Discussion on the Averaging of |Vub| in inclusive semileptonic B Decays
=======================================================================

Present:
--------

R. Faccini, G. Eigen, K. Schubert,  T. Iijima, D. del Re, N. Uraltsev,
F.  Muheim, Y.  Sakai, C.  Schwanda, T.  Onogi,  V.  Luth, H.  Kakuno,
A. Sarti, U.  Langenegger

(no particular order,  except roughly along  the table ...)


Miscellaneous
-------------

o The absence of CLEO is  unfortunate, but we decided to formulate our
  thoughts on this topic anyway so  as to start the discussion and try
  to progress towards a well  understood average of |Vub|. It was felt
  that  the average  will  wiggle  somewhat in  the  beginning due  to
  improvements in the measurements but also due to improvements of how
  errors are treated (as correlated).

o A  table of  (correlated) errors was  discussed. See below  for more
  details.

o  The question as  to when  we should  start with  averaging: General
  consensus that this will be a result that will change over time both
  due to new results and due  to new insights into errors, so there is
  no specific reason not to start as of now.

o It is understood that the  systematic errors will need to be updated
  over time as e.g. branching fractions become better known (and hence
  reduce some of  the correlated errors). This is  obviously also true
  for (excl) b->ulnu branching fractions.

o We actually  did not discuss it in full detail,  but we should think
  about  whether  we  also  average  BF(b->ulnu) or  whether  this  is
  extracted from |Vub|.

o  It  might  be  useful   that  each  experiment  has  some  form  of
  parametrizations  of  errors  as  a  function  of  uncertainty  with
  experiments, so that more than  just a linear scaling to correct for
  chaining errors in input parameters can be done.

o In the parallel session, a request  was made that it would be a good
  idea to have a common repository for code that is used by all people
  (e.g. to calculate a common Vcb). The same applies to Vub (though to
  much  less extent,  as the  formulae are  simpler), but  it  will be
  implemented in the HFAG pages.
  

Correlated Errors
-----------------

1. Detector:

  Geant  (Belle uses  G3, BABAR  G4)  not clear  how much  correlation
  exists.   Not clear  how  much the  simulation  of (e.g.)   hadronic
  interactions affects the  various analyses.  Other experiments apart
  from CLEO have specific cuts against "spurious" tracks and clusters.


2. f+/f0:
 
  This should be trivial to  adjust as it affects only a normalization
  (see below for lifetimes).


3. Background modeling:

  b->clnu: FF(especially  endpoint), BF. Both  BABAR and Belle  do the
  same, vary  the BF by  the uncertainty in  PDG. The FF  variation is
  mostly affecting  the endpoint analyses,  and has not been  done for
  the mX-Q2 analyses. No need to do so seen.

  K yield: The inclusive kaon  yield should be varied by analyses that
  employ kaon veto (or kaon ID in a BRECO/D*lnu tag).

  D BF:  The most relevant BF of  D decays should be  varied, again by
  the PDG uncertainty.


4. b->ulnu modeling:

  JETSET: Certainly common error. Some tuning in progress at the
  experiments. 

  ssbar: Belle will determine a  proposal what to vary how much, which
  will be compared to BABAR's proposal.

  BF: Variation of the inclusive and exclusive BF with the measured
  uncertainties to evaluate the effects on efficiency. 

  Hybrid model  in Belle:  pilnu (rholnu) and  dF/N; BABAR:  ISGW2 and
  dF/N. A proposal was made  to include the configuration of the other
  experiments in  the evaluation of  the systematic. Note: dF/N  is de
  Fazio and.


5. Extrapolation:

  Both BABAR and Belle use dF/N 

  Some discussion  on what to  use. Varying lbar  (or mb) by  too much
  touches unphysical ranges (N.U.).  BABAR currently uses lbar = 0.480
  +/- 0.120  GeV and lone =  -0.300 +/- 0.105 GeV^2.   A correlaton of
  -0.8 between lbar and lone (can  CLEO maybe confirm this or give the
  numbers from  their moments analysis?)   should be included  in this
  evaluation.   These values  correspond to  CLEO's  measurement after
  removing terms  proportional to alpha_s^2  and 1/m_B^3 (this  is the
  most consistent to the order of dF/N).

  An alternative  proposal by  N.U. was to  take the  numerical values
  obtained  by  CLEO  and  use   alpha_s  =  0.3  to  approximate  BLM
  corrections.

  We had a  lengthy discussion on whether we  should use delta(lbar) =
  90 MeV or  whether we should use the  measured uncertainty (actually
  dominated by theoretical uncertainties).  The argument for using the
  experimental instead of last year's  delta(mb) = 90 MeV is to absorb
  some amount of  the error due to e.g. the specific  form of he shape
  function  chosen  in this  way.   The  argument  for not  using  the
  measured errors is that this is  arbitrary and it would be better to
  separate the parameter uncertainty from the model uncertainty.

  However it was felt that the  latter is difficult to quantify, so as
  a first step the proposal was accepted to use (only) the total error
  on the measurements by CLEO (i.e. the numbers above).


6. Theory:
   
  CLEO has an error "bsg"  that is estimated by varying the parameters
  of the SF by 10%. The discussion centered on whether that is a large
  enough error  and whether other  analyses should do a  similar thing
  (e.g. to give an error for the uncertainty on the parametrization of
  the shape function).

  There was a vague consensus that we need to understand a bit more
  how to get a "sensible" estimate of this error. We agreed to talk
  about this with our theory friends.

  N.U. proposed to study higher-order effects in the endpoint analyses
  by shifting the spectrum and study the effect on fu. 


7. BF -> Vub:

 A short discussion on double counting the error on mb and lbar was
 summarized that we need to use a self-consistent set of parameters in
 the extrapolation and he extraction of Vub.

 We did  not reach  a decision on  whether we  want to average  the BF
 separately or  "extract" it  from |Vub|. This  would imply  that CLEO
 gives the full BF in their mX-Q2 analysis.

  
8. tau with f+/f0:

 We will circulate a proposal for what B, B0, B+ lifetimes, errors and
 f+/f0 with errors we want to use.


Action Items
------------

 o Belle will estimate their error with (dlbar, dlone) = (0.120, 0.105)

 o  UL will circulate numbers and errors on 
     tau+
     tau0
     tau
     f+/f0

 o We'll have a phone meeting in May (roughly) to discuss this in more
   detail. We  hope to have inputs  from CLEO before than  and to have
   all three experiments participate in that meeting.