E852 experiment
Analysis of Eta Pi0
system with the decay Eta -> Pi+ Pi- Pi0
********************************************
Comments of
Gary Adams
Date: Thu, 04 Aug 2005 12:06:10 -0400 From:
Gary Adams Dennis, Neal and Suh-Urk:
I looked over Vladimir's outline and I think it is in good
shape to forward to the collaboration, with our endorsement. Please send me your
opinions immediately. When I make the announcement, that
will initiate the collaboration review of the outline. I think the paper will
need lots of details so I recommend PRD rather than a letter. I found only one
point that seems odd to me. In the MDPWA it says that the small waves were put
in as "simple functions of mass". I assume this was done because they
do not look resonant in the PWA fits. I need to get further clarification on that.Regards, Gary
Answer
We use the different fits with
various forms of UNP waves in MDPWA. The results are not sensibilities. Of
cause UNPW do not look resonant except D0, but any form of UNP amplitudes doesn’t
change the parameters BW of P+ and D+ waves. The reasons are the large errors
of UNPW with our bad statistics and UNPW intensities are rather weak in the region of 1.24-1.34 GeV (see Fig.4, Part 2)
********************************************
Date: Mon, 08 Aug 2005 13:39:40 -0400 From: Gary Adams
trc:
I'm not too worried about the mass; we had a pretty big error bar on our
old result. The crystal barrel and obelix masses are
a bit harder to reconcile with the new one. I do think that we will have to
decide on the relative merits of the different resonance fits. If we can
convince ourselves that the MDPWA is the better method in this case, then that
would result in smaller systematic errors. Clearly we will have to make
detailed comparisons with the IU results too. In the end, the only significant
statement in the IU paper is that the t dependence is not consistent with a BW.
If we can't address that question, I don't think there will be much
controversy, but we should try.
Answer
Our fit (we
mean the fit in two t-intervals at bin size 100 MeV
see table-4 and figs 11-12) was done including m<1.1 GeV
range. We found that in this range the contribution of the leakage from 0+ wave in P+ wave is essential. The consistent results of the fit
procedure in |t| < 0.225 GeV/c^2 (fig 11, part-2, table 4), 0.225 < |t| <1.0
(fig 12, part 2, table 4) and whole t-range ( part-2,table 4) can be obtained
only if the leakage from S-wave is included in the calculations. It isn't the
case of IU paper.
In this context, I would like to better understand the motivation for
each wave shape in the MDPWA. Does it just come from the mass dependence in the
fitted PWA amps? I also think that needs to be part of the documentation for
the analysis. Vladimir, have you done the MDPWA for different t bins?
Answer
It's hard and long work. It need to use mine own codes with many changes of parameters.
*******************************************
Date: Mon, 03 Oct 2005 18:14:21 -0400
From: Gary Adams
Vladimir:
I think I am starting to understand your draft a bit better after reading
some of the answers to early trc questions. I still have some questions
and comments on draft 1:
1) p.2
The IU paper should not be referred to as E852since we were not authors.
Agree. It must be a cite and our comments and detailed criticism.
2)p.2, c2, para 5,
The Monte Carlo description implies that we used phase space to correct,
which is not true since each wave is done separately.
The rigth phase space was included in our Monte Carlo calculation and in our PWA and MDPWA. We used standart E852 method (see S.U. Chung , Phys.Rev.D60,
092001(1999)) and we don't anything new. Please, clarify your question.
3)p.3, c1, para 1
The reaction should read eta pi0.
Accepted
4)p.3, c1, para 3,
You say "the unnatural parity waves are poorly determined...."
This is a weak point and it needs to be clarified in the paper.
"the unnatural parity waves are poorly determined" means that we have
small statistics. UNPW are rather small and have so large errors which don't allow us to make MDF,
except D0 wave at mass > 1.1 GeV (see Fig.22, Note part 1, website eta pi0 (MS) paper).
In MDPWA we fit D0 wave with the same resonant parameters as D+ wave simultaneously.
The result of MDPWA is the next: the UNPW affect slightly on parameters of a_2 and
pi_1 resonances.
Fig 4
shows a big P- intensity and I assume that it is going into the
"background" in your MDPWA.
A combination of three waves P0, P- and P+ may give the isotropic background (see, ~e852/secure/analysis/1995/korotkih/index.html , Note 2 "MDPWA of EtaPi0" ) if they have the same mass dependence. This note use preliminary data analysis at 2000 year and we may use them only as the methodical results.
Did you try a resonance in P-? If it is
consistent with a resonance that should probably be shown instead. In
either case we need to understand it better.
Note, that statistic of P- wave as well as P0, D- waves is very bad.
If you look in Fig 22, Note part 1, website eta pi0 (MS) then you can see
that P- wave is not consistent with a resonance in large mass interval 0.8-1.74 GeV.
5)p.3, c2, para 5
I don't understand what this section means: "It is not the same...."
The text is needed to correct. We mean the leakage to the P+ wave from D+ wave in the region mass 1.3 GeV is small,
but the leakage from S0 wave in the region 1.0 GeV is large .
6) p.4, col 1, para 2
I understand from the web that you included a wider mass range for the 100
MeV binned data, but the reason for that is not explained in the
paper. Also, I don't understand why the t-bin test was done with the PWA
instead of the MDPWA. Wouldn't the MDPWA do a better job with the lower
statistics?
100 MeV binned data was used to increase statistic when we divide t ' region on two bin interval. Then we make PWA in each mass ant t' bins. We suppose that it is the most reliable way of analysis. MDPWA is not good for small waves, it is sensitive to the waves with large intensities.Both methods MDPWA and PWA+MDF have own advantages and defects. We don't think the MDPWA do a better job with the lower statistics. MDPWA is better because it solves the ambiguous solution problem and gives the leakage contribution.
You say that the two resonance analysis agree (table 1), but then cast
doubt on one of them (table 2). The pi1 width for the full t range is
around 360 MeV in table 1, but it is 585 in table 2. They have the same
data in them so that tells me that there is a big systematic error
associated with Mass binning. How can that be? Did you average your data
in each bin correctly? Also, what should I assume for the systematic
errors in table 2? Again, it seems to me that the t-bin fits would be
better if we used the MDPWA. Did you try that?
The result in table 2 are with 100 MeV binned data. It is clear that
the spreading of any peak over large mass interval will increase a peak width.
So we have not to compare the widths in table 1 and table 2. The table 2 shows
that our result of resonant mass are stable. The systematic errors in table 2 are from the random selections of ambiguous solution combinations in MDF
We can't do MDPWA in two t' bins because it is hard work and time consuming.
The relationship between the analysis methods is a big issue and I don't
think the motivation or details come through in the paper. The paper jumps
around between analysis methods and I don't think the reader can tell what
we are reporting. Also, I still can't tell what method was used for
putting "simple dependence of mass..." into the MDPWA. Due to the low
statistics I would like to use the MDPWA as our result but right now I
don't understand the MDPWA method well enough. You need to put some more
backup material on the web.
You may see my old notes (see, ~e852/secure/analysis/1995/korotkih/index.html , Note 2 "MDPWA of EtaPi0" ) and our paper S.U. Chung , Phys.Rev.D60,
092001(1999).
7) fig 2 shows wiggles in the acceptance functions. I don't think those
are real. Did you fit some data with a polynomial that had too many
coefficients?
We don't understand question. May be you look original histograms of acceptance without smoothing in Fig14 and Fig.15, Note part 1, website eta pi0 (MS) paper.
8) fig. 4 shows a BG curve under the wave sum. That should be explained in
the text. Is it an isotropic wave in the pwa?
Agree. The background in our analysis is isotropic and fixed.
9) I like the comparison between pwa methods in fig 4 because it
distinguishes between the ambiguous pwa fits. However the P- wave is large
and I can't tell where it comes from. Is the low mass rise coming from
leakage? Also, see the above comment that a resonance might be needed in
that wave. In fig 5 you change to a comparison with the average pwa result
(not explained). I think figs 4 and 5 should use one method for
comparison, probably using all the ambiguous fits.
About P- wave see above. Agree, text must be cleared. We must explain the transition from comparison with ambiguous solutions in fig.4 to average pwa
results in fig.5. But it is not right to show ambiguous solutions in fig.5. The results in fig.5 are obtained with pwa average ambiguous solutions.
10) The phase plots should all use a single phase difference, for example
P+ minus D+.
Agree
11) You need to compare our result with all available published work on eta
pi0, including the IU paper.
Agree, but what do you mean concretely.
12) Fig 6b has too many curves. I suggest keeping the sum and the leakage
and leaving out the resonance.
Don't agree.
In summary, I think we are trying to put too much into one PRL
paper. However I still don't know yet how to condense it around a single
result.
Regards,
Gary
Let's try to condense our result for paper in PRL or Phys.Lett.
Neal promis to work in this direction. We need explain
the difference between Crystall Ball, IU and MS for EtaPi0 in principle.
Not only values of parameters, but also the approaches and interpretation.
If we explain main situation in short paper, then we'll write long paper.
Again we remind that our MSU result and IU on mass EtaPi0 in the whole region of t' are consistent, mass=1270 +/- 14 MeV, (MSU):
mass=1272 +/- 17 MeV (IU, PRD 67, 2003 ). So a main question is a difference with
Crystall Ball: mass=1360 +/- 25 MeV (PL B446,349, 1999 ).
Vladimir K., Ludmila M.
----------------------------------------------
Date: Tue, 25 Oct 2005 11:30:31 -0400
From: Gary Adams
Vladimir, et al:
I have not read draft two yet, so this is a comment on your responses to my
earlier questions.
><4)p.3, c1, para 3, You say "the unnatural parity waves are poorly
determined...." This is
a weak point and it needs to be clarified in the paper.
<3 Oct>"the unnatural parity waves are poorly determined" means that we have
small statistics.
UNPW are rather small and have so large errors which don't allow us to
make MDF,
except D0 wave at mass > 1.1 GeV (see Fig.22, Note part 1, website eta pi0
(MS) paper).
In MDPWA we fit D0 wave with the same resonant parameters as D+ wave
simultaneously. The result of MDPWA is the next: the UNPW affect slightly
on
parameters of a_2 and pi_1 resonances.
I do not agree and I think this is a problem. The MDPWA is not limited by
statistics.
There are ambiguous solutions that span a wide range of values
for the negative reflectivity waves. The MDPWA picks one set, as you have
explained, but if we are going to publish it we need to be able to say what
is needed to make the MDPWA work.
In Note 3 we explain that MDPWA doesnt work with the ambiguous solutions.
The MDPWA doesnt pick any set of ambiguous solutions. A section C2 in our paper S.U. Chung , Phys.Rev.D60, 092001(1999) is devoted to MDPWA and this method was described.
We wrote in Note 3:
Here we present the results of fit which is a mass dependent partial wave analysis (MDPWA).
MDPWA is carried out not separately for each $\eta\pi^0$ mass bin, but all
bins are fitted simultaneously. They are tied together with a mass-dependent
function for each partial wave. A likelikhood function depends on not only the angles, but also on mass of $\eta\pi^0$ system.
The advantages of a MDPWA are the next. The results of such analysis don't depend on the ambiguous solutions.
So it is not necessary to take an average ambiguous solutions or select between them.
We need to show what a "resonance only"
fit of all waves looks like, and then what happens if we add the leakage,
and then what happens if we add the polynomial (?) background. All of
those need to be documented. I still can't tell how you picked the shapes
of the mass dependence.
We wrote in Note 3 what form of mass dependences are used, see formulae (3) - (9). We can try to use other form UNP wave, but it is useless. MDPWA is sensitive to form of wave angular dependence and sligthly to selected mass dependence.
What I see on the web is not enough to give me
confidence in the MDPWA. This justification needs to go in the paper, so I
think PRD is the only option if we publish the MDPWA.
All previous papers: EtaPi-, RhoPi, EtaPrimPi- and other were published in PRL or PL before PRD. We supoose that our result is important and the short publication
is necessary for physical community. Observation of neutral state 1-+ is a clear evidence of Exotic.
No, that is not correct. You have not demonstrated that P- is not
consistent with a resonance. As noted above, there are lots of ambiguous
solutions that the fit can choose from. My question is, where does the fit
you have shown put the P- intensity?
The wide range values of ambiguos solutons don' t allow us to solve what form of mass dependence to use for P-. Other publications, for example Sadovsky with large statistic, show that there is no any resonance in P-. Our forms of mass dependence are used to describe background in these waves.
<3 Oct>100 MeV binned data was used to increase statistic when we divide t '
region on two bin
interval. Then we make PWA in each mass ant t' bins. We suppose that it is
the most
reliable way of analysis. MDPWA is not good for small waves, it is
sensitive to the waves
with large intensities.Both methods MDPWA and PWA+MDF have own advantages
and defects. We don't think the MDPWA do a better job with the lower
statistics.
MDPWA is better because it solves the ambiguous solution problem and gives
the
leakage contribution.
Now I am really confused. If the MDPWA is not reliable then why is it in
the paper? If we have confidence that it is a good method (meaning small
systematic errors) then it is clearly the best way to study the t
dependence because it solves the problem of ambiguous solutions.
Both method (PWA+MDF) and MDPWA are the complement of one another.
We make a mistake if to use only one. It is our reliability of our results.
Yes, we agree that it is necessary to study the t
dependence parameters in MDPWA. Of course, it is necessary to calculate the systmatic errors of MDPWA.
<3 Oct>The result in table 2 are with 100 MeV binned data. It is clear that the
spreading of any
peak over large mass interval will increase a peak width. So we have not
to compare the
widths in table 1 and table 2. The table 2 shows that our result of
resonant mass are
stable. The systematic errors in table 2 are from the random selections of
ambiguous
>solution combinations in MDF.
OK I see two problems in this answer. First, it sounds like you have not
properly averaged your resonance fits over the bin width of each
point. New binning should not change the answer if you have at least a few
channels in the fit. Second, you have not answered my question.
Where does the discrepancy come from? If it is coming from the ambiguous
solutions, then we probably should not publish the t binned fits because
they have large systematic errors. I am very sympathetic to your problem
of not being able to do more work (We all have the same problem.) but our
job at this point is to decide what should be published, and what should be
improved or dropped.
<3 Oct>You may see my old notes (see,
>~e852/secure/analysis/1995/korotkih/index.html , Note 2
>"MDPWA of EtaPi0" ) and our paper S.U. Chung , Phys.Rev.D60, 092001(1999).
OUCH! That old note shows that most of the pi1 signal is from
leakage. Your new fit shows very small leakage. How does this help? I
am more confused than before.
We don't agree with such conclusion.In this work a ratio Leakge to a2
from MDPWA is very small (R=0.018). The results of (PWA+MDF) and MDPWA are consisten (See TABLE III there)
It looks like we are making progress but as you see I still have some big
reservations. I will read the new draft and see if I have any new questions.
Thanks,
------------------------------------------------
Date: Tue, 25 Oct 2005 11:56:36 -0400
From: Gary Adams
Vladimir:
I found only one new issue in draft 2. In your web description of the
MDPWA you have a background contribution from the eta sidebands. That
should be described in the paper. I also had the impression that some
background curves were fitted, but upon closer study I do not see that in
the latest analysis. Am I correct that the background only comes from the
sidebands? In fact, in the paper I can't find a description of how the
side-band data was used in either analysis.
Gary
Agree. We change the text. The background is calculated as Bckg(M)=Data(M)*(Sidebands(M)/Signal(M)) in both
method and fixed.
Vladimir K., Ludmila M.
*******************************************
*******************************************