Análisis de la hipótesis especifica 2 Hipótesis de normalidad

7.5.1 Background information

The development of X-ray fluorescence (XRF) logging techniques has allowed the generation of continuous, high-resolution stratigraphic records from marine and lacustrine split sediment cores. XRF core scanners track down-core changes in the chemical composition of minor, major and trace elements and work through the interaction of x-rays with the surface of a split sediment core. XRF core scanner measurements are therefore non- destructive, consumable costs are relatively low and sample preparation is minimised in comparison to the XRF analysis of discrete samples. Under the influence of incoming x-ray radiation, electrons are excited to a higher energy state, in turn causing the displacement of other electrons to lower energy states to fill the resulting vacancies (figure 7.3). The surplus electromagnetic energy emitted by the displaced electrons has a wavelength(s) and frequency that are characteristic of a specific element. The characteristic electromagnetic spectra follow Moseley’s Law, an empirical law that relates the electromagnetic frequency to the atomic number of an element (Moseley, 1914). The measurement of these wavelengths allows the characterisation and proportion of elemental intensities within the analysed sample to be semi-quantitatively assessed in counts per second (cps). The incident x-ray beam only interacts with small sample volumes and so the emitted x-rays only contain information from a thin surficial layer on the sediment core surface. The response depths of the different elements within the sample depend on the wavelength of the fluorescent radiation so that for lighter elements such as Al and Si, this depth is a few μm whereas for heavier elements such as Fe the depth rises to a few hundred μm (figure 7.3) (Jansen et al., 1998). The relatively shallow penetration of x-rays below the sediment surface means that continuous XRF data may exhibit a pronounced degree of scatter as the analysis are sensitive to small morphological variations in the core surface. Data scatter may also result from variations in sediment porosity, density, water content and grain size (Boning et al., 2007).

The major advantage of scanning XRF over conventional geochemical analytical techniques is that measurements are non-destructive and can be taken at high spatial resolution and obtained directly from the sediment surface, thus negating the need for lengthy sample preparation. The advantages of scanning XRF suggest that this technique has great potential as a tool for palaeoenvironmental research if the results are properly converted into elemental concentrations by means of an appropriately tested calibration model (Weltje and Tjallingii, 2008). In addition to providing a useful device for examining palaeoenvironmental data, XRF core scanners also offer a means of initial (shipboard) correlation between sediment cores as well as other applications such as stratigraphic interpretations including the tuning of sediment sequences to astronomical forcing (Pӓlike et al., 2001), aeolian dust fluxes (Jahn et al., 2003), diagenetic processes (Funk et al., 2004) and the recognition of sedimentological events such as turbidites (Richter et al., 2006). Figure 7.3 Schematic illustration of excitation geometry at the sediment surface - analyser interface in XRF core scanner analysis. The critical depth (Δx) of elements Si, Ca and Fe (not to scale) is linked to the amounts of fluorescence energies emitted by different elements. Inset image show the Avaatech XRF core scanner within the Department of Earth Sciences, University College London. Modified from Tjallingii (2007).

7.5.2 Core preparation and measurement

Analysis of archive core halves was conducted on an Avaatech XRF core scanner housed in the Department of Earth Sciences, University College London (figure 7.3). A detailed technical description of the use of the Avaatech core scanner is given in Richter et al. (2006). The Avaatech core scanner is equipped with a Rhodium X-ray source and thermoelectrically-cooled Peltier spectrometer detector, a 125 μm Beryllium window and has an operating voltage range of 7-50 kV capable of detecting elements from Aluminium to Uranium. The system used a helium-flushed prism in order to minimise radiation scattering between the incoming X-ray beams and outgoing XRF radiation. The analytical precision of the XRF core scanner varies with element intensity that is linked to element concentration, lithology and the physical properties (e.g. grain size, porosity) of the sample material, as well as with changes in hardware settings (Röhl and Abrams, 2000; Westerhold, 2003).

Before scanning XRF analysis can be conducted, samples must be homogenous, dry and have a smooth, flat surface. Split sediment cores, especially older material, are susceptible to cracking as a result of drying out or gas expansion and as such require preparation prior to analysis. The cores were wetted with distilled water and carefully cleaned and smoothed using a flat, rounded steel plate to remove surface irregularities. The surface was then covered with UltraleneTM X-ray transmission foil which was pressed over the smoothed core face using a roller and a brush to remove any remove any trapped air and ensure total contact between the sediment surface and transmission foil. Whilst XRF analyses are not temperature dependent, the cores were warmed to room temperature prior to analysis in order to avoid condensation of water droplets on the underside of the UltraleneTM foil which may interfere with the XRF measurements.

Each core section was scanned three times for 45 seconds per measurement at operating voltages of 10 kV with no filter (for Al, Si, P, S, Cl, K, Ca, Ti, Mn, Fe, Cu, Zn), 30 kV with a Pd-thick filter (for Br, Rb, Sr, Zr) and 50 kV with a Cu filter (for Ba, Pb, U) to measure an extensive suite of elemental intensities. Measurements were made at a resolution of 2 mm with a corresponding slit size of 2 mm and at an electrical current setting of 1.0 mA. Cores were carefully measured prior to analysis to avoid cracks and interruptions in the sediment surface. Despite this, data from the ends of core sections were found to exhibit a large degree of scatter where the sediment surface becomes more disrupted. In these instances, values were removed from the dataset.

7.5.3 Calibration of XRF data

Calibration of raw XRF counts was performed using the WinAxil XRF conversion software and WinBatch software from Canberra. The software uses an iterative least-squares fitting model with a Gaussian function in order to approximate the characteristic fluorescence lines. This enabled raw spectral data to be processed using predefined models to calculate elemental intensities, which are calculated as the peak areas of the processed model. Models were checked prior to calibration where model performance was tested using chi-squared (χ2

) statistical testing to indicate goodness-of-fit where values of χ2 = 1 indicated a perfect fit and values greater than χ2

= 3 were rejected.