This is the webpage for the GPC review of the paper
Single Spin Asymmetry AN in Polarized Proton-Proton Elastic Collisions at sqrt(s) = 200 GeV
Version 1 --- maintained by Kin Yip
Originally from the complete version (http://drupal.star.bnl.gov/STAR/blog/yipkin/2012/apr/26/paper-q-a) which includes implementation of a lot of obvious text corrections and here only those related to "Physics" are included :
Flemming Videbaek :
A few persons/groups have made suggestions to this sentence. Now it's changed to be:
"The Roman Pot stations are located on either side of the STAR interaction point (IP) at 55.5 m and 58.5 m with horizontal and vertical insertions of detectors respectively."
Andrzej Sandacz:
> B) line 211 and Eqs (7), (8), (9)
> L^{eff}, and also L^{eff}_x, L^{eff}_y are not defined explicitely.
> One can try to infer which elements of the transport matrix may correspond to L^{eff} by guessing that
> they must be the largest values, but still distiction between L^{eff}_x and L^{eff}_y is not
> straightforward. The problem arised after replacing Eq. (7) written formerly for the general case in
> terms of symbols, by the present selected example with values of the transport matrix elements for
> a particular store. I understand the aim of the change: to give idea about values of TM elements.
> But for an average reader these precise values probably are only of moderate interest. Thus I would
> propose to go back to the previous version, i.e. general Eq. (7) with symbols, which is more transparent.
Now, we have the transport matrix in both symbols and in real nos. and so L^{eff}_x and L^{eff}_y are clearly defined.
Hal for the Argonne group :
> ** l.168 - maybe change "downstream" -> "away from"??
A few persons/groups have made suggestions to this sentence. Now it's changed to be:
"The Roman Pot stations are located on either side of the STAR interaction point (IP) at 55.5 m and 58.5 m with horizontal and vertical insertions of detectors respectively."
> ** l.225 - "... with p = 100.22 GeV/c the beam momentum." or something like this to explain why p is not 100.00.
Actually, even the total energy is not 100.00 GeV. All of these come from the fact that G-gamma has been set to be 191.5 between 191 and 192 (to avoid integers which might result in resonance). We've changed it to be "p = 100.2 GeV/c" as have the fraction error in the order about 1e-4.
> ** Eq. 11 - this equation assumes perfectly transversely polarized beams. We believe it is possible that due to
> non-ideal beam orbits and magnetic fields or magnet alignments, the beams may have a small longitudinal
> component at the STAR IR even when the rotators are off. This could lead to a (A_LS = A_SL)sin(phi) term in the
> equation. It will (probably) also be negligible in your analysis as you describe ~ line 241. Please consider whether
> you wish to add such a term in the equation and a couple words in the text about it. Alternately, you might
> mention that higher order correction terms are ignored in the equation.
We now mention "higher order correction terms are ignored" under Equation 11, as you suggest.
> ** l.272 - perhaps quote the actual values instead of "~55" and "~10"??
The "collinearity" in reality is slightly different from run to run, not just one single value.
> ** l.283 - do we understand correctly that there may be a false asymmetry that is proportional to the beam polarization, and it has not been ruled out by your tests?
There we try to explain in detail that the false asymmetry is ~ 0.
> ** Fig.6 - The text and error bar run into each other for (d).
> Fig.6 caption - is the vertical dashed line the average of experimental values or just zero or ...??
> Please include in the caption.
We've added "The vertical dashed line indicates where Im($r_5$)=0." at the end of the Fig. 6 caption.
> ** l.299,300 - this sentence is a bit confusing. You use "variable" for both deltaB and Re r_5, we believe. Maybe
> something like "The remaining lines show changes of Re r_5 and Im r_5 when the parameter was varied by +/- 1
> sigma during the fit procedure." We are not sure this is even what is meant. Sorry.
Actually, you've understood perfectly. We've made the changes as you've suggested.
> ** l.308,9 - "... more emphasized in estimating ..." we are confused by this and aren't sure what the
> authors are trying to say.
It just says the AN peak is more sensitive to Im(r5) and thus attracts more attention. This explains why we look at the Im(r5)'s in Fig. 6.
Panjab University :
> Page 6 line no. 222
> After the selection chi**2 < 9, please explain.
Most cuts are at 3 sigma-level and so this so-called chi**2 at 3-sigma is 3**2 = 9.
> Page 7 FIG. 3
> Is it possible to display the distribution of forbidden asymmetry for the five t ranges in
> FIG. 3(a)-(e) instead of showing for the whole range of t in FIG.3f?
>In FIG. 3(a) the error bars are large for two points around phi 80. Similar trend is seen in FIG. 3(e).
>It is not seen for the negative phi values. Any reason.
We feel that the figures are already too busy and it's not good to add more. The larger error b\
ars are because of low statistics and they're related to how close the respective vertical roma\
n pots were moved to the center of the beampipe. During the run, we tried to move the pots as c\
lose to the beampipe as possible without having too much background.
> Page 8 Table 1
> First bin (0.003< -t <0.005), the statistics is less(20%) as compared to other bin. Its width is
> 0.002 as compared to 0.005 and 0.15 for other bins. I think fit in FIG. 4 should be made removing this point.
The point with the smallest -t range is probably the most interesting point in this measurement\
and it contributes a lot in determining the shape.
Janusz from the Cracow group :
> General:
> Explain fully the coordinate system shown in Fig. 1
In the caption of Fig. 1, I now add "Positive y is pointing towards the sky and positive x is pointing to the center of the RHIC ring."
> 144. At very high energy sqrt(s) - -> At high centre of mass energy, sqrt(s),
Now, we define "center of mass energy" for sqrt(s) in "Introduction" (the 1st time that it appears, line 127) and here, we just say "At very high sqrt(s)".
> 154-155. The contribution of the two spin-flip amplitudes, …., to the asymmetry AN is small as indicated by both experimental estimates [] and th. pred.[18]
OK. We have changed it to :
"The contribution of the two double spin-flip hadronic amplitudes, …., to the asymmetry AN is small, as indicated by both th. pred.[18] and experimental estimates [19,20]".
> Section 3
> 168. on each side - -> on West and East side
A few persons/groups have made suggestions to this sentence. Now it's changed to be:
"The Roman Pot stations are located on either side of the STAR interaction point (IP) at 55.5 m and 58.5 m with horizontal and vertical insertions of detectors respectively."
> 190. Give also value of the distance of 10 mm in units of the beam width at RPs.
It's about 10-12 sigma's but it's probably difficult to be exact.
> 214. small correction less than 4 $\mu$rad, the full - -> small, less than 4 $\mu$rad, correction, the full
We've adopted Hal's suggestion : "small corrections of less than 4 $\mu$rad" .
> 218-221. are not very clear, especially the use of “similar’ or “typical”
> replace: are taken from the fits similar to those in Fig. 2 - -> are taken from the fits to data performed for each run. An example is presented in Figs. 2(a) and 2(b).
We've changed to "are taken from the fits to data performed for each data segment. An example is shown in Fig.~2."
> 257. the position of the t = 0 trajectory … - -> the position of the t = 0 elastically scattered proton trajectory or the beam position
"t = 0 trajectory" is an ideal trajectory with no scattering.
> 268-269. The simulation included … optics - -> Simulation of the elastically scattered proton transport through the
> RHIC magnetic lattice and the apertures was performed and the detector acceptance was calculated.
The present form seems cleaner and the acceptance is mentioned in the following sentence.
> 274-275. Assuming the background is unpolarized - -> Assuming that the background is the beam polarization
> independent …
"unpolarized" is probably easier to understand.
>298-302. Table II shows the fitted values of Re r5 and Im r5 together with statistical and total systematic
> uncertainties. Also, the contributionsto the systematic uncertainty are given in this table. They are due to:
> systematic uncertainty on Leff , alignment ……. .They were obtained by changing the value of the considered
> parameter by $\pm$1 standard deviation.
A few people have commented/suggested, we've changed to:
"In Table II, we show the central value of the fit and uncertainties on Re$\:r_5$ and Im$\:r_5$ due to various effects. In the first line of the table, the statistical error to the fit with the central value of the parameters is shown. The remaining lines show changes of Re$\:r_5$ and Im$\:r_5$, when the parameter was varied by $\pm$1$\:\sigma$ during the fit procedure. ......"
> 308-309 Remove sentence: Since the maximum... since it is not needed
This sentence explains why we look at Im(r5), but not Re(r5).
> Present the main result with bold face font
It's difficult to say which nos. are more important in this table. ( And we have changed 1st two lines into one line as Andrzej has suggested. )
> Table II. …. Measurement induced uncertainties (1) – statistical, …. (4). Uncertainties associated with the fit: (5) – the total cross-section …
Changed to "Table II: ..... (1): Statistical Uncertainties. (2)-(4): Systematic uncertainties associated with this measurement.
(5)-(7): Systematic uncertainties associated with ...."
Steven Heppelmann for Penn State U. :
>If I look for an issue it would be the rather large chi-square value on Figure 3b (32.68 for 17 DOF). This is fairly unlikely
>(1-2%) and might suggest that the errors bars on the 17 points should actually be about sqrt[2] larger than the ones
>plotted. In fact, most of the points of Figure 3 have chi2/dof >1 which as a group is somewhat unlikely too.
>Because the statistical error from the fits are only about 1/2 of the error on polarization, I don't think the story would
>change much but I just wanted to comment and ask if some words in the text should be added to indicate that this has
>been considered in the systematic error analysis.
If we look at the variation of point-to-point, it's bigger than statistical variations. That means there
are some point-to-point systematic uncertainties. These variations are likely from the variations of
geometry of t=0 and the uncertainties are factored in systematic uncertainty of delta_t(alignment).
If we add systematic errors to the points before the fits, we'll certainly get better/more realistic chi-sqs.
The other point is that the function we are fitting with is not "perfect".
If we add high order/off-set terms, such as phi0 which we've decided to drop, chi-squares get better.
We have chi2/dof = 11.1/16, 25.8/16, 12.6/16, 23.2/16, and 14.2/16 for the 5-bins (when phi0 is included).
The errors look small partly because we use the same “large scale” for all the plots because
we want to accommodate the with the largest AN scale, ie. 3(a), in which the chi2/dof = 11.32/17 < 1.
For the other 3 fits, chi2/dof~1.18, 1.54 and 1.32 which seem to be reasonable experimental fitting results.
And of course, we've used the method/formula to calculate the error on each bin.
Shan Dong University :
> 1)In line 214,215, "full transport matrix was used", does this mean Eq(8),Eq(9) and L_{x,y}^{eff} are not used for
> the calculation of angles? If so, how to understand the uncertainties of L^{eff} in line 257,263,274 and in table II?
> Or they are only used in estimating the uncertainties?
L^{eff} (for x or y) are just two of the elements (the two > 20 m) in the transport matrix
and so when the transport matrix was used, they were indeed used. These were the two dominating terms in the transport matrix and the other elements are very small
compared to L^{eff}. Conceptually, it's often easier to just consider these two terms
when you try to understand various things in the analysis. So, equations (8) and (9) are
just approximations to help people understand/grasp the main idea and the transport
matrix has been used in the analysis. And indeed, the uncertainties L^{eff} are
essentially the uncertainties of the transport.
> 3) Line 241: "preliminary results of this experiment [20] show that..."
> Are these results not part of this analysis or not further checked using the
> final data sample? It is a bit surprising to cite preliminary results of ourselves
> for the same analysis, as we are the same collaboration or group.
The double-spin asymmetries A_NN/A_SS indeed belong to another set of analysis. Unlike the single-spin asymmetry (A_N) in this paper, there is NO square-root formula for extracting A_NN/A_SS. The square-root formula helps cancel out a lot of luminositybunch variance etc. We therefore need reliable normalization (bunch intensities etc.) for the A_NN/A_SS analysis which is an ongoing effort. We'd like to publish the A_N and r5 results first.
> 4) In line 305, "Re r_5 =0.00167 +/- 0.0063 in line 306 Im r_5=0.00722+/-0.057",
> The rounding of the digits should be consistent and make real sense as in other places
> in the paper.
What we've done is to show 3 significant figures for the measurements and 2-significant figures
for the errors. A "typical" (but not always) rule for displaying experimental uncertainties is to show
1-significant figure less since the uncertainty is an estimate and cannot be more precise than the best
estimate of the measured value. For some errors are too small compared to the dominant one (polarization),
so we've also restricted the significance to 4 decimal points which is ~O(1%) of relative accuracy.
Stephen Bültmann :
> line 171 : only -> almost exclusively
Done !
> line 172 : insensitive -> nearly insensitive
Done !
Last edited by Kin (May 9, 2012)
The pdf file of the paper and tex/figures in tar-gzipped format are attached here.
Running it requires several input files and some of which depend on which run number you're running against.
// ---------- Codes (say "mytest.C") to compile and run once if you have the filename(s) in .file.list --------------
// --- several run-dependent input files are really needed to run it meaningfully.
class StChain;
StChain *chain=0;
void mytest( string output_file = "hist.root" ){
gROOT->LoadMacro("$STAR/StRoot/StMuDSTMaker/COMMON/macros/loadSharedLibraries.C");
loadSharedLibraries();
gSystem->Load("libStBFChain");
gROOT->LoadMacro("${HOME}/MuDstpp2pp/AnalyzeMu.C++");
chain = new StBFChain();
StMuDstMaker *maker = new StMuDstMaker(0,0,"",".file.list","",1000);
AnalyzeMu( maker, output_file ) ;
if ( chain != 0 ) delete chain ;
return ;
}
// -------------------------- End -----------------------------------------------
1. The latest Version 3.2 draft of the paper can be found:
http://drupal.star.bnl.gov/STAR/subsys/pp2pp/anpaper-review/paper-draft/version32may-30-2012
2. Our responses to the comments from the collaboration, with links to responses to individual institutions can be found below. You should see by the name of the file.
http://drupal.star.bnl.gov/STAR/subsys/pp2pp/anpaper-review/response-collaboration-review
We also created a file which contains summary of responses to the comments and major questions. The link to the file can be found on the bottom of the list.