the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Jan 2024
Laser Ablation – ICP-MS measurements for high resolution chemical ice core analyses with a first application to an ice core from Skytrain Ice Rise (Antarctica)
Abstract. For many ice core related scientific questions, like the analysis of fast changing climate signals, the depth resolution of conventional methods of analysis is not sufficient. In this study we present a setup of Laser Ablation Inductively Coupled Plasma – Mass Spectrometry (LA-ICP-MS) for high resolution glacier ice impurity analysis to the sub-millimetre scale. This setup enables chemical impurity ice core analysis to a depth resolution of ~ 80 μm while consuming only minimal amounts of ice. The system was used for simultaneous analysis of sodium, magnesium and aluminium incorporated in the ice structure. In a first case study within the framework of the WACSWAIN (WArm Climate Stability of the West Antarctic ice sheet in the last INterglacial) project, the method was applied to a selection of samples from Skytrain ice core (West Antarctica), on a total length of 6.7 m consisting of about 130 single samples. The main goal of this study was to use the new LA-ICP-MS method to extract meaningful climate signals on a depth resolution level beyond the limits of Continuous Flow Analysis (CFA). A comparison between low resolution CFA data and the high resolution LA-ICP-MS data reveals generally good agreement on the decimetre scale. Stacking of parallel laser measurements together with frequency analysis is used to analyse the high resolution LA-ICP-MS data on the millimetre scale. Spectral analysis revealed that despite effects of impurity accumulation in ice crystal grain boundaries, periodic concentration changes in the Skytrain ice core on the millimetre scale can be identified in ice from 26 ka BP (kilo years before present; 1950 CE) and also from the Last Interglacial (LIG). These findings can open new possibilities for climate data interpretation with respect to fast changes in the last glacial and beyond.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(11638 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(11638 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-3071', Anonymous Referee #1, 14 Feb 2024
General comments
Hoffmann et al. present a new laser ablation inductively-coupled plasma mass spectrometry (LA-ICP-MS) setup for ice core analysis, established at the University of Cambridge. They analyzed several meters of the Skytrain ice core from coastal Antarctica in order to demonstrate the capability of the setup to detect layers thinned beyond the resolution of continuous flow analysis. Setups for LA-ICP-MS ice core analysis have now been established by several groups following an early pioneering phase in the 2000s. The interpretation of the high-resolution LA-ICP-MS ice core signals is not straight-forward but by now the technique has demonstrated to offer new means to investigate both, highly thinned stratigraphic signals and the ice-impurity interaction at the grain scale. However, compared to meltwater analyses only limited data exists so far and clearly there is still much to be learned from ice core analysis by LA-ICP-MS. It is thus of interest to the ice core community to see another group invest in this technique, in the hope that additional data obtained with different setups will also provide additional insight for various ice conditions. To fully demonstrate that the new setup can meet this expectation the work presented here could, and should, be much improved. In its present state the manuscript is not able to fully capitalize on the data, since substantial ambiguities remain about the production of the data and the interpretation. Considering the amount of work that very obviously has gone into the measurements, this is regrettable and hopefully can be remedied. In the following I am trying to provide concrete suggestions on where further work is needed in order to strengthen the manuscript. I will primarily focus on two main areas and neglect other smaller issues at this stage.
The first main area concerns the LA-ICP-MS setup and the analytical details regarding the measurement of ice core samples. Since this is the first paper about this system the analytical components matter and need to be well documented. The following points need to be addressed and clarified:
A fundamental issue seems to be that, based on the statement made in line 160, the laser did not couple reliably to the ice and that settings had to be constantly adjusted in order to achieve ablation. Is this why not one set of parameters by many were used for analysis? One would expect just one settings of ablation parameters being used (apart from preablation) like in previous studies targeting line profiles (e.g. Della Lunga et al., 2017).
In the manuscript the reported fluences vary strikingly between spot sizes: 6-8 J cm-2 (Table 2 states 6 J cm-2) for the “round” spot versus 36-55 J cm-2 (Table 2 states 36-50 J cm-2) for the rectangular spot. Is the laser really capable of providing 36-55 J cm-2 or is this a mistake? How could such a large difference in fluence needed for ablation be caused by the choice of the spot geometry, isn’t the beam homogenized and filtered by a mask to generate various spot sizes and geometries?
There are many instances throughout the manuscript where it is not clear for the reader what was done precisely and how. This needs to be made more traceable in order to understand which settings were actually used for data production. Section 2.3 is an example in this regard, with predominant use of reporting ranges (e.g. 20 – 150 Hz) and “typical” settings (“round spot of typically 150 µm”). These statements do not match the values reported in Table 2, which adds to the confusion.
Considering that fluences of 6-8 J cm-2 were needed for the “round” spot, this is significantly higher compared to previous LA-ICP-MS ice core analysis using a 193 nm excimer laser, where only 2-4 J cm-2 were required (e.g. Müller et al., 2011). This is an interesting point that deserves more attention: Is the large fluence needed due to the setup, or due to the ice? What happens if comparable settings to previous studies are used: No ablation visible or some ablation but not enough signal in the ICP-MS?
A related important point concerns the complete absence of any optical close-up images taken with the laser on-board camera (which should exist in this setup). Considering the general importance of such images, it is hard to understand why they are not mentioned here. Optical close-up images would not only help significantly also for data interpretation (see below) but should also demonstrate how ablation craters look like, in particular when using such high fluences and large spots. In addition, the optical appearance of the surface would be important to show with regards to sample preparation, where interesting differences exist compared to previous studies. After sample preparation in the cold room, the samples seem to be actually stored for a significant amount of time until they are measured. There are some hints that contamination could not be fully avoided (see below). Moreover, it would be of interest how much the ice surface changes visually after the final preparation step until measurement, e.g. by sublimation and associated widening of the grain boundaries on the surface? This could be important to better assess the impact of such matrix-related features on the signals obtained. The overview photos do not provide enough detail for this purpose.
There are discrepancies in the presentation of the washout time. Regarding the wide range in ablation settings used it should be addressed what effect these parameters had on the washout time, e.g. depending on element, spot size, geometry, fluence, etc. The so-called single pulse response (SPR) is typically measured as FW0.1M or FW0.01M (full-width at 10% of the maximum, or 1%, respectively) of single laser shots. In Figure 4, all peaks from a single spot show a FW0.1M in the range of 300-400 ms or more. This is inconsistent with the values reported in Table 3. Why does the DCI change the peak shape into a slow uptake (it seems 238U is almost going into saturation?). Further attention should be devoted to the fact that the washout times differ greatly between 238U and 23Na – what could be causing this, are the washout times different for all elements? It is stated in line 148 that the washout time was determined with a 50 µm spot – the rectangular one with fluence >36 J cm-2? What fluence was actually used for the SPR experiments on NIST glass, the same as for ice?
The calculation of the spatial resolution is important but in spite of making a serious effort, I was unable to understand how this was done. Scan speed seems to have be held constant for different spot sizes and repetition rates. Since a quadrupole ICP-MS was used, was there consideration of synchronization issues, e.g. between acquisition time and repetition rate (van Elteren et al., 2019)? A 1D line profile would be a map with just one line, but can be affected by imaging artifacts nonetheless. This is an additional issue to having a reliable estimate of the washout time, for each element (if different in washout) and ablation parameters used for data production. Then please carefully explain how the resolution was calculated for all settings.
The experiment with the blank ice is important but needs clarification. Again, one would expect this experiment to be conducted with all relevant ablation parameters that were used for data production, because they are likely to influence sensitivity in analysis. Here the “laser settings for the lowest depth resolution” (what are those?) were used. It would be important to show data for all analytes, not just Na. The initial peak (Figure 5) should not be caused by residual material in the sample line, if the helium is already flowing through the sample lines to the ICP-MS with the laser off? Figure 5 shows clear differences in baseline intensity levels, which are not discussed. Is this due to instrumental drift or, if the setup is used for analysis of other, Na-rich matrices, could this be due to remaining contamination within the ablation chamber? Relatedly, was there any correction for instrumental drifts, e.g. by ablating a line on the NIST glass standard before and after each acquisition. If not, how were drifts treated for constructing a single dataset from individual lines measured?
The discussion of the sharp peaks in the blank ice Na data needs to be clarified. If they were noise as proposed by the authors, what is the cause of this noise? They are unlikely caused by the ICP-MS because peaks are absent in gas blank but present when ablating: Evidence for the peaks being related to the ice. The authors speculate that microscopic particles could be the origin, but what about the grain boundaries? Especially relevant for Na. Carefully co-registering the signal in the ICP-MS together with the visual path of the laser crossing a boundary would be immediately helpful to answer this, even if watched by eye only. Considering these peaks, how does the blank ice signal look for the “higher depth resolution” settings? Regarding a potential cause, could it be contamination or an effect of storage and aging of the ice surface? How long has the MQ ice been stored between preparation and analysis? Maybe the occurrence of these peaks could be avoided by decontaminating the sample just before analysis. In line 192, the authors write “The average intensity of the signal retrieved from the blank ice is barely distinguishable from the one with just helium” and then in line 196 onwards “The median of the Na signals from MQ ice is on average about 9% higher than the one of the helium stream”, which is contradictory.
Ultimately, it cannot be ruled out that the peaks are related to remaining contamination on the ice, which is detectable due to the high resolution of LA-ICP-MS analysis. The removal as outliers is not convincingly justified. Visual images taken during ablation could help to test the hypothesis that these features are related to the ice.
The second area concerns signal interpretation as evidence for stratigraphic (annual) layering and the comparison with continuous flow analysis (CFA). The following points need to be addressed and clarified:
To my understanding the manuscript does not show any raw data, but only smoothed datasets. There should be generally showing of raw data (for multiple elements) and the effect of the applied smoothing.
If the goal was to identify annual layers in the LA-ICP-MS line profiles, one would have expected to see data covering a section where CFA is able to clearly resolve them. This data exists for the Skytrain ice core (Hoffmann et al., 2022). There, the LA-ICP-MS vs CFA comparison could show that i) annual layers exist for the investigated element and ii) provide guidance on how to identify them in the LA-ICP-MS data, in particular which elements, and ultimately the demonstration that annual layer detection is possible with this LA-ICP-MS data.
I think the authors have tried something similar but am not certain. Figure 6 shows an exemplary comparison between LA-ICP-MS and CFA data for a section between 83.2 and 84 m depth. On the one hand, this is below the depth of 60-70 m until which the identification of annual layers in Na was possible in CFA (Hoffmann et al., 2022). However, the authors write later in line 275 “The age model in this core section is based on annual layer counting of the CFA Na and Ca signals constrained by absolute age markers, mainly volcanic eruptions at this depth” – so was it possible to count annual layers at this depth in CFA? If so, where are they in Figure 6? On the other, the manuscript states that the expected annual layer thickness based on the age model at this depth is 11 cm, which should be within the range of Na depth resolution (3.8 – 4.7 cm, Hoffmann et al., 2022) and, the CFA Na data in fact shows 7-8 peaks within 80 cm of Figure 6. Yet the authors state in line 220 “This finding indicates, that even at this shallow core depth the sodium CFA signal might not be able to resolve seasonal variations present in the ice.” and in line 228 “This is surprising because, according to the spatial resolution of 4 cm of the CFA ICP-MS, it should in principle be possible to identify features smaller than the roughly 10 cm variations that are visible in the Na data. We attribute this remarkable smoothing of the CFA data mainly to turbulent mixing effects in the sample lines and the ion source of the CFA ICP-MS.”. This would mean that annual layers remain obscure at this depth 83.2 – 84 m and that the CFA resolution previously reported in Hoffmann et al. (2022) was overestimated? I find this confusing and tried to highlight the ambiguities.
Figure 8 is related to this issue, but shows fluorescence Ca from CFA, which has a higher depth resolution of 1.4 cm (Hoffmann et al., 2022). In Figure 8 it is compared to LA-ICP-MS Na, however. I would expect to see the direct comparison to LA-ICP-MS Ca which exists in Figure 6? Notably, also in Figure 8, CFA Ca shows about 8 peaks within 83.2 – 84 m depth.
To reiterate, showing the raw LA-ICP-MS data together with the smoothing is important. At present the smoothing levels appear arbitrary. E.g. in line 235 the authors write “The theoretical expected annual layer thickness at this depth is about 2.1 mm and therefore much smaller than the resolution of the CFA data. Again, to increase readability the LA data were smoothed using a running average of 4.5 mm.” This would mean you are eliminating a potential annual layer signal by smoothing, which could not have been what you intended?
The reasoning that spectral analysis can reveal periodic signals in the LA-ICP-MS data that correspond to annual layer predictions by the present age model is not convincing due to the following reasons: a) the annual layers are unlikely periodic over small depth intervals. The reality is much more complex, some layers can be spaced at smaller, some at greater distance (some could be missing entirely). The mentioned peaks in Figure 8 are illustrating this uneven spacing. b) the age model predictions (especially in the deeper parts) can have large uncertainties (which are not quantified in the text, however), c) it is not clear if the annual layers have even survived interactions with the ice matrix, e.g. given the potential imprint of grain boundaries on the Na signal. An in-depth test case from more shallow depth with annual layers in CFA and some indication on the imprint of the ice matrix (optical laser camera) would help to establish this approach, because at present it is not justified enough. The ambiguity regarding deep ice conditions remains regardless of this, however.
The authors are aware of these limitations and write in the conclusions “Nevertheless, the frequency analysis cannot conclusively and unrelated to other information be used to identify horizontal layering.”, but then one wonders why no other information was used here? The same goes for a more extended use of the parallel line approach. In one case, 8 parallel lines were measured and compared. Figure 10 shows the results, which are encouraging, but the “clear correlation” with a sine wave is an overstatement. There is one section with higher Na intensities and one section with lower intensities, at the bottom and the top section of the 8 lines (Figure 10, left). A longer profile would have been needed to discuss signal periodicity. A second case with 3 parallel lines is mentioned but the original profiles not shown. I would strongly recommend to extend this type of experiment with parallel lines, because it could bring much more clarity to the data especially in the deep ice. Note that I am not asking for full 2D imaging, which was clearly not within the scope of this work or maybe not possible. The importance of measuring parallel lines has been shown before (e.g. Della Lunga et al. (2017) stated that “the averaging of the LA signal between two or more parallel tracks spaced by a few millimetres is not only desirable, but necessary”). The subsampling of 2D images done by Bohleber et al. (2021) showed that the grain boundary imprint can remain dominant to parallel lines up to a resolution of a few 100 microns.
For the parallel line experiment in Figure 10 the data was filtered according to the “theoretical depth resolution of 185 µm”. Some high-frequency variations remain in the stack. The authors write in line 300: “These small scale variations can be caused by (i) incomplete decontamination of the sample surface, (ii) particles that show up as sharp peaks in the ICP-MS signal (iii) actual small scale intensity variations e.g. caused by accumulation of Na in crystal grain boundaries. Based solely on the line data presented here, which of these possible sources is the most dominant cannot be determined.” This could be very problematic, because it would mean that it cannot be ruled out that the signals are influenced by contamination? (See comments on blank ice above).
Section 3.4 greatly suffers from the initially mentioned remaining ambiguities in the data and needs to be revised carefully. In Figure 11, it is not clear how much smoothing was applied, but in any case, the low-pass filtered signal suggests close to twice as many maxima as the sine curve. It is unclear to the reader what these maxima may mean and it is unclear what uncertainty the “theoretical annual layer thickness” has for this 10 cm section. Then line 322 reads “Comparing the 2-4 mm periodicity to the sample picture in Fig. 11, suggests that this periodicity is likely to correlate with the estimated average crystal grain size in this depth. The 2-4 mm periodicity therefore probably does not represent layering in the sense of larger scale horizontal features, but accumulation of sodium in the crystal grain boundaries.” None of this is actually shown. Again, in the absence of imaging the parallel line approach could shed light on the grain boundary imprint, especially if combined with optical images of the laser camera. Notably this imprint could vary distinctly between sections and ice conditions and hence it is not sufficient to discuss this exemparily.
Just one other example of the ambiguity is the discussion of Magnesium in line 334 onwards. After noting detected areas of enhanced power in the spectral analysis, it is stated that “Magnesium is known to show a similar seasonality as sodium (e.g. Curran et al. (1998)).” However, Figure 6 shows no evidence of this potential seasonality. Then it is noted that “During the glacial, the total magnesium content contains a large fraction originating from mineral dust particles (de Angelis et al., 2013), which is not the case for sodium. The maxima in the fourier spectra at very small wavelengths of about 250 μm could hint to effects generated by ablation of such single particles. The soluble magnesium fraction can migrate into crystal grain boundaries (e.g. (Bohleber et al., 2021)). We therefore attribute the larger periods (3-6 mm and 8-12 mm) again to grain boundary effects.” The last sentence is speculative and not supported by data. Observing the laser cross grain boundaries and simultaneously recording Mg intensities would have provided important information here. The text goes on to say that ”It is remarkable, that the magnesium data shows an additional area of enhanced, although as well strongly intermittent power, around ∼ 1.5 mm period length and thus in the range of the annual layer thickness from the age model. In summary, these findings could be an indication that the magnesium is less affected by grain boundary effects than the sodium. In consequence, magnesium might be better suited for identification of annual layers, but this needs to be confirmed by more detailed analysis (parallel lines and signal stacking) in the future.” Also here, no proof for these statements referring to Mg being a better annual layer indicator are presented in the data.
This illustrates how the manuscript suffers from not having established well how an annual layer looks like in the different chemical species observed by LA-ICP-MS and discussing more thoroughly the differences between the chemical species shown in Figure 6. Based on CFA in Figure 8, Ca shows the most promising annual signal, in line with the findings by Hoffmann et al. (2022), although a better comparison with LA-ICP-MS Ca should be made in Figure 8. The “striking alignment” between CFA Ca and LA-ICP-MS Na in Figure 8 is an overstatement. Unfortunately, there is little Ca data presented due to a technical issue, line 241: “The LA-ICP-MS calcium signal suffered from a contamination issue in the ICP-MS at the time of these measurements.” How could a Ca-specific contamination in the ICP-MS have happened? This needs to be explained. Later on in the conclusions, the authors write “The calcium signal however was lost in the noise for most measurements.”
With all this taken into consideration, it is strange that no parallel lines were measured for the deepest ice samples with arguably the most complex conditions, line 362 onwards. This section is again mostly speculative about the origin of the features in the spectral analysis and needs better support by data. The conclusions need to be carefully rewritten with special emphasis on replacing speculative statements with statements that are actually supported by the results.
Specific comments
Considering the amount of major comments, I am only including a few specific comments below. This mostly concerns how references are made to existing literature, which should be extended and placed more carefully.
- Line 155: This needs to be rewritten after a more careful assessment of the washout times, although it is not entirely clear what the authors intend to show with the comparison of washout times to previous studies. Consider that the washout time as single pulse response is determined as FW0.1M or FW0.01M, and not by a single flank of the pulse signal. Otherwise you cannot compare these values to other studies.
- Line 349: Regarding the role of Aluminium and its relationship to particles, the actual relationship between a specific element and features like particles needs to be shown first and should not be assumed a-priori, e.g. some recent data by Bohleber et al. (2023).
- This is not the first study showing that heavily smoothed LA-ICP-MS signals agree with what is measured by CFA. The LA-ICP-MS vs CFA agreement has been shown and exploited in previous works (Sneed et al., 2015; Della-Lunga et al., 2017; Spaulding et al., 2017). The previous studies need to be credited accordingly in the respective statements in the text, of which there are several instances (e.g. lines 226).
References
Bohleber, P., Roman, M., Šala, M., Delmonte, B., Stenni, B., & Barbante, C. (2021). Two-dimensional impurity imaging in deep Antarctic ice cores: snapshots of three climatic periods and implications for high-resolution signal interpretation. The Cryosphere, 15(7), 3523-3538.
Bohleber, P., Stoll, N., Rittner, M., Roman, M., Weikusat, I., & Barbante, C. (2023). Geochemical Characterization of Insoluble Particle Clusters in Ice Cores Using Two‐dimensional Impurity Imaging. Geochemistry, Geophysics, Geosystems, 24(2), e2022GC010595.
Della Lunga, D., Müller, W., Rasmussen, S. O., Svensson, A., & Vallelonga, P. (2017). Calibrated cryo-cell UV-LA-ICPMS elemental concentrations from the NGRIP ice core reveal abrupt, sub-annual variability in dust across the GI-21.2 interstadial period. The Cryosphere, 11(3), 1297-1309.
Hoffmann, H. M., Grieman, M. M., King, A. C., Epifanio, J. A., Martin, K., Vladimirova, D., ... & Wolff, E. W. (2022). The ST22 chronology for the Skytrain Ice Rise ice core–Part 1: A stratigraphic chronology of the last 2000 years. Climate of the Past, 18(8), 1831-1847
Müller, W., Shelley, J. M. G., & Rasmussen, S. O. (2011). Direct chemical analysis of frozen ice cores by UV-laser ablation ICPMS. Journal of Analytical Atomic Spectrometry, 26(12), 2391-2395
Sneed, S. B., Mayewski, P. A., Sayre, W. G., Handley, M. J., Kurbatov, A. V., Taylor, K. C., ... & Spaulding, N. E. (2015). New LA-ICP-MS cryocell and calibration technique for sub-millimeter analysis of ice cores. Journal of glaciology, 61(226), 233-242.
Spaulding, N. E., Sneed, S. B., Handley, M. J., Bohleber, P., Kurbatov, A. V., Pearce, N. J., ... & Mayewski, P. A. (2017). A new multielement method for LA-ICP-MS data acquisition from glacier ice cores. Environmental science & technology, 51(22), 13282-13287.
van Elteren, J. T., Šelih, V. S., & Šala, M. (2019). Insights into the selection of 2D LA-ICP-MS (multi) elemental mapping conditions. Journal of Analytical Atomic Spectrometry, 34(9), 1919-1931.
Citation: https://doi.org/10.5194/egusphere-2023-3071-RC1 -
AC1: 'Reply on RC1', Helene Hoffmann, 05 May 2024
Thank you for your review and your valuable input. We agree that there were some technical inaccuracies and especially the signal interpretation section needed some clarification.
General comments
Hoffmann et al. present a new laser ablation inductively-coupled plasma mass spectrometry (LA-ICP-MS) setup for ice core analysis, established at the University of Cambridge. They analyzed several meters of the Skytrain ice core from coastal Antarctica in order to demonstrate the capability of the setup to detect layers thinned beyond the resolution of continuous flow analysis.
We note that although Skytrain Ice Rise is ‘coastal’ relative to the majority of Antarctic ice cores, it is actually ~700 km inland and as a result has low impurity concentrations (lower than EPICA Dome C for example). The analytical challenge regarding the limit of detection and background is substantial compared to true coastal locations.
Setups for LA-ICP-MS ice core analysis have now been established by several groups following an early pioneering phase in the 2000s. The interpretation of the high-resolution LA-ICP-MS ice core signals is not straight-forward but by now the technique has demonstrated to offer new means to investigate both, highly thinned stratigraphic signals and the ice-impurity interaction at the grain scale. However, compared to meltwater analyses only limited data exists so far and clearly there is still much to be learned from ice core analysis by LA-ICP-MS. It is thus of interest to the ice core community to see another group invest in this technique, in the hope that additional data obtained with different setups will also provide additional insight for various ice conditions.
To fully demonstrate that the new setup can meet this expectation the work presented here could, and should, be much improved. In its present state the manuscript is not able to fully capitalize on the data, since substantial ambiguities remain about the production of the data and the interpretation. Considering the amount of work that very obviously has gone into the measurements, this is regrettable and hopefully can be remedied. In the following I am trying to provide concrete suggestions on where further work is needed in order to strengthen the manuscript. I will primarily focus on two main areas and neglect other smaller issues at this stage.
The first main area concerns the LA-ICP-MS setup and the analytical details regarding the measurement of ice core samples. Since this is the first paper about this system the analytical components matter and need to be well documented. The following points need to be addressed and clarified:
A fundamental issue seems to be that, based on the statement made in line 160, the laser did not couple reliably to the ice and that settings had to be constantly adjusted in order to achieve ablation. Is this why not one set of parameters by many were used for analysis? One would expect just one settings of ablation parameters being used (apart from preablation) like in previous studies targeting line profiles (e.g. Della Lunga et al., 2017).
The laser did couple reliably with the ice. Only for the completely bubble free, totally transparent blank ice samples the laser settings needed to be adjusted, for example lower scanning speed and larger spot size was required. The settings for the glacier ice sample measurements were optimised at the beginning of every measurement session, which usually lasted 3-4 days with several weeks to months in between. During this time, the laser ablation system was not only used for ice but for geological samples and was subject to the normal processes of usage and wear. This led to the need of readjustment of the major settings (laser power, repetition rate) for each measurement session. Within the session, the settings were kept constant, to ensure that each continuous set of 8 or 16 consecutive ice core samples were analysed with the same parameters. We amended section 2.3.3 accordingly, (L 162 ff revised manuscript).
In the manuscript the reported fluences vary strikingly between spot sizes: 6-8 J cm-2 (Table 2 states 6 J cm-2) for the “round” spot versus 36-55 J cm-2 (Table 2 states 36-50 J cm-2) for the rectangular spot. Is the laser really capable of providing 36-55 J cm-2 or is this a mistake? How could such a large difference in fluence needed for ablation be caused by the choice of the spot geometry, isn’t the beam homogenized and filtered by a mask to generate various spot sizes and geometries?
The fluence was 6-8 J cm-2 for all spot geometries and sizes with only minor variations. The high numbers for the rectangular spot geometries were incorrect and based on an erroneous readout by the laser software, which was recently confirmed by the manufacturer. We corrected the values in the manuscript accordingly.
There are many instances throughout the manuscript where it is not clear for the reader what was done precisely and how. This needs to be made more traceable in order to understand which settings were actually used for data production.
We corrected the values in Table 2 and also added the specific laser repetition rates, pre-ablation settings, spot sizes and fluences used for each depth interval in Table 4.
Section 2.3 is an example in this regard, with predominant use of reporting ranges (e.g. 20 – 150 Hz) and “typical” settings (“round spot of typically 150 µm”). These statements do not match the values reported in Table 2, which adds to the confusion.
There were some small errors in the reported ranges and numbers. We corrected them and reformulated parts of section 2.3.3 to describe the procedures more precisely.
Considering that fluences of 6-8 J cm-2 were needed for the “round” spot, this is significantly higher compared to previous LA-ICP-MS ice core analysis using a 193 nm excimer laser, where only 2-4 J cm-2 were required (e.g. Müller et al., 2011). This is an interesting point that deserves more attention: Is the large fluence needed due to the setup, or due to the ice? What happens if comparable settings to previous studies are used: No ablation visible or some ablation but not enough signal in the ICP-MS?
We agree that it is higher than the fluences used in other setups. We found that with the configuration we used (e. g. 40 µm/s scanning speed, 20Hz repetition rate, we needed a fluence above 6 J cm-2 to make the ablation happen. This might be due to the finely microtomed and thus much smoother, more reflective and less uneven surface of the samples compared to other systems.
A related important point concerns the complete absence of any optical close-up images taken with the laser on-board camera (which should exist in this setup). Considering the general importance of such images, it is hard to understand why they are not mentioned here.
We agree that close up pictures can help to assess the surface conditions of the sample and we did provide them in the manuscript (appendix). The quality of the on-board camera pictures of the laser is not significantly better than the high resolution photographs of the samples before and after measurement. See comparison of large scale picture (2 x 5 cm) to example overview picture from Bohleber et al.( 2023) and on-board camera pictures below. We can share selected full resolution pictures of all samples in the supplementary information if needed and add an exemplary mosaic picture of the laser camera in the appendix (see below, 2 x 3.8 mm, vertical 150 µm ablation path on the left)
Overview pictures from different systems compared: A picture of a 2 x 5 cm sample from ~ 100 ka BP from Skytrain ice core on top, grain boundaries and bubbles clearly visible. Bottom overview picture from Bohleber et al. 2023. Note the difference in surface roughness.
Left: 2 x 3.8 mm mosaic picture of the on-board camera of the Cambridge laser system compared to an on-board camera picture from the Venice system (Bohleber et al. 2023) on the right.
Optical close-up images would not only help significantly also for data interpretation (see below) but should also demonstrate how ablation craters look like, in particular when using such high fluences and large spots. In addition, the optical appearance of the surface would be important to show with regards to sample preparation, where interesting differences exist compared to previous studies.
We added an exemplary close up image of the ablation path with the 150 µm round spot in Fig. 3 (see below). The surface in general appears to be smooth, with no visible contamination like for example scratch marks or particles.
Fig. 3 addition: close up image of laser path and grain boundary
After sample preparation in the cold room, the samples seem to be actually stored for a significant amount of time until they are measured. There are some hints that contamination could not be fully avoided (see below).
We cannot imagine how storage alone while the sample surfaces are not in contact with any material would lead to increasing contamination over time. The final preparation step with the sledge microtome creates a perfectly smooth surface, which is not touched by any material until the laser ablation measurement in the cryocell. The handling of the samples during insertion into the cryocell was done in a dry nitrogen purged glovebag to minimise any deposition of dust particles or condensation on the sample surface. We avoided another step of manual surface scraping before insertion into the cell, first because of the risk of damaging the perfectly smooth surface, which could then lead to focusing problems of the laser. Second because the risk of introducing more contamination by another manipulation of the surface and additional potential frosting was considered much higher than the potential theoretical benefits. The only realistic contamination possible are loose dust particles settling on the sample surfaces after the final preparation step. These particles would be removed during pre-ablation at the latest.
Moreover, it would be of interest how much the ice surface changes visually after the final preparation step until measurement, e.g. by sublimation and associated widening of the grain boundaries on the surface? This could be important to better assess the impact of such matrix-related features on the signals obtained. The overview photos do not provide enough detail for this purpose.
The samples were stored no longer than three days at -25°C. Longer term storage was done at liquid-nitrogen temperature, minimising sublimation (see L 89 ff revised manuscript). During the short-term storage time, we observed only minimal widening of the grain boundaries by sublimation, but the effect was very small. We observed grain boundary widths of about 5µm at maximum (see Fig. 3).
There are discrepancies in the presentation of the washout time. Regarding the wide range in ablation settings used it should be addressed what effect these parameters had on the washout time, e.g. depending on element, spot size, geometry, fluence, etc. The so-called single pulse response (SPR) is typically measured as FW0.1M or FW0.01M (full-width at 10% of the maximum, or 1%, respectively) of single laser shots. In Figure 4, all peaks from a single spot show a FW0.1M in the range of 300-400 ms or more. This is inconsistent with the values reported in Table 3. Why does the DCI change the peak shape into a slow uptake (it seems 238U is almost going into saturation?). Further attention should be devoted to the fact that the washout times differ greatly between 238U and 23Na – what could be causing this, are the washout times different for all elements? It is stated in line 148 that the washout time was determined with a 50 µm spot – the rectangular one with fluence >36 J cm-2? What fluence was actually used for the SPR experiments on NIST glass, the same as for ice?
There were some errors in the presentation of the washout experiment. We recently repeated the experiment with the DCI, replotted the data, recalculated the washout times and updated section 2.3.2 accordingly. The original figure was a composition of different measurements for different elements. We updated Fig. 4 after the repeated experiment (see below). In this experiment we included all relevant elements simultaneously. The example figure shows 23Na and 238U. An average washout time of 0.248 +/- 0.035s at 1% full width maximum for all elements was found. The washout times for all species agree within one standard deviation and there were no longer any large discrepancies found between the elements. We also updated Table 3 accordingly.
Fig. 4 revised: New results of the repeated washout experiment for Na and U.
The calculation of the spatial resolution is important but in spite of making a serious effort, I was unable to understand how this was done. Scan speed seems to have be held constant for different spot sizes and repetition rates. Since a quadrupole ICP-MS was used, was there consideration of synchronization issues, e.g. between acquisition time and repetition rate (van Elteren et al., 2019)? A 1D line profile would be a map with just one line, but can be affected by imaging artifacts nonetheless. This is an additional issue to having a reliable estimate of the washout time, for each element (if different in washout) and ablation parameters used for data production. Then please carefully explain how the resolution was calculated for all settings.
The calculation of the spatial resolution was done according to the procedure in Sneed et al. (2015), we present here an example calculation and clarified the paragraph (L 181 ff revised manuscript) Low resolution setting: Spot size d: 150 µm, repetition rate f: 20 Hz, scanning speed vscan: 40 µm/s, acquisition time tacq: 500 ms. The laser fires a 150 µm spot at 20 Hz or 20 shots per second. It moves at 40 µm/s, therefore travels 2 µm per shot. The ICP-MS acquisition time is 500 ms, which equals 10 shots. Adding the distance the laser travels during the washout time therefore the spatial resolution R is:
The experiment with the blank ice is important but needs clarification. Again, one would expect this experiment to be conducted with all relevant ablation parameters that were used for data production, because they are likely to influence sensitivity in analysis. Here the “laser settings for the lowest depth resolution” (what are those?) were used. It would be important to show data for all analytes, not just Na.
We changed Fig. 5 (see below) and included the results for all elements measured during one blank ice experiment for a 1.3 cm long laser line scan. We included all the relevant laser settings in the text and extended section 3.1 (L 195 ff rev. manuscript) accordingly.