Re: [BLASTTALK] Asymmetry

From: Chi Zhang (zhangchi@MIT.EDU)
Date: Thu Jul 29 2004 - 23:38:46 EDT


Hi Genya,

To begin with, I can't use your exact run list. run 9022 has 0 charge, run
9051 is not closed, runs 9041, 9043, 9045 I could not find. you also have
9056-9056. Not that it matter.

I hope the rate is not an issue any more.

So with all the runs left in you list, total charge 24.6kC. I have the
following:
        para perp
Q N+ N- A*100 N+ N- A*100
1.95 2747 1669 -6.9+-1.1 3953 1362 11.4+-0.9
2.29 746 482 -9.1+-2.1 1200 358 15.4+-1.6
                        (-8.7+-3.4) (14.7+-1.2)
2.64 175 163 -22.5+-4.1 389 78 24.8+-2.6
                       (-19.8+-3.4) (23.0+-2.3)
your numbers are in the parenthesis. I have total of 3591 events in the
second and 3rd bins which is 78% of your 4555 (0.53/0.75=71% though).

There is no difference in experimental asymmetries. So it is in the models
and how T20 and T21 are combined into Asymmetries. So let s compare. Take
spin angle 30 degree, T22 aside for the moment:

Q T20 T21 R20 R21 A_theory*100
                        para/perp para/perp para/perp
1.95 -0.601 -0.071 0.355/-0.450 -1.208/0.439 -15/25
2.29 -0.861 -0.127 0.445/-0.477 -1.177/0.296 -26/38
2.64 -1.115 -0.210 0.536/-0.474 -1.126/0.140 -38/50

Note that A = r20*T20 + r21*T21 in both sector, contribution from T22 is
a couple percent. Note also that the experimental asymmetries compare to
A_theory/sqrt(2) because of the Pzz, P2d convention difference.

Pzz's extracted this way are:
Q Pzz*100
        para perp
1.95 65+-6.8 64+-3.3
2.29 50+-5.7 57+-3.4
2.64 84+-13 70+-5.1

There are some almost 2-sigma discrepencies but the worst case
involves the 3rd bin. Funny that the two sectors are consistant at
same Q2, but bin by bin in each sector, the fluctuation exceeds
1-sigma a little bit. When a larger set of data are taken discrepencies
went away. see my email last night.

If you have different T20 and T21, it is you against Dr. Abbott then. I am
in no position to discuss discrepency between Abbott and any other models.
If you have different kinematic coefficients, then we should check.

Now a few comments about your message:
> I agree that the polarization angle appears to be 30 instead of 32...
> But it doesn't affect the measured polarization much...

I don't know what to say, look at the following, with 32 degree, you don't
claim you know Pzz to better than +-0.1, i.e. half the difference between
the two sector.
32 deg: parallel 0.54+-0.05 0.56+-0.08 0.51+-0.08
        perpend 0.73+-0.03 0.75+-0.05 0.72+-0.05
I see you are also using |A_left| + |A_right| which is less sensitive to
spin angle, but you can't claim you have +-0.05 in error, and I think it
is very wrong to use this when discrepencies are not resolved. We have
been doing this since last summer when I did not know how to explain
the discrepancy between left and right. One year passed and we must do
better than that now.

This is like the luminosity case, on one hand you quoted ed rate of
0.75/C which says lumi is <2e31 if you check with ed cross section, on the
other hand you quoted luminosity of 5e31. So I would conclude your were
saying the luminosity is 3.5+-1.5 e31.

> Before presenting my data, I'd like to comment on asymmetry->t20
> conversion.
> t20, A(q) and B(q) present the total set, and since A and B are known
> very well, I
> believe it is a very good approximation that the relative error in
> asymmetry is equal to
> relative error in t20 determination.

This is cute but this is not a real error propagation.

In fact, my projection is(made on July 23rd)
Q dT20 T20 dt20/t20(450kC) dt20/t20(300kC)
1.940 0.011 -0.5845 1.9% 2.3%
2.298 0.016 -0.8425 1.9% 2.3% (2.2%)
2.691 0.027 -1.1130 2.4% 2.9% (2.5%)
With 0.75/C and 0.8 in Pzz, you should do much better than that. I know
where the problem is but you are a smart guy.

So I think discrepency still exists in Pzz. It is further narrowed down to
theory models. The theory for Asymmetry can be break down
into two parts: Tij and the kinematics coefficients. These coefficients
are functions of spin angle relative to 3-momentum transfer. I showed the
theory values of T20, T21 and their kinematic coefficients. I believe my
calculation implemented in DGen faithfully reflect Abbott's
parametrization which is shown on the T20 plots I sent.

I also believe I computed the kinematic coefficients correctly. One can
check self consistancy in this respect by NOT using T21 world data. But
solve T20 and T21 simultaneously from the two asymmetries we measure. To
claim consistancy, the bin used for normalization should be right on the
curve in both T20 and T21. Please see the atatched T20_T21.eps. Top left
is T20, top right is T21.

I could be wrong and I wish I am, so this arguement can be over. But I did
a lot of work to extract T20 and put it together with world data for
comparison. I also did a lot to check my consistancy. It would take more
that just "you have a big error bar" to convince me I am wrong.

Chi





This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:31 EST