Introduction

Sauce glazed ware, one of the most representative Fe-based glazed wares, had played a key role in the history of Chinese ceramics1. It first appeared during the late Eastern Han Dynasty (184–220 CE) in Eastern China (Jiangsu and Zhejiang regions), developed significantly during the Sui and Tang Dynasties (581–907 CE), and matured during the Northern Song Dynasty (960–1127 CE)1. During the Song and Jin dynasties (960–1279 CE), plenty of renowned kilns produced sauce glaze wares, including Ding kilns and Jingxing kilns (Hebei province), Yaozhou kilns (Shaanxi province), Dangyangyu kilns, Duandian kilns, and Xinan kilns (Henan province), Jian kilns (Fujian province), etc.2. The sauce wares produced by these kilns were all based on iron as a colorant. The color of these wares could vary from yellowish- to reddish-brown, and the hue could vary with the angle of the incident light3.

Many studies have been dedicated to the analyses of the chemical composition and the microstructure of ancient Fe-based sauce glazes in an attempt to understand the origin of their characteristic color2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19. It is a commonly accepted viewpoint that the colorant element in sauce glazes is mainly iron ions. Zhang et al. applied XRD on the sauce glazes from Yaozhou kilns manufactured during the Song Dynasty and found that the glaze could be divided into two main layers. The first layer at the surface is around 10 microns thick, reddish-brown in color, and contains well-developed feathery iron oxide crystals. The second is brown and black and contains many hematite (ɑ-Fe2O3) and magnetite (Fe3O4) crystals2. The chemical color of these crystals was believed to be the main factor to the glaze appearance. However, the laboratory X-ray diffraction data have a strong background signal and cannot accurately determine the phase of the iron crystals in the glazes. In 2019, Sciau et al. exploited micro-Raman spectroscopy (μ-RS) to analyze the crystals in a sauce glazed sherd from Ding kilns and observed densely packed dendritic ε-Fe2O3 crystals in micron size distributed at the surface of its glaze4. Subsequently, similar discoveries have also been reported in sauce wares from Chinese northern and southern areas including Yaozhou kilns (Shaanxi province)5,6,7, Qingliangsi archaeological sites (Henan province)8, Xiao kilns (Anhui province)9, Jingdezhen kilns10 and Qilizhen kilns (Jiangxi province)11, Deqing kilns (Zhejiang province)12, and Jian kilns (Fujian province)13,14,15,16,17,18,19,20. However, the crystal chemical color in the glaze is still unable to establish a direct relationship with the appearance of the sauce glaze, which varies from yellowish brown to reddish brown. Therefore, Guan et al.14 and Wang et al.13 detected Al ion substitutions in these ε-Fe2O3 crystals, leading to darkening these crystals. These reports provide a satisfactory relationship between these iron oxides and the yellowish-brown glaze appearance of Yaozhou and Jian wares.

Besides, Ren et al. 6 and Wang et al. 13 further precisely observed the morphology of ε-Fe2O3 crystal dendrites in the glazes of Jian kilns and Yaozhou kilns after etching the surrounding glassy matrix, revealing the origin of the high gloss of the glaze. Moreover, Wang et al. observed that micronsized ε-Fe2O3 and hematite crystals were present in the yellowish-brown and reddish-brown glazes in replicated Yaozhou wares, respectively19. Though the different nature of the crystals, they present a similar plam-leaf shape composed of nanosized flakes, which can coherently diffract light and produce different structural colors at different angles. However, such a situation is whether suitable to explain the color generation of ancient sauce glazes should be verified. Moreover, most analyses focused on the crystals in the glazes surface with deepness of several microns and little known about the distribution of crystals in the whole glaze. Only Wang et al. used focused ion beam-scanning electron microscopy (FIB-SEM) analyzed the sauce glazed wares from Qilizhen kilns (Jiangxi province) and observed three layers of ε-Fe2O3 crystals11.

In this work, typical fragments of sauce wares excavated from different kilns over different Chinese provinces and dated from the Song Dynasty were analyzed. The morphology, structure, and distribution of the crystals in their glazes were systematically documented by combining optical microscopy, X-ray fluorescence spectroscopy (XRF), X-ray diffraction (XRD), micro-Raman spectroscopy (μ-RS), scanning electron microscopy coupled with energy-dispersive spectrometer (SEM-EDS). Moreover, the crystal distribution throughout the entire glaze layer were further analyzed by FIB-SEM. The results obtained with these techniques are used to discuss the origin of the color in the sauce glazes

Methods

Archaeological background

The 78 sherds of typical sauce glazed ware selected and excavated from the seven kilns and an archaeological site located in five different Chinese provinces, shown as Fig. 1. Table 1 gives the sample labels and corresponding provenance, i.e., Hunyuan kilns, Jingxing kilns, Dangyangyu kilns, Xi’an kilns, Lushan kilns, Yaozhou kilns, Yingou archaeological site and Luhuaping kilns, which are located in Shanxi, Hebei, Henan, Shaanxi, and Fujian provinces, respectively. All samples present a relatively homogeneous reddish-brown or yellowish-brown appearance, illustrated in Fig. 2.

Fig. 1
Fig. 1
Full size image

The geographical distribution of kilns from north to south of China: Hunyuan kilns (Shanxi province), Jingxing kilns (Hebei province), Dangyangyu kilns, Xi’an kilns (Ganquanci kilns and Chenggun kilns), and Lushanduandian kilns (Henan province), Yaozhou kilns and the Yingou archaeological site (Shaanxi province), and Luhuaping kilns (Fujian province), abbreviated as HYY, JXY, DYY, GQC, CGY, LSDD, YZY, YG, and LHP, respectively.

Table 1 Archaeological information about the samples
Full size table
Fig. 2: Fragments images.
Fig. 2: Fragments images.
Full size image

Selected samples dated to the Song Dynasty (960-1279 CE) from different kilns: a JXY-01 (Jingxing kilns, Hebei province), b DYY-02 (Dangyangyu kilns, Henan province), XA-GQC-05 (Xinan kilns-Ganquancun archaeological site, Henan province), XA-CGY-10 (Xinan kilns-Chengguan archaeological site, Henan province), YZY-BG-01 (Yaozhou kilns, Shaanxi province), YG-07 (Yingou archaeological site, Shaanxi province), LS-DD-10 (Duandian kilns, Henan province), SJ-LHP-05 (Jian kilns-Luhuaping archaeological site, Fujian province).

Optical images of the glaze surfaces (Fig. 3) show that their color varied from light yellowish-brown to reddish-brown. The samples JXY-01, DYY-02, XA-CGY-10, YZY-BG-01, and YG-07 present few reddish brown spots on the glaze, which is otherwise uniformly yellowish-brown to orange-yellow. The samples XA-GQC-05, LS-DD-10 and SJ-LHP-05 display a much larger number of spots that cover the sample surface uniformly, giving rise to an overall reddish-brown color. Specifically, the spots on the surface of XA-GQC-05 are reddish brown with silver in the center, those on LS-DD-10 are reddish brown with strong gloss under ring light, and those on SJ-LHP-05 range from light brown to black, with a large number of unmelted particles. Besides, the cross-section observations (Fig. S1) show that the glaze is composed of a surface brown layer sitting on an underlying black layer, and the color of bodies varies from brown to black.

Fig. 3: Optical images of samples.
Fig. 3: Optical images of samples.
Full size image

Optical observation of the glaze surfaces of a JXY-01, b DYY-02, c XA-GQC-05, d XA-CGY-10, e YZY-BG-01, f YG-07, g LS-DD-10 and h SJ-LHP-05.

Sample preparations

Etched surfaces were prepared with the alkali etching method to remove the glassy matrix without degrading the embedded crystals in order to avoid its influence on the crystals analysis. The experimental details related to this procedure are described elsewhere6.

Analytical techniques

The samples were systematically investigated by nondestructive optical microscopy, XRF, μ-RS, and XRD prior to the etching of their surfaces. Etched fragments were then analyzed in addition by SEM-EDS. Furthermore, FIB-SEM technique was carried out to investigate the three-dimensional microstructure of the crystals in the glaze surfaces.

Optical observation of the glaze surfaces was performed using a KEYENCE optical microscope coupled with a VHX-7000 surface depth observation system. The images were taken with magnifications ranging from X20 to X2500. The surface coloration and crystal distribution in the glazes were investigated using a ring light sources, respectively.

The elemental composition of the glaze surfaces was measured using a HORIBA XGT-7200 V energy-dispersive X-ray fluorescence spectrometer (XRF). The incident X-ray beam was 1.2 mm in diameter and had an energy of 30 keV. At least three points were analyzed on each glaze to assess their homogeneity, with an acquisition time of 120 s per point.

XRD measurments were performed using a TD-3500 diffractometer. The voltage of 40 kV and current of 30 mA were selected. The diameter of the beam on the sample was around 1 × 10 mm. Each pattern was acquired in 10 h to ensure relevant accumulation with the required angular step size.

μ-RS analyses were carried out using a Renishaw inVia spectrometer. A CW 532-nm solid-state laser was used to provide Raman excitation. A 2-3 μm laser spot size was selected to focus on the particles in micron size. The 1.31 mW laser power was used to avoid any thermal influence on the particles spectra. The Raman range of 100-2000 cm−1 was performed to entirely observe the Raman features of the particles in glazes.

SEM images were obtained using a Hitachi High Resolution Field Emission SEM (Hitachi SU8010). The size of the electron spot was approximately 0.8 nm at 15 kV. The sample surfaces and the cross sections were coated with a thin platinum film to avoid the charge aggregation effects. Backscattered electron imaging was operated at a voltage of 10 kV.

Serial images of the crystals in the glazes were acquired usinga Crossbeam 540 FIB-SEM (Zeiss, Germany). The analysed details were described elsewhere11. The final SEM images were obtained in situ with a resolution of 6 nm at 2 kV accelerating voltage and a 180 pA electron current in backscattered electron imaging mode. For each sauce glaze sample, about 600 slice SE images were obtained. The grayscale contrast segmentation of these images was carried out by Dragonfly software 2024, and the 3D structure information of ε-Fe2O3 crystal was reconstructed and visualized11.

The MetaVue VS3200 noncontact imaging spectrophotometer (X-rite) and the Agilent Cary 7000 universal measurement spectrophotometer (UMS) were used to observe the glaze appearance and color variations at different angles, respectively. For the X-rite and UMS, the size of the measurement spot was selected as 12 and 8 mm, and the resolution was 10 and 1 nm interval, respectively. The reflective spectra with a range of 400–700 nm were recorded. Moreover, using UMS, a set of multi-angle spectra recorded at different incident angles with intervals of 10° ranging from 10° to 60° and scattering reflective spectra recorded at different reflective angles with intervals of 2° ranging from ±10° at an incidence angle of 30° of the glaze surfaces of the samples were obtained using S- and P-polarized incident lights, respectively.

Results

Macro-analyses of elemental and mineral compositions of glazes and bodies

The chemical composition of the glazes was measured by XRF, and the analytical results are illustrated in Table 2. All samples present similar SiO2 (67.3-69.5 wt%) and Al2O3 (15.3-16.7 wt%) contents, except for slightly lower SiO2 (~64.7 wt%) in the glazes of Luhuaping kilns, one of the southern kilns. Na2O, MgO, K2O and CaO in the glazes of all samples also show similar contents, ranging from 0.5-1.0 wt%, 1.2-1.6 wt%, 2.4-3.5 wt% and 2.8-4.0 wt%, respectively, except for slightly higher MgO (~2.2 wt%) and CaO (~5.5 wt%) in the glazes of Hunyuan kilns. The amounts of the colorant elements, i.e., TiO2 and Fe2O3, in the glazes of all samples, vary similarly from 0.6 ~ 0.9 wt% and 5.4 ~ 8.8 wt%. Noteworthy, Fe2O3 contents in the glazes of samples CGY-01 ~ 11 ( ~ 7.5 wt%) and LHP-01 ~ 12 ( ~ 8.8 wt%) are slightly higher than the others.

Table 2 Analytical results (wt%) obtained by lab-XRF on the glaze surfaces of the samples
Full size table

Table 3 shows the elemental composition of the bodies. The wares from the northern kilns and Yingou archaeological site present typical “high aluminum and low silicon” northern regional characteristics, whereas the wares from LHP kilns display typical “high silicon and low aluminum” southern regional features2. All samples contain relatively high K2O (1.6-2.8 wt%) and low CaO (0.2-1.0 wt%), as well as low Na2O and MgO contents, lower than 1 wt%. The colorant elements in the bodies are mainly TiO2 and Fe2O3, ranging from 1.3 to 1.7 wt% and 2.0 to 4.6 wt%, respectively, except for significantly higher Fe2O3 (~11.9 wt%) in the LHP wares. The main reason is that there are many laterite mines in Shuiji County, Nanping City, Fujian Province21, which were rich in iron and exploited for manufacturing ceramics.

Table 3 Analytical results (wt%) obtained by lab-XRF on the bodies of the samples
Full size table

The principal component analysis (PCA) method was further applied to treat the data of Tables 2 and 3. The precise information concerning raw materials of each kiln was illustrated as Fig. 3. The glazes (Fig. 4a) present no apparent regional divergence, only that the higher MgO and CaO in Hunyuan glazes and higher Fe2O3 in LHP glazes distinctively set them apart from the others, in agreement with the results in Table 2. Interestingly, significant regional features were detected in the bodies (Fig. 4b). The significant highest Al2O3 and Fe2O3 contents in DYY and LHP bodies, respectively, set them apart from the others, consistent with Al2O3 ~ 30.8 wt% for DYY and Fe2O3 ~ 11.9 wt% for LHP bodies in Table 3. The relatively higher Al2O3 and Fe2O3 contents in JXY and HYY bodies, respectively, locate them slightly apart from the others, consistent with Al2O3 ~ 27.0 wt% for JXY and Fe2O3 ~ 4.2 wt% for HYY bodies in Table 3. The slightly higher CaO contents in some YZY, YG and CGY bodies set them apart. The difference caused by the CaO amount could be negligible due to its content being less than 1 wt%.

Fig. 4: Two-dimensional correspondence chemical composition analyses.
Fig. 4: Two-dimensional correspondence chemical composition analyses.
Full size image

a The glaze surfaces and b the bodies of the wares from different kilns.

XRD analyses were further carried out to investigate the mineral composition of the glazes, illustrated in Fig. 5. The characteristic peaks from the pattern of samples JXY-01, DYY-02, XA-GQC-05, XA-CGY-10, YZY-BG-01, YG-07 and LS-DD-10 are around 18.95°, 27.05°, 27.86°, 30.22°, 33.25°, 35.33°, 36.82°, 38.45°, 40.57°, 41.70°, 46.22°, 49.57°, 53.23°, 59.07°, 61.42° and 63.50°, which corresponds to the (002), (120), (112), (013), (112), (130), (131), (004), (132), (212), (133), (142), (134), (006), (205) and (330) reflections of ε-Fe2O3 crystal recorded on oil spot and rabbit hair glazes from Jian kilns recorded by Dejoie et al. 15,22,23. Similar features are observed on the other samples. In addition to the characteristic peak of ε-Fe2O3, SJ-LHP-05 also shows the characteristic peaks of quartz24.

Fig. 5
Fig. 5
Full size image

XRD spectra of the glazes of all samples.

It is noteworthy that compared to the reference pattern of ε-Fe2O3 crystal (PDF#89-7047), the XRD patterns recorded on the different samples are shifted from it to different degrees. The peak heights, full widths at half maximum (FWHM) and shifts compared to the reference pattern of six typical diffraction patterns are displayed in Table S1. The results show that the peak positions of samples DYY-02 and YG-07 are at the same positions compared to those of the reference, while the peaks of JXY-01, XA-GQC-05, XA-CG-10 and YZY-01 are shifted toward the higher angles, and the peaks of LS-DD-10 and SJ-LHP-05 are shifted toward lower angles. Such variation might be explained by the doping in ε-Fe2O3 crystals reported by previous works11,12. Besides, the height and FWHM of the six peaks of different samples are varied. The strongest peaks of DYY-02, XA-CGY-10, YZY-BG-01, SJ-LHP-05 and YG-07 were at crystal plane (200), whereas the strongest peaks of JXY-01, XA-GQC-05 and LS-DD-10 were at crystal plane (132), indicating the different crystalline orientation.

Crystals in the bodies were investigated by XRD, and the results are shown in Fig. 6. The spectra of the bodies of all samples present typical patterns of mullite (PDF#42-1501), quartz (PDF#46-1045), and cristobalite (PDF#39-1425) except for CGY-10, which only presents mullite patterns. It is consistent with analytical results in Table S3: the body of CGY-10 presents the highest Al2O3/SiO2 ratio, close to 1:2. The contents of the crystals vary for each ware. The primary phase in DYY-05, YZY-01, and YG-17 is mullite, whereas quartz and cristobalite are mainly detected in HYY-06, JXY-01, GQC-05, LSDD-09, and LHP-11. Interestingly, the ratio of quartz and cristobalite varies for these samples: high quartz and low cristobalite contents for JXY-01, GQC-05, LSDD-09 and LHP-11 and low quartz and high cristobalite contents for HYY-06. Since transforming quartz to cristobalite requires increasing temperature25, the lower ratio of quartz and cristobalite for the HYY-06 body indicates a higher firing temperature.

Fig. 6
Fig. 6
Full size image

XRD patterns of the bodies of the samples.

Micro-analyses of morphology and structure of crystals in the glazes

The glaze surfaces of different samples were observed by SEM, as shown in Fig. 7. The surfaces of JXY-01 (Fig. 7a), DYY-02 (Fig. 7b), XA-CGY-10 (Fig. 7d), YZY-BG-01 (Fig. 7e), and YG-07 (Fig. 7f) are uniformly covered with a layer of dendriform ε-Fe2O3 crystals several with sizes of several to tens of microns. The average sizes were around 3.62, 3.24, 1.98, 3.23 and 2.39 μm, respectively, illustrated in Table S2. Interestingly, XA-GQC-05 (Fig. 7c), LS-DD-10 (Fig. 7g), and SJ-LHP-05 (Fig. 7h) present slightly different morphologies of ε-Fe2O3 crystals with larger sizes of tens of microns. In addition, several polygonal crystals in the center of the ε-Fe2O3 crystals were observed in the surface of the XA-GQC-05 sample. SEM-EDS analyses of the polygonal crystals (Fig. 8) show that the polygons mainly contain Mg and Fe, which could be identified as mafe-Fe spinel.

Fig. 7: SEM observation of the glaze surfaces.
Fig. 7: SEM observation of the glaze surfaces.
Full size image

a JXY-01, b DYY-02, c XA-GQC-05, d XA-CGY-10, e YZY-BG-01, f YG-07, g LS-DD-10, h SJ-LHP-05.

Fig. 8: SEM-EDS analyses of the polygon magnesioferrite crystals in Fig. 4c.
Fig. 8: SEM-EDS analyses of the polygon magnesioferrite crystals in Fig. 4c.
Full size image

a SEM observation and the corresponding elemental distribution: b Mg, c Fe and d Si.

The glaze surfaces were furhter etched with 5 mol L−1 sodium hydroxide solution at 60°C for 6 h in order to remove the surrounding glassy matrix and accurately observe the morphology of the ε-Fe2O3 crystals. Simialr to previous reports6,13, Fig. 9a displays that these dendritic crystals in JXY-01 are composed of numerous lamellar crystals in size of hundreds of nanometers with an offset (d1) of around 110 nm. The dendrites in sample DYY-05 (Fig. 9b) are composed of several comb-organized secondary branches separated by approximately 230 nm (D2), each of which is composed of offset stacked flake-shaped crystals with an offset (d1) of around 110 nm (d2). Similarly, the dendritic ε-Fe2O3 crystals in the other samples (Fig. 9c, e–h) are also composed of secondary branches separated by distance ranging from 170 to 330 nm. The corresponding offset stacked flake-shaped crystals with an offset varying from 100 to 110 nm, displayed as Fig. 9d–h.

Fig. 9: SEM observation of the glaze surfaces after being etched.
Fig. 9: SEM observation of the glaze surfaces after being etched.
Full size image

a JXY-01, b DYY-02, c XA-GQC-05, d XA-CGY-10, e YZY-BG-01, f YG-07, g LS-DD-10, h SJ-LHP-05.

The glaze cross-sections were also precisely analyzed by SEM to observe the overall distribution of ε-Fe2O3 crystals in the whole glaze. The SEM observations (Fig. 10a-f) of JXY-01, DYY-02, XA-GQC-05, XA-CGY-10, YZY-BG-01, YG-07, present similar morphology, i.e., the crystals distributed in multi-layers. The crystals in the first layer, with a size of several to tens of microns, are distributed from the surface to 1-2 μm deep, whereas the crystals in the second and third layers are small to sub-micron scale. Micro-Raman analyses from the glaze surface to in glaze deep (pt01-04 in Fig. 10a) of JXY-01, DYY-02, GQC-05, CGY-01, and BG-01 show that crystals in the first layer, second, and upper part of the third layer present similar Raman features (Fig. 11 curve pt01-03), which are consistent with characteristics of ε-Fe2O3 crystals15 and previous XRD results (Fig. 5). Nevertheless, the Raman spectrum recorded at the glaze in deep (Fig. 11 curve pt04) presents typical Raman features of spinel, in agreement with magnetite26. These results are consistent with previous reports of Yaozhou sauce-glazed wares5. Besides, crystals in LS-DD-10 (Fig. 10g) and SJ-LHP-05 (Fig. 10h) are mainly distributed in the glaze surfaces and continue to tens of microns deep. The corresponding Raman spectra (Fig. 11 curve pt05 and 06) of crystals in the glaze surfaces present typical ε-Fe2O3 features15, whereas the spectrum (Fig. 11 curve pt07) of crystals in deep glaze display magnetite features26.

Fig. 10: SEM observation of the glaze cross-sections.
Fig. 10: SEM observation of the glaze cross-sections.
Full size image

a JXY-01, b DYY-02, c XA-GQC-05, d XA-CGY-10, e YZY-BG-01, f YG-07, g LS-DD-10, h SJ-LHP-05.

Fig. 11: Raman spectra recorded from the glaze surface to in glaze deep of samples.
Fig. 11: Raman spectra recorded from the glaze surface to in glaze deep of samples.
Full size image

a JXY-01 and b SJ-LHP-05, corresponding to the pt01-04 in Fig. 10a and pt01-02 in Fig. 10h.

To precisely observe the morphology of crystals in the three layers, FIB-SEM analyses were performed to extract the accurate information of the crystals. Figure 12a confirmed that these ε-Fe2O3 crystals in the glaze could be divided into three layers. The first layer was composed of dendritic crystals with several microns in size (Fig. 12b). These dendritic crystals present a highly ordered dendritic structure composed of comb-organized secondary branches of hundreds of nanometers along a main trunk with a length of around 3 μm, consistent with SEM observations (Fig. 9e). The second layer is composed of spherical or irregular particles with tens to hundreds of nanometers in size (Fig. 12c), with a thickness around 3-4 μm. The third layer contains dendritic ε-Fe2O3 crystals around 100–200 nm in size (Fig. 12d). Two larger crystals consisting of four or five branches and one large crystal consisting of eight branches in the third layer are extracted, shown as Fig. 12e and f, respectively.

Fig. 12: Multilayer structures of ε-Fe2O3 crystals in sauce glaze.
Fig. 12: Multilayer structures of ε-Fe2O3 crystals in sauce glaze.
Full size image

a Three layers of these ε-Fe2O3 crystals in the glaze, b crystals in the first layer, c crystals in the second layer, d crystals in the third layer, e two larger crystals consisting of four or five branches and f one large crystal consisting of eight branches in the third layer. The whole reconstruction volume is 13 μm (x) × 10 μm (y) × 3.3 μm (z) for YZY-BG-01.

Reflective spectroscopic analyses

The reflective spectra of the glaze surfaces of all samples were recorded and illustrated as Fig. 13. The YZY-BG-01 had an obvious reflective peak around 560 nm, and the color tended to be yellowish brown, while the other samples had an obvious reflective peak around 580 nm, and the color tended to be reddish-brown.

Fig. 13
Fig. 13
Full size image

Reflective spectra (left image) of all samples and the corresponding first derivative curves (right image).

Multi-angle reflection spectra recorded at different reflective angles with intervals 2° ranging from 60 ± 10° at an incidence angle of 30° of the glaze surfaces of samples YZY-BG-01 and LS-DD-10 were analyzed by P-polarized incident lights and S-polarized incident lights, as shown in Fig. 14 and Fig. S2, respectively. Both YZY-BG-01 and LS-DD-10 presents visible angular heterochromatic features under P-polarized incident lights. The spectra of YZY-BG-01 (Fig. 14a) recorded from 60° to 70° present an obvious peak around 592 nm, consistent with the yellowish-brown color. The spectra recorded from 50° to 58° with two reflective peaks positioning around 569 and 612 nm, respectively, corresponding to the reddish-brown color. Conversely, the glaze of sample LS-DD-10 (Fig. 14b) displays different visible angular heterochromaticity, that is, the spectra recorded from 50° to 58° present an obvious peak around 596 nm, consistent with the yellowish-brown color. The spectra recorded from 60° to 70° (Fig. 14b) present two reflective peaks locating around 544 and 614 nm, respectively, corresponding to the reddish-brown color. Interestingly, the glazes of the samples present no obvious reflective peaks with S-polarized incident lights (Fig. S2). Besides, multi-angle reflective spectra of the glaze of sample LS-DD-10 at different incidence angles (10° range from 10° to 60°) under S-polarized light also present no obvious reflective characteristics (Fig. S3a). Conversely, under P-polarized light, the spectra (Fig. S3b) present two obvious peaks locating around 550 nm and 615 nm, respectively, corresponding to reddish-brown color.

Fig. 14: Scattering reflective spectra recorded at different reflective angles.
Fig. 14: Scattering reflective spectra recorded at different reflective angles.
Full size image

The reflective angles with the intervals 2° range from 60° ± 10° at an incidence angle of 30° of the glaze surfaces of samples YZY-BG-01 a and LS-DD-10 b were obtained using P-polarized incident lights, respectively.

Discussion

The glazes of fragments from northern areas present similar elemental compositions (Table 2 and Fig. 4), i.e., 67-69 wt% of SiO2, 15-17 wt% of Al2O3, 2-3 wt% of K2O, 2-3 wt% of CaO, ~1 wt% of MgO, and 6-7 wt% of Fe2O3, except for the Hunyun fragments, which have higher MgO (~2 wt%) and CaO (~5 wt%). Conversely, the glazes of fragments from southern kilns present the highest Fe2O3 contents (~9 wt%) and relatively low TiO2 amounts (~0.6 wt%). Besides, the bodies of all fragments present significant regional characteristics1, i.e., relatively higher Al2O3 (24-30 wt%) and lower SiO2 (62-66 wt%) for fragments produced in northern areas, whereas relatively lower Al2O3 (~20 wt%) and higher SiO2 (~63 wt%) for fragments produced in southern areas. Remarkably, the bodies of fragments excavated from Luhuaping kilns present significantly higher Fe2O3 with content of around 11 wt%, consistent with previous reports27.

The crystal types in sauce glazes are mainly ε-Fe2O3 crystals and a few magnesioferrite and quartz. However, the reflective edge of pure ɛ-Fe2O3 powder is around ~550 nm28, which is not consistent with the reflective edge of the glaze surface (~560 nm and ~580 nm) (Fig. 7). The main reason could be attributed to the iron ion in the ε-Fe2O3 crystal being partially replaced by other ions, confrimed by XRD and Raman results shown as Fig. 5, resulting in a darkening of the crystal color11,12. Besides, the variation of glaze darkness is not only related to the ε-Fe2O3 crystals, but also to the thickness of the crystallization layer and magnetite crystals in deep glaze detected by μ-RS (Fig. 11). FIB-SEM observations (Fig. 7) confirmed that the ε-Fe2O3 crystals in the glaze could be divided into three layers, consistent with the previous reports concerning Qinglizhen wares10. The top layer is composed of ε-Fe2O3 crystals with a large size of hundreds microns, and the middle and third layers are composed of ε-Fe2O3 crystals with a small size of several to hundreds of microns. The thickness of the crystal layers and the proportion of crystals in each layer should also have a significant effect on the color and reflectivity of the glaze17. Besides, the size of ε-Fe2O3 crystals around magnesioferrite crystals (Fig. 7c and Fig. 8) is significantly larger than others. Similar phenomenon is observed in sauce glazes of Yaozhou kilns5. Based previous study29, the corresponding lattice mismatch of spinel and ε-Fe2O3 crystals could be less than 2.80%, which appears favourable for epitaxial growth of ε-Fe2O3 on spinel, which should be further confirmed in our case.

SEM observations (Fig. 9) show the dendritic ε-Fe2O3 crystals in the samples are composed of secondary branches separated by distance ranging from 170 to 330 nm, consistent with previous research concerning sauce glazes from Yaozhou kilns7 and Jian kilns11. Such structure could react with the incident light to generate structural color. According to the coherent diffraction equation, 2nDcosθ = mλ, where n is the crystal reflection coefficient equal to 2.9 for iron oxide, θ is the light incidence angle, λ is the reflected light, D is the reflected crystal face spacing, and m is the natural constant, the wavelength range of the structure color could be calculated. The shift D between two successive secondary dendrites in ε-Fe2O3 crystals are 170 and 230 nm (Fig. 6) for samples YZY-BG-01 and LS-DD-10, respectively. The equation could be simplified to 5.8×(170 or 230 nm)×cosθ = mλ. If m = 1 and θ ranging from 40 to 60°, the corresponding structural color wavelength varies from 493-755 nm and 667-1022 nm for samples YZY-BG-01 and LS-DD-10, respectively. Compared with the scatter-reflectance spectra of YZY-BG-01 (Fig. 9a), the reflective peaks fall within the scope of theoretical calculations, whereas the positions of the scatter-reflectance peaks of YZY-BG-01 (Fig. 9b) are lower than theoretical values. It might be explained by the influence from the nature of ε-Fe2O3 crystals, for instance, ion substitutions in the crystals indicated in previous work11,12 or the size variations observed by SEM (Fig. 4 and Table S2), which should be further confirmed. Based on the results, the glaze color is not only mainly original from the chemical color generating from the ε-Fe2O3 crystal but also slightly the structural color generating from the crystalline structure. The latter well account for the visible heterochromaticity characteristics of these precious sauce glazed wares. Besides, the lamellar crystals with an offset of around 110 nm in the secondary branches of the dendrite crystals could increase the reflection of incident light, giving rise to a high gloss of the whole glaze.

In addition, ε-Fe2O3 was considered a metastable phase synthesized in the laboratory, existing only in nanocrystalline30,31. Interestingly, micrometer-sized ε-Fe2O3 crystals were observed in the ancient sauce-glazed wares. The stability of ε-Fe2O3 crystals appears to be controversial here32,33,34. However, it should be noted that the ε-Fe2O3 crystals detected in ancient glazes are surrounded by the glassy matrix, which possibly benefits the size of the Fe2O3 particle growing larger. Such a phenomenon was demonstrated by Shkurai et al.35, who believed that the Fe2O3 particles aggregate due to the softness of the SiO2 matrix. Besides, much research also demonstrated that ion substitutions in crystals could favor the growth of ε-Fe2O3 crystals36, which may be consistent with the situations of micro-sized ε-Fe2O3 crystals in ancient sauce glazes11,12.

In conclusion, representative Fe-based sauce glaze wares from eight Chinese Northern and Southern kilns were systematically analyzed by OM, XRF, XRD, μ-RS, SEM-EDS, FIB-SEM, and reflective spectroscope. The sauce color is origin from the ε-Fe2O3 crystals in the glaze. On the one hand, the ε-Fe2O3 crystals are distributed in the whole glaze, which could be divided into three layers. The first layer in the glaze surface contains large crystals with sizes of hundreds microns. The second and third layers in the deepened glaze are composed of small crystals with several to tens of microns in size. Such a multilayer could enhance the crystal coloring ability. Besides, magnetite crystals were detected in the glaze, also giving rise to darken the crystal color. In addition, these dendrite ε-Fe2O3 crystals were composed of secondary branches with an offset of ranging from 170–330 nm, which could react with incident light and generate structural color. Moreover, the structural color could vary with changes in the angle of incident lights, leading to the glaze appearing angular heterochromism.