Nanoscale mechanism of UO2 formation through uranium reduction by magnetite

Uranium (U) is a ubiquitous element in the Earth’s crust at ~2 ppm. In anoxic environments, soluble hexavalent uranium (U(VI)) is reduced and immobilized. The underlying reduction mechanism is unknown but likely of critical importance to explain the geochemical behavior of U. Here, we tackle the mechanism of reduction of U(VI) by the mixed-valence iron oxide, magnetite. Through high-end spectroscopic and microscopic tools, we demonstrate that the reduction proceeds first through surface-associated U(VI) to form pentavalent U, U(V). U(V) persists on the surface of magnetite and is further reduced to tetravalent UO2 as nanocrystals (~1–2 nm) with random orientations inside nanowires. Through nanoparticle re-orientation and coalescence, the nanowires collapse into ordered UO2 nanoclusters. This work provides evidence for a transient U nanowire structure that may have implications for uranium isotope fractionation as well as for the molecular-scale understanding of nuclear waste temporal evolution and the reductive remediation of uranium contamination.


Supplementary Notes
Supplementary Note 1. Selected area electron diffraction (SAED) measurement. The reduction of uranium oxides results in phase transitions which can be determined indirectly from crystallographic analysis of the structures in TEM samples. Thus, we endeavored to identify the valence state of U-bearing nanoparticles using electron microscopy and SAED. We performed a fast Fourier transform (FFT) analysis of the HR-STEM images shown in Figures 3 to 5 in the main text. While HR-STEM approaches can directly map structures of phases on the atomic structure, it is not a rigorous crystallographic analysis and can result in the ambiguous or incorrect identification of phases. Robust crystallographic identification of phases can be implemented by either performing Multi-Slice simulations of HRSTEM imaging, or by taking systematic SAED measurements of nanometric regions of sample materials. We chose SAED to complement our FFT analysis of HR-STEM.
Definitive identification of uranium oxide phases requires both the proper indexation of the SAED data and the comparison of diffraction data to standards with known phases and valence states. We prepared TEM specimens from UO2, U3O8, UO3 materials to compare with the 72-hour samples from the same stock used in the EELS measurements and microscopy investigations. We used the same methods for preparation as for the samples used for EELS measurements. The collected solids were dispersed into 70% ethanol solution, sealed in a serum bottle and sonicated for 3 min anoxically. A drop of the sonicated suspension was then deposited onto an ultra-thin carbon grid (Electron Microscopy Sciences CF200-CU-UL; 200 µm square mesh; 3-4 nm carbon foil; copper grid; silicon free) and was immediately transferred into a vacuum desiccator for preservation before the measurement. The sample spent less than 15 mins under ambient conditions before being introduced into the vacuum of the TEM.
We conducted SAED measurements on a JEOL 2100 LaB6 operating at 200 kV. As noted for the EELS measurements, the TEM samples can incur damage from the electron beam. We cooled the samples to ~108K in a Gatan 620 LN2 cooling stage to reduce the effects of radiolysis and degradation of the samples under the electron beam. Colella et al. (2005) observed that the uranium oxide phase could have reduced under the electron beam at these temperatures with extended exposure and total doses. We minimized the total exposure time to a few minutes and limited the beam current on the sample to ~1nA, thus reducing the total dose below that of EELS measurements, ~1x10 25 electrons/m 2 . For the SAED measurements in a low dose configuration, we used spot size #5 and inserted a small 70 μm condenser aperture, which produced beam currents of ~1 nA in a spread beam having a diameter ~ 2 μm (Koehler, parallel illumination condition). Under such microscope conditions, the beam damage was minimized, and most samples were stable and did not change phase or evolve within the 1 to 2 min exposure period. However, we did note changes to the UO3 standard samples, which appeared to be more beam sensitive than the other samples analysed by SAED. During the short exposure, diffraction spots disappeared or changed symmetry, indicating that the UO3 samples may have transformed under the electron beam. Diffraction analysis confirmed that not all diffraction reflections could be ascribed to the crystal structure of UO3-1540845.
Simulated diffraction patterns were generated using JEMS software 1 and compared to the experimental patterns. To improve the analysis and phase identification, we radially integrated the experimental SAED patterns and plotted their diffraction intensity distribution with reciprocal space using the PASAD plugin within in Gatan DigitalMicrograph® software 2 . The following PDF crystal files data were used to generate the simulated patterns and to index the experimental patterns: UO2-1541665, U3O8-2310811 and UO3-1540845 from the Crystallography Open Database. Supplementary Figure 7 shows a montage of SAED patterns, and the profiles of simulated SAED patterns generated from the PDF files were overlaid and compared to the experimental data. Due to the large crystal sizes, the experimental SAED data of the UO2 standard sample have a spotted pattern rather than a powder ring pattern ( Supplementary Figure 7 a,b). The comparison shows that both UO2 standard and nanowires (72-hour) structure SAED fit with the UO2-1541665 PDF file simulated pattern ( Supplementary Figure 7 c,d). A good fit to the PDF file U3O8-2310811 was found for U3O8 standard (Supplementary Figure 7 e,f). As for the UO3 standard, the SAED pattern changed during measurement, and the sample evolved under the electron beam, potentially incurring damage and changing phase and valence state. Analysis of the SAED revealed that there are missing and shifted reflections for the UO3 standard data ( Supplementary Figure 7 g,h). The behavior of this sample under the low-dose conditions of SAED measurements was distinctly different from that of other standards and the 72-hour samples. Though SAED data of UO3 are fraught with artifacts, we present them here to underscore the necessity for rigor when analyzing these materials in TEM that can potentially reduce under the high-energy electron irradiation.
The scattered line plot in Supplementary Figure 8 and indexation of the patterns clearly show that the electron diffraction pattern of the nanowires in 72-hour sample contains reflections that can only be indexed to the UO2 phase, and the pattern matches well the UO2 standards and the PDF file UO2-1541665 data. We, therefore, conclude that the nanowires primarily contain reduced U(IV), which also corresponds well and validates our complementary FFT analysis of HR-STEM images and electron energy loss spectroscopy (EELS) measurements.

Supplementary Note 2. Description and motivation for the defined EELS measurement parameters.
The experimental determination of uranium valence state in oxides complexes follows closely the methodologies implemented in previous studies 3,4 . We adapted those approaches in these experiments due to specific experimental challenges, e.g., radiation damage and the difficulties posed by spatially probing different nanostructures comprising mixtures of uranium IV, V, and VI valence states. In general, investigating uranium reduction reactions with electron microscopy and electron energy loss spectroscopy (EELS) is challenging since the electron beam irradiation at high doses can directly reduce uranium, and the limitations in the EELS spectrometer hardware confine our observations to a limited number of approaches. As such, we developed rigorous methods to avoid experimental artifacts from beam-induced effects and to provide statistically relevant investigations and precise determinations of the valence states through the use of well-characterized uranium oxide standards. In the following, we detail a step-by-step procedure for determining the valence states by calculating the branching ratio between the M4 and M5 edges of uranium and comparing these ratios with well-characterized uranium oxide standards.
The branching ratio is calculated by measuring the integrated counts under the M edges and taking the ratio of M5/(M4+M5). This calculation is performed through a series of operations on the raw spectrum data (Supplementary Figure 9a). First, we subtract the background using the fitting algorithms contained within the Gatan DigitalMicrograph® software using a selection window on the pre-edge (Supplementary Figure 9b). Then, we take a second derivative of the backgroundsubtracted spectrum using routines contained within the Gatan DigitalMicrograph® software package that calculates an approximate second derivative (Supplementary Figure 9c). The parameters of the second derivative calculation averages over an interval defined as 'w+' units wide, which are ideally set to match the edge width, and subtracts half of the averages in the two adjacent "wings" of the edges, having a width defined as 'w-' in the software. The spectrum is then divided by the squared sum of 'w+' and 'w-' to yield an approximation of the second derivative with respect to energy. For spectra measured at a dispersion of 0.25 eV/per channel, we used window widths of 5 and 10 eV, respectively, for 'w+' and 'w-'. In the final step, we determined the integrated counts under the positive peaks of the M4 and M5 edges and extracted the integral counts. Using the extracted integral counts, we calculated the branching ratio, M5/(M4+M5).
We chose to acquire spectra at dispersions of 0.25 eV/per channel which provided sufficient energy resolution to produce well-defined M edges above background and enough range to measure the pre-edge and post-edge backgrounds for both M4 and M5 edges in a single-spectrum. Because the electron loss is higher than 2,000 eV, the drift tube and prism excitations of the spectrometer had to be adjusted to provide the appropriate range between 3,450-3,940 eV. Thus, dual EELS acquisition techniques could not be used in which both the low loss regime with the zero-loss peak and high low loss spectrum with the M edges can be acquired simultaneously. Dual EELS acquisition is useful for calibrating the absolute edge energy position and determining chemical shifts in the edge energies associated with changes in valence. Combined with the use of the monochromator in the Titan Themis, it is possible to measure valence states from chemical shifts precisely and also differentiate them spatially using scanning TEM based techniques. However, the chemical shifts associated with different valence states in uranium are below 1 eV, which cannot be measured with Dual EELS acquisition. Furthermore, the use of high dispersion to observe both edges does not permit sufficient energy resolution for accurately determining the valence states from these methodologies. For these reasons, we adapted the branching ratio measurement schemes of Colella et al. (2005) to locally identify the valence states of different nanostructures within the samples. Colella et al. (2005) also reported on the challenges of measuring valence states of uranium due to electron beam damage that can reduce the uranium in oxides, giving an observable change in the branching ratio. An inherent consequence of measuring valence states from inelastically scattered electrons is that the sample must be ionized by the beam. Furthermore, to obtain appreciable signals at the high loss M edges of uranium (≥ 3,550 eV), we must irradiate the sample with high doses. There is, thus, a trade-off between having sufficient signal-to-noise (SNR) in the measured edges and limiting the amount of beam-induced effects. As the samples in our experiments are dispersed nanostructures, the minuscule sample volume also complicates the EELS measurements, and as a consequence, we must increase the local dose on the sample in the EELS measurements by using electron beam currents of several nano-amperes to obtain welldistinguished edges above background. Also, to obtain well-resolved edges, we used exposure times of 5 s, summed 10 spectra together and acquired a high-quality (HQ) dark reference to improve SNR and energy resolution. In utilizing these acquisition parameters, the samples were exposed for several minutes at high beam currents. These high doses reduce the TEM samples at ambient temperatures. According to the findings of Colella et al. (2005), which were based on the analysis of uranium oxide standards, an increase in the branching ratio signifies the reduction of uranium. To minimize the beam-induced effects that chemically reduce the sample through a combination of heat and ionization damage, we cooled the sample to 108 K in a liquid nitrogen specimen cooling stage and operated the microscope at high tension of 300 kV in TEM mode.
We used diffraction-coupled EELS geometry with small camera lengths of 29 mm to collect a significant amount of the inelastic signal, improving the SNR, and reducing the exposure time and total dose. Diffraction-coupled EELS also allowed to spatially isolate different regions of the sample that contribute to the spectrum data, and thus, we could probe variation in valence states between the different nanostructures. Using the smallest selected area electron diffraction (SAED) aperture of 10 μm, we isolated areas with a diameter of ~200 nm which was sufficient, e.g., to separately probe the valence states of uranium on magnetite particles and the U-containing nanowires (Supplementary Figure 9d,e). Ideally, the illumination of the sample should be parallel for diffraction-coupled EELS. However, we used a slightly converged beam (~1 mrad) for two primary reasons: (1) regions with suitable thickness and particle density were limited in the TEM samples for which we wanted to restrict and control the amount of dose in the surrounding areas that were interrogated in subsequent measurements, and (2) we needed to increase the current in the spectrometer to 2 nA, requiring us to converge the beam. To provide statistically relevant measurements, we used a fixed C2 lens current and illumination area. Having a 2 nA current in 200 nm diameter area gives an approximate dose rate of ~4x10 23 electrons/m 2 /s. We aimed to have the high current illumination of the sample and maintain the measurement time under 2 minutes. This long exposure time includes both the acquisition of spectra and the setup of the measurement. To reduce noise and improve SNR, we used 5 s exposure and summed 10 spectra together using the automated algorithm within Digital Micrograph, and the total acquisition of the spectra was roughly one minute. Thus, we limited measurements to a total dose of ~5x10 25 electrons/m 2 . Though no prior studies have been reported on the critical dose and dose rate, the measured branching ratios in our study suggest that the irradiation conditions under the liquid nitrogen conditions used in our experiments were below the threshold to induce the reduction of uranium oxides.
To provide internally consistent data sets and to account for beam-induced effects, we performed "benchmark" measurements on uranium oxide standards having known valence states at the same microscope conditions and temperatures and under the same dose and dose rate conditions. We obtained the following standards for the three different valence states, uraninite for U(IV), UMoO5 for U(V), and UO3 for U(VI). We performed a minimum of 10 measurements per standard. We chose regions in the TEM samples that were comparably as thin as the nanowire samples such that background signals and EELS edge shapes were similar. Using Gatan DigitalMicrograph® software, we processed the spectra under the same routines and determined the branching ratio the median values from 0.6956, 0.6887 to 0.6640, for the U(IV), U(V), and U(VI) standards, respectively (Supplementary Table 3). As discussed above (Supplementary Note 1), partial U reduction was observed under the beam for UO3, even at LN2 temperatures and low-dose beam conditions. As reduction occurs rapidly in these particularly electron beam-sensitive samples, it is difficult to control and avoid during measurements. As a result, UO3 exhibits the largest range of EELS branching ratio values of all the samples (Fig. 7) and its median value is likely an overestimate. . FT-EXAFS spectra in R space. Uraninite and U(VI) adsorbed on ferrihydrite 5 were used as reference standards. The Fourier-transformed EXAFS signal of samples taken in the early stages of the reduction (from 4 to 24 hours) indicates the appearance of a more crystalline structure as a function of reaction time. The 3.8 Å peak is a distinct feature for U-U pair correlation in the uraninite crystal structure, and its amplitude increases, particularly at 24 h. Sample spectra were collected at I20, DLS. U(IV) reference spectra were collected at B18, DLS. Source data provided as a Source Data file.