Determining the Geologic Provenance of Tiny Obsidian Flakes in Archaeology Using Nondestructive EDXRF

The beginning of the 1960s witnessed the widespread introduction of physical sciences methods and techniques in archaeology. Along with instrumental neutron activation, X-ray fluorescence (XRF) analysis was among the first techniques to be applied to archaeological research,1,2 and the subsequent collaboration between archaeologists, geologists, and chemists provided archaeology with a new ability to employ instrumental measurement techniques to determine, with a great degree of confidence, the parent geological source materials from which prehistoric obsidian artifacts were manufactured.

Obsidian, a natural volcanic glass, was one of the first materials to be subjected to XRF analysis because of its distinctive chemical properties. Since obsidian is, in essence, a supercooled silica liquid, its chemical composition—major, minor, trace, and rare earth elements—is usually quite homogeneous, and researchers quickly discovered that each obsidian “source” (more properly, a distinctive chemical entity) usually possessed a unique combination and concentration of elements (often termed a chemical “fingerprint”) that allowed parent geologic eruptive sources to be distinguished chemically from one another. This chemical distinctiveness allowed researchers to pinpoint the geologic “source” for archaeological artifacts and, by using ancillary dating information, to identify change and continuity in the prehistoric use of particular obsidian sources through time.

Archaeological issues

Despite the benefits of instrumental techniques like XRF for achieving archaeological goals, one of the limitations has been the difficulty of analyzing very small, thin obsidian specimens that typically occur in western North American archaeological sites as chipping waste and artifact manufacturing residue. In addition to taking into consideration X-ray geometry issues, X-ray absorption, and matrix effects,3-7 for quantitative analysis most matrix-correction algorithms require that the sample being analyzed is sufficiently thick so that the measured concentration does not change as a function of sample thickness. Samples that meet this criterion are sometimes referred to as being "infinitely thick" for the elements of interest. Hughes8 showed that relatively large obsidian artifacts, even those biconvex and lenticular in cross-section, were adequate to satisfy these requirements for certain trace elements, but problems were encountered when analyzing very small obsidian flakes (i.e., those <10 mm in diameter and <ca. 1.5 mm thick).

Small obsidian samples (i.e., those not infinitely thick or infinitely thin) pose problems for nondestructive quantitative energy-dispersive X-ray fluorescence (EDXRF) analysis in part because as the specimen becomes smaller and thinner the proportion of X-rays emitted from the sample decreases in relation to those contributed by surrounding media.9 This can be compensated for to a certain extent by using a primary beam collimator but, depending on the amount of collimation, the number of incoming X-rays from the sample can be reduced to the point that analysis times become prohibitive for very small artifacts. Using a 75-kV tube, special filters, and secondary targets, Giauque et al.9 have successfully generated precise quantitative EDXRF data for small obsidian samples, but their customized analytical setup is difficult to replicate with commercially available EDXRF units that typically offer a much more limited set of X-ray tubes and primary and secondary filter options. A different analytical approach has been applied here for the analysis of small obsidian specimens.

Figure 1 - Obsidian samples from Annadel, Lookout Mountain,and Mt. Hicks showing the size range of flakes subjected to EDXRF analysis (see Table 1). Note: The dimensions of the smallest specimen (those at the far right in the figure) analyzed from each source are as follows. Annadel sample G: length = 3.07 mm, width = 3.37 mm,thickness = 0.46 mm, weight = 5.5 mg; Lookout Mountain sample G: length = 5.43 mm, width =3.21 mm, thickness = 0.87 mm, weight = 11.7 mg;Mt. Hicks sample G: length = 4.32 mm,width =3.24 mm, thickness = 0.57 mm,weight = 7.7 mg.

Early in the history of archaeological obsidian characterization, Robert Jack10-12 noted the advantages of using certain mid-Z elements (those between 37 and 41 in atomic number) to fingerprint obsidian sources, pointing out that because these elements are adjacent to one another in the periodic table, very similar excitation and detection efficiencies could be achieved, thus facilitating direct comparison of adjacent element relative intensities. Although Jack and Carmichael's research employed wavelength-dispersive X-ray fluorescence (WDXRF), this same basic approach can be utilized via EDXRF analysis because elements adjacent in atomic number have X-ray emission lines that are close in energy. For this reason, matrix effects, sample thickness, and X-ray geometry issues affect each element in the same way, and ratios of the fluorescent emission line intensities between and among elements are highly stable. Unlike earlier WDXRF analyses, however, those conducted by EDXRF13 are significantly faster, taking as few as 30 live time seconds to generate large numbers of counts for mid-Z elements, due to the high-flux end window X-ray tubes and digital signal processing in use today. The present work, an outgrowth of research conducted over the past few years (e.g., Hughes14,15), was undertaken to illustrate the utility of EDXRF analysis for characterizing small obsidian flakes and to demonstrate the applicability of this research to archaeological studies.


Table 1 - Integrated net peak intensity data for obsidian samples from Annadel, Lookout Mountain, and Mt. Hicks*

Laboratory tests on obsidian samples were conducted using a QuanX-EC™ (Thermo Fisher Scientific, Waltham, MA) EDXRF spectrometer equipped with an end window silver (Ag) target X-ray tube, 50-kV X-ray generator, digital pulse processor with automated energy calibration, and Peltiercooled solid-state detector with 145-eV resolution (FWHM) at 5.9 keV. The X-ray tube was operated at differing voltage and current settings, and with different primary beam filters, to optimize excitation of the elements selected for analysis. In this case analyses were conducted for Rb, Sr, Y, Zr, Nb, Fe, and Mn using the Ka emission line for each element. The analyses were conducted for 30 dead time-corrected seconds, with tube current scaled to compensate for the physical size of the specimen. Background and peak overlaps were subtracted to generate integrated intensity net count rate data for trace elements between 25 and 26 and 37 and 41 in atomic number.

To facilitate controlled comparisons between large obsidian artifacts (those of sufficient physical thickness to generate reliable quantitative composition estimates) and smaller, thinner obsidian samples (of insufficient physical size to generate replicable and reliable quantitative composition estimates), integrated net count data for a suite of trace elements were first generated on suitably large in-house geologic reference standards. Geologic obsidian samples from three well-known western North American obsidian sources—Annadel, located in the North Coast Range of northern California; Lookout Mountain, located in the Casa Diablo area of central eastern California; and Mt. Hicks, located in southwestern Mineral County, Nevada—were collected and an experienced flintknapper used one obsidian cobble from each source to make an obsidian artifact. All manufacturing debris from each cobble was saved, and was sorted by size class before being subjected to EDXRF analysis. Figure 1 illustrates the size range of the obsidian samples used in this analysis.

Figure 2 - Mid-Z spectra of three Annadel obsidian samples. All specimens were detached from the same obsidian cobble, sorted by size. A = the largest specimen,G = smallest specimen.

Integrated net count rate data for large samples were generated, and these served as the baseline control for each obsidian source. Following these analyses, integrated net count rate data were then obtained for very small flakes (many <4 mm in diameter) detached from large geological samples from each obsidian source using the same analysis conditions applied to large samples (see Table 1). Resulting data for each element of interest were compared directly to values obtained from the baseline control groups to determine whether the same substantive result was achieved. As Table 1 shows, although higher numbers of counts were generated on larger, thicker specimens, the ratios of counts between large and very tiny, thin flakes are practically identical (see Figure 2).

Figure 3 - Ternary diagram plot showing the relative net peak intensity proportions of Rb, Sr, and Zr in geologic samples from Annadel, Lookout Mountain, and Mt. Hicks obsidian sources in relation to other regionally significant obsidians. Black triangles plot infinitely thick samples from each source; red triangles plot smaller flakes in Table 1.

Figure 3 illustrates these concordances using Rb/Sr/Zr ratio data, while Figure 4 shows them using Zr/Y versus Rb/Sr ratios. Lower count rates were achieved for very small specimens principally because the X-ray beam excitation area was only minimally filled by the small target samples, which were held in place in a sample cup using very thin (ca. 4 µm) polypropylene film. Use of a primary beam collimator helped to more precisely focus the X-ray beam on each tiny sample, but reducing the beam size dramatically decreased the count rates for very small samples, meaning that counting time would have had to be extended significantly in order to achieve adequate numbers of counts. Given the goals of this study, the count rates/second achieved were adequate to draw distinctions between the trace element compositions of these obsidians.