A high-resolution lens-coupled X-ray imaging detector equipped with a thin-layer transparent ceramics scintillator has been developed. The scintillator consists of a 5 μm thick Ce-doped layer (LuAG:Ce) bonded onto the support substrate of the non-doped LuAG ceramics by using a solid-state diffusion technique. Secondary electron microscopy of the bonded interface indicated that the crystal grains were densely packed without any pores in the optical wavelength scale, indicating a quasi-uniform refractive index across the interface. This guarantees high transparency and minimum reflection, which are essential properties for X-ray imaging detectors. The LuAG:Ce scintillator was incorporated into an X-ray imaging detector coupled with an objective lens with a numerical aperture of 0.85 and an optical magnification of 100. The scintillation light was imaged onto a complementary metal–oxide–semiconductor image sensor. The effective pixel size on the scintillator plane was 65 nm. X-ray transmission images of 200 nm line-and-space patterns were successfully resolved. The high spatial resolution was demonstrated by X-ray transmission images of large integrated circuits with the wiring patterns clearly visualized.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
X-ray microscopy is one of the most powerful tools for non-destructive observation. In particular, X-ray imaging methods such as computed tomography, phase contrast imaging, topography, and laminography provide high spatial resolution when combined with monochromatic X-rays generated by synchrotron radiation sources. In these methods, X-ray imaging detectors play a critical role in determining measurement accuracy. One of the most important performance metrics is the spatial resolving power of the imaging detector. In resolution ranges smaller than a few micrometers, indirect lens-coupled detectors are widely used [1–6]. These consist of a scintillator layer, imaging optics, and an optical imaging sensor. The X-rays are converted within the scintillator layer to visible light and then projected onto the image sensor through magnification optics. Scintillation light emitted outside the focus of the objective lens blurs the X-ray images. To eliminate this blur, the scintillator layer is made very thin. For high spatial resolution imaging, the thickness of the scintillator typically needs to be as thin as 5 μm . Such thin self-standing scintillators, however, do not endure due to their mechanical instability. In practice, thin scintillators are generally attached to various mechanical support layers [3,5]. Due to the imperfection of the scintillator-support interface, X-ray imaging detector systems are generally not capable of resolving deep sub-micron features .
In this Letter, we report the successful development of an X-ray imaging detector, which can resolve 200 nm line-and-space patterns. The detector has a 5 μm thick ceramic scintillator layer of supported by a transparent ceramic layer of non-doped LuAG. The LuAG:Ce and LuAG layers were directly attached by using a solid-state diffusion bonding method without adhesive [8,9]. The imaging capability of the detector was demonstrated by the X-ray transmission images with clear metal wiring structures in a very-large-scale integrated (VLSI) circuit.
The composite consisting of the 5 μm thick LuAG:Ce ceramics scintillator and the 1 mm thick non-doped LuAG ceramics was fabricated using a procedure identical to that used in our previous study for the YAG:Ce/YAG composite . The LuAG:Ce ceramics was reported to have an emission peak at 520 nm and a light yield of 16,000 photons/MeV . Figure 1(a) shows the photograph of the composite. Using optical means, we were unable to detect any reflectivity at the bonded interface. Optical transmissions were recorded for samples without any coating. The illumination direction was set to be perpendicular to the composite surface. The transmission of the composite gave a transmission in the range of 81.6% to 84.2% for the wavelength () range from 450 to 700 nm. At , the transmission was 83.8%. This coincides very well to the transmission of the LuAG bulk material with a refractive index of 1.85, which was measured to be 82.7% at . Note that the dielectric boundary between the air () and the material with theoretically gives a transmission of 83.0% . The agreement between the transmission of the composite and the bulk sample strongly implies that the bonded interface has high transmission and very low reflection.
In the above macroscopic method, scattered and/or diffused light may be recorded as transmitted light. In order to examine microscopic characteristics, a cross-sectional sample of the composite was recorded using a secondary electron microscopy (SEM) device (Hitachi S-4700 SEM). The surfaces of the sample were prepared using a polishing and thermal etching method. The SEM images [Fig. 1(b)] show clear grain boundaries. In addition, the images at the bonded interface and the bulk region show no pores, at least in the length scale of the fluorescence wavelength. A survey for a cross section was conducted, and no pores were detected. This strongly indicates that scattering and/or diffusion of the light was well suppressed.
In order to assess X-ray imaging performance, an indirect X-ray imaging detector equipped with the composite as explained above was assembled. The detector configuration is described in the dashed square section of Fig. 2. An objective lens with a NA of 0.85 transferred the fluorescence images to the tube lens thorough an aluminum reflection mirror and a bandpass filter. The transferred images were then projected to the complementary metal–oxide–semiconductor (CMOS) image sensor with an image format of 6.5 μm square pixels (PCO AG pco.panda 4.2). The optical magnification of the detector was set to 100. The images were sampled at an effective pixel size of in a field of view of at the scintillator plane accordingly. The composite was coated with a reflection film at the X-ray illumination surface and an anti-reflection film on the opposite side. The reflection film effectively doubled the light yield by reflecting the fluorescence light in the direction of the objective lens. Both films were composed of a dielectric multilayer and optimized for the wavelength region of 450–700 nm and for an incident angle from 0 to 58.2 deg, which were matched to the LuAG:Ce fluorescence band with  and the acceptance angle of the objective lens with a NA of 0.85. The fluorescence emission from the non-doped LuAG in  was removed by an optical bandpass filter.
Spatial resolution of the constructed indirect X-ray imaging detector was evaluated using an X-ray beam at the SPring-8 beamline BL29XUL . Figure 2 shows the schematic of the experimental setup. The X-ray beams were monochromatized to be in the order of by the Si (111) double crystal monochromator (DCM). The intensity of the high-order harmonics was attenuated to be in the order of by double X-ray mirrors. The speckle pattern induced by the slits was removed by inserting an X-ray diffuser . The X-ray beams passing through the DCM, the double X-ray mirrors, and the X-ray diffuser were irradiated to the X-ray test chart (NTT-AT, XRESO-100) containing line-and-space and Siemens star patterns. These are made of a 1 μm thick tantalum layer. The X-ray transmission profile of the test chart was delivered to the indirect X-ray imaging detector. The test chart was placed 0.5 mm upstream of the surface of the composite to minimize the signal spread of the X-ray diffraction. The X-ray beam had a photon energy of 7.3 keV and an intensity of . Images were obtained by averaging 100 frames, where each had an exposure time of 200 ms/frame. A microradiograph image was obtained by normalizing the transmitted image with an average of flat-field images. Figures 3(a)–3(c) are the images of the patterns with 200, 400, and 600 nm lines-and-spaces, respectively. Figure 3(a) shows that 200 nm line-and-space patterns were successfully resolved. The line profiles for the patterns shown in Fig. 3(d) for a 200 nm line-and-space gives a modulation of 9%. Here the modulation, , was defined as , where were the maximum and minimum transmittances, respectively, and was the baseline transmittance measured outside the feature region. In the case of 400 and 600 nm lines-and-spaces [Figs. 3(e) and 3(f)], the improved to 45% and 41% for the lines-and-spaces of 400 and 600 nm, respectively.
X-ray images of the Siemens star pattern were recorded to measure the modulation transfer functions (MTFs) at the photon energies () of 7.3, 10.0, 12.0, 14.0, 16.0, and 18.0 keV. For those photon energies, the nominal quantum efficiencies derived from the photoelectric cross section of the scintillator screen were 34%, 36%, 38%, 25%, 18%, and 15%, respectively. The pattern had the smallest feature with a line-and-space of 100 nm. Images taken at 7.3 and 16 keV are shown in Figs. 4(a) and 4(b), respectively. Both the images show fine features. The image taken at 16.0 keV shows blurs, which are noticeable for features with a line-and-space of 1000 nm and more prominent for smaller features. The penetration depth is 22.9 μm for , which is deeper than that of 11.7 μm for . Increased blur may be explained by the fact that the image at was taken with more fluorescence emitting outside the focus of the objective lens (defocus effect). Note that the depth of focus of the lens was , which is shallower than the 5 μm thick scintillator layer. The MTFs were analyzed for line profiles along the concentric circles on the Siemens pattern, as shown in Fig. 4(c). The spatial frequency was defined as , where is the line width. The Siemens pattern had several regions where the patterns were missing [Figs. 4(a) and 4(b)]. Data in the corresponding frequency regions were not present and are marked in Figs. 4(c) and 4(d) with the label . Figure 4(d) depicts MTFs around the cutoff frequencies. All the MTFs showed cutoff frequencies around 2500 line pairs/mm or higher. The cutoff frequencies were obtained by linearly extrapolating the last two data points. For , the cutoff frequency was as high as 2650 line pairs/mm (). This is consistent with the successful observation of 200 nm line-and-space structures [Figs. 3(a) and 3(d)]. In Fig. 4(c), the MTF for 7.3 keV shows higher modulation than the other MTFs around the frequency of 1000–2000 line pairs/mm. This peculiar trend can be understood that the lack of the fluorescence X-rays from Lu atoms in the scintillator gives sharper images for . Note that 7.3 keV is below the Lu and absorption edges at 9.24 and 10.35 keV, respectively. The MTF for shows a modulation drop of around 300–850 line pairs/mm. We have noticed fringes in the X-ray images corresponding to these frequencies, which degrade the modulation. These fringes are probably caused by the X-ray diffraction.
In a lens-coupled detector system composed of an aberration-free lens, the point spread function can be expressed by an airy disk. The cutoff frequency for this type of ideal optical system can be expressed by , where NA is the numerical aperture of the objective lens and is the wavelength of the scintillation light . For an optical system with and , as in this Letter, the cutoff frequency is expected to be 3269 line pairs/mm, as depicted as an arrow B in Fig. 4(d). The observed cutoff frequency had a slight performance drop from 3269 to 2650 line pairs/mm as an arrow A in Fig. 4(d). This drop corresponds to the cutoff line width (the half of the cutoff spatial resolution), degraded from the diffraction limit of 152 to 189 nm. The discrepancy is to be caused by the defocus effect, aberrations in the optical systems, the light emitting from the support substrate of non-doped LuAG, and/or X-rays diffracted by the Siemens pattern.
Image quality was investigated by taking X-ray transmission images of a VLSI circuit chip, which was a CMOS image sensor. The details of the chip were reported elsewhere . Briefly, it was produced by a silicon-on-insulator CMOS process with a technology node of 200 nm. Five metal layers of aluminum were incorporated, where each layer was 600–900 nm thick. The metal layers were connected by a vertical interconnection area (via) made of tungsten. Beneath these metal wiring sections was substrate silicon with a thickness of 500 μm. Figure 5(a) shows the designed layout of metal layers and the vias for a pixel. The aluminum lines had a minimum width of 300 nm. The black squares are the vias. The X-ray transmission image corresponding to the drawing depicted in Fig. 5(a) was measured by adding 255 frames with an exposure time of 600 ms each. An X-ray beam was irradiated at a photon energy of 12 keV with an intensity of . In order to enhance the visibility, the contrast-limited adaptive histogram equalization  was applied after the flat-field correction. The resulting image is shown in Fig. 5(b). The metal wiring patterns and vias were successfully visualized. The contrast of a line profile in the dotted rectangle in Fig. 5(b) is shown as Fig. 5(c). The aluminum wires had a thickness of 600 nm, which is significantly thinner than the silicon thickness of 500 μm. As a result, the metal wires showed a contrast as low as 1%. Here the contrast was defined as . The present detector successfully visualized these low contrast and deep sub-micrometer patterns. This technique will be of interest for applications such as defect detection of VLSIs. Due to the high contrast of the detector, we expect that it will be suitable, even in combination with laboratory X-ray sources.
Previously, the cutoff line width of 300 nm was reported  by using a thin-film scintillator fabricated by a liquid phase epitaxy (LPE) method . In their configuration, the scintillator was placed 6 mm upstream of the objective lens with a NA of 0.7. Our approach gives an opportunity to shield the objective lens against the X-rays through the ceramics substrate, which allowed the scintillator to be closer to the objective lens and made the resulting NA as large as 0.85. In comparison with the LPE method, our approach is also advantageous from the viewpoint of manufacturability.
In our previous work , we employed YAG:Ce instead of LuAG:Ce. For comparison, we measured and analyzed spatial resolving power of a detector with YAG:Ce in the identical manner at , and yielded the cutoff frequency of 2350 line pairs/mm (data not shown). However, the modulation of 10% was obtained only at 480 line pairs/mm, while the present LuAG:Ce gave 10% modulation at 2320 line pairs/mm. The superior performance of LuAG:Ce should be attributed to its larger stopping power .
In conclusion, we have demonstrated an X-ray imaging detector capable of resolving 200 nm line-and-space patterns by using an indirect imaging detector equipped with a 5 μm LuAG:Ce film scintillator [Fig. 3(a)]. The thin scintillator was fabricated using the solid-state diffusion bonding technique. The optical quality of the LuAG:Ce ceramics scintillator was evaluated using optical transmission and SEM measurements [Fig. 1(b)]. These indicated high optical transparency with minimum reflection, diffusion, and scattering. These results revealed that improvement of the optical property of thin scintillator can lead spatial resolving power with a cutoff line width comparable to the diffraction limit of . The high resolving power was proved by the images showing clear deep sub-micron structures of metal layers and vias in the VLSI chip.
Japan Society for the Promotion of Science (JSPS) (26790077).
The authors would like to thank the Konoshima Chemical Co., Ltd. for producing the thin-film scintillators. We also thank SIGMAKOKI Co., Ltd. for manufacturing the camera head and depositing the optical coating onto the scintillators, respectively.
1. A. Koch, C. Raven, P. Spanne, and A. Snigirev, J. Opt. Soc. Am. A 15, 1940 (1998). [CrossRef]
2. S. M. Gruner, M. W. Tate, and E. F. Eikenberry, Rev. Sci. Instrum. 73, 2815 (2002). [CrossRef]
3. K. Uesugi, Y. Suzuki, N. Yagi, A. Tsuchiyama, and T. Nakano, SPIE Conf. Proc. 4503, 291 (2002). [CrossRef]
4. J. Touš, M. Horváth, L. Pína, K. Blažek, and B. Sopko, Nucl. Instrum. Methods Phys. Res., Sect. A 591, 264 (2008). [CrossRef]
5. J. Tous, K. Blazek, M. Nikl, and J. A. Mares, J. Phys. Conf. Ser. 425, 192017 (2013). [CrossRef]
6. T. Martin and A. Koch, J. Synchrotron Radiat. 13, 180 (2006). [CrossRef]
7. H. Graafsma and T. Martin, Advanced Tomographic Methods in Materials Research and Engineering (2008), p. 277.
8. H. Yagi, K. Takaichi, K. Hiwada, K.-I. Ueda, and T. Yanagitani, Jpn. J. Appl. Phys. 45, L207 (2006). [CrossRef]
9. H. Yagi, T. Yanagitani, K. Takaichi, K.-I. Ueda, and A. A. Kaminskii, Opt. Mater. 29, 1258 (2007). [CrossRef]
10. T. Kameshima, T. Sato, T. Kudo, S. Ono, K. Ozaki, T. Katayama, T. Hatsui, and M. Yabashi, AIP Conf. Proc. 1741, 040033 (2016). [CrossRef]
11. T. Yanagida, Y. Fujimoto, Y. Yokota, A. Yoshikawa, T. Ishikawa, H. Fujimura, H. Shimizu, H. Yagi, and T. Yanagitani, IEEE Nuclear Science Symposuim & Medical Imaging Conference (2010), p. 1612.
12. W. J. Smith, Modern Optical Engineering, 3rd ed. (SPIE, 2000).
13. Y. Fujimoto, T. Yanagida, H. Yagi, T. Yanagidani, and V. Chani, Opt. Mater. 36, 1926 (2014). [CrossRef]
14. K. Tamasaku, Y. Tanaka, M. Yabashi, H. Yamazaki, N. Kawamura, M. Suzuki, and T. Ishikawa, Nucl. Instrum. Methods Phys. Res., Sect. A 467-468, 686 (2001). [CrossRef]
15. K. S. Morgan, S. C. Irvine, Y. Suzuki, K. Uesugi, A. Takeuchi, D. M. Paganin, and K. K. W. Siu, Opt. Commun. 283, 216 (2010). [CrossRef]
16. T. Hatsui, M. Omodani, T. Kudo, K. Kobayashi, T. Imamura, T. Ohmoto, A. Iwata, S. Ono, Y. Kirihara, and T. Kameshima, Proceedings of the International Image Sensor Workshop (IISW) (2013).
17. K. Zuiderveld, Graphic Gems IV (Academic Press Professional, San Diego, 1994).
18. J. W. Jung, J. S. Lee, N. Kwon, S. J. Park, S. Chang, J. Kim, J. Pyo, Y. Kohmura, Y. Nishino, M. Yamamoto, T. Ishikawa, and J. H. Je, Rev. Sci. Instrum. 83, 093704 (2012). [CrossRef]
19. M. Nikl, A. Yoshikawa, K. Kamada, K. Nejezchleb, C. R. Stanek, J. A. Mares, and K. Blazek, Prog. Cryst. Growth Charact. Mater. 59, 47 (2013). [CrossRef]