Simultaneous Quantification of Forsterite Content and Minor–Trace Elements in Olivine by LA–ICP–MS and Geological Applications in Emeishan Large Igneous Province

Round 1

Reviewer 1 Report

This m/s presents a new method to analyse major, minor and some trace elements in olivine with one LA-ICP-MS method. The key advantage of the method is that only one instrument is needed and that accurate major element data can be obtained while also analysing some low-abundance elements in a relatively short time. Thus the key audience of this m/s is the kimberlite indicator mineral exploration sector where tens of thousands of olivines are analysed, typically on EMPA, requiring long acquisition times for some of the trace elements.

Currently, the m/s is not written for that target audience and instead, in this reviewer’s opinion at least, an unwarranted straw man is set up with the 100% method against which the new method is pitched. The problem is that when an analyst a priori knows that they will only analyse olivine, then the 100% method can be simplified to something that is very similar to the method proposed here. The key strength of the 100% method is its ability to measure many different types of minerals without prior knowledge of their chemistry. The experiment set out here essentially robs the method its advantage. Therefore, this reviewer feels that the 100% method calculations performed in this m/s need to be redone, with a revised data reduction scheme that only uses Si, Mg, Fe, Mn and Ni. The internal precision of this bespoke 100% for-olivine method needs to then be compared to the internal precision of the method proposed by the authors.

With regard to precision, this reviewer has issues with the calculations performed for Figure 2, which are likely not adequate for the question. It is currently not possible to figure out what the true sources of uncertainty are because the m/s does not explain in which detector mode the data are obtained and how well the pulse count to analogue to Faraday conversions work (how accurate are they). With respect to precision, this reviewer is at a loss to understand why the same integration times were used for all analytes when it is perfectly predictable that some will yield very small signals (e.g. Li) while other (e.g. Ni) will yield very large signals. Same goes for Mg vs Fe. Why was Fe not analysed for longer to obtained better counting statistics.

For these combined reasons, this reviewer recommends that this current m/s should not be accepted without very major revision and a new round of reviews. A list of suggestions for improvement follows.

$$$$$$$$$$$$$$$$$$

 

Line 53: oscillatory zoning is not usually encountered in olivine. What is of interest is to capture diffusion profiles and also avoiding inclusions, serpentinised cracks, etc.

Line 62: replace ‘is’ with ‘was’

Lines 65-74: I feel that this section is somewhat misleading. What the authors want to do is to determine the forsterite content, which is atomic Mg/(Mg+Fe), i.e. a ratio. They mislead the reader that 5% errors on MgO and FeO would propagate to large uncertainty in Mg/(Mg+Fe). But this is not true because the main source of error in LA-ICP-MS is signal instability, which affects absolute values a lot more than ratios. This is very well documented and the readers need to go back and calculate the ratio precision for Mg/(Mg+Fe) in those early studies. When they conclude the paragraph with: “This can largely limit the potential widely application of determining Fo content in olivine using LA–ICP–MS” they are really setting up a straw man.

Lines 125-126: Of course I understand that analytical spots were selected for optical clearness to avoid ‘cracks and inclusions’. One could argue that this step of the procedure takes significant time. It is equally possible to put a grain mount or thin section into a field-emission-gun SEM, get BSE and SE images in automatic mapping mode to find clean analytical spots, as well as getting precise and accurate EDX chemical analyses of the olivine. In most modern LA-units it is then possible to easily import and coordinate the BSE images and select spots on there. My point simply is that analysis time is not the only factor in total method time effectiveness. This needs to be spelled out rather than only focussing on the pure analysis time.

Sections 2.4.1. and 2.4.2. I understand what the authors are doing. But I think the distinction between the 100% method and their method is not as clear cut as the m/s makes out. The point is quite simple. Olivine has a predictably simple chemistry and therefore the method proposed here is essentially pared down version of the 100% approach. I do not understand why Mn and NI were not also used in the calculations?

Lines 187-189: It is unclear to me why the Fo content calculation was not implemented into an Iolite DRS. It isn’t a complex calculation and Iolite easily allows users to define new channels that could calculate equations 1 through to 9.

Lines 215-219: regarding the precision of the Mg/Fe ratio, according to table 1 the dwell time for all analytes was 10 ms. Thus with a signal of 30,000 on 57Fe, only 300 counts are collected. This reviewer is at a loss to understand why the dwell time for this mass was not simply increased to improve statistics? A further question that arises is whether with larger spots sizes, the 25Mg signal would trip into analogue or even Faraday mode? No information is given where the pulse count to analogue cross over was and whether this could introduce an offset and inaccuracy. It is not sufficient to only discuss the 57Fe signal and its counting statistical error.

Lines 231-239: I’m not convinced that the modelled error shown in Fig. 2b and explained in this paragraph is relevant. I can see that the calculation would be correct for a constant error on the Fe/Mg ratio. However, the point is that with increasing fayalite content, the Fe signal would be larger, which would bring down the error. Likewise, the real error would expand strongly when moving to forsterite contents of 95 where the Fe signal would drop considerably and where the substantial background signal (which is not given but should be given) would introduce noise.

The problem here is that the demonstration of precision is made with MongOL with Fo89.5. So per figure 2a, a 44 micron spot yields a 57/25 precision of 1%. However, if the same laser and MS parameters were used to analyse a cratonic harzburgite olivine with Fo94, the Fe signal would be half as large and the error therefore at least 1.5%. This illustrates that the usefulness of Fig. 2b is very limited. The figure should not work with constant ratio errors but with errors as a function of signals on mass 25 and 57.

 

Section 3.2: Precision and accuracy. This section demonstrates that both the 100% and the new approach return highly accurate data, within 0.22 and 0.11%, much smaller than the internal precision. I can appreciate the better reproducibility of the new method but the fact is that both methods are more accurate than precise. Unfortunately, the section does not actually provide a comparison of precision other than showing the error bars in figures 3a and 3c. This section needs to explain how exactly the 100% method was implemented and I’m returning to the straw man that I alluded to before. Surely, if the purpose of an analytical campaign is to solely analyse olivine, then a pared down version of the 100% method could be written (e.g. as a DRS in Iolite). All that would be needed is to sum up Si, Mg, Fe, Ni and Mn. Of course other low signal analytes are going to add uncertainty but this normally more than outweighed by the main advantage of the 100% method, namely that you can apply it to many different minerals. Thus, if the authors want to compare like with like, they need to be fair and modify the 100% method for the purpose of the study (olivine only analyses).

 

Section 3.3: This is largely fine and demonstrates the strength of the method. I understand why the P values were corrected with a constant offset but I am uncomfortable about the Zn data. The 3 very different apparent concentrations from the 3 isotopes raise concern that none might be correct. The offset approach is only appropriate if the calibration value is inaccurate. However, if the offset is due to an unresolved molecular interference, then a constant offset calculation is not appropriate. I suggest that either, the Zn data are abandoned or the nature of the molecular interference is investigated with higher resolution mass spectrometry.

Lines 336-337. This reviewer fails to understand why the same dwell times were used for highly abundant minor elements (e.g. Ni, Mn) while low concentration trace elements (e.g. Li) were analysed with the same 10 ms integration. Unless this is a limitation of the mass spectrometer, I think it is a flaw in the method design that unnecessarily compromises the LODs.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript by Wu et al. reports a new analytical approach for analyzing a set of petrogenetically significant elements of olivine, which is a major source of information on mantle melting and thermal conditions. The relevance of this manuscript is well-founded from multiple studies on excavating minor and trace elements in usually impurity-poor mineral matrixes using improved EPMA/LA-ICP-MS/other analytical protocols, which were published recently, and thus the current study is another important contribution to optimizing research strategies for mantle and igneous petrologists. I found this manuscript quite straightforward, well-written and ready for publication with some minor comments given below.

1. Line 125. Why not checking the clarity of surface by means of SEM to make sure no microinclusions or microcracks are present?
2. Line 137. What does “low trace element” mean? Low-content? Light elements? Please clarify.
3. Line 149. You probably mean “facilitate”, not “facility”. Please check.
4. The text should be re-checked for minor spelling errors, subscript/superscript usage etc.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have made revisions that deal with the majority of the issues I identified in the original review. The fact remains that the study could have been designed better (e.g. adjust dwell times for expected signal; use different dwell times for different laser spot size, and use different fluence to avoid having Fe in pulse count and Mg in analogue mode) but the authors have toned down their claims and have also done a good job with checking out errors arising form counting statistics as a function of Fo content.

 

Some linguistic issues remain:

line 169, rephrase, it should say that the analytical precision directly depends on the Fo content;

line 218: k should be a conversion factor;

line 221: specify whether these are absolute uncertainties (by the way it is better to say uncertainty than error);

line 310: it should be dwell time not dwelling time

line 327; it would be better to say in the filed of diamond indicator mineral research

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx