Sorting Porcelain: Using micro-XRF in a Provenance Study on Sèvres Biscuit Figurines

No matter which type of Cultural Heritage material we are studying, we might come across objects of unknown provenance in need of historical allocation. An easy way to identify the origin and authenticity of an object in doubt is to assess the production marks and artistic style. However, identification based solely on these factors can contain uncertainty. In these instances material analysis can aid clarification. 

In the case of ceramics and porcelains, trace elements such as Rubidium, Yttrium, Strontium and Zirconium can give a clear answer to the origin of the raw material used. Present in the ppm range, their ratios can be markers for a specific lithology. The origin of the raw material can give information not only on the different porcelain manufacturers, that used specific mining sites, but also on the dating, as over time, different deposits were exploited. 

For decades X-ray fluorescence analysis (XRF) has been a well-established technique for the analysis of a wide variety of Cultural Heritage materials. Due to its non-invasive nature and its capabilities of solving diverse research objectives in-situ without a lot of effort or sample preparation, it is well suited to be the analytical technique of choice for provenance studies.

The following case study displays common workflows for conducting provenance studies on porcelain using the ELIO, a portable micro-XRF spectrometer designed for cultural heritage applications. The analysis was conducted in cooperation with Viviane Mesqui, Conservatrice du patrimoine, Sèvres Manufacture et Musée Nationaux, and Philippe Colomban, em. CNRS Research Prof. at Sorbonne University, Paris at the Sèvres Manufacture et Musée Nationaux in Sèvres (Fig. 1). 

Fig. 1: Images of the measurement campaign with the Bruker ELIO in the storage facility of the Sèvres Manufacture et Musée Nationaux.

Images courtesy of Celine Lecomte

The study presented here was conducted on four Biscuit porcelain sculptures (Fig. 2). Biscuit porcelain, or bisque, is an unglazed, white porcelain known for its matte surface. It is and has been widely used in European pottery for sculptural and decorative pieces that do not require a protective glaze, resulting in a delicate appearance that was valued for artistic objects. 

A current acquisition of a biscuit figurine by the Sèvres Manufacture et Musée Nationaux (MNC 2023.4.1) is thought to be from the Sèvres production due to its style and the presence of a lily production mark. However, as there were uncertainties as to whether the object was produced in Sèvres or in Limoges, material analysis was desired for final confirmation. For comparison, the composition of the unknown sculpture was compared to two biscuit sculptures from the Sèvres workshop (MNC 2641 and MNC 21134) and a medallion (MNC 25341) from the Limoges porcelain manufacture. 

Fig. 2: The four biscuit pieces and their attribution. Object in doubt MNC 2023.4.1 is considered to derive from Sèvres due to its style and production mark. To verify this, XRF data was compared to statues from the Sèvres workshop (MNC 2641 and MNC 21134) and one biscuit medallion from Limoges (MNC 25341).

Images courtesy of Prof. Dr. Philippe Colomban

Provenance studies on ceramics, porcelain and glass objects are often conducted using trace elements such as rubidium (Rb), strontium (Sr), yttrium (Y) and zirconium (Zr), as the ratios of their concentration can be a geomarker for the raw material used. 

Quantitative micro-XRF analysis of these elements in light-element matrixes has serval benefits: Although matrix effects influence quantitative results, in the case of materials such as porcelain they do not have such a high impact: As the impact is proportional to the atomic number of the element as well as to its relative abundance in a sample and an average silica glass or a ceramic consist of >90 % of light and very light elements, the influence is negligible. Heavy elements such as Rb, Sr, Y, and Zr are only present in traces in the ppm range. As a result, there is a linear correlation between concentration and intensity (Fig. 3a).

Fig. 3a: Linear correlation between the intensity of high-energy traces and concentration.

Fig. 3b: Calculated information depth in different matrixes for selected emission lines

In addition, these elements show great properties for XRF analysis, as their emission lines are located between 13-18 keV, yielding a high information depth as a result. Information depth is here defined as the depth from which 90% of the generated intensity can escape the sample. Going to numbers, in the case of characteristic K-lines of Rb, Sr, Y, and Zr, these would be around 800 µm in a light element matrix (Fig. 3b). This means, that the intensities we want to correlate to specific concentrations do not depend on surface patina or environmental deposits. Moreover, in the case of the biscuit figurines, no attenuating lead-rich glaze must be taken into consideration. 

XRF analysis was conducted using a portable Bruker ELIO system, equipped with a 4 W Rh-anode X-ray tube and a 50 mm2 SDD. The ELIO’s primary beam is collimated to a fixed spot size of 1 mm. Due to the flexible set-up and the compact measuring head, the system can be easily approached to three-dimensional sculptures. Excitation conditions were 50 kV and 80 µA at 180 s real time. 

To enhance the signal to noise-ratio, the primary radiation was optimized using a filter based on layers of Cu (25.4 µm), Ti (25.4 µm) and Al (254 µm) (ELIO filter no. 3) (Fig. 4 a). To maintain the sensitivity for lower emission lines, point data without filtering of the primary radiation was collected. The homogeneous occurrence of the elements in question was checked by acquiring multiple points. Data was processed using the Esprit Reveal Software for calculating net intensities (cps) using a Fundamental-parameter based forward calculation and correcting for dead time (Fig. 4 b). Net intensites of all four sculptures was exported to Excel and ratios of the different trace elements plotted in binary plots. Intensities of high energy traces (Rb, Sr, Y, Zr) are based on measurements with a filter, while net intensities of lower emission lines (K, Ca and Pb) are extracted from point spectra collected without filtering of primary radiation.

Fig. 4a: Using a strong filter, the peak to background ratio can be enhanced in the energy range between 13-18 keV.

Fig. 4b: Net intensities (cps) used for comparing the spectral data of the four objects were extracted using a Fundamental-parameter based forward calculation.

XRF analysis confirms that the two Sèvres sculptures and the unknown piece are hard paste porcelain based on china clay. In all three cases, the biscuit shows an enrichment in lime (Ca), which is well in accordance with the raw material used in Sèvres. On the other hand, the Limoges medallion has a considerably higher K content (Fig. 5, K/Ca binary plot). 

Studying the various ratio plots of both matrix-relevant compontents, e.g. K and Ca, to high energy traces of Rb, Sr, Y, and Zr, the unknown sample (highlighted in orange) is closely located to the net intensities obtained for Sèvres biscuit (marked light blue). On the other hand, data from the Limoges medallion cannot be assigned to the same group (Fig. 5). XRF data thus unambiguously supports the assumed ascription of the unknown object to the Sèvres manufacture.

Fig. 5: Ratio plots of the various high-energy traces. A clear correlation between net intensities of Sèvres and the unknown sample is notable. Intensities of high energy traces (Rb, Sr, Y, and Zr) are based on measurements with a filter, while net intensities of lower emission lines (K, Ca, and Pb) are extracted from point spectra collected without filtering of primary radiation.

XRF analysis is a quick and reliable tool for provenance studies on a wide variety of Cultural Heritage objects. The matrix effects for high-energy lines are negligible when dealing with light-density matrixes such as ceramics, glass or stone. As the intensity and concentration of materials such as porcelain, stone or glass show a linear correlation, a reference-based type calibration could be used to further assign the net intensities extracted to absolute values. With this information the data obtained can be compared to measurements executed by other research groups that use different equipment. 

Further details about this case study are available in a paper, co-authored by experts from Bruker, published in the journal Ceramics under the title "Chemical and Vibrational Criteria for Identifying Early Sèvres Factory Porcelain Productions".