A NanoSIMS 50 L Investigation into Improving the Precision and Accuracy of the 235 U / 238 U Ratio Determination by Using the Molecular 235 U 16 O and 238 U 16 O Secondary Ions

: A NanoSIMS 50 L was used to study the relationship between the 235 U / 238 U atomic and 235 U 16 O / 238 U 16 O molecular uranium isotope ratios determined from a variety of uranium compounds (UO 2 , UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 · 6(H 2 O), and UF 4 ) and silicates (NIST-610 glass and the Plesovice zircon reference materials, both containing µ g / g uranium). Because there is typically a greater abundance of 235 U 16 O + and 238 U 16 O + molecular secondary ions than 235 U + and 238 U + atomic ions when uranium-bearing materials are sputtered with an oxygen primary ion beam, the goal was to understand whether use of 235 U 16 O / 238 U 16 O has the potential for improved accuracy and precision when compared to the 235 U / 238 U ratio. The UO 2 and silicate reference materials showed the greatest potential for improved accuracy and precision through use of the 235 U 16 O / 238 U 16 O ratio as compared to the 235 U / 238 U ratio. For the UO 2 , which was investigated at a variety of primary beam currents, and the silicate reference materials, which were only investigated using a single primary beam current, this improvement was especially pronounced at low 235 U + count rates. In contrast, comparison of the 235 U 16 O / 238 U 16 O ratio versus the 235 U / 238 U ratio from the other uranium compounds clearly indicates that the 235 U 16 O / 238 U 16 O ratio results in worse precision and accuracy. This behavior is based on the observation that the atomic ( 235 U + and 238 U + ) to molecular ( 235 U 16 O + and 238 U 16 O + ) secondary ion production rates remain internally consistent within the UO 2 and silicate reference materials, whereas it is highly variable in the other uranium compounds. E ﬀ orts to understand the origin of this behavior suggest that irregular sample surface topography, and / or molecular interferences arising from the manner in which the UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 · 6(H 2 O), and UF 4 were prepared, may be a major contributing factor to the inconsistent relationship between the observed atomic and molecular secondary ion yields. Overall, the results suggest that for certain bulk compositions, use of the 235 U 16 O / 238 U 16 O may be a viable approach to improving the precision and accuracy in situations where a relatively low 235 U + count rate is expected.


Introduction
The use of SIMS (secondary ion mass spectrometry) to determine the 235 U/ 238 U isotope ratio of environmental uranium microparticles was first reported in [1]. Since then, SIMS has become a routine technique for determining the 235 U/ 238 U and other isotope ratios of uranium compounds, both on individual particles and on bulk materials (up-to-date summaries can be found in [2][3][4][5][6]).
Determination of the 235 U/ 238 U ratio from uranium compounds by SIMS has typically focused on use of the atomic 235 U and 238 U secondary ions (refer to [6] for a comprehensive summary of actinide analysis by SIMS). While use of the molecular 235 U 16 O 2 and 238 U 16 O 2 secondary ion species has been shown [3] to be a viable means of mitigating metallic molecular isobaric interferences on 235 U and 238 U for SIMS instruments with relatively low MRP, use of the molecular uranium secondary ions has not yet been systematically investigated in SIMS instruments capable of achieving high enough MRP values (e.g., the LG-SIMS instruments) to separate some of the more common molecular isobars (see discussion in [3]) that interfere with the monoatomic 235 U and 238 U. Because certain molecular isobars can be mitigated at high MRP, the primary mechanisms for improved precision and accuracy would be improved count rate by using molecular uranium oxides species ( 235 U 16 O + and 238 U 16 O + ) in comparison to the atomic 235 U + and 238 U + species.
If the NanoSIMS is to be increasingly used for determining the 235 U/ 238 U ratio, it is worthwhile to investigate whether use of the 235 U 16 O/ 238 U 16 O molecular secondary ion ratio can provide better internal precision (as reflected by the degree of scatter associated with the individual cycles comprising a single measurement) and accuracy (as reflected by the deviance, relative to the 'true' value for a given material, of the 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratios calculated by averaging the individual cycles comprising each measurement together) compared to the 235 U/ 238 U ratio. As previously mentioned, there are many sources of uncertainty and a variety of correction regimes that can be factored into the reporting of a final isotope ratio; however, a first principle is that a raw measurement displaying good precision and accuracy will produce a higher-quality final result compared to a raw measurement with low precision that deviates strongly from the 'true' value, or both.
This study shows that the 235 U 16 O/ 238 U 16 O molecular secondary ion ratio can improve the internal precision and accuracy for certain uranium bulk matrices, whereas for other uranium bulk matrices it results in no discernable improvement. Furthermore, for a few of the uranium bulk matrices examined, use of the 235 U 16 O/ 238 U 16 O ratio is associated with an order of magnitude reduction in internal precision and accuracy of the raw measurement as compared to 235 U/ 238 U ratio. This conclusion is achieved by comparison of the 235 U/ 238 U atomic and 235 U 16 O/ 238 U 16 O molecular secondary ion ratios acquired on the NanoSIMS with 235 U/ 238 U values determined by solution multi-collector inductively coupled plasma mass spectrometry (MC-ICP-MS; for the uranium compounds) and reported in the literature (for the silicate reference materials). The possible causes of this inconsistent behavior are discussed in terms of the SIMS matrix effect, carbon-and fluorine-based molecular interferences, as well as other factors known to influence the sputtering and ionization process.

Materials
Materials analyzed in this study: A variety of uranium compounds, all produced at Oak Ridge National Laboratory [17][18][19][20], (UO 2 , UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 ) as well as two silicate reference materials of known uranium isotope composition (the NIST-610 glass as well as the Plesovice zircon [17]). The NIST-610 glass and Plesovice zircons are widely used reference materials in the geosciences. Specific to this study, the uranium isotopic composition of the NIST-610 glass is 235 U/ 238 U = 0.0023956 ± 0.0000005, determined by [21]. The uranium concentration of the Plesovice zircon is described in [22], and the zircon is assumed to a have a natural 235 U/ 238 U ratio of 0.0072549 ± 0.0000008, which is the consensus value reported in [23].
The UO 2 material used was a ≥150 µm diameter sphere that was prepared from UO 3 synthesized via an internal gelation route described in [17]. The UO 3 microspheres, which were not the UO 3 that was analyzed in this study, were reduced to UO 2 by calcination and sintering for 5 h at 600 • C and 3 h at 1700 • C under a reducing atmosphere (Ar-4% H 2 ). The UO 2 F 2 was produced by hydrolysis of UF 6 in a controlled atmosphere environment. UF 6 was released into a chamber at approximately 23% relative humidity at 24 • C. Particulate UO 2 F 2 was collected on vinyl collection surfaces. The matter was calcined overnight under a stream of 10 mL/s N 2 gas at 150 • C, which has previously been shown to produce anhydrous uranyl fluoride [18].
The UO 3 analyzed in this study was prepared by heating UO 2 (NO 3 ) 2 ·6(H 2 O) in air at 225 • C for 12 h, producing an amorphous uranium oxide, but no additional steps were taken to remove residual nitrogen after calcination. The UF 4 was produced by reducing UF 6 in NO x in a two-step process (see [19,20]). First, NOUF 6 (nitrosylium hexafluorouranate) was produced by exposing frozen UF 6 to liquid N 2 O 4 . The NOUF 6 was then converted to UF 4 by devolatilizing NO x compounds at 350 • C for 120 min in the presence of H 2 gas. No additional information is available about the origin of the UO 2 (NO 3 ) 2 ·6(H 2 O). Raman analysis of the materials on the mount containing the uranium compounds was performed to confirm their bulk chemistry.

Sample Preparation for NanoSIMS Analysis
For NanoSIMS analysis, the materials described above were potted in epoxy resin (Buehler EpoThin2). Two mounts were made. One mount contained all the uranium compounds (UO 2 , UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 ), whereas the other contained the NIST-610 glass and the Plesovice zircon reference material. After allowing the epoxy to cure for approximately 1 week under ambient conditions, the mounts were ground and polished using successively finer silicon carbide-based abrasives, with a final polishing step using 0.25 µm diamond paste. An important difference between the two mounts was that the uranium compounds were polished using mineral oil as a lubricant, whereas the mount containing the NIST-610 glass and Plesovice zircon were polished in water. Mineral oil lubricant was selected for the uranium compounds because of the solubility of uranyl fluoride and uranyl nitrate in water. Methyl alcohol and acetone were used as cleaning agents for both mounts. Following a final cleaning step, and immediately before introduction into NanoSIMS airlock system, both mounts were coated with 100 nm of gold using a Cressington ® 208 HR sputter coater outfitted with an MTM-20 thickness controller.
Representative scanning electron microscope (SEM) images, collected using a Zeiss ® Merlin TM field-emission SEM, of two (the UO 2 and UO 2 F 2 ) of the uranium compounds are provided in Figure 1. Images were initially collected using a secondary electron detector with an accelerating voltage of 10.0 kV and a beam current of 1.1 nA. The high-resolution images displayed in Figure 1 were then collected using an in-column detector, with an accelerating voltage of 1.5 kV and a beam current of 80 pA. There are two images for each of the uranium compounds shown. In the left panel, a relatively wide field of view is shown, whereas in the right panel, a close-up image of NanoSIMS analytical craters made using a static 200 pA primary beam (see Section 2.3 for further details pertaining to NanoSIMS analysis) is provided. It is important to note, however, that these craters were made as part of a series of experiments that preceded the isotope ratio measurements detailed in this study. Following acquisition of the images shown in Figure 1, the mount was repolished and recoated with 100 nm of gold before being introduced back into the NanoSIMS for the analyses described in this paper. However, the pit geometries and material surfaces should be representative of the data acquired in this study because the same 200 pA primary beam conditions were used, and the material preparation steps were identical. Immediately following the NanoSIMS isotope data acquisition reported in this study, the mount containing the uranium compounds was removed from the instrument. A portion of the uranium materials were then excavated from the mount for solution isotope analysis (see Section 2.4).
Following the sample processing and characterization procedure described, and in order to specifically assess questions over the role of sample preparation on the 235 U/ 238 U vs. 235 U 16 O/ 238 U 16 O ratios that arose during data interpretation, UO 2 taken from the same batch described in Section 2.1 was mounted in epoxy using the preparation described above, crushed in an agate mortar and then dispersed onto a sticky carbon tab (9 mm Leit ® tabs), as well as pressed into high purity Indium metal. Prior to analysis (in their own session; see Section 2.3) on the NanoSIMS, the UO 2 mounted in epoxy and pressed into Indium were coated with 100 nm Au, whereas the UO 2 dispersed onto the sticky carbon tab was not. (Note, craters were made before the analyses conducted in this study, and the mount containing these materials was further polished before NanoSIMS analysis).
Following the sample processing and characterization procedure described, and in order to specifically assess questions over the role of sample preparation on the 235 U/ 238 U vs. 235 U 16 O/ 238 U 16 O ratios that arose during data interpretation, UO2 taken from the same batch described in Section 2.1 was mounted in epoxy using the preparation described above, crushed in an agate mortar and then dispersed onto a sticky carbon tab (9 mm Leit ® tabs), as well as pressed into high purity Indium metal. Prior to analysis (in their own session; see Section 2.3) on the NanoSIMS, the UO2 mounted in epoxy and pressed into Indium were coated with 100 nm Au, whereas the UO2 dispersed onto the sticky carbon tab was not.

NanoSIMS Analysis and Data Processing
The isotopic measurements reported in this study were performed using a NanoSIMS 50 L at Oak Ridge National Laboratory in November 2018. An additional session focused solely on UO2 mounted on different substrates was conducted in March of 2019. The NanoSIMS platform, including important differences relative to other SIMS instruments, is described in [12]. An important note is that the instrument used in this study is equipped with the new Hyperion-II radio frequency plasma oxygen ion source (described in [24,25]), which was used to generate the O − beam used in this study. For isotopic measurements, the NanoSIMS primary column was tuned to achieve a ~200 pA beam of O − with a diameter of ~1 µm at the sample surface ( Figure 1). This tuning was achieved primarily through the use of the L1 and L0 lenses in addition in combination with the D1 aperture. This beam was used for all analyses and tuning during the November 2018 session, with analyses performed using a static primary beam on the surface of the sample. The only exception to the beam current condition were analyses performed on the UO2 sphere, which were made using progressively lower beam currents (200 pA, 150 pA, 100 pA, 50 pA, and 20 pA). These lower beam currents were achieved by lowering the value on the L1 lens.
While the instrument was operated at an MRP of ~7000 (defined as M/∆M and achieved through the use of entrance and aperture slits as well as adjustments to the quadrupole lens), the physical properties of the NanoSIMS magnet means that it is impossible to collect 235 U and 238 U in the same magnetic field. It is also impossible to collect the 235 U 16 O and 238 U 16 O in the same magnetic field. Therefore, magnetic peak hopping was used. During each analysis, the magnet was stepped through four fields. The first field was used for settling the magnet, the second field contained 235 U, the third field contained 235 U 16 O, and the fourth field contained 238 U, 238 U 16 O, and 238 U 16 O2. Please note that 235 U Figure 1. SEM images of two of the uranium compounds (UO 2 and UO 2 F 2 ) on the epoxy mount, showing representative 200 pA analytical craters. (Note, craters were made before the analyses conducted in this study, and the mount containing these materials was further polished before NanoSIMS analysis).

NanoSIMS Analysis and Data Processing
The isotopic measurements reported in this study were performed using a NanoSIMS 50 L at Oak Ridge National Laboratory in November 2018. An additional session focused solely on UO 2 mounted on different substrates was conducted in March of 2019. The NanoSIMS platform, including important differences relative to other SIMS instruments, is described in [12]. An important note is that the instrument used in this study is equipped with the new Hyperion-II radio frequency plasma oxygen ion source (described in [24,25]), which was used to generate the O − beam used in this study. For isotopic measurements, the NanoSIMS primary column was tuned to achieve a~200 pA beam of O − with a diameter of~1 µm at the sample surface ( Figure 1). This tuning was achieved primarily through the use of the L1 and L0 lenses in addition in combination with the D1 aperture. This beam was used for all analyses and tuning during the November 2018 session, with analyses performed using a static primary beam on the surface of the sample. The only exception to the beam current condition were analyses performed on the UO 2 sphere, which were made using progressively lower beam currents (200 pA, 150 pA, 100 pA, 50 pA, and 20 pA). These lower beam currents were achieved by lowering the value on the L1 lens.
While the instrument was operated at an MRP of~7000 (defined as M/∆M and achieved through the use of entrance and aperture slits as well as adjustments to the quadrupole lens), the physical properties of the NanoSIMS magnet means that it is impossible to collect 235 U and 238 U in the same magnetic field. It is also impossible to collect the 235 U 16 O and 238 U 16 O in the same magnetic field. Therefore, magnetic peak hopping was used. During each analysis, the magnet was stepped through four fields. The first field was used for settling the magnet, the second field contained 235 U, the third field contained 235 U 16 16 O, were measured on the same detectors in the different magnetic fields. Each analysis consisted of 20 cycles through the four magnetic fields, with a 10 s/cycle count time on each mass of interest. These conditions were used for all the materials. Between materials, minor tuning of the secondary ion extraction lens and secondary ion beam steering plates was done to compensate for differences in the sample height. An important note is that no formal presputtering routine was built into the analytical sequence. This was because the secondary ion signal was observed to achieve a maximum within a few seconds of the primary beam hitting the sample surface, so it was decided that no presputtering was necessary for implantation purposes. Furthermore, there is a short period (<10 s) at the beginning of each analysis during which the primary beam dwells on the sample surface before the magnet begins cycling. As described below, the first two cycles of data collected during each analysis were excluded from consideration.
The raw data (counts observed at each cycle of data) was exported into Microsoft ® Excel for processing. To generate 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratios for each analysis, the counts of 235 U and 235 U 16 O observed during each cycle of data (lasting 10 s) were linearly extrapolated to the timestamp of 238 U, 238 U 16 O, and 238 U 16 O 2 (which were all collected in the same magnetic field). This was necessary as the secondary ion intensity was observed to change during each analysis (lasting approximately 45 min). In contrast, the 44 ns deadtime was found to be inconsequential because of the relatively low count rates obtained in this study (typically <10,000 cps), so a deadtime correction was not applied. Using this approach, a 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratios were generated for each cycle of the 20 cycles of data during a single analysis. The cycle-by-cycle values for the final 18 cycles of data collected during each analysis were then averaged together, with the standard deviation associated with this average used as the 1 standard deviation (1σ) uncertainty for each analysis.
For the UO 2 that was prepared specifically to assess questions over the role of sample preparation on the 235 U/ 238 U vs. 235 U 16 O/ 238 U 16 O ratios that arose during data interpretation, an additional session on the NanoSIMS was conducted in March of 2019. In this session, analytical conditions were nearly identical (e.g., same mass table/detector configuration, same analysis duration, same data processing approach) to those described above, except that a 100 pA primary beam was scanned over a 5 × 5 µm area for each analysis (as opposed to being used in static mode). Using this approach, data was collected for the UO 2 mounted in epoxy and dispersed onto the sticky carbon tab, but an operational issue prevented us from carrying out a full set of measurements on the UO 2 pressed into Indium. Additionally, a series of scans using the Cameca ® Bar Graph software from mass 248 to 258 were conducted on the UO2 in epoxy, on the sticky carbon tab, and pressed into Indium using the 100 pA primary beam scanned over a 5 × 5 µm area.

Solution MC-ICP-MS
Following NanoSIMS analysis, stainless steel tweezers were used to excavate a portion of each uranium compound, and then transfer the excavated material into 0.5 mL of 4 M HNO 3 . After 72 h of digestion at ambient conditions, the major uranium ratio for these solutions were determined on a Thermo Scientific NeptunePlus MC-ICP-MS. The instrument was equipped with the nuclear package and a jet interface. During analysis 235 U was placed on the L5 Faraday cup connected to a 10 13 Ω amplifier, and 238 U was located on L4 Faraday cup connected to a 10 11 Ω amplifier. The amplifiers were calibrated electronically with 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 ã 52 µL min −1 Elemental Scientific Inc. integrated PFA nebulizer into a quartz Elemental Scientific Inc. Apex Omega. The instrument sensitivity was~1.4 V of signal per nanogram of uranium. The unknown samples were analyzed using standard sample bracketing method with 2% HNO 3 washout blanks placed before each standard/sample. Mass fraction corrections were determined with New Brunswick Laboratory Program Office Certified Reference Material U010. Unknown solutions were corrected for mass fractionation, instrument blank, baseline, and gain using established protocols in the Nuclear Analytical Chemistry and Isotopics Laboratory at Oak Ridge National Laboratory. Controls of IRMM 183, IRMM 184, and Oak Ridge National Laboratory WRM were analyzed throughout the sequence to monitor instrument operation. Final calculated isotopic values were provided with Guide to Uncertainty in Measurement-compliant uncertainties.

Results
The ratios reported in Table S1, which were determined using NanoSIMS as described in Section 2.3, are uncorrected for mass fractionation. Additionally, the 1σ uncertainties associated with these ratios are simply the standard deviation associated with the final 18 cycles of data collected during each analysis. While it is common to apply a standard reference material-based mass fractionation correction, as well as to factor the variability of standards measurement into the determination of uncertainty for unknowns, when reporting isotope data, applying correction factors, and combining uncertainties would complicate the aims of this study, which are to understand the effects of different analytical approaches on the raw data.
In Figure 2, the percent uncertainty, calculated as 1σ associated with the average of the ratios observed for each cycle of data for each analysis, on the 235 U/ 238 U ( Figure 2a (Figure 2b) observed during each analysis of the UO 2 sphere's polished cross section. As described in Section 2, a variety of beam currents were used to analyze the UO 2 , resulting in a wide spread in total observed counts for this material. In contrast, the other materials were only analyzed using the 200 pA beam current, resulting in a much narrower spread in count rate. Therefore, the type of plot shown in Figure 2 is only representative of the UO 2 sphere. In Figure  The ratios reported in Table S1, which were determined using NanoSIMS as described in Section 2.3, are uncorrected for mass fractionation. Additionally, the 1σ uncertainties associated with these ratios are simply the standard deviation associated with the final 18 cycles of data collected during each analysis. While it is common to apply a standard reference material-based mass fractionation correction, as well as to factor the variability of standards measurement into the determination of uncertainty for unknowns, when reporting isotope data, applying correction factors, and combining uncertainties would complicate the aims of this study, which are to understand the effects of different analytical approaches on the raw data.
In Figure 2, the percent uncertainty, calculated as 1σ associated with the average of the ratios observed for each cycle of data for each analysis, on the 235 U/ 238 U ( Figure 2a Figure  2b) observed during each analysis of the UO2 sphere's polished cross section. As described in Section 2, a variety of beam currents were used to analyze the UO2, resulting in a wide spread in total observed counts for this material. In contrast, the other materials were only analyzed using the 200 pA beam current, resulting in a much narrower spread in count rate. Therefore, the type of plot shown in Figure 2 is only representative of the UO2 sphere. In Figure (Figure 5a,b) and UO3 (Figure 6a,b). In Figures 7 and 8, each analysis of 235 U/ 238 U and 235 U 16 O/ 238 U 16 O and its associated 1σ uncertainty for the UF4 (Figure 7a,b) and UO2(NO3)2·6(H2O)3 (Figure 8a,b) is shown. In Figure 9, each analysis of 235 U/ 238 U and 235 U 16 O/ 238 U 16 O and its associated 1σ uncertainty for the NIST-610 glass (Figure 9a,b) is shown relative to the reference 235 U/ 238 U of 0.00239, which is the average LG-SIMS value reported in [21]. In Figure 10, each analysis's 235 U/ 238 U and 235 U 16 O/ 238 U 16 O and its associated 1σ In Figures 5 and 6, each analysis of 235 U/ 238 U and 235 U 16 O/ 238 U 16 O and its associated 1σ uncertainty is shown relative to the solution MC-ICP-MS values for the UO 2 F 2 (Figure 5a,b) and UO 3 (Figure 6a,b). In Figures 7 and 8 (Figure 7a,b) and UO 2 (NO 3 ) 2 ·6(H 2 O) 3 (Figure 8a,b) is shown. In Figure 9, each analysis of 235 U/ 238 U and 235 U 16 O/ 238 U 16 O and its associated 1σ uncertainty for the NIST-610 glass (Figure 9a,b) is shown relative to the reference 235 U/ 238 U of 0.00239, which is the average LG-SIMS value reported in [21]. In Figure 10, each analysis's 235 U/ 238 U and 235 U 16 O/ 238 U 16 O and its associated 1σ uncertainty for Plesovice zircon reference material (Figure 10a,b) is analyzed using the 200 pA primary beam current is shown relative to the natural 235 U/ 238 U of 0.00726, which is the consensus value reported in [23]. For both materials, the uncertainties reported in the literature are far below the internal precision level achieved via the NanoSIMS.  For the UO 2 sphere, the average NanoSIMS 235 U/ 238 U ratio and its associated 1σ uncertainty for the 200 pA, 150 pA, 100 pA, 50 pA, and 20 pA primary beam currents were 0.00223 ± 0.00004, 0.00221 ± 0.00002, 0.00222 ± 0.00003, 0.00222 ± 0.00005, and 0.00223 ± 0.0005, respectively. These values compare well with the solution value of 0.00219 ± 0.00002. For the various beam currents, the average of the 1σ uncertainties associated with the individual measurements made using each beam current are 0.00008 for the 200 pA and 150 pA settings. The 100 pA, 50 pA, and 20 pA settings produced slightly higher average uncertainties of 0.00010, 0.00013, and 0.00023, respectively. For the 235 U 16  For the UO2 sphere, the average NanoSIMS 235 U/ 238 U ratio and its associated 1σ uncertainty for the 200 pA, 150 pA, 100 pA, 50 pA, and 20 pA primary beam currents were 0.00223 ± 0.00004, 0.00221 ± 0.00002, 0.00222 ± 0.00003, 0.00222 ± 0.00005, and 0.00223 ± 0.0005, respectively. These values compare well with the solution value of 0.00219 ± 0.00002. For the various beam currents, the average of the 1σ uncertainties associated with the individual measurements made using each beam current are 0.00008 for the 200 pA and 150 pA settings. The 100 pA, 50 pA, and 20 pA settings produced slightly higher average uncertainties of 0.00010, 0.00013, and 0.00023, respectively. For the 235 U 16 O/ 238 U 16 O ratio, average ratios and associated 1σ uncertainties, for the 200 pA, 150 pA, 100 pA, 50 pA, and 20 pA primary beam currents were 0.00218 ± 0.00001, 0.00217 ± 0.00001, 0.00218 ± 0.00001, 0.00219 ± 0.00007, and 0.00218 ± 0.0002, respectively. These values align more closely with the solution value 0.00219 ± 0.00002 instead of the 235 U/ 238 U ratios. For the various beam currents, the average of the 1σ uncertainties associated with the individual measurements made using each beam current are 0.00003 for the 200 pA, 150 pA, and 100 pA settings. The 50 pA and 20 pA settings produced slightly higher average uncertainties of 0.00007 and 0.00010, respectively.
While solution data was not obtained for the NIST-610 glass and Plesovice zircon reference materials analyzed in this study, there is an LG-SIMS 235 U/ 238 U ratio available for the NIST-610 glass published in [21], whereas the Plesovice zircon (described in [22]) can be assumed to have a natural 235 U/ 238 U ratio. For the NIST-610 glass, which was only analyzed using the 200 pA beam current, the average 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratios and associated 1σ uncertainties were 0.00283 ± 0.00028 and 0.00340 ± 0.00020. Considering uranium occurs at the parts per million level in the NIST-610 glass and that only a 200 pA primary beam current was used, both values compare well with the published 235 U/ 238 U ratio of 0.00239 [21] (note the uncertainty of the published ratio is in the seventh decimal place). For the NIST-610 glass, the average 1σ uncertainty of the individual 235 U/ 238 U ratios is slightly higher at 0.00084 compared to the 235 U 16 O/ 238 U 16 O ratios at 0.00067. For the Plesovice zircon,      LG-SIMS value reported in [22]. Uncertainty of the value in [22] is considerably lower than the NanoSIMS results and is therefore not shown here. LG-SIMS value reported in [22]. Uncertainty of the value in [22] is considerably lower than the NanoSIMS results and is therefore not shown here.  16 O ratio and its associated 1σ uncertainty, for the NIST-610 glass. The reference 235 U/ 238 U ratios (red line) for the NIST-610 glass is the average LG-SIMS value reported in [22]. Uncertainty of the value in [22] is considerably lower than the NanoSIMS results and is therefore not shown here.  [23]. Uncertainty of the value in [23] is considerably lower than the NanoSIMS results and is therefore not shown here.  [23]. Uncertainty of the value in [23] is considerably lower than the NanoSIMS results and is therefore not shown here.
While solution data was not obtained for the NIST-610 glass and Plesovice zircon reference materials analyzed in this study, there is an LG-SIMS 235 U/ 238 U ratio available for the NIST-610 glass published in [21], whereas the Plesovice zircon (described in [22]) can be assumed to have a natural 235 U/ 238 U ratio. For the NIST-610 glass, which was only analyzed using the 200 pA beam current, the average 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratios and associated 1σ uncertainties were 0.00283 ± 0.00028 and 0.00340 ± 0.00020. Considering uranium occurs at the parts per million level in the NIST-610 glass and that only a 200 pA primary beam current was used, both values compare well with the published 235 U/ 238 U ratio of 0.00239 [21] (note the uncertainty of the published ratio is in the seventh decimal place). For the NIST-610 glass, the average 1σ uncertainty of the individual 235 U/ 238 U ratios is slightly higher at 0.00084 compared to the 235 U 16 16 O ratios. Data from the UO 2 prepared specifically for this session can be found in its own section at the end of Table S1, and was not included in the UO 2 results described above. The mass scans described in Section 2.3 can be found in Table S2. In Figure 11, it is possible to compare the 235 U/ 238 U vs. 235 U 16 O/ 238 U 16 O for the UO 2 dispersed onto the sticky carbon tab (Figure 11a) versus the UO 2 mounted in epoxy ( Figure 11b); As noted in Section 2.3, a full dataset was not acquired for the UO 2 pressed into Indium. In Figure 11, the 235 U/ 238 U ratios, and their complimentary (e.g., from the same analysis) 235 U 16 O/ 238 U 16 O ratios, are grouped together and provided side-by-side in identical order (as opposed to the vertically stacked format used in Figures 3 and 5, Figures 6-10). Mass scans ranging from mass 248 to 258 are provided as insets on Figure 11 (a full dataset from these mass scans can be found in Table S2, in addition to the mass scan data from the UO 2 pressed into Indium), and it is noteworthy (see discussion Section 4.3) that the 238 U 12 C peak at mass~250 is visible in the mass scans conducted on the UO 2 dispersed onto the sticky carbon tab as well as mounted in epoxy. For the UO 2 analyzed during this additional session, both the 235 U/ 238 U and 235 U 16 16 O data used to construct these plots were collected in a separate session (March 2019) and can be found in their own section at the end of Table S1, while the mass scan data can be found in Table S2.
Therefore, it is still worthwhile to consider additional origins for the highly unsTable 2 35 U 16 O/ 238 U 16 O ratios displayed by some of the uranium compounds. In Figure 12, where the 1σ uncertainty of the 235 U 16 16 O ratios that are more precise than the 235 U/ 238 U ratios. This observation might favor the scavenging scenario described at the start of this section, because it could be used to state that analyses where there was less available oxygen ions to form the U-O molecular ion display poorer precision and accuracy in the 235 U 16 O/ 238 U 16 O ratio ; however, this possibility is not supported by the fact that the two materials with the highest concentration of reactive ions and lowest uranium concentrations (the NIST 610 glass and the Plesovice zircon) align with the results from the UO2, which is the only uranium compound not containing other elemental constituents.  16 O for UO 2 that was mounted in epoxy (a) as well as dispersed on a sticky carbon tab (b). Mass scans, from mass 248 to 258, acquired for both sample preparation approaches are shown as insets. While the 238 U 12 C peak is visible in both scans, the 238 U 12 C peak is~1000 counts/second for the UO 2 dispersed on the sticky carbon tab whereas it is 100 counts/second for the UO 2 mounted in epoxy. The 235 U/ 238 U and 235 U 16 O/ 238 U 16 O data used to construct these plots were collected in a separate session (March 2019) and can be found in their own section at the end of Table S1, while the mass scan data can be found in Table S2.

Previous Investigations into the Controls on Secondary ion Yields during SIMS Analysis
In the broader SIMS community, there are many questions about the sputtering and ionization processes (see discussion by [13]). Although models that produce testable predictions based on the relative yields between secondary monatomic and molecular secondary ion species have been developed (e.g., the local thermodynamic equilibrium model of [26]), the reality is that the secondary ion yields are dependent on so many factors (e.g., composition of the target, the particular element in question, the nature of the primary beam, the secondary ion extraction geometry of the instrument itself, etc.) that it has proven difficult to develop a unified sputtering and ionization model that fully explains secondary ion yield observations. Uncertainty in the debate over what actually controls ionization [27][28][29][30][31] is compounded by the fact that only a small fraction of the sputtered material, which is a mixture of both monatomic, molecular, and polyatomic species, acquires a charge such that it can be extracted and transmitted through the SIMS instrument's ion optical system. Similarly, ions that are part of the primary beam itself are implanted into material being sputtered, further complicating the development of a unified theory of ionization. While there are still uncertainties about the sputtering and ionization processes, a basic observation of SIMS data is that there are sometimes relatively fixed relationships between the secondary ion production rates of monatomic elemental species and their respective molecular elemental oxide secondary ions.
For example, [32]  For any isotope ratio determination, the total uncertainty is a combination of uncertainty associated with the count rate, the behavior of standards, the robustness of the mass fractionation correction regime, inter-detector calibration issues, as well as other factors. For SIMS, as with all mass spectrometric isotope ratio determinations, there are a variety of approaches to constraining and combining these uncertainties (see discussion by [4]). However, it is indisputable that two major contributors to the precision and accuracy of any SIMS measurement are (1) the stability of the isotope ratio, as reflected by the standard deviation, during each analysis, and (2) the deviance of the raw (e.g., uncorrected) ratio relative to the 'true' value for the matrix in question. For the 235 U/ 238 U ratio, both are known to vary proportionally with the count rate of 235 U (when 235 U is the minor isotope). For all the materials examined in this study, the secondary ion count rates of 235 U 16 O and 238 U 16 O were considerably higher than the 235 U and 238 U (exceptions are discussed in Section 4.3).
As described in Section 2, the UO 2 was analyzed using a variety of primary beam currents ranging from 200 pA down to 20 pA. This resulted in a wide range of total counts for 235 U and 235 U 16 O that enable a robust assessment of the effect of using the 235 U 16 O/ 238 U 16 O compared to the 235 U/ 238 U ratio. In Figure 2, which is a plot of the percent uncertainty, calculated as 1σ associated with the average of the ratios observed at each cycle of data during each analysis, on the 235 U/ 238 U (Figure 2a (Figure 3b) and its associated 1σ uncertainty for the UO 2 sphere. In Figure 3, the individual analyses are grouped according to the five primary beam currents (200 pA, 150 pA, 100 pA, 50 pA, and 20 pA) that were used to analyze the UO 2 sphere and are also shown relative to the solution MC-ICP-MS value. Clearly, there is a reduction in the standard deviation associated with each measurement. In Figure 3, it is also evident that the raw isotope ratios fall closer to the solution MC-ICP-MS value when the 235 U 16  For the UO 2 analyzed in this study, use of the 235 U 16 O/ 238 U 16 O ratio appears to result in improved accuracy and precision. However, because the atomic-to-molecular secondary ion formation rate and general behavior is matrix dependent, it is necessary to consider whether this behavior is present in the other matrices examined in this study. While the UO 2 was the only material that was investigated using multiple primary beam currents, the comparison of the 235 U/ 238 U versus the 235 U 16 O/ 238 U 16 O acquired using only a 200 pA beam current should still provide a preliminary indication of whether the observations made on UO 2 can be extended to the other uranium compounds. In Figures 5-8 (Figure 5a,b), UO 3 (Figure 6a,b), UO 2 (NO 3 ) 2 ·6(H 2 O) (Figure 7a,b), and UF 4 (Figure 8a,b), it is clear that the observations made on UO 2 cannot be extended uniformly to the other uranium compounds. This is because for many of the analyses, there is considerably higher standard deviation when the 235 U 16 O/ 238 U 16 O ratio is used. Additionally, when the NanoSIMS data is compared with the MC-ICP-MS 235 U/ 238 U ratios (UO 2 F 2 in Figure 5, and UO 3 in Figure 6, UO 2 (NO 3 ) 2 ·6(H 2 O) in Figure 7, and UF 4 in Figure 8), it is clear that there is considerably more scatter relative to the solution value when the 235 U 16 O/ 238 U 16 O ratios are used for all of the uranium compounds except the UO 2 . The scatter also appears decoupled from the 235 U/ 238 U ratios (e.g., for some analyses, the 235 U 16 O/ 238 U 16 O ratio deviates strongly from the solution value, but the 235 U/ 238 U ratio remains stable).
Therefore, it appears that use of the 235 U 16 O/ 238 U 16 O ratio would result in improved precision and accuracy for certain bulk chemistries (e.g., UO 2 ), but not others (e.g., UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 ). The reasons for this behavior will be discussed in the following section, but at this point it is also worth considering the behavior of matrices containing only trace levels of uranium. In Figures 9  and 10, each analyses of 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratio, and their associated 1σ uncertainties, for the NIST-610 glass (Figure 9a,b) and Plesovice zircon (Figure 10a,b) are shown relative to the reference values (refer to figure captions) for these materials. For the NIST-610 glass, there is a slight improvement in the uncertainty of the individual measurements when the 235 U 16 O/ 238 U 16 O ratio is used, but this is accompanied by a slight shift further away from the 'true' 235 U/ 238 U ratio reported in [22]. In contrast, the data from the Plesovice zircon indicates a marked improvement in the uncertainty associated with the individual measurements as well as a decrease in the degree of deviance relative the natural 235 U/ 238 U ratio used as the 'true' value. The observation that the 235 U 16 O/ 238 U 16 O ratios have marked improvement for these two materials is an important observation considering that uranium is only present at a trace phase within these two matrices (ppm levels)-this is also evident by the considerably lower count rates for the atomic and molecular uranium secondary ion species relative to the uranium compounds (Table S1).
In summary, the data indicates that use of the 235 U 16 O/ 238 U 16 O ratio results in higher precision and better accuracy for the UO 2 as well as the two silicate reference materials. In contrast, for the UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 it is clear that the 235 U 16 O/ 238 U 16 O ratio is associated with considerably higher scatter (both within and between individual analyses) compared to the 235 U/ 238 U. The implications of this observation for SIMS isotope ratio measurements will be discussed in the final section of the discussion. Now, the discussion will shift to a discussion of the possible mechanisms responsible for producing the sporadic results regarding improvement in the accuracy and precision when the 235 U 16  As described in the preceding section, an observation from the dataset is that use of the 235 U 16 O/ 238 U 16 O compared to the 235 U/ 238 U ratio results in improved accuracy and internal precision for the UO 2 and two silicate reference materials. In contrast, the accuracy and precision are considerably worse for the UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 when the 235 U 16 O/ 238 U 16 O is used. Three possible origins for this behavior will be explored in this discussion: (a) the SIMS 'matrix effect', (b) fluorine-and carbon-based interferences, and c) the effect of sample topography.
If only data from the uranium compounds were available, it would be reasonable to attribute the decrease in precision and accuracy to some type of matrix effect related to the presence of fluorine and nitrogen in the bulk matrices of the UO 2 F 2 , UO 2 (NO 3 ) 2 ·6(H 2 O), UO 3 , and UF 4 . The UO 3 examined in this study was prepared by calcination of UO 2 (NO 3 ) 2 ·6(H 2 O), but no additional steps were taken to remove residual nitrogen. Therefore, there is presumably residual nitrogen present in the UO 3 that was examined in this study. In such a scenario, it would be possible that nitrogen and fluorine contributed to scavenging of the oxygen ions available at the sputtering site, thus destabilizing the formation rate of the UO molecular ion. However, oxygen scavenging at the sputtering site does not explain why the results from the two silicate reference materials mirrored those from the UO 2 in that the 235 U 16 O/ 238 U 16 O resulted in improved accuracy and precision relative to the 235 U/ 238 U ratio. In considering that uranium is only present as a trace phase within these two silicate reference materials, there would have been an abundance of reactive ions available at the sputtering site to scavenge oxygen ions.
For the two fluorine bearing uranium compounds, it is also possible that 19 F + combined with 235 U + to form the 235 U 19 F + molecular secondary ion (mass = 254.0423262) that would require an MRP of (defined as M/∆M) of 75,353, well above what as used in this study, to be separated from the 238 U 16 (Figure 5b), the individual 235 U 16 O/ 238 U 16 O ratios are consistently biased above the solution ICP-MS and NanoSIMS 235 U/ 238 U ratios (Figure 5a), whereas for the UF 4 (Figure 7b), the 235 U 16 O/ 238 U 16 O ratios are highly scattered but do tend to agree with the NanoSIMS 235 U/ 238 U ratios (Figure 7a), which are also biased below the solution ICP-MS value. Therefore, there does not appear to be a systematic or significant contribution from the 235 U 19 F + on the 238 U 16 It is also necessary to consider the possibility that 238 U 13 C + (mass = 251.054138) and 238 U 12 C 1 H + (mass = 251.058608) molecular ions interfered with the observed 235 U 16 O + secondary ion signal (mass = 251.038837622). Because MRP values of 16,408 and 12,698 would be needed to separate the potential 238 U 13 C + and 238 U 12 C 1 H + interferences on 235 U 16 O + , respectively, and the fact that carbon from the epoxy or polishing compound may have contaminated, and then become trapped, in the uranium compounds with irregular surfaces, it is fully possible that a portion of the 235 U 16 O + signal was derived from carbon-based molecular interferences. 238 U 13 C + and 238 U 12 C 1 H + would be expected to increase the observed 235 U 16 16 O ratio, so understanding the interplay between these two potential interferences is not straightforward.
While the effects of fluorine-and carbon-based molecular interferences on the measured 235 U 16 O/ 238 U 16 O ratios cannot be ruled out, fully evaluating this potential mechanism as the cause of destabilization of the 235 U 16 O/ 238 U 16 O ratios is complicated by the fact that there are no data available to independently assess the rate at which these potential molecular interferences were formed. While it may theoretically be possible to monitor the intensity of the 238 U 12 C molecular species, and then use this information to subtract the inferred 238 U 13 C contribution to the 235 U 16 O signal, this would require a different detector and magnetic field setup than what was used in this study, and therefore no such data are available. It would also be necessary to account for the 238 U 12 C 1 H hydride formation rate. To more fully understand the magnitude of the 238 U 13 C contribution to the 235 U 16 O signal, an additional analytical session was conducted in March 2019 whereupon data collected from UO 2 dispersed onto a sticky carbon tab could be compared with data collected from UO 2 mounted in epoxy.
Examination of the insets on Figure 11 reveals that the 238 U 12 C molecular species can be observed in both sample preparation approaches (the sticky carbon tab as well as the epoxy). Examination of a mass scan conducted on UO 2 pressed into Indium (provided in Table S2) reveals no signal at the 238 U 12 C molecular species, suggesting that the source of carbon is indeed related to the sample preparation (e.g., contamination introduced during grinding/polishing) and sticky carbon Table For the UO 2 on the sticky carbon tab, the 238 U 12 C molecular species achieves a maximum of~400 c/s (Table S2), whereas for the UO 2 in epoxy, it achieves a maximum of~65 c/s (Table S2). Using the natural abundance of 13 C (1.07%), the expected count rate for the 238 U 13 C interference on 235 U 16 16 O ratio would be expected to exhibit a few % positive deviation relative to the 235 U/ 238 U. In contrast, where only trace amounts of carbon are present (as would be expected for the UO 2 mounted in epoxy), the effect would be considerably less. This prediction is born out in Figure 11, where it can be seen that some of the 235 U 16 O/ 238 U 16 O values from the UO 2 dispersed onto the sticky carbon tab are positively deviated by anywhere from few % up to~20% relative to the 235 U/ 238 U, whereas for the UO 2 in epoxy, the 235 U 16 O/ 238 U 16 O uniformly displays better precision and accuracy despite the fact that trace amounts of carbon are present. However, the fact that several of the analyses on the sticky carbon tab produce 235 U 16 O/ 238 U 16 O with better precision and accuracy compared to the 235 U/ 238 U suggests that the process of forming the U-C molecular species is not consistent from analysis to analysis despite having an abundance of carbon available at the sputtering site. Furthermore, the effect of this interference would be expected to be far worse on the carbon sticky tab in comparison to the epoxy, even under circumstances where an uneven sample surface may have trapped some epoxy and/or abrasive particles during preparation. Therefore, it is not clear whether the carbon-based molecular interferences can serve as the sole explanation for why the compounds shown in Figures 5-8 16 O ratios that are more precise than the 235 U/ 238 U ratios. This observation might favor the scavenging scenario described at the start of this section, because it could be used to state that analyses where there was less available oxygen ions to form the U-O molecular ion display poorer precision and accuracy in the 235 U 16 O/ 238 U 16 O ratio; however, this possibility is not supported by the fact that the two materials with the highest concentration of reactive ions and lowest uranium concentrations (the NIST 610 glass and the Plesovice zircon) align with the results from the UO 2 , which is the only uranium compound not containing other elemental constituents.
Therefore, it is necessary to consider other reasons for worsening precision and accuracy at lower average uranium oxide molecular secondary ion formation rates. Another possibility that can be considered with the available data are the effects of topography. As discussed earlier regarding the potential for fluorine-and carbon-based molecular interferences, it is possible that carbon from the epoxy may have become trapped in the irregular surfaces of the some of the uranium compounds. However, it is also necessary to consider that topography can also have a direct effect on secondary ion yields [35]. Therefore, it is necessary to consider other reasons for worsening precision and accuracy at lower average uranium oxide molecular secondary ion formation rates. Another possibility that can be considered with the available data are the effects of topography. As discussed earlier regarding the potential for fluorine-and carbon-based molecular interferences, it is possible that carbon from the epoxy may have become trapped in the irregular surfaces of the some of the uranium compounds. However, it is also necessary to consider that topography can also have a direct effect on secondary ion yields [35].
Examination of Figure 1 reveals that the UO2 produced a flat surface when the mount was polished, whereas the UO2F2 did not. Although SEM images of the two silicate reference materials were not acquired as part of this study, both materials are routinely analyzed by SIMS and are known to produce highly polished and flat surfaces when mounted in epoxy and polished. While SEM images of the UO3, UO2(NO3)2·6(H2O), and UF4 are not shown, their surfaces are more similar to the UO2F2 than the nicely polished and flat surface of the UO2. Therefore, another basic observation from the data is the uranium compounds with poor surface preparation produced considerably worse 235 U 16 O/ 238 U 16 O in comparison to the 235 U/ 238 U ratios. The effect of topography on the precision and accuracy of isotope ratios determined by SIMS has been investigated by several authors (see discussion by [35]), with the general consensus being that irregular surfaces can induce spot-to-spot variation at the per mil level. For the results of this study, it is important to note that the average levels of accuracy and precision associated with the 235 U/ 238 U ratios of the UO2F2, UO3, UO2(NO3)2·6(H2O), and UF4 are not considerably worse than for the UO2. Rather, the difference is seen in the 235 U 16  It is possible that the irregular surface produced by mounting and polishing the UO2F2, UO3, UO2(NO3)2·6(H2O), and UF4 results in irregular UO + production, whereas the effect on the U + ions is not as drastic. This lack of consistency in the UO + production rate is evident in Figure 12, where it is clear that the materials for which use of the 235 U 16  Examination of Figure 1 reveals that the UO 2 produced a flat surface when the mount was polished, whereas the UO 2 F 2 did not. Although SEM images of the two silicate reference materials were not acquired as part of this study, both materials are routinely analyzed by SIMS and are known to produce highly polished and flat surfaces when mounted in epoxy and polished. While SEM images of the UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 are not shown, their surfaces are more similar to the UO 2 F 2 than the nicely polished and flat surface of the UO 2 . Therefore, another basic observation from the data is the uranium compounds with poor surface preparation produced considerably worse 235 U 16 O/ 238 U 16 O in comparison to the 235 U/ 238 U ratios. The effect of topography on the precision and accuracy of isotope ratios determined by SIMS has been investigated by several authors (see discussion by [35]), with the general consensus being that irregular surfaces can induce spot-to-spot variation at the per mil level. For the results of this study, it is important to note that the average levels of accuracy and precision associated with the 235 U/ 238 U ratios of the UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 are not considerably worse than for the UO 2 . Rather, the difference is seen in the 235 U 16 O/ 238 U 16 O ratio. If sample topography is to blame for this behavior, then it would appear that the topographical influence on the 235 U 16 O/ 238 U 16 O is an order of magnitude greater than for the 235 U/ 238 U.
It is possible that the irregular surface produced by mounting and polishing the UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 results in irregular UO + production, whereas the effect on the U + ions is not as drastic. This lack of consistency in the UO + production rate is evident in Figure 12, where it is clear that the materials for which use of the 235 U 16 (Table S1). A similar effect is observed for the two silicate reference materials that were examined, even though the other uranium compounds examined exhibit less consistency (e.g., Figure 12). Therefore, in a situation where low count rates are an anticipated issue (e.g., a small amount of material is available or a high spatial resolution measurement with a low primary beam current is needed) and there is limited information about the chemistry of the material, a prudent approach would be to collect both the 235 U 16 16 O ratios reported in this study have not been corrected for any type of mass fractionation or other instrument specific bias. Therefore, it is not possible to evaluate which ratio (e.g., 235 U/ 238 U versus 235 U 16 O/ 238 U 16 O) is most amenable to the type of correction regimes that are typically used in isotope ratio mass spectrometry. The basic conclusion outlined in the preceding section will likely apply to any type of correction regime that is applied, but a logical next step is to further study the behavior of the 235 U/ 238 U versus 235 U 16 O/ 238 U 16 O ratio in a suite of uranium compounds with certified uranium isotope compositions. During such an approach, it would also be worthwhile to determine whether the observations made in this study hold true at different 235 U/ 238 U ratios because all the materials examined in this study have depleted-to-natural 235 U/ 238 U ratios. Other important variables to test would be the effects of different sample surface morphologies and substrates, in addition to a systematic assessment of the potential interference contributions outlined in Section 4.3. In summary though, the basic observation that the raw 235 U/ 238 U and 235 U 16 O/ 238 U 16 O ratios are both quite close to the solution MC-ICP-MS results means that both ratios should be amenable to the application of a reference material-based correction scheme. Therefore, selection of the ratio with the lowest internal precision (e.g., 1σ value) should yield the 'best' result in count-rate-limited situations.
While this study was expressly focused on the 235 U/ 238 U determination, the possibility of taking advantage of the greater U-O molecular generation in comparison to the U elemental secondary ion production to improve accuracy and precision may also extend to other uranium isotope ratios. For example, the 236 U/ 238 U and 234 U/ 238 U ratios are important indicator of whether samples from a variety of matrices contain anthropogenically perturbed U (see discussion by [36][37][38]). Because the natural 234 U/ 238 U is~0.000055, and <10 −10 for 236 U/ 238 U [37], the issue of low secondary ion count rates for 235 U that formed the basis of this investigation become much more severe for determination of the 234 U/ 238 U and 236 U/ 238 U by SIMS. Furthermore, the 236 U/ 238 U determination is compounded by the formation of 235 U 1 H secondary ions (see discussion by [4]). While it is likely that the secondary ion count rates of the 236 U 16 O (mass = 252.04077 amu) and 234 U 16 O (mass = 250.035861 amu) molecular species would exceed those of the 236 U and 234 U elemental secondary ions during SIMS analysis, MRPs (defined as M/∆M) of 18,836.6 and 16,755.8 would be needed to separate interferences from 238 U 14 N (mass = 252.053857 amu) and 238 U 12 C (mass = 250.050783), respectively. Furthermore, examination of the mass scans provided as insets in Figure 12 reveals elevated secondary ion counts at masses~250 and~252, suggesting that a first step in exploring use of the elemental oxide ratios for the 236 U/ 238 U and 234 U/ 238 U determination would be to understand whether a suitable sample preparation route can be developed such that the contribution of nitrogen and carbon to the analytical volume is eliminated.

Conclusions
Comparison of the 235 U/ 238 U versus the 235 U 16 O/ 238 U 16 O ratios measured by NanoSIMS from a variety of uranium compounds and silicate reference materials indicate the following: (1) For the UO 2 and silicate reference materials, which display internally consistent relationships between their UO + and U + secondary ion production rates, use of the 235 U 16 O/ 238 U 16 O ratio results in improved accuracy and precision compared to the 235 U/ 238 U ratio.
(2) For the UO 2 F 2 , UO 3 , UO 2 (NO 3 ) 2 ·6(H 2 O), and UF 4 , which have variable internal relationships between their UO + and U + secondary ion production rates, use of the 235 U 16 O/ 238 U 16 O ratio is associated with considerably worse accuracy and precision compared to the 235 U/ 238 U ratio.
(3) Preliminarily, the observed different behavior between the UO 2 and silicate reference materials versus the other uranium compounds appears attributable to the effects of sample topography, although the effects of fluorine-and carbon-based molecular interferences cannot be fully ruled out.
(4) Additional experiments are necessary to determine whether the observations made in this study can be applied to additional uranium compounds, other isotope compositions, different sample morphologies, and varied substrates.
Supplementary Materials: The following are available online at http://www.mdpi.com/2075-163X/9/5/307/s1, Table S1: NanoSIMS analytical results, Table S2: NanoSIMS mass scans. Funding: Notice: This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan). SEM analyses were conducted at the ORNL Center for Nanophase Materials Sciences, which is a DOE Office of Science user facility.