Fe speciation in multiphase assemblages of black and white porcelain (Jin dynasty 1115–1234 CE) using micro-X-ray absorption spectroscopy

Abstract

Black and white porcelain is a highly heterogeneous multilayer ceramic system, whose technological information and microstructure are influenced by the composition of the raw materials and complex reactions that occur during the firing process. This study performed several synchrotron radiation-based analytical techniques (µ-XRF mapping, µ-XANES spectroscopy, µ-XANES mapping, and SIXES) to analyze the spatial distribution of major elements in the stratigraphic layers, as well as the spatial distribution of iron speciation. The results show that the pigment sources are diverse, including iron-rich glazes, iron-bearing clay minerals, and non-clay iron minerals. Pigment particles, mainly composed of hematite, play a key role in the formation of decorations. In addition, the valence state of iron among layers is not stable and even exhibits differences at the micrometre scale. Our results also indicate that the multi-technique approaches based on synchrotron-radiation have significant advantages in studying complex layered ceramic systems, especially the µ-XANES mapping and SIXES techniques, which provide information on iron species and their local variations while maintaining micrometre scale spatial resolution.

---

1 Introduction

Complex, layered and chemically heterogeneous systems, such as glazed ceramics, typically undergo reactions at interfaces, diffusion fronts and redox gradients that occur over length scales from only a few to several hundred micrometers. Elucidating these processes requires analytical methods that can map elemental distributions and determine oxidation states in complex multiphase assemblages. Conventional laboratory analytical approaches often lack the spatial resolution or chemical sensitivity required to capture these subtle changes, especially when multiple crystalline phases coexist with glassy components. Therefore, the mechanisms that control the phase formation and transformations of such composite materials remain incompletely understood.

Black and white porcelain is a type of decorative porcelain with a complex, heterogeneous, multilayered structure, which was produced on a large scale in northern China during the Jin dynasty (1115–1234 CE). Its typical cross-section consists, from the inner layers to the outer ones, of a coarse ceramic body formed from clay-based raw materials, covered with a thin layer of white slip usually prepared from high quality kaolinitic clay, sometimes referred to as “makeup clay”.^1–3 On top of it, a black decoration is added using layers of Fe-based pigments derived from various sources, most commonly Fe-bearing minerals such as magnetite (Fe₃O₄), hematite (α-Fe₂O₃) and goethite (α-FeOOH),^4,5 and transparent glazes, producing a strong contrast between the black patterns and the white background. In contrast to monochrome ceramics, which typically comprise a relatively simple stratigraphy of body, slip, and glaze, black and white ceramics exhibit a more complex multilayered structure due to the presence of an additional iron-rich pigment layer. This configuration introduces multiple reaction interfaces and heterogeneous sources of iron, thereby promoting coupled processes of dissolution, diffusion, and recrystallization during firing. Due to differences in raw material resource availability and technological development, workshops in different areas gradually developed distinct production techniques and decorative styles. Therefore, the stratigraphy of the pigmented and non-pigmented layers, the interactions between the layers during firing, and the microstructure and optical properties of black and white porcelain are extremely sensitive to the nature of the raw materials used, pigment preparation and firing conditions.

In recent years, researchers have gradually paid more attention to how black and white porcelain was produced. During the manufacturing process, and more specifically during the high temperature steps of the firing, the iron-based crystals undergo profound transformations. Various dissolution and recrystallization processes occur both during the heating and cooling steps leading to modifications in the valence state and the distribution of iron ions especially in the glaze. Furthermore, additional chemical interactions occur at the interfaces, leading to specific crystallizations that can provide key information regarding both the raw materials used and the firing protocol.

The chemical and structural characterization of these thick, multilayer and heterogeneous materials requires at the same time a high spatial resolution over large area. Standard laboratory techniques can provide significant information, such as elemental composition and its distribution as well as the identification of crystalline phases, but they cannot allow us to study Fe speciation and may lack either the sufficient spatial resolution or field of view. In addition to the crystalline phases containing iron, dissolved ferrous and ferric ions are also present in the glass matrix with a variety of possible coordination environments, such as tetrahedral coordination,^6–8 hexa-coordination,^9–11 or octahedral coordination.^7,12,13 Therefore, this poses difficulties in revealing the firing process and color-formation mechanism of black and white porcelain.

Synchrotron radiation-based analytical techniques, especially µ-XRF imaging and µ-XANES, are powerful tools for characterizing the stratigraphic distribution of elements and for probing the oxidation states and coordination environments of transition metals at the microscale.¹⁴ In the case of heritage materials, which are highly structured mixtures, several studies have shown the benefit of µ-XAS analyses, which allow us to discriminate properly the different layers.^6,15–20 These techniques have been successfully applied to heterogeneous systems such as heritage ceramics from diverse origin, revealing the evolution of pigment sources,^21–24 processing sequences,^{15,16,18,25,26} and coloring mechanisms.^20,27–30 While µ-XANES point analyses are generally the standard procedure to provide information on oxidation state and coordination of one selected element, Selectively Induced X-ray Emission Spectroscopy (SIXES) allows us to obtain the spatial distribution of the chemistry of the targeted element.³¹ This approach has been successfully applied, notably in paintings, to explore the chemistry behind pigment degradation.^24,32 In these studies, the targeted elements (S, Cr, and Cd) exhibit marked pre-edges and significant differences in XANES spectral features, which make the selection of the energies to obtain a chemical contrast fairly straightforward.

In this study, we demonstrate that SIXES is capable of discriminating between iron species with similar oxidation states at the Fe K-edge, extending its applicability beyond systems characterized by strongly contrasting XANES signatures. When associated with µ-XRF imaging and punctual µ-XANES analysis, SIXES can indeed be a powerful tool for studying the distribution of Fe-based phases in the complex layered structure of black and white porcelain with an adequately large field of view. The results provided a crucial insight into the role of Fe both as a proxy of the manufacturing process and as the coloring agent leading to the aesthetic aspects of black decorations. Moreover, it enabled us to uncover key information about production techniques, and to elucidate the optical properties of coloring elements and shed new light on the raw material selections and firing conditions in black and white porcelain.

2 Experimental details

2.1 Sample description

Seven black and white porcelains produced by private workshops in Shanxi province are the main focus of this study, together with one black-glazed porcelain (GZ01) that serves as a reference sample. According to stratigraphic and archaeological classification, these samples were mainly produced during the Jin dynasty (1115–1234 AD) in Hejin Kilns (GZ-BD01), in Xiangning Kilns (TGD-BD04, TGD-BD06, TGD-BD08, TGD-BD14) and in Huozhou Kilns (HZ-BD01 and HZ-BD02). The sample photos are shown in Fig. 1. Their common characteristic is that the gray coarse body is covered with a layer of white slip, and the glaze and decorative areas are glossy.

Fig. 1 Photographs of the 8 studied sherds: (a) GZ-BD01; (b) TGD-BD04; (c) TGD-BD06; (d) TGD-BD08; (e) TGD-BD14; (f) HZ-BD01; (g) HZ-BD02; (h) GZ-01.

2.2 Thin cross-section sample preparations

Since the glaze is thin and layered, it is essential to carefully prepare a thin cross-section of the samples to preserve as much as possible the original stratigraphy of the sample. As such, the preparation method of thin section samples used here is similar to that used in transmission microscopy. The detailed sample preparation process can be found in the study by Tian et al.²⁰ In this study, the final thickness of the samples after polishing was between 40 and 100 µm. The advantage of this preparation approach is that it protects the very superficial layer of glaze from damage and provides better in-depth information.

2.3 Synchrotron radiation investigations

The synchrotron-based measurements were performed at the ID21 beamline of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). Undulator-generated X-rays were monochromatized using a Si (311) double-crystal monochromator. The beam was focused by Ni-coated Kirkpatrick–Baez mirrors to ∼1.0 (h) × 0.7 (v) µm² at the sample. Incident intensity (I₀) was monitored with an upstream photodiode, and transmitted intensity with a downstream photodiode for transmission measurements; fluorescence photons were recorded with a silicon drift detector positioned at ∼90° to the incident beam. When required, attenuators were inserted to keep the detector dead-time within the right range. All measurements were performed under vacuum to minimize scattering and air-induced reabsorption of fluorescence and to limit reactions between the metal centre and ambient gases.³³ Energy calibration used an iron metal foil, with the first inflection point of the Fe K-edge set to 7111.32 eV (corresponding to the maximum of the first-order derivative).

2.3.1 Acquisition of µ-XRF map images. Two-dimensional µ-XRF maps were acquired by raster-scanning the samples in the vertical mounting geometry, while I₀ was continuously monitored upstream to correct for temporal variations in incident flux. During mapping, attenuators were put in the beam as needed to control flux and maintain the detector dead-time in the desired range (∼30%).

2.3.2 Acquisition of µ-XANES spectra. XANES point spectra, collected in both fluorescence and transmission modes, were acquired with a step size of 0.15 eV and a dwell time of 0.04 s. The spectra were collected along profile lines from the body to the glaze surface to track changes in iron species across the depth of the glazes. Reference spectra of hematite (α-Fe₂O₃), magnetite (Fe₃O₄), wustite (FeO), pseudobrookite (Fe₂TiO₅) and ε-Fe₂O₃ were measured in transmission mode using an unfocused beam (∼300 × 300 µm²) with a dwell time of 100 ms per point.

2.3.3 Acquisition of µ-XANES mapping and selectively induced X-ray emission spectroscopy (SIXES) mapping. To further explore the distribution of different Fe-species, µ-XRF mapping of areas of interest was implemented at several energies. Depending on the number of energies selected, this hyperspectral approach allows us either (i) to fully reconstruct a 2D-XANES map when maps are measured at enough different energies or (ii) to discriminate specific well-defined species by selecting carefully a few characteristic energies. In the rest of this paper, the first case is referred to as µ-XANES mapping, while the second is referred to as SIXES mapping.

µ-XANES maps were acquired over areas within the pigmented layers of TGD-BD06 (336 × 178 µm²) and HZ-BD01 (219 × 126 µm²). In this case, maps were recorded around the Fe K-edge with energies ranging from 7085 eV to 7240 eV with a step size of ∼1 × 1 µm². Such acquisitions generated hyperspectral images of about 4000 XANES spectra, with each pixel corresponding to one spectrum.

SIXES maps are implemented on areas across the pigmented layers of TGD-BD14 and GZ-BD01. 13 energies were selected by comparing the features of reference spectra (7080, 7090, 7100, 7114, 7125, 7131, 7133, 7136, 7146, 7167, 7195, 7240, and 7258 eV). Flat-field corrections and image re-alignment were implemented on the different maps. After pre-processing, the maps acquired at different energies were compared to identify the features of the different minerals present.

2.3.4 Analysis of data. Initial XANES data extraction and preprocessing were performed using in-house scripts. Fe K-edge XANES spectra were then normalized and analyzed using Orange Data Mining software.³⁴ PCA of the smoothed second derivatives partitioned the spectra into clusters, allowing us to average data according to the clusters. Pre-edge regions of the averaged spectra were then fitted. This approach is used because the pre-edge energy and intensity are sensitive to the Fe oxidation state and coordination geometry.^7,35 The pre-edges were fitted using Igor Pro software with Gaussian functions to obtain the centroid and integrated intensity. These values were compared with reference spectra to infer Fe valence and coordination. To improve accuracy, the fitting range was restricted to ∼7108–7117 eV following Fiege et al.,³⁶ with careful background subtraction to avoid biasing intensities. For estimating Fe³⁺/ΣFe, we considered the two approaches of Wilke et al.³⁷—the intensity ratio I(Fe³⁺)/[I(Fe³⁺) + I(Fe²⁺)] and the centroid (C). In line with Cottrell's high-precision calibration for silicate glasses,³⁸ we adopted the centroid method to determine Fe³⁺/ΣFe in the glass matrix (eqn (1)). It is worth noting that this method is heavily matrix-dependent, so this is an approximation and has to be taken with caution.

Fe³⁺/ΣFe × 100 = [−0.028 + (0.000784 + 0.00052(7112 − C))^0.5]/0.00026, (7112 eV ≤ C ≤ 7113.5 eV) (1)

µ-XANES maps were also analyzed with Orange Data Mining: after image realignment and spectral fitting, spectra were extracted and processed as in the point analyses, using K-means and PCA for clustering. The resulting classes were displayed on a scatter plot, and cluster-averaged spectra were computed to derive the Fe chemistry and valence distributions.

PyMCA software was used to batch fit the XRF spectra to generate elemental maps (Al, Si, K, Ca, Fe, Ti, and Mn), and to normalize those maps to the incident beam intensity.³⁹ SIXES maps were also visualized by using PyMCA's ROI imaging tool.

3 Results and discussion

3.1 Elemental distribution

Fig. 2 shows the elemental distribution of Fe, Al, and Ca in the cross-sections of the decorations. Preliminary investigation revealed a large number of anorthite crystals and iron-bearing particles in the pigmented areas of most samples.⁴⁰ These stratigraphies provide important clues regarding the application and raw materials used as pigments. The anorthite-rich layer between the pigment and upper glaze layer in GZ-BD01 and TGD-BD04 is consistent with the use of a clay-bearing Fe pigment also rich in Al. The local removal of slip in GZ-BD01 suggests an additional carving or scraping technique. In TGD-BD06, the coexistence of Fe particles with abundant anorthite crystals within the pigment layer revealed that the raw pigment contained Fe-bearing minerals and a Ca-bearing component, with less clay. In contrast, TGD-BD08 shows broadly homogeneous Ca and Fe throughout the glaze, implying coloration by Fe-rich glaze rather than an iron-particulate pigment layer. HZ-BD01 and HZ-BD02 display pigment layers that are rich in Fe (only hematite) but lack Ca and Al, and there is a slight diffusion of iron into the glaze layer. This suggests that the raw pigments are made of iron-ore-dominated pigments without kaolin-based clay, whereas TGD-BD14 is Fe-rich but Al/Ca-poor, indicating a mineral-dominated Fe pigment with minor kaolin/feldspar components. The black-glazed reference GZ-01 differs by showing Fe dispersed in the glaze matrix and anorthite confined near the interface between the glaze and body.

Fig. 2 Samples differentiated by color in an optical microscopy image of a cross-sectional pigment area, along with the distribution of Al, Ca, and Fe elements obtained in X-ray fluorescence mode.

3.2 Iron valence and coordination analysis

PCA processing of the µ-XANES point analyses of pigmented areas, as illustrated by the PC1/PC2 scatter plot in Fig. S1 (SI), reveals that samples GZ-BD01, GZ01, TGD-BD04, TGD-BD08 and TGD-BD14 form a single cluster, which suggests a relatively homogeneous chemical environment for Fe in these samples, while points measured on TGD-BD06, HZ-BD01, and HZ-BD02 each divide into two distinct clusters, indicating heterogeneity in the Fe environments in these cases. It is important to note here that XANES spectra record the average Fe environment in the measured volume, so one spectrum can contain features of both Fe-rich crystallized particles and Fe ions of the glassy matrix. Hence, homogeneous chemical environment does not mean that Fe has a unique chemical environment, but rather several co-existing in a homogeneous manner in the pigmented layer.

The average XANES spectra of clusters are shown in Fig. 3. Samples GZ-BD01, GZ01, TGD-BD08, and TGD-BD14 exhibit pre-edge centroid and absorption edge positions similar to those of hematite, confirming the dominance of Fe³⁺ in the pigmented layer. However, their comparatively more intense pre-edge features and broader white lines suggest that Fe³⁺ is located in a lower-symmetry environment, likely associated with disordered crystals or with ions in glassy matrices. In contrast, the spectra of TGD-BD04, TGD-BD06_cluster2, HZ-BD01_cluster1, and HZ-BD02_cluster1 closely match the reference hematite spectrum, in agreement with Raman spectroscopy, which confirms the presence of abundant hematite crystals in these regions (Fig. S2). It is worth noting that these hematite crystals are not structurally ideal and may exhibit lattice distortions, possibly caused by high-temperature firing and likely elemental substitution.

Fig. 3 Average Fe XANES spectra of the pigment area after PCA analysis: (a) GZ-BD01 and GZ-01; (b) TGD-BD04, TGD-BD06, TGD-BD08 and TGD-BD14; (c) HZ-BD01 and HZ-BD02.

The results of the fitting of the pre-edge features of the Fe K-edge spectra recorded on the pigmented area and the upper glass matrix are shown in Fig. 4a and b, respectively. The colored solid points represent the samples of this study, the black hollow points represent the reference spectra (hematite, magnetite, wustite, maghemite, and Fe₂(SO₄)₃), and the gray circles represent the corresponding centroid and integral intensities of Fe²⁺ and Fe³⁺ in the various geometrical sites of the model compounds obtained by Wilke et al.^13,41 All centroid positions are near ∼7113.5 eV, suggesting a predominance of Fe³⁺ in both the pigment and the upper glaze, although some Fe²⁺ is likely present. The pre-edge features of the glaze layer are represented by error bars rather than clustered points due to the variations in the Fe³⁺/Fe²⁺ ratio within the glaze layers.

Fig. 4 Scatter plot of the centroid and integrated intensity of the pre-edges of Fe XANES spectra of the (a) pigment area and (b) upper glaze.

In the pigmented area (Fig. 4a), the pre-edge features of the spectra corresponding to HZ-BD01_cluster1, HZ-BD02_cluster1, TGD-BD06_cluster1, and TGD-BD04 closely match those of the reference hematite, further confirming the presence of hematite crystals. In contrast, the pre-edge features of the other spectra are located between those of the magnetite reference and the ^[5]Fe³⁺ and ^[6]Fe³⁺ model compounds from Wilke et al. studies.^41,42 Besides, the centroid of the pre-edge in the pigmented area of GZ-01 is located around 7113.0 eV, suggesting a significantly lower Fe³⁺/Fe²⁺ ratio.

In the glaze layers (Fig. 4b), the integrated intensity of the pre-edge increases compared to the pigmented layer, indicating that the coordination number of iron decreases. According to the model of Wilke et al., it seems to be approaching an average 5-coordinate environment. Fe³⁺ is usually in a tetrahedral environment in silicate glass,^9,43 but high concentrations of alkali and alkaline earth ions may cause Fe³⁺ to switch from a 4-coordinated to a 5-coordinated or even a 6-coordinated environment, which is consistent with the previously discussed results.⁴⁴

Table 1 shows the estimated Fe³⁺/ΣFe ratio of each layer of the stratigraphy calculated using eqn (1) in 2.3.4. The black and white porcelain samples seem dominated by Fe³⁺, though a small amount of Fe²⁺ is present, suggesting that the firing atmosphere was more strongly oxidizing. TGD-BD04 exhibits the highest Fe³⁺/ΣFe, suggesting that it was likely fired under the most strongly oxidizing atmosphere among the studied corpus. The Fe³⁺/ΣFe ratios in each layer of the glazes of TGD-BD08, HZ-BD01, and HZ-BD02 are relatively comparable, indicating that the atmosphere seems stably controlled throughout the whole firing stages. Conversely, GZ-BD01, TGD-BD06, and TGD-BD14 have higher concentrations of Fe³⁺ in the bodies than in the other layers, which indicates that the samples were fired in an oxidizing atmosphere during heating but a slight reduction process may have occurred during the later stages of heating. Besides, the black glaze GZ-01 exhibits significantly lower Fe³⁺/∑Fe ratios in the body (49.7%) and the pigment-containing glaze layer (49.0%) compared to the upper glaze (100%). This distribution suggests that GZ01 may have been fired in a neutral or mildly reducing atmosphere during the heating stage followed by an oxidizing atmosphere during cooling, resulting in complete oxidation of the outer glaze while partial Fe²⁺ was retained in the inner layers.⁴⁵

Table 1 Fe³⁺/ΣFe (%) average of each component of the samples obtained by pre-edge fitting and estimation

Sample layers	GZ01 (black glaze)	TGD-BD08	HZ-BD01	HZ-BD02	TGD-BD04	GZ-BD01	TGD-BD06	TGD-BD14
Body	49.69	85.30	79.19	84.40	—	100*	90.99	100*
Pigment	49.01	77.64	82.52	81.41	100*	78.40	89.29	86.63
Glaze	100*	—	75.80	80.81	100*	72.17	83.03	70.92

---

XANES spectra were also acquired as profiles across the stratigraphy of the samples, from the surface to the bottom of the glaze layer in HZ-BD02, TGD-BD04, TGD-BD006 and TGD-BD14. The centroid positions of the Fe pre-edge features are plotted against the corresponding depth in Fig. 5. This figure shows that the Fe³⁺/Fe²⁺ ratio in depth across the glaze is not constant but changes gradually with depth, which is attributed to the diffusion of electronic holes and ions that generate oxidation gradients during firing.⁴⁶ In HZ-BD02, TGD-BD04, and TGD-BD14, the centroid position gradually increased with distance from the glaze surface, indicating a gradual increase in Fe³⁺ and a gradual decrease in Fe²⁺, close to the glaze surface. In contrast, in TGD-BD06, the centroid position decreases with increasing depth, Fe²⁺ gradually increases and Fe³⁺ gradually decreases.

Fig. 5 Scatter plot of the centroid position of pre-edges and the corresponding depth in the cross sections of HZ-BD02, TGD-BD04, TGD-BD06 and TGD-BD14.

The valence state of iron in the glass network at the surface of glazes is usually related to the atmosphere conditions during the cooling stage. As such, if cooling is done in an atmosphere more oxidizing than the prior, an oxidation gradient should appear with Fe being more oxidized at the surface than deeper. In contrast, if the cooling stage is more reducing than the prior steps of the firing, Fe should appear more reduced at the surface. Fe appears less oxidized at the surfaces of HZ-BD02, TGD-BD04, and TGD-BD14, which suggests that they may have been exposed to a less oxidizing atmosphere, while the deeper regions experienced stronger oxidation. In contrast, TGD-BD06 shows the reverse trend, so this sample was likely exposed to a more oxidizing atmosphere. A higher Fe³⁺/ΣFe ratio in the bodies indicates that they experienced a more oxidizing process (Table 1). During firing, the combustion of organic components in the body and white slip, as well as the calcination of minerals (such as carbonates), leads to an increase in open porosity.⁴⁷ The expanded pore network promotes the diffusion of oxygen into the pigment and glaze layers, thereby further promoting the oxidation of iron ions. Therefore, the Fe³⁺/ΣFe ratio in the pigmented and glaze layers may be influenced by both the external firing atmosphere and oxygen diffusing upward from the bodies.

Fig. 6 shows µ-XANES maps and XANES spectra of the pigmented areas of HZ-BD01 (a–d) and TGD-BD06 (e–h). Fig. 6a and e show the maps of edge jump, which are correlated to the concentration of Fe (red area is rich in Fe). Fig. 6b and f show the Fe valence state maps, red areas contain higher Fe³⁺, while blue areas contain less Fe³⁺ but more Fe²⁺. Fig. 6c and g show the spatial map of clusters obtained by K-means and PCA analysis of the XANES spectra for all pixels (4 clusters in HZ-BD01 and 3 clusters in TGD-BD06). The average spectra of the clusters are shown in Fig. 6d and f, respectively.

Fig. 6 µ-XANES analysis of the Fe K edge of HZ-BD01 (a–c) and TGD-BD06 (e–g). (a and e) Edge jumps (correlated to Fe concentration), (b and f) edge energy map at 7120 eV, the red region contains more Fe³⁺, while the blue region contains more Fe²⁺. (c and g) Image segmentation obtained by PCA and K-means clustering and (d and h) the corresponding average XANES spectra, showing 4 clusters in HZ-BD01 and 3 clusters in TGD-BD06.

In HZ-BD01 (Fig. 6a–d), Fig. 6a shows a significant enrichment of iron in the pigment layer, with a small amount of iron beneath the glaze and pigment layers, indicating some degree of pigment dissolution. Fig. 6b shows that iron in the pigment layer exists mainly in the form of Fe³⁺. The glaze layer and the layer beneath the pigment contain less Fe²⁺ and more Fe³⁺. The Fe³⁺ distribution near the slip layer is uneven, indicating the presence of local redox non-uniformity at the micrometer scale within the layer. In Fig. 6c, the K-means plot divides the entire measurement area into 4 clusters. The pigment layer, glaze layer, and the layer beneath the pigment each show only one cluster, indicating a similar iron environment within each layer. But the area near the slip layer contains two clusters, suggesting a greater variability in the iron environment. This might be due to interlayer reactions, which lead to greater variability in the local environment.

In TGD-BD06 (Fig. 6e–h), the measurement area is acquired within the pigment layer. In the edge jump map, the red spots are rich in Fe, the green area is the Fe-rich regions in the matrix, and the blue area corresponds to Fe-poor regions (Fig. 6e). Fe-rich particles contain higher levels of Fe³⁺. The surrounding glass matrix has slightly lower levels of Fe³⁺ and higher levels of Fe²⁺. All pixels were grouped into three clusters (Fig. 6g), and the average spectrum of cluster 1 is consistent with hematite. Clusters 2 and 3 have similar spectral shapes, but the absorption of cluster 3 is slightly higher at around 7120 eV, indicating that it is relatively more reduced (with a higher Fe²⁺ ratio) and corresponds to the Fe-poor matrix. This indicates micrometer-scale redox heterogeneity within the pigment layer, which may result from heterogeneous pigment dissolution and locally limited diffusion, together with spatial variations in oxygen fugacity during firing.

3.3 Fe speciation and distribution in pigmented areas

Fig. 7 and 8 show the results of the SIXES maps in the pigmented layers of TGD-BD14 and GZ-BD01, respectively. Two areas (area 1 and area 2) were selected in each sample to observe the nature and the spatial distribution of Fe-based crystals on the micrometer scale. The images shown in both figures are generated by calculating the differences in the intensities of the Fe K-emission lines at energies corresponding to the main features of the reference spectra of iron-bearing oxides, which have been found in the pigmented area by Raman microscopy (Fig. S3).

Fig. 7 SIXES images of (a) area 1: glaze layer and pigment; (b) area 2: pigment and white slip of TGD-BD14 acquired at multiple energies and stacked together. The reference spectra (c) are used to determine the differences in the spectra of hematite and pseudobrookite at certain energies. The differences between images acquired at 7133 eV (B) and at 7131 eV (A) highlight hematite positively (red) and pseudobrookite negatively (blue) in the pigment layer.

Fig. 8 SIXES images of two pigment areas in GZ-BD01 (area 1: panels a–c; area 2: panels d–f). The reference spectra in (g) are used to determine the differences in the spectra of hematite, pseudobrookite and ε-Fe₂O₃ at certain energies. (a and d) show the distribution of hematite by B-A, (b and e) highlight the distribution of ε-Fe₂O₃; (c and f) show the distribution of pseudobrookite by F-D.

In TGD-BD14 (Fig. 7), area 1 includes the glaze and pigment layers (Fig. 7a), and area 2 includes the pigment layer and the white slip (Fig. 7b). By substracting the map collected at 7133 eV (feature B) from the map collected at 7131 eV (feature A), hematite (red area) and pseudobrookite (blue area) are highlighted positively and negatively, respectively. The results show that the pigment layer is dominated by hematite, which is widely distributed. Pseudobrookite occurs as large crystals but in small quantities. In the glaze layer, however, there are no iron-bearing crystals (Fig. 7a). Iron-bearing particles are indeed restricted to the pigment layer and are rarely observed in the glaze or white slip, suggesting limited dissolution of the pigment and diffusion of ionic Fe into these layers, this might be attributed to the high Al content in the pigment, which acts as a refractory framework to increase local viscosity. As such, the decoration color is mainly controlled by the hematite in the pigment layer.

In GZ-BD01 (Fig. 8), the investigated areas are within the pigment layer. The reference models are hematite, pseudobrookite and ε-Fe₂O₃. Six energies (labeled A–F), corresponding to the greatest differences between the features of the reference spectra of hematite, pseudobrookite and ε-Fe₂O₃ are displayed in Fig. 8g. Differences between maps were used to highlight specific phase distributions: B-A highlights hematite, C-E highlights ε-Fe₂O₃, and F-D highlights pseudobrookite. The distribution of hematite is shown in Fig. 8a and d, ε-Fe₂O₃ is shown in Fig. 8b and e, and pseudobrookite is shown in Fig. 8c and f. Areas 1 and 2 both show that hematite occurs in larger sizes and greater abundance, whereas ε-Fe₂O₃ appears as smaller spots distributed along hematite rims and more broadly within the glassy matrix. Pseudobrookite is only locally present and occurs as coarser grains. Within the pigment layer, hematite and ε-Fe₂O₃ constitute the dominant phases, consistent with an Fe³⁺-rich environment.

3.4 Color mechanism of decorative patterns

The analytical results indicate that the pigments of the black and white porcelain are mainly composed of Fe³⁺-bearing oxides, iron generally being in octahedral coordination in the structures. Hematite is the predominant crystal phase responsible for color formation.

At high temperatures, some hematite crystals diffused into the surrounding glass matrix and the overlying glaze. In these regions, iron is predominantly present as Fe³⁺ with a reduced coordination number, accompanied by a minor proportion of Fe²⁺. Previous studies have shown that Fe³⁺ absorbs UV light regardless of whether it is in a tetrahedral or an octahedral environment.⁴⁸ Due to the deformation of the Fe ion coordination field, the absorption peak can expand and even reach the visible light area, making the glass matrix appear brownish-yellow.¹¹ Besides, Ti in the glassy matrix is predominantly present as Ti⁴⁺, as shown in Fig. S4. Although Ti⁴⁺ itself is optically colorless, its high field strength and strong polarizing power can modify the local electric field and coordination symmetry around Fe³⁺, thereby enhancing the chromogenic capability of iron ions.^49,50 Therefore, the Fe-containing glassy matrix of the black and white porcelain sample appears brownish-yellow.

When the pigmented layer is thick and contains densely packed pigment particles, the particles will be the main contributing factor to the color of the patterns, giving them an opaque black color. When the pigment layer is thin, the glaze layer provides a more significant contribution to coloring, giving the patterns a brown color. This explains why the patterns in the thinner areas and edges of the patterns appear brown, particularly in samples from the Xiangning (TGD) and Huozhou (HZ) workshops.

The coloring mechanism of the black glaze in GZ01 differs from that of black and white porcelain, as it can be primarily attributed to Fe³⁺ ions dispersed within the glaze layer instead of iron crystals. As such, when the glaze is relatively thin, a brownish tone is observed, whereas a thicker glaze absorbs more visible light, resulting in a darker appearance, up to black.

The decoration of TGD-BD04 is different from the others, as it displays a darker brown pigment near the upper glaze, and red layer below it, as shown in Fig. 9a. µXRF-maps of the area highlighted by a red rectangle are shown in Fig. 9b. Both the brown and red areas are rich in iron (especially in crystals) and contain small amounts of K, along with some Ti-bearing particles. Fig. 9c shows the average Fe K-edge XANES spectra obtained in these two areas. Both spectra display the characteristic features of hematite. However, it is noteworthy that the hematite features appear significantly sharper in the spectrum of the red area than in that of the brown areas, suggesting that the Fe environment is closer to hematite in the red area. Further analysis of the pre-edge features, shown in Fig. 9d, and the fitting results shown in Fig. 9e, shows that the red areas have a much higher t_2g/e_g ratio (0.99) than the brown areas (0.31). According to previous studies,⁵¹ the decrease in the t_2g/e_g ratio is usually associated with an increased degree of aberration in the coordination environment of Fe³⁺ in octahedral sites. Such structural distortion would induce a decrease in t_2g/e_g ratio, which leads to changes in the visible light absorption behavior. Specifically, a lower t_2g/e_g ratio shifts the absorption wavelength towards the long-wave direction and enhances the absorption of green and yellow light, giving a complementary purplish-black appearance, while an elevated ratio absorbs predominantly green-band light, giving a bright red color.^51,52

Fig. 9 (a) Cross-sectional optical micrograph of TGD-BD04 showing the red and brown areas; (b) XRF-map of the area outlined in red in (a); (c) Fe K-edge XANES spectra acquired from the red and brown areas, respectively; (d) pre-edge fitting results of brown area (left) and red area (right); (e) results of Gaussian function fitting of the pre-edge.

4 Conclusion

This study investigated six representative black and white porcelains alongside one black glazed ware produced by private workshops in southern Shanxi during the Jin dynasty. By investigating the morphology and valence state of iron using synchrotron-based X-ray absorption spectroscopy (µ-XANES), in combination with cross-sectional microstructural analysis of the decorated areas, this study provides new insights into the technological characteristics of black and white porcelain production and the color mechanisms of decorative patterns.

The production of black and white porcelain demonstrates a diverse range of raw material combinations. Pigments are derived from three main sources: iron-rich ores, iron-rich clays, and iron-rich glazes, while glazes can be categorized into Ca-rich and Ca-poor types based on the formation of anorthite. This flexible pairing of pigments and glazes is a typical trait of private workshops and reflects how different workshops utilized local resources. Besides, the technique involving black and white porcelain is under-glazed decoration. During firing, significant elemental diffusion and interfacial reactions occur: calcium ions migrate from the glaze into the pigment layer, while iron ions diffuse from the pigment into the glassy matrix. This chemical evolution is not only a hallmark of high-temperature firing but also creates the distinct visual depth and strong interlayer bonding characteristic of black and white porcelain.

The µ-XRF and SIXES maps further revealed a targeted pairing strategy in our samples: highly fluid Ca-rich glazes are coupled with high-clay pigments, while viscous Ca-poor glazes are paired with iron-rich ores. This adjustment minimizes pigment dispersion at high temperatures, ensuring sharp black and white patterns. Such intentional material combinations reflect the craftsmen's profound understanding of raw material properties and their mastery in adapting techniques to materials.

The pre-edge feature of µXANES determined the average valence state and coordination of iron across the stratigraphy. Fe is present mostly as Fe³⁺ ions in an octahedrally coordinated environment, which indicates that the samples were fired in an oxidizing atmosphere, but the slight differences in oxidation levels between samples indicate variations in kiln atmosphere control.

The color of the black decorations results from the contributions of iron-bearing particles and iron ions of the glass matrix. Due to the differences in the nature of Fe-based crystals, their grain size and their concentration in black decorations, a variation in color tones and shades can be observed.

In conclusion, this study exemplifies the necessity to use synchrotron radiation-based micro-analyses to study the nature of this type of decoration. It not only allows an in-depth study of the valence and coordination of the chromogenic elements, but also accurately reflects their chemical speciation and spatial distribution over relevant fields of view and with relevant spatial resolution. In particular, SIXES provides a practical and rapid approach for speciation mapping, as it allows us to assess the spatial distribution of phases using only a few carefully selected energies.

Author contributions

Minli Wang: conceptualization, investigation, data acquisition, formal analysis, visualization, writing – original draft; Christophe Faulmann: formal analysis, methodology, validation, writing – review and editing; Clément Hole: data acquisition, software, writing – review and editing; data curation; Fen Wang: investigation, methodology; Chenchen Luo: formal analysis; Tian Wang: conceptualization, investigation, funding acquisition, resources; Philippe Sciau: supervision, project administration, conceptualization, methodology, writing – review and editing.

Conflicts of interest

There are no conflicts to declare.

Data availability

All primary data supporting the findings of this study, including processed µ-XRF maps, µ-XANES spectra and maps, and microstructural images, are available within the article and its supplementary information (SI). The raw synchrotron datasets are available from the ESRF data repository at https://doi.esrf.fr/10.15151/ESRF-ES-1306925003. Supplementary information is available. See DOI: https://doi.org/10.1039/d6ja00035e.

Acknowledgements

We acknowledge the European Synchrotron Radiation Facility (ESRF) for providing beam time at the ID21 beamline for the proposal number HG-226. In agreement with the ESRF data policy, the raw data related to this study can be found here: DOI 10.15151/ESRF-ES-1306925003. This work was supported by the “Cultural Relics Scientific Identification and Research Innovation Team “project under the Shaanxi Province Sanqin Talents Special Support Program, the National Natural Science Foundation of China (No. U2574210, 52272019, and 52272020), the Central Government Guides Local Science and Technology Development Fund Project (No. 2024ZY-JCYJ-04-06), the Shaanxi Province Technology Innovation Guidance Special Project (No. 2024QY-SZX-04), and China Scholarship Council (CSC).

References

1. D. Qin, Huaxia Archaeol., 2018, 58–74.
2. Z. Zhang, Cizhou Kiln Porcelain Pillows, People’s Fine Arts Publishing House, Beijing, 2000.
3. F. Wang, Pottery and Porcelain of Yaozhou Kiln, Shaanxi University of Science and Technology Press, Shaanxi, 2000.
4. N. Wood, Chinese Glazes: Their Origins, Chemistry and Re-creation, A & C Black, London, 2007.
5. D. Qin, Cultural Relics, 1994, 9, 48–55.
6. S. Quartieri, M. Triscari, G. Sabatino, F. Boscherini and A. Sani, Eur. J. Mineral., 2002, 14, 749–756.
7. M. Wilke, F. Farges, P.-E. Petit, G. E. Brown and F. Martin, Am. Mineral., 2001, 86, 714–730.
8. L. Galoisy, G. Calas and M. A. Arrio, Chem. Geol., 2001, 174, 307–319.
9. F. Farges, Y. Lefrère, S. Rossano, A. Berthereau, G. Calas and G. E. Brown, J. Non-Cryst. Solids, 2004, 344, 176–188.
10. F. Farges, S. Rossano, Y. Lefrre, M. Wilke and G. E. Brown, Phys. Scr., 2005, 957–959.
11. V. Vercamer, G. Lelong, H. Hijiya, Y. Kondo, L. Galoisy and G. Calas, J. Non-Cryst. Solids, 2015, 428, 138–145.
12. L. Bugaev, F. Farges, E. Rusakova, A. Sokolenko, Y. Latokha and L. Avakyan, Phys. Scr., 2005,(T115), 215.
13. M. Wilke, F. Farges, G. M. Partzsch, C. Schmidt and H. Behrens, Am. Mineral., 2007, 92, 44–56.
14. A. Bianconi, J. Garcia, A. Marcelli, M. Benfatto, C. R. Natoli and I. Davoli, J. Phys. Colloq., 1985, 46, 101–106.
15. P. Sciau, Y. Leon, P. Goudeau, S. C. Fakra, S. Webb and A. Mehta, J. Anal. At. Spectrom., 2011, 26, 969.
16. M. Matsunaga and I. Nakai, Archaeometry, 2004, 46, 103–114.
17. L. Lühl, B. Hesse, I. Mantouvalou, M. Wilke, S. Mahlkow, E. Aloupi-Siotis and B. Kanngiesser, Anal. Chem., 2014, 86, 6924–6930.
18. I. Cianchetta, K. Trentelman, M. S. Walton, J. Maish, A. Mehta and B. Foran, J. Am. Ceram. Soc., 2016, 99, 1792–1801.
19. A. Gianoncelli, G. Kourousias, S. Schöder, A. Santostefano, M. L'Héronde, G. Barone, P. Mazzoleni and S. Raneri, Appl. Sci., 2021, 11, 8052.
20. T. Wang, T. Zhu, Z. Feng, B. Fayard, E. Pouyet, M. Cotte, W. De Nolf, M. Salome and P. Sciau, Anal. Chim. Acta, 2016, 928, 20–31.
21. R. Bugoi, B. Constantinescu, E. Pantos and D. Popovici, Powder Diffr., 2008, 23, 195–199.
22. L. Monico, K. Janssens, M. Cotte, S. M. Webb, F. Vanmeert, V. Gonzalez, G. Van Der Snickt, K. Keune, B. G. Brunetti, G. Falkenberg, J. Garrevoet, A. Van Loon, M. Vermeulen, M. Maguregui, S. De Meyer, E. A. Clerici, F. T. H. Broers, L. Cartechini, N. De Keyser, F. d'Acapito, F. Meirer, A. Romani, F. Rosi, S. C. Marri, D. Comelli, N. Deleu, I. Fazlic, M. Ghirardello, C. Holé, S. Pérez-Diez, M. Thoury and C. Miliani, Riv. Nuovo Cimento, 2025, 48, 315–434.
23. M. S. D. Río, A. Sodo, S. G. Eeckhout, T. Neisius, P. Martinetto, E. Dooryhée and C. Reyes-Valerio, Nucl. Instrum. Methods Phys. Res., Sect. B, 2005, 238, 50–54.
24. G. Van der Snickt, J. Dik, M. Cotte, K. Janssens, J. Jaroszewicz, W. De Nolf, J. Groenewegen and L. Van der Loeff, Anal. Chem., 2009, 81, 2600–2610.
25. C. Holé, C. Dejoie, G. Wallez, Z. Ren, T. Wang and P. Sciau, J. Phys. Chem. C, 2025, 129, 12150–12158.
26. F. Meirer, Y. Liu, E. Pouyet, B. Fayard, M. Cotte, C. Sanchez, J. C. Andrews, A. Mehta and P. Sciau, J. Anal. At. Spectrom., 2013, 28, 1870.
27. Y. Chen, R. Wen, L. Wang and M. Zhang, Anal. Methods, 2022, 14, 541–548.
28. F. Gherardi, C. Hole, E. Campbell, M. Cotte, R. Tyson and S. Paynter, JPhys Photonics, 2024, 6, 025001.
29. L. Wang and C. Wang, J. Anal. At. Spectrom., 2011, 26, 1796.
30. C. Xing, X. Deng, Y. Feng, L. Li, D. Chen, S. Chu, Y. Mu, L. Wu, L. Yan, L. Zhao, X. Feng, Y. Tao, Y. Li, S. Wu and G. Chang, J. Eur. Ceram. Soc., 2025, 45, 116892.
31. K. Sakurai, A. Iida, M. Takahashi and Y. Gohshi, Jpn. J. Appl. Phys., 1988, 27, L1768.
32. L. Monico, G. Van der Snickt, K. Janssens, W. De Nolf, C. Miliani, J. Verbeeck, H. Tian, H. Tan, J. Dik, M. Radepont and M. Cotte, Anal. Chem., 2011, 83, 1214–1223.
33. L. Monico, M. Cotte, F. Vanmeert, L. Amidani, K. Janssens, G. Nuyts, J. Garrevoet, G. Falkenberg, P. Glatzel, A. Romani and C. Miliani, Anal. Chem., 2020, 92, 14164–14173.
34. J. Demšar, T. Curk, A. Erjavec, Č. Gorup, T. Hočevar, M. Milutinovič, M. Možina, M. Polajnar, M. Toplak, A. Starič, M. Štajdohar, L. Umek, L. Žagar, J. Žbontar, M. Žitnik and B. Zupan, J. Mach. Learn. Res., 2013, 14, 2349–2353.
35. G. E. Brown Jr, F. Farges and G. Calas, in Structure, Dynamics, and Properties of Silicate Melts, ed J. F. Stebbins, P. F. McMillan and D. B. Dingwell, De Gruyter, 1995, pp. 317–410.
36. A. Fiege, P. Ruprecht, A. C. Simon, A. S. Bell, J. Göttlicher, M. Newville, T. Lanzirotti and G. Moore, Am. Mineral., 2017, 102, 369–380.
37. M. Wilke, G. M. Partzsch, R. Bernhardt and D. Lattard, Chem. Geol., 2005, 213, 71–87.
38. E. Cottrell, K. A. Kelley, A. Lanzirotti and R. A. Fischer, Chem. Geol., 2009, 268, 167–179.
39. V. A. Solé, E. Papillon, M. Cotte, Ph. Walter and J. Susini, Spectrochim. Acta, Part B, 2007, 62, 63–68.
40. M. Wang, C. Faufmann, F. Wang, T. Wang and P. Sciau, J. Raman Spectrosc., 2024, 55, 1236–1246.
41. M. Wilke, C. Schmidt, F. Farges, V. Malavergne, L. Gautron, A. Simionovici, M. Hahn and P.-E. Petit, Chem. Geol., 2006, 229, 144–161.
42. M. Wilke, F. Farges, G. M. Partzsch, C. Schmidt and H. Behrens, Am. Mineral., 2007, 92, 44–56.
43. B. Hannoyer, M. Lenglet, J. Dürr and R. Cortes, J. Non-Cryst. Solids, 1992, 151, 209.
44. P. A. Bingham, O. M. Hannant, N. Reeves-McLaren, M. C. Stennett and R. J. Hand, J. Non-Cryst. Solids, 2014, 387, 47–56.
45. C. Holé, L. Leber, Q. Bruel, C. Muller, H. A. S. Orduz, T. Wang, P. Sciau and M. Cotte, J. Am. Chem. Soc., 2025, 147, 42433–42440.
46. C. Holé, Z. Ren, F. Wang, J. Zhu, T. Wang and P. Sciau, Mater. Today Commun., 2022, 33, 104329.
47. F. Wesenauer, C. Jordan, M. Azam, M. Harasek and F. Winter, Materials, 2021, 14, 4942.
48. R. Kukkadapu, H. Li, G. Smith, J. Crum, J. Jeoung, W. Poisl and M. Weinberg, J. Non-Cryst. Solids, 2003, 317, 301–318.
49. F. Farges and G. E. Brown, Geochim. Cosmochim. Acta, 1997, 61, 1863–1870.
50. Y. Deng, Y. Zhou, Y. Yang, X. Shi, K. Zhang, P. Zhang and W. Yang, Adv. Mater. Sci. Eng., 2018, 2018, 5085031.
51. N. Nurdini, M. M. Ilmi, E. Maryanti, P. Setiawan, G. T. M. Kadja and Ismunandar, Heliyon, 2022, 8, e10377.
52. E. Maryanti, M. M. Ilmi, N. Nurdini, P. Setiawan, Y. M. Syah, C. Saiyasombat, G. T. M. Kadja and Ismunandar, Archaeol. Anthropol. Sci., 2022, 14, 122.

---