THERMOFISHER SCIENTIFIC QUANT'X ANALYSIS AND INSTRUMENTATION

The following is the analytical trajectory used in the EDXRF analysis of archaeological and source standard obsidian in the EDXRF lab at Berkeley and now Albuquerque. 

The ThermoFisher Scientific Quant'X EDXRF spectrometer methods:                    

All archaeological samples are analyzed whole completely non-destructively. The results presented here are quantitative in that they are derived from "filtered" intensity values ratioed to the appropriate x-ray continuum regions through a least squares fitting formula rather than plotting the proportions of the net intensities in a ternary system (McCarthy and Schamber 1981; Schamber 1977). Or more essentially, these data through the analysis of international rock standards, allow for inter-instrument comparison with a predictable degree of certainty (Hampel 1984).            

The spectrometer is equipped with a peltier air cooled solid-state Silcon Drift Detector (SDD-Hitachi Vortex 60EX) X-ray detector, with an ultra-high-flux end window bremsstrahlung rhodium (Rh) x-ray target with a 12.5 μm brazed beryllium (Be) window and an 8.8 mm collimator (3.5 mm for samples ≤12 mm in largest diameter), an x-ray generator that operates from 4-50 kV/0.02-1.0 mA at 0.02 increments, using an IBM PC based microprocessor and WinTrace^TM 7.2 software.  The spectrometer is equipped with a 2001 min^-1 Edwards vacuum pump for the analysis of elements below titanium (Ti).  Data is acquired with a pulse processor and analog to digital converter.  The Vortex 60EX SDD detector provides a resolution <130 eV FWHM at Mn Ka₁ typical, with higher count rates than PCD (Si/Li) detectors previously used in ThermoFisher lab EDXRF systems.

TRACE ELEMENT ANALYSES

For Ti-Nb, Pb, Th elements the mid-Zb condition is used operating the x-ray tube at 35 kV, using a 0.05 mm (medium) Pd primary beam filter in an air path at 100 seconds livetime to generate x-ray intensity Ka₁-line data for elements titanium (Ti), manganese (Mn), iron (as Fe^T), cobalt (Co), nickel (Ni), copper, (Cu), zinc, (Zn), gallium (Ga), rubidium (Rb), strontium (Sr), yttrium (Y), zirconium (Zr), niobium (Nb), and La₁-line data for lead (Pb), and thorium (Th).  Not all these elements are reported since their values in many volcanic rocks is very low. Trace element intensities were converted to concentration estimates by employing a least-squares calibration line ratioed to the Compton or Rayleigh background established for each element from the analysis of international rock standards certified by the National Institute of Standards and Technology (NIST), the US. Geological Survey (USGS), Canadian Centre for Mineral and Energy Technology, the Japanese Geological Survey, and the Centre de Recherches Pétrographiques et Géochimiques in France (Govindaraju 1994). Line fitting is linear (XML) for all elements but Fe where a derivative fitting is used to improve the fit for iron and thus for all the other elements.  When barium (Ba) and cerium (Ce) data are acquired, the Rh tube is operated at 50 kV and .25-1.0 mA in an air path at 100 seconds livetime to generate x-ray intensity Ka₁-line data, through a 0.630 mm Cu (thick) filter ratioed to a portion of the bremsstrahlung region (see Davis et al. 2011).  Further details concerning the petrological choice of these elements in Southwest obsidians is available in Shackley (2005; also Mahood and Stimac 1991; Hughes and Smith 1993; Shackley et al. 2018). A suite of 20 specific standards used for the best fit regression calibration for elements Ti- Nb, Pb, ,Th, and Ba include G-2 (basalt), AGV-2 (andesite), GSP-2 (granodiorite), SY-2 (syenite), BHVO-2 (hawaiite), STM-1 (syenite), QLO-1 (quartz latite), RGM-1 (obsidian), W-2 (diabase), BIR-1 (basalt), SDC-1 (mica schist), BCR-2 (basalt), TLM-1 (tonalite), SCO-1 (shale), NOD-A-1 (manganese), NOD-P-1 (manganese) all US Geological Survey standards, NBS (NIST)-278 (obsidian) National Institute of Standards and Technology, BR-N (basalt) from the Centre de Recherches Pétrographiques et Géochimiques in France, and JR-1 and JR-2 (obsidian) from the Geological Survey of Japan (Govindaraju 1994).

Very Small Samples: When a sample is ≤12 mm in largest dimension, the tube voltage (kV) can be increased to 45 kV and automated mA based on mass, and a 3.5 mm beam collimator used.  If the configuration of the flake (concave= not good, flat=good) this often will yield results that allows for relatively confident assignment to source (see Davis et al. 2011; Hughes 2010).  Success down to flake sizes < 9 mm is typical.  In SDD systems Rb is often most affected by small sample sizes so assignment to source may require use of other trace elements.

            The data from the WinTrace™ software are translated directly into Excel for Windows software for manipulation and on into SPSS ver 27 or JMP ver 17 for Windows for statistical analyses when necessary. In order to evaluate these quantitative determinations, machine data were compared to measurements of known standards during each run (Table 1).    An appropriate standard (i.e. RGM-1, NBS-278, G-2, AGV-2) is analyzed during each sample run for artifacts to check machine calibration.  Other appropriate standards from the above list are used for other volcanic or plutonic rocks.

DETECTION LIMITS

Detection limits for EDXRF have been calculated at 6σ here

"Matrix" issues in the analysis of standards:  Recently, a number of scholars have questioned the validity of using pressed powder pellets of international standards for empirical calibration and data checking, called a "matrix issue" by some questioning the potential differing analytical results between pressed powder and whole rock samples (Jeff Ferguson, personal communication, 2010; cf. Shackley 2011a).  This potential problem was tested in detail in Kathleen Davis et. al.'s (1998, reprinted 2011:45-64) study of EDXRF on obsidian and found to not be an issue (see also Shackley and Hampel 1992).  Hermes and Ritchie (1997) derived similar conclusions using EDXRF with archaeological felsites. Table 1 exhibits the analysis of one of my pressed powder pellets of USGS RGM-1 and a flake from the same USGS collected boulder sent by Steve Wilson of the USGS.  As one can see, the variability is within 1% or less in most cases, and some of this variability is inherent in the native variability within the single 200 kg boulder collected at Glass Mountain by USGS, not differing matrices.  This issue is well examined by Ron Jenkins and seems to be misunderstood in the discipline (c.f. Jenkins 1999:167-172).  While it is true that his discussion hinges on the assumption of homogeneity, one must assume that for purposes of non-destructive EDXRF that pressed powder standards are heterogeneous while obsidian is homogeneous.  This is, indeed, not the case for obsidian studies in general.  Jenkins explains:

Primary absorption occurs because all atoms of the specimen matrix will absorb photons from the primary source.  Since there is a competition for these primary photons by the atoms making up the specimen [pressed powder versus glass], the intensity/wavelength distribution of these photons available for the excitation of a given analyte element may be modified by other matrix effects (Jenkins 1999:168; emphasis mine).

In the case of pressed powder pellets, the "competition" involves air molecules locked in the powder matrix that is not analyzed in obsidian studies and of course not present in volcanic glass, particularly with XRF and ICP-MS.  While it is true that the borate mix used in the pressed powder can yield some boron, this element is undetectable by XRF and so light (Z=5) as to not cause peak overlap, especially when using any tube filtering.  Perhaps more important is the issue of infinite thickness, particularly when Z≥51 and high tube voltages are required.  In this case, some of the x-rays are radiated through the sample, and ratioing to the Compton or bremsstrahlung regions can yield complex results, as can be seen in the Table below for Ba, and evidently a problem for other analysts as well as seen by the GeoRem values in Table 1.  In this case, careful calibration can mitigate some of this error.  Parenthetically, only recently have PXRF products been able to reach 50 kV in order to adequately move electrons out of orbit for those atoms Z≥51, despite opinion to the contrary (see Jenkins 1999:9-12). 

As I've noted in print and in discussions with many of my colleagues, not using international standards, severely hinders the potential for establishing validity in obsidian research - it is nearly impossible to compare results across laboratories (Shackley 2010, 2011a; Speakman and Shackley 2013).  Using the "matrix" issue as an excuse is unacceptable and just spitting in the wind.

Table 1. X-ray fluorescence concentrations for selected trace elements of USGS RGM-1 pressed powder pellet (n=52 runs; 1/19/2023) on the Si/Li detector instrument, and whole rock flake from original USGS boulder (n=20 runs; 1/20/2023). ± values represent first standard deviation computations for the group of measurements. All values are in parts per million (ppm) as reported in Govindaraju (1994), USGS, GeoRem, and this study. RGM-1 is a U.S. Geological Survey obsidian standard obtained from Glass Mountain, Medicine Lake Highlands Volcanic Field, northern California.  Pressed powder and whole rock elemental analysis of USGS RGM-1 obsidian standard on the ThermoScientific Quant'X and USGS, Govindaraju (1994), and GeoRem recommended values.

Back to the XRF Lab Home Page

Revised:02 August 2024 13:35 -0600