DST software innovations were needed to cope with the problem of radiation damage in the scintillators. The radiation damage is known [44] to result in significant absorption of light in the scintillators and observable (factor of 2 in our device) dependence of the signal amplitude on the location of primary track. Hitherto, this problem had not been addressed in the NA44 DST production software.
The correction method chosen consisted in constructing a product of the two PMT amplitudes for each tower. The way it affects usage of the calibration constants is a subject of a special discussion and therefore is separated in Appendix A. In brief, to recalibrate, one must multiply tower's energy by a factor which depends on the attenuation length and therefore has to be measured. The measurement was performed by selecting the tower with maximum amplitude (to be able to ignore the effects of threshold and pedestal subtraction, the random details of propagation of shower tracks to the neighboring towers and sharing of light between the towers), then averaging separately the sum and the product of the tower's two PMT signals, and taking the ratio of sum to the square root of product. Only single track events were considered. Under some model assumptions, one can relate this ratio to the attenuation length in the scintillator.
I estimated the attenuation length to be between 5 and 7 cm for most of the towers in 1996. It was seen to be systematically shorter in the EM section (which had no Cu plates in it [32]). 5 out of the 9 UCAL stacks were U/Cu/scintillator, and 4 were U/scintillator. The pure U stacks showed better performance and therefore seemed to have been restacked with newer scintillator. Comparing the sum/ ratios between 1995 and 1996, I noticed a systematic increase of about 5-7% in 1996, which indicated continuing deterioration .
A potential danger associated with using the product is that of losing the signal altogether if at least one of the PMTs gives no signal, as may happen due to the attenuation. However, the counterargument is that this never happens for (the nominal spectrometer momentum) electrons in the EM section and hadrons in the hadronic section. Therefore, the possibilities of identifying the electromagnetic events by high EM signal, selecting hadrons by high signal in the hadronic section and vetoing any background by low signal remain unaffected. I studied the issue quantitatively, selecting (by Cherenkovs, C2 at 14.7 PSI) a sample of electrons and a sample of protons and kaons in a 4 GeV positive setting, and found that for the true electrons the inefficiency due to making a product instead of a sum is , which is comparable with the inefficiency due to non-interacting (the non-interacting probability is ). For and , such kind of inefficiency in the hadronic section is less than due to the larger signal from these particles.
Out of a variety of other possible correction methods, the following two were tried:
|