2010 international workshop on XRF core scanning

Rik Tjallingii1, Y. Hamann2, D. Garbe-Schönberg3, G. Jan Weltje4, U. Röhl5 and workshop participants

1NIOZ-Royal Netherlands Institute for Sea Research, Texel, The Netherlands; Rik.Tjallingii@nioz.nl
2Geological Institute, ETH, Zürich, Switzerland
3Institute of Geosciences, CAU University of Kiel, Germany
4Department of Geotechnology, Delft University of Technology, The Netherlands
5MARUM-Center for Marine Environmental Sciences, University of Bremen, Germany

Texel, The Netherlands, 8-10 September 2010

Over the last decade, X-ray fluorescence (XRF) core scanning has become an estab-lished method for non-destructive and fast acquisition of sediment compositions (i.e., element count rates) directly at the surface of split cores. State-of-the-art core scanners can measure element intensities at sub-millimeter resolution that allow detailed recording of compositional variations in finely laminated and even varved sediments. Core-scanning data are widely applied to paleoceanographic and paleoclimate reconstructions on timescales ranging from seasonal to millions of years. New developments in data processing and calibration techniques have increased the need to exchange experiences among users at various laboratories equipped with an XRF core-scanner. Therefore, a three-day workshop was held at the Royal Netherlands Institute for Sea Research to discuss technical aspects and application challenges of XRF core scanning, in particular Avaatech scanners, in the wider field of paleoceanography.

On the first day of the workshop, leading researchers and laboratories gave an overview on applications of geochemistry to scientific problems and on the quality of geochemical data generated by XRF core scanning. The quality of XRF core-scanner data is commonly evaluated by comparing core-scanner records with destructive analyses of discrete samples (e.g., Inductively Coupled Plasma (ICP)-Optical Emission Spectroscopy or ICP-Mass Spectroscopy). Geochemical data are closed-sum data that are intrinsically correlated and cannot be directly quantified on an element-by-element basis. Geochemical data are therefore often represented as element ratios in order to interpret down-core composition variations in terms of changes in climate and environment, sediment transport mechanisms, or diagenetic conditions. Additionally, element intensities from XRF scanners are not solely related to element concentrations, but are also affected by down-core variations of physical sediment properties (size distribution, density, water content), as well as absorption and enhancement effects, and measurement geometry. It was shown that the log-ratio representation of XRF count rates and concentrations allows effective minimization of the noise caused by these down-core variations, and allows enhancement of the signal-to-noise ratio by means of appropriate multivariate filtering techniques. Log-ratio calibration permits rigorous quantification of the precision of XRF core-scanner data based on replicate measurements, which paves the way to fully quantitative applications of XRF core scanning.

Figure 1: Comparing count rates and goodness-of-fit statistics of element silicon (Si) and iron (Fe) measured on certified geochemical reference standards (x-axis; e.g., http://georem.mpch-mainz.gwdg.de/) with an Avaatech core scanner equipped with A) a pin-diode detector (X-PIPS) and B) a silicon-drift detector (SDD). The newly developed SDD detector increases the count rate (black) but also chi-square statistics (red) for Si and Fe due to higher sensitivity of this detector. The relative standard deviation (blue) decreases indicating better signal-to-noise conditions for measurements acquired with the SDD detector. The relative standard deviation is calculated as D-Area/Element-Area. For practical reasons the chi-square and relative standard deviation are plotted on a logarithmic scale.

The second day was dedicated to the discussion of the mathematical transformation of XRF spectra into elemental count rates by least-squares fitting of the characteristic X-ray peaks. Practical problems concerning data processing and goodness-of-fit parameters (e.g., chi-squared χ2) were presented by members of the MARUM XRF core scanner laboratory of the University of Bremen, Germany. Many technical issues were discussed in a lively debate between XRF core-scanner users, specialists in XRF acquisition, and specialists in XRF spectrum evaluation. One of the key points in this discussion was that the increased efficiency of recently developed digital XRF detectors significantly reduces measurement times and increases the signal-to-noise ratio. However, this increased sensitivity of digital detectors also brings out the complexity of XRF spectra, which may result in strongly increased c2 statistics suggesting a poor spectrum fit. A suitable alternative approach to the use of c2 statistics is to express the goodness-of-fit in terms of relative errors (i.e., the standard deviation as a proportion of the element intensity; Fig. 1). As a rule of thumb, elements displaying negative count rates or relative errors in excess of 10% are considered to be below the detection limit.

The third day was devoted to complementary non-destructive scanning tools, which are optional for the latest scanners (e.g., visible-light and UV digital line-scan cameras, magnetic susceptibility sensors, radiograph imagery), and their applications to sediment and coral-core analysis. In addition, laser-ablation ICP-spectroscopy was presented as a complementary destructive chemical technique. In a final discussion, the workshop participants expressed the need for an electronic information platform to share practical experience on sample preparation, measurement techniques, data processing, technical solutions and preventive maintenance.

The next international workshop on XRF sediment core scanning will be held in two years time. More information about current developments concerning the electronic information platform and the workshop, including some of the presentations, is available at: www.nioz.nl/xrfworkshop