DATA
Correction of X-ray tube aging effects and quantitative estimation of elemental composition using the ITRAX XRF core scanner: A case study of Japan Sea sediments
2025 Volume 59 Issue 4 Pages 129-143
Details
2025 Volume 59 Issue 4 Pages 129-143
The XRF core scanner, which facilitates non-destructive, high-speed, and high spatial resolution measurements, has recently been extensively used to measure element variability in sediments. Although identifying the precision of the measurement and concentration of the elements in the sample is essential to elucidating the variations recorded in the sediments, the element peak area count ratios are most commonly used to report the results of XRF core scanners in many previous studies, and element concentration is not discussed. This is mainly because quantitative calibration of element concentration necessitates numerous additional analyses of discrete samples. Because XRF core scanner measurements are often conducted before discrete sample analysis, if approximate element concentrations in the sediment can be estimated from XRF core scanner data just after measurement, sampling intervals should be determined for further analysis.
In this study, we developed a new database on the aging of X-ray tubes, and propose a method for estimating element concentrations in the wet sediments from XRF core scanner measurement without any additional measurement. The ITRAX XRF core scanner at Kochi University, the first ITRAX installed in Japan in 2014, was used in this study. Mo X-ray tubes were used at 30 kV, 55 mA with XRF exposure time of 10 s or 32 s. First, the aging effect of the X-ray tube was monitored and its correction method was proposed. Subsequently, the precision of the ITRAX XRF measurements was evaluated and the relationship between the element peak area count by ITRAX and element concentration was analyzed. Finally, we propose a new method for applying these databases constructed in this study to any results measured by ITRAX XRF core scanner at Kochi University. Using this method, anyone can estimate approximate element composition of the sediment from their XRF core scanner data obtained at Kochi University without any additional measurement. This method even enables estimation of element concentrations of sediments just after the ITRAX measurement.
In recent decades, X-ray fluorescence (XRF) core scanners have been extensively used for the analysis of sediment cores in numerous academic fields, including paleoceanographic and sedimentological studies (Rothwell and Croudace, 2015a) because they facilitate nondestructive and high-resolution measurements with high-speeds. The ITRAX XRF core scanner is one of the most popular and widely used XRF core scanners (Croudace et al., 2006; Rothwell and Croudace, 2015a). It is manufactured and sold by COX Analytical Systems (Croudace et al., 2006). Optical images, radiographic images, and XRF spectra can be simultaneously obtained along the same measurement line. The flat-beam technology developed for the ITRAX core scanner enables us to obtain high-resolution radiographs and XRF measurements with the highest vertical resolution of 0.2 mm, which corresponds to the X-ray beam width (Croudace et al., 2006). Long (up to 180 cm) and thick (up to 12 cm diameter) samples, such as half-round cores and U-channel samples, can be directly measured using the ITRAX core scanner (Croudace et al., 2006).
XRF core scanner data of wet sediment cores are often expressed as element peak area counts derived from XRF spectra (Rothwell and Croudace, 2015b). Calibrating count data to element concentration data in sediments is not straightforward because ITRAX count data reflect element concentrations and are influenced by variations in particle size, porosity, and water content (Croudace et al., 2006; Kido et al., 2006; Weltje and Tjallingii, 2008; Weltje et al., 2015; Chen et al., 2016; Katsuta et al., 2019). In addition, because the ITRAX uses narrow beam width to obtain high spatial resolution, the excitation volume of the XRF measurement is small, resulting in smaller counts in the XRF spectra. Therefore, the advantages of the ITRAX such as being non-destructive, high-speed, and high-resolution are counterbalanced by disadvantages such as the large measurement error of XRF data compared with the conventional wavelength-dispersive X-ray fluorescence (WD-XRF) method.
Several previous studies have focused on the quality improvement of XRF data obtained by ITRAX and other XRF core scanners (Weltje and Tjallingii, 2008; Weltje et al., 2015; Ohlendorf et al., 2015; Chen et al., 2016; Gregory et al., 2019; Dunlea et al., 2020; Mondal et al., 2021). Weltje and Tjallingii (2008) proposed a log-ratio calibration model. Log-ratio calibration is advantageous over the linear calibration because theoretically, the log-ratio of peak area counts has a linear relationship with that of concentrations, even when matrix and specimen effects (measurement geometry and specimen homogeneity) exist (Weltje and Tjallingii, 2008). By further developing the log-ratio calibration method, Weltje et al. (2015) proposed a multivariate log-ratio calibration (MLC) model, which allows for the unbiased prediction of geochemical compositions. The usefulness of the MLC model has been confirmed by other studies, and further development of the method has been proposed (Chen et al., 2016; Gregory et al., 2019). Although their method improved the quantitativeness of the results, a number of replicate XRF scans and a large number of measurements of discrete samples taken from the scanned cores, including replicate measurements, were required (Weltje and Tjallingii, 2008; Weltje et al., 2015). Therefore, to avoid a time-consuming procedure to obtain quantitative calibration equation, numerous studies have used peak area counts as a semi-quantitative method (e.g., Kuroda et al., 2021; Hsiung et al., 2021). In addition, Dunlea et al. (2020) proposed the use of log-ratio calibration, comparing XRF scanning results among seven different XRF core scanner machines. Mondal et al. (2021) compared several existing correction methods and concluded that dividing element peak area counts by the coherent/incoherent X-ray scatter ratio was the most effective approach, using the same XRF core scanner machine as that used in this study.
The aging of the X-ray tube caused a reduction in the X-ray intensity emitted from the X-ray tube, and thus significantly affected the quality of data interpretation. Ohlendorf et al. (2015) showed that the aging effect of the X-ray tube causes a significant decrease in the intensity of XRF spectra and peak area counts of each element calculated from the XRF spectra within 1200 h of Mo X-ray tube lifetime. They pointed out that the aging effect might lead to a misinterpretation of the down-core variation of the element concentration, particularly, when scanning cores under varying Mo X-ray tube ages. However, a corrective method for this effect has not been developed (Ohlendorf et al., 2015), as studies using long sediment cores remain relatively limited compared to the larger number of studies based on shorter cores (Rothwell and Croudace, 2015a).
In this study, we attempt to establish a new quantitative method for the measurement of major and minor elements using an ITRAX XRF core scanner installed at Kochi University. First, the aging effect of the X-ray tube was monitored and a correction method was established. Subsequently, the precision of the ITRAX XRF measurements was evaluated and the relationship between the element peak area count and element concentration was analyzed. Finally, we propose a new method for applying the database constructed by the measurement described above to calculate element concentration for any results of the core samples previously measured by ITRAX XRF core scanner at Kochi University without any additional measurement. This method even enables estimation of the element concentrations of sediments just after the ITRAX measurement.
To monitor the aging effect and evaluate measurement precision of the ITRAX at Kochi University, standard samples made from Japan Sea sediment (Dunlea et al., 2020) were measured. The Japan Sea sediment standards were prepared from the Quaternary Japan Sea sediments of four characteristic lithofacies, and named “light layer,” “dark layer,” “calcareous,” and “siliceous.” Each sample was dried, powdered, homogenized, and pressed into pellets (Dunlea et al., 2020).
A reference glass standard provided by COX Analytical Systems as a standard ITRAX accessory to monitor the ITRAX measurement conditions was also used for monitoring the aging effect of X-ray tubes.
Seven commercially available sediment standard reference materials: JMS-1, JMS-2, JLk-1, JSd-1, JSd-3, MESS-4, and JG-1 (Imai et al., 1995, 1996; Terashima et al., 2002; Supplementary Table 1) were measured with ITRAX to estimate the element concentrations from peak area counts calculated from XRF spectra (see Section “Estimation of the Element Concentration” for details). These standard reference materials were provided as fine powders and pressed into pellets before the measurements. A press power of 20 t was used in this study.
The Japan Sea sediments drilled at Site U1424 by the Integrated Ocean Drilling Program (IODP) Exp. 346 (Tada et al., 2015) were also used to compare the element counts from ITRAX measurement with element concentrations analyzed by WD-XRF. The 64 discrete samples of Site U1424 with various lithologies, densities, water contents, and total organic carbon content (Tada et al., 2015; Seki et al., 2019) were obtained from the Quaternary intervals measured by ITRAX (Seki et al., 2019). The fused disc was made from the discrete samples, and the element concentrations were measured using the conventional XRF method in Kochi University.
ITRAX measurement settingsIn this study, an ITRAX installed at the Marine Core Research Institute (MaCRI), Kochi University, Japan, which was installed in 2014 as the first ITRAX in Japan, was used. We measured each sample using the standard measurement procedure of ITRAX (Croudace et al., 2006; Löwemark and Itrax operators, 2019). The plastic wrap provided by COX Analytical Systems as a standard ITRAX accessory was not used for dry pressed samples. The automatically calculated peak area counts from the XRF spectrum by Q-spec software were used in this study. We selected 27 elements (Al, Si, P, S, Cl, Ar, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Ga, As, Br, Rb, Sr, Y, Zr, Ba, Ta, W, and Pb) that were detectable in the XRF spectra of the Japan Sea sediments (Supplementary Fig. 1).
Pressed pellets of the Japan Sea sediment standard samples (“light layer,” “dark layer,” “calcareous,” and “siliceous”) with a diameter of 2 cm were scanned by the ITRAX under the conditions of 0.2-mm intervals, 15 lines, and 10 s and 32 s XRF exposure time (Supplementary Fig. 2). The measurements were repeated five times. The commercially available sediment standard reference materials (Supplementary Table 1) were measured on the same settings as those of the Japan Sea sediment standard samples, with 10 s XRF exposure time. The archive halves of Site U1424 were measured by ITRAX with 2-mm steps and 10 s XRF exposure time (Seki et al., 2019). The commercially available sediment standard reference materials and Site U1424 cores were measured shortly after the installation of second Mo X-ray tube to obtain the XRF spectra without the X-ray tube aging effect (see Section “Aging Effect of the X-ray Tube” for details). Mo X-ray tubes were used at 30 kV and 55 mA for all XRF measurements except reference glass standard measurement. A reference glass standard was measured at one point with 30 kV, 30 mA, and 100 s exposure time.
The intensity of the XRF spectra and peak area of each element were basically proportional to the concentration of each element. However, the intensity of the XRF spectra also depends on the intensity of the X-ray beam generated by the X-ray tube. The X-ray beam intensity of the X-ray tube gradually decreases, and the peak area counts of each element decreases as the X-ray tube ages (Nakai, 2005; Ohrendorf et al., 2015). This aging effect becomes significant when total duration of the measurement is long, for example, when many sediment cores covering a long sedimentary sequence are analyzed or the measurement of one sedimentary sequence is segmented into several intervals. However, little attention has been paid to the aging effect of X-ray tubes in previous studies, probably because many studies finish their measurement within a month and the X-ray tube aging effect is not significant. The other reason may be that the count ratio or log ratio is used in numerous studies because the X-ray tube aging effect is believed to have been canceled. Ohlendorf et al. (2015) reported the aging effect of an X-ray tube in the case of an ITRAX core scanner. To monitor the aging effect, they measured a reference glass standard with known composition after scanning of each core. They reported that relatively heavy elements (Sr and Zr) are more affected by the aging effect than relatively light elements (Si, Ca, and Fe), that is, the decline in counts caused by the X-ray tube aging effect was greater for heavy elements than for lighter ones. They also reported that Rayleigh scattering (coh = coherent scatter at the same energy as that of the tube anode radiation) and Compton scattering (inc = incoherent scatter at a lower energy than that of the tube anode radiation) decreased by ratios similar to those of heavier elements (Ohrendorf et al., 2015). Based on these observations, they proposed that the aging effect on heavier elements (Sr, Zr) can be eliminated by dividing the element peak area by the coh or inc peak area. For lighter elements (Si, Ca, and Fe), they proposed the use of element ratios with comparable atomic numbers for which the aging effect is similar (e.g., Si/Ti ratio, sometimes interpreted as a proxy of the biogenic silica content; Rothwell and Croudace, 2015b) to correct the aging effect (Ohrendorf et al., 2015).
In this study, we propose a method for correcting the aging effect. We also report the aging effects on all the elements measured in this study because the aging effects were evaluated for only 5 elements: Si, Ca, Fe, Sr, and Zr in previous studies (Ohrendorf et al., 2015), which are not sufficient for establishing quantitative analysis of sediment composition methodology for the aging effect.
Monitoring the aging effectTwo Mo X-ray tubes were used to monitor the aging effect. The first Mo X-ray tube was used for 5663 h. The second Mo X-ray tube was still in use when we finished the measurements of the Japan Sea sediment standard samples in this study (Supplementary Table 2).
The results of pressed pellets of the “light layer” standard samples measured with 32 s XRF exposure time measured for time intervals ranging from a week to a few months was used in this section (Fig. 1, Supplementary Fig. 3, and Supplementary Table 2), because all results for the four Japan Sea standard samples (“light layer,” “dark layer,” “calcareous,” and “siliceous”) showed a similar decreasing trend over time, and the “light-layer” standards were measured most frequently (n = 19). The measurements were repeated five times. All peak area counts of each standard were averaged, and the standard deviations were calculated. The peak area counts of each element and the standard deviation were plotted against the total operation time of the X-ray tubes. In addition to inc and coh, the effects of aging on the 27 elements were monitored (Fig. 1 and Supplementary Fig. 3). All 27 element count data are presented in Supplementary Table 2.
Aging effect monitored using “light layer” standard sample (see text for detail). (a) Si, (b) K, (c) Ti, (d) Fe, (e) Br, and (f) Sr. Diamond: measured by first Mo X-ray tube, used for total 5663 hours. Triangle: measured by second Mo X-ray tube.
Although most counts of the elements decreased with operation time, W, Ta, Cr, and Ar show a trend that differs from that of the other elements with increasing X-ray tube age. An increase in W count was observed as the X-ray tube operating time increased (Supplementary Fig. 3t). Because the filament inside a Mo X-ray tube is composed of W, W is gradually evaporated and precipitated inside the X-ray tube, and the W peak area in the XRF spectra increases with aging of the X-ray tube (Nakai, 2005; Ohrendorf et al., 2015). Therefore, the increase in W count is attributed to another effect of tube aging, showing that the aging effect is not simply represented by the decrease in X-ray intensity. “Ta” counts showed a trend similar to that of W count (Supplementary Figs. 3s and 3t). This is possibly because the “Ta” peak overlaps with the Compton scattering peak of W. Therefore, the “Ta” count might be contaminated with the Compton scattering peak count of W and increased owing to the increase in W count (Supplementary Figs. 3s and 3t). Moreover, Cr count originating from the X-ray tube is sometimes observed in X-ray spectra because Cr is contained in the filaments as an impurity (Nakai, 2005). In the case of the first Mo X-ray tube, the Cr count decreased with increasing tube age similar to the counts of other elements reflecting the decrease in the X-ray intensity owing to the aging effect of the X-ray tube (Supplementary Fig. 3h). In contrast, the Cr count increased with increasing tube age in case of the second Mo X-ray tube (Supplementary Fig. 3h). This indicates that the Cr peak increase with tube aging is observed only in the second X-ray tube owing to the Cr contained in the X-ray tube as impurity (Supplementary Fig. 3h). The Ar count increased with tube aging only in the second Mo X-ray tube (Supplementary Fig. 3e). However, the Ar count does not necessarily reflect elemental variability in the sediments but mainly reflects Ar in the air between the sample surface and XRF detector. Therefore, we did not use the Ar count to interpret the element variability in the sediment cores. In summary, the W, Ta, Cr, and Ar counts in this study were not used owing to the reasons above. The counts of the other 23 elements, coh, and inc decreased over time because of the decreasing X-ray intensity caused by X-ray tube aging, particularly with respect to lighter elements (Fig. 1 and Supplementary Fig. 3).
Difference in aging effect for each elementThe measurement results of the first Mo X-ray tube were used to evaluate the aging effect in the following discussion because the second Mo X-ray tube was relatively young and the aging effect was insignificant in certain elements (Fig. 1, Supplementary Fig. 3, and Supplementary Table 2). For example, Br and Sr exhibit significant reductions in the second Mo X-ray tube within the operational time of 0–3000 h, while Si, K, Ti, and Fe do not show similar decreasing trends (Fig. 1). Although the second Mo X-ray tube also seems to exhibit the aging effect on certain elements, we use the first Mo X-ray tube, which has longer records, in the following discussion. Figure 2 shows the percentage decrease in the counts of 16 elements (Al, Si, S, Cl, K, Ca, Ti, Mn, Fe, Ni, Zn, Br, Rb, Sr, Zr, and Ba; the elements with a standard deviation of <10%) with the aging of the first X-ray tube. The percentage decrease in the count of each element was calculated using the counts at the end of the X-ray tube lifetime (5636 h) divided by the counts near the time of the tube installation (661 h). Figure 2a shows the percentage decrease in counts during the lifetime versus the atomic number for the 16 elements. The overall trend, particularly for relatively light elements (atomic number lower than 35), is similar to the trend previously reported by Ohrendorf et al. (2015); however, the percentages in this study show relatively high values (Fig. 2a). This may be owed to the difference in the X-ray tube lifetime, the tube voltage, or individual difference of X-ray tube.
Percentage of the peak area counts at the end of first Mo X-ray tube lifetime (5636 h) divided by counts at the early stage of lifetime (661 h) versus (a) atomic number, and (b) energy of X-ray fluorescence. Closed circles: elements either K or L lines are observed. Open circles: elements where both K and L lines of fluorescence are observed (Rb, Sr, Zr). Diamonds: data obtained from reference glass standard measurement (Ohlendorf et al., 2015).
For relatively heavy elements (atomic number exceeding 35), the trend of the percentage decrease in counts differs from that of lighter elements (Fig. 2a). The reason for this difference comes from variation of fluorescence lines. For lighter elements (atomic numbers below 35), only K lines were observed in the XRF spectra, whereas L lines were observed for heavier elements (Supplementary Fig. 1). For Rb, Sr, and Zr (atomic numbers 37, 38, and 40), both K and L lines were observed in the XRF spectra. Only the L line was observed for Ba (atomic number: 56).
In Fig. 2b, the percentages of the decrease in counts were plotted against the energy of the main peak of each element in the XRF spectra (Fig. 2b). Because the percentages of the decrease in counts for Rb, Sr, and Zr are affected by both the K and L lines, the calculated percentages of the decrease in the counts of these 3 elements were plotted for two energies (Fig. 2b).
Figure 2b shows that the percentages of the decrease in counts decreased as element energy increased where only the K or L lines were observed. Notably, Ba (4.5 keV), where only the L line was observed, yielded the same percentage as Ti (4.5 keV), where only the K line was observed. Therefore, the difference in line has no effect on the relationship between the aging effect and fluorescence X-ray energy.
Based on the trend described above, the percentage decrease in counts of Rb, Sr, and Zr, where both K-line and L-line were observed, was estimated to be ~50% for the K-line energy and ~70% for the L-line energy (Fig. 2b). Therefore, the reduction percentages for these 3 elements (~60%) show the mean values of the K line (~50%) and L line (~70%; Fig. 2b). Based on these results, we conclude that the difference in the percentages of the decrease in counts with the aging of Mo X-ray tube depends on the energy in the XRF spectra and not on the atomic number, as suggested in a previous study (Ohlendorf et al., 2015).
Correction of the aging effectThe element/coh ratios were calculated and compared with the X-ray tube operation time (Fig. 3, Supplementary Fig. 4) because the normalization with coh is recommended by previous study (Ohlendorf et al., 2015) as the correction method of the X-ray tube aging effect. Although the inc is also proposed as a divider in previous research (Ohlendorf et al., 2015), it was not used in this study because inc is affected by other factors, such as the water content, organic matter content, and porosity of the sediment (Ohrendolf et al., 2015). Therefore, although division by inc works well for dried sediment, its effectiveness is lower for wet sediment compared to division with coh. Figure 3 shows that the aging effect on Br, Rb, and Sr was diminished by normalization with the coh count, resulting in a consistent count ratio over 8000 h within the analytical error. However, the aging effect on other elements remains (Fig. 3, Supplementary Fig. 4). This is because the percentage of the decrease in the counts with the X-ray tube aging for the elements other than Br, Rb, and Sr differs from that of coh (55% ± 1%; Fig. 2). Therefore, the Br/coh, Rb/coh, and Sr/coh ratios are applicable for the correction of the aging effects; however, this normalization is not applicable to other elements (Fig. 3, Supplementary Fig. 4).
Peak area counts of each element divided by the counts of coh using “light layer” standard sample. (a) Si/coh, (b) K/coh, (c) Ti/coh, (d) Br/coh, (e) Rb/coh, (f) Sr/coh. Diamond: measured by first Mo X-ray tube, used for total 5663 hours. Triangle: measured by second Mo X-ray tube.
The aging effect can also be corrected by monitoring the reduction rate of an element’s count by the X-ray tube aging effect using standard samples. Based on the results shown in Fig. 1, Supplementary Fig. 3, and Supplementary Table 2, the decrease rate of an element’s count owing to the aging effect at each tube operating time can be calculated for each element. The effect of aging on the measured element count for each sample was corrected using the calculated results (Seki, 2017).
Supplementary Fig. 5 and Supplementary Table 3 show the results of reference glass standard measured at ITRAX in Kochi University. Because the XRF spectra of reference glass standard is routinely measured to check the XRF performance of the ITRAX (Löwemark and Itrax operators, 2019), it provides the most frequent and long record about the X-ray tube aging effect. Therefore, we recommend ITRAX users to calculate the X-ray tube aging effect for their measurement by using the monitoring database of the reference glass standard (Supplementary Fig. 5 and Supplementary Table 3).
To evaluate the analytical error of the ITRAX measurement, repeated measurement results of four Japan Sea standard samples (“light layer,” “dark layer,” “calcareous,” and “siliceous”), measured by 10 s XRF exposure time were used. An XRF exposure time of 10 s was selected because it is the shortest exposure time frequently used for the measurement of sediments (e.g. Löwemark and Itrax operators, 2019; Seki et al., 2019). In addition, a 10-s exposure time is expected to be highly practical for minimizing the measurement duration. To evaluate the precision of the repeated ITRAX measurements that were not affected by sample inhomogeneity, the results of five repeated measurements were averaged for each of the measured 15 areas within a standard sample (Supplementary Fig. 2). The results of five-times repeated measurement on the four Japan Sea standard samples on 11 different dates in 2016 (Supplementary Table 4) were used.
Figure 4, Supplementary Fig. 6, and Supplementary Table 4 show the results of repeated measurements for the four standard samples. The averages (horizontal axis) and standard deviations (vertical axis) of the peak area counts of 165 measurements (measurements of 15 areas were repeated 11 times in different days) were plotted for each standard sample (Fig. 4 and Supplementary Fig. 6). The total average of the “peak area average (horizontal axis)” and “standard deviation (vertical axis)” of 165 measurements for each standard sample are also shown in Fig. 4 and Supplementary Fig. 6.
Standard deviations and average peak area counts of five times repeated measurements of four Japan Sea standard samples. Horizontal axis is the average of peak area counts from measurements repeated five times. Average of each standard sample is also shown (see text for details). STD: standard deviation. Ave.: Average. Theoretical standard deviation (σ) line is also shown (see text for detail). (a) Si, (b) Cl, (c) K, (d) Ti, (e) Fe, (f) Br.
Theoretically, the standard deviation of the peak area counts of elements obtained by XRF measurements can be described as follows (Croudace et al., 2006; Nakai, 2005):
standard deviation = √N,
where N reflects the measured peak area counts.
Despite suggesting the use of this theoretical value as measurement precision for the ITRAX, Croudace et al. (2006) proposed other potential error sources such as the surface conditions of the samples. XRF measurements are reliable if the measured standard deviation is roughly below 2√N, because a higher standard deviation indicates lower S/N ratios (Nakai, 2005). To verify the results obtained in this study, theoretical lines [standard deviation (vertical axis) = √N, 2√N, and 3√N; N = peak area count (horizontal axis)] are shown in Fig. 4 and Supplementary Fig. 6 for comparison. The averages of most elements are plotted below the 2√N line. However, for elements with small peak area counts (<500 counts; P, V, Cu, Ga, As, Y, Ba, Ta, and Pb), the standard deviations are larger than 2√N (Supplementary Fig. 6). An exception is Zr, which has larger peak area counts than the above-mentioned 9 elements, but its standard deviation exceeds 2√N (Supplementary Fig. 6q). Additional studies are required to determine the precision of these 10 elements when larger counts are obtained in the measurements. For Si (Fig. 4a), Cl (Fig. 4b), and K (Fig. 4c), the average standard deviation is as small as the theoretical value √N. Based on the results for the five repeated measurements shown in Fig. 4 and Supplementary Fig. 6, the standard deviations differ according to element even when the peak area counts are the same. Therefore, in this study, we estimated the standard deviation of each element using the following equation:
σ = αelement√N,
where σ is the standard deviation of the peak area count, αelement is a factor specific to each element, and N reflects the measured peak area counts.
Parameter αelement was calculated for each element based on the following procedure. First, the average count and standard deviation were calculated for five repeated measurements in each area (Supplementary Table 4). Subsequently, the standard deviation was divided by the square root of the average count (= σ'/√Nave) for each of the five repeated measurements (165 measurements in total). The σ'/√Nave for all 165 measurements were then averaged for each element to calculate αelement (Table 1). Parameter σ for each element calculated based on αelement is shown in Fig. 4 and Supplementary Fig. 6. The σ value can be used as the standard deviation of the peak area count obtained from the sample measurements (e.g. Seki et al., 2019). These σ values for each element represent potential analytical uncertainties. Therefore, these uncertainties can be applied to other ITRAX data if they were obtained with the same measurement settings as in this study, because the σ value is not affected by X-ray tube aging.
Calculated precision of XRF measurement using ITRAX XRF core scanner
| αelement | average peak area counts maximum | average peak area counts minimum | |
|---|---|---|---|
| Al | 1.6 | 226 | 79 |
| Si | 1 | 1852 | 619 |
| P*1 | 2.5 | 51 | 6 |
| S | 1.3 | 591 | 136 |
| Cl | 1 | 1735 | 620 |
| Ar | 1.3 | 346 | 145 |
| K | 1 | 6099 | 3042 |
| Ca | 1.1 | 131579 | 1243 |
| Ti | 1.2 | 4254 | 898 |
| V | 2.7 | 283 | 4 |
| Cr | 1.9 | 441 | 15 |
| Mn | 1.4 | 1774 | 429 |
| Fe | 1.2 | 205273 | 37191 |
| Ni | 2 | 586 | 163 |
| Cu | 2.9 | 511 | 151 |
| Zn | 1.8 | 1618 | 474 |
| Ga | 3.7 | 488 | 31 |
| As*2 | 7.2 | 378 | 6 |
| Br | 2.1 | 3725 | 414 |
| Rb | 3.2 | 2847 | 333 |
| Sr | 1.7 | 16708 | 1129 |
| Y | 4.8 | 492 | 48 |
| Zr | 2.9 | 1785 | 451 |
| Ba | 2.3 | 392 | 30 |
| Ta*1 | 2.4 | 605 | 35 |
| W | 1.5 | 1783 | 49 |
| Pb*3 | 6.2 | 286 | 13 |
| inc | 2 | 47001 | 13351 |
| coh | 1.6 | 13809 | 5772 |
For 4 elements, αelement was calculated from measurement results of (*1) light-layer, dark-layer, and siliceous standard samples, (*2) dark-layer and siliceous standard samples, (*3) light-layer and calcareous standard samples, because no counts were detected during some measurement of other standard samples.
To estimate the element concentration from the peak area counts measured by the ITRAX, the pressed commercial standard reference materials (Subsection “Materials used for the measurement” and Supplementary Table 1) were measured with the ITRAX (see Subsection “ITRAX measurement settings”). The 15 areas of each pressed sample (Supplementary Fig. 2) were measured five times, and all the results (15 × 5 = 75 results) were averaged. The standard deviations of the 75 results were calculated for each standard reference material to obtain a calibration curve for each element. The average counts and standard deviations measured using the ITRAX on the standard materials are shown in Supplementary Table 5 and plotted versus certified values in Fig. 5 and Supplementary Fig. 7. The calibration line for each element, calculated from these results, is also shown in Fig. 5, Supplementary Fig. 7, and Table 2. Among the 27 elements measured with the ITRAX, Cr, W, and Ta were excluded from the discussion because impurities originating from the X-ray tube could contribute to the peak area counts of these elements. Ar was also excluded because the Ar counts mainly reflects Ar in the air between the sample surface and the XRF detector. The correlation coefficient (r2) was >0.9 for the 15 elements (Si, P, S, K, Ca, Ti, Mn, Fe, Ni, Cu, As, Br, Rb, Y, and Ba; Fig. 5 and Table 2), whereas r2 was <0.9 for 8 elements (Al, Cl, V, Zn, Ga, Sr, Zr, and Pb), mainly owing to small counts and the large standard deviation (Supplementary Fig. 7 and Table 2). Therefore, the correlation between the counts and concentrations of the 15 elements (r2 > 0.9) within the concentration range covered by the standard materials was confirmed (Fig. 5 and Table 2). Because concentrations of seven standards are not evenly spread for all elements, more analysis may need for some elements (e.g., P, Br, and Y) to finetune the calibration curve in future studies. The linear relationship between the peak area counts and the concentrations of 15 elements (r2 > 0.9) suggests that the difference in the matrix between the seven standard materials (marine sediment, lake sediment, river sediment, and granodiorite; see Supplementary Table 1 for details) may be negligible, although further analyses are needed for verification.
Calibration line for concentration of each element obtained from standard reference material measurement. The average counts and standard deviations measured by the ITRAX are plotted versus certified values. (a) Si, (b) P, (c) S, (d) K, (e) Ca, (f) Ti, (g) Mn, (h) Fe, (i) Ni, (j) Cu, (k) As, (l) Br, (m) Rb, (n) Y, and (o) Ba.
The list of correlation coefficient and the calibration equations between concentration and count
| calibration equation* | unit of x | correlation coefficient (r2) | |
|---|---|---|---|
| Al | y = 11x + 106 | % | 0.61 |
| Si | y = 93x – 753 | % | 0.99 |
| P | y = 120x + 14 | % | 0.96 |
| S | y = 310x + 16 | % | 1.00 |
| Cl | y = 371x + 36 | % | 0.85 |
| K | y = 2719x – 375 | % | 0.96 |
| Ca | y = 4189x – 39 | % | 1.00 |
| Ti | y = 9157x + 512 | % | 0.99 |
| V | y = 1.5x + 57 | μg/g | 0.88 |
| Mn | y = 27121x + 490 | % | 1.00 |
| Fe | y = 33473x + 42866 | % | 0.94 |
| Ni | y = 4x + 247 | μg/g | 0.98 |
| Cu | y = 7x + 29 | μg/g | 0.91 |
| Zn | y = 8x + 484 | μg/g | 0.73 |
| Ga | y = 16x + 130 | μg/g | 0.59 |
| As | y = 18x – 117 | μg/g | 0.99 |
| Br | y = 16x – 16 | μg/g | 1.00 |
| Rb | y = 25x – 562 | μg/g | 0.94 |
| Sr | y = 13x + 1095 | μg/g | 0.75 |
| Y | y = 12x + 22 | μg/g | 0.99 |
| Zr | y = –0.0048x + 1332 | μg/g | 3.0*10–6 |
| Ba | y = 0.35x + 21 | μg/g | 0.99 |
| Pb | y = 4x + 40 | μg/g | 0.69 |
*y = peak area counts measured by XRF core scanner, x = concentrations (certified values).
The peak area counts of light elements are easily affected by the water in the sediments and the thin water film between the sediment surface and plastic wrap because low-energy fluorescence is easily absorbed by water (e.g., Kido et al., 2006; Chen et al., 2016; Katsuta et al., 2019). Therefore, certain previous studies attempted to correct the water effect of XRF data. For example, Kido et al. (2006) and Katsuta et al. (2019) proposed a water content correction method using X-ray transmission data. Chen et al. (2016) implemented a water content correction method into the MLC method.
In this study, standard materials were prepared in the form of pressed dry powders, whereas the core samples (e.g., archive-halves or U-channels) are wet. To evaluate the water absorption effect, the element concentrations of discrete samples (64 samples) taken from 1-cm thick intervals of IODP Site U1424 and measured by the conventional XRF method were compared with the ITRAX results obtained at the same stratigraphic intervals. ITRAX measurements of U1424 cores were conducted using the second Mo tube when it was new, therefore the results were not affected by the X-ray tube aging effect. Because ITRAX measurements of U1424 cores were conducted at 2-mm intervals (Seki et al., 2019), averages and standard deviations of five sequential measurements equivalent to 1-cm thick intervals were calculated to compare with the results of discrete samples (Supplementary Table 6). Comparison of the results for the 8 elements reveals a difference between the calibration curves obtained for dry samples (black lines in Fig. 6; based on the comparison between ITRAX measurements of pressed standard materials and their certified concentration values) and wet samples (red lines in Fig. 6; based on the comparison between the ITRAX measurements of wet U1424 cores and conventional XRF measurements of discrete samples from U1424 cores; also see Supplementary Table 6). Because lighter elements (Al and Si) have more difference than heavier elements (Ti and Fe) (Fig. 6), this difference can be attributed to the absorption of X-ray fluorescence by interstitial water in the core samples and by a thin water film that formed between the sediment surface and the plastic wrap (Kido et al., 2006). Although the absorption rate by interstitial water should differ depending on the water content of sediments, the results shown in Fig. 6 suggest that equations obtained in this study (Fig. 5) could be used to roughly estimate concentration of elements for relatively heavy elements (Ti and Fe).
Comparison of regression lines for 8 elements obtained from measurement of dry samples (Black closed circle; pressed dry powder of standard reference material) and wet samples (red open square; wet core samples of U1424 cores, element concentration defined by conventional XRF measurement). (a) Al, (b) Si, (c) P, (d) K, (e) Ca, (f) Ti, (g) Mn, (h) Fe.
In this section, the method for utilizing the data shown in this study is described (Fig. 7). To apply this method to the ITRAX XRF core scanner data measured at Kochi University, the measurement date, “XRF voltage and current” and “Tube” from “document.txt” file in the results should be checked first. If “XRF voltage and current” are 30 kV and 55mA (same with this study), respectively, and “Tube” is “Mo”, all data shown in this study can be used to estimate concentrations from the ITRAX results. Second, using the counts of each element contained in the “result.txt”, counting error of each element can be calculated from Table 1, Supplementary Table 4, Fig. 4 and Supplementary Fig. 6 (Fig. 7). Third, “XRF exp. time” is checked in the “document.txt” file. As element peak area counts have linear relationship with “XRF exp. time” (Supplementary Table 7), if “XRF exp. time” is not 10 seconds, all peak area counts contained in “result.txt” should be calculated by following equation before next step (Fig. 7) to compare with the data shown in this study.
Method for calculating concentration of each element from XRF data measured by ITRAX at Kochi University.
Calculated counts = original counts * 10/“XRF exp. time”
Then, estimate the aging effect from Fig. 1, Supplementary Fig. 3, and Supplementary Table 2, and correct aging effect according to the instruction in the Subsection titled, “Correction of the aging effect.” Finally, the concentration of each element was estimated using Fig. 5 and Supplementary Fig. 7 (Fig. 7).
Despite the large error in the XRF core scanner results compared to conventional XRF or ICP measurement, the calculation described above provides useful information for interpretating the approximate elemental composition of measured samples. From the estimation of the measurement error (Table 1, Supplementary Table 4, Fig. 4, and Supplementary Fig. 6), noise and true variation mixed in the ITRAX data could confidently be separated. Because ITRAX measurement can be conducted shortly after the retrieval of cores, rough calculation of element concentration can be used for the planning of further analysis. Therefore, utilization of the dataset in this study can improve the usefulness of the data obtained by ITRAX.
In this study, the factors that affect the peak area counts of elements measured by the ITRAX other than element concentrations (measurement settings and X-ray tube aging effect) were evaluated, and the possibility of the quantitative analysis of chemical composition of the sediments using ITRAX was discussed. In addition, to evaluate chemical composition of sediment samples, a database on the aging effect, precision, and calculation curve for each element was presented.
The aging effect of the X-ray tube was monitored by repeated measurements of standard sample, and correction method for this effect were established. It was demonstrated that the aging effect of the X-ray tube exerts a significant influence on the peak area counts of each element calculated from XRF spectra particularly when time interval of ITRAX measurements spanned a long period. The aging effect on elements such as Br, Rb, and Sr could be cancelled by dividing the peak area counts of each element by the coh counts (Fig. 3). However, this normalization method cannot be applicable to other elements (Fig. 3). Another method for correcting the aging effect is to monitor the decreasing rate of an element count by frequently measuring standard samples and correcting the aging effect using the monitored decrease rate for each element. In this study, the reduction rate of an element count was obtained by measuring the “light layer” standard samples (Fig. 1, Supplementary Fig. 3, and Supplementary Table 2). The decrease rate of an element’s count calculated by measurement of the reference glass standard (Supplementary Fig. 5 and Supplementary Table 3) can also be used to correct the peak area counts for the aging effect.
The precision of the ITRAX XRF core scanner measurement (10 s XRF exposure time) was estimated from measurements of Japan Sea sediment standard samples (“light layer,” “dark layer,” “calcareous,” and “siliceous”) performed five times. The estimated standard deviation of the ITRAX measurements for each element (Fig. 4, Supplementary Fig. 6, Table 1, and Supplementary Table 4) could be applied for other ITRAX measurement results.
Using the commercially available standard reference materials, the calculation equation to estimate the elemental concentrations from the peak area counts measured by ITRAX were established for 15 elements when samples are dry (Fig. 5). Although the XRF absorption effect of the interstitial water and the water film between sediment surface and plastic wrap is not negligible in the case of wet sediments (Fig. 6), the database summarized in this study (Fig. 5, Supplementary Fig. 7, and Supplementary Table 5) provides useful information to estimate the approximate concentrations from other ITRAX measurement results.
We thank Mr. Takuya Matsuzaki, Mr. Shinsuke Yagyu, Ms. Michiko Kawamura, Dr. Lallan P. Gupta, and the staff at the Kochi Core Center for their laboratory assistance. We appreciate Ms. Kaz Mitake for her help with the laboratory and calculations. We also thank Mr. Akinori Karasuda for his assistance with the laboratory. We appreciate two anonymous reviewers for their constructive comments.
This research study used samples provided by the Integrated Ocean Drilling Program (IODP). The study was performed under the cooperative research program of the Marine Core Research Institute (MaCRI) at Kochi University (16A035, 16B031, 17A011, 17B011, 18A048, 18B045, 20A034, 20B031, 22B070). This work was supported by JSPS KAKENHI Grant Numbers 25•9053, 22J40021, and 22KJ149800 for Arisa Seki and 16H01765 for Ryuji Tada. This work was also supported by a Sasakawa Scientific Research Grant for Arisa Seki from the Japan Science Society. Funding for this research was also provided by Fujiwara Natural History Foundation.
