A preliminary investigation into the use of molecular oxide and hydride secondary ion relationships for improvement of the 236U/238U determination on a NanoSIMS 50L

A NanoSIMS 50L is used to investigate uranium molecular (235U16O, 236U16O, 238U16O, 235U1H, 238U1H, 236U16O1H, and 238U16O1H) and elemental (235U, 236U, and 238U) secondary ion production during sputtering of synthetic UO2 and the NIST-610 standard to determine if: (1) the 236U16O/238U16O molecular oxide ratio performs better than the 236U/238U elemental ratio, and (2) there is co-variance between the molecular hydrides and oxides. Despite an order of magnitude greater abundance of 236U16O secondary ions (compared to 236U), the 236U16O/238U16O ratios are less accurate than the 236U/238U ratios. Further work is needed before the higher count rate of the 236U16O secondary ion can be used to obtain a better 236U/238U ratio. The second objective was undertaken because correction for the interference of 235U1H on the 236U secondary ion species typically utilizes the 238U1H/238U ratio. This becomes problematic in samples containing 239Pu, so our aim was to understand if the hydride formation rate can be constrained independently of having to measure the 238U1H. We document correlations between the hydride (238U1H and 238U16O1H) and oxide (236U16O) secondary ions, suggesting that pursuing an alternative correction regime is worthwhile.

the hydride formation rate to the observed 236 U/ 238 U using the formula 236 U/ 238 U = (( 236 U + 235 U 1 H)/ 238 U) − ( 238 U 1 H/ 238 U) × ( 235 U/ 238 U) 2 to obtain a corrected 236 U/ 238 U value. While this correction regime has been shown to work for uranium bearing samples that do not contains plutonium, the presence of 239 Pu within the sample makes the correction regime unusable due to the fact that an MRP of 37,056 is necessary to separate 238 U 1 H from 239 Pu.
While the 238 U 1 H/ 238 U based hydride correction regime has been shown to work for samples that do not contain Pu, other possible issues relate to the fact that a comparatively high count rate of the 238 U 1 H is then being applied to the much lower signal at mass 235 U 1 H + 236 U by way of the 235 U/ 238 U ratio. In theory, proper detector background and deadtime correction regimes should mitigate any issues related to the drastic differences in count rate between the various secondary ion species, but in practice, these differences could influence the 236 U/ 238 U determination. This is especially true in situations where the 236 U and 238 U 1 H are being measured at different times during an analysis (e.g. during magnetic peak hopping) and the molecular hydride formation rate is unstable over the course of that analysis. On Fig. 1, which is a compilation from the literature of reported 236 U/ 238 U ratios plotted as a function of the reported 235 U/ 238 U ratio, it can be seen that there is a broad increase in the reported 236 U/ 238 U ratios with increasing 235 U/ 238 U. Furthermore, this increase (which can be modeled utilizing a power-law curve) broadly follows the hypothetical 235 U 1 H/ 238 U ratios calculated at any given 235 U/ 238 U ratio assuming a 1% hydride formation rate. While there are obviously many nuances specific to the data from each study, the existence of this relationship between the reported 236 U/ 238 U and 235 U/ 238 U ratios suggests that, broadly speaking, the 236 U/ 238 U ratios is not adequately corrected for the 235 U 1 H interference during SIMS analysis.
The purpose of this study is to investigate two possible avenues for improvement of the 236 U/ 238 U ratio. One possible avenue is in use of the 236 U 16 O/ 238 U 16 O ratio instead of the 236 U/ 238 U ratio. From other studies (e.g. 5 ) it is known that the 23x U 16 O molecular secondary ions are typically detected in greater abundances as compared to the 23x U elemental secondary ions. Therefore, it is possible that the increased count rate of the 236 U 16 O relative to the 236 U results in improved precision and accuracy. Use of the 236 U 16 O molecular ion, as opposed to the 236 U elemental secondary ion, to determine the abundance of 236 U, relative to the other isotopes of U in a sample, has not yet been reported in the scientific literature. The second avenue is more exploratory in nature and involves an assessment of whether the pertinent (e.g. 238  has not yet been attempted, or even explored, in the scientific literature. These goals are achieved by examination of UO 2 with a known uranium isotopic composition, as well as the NIST-610 standard whose isotopic composition has previously been documented, on a variety of different substrates and analytical conditions. In the end, we show that both avenues for improvement are viable if a variety of issues can be resolved.

Results: NanoSIMS data processing and presentation
The raw data from the NanoSIMS, consisting of the total number of counts observed during each of the 6.5 s cycles, was exported from the instrument and processed in Microsoft Excel. The raw data was uncorrected for the electronically programmed 44 ns electron multiplier deadtime, so this correction was applied to the raw data followed by linear interpolation of the secondary ion count rates during each cycle to the time of the magnetic field (B7) containing that cycle's 238 U, 238 U 16 O, and 238 U 16 O 2 secondary ion data. This was done to account for any drift in the secondary ion count rates over the course of each cycle. The linearly interpolated and deadtime corrected secondary ion counts were then utilized to calculate the various ratios of interest at each cycle. These cycle-by-cycle ratios were then averaged over the 20 cycles of data collected for each analysis. These ratios are reported in supplementary table 2a and utilized in plots 2 through 9. The error reported for each ratio is the within-run uncertainty on each ratio calculated as standard deviation of the mean value of the 20 cycles of data.

Discussion
The primary goal of this study is to investigate two potential avenues for improvement of the 236 U/ 238 U ratio determination by SIMS. One potential avenue is by use of the 236 U 16  H molecular hydride secondary ion production rates and various other combinations of the uranium elemental and molecular oxide ratios. If such co-variance were to exist, it could mean that it may ultimately be feasible to constrain the uranium hydride secondary ion formation rate taking place during an analysis independently of the assumption that the entire signal at mass 239 is composed entirely of 238 U 1 H. As discussed in the introduction, the approach of utilizing the 238 U 1 H/ 238 U ratio to correct for the contribution of 235 U 1 H to the 236 U/ 238 U ratio has been shown to work in samples that do not contain Pu, but this approach cannot work for mixed U-Pu samples. Therefore, consideration of alternative means of constraining the U-hydride secondary ion formation rate during an analysis is a worthwhile endeavor. It is first necessary to consider whether the uranium isotopic data collected in this study is of high enough quality to investigate these possibilities. This can be ascertained on the basis of whether the isotope ratios behave according to the expected relationships between minor isotope count rate, within-run uncertainty, and deviation between the measured and true values. In Fig. 2, the fractional 1σ within-run uncertainty (Fig. 2a) and observed/true (Fig. 2b) versus the total number of 235 U counts for the 235 U/ 238 U ratio associated are shown for each analysis of the UO 2 and NIST-610 glass. As can be seen, the within-run 1σ (Fig. 2a) and degree of deviance ( Fig. 2b) both decrease exponentially with increasing count rate of the minor isotope ( 235 U). The same is true of the 235 U 16 O/ 238 U 16 O within run 1σ (Fig. 3a) and observed/true (Fig. 3b) 7 for the NIST-610 glass of 235 U/ 238 U = 0.0023856 (7). Similar behavior is seen in the uncorrected 236 U/ 238 U (Fig. 4) and 236 U 16 O/ 238 U 16 O ratios (Fig. 5) ratios, as well as the 234 U/ 238 U ratio (not shown in figures, but data is provided in supplementary table 2). The fact that minor isotope count-rate appears to be the major contributing factor to the precision www.nature.com/scientificreports/  www.nature.com/scientificreports/ and accuracy of the observed isotopic ratios indicates that the U isotope data collected in this study behaves in-line with what would be expected for mass spectrometric isotope data. Therefore, the dataset can be used to further investigate the two possible avenues for improvement of the 236 U/ 238 U ratio determination outlined in the preceding paragraph. It is also important to consider the factors impacting hydride formation rate during the analytical session within-which the data presented in this study was collected. 235 U 1 H hydride formation rate as it applies to the 236 U/ 238U determination on micrometer sized particles by SIMS has recently been explored by Simons and Fassett (2017), with a basic observation being that the composition of the analytical substrate upon which the particles are sitting exerts far more control over the hydride formation rate than the residual vacuum within the analytical chamber. The data presented in this study was collected during the course of a single analytical session, wherein the vacuum level within the analysis chamber remained at ~ 4.5 × 10 -9 mbar. Examination of supplementary table 2a reveals that the 238 U 1 H/ 238 U ratios observed in this study range from as low as 0.00001 to as high as 0.03.
However, it is equally important to note that the 238 U 1 H/ 238 U ratios for the polished UO 2 and NIST-610 glass reference material are all below 0.0003, whereas the values for the UO 2 dispersed onto the various substrates are an order of magnitude higher. This could indicate that residual volatiles adhered to the irregular surfaces created by dispersing the finely crushed UO 2 particles onto the substrates contributed to hydride formation, and/or that the substrates themselves contribute substantially to hydride formation (as suggested by Simons and Fassett, 2017). In either case, obtaining a dataset displaying a range of hydride formation rates is consistent with the goals of our study, which is to explore various hydride correction regimes. An equally important note is that there does not appear to be any statistically significant (e.g. outside of within-run uncertainty) differences in the 235 U/ 238 U or 235 U 16 (supplementary table 2b). In contrast, the data from the un-greased substrates are pulled, on average, towards values lower than the solution value. In some cases, the corrected 236 U/ 238 U ratios are ≤ 0. For the polished UO 2 mounted in epoxy, the correction regime  There are several possible reasons for this. One may be that the hydride formation rate at the time of the 238 U 1 H acquisition is higher than that at the 236 U acquisition. Examination of the secondary ion count rate data does indicate that for some of the analyses (full dataset provided in the electronic appendix), the 238 U 1 H does increase as a function of time during the analytical session whereas in others, it remains stable or decreases. This variability in the count-rate as a function of time during the analysis is the reason why the raw secondary ion count-rates were linearly interpolated to the time of the 238 U acquisition prior to additional data processing. However, the reality is that the very low 236 U count rate (sometimes < 1 c/s) complicates the linear interpolation method. Equally problematic is the fact that the hydride formation rate, as constrained by monitoring the relatively strong signal at mass 238 U 1 H, must then applied to the considerably lower signal at mass 235 U to infer the number of 235 U 1 H ions being produced at the time of the 236 U acquisition. While other factors may be important, the 236 U/ 238 U ratio overcorrection is most likely related to the large difference between the count rates at masses 235, 236, and 239. This is especially true for those analyses where the 238 U 1 H increases as a function of time whereas, the 236 U and 235 U count rates are too low to exhibit a noticeable increase or decrease as a function time during any of the analyses.
While the 236 U count rate is sometimes < 1 c/s for the analyses conducted in this study, the 236 U 16 O count rates were typically > 10 c/s. This higher count rate results in a smaller within-run 1σ (Fig. 4a vs 5a), but as noted earlier the raw 236 U 16 O/ 238 U 16 O ratios exhibit more scatter as compared to the raw 236 U/ 238 U ratios (Fig. 4b vs  5b). As with the 236 U/ 238 U ratio, it is necessary to account for an interference on the 236 U 16 16 O ratios are associated with better withinrun precision is promising. If the sources of uncertainty introduced by the hydride correction regime can be resolved, it may ultimately be feasible to take advantage of the improved 236 U 16 O count rate. However, this will require a better understanding of the molecular oxide and hydride formation dynamics, which will be discussed in the subsequent section.
Evidence for coupling between the U molecular hydride and oxide secondary ions. As discussed in the preceding section, developing a better understanding of molecular oxide and hydride secondary Table 1. NanoSIMS Analytical conditions utilized in this study. All analyses of UO 2 were conducted with a 200 pA Oprimary beam rastered over a 5 × 5 µm area. In contrast, the NIST-610 glass was analyzed using a scanning 2 nA primary beam. All analyses consisted of 6.5 s per cycle at each mass station, with a total of 20 cycles of data for each analysis. This discussion will now seek to understand whether the secondary ion molecular U hydride formation rate can be constrained independently of the signal at mass 239 ( 238 U 1 H) or 255 ( 238 U 16 O 1 H). While there is still considerable uncertainty in the SIMS community over the mechanisms governing molecular secondary ion formation, co-variance between different elemental and molecular oxide species has been documented and is actually the basis for certain inter-element correction regimes utilized in other disciplines (see discussion by 8 ). Therefore it is possible that there will be some co-variability between either the 238 U 1 H/ 238 U and/or 238 U 16  In practice, such an approach would obviously require the existence of matrix matched reference materials as well as some other analytical considerations. However, a first step is simply to examine whether there is any evidence of this co-variability. The dataset collected in this study allows us to answer this question since the UO 2 was examined across a range of different substrates and preparation routes that presumably influence both the hydride and non-hydride molecular secondary ion formation rates. Therefore, if co-variability exists, it should be evident in the dataset collected in this study.
In Fig. 6a, it can be seen that there is a weak positive correlation between the uncorrected 236 U/ 238 U and the observed 238 U 1 H/ 238 U ratio. However, the vast majority of the datapoints are clustered towards lower 238 U 1 H/ 238 U ratios, whereas the data points with higher 238 U 1 H/ 238 U are also those with the highest levels of within-run uncertainty on the uncorrected 236 U/ 238 U. The lack of a strong positive correlation between the uncorrected 236 U/ 238 U and the observed 238 U 1 H/ 238 U ratio is not surprising given the observations made in the preceding section, that correction of the 236 U/ 238 U using the 238 U 1 H/ 238 U ratio did not uniformly improve the 236 U/ 238 U determination. Unsurprisingly, in Fig. 6b there is no positive correlation between the uncorrected 236 U/ 238 U ratios and the observed 238 U 16 O/ 238 U ratio. As discussed earlier, it is likely that scatter in the 236 U/ 238 U ratios induced by the low count rate of the 236 U secondary ion supersedes any relationships that may exist between the 235 U 1 H and 238 U 1 H hydrides (which should be one-to-one). It is also unclear whether the magnitude of the hydride interference induced shift in the uncorrected 236 U/ 238 U ratios, away from their true values, co-varies with the rate of formation of a non-hydride molecular oxide secondary ion (Fig. 6b).
In contrast to the lack of any correlations between the 236 U/ 238 U and the observed 238 16 O/ 238 U ratio. This implies that formation of the hydrogen-bearing molecular oxide secondary ion is at the expense of the 238 U 16 O molecular oxide, which is confirmed by examination of Fig. 8a where it can be seen that the 238 U 16 16 O. Both goals require a practical assessment of how the various secondary ion species behave. As explained above, molecular oxide and hydride production rates do appear coupled. In theory, this coupling could be applied to the analysis of an unknown by taking an approach similar to the reference materials based Scientific Reports | (2020) 10:12285 | https://doi.org/10.1038/s41598-020-69121-9 www.nature.com/scientificreports/ calibration curves that are widely utilized in SIMS. For example, in considering that the 238 U 1 H and 238 U 16 O 1 H production rates are coupled to the UO/U ratio, one approach might be to establish this relationship using standards of known 236 U/ 238 U and that are matrix matched to the unknown in question, and then use the curve to apply a correction to the observed 236 U/ 238 U or 236 U 16 O/ 238 U 16 O during an analysis of the unknown based on that analyses' observed 238 U 16 O/ 238 U ratio. The dataset does not currently exist to further evaluate such an approach. Application of a correction regime derived from the UO 2 dataset cannot be appropriately applied to the NIST-610 measurements since these were made using a different primary beam condition. However, the observations made in this study certainly give promise to the possibility that a correction regime for hydride formation independently of having to measure the 238 U 1 H species is worth pursuing.

Conclusions
Consideration of the NanoSIMS data for the 235    H) formation rates as a function of the non-hydride molecular oxide production rates taking place within a particular analysis. This conclusion is based on the existence of co-variance between the molecular hydride and oxide production rates. However, more work is needed to further assess this possibility.

Methods
Description of materials and mounting techniques. The UO 2 analyzed in this study was produced by calcination of UO 3 (produced via internal gelation as described in 9 ) spheres at 600 °C for 5 h followed by sintering at 1,700 °C for 3 h in a reducing atmosphere (Ar w/4% H 2 ). A large microsphere (≥ 500 µm diameter) of this sintered material was randomly selected from the batch and coarsely crushed before being crudely dispersed onto the various substrates utilized for NanoSIMS analysis by using a pair of stainless-steel tweezers. A small shard of the crushed material was also routed for uranium isotopic measurements via solution multi collectorinductively coupled plasma-mass spectrometry (MC-ICP-MS) which will be described in the subsequent paragraph. For NanoSIMS analysis, the substrates consisted of a polished carbon planchet (Ted Pella, Inc), a silicon wafer (Nova Electronic Materials), high purity (> 99.99%) platinum foil (Aldrich), a polished aluminum billet, and a carbon sticky tab (Ted Pella, Inc). Each of the substrates was observed to produce a 238 U secondary ion signal ≤ than the detector background of 0.01 c/s (averaged over 5 min). Each of the substrates (except the sticky tab) were prepared to receive the UO 2 in two ways. One way was completely bare such that the particles were only adhered to surface electrostatically. The other way was with a thin coating of Apiezon-L grease. The grease coating was applied by smearing a small quantity of it onto the substrate followed by use of a cotton cleanroom wipe to smoothen and remove the bulk of the grease such that only a thin veneer of grease remained on the substrate.  www.nature.com/scientificreports/ In addition to the crushed UO 2 mounted on the various greased and ungreased substrates, a cross-sectioned and polished sphere taken from the same batch was also analyzed. This sphere was mounted in Buehler Epothin2 epoxy followed by use of silicon-carbide and diamond based abrasives (down to ¼ µm grit) to produce a flat surface that was then coated in 50 nm of Au using a Cressington 208 HR sputter coater outfitted with an MTM-20 thickness controller prior to analysis on the NanoSIMS 50L. A shard of the NIST-610 glass reference material was also analyzed. It was prepared in the same way as the cross-sectioned and polished UO2. The U isotopic composition of the NIST-610 glass, determined via multiple techniques, has previously been reported by 7 as follows: 234 U/ 238 U = 0.00000945(5), 235 U/ 238 U = 0.0023856 (7), and 236 U/ 238 U = 0.00004314(4). NanoSIMS analysis. The various sample formats (greased and ungreased substrates containing crushed UO 2 as well as the polished UO 2 and NIST-610 glass) were all analyzed using a NanoSIMS 50L (described in 10 ) at Oak Ridge National Laboratory in August of 2019. The instrument used in this study was equipped with the Hyperion-II radio-frequency plasma oxygen ion source (described in 11 ). For the U isotopic measurements, the NanoSIMS primary column was tuned through use of the L1 and L0 lenses and D1 aperture to achieve a ~ 1 µm diameter 200 pA beam of O− ions on the sample surface. The exception to this beam current is for analyses on the NIST-610 glass, which only contains ppm levels of U. For these analyses, the L1 and L0 lenses were tuned to achieve a 2 nA beam. The NanoSIMS 50L entrance and aperture slits were tuned, in conjunction with the quadrupole lens, to achieve a mass resolving power of ~ 7,000 (m/∆m) at ~ 40% relative transmission. While the NanoSIMS 50L used in this study is equipped with seven moveable detector positions (each one consisting of an inter-changeable electron multiplier and Faraday cup), mass dispersion is insufficient at the mass ranges used in this study to be able to analyze the various U isotopes within the same magnetic field. Therefore, individual analyses were conducted using a peak hoping approach with the detector and magnetic field configuration outlined in Table 1. The secondary ion imaging capability was utilized to identify the locations where individual analyses were performed within a broader region of interest (typically 50 × 50 µm area). Following a minor tuning of the secondary ion extraction and steering optics to account for slight topographical variations between www.nature.com/scientificreports/ analysis positions, data were collected as the primary beam was scanned over a 5 × 5 µm area using the Cameca 'isotopes' acquisition mode for twenty cycles (each cycle lasting 6.5 s; total analysis time ~ ≈ 70 min including magnet cycling and settling times). Prior to starting the analyses reported in this study, detector noise levels were observed to be within the instrument's factory specifications (< 0.01 c/s averaged over five minutes). The data exported from the NanoSIMS for processing (described in "Discussion") was in the form of counts/cycle and was uncorrected for detector deadtime.
Solution MC-ICP-MS. As mentioned in "Use of the 236 U 16 O/ 238 U 16 O vs the 236 U/ 238 U ratio", an aliquot (consisting of a single shard) of the crushed UO 2 was also analyzed via MC-IPC-MS to obtain its uranium isotopic composition. The UO 2 was digested in 0.5 mL of 4 M HNO 3 for ~ 2 weeks at ambient conditions before analysis on a Thermo Scientific NeptunePlus MC-IPC-MS equipped with a jet interface and nuclear package. During analysis 234 U, 235 U, 236 U, and 238 U were placed on adjacent Faraday cups connected to 10 11 and 10 13 Ω electronically calibrated amplifiers using a 0.3 gain calibration card. The tau factor as well as a 20-min Faraday cup baseline measurement were performed. Sample solutions were introduced with a ~ 52 µL min −