Introduction

In ancient China, the technology of mercury gold gilding is well known as fire gilding (鎏金)1, and it is an age-old process which began to appear from the Warring States Period (476–224 BC) carried out by craftsmen for applying a thin gold layer onto the surface of metal objects, especially bronze. The gold would dissolve into the mercury when the temperature nearly reaches the vicinity of 400 °C, the Au–Hg amalgam was prepared with roughly the consistency of butter. The thermal process produced a well-bonded layer of gold with a carriable amount of residual mercury, whose thickness could range from a few tenths of micron to about one micron2. This technological sequence produces a gilding layer with peculiar features, and the most relevant characteristic is the presence of residual mercury3. By using these amalgam-based methods, a different object could be coated not only for decoration, but also to simulate solid gold or silver object4.

Furthermore, during the long-term interaction with the surrounding burial environment, the specific corrosion products are formed with peculiar structures, they often differ from the compounds naturally grown on the artefacts, this phenomenon is generally attributed to the common degradation processes5,6. In the report of the heavy metal elements distribution in the farmland soils of 6 counties in the western of the Chengdu Plain from Sichuan environmental monitoring center, the primary heavy metals identified are V, Cd and Ni, and As, Co, Cr, Mn, Hg, Mn, Pb and Se present higher than the standard values7,8. Scholars have studied the corrosivity of soil leaching solution in Sichuan plain areas, it is reported that the content of SO42− and Cl− are much higher than the standard values in soils leaching solution, and the conductivity is with a high value (about 332.0 μS cm-1) indicates an easy corrosion9.

In recent years, many studies concerning the characteristics of this gilt-bronze have been published, including research into the microstructure and the componential of the gilding techniques on the Han Dynasty (202 BC–AD 220) bronze fragments10, the corrosion of fire-gilded bronze surface by XPS11, and the surface of naturally corroded gold and silver coated copper-based artefacts by SEM and XPS2.

The surfaces and the interfaces of metal artefacts are key to studying the properties and chemical/physical changes of the ancient materials12. So far, several approaches have been described to probe the depth distribution of elements in XPS measurement, such as angle-resolved XPS (ARXPS), reflected electron spectroscopy (RES), peak-shape analysis (PSA), and ion beam etching techniques13. It has been reported that the ARXPS and RES as non-destructive methods can only give the overall chemical information with the probing depth of 1–4 nm14. Similarly, the PSA techniques can only probe the composition distribution up to 6–8 nm below the surface of the materials, which was on basis of the electron-transport theory through analysing the correlation between the peak shape and XPS signal intensity15. On the other side, ion beam etching has been reported as an important tool for XPS to study composition gradients in the depth, which was not specific to any one kinds of materials13,16. As such, in this paper, ion beam etching XPS has been chosen to investigate the in-depth chemical profile of the gilded layers.

The main aim of this work is to study the ionic migration of the major component gilded layers in the naturally corroded gilt-bronze, which was excavated from the Dafenbao Cemetery (大坟包墓地), Sichuan Province, China. In this paper, we carried out the valence distribution of elements in different part of the gilded layer by using argon (Ar) IBE-XPS, SEM with EDX and element mapping. Furthermore, the ultimate purpose of this study is to investigate the chemical and valence features of the ancient Chinese gilt-bronze artefacts, providing an effective ionic migration process for the major elements extant in the gilded layers, as well as the interactions occurring between the gilt-bronze and the surrounding burial environments.

Methods

Archaeology materials

The three naturally corroded gilt-bronze samples were excavated from the Dafenbao Cemetery, Jiangkou Town, Pengshan County. The cemetery is located about 100 m to the west of Minjiang River. The outcrop layer in Pengshan County is relatively new and the quaternary system is widely distributed on the surface of the soil, the cretaceous is found in Jiangkou Town and older strata are buried deep in the belly17. The artefacts were all dated to the Eastern Han Dynasty (AD 25–220). Due to the high humidity of the burial environment, the gilt-bronze artefacts were significantly corroded, thus, there was not any metal extant in the gilt-bronze samples, and the gilding layer fell off by touch. The three samples were prepared from the bottom of the artefacts where the gilded layers still can be clearly observed on the bronze, shown in Fig. 1.

Fig. 1: Sampled artefacts and sampling spots.
figure 1

The samples from the artefacts and the places gilded layers with bronze where were sampled: a sample 1, b sample 2 and c sample 3.

Full size image

Before the experiment, a surface cleaning procedure was carried out using distilled water to wipe off the soil and then washing with ethyl alcohol to remove any carbon contamination.

Instruments and methods

Ion beam etching XPS was chosen for the gilded layers so that the layers could be analysed to a greater depth. To investigate the interlayers and interfaces between the gilded layer and bronze, study the ions, the content, and the ion valence and the ionic migration in the gilded layer, we conducted IBE-XPS analysis in the surfaces of the gilded layer (0 min), in the interlayers of the gilded layer (3, 5, 10 min), and the part near the joint interfaces between the bronze and the gilded layer (30 min), aiming to answer the questions mentioned above. Since the IBE-XPS analysis needs to cut the gilt-bronze relics to get an area larger than 3 × 3 mm, it is a destructive analysis even though the damage is very small. It is difficult to find the balance between conservation and scientific analysis of the artefact. Therefore, we have tried our best and carried out parallel experiments using three samples that were excavated in the same grave, the same type, cut from the same area from three objects in order to explain the information of the gilded layers.

Chemical and metallurgical features were observed by means of Zeiss Evo18 scanning electron microscope (SEM), energy-dispersive spectroscopy (EDX) and element mapping was carried out with Oxford Ultim Max, the acceleration voltage was set to 15 keV, without any gold or carbon coating.

The XPS investigation was performed by using an AXIS Supra (Kratos), a full spectrum was acquired with monochromatic Al kα (1486 eV) X-rays at 5 mA emission, 15 kV HT, 160 eV pass energy and the analysis area of 700 × 300 μm. And the element spectrum was acquired with the X-rays at 10 mA emission, 15 kV HT and 20 eV pass energy. The films were etched using a 45°-inclined argon ion beam in the XPS for up to 15 min at a power of 4 kV, 140 μA, the etched area was 3 × 3 mm. The instrument was calibrated to gold metal Au 4f (84.0 eV).

XPS data was analysed by using the CasaXPS programme, peaks were fitted with a Shirley background prior to component analysis.

Results

Characterization of the gilded layers surface by electron microscopy, EDX spectroscopy and element mapping on the samples

The element mapping images of sample 2 are shown in Fig. 218, the elements Au, Ag, Hg, Cu, Sn and O were detected. Among them, Hg was commonly found in the gilded layer, which is a characteristic feature of mercury gilding. The shape of the peeled-off gilded layer can be clearly profiled, proving the Au and Hg were only extant in the gilded layer. Ag was detected in the gilded layers, but the edge of the lack part of the gilded layer was not clearly defined, it may have been unintentional due to the underdeveloped gold purification technology of ancient China, or on purpose to adjust the colour of the layer by adding the silver into the gilding mercury amalgam19. The lack of places of the gilded layer shows a matching shape of the Sn, we can infer that Sn mainly exists in bronze.

Fig. 2: SEM, EDX and element mapping images.
figure 2

The images of SEM, EDX and element mapping of sample 218.

Full size image

Many studies into bronze corrosion mechanisms have reported that the sequence of corrosion is Pb > Cu > Sn20. It is worth noting that, the degradation of gilt-bronze is induced by galvanic coupling between bronze and gold11. The metal of the bronze was entirely corroded but the gilded layer was in a better condition, this may be due to the electrochemical reaction between the gilded layer and the bronze. Moreover, due to the oxidation and corrosion of the metal core in the burial, the crystals constituting the corrosion layer nucleated and grew on the surface of the bronze, mainly in the form of copper oxides, copper carbonates, and copper chlorides at the expense of core metal and leading to copper dissolution21. Therefore, the irregular distribution of Cu was related to the corrosion of the bronze alloy, shown in Fig. 2 element Cu. The detection of Sn suggests that the bronze was alloyed with Sn, and the appearance of Sn (in Fig. 2, element Sn) on the surface of the gilded layer may come from the corrosion and the Sn migration from the bronze under the gilded layer. On the other hand, the mass distribution of O suggests that both the bronze and gilded layers were all significantly deteriorated and oxidized.

The SEM-EDX data of the three samples are listed in Table 1, the wt% sigma of EDS was below 3% (signed in Table 1), and the SEM and EDX images of samples 1 and 3 are shown in Fig. 3. The elements Au, Ag, Hg, Cu, Sn, O can be observed in the surface of the gilded layers. Hg was commonly found in the gilded layer of all three samples, which suggests that all the samples were gilded by the fire gilding techniques.

Fig. 3: SEM and EDX images.
figure 3

The images of SEM and EDX of a sample 1 and b sample 3.

Full size image
Table 1 The bulk element compositions on the surface of the samples
Full size table

To estimate the etched time of the gilded layer, the thickness of the gilded layer should be detected. Since the gilt-bronze artefacts were severely corroded, the gilded layer samples which we used in this experiment were all blurry with a covered white oxide object, thus, the thickness of the gilded layer could not be observed from these cross-section samples. In addition, all the materials are vaguer after polishing the intersecting surface of the gilt-bronze samples. Therefore, the thickness data of the gilded layers were from the peeled-off layers and from the other gilt-bronze artefacts, which were from the same grave and type as the three samples we prepared in this investigation.

As shown in Fig. 4, the thickness of the peeled-off layers ranges from 441.5 to 1.143 nm. Considering the average etching time of the objects22 the maximum etching time was set to 30 min.

Fig. 4: SEM images of the peeled gilded layers.
figure 4

The thickness of the peeled gilded layers: (a), (b), and (c).

Full size image

Characterization of the gilded layers by ion beam etching XPS

The wide survey scan spectrums of the three samples are shown in Fig. 5. On account of the complex burial environment, and the gilt-bronze objects have been severely corroded, elements Au, Ag, Cu, Sn and Hg can be observed in the spectrums (Fig. 5, Table 1). However, the fire gilding technique was involved in making an Au–Hg amalgam of gold mixed in mercury which can spread easily on the surface of objects and form a thin gold film after the mercury was evaporated using a charcoal fire23.

Fig. 5: XPS survey spectrums of the three samples.
figure 5

a Sample 1, b sample 2, c sample 3, and d the percentage composition of the element Cu/Sn/Ag/Au/O in the gilded layers in different etching times.

Full size image

Since Hg was the solvent of the Au-Hg amalgam, this paper focuses on the other major elements in the gilded layer during the long burial time. In the wide survey spectra of the three samples (in Fig. 5a–c), Au, Ag, Cu and Sn can be observed clearly with different etching times and were chosen to be prepared High for resolution XPS spectrums analyses.

In Fig. 5a–c, the peaks in XPS survey spectrums have been signed, the peaks mainly belong to elements Au, Ag, Cu and Sn. In spite of a surface cleaning procedure carried out using distilled water to wipe off the soil and then washing with ethyl alcohol to remove any carbon contamination before the experimental, the elements C and Si are detected in the surfaces of the gilded layers of the XPS survey spectrums (signed in sloping shape in Fig. 5). The percentage composition of the 5 elements (Au, Ag, Cu, Sn, O) are shown in Fig. 5d, it is worth mentioning that the percentage composition was only consisted of the 5 elements but not the whole compounds of the gilded layers. Besides, the surfaces (0 min) of the three samples are all severely oxidated with a high percentage content of O1s peaks, and the other elements were relatively low. The percentage content of each sample and intensity variation trends were similar when the etching time came to 3, 5 and 10 min, confirming that these parts were the interlayers. Due to the intergranular corrosion of the interface between the gilded layer and bronze, the percentage content of O1s at 30 min raised. The details are described below.

Oxidation and corrosion resistance are the hallmarks of Au, making it a popular material for millennia. Based on the stability of gold, the HR-XPS of gold was set as important reference data to initially study the gilded layers. The values of the binding energy, full width at half maximum (FWHM) and relative area of the component peaks used to fit the data are compiled in Table 2. The corresponding Au4f XPS spectrums of sample 1 (Fig. 6a) and sample 3 (Fig. 6c) confirmed a stable clean Au extant in the gilded layers without any oxidation in all the etching times. But in the images of sample 2, the XPS spectrum has been changed with new features besides the Au 4f5/2 and Au 4f7/2 at the etching time of 30 min, shown in the enlarged image in Fig. 6b. A well-defined peak appears at 86.5 eV, this peak is more likely assigned to the oxidation of ionic Au species24,25. This additional component is attributed to the presence of low-coordinated atoms that appear on the Au interface as a result of interface restructuring during oxidation. An additional feature at low binding energy at 83.9 eV26, and a wider FWHM data 1.1 (shown in Table 2 and font bold), also means that this area was oxidated and the intergranular corrosion occurred. This additional component is attributed to the presence of low-coordinated atoms that appear on the Au interface as a result of interface restructuring during oxidation27.

Fig. 6: High-resolution XPS spectrums and linear graph of Au.
figure 6

High-resolution XPS spectrums for a sample 1, b sample 2, c sample 3 with fitting, d the linear graph of the fitting area of the Au 4f7/2 peak for the three samples.

Full size image
Table 2 The results of the Au 4f7/2 and Ag 3d5/2 peaks, the binding energy, the FWHM and the area of the fitting peaks
Full size table

As a consequence, the etching at 30 min of sample 2 was mostly positioned at the joint place between the gilded layer and bronze, which achieved the designated objectives. The significant data of sample 2 was set to be the priority object. However, due to the unstable burial environment of the gilt-bronze, sample 1 and sample 3 were still valuable in comparison work.

Furthermore, the linear ship of the fitting area for the Au 4f7/2 peak which was relevant to the content of gold was shown in Fig. 6d. Thus, the results suggest that all three samples have a lower content on the surface of the gilded layer (etching 0 min), the content rises in the interlayers with a similar intensity (etching 3, 5 and 10 min). The content of the Au was reduced at etching 30 min, and the intensity was almost identical to the surface, it is a signal for the end of the gilded layers and a strong proof that the etching has reached the joint areas at 30 min.

The results of Figs. 6 and 7 combined with Table 2 suggested that the majority of the gold and silver atoms should exist in a nonvalenced valent state for the prepared samples in the interlayers of the gilded layer28. The surface chemistry, including stoichiometry and contaminations of the etching time in 0 and 30 min were dramatically changed after long-term burial29, the oxidation was expected to contribute to the Au 4f7/2 and Ag 3d5/2 peak shifts to higher binding energies30. In addition, the silver was extant to be Ag (I) oxides (Ag2O), since the binding energy of Ag 3d5/2 was near the 367.8 eV ± 0.131. In the etching at 30 min of sample 2 (Fig. 7b enlarged image), the binding energy Ag 3d5/2 showed a complicated shift to 367.8 eV with a dependent peak in 370.2 eV, we inferred that it was also the signal that the ending part intergranular corrosion occurred of the gilded layer, and it showed a very low content of silver with the contamination of the burial soil environment.

Fig. 7: High-resolution XPS spectrums and linear graph of Ag.
figure 7

High-resolution XPS spectrums for a sample 1, b sample 2, c sample 3 with fitting and d the linear graph of the fitting area of the Ag 3d5/2 peak for the three samples.

Full size image

In Fig. 7d, the Ag shows a high content intensity on the surface of the gilded layer. Then the content reduced with the etching going on, in samples 1 and 3 the slope of the silver linear graph was similar in the etching at 3, 5, 10 and 30 min. But in sample 2 the content of Ag was only similar in the etching at 3, 5 and 10 min in the interlayers, the content of Ag plummeted at 30 min, compared to the intensity before. In ancient China, due to the limitations in gold purification techniques, natural gold often contained a certain percentage of silver, and Ag was usually detected in trace elements of bronze researches11. Noticeably, the samples we have analysed are severely corroded and unable to obtain trace elements data from the bronze, as a result, we cannot rule out the possibility of trace content of Ag from the bronze. The source of the silver requires further investigation in future studies. Therefore, in this paper, we only focus our discussion on the silver observed within the gilded layers.

Combined with the Au analysis above, the mechanism of the silver migration process in the gilded layers may be that, the silver would migrate at a uniform rate into the interlayer with some Ag (I) ingredients remaining, then collecting on the surface of the gilded layers and oxidating to Ag (I) with a higher value.

Through studying the corrosion of the ancient Chinese gilt-bronze artefacts, we can see that the gilded layer has not suffered too much corrosion, but the substratum bronze was severely corroded with the intergranular and α-phase corrosion, the principle corrosion products were identified as Cu2O1. But on the surface of the ancient gilt-bronze, it has been confirmed that the main exogenous element from the burial environment is Cl with O, C and a limited contribution of Si and Al32.

The images in Fig. 8a–c of the Cu 2p3/2 high-resolution XPS spectrums suggested that the main valence was Cu (II) with the significant satellite peak shown as red in the images on the surface (etching time 0 min) of the gilded layers. The main peak of CuO is accompanied by a satellite peak at about 9 eV higher binding energies, and this is absent in the case of Cu2O33. The d shell of Cu2O is full hence the screening via a charge transfer into the d states is not possible and this is the reason for the absence of a satellite peak in Cu2O spectra34.

Fig. 8: High-resolution XPS spectrums and linear graph of Cu.
figure 8

High-resolution XPS spectrums for a sample 1, b sample 2, c sample 3 with fitting and d the linear graph of the fitting area of the Cu 2p3/2 peak for the three samples.

Full size image

As the etching proceEDX, the Cu 2p3/2 XPS main valence of the interface was Cu (I), with evidence of a binding energy value at 932.5 eV ± 0.1, shown in Table 3 and peak 1. The binding energy of the main peak Cu 2p3/2 photoemission signal could be subjected to an overlapping of the Cu (0) and Cu (I) oxidation states, the binding energy value is ambiguous for Cu. Cu (0) and Cu (I) species are difficult or even impossible to distinguish using the Cu2p XPS spectrum35. But in Archaeological materials, which have been buried for thousands of years, it is unlikely to be Cu (0).

Table 3 The results of the Cu 2p3/2 peaks, the binding energy, the FWHM and the area of the fitting peaks
Full size table

As shown in Fig. 8d, the linear graph of the fitting area of the Cu 2p3/2 peak prepares an opposite migration process than the silver. The low content on the surface of the Cu may be due to the artefact’s burial environment, the Cu (II) is very easily released in the soil36.

Figure 9 presents the Sn3d XPS spectrums and the linear ship of the content in different etching times for the three samples. The values of the binding energy, FWHM and relative area of the component peaks used to fit the data are compiled in Table 4. The components Peak 1 at 486.5 eV ± 0.1 and Peak 2 at 485.5 eV ± 0.1 of Sn 3d5/2, are assigned to oxidized Sn (IV) and metallic Sn (II), respectively37,38. Besides, due to the binding energy of Sn3d shifts of between 1 and 2.5 eV during the formation of surface tin oxides39,40, the increased value of the binding energy at 487.1 eV of Sn 3d5/2 on the surface (etching time 0 min) for the three samples attributed to the oxidation.

Fig. 9: High-resolution XPS spectrums and linear graph of Sn.
figure 9

High-resolution XPS spectrums for a sample 1, b sample 2, c sample 3 with fitting and d the linear graph of the fitting area of the Sn 3d5/2 peak for the three samples.

Full size image
Table 4 The results of the Sn 3d5/2 peaks, the binding energy, the FWHM and the area of the fitting peaks
Full size table

Furthermore, gradually increasing linear ships of the Sn contents were apparent in Fig. 9d, confirming that the Sn was initially from the bronze and migrated to the surface of the gilded layer during the long burial time.

Discussion

In this work, the gilded layers were studied by SEM and ion beam etching XPS, which were sampled from three naturally corroded gilt-bronze artefacts from the Eastern Han Dynasty. The SEM-EDX showed that fire-gilding techniques were used to gild the objects, and the main elements of the gilded layers were Au, Ag, Cu and Sn with heavy oxidation.

This study marks the first time that ancient Chinese gilded layers have been analysed by IBE-XPS in the interlayers and surfaces, inferring the ionic migration process. And it is the first time this method was applied to gilt-bronze objects excavated from the Han Dynasty. Since the oxidation was expected to contribute to shifting to higher binding energy, the wide survey XPS spectrums showed that the surface of the gilded layers contained more O1s content. Combined the major component result of the EDX, Au4f, Ag3d, Cu2p and Sn3d were chosen to prepare the HR-XPS analyses. Provided new data information for the study of gilt-bronze corrosion, especially the surfaces and interlayers of the gilded layers.

In summary, the HR-XPS results showed the Au (0) was stable, the etching time at 30 min for Au4f was confirmed to have reached the joint interface between the gilded layer and the bronze of sample 2, and assigned to be Auδ+. Due to the galvanic coupling reaction between bronze and gold, the mechanism of silver migration process in the gilded layers was as follows: The silver would migrate at a uniform distribution through the interlayer with Ag (I) ingredients remaining, then collecting together on the surface of the gilded layers and oxidating to Ag (I), shifting to a higher binding energy value. The copper and tin detected in the gilded layers originated from the migration of the bronze. The copper in the gilded layers was extant in Cu (II) and Cu (I) form, distributed across the surface and the interfaces of the gilded layers, respectively. The gradually increasing linear ships of the Sn contents indicate that the Sn originated from bronze, and exists largely as Sn (IV) with some Sn (II) in the interlayers. Moreover, the Cu (II) and Sn (IV) are very easily released into the soil, which demonstrates a low content in the surface of the gilded layers in the HR-XPS spectrums.