3.1 Sample selection
Silcrete raw material samples were collected by Dr. Kyle Brown during surveys in the Mossel Bay region on the Western Cape of South Africa from three primary outcrops, D9-1, E3-1, and I14-2 (Fig. 1; referred to as D9, E3, and I14 throughout the paper), within ~ 80 km of the Pinnacle Point archaeological site complex (Marean 2010; Brown et al. 2012). These silcrete sources are very fine-grained matrix supported silcretes and contain scattered quartz grains throughout the matrix (Murray et al. 2020). High quality silcrete of this nature are common in the archaeological record of South Africa and Australia (Holdaway et al. 2004; Domanski and Web, 2007; Porraz et al. 2013; Webb et al. 2013; Doelman et al. 2015). Source I14 is a part of the Bokkeveld formation of the Cape Town supergroup whereas sources D9 and E3 are of the Grahamstown formation. The nodules collected from all three sources are variable in shape and range from rectangular to globular in shape and angular to sub-angular in roundness. The silcrete utilized for this study derives from a dedicated research assemblage of toolstone maintained at Arizona State University for method development. The silcrete specifically used in our experiment was originally collected, heat-treated, and knapped by Kyle Brown and Simen Oestmo (Oestmo 2017).
3.2. Heat treatment experiments
The heat treatment process used for this research is described in detail by Murray et al. (2020, 2022) and Oestmo (2017), so only summary information is provided here. Silcrete nodules from each source were cut to create paired 7 x 7 x 15 cm blocks to get one heated and one unheated block from the same piece of rock. A total of 16 blocks were precut to minimize variance for the following heat-treatment procedure and flaking experiment (Oestmo 2017). Source D9 yielded three heated and three unheated blocks, I14 produced a single pair of one heated and one unheated, and E3 produced four heated and four unheated blocks. The eight silcrete blocks designated for heat treatment (3 blocks from D9, 1 block from I14, 4 blocks from E3) were heated using temperature and duration specifications outlined by Brown et al. (2009). Samples were heated in an electric kiln fitted with an external J-Kem programmable temperature controller (Model 360/Timer-K) and Digi-Sense DualLogR thermocouple thermometer. The temperature of the furnace was ramped to 350º C over five hours. 350º C has been shown to be an optimal temperature for the silcretes surrounding Pinnacle Point (Brown et al. 2009). This temperature was held constant for 12 hours and then dropped slowly to 40º C. The cooled blocks were subsequently knapped to produce the flakes that were analyzed for geochemical signatures in this study (Oestmo, 2017).
3.3 Sample preparation
We selected 30 silcrete flakes (10 per source – 5 heated, 5 unheated) for ICP-MS analyses (solution and laser) from three South African silcrete sources (D9, E3, I14). We used a rock hammer to break off two 1 cm chunks from each flake (one for solution ICP-MS and one for LA-ICP-MS). Sub-samples (1 cm chunk) for LA-ICP-MS were embedded in epoxy (Fig. 2) and ground flat using 120 and 240 grit SiC abrasive pads. Sub-samples (1 cm chunk) for solution ICP-MS were pulverized down to a particle size of ten micrometers using a mortar and pestle. These samples were then added to a silicon nitride vial and milled for 5 minutes using a SPEX CertiPrep 8000D Ball Mill housed at Arizona State University to produce a fine powder. Between samples, we cleaned the vials by milling ashed quartz (i.e., fine sand with all organics removed) for 2 minutes and wiping down the vial using distilled water and ethanol. Prior to destruction, each flake was scanned with a Variable Inc. Spectro 1 Pro hand-held colorimeter to record quantitative color measurements (Murray et al., 2022). Further, we used a silicone peeling compound to permanently record the surface roughness of these flakes (Murray et al., 2020).
100 mg of each powdered silcrete was weighed out and digested in 3 mL HNO3, 1 mL HF, and 1.5 mL HCl in PFA vessels overnight on a hot plate (100 C). We then added 1 mL of HNO3 and HF to further digest the samples due to precipitates in the solution. All acid reagents were distilled concentrated acids, equivalent to commercial trace metal grade acids. Once fully digested, the solutions were dried down and taken up into solution with HNO3 and HCl to further drive off any residual HF. From here, we completed two dilutions to ensure that all samples were on the calibration curve (some elements were over the calibration curve so having a diluted sample helped generate better results). The first dilution consisted of extracting 0.5 mL of stock and bringing it up to 15 mL of solution with 0.32 M of HNO3 to create aliquot A. We then extracted 1.0 mL of aliquot A and brought it up to 10 mL of solution using HNO3 to create aliquot B. Three digestion blank solutions were taken through the dissolution process as well and used for instrument calibration.
3.4 Solution ICP-MS Analyses
The main goal of using solution Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) for this study is to measure elemental composition in a representative sub-sample of each silcrete specimen included in this study. Silcrete is a very heterogeneous material so a bulk analysis using digested rock powder is more likely to yield representative geochemical data. Therefore, solution-based ICP-MS was determined to be the best analytical technique to develop an overall understanding of the geochemical signature. The ICP-MS analyses were completed at Arizona State University at the Metals, Environmental, and Terrestrial Analytical Laboratory (METAL) clean chemistry facility. Samples were analyzed using a single quadrupole ICP-MS (ThermoFisher Scientific iCAP Q, with CCT option). Prior to sample measurement, the instrument passed all calibrations (mass and cross-calibration) and performance reports (i.e., sensitivity, stability, oxide production ratio, and doubly charged production ratio prior to sample measurement). Calibration standards consisted of single-element certified ICP solutions, and all samples were bracketed by calibration standards. A total of 68 elements were measured (see Table S1-1). To evaluate potential instrumental drift and plasma suppression, an internal standard solution of 200 ppb Sc, Ge, Y, In and Bi was mixed online with samples. USGS G-2 standard was also run alongside samples for Quality Control (see Table S1-2). Secondary standards (“Leonard VII”) and three blank solutions were run every five samples and showed good accuracy and reproducibility (see Table S1-3). Most elements were run in CCT-KED (“Collision Cell Technology-Kinetic Energy Discrimination”) mode with high purity He as the collision cell gas.
3.5. LA-ICP-MS
In addition to solution ICP-MS, this research tested the viability of laser ablation ICP-MS (LA-ICP-MS) as an alternative method that does not require laborious chemical preparation. Because silcrete can be difficult to digest for solution ICP-MS, we explore this approach as another option for geochemically characterizing samples. Additionally, unlike non-destructive portable X-ray fluorescence, minimally destructive LA-ICP-MS provides low detection limits and a wide range of measurable analytes similar to solution ICP-MS.
LA-ICP-MS were carried out at the Liverpool, New York, USA facility of Eurofins EAG Laboratories out using an Applied Spectra 266 nm Tandem LIBS/LA system coupled to a Thermo Scientific iCAP TQ (ICP-QQQ-MS). Each specimen was analyzed with five scan line ablations in an effort to acquire representative data that accounts for intra-sample heterogeneity in a comparable manner to solution ICP-MS results (Fig. 2); this is sometimes referred to as pseudo-bulk analysis. A key difference from the solution ICP-MS analyses is that LA-ICP-MS employed a semi-quantitative calibration. LA-ICP-MS is a highly matrix-sensitive technique, and fully quantitative concentrations require the use of both a matrix-matched calibrant and an internal standard for data normalization. For this work, LA-ICP-MS calibration used multi-element doped silicate glass reference materials NIST SRM 612 and 610, without the use of internal standardization. The specific parameters of LA-ICP-MS analyses may be found in Tables 1 and 2 below.
Table 1
Laser parameters for LA-ICP-MS analyses
Ablation pattern and length | Scan line (five per sample), 8–10 mm each |
Cleaning ablation spot size and scan speed | 200 µm spot at 250 µm/second scan |
Data collection spot size and scan speed | 100 µm spot at 50 µm/second scan |
Fluence and repetition rate for all ablations | 0.6 mJ/cm2 and 10 Hz |
Background (gas blank) signal window | 30 secs, prior to each data collection ablation |
Carrier gas and flow rate | 0.8 L/min He |
Table 2
Mass spectrometer parameters for LA-ICP-MS analyses
Plasma power | 1250 W |
Auxiliary gas | 1.2 L/min Ar |
Dwell time/mass | 0.1 seconds |
Analysis mode | Kinetic Energy Discrimination (He KED) |
Masses analyzed | Li7, Na23, Mg24, Al27, Si29, K39, Ca44, Sc45, Ti48, Mn55, Fe56, Fe57, Cu63, Cu65, As75, Sr88, Mo95, Mo98, La139, Ce140, Tl205, Pb207, Pb208, and Th232 |
A shorter list of analytes was measured by LA-ICP-MS than by solution ICP-MS. This is due to coupling a transient signal from LA with sequential mass spectrometry by quadrupole ICP-MS. A smaller number of analytes will typically yield lower measurement uncertainty than a longer list if all other parameters are identical. Increasing the duration of each ablation (typically by ablating a longer scan line) can mitigate this issue but sample dimensions, or the need to minimize damage, provide a constraint. The analyte list measured by LA-ICP-MS was determined by exploratory statistical interpretation of the already extant solution ICP-MS results and selecting elements likely to prove useful for silcrete source discrimination.
LA-ICP-MS data reduction was carried out in Iolite v4 (Elemental Scientific Lasers) using the Trace Elements data reduction scheme (Fig. 3). All results have been background corrected using the pre-ablation “gas blank” background signal. Calibration was carried out using measurements of NIST SRMs 612 and 610, analyzed periodically throughout the analysis run in a sample-standard bracketing approach to account for instrumental drift. As noted above, the resulting concentrations are semi-quantitative because of the lack of internal standardization. This was a deliberate choice because measuring an internal standard element by a secondary technique can add time and be cost-prohibitive to undertaking a source discrimination or provenience project.
A more economical approach that is suitable for LA-ICP-MS-based provenience studies is “Multi-Element Ratio Fingerprinting” (MERF), in which semi-quantitative concentrations for all analytes of interest are standardized to the semi-quantitative concentration of a matrix element, in this case, silicon, measured in each ablation. The resulting ratio data have functionally been corrected for differences in ablation yield between replicate ablations of a specimen, which is especially important for a heterogeneous material like silcrete. Furthermore, an additional data processing step, used here, can optionally be added. A log10 transformation of the Element X ppm/Si ppm ratio data helps to ensure that large differences in magnitude among the variables (e.g., % weight vs. sub-ppm) do not bias the results of statistical analyses.
Generally, the downside to the MERF approach is that the resulting data cannot readily be compared to results obtained using other techniques and are generally unsuitable for meta-analyses. However, the semi-quantitative ratio data produced here can always be reprocessed in Iolite at a later date if an internal standard element is subsequently measured by a secondary technique like SEM-EDS or EPMA. This makes the MERF approach compatible with both pilot studies and subsequent follow-up work using a range of instrumental chemistry techniques to yield fully quantitative data. Figure 3, step one summarizes the data processing steps used for the LA-ICP-MS MERF section of this project.
3.6 Statistical analyses
The geochemical data sets generated by solution ICP-MS, and separately by LA-ICP-MS, were each analyzed by Principal Component Analysis (PCA) to determine whether specimens could be grouped by geological source (D9, E3, I14) or treatment state (heated vs. unheated) using JMP Pro 16 (JMP Statistical Discovery LLC) statistical software. A general workflow for both solution and LA-ICP-MS data reduction and visualization is shown in Fig. 3.
For the solution ICP-MS data, we conducted a Principal Component Analysis (PCA) to determine whether the geological source and treatment state can be determined with minimal reduction steps. To identify which elements were most informative for separating sources and heat treatment, we conducted a predictor screening using the bootstrap forest partitioning platform in JMP Pro 16. We examined the top 20 elements and used a backwards stepwise approach to determine which five elements maximized group separation while increasing or maintaining the cumulative percentage of data set variation explained by Principal Component 1 (PC1) and Principal Component 2 (PC2). PC1 and PC2 for the top five elements were plotted and 95% density ellipses were calculated for each grouping (e.g., D9 heated, D9 unheated). To determine other elements useful for source discrimination, we also conducted an exploratory data analysis using R (R Core Team 2022). Specifically, we used the following packages: ggplot2 (Wickham 2016), GGally (Schloerke et al. 2021), gridExtra (Auguie 2017), and ggcorrplot (Kassambara 2023). The overall workflow is shown in Fig. 3.
For LA-ICP-MS data, visualization was similar to the solution ICP-MS methods (Fig. 3, Step 2). We conducted a predictor screening using the bootstrap forest partitioning platform in JMP Pro 16. However, screened predictors encompassed all elemental variables measured by LA-ICP-MS (n = 68) in their log10 transformed and Si-standardized ratio data format. Additionally, five multivariate outlier measurements were excluded from the data set. Similar to solution ICP-MS methods, we examined the top 20 elements to determine which five elements maximized group separation. Lastly, we conducted a PCA using the top five elements best for deciphering LA-ICP-MS data and applied it to the solution ICP-MS data, testing whether we will get similar results using both methods.