Hi all,

 this is our (Daniele,Riccardo and Giancarlo) proposal for the answer
to the review committe. I marked with a * what it should be corrected in
the BAD. If you have something to add reply to this mail and I will
prepare an answer to post to the Review HN.

 Daniele

==========================

>Dear authors of BAD#347

>The reviewers (Marie-Helene Schune, Stephane Willocq, Franz Muheim)
>have discussed your analysis and would like to
>have a phone meeting with you on
>        Wednesday 17 April
>        8 am to 10 am (SLAC time).
>Urs, can you please arrange a phone conference for this slot. Thank you.
>
>First we  would like to congratulate the authors on a well written
>BAD which was quite easy to read.
>This is an important analysis and will deliver a very important physics
result.
>
>Below we list a set of questions and concerns we have with this analysis.
>We would like to ask the authors if they have plans to also write a
>physics note/ conference paper on the timescale of DPF,
>i.e. within the next few weeks?
>Such a note would get more widely read in the collaboration and we
>feel that this would be worthwile.
>
>Section 1.1 and 1.2
>===================
>
>Could this section be slightly  expanded, i.e. in the 3rd paragraph of
>page 6, it would be useful to add a figure with the hadronic mass for
>b->ulv and  b->clv (generator level).

* we will do that. We should add also a reference to the Combo BAD in
which
there is a detailed description of the Vub generator and we don't want to
duplicate it in our BAD.

>The choice of symbols M_sl and M_u is unfortunate.
>In many HEP brains, variables of the type M_a are automatically
>allocated to masses and thus this can easily cause confusion.
>We suggest to use something along the lines of
>N_sl,obs and N_sl,t and  N_u,obs and N_sl,t, where "obs" indicates
observed
>and $t$ indicates "true", respectively.

M = measured, sl = semileptonic, u = b->ulnu. We could replace M with
N_obs but we are not sure that N_sl,obs is easier to be understood.
We therefore prefer to stick to the present convention.

>page 7
>The error on the b quark mass, and its effect
>in formula 6 needs to be consistent.
>The size of the error on m_b is an ongoing debate.
>The error of  +-0.035(m_b) given in formula 6 corresponds
>to an error of 60 MeV on m_b.
>At the CKM workshop, it was quoted that
>   m_b = 4.58 +-0.09 GeV
>But Marco Battaglia from the LEP Vub working group says now
>that the agreement (between who?) at the CKM workshop is
>   m_b = 4.58 +-0.10 GeV
>which in formula 6 translates into  +-0.052(m_b).
>
>In addition, the lastest update by Uraltsev (hep-ph/09905520)
>includes 1/mb^3 corrections, so that the term +- 0.0030(1/mb^3)
>can be dropped.

thanks for the good comments. The way we get total number of b->ulnu
events and theoretical error is still under discussion.

>This  analysis proposes to measure the ratio of branching ratios
>B(B->Xulnu)/ B(B->Xlnu), and this is explained qualitatively.
>Can this argumentation be quantified?
>The cost of this method is an additional 4.0% error due to
>the measurement of  B(B->Xlnu) in BaBar.

We would suggest to use the PDG error on BR(B->Xlnu) is ~2%. At the moment
the systematic error due to the mes fitting technique is of the same order
of magnitude and the fit effects cancel out
since they appear in the numerator and the denominator of the ratio of
BR's. If we measure an absolute BR the error on the total number of
reco'ed B should be  larger since the purity is very low (see fig 1
plot 1) and it is difficult to parametrize the background on MC since the
modes that are contributing are different between data and MC.

>The systematic due to limited knowledge of the individual
>B->D, D*, D** l nu branching ratios also affects the denominator
>and not only the numerator (b->clv Background) in equation 4.

In the denominator we are sensitive to the difference in the P* spectrum.
The effect of different BR for D, D*, D** on the P* cut should be small.


>Section 2
>=========
>
>We have a few questions regarding Figures 1 and 43 - 46.
>How relevant are these results? These are not used
>in the extraction of the result?

Those plots are there to show the sample, its purity, and the difference
in
purity among various MC.

>We note that the  superblock purities in Figure1, do not agree with the
>definition in the text, i.e. Superblock 1 has 73% purity,
>Superblock 1 and 2 have 40% combined purity
>How is the a priori purity determined?

Part of the collections were produced with a different release
(different PID,...) and the definition of the purity is slightly different
from the one I used to create the collections. Here S and B are
determined with a cut M_es>5.27, the reduced collections had M_es in 5280
+- 2.5 sigma(=2.85 MeV) interval.

>What is the justification for using Xbal function for the signal?

because of the presence of neutrals in the reconstructed sample. The fit
works fine, see for instance at the plots 9, 10 on page 9 (and in the
cocktail most of the modes with many neutrals are not present)

>How many sigma wide is the DeltaE cut?

~2 sigmas. This cut is mode dipendent (see BAD 271).

>Wouldn't DeltaE be better to choose best candidate ?

B candidates within the same B and D mode are chosen on the basis of the
deltaE. This
way of selecting the best B is not appropriate once we have modes that
have very different purities. The selection would be biased since the
probability to get a low DE is higher for modes with a large combinatoric
background

>in BAD 347 D and D* are seeds
>in BAD 271 D and D* are pre-seeds,
>and seed is D(*)+x where x is an element from piKList,
>please clarify.

* we will put pre-seed everywhere.

>Why no 2H or 4 H in table 3 of BAD271?

charge conservation

>How are the systematic errors on the event yields determined?

the systematics section has been drafted but not yet handed to you.

>What are the main contribution to the backgroundsin the different
>Superblocks? Is it from continuum events or from other BBbar events.

to have an idea, look at

 http://www.slac.stanford.edu/~ursl/vera/plot2-mesallevents.eps

>page 10 Monte Carlo samples.

>What is the difference between breco-cocktail and
>augmented breco-cocktail samples?

breco-cocktail is the official cocktail that includes the modes used
by the hadronic breco mixing. the augmented one contains more modes
since we are using SemiExcl reconstruction. In detail

- three Dc modes are included are included: Dc->Kspipi0,Dc->Kspipipi,
Dc->kpipipi0
- D0 in B+->D0X can go in the usual 4 modes(kpi,kpipi0,kpipipi,kspipi)
and the same for Dstar0->D0

>Does any of the cocktail MC have a preselection for leptons on the
>semileptonic side?

no

>The Branching fraction for b->ulv are different for charged
>and neutral Bs.
>
> B0 -> u e nu:  0.932e-3 (only low mass exclusive states)
> B- -> u e nu:  1.017e-3 (only low mass exclusive states)
>
> B0 -> u e nu:  1.351e-3 (hybrid MC)
> B- -> u e nu:  1.346e-3 (hybrid MC)

* agreed, we will put two numbers in the BAD.

>                                             Section 3
>=========
>
>Is Figure 3 a place holder? Please plot muon efficiency and fake rates.

we have it. we will put it in the BAD.

>Section 3.3 should be moved to after section 3.7

* There must have been some confusion in the final editing. The intended
order of the section was:

3.3   reconstruction of the hadronic system and the missing mass
  3.3.1 track selection
  3.3.2 charged kaon identification
  3.3.3 neutral selection
  3.3.4 kinematic fit application
3.4   event based cuts...

the logic was to introduce MX and M_nu before using them in the event
based selection.
3.3.n are just ingredients that go in the definition of MX and M_nu

>p14. Equation 7 has a LaTeX problem

* correct

>The "ChargedTrack" list contains a lot of fake tracks. Why was this list
>chosen? Has the "GoodTrackVeryLoose" list been tried?

we are investigating this possibility

>Please add the Figure for the Ks selector.

* we'll do that

>Are you using the "CalorNeutral" list for neutral objects?

* yes, we wil remove the other two lists from the BAD.

>The section about kinematic fitting is not easy to understand.
>Appendix C would benefit by adding more text to explain the technique.
>For example, not all symbols are defined.

* we will improve this

>It needs to be stated that the X system is the hadronic
>part of the semileptonic decays.

* we will add a sentence there.

>Figure 5, please define what is plotted,
>eg. residual DeltaMB_reco = sqrt(EB^2_reco - pB^2_reco) - MB(PDG)
>    residual DelatMB_Xlnu = ???

ok

>Why is the resolution on MB_reco after kinematic fitting 8.6 MeV,
>which is much larger than the typical MES resolution of
>3 - 4 MeV (Figure 1)?

that resolution of 3-4 MeV should be compared with the sigma from a
gaussian fit to that distribution and not to the RMS.
If it turns out to be larger we will start to be worried.

>Before Figure 6 & 7 we would like to see the reconstructed
>hadronic Mass MX for b->ulv and b->clv events.

* we will show a comparison between reconstructed and fitted Mx for
b->ulnu and b->clnu events respectively after the kinematic fit description.

>Figures 6 & 7: sign of DeltaM definition is not the usual
>        deltaM = M_X(measured, fitted) - M_X(true)

>Please explain why BIAS = MEAN*RMS. Why not BIAS = MEAN/RMS?

Oliver, can you please comment on this?

>Do plots 8 to 13 have a sideband subtraction?
>For the right-hand side of Figures 8 - 13, it would be useful to
>add other backgrounds, such as cascade leptons and maybe
>lepton fakes. Especially for Figure 9.

no, we don't subtract the sidebands here.
The background for the cocktail MC after the cuts is very very small.

>How is it possible to have 0 leptons in Fig 8?

this is before the cut on that variable, actually all the cuts are applied
except the plotted one. This applies to plots 8-13 as well.

>Could you please add a Figure of
>the momentum of the B in the Upsilon4S rest frame
>and of the corresponding resolution (magnitude, polar and azimuthal
angle)
>in order to demonstrate the gain due to
>boosting into B frame?

we will verify this.

>It is not clear why not requiring a sign correlation for Neutral B events
>is better?

we do actually use the sign correlation correcting properly the mixing
effect. In section 4.1.2 we describe this.

>Comparison of Figures 12 and 13  indicates that the Ks purity is quite
low.
>This can be improved by adding additional requirements,
>for example distance from the  B vertex, ...

the cut on the Ks here is very loose (actually no cut on the Ks mass, we
just took the list from Beta). Now we apply a cut at three sigmas
(+-12MeV) and the situation improves a lot.

>Figure 11
>Caption: The right tail in the b->clv plot is mainly
>due to missing particles (2nd neutrino, KL).

the effect of the 2 neutrino is not so crucial as far as the right tail is
concerned. If you compare the signal depleted plot (that actually contains
the 2nd neutrino effect) for Mnu^2 on page 30 there is no right tail. The
effect is determined by the loss of one charged kaon (since we request
charged kaons in the signal enhanced sample and b->clnu events fall there
if the kaon is missing) and Kl, actually not reconstructed at all.

>what is the MM2 resolution in the core, eg with a fit to 2 Gaussian?
>This is a useful quantity to compare with other analyses.

we will try this

>MM2 is a 1-dim representation of 2 measurements: Emiss and pmiss.
>Have other cuts, e.g. on |Emiss-Pmiss| < xxx or |MM2/Emiss| < xxx
>been investigated?

>Table 5:
>It is not clear what the starting point for calculating this efficiencies
>is? Is it a B-reco candidate plus a lepton? Is there already a lepton
momentum
>cut?

a B-reco candidate plus one or more leptons with a momentum greater than
(Urs can you please put a number?)

>Before the Data/Monte Carlo comparison
>we feel it would be nice to show the equivalent of Figure 1, but with
>a lepton tag.

* yes, you are right, probably it is useful to show the effect of the
lepton cut.

>It would also be useful to show  a Data/Monte Carlo comparison
>before splitting into b->u enhanced and depleted samples.
>This allows to separate data/MC disagreements due to the
>kaon veto from the other cuts.

we are above all interested in data-MC comparison after the lepton cut and
including the cut on the kaon (we actually do the measurement on this
sample). The kaon efficiency disagreement is shown in the first row.

>It would also be interesting to give the agreement between number
>of data and Lumi normalized MC events at all stages.
>In this comparison we would like you to show the different
>components in the Monte Carlo, e.g. signal, b->clv, cascade, fakes

the way we extract the final number (ratio of BR) does not depend
on lumi normalization problems, which depend on breco effciency, lepton ID
and so on.

>Figure 14
>The muon spectrum is much harder in data than in MC. Any explanation?

* the plots are normalized to the same area. Probably the high momenta
region is compatible and the region at low p has differences (due to
muon efficiency discrepancy at low energy between data-mc). We will
add plots normalized to the same number of reco'ed B's.

>Figure 15
>The hadron spectrum is softer in data than in MC. Any explanation?

we see a problem just in the first bin. We will check the effect
of goodtrackloose-like cuts on this plot. We are also going to explore the
possibility of higher rate of tracks splitted into an SVT only and a DCH
only track in data compared to MC.

>Figure 17:
>the first two columns seem to indicate that the effect of the cuts
>on the MX distribution is to suppress low MX.
>Since MX is the final signal variable we would like to see
>a plot or plots for both, the signal and background efficiency versus Mx
>after each of the six cuts listed in table 6.

we will add the signal efficiency as a function of MX. There are a couple
of remarks to be done:
the bulk of the low Mx events that are cut away are background events (the
M_nu cut removes the events where a particle was lost); the dependence of
the efficiency on MX will be accounted for in the systematics from
Vub multiplicity.

>Figure 18:
>When compared with Figure 11, it seems that b->u enhanced and b->u
depleted
>plots have been switched, i.e. the b->u enhanced plot has a tail
>for positive MM2, which is typical for b->clv.

Figs 11 are the b->ulnu contribution and b->clnu contributions both on
the  b->u (signal) enhanced sample.
Fig 18 shows the b->ulnu (signal) enhanced and  b->ulnu (signal) depleted
samples. b->ulnu (signal) enhanced contains at 90%  b->clnu events.

>How does varying the track criteria, e.g. using "GoodTracksVeryLoose"
>affect the difference in the Q=0 bin between data and MC?

it is on the list of things to do

>Figure 19
>Is this the relative error on the signal yield/Branching ratio or on
>Vub?

* a typo here. all these numbers (in all optimization plots) show the
relative error on the ratio of BR's.

>Please state in the text what is the lower limit on MM2 for the
>asymmetric cut.

* ok

>Please explain Figures 21 and 23? What is the difference? Why is
>there a jump between 0.9 - 1.0 GeV in Figure 21.

it should be descibed in the text.  Figure 21 is relative to the
cut on the most energetic lepton. Figure 23 shows the dependency on
cut on the energy of the lepton in the counting of the total number
of leptons. These to cuts could be in principle different.

>Table 6 and Figure 22
>We would like to see this optimisation also for the different
superblocks.

We already optimize  the required level of purity

>Section 3.12
>This optimisation study seems very convolved to us.
>What is the rationale for not doing a
>a comparison of variables like MX mass resolution before kinematic
fitting,
>total neutral energy resolution, MM2?

Mx resolution optimization is already there. Since Mx is the variable used
to extract the signal no energy resolution or MM2 study is actually
needed.
Why before the kin fit?

>Figure 30 is too small

* ok


>Section 4
>=========

>Figure 34 and 41 larger please.

* difficult to enlarge them. If you like we can put the chi^2 on each fit.

>Equation 12
>Why are there no efficiencies in this equation, please explain.

actually, after some calculations, the only efficiency present
multiplies the factor Nb and then it is irrelevant (already included
in the epsilon_b we use to get the ratio of BR's)

>Equation 13 and 14, page 45
>Why are the two samples fitted separately ?  A fit to both sample
>would allow to extract M(b->ulv) and M(b->clv) simultaneously.

yes, this is a possibility we are exploring at the moment expecially in
the fit that takes into account different multiplicity categories.

>Figure 36
>Why is the fit not identical to the histogram?

two reasons:

1) the signal Mx model (to get the signal tail for Mx>1.6GeV) is obtained
on a larger statististics with repect to the fitted one and the same for
the "other" component
2) since each bin corresponds to a Mes fit, it is not mandatory that
component 1+2+3 gives exactly the total (mes fit result for the signal can
have small fluctuations)

>Figure 38 is note very clear. Maybe two  1-dim figures will be more
clear.

* we will add the corresponding table.

>page 49 Equation 15
>Is this a typo?
>In the Monte Carlo we have
>B(B->Xclnu) = 10.4  %
<B(B->Xulnu) = 1.2 e-3 average of different Vub mix
>the ratio is about  1%

this is a typo, you are right. This number should be 1.15% instead of
11.5%

>And finally a few suggestion for systematics
>- Separate neutral and charged B.
>- Separate Muons and electrons
>- vary B->D,D*,D** l nu branching ratios
>- vary Vub model for signal efficiency

please, have a loook to the web page we presented last monday

http://www.slac.stanford.edu/~daniele/vub/systematics.html



                                       the authors