Multi-element X-ray movie imaging with a visible-light CMOS camera

A visible-light camera was used to resolve X-ray fluorescence spectra. Following the installation of a micro-pinhole, simultaneous multi-element X-ray fluorescence movie imaging was conducted in a synchrotron facility.


Introduction
Movies are useful as a record of the overall progress of chemical reactions. Generally, a movie captured in visible light cannot reveal the elemental composition of a reaction. By contrast, X-ray fluorescence (XRF) can aid the identification of elements with quantitative information such as average concentration, and XRF imaging can reveal their spatial distribution. Therefore, a movie comprising a continuous series of XRF images can be a powerful tool for exploring multi-element reactions such as ion exchange, composition segregation, selective precipitation and formation of chemicalgradient materials (Zhao & Sakurai, 2017a).
Full-field XRF imaging is an efficient way to collect XRF images. The technique requires an X-ray area detector that also possesses spectroscopic capability. Thus far, several options exist including a pn charge-coupled device (CCD) (Strü der et al., 2001;Scharf et al., 2011) and single-photoncounting CCD (Romano et al., 2014(Romano et al., , 2016Zhao & Sakurai, 2017b). The present short communication indicates the costefficient alternative of a visible-light complementary metaloxide semiconductor (CMOS) camera used for direct detection of X-rays (Janesick, 2001;Holst & Lomheim, 2007). In previous work, it was found that the camera could resolve XRF spectra using laboratory X-ray sources (Zhao & Sakurai, 2017c). Here, the camera operation is performed in a synchrotron facility and its multi-element movie imaging capability with potential for high spatial resolution is illustrated.

Experimental
The visible-light CMOS camera (PCO.edge 5.5, PCO AG) used in the present work has 2560 Â 2160 pixels with a pixel ISSN 1600-5775 pitch of 6.5 mm Â 6.5 mm. The sensor was electrically cooled down to 5 C. For full-field X-ray imaging, the optical window of the camera was replaced with a 25 mm-thick beryllium X-ray window. A pinhole was drilled onto 50 mm-thick tungsten foil, which was then installed in front of the X-ray window. To measure the X-ray photon energies, the camera was operated in single-photon-counting mode, and all camera images were processed using an integrated-filtering method (Zhao & Sakurai, 2017c) for charge-sharing correction.
Experiments were conducted in the BL-14B Photon Factory of KEK, Japan. The primary X-ray beam passed through an Si(111) double-crystal monochromator and was collimated to a width of 1 mm and height of 10 mm. The energy used in the present work was 13.5 keV. In the imaging experiments, samples were positioned vertically with a glancing angle of approximately 5 between the primary X-ray beam and the sample surface (Fig. S1 of the supporting information).
Two additional modifications were necessary for applying the camera to synchrotron experiments. First, the top part of the camera housing was changed from aluminium to 2 mmthick brass for shielding against high-energy X-ray scattering. Second, the double-crystal monochromator was detuned to eliminate higher-order harmonics that contribute strongly to noise collected by the camera sensor.

Results and discussion
The spatial resolution of the full-field XRF imaging system was tested using a pinhole diameter of 10 mm. The distances from the pinhole to the target and camera sensor were 1.65 mm and 14 mm, respectively. The resolution target was a chromium pattern coated on silica glass [ Fig. 1(a)]. The total measurement time was 6 h. Subsequent full-field XRF analysis [ Fig. 1(b)] shows the spectral peaks of chromium from the target. XRF contamination of copper and zinc, which originated from the brass housing of the camera, was also observed. The XRF image of chromium [ Fig. 1(c)] indicates a spatial resolution of 15 mm or even better for the system.
To demonstrate multi-element X-ray movie imaging, the growth process of zinc dendrites during electro-deposition was recorded. The reaction occurred in a thin container with dimensions of 30 mm (W) Â 20 mm (L) Â 1 mm (D). The container was filled with 0.1 M ZnCl 2 solution, and two copper electrodes were immersed (Fig. S1). The distance from the upper tip of the cathode to the lower horizontal anode wire was 15 mm. The electric potential between them was 3 V. The front surface of the container was made of a 50 mm-thick polyester film to allow the X-rays to pass through. In this experiment, a 50 mm pinhole was used to collect a stronger XRF signal although this reduces the spatial resolution. The distances from the pinhole to the reaction cell and camera sensor were 8 mm and 14 mm, respectively. After connecting the circuit to initiate the reaction, the experiment continued for 1 h, as did the movie recording. In the full-field XRF spectra [ Fig. 2(a)], most events in the Zn K peak came from the reaction cell (not the brass shielding), and therefore the spectral imaging of Zn K showed the zinc distribution in the reaction. In this way, images of zinc were generated every 2 min, and a continuous movie was obtained [ Fig. 2(b)]. These XRF movie frames confirm the visual observation that zinc dendrites first appeared on the upper tip of the cathode and then gradually grew downward; they also show what visible light cannot: the zinc dispersed in the solution and was gradually exhausted.  In the spectra shown in Figs. 1(b) and 2(a), the small peak of Si K, originated from the Si-based camera sensor. The small peaks of Ti K and Fe K are assigned to trace environmental contaminants. Furthermore, in this work the XRF spectra have a slight spectral contamination of zinc and copper from the brass camera housing. In future, this contamination can be easily eliminated by optimizing the housing shape or coating other high-Z 'XRF-silent' elements. Meanwhile, the present work shows that the spatial resolution of full-field XRF imaging can reach 15 mm by using a pinhole of a sufficiently small size. When employing this simple approach, the XRF intensity collected by the pinhole becomes weaker as the pinhole becomes smaller. Consequently, a larger pinhole was adopted in the movie experiment as a compromise. To reduce this problem, a higher-flux synchrotron X-ray beam may be used to compensate the intensity loss, or the pinhole may be replaced by other X-ray imaging optics such as a collimator plate (Sakurai, 1999;Sakurai & Eba, 2003;Eba et al., 2016). Other approaches explore the use of polycapillaries (Scharf et al., 2011), modified uniformly redundant array masks (Haboub et al., 2014) and potentially other devices in this context.
The silicon-based image sensor of the CMOS camera was estimated to be thinner than 10 mm by measuring and comparing its absorption efficiency for X-rays of different energies. Therefore, the visible-light CMOS camera has a lower detection efficiency for high-energy X-rays compared with many professional X-ray area detectors (Strü der et al., 2001;Henrich et al., 2009;Dinapoli et al., 2011;Blaj et al., 2016). However, its low cost and contemporary availability provides an excellent opportunity for many researchers to establish their own setups. Moreover, considering its unique advantages of a large pixel count and small pixel size, it is clear that the CMOS camera has potential as the XRF detector of choice for visualizing chemical diffusion in various reactions.