Minutes of the NFS-meetings held on 14/6/94, 28/6/96, 12/7/94, 26/7/94:
=======================================================================
In these meetings we had the following reports:
* Bernd Pagels : pi0pi0 analysis: estimation of pionic background
* Olaf Behnke : Significance of A_000
K0b/K0 Normalisation in 2pi0 analysis
* Marion Schaefer : Reconstruction of K0-->3pi0 using a 19C fit
Status of K0-->3pi0 analysis
pi0pi0 analysis: ways to estimate pionic background: (B.Pagels)
===================================================
We studied two different methods to estimate the pionic background in our
pi0pi0 data:
(1) Fitting the de/dx-distribution of the kaon candidat.
From our data the de/dx-distributions of the kaon candidat were produced
in 10 momentum bins and 12 lifetime bins after all analysis cuts except the
de/dx cut. Each of these distributions was fitted with the function:
F(de/dx) = f_kaon(de/dx) * (1-a) + f_pion(de/dx) * a
f_kaon(de/dx) and f_pion(de/dx) denote de/dx reference distributions in a
given momentum bin for kaons and pions (thanks to Christos T.). a is
the fit parameter, which is a measure of the pionic contribution to the
de/dx distribution.
From these 120 fits the lifetime dependent pionic contribution to our data
was obtained. The results are:
(a) We have 0.24 % pionic background in the lifetime interval [-10,20] taus.
The pionic backround is in the important lifetime interval [5,10] taus
less than 2.5%.
(b) The pionic background peaks between 0 and 5 taus.
(c) The systematical error is not known, therefore the second method is
an important crosscheck !!!
(2) Taking events, where the kaon candidat passes the defect (inefficient)
PID22 (run periods p21-p23).
Now the de/dx-distributions of the kaon candidat (after all analysis cuts
exept the de/dx cut) show due to the inefficiency of the cherencov of PID22
a visible pionic shoulder at low de/dx values. Thus pionic events can be
selected with a cut de/dx<1.75.
The result:
(a) The lifetime dependent pionic contribution agrees very good with the
one of the previous method!
(b) 50% of the pionic events sample seems to originates from the
annihilation:
p pbar ---> pi+ pi- pi0 pi0
A special c-fit was used to show this. This channel is responsible for
the peak in the lifetime distribution between 0 and 5 taus.
Significancetests for A000: (o.Behnke)
===========================
A simulation of the A000 asymmetry with 10**12 k0+k0b events
between 0 ts and 20 ts was done under different conditions:
1. perfect K0 Lifetime resolution
2. resolution from 7c-fit (used by Marion, which gave a histogram of the
resolution to me which was used in the simulation)
3. without/with different backgrounds
The K0b/K0 normalisation used was always 1.
The simulations were fitted taking as fit function
exactly the same function which was used for the simulations itself
(of course now with some paramters fitted free!).
The fitted paramters are
Real part of eta000
Imaginary part of eta000
K0b/K0 Normalisation
Backgroundparamters (if there was background simulated)
The G O A L of all this procedure is to find out
with what precision one can measure the real and imaginary
part of eta000.
The results are presented in a big table hopefully you can
find in the folder. Here only a very few should be mentioned:
P R E C I S I O N S from fit
Delta(RE(eta000)) Delta(IM(eta000))
perfect resolution / no background 0.031*10**-4 0.061*10**-4
7c-fit resolution / no background 0.145*10**-4 0.169*10**-4
perfect resolution / no background fixed 0.051*10**-4
7c-fit resolution / no background fixed 0.085*10**-4
The results can be easily scaled to another statistics
with the square root of events.
The results quoted are maybe helpful to estimate
in an optimistic view of background what can be
expected from Marions and Thomas analyses.
Starting to study K0b/K0 Normalisation in 2pi0 analysis: (O.Behnke)
========================================================
A first attempt was done to estimate the shifts
of phi00 due to a possible time dependant K0b/K0 normalisation.
*
For two track events it may be a good
assumption that the K0b/K0 normalisation does
for a fixed kinematics of K+ pi- not depend on the K0-decaytime.
If one had perfect time resolution for the k0-decay it should
be possible then to study all normalisation dependancies
as function of purely K0 kinematical quantities.
That means for a given K0-kinematic one can integrate out
all the kinematic of the charged tracks which can contribute
to this K0-kinematics.
From the detector symmetry one can
conclude that Pk0 and Pk0z/Pk0 are two suitable variables
of the K0-kinematics in terms of normalisation.
*
A time dependant normalisation can now be induced by the combination
of two effects:
1. the normalisation is dependant of the K0 kinematics
2. the acceptance as function of the lifetime is different
for different K0 kinematics
*
From the data and MC it was tried to estimate how big the effects
1. and 2. can be:
The normalisation as function of Pk0 and pk0z/pk0 are flat
within 10% (for short lifetimes [0,4]).
At higher lifetimes there is an enhancement of low momenta
in the data (also MC) due to the different acceptance
and to resolution effects. The resolution of events with low
momenta is worse and there are relatively more events shifted
to higher lifetimes.
*
With a simple model it was calculated what time dependance
in the normalisation effects 1. and 2. can introduce (see plots!!!)
*
A simulation of the asymmetry was done where we introduced
a time dependant normalisation as alpha(t) = alpha(0) + beta*t.
From the model and the data we get beta<0.0005.
If one fits the asymmetry assuming no time dependancy
one gets back a systematic shift in the order of 1 degree.
This is only a beginning, study will be continued!
Reconstruction of K0 --> 3 pi0 using a 19C fit: (M.Schaefer)
===============================================
(MEETING ON 31.5.1994)
In addition to the constraints exploiting momentum/energy
conservation and the constraints on the 3 pi0 masses and the K0 mass
( 7C fit) I have included in the C-fits
the constraints arising from the measured directions
of the photons in the calorimeter, resulting in a 19C fit.
By means of MC data I have analyzed the differences between the old
7C fit and the new 19C fit. When comparing both fits one should keep
in mind that the information from the shower angle measurements is
rather poor, i.e. the RMS for both theta and phi is of the order of
150 to 250 mrad for low energetic photons.
Concerning the vertex resolution one finds only a minor improvement with
the 19C fit, the RMS improves by about 10 % . The most effective cut
against the tails in the vertex resolution is a cut on the K0 momentum,
for example a cut of p(K0) > 400 MeV/c leads to an almost complete
suppression of the tails.
Concerning statistics, one finds that about 35 % of events that have
a 7C fit probability of < 0.1 , are shifted to probabilities > 0.1
by the 19C fit (recall that the prob. distributions of these fits are
not flat mainly due to the poor energy resolution of the calorimeter).
How does the background behave under the 19C fit?
1.) Background from K0s --> 2pi0 + 2 fake photons:
This channel was studied with the 2pi0 MC. Without any background
rejection cuts (like the cut on the "6 gamma pseudomass", which is
the 6 gamma inv. mass calculated at t/taus=0) , the 19C fit accepts
slightly more events than the 7C fit. Quantitative estimates are
difficult due to the limited statistics of the MC data available.
2.) Background from K0s pi0 :
Here the 19c fit does not seem to have a larger acceptance than the
7c fit, but again we need higher statistics to draw quantitative
conclusions.
What remains to be determined is the behaviour of the pionic background
under the 19c fit.
Summary on the K0 to 3 pi0 analysis: (M.Schaefer)
===================================
(presented in the last two meetings)
In the K0 --> 3 pi0 analysis the dominant sources of background arise
from
1.) the reaction p pbar --> pi+ pi- 3pi0 , where one of the charged
pions is identified as a kaon
2.) the golden channel K0S --> 2 pi0 plus 2 fake photons in the
calorimeter
3.) golden plus pi0 at the primary vertex.
While background sources 2. and 3. have been studied with MC data (how-
ever with poor statistics), the pionic backgr. has to be determined from
the data itself.
For this purpose I have analyzed 6 gamma data from period 20 to 23 with
the inefficient Cerenkov counters. Requiring that the charged kaon
candidate has traversed PID no. 18, 19 or 22 and cutting on the sigma
(dE/dx) of the kaon to be less than -2.0 , one has an almost pure sample
of pionic events. The lifetime distribution of this background is then
obtained by passing the events through the full 3 pi0 analysis
procedure including all cuts except the ones on dE/dx.
In order to get an absolute estimate of the total background present in
the 6 gamma data, I have performed a lifetime fit: the lifetime distr.
of the 3 backgr. contributions and the one of the K0L --> 3 pi0 signal
(from MC) are fitted to the data leaving all normalisations as free
parameters. For P25 data, the fit gives about 4500 signal events, which
is in nice agreement with the expectations calculated from the # of
K0S --> 2pi0 events using the known br. ratios and efficiencies.
In the decay time region between 0 and 10 taus, the pionic background
turns out to be 3 times larger than the sum of fake and golden + pi0 bgr
The shape of the total background can be parametrized by an exponential
(with slope -0.2) plus a constant.
Concerning the pseudomass, I used at cut < 600 MeV/c2 which has a flat
acceptance for the signal and is 95 % efficient. While this cut rejects
fake and golden + pi0 backgr. mainly in the negative lifetime region,
it cuts roughly 50% of pionic events for t>0.
Using the analysis presented so far, I calculated the expected error on
Im(eta_000): assuming 10.000 K0L --> 3 pi0 events reconstructed and
taking the background distribution found by the lifetime fit above, a
one parameter fit for the asymmetry A_000(t) results in a
sigma(Im(eta_000)) of 0.11. For the same number of events but without
background, the sigma would go down to 0.067.
Next meeting:
=============
***********************************************************************
* We agreed to have the next NFS-meeting on Tuesday 09/8/94 at 10.30 h *
* in the CPLEAR meeting room. ======= *
***********************************************************************