Full-field energy-dispersive powder diffraction imaging ... - IUCr Journals

research papers Journal of

Applied Crystallography

Full-field energy-dispersive powder diffraction imaging using laboratory X-rays

ISSN 1600-5767

Received 7 November 2014 Accepted 14 January 2015

Christopher K. Egan,a* Simon D. M. Jacques,a,b Matthew D. Wilson,c Matthew C. Veale,c Paul Seller,c Philip J. Withersa,b and Robert J. Cernika a School of Materials, University of Manchester, Manchester, M13 9PL, UK, bResearch Complex at Harwell, Rutherford Appleton Laboratory, Didcot, Oxfordshire OX11 0FA, UK, and cScience and Technology Facilities Council, Rutherford Appleton Laboratory, Harwell Campus, Oxfordshire OX11 0QX, UK. Correspondence e-mail: [email protected]

# 2015 International Union of Crystallography

A laboratory instrument with the ability to spatially resolve energy-dispersed X-ray powder diffraction patterns taken in a single snapshot has been developed. The experimental arrangement is based on a pinhole camera coupled with a pixelated spectral X-ray detector. Collimation of the diffracted beam is defined by the area of the footprint of a detector pixel and the diameter of the pinhole aperture. Each pixel in the image, therefore, contains an energydispersed powder diffraction pattern. This new X-ray imaging technique enables spatial mapping of crystallinity, crystalline texture or crystalline phases from within a sample. Validation of the method has been carried out with a back-toback comparison with crystalline texture mapping local to a friction stir weld in an aluminium alloy taken using synchrotron radiation.

1. Introduction X-ray imaging and computed tomography is a cornerstone of the nondestructive evaluation of materials and structures, enabling the internal morphology of objects to be probed in two and three dimensions (Flannery et al., 1987). There is an increasing trend of correlating data from multiple length scales and combining insights from different techniques to understand physical and chemical processes happening inside materials (Maire & Withers, 2014). For example, researchers have correlatively combined X-ray tomography data with twoand three-dimensional electron microscopy, electron backscatter diffraction and energy-dispersive X-ray spectroscopy data to fully understand the pitting corrosion of stainless steel (Burnett et al., 2014). Elsewhere, researchers have been able to develop instrumentation that enables two- and threedimensional chemical imaging of materials via X-ray fluorescence (XRF) and X-ray absorption edge spectroscopy (Boone et al., 2014; Egan, Wilson et al., 2014; Schroer et al., 2003). Another active area of research is the deployment of scanning beam techniques like XRF and, in particular, X-ray powder diffraction (XRPD) using synchrotron radiation. These scanning beam methods can be used to build two- and threedimensional images of an object’s chemical or crystalline structure (Bleuet et al., 2010, 2008). Recent applications include three-dimensional imaging of transition metals in biological specimens (Bourassa et al., 2014), chemical imaging of operational catalytic reactors (Jacques et al., 2011) and the study of human teeth biomineralization (Egan et al., 2013). An advancement on scanning beam methods was recently proposed whereby full-field XRPD images of an object could be taken in a single snapshot using synchrotron radiation, with J. Appl. Cryst. (2015). 48, 269–272

every pixel of the image containing an energy-dispersive X-ray powder diffraction pattern (Egan, Jacques et al., 2014). This method removes the requirement of a scanning beam and was found to significantly improve the speed of acquisition, particularly for three-dimensional imaging. A major limitation from a throughput and access point of view is that a large-scale synchrotron facility is required to produce high-flux radiation, inhibiting wide-scale adoption. To this end, we have modified this existing synchrotron setup and built a laboratory system suitable for performing X-ray powder diffraction imaging using easily accessible and cheaper laboratory X-ray systems. Energy-dispersive powder diffraction using laboratory sources has a long-standing history, particularly in the security sector for the identification of illicit drugs and explosives (Cook et al., 2009; O’Flynn et al., 2013) as well as in biomedical biopsy examination (Kidane et al., 1999). The extension of such methods to imaging using novel collimator designs and scanning beams has also been reported (Harding et al., 1987; Harding & Schreiber, 1999). We propose a novel experimental design that enables full-field images of an object’s crystalline structure to be acquired in a single snapshot using a simple collimator and pixelated energy-dispersive detector.

2. Methods The experimental arrangement is shown in Fig. 1. In our case a tungsten target microfocus X-ray tube operating at 160 kV and 60 W irradiates a sample with a cone beam. A pinhole aperture of diameter 500 mm is positioned at a low angle of 2 to the incident beam. The pinhole projects energy-dispersed powder diffraction patterns onto a pixelated spectral X-ray doi:10.1107/S1600576715000801

269

research papers detector. The collimation of each ray path is controlled by the diameter of the pinhole aperture and the areal footprint of the corresponding individual pixel. This is described by the equation dþa ¼ 2 tan1 ; ð1Þ 2f where is the spread in the diffraction angle, d is the size of an individual pixel, a is the pinhole diameter, and f is the distance between the detector pixel and the pinhole. This equates to an energy broadening governed by E ¼ cot : E

ð2Þ

Using a pinhole–detector mean 2 angle of 8 to the direct source beam and a pinhole-to-detector distance of 360 mm gives a mean energy broadening on the detector of 2.5%. A geometric magnification of 2 was used in operation, giving a field of view of about 1 1 cm. The X-ray detector is a HEXITEC 80 80, which has 6400 pixels on a 250 mm pitch. Each pixel measures the energy of every recorded photon to provide spectroscopic X-ray images (Seller et al., 2011). The average spectral energy resolution of each pixel is 800 eV full width at half-maximum (FWHM) at 60 keV. A charge sharing discrimination algorithm is implemented to correct for interpixel sharing of electrons due to photons interacting close to neighbouring pixels. This is implemented by inspecting indi-

vidual frames of data: if, in the same frame of data, neighbouring pixels simultaneously register an X-ray event, then they are removed from the processed data. For more detailed information on the implementation of charge sharing algorithms see Veale et al. ( 2014). The Bremsstrahlung radiation from the source had a useful energy range approximately between 20 and 100 keV with an estimated flux of 4 106 photons s1 mm2 keV1 at 40 keV at a distance of 1 m from the source. As such, with the available flux much weaker than that of the synchrotron, image exposure times were extended to 30 min per image. An additional 2 correction is required to account for the projected spread of the Bragg angle across the face of the detector. We used a flat-field illumination from a fine-grained aluminium plate to build a detector 2 map based upon the position of the principal Bragg peaks. A peak search algorithm is used to correlate X-ray energy and d spacings to find the angle of diffraction as seen by each pixel of the detector. This is then used to shift individual pixel energy scales using the equation E d sin = ˚ . The raw data from the flat-plate sample, 6.19926 keV A before angular calibration, appear as diffraction rings which scroll across the face of the detector as one moves up and down in X-ray energy, as shown in Fig. 2 for example. After calibration, the diffraction rings become spread out such that, as one moves to a particular Bragg peak at a particular energy (e.g. the 111 aluminium peak at 36.5 keV), the whole detector appears uniformly white as all the pixel energy scales are aligned. Further details on angular calibration procedures can be found in the work by Egan, Jacques et al. (2014); in particular, see the electronic supplementary materials for that paper.

3. Results and discussion In order to fully assess the transition from the synchrotron to the laboratory, we performed a back-to-back comparison test using baseline data collected at beamline I12 of the Diamond Light Source, the results of which were published in a previous paper (Egan, Jacques et al., 2014). A brief summary of the results of that experiment is as follows: the sample studied was a 5 mm-thick aluminium alloy (AA7050-T6) plate which was bead-on-plate friction stir welded down its centre line in order to assess the microstructural properties of the alloy after

Figure 1 (a) Plan view of the experimental arrangement for full-field energydispersive powder diffraction imaging. The pinhole aperture is set off-axis to the X-ray beam such that the incident beam does not clip the aperture. This pinhole projects onto a pixelated energy-dispersive X-ray detector. The collimation of a diffracted ray path is defined by the areal footprint of a pixel on the detector and the diameter of the pinhole. (b) Detail of an individual ray path diffracted from a voxel inside the sample, which then travels on through the pinhole aperture and hits a pixel on the detector. These two straight lines define the 2 angle for that voxel. (c) Line drawing showing the apparent broadening () defined by a pinhole aperture of diameter a and a pixel of size d, which are separated by a distance f.

270

Christopher K. Egan et al.

Figure 2 Projected diffraction rings from a fine-grained uniformly flat aluminium plate used to calibrate the 2 distribution across the detector. As the X-ray energy increases, the diffraction rings move to lower 2 angles. A peak search algorithm is used to correlate X-ray energy and d spacings to find the angle of diffraction as seen by the detector.

Full-field energy-dispersive powder diffraction imaging

J. Appl. Cryst. (2015). 48, 269–272

research papers welding. A cross section (approximately 5 5 40 mm) of the weld was cut out of the plate and taken to the synchrotron for analysis. Using our pinhole projection method at the synchrotron, we were able to successfully map out the preferred orientation across the weld section by mapping the relative intensities of three principal diffraction peaks. We were also able to show the distribution of alloy precipitates across the weld section based upon weaker diffraction peaks not corresponding to the matrix alloy. In the laboratory experiment presented in this paper, we use the synchrotron data as baseline information to compare with data collected using the laboratory system; the results are shown in Fig. 3(a). First we look at the preferred orientation across the weld section. The scattered intensity from the 111, 200 and 220 Bragg peaks from face-centred cubic (f.c.c.) aluminium are coloured as red, green and blue, respectively, and overlaid as an RGB image. A fine-grained nontextured material would appear white in this image. Any preferred orientation appears as a colour with red regions showing the (111) planes to be

Figure 3 (a) Preferred orientation in an aluminium alloy friction stir weld expressed as a colour map with red indicating the 111, green indicating the 200 and blue indicating the 220 diffraction peaks such that white would represent no preferred texture. The rotating side (RS) of the weld section is shown. Data collected using the synchrotron setup are shown above and compared with data collected using the laboratory arrangement. The approximate outline of the friction stir welding tool is shown as a solid white line. The dashed line represents the spatial segmentation of precipitates across the weld section, which was observed at the synchrotron but was not observable in the laboratory data owing to a weak signal. (b) Typical energy-dispersed powder diffraction pattern averaged over 3 3 pixels for the laboratory setup with the Al and other diffraction peaks labelled. (c) Equivalent diffraction pattern from synchrotron data of the same sample, showing much narrower peak widths and generally better quality data. Note that the energy scale in this data set is much higher owing to the lower mean angle of diffraction utilized at the synchrotron. (d) Photograph of the friction stir weld sample showing the approximate area of study. J. Appl. Cryst. (2015). 48, 269–272

preferentially aligned parallel to the weld-plate surface (h111i//ND). The main point of these data is to compare the laboratory and synchrotron images; we can see that there is a reasonable match, albeit with some image blurring. A pinhole camera has an inherent geometric blurring due to beam divergence, which is governed by the equation b ¼ dð1 þ MÞ, where d is the pinhole diameter and M is the geometric magnification. In this case, we used a geometric 2 magnification; this gives b = 1500 mm, which equates to six pixels on the detector. Back-projecting this onto the sample accounting for the geometric magnification gives an on-sample blurring of approximately 750 mm. Despite this additional blurring, the laboratory data match well to the synchrotron data and clearly show the spatial variation in preferred orientation across the weld section. Fig. 3(b) shows an energy-dispersed powder diffraction pattern from the aluminium alloy sample averaged over 3 3 pixels on the detector. The pattern matches well to that of f.c.c. aluminium with the peak indexes shown. In this pattern, we also observe Cd and Te XRF signals which arise from detector self-emission and are always present. These can be easily separated from the diffraction signal. Analysis of the diffraction peak widths shows them to be rather broad, averaging around 3 keV FWHM, equating to about 5% energy broadening. This compares unfavourably with data collected at the synchrotron which have diffraction peak widths of around 1 keV, only slightly above the energy resolution of the detector at 0.8 keV. The peak broadening effect observed in the laboratory data is caused by the reduced effective collimation arising from a shorter pinhole-to-detector distance and larger pinhole aperture. In addition, there is a broadening associated with the source-beam divergence across the sample thickness. With a source-to-sample distance of 200 mm, the beam divergence of a ray path across a sample of thickness 5 mm using the current setup can be up to 0.2 , which equates to a 5% energy broadening. This peak broadening unfortunately limits the method, in its current form, to studying crystalline materials that have large lattice parameters and, therefore, widely spaced Bragg peaks. To mitigate this, it would be necessary to move the sample further away from the source and/or study thinner samples. For example, a sample of 1 mm thickness which is positioned 400 mm from the source would give a through thickness broadening of 0.5%, much better than the energy resolution of the detector. An unfortunate downside of the observed peak broadening was that we were unable to observe the scattering signal from precipitates in the sample because they were hidden underneath the aluminium Bragg peaks. Despite the above highlighted limitations associated with low spatial resolution and diffraction peak broadening, the ability to image and identify specific crystalline phases within what is essentially an X-ray radiograph is highly promising. In particular, the transfer of the method from the synchrotron to the laboratory has significantly improved end-user opportunities where access to large-scale synchrotron facilities is limited. An additional benefit of this method, which has been previously demonstrated (Egan, Jacques et al., 2014), is the possibility to perform three-dimensional powder diffraction



271

research papers tomography by rotating the sample, recording projections and reconstructing the three-dimensional data. The resultant computer reconstructed volume contains the three-dimensional morphology of the sample with each voxel containing an X-ray powder diffraction pattern. Improvements in the method are envisaged, for example by increasing the flux of the X-ray source. In this experiment we used a low-powered microfocus X-ray tube (60 W). This type of tube is not designed for high flux; however, high-powered X-ray tubes (typically 1–2 kW) are widely available which could increase the available photon flux by approximately 10. This development will enable shorter data acquisition times and higher spatial resolutions (with smaller pinhole apertures). Further improvements can be made by moving the detector to lower 2 angles, thereby moving the Bragg peaks to higher X-ray energies. A combination of using a smaller-diameter pinhole (facilitated by a higher-power X-ray tube) coupled with moving the sample and detector further away from the source will reduce energy broadening and Bragg peak widths. On the basis of these improvements, we expect to be able to achieve spatial resolutions of around 50–100 mm and exposure times of around 5–10 min, with peak widths of around 1.5 keV. Potential scientific applications of this method might include the following: time-lapse studies of working catalytic reactors to understand the factors that influence the distribution and nature of the active phase during preparation and industrial operation (Jacques et al., 2011); mineral mapping in geological samples and three-dimensional visualization of rock core samples (Mees et al., 2003); and the study of the time evolution of corrosion in metallic alloys (Knight et al., 2011).

4. Conclusions We have built a laboratory instrument with the ability to rapidly image spatial variations of crystallinity within a sample. On the basis of previous experiments using synchrotron radiation, we have developed a pinhole camera system that enables energy-dispersed powder diffraction patterns to be spatially mapped over a sample. The method in its current form has a spatial resolution of approximately 600 mm with powder diffraction patterns with peak widths of around 3 keV. We performed a back-to-back comparison test using data collected at the synchrotron to validate the method. Some limitations still need to be overcome; in particular, the installation of a high-powered X-ray tube would facilitate improvements in spatial resolution, reduce data-acquisition times and produce narrower peak widths. Nevertheless, we have demonstrated the ability to perform chemical and crystallographic imaging of materials in the laboratory.

272


The Manchester (Henry Moseley) X-ray Imaging Facility was funded, in part, by the EPSRC (grants EP/F007906/1, EP/ F001452/1, EP/I02249X/1 and EP/H046577/1).

References Bleuet, P., Lemelle, L., Tucoulou, R., Gergaud, P., Delette, G., Cloetens, P., Susini, J. & Simionovici, A. (2010). Trac-Trend. Anal. Chem. 29, 518–527. Bleuet, P., Welcomme, E., Dooryhee, E., Susini, J., Hodeau, J. L. & Walter, P. (2008). Nat. Mater. 7, 468–472. Boone, M. N., Garrevoet, J., Tack, P., Scharf, O., Cormode, D. P., Van Loo, D., Pauwels, E., Dierick, M., Vincze, L. & Van Hoorebeke, L. (2014). Nucl. Instrum. Methods Phys. Res. Sect. A, 735, 644–648. Bourassa, D., Gleber, S.-C., Vogt, S., Yi, H., Will, F., Richter, H., Shin, C. H. & Fahrni, C. J. (2014). Metallomics, 6, 1648–1655. Burnett, T., McDonald, S., Gholinia, A., Geurts, R., Janus, M., Slater, T., Haigh, S., Ornek, C., Almuaili, F. & Engelberg, D. (2014). Sci. Rep. 4, 4711. Cook, E., Griffiths, J., Koutalonis, M., Gent, C., Pani, S., Horrocks, J., George, L., Hardwick, S. & Speller, R. (2009). SPIE Defense, Security, and Sensing, pp. 73100I-73100I-73110. International Society for Optics and Photonics. Egan, C. K., Jacques, S. D., Connolley, T., Wilson, M. D., Veale, M. C., Seller, P. & Cernik, R. J. (2014). Proc. R. Soc. London Ser. A, 470, 20130629. Egan, C. K., Jacques, S. D., Di Michiel, M., Cai, B., Zandbergen, M. W., Lee, P. D., Beale, A. M. & Cernik, R. J. (2013). Acta Biomaterial. 9, 8337–8345. Egan, C. K., Wilson, M. D., Veale, M. C., Seller, P., Jacques, S. D. & Cernik, R. J. (2014). Nucl. Instrum. Methods Phys. Res. Sect. B, 324, 25–28. Flannery, B. P., Deckman, H. W., Roberge, W. G. & D’Amico, K. L. (1987). Science, 237, 1439–1444. Harding, G., Kosanetzky, J. & Neitzel, U. (1987). Med. Phys. 14, 515– 525. Harding, G. & Schreiber, B. (1999). Radiat. Phys. Chem. 56, 229–245. Jacques, S. D. M., Di Michiel, M., Beale, A. M., Sochi, T., O’Brien, M. G., Espinosa-Alonso, L., Weckhuysen, B. M. & Barnes, P. (2011). Angew. Chem. Int. Ed. 50, 10148–10152. Kidane, G., Speller, R., Royle, G. & Hanby, A. (1999). Phys. Med. Biol. 44, 1791. Knight, S., Salagaras, M. & Trueman, A. (2011). Corrosion Sci. 53, 727–734. Maire, E. & Withers, P. J. (2014). Int. Mater. Rev. 59, 1–43. Mees, F., Swennen, R., Van Geet, M. & Jacobs, P. (2003). Geol. Soc. London Spec. Publ. 215, 1–6. O’Flynn, D., Desai, H., Reid, C. B., Christodoulou, C., Wilson, M. D., Veale, M. C., Seller, P., Hills, D., Wong, B. & Speller, R. D. (2013). Crime Sci. 2, 1–6. Schroer, C., Kuhlmann, M., Gu¨nzler, T., Lengeler, B., Richwin, M., Griesebock, B., Lu¨tzenkirchen-Hecht, D., Frahm, R., Ziegler, E. & Mashayekhi, A. (2003). Appl. Phys. Lett. 82, 3360–3362. Seller, P., Bell, S., Cernik, R. J., Christodoulou, C., Egan, C. K., Gaskin, J. A., Jacques, S., Pani, S., Ramsey, B. D., Reid, C., Sellin, P. J., Scuffham, J. W., Speller, R. D., Wilson, M. D. & Veale, M. C. (2011). J. Instrum. 6, C12009. Veale, M., Bell, S., Duarte, D., Schneider, A., Seller, P., Wilson, M. & Iniewski, K. (2014). Nucl. Instrum. Methods Phys. Res. Sect. A, 767, 218–226.


J. Appl. Cryst. (2015). 48, 269–272