[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: comment on simulation section




Hi Mihajlo,

I am a little confused as to why the vertex does not affect the
reconstructed masses. Is it not used to determine the photon angles? You
need the vertex and the impact position to determine the direction of the
photons.

Thanks, Elton.




Elton Smith
Jefferson Lab MS 12H5
12000 Jefferson Ave
Suite # 16
Newport News, VA 23606
elton@jlab.org
(757) 269-7625
(757) 269-6331 fax

On Fri, 8 Feb 2008, Kornicer,  Mihajlo wrote:

>
> Hi all,
>
> regarding the issue of vertex error:
>
> >> page 11, bottom line, vertex errors
> >> Assuming the vertex position error is the 'size of the target  (30cm)' for
> >> reconstruction seems to be very pessimistic. Does this uncertainty  in the
> >> vertex position influence all the invariant mass plots? Should it be
> >> updated (even artificially) by assuming a charged particle z-position
> >> resolutions?
> >
> > What matters for the sake of the invariant mass plots is whether the
> > true vertices are spread in z throughout the target.  Maybe Mihajlo
> > can provide a quantitative answer to how much, if any, broadening of
> > the peak, gets induced by the spread in actual vertex position.
> >
> > To clarify what we did:  In reconstruction all vertices are assumed
> > to  originate from the center of the target (of course this will be
> > refined once an event vertex can be better determined from charged
> > particles).  The uncertainty assumptions are needed to construct the
> > error matrix which is important for kinematically fitting the the
> > event -- here we went with what was the actual uncertainty currently
> > in the simulation, i.e. the physical size of the target.
>
> As Matt pointed out, uncertainty in vertex position does not influence
> the mass peak position or width. The vertex is also not used in fitting,
> if I understand the mass-constrain two-gamma fitter correctly.
> Thus the size of the vertex-error only affects overall scale of error matrix
> and the distribution of confidence levels. If 30 cm looks to big,
> it is safe to use the expected vertex-resolution instead
> (few centimeters I guess).
>
> If one moves the vertex, say upstream by 10 cm, without recalibrating
> photon momenta the reconstructed pi0 (eta) mass and width change
> insignificantly in the forward and by <10% in the barrel.
>
> Cheers,
> Mihajo
>
>
>
>
> >
> > Hi Elton,
> >
> > Thanks much for the comments.  I have also earlier this evening
> > talked  with Elke on the phone about her comments and we agreed on
> > places  where clarity can be improved.  In making revisions, I will
> > try to  incorporate your other comments on the text.
> >
> > I want to quickly try to address a few of the substantial ones:
> >
> >> page 11, bottom line, vertex errors
> >> Assuming the vertex position error is the 'size of the target  (30cm)' for
> >> reconstruction seems to be very pessimistic. Does this uncertainty  in the
> >> vertex position influence all the invariant mass plots? Should it be
> >> updated (even artificially) by assuming a charged particle z-position
> >> resolutions?
> >
> > What matters for the sake of the invariant mass plots is whether the
> > true vertices are spread in z throughout the target.  Maybe Mihajlo
> > can provide a quantitative answer to how much, if any, broadening of
> > the peak, gets induced by the spread in actual vertex position.
> >
> > To clarify what we did:  In reconstruction all vertices are assumed
> > to  originate from the center of the target (of course this will be
> > refined once an event vertex can be better determined from charged
> > particles).  The uncertainty assumptions are needed to construct the
> > error matrix which is important for kinematically fitting the the
> > event -- here we went with what was the actual uncertainty currently
> > in the simulation, i.e. the physical size of the target.
> >
> >> page 15, figure 11, question
> >> Do we understand why the resolutions in the FCAL are much better  than the
> >> resolutions in the BCAL? In the introduction Alex plots pi0 and eta  widths
> >> and there seems to be little difference between FCAL and BCAL. What is
> >> different? If we know this we should add a note to the text. (Could  it be
> >> the vertex resolution? see above).
> >
> > This is a fascinating question that we picked up on late last week
> > and  Mihajlo has worked on to sort out.  We think this is actually
> > just due  to the kinematic "bias" for the types of events FCAL and
> > BCAL accept.   While both have comparable energy resolution, and Alex
> > has shown this  is the key factor in determining pi^0 and eta width,
> > the BCAL is  biased towards catching softer more forward particles.
> > In that front  part of the BCAL leakage is an issue and the energy
> > resolution in  probably worse.  The FCAL typically gets nice high
> > energy photons that  can be well measured.
> >
> > The term that drives the mass resolution goes like:  A*sqrt( 1/E1 +
> > 1/ E2 ) where E1 and E2 are photon energies and A is the statistical
> > term  in energy resolution (the number that is roughly the same for
> > BCAL and  FCAL).  For this eta pi^0 channel Mihajlo made a plot of
> > this term in  the BCAL and FCAL for etas and pi0s:
> >
> > http://dustbunny.physics.indiana.edu/~mikornic/GlueX/CalRew08/sigmaM_Afactor_pi0.eps
> > http://dustbunny.physics.indiana.edu/~mikornic/GlueX/CalRew08/sigmaM_Afactor_eta.eps
> >
> > In both cases the BCAL with comparable energy resolution to FCAL is
> > biased towards more poorly measured pi^0's and eta^'s.  So, I think
> > the interesting bottom line is that one doesn't get comparable eta
> > and  pi^0 widths in the BCAL and FCAL by simply matching their
> > average  energy resolutions.  One has to consider several other
> > things like:   position dependence of energy resolution in BCAL and
> > the fact that  BCAL is biased towards softer photons.  When these are
> > considers one  would like a much better energy resolution in the BCAL
> > to match  reconstructed pi^0 width in the FCAL.  Of course this
> > statement  probably also depends somewhat on the physics channel.
> >
> > Mihajlo's results are rather fresh and we should compare also with
> > Alex's parametric studies to try to understand  exactly what
> > differences are (after all Alex is likely simulating real event
> > kinematics also).  We will talk tomorrow and try to address this
> > somewhat softly in the text.  It is clear we will want to have a
> > prepared answer since this is kind of an obvious question.
> >
> >> page 23, figure 20 (right)
> >> It is hard to see the data symbols. If I read them correctly there  seems
> >> to be substantial false P-wave found for masses around 0.9 GeV  (about 20%)
> >> which seems to be quite large. Symbols in different colors would  help see
> >> what is plotted, but there should be a comment on the amount of  false
> >> resonant P-wave.
> >
> > Noted, this plot was meant to be "proof of principle" that we
> > actually  have amplitude fitting code in such a shape to begin to
> > look at these  things.  I agree there appears to be some P-wave
> > leakage and all of  this needs further study.  I was hoping to just
> > demonstrate that our  technology (both in generating and fitting
> > amplitudes) has matured  enough to be at a stage to begin to do these
> > types of tests.  I can't  say anything more substantive about P-wave
> > leakage other than really  what is in the caption and text: that this
> > is the type of thing we are  going to be looking at.  If you feel
> > this plot is a distraction, I'm  happy to remove it.
> >
> > Cheers,
> >
> > -Matt
> >
> >
> >
>
>
>
>