Geant4-based calibration of an organic liquid scintillator

A light-yield calibration of an NE 213A organic liquid scintillator detector has been performed using both monoenergetic and polyenergetic gamma-ray sources. Scintillation light was detected in a photomultiplier tube, and the corresponding pulses were subjected to waveform digitization on an event-by-event basis. The resulting Compton edges have been analyzed using a Geant4 simulation of the detector which models both the interactions of the ionizing radiation as well as the transport of scintillation photons. The simulation is calibrated and also compared to well-established prescriptions used to determine the Compton edges, resulting ultimately in light-yield calibration functions. In the process, the simulation-based method produced information on the gain and intrinsic pulse-height resolution of the detector. It also facilitated a previously inaccessible understanding of the systematic uncertainties associated with the calibration of the scintillation-light yield. The simulation-based method was also compared to well-established numerical prescriptions for locating the Compton edges. Ultimately, the simulation predicted as much as 17% lower light-yield calibrations than the prescriptions. These calibrations indicate that approximately 35% of the scintillation light associated with a given gamma-ray reaches the photocathode. It is remarkable how well two 50 year old prescriptions for calibrating scintillation-light yield in organic scintillators have stood the test of time.


Introduction
Due to relatively high detection e ciency, strong inherent gamma-ray rejection properties, and fast scintillation pulses, organic liquid scintillators are typically employed to detect fast (MeV) neutrons in mixed neutron and gamma-ray elds. The aromatic organic liquid scintillator NE 213 [1] was originally introduced in the 1960s [2] and poses a nonnegligible health risk. However, the excellent intrinsic neutron/gamma-ray pulse-shape discrimination characteristics and high fast-neutron detection e ciency continue to make NE 213 an excellent choice for fast-neutron applications. In this paper, the scintillationlight yield of the more recent NE 213A version of the liquid is calibrated using the Compton edges in measured energy distributions from a set of gamma-ray sources. This effort has been undertaken as the rst step in a systematic program of parametrizing the scintillation-light yields of some recently developed organics and oils. The analysis of the data has been greatly facilitated by a GEANT4 simulation of the detector apparatus which models the interactions of gamma-rays and secondary electrons as well as the scintillation photon transport.

Gamma-ray sources
Above gamma-ray energy E γ ∼ 100 keV, the scintillation-light yield produced in organic liquids by atomic electrons freed by interactions with incident gamma-rays is very close to linear [3,4]. The low average Z value typical for organics results in the gammaray/electron interactions being dominated by Compton scattering. Above E γ = 1.022 MeV, pair production takes over and dominates by ∼5 MeV. Measured Compton edges located at energy E CE may be evaluated to calibrate the scintillation-light yield of a detector. Table 1 summarizes the radioactive sources used in this work.  [5]. Table 2 presents some of the well-known properties of NE 213A.  From the right, the cell (light gray) and the µ-metal shielded PMT and base housing (black). Contacts for signal and high voltage (gray) extend to the left from the base of the housing. For interpretation of the references to color in this gure caption, the reader is referred to the web version of this article.

Signals, electronics, and data acquisition
The operating voltage of the detector was set at −2 kV, a voltage employed for this detector in previous VME setups [13][14][15][16][17]. At this voltage, a 1 MeV ee signal had a risetime of ∼5 ns, an amplitude of ∼900 mV and a falltime of ∼60 ns. The data-acquisition system was based on a CAEN VX1751 Waveform Digitizer [18] with a 10 bit ADC and an analog input bandwidth of 500 MHz. The digitizer was con gured for a 1 µs acquisition window with 10 9 samples per second over a −1 V dynamic input range. The voltage resolution was ∼1 mV. In order to preserve the −2 kV operating voltage used in the previous investigations, it was necessary to attenuate the analog signals from the detector by 16 dB using a CAEN N858 dual attenuator module [19]. Figure 2 shows a typical waveform.
The internal falling-edge threshold was set to −25 mV. The waveform of each pulse was analyzed using a suite of analysis software [20] developed in-house. Analysis of the data was performed using the Python-based [21] code libraries pandas [22], SciPy [23], and numpy [25], where the signal baseline was rst subtracted so that the charge corresponding to each scintillation pulse could be determined by integration. The event-timing marker was obtained using a standard zero-crossover method [4]. Voltage sampling was started 25 ns before the event-timing marker and extended to 475 ns after the event-timing marker. Integration was performed o ine over this 500 ns window which will be required for neutron/gamma-ray pulse-shape discrimination, resulting in an o ine software-based charge-to-digital conversion. The conversion was calibrated to 6.35±5.5% fC/QDC channel using a charge-injection circuit.

GEANT4 simulation
The response of the detector to gamma-rays was simulated using a C++ Monte Carlo model developed with the GEANT4 toolkit [26]. GEANT4 version 4.10.04 [27] patch 03 (8 February 2019) was employed, with a physics list based on the hadronic class FTFP_BERT_HP and electromagnetic physics classes G4EmStandardPhysics and G4EmExtraPhysics, using a procedure similar to that reported in Ref. [32]. The resulting model was used to simulate the gamma-ray response by modeling the gamma-ray interactions in the detector and tracking the secondary electrons and scintillation photons [28] that they produced.

Measurement
The calibration sources were systematically placed in front of the NE 213A detector which was aligned so that the cylindrical symmetry axis of the detector pointed at the source. Sources with an activity below 1 MBq ( 22 Na, 137 Cs, 232 Th) were placed at a distance of 45 cm from the face of the unshielded detector, while the distance was increased to 200 cm for sources with an activity above 1 MBq ( 60 Co, AmBe). Hydrogen-rich materials were removed from the vicinity of the setup to minimize the production of 2.22 MeV gamma-rays from neutron capture during the AmBe irradiations. A typical run time was 1 hour. Prior to data collection, background was investigated using a 1.5 inch LaBr 3 (Ce) gamma-ray detector. Gamma-rays from the de-excitations of 40 K (1.46 MeV) and 208 Tl (2.61 MeV, 583 keV, 510 keV) were observed. As count rates were on the order of a few 100 Hz, deadtime was very low, so that the room background could be subtracted from the source measurement after a straightforward realtime normalization.             light-yield gradient employed in the GEANT4 simulation and the ∼80% relative gain, ∼35%

Results
of the scintillation light produced by a gamma-ray reaches the photocathode.

Summary and Discussion
A scintillation light-yield calibration of an NE 213A organic liquid scintillator detector ( Fig. 1) has been performed using single-energy and double-energy gamma-ray sources.
An event-by-event waveform-digitization (Fig. 2) of the scintillation signals resulted in measured Compton-edge distributions. Interpretation of the Compton-edge distributions used a GEANT4-based simulation which models the interactions of ionizing radiation and the transport of scintillation photons produced along particle tracks. Simulations employed a 1700 photon per MeV ee light-yield gradient, tuned to the data using relative PMT gain as a scaling parameter, and matched to the Compton edges by applying an additional smearing (Fig. 3). The relative gain function determined in this manner was linear over the ∼5 MeV ee range considered at a photoelectron multiplication of (3.27 ± 0.07) · 10 6 , corresponding to a (1388 ± 31) scintillation photon per MeV ee light-yield gradient (Fig. 4).
Charge distributions were determined as a function of electron energy by enforcing very strict cuts in the simulation around the upper edge of the recoiling electron energy-loss spectrum as well as considering the well-established prescriptions of Knox and Miller and Flynn et al. (Fig. 5) These restricted simulated distributions facilitated an evaluation of the intrinsic detector resolution, which was determined to be ∼18% at ∼1 MeV ee and to fall o ∼1/ √ E CE (Fig. 6). An advantage of the simulation approach over the prescriptions is that it allows for the unfolding of spectra from radioactive sources emitting more than one gamma-ray, even if the energy separation of the gamma-rays is small. To demonstrate this advantage, the entire simulation and analysis procedure was successfully repeated for two such sources, 22 Na (759 keV gamma-ray separation) and 60 Co (160 keV gamma-ray separation) (Fig. 7). Linear light-output calibrations were then determined (Fig. 8)