Here, we report on a new record in the acquisition time for fast neutron tomography. With an optimized imaging setup, it was possible to acquire single radiographic projection images with 10 ms and full tomographies with 155 projections images and a physical spatial resolution of 200 µm within 1.5 s. This is about 6.7 times faster than the current record. We used the technique to investigate the water infiltration in the soil with a living lupine root system. The fast imaging setup will be part of the future NeXT instrument at ILL in Grenoble with a great field of possible future applications.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
High sensitivity to hydrogen, lithium and their compounds, as well as isotope sensitivity are some of the key properties making neutrons a versatile probe for diverse imaging applications . Neutron imaging is increasingly used to study distribution and transport of hydrogen rich fluids in geomaterials and engineered porous media . Specifically, 2D neutron radiography is a well-established tool to address water transport phenomena in plants, e.g. root uptake , axial transport in the xylem , external water transport in an atmospheric epiphyte  or the formation of xylem embolisms . Its 3D counterpart (neutron tomography) is an ideal approach for studying heterogeneous media but the acquisition is often too slow to study transport processes such as the water imbibition of soil aggregates , water transport in sandstones  or rocks , drying of concrete  and water transfer in root-soil systems  in 3D. Either studies were limited to 2D or focused on non-dynamic situations, for example Moradi et al. applied neutron tomography to study quasi-steady water gradients at root-soil interfaces under different soil moisture conditions . Neutron tomography has been successfully applied in energy material research to optimize the water management of fuel cells , to visualize the lithium distribution in lithium-air batteries  or to analyze hydrogen storage in hydride-graphite composites . Until recently, the slow acquisition speed (≥ 1 hour per neutron tomogram) has been one of the major limitations of neutron tomography impeding 3D observations of dynamic processes such as fluid transfers in porous systems. Recent technological improvements of neutron imaging stations along with the development of sCMOS detector technology have opened new venues for 3D imaging of fast processes at high spatiotemporal resolution. A key requirement for fast measurements with high spatial resolution is high neutron flux at the instrument, which allows maintaining sufficient signal to noise ratios even at high collimation, and low exposure times. Previous works have demonstrated that given suitable neutron flux conditions, modern imaging stations can perform tomographic scans in less than one minute while maintaining image quality sufficiently high to rigorously measure dynamic processes. This was for example recently demonstrated at the neutron imaging station CONRAD II (Helmholtz Centre Berlin for Energy and Materials, Berlin/Germany) where the fluid flow in a porous rock was investigated at a spatiotemporal resolution of 110 µm/pixel and 1 min per tomogram . Using the same instrument but even higher acquisition speed, the water imbibition in a rooted soil column was visualized with a time resolution of 10 s per tomogram at the same facility . These experiments are essential keystones towards real-time tomography. Despite these achievements, ultimately, the limiting factor is neutron flux, i.e., the development of more powerful neutron sources hold great further potential for ultra-fast tomography. A new imaging station located at Institut Laue-Langevin (ILL, Grenoble/France) and named NeXT (Neutron and X-ray Tomography) enjoys, to the best of the authors’ knowledge, the highest cold neutron flux for imaging purposes in the world. In this paper, we present first results from a pilot experiment demonstrating the potential for improvement in terms of acquisition speed and image quality of time resolved neutron tomography. The experiment was designed to study water imbibition patterns in rooted soil columns using deuterated water as tracer substance.
2. Experimental setup
2.1 Neutron imaging
The test was performed at NeXT-Grenoble, a recently developed neutron and X-ray tomography station realized at the Institut Laue-Langevin in collaboration with the Universite Grenoble Alpes . The NeXT instrument -illustrated in Fig. 1- is located at the end of a curved m = 2 neutron guide with a maximum radius 4 km, to reduce unwanted epithermal and fast neutrons as well as gamma which would derive from a direct view on the reactor. The flux at the end of the neutron guide is 1.2·1010 n·cm−2·s−1 and the neutron spectrum peaks at 2.8Å, as highlighted in Fig. 1(b). The tomograph, which currently shares its beam time with a Rainbows reflectometer, is placed at 10 m distance from a set of circular pinholes with diameters ranging from 1.5 to 30 mm. Using the largest pinhole the L/D ratio was 333 and the flux at measuring position 3.8·108 n·cm−2·s−1. The detector system consisted of a 200 µm thick 6LiZnS:Ag scintillator screen, which converts impinging neutrons to visible light, and a sCMOS camera (Hamamatsu Orca 4V2) in combination with a high aperture Canon photo lens (focal length 50 mm, aperture f./1.2). The effective pixel size after binning (2 × 2) was 100 µm yielding a spatial resolution of 200 µm as determined by the test pattern (Siemens star) shown in Fig. 2. In contrast to the conventional acquisition of neutron tomograms (stepwise sample revolution over 180° or 360°) we rotated the sample at constant speed of 0.33 rps while images were taken continuously with an exposure time of 0.01 s yielding a series of 310 radiographic projections for a complete sample rotation. Consequently, the total acquisition time per tomogram was 3 s for a 360° reconstruction and 1.5 s for a 180° reconstruction. The maximum motion blur occurring at this distal edge of the sample (Ø 27 mm) due to the continuous rotation is 137 µm when 310 projections are captured over 360° . As this value is well below the spatial resolution it does not substantially deteriorate the image quality and can be tolerated.
2.2 Imbibition test setup
Water infiltration in soil is affected not only by the inherent heterogeneities of soil, but also (and predominantly) by the interaction with plant roots and their water uptake. Plant roots are known to modify the hydraulic properties of the soil surrounding them, the so-called rhizosphere . Taking into account the complex three-dimensional shape of root system architectures and the propagation speed of water injected into unsaturated soil, high-speed neutron tomography is essential to capture imbibition patterns in root-soil systems. Figure 3 illustrates the principle of the set-up used for the water infiltration experiment. We employed an 11 days old white lupine plant (Lupinus albus), grown in a plant growth chamber under the environmental conditions detailed in . The plant container made of quartz glass (Ø 27 mm and a height of 100 mm) was filled with sandy soil collected at the Chicken Creek catchment near Cottbus. At mid height, a 1 cm thick horizontal layer of coarse sand compartmented the soil column. This layer acted as capillary barrier between the soil compartments while still allowing roots to grow across. During the imaging experiment, the plant sample was placed and secured on an aluminum dish serving as water reservoir for irrigation with deuterated water (D2O). The dish was mounted on a manipulation stage as close as possible to the scintillator. At the start of the experiment, the plant sample was watered from below with 4 ml of D2O, which was injected into the aluminum dish holding the plant container by a syringe pump (Fresenius Pilot C) at a rate of 100 ml/h. Given the high neutron contrast between hydrogen and its isotope deuterium (despite the high chemo-hydraulic similitude of their compounds normal and heavy water), D2O can be used as tracer for tracking the replacement and transfer of normal water (H2O) which partially occupies the pores in the soil column . This also helps to trace the local root uptake while keeping the overall neutron transmission of the soil sample at a sufficient level for high signal to noise ratios . The ascending water front was measured by means of a 4D image set (a time-lapsed series of 3D volumes). The tomograms were reconstructed and rendered in 3D using the software tools Octopus (Inside Matters, Gent/Belgium), IDL (Harris Spatial Resolution, Broomfield/USA) and VGSTUDIO MAX (Volume Graphics, Heidelberg/ Germany).
3. Results and discussion
We report on a new record for high-speed tomographic measurements performed at the NeXT instrument at ILL with acquisition times of 1.5 s for a full 180° tomography with 155 single radiographic projections and a voxel size of 100 µm × 100 µm × 100 µm . This improves the existing record of acquisition speed for neutron tomography by a factor of 6.7 compared to the previous record reported in  while maintaining a comparable spatial resolution. The technique was used to investigate the water uptake in a lupine root system (Fig. 4). Figure 4(a) shows a radiographic projection of the root-soil sample. Beside the basic interest of pushing the boundaries of the technique, this high acquisition speed is essential in this context given the high intrinsic speed of the process itself. The image intensity (gray values of pixels) reflects the local neutron transmission of the sample. As shown in Fig. 4(a), the root system is clearly visible as the dark ramified structure. Conversely, roots appear bright in Fig. 4(b) because the pixels in the reconstructed tomographic image display the local attenuation coefficients of the reported slice. In Fig. 4(c), a 3D-rendered view of the root system is presented together with the qualitative water distribution in the surrounding soil. In order to distinguish between roots and soil water we adjusted a set of global thresholds segmenting the 3D image. The voxels with highest attenuation values representing the root tissue (corresponding to a local water content θ > 0.5 cm3/cm3) were rendered in yellow. Voxels with medium attenuation values represent soil with a water content range of about 0.1 cm3/cm3 < θ < 0.5 cm3/cm3. The remaining voxels representing the soil with low water content, the container wall as well as the aboveground air space that were here rendered transparent. Figure 4(d) shows the segmented root system as retrieved by semi-automated region growing algorithms implemented in the software VGSTUDIO MAX. Note that with respect to the level of detail to which the root system could be retrieved, the quality of the high-speed tomogram appears sufficient to study local water uptake of the taproot and first order lateral roots, though not second order laterals.
Table 1 summarizes some key parameters describing the imaging conditions and quality of this test measurement. For comparison, the corresponding values for the prior speed record from CONRAD II at HZB  are reported in the second column. The signal-to-noise ratio (SNR = (signal intensity - background intensity) / standard deviation of signal)) was exemplarily calculated for the radiographic projection shown in Fig. 4(a) and its corresponding flatfield image. The signal intensity was determined for a 20 × 50 mm2 soil-filled region in the central part of the sample and the standard deviation for a 7 × 7 mm2 region at the center of the flatfield. The contrast between taproot and soil was calculated for the radiographic projection and the according vertical tomographic slice after 3D reconstruction using Eq. (1).4(a) and 4(b)). Table 1 documents that the improvement of speed did not compromise image quality. Even though using smaller pixel size (nominal higher spatial resolution) the signal to noise ratios (SNRs) are higher than in the previous high-speed measurement performed at CONRAD II . The contrast between soil and roots is slightly smaller due to the the higher soil water content (root-soil contrast intrinsically decreases with increasing soil water content, as the difference in attenuation between them becomes smaller). The great gain in temporal resolution can be attributed to the tenfold higher neutron flux yielding a much better imaging statistics (SNR) for high-speed imaging. Note the limiting factor in this measurement was the acquisition speed of the camera rather than the signal-to-noise ratio, which would have allowed even faster acquisition. Likewise, also applications not requiring high-speed acquisition can benefit from high flux conditions. It allows refinement of spatial resolution while preventing the slowdown of acquisition .
Fast tomography is important for studies of dynamic processes because any structural or morphological changes during tomographic data acquisition may cause artifacts in the reconstructed tomographic image. Image acquisition should be sufficiently fast to ensure only sub-pixel changes of the process. In Fig. 5, a time series of 3D-rendered images are presented to illustrate the dynamic water transfer in the soil-root system during the infiltration experiment with deuterated water. For the sake of improved image statistics, 310 radiographic projections (covering an angular range of 360 instead of 180 degrees) were used to reconstruct the tomographic images. Note that the resulting time resolution of 3 s/tomogram is still sufficient to capture the D2O infiltration in soil. D2O itself is comparatively hard to observe in the neutron images due to its lower attenuation coefficient (µH2O = 5.4 cm−1 vs. µD2O = 0.1 cm−1) but it is ideal to visualize the displacement of light water (H2O). Upon injection into the aluminum dish (see Fig. 3(a)) D2O enters the plant container quickly through the porous bottom plate, forming a water front at the bottom region of the soil column (see Fig. 5 at t = 9s). As D2O invades the pore system of the soil matrix it displaces present light water (H2O), pushing it upwards in the soil column (15 s ≤ t ≤ 75 s) until the water front arrives at the hydraulic barrier where it accumulates and shows at its upper border a sharp boundary (t = 90 s). The complete time series is provided as a video in the supplementary materials (Visualization 1). Compared to image series presented in  the boundary of the proceeding water front appears sharper, i.e. less affected by motion blurring, due to the better time resolution. Interestingly, the propagation speed of the water in this infiltration experiment has been somewhat faster (90 s vs. 140 s to arrive at the hydraulic barrier, c.f. video 1 in ), favored by the higher initial water content of soil (0.15 cm3/cm3 vs. 0.1 cm3/cm3, see Table 1) that improves the wettability and permeability promoting faster propagation of the water front. An influence of roots on the water imbibition pattern is clearly observable in Fig. 5 (45 s ≤ t ≤ 90 s). While soil water (H2O) is displaced quickly in the bulk soil, the D2O infiltration of soil around the taproot proceeds more slowly, indicating reduced water mobility in the rhizosphere during wetting, which was identified earlier and can be explained by a hydraulic modification in the rhizosphere . A more detailed analysis is beyond the scope of this work and will be subject of future studies.
We demonstrated the feasibility and potential of high-speed neutron tomography performed at NeXT/ILL and achieved with 1.5 s a new record for the acquisition time of a full tomographic image (with 155 radiographic projections at a physical spatial resolution of 200 µm). The extraordinary flux conditions maintained at the facility resulted in sufficient signal to noise ratios to perform analyses, even at such record speed. This indicates that even further acceleration of tomographic acquisition into the sub-second range is possible without compromising spatial resolution by binning. Fast acquisition of tomograms usually results in a big data output, e.g. capturing neutron tomograms over 45 min at record speed yielded 240,000 radiographic projections generating a data amount of 0.5 TB. Including post-processing the memory requirement for such a single measurement is about 1.5–2 TB. Hence, routines for handling big data are necessary, which may be adapted from existing strategies for processing fast synchrotron X-ray measurements. The present study demonstrated the feasibility of ultra-fast neutron tomography to visualize 3D flows in porous media suggesting further applications in the field of soil, material and geosciences as already discussed in .
Deutsche Forschungsgemeinschaft (396368046, OS 351/8-1, TO 949/2-1); Horizon 2020 Framework Programme (N°731096).
We thank our student assistant, Boyana Kozhuharova, for assisting the data analysis at University of Potsdam.
1. N. Kardjilov, I. Manke, R. Woracek, A. Hilger, and J. Banhart, “Advances in neutron imaging,” Mater. Today 21(6), 652–672 (2018). [CrossRef]
2. E. Perfect, C. L. Cheng, M. Kang, H. Z. Bilheux, J. M. Lamanna, M. J. Gragg, and D. M. Wright, “Neutron imaging of hydrogen-rich fluids in geomaterials and engineered porous media: A review,” Earth-Sci. Rev. 129, 120–135 (2014). [CrossRef]
3. A. Dara, B. A. Moradi, P. Vontobel, and S. E. Oswald, “Mapping compensating root water uptake in heterogeneous soil conditions via neutron radiography,” Plant Soil 397(1-2), 273–287 (2015). [CrossRef]
4. U. Matsushima, W. B. Herppich, N. Kardjilov, W. Graf, A. Hilger, and I. Manke, “Estimation of water flow velocity in small plants using cold neutron imaging with D2O tracer,” Nucl. Instrum. Methods Phys. Res., Sect. A 605(1–2), 146–149 (2009). [CrossRef]
5. W. B. Herppich, C. E. Martin, C. Tötzke, I. Manke, and N. Kardjilov, “External water transport is more important than vascular transport in the extreme atmospheric epiphyte Tillandsia usneoides (Spanish moss),” Plant, Cell Environ. 42(5), 1645–1656 (2019). [CrossRef]
6. C. Tötzke, T. Miranda, W. Konrad, J. Gout, N. Kardjilov, M. Dawson, I. Manke, and A. Roth-Nebelsick, “Visualization of embolism formation in the xylem of liana stems using neutron radiography,” Ann. Bot. 111(4), 723–730 (2013). [CrossRef]
7. A. Carminati and H. Flühler, “Water Infiltration and Redistribution in Soil Aggregate Packings,” Vadose Zone J. 8(1), 150–157 (2009). [CrossRef]
8. S. A. Hall, “Characterization of fluid flow in a shear band in porous rock using neutron radiography,” Geophys. Res. Lett. 40(11), 2613–2618 (2013). [CrossRef]
9. C. L. Cheng, E. Perfect, B. Donnelly, H. Z. Bilheux, A. S. Tremsin, L. D. McKay, V. H. DiStefano, J. C. Cai, and L. J. Santodonato, “Rapid imbibition of water in fractures within unsaturated sedimentary rock,” Adv. Water Resour. 77, 82–89 (2015). [CrossRef]
10. F. C. de Beer, W. J. Strydom, and E. J. Griesel, “The drying process of concrete: a neutron radiography study,” Appl. Radiat. Isot. 61(4), 617–623 (2004). [CrossRef]
11. A. B. Moradi, J. W. Hopmans, S. E. Oswald, M. Menon, A. Carminati, and E. Lehmann, “Applications of Neutron Imaging in Soil–Water–Root Systems,” in Soil–Water–Root Processes: Advances in Tomography and Imaging, S. H. Anderson and J. W. Hopmans, eds. (The Soil Science Society of America, Inc., 2013), pp. 113–136.
12. A. B. Moradi, A. Carminati, D. Vetterlein, P. Vontobel, E. Lehmann, U. Weller, J. W. Hopmans, H.-J. Vogel, and S. E. Oswald, “Three-dimensional visualization and quantification of water content in the rhizosphere,” New Phytol. 192(3), 653–663 (2011). [CrossRef]
13. H. Markötter, I. Manke, R. Kuhn, T. Arlt, N. Kardjilov, M. P. Hentschel, A. Kupsch, A. Lange, C. Hartnig, J. Scholta, and J. Banhart, “Neutron tomographic investigations of water distributions in polymer electrolyte membrane fuel cell stacks,” J. Power Sources 219, 120–125 (2012). [CrossRef]
14. J. Nanda, H. Bilheux, S. Voisin, G. M. Veith, R. Archibald, L. Walker, S. Allu, N. J. Dudney, and S. Pannala, “Anomalous Discharge Product Distribution in Lithium-Air Cathodes,” J. Phys. Chem. C 116(15), 8401–8408 (2012). [CrossRef]
15. C. Pohlmann, K. Herbrig, Ł Gondek, N. Kardjilov, A. Hilger, H. Figiel, J. Banhart, B. Kieback, I. Manke, and L. Röntzsch, “In operando visualization of hydride-graphite composites during cyclic hydrogenation by high-resolution neutron imaging,” J. Power Sources 277, 360–369 (2015). [CrossRef]
16. E. Tudisco, M. Etxegarai, S. A. Hall, E.-M. Charalampidou, G. D. Couples, H. Lewis, A. Tengattini, and N. Kardjilov, “Fast 4-D Imaging of Fluid Flow in Rock by High-Speed Neutron Tomography,” J. Geophys. Res.: Solid Earth 124(4), 3557–3569 (2019). [CrossRef]
17. C. Tötzke, N. Kardjilov, I. Manke, and S. E. Oswald, “Capturing 3D Water Flow in Rooted Soil by Ultra-fast Neutron Tomography,” Sci. Rep. 7(1), 6192 (2017). [CrossRef]
18. A. Tengattini, N. Lenoir, E. Ando, B. Giroud, D. Atkins, J. Beaucour, and G. Viggiani, “NeXT-Grenoble, the Neutron and X-ray Tomograph in Grenoble,” (2019 (submitted)).
19. A. Carminati, A. Moradi, D. Vetterlein, P. Vontobel, E. Lehmann, U. Weller, H.-J. Vogel, and S. Oswald, “Dynamics of soil water content in the rhizosphere,” Plant Soil 332(1-2), 163–176 (2010). [CrossRef]
20. M. Zarebanadkouki, Y. X. Kim, A. B. Moradi, H.-J. Vogel, A. Kaestner, and A. Carminati, “Quantification and Modeling of Local Root Water Uptake Using Neutron Radiography and Deuterated Water,” Vadose Zone J. 11(3), 0 (2012). [CrossRef]
21. C. Tötzke, N. Kardjilov, N. Lenoir, and A. Tengattini, “Tracking 3D water flow and root uptake in soil by means of ultrafast neutron tomography,”Institut Laue-Langevin (ILL), ed. (2018). [CrossRef]
22. N. Kardjilov, A. Hilger, I. Manke, R. Woracek, and J. Banhart, “CONRAD-2: the new neutron imaging instrument at the Helmholtz-Zentrum Berlin,” J. Appl. Crystallogr. 49(1), 195–202 (2016). [CrossRef]
23. A. Tengattini, N. Kardjilov, N. Lenoir, H. Markötter, A. Hilger, T. Arlt, and I. Manke, “High-speed sub-4 µm neutron imaging with a compact and flexible detector,” (2019 (submitted)).