Experimental results of the first ANET resolution campaign with the compact neutron collimator

Abstract

This paper presents the first determination of the improvement in beam divergence when including the ANET Compact Neutron Collimator on an existing thermal neutron beam, obtained with a measuring campaign at the LENA Mark-II TRIGA reactor in Pavia. This novel collimator consists of a sequence of collimating and absorbing channels organised in a chessboard-like geometry. It has a scalable structure both in length and in the field of view. It is characterized by an elevated collimation power within a limited length. Its scalability and compactness are added values with respect to traditional collimating system. The prototype tested in this article is composed of 4 concatenated stages, each 100 mm long, with a channel width of 2.5 mm, delivering a nominal L/D factor of 160. This measuring campaign illustrates the use of the ANET collimator and its potential application in ne

Keywords

Neutron imaging neutron collimation compact neutron collimator

1. Introduction

This paper is dedicated to the study and characterisation of a Compact Neutron Collimator (CNC) developed within the framework of the ANET project [2] and is an extension of the first experimental campaign results article [5]. The measurement campaign has been performed at the LENA 250 kW Mark-II TRIGA reactor in Pavia (Italy). The thermal neutron beam provided by the LENA facility is poorly collimated and thus it can be used as a test-bench to verify the ANET CNC collimation performances and its possible application for neutron imaging. To measure the image resolution, two reference test objects have been used, i.e. a gadolinium Siemens star and a gadolinium bar pattern developed at the Paul Scherrer Institute [4]. These two devices allow to determine the image resolution in a range between 25 μm and 1000 μm. In the following sections the measurement set-up and methodology are described and a demonstration of the impact of the ANET CNC in terms of image resolution is illustrated.

2. Experimental set-up and methodology

The experimental set-up used is composed of 4 main stages (Fig. 1): the LENA thermal neutron source, the ANET CNC coupled with a Physik Instrumente Stewart platform, the test object and a commercial neutron camera ensemble [1,3–5].

Fig. 1.

Schematics of the neutron imaging setup with the ANET collimator.

Fig. 2.

Geometrical representation of a single collimation channel. For the definition of the single quantities refer to the text.

In Fig. 2 a single collimation channel is shown and the relevant variables are defined:

D: Diameter of the collimator channel.

L: Length of the collimator channel.

x: Distance of the object from the collimator.

l: Distance of the object from the scintillator.

$σ_{pen}$ : geometrical resolution due to the penumbra blurring.

The ANET CNC, as explained in ref. [5], is a multi-channel device with air collimation channels alternated to highly absorbing channels made of a mixture of $B_{4} C$ and PLA. It is operated following a dynamic acquisition protocol to deliver a clean neutron radiography, whithout the possible artefacts due to the discrete structure of the collimator. To this end, the CNC is mounted on a Stewart platform, capable of moving it through 6 independent degrees of freedom with a translational precision of 10 μ m and a rotational precision of 12.5 μrad. These levels of accuracy are required during the preliminary alignment procedure and the main image acquisition, as explained in the following section.

The sample object, in front of the collimator, is moved in and out of the field of view by mean of a small linear stage. The image obtained when the sample object is out of the beam defines the so-called “Open Beam Image”.

The neutron camera is a commercial device [3], composed by a 14-bit Sony CCD coupled to a 25 mm f/1.4 lens and a 400 μm scintillator screen made of ZnS(Cu) mixed with ⁶LiF at a mass rate 2:1. The camera is cooled at 20 degrees below the room temperature to minimise the electronic noise, measured at 340 pixel counts (out of the 14 bit of dynamic range of the camera) during 900 s of integration time. The set-up offers a field of view of 100 × 100 mm²

Any neutron radiography (I) should be properly normalised using an open beam image ( $I_{OB}$ ) and a dark field image ( $I_{DF}$ ) under the same exposure conditions. The quantities I, $I_{OB}$ and $I_{DF}$ are expressed in pixel counts. The normalized neutron radiography $F_{rad}$ is defined as follows: $\begin{matrix} (1) & F_{rad} = \frac{I - I_{DF}}{I_{OB} - I_{DF}} \end{matrix}$

2.1. The CNC alignment procedure

The ANET CNC prototype has a field of view of 50 mm × 50 mm and a nominal L/D factor of 160. It is 400 mm long, and each collimation channel is 2.5 mm wide, corresponding to a maximum neutron angular divergence of 0.36 degree (or 6 · 10⁻³ rad). Given these constraints, the alignment of the collimator w.r.t. the beam axis is crucial.

The PI Stewart Platform specifications allow for the angular fine tuning of the collimator alignment with the required precision. The static image of the collimator before and after the alignment is shown in Fig. 3. The image is taken with the neutron camera described in the previous section. The chess-board image reflects the collimator structure. The image after the alignment shows a good level of uniformity in the inner part while the effects of the conical shape of the Pavia LENA beam are more visible at the edges and the corners. This results in an effective field of view with 40 mm diameter that is shown as a circle in Fig. 3 (right).

Fig. 3.

Neutron image of the collimator before the alignment (left) and after the alignment (right). The red circle indicates the effective field of view.

2.2. The CNC method of operation

In order to properly use the ANET CNC avoiding the artefacts induced by the chessboard structure in the final image a dynamic image acquisition approach has been developed. In order to optimise the image quality, whilst minimising the data processing, a long exposure radiography with the collimator moving continuously is taken. The time of exposure is chosen to maximize, in the region of interest, the CCD pixel counts without reaching pixel saturation. This, of course, depends on the neutron source intensity. In the Pavia conditions, the exposure time without the collimator was set to 60 s, and 900 s when the ANET CNC was included in the set-up. Along this same time period, the Stewart Platform moved the CNC following a specific pattern confined within 5 × 5 mm² area, equivalent to two air channels and two absorbing rods. Regarding the determination of the optimal pattern, extensive studies have been performed, following different shapes (i.e. square, triangular, random,…) and different sizes. The “saw” profile shown in Fig. 4 turned out to be the optimal one. In Fig. 4, ω is equivalent to the scan range (5 mm) divided by half of the number of divisions. This results in a value of 47.6 μm, smaller than the size of the projected pixel on the scintillator, which is 47.9 μm. The overall result (Fig. 4 right) is a smooth, pattern-less image. The instrument not only follows the pattern, but the process also includes certain hardware perks to reduce the systematic errors induced by the micrometric movements:

The movement speed is corrected at each vertex in order to compensate for the statistical delay created by the change of velocity.

Since the clock controlling the motors is at the order of the milisecond, there is an “inclusion radius” on the vertexes, coherent with the size of the motor step given the clock.

The movement at the vertexes is round, not sudden, in order to smooth the path and reduce the impact on the change of velocity.

Those three elements, once applied, greatly reduce the dis-homogeneity on the image generated by the movement of the CNC. The stability of the system can be appreaciated through the OB measurement. The uniformity of the OB image is an important factor as it determines the reproducibility and quality of the normalisation procedure described in equation (1). This can be evaluated by calculating the ratio of different OB images and extracting the intensity distribution of the pixels within the field of view. The resulting histogram in Fig. 5 shows a mean of 0.999 with a standard deviation of 0.011, implying a uniformity and repeatability within

\pm 1 %

Fig. 4.

Neutron image of the collimator (left). Example of dynamic pattern (centre). The axis represent vertical and horizontal directions in space, respectively. The pattern has also been tested orthogonal to what is shown in the example figure, with identical results. Image of the open beam with the dynamic pattern (right). The red circle indicates the effective field of view.

Fig. 5.

Histogram of the ratio of two open beam images.

3. Improvement in resolution due to the ANET CNC

To evaluate the system spatial resolution, a set of Siemens star and bar pattern has been used [4], evaluated through optical means. These instruments cover a resolution ranges of 25–250 μm and 50–1000 μm respectively. The instrumental error on each device has been considered as half a step from each resolution indicator.

Keeping the collimator and the scintillator at a distance of 100 mm from each other, different measurements were taken with the Siemens star and the bar pattern at variable distances from the scintillator. The measurements were repeated with and without the ANET collimator. The impact of the ANET collimator can be appreciated in Fig. 6. The radiography images taken with the collimator show an evident improvement in the spatial resolution with respect to the same image taken without the ANET CNC.

Fig. 6.

Table showing the normalised measurements of the Siemens star with and without the collimator at 40 mm and 20 mm from the scintillator respectively.

To be more quantitative and to describe the behaviour of the spatial resolution as a function of the distance l from the scintillator, it is worth to extract the relation between the geometrical resolution $σ_{pen}$ and the quantities x and l defined in Fig. 2. From geometrical considerations (see Fig. 2) it can be easily derived: $\begin{matrix} (2) & σ_{pen} = \frac{l D}{L + x} \end{matrix}$ The total resolution $σ_{total}$ can be calculated through [4]: $\begin{matrix} (3) & σ_{total} = \sqrt{σ_{pen}^{2} + σ_{\det}^{2}} \end{matrix}$ Where $σ_{\det}$ is the contribution due to the neutron camera detector ensemble, which is related to the inverse of the Nyquist frequency, defined as the minimum measurable separation between two lines and numerically equivalent to the double of the pixel size projected at the scintillator plane. In the case of the neutron camera used for this experiment $σ_{\det} =$ 96 μm.

The total resolution appreciable in a measurement is then a combination of the beam collimation quality and the contribution of the detecting system. The comparison between the total resolution measured at LENA in Pavia with and without the ANET collimator at various distances between the test object and the scintillator is shown in Fig. 7.

Fig. 7.

Comparison between the resolution obtained with and without the ANET collimator as a function of the distance of the Siemens star from the scintillator.

The image resolution improvement due to the ANET collimator is mighty and it becomes more relevant with increasing distance. The total resolution values with ANET CNC range between 100 and 300 microns.

The measured points behaviour can be described by the simple geometrical model derived from equation (3) The error associated with the model is half a pixel, as the resolution true value may vary within this range. The comparison is shown in Fig. 8: the theoretical curve and the data points are in good agreement.

Fig. 8.

Comparison between the resolution obtained with the ANET collimator and that expected from the theoretical model.

4. Conclusions

This paper shows the results obtained with the ANET CNC compact collimator prototype during the measuring campaign at the LENA Mark-II TRIGA reactor, during summer 2021. The main objective of the campaign was to validate the theoretical performance of the device on an experimental set-up.

The measurement serves as a proof of concept that the ANET multi-channel 2D collimator employing a dynamic pattern acquisition technique can serve as a core part of an imaging setup.

The resolution measurements follow the theoretical expectations and demonstrates the ANET CNC capability to improve beam lines with limited collimation power.

The inclusion of the ANET CNC on the LENA channel-B has rendered an uncollimated beam capable of performing neutron imaging to an acceptable degree. This brings new uses to the facility and opens the door to future developments of neutron radiography even in small and medium-sized sources not originally projected for this kinds of applications.

Future campaigns with higher intensities and different neutron beam facilities will be useful to demonstrate the ANET collimator portability and performances.

References

https://www.physikinstrumente.com/en/technology/parallel-kinematics/multi-axis-positioners/, visited on 10-August-2021.

Bedogni

et al., Design of a novel compact neutron collimator, Journal of Instrumentation16(08) (2021), P08055. doi:10.1088/1748-0221/16/08/P08055.

Hewat

A.W.

, Inexpensive neutron imaging cameras using CCDs for astronomy, Physics Procedia69 (2015), 185–188. doi:10.1016/j.phpro.2015.07.026.

Kaestner

A.P.

et al., Samples to determine the resolution of neutron radiography and tomography, Physics Procedia88 (2017), 258–265. doi:10.1016/j.phpro.2017.06.036.

Sans-Planell

et al., First results with the ANET compact thermal neutron collimator, Journal of Instrumentation16(11) (2021), P11025. doi:10.1088/1748-0221/16/11/P11025.