Sourcing obsidian: a new optimized LA-ICP-MS protocol

Laser Ablation-Inductively Coupled Plasma-Mass Spectrometry [LA-ICP-MS] is one of the most successful analytical techniques used in archaeological sciences. Applied to the sourcing of lithic raw materials, it allows for fast and reliable analysis of large assemblages. However, the majority of published studies omit important analytical issues commonly encountered with laser ablation. This research presents a new advanced LA-ICP-MS protocol developed at Southern Cross GeoScience (SOLARIS laboratory, Southern Cross University, Australia), which optimizes the potential of this cutting-edge geochemical characterization technique for obsidian sourcing. This new protocol uses ablation lines with a reduced number of assayed elements (specific isotopes) to achieve higher sensitivity as well as increased precision and accuracy, in contrast to previous studies working with ablation points and an exhaustive list of measured isotopes. Applied to obsidian sources from the Western Mediterranean region, the Carpathian basin, and the Aegean, the results clearly differentiate between the main outcrops, thus demonstrating the efficiency of the new advanced LA-ICP-MS protocol in answering fundamental archaeological questions. Statement of significance Our new LA-ICP-MS protocol, specifically tailored for the geochemical sourcing of obsidian artefacts in the Western Mediterranean area, was developed at SOLARIS (Southern Cross GeoScience, Southern Cross University, Australia) with a top-of-the-range Agilent 7700x ICP-MS coupled to a an ESI NWR 213 Laser Ablation System. Taking into account the common analytical issues encountered with the LA-ICP-MS technique, we focused on two parameters: the use of ablation lines instead of ablation points, and the development of a reduced list of measured isotopes. The use of ablation lines aims to compensate for any sample heterogeneity, achieve a higher count rate as well as a better signal stability, and also reduce laser-induced elemental fractionation. The measured isotopes have been carefully selected amongst the most efficient to discriminate between the different obsidian sources. This shortened list of isotopes achieves precise and accurate measurements with a higher sensitivity, and with the use of ablation lines, contributes to enhancing the potential of this geochemical characterization technique for obsidian sourcing. Data availability The LA-ICP-MS results for the obsidian geological samples from the Mediterranean area are available as


Introduction
Geochemical characterization methods currently used for obsidian sourcing studies in archaeology include: X-Ray Fluorescence spectroscopy [XRF] (Carter and  were developed disregarding some of these issues. Many studies still use discrete ablation points, despite the fact that the use of ablation lines and rasters is a well-established means of overcoming elemental fractionation (Jackson 2001), which is one of the main issues of LA-ICP-MS analysis. Lines and rasters also allow for a higher count rate, achieve better signal stability and help compensate for sample heterogeneity . Most obsidian sourcing studies were also assaying up to 30 isotopes, when only a handful of these isotopes are typically used to discriminate between the obsidian sources and attribute the artefacts to those sources (see e.g. Here we present, validate, and explain the rationale underlying a protocol designed to optimize the LA-ICP-MS technique for obsidian sourcing. Geological and archaeological obsidian samples were analysed as a means of testing this new protocol, which improves analytical sensitivity, accuracy, reliability, and efficiency (i.e. swiftness in regard to the aforementioned factors) by focusing on two main changes: (a) the use of a reduced list of assayed isotopes, and (b) the use of ablation lines instead of ablation points, as advised in earlier methodological studies.

V1 and V2 protocols
The hypothesis explored here is that a reduced number of assayed isotopes can achieve a better sensitivity. This led to the development and comparison of two different protocols: one commonly found in the literature (named V1) employs an exhaustive list of measured isotopes, the secondoptimized (V2)employs a reduced list of isotopes. The instrumental settings used for both protocols are summarized in the Table 1.

Laser ablation parameters
As previously mentioned, the use of ablation lines in LA-ICP-MS analyses has been proven to reduce element fractionation, correct for sample heterogeneity and achieve higher count rates . To our knowledge, such an ablation protocol has rarely been applied to obsidian sourcing (although see e.g. In this study, we opted to use ablation lines in order to optimize the LA-ICP-MS technique. With our protocol designed for both geological and archaeological obsidian samples, the ablation settings have been tailored specifically for each sample type. The same instrumental parameters were utilized in both cases (see Table 1).

Geological samples
The geological samples were cut and embedded in an epoxy resin (Epofix, Struers), then polished down to ¼ µm (using a polycrystalline diamond solution). Before analysis, the geological samples were cleaned in distilled water in an ultrasonic bath for five minutes, then rinsed consecutively with running tap water, distilled water, and alcohol. On these polished sections, an ablation line of 1.2 mm with a scan speed of 10 µm/sec achieved a 2:15 min signal, and a spot size of 60 µm width and 5 µm depth was used to attain the best possible results. A laser output of 40% [energy per pulse ≈ 0.044 mJ] was selected.

Archaeological samples
For the archaeological samples, the protocol was adapted to minimize the impact of ablation and thus maximize the preservation of the artefact. Accordingly, the ablation line was reduced to 40 µm wide (thinner than human hair) and 0.6 mm long, making it barely visible to the naked eye and considered as virtually non-destructive. The depth of the line was increased to 10 µm in order to make up for any geochemical surface alteration (often present on artefacts; see Poupeau et al. 2010). To compensate for a loss of signal due to the shorter and narrower ablation line, the scan speed was lowered to 5 µm/sec and the output amplified to 80% [energy per pulse ≈ 0.389 mJ] instead of 40% as with the geological samples. Preparation of the archaeological samples before analysis involved cleaning in distilled water in an ultrasonic bath for five minutes, followed by successive thorough rinses of distilled water, alcohol, and acetone.

Sensitivity: V1 vs. V2 protocol
In order to compare the sensitivity of our V1 and V2 protocols, a series of measurements were obtained on the same day, under similar plasma conditions on the NIST 613 SRM. For all of the isotopes common to both protocols ( 66 Zn, 85 Rb, 88 Sr, 89 Y, 90 Zr, 93 Nb, 133 Cs, 137 Ba, 146 Nd, 147 Sm, 208 Pb, 232 Th, and 238 U), a simple comparison of the raw counts shows that higher count rates were achieved with the second protocol (Table 2), and so a higher sensitivity (raw count rate/expected concentration in ppm) was established. Indeed, since fewer isotopes are selected in the V2 protocol but the total acquisition time per line stays the same (2:15 min), each isotope signal will be acquired for a longer period (2:15 min divided by 15 instead of 30). Therefore, higher count rates were achieved, resulting in higher sensitivity. Mazet et al., in prep.) was analyzed with the V2 protocol during a total of 25 runs. In order to assess the accuracy, precision, and reproducibility of our analyses,  Table 3. For 232 Th, the relative error does not exceed 6%, and for the majority of isotopes the relative error is below 5%, and less than 3% for five of them ( 85 Rb, 88 Sr, 137 Ba, 208 Pb, and 238 U). To further our assessment of the V2 protocol accuracy, we also compared the relative error obtained on the same number of measurements (n=8) on the NIST 613 standard between the V1 and V2 protocol. For the majority of isotopes assessed, the relative error here again calculated against the reference values of the GeoRem database is lower with the V2 protocol results than the V1 protocol results (see Table 3). This new protocol is therefore producing accurate results while achieving higher sensitivity for isotope discrimination.

Precision
To compare the precision of the analysis between the exhaustive (V1) and optimized (V2) protocols, the standard error of the mean was calculated for each of the 13 isotopes assayed in both protocols (8 measurements). The results are presented in Table 4 and show, for each isotope, a considerably lower standard error of the mean for the V2 protocol as well as a lower standard deviationi.e. a higher precision of the measurements. This clearly reflects that a smaller number of isotopes assayed multiplies the measurement points, consequently increasing the precision.  The same conclusion would be made if it was possible to compare our data to previous studies using several ablation points (data unavailable/unpublished), since an ablation line is in fact constituted of a series of points, i.e. about 70 to 80 in our V2 protocol, a quantity difficult to reach in a reasonable time with punctual ablation ICP-MS analysis protocols. As demonstrated in Table 4, only the 66 Zn isotope, which may have interferences with polyatomic structures (e.g. 50 Ti 16 O; see Evans and Giglio 1993), presents a higher standard error of the mean than for the V1 protocol.

Reproducibility
The reproducibility of the analyses through time was also assessed and represents a crucial factor in archaeological studies, particularly to sourcing studies. Using the same international standard (NIST SRM 613) the evolution of the 66 Zn, 88 Sr, 133 Cs, 137 Ba, and 146 Nd contents was observed over a 6 month period, as illustrated in Fig. 1 (23 measurements represented). The variations frequently remain within a 2s range, thus attesting the repeatability of these measurements.

Matrix-induced effect and comparison to a common protocol
The BCR-2G standard (glass, basaltic composition; USGS, 2014) from the U.S. Geological Survey (USGS) was  analyzed several times to control for matrix-induced effects. The obtained average composition was compared against the USGS and GeoRem reference values, as well as against the values obtained by Barca et al. (2007) with LA-ICP-MS (see Table 5). The accuracy was assessed as the relative error between the measured values and the reference values from the GeoRem database. Accurate results were obtained and the relative error remains systematically below 10%, except for the zinc content which appears problematic. Comparing this study with the ablation point and exhaustive isotope list protocol (

Sources discrimination and provenance attribution of artefacts
The viability of a specific method for obsidian sourcing does not only lie on its reliability (in which we entail   sensitivity, precision, accuracy, and reproducibility; see e.g. Hughes 1998; Frahm 2012 for discussion), but also on its validity, i.e. its ability to distinguish between the relevant obsidian sources and to attribute obsidian artefacts from an assemblage to a specific source. The concept of source is defined in this context as a specific geochemical signature and not as a geographical location (see Hughes and Smith 1993). The primary known obsidian sources of the Western Mediterranean area, Carpathian basin, and Aegean area (Fig. 2) were considered in this study to assess the validity of the V2 protocol for obsidian sourcing: Sardinia (sub-types SA, SB1, SB2, and SC; Tykot 1997), Lipari (Pichler 1980   clearly distinguished from one another, thus confirming the validity of the V2 protocol in the geographical area considered. The validity of our protocol on the archaeological level, i.e. its capacity to attribute each artefact of an assemblage to a specific source, was assessed through the analysis of 538 archaeological samples from the Tyrrhenian area (Neolithic period).  Table 6 and are in fairly good agreement. Only the measured 88 Sr content for the SC group is slightly lower than in the other studies, i.e. 82-106 ppm (taking into consideration 1 standard deviation) while other laboratories report values ranging from 95 to 167 ppm. This difference could eventually be explained by a difference in source sampling.

Conclusions
This study demonstrates that the new LA-ICP-MS protocol developed at Southern Cross University improves analytical reliability, validity and efficiency when applied to identifying obsidian provenance in the Western Mediterranean.
Analysis of the NIST SRM 613 international standard using the enhanced protocol (V2) demonstrated improved ability to obtain accurate and precise measurements with a higher sensitivity and within a very limited time frame (3 to 5 punctual measurement of about 60 s are usually used in previous studies, where our protocol produces a series of 70 to 80 measurement points in 2:15 min). Comparing the data obtained on the BCR-2G basalt standard (USGS) by a standard protocol using ablation points and an exhaustive list of isotopes (Barca, De Francesco, and Crisci 2007), our optimized protocol using lines and fewer isotopes obtained better or comparable results, when considering the accuracy of the measurements -V1 analysis was more accurate than V2 for only 4 of 14 isotopes. Furthermore, when the V2 protocol is applied to the Mediterranean obsidian sources, differentiation between sources is particularly distinct, thus confirming the validity of the optimized protocol (V2) as a sourcing tool in obsidian provenance research. Further study is required to investigate the rather low precision and accuracy results of the 66 Zn isotope, as well as the application of the V2 protocol rationale to further obsidian sources in the Mediterranean area (e.g. Near East).
In conclusion, the use of a refined LA-ICP-MS protocol tailored specifically to the target material is a demonstrably effective means of optimizing this cutting-edge geochemical characterization technique. In obsidian sourcing, it is particularly important for a meticulous selection of isotopes to be measured in order to discriminate between the sources of a particular geographical area: the more judiciously selected the list of isotopes, the better results.

Conflict of interest statement
The authors confirm there are no conflicts of interest.

Author biographies
Marie Orange is a Ph.D student within Southern Cross GeoScience, Southern Cross University, Australia. Her research focuses on obsidian trade in the Western Mediterranean during the Neolithic period.
Dr. François-Xavier Le Bourdonnec is an Associate Professor of Archaeological Sciences at Bordeaux Montaigne University. His work deals with circulation and economy of prehistoric lithic raw materials.
Dr. Anja Scheffers is a professor at Southern Cross University, Australia. Her research focuses on how coastal environments have changed in the past. She is particularly interested in processes that shape and modify coastal landscapes over a variety of length and time scales and the coupling and feedback between such processes, their rates, and their relative roles, especially in the contexts of variation in climatic and tectonic influences and in light of changes due to human impact.
Dr. Renaud Joannes-Boyau is a Senior Research Fellow at Southern Cross University, Australia, in charge of the ESR dating and Laser-Ablation ICP-MS laboratories. His research involves the application of physical techniques to archaeological problematics, in particular the direct dating of fauna and hominid fossil remains as well as the investigation of isotopic signature in fossil teeth and bones to reconstruct dietary changes and diagenetic processes.