Abstract
Grating based X-ray differential phase contrast imaging (DPCI) allows for high contrast imaging of materials with similar absorption characteristics. In the last years’ publications, small animals or parts of the human body like breast, hand, joints or blood vessels have been studied. Larger objects could not be investigated due to the restricted field of view limited by the available grating area. In this paper, we report on a new stitching method to increase the grating area significantly: individual gratings are merged on a carrier substrate. Whereas the grating fabrication process is based on the LIGA technology (X-ray lithography and electroplating) different cutting and joining methods have been evaluated. First imaging results using a 2×2 stitched analyzer grating in a Talbot-Lau interferometer have been generated using a conventional polychromatic X-ray source. The image quality and analysis confirm the high potential of the stitching method to increase the field of view considerably.
Introduction
Differential phase contrast imaging (DPCI) based on X-ray grating interferometry is a promising imaging method exploiting three sources of contrast: absorption, phase and dark field (small-angle scattering) [22, 45]. It allows for high contrast imaging of materials with similar absorption characteristics, which is of great interest in medical and materials sciences, in food industry and security screening [4, 47]. The available grating area limits the field of view, restricting studies to small objects like small animals [41, 42] or parts of the human body like joints, liver tissue, arteries or breast tissue [10, 46]. Studies showed particularly high potential for mammography. It can strongly profit from the extended imaging possibilities as the small-angle scattering contrast reveals micro-calcifications, which might be related to aggressive tumor types [20, 34]. In order to capture the whole sample, the minimum field of view needs to be in the order of 200 mm×200 mm, which is currently not feasible due to the size limitations of the gratings (50 mm×50 mm or 70 mm in diameter) fabricated by the LIGA process (X-ray lithography and electroplating) [26]. Therefore, recent studies scanned breast specimens and stitched the captured images [1, 37], which is not feasible for clinical application regarding time and dose. Another approach uses scanning type mammography [34].
Key elements of a Talbot interferometer are the phase and the analyzer grating. Especially the latter’s fabrication is highly challenging as the investigation of large objects at energies >25 keV demands microstructures with heights of more than 50μm and aspect ratios up to 100 on large areas. Current fabrication processes like Silicon etching [2, 5] and LIGA technology [14, 40] struggle to fulfill all of these demanding requirements. For the LIGA technology, the currently available mask size with a circular area of 70 mm in diameter is the main limiting factor.
In case of the LIGA technology, different approaches have the potential to increase the field of view in DPCI, e.g. larger masks, stepped exposure, layer deposition or stitching individual gratings.
Increasing the mask size depends on the stability of the mask membrane. Currently the X-ray mask consists of a thin and fragile 2.5μm thick titanium membrane carrying the gold absorbers [23]. Increasing the mask area leads to deflections and waviness of the membrane and the risk of membrane rupture rises significantly. As a consequence the absorbers are randomly tilted, so a well-defined patterning is no longer possible.
Stepped exposure is a standard technique in UV lithography. The transfer to X-ray lithography hasn’t been established up to now at the X-ray beamlines as an expensive re-design of the lithographic scanner would become necessary.
The grating fabrication by layer deposition was shown by different groups [13, 18]. In this case gratings are built up by depositing layer by layer (lamella by lamella). It allows to create ultra-high aspect ratio gratings with very long lamellas in the range of some 10 cm. Unfortunately, the layers have to be built up sequentially. In order to fabricate a grating of 200 mm with a period of 2μm 200.000 cycles become necessary (100.000 absorber layers, 100.000 spacer layers). Using a slightly different approach with a so called staircase substrate, Lynch et al. [18] reduced the number of cycles tremendously, providing a relatively large area of 20 mm×20 mm. As a consequence the aspect ratio is reduced significantly.
Stitching individual gratings offers the possibility to realize larger absorption grating areas with high aspect ratios up to 200 mm×200 mm. Therefore, we developed and analyzed the stitching possibilities as an add-on fabrication step to the standard LIGA based grating fabrication process [14, 33].
Grating fabrication
LIGA process
High quality X-ray absorber gratings with an area of 50 mm×50 mm or 70 mm diameter are produced by means of the LIGA process [14, 33]. A polymer template is patterned by X-ray lithography and subsequently filled with gold by electrodeposition.
The X-ray gratings are fabricated on 525μm or 200μm thick 4 inch Silicon wafers, covered by a thin conductive top layer of titanium to enable later electrodeposition. An epoxy based negative resist layer (mr-X, micro resist technology, Berlin) is deposited by spin coating [16]. After prebaking at 95°C, the resist layer is patterned by X-ray exposure using synchrotron radiation and a mask fabricated by e-beam lithography. In a subsequent post exposure bake at 75°C, the exposed resist is cross-linked. The unexposed parts are dissolved in a developer bath using propylene glycol methyl ether acetate (PGMEA) at room temperature, followed by a rinsing step using an isopropyl alcohol bath at room temperature. The resulting cavities are filled with gold by electroplating at 55 °C using a sulfide gold electrolyte [3, 33].
Stitching process
The stitching method is designed as an add-on to the already existing fabrication chain. Three steps are executed: removing the unstructured areas by cutting, grating alignment and fixation on a carrier substrate (Fig. 1).
Cutting away unstructured areas
Removing the unstructured areas of a grating substrate has to fulfill several boundary conditions: High parallelism (rectangularity) of the tile edge to the lamellas Small mechanical impact on the grating structures Temperature during cutting less than 60°C to avoid structure damage
Several techniques like laser cutting, wafer sawing, etching or water-beam cutting can be taken into account. Due to the easy availability we made preliminary tests with a wafer saw DAD 3430 Disco with a 45μm thick diamond-covered blade (grain size: 4–6μm) at a speed of 30.000 rpm and a feed of 0.8 mm/s. The saw was aligned to the lamellas of the gratings using a standard light microscope. Our analysis showed that the angle between the sample edge and the lamella direction was in the range of 0,5 mrad to 1 mrad. Also the deviation of the edge position from an ideal line was less than 14μm for more than 95% of the whole edge. Both values were assumed to be sufficient to perform first tiling experiments using wafer sawing to cut the samples. In order to minimize the inter-grating gaps, the cutting line was positioned 1 mm away from the border in the structure field to remove usually partly over-electroplated grating structures at the borders.
As shown in Figs. 2 and 3, wafer sawing allows for sufficient high parallelism between lamellas and cutting edge; the deviation angles between lamellas and cutting edge of the 32 investigated edges were all <1 mrad. For the evaluation, panorama images of the cutting edges recorded by a standard light microscope were analyzed. The roughness of a cutting edge was characterized by its standard deviation σ and the maximal peak-to-valley-distance. Figure 3 shows on the one hand a smaller mean roughness for the parallel edges compared to the vertical ones, but a higher mean value of the maximum peak-to-valley distance for the parallel edges. This is due to a smoother cutting of the parallel edges, but the fiber-like break-outs of single lamellas caused large peak-to-valley distances. For vertical edges, notches appeared more often, increasing the mean edge roughness. The edge quality tended to be higher for the thicker 525μm Si wafers compared to the 200μm ones.
The diagram in Fig. 2B shows a typical cutting edge profile. At the corners the cutting edge deviation often increased, as also the light microscopic image of the section of a cutting edge (Fig. 2A) illustrates. The diagram in Fig. 2C gives the histogram of the cutting tolerance which is related to the mean edge roughness; positive values mean mainly particles on the edge, whereas negative values indicate notches.
Assembling
We investigated different joining techniques taking the following boundary conditions into account: Homogenous and continuous connection to avoid scattering Temperatures <60 °C to avoid structure damage Temperature stability X-ray stability
To avoid an additional carrier substrate and thus to avoid additional absorption, clamping would be the most promising frictional connection principle for this application. It requires at least one accessible edge, which is available when gratings are stitched in up to two columns or rows (1×n, 2×n, n×1 and n×2). These stitched gratings could become of interest for tomography applications, which may need stitching the gratings only along the rotational axis.
Form closure methods do not fulfill the boundary condition of a homogenous and continuous connection all over the field of view because of required support structures extending into the field of view. Therefore, for large area gratings which encompass more than two rows or columns, the grating tiles need to be fixed to a carrier substrate. Among the substance-to-substance bonds, adhesive bonding provides a safe connection at relative low temperatures (<60°C). It is an advantage that adhesives can be designed according to the application needs. We excluded high temperature bonding techniques like welding and brazing due to the temperature regime.
In the very first experiments we chose a 625μm thick 6”-Silicon wafer as carrier substrate and a standard Epoxy resin as glue. The grating tiles were put edge to edge without further alignment. To minimize pitch and roll errors due to height variations of the adhesive layer, the adhesive was only applied along the outer edges of the tiles.
Grating performance
We tested a first stitched grating consisting of four tiles with an overall edge length of 96 mm as analyzer grating (G2) in a Talbot-Lau interferometer (Department of Physics and Institute of Medical Engineering, Technische Universität München). The set-up is symmetrical with a phase grating (G1) with an area of 50 mm×50 mm, a period of 5,4μm and a gold height of 5.2μm leading to a PI-Phase shift. The absorber gratings period was 5.4μm as well; the gold height was 70μm to fit the existing setup configuration with a design energy of 27 keV. The wafer was mounted in front of the installed photon-counting detector (Pilatus II, Dectris, Switzerland) with a pixel size of 172μm. Further general information on the imaging system can be found in [7, 46].
The alignment of the gratings utilized in the interferometer is crucial for its imaging performance. Due to the inter-tile rotational error around the yaw axis, the stitched grating was aligned to achieve the most reasonable impression for the image quality over the total grating area.
To determine the gratings’ quality and their assembly we performed flat-field measurements to obtain visibility maps. The visibility is an important performance factor which is directly associated to the imaging capabilities of the set-up. Grating defects or inhomogeneities and grating misalignments affect the visibility and cause lower values or variations within the recorded maps. To examine the influence of the gaps between the single tiles on radiographic images, we acquired absorption-, phase- and dark-field-contrast data of a frog (ex-vivo) at a tube voltage of 40 kVp.
Results
Figure 4 shows an example of four (2×2) stitched grating tiles providing a grating area of 96 mm×96 mm and the corresponding visibility maps of the total grating area. We choose this example as it shows one grating with lower visibility and almost the same visibility for three gratings in the range of 36% to 40% with a low standard deviation of less than 2% which is the usual variation of single gratings. This demonstrates a very good alignment of the three grating tiles with respect to each other. The lower right tile’s visibility is reduced to 29.4%, which in a first guess might be attributed to an angle misalignment of this grating. Theoretical considerations lead to the conclusion that a decrease in visibility in the range of 20% would result in the configuration of this experiment (Pixel size: 172 mm, period 5.4μm) only from an angle misalignment of around 60 mrad. As we measured a misalignment to the other gratings of around 1 mrad only the decrease in visibility is attributed to a lower grating quality which is also indicated by a higher standard deviation in visibility.
The inter-grating gaps affected a single pixel row (Fig. 4C); visibility was reduced to approx. 6% in this region. Thus, the extraction of all three contrast modalities was still possible in the full field of view (Fig. 5). Although the size of the image is only 32 mm in vertical direction due to the limit of the detector size, the influence of the stitching process on the image can be evaluated without causing overlapping effects in case of additional image stitching. The projection images of the small frog reveal only a small impact of the stitching on the image: the affected vertical and horizontal pixel row becomes best visible in the dark-field image, less in the phase image whereas it disappears in the absorption image, as the arrows in Fig. 5 indicate. The assembly was mechanically stable; no drifting was observed.
Conclusions
This first proof-of-principle demonstrates the great potential of stitching of X-ray gratings to increase the field of view significantly. We demonstrated that cutting and assembly can be done within the required precision. Being an add-on to the previous established grating process without causing any changes is a main advantage of the stitching process. This also implicates that the process could be readily used for gratings fabricated by any other method.
First imaging results of a frog using a 2×2 stitched analyzer grating in a Talbot-Lau interferometer show a high image quality and the visibility is homogenous all over the field of view apart from the inter-grating gaps. In the affected pixel row the visibility was reduced, but not zero.
Nevertheless, the implementation into a fabrication process requires further development; alignment tolerances have to be identified depending on the pixel size and an alignment method has to be developed to ensure maximum alignment precision for all tiles. Low absorbing materials suitable for carrier substrates and an adapted adhesive have to be found. Furthermore, due to the increased grating area the shadowing problem of planar gratings in X-ray tube set-ups with cone-beam geometry has to be solved [43].
Footnotes
Acknowledgments
The authors want to thank the cleanroom team of IMT, especially Alexandra Karbacher, Christin Straus, Julia Wolf, Barbara Matthis and Martin Börner, for their immense support during grating fabrication and Klaus-Martin Reichert for his support in image acquisition. Furthermore, we thank Abel Gil Villalba, Harsh Patel and Tillmann Volz.
This work was carried out with the support of the Karlsruhe Nano Micro Facility (KNMF, www.knmf.kit.edu), a Helmholtz Research Infrastructure at Karlsruhe Institute of Technology (KIT, 
). It is part of the Helmoltz virtual institute NXMM. Jan Meiser is a member of the Karlsruhe School of Optics and Photonics (KSOP).