Experimental procedure

Analysis was performed by LA-ICP-MS at the Centre Ernest-Babelon, IRAMAT, at the CRNS facility, Orleans, France, under the supervisor of Dr Bernard Gratuze. There were two campaigns of analysis, the first in July 2013 with 96 samples and the second in February 2015 with 196 samples, making a total of 292. The second campaign used a slightly different analytical arrangement with a new laser system, so the two campaigns will be described separately.

5.3.2.1 Campaign 1

The LA-ICP-MS utilised a VG UV-laser connected to a Thermo Fisher Scientific Element XR mass spectrometer. The mass spectrometer was equipped with a three stage detector containing a dual mode (counting and analog) secondary electron multiplier (SEM) which accommodated a linear dynamic range of over nine orders of magnitude

associated with a single Faraday collector. This setup allows the analysis of major, minor, and trace elements in a single run regardless of their concentrations and their isotopic abundance. This is particularly important as in contrast to solution ICP-MS the dilution of samples is impossible with laser ablation. The ablating beam was generated by a Nd YAG pulsed laser operating at a wavelength of 266 nm and frequency of 7 Hz (5 Hz was used on very thin samples to reduce burn depth), running at 3-4 mJ. The laser was run for 20 seconds of pre-ablation (to burn away possible surface contamination), followed by 50 seconds ablation and collection. An argon stream (1.2 l/min) transports the material to a plasma torch for atomisation and ionisation before quantification in the mass spectrometer. Ablation scars were typically 70-100 μm in diameter and typically 400 μm in depth, but dependant on laser frequency.

The sample feeder contained ten 3 cm chambers, each able to hold around 16 samples depending on size. The samples were secured in place by putty (Figure 5.9). Nine chambers were used for samples and the final chamber utilised for the standards.

Samples were positioned on end to present the greatest depth and a fresh fractured edge was selected to avoid dirt and corrosion. Between samples a 1 minute blank was run, and between chambers a longer blank was conducted to flush out the system and measure background element levels. The background count was subtracted from the sample count for each session to remove noise.

Figure 5.9. Example of the samples held within one cell from Campaign 1.

Due to the small spot size, LA-ICP-MS requires a high level of sample homogeneity.

This was tested by analysing two sites from each sample. Differences were found to be

<10 relative percent (R%) between sites for most elements. During data collection a live count was observed so that elemental spikes or drop-offs due to inclusions or corrosion could be quickly recognised. If seen, these analyses were either adjusted to remove the problematic collection period or the results were discarded and retaken. A final test of sample homogeneity and to compare the analytical procedure against a more established technique, was to analysis the samples of Campaign 1 with EPMA and compare the results to that of LA-ICP-MS.

EPMA analysis used a WDS detector and was performed at the Wolfson Laboratory at UCL by Kevin Reeves on all the Campaign 1 samples. A total of 24 elements were measured using an average of 7 areas per sample at 800 times magnification. The EPMA was run at a 15 kV accelerating potential. A comparison of the major and minor oxides between the LA-ICP-MS and EPMA proved very favourable with an average R%

Difference between the results of <5% in MgO, Al2O3, SiO2, CaO, MnO, Na2O, Cl, Fe2O3

(Table 5.3; see Figures 5.10 and 5.11). The resultant differences between the techniques in these elements appear to suggest random scatter rather than systematic difference, however, the slightly larger negative value for Na2O (-4.67%) might suggest a systematic under-reporting by LA-ICP-MS (or over reporting by EPMA). Slightly larger differences are seen in P2O5 (7.19%), K2O (6.05%) and TiO2 (10.57%). For these elements this may well reflect increased inaccuracy in the EPMA at lower abundances due to the lower detection limits of the technique, resulting in underestimation.

Increased dispersion at lower concentrations can be seen in Figure 5.11. On the whole, this demonstrates very close similarities between the techniques and signals that that LA-ICP-MS is a reliable technique for major and minor elements quantification with comparable results to EPMA, and better results at lower concentrations.

In terms of the potential heterogeneity within the samples, the similarity between techniques shows that LA-ICP-MS was producing results representative of the bulk composition of the vessels. This confirms that the samples are homogenous and that useful data is being produced. There is only a single sample, AH 3746-06, that demonstrates large scale differences between the techniques, however this was caused by error during sample preparation, meaning proper analysis could not be conducted and so this datum will be ignored. Some others samples demonstrate small

Table 5.3. Inter-comparison of the major and minor oxides using LA-ICP-MS and EPMA of the samples from Campaign 1. Relative percentage (R%) differences demonstrate close correspondence between the two techniques. Values in weight %. N = 92 samples.

Na2O MgO Al2O3 SiO2 P2O5 Cl K2O CaO TiO2 MnO Fe2O3

LA-ICP-MS Mean 13.16 1.73 2.26 70.80 0.17 0.83 1.16 8.65 0.14 0.289 0.65

EPMA Mean 13.80 1.75 2.31 69.62 0.16 0.79 1.24 8.95 0.13 0.293 0.63

Average Difference (LA-ICP-MS - EPMA) -0.64 -0.02 -0.04 1.18 0.01 0.04 -0.07 -0.30 0.01 -0.005 0.02

Relative % Difference -4.67 -0.88 -1.90 1.70 7.19 4.46 -6.05 -3.33 10.57 -1.61 3.44

Standard Deviation 0.56 0.08 0.12 0.74 0.02 0.05 0.07 0.41 0.02 0.02 0.07

Coefficient of Variation 4.27 4.64 5.32 1.05 12.03 5.43 5.88 4.77 13.15 6.73 10.48

Figure 5.10. Comparison of the major and selected minor oxides analysed by LA-ICP-MS in Campaign 1 and EPMA. Weight %. Log scale. X=Y line and 10%

relative % boundaries are marked. Close similarity between the techniques are shown with most oxides within 10%R variation.

Figure 5.11. Comparison of the minor oxides from LA-ICP-MS of Campaign 1 and EPMA. Weight %. Log scale. X=Y line and 10% relative boundaries marked.

Close similarity is mainly shown, although with a slight underestimation of potash by EPMA and overestimation in other oxides of lower abundances, probably linked to lower detection limits of the EPMA technique.

scale differences in one or two oxides; for example, AH 374613 shows a difference of -20 R% CaO and -18 R% Al2O3; SEP 3791-08 and -15 indicates differences of -18 R% and 6.6 R% in CaO and Fe2O3 respectively; RAM 5947-14 showed a difference -28 R% in Fe2O3. These are exceptions and for most samples a very close correlation is seen between both techniques.

This comparison demonstrates that LA-ICP-MS is suitable for major and minor oxide analysis with results comparable or better than that produced by EPMA, and that the samples are homogeneous with the LA-ICP-MS results representative of the bulk sample. The EPMA data is provided in full in Appendix E.

5.2.3.2 Campaign 2

An equipment upgrade during the winter of 2014 meant that the analysis in February 2015 benefited from a new laser and gas collection system. The mass spectrometer remained unchanged but the new ablation device was a RESOnetic Resolution M50e.

This was an excimer laser produced by argon fluoride at 193nm wavelength. Power was set to 4 mJ and 7 Hz pulse rate. This system used a dual gas collection with helium released at the base of the chamber carrying the ablated material to an argon stream that transported the material to the plasma torch. Use of helium has been proved to significantly increase analyte signal intensities (Eggins et al 1998, 286). Helium flow rate was 0.6 l/min and argon at 1.2 l/min. Ablation was set for 1min 10 seconds, with 50 seconds collection time. Spot size was set to 100μm for most samples, and was only reduced when saturation occurred (reduction to 70μm at lowest). The main causes of saturation were elements with only one usable isotope, e.g. Mn, Cu, Al and Mg.

A single sample tray was used, which could hold up to 100 small 1-5mm samples along with reference standards (Figure 5.12). Blanks were run between every 15-20 samples.

During Campaign 2 only one analysis was performed per sample. The high precision of the system and the homogeneity of the samples as demonstrated from Campaign 1 allowed this to be considered acceptable. Nonetheless, methods were employed to check for heterogeneity; live counts were observed during analysis to monitor drop-offs, element spikes or saturation and if evident those results were discarded and the

sample re-run. Additionally, after the data was processed, any results considered unusual were flagged and the analysis re-run. Repeats were performed on two sample (TIB 5583-01; JER 5124-09) due to their slightly unusual composition, but near identical analyses were produced in each case. Two samples (RAM 6490-07 and -08) were repeated as they had reduced amounts of trace elements, this was found to be due to saturation of the beam and the inaccurate reading was discarded. A further five samples (TIB 5583-16; NS 6362-01; -05; HB 3032-18; JER 5124-22) were analysed twice to test agreement between analyses. Very close agreement is seen in major and minor oxides with average relative variation <3% for all except P2O5. Some larger differences were seen in certain oxides in individual results, such as P2O5 in JER 5124-22. For the trace elements, most showed <10% relative difference, exceptions at 10-20% included PbO, WO, Tm2O3, SnO2, In and ZnO. A further 11 elements had larger variations. The data for these comparisons are shown in Appendix F.

Figure 5.12. Tray of samples for analysis as part of Campaign 2. Standards are at the top and right.

5.3.3 Calibration

For both campaigns the same calibration and quantification methods were used.

The full details of the quantification methods are published in Gratuze (1999), so will only be discussed briefly here. Calibration was performed using five reference standards; NIST610, Corning B, C and D, and APL1 (in-house standard with a few specific elements – Cl, Na, Mg, Al, P, K, Ca, Mn, Fe – determined using Fast NAA; for more details on standards see Gratuze 2013). After a 2-3 hour warm up period all the standards were run and then run periodically throughout the session (Campaign 1 every 2 chambers; Campaign 2 every 20-30 analyses) to correct for drift. The standards were used to calculate the response coefficient (k) for each element (Gratuze 1999, 873) which allows the counts to be weighted. The calculated values were normalised against ²⁹Si, the internal standard, to produce a final percentage. A correction for isotopic abundances also has to be performed. Details of calculations used to produce the results are given by Gratuze (2013; 2016, 183). Corning A and NIST612 were analysed independently of calibration to provide comparative data. During the running of blanks the background levels were measured and this was subtracted from the raw counts which increased detection limits by the removal of noise. The end results are reported as element ppm, which are then converted to percentages and oxides as required. A total of 58 of elements were recorded.