Introduction

Degradation of materials is a ubiquitous yet harmful complex process, the rate and extent of degradation usually depend on the composition and microstructure of the material and environmental factors1,2,3. Ancient cultural relic materials are no exception4, since they often undergo long burial periods prior to archeological excavations, lasting from hundreds to thousands of years. During this period, various physical and chemical interactions occur between these materials and the natural environment. The degradation process may be slow, but it is unavoidable5,6.

Chinese porcelain has always been an important trade commodity on the ancient Maritime Silk Road, and a large number of ceramics have also been discovered in underwater archeology in recent years. These precious cultural heritages serve as evidence of human cultural exchanges and economic trade, as well as possess high historical, cultural, and research value. Although porcelains are generally considered to have good corrosion resistance and can be maintained for a long time, they often suffer various degrees of damage in harsh marine environments, such as contamination, lower gloss, surface roughening, flaking, crazing, and even breakage7,8,5h), indicating that Fe is probably from the external environment.

Fig. 5: The contaminants in sample No. 06.
figure 5

a and b are OM and SEM images of the polished cross section, respectively; c yellow substances in a crack: rectangular area in b; d the contaminants in a crack near glaze-body interface: rectangular area in b; e–h are the EDS map** results of d.

Full size image
Table 3 Chemical compositions of the marked positions shown in Fig. 5 (in wt%).
Full size table

For sample No. 07, most of the glaze layer is clear and transparent and has many bubbles, as shown in the OM and SEM images (Fig. 6a and b). A few irregularly shaped holes were found near the surface. A small number of needle-like or short column-like anorthite crystals were distributed at the glaze-body interface, and short column-like anorthite crystals were also observed near the irregular-shaped holes, as shown in Fig. 6c and d (EDS analysis results are shown in Table 2).

Fig. 6: OM and SEM images of sample No. 07.
figure 6

a and b are OM and SEM images of the polished cross section, respectively; c a bubble and a few irregular shaped holes near the surface; d short column-like crystals in the glaze.

Full size image

When observed by the naked eye, it was evident that sample No. 07 contained many white spots on the ridges and rim, as well as a deteriorated glaze. The OM and SEM images of the region with white spots are shown in Fig. 7a and b, respectively. It can be observed that there are numerous craters on the glaze surface, and the residual glaze in the craters presents a porous microstructure, as shown in Fig. 7c and d, likely formed after a long period of seawater corrosion. In the magnified SEM images (Fig. 7e and f), densely spaced plate-like crystals and submicron porous microstructures in the surrounding glaze were observed. Although the composition is difficult to acquire because the EDS test signal of the crystal inside the crater was blocked, it can be inferred that these plate-shaped crystals are anorthite, based on the Raman spectrum acquired at the crater position (Supplementary Fig. 7). In addition, the nearby porous microstructure is considered to be a corroded phase-separated structure because previous studies showed that it is always accompanied by anorthite in the ancient Chinese calcium glaze17. Therefore, the vessel shape of porcelain is also an important consideration. The corrosion state at the ridges and rim of No. 07 is more serious than in other areas, and it has even completely peeled off. The thickness of the glaze at the ridges and rim is often thinner than that of other areas; consequently, it is more likely to be affected by the composition of the body during the firing process to form an anorthite crystal-phase separation structure, which exacerbates the corrosion in these areas (Fig. 7).

Fig. 7: Corrosion morphology of sample No. 07.
figure 7

a and b are the OM and SEM images of the seriously corroded region on the glaze surface, respectively; c a crater on the glaze surface; d the plate-like crystals and nearby phase-separation structure in a crater; e and f are the framed area in d: the porous glass phase, BSE backscattered electron image, SE secondary electron image.

Full size image

Matte glazes

Sample Nos. 08, 09, and 10 were all matte glazes, and they were corroded more severely than the transparent glaze. The XRD diagram of the glaze surface of No. 08 (Fig. 8) shows that the diffraction peaks of quartz, diopside, and anorthite are very prominent. Many long rod-like diopside crystal clusters spread on the glaze surface (Supplementary Fig. 8 and Note 1).

Fig. 8: XRD diagrams acquired on the glaze surface of Nos. 08 and 10.
figure 8

The diffraction peaks of quartz and anorthite exist both in the two samples, while diopside exist only in No. 08.

Full size image

It can be seen in the OM image of the polished cross section (Fig. 9a) that many cracks penetrated the glaze layer and extended into the body. The SEM image (Fig. 9b) shows that in addition to many unmelted quartz grains and bubbles of different sizes, a variety of crystals are densely distributed in the glaze. Figures 9c and e show the magnified SEM images of the glaze-body interface, where there are densely spaced short column-like crystals, and the EDS analysis results (Table 2) show a chemical composition compatible with anorthite, which is in agreement with the XRD conclusions. In the area proximal to the glaze surface (Fig. 9d), there are more crystals <10 μm, which have a rhomboid or irregular quadrilateral cross section, in addition to short column-like anorthite crystals. The composition of position P1 in Fig. 9d (the EDS analysis result is in Table 2) is compatible with diopside, supporting the XRD results, indicating that the glazes contain diopside crystals. In addition, a small number of crystals with a high content of P and Ca were found in the glaze (Fig. 9f), which is presented as high-contrast regions in the backscattered electron image. The corrosion morphologies of sample Nos. 08 and 09 are shown in the supplementary material (Supplementary Fig. 8, Fig. 9, Note 1, and Note 2). Figure 9e and Supplementary Fig. 9c showed that the droplets with higher contrast were dispersed in the matrix. According to previous studies on the crystallization-phase separation structure in the ancient glaze, it can be inferred that the droplet phase and matrix are rich in CaO and SiO2, respectively17,40. The phase separation structure in the glaze is more prone to corrosion, and many neoformed microstructures, mainly quintinite-group minerals, formed in the cracks and pores of No. 0941.

Fig. 9: OM and SEM images of sample No. 08.
figure 9

a and b are the OM and SEM images of the polished cross section, respectively; c and d are the magnified SEM images of the glaze-body interface and the area close to the glaze surface, respectively; e SEM image of the rectangle region in c: short column-like anorthite crystals and phase-separation structure; f SEM image of the rectangle region in d: crystals with elliptical cross section.

Full size image

Many diopside and anorthite crystals are distributed on the surface and inside of the glaze of Nos. 08 and 09, along with the phase-separated structure. The crystallization-phase-separation structure is a relatively common structural feature in ancient Chinese high-fired calcium glazes, and its formation is simultaneously affected by kinetic factors such as the firing schedule and thermodynamic factors determined by the chemical composition of the glaze17,40. The uneven distribution of these two crystals indicates the complexity and heterogeneity of the glaze microstructure. In terms of chemical composition, the MgO content in the glaze of No. 08 was higher than that of all the transparent glazes (Supplementary Fig. 2) and higher than most Longquan celadon glazes during the Yuan dynasty42. Many new microstructures (relative to the pristine glaze microstructure) were formed owing to the serious corrosion of No. 09 under natural conditions, leading to observed changes in the chemical composition, such as the MgO content in the glaze of No. 09 reached 16.31 wt%, while the corrosion of the glaze caused the low SiO2 content.

From the perspective of physical chemistry, these glazes can be approximately represented by the CaO–MgO–Al2O3–SiO2 quaternary system. When the content of Al2O3 was approximately 15 wt%, pyroxene and plagioclase were obtained as the main crystalline phases43. Considering the composition, these Longquan celadon glazes are not fully crystalline, allowing for the formation of a transparent glaze at sufficiently high firing temperatures. During the firing process, the Ca–Mg silicate crystalline phase can be formed at approximately 1000 °C, and with increasing temperature, the crystals gradually melt44. However, the firing temperatures of Nos. 08 and 09 are only 1215 °C and 1190 °C (Supplementary Table 3), which are significantly lower than those of Nos. 04 and 06 (1245 °C and 1265 °C, respectively), and they exhibit transparent glazes. The lower firing temperature kept the pyroxene and feldspar crystals precipitated during the heating process without being completely melted. At the same time, the crystallization during the heating process is also reflected in the uneven distribution of crystals in the glaze. Therefore, chemical composition and firing schedule together affected the microstructure of matte glazes. This type of celadon ware with a matte glaze was not deliberately made by the kiln workers but was more likely to be low-quality products that appeared in the mass-production process. Although they retained market value in overseas trade, the selection of raw materials and the control of the firing process were not refined.

The glaze of No. 10, with poor transparency, was seriously corroded, and it contained many white spots, as observed from OM (Fig. 10a). The SEM image of the polished cross section (Fig. 10b and c) shows a crack and a nearby hole in the glaze, in which there are densely spaced plate-like crystals (Fig. 10d). The glass phase around the crystal was corroded, exhibiting a porous microstructure. It can be observed by comparing the secondary electron image (Fig. 10e) and the backscattered electron image (Fig. 10f) that the phase-separated droplets with higher contrast were corroded to form a porous microstructure. In another position, there are many interconnected craters near the glaze surface (Fig. 10g), in which many anorthite crystals (EDS result is listed in Table 2 and Raman spectrum is shown in Supplementary Fig. 7) and porous microstructures, left by the nearby phase-separated droplets being corroded, can be discovered (Fig. 10h and i). The corrosion microstructure inside the glaze layer also reflects the role of cracks in the corrosion process. The cracks provided a fast channel for seawater to enter the glaze, thereby accelerating the corrosion process.

Fig. 10: Corrosion morphology of sample No. 10.
figure 10

a and b are the OM and SEM images of the polished cross section, respectively; c SEM image of the rectangle region in b: the corrosion hole near a crack; d SEM image of the plate-like crystal in the hole and the corroded glass phase nearby; e and f are the secondary electron image and backscattered electron image of the porous glass phase in d, respectively; g SEM image of the polished cross section: the area with many interconnected craters near the glaze surface; h SEM image of a crater in the glaze; i SEM image of the anorthite crystals and the corroded glass phase nearby in a crater.

Full size image

From the perspective of microstructure, the matte glaze in Longquan celadon is more complicated. Many crystalline phases and phase-separated structures could also have a significant impact on the corrosion process. Therefore, one cannot only focus on the corrosion of the glass phase in the glaze. As shown in Supplementary Fig. 9, Fig. 9e, and Fig. 10f, the CaO-rich droplet phase with higher contrast was dispersed in the SiO2-rich matrix with lower contrast in the glazes of Nos. 08, 09, and 10. More nonbridging oxygen atoms in the CaO-rich phase made the silica-network structure more open and unstable, so it was more likely to be corroded by seawater. The dissolution of the CaO-rich droplets caused the formation of a porous microstructure, leading to an increase in the local specific surface area and resulted in more seawater entering into the local submicron pores, accelerating the disintegration of the SiO2-rich continuous phase41. Finally, many large craters with a size of tens of microns were formed with the shedding of the anorthite crystals, which seriously damaged the esthetics of glaze. A schematic of the process is shown in Fig. 4. In addition, a large number of diopside crystals on the surface of the glaze did not seem to be severely corroded (Supplementary Fig. 8), which may be attributed to its good chemical stability45. It is worth noting that the glazes with a large number of crystals such as Nos. 08 and 09 are similar in microstructure to glass ceramics, all of which belong to the multiphase systems, despite the difference in crystal size and homogeneity. The glass-ceramic system is considered to have unique advantages in immobilizing radioactive fission products for long-term disposal, but the multiphase system also makes it more difficult to study the corrosion resistance46,47. This type of ancient glaze may be used as an archeological analog to study the long-term corrosion behavior of multiphase systems in natural environments. Within our study, the results indicate that phases with poor chemical stability are more likely to be preferentially corroded.

Furthermore, the corrosion of porcelain occurs not only in the glaze but also in the body. The body of ancient porcelain in China is mainly composed of crystalline phases such as quartz and mullite, as well as a glass phase and a small number of pores16,48. They can also be corroded under the reaction of prolonged exposure to seawater. On the one hand, the cracks penetrating the glaze layer extended into the body, so the seawater could directly enter the inside of the body, bringing in external species and forming new substances during the corrosion process (Fig. 5g); however, the body may be directly exposed to the external marine environment when porcelain wares are shattered or the glaze is completely corroded, providing different corrosion paths, as shown in Supplementary Fig. 10.

In conclusion, this study yielded insights into the degradation process of several ancient glazes with similar compositions but different microstructures under the same conditions and discussed several factors that affect the long-term degradation behavior of the glazes in the marine environment. Based on a comparative analysis of the different samples, the following conclusions were obtained: (1) the chemical composition of ancient porcelain glazes may change after long-term corrosion in a marine environment, reflecting the alteration of the pristine microstructure and the formation of new microstructures; (2) the formation of a large number of crystals in the matte glaze is related to the composition and the lower firing temperature; (3) the degradation of the glazes is affected by the physical damage as well as the chemical corrosion. Due to the severe physical damage, the glaze surface became rough and lackluster, which also promoted the chemical corrosion process. The combined effect of them may lead to more severe degradation. (4) The transparent glaze mainly experienced the process of dissolution and reprecipitation, resulting in the formation of a corrosion layer of spherical aggregates; (5) the severe corrosion of the matte glaze resulted from the lower chemical stability of the anorthite crystallization-phase-separation structure. Among them, the CaO-rich droplet phase and the crystal-glass interface layer dissolved more easily and were preferentially corroded in the marine environment; (6) the corrosion process of porcelain in the marine environment is affected by internal factors, such as the glaze composition, microstructure, and shape of the ware, in addition to external environmental factors.

Methods

Analytical techniques

The chemical compositions of the bodies and glazes were quantitatively examined using EDXRF (energy-dispersive X-ray fluorescence, Bruker M4 TORNADO). The body was tested on a polished cross section, and the glaze was tested on both the natural surface and polished cross section. Bruker M4 TORNADO was equipped with a 30-W (50 V, 600 μA) X-ray tube, and the diameter of the X-ray focus was 25 μm. For quantitative analysis, the calibration process involved several steps. First, a series of 13 standard samples was formulated using standard geological minerals. Calibration curves were created for each element in the standard series, representing the relationship between the intensity of the element characteristic peak and the element concentration. The EDXRF data were calibrated according to the calibration curves. Analysis was carried out using the software accompanying the instrument.

X-ray diffraction (Bruker D8 DISCOVER) using Cu Kα radiation (40 kV, 40 mA) was applied to investigate the phase components of the glaze surface. Analyses were performed directly on the surface of the fragments.

Macroscale morphologies of the glaze surfaces and polished cross sections were observed using an optical microscope (Keyence VHX-2000). The micromorphology and semiquantitative microarea chemical compositions of the glaze surface and polished cross section of the samples were characterized using field-emission scanning electron microscope (FEI Magellan400) equipped with EDS. Before the SEM–EDS analysis, the samples were coated with platinum (for No. 06) or carbon (for all other samples).

The firing temperature of the body was measured using a dilatometer (NETZSCH DIL 402 C). To prepare the test samples, the glazes were grounded, and the bodies were cut into pieces of 5 × 5 × 25 mm3. Due to sample-size limitation, only some of the samples that met the requirements were tested. The heating rate of the dilatometer was 5 °C/min, and the approximate value of the firing temperature of the sample was obtained according to the inflection point in the thermal expansion curve49,50. The firing-temperature results are listed in Supplementary Table 3.

Micro-Raman spectroscopy (Horiba XploRA one) was applied to examine the glaze on both the natural surface and polished cross sections. The Raman spectral signal was recorded from 100 cm−1 to 3500 cm−1, using a 532 nm excitation radiation source (25 mW). For glass phase, each Raman spectrum was acquired in 10 accumulation cycles, and the integration time for each one is 20 s. The wavenumber calibration of the peak position was at a ×50 objective with a piece of crystal silicon. A confocal microscope (Olympus BX41) was employed to focus the laser beam on the test areas with ×10 and ×50 objectives. The baseline was subtracted using the LABSPEC software, and the smooth deconvolution of the spectra was performed using the Origin software51,52.

Celadon samples

In this study, 11 typical celadon shards with varying degrees of degradation from the Dalian Island shipwreck of the Yuan Dynasty, provided by the Fujian Museum, were investigated. The samples were immersed in deionized water for desalination after the excavation.

The bodies of 11 samples were all gray, and the glaze colors were green or yellow-green. There are large discrepancies in appearance and preservation of the samples, which can be divided into four categories: bright as new (Nos. 01–05), lacking luster (Nos. 06 and 07), a matte glaze (Nos. 08–10), and underfired (No. 11). Sample photos, appearance descriptions, and the analysis results of the firing temperatures are listed in Supplementary Table 3.