Minutes of the NFS-meetings held on 19/11 and 23/11/1993:
=========================================================
Reports from the following people have not yet been summarized:
----------------------------------------------------------------
- Bernd Pagels
- Marion Schaefer
Reports from the following people are summarized below:
-------------------------------------------------------
- Olaf Behnke
- Andreas Schopper
- Marcin Wolter
Status-report about tuning the errors of the showerangles: (O.B)
================================================================
I selected with Guidos program
K+K-pi0 events. The photons of these pi0's
are well suited to make tests with their
calo-measured showerangles (named pansho, tansho):
- You know the true angles, because the photondirections
must go from the annihilationvertex to the impactpoints.
- You know the true energy from 3c-fit
- expect low contaminations from fake photons
- as the photons are coming near from the center
of the detector you have all the range of different true theta
but (this is a drawback) practically only zero true phi.
With a datasample of roughly 10000 photons, for the
first time the resolution of the measured showerangles
in the Calorimeter was studied. Making a cut abs(theta true) <10 grad
a comparison of the resolutions with the results of Christoph Felder
for the prototype ( with phi_true = theta_true = 0.) could be done.
It was found very well agreement, the only deviations were for
lowenergetic photons (egamma-true <100 MeV), where the calo-data showed
a worse resolution.
To demonstrate the necessity of tuning the errors of the
showerangles i showed plots of the quantities
(pansho - phi_true)/dpasho (*)
and
(tansho - theta_true)/dtasho (**)
for different measured photonenergies. Dpasho and dtasho
are the showerangles-fit's variances for pansho and tansho.
So if everything is correct (*) and (**) should be gaussians
with sigma = 1.
What you observe is instead:
- gaussian fits of (*) -> sigma significant bigger than 1
(**) -> sigma significant smaller than 1
- significant photonenergy-dependance of the sigmas:
bigger energies -> bigger sigmas
Conclusion: the origin of the effects on (*) and (**)
is poorly understood right now, but anyway you have to
corect (=rescale) the dpasho and dtasho,
so that you get proper (*) and (**).
To obtain the best corrections, this means especially
to find out now which photon-variables are the best suited ones
(like energy, number of hits, showerlength).
So this is in work now.
About the use of the showerangles for pi0pi0 reconstruction: (O.B)
==================================================================
The 14c-fit should be the optimal way
to involve also the showerangles in our
pi0pi0 reconstruction. If this is not
yet available one can think of other
simpler methods for using them and test
first, if the showerangles are good for
anything or not.
Idea:
-----
make a simple constraint-fit
of the showerdirections, that they
must all meet in one point on the
K0-flight axis (where I take the
direction from 1c-fit).
Implementation:
---------------
You scan testpoints P_i (=: point i) along the 1c-fit
K0-direction axis (take of course the axis, which goes through the
annihilationvertex ). For each of these
testpoints you draw the four lines to
the photon-impactpoints and calculate the
respective angles phi_p_i and theta_p_i
in the Calorimeter. Then you can calculate
a chi-square:
chis = sum over 4 gammas [ ((phi_p_i-pansho)/dpasho)**2 +
((theta_p_i-tansho)/dtasho)**2 ]
with pansho, tansho, dpasho and dpasho are the
measured Calo-showerangles and their variances from fit.
The testpoint p_i with the smallest chis (call it chi_min)
you take as your decayvertex reconstructed by the showerangles.
As you expect a resolution of several centimeters it's
sufficient to make this scan in steps of one cm. I also
requested the points to lie inside the detectorvolume.
From the minimum point you can make another scan around
it and find the points where the chis is chis_min + 1.
This gives you a variance (call it var_lamda_sho) for
this decayvertex.
Using it:
---------
You can compare then
the so found K0-decayvertex with the one
from 6c-fit. Of course you expect a much
better resolution of the 6c-fit vertex, so
you cannot expect a improvement of the
resolution. But still you can make consistency
checks. For instance you can look what is the value
of the chis you would find at the 6c-fit vertex
compared to the chi_min. For to big differences
you can suspect the event to be background and
throw it away.
Testing on MC-Data:
-------------------
All this method described above was applied first
on MC-2pi0 data. The central resolution of
the showerangles-decayvertex is almost a factor 10
worse than the one from 6c-fit. (oh s...)
By making an asymmetric cut (lamda_6c - lamda_sho) < 1*var_lamda_sho
i showed that you can get rid of some parts of:
- events shifted due to the tails in 6c-fit resolution
from low-lifetimes to the asymmetry-region
- events where the 6c-fit found a wrong gamma-gammapairing
and which gives a rather flat lifetimedistribution
conclusion:
-----------
The showerangles maybe are not the sharpest sword, but
they could be of importance.
After finishing calibration on the error of the showerangles
this can then be more seriously tested on real data
than two month before.
NFS-data production of 1992: (A.S.)
===================================
We received at CERN all the NFSFIT data from the outside institutes,
except from Rutherford Lab.
The NFSFIT production includes the Nakada-method, 1C-, 5C-, and 6C-fit
on the 2 track events with 4 detected photons with good Z-info.
The data of CERN, PSI, Basel and Lyon represent approx. 90% of the
total 1992 data.
The NFSFIT data were reprocessed on the new alpha-300 to apply the
LOOP method as well as probability cuts (1C>0.01, 6C>0.01) and
the Q-cut (Q<0.6). The LOOP method was only applied on data with
a K0 momentum bigger than 400 Mev/c.
The reprocessing took approx 6 days on the alpha and the data were
stored on 2 Exabytes.
The total number of events is summarized in the following table:
NFSFIT data 5'380'605 events
NFSLOOP data 1'414'935 events
after 1C and 6C
with 0.1 prob cut 841'221 events
---
The data as shown in the lifetime distribution for the status report
(after various analysis cuts) consisted of 336'500 events.
Status of the Monte Carlo production: (M.W.)
============================================
PI0PI0 channel - CSF farm
8 milions events generated which corresponds to 70000 events after
standard cuts (Prob 1C fit>0.1, Prob 6C fit>0.1, Qcut<0.6)
3PI0 background - ETH ALPHA cluster
3 milions events generated which corresponds to 7500 event reconstructed
as PI0PI0 events after standard cuts.
Dalitz decays - ETH ALPHA cluster
PI0PI0 800000 events generated - 21000 minraw events
3PI0 400000 events generated - 23000 minraw events