A novel energy resolved neutron imaging detector based on TPX3Cam for the CSNS
Jianqing Yang, Jianrong Zhou, Xingfen Jiang, Jinhao Tan, Lianjun Zhang, Jianjin Zhou, Xiaojuan Zhou, Wenqin Yang, Yuanguang Xia, Jie Chen, XinLi Sun, Quanhu Zhang, Zhijia Sun, Yuanbo Chen
AA novel energy resolved neutron imaging detector based onTPX3Cam for the CSNS
Jianqing Yang , Jianrong Zhou , Xingfen Jiang , Jinhao Tan , Lianjun Zhang ,Jianjin Zhou , Xiaojuan Zhou ,Wenqin Yang ,Yuanguang Xia , Jie Chen , XinLiSun , Quanhu Zhang , Zhijia Sun , Yuanbo Chen
1. Xi’an Research Institute of Hi-Tech, 710025, Xian,China.2. Spallation Neutron Source Science Center, Dongguan, 523803, Guangdong, China;3. State Key Laboratory of Particle Detection and Electronics, Institute of High Energy Physics, Chinese Academy of Sciences,Beijing, 100049, China;4. University of Chinese Academy of Sciences, Beijing 100049, China5. Harbin Engineering University, Harbin, Heilongjiang, China, 150000
Abstract:
The China Spallation Neutron Source (CSNS) operates in pulsed modeand has a high neutron flux. This provides opportunities for energy resolved neutronimaging by using the TOF (Time Of Flight) approach. An Energy resolved neutronimaging instrument (ERNI) is being built at the CSNS but significant challenges forthe detector persist because it simultaneously requires a spatial resolution of less than100 μm, as well as a microsecond-scale timing resolution. This study constructs aprototype of an energy resolved neutron imaging detector based on the fast opticalcamera, TPX3Cam coupled with an image intensifier. To evaluate its performance, aseries of proof of principle experiments were performed in the BL20 at the CSNS tomeasure the spatial resolution and the neutron wavelength spectrum, and performneutron imaging with sliced wavelengths and Bragg edge imaging of the steel sample.A spatial resolution of 57 μm was obtained for neutron imaging by using thecentroiding algorithm, the timing resolution was on the microsecond scale and themeasured wavelength spectrum was identical to that measured by a beam monitor. Inaddition, any wavelengths can be selected for the neutron imaging of the given object,and the detector can be used for Bragg edge imaging. The results show that ourdetector has good performances and can satisfy the requirements of ERNI fordetectors at the CSNS.Key words: Neutron imaging, Energy resolution, TOF, Spatial resolution, Bragg edge *Corresponding author : Jianrong Zhou, [email protected], Dongguan, China**Corresponding author : Quanhu Zhang, [email protected], Xian, China***Corresponding author : Zhijia Sun, [email protected], Beijing, China . Introduction
As a visible and non-destructive method of inspection, traditional neutronimaging evaluates the attenuation of a polychromatic neutron beam through a sample.However, the properties of the sample are integrated over the entire spectrum of thebeam. Recent years have witnessed the development of energy resolved neutronimaging, which means that neutron imaging can be performed using a spectrumlimited to a short energy band [1]. This is useful for analyzing energy dependence ofthe sample on neutrons through the contrast in neutron images of distinct energies,and the crystallographic information can be obtained using the Bragg edge imagingfor crystal samples [2]. In principle, there are two ways of conducting energy resolvedneutron imaging experiments. At the reactor sources, monochromator devicesincluding velocity selectors [3] and double crystal monochromators [4] are availableto select neutrons of specific energies. At the spallation sources, the TOF approach isa method for measuring the velocities of neutrons, i.e. energy or wavelengthaccording to the de Broglie wave equation based on their flight distances as well asthe TOF between the source pulse and the neutron arrival at the detector. Comparedwith reactor sources, the spallation sources have a significant advantage [5][6]because full energy scanning neutron imaging and Bragg edge imaging of the samplecan be implemented in one experiment by using the TOF approach to improve theexperimental efficiency. However, there are significant challenges for the detector inthis context due to the requirements of high timing and spatial resolutions. At present,the spallation sources in operation include the JSNS (Japan Spallation Neutron Source)in Japan, the ISIS in the UK, the SNS (Spallation Neutron Source) in USA and theCSNS in China. The energy resolved neutron imaging system, RADEN has beenconstructed at the JSNS of J-PARC. The detectors applied to it include the μNID,nGEM and the LiTA12 [7] whose timing resolutions are on the dozens or hundreds ofns scale. However, their spatial resolutions are over 100μm. The IMAT imaginginstrument has been installed in the ISIS [8]. The conceptual design of the energyresolved neutron imaging instrument, VENUS has been completed at the SNS and itis being constructed [9]. The MCP neutron counting detector based on Timepix3 chipreadout has been chosen as the imaging detector for the ISIS and SNS. It wasdeveloped by Tremsin A.S. from the University of California, Berkeley and can reacha sub-15 μm spatial resolution for both thermal and cold neutrons [10] and asub-microsecond timing resolution [11]. It is thus superior to the detectors applied tothe JSNS in terms of spatial resolution.The CSNS is a high-flux pulsed neutron source (25 Hz) that is mainly applied toneutron scattering, imaging, and other kinds of neutron science research [12]. TheERNI is currently being built at the CSNS with the goal of characterizing andanalyzing the 3D distribution of the microstructures, defects, stresses and evendynamic processes of materials and devices. For the ERNI, a detector needs to have amicrosecond-scale timing resolution and a spatial resolution of less than 100 μm atthe same time. The timing resolution enables the detector to distinguish neutronenergy by using the TOF approach and the high spatial resolution can be used tonalyze the internal microstructure of samples in a number of applications. TheTPX3Cam is a recently developed camera with a high timing resolution that isdifferent from a CCD camera as it is based on the Timepix3 chip. It can record thecoordinates (x,y), the time of arrival(TOA), with a granularity of 1.56ns, and timeover threshold (TOT) with a granularity of 25 ns for the pixels fired almost at thesame time [13]. The energy resolved neutron imaging detector can be realized byusing the standard CCD camera-based neutron imaging detector, in which theTPX3Cam replaces the CCD camera without a timing resolution. However, the imageintensifier needs to be coupled to the TPX3Cam in order to gain single photonsensitivity and overcome the high threshold of timepix3 chip. Compared with theMCP neutron counting detector, the TPX3Cam is protected from direct radiationdamage because a reflective mirror is used to deflect the scintillation light to thedirection vertical to that of the neutron beam. A high timing resolution can be realizedconsidering the time precision of the TPX3Cam and the high spatial resolution can beobtained by the centroiding alrorithm [14][15].This paper develops the prototype of an energy resolved neutron imagingdetector based on TPX3Cam. Proof of principle experiments were performed in theBL20 at the CSNS to evaluate its performance and the results show that it is a viableoption for use of the ERNI at the CSNS.
2. Experimental setup
A schematic diagram of the energy resolved neutron imaging detector is shownin Fig. 1. At the CSNS, a pulsed neutron beam was produced by the spallationreaction, with a repetition frequency of 25 Hz. The signal T0 from the acceleratorrepresents the time of generation of the pulsed neutron beam. It was used as thetrigger for TPX3Cam and was recorded for the calculation of the TOF. The pulsedneutrons transmitted through the sample and the transmitted neutrons were convertedinto light through the scintillation screen. The Timepix3 chip of TPX3Cam had a highdetection threshold, about 1000 photons per pixel hit according to the specification inthe manual of TPX3Cam. Thus, an image intensifier was needed to enhance thescintillation light induced by the neutrons. The scintillation light was deflected intothe optical lens by a reflective mirror, enhanced by the image intensifier, and finallyreached the Timepix3 chip. The coordinates (x,y), TOA, and TOT of each fired pixelwere measured simultaneously. The TOF was calculated according to T0 and TOA,and represented energy or wavelength of the neutron. igure 1 Schematic diagram of detector
Fig.2 shows that the detector was mainly composed of a darkroom, scintillationscreen, 45° reflecting mirror, Schneider lens, 18 mm Photonis image intensifier,TPX3Cam and a movable platform. To protect the detector from radiation damage,the boron–aluminum alloy plates were installed on the inner wall of the darkroom.The scintillation screens commonly used for neutron imaging included 50 μm thick LiF/ZnS and 10 μm thick GOS (Gd O S (Tb/ LiF)). Compared with LiF/ZnS (Cu),GOS has better performance in terms of spatial resolution for two reasons. First, it canbe made much thinner owing to higher cross-section of absorption of gadolinium.Secondly, the mean free path (MFP) of secondary particles created by GOS is muchsmaller than that generated by LiF/ZnS. Thus, the GOS scintillation screen was usedin the spatial resolution test to limit the light spot size to about 10 μm [16][17].However, the light output of LiF/ZnS was almost 100 times higher than that of GOS.Thus, a LiF/ZnS scintillation screen was used to reduce the experimental time whenthe wavelength spectrum of the neutrons was measured and Bragg edge imaging wascarried out. In addition, the timing resolution of the detector was mainly affected bythe scintillation screen and the TPX3Cam, and needed to be evaluated. The lightdecay times of the LiF/ZnS and GOS scintillation screen were on the microsecondscale[18-21]. CSNS has a high neutron flux, the detector is necessary to have a goodtiming resolution. The timing resolution of this detector is the minimum time intervalrequired to separate two subsequent neutron events and it is mainly affected by thescintillation screen and the dead time of single pixel for Timepix3 chip. To analyze theeffect of scintillation screen on this detector’s timing resolution, the maximum timedifference of fired pixels for a single neutron event can be obtained by calculating themaximum difference of measured TOA for these pixels.The results showed themaximum time difference was within sub-microsecond when LiF/ZnS or GOSscintillation screen was used. This is due to the fact that the threshold is high enoughfor Timepix3 chip to cut out the photons from longer tail of the decay curve. Inaddition, when two neutrons hit on the same position of the scintillation screen within1μs, they are not able to be distinguished because the dead time of single pixel was1μs for Timepix3 chip. So, the detector has a timing resolution of microsecond scale.Considering the time fluctuation caused by the moderator is on the order oficroseconds, the detector can meet the requirement at CSNS.
Figure 2 Setup of the detector
The peak wavelengths of the scintillation light were 530 nm for LiF/ZnS and549 nm for GOS. Thus, the Hi-QE Blue photocathode was chosen for the imageintensifier because it is compatible with the scintillation light wavelength. Meanwhile,the double MCPs were selected to increase the gain to about 90, 000 ph/ph and a fastP47 phosphor screen was chosen, with fast rise and decay times of 7 ns and 100 ns,respectively. In the measurements, 10 Gb optic fiber Ethernet was used for theTPX3Cam, and the maximum counting rate was 80 Mhits/s.
3. Wavelength spectrum using the TOF approach
To verify the capability of the detector in terms of the timing resolution, thewavelength spectrum was firstly measured at the BL20 with a decoupled poisonedhydrogen moderator using the TOF approach. The T0 signal needed to be connectedto TPX3Cam, and the trigger mode of TPX3Cam was set toPEXSTART_TIMERSTOP, which means that the rising edge opened the shutter and itclosed after the duration of exposure expired. The TOF spectrum of the empty beamwas measured in the BL20 at the CSNS, and was used to obtain the neutronwavelength spectrum. It was calibrated according to the detection efficiency of thescintillation screen for neutrons of different wavelengths. The calibrated spectrum ofthe wavelengths of the neutrons was compared with that measured by an He beammonitor (ORDELA 4562N) [22] under the condition that these two spectra werenormalized to the largest count. The results in Fig. 3 show that the spectral shapemeasured by the TPX3Cam detector was almost identical to that measured by thebeam monitor. igure 3 Neutron wavelength spectrum of CSNS BL20
4. Spatial resolution
Spatial resolution was defined as the minimum distance between light spots thatcan be distinguished. A high spatial resolution is a desirable feature of a neutronimaging detector for observing fine details of an object. For the traditional CCDcamera, the pixel size is usually less than 10 μm, or between 10 μm and 20 μm. TheTPX3Cam has a larger pixel size of 55 μm, but the centroiding algorithm can be usedto improve its spatial resolution because the TOA, TOT, and position coordinatescould be independently measured for all fired pixels. To test the centroiding algorithm,experiments on spatial resolution were carried out, including an optical test and aneutron beam test. There were two reasons for carrying out the optical spatialresolution experiments. First, the intrinsic spatial could be gotten by the optical test.Secondly, the experimental time allocated to the authors for the BL20 was short. Theoptical and neutron spatial resolutions were obtained when the isotopic light sourcewhich is a tritium tube was used, and the neutron beam experiment was carried outrespectively. For the isotopic light source, the principle of luminescence is that βdecay of the tritium gas will occur, the produced electrons excite the phosphor powdercoated on the inner wall of tritium tube, and the light will be produced. The reasonwhy the tritium tube was used is that the intensity of the produced light is very stablebecause the half-life of tritium is 12.43 years, and it is also very weak and suitable forthe intensifier.4.1 Centroiding algorithmIn the optical and neutron beam tests, the lens and image intensifier enlarged thelight spot from the isotopic light source or the scintillation light. Multiple pixels werefired within sub-microseconds for a given light spot, from either the isotopic lightsource or a neutron event. This can lead to inaccuracies in determining the position oflight or the neutron, and degrades the spatial resolution. However, the centroid of thelight spot could be reconstructed, and the spatial resolution would be improved usingthe centroiding algorithm, which included clustering and reconstruction. In theTPX3Cam data, the TOA, TOT, and the coordinates (x,y) were recorded for everyfired pixel. To implement the algorithm, all the fired pixels were sliced according tohe T0 trigger signal, and pixels in one period of T0 were regarded as a processed unitfor clustering and reconstruction. The principle of clustering was to determinewhether any pair of fired pixels were correlated in time and space or not using thetime of arrival and position coordinates. The two fired pixels were assigned the samecluster ID if they were adjacent to each other and the interval between them was lessthan one microsecond. Once all the pixels in a T0 period had been clustered,reconstruction would be done, whereby the centroids were calculated according to thecoordinates (x,y) and TOT of the fired pixels with the same cluster ID. The centroidswere regarded as the centers of initial light spots or the positions of neutron events.4.2 Optical spatial resolutionThe optical resolution tests were firstly carried out in the darkroom using theisotopic light source on the USAF (US Air Force) test patterns before the neutronimaging experiment to verify the validation of the centroiding algorithm and obtainthe intrinsic spatial resolution. The tests were performed at an optical magnification of2.6. To eliminate the non-uniformities of the light source intensity, the test patternsimage was normalized by the equivalent image with no USAF test pattern present inthe light source. The results of optical imaging are shown in Fig. 4 (a) and 4 (b). Toanalyze the spatial resolution, all the elements of group 4 were projected into thevertical direction (shown in Fig. 4 (c)). Element 4 of group 4 was resolvedcorresponding to 23 lp/mm (44 μm resolution). To improve the spatial resolutionfurther, the centroiding algorithm was used to process the results of optical imaging ofthe test patterns, and these results are shown in Fig. 5 (a) and Fig. 5 (b), where theresolution was significantly improved. Fig. 5 (c) shows that the element 5 of group 5was resolved and it corresponded to 57 lp/mm (18 μm resolution).
Figure 4 (a–b) Optical imaging of USAF test pattern before the centroiding algorithm (c)Projections of the Groups 4 before the centroiding algorithm, Element 4 of Group 4 is resolvedcorresponding to 23 lp/mm (44 μm resolution)igure 5 (a–b) Optical imaging of USAF test pattern after the centroiding algorithm (c)Projections of the Groups 5 after the centroiding algorithm. Element 6 of Group 5 is resolvedcorresponding to 57 lp/mm (18 μm resolution) igure 6 (a) Neutron imaging of Siemens-Star test object before the centroiding algorithm (b)MTF curve before the centroiding algorithm (c) Neutron imaging of Siemens-Star test objectafter the centroiding algorithm (d) MTF curve after centroiding algorithm.
The neutron images of the Siemens Star test object are shown in Fig. 6 (a), whichshows that the resolution was between 50 μm and 100 μm. There are 128 line pairs inthe Simens Star test object and the density of line pairs decreases with the increase ofthe radius from the center to the boundary of the test object. The light intensityvariation of the line pairs at the every radius of the test object can be fitted by the sinefunction. The corresponding modulation transfer function (MTF) of different linespair densities can be obtained by dividing the amplitude the fitted sine function by thebias level of the fitted sine function. To analyze the spatial resolution quantitatively,the MTF curve was calculated (shown in Fig. 6 (b)). When the MTF was equal to10%, the spatial resolution was 12 lp/mm (84 μm). Fig. 6 (c) shows the result of Fig.6 (a) after the centroiding algorithm. It is clear that the spatial resolution of this imagewas better than that before being processed by the centroiding algorithm. The MTFwas calculated and it is shown in Fig. 6 (d). When MTF was equal to 10%, the spatialresolution of neutron imaging was 18 lp/mm (57 μm), and was worse than the opticalspatial resolution. This can be attributed to two reasons. First, the neutron flux wasonly 10 n/cm ·s, and needed to be increased to improve the contrast ratio and spatialresolution of the neutron image. Secondly, the intrinsic spatial resolution of the GOSscintillation screen was 25 μm and it was a cause for the degradation of neutronspatial resolution compared to optical spatial resolution.
5. Neutron imaging of the total and sliced wavelengths for a CSNS Cd object
The peak cross-section of
Cd, constituting 12.2% of naturally occurring Cd,was ~20,000 barns for thermal neutrons [23]. There was a strong energy/wavelengthdependence in the transmission of the neutrons through cadmium. A CSNS patternobject of 35 × 60mm was manufactured by using a Cd plate to validate the neutronimaging of the sliced wavelengths. On the whole, the cross-section of neutronsthrough cadmium increased with the wavelength in the range of 0–4 Å. The neutronimages of total and sliced wavelengths are shown in Fig.7 from which it was clear thatthe image contrast depended on the chosen neutron energy/wavelength, and graduallyenhanced with the wavelength for the CSNS pattern object.
Figure 7 (a) CSNS Cd object, (b–f) Neutron images of CSNS Cd object at 0–4 Å, 0–1 Å, 1–2, 2–3 Å and 3–4 Å respectively
6. Bragg edge imaging of a steel sample
Neutrons of different wavelengths varied in transmitted intensity through thepolycrystalline samples due to diffraction. This led to a characteristic Bragg edgetransmission spectrum for them. A steel sample was placed on the incidence windowas shown in Fig. 8(a). It was a homogeneous polycrystalline samples. The results ofneutron imaging of the sample are shown in Fig. 8(b).
Figure 8 (a) Steel sample on incidence window (b) Neutron image of the steel sample(c) The Bragg edge transmission spectrum
In the TOF spectrum, the maximum time of flight was 40 ms and the number ofbins was set to 3000. Thus, the width of each bin was 13 μs. The wavelength spectraof the empty neutron beam and the transmitted neutrons were obtained according tothe measured TOF spectrum. Based on these two spectra, the transmission wavelengthspectrum of steel sample was calculated. The result is shown in Fig. 8 (c), where thecorresponding lattice planes are marked. The five measured bragg edges were 4.04,2.86, 2.32, 2.02 and 1.80 Å. The results were almost in accordance with those in Refs.[24-27]. This demonstrates that the detector can be used for the Bragg edge imaging. . Conclusion and outlook
This study constructed an energy resolved neutron imaging detector based on theTPX3Cam. The results verified the high timing resolution and spatial resolutions ofthis detector. It can be applied as an instrument detector for ERNI. In future work ofthe area, the experiments should be performed in the SANS (Small Angle NeutronScattering Instrument) which has a higher flux than those of BL20, to further improvethe spatial resolution neutron imaging. A GOS scintillation screen with a higherspatial resolution and an optical lens with a higher magnification than that employedhere should be used.5.
Acknowledgments
This work was supported by the National Key R&D Program of China (GrantNo. 2017YFA0403702), the National Natural Science Foundation of China (Grant No.U1832119, 11635012 and 11775243), Youth Innovation Promotion Association CAS,and Guangdong Basic and Applied Basic Research Foundation (Grant No.2019A1515110217).6.