- effects of the 6MeV mass shift on the formfactors will be investigated

  - inconsistencies in the parameter r values from different fit
approaches will also be investigated. Alan has pointed to a 6% difference

Parameter r was defined before normalizations and integration of signals so as defined it does not need to be equal to the ratio of events in two peaks. Since this parameter is not used in the analysis (eq. 11 has been corrected) the discrepancy is no issue.

ok, here they are - will be updated in the note. 
parametrizing the dependence with 2nd order polynomial. 

Comparison of the kinematic variables before/after weighting. 
The ratio of muon pt spectra in data and MC (plot labelled ptle) was used for the weighting. 
take good D* candidates after mass cuts but before additional lifetime cuts. the color coding 
of histograms is: blue = data, red = MC before weighting, green = MC after weighting 

D* cuts  

D* cuts && nCFT>1

D* cuts && nCFT>1 && nSMT>1

D* cuts && nCFT>5 && nSMT>1 

as one can see data and MC are  different by (94.1-93.2)/94.1 = 0.0096 (i.e ~1%). 
this correction will be used to scale down the total signal efficiency

     * sensitivity of the efficiency ratio (Eq(7))  to some scaling
factors will be established, in particular the contribution from the
significance cut with respect to the PV for pi** (expected to be small,
based on previous studies)

The ratio of IP significances (wrt primary and secondary vertices) is used as selection
in the analysis. Thus, scaling them simultaneously will cancel out with good precision.

Scaling IP (wrt secondary vertex) error only (and thus grossly overestimating the effect) :
IP error scaling factor +20% : relative eff change - 5.2%
                        +40% :                     -13.9% 

  - mass resolution of 6.5 MeV (from MCarlo) or larger (by 30% as AZ suggested)

        * fits to the data with 8.5 MeV look visibly worse, however
the change in chisq is only 43-->46 (see Andrei's new plots)

fits for mass resolution 8.5 MeV +- 30%. we think this overestimates the mass resolution, see reasoning
below why the mass resolution is better than 8.5 MeV.
8.5 MeV
8.5 MeV + 30%
8.5 MeV - 30%
D1 width free, D*2 width 6.5 MeV


        * Rick suggested to establish the data-MC difference in the
mass resolution by using the psi(2S)-->J/psi pipi channel with the
J/psi mass constraint. The experience from the X(3872)
paper indicates that the difference is smaler than 30%.


a) psi(2S) -> J/psi pipi comparison has been done by Brad and Co. (X(3872) analysis) 

         Width
MC       8.9 MeV
Data    10.4 MeV  

difference in mass resolution 14.4%

b) B+ -> J/psi K+   comparison by Ralph (Bs -> mumu analysis)
       Mass        Width
MC     5.279 GeV    37 MeV
Data   5.273 GeV    41 MeV

difference in mass resolution 9.8%

we think that the case b) is more adequate for the D** case since there is one track (K+) and 
the J/psi mass is used as constraint. Therefore we propose to scale the mass resolution by +10% and vary it
+-10% from this value for the systematic error estimate.

        * Andrei reported that some other methods to extract this
number from MC indicated that the MC  resolution could be actually better, as good as 5 MeV

mass resolution has been estracted from MC using two methods
a) plot (reconstructed mass - generated mass), fit with Gauss 
b) plot reconstructed mass, fit with Breit-Wigner convoluted with Gauss. Sigma of Gaussian is a fit parameter.
Width of BW is fixed to the generated value.

the results are (they have changed compared to the note4324 v3.2 because i found that i did not scale the histogram
errors together with contents. so the errors changed, the fits changed and the agreement between two methods 
is better now)

method    D1(2420)  D*2(2460)
a)        5.7 Mev   6.5 Mev
b)        5.1 Mev   5.5 Mev


another set of numbers for the weighting based on the D meson pt
method    D1(2420)  D*2(2460)
a)        5.4 Mev   7.0 Mev
b)        5.0 Mev   6.4 Mev


in principle the two methods should give the same results but some discrepancy is observed. 
the discrepancy is not fully understood though now it almost within errors. 
we propose to average two methods and use the difference 
between these two estimates as additional systematic error on the resolution. then and also taking 
into account scaling by 10% from data/MC comparison we have
            
               D1(2420) D*2(2460)
resolution     5.9 MeV  6.6 MeV  
for ptD weight 5.7 Mev  7.4 MeV   (assigned +-20% error covers this difference) 

we propose to vary the resolution by 20% (as currently in the note) to estimate the associated systematic error. 
this variation should still cover the uncertainty on the difference between a) and b) and the uncertainty of teh data/MC
scaling. 

the new fit assumes different mass resolutions for D1 and D*2 (as above). 
also this fit uses the data sample with removed double events (total 2.3% sample reduction)

Dstpi2_v33.eps

the final numbers changed very little - will update later...

  - fixed parameters of the fit Belle vs CDF vs PDF

         * the consensus was to stick with the Belle values
         * results with the PDG value for the D2* width will be added in a footnote

will get to it

  - effects of the 6MeV mass shift on the formfactors will be investigated

will get to it

  - inconsistencies in the parameter r values from different fit
approaches will also be investigated. Alan has pointed to a 6% difference

will get to it

  - D** single- double pi decay modes. a saturation by a single pi decay mode was assumed.  
Andrei will provide more justification for this.

double pion mode is for ex. narrow D**-> D* pi pi. As we don't reconstruct it the Br
we measure could be underestimated. however it is the usual practice to assume this decay rate is zero. 

except from the lifetime ratio note 4280 describing the argument follows. 
as one can see the DELPHI measurement is compatible with 0 (line 4 below)

with more data we should explore this mode and try to do the measurement ourselves.

  -  Interpretation of results

      Alice pointed out at the potential differences in the B meson compositions at the Tevatron and LEP,
as discussed at the previous meeting (Bs->uu)

i guess this question is about using LEP data for f_d & f_u. since we normalize to the semileptonic D* signal
we dont' need these numbers for the measurement itself. however it's true that we use the D* Br from 
PDG (measured by LEP) to calculate the absolute Br. it looks as statistics increase and we start to 
compete with LEP in precision this question will be coming up frequently so may be it's worth to think 
about general party line re this issue. to me the fact that we use things approved by PDG is good enough. 

re possible Bs contribution :
the only reason Bs production could matter are modes Bs -> mu nu Ds** with Ds** -> narrow D**K which in principle 
may exist. They were never observed and are not present in generators. Estimate of the Bs contribution 
into the semileptonic B->mu nu D* X sample is 2+-1%. this mode is possible when Bs-> mu nu Ds** with
Ds** -> D*K and the estimate was elaborated in d0note 4280, page 9. so the expected contribution 
to the narrow D** sample should be even smaller.

  - Interference effects

* narrow states - the overall efficiency of 50% is primarly due to the
significance cuts, therefore the assumption about
full integration over angles may be justified.
Nevertheless, Andrei will look at the effects of kinematic cuts
on integration over spherical harmonics.

here it's important that the efficiency for D** decay angles is the same for D1 and D*2. 
presumably this efficiency could be different only because of different pt spectra for D1 and D*2. 
purely geometrical effects should be the same for D1 and D*2. the plot below shows distribution on angle between 
the D** flight direction and direction of D* in the D** rest frame for D1 (points) and D*2 (green histo).
the dip at 1 corresponds to the case when pi** is emitted in the opposite direction to the D** flight so 
it has small pt and goes under the threshold. 




the plot below shows the same cos() for D1 and three values of 
the pi** pt cut : 0.18 (green), 0.3(bleu, used in analysis) and 0.5(red) Gev cut. 


the above information shows that the effect if small if any but we still need to quantify it.

* narrow-wide states there were no objections to the current estimate for the size of
this effect

  - a PRL draft v 1.0 is ready, it will be distributed to the
B-physics group and the EB only after conclusion of some of the tests
mentioned above.

  - our next meeting is scheduled for Jan. 13, 2005.