Introduction

Portable X-ray fluorescence spectrometry (pXRF), owing to its in-situ, non-destructive, cost-effective, and convenient characteristics, has been widely applied in fields such as mineral resource exploration1,2,3, cultural heritage archaeology4,5,6, and soil quality evaluation7,8. In archaeology, the geochemical data in artifacts are valuable for identifying material types9, tracing raw material sources10,11, and reconstructing early human activity and trade routes12,13,14,15. Although inductively coupled plasma mass spectrometry (ICP-MS) and wavelength-dispersive X-ray fluorescence spectrometry (WD-XRF) are highly sensitive in elemental analysis, they are relatively destructive, costly, and time-consuming, limiting their widespread use on precious archaeological samples16. By contrast, pXRF has become one of the commonly employed non-destructive techniques for elemental analysis in archaeology17,18,19,20, providing an important complementary approach for the compositional study of cultural heritage materials worldwide.

Ancient jade and stone artifacts, as important subjects of archaeological research, have attracted considerable scholarly attention owing to their material properties and cultural significance21,22,23. pXRF is commonly employed for the rapid and non-destructive analysis of major elements in jade and stone artifacts, providing support for the determination of material types. However, its accuracy in the elemental analysis of jade and stone artifacts remains debated, and further systematic evaluation and methodological correction are needed24,25. Yuan et al.26 used pXRF to analyze nine jade artifacts from Tomb No. 1 at Hougudui, Gushi County, Henan Province, and found MgO, CaO, and SiO2 contents ranging from 10.5–12.4%, 10.1–12.6%, and 73.6–76.7%, respectively. Based on Raman spectroscopy characteristics, the samples were identified as being composed of actinolite–tremolite minerals. However, they did not discuss the discrepancies between the pXRF results and the theoretical mineral compositions (MgO, 24.8%; CaO, 13.8%; SiO2, 59.2%), nor did they explore possible causes. Yang et al.27 analyzed 11 jade artifacts from the Dawenkou culture site in Shandong using portable infrared spectroscopy and pXRF. Infrared spectroscopy results indicated that the artifacts were composed of serpentine minerals. pXRF showed SiO2 contents of 34.2–53.5% (mean: 44.1%), slightly higher than the theoretical value of 43.4%, and MgO contents of 21.6–44.8% (mean: 36.9%), lower than the theoretical 43.6%. The authors hypothesized that the discrepancies in Mg and Si contents relative to the theoretical values may be related to element migration caused by long-term weathering; however, this explanation has yet to be substantiated.

For jade artifact samples, pXRF measurements often exhibit certain deviations, which previous studies have largely attributed to elemental migration during weathering. Consequently, dedicated correcting studies for such data remain limited. However, if these deviations are not induced by external factors such as surface alteration, potential systematic errors may compromise the accuracy of lithological identification of jade artifacts. To improve the accuracy of pXRF analysis, correction studies have been conducted on other materials, such as mudstone. Rowe et al.28 noted that mudstone has a complex chemical composition and that different matrices can significantly interfere with X-ray fluorescence signals, causing deviations in quantitative analysis. They developed a correction system incorporating various international standards and internal reference materials to improve pXRF accuracy in mudstone geochemical analysis. Although the correction system covers a wide range of elemental concentrations and is optimized for mudstone’s chemical diversity, it has limitations when applied to pXRF correction of jade and stone artifacts under specific burial conditions. As “non-standard” materials, these artifacts lack certified reference materials (CRMs) for correction. Even with matrix-matched CRMs, subtle matrix differences between the reference materials and actual samples may still lead to analytical deviations.

To address this issue, using the analyzed samples themselves as correction standards is more likely to reflect the elemental composition of their specific matrix, thereby improving analytical accuracy. The correction curve directly represents the response behavior of the actual matrix without relying on the “approximate matching” of CRMs. For example, Conrey et al.29 argued that archaeological materials like obsidian and flint lack suitable solid reference materials, so they used WD-XRF determined chemical compositions of the samples themselves as correction benchmarks. The results indicate that matrix-matched correction can effectively reflect the true characteristics of the sample matrix, reducing interference caused by mismatches with CRMs. However, this study constructed and validated the correction model using WD-XRF data from the same batch of samples, without employing independent samples for external validation, which limits the model’s reliability.

Due to the non-destructive nature of jade and stone artifacts, the conventional powder preparation required for laboratory XRF analysis is difficult to implement, resulting in a lack of standard reference data and posing challenges for pXRF correction studies. To systematically evaluate the accuracy and correction issues of pXRF in the non-destructive elemental analysis of jade and stone artifacts, this study focused on jade and stone artifacts unearthed from the Sanxingdui and Lianhe sites of the Shang–Zhou period in the Chengdu Plain. After confirming sample lithology through mutual verification of multiple methods, including macroscopic examination, thin-section microscopy, scanning electron microscopy with energy-dispersive spectroscopy (SEM-EDS), micro laser Raman spectroscopy (LRS), WD-XRF, and pXRF analysis, we analyze the main factors influencing the accuracy of pXRF test data by comparing the pXRF and WD-XRF results for the same samples. Simultaneously, a correction model was constructed based on the WD-XRF data of the samples themselves to ensure matrix consistency, thereby maximizing the accuracy of the measurements. Finally, the reliability of the corrected pXRF data was further evaluated using independent sample validation. Given that jade and stone artifacts unearthed from sites in the Chengdu Plain since the Shang–Zhou period share similar burial ages and relatively minor variations in environmental conditions (e.g., temperature, precipitation), the obtained specific correction equation is expected to enable effective correction of pXRF data for jade and stone artifacts, even when the total number of such precious archaeological objects is limited and destructive analysis is not permitted, thereby providing valuable reference for the determination of their material types.

The Sanxingdui site (104°11′24″–104°13′12″E, 30°59′38″–31°06′33″N), located on the southern bank of the Jian River in the northwestern part of Guanghan County, Sichuan Province, China, is a significant cultural site from the Shang and Zhou periods (4.4-2.9 ka B.P.) in the Chengdu Plain (Fig. 1a, b). The site covers an area of 12 square kilometers and is considered one of the most remarkable archaeological discoveries of the 20th century30. The unearthed jade and stone artifacts are abundant, finely crafted, and regionally distinctive (Fig. 1d), providing crucial physical evidence for exploring the origins and development of the ancient Shu civilization31,32,33. In 2020, archaeologists discovered the Lianhe Site on the northern bank of the Jian River, approximately 20 kilometers northeast of the Sanxingdui Site. The Lianhe Site (5.0–3.2 ka B.P.) is situated in Group 11 of Yangguan Village, Nanfeng Town, Guanghan County, Sichuan Province (Fig. 1a, b), with central geographic coordinates of 104°14′54.84″E and 31°1′1.24″N, covering an area of approximately 17,000 m2. In addition to pottery and porcelain, the site also yielded abundant jade and stone artifacts from the Shang and Zhou periods (Fig. 1c). These artifacts exhibit diverse shapes and functions, and the raw materials used are notably varied34. It is noteworthy that the jade and stone artifacts unearthed from the two sites serve as tangible materials, providing important and representative samples for assessing and enhancing the accuracy and reliability of pXRF analysis in jade and stone research. The 23 jade and stone artifacts from the Lianhe site and the 13 from the Sanxingdui site together form the core materials for this study. The types, shapes, sizes, and colors of the samples, along with the analytical methods employed, are presented in Table 1 and Fig. 2.

Fig. 1: Location of study sites and cultural remains.
Fig. 1: Location of study sites and cultural remains.
Full size image

a Chengdu Plain where the sites are located. b Locations of the Lianhe and Sanxingdui sites (base map from https://www.geocloud.cgs.gov.cn/). c Ash pit excavation at the Lianhe site. d Jade and stone artifacts from the Sanxingdui site.

Fig. 2
Fig. 2
Full size image

Macroscopic characteristics of the test samples.

Table 1 Testing methods and lithological naming for jade and stone artifact samples from the Lianhe and Sanxingdui sites
Full size table

Methods

Macroscopic examination and microscopic thin-section identification

All samples underwent macroscopic examination, with preliminary lithology naming based on macroscopic features such as color, mineral composition, degree of crystallization, and grain size. A polarizing microscope was used to observe and identify the structure and mineral composition of thin sections of the artifacts, providing precise lithology naming for the samples. The work was carried out in the Sedimentary Geology Laboratory at Chengdu University of Technology using a Nikon LV100POL polarizing microscope from Japan, equipped with a 50 W halogen lamp. A total of 29 rock fragment samples were identified.

Scanning electron microscopy with energy-dispersive spectrometry

The scanning electron microscope allows for detailed observation of the microstructures of samples, and when combined with energy-dispersive spectroscopy, enables the identification of mineral compositions. In this study, four samples from which small fragments could be obtained were analyzed at the State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation, Chengdu University of Technology. The instruments used were a Quanta 250 FEG field emission environmental scanning electron microscope (resolution of 1.2 nm) from FEI, USA, and an INCA X-max 20 energy-dispersive spectrometer (energy resolution ≤129 eV) manufactured by Oxford Instruments, UK, operated at an accelerating voltage of 20 kV.

Micro laser Raman spectroscopy analysis

Confocal micro laser Raman spectroscopy (LRS) is an effective analytical technique for characterizing and identifying the structural and compositional features of artifacts through microscopic analysis, and it is applicable to both homogeneous and heterogeneous samples. In this study, the method was employed to determine and identify the mineral types of seven representative samples. The analyses were conducted using a LabRAM Soleil Raman Microscope manufactured by HORIBA France SAS.

Wavelength-dispersive X-ray fluorescence spectrometry

WD-XRF is a highly effective XRF method for measuring and quantifying low-Z elements (i.e., elements with lower atomic numbers, such as Mg, Al, Si, K, Ca, Ti, Mn, and Fe)35. Moreover, it is one of the most commonly used analytical techniques in pXRF correction studies25,29. In this study, WD-XRF was employed to determine the major elemental compositions of the jade and stone artifacts, including SiO2, Al2O3, Na2O, and Fe2O3, and others. The analyses were performed using a Rigaku ZSX Primus II X-ray fluorescence spectrometer (4 kW, Japan). A total of 16 rock fragment samples were tested, with all measurements conducted at Qingdao Sibada Analytical Testing Co., Ltd., China.

Portable X-ray fluorescence spectrometry

Elemental composition analyses were conducted on all jade and stone artifact samples using the Thermo Fisher Scientific Niton XL3t GOLDD+ portable XRF (pXRF) analyzer (Table 1). The instrument is equipped with a 2 W Ag anode X-ray tube and operates within a range of 6–50 kV and 0–200 μA. Measurements were performed in the “TestAllGeo” mode, covering approximately 30 elements from magnesium (Mg) to uranium (U). All analyses were conducted using an 8 mm diameter spot size (analyzed area ~ 50 mm2).

The range of elements detectable by pXRF is influenced by the measurement duration, with longer acquisition times generally improving analytical precision. Therefore, a measurement time of at least 120 s is typically recommended36,37,38,39. In the practical measurements, it was observed that the analytical results for elemental concentrations stabilized when the acquisition time reached 180 s; therefore, a measurement duration of 180 s was adopted for each test in this study. For pXRF measurements, block samples with fine grains and minimal visible heterogeneity were selected. Whenever possible, analyses were performed on smooth and flat surfaces to ensure close contact between the instrument window and the sample surface, thereby minimizing the influence of sample heterogeneity and surface geometry on the results. In the subsequent data processing, the raw concentrations of Si, Ti, Al, Fe, Mn, Mg, Ca, K, and P were converted into their corresponding oxides, and the oxide contents were normalized to a total of 100% to facilitate comparison. The analyses were conducted at the Sichuan Provincial Cultural Relics and Archaeology Research Institute.

Results and discussion

Multi-method lithological correction of samples

Based on macroscopic and microscopic characteristics, geochemical data, and Raman spectroscopic results, this study identified previously unclassified jade and stone artifacts from the Lianhe site, revealing that they primarily consist of the following lithological types:

Serpentinite samples: Samples LH-17, LH-18, and LH-19 exhibit a green or light green color, with a cryptocrystalline to microcrystalline structure, homogeneous composition, and can be scratched with a knife (Fig. 2). SEM images show that sample LH-17 is composed of a significant amount of foliated minerals (Fig. 3a). EDS analysis indicates that the sample is primarily composed of O (49.7 Wt%), Mg (20.8 Wt%), and Si (19.5 Wt%). Comprehensive identification suggests that the sample is composed of serpentine minerals. Samples LH-18 and LH-19 exhibit similar Raman spectral features, with distinct characteristic absorption peaks appearing near 135 cm⁻¹, 229 cm⁻¹, 375 cm⁻¹, 455 cm⁻¹, and 688 cm⁻¹. These are typical spectral features of serpentine (Fig. 4a). Based on the above characteristics, the lithology of samples LH-17, LH-18, and LH-19 is identified as serpentinite.

Fig. 3: Scanning electron microscopy (SEM) and energy-dispersive spectrometry (EDS) analyses of the samples.
Fig. 3: Scanning electron microscopy (SEM) and energy-dispersive spectrometry (EDS) analyses of the samples.
Full size image

a Leaf-shaped serpentine aggregates observed in sample LH-17. b Barium feldspar and (c) quartz identified in sample LH-23. Panels 1, 2, and 3 show the corresponding EDS spectra for (a), (b), and (c), respectively.

Fig. 4: Raman spectra of the samples.
Fig. 4: Raman spectra of the samples.
Full size image

a Serpentine identified in sample LH-19. b Tremolite identified in sample LH-21. c Quartz inclusion in sample LH-22. d Quartz in sample LH-23 (excessive laser energy caused local heating, resulting in a shift of the quartz Raman peak toward lower energy).

The pXRF test results (Table 2) show that the chemical composition of samples LH-17, LH-18, and LH-19 is dominated by SiO2 and MgO, which is consistent with the primary chemical composition of serpentine minerals. Specifically, SiO2 ranges from 58.3% to 67.9%, and MgO ranges from 23% to 25.1%. Compared with the theoretical chemical composition of serpentine (SiO2 44.1%, MgO 43.6%)40, the pXRF measurements exhibit noticeable deviations, characterized primarily by elevated SiO2 values and reduced MgO values.

Table 2 pXRF test results of jade and stone artifact samples from the Lianhe and Sanxingdui sites (wt.%)
Full size table

Tremolite samples: Samples LH-20 and LH-21 exhibit a milky white to yellowish-green color. No effervescence was observed in sample LH-21 upon the addition of dilute hydrochloric acid, indicating the absence of a detectable carbonate reaction. The two samples exhibit similar Raman spectral characteristics, with distinct absorption peaks observed at 179 cm⁻¹, 223 cm⁻¹, 370 cm⁻¹, 396 cm⁻¹, 675 cm⁻¹, and 1062 cm⁻¹, which are typical spectral features of tremolite minerals (Fig. 4b). Based on the macroscopic characteristics and Raman spectral results, the samples can be identified as tremolite rocks.

The pXRF test results (Table 2) show that the chemical composition of samples LH-20 and LH-21 is predominantly composed of SiO2, MgO, and CaO, which aligns with the main chemical composition of tremolite. Specifically, SiO2 ranges from 67% to 75.7%, MgO from 12.6% to 17.4%, and CaO from 9.6% to 13.4%. Compared with the theoretical chemical composition of tremolite (SiO2 59.2%, MgO 24.8%, CaO 13.8%)40, the measured results also show a higher SiO2 content and a lower MgO content.

Obsidian samples: Sample LH-22 is black with a cryptocrystalline structure, displaying a conchoidal fracture (Fig. 2). The Raman spectrum shows a diffuse envelope peak around 400–1200 cm⁻¹, indicative of a glassy state, characterized by broad peak shapes and low intensity, suggesting that the sample is vitreous. Additionally, a relatively high-intensity vibrational peak is observed at 465 cm⁻¹, suggesting the possible presence of minor quartz inclusions in the sample (Fig. 4c). Based on the comprehensive analysis of its macroscopic characteristics and Raman spectral results, the lithology of the sample is identified as obsidian.

pXRF analysis (Table 2) indicates that the primary chemical component of sample LH-22 is SiO2 (95.7%), followed by Al2O3 (2.4%), with the remaining oxides mostly present at concentrations below 1%. Previous studies indicate that typical obsidian compositions contain approximately 70–77% SiO2 and 12–14% Al2O341. In comparison, sample LH-22 exhibits a higher SiO2 content and a lower Al2O3 content.

Acid-extrusive rock sample: Sample LH-23 is light gray in color and appears relatively loose in texture upon macroscopic observation. SEM-EDS analysis reveals a distinct porphyritic texture, with quartz phenocrysts distributed within the groundmass (Fig. 3c). The sample exhibits well-developed dissolution pores, with minor occurrences of barium feldspar (Fig. 3b), as well as abundant moldic pores formed by feldspar dissolution and clay minerals formed during later alteration. Raman spectroscopy further confirms quartz as the dominant mineral component (Fig. 4d). The rock contains a low proportion of mafic minerals, and its mineral assemblage is mainly composed of quartz and feldspar. Based on macroscopic observation, SEM-EDS, and Raman spectroscopy results, the lithology of this sample is identified as an acid-extrusive rock.

The pXRF data (Table 2) show that the chemical composition of sample LH-23 is dominated by SiO2 (94.1%), followed by Al2O3 (3.3%), with the contents of other oxides all below 1%. In terms of mineral composition, in addition to quartz, the sample also contains a certain amount of feldspar (KAlSi3O8). Therefore, its chemical composition would theoretically be expected to include appreciable amounts of Al and K. However, the actual results show an anomalously high SiO2 content, while the concentration of K2O is lower than theoretical expectations.

In summary, the pXRF results for these jade and stone artifacts show certain deviations in the contents of SiO2, MgO, Al2O3, and K2O compared to the theoretical chemical composition of the constituent minerals, with the discrepancies being particularly pronounced for SiO2 and MgO.

Main factors affecting the accuracy of pXRF data

The factors affecting the accuracy of pXRF data include sample homogeneity, flat surfaces, natural or fresh surfaces, and instrument system errors42,43. In cases where destructive analysis is not feasible, relatively flat and homogeneous sample surfaces were preferentially selected for pXRF analysis to minimize the influence of sample heterogeneity and surface morphology on the results. Nevertheless, sample weathering and potential systematic errors of the instrument may still represent significant factors affecting the accuracy of pXRF data.

WD-XRF analysis typically requires preparing samples into homogeneous powders, which are then fused at high temperatures to form glass discs. This preparation method enables more accurate detection of elemental concentrations representative of the unweathered interior of the sample. In contrast, pXRF analysis primarily probes the near-surface composition, making the results more susceptible to the effects of weathering. To investigate the influence of weathering on the pXRF results, we compared the WD-XRF data with the pXRF data for 16 samples (Fig. 5, Table 3). The results indicate that, in pXRF analyses, 6 samples exhibited slightly lower CaO contents compared with the WD-XRF data, while the remaining 10 showed slightly higher values. For MgO, K2O, Al2O3, and Fe2O3, the pXRF data were generally lower than those obtained by WD-XRF; in contrast, the SiO2 contents were consistently higher (Fig. 5). During the weathering process, the migration ability of the major rock-forming elements in the rocks follows this order: Ca> Sr > Na > Mg > K > (Si, P, Mn) > (Fe, Ti, Al)44,45,46. Although the MgO and K2O contents obtained from pXRF analysis are slightly lower than those measured by WD-XRF, it is worth noting that the contents of Al2O3 and Fe2O3, which are largely unaffected by migration, also show lower values in the pXRF data. This phenomenon suggests that weathering is not the primary factor of the deviation observed in the pXRF results. Since the near-surface elemental data of jade and stone artifacts obtained by pXRF do not exhibit the regular compositional variations expected from weathering when compared with the internal elemental data obtained by WD-XRF, instrument-related systematic error emerges as the primary factor to be considered. Such an error can be mitigated by establishing a mathematical correction model based on jade and stone artifacts as archaeological materials, thereby bringing the analytical results closer to the true values.

Fig. 5
Fig. 5
Full size image

Comparison of the elemental contents of 16 samples analyzed by WD-XRF and pXRF.

Table 3 XRF and pXRF test results for jade and stone artifact samples from the Lianhe site (wt.%)
Full size table

Correction method

To further enhance the reliability and accuracy of pXRF analyses, this study established a mathematical correction model based on data from 16 samples that were analyzed using both WD-XRF and pXRF. The sample size meets the basic requirements of econometric analysis, thereby ensuring the feasibility of linear regression analysis. Specifically, WD-XRF results were used as the reference, with the pXRF data for each element plotted on the X-axis and the corresponding WD-XRF data on the Y-axis. Regression fitting was then performed to derive the functional expressions and correlation coefficients. Since pXRF is unable to detect sodium (Na) due to its low atomic number, Na was excluded from the WD-XRF dataset, and the concentrations of the remaining major elements were normalized to a total of 100%.

The correction curves for the oxides MgO, Al2O3, SiO2, P2O5, K2O, CaO, TiO2, MnO, and Fe2O3 are shown in Fig. 6. The univariate linear regression analysis indicates that the coefficients of determination (R2) for the major elements in jade and stone artifacts vary between the two methods, ranked from highest to lowest as follows: CaO > TiO2 > Fe2O3 > SiO2 > MgO > Al2O3 > K2O > P2O5 > MnO (Table 4). In correlation analysis, a coefficient of determination greater than 0.70 is considered indicative of a significant correlation47,48. In this study, the R² values for CaO, TiO2, Fe2O3, SiO2, MgO, Al2O3, and K2O all exceed 0.7, indicating a strong correlation between the deviations observed in pXRF and WD-XRF measurements for these oxides.

Fig. 6
Fig. 6
Full size image

Correction curves for the oxides MgO, Al2O3, SiO2, K2O, P2O5, CaO, TiO2, MnO, and Fe2O3 in jade and stone artifact samples (using laboratory WD-XRF data as reference values).

Table 4 Correction equations and coefficients of determination for oxides
Full size table

The coefficients of determination for MnO and P2O5 are the lowest. Previous studies have shown that even under full vacuum conditions, reliable detection and quantification of phosphorus (P) using a benchtop ED-XRF is challenging because most detectors exhibit low sensitivity to P35. In addition, the concentrations of MnO and P2O5 in the jade and stone artifacts are both below 1%, which may prevent pXRF from accurately measuring these elements, resulting in significant discrepancies with WD-XRF data and limiting the ability to establish a reliable correlation between the two datasets. Although the coefficients of determination (R2) for MnO and P2O5 are relatively low, these elements occur at very low concentrations in the samples (often below the detection limit) and do not play a decisive role in determining the lithology of the jade and stone artifacts. Therefore, they are not the focus of this study. In contrast, deviations observed for major elements that are critical for lithological identification of jade and stone artifacts (e.g., SiO2, MgO, CaO) between pXRF and WD-XRF measurements show strong correlations, indicating the applicability of the correlation equations established in this study. Based on these mathematical correction equations, the raw pXRF data of jade and stone artifact samples can be corrected to yield elemental compositions that are closer to reference values.

pXRF data correction of samples from the Lianhe site

To evaluate the reliability of the pXRF data after correction, the corrected pXRF results for the jade and stone artifact samples were compared with the theoretical chemical compositions of samples whose lithology had been previously determined using other analytical methods.

The corrected pXRF data (Table 5) show that the SiO2 content of serpentinite samples (LH-17, LH-18, LH-19) ranges from 46.7% to 54%, while the MgO content from 33.2% to 36.3%. For the tremolite samples (LH-20, LH-21), the SiO2 content ranges from 55% to 64.3%, MgO from 19.2% to 25.7%, and CaO from 8.7% to 11.7%. Compared with the theoretical chemical compositions of the primary minerals (serpentine: SiO2 44.1%, MgO 43.6%; tremolite: SiO2 59.2%, MgO 24.8%, CaO 13.8%)40, the elemental contents of the corrected serpentine and tremolite samples are closer to the theoretical values, indicating that the mathematical correction improves the accuracy of the results. Additionally, the CaO content in serpentine samples is consistently below 1%, whereas in tremolite samples it generally exceeds 8%, reflecting the difference in calcium content between tremolite {Ca2Mg5[(Si4O11)]2(OH)2} and serpentine {(Mg, Fe)6[Si4O10](OH)8}. Therefore, the CaO content obtained from pXRF analysis can serve as a valuable elemental indicator for distinguishing between serpentine and tremolite lithologies.

Table 5 Corrected pXRF data of jade and stone artifact samples from the Lianhe site and Sanxingdui site (wt.%)
Full size table

The corrected SiO2 content of the obsidian sample (LH-22) is 89.5%, with Al2O3 as the secondary component at 5.6%, while the contents of other oxides are generally low, mostly below 1%. After correction, the elemental composition of LH-22 is dominated by SiO2 and Al2O3. The SiO2 content remains higher than the typical range reported for obsidian in the literature (70–77%), which may be related to the presence of minor quartz inclusions detected by Raman spectroscopy (Fig. 4c). Meanwhile, the increase in Al2O3 content after correction brings the pXRF data closer to the characteristic elemental composition of obsidian.

The corrected contents of SiO2, Al2O3, and K2O in the acidic extrusive rock sample (LH-23) are 88.1%, 6.5%, and 1.1%, respectively. After correction, the proportions of Al2O3 and K2O both increase. The corrected pXRF data are more consistent with the elemental characteristics corresponding to the mineral composition of the acidic extrusive rock (e.g., quartz and feldspar) described above.

pXRF data correction of samples from the Sanxingdui site

In addition to the jade and stone artifact from the Lianhe Site, this study also uses the mineral composition data (Table 6) and major element composition data (Table 2) of the jade and stone artifact samples from the Sanxingdui site to verify the reliability of the pXRF data correction results. To minimize instrumental bias, the chemical composition data of the Sanxingdui jade artifacts were acquired using the same instrument and under identical analytical conditions as those applied to the Lianhe samples, thereby enhancing the comparability between datasets.

Table 6 Mineral composition of jade and stone artifact samples from the Sanxingdui site
Full size table

The mineral composition identification results indicate that eleven samples (e.g., SXD-20, SXD-52, SXD-11) are dominated by serpentine minerals (Table 6), with their volume percentage mostly exceeding 75%, thereby being classified as serpentinite. pXRF analytical data show that the SiO2 content in these serpentinite samples ranges from 48.2% to 65.6%, while the MgO content from 23.3% to 30.9% (Table 2). Compared with the theoretical chemical composition of the primary mineral (serpentine: SiO2 44.1%, MgO 43.6%)40, the pXRF results exhibit a marked tendency toward elevated SiO2 and reduced MgO values. After applying the correction equations (Table 4) to the pXRF data of jade and stone artifacts from the Sanxingdui site, the corrected SiO2 contents of the serpentinite samples range from 37.5% to 52.2%, and MgO from 33.6% to 42.6%. Compared with the uncorrected values, these results are closer to the theoretical compositions of the primary minerals (Table 5). Minor variations in mineralogical composition among the samples may account for differences in their chemical compositions. For instance, the elevated Fe2O3 content observed in sample SXD-20 may be related to the presence of basic volcanic glass, as identified in the microscopic thin-section analysis (Table 6; Fig. 7a, b). Basic glass typically contains FeO, Fe2O3, MgO, and CaO, which is consistent with the relatively higher Fe2O3, MgO, and CaO concentrations in the pXRF data of SXD-20 compared with other serpentinite samples.

Fig. 7: Macroscopic and microscopic characteristics of the samples.
Fig. 7: Macroscopic and microscopic characteristics of the samples.
Full size image

a, b Sample SXD-20. c Sample SXD-72. d Sample SXD-79. a Under plane-polarized light. b–d Under cross-polarized light.

Samples SXD-72 and SXD-79 were identified as tremolite rock by thin-section analysis (Table 6). Sample SXD-72 appears pale yellowish-white and translucent. whereas sample SXD-79 is light grayish-white with a greasy luster and is less transparent. Both samples are overall homogeneous, with grains too fine to be discerned macroscopically. Thin-section observations reveal that the tremolite minerals in both samples are predominantly colorless to pale yellow, forming fibrous aggregates with contents approaching 100%, and exhibiting second-order blue-green interference colors (Fig. 7c, d). pXRF analytical results show that samples SXD-72 and SXD-79 contain 66.7–67.3% SiO2, 14.7–15.9% MgO, and 14.6–15.1% CaO. Compared with the theoretical chemical composition of tremolite (SiO2 59.2%, MgO 24.8%, CaO 13.8%)40, these results exhibit certain deviations, particularly with higher SiO2 and lower MgO contents (Table 2). After correction, the SiO2 content in samples SXD-72 and SXD-79 ranges from 55.2% to 56.3%, MgO from 22.2% to 23.7%, and CaO from 12.9% to 13.3% (Table 5). Compared with the theoretical values, the corrected pXRF data are more consistent with the chemical composition of tremolite. These results demonstrate that the established pXRF correction equations are likewise applicable to the analysis of jade and stone artifacts from the Sanxingdui site.

Existing issues and outlook

Even after correction, pXRF data retain certain inherent limitations. The most prominent among these is its inability to effectively detect elements with low atomic numbers (particularly those with atomic numbers ≤11), such as sodium (Na). This limitation is especially evident in studies of granitic rocks rich in sodic feldspar. Since sodic feldspar plays a critical role in determining mineral composition and lithological classification in such rocks, the absence of Na measurements constrains comprehensive compositional characterization and, to some extent, compromises the accuracy of lithological identification in granitic rocks.

Building upon the lithological identification of jade and stone artifacts from the Lianhe and Sanxingdui sites through multiple analytical approaches, this study focused on evaluating the data accuracy of pXRF in non-destructive elemental analysis and addressing issues of correction. The systematic analysis demonstrates that the pronounced discrepancies between pXRF measurements and theoretical compositions are primarily attributable to instrumental systematic error, rather than to elemental migration induced by millennia of weathering, as traditionally assumed. Based on the WD-XRF and pXRF measurements, this study established a mathematical correction equation that can be used to correct the pXRF elemental analysis results of jade and stone artifacts from the Shang–Zhou period at the Sanxingdui and Lianhe sites in the Chengdu Plain, thereby supporting lithological identification. Theoretically, this correction equation may also be applicable to jade and stone artifacts from other sites in the Chengdu Plain with similar chronological periods and comparable burial conditions and post-depositional alteration processes. However, apart from the Sanxingdui site, pXRF data for jade and stone artifacts from other sites remain limited. As more data become available, this correction equation is expected to provide a valuable reference for pXRF analyses of jade and stone artifacts across a broader range of sites in the Chengdu Plain. This correction approach allows researchers to obtain reliable geochemical data without the need for destructive sampling to acquire standard reference materials, thereby supporting the scientific identification of jade and stone artifact material types. In the context of the increasing application of pXRF in jade and stone artifact research, the establishment of such a correction equation holds significant importance.