A NanoSIMS 50 L Investigation into Improving the Precision and Accuracy of the ²³⁵U/²³⁸U Ratio Determination by Using the Molecular ²³⁵U¹⁶O and ²³⁸U¹⁶O Secondary Ions

Round 1

Reviewer 1 Report

In this paper the authors propose that uranium isotopic ratios based on uranium oxide ions may significantly improve isotopic ratio precision and they demonstrate with samples of various chemical forms of uranium that the improvement does occur for some samples but not others, i.e., for some of the chemical forms there is considerably more scatter in the oxide ratios than in the elemental ratios.  Further they observe that significant high isotopic bias occurs with oxide ratios for the samples that have large ratio scatter.  In their conclusions they preliminarily attribute this behavior to sample topography because the samples that exhibit the more expected behavior have smooth flat surfaces whereas the others have irregular surfaces.

The fundamental flaw of this paper is that the authors have failed to consider the effect of mass spectral interferences from both carbon and fluorine on the uranium oxide ions.  Although the mass resolving power used in the study is more than sufficient to exclude the most likely interferences with uranium elemental ions, it is not sufficient for the oxides.  For example, U238+13C requires a resolving power greater than 16,000 to be separated from U235+16O, and U238+12C+H requires greater than 12,000.  U235+19F requires greater than 74,000 to be separated from U238+16O.  Although minor uranium isotopes were not considered in this study they are essential for complete isotopic characterization of uranium, and U238+12C requires a resolving power greater than 16,000 to be separated from U234+16O.  This referee considers it very likely that the irregular surfaces described in the paper have incorporated some residual carbon-containing epoxy that produces variable levels of the interfering species from spot-to-spot resulting in variable and biased isotopic ratios.  The smooth polished surfaces of several of the samples are less likely to incorporate residual epoxy and therefore give more consistent oxide ratios.  I suspect that the oxide ratios from the UF4 sample would be biased low if it were smoothly polished and eliminated the epoxy but that the interfering carbide ions at 251 contribute more than the interfering fluoride ions at 254 for the sample in this study.  The approach proposed in this paper might be useful if it could be guaranteed that the samples contained no carbon or fluorine.  However, in the real world of nuclear forensics and environmental samples such as practiced by the IAEA and its NWAL where samples are often contaminated with environmental materials that level of cleanliness would be virtually impossible to guarantee. Hence the use of uranium elemental ions is a much safer approach.

If the authors want to further investigate the presence or absence of interferences from their samples they could make broader mass scans to look for isolated peaks from U235+12C at mass 247 and U238+19F at mass 257.

Author Response

Response: In order to address the reviewer’s concerns regarding the possibility of molecular secondary ion interferences, we have added a discussion of the possibility that ²³⁵U¹⁹F (mass = 254.0423262) contributed to the observed ²³⁸U¹⁶O (mass = 254.045697622) secondary ion signal, and ²³⁸U¹³C (mass 251.054138) and/or ²³⁸U¹²C¹H (mass = 251.058608) contributed to the ²³⁵U¹⁵O (mass = 251.038837622) secondary ion signal.  This discussion has been added to section 4.3, which explores possible origins of the unstable ²³⁵U¹⁶O/²³⁸U¹⁶O ratio relative to the ²³⁵U/²³⁸U ratio for some of the uranium bearing compounds.  However, it is important to emphasize that the fluorine- and carbon-based molecular interferences would be expected to have opposite effects on the ²³⁵U¹⁶O/²³⁸U¹⁶O. This is because fluorine-based interferences would be expected to lower the ratio, whereas the carbon-based interferences would be expected to raise it.  In and of itself, this complicates an assessment of the importance of the molecular interferences on the interpretation of the dataset. We believe this is recognized by the reviewer, since the reviewer mentions the competing effects of the fluorine- versus carbon-based interferences on the results from the UF₄.  Based on our analysis of the results within the context of the fluorine- and carbon-based molecular interferences (now found in section 4.3), we do not see evidence of the ²³⁵U¹⁹F interference because the ²³⁵U¹⁶O/²³⁸U¹⁶O ratios from the UO₂F₂ and UF₄ (the two fluorine bearing uranium compounds analyzed in this study) do not display an obvious negative bias relative to the ²³⁵U/²³⁸U ratios collected at the same time. The reviewer mentions that the UF₄ ²³⁵U¹⁶O/²³⁸U¹⁶O are biased low relative to the solution values, but an important observation from this sample is that the ²³⁵U/²³⁸U ratios are as well.    In contrast, the carbon-based molecular interferences would be expected to produce ²³⁵U¹⁶O/²³⁸U¹⁶O ratios that are positively biased with regard to the ²³⁵U/²³⁸U. Examination of the data reveals that some of the individual analyses do display this behavior, which would be consistent with the possibility that particles of epoxy became trapped within the irregular surfaces of some of the samples, thus producing transient positively biased ²³⁵U¹⁶O/²³⁸U¹⁶O values.  Therefore, while it is difficult to fully assess the extent to which these interferences may have influenced the observed secondary ion yields, we agree that the issue is worth discussing in the paper. 

              The reviewer posits that monitoring of the ²³⁵U¹²C (mass = 247.043923) or ²³⁸U¹⁹F (mass = 257.0491862) might provide a means of correcting any possible molecular interferences. It is important to note that on the NanoSIMS, which uses very low beam currents in comparison to most other SIMS instruments, the ²³⁵U signal is already very low (typically in the tens of counts per second during analyses of depleted uranium). When coupled with the fact that carbon does not ionize well using a negatively charged oxygen primary beam (on the NanoSIMS, carbon measurements are typically made using a positively charged cesium primary beam), the secondary ion signal at ²³⁵U¹²C would be very low.  Given the expectedly low signal and large differences in ionization efficiency between uranium and carbon, extrapolating the measured secondary ion intensity of ²³⁵U¹²C into a correction factor for the ²³⁸U¹³C and ²³⁸U¹²C¹H molecular ions would not be straightforward.  Further complicating the matter is that the ²³⁸U¹²C¹H interference correction would require the development of a hydride formation rate assumption, which is a matter of debate in the SIMS literature for more simple molecules like the ²³⁵U¹H interference on ²³⁶U. Measurement of the ²³⁸U¹⁹F molecular ion is more straightforward as there is typically a good signal at this mass when fluorine bearing uranium compounds are sputtered. However, as discussed above- the magnitude of the interference appears negligible for the two fluorine bearing compounds analyzed in this study (the ²³⁵U¹⁹F interference on ²³⁸U¹⁶O would be expected to lower the ²³⁵U¹⁶O/²³⁸U¹⁶O ratio relative to the ²³⁵U/²³⁸U ratio collected during the same analysis, whereas the data from the UO₂F₂ and the UF₄ displays the opposite behavior).  Certainly, a correction based on the observed signal at mass 257.0491862 could be developed, but the data we have suggests that application of such a correction factor (effectively subtracting counts from the ²³⁸U¹⁶O secondary ion intensity) might further positively bias the data, thus pushing the ²³⁵U¹⁶O/²³⁸U¹⁶O values even further from the solution ICP-MS and NanoSIMS ²³⁵U/²³⁸U ratios. 

              The issue of how to deal with molecular interferences during SIMS analysis a matter of ongoing debate in the literature, so the reviewer is not incorrect to bring these issues up. However, the path to making a measurement-based correction is not as straightforward as the reviewer is suggesting. For example, and as discussed in the paper, some studies have suggested that the ²³⁵U¹⁶O/²³⁸U¹⁶O ratio is preferable to the ²³⁵U/²³⁸U in situations where there is the potential for metallic contamination to the samples of interest and a requisite MRP for total separation is not achievable.   As stated above, we have now explored the molecular interference issue in the paper discussion section, and tried to understand the magnitude of this effect using the available data. However, it is still important to note that the other possible mechanisms that were already discussed in the paper are still equally valid (e.g. that a matrix effect, separate from the molecular interference effect, or the influence of topography on secondary ion yields played a role in controlling the stability of the ²³⁵U¹⁶O/²³⁸U¹⁶O ratio determination).  With regard to the reviewer’s assertion that the technique utilized in this paper would never be utilized for ‘real world’ samples, it is important to emphasize that we never suggested that our technique would supplant existing protocols. However, we have modified the language in the abstract to more clearly convey that use of the ²³⁵U¹⁶O/²³⁸U¹⁶O has the potential to improve precision and accuracy in certain situations and for certain bulk matrices, whereas in the previous version the language was a little more forceful and could be interpreted to mean that we were suggesting use of the ²³⁵U¹⁶O/²³⁸U¹⁶O is a better option.  In closing, we thank the reviewer for bringing up the molecular interference issue, and feel that the paper has benefited by including a discussion of this topic.

Reviewer 2 Report

This manuscript concerns the application of NanoSIMS to the analysis of uranium and specifically addresses how the choice of species analyzed (atomic vs diatomic or polyatomic ion) affects precision. The authors have done a fine job of addressing the analytical issues underlying this approach and the paper provides a useful contribution to SIMS analysis of uranium bearing samples.

I do not have concerns requiring major revisions of the manuscript, but would like to see the following minor issues addressed.

"Beam footprint" is not a term I was immediately familiar with. My vague recollection is that this is a term well accepted in the SIMS community, but it might require a brief definition in this manuscript so as to not lose or confuse a broader readership.

Line 146, remove right parenthesis.

Line 147, change to “…have a natural 235U/238U ratio of 0.00726,…”

Line 149, “…both materials is in the sixth decimal place and is therefore insignificant for discussion of the NanoSIMS results.” While I largely agree, I always want to see careful uncertainty assessments. IN this instance, it would be good to quantify just how insignificant the uncertainty of the standards is. That is, please provide an analysis that determines roughly what fraction of the total uncertainty is driven by uncertainty in the accepted values for the standards.

Line 156, “…polished using successively coarser silicon…” I think you meant to say “successively finer”, correct?

Line 164, “quartz-modulated thickness controller”.  Is this the common terminology for this device/instrument? I vaguely recall these being referred to as “quartz crystal microbalance” or “quartz crystal oscillators”, but I could be wrong here. Please confirm.

Line 206, “I don't understand how you can't measure 25/28 simultaneously if you can measure 28 with the various UO species cited. Must be a detector configuration issue rather than a magnet radius issue. Right?”

Line 221. This seems like a larger analytical issue for precision (and accuracy!) than which species is selected. Show typical, best, and worst, signal vs time plots to enabel the reader to gauge this effect (the noise power spectrum of the signal source must be assessed against dwell times).

Line 227. polyatomic ion interferences would seem to be an issue for such analyses, especially at the lowest signal levels encountered. Also, while not part of this study, 234 and 236 are of great interest in analysis of anthropogenic uranium particles. Interferences are expected to be different and more important for these. A survey of polyatomics observed for the various materials studied here would provide a better justification for the mass resolution used, the lack of corrections at that resolution and would presage the likely magnitude of such effects in future studies that address 234 and 236.

Line 240. Please give the measured sample uptake rate versus the nominal 50 uL/min. This is necessary to assess SUE which in turn provides an indication of the quality of instrument operation and tuning (n.b., actual uptake rates can be much different than nominal especially for low flow nebulizers).

Line 254. what does "factor the behavior of standards into the determination of uncertainty" mean? If it means "factor the variability of standards measurements into the determination of uncertainty for unknowns", then this latter phrasing is preferred (or something like it) as it is more specific regarding the "behavior" of interest.

Fig 4a. panel a shows bias. panel b shows less bias and opposite sign. These suggest an interference or detector gain calibration error. Please address this.

Line 288. do you mean "internal precision" here? Since you have explicitly stated that total uncertainties are not the primary focus of this paper.

Line 298. I would like to have seen a discussion here about the potential of NanoSIMS and the method developed by the authors. We need more direct comparisons of such methods. So, why not consume sufficient sample or analyze large enough particles that such direct comparisons are possible, e.g., compare measurements using the same amount of analyte consumed? Even if such comparisons are only among/between various SIMS methods, it would be useful to gauge if the present effort represents an improvement and if so, how much. Please address this issue briefly, ideally using measurements or observations from your laboratory’s instruments.

Fig 5 caption. Is "analysis's" correct? Better: "Plots of UO2F2 replicate isotopic analyses for (a) ..." Similarly for Fig 6 caption.

Line 382. "dependent on so many factors...that it has proven difficult..."

Line 388. "...because only a small fraction of those species acquire a charge...has led to uncertainty over what actually controls ionization."  I suspect that the ionization mechanism would be comparably uncertain even if secondary ion yields were much higher. So, I don't quite understand the connection implied by low SUE and uncertainty in the mechanism. Modify or explain.

Line 389. meaning "produced by" or "part of"? If you want to include ionization by the primary beam apart from interactions with the surface (primary beam hitting lens elements, ionization gas phase species), that's fine, but please be clear. Instead of "originating from" do you mean "part of"?

Line 393. Do you mean fixed relationships in the observed ion beam intensities for these species or do you mean between the monoatomic ions and related sputtered neutral oxides?

Line 394. suggest striking "the relationship"

Line 401. I did not know this: Pb/U relative sensitivity factors can be determined from RSF for U, UO, and UO2. Really? Is this the preferred method? That is, it seems that better accuracy could be obtained if a more direct determination of RSFs were made. I appreciate that this is not always possible and requires more analytical effort, but apart from these issues, please comment on this.

Author Response

Reviewer #2 and Author Responses:

Reviewer #2: This manuscript concerns the application of NanoSIMS to the analysis of uranium and specifically addresses how the choice of species analyzed (atomic vs diatomic or polyatomic ion) affects precision. The authors have done a fine job of addressing the analytical issues underlying this approach and the paper provides a useful contribution to SIMS analysis of uranium bearing samples.

              I do not have concerns requiring major revisions of the manuscript, but would like to see the following minor issues addressed.

Response: We appreciate that reviewer #2 views the paper as a useful contribution and are grateful for the detailed comments and suggestions.

Reviewer #2: "Beam footprint" is not a term I was immediately familiar with. My vague recollection is that this is a term well accepted in the SIMS community, but it might require a brief definition in this manuscript so as to not lose or confuse a broader readership.

Response:  In line 57, which is the first usage of ‘beam footprint’, the term is defined as the ‘the size of the primary beam impinging on the sample surface’. 

Reviewer #2: Line 146, remove right parenthesis.

Response: This is now line 148; the right parenthesis has been removed. 

Reviewer #2: Line 147, change to “…have a natural 235U/238U ratio of 0.00726,…”

Response: This has been changed (now line 149).

Reviewer #2: Line 149, “…both materials is in the sixth decimal place and is therefore insignificant for discussion of the NanoSIMS results.” While I largely agree, I always want to see careful uncertainty assessments. IN this instance, it would be good to quantify just how insignificant the uncertainty of the standards is. That is, please provide an analysis that determines roughly what fraction of the total uncertainty is driven by uncertainty in the accepted values for the standards.

Response: The uncertainty associated with each of the silicate materials is now provided, and the phrase stating that the uncertainty values associated with these two materials are ‘insignificant for this study’ has been removed.  Additionally, the paragraph containing the uranium isotope ratios of the two silicate reference materials has been moved to the to top of this section (now lines 129 – 133).  I agree with the reviewer that the issue, of how the uncertainty associated with the standards factors into the uncertainty reported for unknowns, is an important consideration in isotope ratio mass spectrometry. However, for the purpose of this study, the goal is not to report ‘final’ isotope ratios with expanded uncertainties, but rather, to consider which analytical approach produces the measurement with the best internal precision and accuracy. As described in section 4.1 of the paper, while there are many sources of uncertainty that get factored into the expanded uncertainties typically reported for final isotope ratios, a stable measurement displaying good within-run statistics (in the case of this study, this is quantified by the standard deviation on the individual cycles of data comprising each measurement) will yield a lower uncertainty than a measurement that has poor within run statistics- all other sources of uncertainty being equal.  This is all discussed in section 4.1, so I want to avoid bringing this discussion into this section, which is expressly focused on simply describing the materials that were used in this study.   

Reviewer #2: Line 156, “…polished using successively coarser silicon…” I think you meant to say “successively finer”, correct?

Response: Thank you for pointing out this error- it has been fixed (now line 157)

Reviewer #2: Line 164, “quartz-modulated thickness controller”.  Is this the common terminology for this device/instrument? I vaguely recall these being referred to as “quartz crystal microbalance” or “quartz crystal oscillators”, but I could be wrong here. Please confirm.

Response: The phrase ‘quartz-modulated thickness controller’ has been removed, and instead the model of the thickness controller (Cressington MTM-20) is provided (now line 165).

Reviewer #2: Line 206, “I don't understand how you can't measure 25/28 simultaneously if you can measure 28 with the various UO species cited. Must be a detector configuration issue rather than a magnet radius issue. Right?”

Response: On the NanoSIMS 50L, it is possible to get single-mass unit spacing on adjacent detectors if the detectors are positioned towards the high radius side of the multi collection array and the highest mass of interest is relatively low (e.g. up to around mass 55, in our experience).  However, when higher masses are to be measured, it is impossible to get single mass unit spacing on adjacent detectors regardless of their position along the multi collection array (for example, attempting to look at the ²³⁸U, ²³⁸U¹⁶O, and ²³⁸U¹⁶O₂ secondary ions on detectors 5, 6, and 7 while simultaneously looking at ¹⁶O and ¹⁸O on detectors 1 and 2). Furthermore, at high masses (e.g. for the Actinides), it is impossible to get below a five mass unit spacing on adjacent detectors (e.g. ²³⁵U and ²³⁸U cannot be measured in the same magnetic field). This is largely a function of the dispersion through the magnet. 

Reviewer #2: Line 221. This seems like a larger analytical issue for precision (and accuracy!) than which species is selected. Show typical, best, and worst, signal vs time plots to enable the reader to gauge this effect (the noise power spectrum of the signal source must be assessed against dwell times).

Response: Thank you for pointing this out- In the earlier version of the paper, it was not specified that only the final 18 cycles of data collected during each analysis were used for calculations. This is now explicitly mentioned in the methods section (lines 222 – 244).  Basically, the signal is unstable at the start of each analysis, so removing the first two cycles from consideration accounts for the contribution that this period would have on the precision and accuracy while also serving as a sort of pre-sputter (since no formal pre-sputter routine was built into the analytical sequence, as explained in the paper).  Additionally, some minor wording changes have been made- for example, instead of stating that the count rate was observed to always decrease during the course of each analysis, it is now simply stated that the secondary ion intensity was observed to change.  I am wary of including a series of figures illustrating the change in secondary ion intensity as a function of time during each analysis, since this will trigger the need to include a discussion of the fundamental controls on secondary ionization behavior. While this topic is already broached in the discussion, it is very complicated and would increase the length of the paper.  Furthermore, I will say that we are currently working on a paper that addresses this very issue, and the consideration of secondary ion intensity versus time is one of the pieces of information used in the companion study.  

Reviewer #2: Line 227. polyatomic ion interferences would seem to be an issue for such analyses, especially at the lowest signal levels encountered. Also, while not part of this study, 234 and 236 are of great interest in analysis of anthropogenic uranium particles. Interferences are expected to be different and more important for these. A survey of polyatomics observed for the various materials studied here would provide a better justification for the mass resolution used, the lack of corrections at that resolution and would presage the likely magnitude of such effects in future studies that address 234 and 236.

Response: The reviewer is correct in stating that the polyatomic ion interferences must be accounted for in interpreting the results. Because the other reviewer’s comments were almost exclusively focused on this issue, please refer to my response to the other reviewer, as well as the resultant changes to the paper that were made in response to the other review. 

Reviewer #2: Line 240. Please give the measured sample uptake rate versus the nominal 50 uL/min. This is necessary to assess SUE which in turn provides an indication of the quality of instrument operation and tuning (n.b., actual uptake rates can be much different than nominal especially for low flow nebulizers).

Response: According to the ICP-MS technician that ran the samples, the actual uptake was ~52 uL/min. This has been changed in the manuscript. 

Reviewer #2: Line 254. what does "factor the behavior of standards into the determination of uncertainty" mean? If it means "factor the variability of standards measurements into the determination of uncertainty for unknowns", then this latter phrasing is preferred (or something like it) as it is more specific regarding the "behavior" of interest.

Response: The proposed wording change has been made. 

Reviewer #2: Fig 4a. panel a shows bias. panel b shows less bias and opposite sign. These suggest an interference or detector gain calibration error. Please address this.

Response: The reviewer is correct that two panels (²³⁵U/²³⁸U and ²³⁵U¹⁶O/²³⁸U¹⁶O) display contrasting behavior relative to the solution value. The reviewer is correct in that this phenomenon could be either a molecular interference or gain calibration error (in addition to the other possibilities that were already mentioned in the paper – notably the effect of sample topography).  The issue of molecular interferences and their possible effect on the data is now addressed in response to the other reviewer’s comments, and for the most part can be found in section 4.3 of the paper.

              On the NanoSIMS, inter-detector calibration while using the electron multipliers is achieved through a Cameca^® software program called ‘pulse height distribution’, that allows the user to scan the output pulse amplitude of the incoming secondary ions as a function of the voltage applied to the electron multiplier in question. Similarly to the electron multiplier gain (which is more strictly defined as the ratio between the detectors electron output current and the secondary ion input current), the pulse height distribution depends on the secondary ion to electron pulse conversion efficiency at the first dynode as well as other dynode amplification effects.  Aging of an electron multiplier tied to the secondary ion flux results in both a decrease in the pulse height distribution as well as the ion to electron conversion efficiency.  Therefore, optimizing the pulse height distribution by increasing the voltage applied to the electron multiplier also compensates for the decrease in gain associated with electron multiplier aging.

              In practice, a range of voltages are applied to each detector individually until an optimum pulse height distribution is found, with the aim being to have the maximum pulse height on all the detectors occur at roughly the same position. Using the same program, the threshold voltage for each detector is also set. In principal this works quite well to eliminate inter detector issues that arise from aging etc, but in practice the scans must be done using high count rates (> 500 k c/s) to get a good handle on the voltage that creates the optimum pulse height position. Even with high count rates, there is still some variability and subjectivity in finding the optimum pulse height position. Therefore, using this program does not fully eliminate variability resulting from inter-detector differences in the way that incoming ions are converted into counts.  This highlights the importance of running standards in situations where the goal is to get the most accurate isotope ratio, since the standard-based correction would in principal account for inter-detector calibration issues (including the gain). This is especially true if the standards are matrix matched and yield approximately the same secondary ion count rates as the unknowns, so that a final correction factor based on the behavior of the standards could eliminate deviation of the unknown’s isotope ratios due to differences in detector efficiency. 

Reviewer #2: Line 288. do you mean "internal precision" here? Since you have explicitly stated that total uncertainties are not the primary focus of this paper.

Response: The reviewer is correct; this has been changed. 

Reviewer #2: Line 298. I would like to have seen a discussion here about the potential of NanoSIMS and the method developed by the authors. We need more direct comparisons of such methods. So, why not consume sufficient sample or analyze large enough particles that such direct comparisons are possible, e.g., compare measurements using the same amount of analyte consumed? Even if such comparisons are only among/between various SIMS methods, it would be useful to gauge if the present effort represents an improvement and if so, how much. Please address this issue briefly, ideally using measurements or observations from your laboratory’s instruments.

Response: This is actually a very complicated topic, which can be partly addressed by looking at figure 2.  Basically, in figure 2 it can be seen that the within run precision gets better as count rate increases, which in turn corresponds to increasing beam current.  Presumably, this increasing beam current also corresponds to an increase in the total volume of material that has been sputtered.  Now, if there was raw count data for uranium compounds that had been analyzed using the other SIMS instruments readily available in the literature, a direct comparison would be possible. However, this level of detail (e.g. raw counts observed during each analysis) is simply not typically reported. Therefore, it is impossible to compare with other machines using the available data. Another type of comparison might be made on the basis of total material sputtered (which could be calculated from either the volume of the analytical craters or, in the case of particles of known mass/volume, the fraction of a particle that has been sputtered) in conjunction with assumptions about ionization efficiency, transmission rates, and detector efficiency (which in and of itself would be quite difficult). In practice, however, attaining these volumes and masses is quite difficult- there are actually a few papers that attempt to produce particles of uniform size (and hence uniform U content), and then sputter them totally away to get an absolute secondary ion yield. However, raw counts are not reported in these papers, so I am not sure how to compare the existing data with similar data (which would require crater volume calculations) from this study.  Figure 2 provides a good visual depiction of the improved internal precision as count rates increase- therefore, someone trying to understand the capability of the NanoSIMS in relation to other SIMS instruments could theoretically use figure 2 as a starting point.  However, making a direct comparison would actually be quite difficult and likely require a study designed explicitly for said purpose. 

Reviewer #2: Fig 5 caption. Is "analysis's" correct? Better: "Plots of UO2F2 replicate isotopic analyses for (a) ..." Similarly for Fig 6 caption.

Response: I have double checked the grammatical correctness of ‘analysis’s’. If the journal editorial office has a particular style that they would rather see used, I am happy to make the change. 

Reviewer #2: Line 382. "dependent on so many factors...that it has proven difficult..."

Response: Now line 417; ‘so’ has been added. Thank you for pointing this out.

Reviewer #2: Line 388. "...because only a small fraction of those species acquire a charge...has led to uncertainty over what actually controls ionization."  I suspect that the ionization mechanism would be comparably uncertain even if secondary ion yields were much higher. So, I don't quite understand the connection implied by low SUE and uncertainty in the mechanism. Modify or explain.

Response: The sentence has been modified. Now, it is stated that the small fraction of ionized material (compared to the amount of material that is sputtered) acquires a charge compounds the uncertainty, as opposed to implying that it is one of the main drivers of uncertainty. 

Reviewer #2: Line 389. meaning "produced by" or "part of"? If you want to include ionization by the primary beam apart from interactions with the surface (primary beam hitting lens elements, ionization gas phase species), that's fine, but please be clear. Instead of "originating from" do you mean "part of"?

Response: “Originating” has been changed to “are part of”. 

Reviewer #2: Line 393. Do you mean fixed relationships in the observed ion beam intensities for these species or do you mean between the monoatomic ions and related sputtered neutral oxides?

Response: In this case, it is the secondary ion beam intensities. Since the neutrals could not get extracted since they carry no charge, I am not sure how they would actually be compared to the ionized elemental species.  I can see why the reviewer is asking this, though, so the sentence has now been modified to specify molecular elemental oxide secondary ions.

Reviewer #2: Line 394. suggest striking "the relationship"

Response: This has been removed. 

Reviewer #2: Line 401. I did not know this: Pb/U relative sensitivity factors can be determined from RSF for U, UO, and UO2. Really? Is this the preferred method? That is, it seems that better accuracy could be obtained if a more direct determination of RSFs were made. I appreciate that this is not always possible and requires more analytical effort, but apart from these issues, please comment on this.

Response: Using the Pb/U versus U/UO (or some other combination of Pb, U, UO, and even UO₂) is how all of the Pb-U geochronological measurements by SIMS are done.  When it comes to U-Pb geochronology, there is simply no other way. People have tried to independently determine a Pb/U RSF factor and then apply it to the Pb-U SIMS measurement, but the result are ages that are highly imprecise and of low accuracy. In contrast, using the Pb/U versus U/UO interelement calbrations produces SIMS ages that have comparable precision and accuracy to ages determined via TIMS.  The references provided in the paper provide a good starting point if the reviewer is interested in reading about the Pb/U versus U/UO interelement calibration used in U-Pb dating by SIMS. 

Reviewer 3 Report

This manuscript is well written and has a logical flow. However, the paper mentions molecule generation and mass resolving power and would therefore benefit from a (series) of mass spectra that demonstrate the improved accuracy and precision by using the atomic ratio vs. the molecular ratio. 

The paper would also benefit from a discussion on how its findings affect applications in the area of nuclear safeguards, nuclear forensics and the environmental transport of actinides in the environment, as these are listed as the research key words.

Author Response

Reviewer #3 and Author Responses:

Reviewer #3: This manuscript is well written and has a logical flow. However, the paper mentions molecule generation and mass resolving power and would therefore benefit from a (series) of mass spectra that demonstrate the improved accuracy and precision by using the atomic ratio vs. the molecular ratio.

Response: We have now included an additional dataset (including a series of mass scans) that allows us to compare the magnitude of the carbon molecular interferences that were raised by reviewer 1; this is achieved by conducting mass scans and full isotope ratio measurement on UO₂ that was dispersed onto a sticky carbon tab as well as mounted in epoxy. These data are summarized in figure 12 (including the mass scans), but it is important to note that the full mass scan data can also be found in supplementary table 2.  We use this new dataset to expand the discussion regarding the use of molecular species, and the potential pitfalls of this approach.

Reviewer #3: The paper would also benefit from a discussion on how its findings affect applications in the area of nuclear safeguards, nuclear forensics and the environmental transport of actinides in the environment, as these are listed as the research key words.

Response: We have expanded the discussion (see lines 728 to 743 in section 4.4) to include a preliminary assessment of whether the approach described in this paper might also be used to improve SIMS measurement of other uranium isotope ratios (e.g. ²³⁶U/²³⁸U and ²³⁴U/²³⁸U) used to understand whether a particular sample contains anthropogenically perturbed uranium. The mass scans and additional dataset described in response to this reviewer’s first point illustrate quite clearly that molecular interferences might pose a bigger problem for these other uranium isotope ratios. We go onto to discuss the possibility that if the interferences can be mitigated through appropriate sample preparation strategies, the approach described in this paper may be a viable way to overcome the fact that ²³⁴U and ²³⁶U typically occur at very low concentrations and are therefore notoriously difficult to measure by SIMS.

Round 2

Reviewer 1 Report

The authors have made an effort to address the point I raised about molecular ion interferences being the likely cause of the large variability and large isotopic biases in the isotopic ratio of oxides for many of the samples in this study.  However, I still cannot recommend publication of this article because I think it is still misleading and it leaves major unanswered questions.

The article is misleading because it gives equal weight to topographic effects as an explanation for the isotopic variability.  Reference 35 that is cited as support for the statement shows only per mil effects of increased variability in oxygen isotopic analysis.  The effects in this article are HUGE compared to that reference.  Isotopic effects (mass bias and variability) are generally small in SIMS and scale roughly with the relative mass difference.  The mass bias between U235 and U238 is typically around 1% in conventional SIMS measurements, and topography effects on isotope ratios would likely create variability of the same order.  Yet the mass biases of the oxide ratios for most of the materials studied in this work are orders of magnitude larger than that (see Figures 5, 6, 8 and 9).  Furthermore, matrix effects are known to be significant for elemental ratios in SIMS, and it is not surprising that the UO+/U+ ratios could vary significantly depending on the chemical form of uranium, and also be sensitive to topography.  However, the isotopic ratios should be much less affected and again I predict that they would be in the percent range at most if that were the only effect.  This leaves molecular ion interferences as the ONLY mechanism of the three proposed in the paper that can produce such large apparent bias effects because interferences can selectively affect one of the isotopes but not the other.

If the interference hypothesis is true, then the paper is also misleading to imply that UO2 as a chemical form is immune from these effects and can benefit from the use of oxide ratios because this specific sample happens to be free of contaminating elements.  If this same UO2 sample were ground into small particles that were mounted in epoxy I predict that it would suffer the same fate.

Clearly the major unanswered question is whether direct evidence of molecular interferences can be detected.  This is so important that the authors should not wait for a follow-up study but should do the assessment now by taking mass spectra of the suspect samples at higher sensitivity through the UO region to see if direct evidence of UC or UCH interferences can be detected (such as 235U+12C).  The 238U+12C interference would create a very large perturbation on the 234U+16O signal if that can be measured.  As I pointed out in my original review, the minor isotopes of uranium must also be measured to provide the most useful forensic information.

Another question to ponder is why the U235/U238 measurement for UF4 (Figure7) is about 20% lower than the MC-ICPMS measurement.  I suspect that either the SIMS or ICPMS measurement is in error, and that issue should definitely be cleared up.

Finally, I don’t see how the results of this study can be of practical use.  If I have an unknown sample, how can I tell a priori whether to make an isotopic measurement on U+ or UO+?  Do I need to make the measurement on both to see which gives the greater precision?  In a sample-limited situation I don’t think that would be practical.

Author Response

Reviewer #1 and Author Responses:

Reviewer #1: The authors have made an effort to address the point I raised about molecular ion interferences being the likely cause of the large variability and large isotopic biases in the isotopic ratio of oxides for many of the samples in this study.  However, I still cannot recommend publication of this article because I think it is still misleading and it leaves major unanswered questions.:              

              The article is misleading because it gives equal weight to topographic effects as an explanation for the isotopic variability.  Reference 35 that is cited as support for the statement shows only per mil effects of increased variability in oxygen isotopic analysis.  The effects in this article are HUGE compared to that reference.  Isotopic effects (mass bias and variability) are generally small in SIMS and scale roughly with the relative mass difference.  The mass bias between U235 and U238 is typically around 1% in conventional SIMS measurements, and topography effects on isotope ratios would likely create variability of the same order.  Yet the mass biases of the oxide ratios for most of the materials studied in this work are orders of magnitude larger than that (see Figures 5, 6, 8 and 9).  Furthermore, matrix effects are known to be significant for elemental ratios in SIMS, and it is not surprising that the UO+/U+ ratios could vary significantly depending on the chemical form of uranium, and also be sensitive to topography.  However, the isotopic ratios should be much less affected and again I predict that they would be in the percent range at most if that were the only effect.  This leaves molecular ion interferences as the ONLY mechanism of the three proposed in the paper that can produce such large apparent bias effects because interferences can selectively affect one of the isotopes but not the other.

Response: At no point does the paper claim that use of the molecular isotope ratio will always be a better option than the elemental isotope ratio, so the reviewer’s statement that the paper is misleading seems like an overblown criticism.  Instead, we proposed a hypothesis (that use of the molecular species could improve precision and accuracy) and then tested it using a variety of materials. For some of these materials, the hypothesis was validated whereas for others it was not. We have presented our full dataset, including data that clearly refutes the hypothesis. The point that the UO₂ might behave similarly to the other uranium compounds if it were ground into powder is valid (see the next comment), but it is quite a stretch to say that we have mislead the paper’s potential readership on this basis since we openly discuss the fact that the UO₂ occurs in a different physical format in comparison to the other uranium compounds. Furthermore we already proposed that this difference in physical format was the primary reason for the difference in behavior of the data.

              In order to further test the magnitude of carbon-based interferences, we have now included an additional dataset whereupon we compare the results (including mass scan data; see figure 12 of the revised paper) for UO₂ mounted in epoxy as well as dispersed onto a sticky carbon tab. While the ²³⁸U¹²C peak is clearly visible in the mass scans, the results suggest that the magnitude of the interference from ²³⁸U¹³C on ²³⁵U¹⁶O is insufficient to account for the more extreme variations displayed by some of the samples.  In the paper, we clearly state that topography has been shown to influence isotope ratios at the permil level, whereas the variations observed in our study are at the percent level; so again, the reviewer’s claim that we are trying to mislead the readership is not warranted. In the end, we feel that we have taken a balanced approach in this paper – we have presented a full dataset (including results that refute our hypothesis), and then discussed three possible mechanisms to explain the observations. 

Reviewer #1: If the interference hypothesis is true, then the paper is also misleading to imply that UO₂ as a chemical form is immune from these effects and can benefit from the use of oxide ratios because this specific sample happens to be free of contaminating elements.  If this same UO₂ sample were ground into small particles that were mounted in epoxy I predict that it would suffer the same fate.

Response: We have now included a dataset that allows us to directly assess the role of carbon-based interferences in deviating the molecular isotope ratio. This is achieved by comparison of data from UO₂ mounted in epoxy with data from UO₂ dispersed onto a stick carbon tab (see figure 12 of the revised paper). First and foremost, the carbon molecular species ²³⁸U¹²C is observable in mass scans acquired from both preparations. Therefore, it is misleading of the reviewer to have assumed that the polished UO₂ did not contain any carbon contamination. Even for the UO₂ dispersed onto the sticky carbon tab, the ²³⁸U¹²C peak is relatively minor compared to the ²³⁸U¹⁶O, and calculations demonstrate that the ²³⁸U¹³C could, at most, account for a few percent of the inferred ²³⁵U¹⁶O signal.  We show that, even when there is an abundance of carbon present within the analytical volume, the magnitude of the expected deviation cannot account for the extreme deviations observed for some of the compounds.  Furthermore, the data for the UO₂ embedded in epoxy and for which the ²³⁸U¹²C signal is present (consistent with contamination by carbon), the molecular isotope ratios still result in improved precision and accuracy compared to the elemental ratios. 

Reviewer #1: Clearly the major unanswered question is whether direct evidence of molecular interferences can be detected.  This is so important that the authors should not wait for a follow-up study but should do the assessment now by taking mass spectra of the suspect samples at higher sensitivity through the UO region to see if direct evidence of UC or UCH interferences can be detected (such as 235U+12C).  The 238U+12C interference would create a very large perturbation on the 234U+16O signal if that can be measured.  As I pointed out in my original review, the minor isotopes of uranium must also be measured to provide the most useful forensic information.

Response: Detailed mass scans have now been conducted (see figure 12 and supplemental table S2) on material specifically prepared to address this issue.  While there is evidence of molecular interference (which was not something we ever denied- quite the opposite, in fact), the magnitude of the interference does not appear to be sufficient to explain the large deviations exhibited by some of the compounds investigated. Regarding the ²³⁴U determination- we have now included a discussion of the effect of molecular interferences on other uranium isotope ratio measurements beyond the ²³⁵U/²³⁸U (see discussion section 4.4; specifically lines 728 to 743).

Reviewer #1: Another question to ponder is why the U235/U238 measurement for UF4 (Figure7) is about 20% lower than the MC-ICPMS measurement.  I suspect that either the SIMS or ICPMS measurement is in error, and that issue should definitely be cleared up.

Response: We have double-checked the veracity of the solution MC-ICP-MS measurement and have no reason to believe it is in error. Matrix effects can impact many different types of mass spectrometric isotope ratio determination, so it may be that a matrix effect has biased either the solution or SIMS measurement.  Since there are no systematic studies assessing the magnitude of the matrix effect across a wide variety of uranium compounds (including UF₄), there is really no way to assess the precise origin of the disagreement between the SIMS and solution MC-ICP-MS data for this particular compound.  

Reviewer #1: Finally, I don’t see how the results of this study can be of practical use.  If I have an unknown sample, how can I tell a priori whether to make an isotopic measurement on U+ or UO+?  Do I need to make the measurement on both to see which gives the greater precision?  In a sample-limited situation I don’t think that would be practical.

Response: As we explained in response to this reviewer’s comments on the first version of the paper, the conceptual framework of the paper is an exploration, not a methodological development or proposed best practice. At no point are we proposing a fundamental shift in the way that uranium isotope ratio measurements are conducted by SIMS.  Our study shows that use of the molecular species warrants further consideration- that is it.