Coincidence resolution time of two small scintillators coupled to high quantum-efficiency photomultipliers in a PET-like system

The lower limit of the time resolution for a positron emission tomography (PET) system has been measured for two scintillator types, LYSO:Ce and LuAG:Pr. Small dimension crystals and ultra bi-alkali phototubes have been used in order to increase the detected scintillation photons. Good timing resolutions of 118 ps and 223 ps FWHM have been obtained for two LYSO and two LuAG, respectively, exposed to a 22Na source.


Introduction
Lowering of the coincidence resolution time (CRT) in the detection of the two 511 keV γ-rays from e + e − annihilation events considerably improves the signal-to-noise ratio of PET devices [1].The timing performance of a scintillator-detector system is caused by many parameters, some related to the scintillator (rise time, decay time, optical photon yield, geometry, and surface finish), some to the photodetector (quantum efficiency QE, transit time spread TTS, single electron response SER).A further contribution derives from time pick-off methods, e.g.leading edge (LED) or constant fraction discrimination (CFD) and full digitization techniques.
In the present work we have investigated the lower CRT value that can be obtained by using recent photomultipliers (PMTs) from Hamamatsu with higher QE values, and the most used scintillator in standard PET, Proteus LYSO crystals.For comparison the CRT has also been measured for Furukawa LuAG:Pr scintillators.Contributions from optical photons propagation [2] have been minimized by using small size crystals, at the cost of a low detection efficiency.The LED method has been adopted because it performs [3] the best CRT value at lower thresholds with usual laboratory electronics.CRTs were measured for two 3x3x5 mm 3 LYSO and two 3x3x3 mm 3 LuAG:Pr crystals, respectively.All crystal surfaces were optically polished, with five surfaces wrapped in many layers of Teflon.The experimental set up includes two couples of back-to-back scintillator+PMT pointing towards a 22 Na source, plus NIM/CAMAC signal processing electronics.Each PMT signal is resistively split to get charge and time information.The time is picked up at a given threshold Q in a CAEN N841 LED.The event trigger is defined by the coincidence of signals over threshold in both detector arms.This also provides the start for a CAEN C414 TDC with 25 ps resolution and the integration gate for a LeCroy 2249W ADC with 0.25 pC/count.TDC time stops were taken from the two individual discriminated signals, suitably delayed to enter the TDC full scale.

Materials, methods and measurements
Signals for both scintillators (Fig. 1a) at the 511 keV photopeak from 22 Na show different average heights (V 511 keV = 5 V for LYSO, 2.6 V for LuAG) and decay times [4] (T 511 keV d = 40 ns for LYSO [5], 20 ns for LuAG [6]), but display comparable rise times (T 511 keV r = 4.4 ns for LYSO, 4.2 for LuAG), mostly dominated by PMT and electronics shaping.The ratio of the charge for the 511 keV signals to the single photoelectron charge Q 1pe , measured from SER, provides the photoelectron (pe) yield N pe = 4240±150 for LYSO [7] and N pe = 1727±182 for LuAG [8] crystals, respectively.The different N pe values are due to the different scintillator light yields and peak emission (310 nm for LuAG, and 420 nm for LYSO).Spectra from these two crystals exposed to γ-rays from a 22 Na source are displayed in Fig. 1b.The energy resolutions for the 511 keV photopeaks are: σ/E=3.16%for LYSO, and σ/E=3.38%for LuAG.After the quadratic subtraction of the statistical contribution (σ/E) stat = ENF/N pe ), the intrinsic resolution (σ/E) 511keV intr quotes 2.65% for LYSO, and 2.03% for LuAG.

Coincidence time resolution results
The charge correlation plot (Fig. 2a) provides the accurate selection of coincident photoelectric events of 511 keV γ-rays from 22 Na in both crystals against Compton, intrinsic radioactivity and higher energy radiation from source.The spectrum (Fig. 2b) of time differences between the two detector arms, gated on photoelectric effect (arrow in Fig. 2a), measures the CRT timing.Their standard deviations σ CRT were measured at several discrimination thresholds (Q) for both LYSO and LuAG crystals.The σ CRT as a function of Q (Fig. 3) is generally explained by the quadratic sum of independent contributions from scintillators, photodetectors, signal transport, and handling electronics.The electronics contribution σ 0 , measured by the time spread (16 ps) of the TDC start signal, weakly affects the overall measured resolutions.The scintillation contribution σ phot due to the underlying photon production statistics, depends on the scintillators.If written as σ phot = c 1 (QT r T d /N pe ), it accounts for the high-Q values of σ CRT .Q is expressed in pe, 1 pe ≈ 11mV.The constant c 1 should be the same for both scintillators and depends on the PMT.The contribution of photodetector noise is σ noise = σ rms /(dV/dt) Q , where (dV/dt) Q is the signal slope at threshold Q, and the σ T T S = (T T S /2.35)/ N peQ , being N peQ the number of detected pe at threshold Q.Both these strongly influence data at low-Q values with a 1/ √ Q dependence, and a constant value at high Q-thresholds.The data in Fig. 3

Summary
The CRT values of two LYSO and two LuAG:Pr scintillators in a PET-like geometry have been measured.Small size crystals have been employed in order to neglect the contribution from light transport.Two R7600U-200 PMTs with UBA photocathode have been used.The PMTs have preliminarily been characterized with a short-pulse pico-laser.Their performances agree with Hamamatsu specifications.Owing to UBA high QE PMTs and to the small crystal sizes, a high number of pe is measured for 511

Figure 2 :
Figure 2: (a) Correlation of signal charges for the two PMT+LYSO detectors with 22 Na source in between.(b) Gaussian fitted time difference spectrum of correlated 511 KeV γ-rays detected in LYSO crystals, at Q-threshold=5 mV, yielding σ CRT =50 ps (FWHM= 118 ps) time resolution.
Two metal package ultra bi-alkali (UBA) R7600U-200 PMTs with the E5996 resistor chain from Hamamatsu Photonics have been used (QE = 43% @ 350 nm, rise time 1.4 ns).The PMTs were The associated noise was σ rms = 1.5 mV.At half pulse height, an average SER slope dV 1 /dt = 17 mV/ns was measured.Single photon charge Q 1pe = 1.25 pC with σ Q1pe = 0.64 pC was measured, yielding an excess noise factor ENF = 1.26.Time-walk corrected (owing to the charge-time correlated measurements) TTS values were 265 ps and 390 ps for the two PMTs, in good agreement with Hamamatsu specifications (350 ps).