Photon Veto

Photon detection capability with an inefficiency of is a key strategy in suppressing backgrounds from decays and other radiative processes. Photons accompanying kaon decays are detected in the sr array of electromagnetic shower counters consisting of alternating layers of 1-mm-thick lead and 5-mm-thick scintillator. Roughly two thirds of the coverage is obtained from the barrel assembly, which surrounds the range stack, 2 in azimuth and 1.90 m axially. Most of the remaining solid angle is subtended by the two endcaps located upstream and downstream of the drift chamber. These systems are augmented by the capability to identify photon conversion elsewhere in the detector, particularly in the target and range stack, and to some extent, in the degrader, using the beam counters.

The barrel is constructed in 24 eight-module sectors as shown in Fig. 2 and 10, supported in a frame with a web-like cross-section made of 1.5-mm-thick stainless steel and fastened to the four aluminum plates supported on the magnet yoke. The azimuthal boundaries of each sector are tilted so that the inter-sector gaps of inert material (air, wrapping, and stainless steel) do not project back to any part of the target. The center divider in each sector splits each layer into halves so that the azimuthal segmentation is 48-fold. The four radial modules add up to a total of 14.3 radiation lengths, and are made up of 16, 18, 20, and 21 lead-scintillator layers, respectively with increasing radius. The 3600 scintillator panels in 75 sizes of increasing width were fabricated of BC408 by Bicron Corp.

Each module is epoxied to a 150-mm-long acrylic mixer block on each end. The light collected from the scintillator layers in each of the 384 mixer blocks is coupled via a flexible wafer made of silicone dielectric gel (Dow Corning SYLGARD 527), to a 67-mm-diameter, 850-mm-long UVT acrylic lightpipe fed through a 73-mm-diameter hole in the magnet endplate steel to a 77-mm EMI 9821KB PMT mounted on the upstream or downstream end of the magnet. Geometrically, the fraction of scintillator light reaching the photocathode varies from 30% for the inner modules to 17% for the outer modules, and produces 10 photoelectrons per visible MeV energy loss in the scintillator. The PMTs are individually shielded from the <15-G fringe field with -metal sleeves. Each PMT signal is connected to a passive splitter with outputs to LRS CAMAC 4300B FERA ADC and LRS 1879 Fastbus TDC systems, and to the energy-sum units used in the fast trigger.

In the regions upstream and downstream of the drift chamber covered by the two endcap photon-veto assemblies [18], the lead-scintillator layers are oriented perpendicular to the detector axis so that all photons originating from the target will convert with high probability, leaving acceptable visible energy. Both endcaps surround a 200-mm-diameter hole on the beam axis allowing for the beam upstream, and the target assembly downstream. Each endcap is segmented into 24 azimuthal modules with 66 petal-shaped lead-scintillator layers making up 12.4 radiation lengths. The 3168 scintillator pieces were machined from NE104 (Nuclear Enterprises, Scotland) using a diamond-tipped cutter which left a surface finish that required no polishing to obtain adequate internal light reflection. The light output of each sheet of NE104 was tested using a Ru radioactive source, and the material in each module was selected to insure reasonably uniform response from module to module.

In order to transport the light axially with a minimum of inert material in the crucial volume, light from the scintillator layers in each module is absorbed and re-emitted by a fluorescent wavelength-shifter bar which covers most of the outer edge [19]. The wavelength shifter is viewed by a 52-mm EMI 9954KB PMT via an adiabatic lightguide, coupled with a flexible wafer made of silicone dielectric gel to a 1.3-m-long, 41-mm-diameter UVT acrylic lightpipe through a hole in the magnet endplug steel. Each PMT is locally shielded against the small fringe magnetic field. The wavelength shifters were made of 6.5-mm-thick acrylic doped with BBOT (Bicron Corp.), chosen for its very fast light output decay time of 1.6 ns. The emission and absorption properties of NE104 and BBOT are well matched for our geometry and give 8--10 photoelectrons per visible MeV in the modules. The attenuation lengths of the NE104 and BBOT are approximately 400 and 800 mm, respectively. Other tests showed that the output from NE104 would not degrade by more than 20% due to radiation damage over the expected life of the experiment.

The 24 modules in each endcap are enclosed in a welded 700-m-thick stainless steel frame. A detailed view of a module is shown in Fig. 18.

 
Figure 18: Isometric schematic showing the structure of a photon veto endcap module.  

An air gap between the scintillator and the wavelength shifter is maintained by two 300-m-diameter copper wires extending the full length of the module. The air gap minimizes reabsorption by the primary scintillator layers of light reflected internally in the wavelength shifter. The modules are supported in the frame in such a way that uniform pressure applied by small springs holds the wavelength shifters in place. Each endcap assembly (750 kg) is supported in a 4.8-mm-thick aluminum cylinder bolted to the magnet steel endplug.

The relative quantum efficiencies for the PMTs were measured and tubes were selected to further balance the output. The PMT signal cables feed amplifiers in the counting room. One amplifier output is connected to a passive splitter which distributes signals to the ADC system, the discriminators feeding the TDC system, and the summing units used for the fast trigger. The second output is summed in groups of 4 separated modules and the sum is fed to a TD channel.

Calibration of the barrel is performed by comparing the energy depositions from cosmic ray data with the calculated energy loss of minimum-ionizing radiation in the known scintillator thickness. Even though the trigger and analysis constraints eliminate events passing through the scintillator at shallow angles, cosmic muons have a broad asymmetric spectrum due to large Landau fluctuations. In addition, there is a large end-to-end variation in light output (factor 2.3--7.3) because the 1.2-m average attenuation length is short compared to the 1.9-m active length. (The attenuation length was determined for the barrel in a similar manner to the range stack.) The position dependence of this effect can be eliminated by considering the geometric mean of the light output from both ends.

Photon vetoing is based on visible energy, i.e., energy deposited in the scintillator. However, total energy calibration is useful as a check (see below), and for other physics measurements involving photons. To turn calibrated visible energy in the barrel into actual photon energy, a correction is made for the invisible energy deposited mostly in the lead but also in the support structure and wrapping materials. This correction was calculated using a Monte Carlo simulation of the passage of electrons and photons through matter using the program EGS [20]. Three energies, 20, 100, and 200 MeV, were run under various conditions, and gave a typical visible energy fraction . The tilted support structures show up as 15% dips in as a function of azimuth of entry, as shown in Fig. 19.

 
Figure 19: Azimuthal dependence of the visible energy fraction in the barrel photon veto, correlated with the support structure shown at the top.  

For the endcaps an initial crude calibration was obtained using cosmic ray muons, although, unlike the barrel, the geometry and orientation of the endcaps in the detector makes a clean trigger, for which the energy deposition is known, impossible. Proper calibration is done in beam with a special trigger to use the 152-MeV monoenergetic muons from decay which stop in the first third of the endcap, after losing energy in the target and other parts of the detector. The trigger and analysis select events for which the muon energy is contained mostly in one module in order to provide an easily recognizable peak. The expected visible energy is determined by a Monte Carlo calculation. The HV was adjusted to balance the fit peaks in all modules and establish the required dynamic range, normalized to the fit Monte Carlo spectrum. The systematic uncertainties on the gains set in this manner are estimated to be about 10%.

Since this calibration relies heavily on a Monte Carlo calculation, an independent check was carried out by reconstructing the energy of the neutral pion in decay. The kinematics are strongly biased against both photons hitting the endcaps when the is accepted in the range stack. However, by using the more reliably calibrated barrel, it is possible to use events where one photon is in the barrel and one in an endcap, and add their measured energies to reconstruct the energy. Taking into account the visible energy fraction , which should be similar for barrel and endcap systems, the consistency of the endcap energy calibration could be checked. As a test of the overall photon energy calibration and the determination of , the summed energy from the barrel and the endcaps of the two photons from in clean triggers was measured and is shown in Fig. 20.

 
Figure 20: Summed photon energy corrected for the visible energy fraction from events. The energy in such events is shown by the vertical dashed line.  

The peak is close to the expected value of 245.6 MeV, the energy.

events are also used to determine the time calibration of the barrel and endcap in the TDCs, since the in the range stack defines the event time reference. The time resolution is limited by the 2-ns time-bucket width in the pipeline TDCs.

The photon-veto system performance is optimized by simultaneously minimizing two characteristics --- the accidental rate and the photon inefficiency. The accidental rate causes good events to be rejected due to random uncorrelated hits, resulting in a loss in acceptance. The photon inefficiency limits the primary function of the system to reject backgrounds. In our photon-veto system, the parameters to optimize are the low-energy threshold and the width of the coincidence-resolving-time window.

For a given energy threshold, the accidental rate is reduced by narrowing the vetoing time window. This is determined by discriminator pulse widths in the trigger and by the time resolution of the detector and electronics in the off-line analysis, which uses leading-edge timing, rather than overlap. To study these effects, we use data and plot the leading-edge time of all individual energy depositions in the barrel and endcaps with respect to the measured time of the . These plots show a peak corresponding to photons from s and a background due to accidentals. Fig. 21

 
Figure 21: Barrel and endcap (a) time resolution and (b) accidental rate as a function of visible energy.  

shows the time resolution and accidental rate inferred from such plots as a function of visible energy. The barrel benefits in resolution from being read out at both ends, and in accidental rate from being farther from the beam than the endcaps.

In the trigger, pulse widths make the effective time window for vetoing tens of nanoseconds. Visible-energy vetoing thresholds are 5 MeV for the sum of barrel energy and 10 MeV for the sum in each endcap. Off-line, time windows are reduced to as low as 12 ns and thresholds are typically 1 MeV for the entire detector.

The inefficiency of the photon-veto system can be measured directly in terms of detection in decay (see below). However, in order to guide the design of the detector and to study its performance, Monte Carlo simulations were performed. Inefficiency comes from three sources: escape of photons before conversion, sampling fluctuations in which little or no energy is deposited in scintillator, and photonuclear interactions in which no charged products are detected. The Monte Carlo program uses the EGS package [20] for electromagnetic interactions, and photonuclear reactions are simulated using PICA [21]. Other routines propagate the reaction products, including neutrons, through the detector, and describe the final-state nuclear de-excitation by a simple evaporation model. The effects of the last few MeV of nuclear excitation are studied separately using various assumptions about the cascade of low-energy photons. For our low detection thresholds, the inefficiency is quite sensitive to these assumptions, giving estimates with an uncertainty of about an order of magnitude.

Nonetheless, the simulations indicate that detection inefficiencies for single photons in events vary from to for photon energies between 20 and 225 MeV. Sampling fluctuations dominate the loss of low-energy photons (<50 MeV), while photonuclear interactions dominate the loss of high-energy photons. For the latter, photonuclear losses are larger than escape losses (the next-largest effect) by about a factor of five.

The primary function of the photon detection system --- rejecting photons from s produced by kaons decaying at rest --- benefits from several factors beyond simple detector efficiency: There are two photons in each event; both have energies greater than 20 MeV and one always has energy greater than 122 MeV; the high-energy photon is constrained by momentum conservation and our selection to penetrate the most efficient sections of the photon detector, including the range stack which is one radiation length thick with little inert material. The detection inefficiency measured by observing the magnitude of the 205-MeV/c momentum peak surviving the analysis [9] is 1 -- 3 , depending on analysis cuts, and is consistent with Monte Carlo simulations.