A road map to the background to D0>pi-K+


 *Prior to dE/dx*....

 1)k-pi+, pi+pi-, and K+K- vetoes are not so important in the
   signal region, *but*, they are important to understanding
   the sideband regions of the m_d plot.  If *only* a cut on
   momentum ratio is made, then the m_d sidebands are filled with
   k-pi+ that is double misid'd, and so those sidebands are not
   so helpful.

   The mass vetos, particularly the pi+pi- and K+K-, also shape
   the sideband regions, scalloping them outboard of m_d+60 MeV (pipi),
   and m_d-60 MeV (KK).  60 MeV=0.5*(m_k^2-m_pi^2)/m_d.

 2)*prior* to dE/dx... it appears that the next most important backgrounds
   are: real D*+, with the D0 going to a PV mode, and the V going to
   a pi0 and a charged: examples are: D0>K- rho+, rho+>pi+ pi0;
   D0>K*- pi+, K*->K- pi0  (when the pi0 is slow, these form the
   classic `satellite' peak of D0 decay), and then, D0>pi- rho+ and
   D0>rho- pi+.

   These backgrounds enter by: loss of the pi0, and then *promotion*
   in mass of the remnant by misidentification of the somewhat slow
   pi to a K; for the Kpipi0, a double misid is needed, and there
   is a mass gain only if the demoted K is stiffer than the promoted pi.

   The K- producing modes are high branching ratio, but are more
   separated in kinematics, and more susceptible to dE/dx, than
   the rho pi modes.

   Trksim MC has indicated that both PV modes with slow pi0 fall
   near the signal region; the rate is hard to estimate, because
   the modelling of dE/dx has to be accurate to nail the rate.

   Trksim also indicates that these backgrounds make a broad hump
   in delta-M.  My toy simulator has not confirmed this result.

   dE/dx is a prominent line of defense from these backgrounds.
   The dE/dx calibration in the data has been greatly improved, 
   but our MC modeling has not kept up.  In principle, the MC
   modeling is now much easier, because of the `beta-gamma
   Universal' calibration technique employed in the new data
   calibration.  However, someone undertake the MC modeling effort.

   An important cross check on the dE/dx comes from: keeping all
   *kinematic* particle assignments of pi-K+, and then studying the
   shapes of background in MD and delta-M for dE/dx cuts that
   favor K-pi+, pi-pi+, and K-K+.  In principle, the MC and
   a good understanding of dE/dx id/misid effiency, crosschecked
   with these plots, allows total control of the PV background.
   
   Reconstructing the pi0 is another prominent line of defense.

 3)*After* dE/dx cuts, there is a prominent background
   to D0>pi-K+ that consists of a true D0 and a random slow pi-.
   We had not clearly understood that this background was indeed
   present, earlier, but now that we are aware of it, it clarifies
   two observations:

         a)That the optimization of dE/dx cuts, in which we
           used the `delta-M' sidebands, indicated that we
           only needed `mild' dE/dx cuts.  Since the delta-M
           sidebands were filled with real D0's with the same
           particles as the signal (indeed, we had noticed
           this!), it was not possible to separate them with
           dE/dx from the signal!

         b)That the *level* of the delta-M sideband background
           seemed to be *universal* between Argus, CLEO 1.5,
           CLEO 2, and CLEO 2/SVX was consistently about
           0.2%/MeV/(Right Sign).  Since all 4 experiments operate
           near root s of 10 GeV in e+e-, then all should have
           the same probability for a D0 to pick up a stray pi-.

           Earlier, it had been confusing that the *level* of
           CLEO 2/SVX's background did not scale with the
           resolution in M_D: sigma_M_d=6.5 MeV in CLEO 2/SVX,
           sigma_M_D=12 MeV in CLEO 2; but the delta-M sideband
           background levels in the two experiments were almost the same.

   In the end, this background must be subtracted, but we win about
   a factor of 3 in its suppression, over CLEO 2, because our delta-M
   resolution (250 keV) is about 1/3 of CLEO 2's (770 keV).


 4)Probably the highest priorities should be:
    a)getting good dE/dx simulation everywhere.  Ideally, the
      package would be callable by both TRKSIM and CLEOG; the
      running with hit number needs to be attended to in CLEOG;
      this has to be glossable in TRKSIM.  This is a long-term
      project.

    b)Getting good simulation of the D0>PV.  The D0>Kpipi0 has
      a good generator; we need to adapt it for pi+pi-pi0.
      However, the way to approach our uncertainty on how much
      rho-pi+ and rho+pi- is to generate *pure* samples of those,
      but *one the Dalitz plot*, not using CLEO's silly `quasi-2-body'
      approach.  The point is: the edge of phase space shapes those
      modes, and we have to get that right.  This is a long-term
      project.

    c)reconstructing the pi0 for the PV.  This is a quicker return
      item.

    d)Doing some studies that tell us the level of the D0 + random
      slow pi- background; and doing some estimates of how much of
      our current peak is from this.