Major element analysis of geological samples with wavelength Dispersive X-ray Fluorescence (WDXRF) spectrometry using glass disks and pressed powder pellets

ABSTRACT A rapid and accurate method for total analysis of major elements in soils, sediments and rock samples has been established by wavelength dispersive X-ray fluorescence spectrometry (WDXRF) using fused glass disks and pressed powder pellets. Thirty geological standard reference materials were used for the construction of the calibration lines. Whereas the widely adopted devices for glass disk preparation were using a high frequency heating furnace and Pt dish for fusion as well as for casting, an electric furnace and Pt crucible were used in this experiment and the molten mixture was finally poured into Pt mold. Although this extended the time needed for disk preparation, the mixing of samples with the borate flux and the dissolution of samples were more complete in this method. The correlation coefficients of the linear calibration lines ranged from 0.9981 (SiO2) to 0.9999 (K2O and CaO). Linear regression analyses between the recommended and calculated values obtained by using the above calibration lines had revealed that the slope is very close to unity, with the exception of P2O5, and the intercept is negligibly small for all constituents. In contrast, the correlation coefficients of the calibration lines obtained by using the pellets ranged from 0.9595 (Al2O3) to 0.9989 (CaO). Despite of these somewhat disappointing results, however, the pellet method with WDXRF is still appeared to be an attractive technique for making rapid semi-quantitative determinations of elemental composition of samples, especially for K2O, CaO, TiO2, MnO and Fe2O3, when considering its easier pre-treatments and faster measurements.


Introduction
X-ray fluorescence spectrometry (XRF) is widely used for determining chemical composition of geological samples.There are two contrasting methods for sample preparation for XRF.One is the use of pressed powder pellets and the other involves the use of glass disks.The first is simple, nondestructive, less time-consuming, and does not need skillful techniques but is more prone to being affected by mineralogical and grain size effects.In contrast, the latter method takes longer to complete and needs a special equipment (bead sampler or fluxer), a Pt crucible and mold, and chemicals (flux), as well as a specialized technique for sample preparation.For the preparation of glass disks, various dilution ratios ranging from 1:1 to 1:300 have been used (Amosova et al., 2016;Gazulla et al., 2012;Nakayama & Nakamura, 2012;Nakayama et al., 2007Nakayama et al., , 2012;;Ogasawara et al., 2018;Xue et al., 2020;Yamasaki, 2014).Lower dilution ratios were employed for both major and trace elements analysis as higher dilution ratio results in lower fluorescent X-ray intensities and consequently makes trace element analysis difficult even with a recently developed high powered X-ray tube (4 kW).However, a higher dilution ratio of 1:9 (or 10 times of dilution) was adopted in this study because it has been shown that around 25 trace elements can be successfully analyzed simultaneously with polarizing energy dispersive X-ray fluorescent spectrometer (EDXRF) installed in our laboratory using the much simpler and quicker pressed powder pellet method.As the instrument is equipped with an Sc/W anode X-ray and a liquid-nitrogen-cooled Ge solid-state high-resolution detector with a Be (8 μm) window, even heavy rare earth elements (from Eu to Lu) can be analyzed using the more sensitive K lines (Takeda et al., 2011;Yamasaki, 2018;Yamasaki et al., 2011Yamasaki et al., , 2013Yamasaki et al., , 2015)).
An extra pure, granular pre-fused flux containing 0.5% of LiBr as a non-wetting agent was used in this study because it is more homogeneous, easier to weigh, and more anhydrous and less hygroscopic.Using this flux also avoids the tedious steps involved in flux preparation.Instead of using a high-frequency heating furnace and Pt dish for fusion as well as for casting for glass disk preparation, as commonly adopted by previous studies (Nakayama et al., 2007(Nakayama et al., , 2012;;Ogasawara et al., 2018;Yamasaki, 2014), an electric furnace and Pt crucible were used in this work, with the molten mixture being finally poured into a non-wet Pt mold (casting dish) also heated in the furnace with the crucible (Figure 1).Although this extended the time needed for disk preparation, the mixing of samples with the borate flux and the dissolution of samples were more complete in this method.Glass disks and the pellets were prepared incorporating 30 geological standard reference materials (Table 1).As matrix effects were automatically corrected by the build-in software (Kimura & Yamada, 1996), the differences between disks and pellets were presumed to be attributable to the mineralogical and grain-size effects.

Wavelength dispersive X-ray fluorescence spectrometer
A ZSX Primus IV (Rigaku Corporation Inc., Tokyo, Japan) equipped with an end-window 4 kW Rh X-ray tube together with an artificial multilayer film crystal for Na and Mg was used in this study.The instrument was calibrated for 10 major analytes using 30 geological standard reference materials (19 rock samples, 4 sediments and 7 soil samples).These reference materials are listed in Table 1, together with their basic information and the producers (Bowman et al., 1979;Imai et al., 1995Imai et al., , 1996Imai et al., , 1999;;National Bureau of Standards & Technology, 1993a, 1993b, 2002;Terashima et al., 2002).The measurement conditions are given in Table 2.As it was necessary to take longer time to accumulate enough counts for Na, Mg and Al, the total time needed for disk samples was around 1.5 times longer than that of pellet samples.

Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES)
An Agilent 5100 (Agilent Technologies, Santa Clara CA, USA) was used in this study.Working standards were prepared from a series of SPEX CertiPrep Multi-elements standards (Quality Control Standard 23, Catalog Number QC-23) and single element standards (Ti, Catalog Number PLTI9-2Y) supplied by Spex Industries, Inc. (Edison, New Jersey, USA).The emission intensities for each element were measured by radial mode at two (Na and K) or three prominent lines (all the other elements) by using the simultaneous multi-wavelength capability of the instrument.Then, the results obtained by the most suitable analytical lines for each element (Table 3), were selected after considering such factors as signal intensities, interferences, background levels, and the reproducibilities of the standard solutions analyzed after every 10-20 real samples.

Preparation of glass disks
All the samples, except for standard reference materials, were pulverized for 10 minutes in a planetary ball mill (pulverisette 7; Fritsch GmbH, Idar-Oberstein, Germany) at 450 rpm.Glass disks were prepared by using a Claisse LeNeo Fluxer (Claisse S. A., now Malvern Panalytical B. V., Lelyweg 1, Almelo, The Netherlands).An amount of 0.600 g of finely ground sample and 5.400 g of the extra pure lithium borate flux (49.75% of Li 2 B 4 O 2 , 49.75% of LiBO 2 and 0.5% of LiBr), catalog number of C-0621-60) provided by the above manufacturer were thoroughly mixed in a disposable plastic tray and then quantitatively transferred to a 10 mL capacity non-wet Pt crucible of 2 cm in diameter and 4 cm in depth (alloy of Pt with 5% Au, 30 g).The crucible and the mold (3.5 cm in diameter and 0.5 cm in depth, 27 g) were then heated in a furnace at 1050 ℃ with the time distribution given in Table 4.The molten mixture was finally poured into a non-wet Pt mold also heated at 1050 ℃ (Figure 1).The most noticeable advantage of this method is the high durability of Pt crucible.It has been possible to prepare more than 2000 disks since 2017 without recasting.

Preparation of pellets
The pulverized sample was put into a polyvinyl chloride ring (25 mm in inner diameter and 5 mm in height) and  pressed using dies made of tungsten carbide and an electrically operated press at 200 kN (NPa SYSTEM CO., LTD., Warabi, Saitama, Japan).The surfaces of dies contacting the sample were covered with a polypropylene thin-film manufactured by Chemplex Industries, Inc., (Palm City, Florida, USA.Catalog Number 425) to prevent possible contamination.

Sample dissolution
A specially designed heat block (DigiPREP MS, SCP SCIENCE, Quebec, Canada) and digestion vessels made of PTFE were used.In this apparatus, it is possible to treat up to 48 samples simultaneously with the previously programmed schedule.The sample temperature was precisely controlled via a corrosion-resistant probe.An amount of 0.500 g of finely ground sample was first treated with 4 mL of HClO 4 -HNO 3 (1:1 mixture) for 45 min.at 120°C and then with 4 mL of HClO 4 -HF (1:1 mixture) twice for 45 min at 170°C.The residue was heated with 5 mL of HNO 3 and dissolved by adding 20 mL of H 2 O at 95°C for 3~4 hours and finally made up to 100 mL.

X-ray intensities as influenced by the dilution ratios
Table 5 shows the ratios of the fluorescent X-ray intensity of the 1:9 glass disk of JA-1 against that of the 1:2 (2.000 g of sample and 4.000 g of flux).The decrease in the X-ray intensity due to dilution by the flux is less than those calculated by the dilution ratio (0.300) because both the penetration depth (a function of the energy of the incident X-rays from the instrument), and the escape depth (the depth from which fluorescent energy can escape the sample and reach the detector) increase with the dilution ratios as shown by Xue et al. ( 2020).These trends were more noticeable on heavier elements.

Calibration lines obtained by glass disks
Table 6 lists the correlation coefficients, the root mean square error (RMSE), and concentration ranges of the calibration lines.More detailed discussion on the RSMS where Xa is the content in the reference material, Xb is the content calculated using regression equation, and N is the number of reference materials.
and the limit of detection has been shown elsewhere (Borkhodoev & Ya, 2016).The number of samples used for the calculation for MnO was 29 rather than 30 because the content of MnO in NIST 2710 (1.31%) is around five times higher than that of the second-highest sample JLk-1 (0.266%).When this reference material was included for calculation, the correlation coefficients for MnO was 0.9999.In spite of large number of samples with a wide concentration range, as well as their different chemical and mineralogical properties, the correlation coefficients for all of the analytes were more than 0.9980, clearly indicating that excellent calibration lines were established.Linear regression analyses between the recommended and calculated values obtained by using the above calibration lines were shown in Figure 2. The slope is very close to unity, with the exception of P 2 O 5 , and the intercept is negligibly small for all constituents.It can be concluded, therefore, that the proposed method assures accurate and dependable results.

Repeatability of the glass disk measurements
Table 7 shows the repeatability of the disk measurements.More detailed discussion on these and related topics has been shown elsewhere.Eight successive measurements of the glass disk of JB-1a were carried out within 2 hours by the same person.The relative standard deviations (RSD) are less than 0.7%.The higher values of Na 2 O and MnO are considered to be due to low fluorescence yield and low content respectively.

Comparison with results obtained by using pellets
In contrast, the correlation coefficients for pellets ranged from 0.9577 (MgO) to 0.9989 (CaO), being lower for elements of lower atomic number, especially Na, Mg and Al (Table 8), which is interpreted as an effect of the lower escape depth (the depth from which fluorescent energy can escape the sample and reach the detector) for lighter elements (Ichikawa & Nakamura, 2016).Despite of these discrepancies and little satisfying results, however, the pellet method with WDXRF still appeared to be an attractive technique for making rapid semiquantitative determinations of elemental composition of samples, especially for K 2 O, CaO, TiO 2 , MnO and Fe 2 O 3 , when considering its easier pre-treatments and faster measurements.

Comparison with results obtained by acid digestion and ICP-AES
Figure 3 shows the results for the WDXRF glass disk compared with those for chemical digestion and ICP-AES for soil samples.The samples were selected in such a way as to cover a wide range of chemical as well as mineralogical compositions.As can be predicted from the excellent calibration lines for WDXRF glass disk method, the correlation coefficients of the linear regression analysis range from 0.9978 (Fe 2 O 3 ) to 0.9998 (K 2 O) and the intercepts are negligibly small for all the analytes, although the slopes for Al 2 O 3 , K 2 O and MnO are around 0.95.In addition, whereas the concentration ranges of MgO, MnO and Fe 2 O 3 are considerably wider than those of the calibration lines, the very high level of agreements between the methods suggests that considerable extrapolation could be allowed in the proposed WDXRF glass disk method.

Conclusions
A rapid and accurate method for total analysis of major elements in soils, sediments and rock samples has been established by wavelength dispersive X-ray fluorescence spectrometry (WDXRF) using fused glass disks and pressed powder pellets.Linear regression analyses between the recommended or certified values and the calculated values obtained by using glass disks have revealed that the slope is very close to unity, with the exception of P 2 O 5 , and the intercept is negligibly small for all constituents.In contrast, the correlation coefficients of the calibration lines obtained by using the pellets ranged from 0.9595 (Al 2 O 3 ) to 0.9989 (CaO).Despite of these discrepancies and little satisfying results, however, the pellet method with WDXRF still appeares to be an attractive technique for making rapid semi-quantitative determinations of elemental composition of samples, especially for K 2 O, CaO, TiO 2 , MnO and Fe 2 O 3 , when considering its easier pre-treatments and faster measurements.

Figure 1 .
Figure 1.Pouring the melt to the mold (left), and the obtained disk and the mold (right).

Figure 2 .
Figure 2. Relationship between the recommended or certified values and the calculated values.

Figure 3 .
Figure 3.Comparison with results obtained by acid digestion and ICP-AES.

Table 1 .
Standard reference materials used for the calibration

Table 2 .
Instrumental conditions for X-ray fluorescence analysis a PX25, Artificial multilayer film.b PC, Gas flow proportional counter.c PET, Pentaerythritol (C5H12O4).d Ge, Germanium.e LiF, Lithium fluoride.f SC, Scintillation counter.g Detection Time, Figures in parentheses show those for pellet.h BG, Background.

Table 3 .
Analytes and analytical lines for ICP-AES a I: Neutral line; II: Singly ionized line.

Table 4 .
Time distribution of the fusion procedures

Table 6 .
Analytical figures of merit for major constituent calibration obtained by glass disks a N, Number of reference materials used for calculation.b R, Correlation coefficient of the linear regression lines; c RMSE, Relative mean square error.RMSE

Table 7 .
Repeatability of the glass disk measurements a Relative standard deviation.

Table 8 .
Analytical figures of merit for major constituent calibration obtained by pellets Number of reference materials used for calculation.bR, Correlation coefficient of the linear regression lines; c RMSE, Relative mean square error.RMSE = (Σ(Xa-Xb)2/(N-2))1/2,where Xa is the content in the reference material, Xb is the content calculated using regression equation, and N is the number of reference materials.
a N,