Comments on Markovic et al. gchron-2024-17

This manuscript presents an impressive body of work on micro-sampling that should be of interest to ID-TIMS geochronologists and possibly more broadly to geochemists.

Except for the absence of discussion of possible complications of alpha recoil, mentioned below, I can see no major problems with the manuscript, which does a good job of explaining the methodology and evaluating possible biases. A few suggestions for improvement are given below.

If the authors choose not to indent paragraphs you should leave a space between them since otherwise the end of a paragraph is sometimes ambiguous .

The images shown as Supplementary data files should have captions explaining each of them in detail. There are a lot of complex image figures so it would help the reader if the authors could refer to specific subfigures that illustrate the various features mentioned in the text (e.g. line 221-222). Fig 5 and 6 would be clearer if the sub-figures were labelled with the appropriate sample. If a figure is not broadly comprehensible to the reader without the caption it hasn’t been composed well enough. The caption should be there to provide more detail.

Do you think it really necessary to perform CA on most zircon samples? CA is very efficient for removing internal altered zones from zircon, which affect only domains that have achieved significant radiation damage. On the other hand, if there is sufficient radiation damage even annealed unaltered zircon will dissolve during the leaching process and Pb can be preferentially leached (a fact that I discovered to my dismay early on and that we tried to explain in Das and Davis 2010 doi:10.1016/j.gca.2010.06.029). I still think that air abrasion is the best method for very old (Archean) zircon. What the authors have developed is similar to air abrasion but much more refined in that one can select internal domains instead of just removing the rims of grains.  CA in principle could leach Pb, since it all resides in damaged sites and this was a problem until Mattinson (2005) proposed annealing the sample. In this case, annealing seems to be carried out prior to machining, which means that ion damaged surfaces are exposed to HF during the CA wash. In fact that authors state that the wash is used to dissolve these sites (line 162). However, it is not clear to me whether Pb could be selectively leached compared to U. For example, I would suggest that the distinct discordance shown for Archean standard OG-1 in Fig 5C may have been induced by the CA wash procedure. Any differential leaching will become increasingly evident as the surface to volume ratio increases (i.e. as the sample becomes smaller).

Figs. 6 and 7 show that machined samples can exhibit discordance. This is significant for Mud Tank zircon and much less for GZ7 zircon. This seems surprising since Mud Tank should have much lower radiation damage. In addition to the figure that the authors use to illustrate this, it might be interesting to plot discordance versus surface to volume ratio if this can be approximated. As mentioned above, the discordance might have been induced by the CA wash, not necessarily the machining procedure. It might be interesting, for a future study, to try analyzing PFIB machined zircon without applying the HF wash to see if this mitigates the discordance problem. Femtosecond laser machined samples would probably show significant Pb loss from the melt fraction, but this should be removed by the CA leach.  

One significant consideration that should be discussed in the manuscript is the effects of alpha recoil when dealing with tiny samples. The authors should read the excellent Rohmer (2003) paper (DOI 10.1007/s00410-003-0463-0) if they haven’t already done so. The problem is that a radiogenic ²⁰⁶Pb atom gets displaced from its parent ²³⁸U atom by an average distance on the order of about 50 microns due to the 8 alpha recoil events in the 238U decay chain (Davis and Davis 2018 doi.org/10.1002/9781119227250.ch11) and probably a similar amount for ²⁰⁷Pb. Since this is comparable to machined sample sizes, the Pb that is sampled will not necessarily be representative of the parent U in the sample. At first glance one might think this to be unimportant as long as the sample is distant from a crystal face. However, zircon typically shows micron-scale oscillatory zoning (e.g. LCT-A in Fig S13), where a significant proportion of the Pb in a thin high-U zone will have been displaced outside of it. For example, in my experience, CA treatment of Archean oscillatory zoned zircon resulted in a comb-shaped sample where the high-U zones were dissolved out. The results of ID-TIMS analysis gave data above concordia with a 207Pb/206Pb age slightly too young because of implantation of Pb from the original adjacent high U zones and the difference between average recoil distances for 238U-chain and 235U-chain nuclides. This could be a fundamental limitation on sample size and should be discussed.  It might also provide an approach to more accurately measure both ²³⁸U and ²³⁵U recoil distances.

Lines 361-365: The author’s discussion of the sensitivity of their results to U blank is valid. It might also be due to variability in the assumed isotopic composition of the Pb blank. However, the factors outlined above may provide a more meaningful explanation for discordance. I think that this paragraph should be shortened.

Fig 8: I have trouble relating the symbols to the legend. What is the difference between the red symbols with the dot in the middle and those without? Why not just present the symbols in the legend as they are on the figure instead of colour coding the sample names?

Line 41: ‘enable for testing’, omit ‘for’

Line 143: Should read: ‘performed at a 30 Kv ion bean voltage’

Line 219: ‘Figs 3, 4(I-M) and S12’

Line 221-222: Refer to specific sub-figures that show the features mentioned in the text.

Line 253: ‘U-Pb isotope analyses were carried out on…’ (?)

Line 294: In addition to giving the size range of samples, it might be more meaningful to give their surface to volume ratios.

Line 465: I find the authors’ reference to volume imaging of zircon to be interesting. I once tried to do this with a UV microscope meant for biological samples but the wavelength was too long to stimulate CL. It might be possible with a short wavelength source and a confocal microscope to produce a tomographic CL image of zircon assuming that UV can penetrate far enough into the grain. Where there is sufficient damage to affect the index of refraction (e.g. Archean zircon), it might be possible to do it optically. I hope that the authors will consider the possibility further.

Don Davis

The authors present a detailed description and analysis of micro-sampling techniques for high-precision CA-TIMS U-Pb zircon geochronology. I agree with the authors that combined micro-sampling and CA-TIMS analysis is the future of high-precision U-Pb dating and consider the submitted manuscript to be an important contribution. The new techniques are exciting and the authors did a nice job of critically analyzing the impacts of PFIB and laser micro-sampling on U-Pb systematics. The study is state-of the-art and the results are presented clearly, with informative figures. I recommend publication of the article, potentially after some additional analyses, as discussed below.

General comments:

1. My first comment is that the authors should more fully recognize the previous work that has been done on laser micro-sampling for CA-TIMS U-Pb zircon geochronology. Specifically, Jim Crowley and Mark Schmitz developed and have been successfully employing laser micro-sampling at Boise State for several years. Unfortunately, they have not published the same type of detailed description of their work; however, examples of their laser micro-sampling can be found in Crowley (GSA abstract, 2018), Kovacs et al. (Engineering Geology, 2020) and Rioux et al. (JMG, 2023). The authors briefly cite the Kovaks study, but group it in with coarse mechanical micro-sampling techniques, and largely dismiss its importance:

“Over the last decades, researchers have increasingly sectioned zircon with mechanical tools such as a scalpel or using a nanosecond laser for ID-TIMS analyses, although such sampling has been coarse and largely neglected the requirement of textural homogeneity of isolated fragments (e.g., Kovacs et al., 2020; Samperton et al., 2015).”

The Boise technique uses thin (~30 microns) doubly polished zircon slabs to fully characterize the zooming patterns and chemistry of the zircon grains, before cutting the grains with a laser. Notably, the spatial resolution of the technique is similar to the current study, with the smallest micros-sampled fractions having weights of 0.1–0.5 micrograms (Crowley, 2018; Kovacs et al. appendix, 2020). 

Rather than dismissing this prior work, it would strengthen the manuscript to recognize this alternate micro-sampling technique and discuss the similarities-differences and pros-cons of each method. This prior work in no way takes away from the current study, which is of high quality and very detailed. Laser and PFIB micro-sampling of zircon is a new field and it is highly beneficial to have multiple labs experimenting with different methods.

2. Secondly, I have some general comments and questions on the origin of the scatter in the micro-sampled zircon dates. The authors did a nice job critically testing whether beam damage impacted the U-Pb systematics and working to understand the scatter in the data. The authors’ conclusion that the scatter in the data is not directly related to beam damage seems reasonable, and is expected given that the damaged lattice is likely removed during the chemical abrasion step. As a clear test of this, I liked that the authors analyzed both large and small shards of mechanically broken zircon (i.e. not PFIB or laser cut). I would like to see a larger dataset of analyses of very small mechanically broken shards for both samples, which have volumes similar to the PFIB and laser micro-sampled zircon fragments. Such a data set would allow for a direct comparison between mechanically sampled grains and PFIB and laser micro-sampled grains. If the two datasets do show similar dispersion, it would more definitively rule out beam related lattice damage/Pb loss.

The authors argue that the observed scatter in the micro-sampled fragments is most likely related to variable U blanks, which seems very possible. My most significant concern with this conclusion is that the percentage of impacted grains seems to be inconsistent with the U blank data reported in Fig. S14. The figure shows that over a 1.5 year period, 16 out of 20 U blanks had values of ~0.3 pg; however, if all of the scatter in the micro-sampled zircon dates is due to variable U blanks, Figure S14A suggests it would require U blanks >0.5 pg for 9 out of 13 analyses. The authors should discuss why the U blanks might be more variable in the actual analyses than in their blank measurements. It might also be informative to run an entire batch of blanks to better understand the variability of the U blanks within a single batch. If there is significant variability, it would suggest that the authors should be using a higher U blank value with larger uncertainties. Even based on the current data and text, the lab appears to be applying a U blank of 0.07 ± 0.02 pg (Fig. S14A), whereas measured U blanks have an average of 0.32 ± 0.08 pg (Fig. S14B). 

I am also surprised that the U blanks are so high. U blanks are typically an order of magnitude smaller than Pb blanks, as would be expected, given the much high concentrations of Pb in the environment. The U blank data presented in Figure S14B suggest that the U blanks during the time of this study are similar to or higher than the Pb blanks. This is an unexpected result, and as the authors note, raises concerns about memory effects in the micro-capsules.   

Overall, a key observation of the study is that micro-sampled zircon fragments show greater variability in dates than larger “control” fragments and appear to plot along Pb-loss or U-gain discordia. The authors provide a robust discussion of this dispersion, but the current data appear to be inconclusive. To directly test between beam damage, U-gain, and micro-scale variability within the grains, it would be interesting for the authors to complete additional analyses of mechanically generated (i.e., not laser or PFAB cut) micro-fragments of Mud Tank, GZ7—as I outlined above—and a large number of blanks. If the mechanical micro-fragments show a similar dispersion to the PFAB and laser microsamples, it would provide strong evidence that the dispersion is not related to beam damage. The co-analyzed U blanks would test whether any observed dispersion can be explained by excess U blanks. I encourage the authors to complete such a test before final publication of this manuscript. 

Detailed comments:

Lines 298–300: “Note that our analyses yield dates up to 20 Ma younger than the ones published in the literature (Black and Gulson, 1978; Horstwood et al., 2016; Gain et al., 2019).” 

The authors should discuss this in more detail. Why are their dates 20 Ma younger?

Lines 462–465: “Beyond zircon, PFIB and femtosecond laser machining may substitute microdrilling as a more precise method for obtaining texturally controlled aliquots of complex samples for isotopic analyses, as well as being applied to microsampling of other U-bearing accessory minerals such as titanite, rutile, apatite and baddeleyite, and to other radiogenic or stable isotope systems.”

I recommend changing “substitute” to “replace”.

Figure 6: It would be useful to add labels to the concordia figure indicating which graph is Mudtank versus GZ-7.