are <1.5
sigma apart
> 6) Page 13 2 lines after the R_u result, the double ratio between R_u
> for B+ and B0 is quoted, but no discussion is given of what this
> result means, or why it is interesting. Is it expected to be unity
> ? Some interpretation would be helpful, if the result is
> interesting enough to derive and quote.
RF> I would reply to the referee that " this a by-product of the way we do
the analysis
and can be partly obtained from the other information present in the
paper. We add it because we
have all the information to compute the correct error, but we do not want
to give too much emphasis on the result:
the naive calculations would predict a number close to 1".
> Minor points
> ------------
>
> a) Abstract line 5:
> 'hadronic mass distribution' is ambiguous in this context. I
> suggest something like 'mass of the hadronic system accompanying
> the lepton'.
We replaced this with your suggestion.
> b) Page 9 line 3:
> 'estimated as the fraction of signal events with m_ES>5.27 GeV'.
> This is difficult to understand without the context of the fit
> shown in Figure 2 - moving the figure reference and mention of the
> fit earlier would help.
We rearranged the text so that the \mes fit is described where it is
used for the first time.
> c) Page 9, last paragraph line 3:
> Presumably the m^2_miss<0.5 GeV^2 requirement is applied before the
> kinematic fit setting it to zero - it would be clearer to mention
> that this value is calculated before the fit.
We changed the ordering of the cut description which takes care of
your point.
> d) Page 9, last two lines:
> The pi_s rejection is a little obscure. Is this test applied to all
> pions (which are then used to calculate the neutrino mass), or only
> those accompanied by an exclusive D0 ? The text could be clearer on
> the procedure which is actually followed.
RF> a possible replacement for the whole sentence:
"We suppress B->D*lnu background with D*->D0pi+ by looking for pions of
opposite
charge of the lepton. Exploiting the fact that in the D* decay the pion is
almost at rest the
kinematic of the D* can be reconstruced without reconstructing the
D0(ED*~...).
We therefore estimate the missing mass of the event as ... and require it
to be ...<-3."
> e) Page 10, after the equation:
> Several numbers are quoted with errors. What are the errors - do
> they include both statistical and systematic effects, or only a
> subset? This should be stated.
We have made it clear that these errors refer to the statistical
uncertainty.
> f) Page 11, line 3:
> What is the error on the 6.8% lepton background?
RF> 6.8+/-1.2 %, but we don't feel like quoting it otherwise we would have
to add many more errors
> g) Page 11, paragraph 2 line 6:
> 'the first bin is extended to m_X < 1.55 GeV' This is hard to
> understand - extended from where ? Presumably you want to say that
> it covers the whole region from 0-1.55 GeV ?
We changed the wording.
> h) Page 12 line 3/4:
> plurals: '... and kaons *are* estimated...
> by varying *efficiencies* by...'.
We changed according to your suggestion.
>
----------------------------------------------------------------------------
> Report of Referee B -- LG9660/Aubert
>
----------------------------------------------------------------------------
>
> This is certainly an important result pioneering a new method of
> measuring a crucial parameter in the Standard Model of particle
> interactions. All that is said in the first paragraph is correct. It
> deserves prompt publication.
>
> On the other hand this letter is very difficult to read leading to a
> great deal of confusion about what is actually done. Perhaps this
> results from presenting a very condensed version of a result that
> should be described in a longer format. This letter would be much more
> clear if it could reference a longer, more complete description of the
> analysis. Failing that the authors need to address the specific points
> I make below to clarify and improve the reader's ability to understand
> the analysis.
We are in the process of preparing a longer PRD providing more details
of the analysis.
> Footnote 5) is overstated. The extraction of the charmless
> semileptonic branching fraction has little to do with quark-parton
> duality. On the other hand the extraction of Vub depends on duality.
We moved the sentence of the footnote to the main text in the
conclusions after the extraction of \vub (where it is relevant).
> Not enough information is given about DeltaE used in the selection of
> B_reco. A plot of some sort would be most useful, presumably DeltaE
> after a selection on m_ES comparing data and simulation, but failing
> that some words on the average resolution on DeltaE and the quality of
> the agreement between the simulation and data.
RF> DeltaE windows range from 35-75 MeV. We can add this info and comment
back to the reviewer that data-MC agreement on this
(which is not bad) is not of interest to this analysis
> It would be valuable to know what fraction of the 1097 B decay modes
> are actually used after the requirement on purity.
We have included this information in the text now.
> I find "The purity of events with a high momentum lepton..." to be
> imprecise and confusing. I suggest "For B's decaying semileptonically
> to a high momentum electron or muon recoiling against B_reco the
> purity of the B_reco selection is 67% (see Figure 2a)."
We have formulated a more precise statement.
> "Prompt" is jargon well known to those interested in semileptonic
> meson decays, but is otherwise confusing. It is not defined in the
> letter. The sentence, "For charged..." should be rewritten to
> eliminate "prompt" or "prompt" should be defined on its first usage. I
> suggest, "For charged B_reco candidates, we require the charge of the
> selected lepton to be that expected from the decay of a b-quark with
> the proper flavor for a B meson recoiling against the candidate."
We have removed "prompt" and use "primary" instead. This is well
established (does this exclude it from jargon?) in semileptonic
papers, also in PRL.
> The sentence, "For neutral..." has me completely confused. I think the
> authors are trying to say that when signal fraction and backgrounds
> are used later in the analysis they take into account the known rate
> of mixing of neutral B mesons which can result in either charge of
> lepton appearing in B's recoiling against B_reco. If this is what they
> mean then they should say it later since it is of no consequence at
> this point. If they mean something else then they must rewrite this
> sentence to make it clear.
At this point in the paper we describe the selection of leptons used
in the further analysis. It is here that events with the wrong charge
flavor correlation are removed (for B+ events). Therefore, we feel
that this is the adequate place to describe that we retain both charge
flavor combinations for B0 events.
> I am confused on the selection sequence for B->X_u l nu candidates.
> Prior to the paragraph beginning "To select..." the neutrino mass
> equivalent to the missing mass is fixed to zero by the kinematic fit
> for m_X. Then the missing mass is required to be consistent with zero.
> I think the authors are trying to draw a distinction between the
> neutrino four vector which they require to have zero mass in the
> kinematic fit and thus no longer is equal to the missing mass four
> vector which for B->X_u l nu candidates is selected to have mass
> consistent with zero. They do this with the subscript nu and miss. If
> this is what they mean then this distinction needs to be made more
> clear. Perhaps the order of these should should inverted to make it
> clearer.
We changed the ordering of the cut description.
> Is the reported resolution on m_x for candidates that also pass the
> B->X_u l nu selection criteria? As written it appears to not be so in
> which case what is point of giving this number? The fit to the m_X
> distribution is done on those events passing the B->X_u l nu
> selection.
The quoted number is the number after all analysis cuts. Due to the
reordering of the cut description, this should be clearer now.
> The language describing the D*lnu partial reconstruction is unclear.
> Instead of "...and require for the neutrino...", it would more clear
> to say "...a
> nd eliminate if the neutrino..." At least I think that is
> what is done. If not then the authors need to tell us what they do
> with the recovered D*lnu events.
We have mostly followed your suggestion.
> Figure 1 is confusing. Is it before or after a detector simulation? It
> is hard to tell if any resolution in the figure agrees with what is in
> the text as the text gives the average resolution on m_x which is
> clearly much larger than what is apparent in the figure for low values
> of m_x. Also "Signal MC" in the caption is jargon. Better is "MC
> Simulation of the B->X_u l nu signal".
In the caption we state that figure a) is the *generated* distribution
and that figure b) is the *measured* distribution. We feel that this
clearly expresses that figure a) is before simulation and figure b) is
after simulation.
We have changed the jargon phrase.
> It would be good to report the effect of the selections made going
> from the initial sample of B->X_u l nu candidates. Do the fractions
> cut away by the various selections (1 lepton, charge, D*lnu partial
> reconstruction, kaon veto) agree with the predictions of the
> simulation? If so this would strengthen the belief that the background
> is well modeled.
We now describe (in the systemnatics section) the very good agreement
of data and MC simulation in the fraction of events passing the
subsequent application of selection criteria.
> Does the shape of m_ES distribution for the B_reco signal agree with
> the prediction of the simulation? Again if so it would strengthen
> belief that simulation is a good model of the data.
RF> the resolution is the same but there is a small offset due to the
different beam energy in simulation.
It should be stressed that this resolution is dominated by beam
characteristics and that we
let the resolution parameters float and therefore this is not the right
quantity to look
for appropriate MC simulation
> Table I and the description of the fit to Figure 3a is very confusing.
> First when N_sl is first mentioned in the text, Table I is not
> referenced. Thus I was left trying to figure out if the N_sl in the
> table was the same as the one in the text. I think it is, but
> reference to the table in the text when the fit to m_ES to extract
> N_sl is explained is necessary to clarify. Second the first three rows
> of the table leads one to believe that the fit is only performed on
> the subsample that passes the indicated m_X cut. Careful reading of
> the text indicates that this is not the case, but rather the cut only
> defines, I think, the size of the first bin in which N_u and N_c are
> reported. This is confusing as the table leads one to believe that the
> first three fits are performed on different subsamples and even after
> reading the text many times I am not sure that m_X < y means what I
> think it does. The meaning of m_X in the table needs to be made clear.
We now make it explicit that N_{sl} in the text and the table are the
same. We changed the wording for the definition of the first \mX bin,
in the text and in the table.
> The entire description of the fit could be made more clear by adding
> some guidance as to how the fit to the m_X distribution works. N_c is
> essentially determined by the data in the m_X bins above 2.0, and is
> extrapolated into the low region by the shape of the m_X for this
> background as predicted by the simulation. Saying this explicitly
> would aid understanding the discussion of the systematics.
**
> It would very much aid understanding if the m_X distribution for the
> control sample is shown comparing data and simulation. At least
> characterize in words the agreement between data and simulation for
> high values of m_X. At the moment the authors are asking to accept on
> faith that they have a good understanding of the m_X distribution of
> the B->clnu background.
The good agreement between data and MC simulation both in the signal
and control samples is expressed in a sentence we added to the
systematics section.
> How is the 1% error for track finding efficiency applied? Is it 1% per
> track or is 1% the overall effect of the uncertainty? It is not clear
> that the identification efficiency errors are reasonable. The authors
> need to give some evidence that the variations is ID efficiencies and
> mis-ID rates given in the paper are reasonable.
The tracking error is estimated by randomly removing tracks from the
analysis corresponding to the uncertainty in the track finding
efficiency. We state this now in the text.
The justification for the uncertainties for the particle
(mis)identification are detailed in the two papers we quote for this
purpose.
> No error seems to appear for the choice of m_X. Why not? While the
> variation in R_u shown for the choice of m_X in Table I is probably an
> over estimate of the error, the Table shows a trend, the highest R_u
> for the lowest m_X and the lowest for the highest, that seems to
> indicate that the choice of m_X and the fit value of R_u are
> correlated. At least the authors should comment on the trend and
> explain why the choice of m_X does not contribute to the systematic
> uncertainty.
RF> The trend on mX is compatible within the uncorrelated errors.
Nonetheless we
believe no error needs to be assigned because a dependence on mX of the
measurement
would be due either to detector/background effects, and this would be
covered by the experimental
systematics) or by the fact that the theory model is incorrect, which is
in the theoretical error.
It would therefore be double counting...
> There is no comparison with other, exclusive measurements of V_ub. One
> PRL 90, 181801 (2003) is even by the authors collaboration. Besides
> comparisons with other inclusive measurements something should be said
> about the exclusive measurements.
We have added a comparison to exclusive results to the conclusion.
|
