Application of X-ray fluorescence mapping in turbiditic sandstones, Huai Bo Khong Formation of Nam Pat Group, Thailand

3.1 XRF mapping

Two to three samples were collected from each sandstone bed of the Huai Bo Khong Formation of the Nam Pat Group. The thickness of each bed ranges from 6–14 cm. The collected samples have the same provenance but different sedimentary structures and bed thickness. Beds that show soft-sediment deformation were excluded from this study because the elemental distribution has been disturbed. Each sample is assigned a set of specific numbers that is indicative of its location. The first and second sets of numbers indicate bed number and the order of sample, respectively, from a particular bed. For example, NP05-3 is the third sample collected from the fifth sandstone bed.

Each sample is cut into a thin slab perpendicular to bedding and analyzed under the µXRF microscopy (Horiba XGT-7000 X-ray Analytical Microscope) at the Department of Geology and Geophysics at Texas A&M University, Texas, USA. The Horiba -XRF uses Rh as a target material for generating X-rays, an accelerating tube voltage of 50 kV and a filament current of 1 mA. Fluorescence data are acquired at a spatial resolution of 100 µm over a scanning area of 5.12 cm × 5.12 cm at 2 accumulations. The µXRF produces spatial resolution for flat samples and the resulting fluorescence maps are visualized and analyzed using ImageJ software. Vertical intensity profile of each element is acquired by using the “Plot Profile” option under the “Analyze” tool. The values of the vertical intensity profile are the input data for the elemental variation analysis. The elements detected in the sandstone samples are Al, Ca, Fe, K, Mn, Ni, Rb, S, Si, Sr, Ti, and Zn. The µXRF scanning results of this study should be regarded as semi-quantitative only because they have not been calibrated with quantitative geochemical analysis such as inductively coupled plasma-atomic emission spectroscopy (ICP-AES).

3.2 Elemental variation analysis

Principal component analysis (PCA) is a statistical method that uses an orthogonal transformation to convert a set of observations with many variables, which in this case is the elements of interest, into a smaller set of variables while retaining trends and patterns of the data. In other words, PCA reduces the dimensionality of the variables. Eigenvectors of a correlation matrix representing the strength of the associated variables are termed principal component (PC). Elements with high correlations with the PCs explain the multivariate data. Kaiser criterion applied and PCs that have Eigenvalues greater than one are retained [22]. A scree plot [23], which is a graphical technique showing the eigenvalues of PCs, is also used in conjunction with the Kaiser criterion to determine the number of PCs to retain. Elements with high factor loading in the previously determined PCs have high correlation with the PCs and explain the multivariate data. PCA is conducted on “R” statistical software.

4.1 In situ elemental distribution and XRF map

Examples of µXRF scanning results are shown in Figure 4. Spatial distributions of K- and Fe-bearing minerals are clearly visible in Figure 4b and c, respectively. Fractures filled with Ca-bearing minerals are indicated by white areas in Figure 4d. Vertical intensity profile of each element is derived from elemental map. Figure 5 depicts an example of vertical intensity profile of Fe count from the µXRF scanning result shown in Figure 4c; the Fe count is increasing upwards which corresponds to Fe-richness. These vertical intensity profiles are further used for PCA in order to determine a set of elements that explain the trends of the multivariate data. The mineral composition of the Huai Bo Khong Formation was petrographically identified in Moonpa and Motanated [18]. Ca intensities are contributed by plagioclase and carbonate cement. Fe intensities are from clinopyroxene and opaque minerals. K-bearing minerals are orthoclase and microcline, and Si-bearing minerals are quartz, feldspar, and clinopyroxene.

Figure 4

Examples of µXRF elemental distribution maps of NP09-1 sample. The scanning area is 5.12 cm × 5.12 cm. Individual element maps in (a)–(d) are Si, K, Fe, and Ca, respectively. Black zones indicate that element was not detected.

Figure 5

Vertical intensity profiles of Figure 2c. Horizontal and vertical axes are Fe counts and vertical height of the sample (cm), respectively.

4.2 PCA

PCA is applied to the vertical intensity data of the 11 elements, excluding Ca, which are detectable under the µXRF scans from 16 sandstone samples. Ca is not included in the PCA because Ca-bearing minerals are often found in fractures. The samples are scanned separately; their elemental intensity profiles and sedimentary structures are then analyzed. Thus, the input data for PCA are in four conditions – unnormalized and normalized against Al, K, and Si – in order to assess the efficiency of normalization when selecting proxies for sedimentary structure observation. The three elements are used for normalization because they are found in the common minerals, quartz and feldspar, of the sandstone samples. Table 1 summarizes PCs and their cumulative variances that meet the Kaiser criterion, where eigenvalue is greater than 1.00. The scree plot of eigenvalues of each PC is shown in Figure 6. In all samples, the line begins to flatten out at PC2; therefore, PC1 should be retained. Al, Fe, K, and Si are variables with high factor loadings. Spatial Al, Fe, K, and Si distribution maps from µXRF scans of NP05-3, NP08-1, NP09-1, and NP12-1 are illustrated in Figure 7. These samples have different sedimentary structures including cross-lamination (NP05-3), massive (NP08-1 and NP12-1), and massive with 2 cm thick mudstone cap (NP09-1). The sedimentary structures are clearly visible in the Fe distribution maps (Figure 7b, f, j, and n).

Figure 6

An example of the scree plot of sample NP01-1. The slope of the line starts to flatten out at PC2.

Figure 7

µXRF elemental distribution maps of NP05-3, NP08-1, NP09-1, and NP12-1 samples. The scanning area is 5.12 cm × 5.12 cm for all sample except for that of NP08-1 which is 5.12 cm × 3.76 cm. Each row contains elemental distribution maps including Al, Fe, K, and Si of each sample. (a) NP05-3 Al, (b) NP05-3 Fe, (c) NP05-3 K, (d) NP05-3 Si, (e) NP08-1 Al, (f) NP08-1 Fe, (g) NP08-1 K, (h) NP08-1 Si, (i) NP09-1 Al, (j) NP09-1 Fe, (k) NP09-1 K, and (l) NP09-1 Si.

When comparing the first PCs in Table 1, PC1 of K-normalized data accounts for the highest variation in the dataset. When the factor loadings of PC 1 from K-normalized elemental data are compared with each other (Table 2), the common variables with high factor loadings among the 16 sandstone samples are Al/K, Fe/K and Si/K.

5.1 Evaluation of the discriminant analysis

The objective of this research is to apply statistical technique of discriminant analysis to determine element intensity or element intensity ratios that can be used as proxies to detect various sedimentary structures in sandstone. The sandstone samples are from the same formation and derived from the same provenance; thus, diagenesis, changes in grain size, and hydraulic sorting can be the factors controlling the geochemical variation in clastic rocks [24,25]. This research evaluates the geochemical variation through spatial elemental maps. From the spatial Al, Fe, K, and Si distribution maps obtained from µXRF scans of NP05-3, NP08-1, NP09-1, and NP12-1 (Figure 7), the sedimentary structures are clearly visible in all Fe distribution maps (Figure 7b, f, j, and n) and in some Si distribution maps (Figure 7l and p). From the PCA, common elemental ratios that are high loading factors across the 16 samples are Fe/K and Si/K intensity ratios (Table 2). Thus, the vertical Fe, Fe/K, Si, and Si/K of NP05-3, NP08-1, NP09-1, and NP12-1 are plotted (Figures 8–11) in order to determine the relationship between the element intensity or element intensity ratios and sedimentary structures.

Figure 8

Vertical Fe count of samples NP05-3, NP08-1, NP09-1, and NP12-1 are plotted in solid black, solid gray, dashed, and dotted lines, respectively.

Figure 9

Fe/K vertical intensity profiles of samples NP05-3, NP08-1, NP09-1, and NP12-1 are plotted in solid black, solid gray, dashed, and dotted lines, respectively.

Figure 10

Vertical Si count of samples NP05-3, NP08-1, NP09-1, and NP12-1 are plotted in solid black, solid gray, dashed, and dotted lines, respectively.

Figure 11

Si/K vertical intensity profiles of samples NP05-3, NP08-1, NP09-1, and NP12-1 are plotted in solid black, solid gray, dashed, and dotted lines, respectively.

Samples NP08-1, NP09-1, and NP12-1 are similar in that both NP08-1 and NP12-1 are massive sandstones and the bottom portion of NP09-1 is massive. However, the ranges and trends of Fe count of each sample are different from one another, even when they have similar sedimentary structures (Figure 8). On the other hand, the trends of Fe/K profiles of NP08-1 and NP12-1, gray and dotted lines in Figure 9, and that of the massive sandstone portion (lower part) of NP09-1, dashed line in Figure 9, are similar. The ranges and trends of Si count of massive sandstone samples NP08-1 and NP12-1 are similar to one another (Figure 10). The trend of Si profile of the massive sandstone with mudstone cap is different from those of the other samples, in that the Si count of mudstone cap is significantly lower that that of the bottom massive sandstone. When the magnitude is ignored, the trends of Si profiles of cross-bedded sandstone (NP05-3) and massive sandstone (NP08-1) are similar. Additionally, the trends of Si/K profiles of massive sandstones NP08-1 and NP12-1 and the massive sandstone portion of NP09-1 are overlying each other.

This study also validates the use of statistical technique of discriminant analysis in determining the element intensity or element intensity ratios that can be used as proxies to detect sedimentary structures of the other sandstone samples from the same formation in the same study area. The vertical elemental intensity data from µXRF scans of sandstone samples NP11-1, NP11-2, NP13-1, and NP13-2 were excluded from the PCA; they were later used as testing datasets. From the spatial Al, Fe, K, and Si distribution maps of NP11-1, NP11-2, NP13-1, and NP13-2, the sedimentary structures are clearly visible in all Fe distribution maps (Figure 12b, f, j, and n). However, the horizontal laminae observed in NP11-1 and NP11-2 are not clearly visible in the spatial Si distributon maps (Figure 12d and  h). Figure 13 illustrates the spatial Fe distribution maps and vertical Fe and Fe/K profiles of NP11-1, NP11-2, NP13-1, and NP13-2. The trends of Fe and Fe/K profiles of NP11-1 and NP11-2, both massive sandstone with horizontal lamination, are similar (Figure 13b and d). In these two samples, the normalization results in a more pronounced variation of Fe/K between the massive and the horizontal lamination parts. Additionally, in the massive sandstone samples (NP13-1 and NP13-2), changes in hydraulic sorting that does not result in changes in sedimentary structures are reflected in both Fe and Fe/K profiles. The controls on geochemical variation are explained in the next subsection.

Figure 12

µXRF elemental distribution maps of NP11-1, NP11-2, NP13-1, and NP13-2 samples. The scanning area is 5.12 cm × 5.12 cm for all sample. Each row displays Al, Fe, K, and Si elemental distribution maps of each sample. (a) NP11-2 Al, (b) NP11-2 Fe, (c) NP11-2 K, (d) NP11-2 Si, (e) NP13-1 Al, (f) NP13-1 Fe, (g) NP13-1 K, (h) NP13-1 Si, (i) NP13-2 Al, (j) NP13-2 Fe, (k) NP13-2 K, and (l) NP13-2 Si.

Figure 13

(a), (c), (e), and (g) are Fe µXRF elemental distribution maps of samples NP11-1, NP11-2, NP13-1, and NP13-2, respectively. (b), (d), (f), and (h) are Fe and Fe/K vertical intensity profiles of samples NP11-1, NP11-2, NP13-1, and NP13-2, respectively. Fe intensity profile is plotted in solid black line; Fe/K intensity profile is plotted in dotted lines.

The research results show that PCA is able to identify the elements and elemental ratios with high loadings that explain the multivariate data; therefore, they should be retained and applied as proxies in the assessment of hydraulic sorting in sandstones from the same provenance. The sedimentary structures are shown in spatial elemental distribution maps of elements and elemental ratios with high loadings. The results from this research is in agreement with that from Pe-Piper et al. [26] where a few key elements identified from PCA are useful for discrimination technique. Additionally, normalization, which in this study is K-normalized, makes the distinction between different intervals become apparent [27] (Figure 13).

5.2 Controls of elemental variation

From the Table 2 and Figures 10 and 11 of Moonpa and Motanated [18], the majority of grains in these turbiditic sandstones is mafic volcanic rocks such as basalt. Fe-rich minerals belonging to amphibole and pyroxene mineral groups are also common. The specific gravity of these mineral groups is relatively higher than that of quartz, plagioclase, orthoclase, microcline, and albite. Because the sandstone samples are derived from the same provenance, diagenesis, changes in grain size, and hydraulic sorting can be the factors controlling the geochemical variation in clastic rocks [24,25]. The sandstones deposited during the Triassic and diagenesis processes such as cementation, compaction, recrystallization, or mineral replacement or dissolution could have affected the chemical composition and texture of the samples. From the petrographic study of the Nam Pat sandstones in Moonpa and Motanated [18], only a few parts of mafic minerals underwent a diagenesis process and altered to epidote, and few alteration minerals such as sericite, kaolinite, and chlorite were observed. Therefore, diagenesis is less likely to be the variable affecting the geochemical variation between the samples because they are from the same locality that have been minimally and similarly diagenesized.

Each bedform develops under certain flow conditions of mean flow velocity and median sediment size [28]. Banerjee [29] and Summer et al. [30] experimentally studied sedimentary structures deposited from decelerating turbidity currents. When the flow velocity is decreasing, the bedform changes from graded bed to planar lamination, cross-lamination, and suspension blanket, respectively. Different sedimentary structures in samples NP09-1, NP11-1, and NP11-2 are clearly observed on the µXRF elemental distribution maps (Figures 7 and 12). Fe and Fe/K intensities of the horizontal lamination portions of samples NP11-1 and NP11-2 are greater than those of the massive portions (Figure 13a–d). In other words, Fe-bearing minerals are concentrated in the horizontal lamination portions when the turbidity currents are decelerating. Thus, changes in hydraulic conditions of the flow that produce different bedform are observed on the µXRF elemental distribution maps. Additionally, hydraulic evolution of the turbidity currents is detected through the µXRF analysis even when there is no change in the resultant deposits. For example, in the massive sandstone samples NP12-1, NP13-1, and NP13-2, changes in hydraulic sorting that do not produce changes in the resultant deposits are reflected in both Fe and Fe/K profiles. The semi-quantitative changes in Fe and Fe/K profiles observed in the massive sandstones of this study are similar to the results from the study of the variation in abundance and grain size of hydraulically fine heavy (zircon) and hydraulically coarse light (feldspar) minerals of the T_a division of Bouma sequence where settling velocity of grains of contransting density changes with sediment concentration [31]. The elemental intensity data from µXRF scans are able to show the evidence of hydraulic sorting even when there is no change in the resultant deposit. In order to determine whether the hydraulic sorting results in variation in abundance, grain size, or both, petrographic study should be investigated.

This research results show that PCA is a useful technique in determining the element intensity ratios that are diagnostic of elemental variation, which is visually observed in sedimentary structures. The methodology demonstrated here is tested to be effective on turbiditic sandstone samples derived from the same provenance. The methodology here could be applied to in situ µXRF scans where the observation of sedimentary structures or grain size measurement is inaccessible such as in boreholes. However, variation in element or elemental ratio could not be applied for turbiditic sandstone correlation across multiple samples from the same formation or provenance. The XRF core scanning results have to be calibrated with those obtained from quantitative geochemical analysis, such as inductively coupled plasma-atomic emission spectroscopy (ICP-AES), in order to quantify the elemental concentration [2,6]. Future research directions should systematically investigate sandstones with different depositional processes or those derived from different types of parent materials.