Introduction

Test runs taken in 2004 for the ZDC SMD as a local polarimeter found a very high analyzing power in 200 GeV collisions, about 8%.  In 2009 the ZDC SMD was commissioned for 500 GeV running and also found an analzying power at 500 GeV of about 8%.  The 200 GeV running with the ZDC SMD this year, however, found a much smaller analyzing power <4%.  We would like to understand the drop in analyzing power at 200 GeV.  The trigger conditions and thresholds were not exactly the same in both years.  We suspect the large drop is because the thresholds were set too high in the 200 GeV running.

The trigger conditions for 2004 are detailed here.  The trigger conditions for 2009 are detailed here.  In 2004 there was a TAC condition on the ZDC, which requires a hit above a threshold.  We understand that this threshold was placed on the sum of the calorimeter modules - and plots of ADC spectra from the modules seem to confirm this.  The threshold was 5.  in 2009 there were thresholds on both the front and back modules of 40. 

 The argument for changing the trigger condition to include a threshold on the front back went as follows: we think the partcile which produces the large analzying power is a high x_F neutron.  By requiring energy in the front and the back we will preferentially select hadronic particles.  An email from Carl to this effect is here.  It's possible the 2004 data sample contained a large amount of contamination from EM particles.  (Side note:  What is we find that the EM particles also have a high analyzing power?  Wouldn't that be interesting?)

 We cannot compare directly the thresholds in 2009 and 2004 because of fundamental changes to the hardware.  Spoke to Xhangbu and Gang about how best to do this.  The suggestion was to compare slopes in the SMD to get an idea for how much gain has really changed. 

ZDC SMD Gains in 2004 vs 2009

Below if a picture of ADC-ped spectrum for each slat for 2004 (blue) and 2009 (pink).  You can see a dramatic difference in slopes.

Below is a table of Gain = 1/slope for each slat for both years.

Slat    2004    2009    Ratio[2004/2009]
0       7.174   24.502  0.293
1       8.200   34.930  0.235
2       7.726   42.820  0.180
3       7.041   37.322  0.189
4       6.537   36.587  0.179
5       7.086   34.012  0.208
6       7.507   29.031  0.259
7       5.090   21.023  0.242
8       5.935   24.803  0.239
9       6.058   26.767  0.226
10      5.443   29.618  0.184
11      6.010   30.110  0.200
12      6.168   35.836  0.172
13      6.671   45.879  0.145
14      6.190   41.552  0.149
15      8.833   28.261  0.313
16      10.583  38.608  0.274
17      9.769   46.787  0.209
18      8.293   38.592  0.215
19      7.627   33.939  0.225
20      8.681   31.628  0.274
21      8.623   27.064  0.319
22      7.585   24.096  0.315
23      7.195   25.938  0.277
24      7.122   30.072  0.237
25      6.459   31.648  0.204
26      6.774   34.181  0.198
27      8.699   40.535  0.215
28      8.343   43.106  0.194
29      7.798   38.216  0.204
 

The average difference in the gains is 23%.

Looking at this we realized at least one problem.  We select on hit per event by requiring a coincidence in vertical and horizontal slats above a threshold after ped and gain corrections.  In 2004 this value was 6.  We never changed this in 2009.  If we want to do the equivalent analysis the threshold should have been changed to 26!  (And maybe higher in 500 GeV data).  To test if this is problem we are running the 2009 data again with a new threshold of 26 and reprocessing the 2004 data with a threshold of 1.

Tests

1a.Increase thereshold for good hits in 2009 data to 26.  We see the left-right asymmetries change as follows:

East:  0.0149(0.0004) -> 0.0168(0.0005)  (~13%)

West: 0.0162(0.0003)-> 0.0177(0.0004)   (~10%)

This is not big enough to account for the difference

1b.  Lower threshold for good hits in 2004 data to ~2009 levels (ADC=1, almost no cut).  The L-R asymmetries change as follows:

East: 0.0155(0.0010)-> 0.0106(0.0005)  (32%)

West: 0.0130(0.0009)->0.0092(0.0006)   (30%)

This is a pretty large difference, bigger than raising the 2009 threshold.  But this might not be totally fair since an ADC-ped threshold of 1 is really close to the pedestal.  I would imagine something between the 10 and 30 % is the right number.

2.  Put threshold on front and back modules in 2004 data.  I checked the gains for the calorimeter modules in 2004 vs. 2009 and find:

Cal     2004    2009    Ratio[2004/2009]
East       16.797  26.887  0.625
West       15.890  19.577  0.812
Average Ratio is 0.718
   a.This would imply that the threshold of 40 would be 29 (?).  I attempted to analyze the data with this threshold, but it really destroyed the stastics avaialable and I couldn't make a conclusion. 

   b.  Returned ADC good hit value to default of 6 and try various thresholds on front and back ZDC calorimeter modules.

        Threshold                   East LR Asymmetry                    West LR Asymmetry

        Sum=5 (Used)             0.0155(0.0010)                        0.0130(0.0009)

        Front, Back = 5            0.0131(0.0020)                         0.0138(0.0019)

        Front, Back = 10          0.0117(0.0035)                         0.0145(0.0036)

        Front, Back = 15          0.0050(0.0058)                         0.0121(0.0064)

        Front, Back = 20          0.0058(0.0098)                         -0.0099(0.0110)

East and west seem to have different tolerances for this - or maybe its just unequal statistics, I recall that the west side usually saw more events for much of this year's running.

Conclusion

About 10-20% of the difference in analyzing power can be accounted for by having a software threshold on good hits which was too low.  The rest is most likely due to setting the thresholds on the zdc calorimeters too high, but I could not verify this rigorously.