Application of XRF Scanning to Different Geological Archives

Abstract

High‐resolution XRF scanning is widely used on marine and lake sediment cores as a means of rapidly acquiring elemental concentrations from closely spaced intervals with no damage to the samples. Therefore, guidance on how to obtain reliable datasets and select suitable step sizes for different geological media are of great importance. Here we apply this efficient analytical method to loess, stalagmite, and tridacna samples. The results show that 11 elements (Al, Si, K, Ca, Ti, Mn, Fe, Rb, Sr, Zr, and Ba), eight elements (Ca, Si, K, Fe, Cu, Ni, Sc, and Sr) and seven elements (Ca, Sr, Cu, Fe, Sc, Ni, and Si) can be robustly detected by this method for loess, stalagmite and tridacna archives respectively, demonstrating the capacity to reconstruct high‐resolution paleoclimate changes. For loess cores, efficiency emission decay of the X‐ray tube, water content, matrix, and grain size effects are the main factors influencing elemental intensities. The efficiency emission decay of the X‐ray tube, scanning path, scanning interval, and radiation area have a significant effect on element intensities of stalagmite and tridacna samples. Based on our investigations, we suggest optimal resolutions for scanning these three archives for millennial to seasonal‐scale paleoclimatic reconstructions. We compare our analyses with existing results from traditional (discrete) analyses, demonstrating similar scales of variability. Our results suggest that geochemical proxies measured by XRF scanning, such as Rb/Sr, Zr/Rb, and Ca/Ti ratios of loess and the Sr/Ca ratio of speleothem and tridacna, can be effectively used to reconstruct high‐resolution paleoclimate changes.

Introduction

With evolving technology, the generation of high-resolution paleoclimate reconstructions is providing critical insights into tipping points and threshold responses in Earth's climate system. XRF scanning makes it possible to produce near continuous geochemical elemental analyses of sediment sequences. Since XRF core scanning techniques developed in late 1990s (Jansen et al., 1998), it has primarily been used to scan marine sediment cores. Due to its convenient, fast and nondestructive advantages, high-resolution XRF core scanning has gradually been utilized to study the geochemical characteristics of lake sediments, loess, peat, and stalagmites as well (Finné et al., 2015; Francus et al., 2009; Kasper et al., 2012; Kylander et al., 2011; Liang et al., 2012; Löwemark et al., 2011; Sun et al., 2016; Tan et al., 2018; Yancheva et al., 2007; Ziegler et al., 2008; Zolitschka et al., 2002).

Many significant findings regarding the characteristics of abrupt climate change and the underlying dynamics have been obtained based on high-resolution elemental records through XRF scanning (Haug et al., 2001; Hodell et al., 2008; Lamy et al., 2004; Spofforth et al., 2008; Sun et al., 2016; Tan et al., 2018). Lamy et al. (2001) used XRF scanning Fe counts and clay mineral parameters from a high-accumulation marine sediment core in the Chilean continental slope to reconstruct rainfall variability since the 7,700 cal yr B.P. These indicators showed pronounced rainfall variability on centennial to millennial time scales. Yancheva et al. (2007) produced a record of Ti counts with nearly annual resolution in Lake Huguang Maar, southeast China, which documented East Asian winter monsoon changes on centennial to millennial timescales over the past 16,000 years. Channell et al. (2012) distinguished Heinrich-like layers associated with IRD input using XRF scanning Ca/Sr and Si/Sr ratios at Sites U1302/03 and U1308. Past storm surge deposits and the timing of past surge events were also inferred from XRF scanning results (Swindles et al., 2018; Tsobanoglou et al., 2010).

More recently, researchers have succeeded in applying XRF scanning to speleothem archives (Li et al., 2019; Tan et al., 2018), corals (Ellis et al., 2019), mumiyo mounds (Berg et al., 2019), rocks (Peti et al., 2019), sedimentary nodules (Croudace, Teasdale, & Cundy, 2019), archeological artifacts (Croudace, Löwemark, et al., 2019), tree rings (Hevia et al., 2018; Sánchez-Salguero et al., 2019) and melt segregations (Alexandrin et al., 2018). Most researchers focus on elemental results with application to paleoclimatic reconstruction. However, little work has been done to test whether experimental methods influence the output of elemental results. The influencing factors resulting from parameter settings (current, resolution, radiation area and scanning path selection, etc.) are not often evaluated. The types of geological material suitable for XRF scanning analysis are rarely assessed.

Here we further these efforts using XRF scanning on loess, stalagmite, and tridacna samples to illucidate which elements can be robustly obtained by XRF scanning, focusing on reproducibility and the major factors influencing the results for each archive. We approach this by comparison with existing records (such as quantitative element content ratios, mean grain size (MGS), magnetic susceptibility (MS), and isotopes data) from these archives, across a range of timescales. We seek to assess optimal settings, confirm which elemental data are reliable, and explore the potential to reconstruct paleoclimatic changes at different timescales using these archives.

Materials and Methods

We use a fourth-generation Avaatech XRF scanner with the latest variable optical XRF technology and high scanning resolution (10 cm–0.1 mm) to conduct our experiments. The apparatus is an Oxford Instruments 100 W Neptune, equipped with a rhodium (Rh) X-ray tube and a SGX Silicon Drift Detector. The basic unit of the XRF core scanner measurement is total counts or counts per second (cps). This unit manifests the elemental intensity which is proportional to the chemical concentrations (Tjallingii et al., 2007). The scanner can measure different archives up to 160 cm long, 14 cm wide, and 7 cm thick. Additional information about technical specifications details of the scanner can be found at http://www.avaatech.com/. The XRF core scanner provides results for the following elements under three different voltages: First run (10 kV)—Mg, Al, Si, P, S, Cl, Ar, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Rh, and Ba. Second run (30 kV)—K, Ca, Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Br, Rb, Sr, Y, Zr, Mo, Rh, Pb, and U, and third run (50 kV)—Br, Rb, Sr, Y, Zr, Mo, Rh, Ag, I, and Ba. The XRF core scanner is set for different voltages with dead time around 30%, which can affect the output counts from the pulse processor. One monitor and three standard samples are used to correct elemental counts. Average analytical errors on the determination of elements are less than 5% (n = 20), except Ba (9.0%) for loess (Guo et al., 2021), Ni (17.9%) and Sc (5.3%) for stalagmite, and Fe (6.7%) for tridacna.

In order to assess the reliable elements, influencing factors and potential application of elemental results to reconstructing paleoclimate changes, two loess cores, four stalagmite and four tridacna samples are chosen for XRF scanning. Detailed position information, setting and method for each sample are shown in Figure 1 and Table 1. Associated preparation and methods for each archive are introduced in the following section.

Figure 1. The location of loess (red dots), stalagmite (purple stars), and tridacna (yellow triangles) samples.

Table 1 Selected Samples of Loess, Stalagmite and Tridacna, as Well as Detailed Methods and Settings for Each

Archive	Name	Method	Step size	Radiation area	Voltage and current
Loess	Linxia	CXRF	1 cm	1.2 cm2 (10*12 mm)	10 kV/0.075 mA and 50 kV/0.070 mA
		DXRF	2 cm	1.2 cm2 (10*12 mm)	10 kV/0.075 mA and 50 kV/0.070 mA
		Two times CXRF scanning in different paths for 33H of core B (10 mm offset from center)	1 cm	1.2 cm2 (10*12 mm)	10 kV/0.075 mA and 50 kV/0.070 mA
	Zhengzhuangcun	CXRF	5 mm	25 mm2 (5*5 mm)	10 kV/0.075 mA and 50 kV/0.070 mA
		CXRF	1 cm	1.2 cm2 (10*12 mm)	10 kV/0.075 mA and 50 kV/0.070 mA
Stalagmite	RM8-2	Three times CXRF scanning in different paths	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.135 mA and 30 kV/0.030 mA
	C9	CXRF	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.135 mA and 30 kV/0.030 mA
		CXRF	0.2 mm	4 mm2 (2*2 mm)	10 kV/0.135 mA and 30 kV/0.030 mA
		CXRF	0.5 mm	4 mm2 (2*2 mm)	10 kV/0.135 mA and 30 kV/0.030 mA
	F16-1	CXRF	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.135 mA and 30 kV/0.030 mA
	XL16	CXRF	0.2 mm	4 mm2 (2*2 mm)	10 kV/0.135 mA and 30 kV/0.030 mA
Tridacna	A5	Two times CXRF scanning in different path	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.085 mA and 50 kv/0.500 mA
		CXRF	0.2 mm	4 mm2 (2*2 mm)	10 kV/0.085 mA and 50 kv/0.500 mA
	MD-4	CXRF	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.085 mA and 50 kv/0.500 mA
		CXRF	0.2 mm	4 mm2 (2*2 mm)	10 kV/0.085 mA and 50 kv/0.500 mA
	HYD-1	CXRF	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.085 mA and 50 kv/0.500 mA
	A87	CXRF	0.1 mm	1 mm2 (1*1 mm)	10 kV/0.085 mA and 50 kv/0.500 mA

Loess

Linxia and Zhengzhuangcun loess cores were retrieved from the northwest and middle-south Chinese loess Plateau (CLP), respectively (Figure 1). After loess cores were split into two parts along the center, we carefully pressed and smoothed the half cores' fresh surface to form an even surface using a plastic blade, which minimizes errors caused by surface irregularities and minor cracks. High-definition of photographs of the half cores obtained from the scanner are shown in Figure 2a. Powder samples from the Linxia profile were prepared in 2 cm intervals (about 2 g), dried at 40°C overnight, grounded to 200 mesh size (<75 µm) in an agate mortar with a pestle, and then pressed into a plastic sheet to make slide samples (4 × 4 × 0.3 cm), creating a flat, homogeneous and reproducible surface. Finally, the plastic slide samples were placed on a wood pallet (Figure 2a). Before XRF scanning, all the samples were covered with a Ultralene film (4 μm thick) to avoid contamination of the measurement prism on the core scanner.

Figure 2. Loess, stalagmite, and tridacna samples. (a) Core and slide samples of Linxia loess profile; (b) Four stalagmites samples (RM8, RM8-2, F16-1, C9, and XL16); and (c) Three tridacna samples (HYD-1, A5, and MD-4). Red dash lines denote the sampling path for the plasma - optical emission spectrometry measurement, and other color lines represent different XRF scanning paths.

The archive halves of samples were scanned with an irradiated area of 1.2 cm² (10*12 mm) and 25 mm² (5*5 mm). The system is set to perform two consecutive runs of the same section. The analytical conditions of the core scanner are as follows: first scan 10 kV and 0.075 mA, second scan 50 kV and 0.070 mA (Table 1). Forty meter-long Linxia loess cores and powder samples were analyzed using continuous XRF (CXRF, scanning in August 2017) scanning and discrete XRF (DXRF, scanning in October 2018) scanning at 1 and 2 cm intervals, respectively. One section (33H of core B) from Linxia profile was scanned 2 times in different parallel paths a few hours apart. Zhengzhuangcun (4 m long) U-channels were measured by CXRF scanning at 1 cm and 5 mm step sizes (Table 1). The samples were scanned lengthwise along the center of the split surface using the Avaatech XRF scanner at the Institute of Earth Environment, Chinese Academy of Sciences (IEECAS).

Stalagmite

Four stalagmites were selected for XRF scanning analysis, including XL16, RM8-2, F16 and C9, sampled in Shaanxi province, Hunan province, Central Asia Kyrgyzstan, respectively (Figure 1). As for these samples, RM8-2 was aragonitic, whereas samples C9 and F16-1 were calcite (Figure 2b). After the stalagmites were split into two or several parts along the growth axis, the surface was polished and then photographed. We noted that XRF scanning was not suitable for samples with constantly varying growth paths or too many voids on the sample surface. Although there was no contamination, we still covered the surface of samples with a 4 μm thick Ultralene film.

The archive pieces of samples were completely scanned with different step sizes and irradiated area. The analytical conditions of these scans are as follows: first run 10 kV and 0.135 mA, second run 30 kV and 0.03 mA. RM8-2 was measured at 0.1 mm step size with 1 mm*1 mm radiation area along three different paths few days apart (as was shown in Figure 2b) to evaluate the reproducibility along different paths. Sample C9 was scanned 3 times along the same path at 0.1, 0.2 and 0.5 mm step size with 1*1 mm and 2*2 mm radiation area so that the consistency of results in different resolutions can be assessed. F16 and XL16 samples were tested at 0.1 and 0.2 mm intervals with 1*1 mm and 2*2 mm radiation areas. Finally, to examine whether these results are reliable, the scanning data were compared with inductively coupled plasma - optical emission spectrometry (ICP-OES) results for samples of XL16 and C9 (Li et al., 2019).

Tridacna

Four tridacna samples were chosen for testing the fidelity of XRF scanning results. A5 and A87 were collected from Paracel Islands. HYD-1 and MD-14 were taken from Scarborough Shoal and Fiery Cross Reef, respectively (Figure 1). A5 and MD-14 were aragonitic, whereas samples HYD-1 and A87 were calcitic (Figure 2c). The pretreatments for tridacna samples were the same as that for the stalagmite samples. The system was set to perform two consecutive runs of the same section. Analytical conditions of the core scanner were as follows: first run 10 kV and 0.085 mA, second run 50 kV and 0.50 mA.

Samples A5 and MD-14 were both measured at 0.1 mm (with 1*1 mm radiation area) and 0.2 mm (with 2*2 mm radiation area) step size. Samples HYD-1 and A87 were scanned at 0.1 mm interval. We scanned sample A5 twice along different paths at 0.1 mm (1*1 mm radiation area) step size so that results from different paths could be compared. The scanning paths were shown in Figure 2c. Finally, to address whether scanning XRF results were robust, selected scanning elements data were compared with ICP-OES results for samples A5 and A87.

Results and Discussion Loess Elemental Results

XRF scanning results show that seven major elements (Al, Si, K, Ca, Ti, Mn, and Fe) and four trace elements (Rb, Sr, Zr, and Ba) display high signal peaks relative to the background and exceed the threshold criteria (generally >100 cps). The intensities of 11 elements measured by CXRF range from 150 to 25,500 cps (Figure 3 and Table 2). For these 11 elements, Ca, Ti, Mn, Fe, Rb, Sr, and Zr display distinctive loess-paleosol alternations (glacial-interglacial variations) similar to that of MGS and MS records; whereas Al, Si, and K manifest weaker loess-paleosol contrast than that of other seven elements. The easily mobilized elements Ca and Sr are intensively affected by precipitation-induced leaching during pedogenic processes, exhibiting high counts in coarse loess layers and low counts in fine particle paleosol layers. However, mobile Rb, Sr, Fe, Mn, Ba and relative stable Zr, Ti, mostly enriched in fine-grained minerals during the pedogenic processes, display an inverse glacial-interglacial variations pattern with higher intensities in paleosol layers and lower intensities in loess layers. In addition, the intensities of Al, Si, and K show unclear loess-paleosol alternations, especially for DXRF results.

Figure 3. Comparison of DXRF (black curve with 2 cm resolution) and CXRF (blue curve with 1 cm resolution) element content for the Linxia profile. The brown and purple curves are mean grain size and magnetic susceptibility of Linxia profile, respectively.

Table 2 Descriptive Statistics for the 11 Elements Measured by the CXRF and DXRF Methods

Element-CXRF	Al/cps	Si/cps	K/cps	Ca/cps	Ti/cps	Mn/cps	Fe/cps	Rb/cps	Sr/cps	Zr/cps	Ba/cps
Max.	1336.89	12221.78	5279.65	29636.35	1539.49	433.44	14713.79	272.33	754.40	753.45	350.60
Min.	324.35	3968.22	2754.74	2900.74	776.17	175.09	7409.26	147.00	345.13	484.60	237.07
Average	907.63	9107.66	4316.99	14924.14	1151.45	269.33	10836.24	205.98	526.42	642.38	303.93
SD	148.76	1303.25	398.61	3890.65	129.98	39.48	1273.22	19.16	55.20	34.98	19.78
CV	16.39	14.31	9.23	26.07	11.29	14.66	11.75	9.30	10.49	5.45	6.51
Element-DXRF	Al/cps	Si/cps	K/cps	Ca/cps	Ti/cps	Mn/cps	Fe/cps	Rb/cps	Sr/cps	Zr/cps	Ba/cps
Max.	1243.00	9123.17	4586.14	28750.76	1463.06	354.87	13632.62	218.34	493.82	728.22	373.05
Min.	595.25	6549.68	3394.68	3212.45	935.84	195.98	7892.23	143.98	260.78	501.98	195.12
Average	838.14	7888.90	3925.67	16158.54	1125.00	254.37	10162.86	172.08	382.61	597.98	289.11
SD	122.33	377.66	233.99	4189.07	102.44	31.63	1024.31	13.68	38.57	36.75	35.50
CV	14.60	4.79	5.96	25.92	9.11	12.43	10.08	7.95	10.08	6.15	12.28

Note. SD and CV represent standard deviations and coefficients of variation, respectively.

Influencing Factors

XRF scanning results indicate that 11 elements (Al, Si, K, Ca, Ti, Mn, Fe, Rb, Sr, Zr, and Ba) are characterized by high signal peaks relative to the background and thus appropriate for paleoclimate research. To assess the reproducibility of these elements, we rescanned section 33H (1.50 m) of LXB twice at 1 cm step size along different paths (center and 10 mm offset from center) using CXRF. Visual inspection indicates that the results are reproducible for Ca, Ti, Fe, Rb, Sr and Zr (Figure 4). According to the correlation analysis among these three datasets, Ca is the most consistently reproducible element in the CXRF results with high coefficients of determination (R² = 0.94–0.96, n = 148), then followed by Sr (R² = 0.85–0.92), Fe (R² = 0.71–0.82), Ti (R² = 0.60–0.80), Rb (R² = 0.44–0.75), and Zr (R² = 0.43–0.62). Because of the indistinct variations and noisy signals, the correlations among the three scans for Mn, Al, Si, and K are relatively low (R² = 0.01–0.39), except for the K of two latter rescanning (R² = 0.70). In addition, Ba, Zr, and Sr XRF intensities decrease progressively through the latter two repeated scans, but variation trends of the elements are well correlated. The offsets are mainly due to efficiency emission of the two different X-Ray tubes (scanned at different times) as well as the changing water content for 33H over time.

Figure 4. Repeated scanning measurements for the LXB 33H at 10 kV 0.075 mA and 50 kV 0.070 mA; 10 and 20 s respectively. Green (August 11, 2017), blue (in the afternoon of November 5, 2019) and orange (midnight, November 5, 2019) curves are 11 elemental counts and 3 element ratios measured by CXRF method in different times.

The first X-Ray tube was installed on August 5, 2017 (the green line data was obtained on August 11, 2017). The second new X-Ray tube was changed on May 3rd, 2019 (Orange and blue data measured on November 5, 2019). The greatest changes are observed for the lowest initial intensities. Similar patterns are also apparent for Al, Si and K. For light elements, the detection depth is progressively smaller than regular detective depth (2–4 mm), for example 1–2 mm for Sr, and 0.05 mm for Al (Richter et al., 2006). The results of Sr, Al, Si and K are easily influenced by water content and water film between the core and 4 μm thick Ultralene film. For heavy elements, such as Zr and Ba, the detection limit increases because the XRF detector is slightly less efficient at the high wavelengths of their emitted radiation (Richter et al., 2006). Standards measured every three days or weekly are used to correct the element counts to account for the observed offsets. Element results shift to same variation level after correction (Figure 4, the first scanned Al, Si and K are divided by decay factor 6.36, 3.84 and 1.39, respectively). Meanwhile, profile shapes (millennial-scale signals) of element ratios (Rb/Sr, Ca/Ti and Zr/Rb) are also consistently reproducible in repeat scans, which lends confidence that the ratios can be used in reconstructing paleomonsoon changes.

In order to ascertain which method produces higher quality results, standard deviations (SD) and coefficients of variation (CV) are assessed. Compared with CXRF results, the intensities of elements measured by the DXRF method show more similar trends and exhibit significant variability at glacial-interglacial cycles (Figure 3). Elements measured at 10 kV for the CXRF method, which are more likely affected by variations of interstitial water, presence of a water film, grain size and surface mineral inhomogeneity (Hennekam & Lange, 2012; Jansen et al., 1998; Liang et al., 2012; Tjallingii et al., 2007) yield more noisy signals and lower precision, with lower SD and CV. This is expected given that the DXRF method employs dried and fully homogenized powder samples, which results in removing or reducing the influence of water content, surface mineral inhomogeneity and grain size effects. Thus, the DXRF method (Table 2) produces results with higher quality and better precision, exhibiting relatively lower variation amplitude, average values, SD (except for Ca) and CV, especially for light elements such as Si (4.79), K (5.96) and Ti (9.11). Elements measured at 50 kV by DXRF, which are more stable and less influenced by surface roughness and physical properties of core sediments (Guo et al., 2020; Jones et al., 2019), have slightly smaller variation amplitude, lower average values, higher SD and lower CV, except for Ba (CV = 12.28).

In addition, comparisons of elemental results from scanning XRF with the elements measured by traditional quantitative XRF also suggest that intensities of 11 elements can be reliably detected by the CXRF and DXRF methods (Guo et al., 2021; Sun et al., 2016). Irrespective of which XRF scanning method is used, the results for seven elements (Ca, Fe, Ti, Mn, Zr, Rb, and Sr) are accurate and credible for paleoclimatic reconstruction, showing high correlation (R² > 0.7, n = 100) with those measured by conventional XRF method (Guo et al., 2021). According to our analysis and previous comparison experiments, both the CXRF and DXRF methods are effective for high-resolution elemental analysis of loess cores and discrete samples. The results of Ca, Fe, Ti, Mn, Rb, Sr, and Zr are accurate and credible. Determination of Al, Si and K elements are not robust and should be approached with caution.

The sedimentation rate (SR) across the CLP is gradually reduced from the northwest to southeast. Based on previous research on SR of loess profiles in the CLP (Ma et al., 2017), these loess sections can be classified into three types, including type A (SR > 12 cm/ka, in the northwest of CLP), type B (SR varies between 12 and 6.5 cm/ka, in the middle of CLP), and type C (SR < 6.5 cm/ka). According to the age model of Linxia profile, SR of Linxia profile varies from 12.10 to 81.67 cm/ka with corresponding resolution ranging from 12.2 to 82.7 year/cm. Therefore, we could utilize 1–2 cm and 2–10 cm scanning intervals to record centennial scale and millennial scale variations. The resolution of type B loess profiles ranges from 83.3 to 153.8 years/cm. Therefore, we could select 0.5–1 cm and 1–5 cm step size to document centennial-scale and millennial-scale climate changes, respectively. The resolution of type C loess profiles is more than 153.8 years/cm. Thus 0.2–0.5 cm and 1–2 cm step size could be applied to reflecting centennial-scale and millennial-scale changes.

Different scanning resolution (1 and 0.5 cm) results of the Zhengzhuangcun loess cores (4 m long), located in the middle of the CLP, shows that 10 detected elements exhibit the same trend and amplitude variations, which confirms that the scanning results are repeatable and reliable (Figure 5). According to Wang et al. (2019), the SR of Zhengzhuangcun profile varies from 6.3 to 20 cm/ka. Therefore, Zhengzhuangcun profile belongs to the type B loess and 0.5–1 cm step size is the proper choice for millennium scale elemental analysis. Several abrupt climate changes during the Holocene and the Last Glacial (such as Younger Dryas, H₁ and H₂) are recorded by the Ca/Ti ratio, MGS and MS at 1.0 cm resolution (Wang et al., 2019).

Figure 5. Comparison of different scanning resolution results for the Zhengzhuangcun profile. The black dotted line is for 0.5 cm resolution. The red dotted line is for 1 cm resolution.

Comparison With Other Proxies

Comparison of DXRF derived element ratios with MS and grain size records indicates that the XRF-based records have great potential for elucidating rapid fluctuations of the East Asian monsoon (Guo et al., 2020; Liang et al., 2012; Sun et al., 2016). The MGS of the Linxia core reflects distinct orbital to millennial-scale winter monsoon oscillations. The MS record exhibits glacial-interglacial fluctuations associated with loess-palaeosol alternations (Figure 6). To assess the paleoclimatic implications of high-resolution element ratios, Ca/K and Zr/Rb, determined by DXRF are compared with MS and MGS of the Linxia core (Guo et al., 2019, 2020), Chinese speleothem δ¹⁸O (Cheng et al., 2009, 2016), and stacked benthic δ¹⁸O (Lisiecki & Raymo, 2005). Zr/Rb exhibits significant glacial-interglacial fluctuations similar to those of the grain-size record, illustrating its capacity to document rapid winter monsoon changes. Ca/K is sensitive to chemical weathering intensity induced by changes in summer monsoon strength. Notably, Ca/K exhibits a parallel pattern of fluctuations to that of MS at the orbital timescale, but is still sensitive to subtle climate changes during glacial times with evident precession-scale and millennial-scale fluctuations. As is shown in Figure 6, Ca/K, Zr/Rb and MGS can resolve warming events during glacial-interglacial cycles. In addition, significant cooling signals for the last climatic cycle (Channell et al., 2012; McManus et al., 1994) and the penultimate climatic cycle are also very distinct in loess proxies, corresponding well with weak summer monsoon events recorded by speleothem, similar to North Atlantic Ocean records.

Figure 6. Sedimentation rate (black) of the Linxia profile, comparison of Ca/K (brown) and Zr/Rb (green) determined by DXRF, magnetic susceptibility (pink) and mean grain size (orange) for the Linxia core (Guo et al., 2019, 2020) with Chinese speleothem δ18O (purple, Cheng et al., 2009, 2016) and stacked benthic δ18O (blue, Lisiecki & Raymo, 2005). The numbers of 1-5e of stacked benthic δ18O record represent different marine isotope stages. S0 and S1 are the first and second paleosol layer of Linxia profile.

Stalagmite Elemental Results

Although the stalagmites are mainly made of calcium carbonate, seven elements can be detected by CXRF method for stalagmite C9, including Si, K, Fe, Cu, Ni, Sc, and Sr. This is because, except for K (∼3.31 keV), the characteristic peaks of Si (∼1.74 keV), Sc (∼4.09 keV), Fe (∼6.4 keV), Cu (∼8.04 keV), Ni (∼7.47 keV) and Sr (∼14.2 keV) are far away from the interference of Ca (∼3.7 keV) and diffraction peaks (Richter et al., 2006; Scroxton et al., 2018). The spectral peaks produced by Ca and the pseudo-peaks produced by the diffraction of carbonate crystalline matrix do not interfere with identifying characteristic peaks of these elements. Among these elements, the metallic elements (Fe, Cu, Ni and Sc) show synchronous fluctuations on long time scales (Figure 7). Most small peaks, on short time scales, are confused with the background noise. The intensities of Ca and Si are relatively stable with average value of 175 and 72,000 cps. K and Sr display two clear peaks in upper part and decreasing trend in following part.

Figure 7. XRF scanning results of calcite stalagmite C9 (Ca in blue, Sr in orange, K in claybank, Sc in brown, Ni in verdancy, Cu in yellow, Si in purple and Fe in dark blue of A5).

Unlike C9, not all stalagmites yield detectable signals for these eight elements. Our successful samples are “dirty” stalagmites with clear colored lamina (such as C9 and F16-1). If the stalagmite grows in an extremely “clean” environment and is mainly composed of calcium carbonate, other elements are below the detection limit.

Influence Factors

For these detected elements, Sr/Ca ratio of speleothem is generally interpreted as local rainfall variations or precipitation-evaporation conditions (Cruz et al., 2007; Cui et al., 2012; Fairchild et al., 2001; Huang & Fairchild, 2001; Johnson et al., 2006; Treble et al., 2003). Therefore, we specially choose Sr/Ca to discuss its potential influence factors. Previous studies show that flowing paths, varying time of water-rock interaction, composition differences of the soil and bedrock, growth rate and variable extent of prior calcite/aragonite precipitation are the major factors (Fairchild & Treble, 2009; Tan et al., 2014; Wong et al., 2011). In addition, experimental methods could also affect the output results.

Sr and Ca intensities of RM8-2 obtained by three different scanning paths are shown in Figure 8a. Sr and Ca counts range from 150 to 300 cps, 20,000 to 70,000 cps, respectively. The records of parallel paths 3 and 2 display similar variation in the long-term trend, resulting from growth axis measurements. However, output data of path 1 is measured along the diagonal direction of growth axis, showing more variations. The Sr/Ca ratio varies from 0.0027 and 0.00465 with an average value of 0.00335 for path 2, ranging from 0.00317 and 0.00491 with an average value of 0.00377 for path 3, and changing between 0.00397 and 0.00663, with an average value of 0.00472 for path 1 (Li et al., 2019). This kind of counts or ratio discrepancy among three the paths are due to differences between scanning paths (path 1, 2 and 3 were scanned in April 16, 21 and 1, 2018, respectively).

Figure 8. (a) Comparisons of XRF scanning Sr, Ca and Sr/Ca ratio among three scan paths of aragonite stalagmite RM8-2. The corresponding scan path 1 (purple), 2 (yellow) and 3 (green) conducted for RM8-2 are shown in Figure 2b. Orange, blue and black curves are scan results of path 1, 2 and 3, respectively. The light gray solid lines are raw data for Sr/Ca and orange, black and blue curves are 9-points moving average datasets. (b) Comparisons of scanning Sr, Ca and Sr/Ca ratio records of stalagmite C9 in 0.1 mm (blue), 0.2 mm (orange) and 0.5 mm (black) step size with Sr/Ca ratio of C9 measured by plasma - optical emission spectrometry method (green).

Based on correlation analysis, the general trend of Sr/Ca variations is basically the same for three paths. The corresponding correlation coefficient is 0.527 (n = 459, p < 0.01) for path 1 and path 2, 0.545 (n = 459, p < 0.01) for path 2 and path 3 and 0.651 for path 1 and path 3 (n = 459, p < 0.01), which are positively correlated and reflect common long-term trends. However, most smaller-scale variability among the three paths is poorly correlated and indicates that scanning path will affect their results, especially for short-term signals. Scanning off the growth axis can produce very disparate results. Decadal and annual banding pinch out from the center to the edge of the stalagmite, making it harder to distinguish the actual depth and timing of deformed, overlapped and crossed lamina in the outer part of stalagmites. Therefore, it is better not to scan the outer part of samples. The same laminae scanned in different paths of center parts have different depths, thus depth corrections should be made for different paths to enable reliable proxy comparison. In addition, if stalagmites are characterized by complex growth paths or several voids on sample surface, the sample should be segmentally measured along varying growing paths.

The results of Sr and Ca for C9 obtained by CXRF with different scan intervals are shown in Figure 8b. These three records (blue, orange and black line represent the 0.1, 0.2 and 0.5 mm step size records, respectively) all exhibit similar long-term trends. The Ca and Sr intensities vary from 5,000 to 65,000 cps and 50 to 200 cps, respectively. Sr/Ca ratios in three different step sizes show parallel trends as well, ranging from 0.00083 to 0.00378 with an average value of 0.0018. The correlation coefficients are very high between each other with 0.945 (n = 125, p < 0.01) for 0.1 and 0.2 mm interval, 0.927 (n = 125, p < 0.01) for 0.2 and 0.5 mm interval, and 0.937 (n = 125, p < 0.01) for 0.1 and 0.5 mm interval (Li et al., 2019).

As to the three different step size scanning results, we apply the same irradiated area, which is set to 2 × 2 mm; under the same conditions, large irradiated areas smooth the random noise signals. If we want to produce higher resolution data, we should reduce the scanning interval and narrow irradiated area at the same time. However, higher resolution does not necessarily yield better results. The actual situation depends on specific growth characteristics of stalagmite samples. Compared with Sr/Ca ratio obtained by ICP-OES with 1 mm step size, the XRF scanning Sr/Ca ratio displays generally similar in long-term trends, but reveals small-scale variability reflecting background noise.

Comparison With Other Proxies

Elemental concentrations in stalagmites are mostly measured by ICP-OES and inductively coupled plasma mass spectrometry methods. The expense and analytical time of these two measurements impede the production of higher resolution elemental data. XRF scanning makes it possible to obtain high-resolution elemental data with rapid and non-destructive analysis. For aragonite stalagmite XL16, Sr/Ca ratio measured by different methods and isotope proxies (Tan et al., 2018) are displayed in Figure 9. Sr/Ca ratio was between 0.00212 and 0.00488, with an average value of 0.0033. Sr/Ca shows a slow increase trend during 4,800–3,400 years BP, and two decrease-increase cycles with multiple fluctuations during the period of 3,400–1,050 years BP. The XRF scanning data are more consistent with the results of ICP-OES Sr/Ca result (R² = 0.79, whereas the R² among XRF scanning Sr/Ca, δ¹³C and δ¹⁸O are less 0.2), which confirms that XRF scanning can provide a reliable Sr/Ca ratio record for stalagmites, taking the ICP-OES results as the ground-truth datasets.

Figure 9. Comparisons of Sr/Ca ratio of aragonite stalagmite XL16 measured by CXRF scanning method (in orange) with plasma - optical emission spectrometry Sr/Ca ratio (Li et al., 2019, in purple), δ18O (Tan et al., 2018, in blue) and δ13C record (Tan et al., 2018, in violet).

In order to record millennial-centennial abrupt climate changes, the resolution should be 20–200 year/mm at least. Based on dating results, the average resolution is 14 year/mm for stalagmite XL16. Consequently, 0.5–1 mm and 2–10 mm scanning intervals could be selected to record centennial-scale and millennial-scale variations, respectively. Generally, the annual laminae thickness of stalagmites varies from 0.1 to 0.01 mm with the growth rate ranging from 10 to 100 years/mm (Baker et al., 1993; Shopov et al., 1994). Therefore, proxies with 0.5–5 mm scanning interval for most stalagmites can resolve centennial and millennial timescale climate changes.

Tridacna Elemental Results

Similar to stalagmites, tridacna is mostly made up of calcium carbonate. They usually inhabit shallow water of coral reefs in warm seas of the Indo-Pacific region. The skeletons of annually banded Tridacna gigas or fossil giant clam are widely applied to paleoclimate research (Aharon, 1991; Aharon & Chappell, 1986; Elliot et al., 2009; Jones et al., 1986; Mcgregor et al., 2013; Watanabe & Oba, 1999; Yan et al., 2011, 2014, 2017). The δ¹⁸O and Sr/Ca of Tridacna spp. Clams have been used for short-term paleoclimate reconstructions, such as the investigation of past temperature change, seasonality variability, and ENSO variability. Here, we utilize XRF scanning method to analyze tridacna samples.

Ca and Sr can be robustly measured by CXRF, similar to stalagmites. Some other metals (such as Cu, Fe, Sc and Ni) can be detected under different voltages as well (Figure 10). Si in 10 kV can be resolved in some samples (such as MD-4 and A87). The characteristic peak of Si is ∼1.74 keV; not close to the interference of diffraction peaks for Ca (the emission X-ray peak of Ca is at ∼3.7 keV). If the concentration of Ca is not too high, Si can be successfully detected by XRF scanning. The content of Ni and Sc are relatively stable with minor change downward, which mainly reflects noise. The average counts of each element are 120 cps for Ni, 68,000 cps for Ca and 360 cps for Sc. Fe ranges from 240 to 360 cps. Cu, Fe and Sr show multi-cycle variations against the depth, but Cu and Fe show shorter-period cycles compared to that of Sr. The content of Sr varies from 1,200 to 1,560 cps with evident similar large amplitude variations in the lower part of A5 and narrow range fluctuations for the upper 30 mm. These metal elements (Fe and Cu, for instance) need further examination and may be related to tridacna's life activity. These shorter period cycles can be compared with seasonal proxies if possible, which may help determine the cause this scale variations.

Figure 10. XRF scanning elemental results (Si in black of MD-4, Ni in brown, Cu in green, Sr in purple, Fe in orange, Ca in blue and Sc in magenta of A5) of tridacna MD-4 and A5.

Influence Factors

Among these elements, Sr/Ca ratio of tridacna is widely used to reconstruct the past sea surface temperature and seasonality changes. According to research on calcitic stalagmites, the scanning path, settings of current and voltage, scanning step size and radiating area are important factors. Therefore, sample A5 was scanned along two different paths along the vertical direction of annual laminas (on October 30, 2018 and May 29, 2019, respectively. A new X-Ray tube was installed in early May of 2019). Path1 was measured twice with 0.1 and 0.2 mm intervals (on October 29, 2019 and May 29, 2019, respectively).

Sr and Ca content results of A5 obtained by two different scanning paths are shown in Figure 11a. The content of Ca for path 1 is higher (ranging from 65,000 to 80,000 cps) than that for path 2 (from 64,000 to 71000cps). However, the concentration of Sr for path 1 (average vale 1,607 cps) is lower than that for path 2 (average vale 1,424 cps). This is mainly attributed to efficiency emission decay of the two different X-ray tube. According to correlation analysis, they are poorly correlated (R² = 0.0006, p < 0.01). Due to irregular growth laminates and varying lengths of each path, the same depth of different scan paths does not necessarily correspond to the same layer (minimum of path 1 and path 2 connected by dash line correspond to the same laminates; Figure 11a). Therefore, Ca and Sr manifest similar change trends with different values. Hence it is important to perform depth corrections for different paths. Similar to Sr content variations, Sr/Ca manifests unclear interannual variation of the upper 30 mm laminates but very evident changes in lower 60 mm, possibly because the growth rate of Tridacna changes sharply from juvenile phase to adulthood (Elliot et al., 2009; Jones et al., 1986; Romanek & Grossman, 1989; Watanabe et al., 2004).

Figure 11. (a) Comparisons of XRF scanning Sr, Ca and Sr/Ca ratios between two scan paths of A5 (the scan paths are shown in Figure 2c). Green and blue curves are scan results of path 1 and path 2. The dashed lines represent correlation between two paths. (b) Comparisons of XRF scanning Sr, Ca and Sr/Ca ratio in different scan intervals for A5. The orange and blue lines are scan results in 0.1 and 0.2 mm intervals, respectively.

We also scanned path 1 twice in different step sizes. For 0.1 mm interval scan, the irradiated area is set to 1 × 1 mm, while for 0.2 mm interval scan, we use 2 × 2 mm irradiated area. The Sr and Ca intensity results are displayed in Figure 11b. Sr ranges from 1,390 to 1,500 cps and from 1,340 to1540 cps for 0.2 and 0.1 mm interval records, respectively. The average count of Ca in 0.2 mm step size is higher than that of 0.1 mm step size. These Sr/Ca ratio records show similar long-period variations but different short-term fluctuations. The interannual cycles are continuous in the lower 60 mm for 0.1 mm scan interval, but these cycles are not resolved at the 0.2 mm scanning step size, which mainly results from different scanning intervals and irradiated areas applied for the two scans. In addition, the actual scanning intervals of two scans are 1 and 2 mm (width of irradiated areas) rather than 0.1 and 0.2 mm. The boundary of each annual layer is bent with width of each layer varying from place to place. The results of two scans are values obtained from 1 or 4 mm² irradiated areas of the same scan path. Therefore, Sr/Ca ratio curves in different intervals display different variation characteristics in annual timescale. For interannual timescale geological records, the choice of irradiated area is the same important as scanning interval.

Comparison With Other Proxies

The scanning Sr/Ca ratio shows clear interannual variability and has potential to reconstruct paleotemperature changes. Here we compare the scanned results of A5 and A87 with Sr/Ca measured by ICP-OES method (Figure 12). These two samples are scanned with 0.1 mm interval and 1 × 1 mm radiation area. Sr/Ca ratio of A5 determined by XRF scanning and ICP-OES fluctuates from 0.019 to 0.023 and from 1.21 to 1.87 with average value of 0.020 and 1.54, respectively. Scanning Sr/Ca ratio of A87 ranges from 0.004 to 0.017 with average value of 0.012, and ICP-OES Sr/Ca ratio varies from 1.23 to 1.72 (average value 1.52). The growth rate of A5 changes quickly with a decreasing trend from juvenile phase to adulthood. However, the growth rate of A87 shows an opposite slow variation trend with wider annual laminae (>2 mm) in the juvenile phase and narrower annual laminae (<2 mm) in adulthood. The annual cycles for A5 recorded by scanning Sr/Ca ratio are persistent in lower 80 mm and relatively unclear for upper 20 mm. Continuous annual peaks documented by ICP-OES Sr/Ca ratio of A87 correspond well with cycles of Sr/Ca ratio determined by XRF scanning method in lower 80 mm. The correlation coefficient of Sr/Ca ratios measured by XRF scanning and ICP-OES methods are 0.548 (n = 1,000) and 0.511 (n = 654) for A5 and A87, respectively, which are positively well correlated with each other. Therefore, we confirm that XRF scanning method can yield reliable Sr/Ca data for Tridacna, especially for samples with low growth-rate (annual laminae >2 mm).

Figure 12. Comparisons of Sr, Ca and Sr/Ca ratio of tridacna A5 and A87 measured by XRF scanning method with ICP-OES Sr/Ca ratio.

Conclusion

The XRF scanning method is applied to three types of geological archives (loess, stalagmite, and tridacna) to test the feasibility for multi-scale paleoclimate research. Our results reveal that for loess cores and powder samples, 11 elements (Al, Si, K, Ca, Ti, Mn, Fe, Rb, Sr, Zr, and Ba) measured by CXRF and DXRF methods are reliably acquired for orbital- to millennial-scale paleomonsoon reconstruction. Elemental intensities can be influenced by efficiency emission decay of the X-ray tube over time, water content, matrix and grain size effects. Scanning path and step size have little impact on the results, but care should be taken to pretreat the surface of cores before scanning. Our scanning elemental ratios correspond well with other loess proxies (e.g., MGS and MS), illustrating the capacity to generate reliable monsoon proxy records.

The scanning results of stalagmite and tridacna samples show that 8 elements (Ca, Si, K, Fe, Cu, Ni, Sc and Sr) and seven elements (Ca, Sr, Cu, Fe, Sc, Ni and Si) respectively are robustly detected. Among these elements, only Sr/Ca ratio could be used to reconstruct paleoclimate changes in millennial to interannual timescales. Comparative results reveal that efficiency emission decay of the X-ray tube, matrix, scanning path, step size and radiation area have significant effects on element intensities. Comparisons among different scanning paths and step sizes suggest that scanning paths should be selected along or parallel to the growth axis. Depth correction is necessary for different paths to remove the influence of irregular layers. Radiation area should be altered according to scanning interval, especially for tridacna research on interannual-seasonal timescale changes. In addition, scanning Sr/Ca ratio of stalagmites compares well with Sr/Ca ratio determined by ICP-OES and other isotopic proxies (δ¹⁸O and δ¹³C), which shows similar trends at centennial and millennial timescales. For tridacna samples, scanning Sr/Ca is consistent with Sr/Ca ratio measured by ICP-OES method with parallel variation trends at interannual to seasonal timescale. These results confirm that the application of XRF scanning can be extended robustly from marine and lake cores to loess, speleothem and tridacna samples. With proper sampling and scanning strategies, XRF scanning of geochemical features for different geological archives can make an enormous contribution to reconstruct palaeoclimate variability across a range of timescales.

Acknowledgments

We thank Huimin Fan and Xing Cheng for lab assistant. This work was supported by funds from the National Natural Science Foundation of China (41525008), the National Key Research and Development Program of China (2016YFA0601902), and State Key Laboratory of Loess and Quaternary Geology, IEECAS (SKLLQG1808).

Data Availability Statement

Datasets for this research are included in this paper (its Supporting Information) which can be downloaded from the Zenedo open access platform (https://doi.org/10.5281/zenodo.3820931).