Nuclear Inst. and Methods in Physics Research, A Application of compact laser-driven accelerator X-ray sources for industrial imaging

X-rays generated by betatron oscillations of electrons in a laser-driven plasma accelerator were charac- terised and applied to imaging industrial samples. With a 125TW laser, a low divergence beam with 5 . 2 ± 1 . 7 × 10 7 photons mrad −2 per pulse was produced with a synchrotron spectrum with a critical energy of 14 . 6 ± 1 . 3keV . Radiographs were obtained of a metrology test sample, battery electrodes, and a damage site in a composite material. These results demonstrate the suitability of the source for non-destructive evaluation applications. The potential for industrial implementation of plasma accelerators is discussed.


Introduction
Advances in industrial methods, such as additive manufacturing (AM), are enabling the fabrication of better and more complicated products than are achievable with traditional manufacturing. Environmental sustainability is a major incentive to develop less wasteful processes, new materials, and energy storage solutions. For example, many aircraft and automotive components are now produced using lightweight fibre reinforced composites to improve fuel efficiency. In parallel, the more widespread adoption of electric vehicles is driving investment and innovation in battery technologies. Rapid growth in these sectors means that in some cases demand is outstripping supply and there is a need for industry to increase productivity. Progress in manufacturing needs to be accompanied by improvements in the product inspection tools employed for metrology and quality control. X-ray computed tomography (XCT) is a powerful technique because it allows multiple GeV within centimetres [2] and emitting on-axis betatron radiation as they oscillate [3,4]. For electrons with energy 2 , the characteristics of the radiation are determined by the wiggler strength parameter = , where is the amplitude and = 2 ∕ is the wavenumber associated with betatron oscillations of wavelength . For self-injected electron bunches, where the accelerated electrons are swept up from the plasma itself, the parameter obeys > 1 and the radiation has the form of a broad band on-axis synchrotron spectrum [5,6] with a peak close to the critical energy The number of photons can reach 10 10 photons per pulse in a beam with 10 mrad divergence and critical energies in the range 10-50 keV [7,8], 10% stability [9] and pulse lengths of <100 fs [10,11].
The small size of the laser-driven accelerator is a key advantage compared to conventional synchrotron light sources and the technology has the potential to be used for a broad range of applications [1]. Tomographic imaging of biological samples [7,8,12,13], time resolved radiography of high energy density plasma [14] and imaging of AM objects [15] and complex microstructures [16] has been reported. The short pulse duration enables radiographic snapshots of fast moving parts with no motion blur. In addition, the source is suitable for x-ray absorption spectroscopy with exceptional time resolution [11]. Because the x-ray source size is of order of 2 ≈ 1 μm, high resolution imaging can be conducted with high x-ray flux, avoiding the trade off between source size and power encountered with conventional x-ray machines. Furthermore, within 1 m the beam has a transverse coherence length of 10's μm meaning that phase enhancement can be obtained with a compact imaging arrangement [17]. Phase contrast provides superior image quality for low density objects that only weakly attenuate x-rays and better distinction between items made of similar materials [18]. This technique has been demonstrated on biological samples using plasma based accelerators [17,19].
In this work, we demonstrate the feasibility of using laser-betatron radiation for high resolution NDE of industrially relevant components. The laser generates x-ray energies in the 10's keV range, ideal for low density polymer and carbon-fibre composite materials. We characterise the properties of the x-ray beams and present example images of samples relevant to three particular areas of importance to industry: accurate metrology, battery development and composite manufacturing. We will discuss the potential of these sources for high frame rate industrial XCT as the underpinning laser technology develops to higher repetition rate operation in the next few years.

Imaging set-up and x-ray source
The beam line apparatus used to produce the x-ray beam can be seen in Fig. 1. The Gemini Ti:sapphire laser, providing pulses of duration 49 ± 3 fs and 7.2 ± 0.4 J on target, was focused with an off-axis f /40 parabola into a gas cell inside a vacuum chamber. Orthogonal to the main beam, a second laser pulse was used to measure the plasma electron density interferometrically. After the cell, the laser beam, electron beam and x-ray beam transited a 25 μm polyimide tape, which functioned as a plasma mirror deflecting most of the undepleted laser energy [20]. A 13 μm aluminium foil protected the sample from any remaining laser light and also acted to filter out 93% of the energy of a 14.6 keV-synchrotron spectrum below the k-edge (1.6 keV). Electrons were deflected off axis with a permanent magnetic dipole and observed on a scintillating screen to measure their energy spectrum. The x-ray beam exited the vacuum chamber through a 250 μm polyimide window and propagated through 2050 mm of air onto an indirect detection xray camera. This consisted of 150 μm thick caesium-iodide scintillator fibre-coupled to a 2048 × 2048 pixel CCD (Andor iKon-L 936) of pixel size 13.6 μm.
The gas cell length and gas pressure were scanned to determine the optimal parameters for generating stable x-ray beams. The most reproducible beams with the highest x-ray flux were measured at a plasma density of 4.4 ± 0.4 × 10 18 cm −3 with an average peak energy of the electrons of 435±7 MeV. The corresponding plasma wave wavelength implies a maximal x-ray pulse duration ∕ ≈ 50 fs. The x-ray spectrum was measured using 12 materials in a filter pack configuration [21]. By comparing the transmission of the different materials relative to each other and finding the best fit synchrotron spectrum, the critical energy Eq. (1) was found to be 14.4 ± 1.4 keV for photons in the range 0.1 − 300 keV. The emitted x-ray flux from the source was determined to be 5.2 ± 1.7 × 10 7 photons mrad −2 per pulse and the transverse source  size was inferred from the betatron parameters to be 2 = 2.3 ± 0.3 μm (Eq. (1)).
The divergence of the x-ray beam (∼ 10 mrad) allows geometrical magnification. The beam line included three sample stages located at different distances in order to vary the x-ray field-of-view, as indicated in Fig. 1. Samples outside vacuum (position 3) could be positioned anywhere from the rear wall of the chamber to directly in front of the camera. Inside the vacuum chamber, a stage was placed close to the exit window (position 2) yielding a magnification M = 2.2, well suited for cm-scale objects. The highest magnification was M = 10.2 with the sample placed 370 mm from the source (position 1). Because this stage was located before the electron spectrometer magnet, an additional kicker magnet was inserted in this case to prevent the electron beam from striking the sample. The magnification was measured at position 1 using a JIMA RT RC-02 resolution target and in configuration 2 using a gold foil with a pattern of horizontal and vertical apertures. The resulting radiographs are shown in Fig. 1.
Radiographs of industrial samples taken with the laser-betatron source are presented in Section 3. To provide context for each particular application, we also display 3D reconstructions of the same objects obtained using the commercial XCT scanners currently being used. In this work, it is not our intention to directly compare image quality between the x-ray sources. Rather, our aim is to demonstrate the capability for high quality imaging of such objects, with a view to offering advanced XCT using this technology in the future. The laserbetatron images in Figs. 2 and 3 were obtained with the samples placed outside the vacuum chamber with magnification of M = 1.5; the sample shown in Fig. 4 was imaged at position 2 (M = 2.2). The image resolution was detector-limited by the point spread function of the scintillator to (78 μm / M ) [8]. For each acquisition 10 single xray images were integrated and an average flat field beam x-ray profile was background subtracted as well as a darkfield image. Clusters of hot pixels, due to bremsstrahlung, were removed using an adapted median filter, taking into account that bremsstrahlung affects multiple pixels.

Dimensional XCT
Accurate metrology is a critical aspect of the manufacturing process to confirm that industrial components meet the required tolerances. This is particularly important for parts produced by AM, which have complicated internal structure that must be imaged in a non-destructive way. As an emerging technology, there is a need for new approaches to quality control and the definition of industry standards. 4 Typical lab-based x-ray machines consisting of a cone beam have a plethora of measurement uncertainties in the scanning procedure. These include variations in operator selected parameters; geometric alignment of the system from the source, detector, and manipulator; and the environment [22]. Post-acquisition, the typical Feldkamp algorithm (FDK) [23] reconstruction introduces further error including approximations from the cone-beam. In an effort to assess the measurement uncertainty in industrial XCT, there have been studies to understand the impact of various parameters and to evaluate performance [24][25][26][27]. Results from the entire process carried out on different machines in different locations have also been compared to assess consistency, as reported in numerous round robin tests [28][29][30].
An image using the laser-betatron source of a plastic test object produced for performance verification is shown in Fig. 2(b). These samples are made with a number of representative geometries designed to test the dimensional measurement accuracy of x-ray imaging systems. Dimensions such as sphere diameter, internal/external diameters of cylinders, centre-to-centre and plane-to-plane distances are typical examples of features that are assessed. In Fig. 2(a), a tomographic reconstruction of the sample is shown where these distances are evident in a 3D context. The accuracy and uncertainty of such a measurement is intrinsically linked to the quantity and quality of the data that can be captured. The radiograph displays exceptional sharpness with minimal noise due to the small source size and high photon count. This level of fidelity could potentially produce an unrivalled maximal permissible error across all measurements with sufficient radiographs and a high level of geometric calibration. The near-parallelism of the laser-betatron x-ray beam is beneficial as this limits imaging artefacts arising from cone beams. Additionally, the ability to tune the x-ray spectrum to higher energies would allow highly penetrating XCT of AM products made with dense materials such as super-alloys used in the automotive and aerospace sectors.

Battery technologies
Better understanding of electrochemical processes is essential to guide innovative product design enabling the transition to low-carbon technologies. 5 The manufacture of Li-batteries is in a constant state of development with new chemistries regularly being trialled, with assembly lines eventually needing to adapt. New processes are rarely right-first-time and can result in defects within the cell. While the manufacturing process is iterated, identification of these problems is paramount not only to maintain optimal performance but also for safety [31]. XCT is being exploited to investigate these issues but with manufacturing lines creating 6-10 cells per minute, a high speed solution is required if every cell is to be inspected.
The laser-betatron source is well suited to provide rapid NDE of battery components, as shown by the example image of a pouch cell in Fig. 3(b). Pouch cells consist of a gel layer, alternating between cathode and anode layers. The electrodes must be sufficiently separated with no contact between them to avoid shorting, and can also suffer from other manufacturing defects such as delamination. These individual layers 5 gov.uk/government/publications/clean-growth-strategy can be observed in a full tomographic reconstruction such as the one shown in Fig. 3(a). Another potential site for quality issues is the tab area visible at the top of the battery. This region is checked as part of a typical inspection process because poor welding of the tabs to the anode and cathode can result in a defective cell. The radiograph, centred on the tab area, highlights the quality achievable with the laser-betatron source, with phase enhancement aiding the distinction between components. For electrodes composed of weakly absorbing materials, such as graphite, phase imaging is necessary to achieve sufficient contrast [32].
Improvements in laser repetition rate will also enable operando XCT and time-resolved x-ray absorption spectroscopy (XAS) to examine the activity of battery cells during charge cycles. These methods allow visualisation of the changes in microstructure [32,33] and mapping of lithium concentration [34,35], important for understanding battery degradation. However, because of the very high average flux required, such studies can only be currently carried out using synchrotron light sources. A key advantage of the laser-produced x-ray pulse is its short duration and inherent synchronisation with the drive laser. The capability to measure ultrafast transitions would be particularly useful for pump-probe investigations on the femtosecond scale, for example studies of new photovoltaic materials [36]. Laser-betatron XAS measurements with <100 fs temporal resolution have already been reported [11,37], demonstrating the suitability of the source for this application.

Composite manufacturing
To meet the huge rise in demand for composite products over a broad range of industries, it is essential to increase rates and efficiency in composite manufacturing. 6 NDE is commonly used to assess design features, test manufacturing methods, and inspect the effects of mechanical testing and damage that has occurred over the service lifetime of a part [38]. Detailed XCT information, along with finite element modelling [39], can improve efficiency through better design, and the definition of acceptance tolerances.
A major concern is the evolution of defects during manufacture. To be able to understand their formation, it is important to conduct in-process XCT to track individual features over time. This was highlighted by a study where samples were imaged with fast CT scans (7 min acquisition time) throughout the cure process inside the cabinet of an industrial scanner (Nikon XTH-320) [40]. The results showed that, although an initial gap between carbon fibre tows filled as the sample heated, small voids emerged that would affect the properties of a finished product. This phenomenon was not detected in other studies using multiple partially cured samples [41], likely because of unavoidable variation in samples [42,43]. However, the long scan times necessary for high resolution and low noise images present difficulties for studies of fast changing samples. This is particularly true where the research requires distinguishing carbon fibre reinforcements from carbon-based polymer resins. Although this can be achieved with industrial scanners, scan times of order 4 h preclude this type of in-process XCT study. Synchrotron imaging can be adopted for uninterrupted in-situ studies, including mechanical tests of cured composite [44] and compression [45] and debulk [46] of uncured composite, but this is not a practical solution for regular use by industrial composite manufacturers because of availability and cost constraints.
These problems could be overcome by employing laser-driven x-ray sources. As an example, we show in Fig. 4(b) a radiograph, obtained with the laser-betatron source, of a composite test sample. This was a semi-circular prism cut from a cured cylinder made up from an array of small diameter unidirectional IM7 carbon fibre reinforced epoxy resin matrix rods, embedded in a second epoxy resin. XCT is used to assess the effect of a kink-band failure, initiated by impact and propagated by compressive end loading. This is visible in the tomographic reconstruction shown in Fig. 4(a). The radiograph exhibits good contrast between the carbon fibre and the resin, highlighting the benefit of the x-ray phase enhancement produced with the laser-betatron source. The layers visible in the image are carbon fibre tows that typically have a diameter of order 200 μm. Development of this technology to deliver fast scanning at high resolution would address larger scale challenges, such as imaging consolidation in corners, or inspecting full scale parts while applying heat, vacuum and/or pressure to the part. In this way, quality assurance and control could be performed before the heat and pressure is applied to cure the resin, reducing scrap, and saving energy, cost and time.

Discussion
The demands of state-of-the-art industrial NDE, such as high resolution, fast scan speed and element specific analysis, are difficult to meet with existing x-ray technology. Laser-based radiation sources produced with a plasma accelerator have ideal properties for addressing these challenges. Recent improvements in reliability and repetition rate of high power lasers make it feasible to produce these compact x-ray devices for commercial deployment in industrial environments. At the laser energy used here of 7 J (125 TW laser power), commercial products are available operating at 5 Hz (e.g. Thales Quark200; Amplitude Pulsar) and would increase the average x-ray flux to above 10 11 ph s −1 . Using diode-pumped solid state technology the repetition rate could be scaled up further [47][48][49][50]. Improvements in x-ray beam consistency have been demonstrated by reducing pulse-to-pulse fluctuations in the laser and gas target performance [51] and through studies of the stability of electron injection mechanisms into the accelerator [9]. Using higher power drive lasers [6,52], the photon energy should be enhanced to the 100's keV to MeV range that will be needed for inspection of large battery packs and full-scale composite products such as wind turbines, although this also increases the size and cost of the system. The laser requirements might be relaxed by adapting target design to increase the betatron photon energy [53].
One of the benefits of adopting laser-driven technology is the ability to drive different, synchronised secondary sources with the same laser. Moderate adaptations to the system can produce other particles that can be utilised for complementary sub-surface inspection techniques such as neutron imaging [54] and positron annihilation lifetime spectroscopy [55]. Therefore, a single machine could be used to deliver a powerful inspection capability with multi-modal, multi-scale imaging.
To be viable as a commercial product, considerable R & D effort is needed to ensure robust performance over long periods and to minimise the size and cost of the system. Product design could be tailored to meet the specific demands of the end-user, with a range of systems dependent on space and budget constraints. The compact design of modern high power lasers means that it would also be possible to integrate them into portable devices [56]. An important consideration is the radiation shielding necessary for the electron beam. In an industrial environment this could be constructed in a similar way to free-electron laser and synchrotron facilities where the electron beam is deflected into a heavily shielded beam dump in the ground while the x-ray beam propagates into an end-station with relatively light shielding.
Although a product based on this technology would be more costly and complex than conventional x-ray machines, it would offer advanced NDE tools that are currently not available in industrial or lab-based settings. In particular, micron-scale resolution tomography with fast scan speed, and ultrafast x-ray absorption spectroscopy could be applied to in situ inspection and product development. For example, using a laser operating at 100 Hz, high frame rate imaging of metrescale objects should be possible at less than 1 min per square metre. Hard x-ray chemical tomography could also be conducted following the methods used on XAS synchrotron beamlines [57].

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.