The paper “Calibration methods for laser ablation Rb–Sr geochronology: comparisons and recommendation based on NIST glass and natural reference materials” by Glorie et al provides a comparison of calibration approaches for LA-ICP-MS/MS Rb-Sr dating. As Glorie et al. show in their paper, many different protocols are currently in use and there is no consensus on a best-practice approach. Importantly, they demonstrate that calibration against Mica-Mg, a very commonly used reference material for Rb-Sr dating, can produce inaccuracy in the Rb-Sr dates. This paper is well-written and overall well presented, although some of the figures present more as drafts and could benefit from some final editing/formatting.

This paper constitutes an important contribution to the Rb-Sr community as it highlights significant pitfalls of some commonly used data reduction strategies. My main comment is that there is no code/software package provided to apply the data reduction protocols recommended by the authors, making it hard for the inclined reader to replicate the approach recommended in this work. I would strongly encourage the authors to consider either i) expanding the description of their data reduction algorithm (including equations), ii) providing a reference to a detailed protocol, iii) or sharing the in-house code the authors used, which would greatly benefit the practical usefulness of this work.

Detailed Comments:

Line 23: “as secondary RM to correct for matrix-dependent fractionation” The term secondary RM is misleading here, as it refers to a RM used for calibration purposes and not one used for QA/QC. This terminology is also inconsistent with Table 1 where “secondary RM” is used in the “traditional” way.  

Line 67: see comment above. In these two sentences “secondary RM” is actually referring to two different things. Perhaps the RM that is used to correct for matrix-dependent fractionation could be termed “matrix-calibration RM” or something similar to avoid confusion.

Lines 86-87: see comments above regarding terminology.

Lines 173-175: “using a customized data reduction algorithm that calculates error correlations from the raw isotopic ratios for each sweep in an analysis, in the same way as for U-Pb data reduction”. It would be useful to describe this in more detail or provide a reference for the interested reader. As you show in Table 1, many publications do not report error correlation at all or use an estimate. Having a best practice approach described somewhere – or even a code publicly available - would be very useful for the community. Since rho affects the precision of isochron ages, I believe this would be a useful addition to this paper.

Line 176: “(1) correcting down-hole fractionation (DHF) against the primary RM” see comment above. To my knowledge there is no publicly available software to do this (such as a DRS for iolite or similar).

190-192: “All mica samples (including biotite samples and MDC phlogopite) were ablated with the laser ablating parallel to cleavage. The Bundarra samples were analysed in thin section and the Taratap sample as a rock block.” From the petrographic description it sounds like biotite does not have a preferred orientation in the Bundarra and Taratap samples. Did you specifically target grains with the cleavage perpendicular to the surface for analysis in these samples? And why? Please clarify.

Lines 204-205: “Session-dependant correction factors (CF) were calculated from the measured 87Rb/87Sr ratio for MDC and Mica-Mg and compared to the reference value” Are these fractionation factors stable within an analytical session? If not, using a spline rather than a constant correction factor may yield better results. EDIT: This is clarified in section 4.2. It may be useful to add half a sentence to line 204 to justify the use of a constant correction factor.

Line 265: “As shown, analytical protocol (A) involving NIST as primary” Specify which NIST SRM.

Figures and Tables:

At least in the low -esolution version, some axes labels/legends are not readable (e.g. Figure 2) or seem to be cut-off on the side (e.g. Figure 4). Spelling is inconstant between and within Figures (e.g., “Mica-Mg”, “MicaMg”, “MicaMG”)

Table 1: What does “estimated formula” mean?

Figure 2: Please add mass numbers to axes labels

Figures 3&5: Color ramp on the side not labelled.

Figures 6&7: What is the uncertainty level of the error bars?

Review of GChron 2023-1915

Glorie et al.

Nick Roberts

This paper presents some issues with normalisation of LA-ICP-MS/MS Rb/Sr geochronology, namely, differences in downhole fractionation of different reference materials. The paper provides guidance on what gives the best data in terms of accuracy and precision using the most common RMs currently in circulation.

This contribution is most welcome. Clearly different strategies are in use by different labs, and highlighting potential issues with accuracy is needed as this method is rapidly growing.

It took the zircon community many years to try and adopt some of the best practise, and still there is massive variation in approach, and the methods and guidance continue to evolve. Thus, this paper will no doubt be surpassed in the future, but is highly relevant right now.

Although the data presented and the general message is fine, I feel some of the details need better communication and clarification.

The first major point is about how primary and secondary RMs are used, and the second is about uncertainty propagation. Both of these are highly relevant to other LA based geochronology methods, and are issues that equally need addressing and highlighting elsewhere.

Normalisation

As with U-Pb dating of common-lead bearing minerals (e.g. calcite), there are three main options for the normalisation of Pb/U ratios – 1) normalise to something ‘homogeneous’ such as NIST, and then conduct a secondary normalisation to a heterogeneous (but isochronous) matrix-matched mineral, 2) normalise directly to a heterogeneous matrix-matched mineral, or 3) normalise to something like NIST and then just check for accuracy using a matrix-matched mineral. The third should be avoided, but I have seen it done. Options 1 and 2 sound simple enough; however, there is some nuance here that is very important, and should be clarified. One method involves a downhole correction to NIST for the Pb/U ratios, and then just re-normalising the mean ratios offline to a secondary RM. The alternative is to use a downhole correction to the matrix-matched mineral. Again, I have seen both options used. The issue is that ablation data can be split up in Iolite and LADR type programs, such that a shorter ablation interval will have a different mean offset than the total ablation (leading to inaccurate results). I acknowledge that the authors know all this, and mention some of these issues; however, it is not clear of the exact procedure adopted here throughout. I would argue the terminology of primary and secondary RMs should be equally clear and consistent. If something is normalised first to NIST, and then to a matrix-matched mineral, this matrix-matched mineral is still a primary RM.

A figure outlining the data reduction protocol might be useful.

Uncertainties

Error correlations are not trivial calculations, and thus some description of how these are calculated by LADR might be useful.

On appearance, the uncertainty propagation seems reasonable; however, it is not entirely clear nor presented in a single section.

Is the reproducibility of the primary RM propagated as excess variance, or simply as its relative uncertainty? And on to the datapoints or the isochron? Best practise guidelines from the U-Pb community are to propagate the primary RM ratio uncertainties as ‘excess variance’ onto the datapoints. The relative age/isochron uncertainty of the primary RM should then be additionally propagated onto the age/isochron uncertainty (although admittedly this step is not part of the published guidelines in Horstwood et al., 2016).

The authors highlight some issues with the use of heterogeneous RMs, namely that if they are used for uncertainty propagation (i.e. increase the datapoint uncertainties to achieve a MSWD of 1 on the primary RM), then datapoints will have overestimated uncertainties.

There are issues with the current simplistic way of conducting data reduction, of which this is one. Arguably, datapoint uncertainties should not be estimated at all using heterogeneous RMs. Current data reduction programmes generally do not allow for such nuance.

I feel the authors could state these issues more clearly, rather than just listing which combination of RMs gives the most precise and accurate final data.

Data

I would expect a full analytical protocol table to be included.

I would expect to see at least one signal as cps or mV in the table.

Normally these tables also include concentration estimates too, although I realise these have a large uncertainty due to matrix-matching and the potential heterogeneity of the materials.

Figures 3 and 5 – The scale could be misleading, as there is no gradation in the data, they are discrete classes.

Figure 6 and 7 – These are lifted from excel by the looks of it, and could be tidied up a little.

Line 103 – state confidence limit for reference ages and ratios – i.e. 2s.

Line 176 – Presumably a background correction was calculated first?

Line 183 – What is the source of the ppm data?

Line 186 – At what stage is this done? Do you mean NIST-610 was used as the primary RM for both Rb/Sr and Sr/Sr ratios, followed by a second correction of the Rb/Sr ratios to the secondary RM?

Line 191 – see above comment on orientation.

Line 192 – surely not all micas are orientated the same way in a rock block? One should be explicit and honest – i.e. the micas were presumably orientated at varying angles, but those analysed have their cleavage close to vertical? Can the range in angles be estimated?

Line 210  - uncertainty propagation uses the relative uncertainty as % - i.e. not as excess variance?

Line 216 – can a supplementary figure be used to compare DHF across different sessions?

Line 228 – Are these with or without drift correction – please state?

Line 231 – Variations will also be due to drift in instrumental conditions that affect mass bias, such as plasma ionisation.

Line 240 – Right, so clarifying this earlier on might make it clearer for the reader.

Line 302-305 – I think there is more nuance here than is discussed – see above.

Line 315 – I would start with a simple phrasing – the datapoint uncertainties will be overestimated – then list the problems this may cause.

Line 317-323 – Some of this would be fixed if a more comprehensive data reduction strategy was employed, where uncertainty propagation and DHF are not necessarily directly linked – see above.

Line 324 – 342 – This gets very wordy. It might be better visualised with a figure in addition to the text. The message is that using RMs with heterogeneous ratios, or that cannot be measured precisely, leads to inaccurate propagation of uncertainties. This message could be clearer and stronger.