micro x-ray fluorescence, micro-spectrometry, semi-quantitative analysis, archaeometry, polycapillary lenses, 2D XRF mapping

Advances in x-ray techniques, including x-ray optics, have paved the way to obtain challenging results in several research fields thanks to the improvement in terms of spatial resolution. This is particularly true for x-ray fluorescence (XRF), where the combination of conventional x-ray sources with polycapillary optics has permitted to have high flux and high focused beams.

However, XRF spectroscopy applied to archeological samples at a lab scale is mainly dedicated to qualitative studies. At the same time, quantitative analysis still remains a strong hurdle mainly due to important matrix effects that affect the signal related to the chemical components under evaluation. In this respect the adoption of x-ray optics on both the source and the detector represents a way to improve the signal to noise ratio, necessary for quantitative analysis.

At LNF XLab Frascati the expertise, gained on x-ray techniques and on polycapillary lenses, has allowed researchers to carry out advanced µXRF studies. RXR (Rainbow X-ray), is the experimental station dedicated to 2D/3D XRF micro-imaging and TXRF analysis, being equipped with 2 detectors of different energy efficiency (covering a full spectrum from 800 eV to 25 keV) and working in confocal mode with the source coupled with a full-lens and both the detectors combined with dedicated half-lenses. This report aims in depicting the RXR potentialities through the results obtained in 2 case studies dedicated to carry out a semi-quantitative analysis of 2 different artifacts (an ancient book, a Buddhist scroll) by µXRF characterization.

XRF spectrometry is a consolidated non-destructive technique to analyze the elemental composition of various samples that provides both qualitative and quantitative information. It is particularly useful in the field of archaeometry, the scientific branch of the archaeology consisting on the chemical and physical characterizations study of artifacts, given its non-destructive property. In addition, XRF analysis deserves particular attention because of simple instrumentation and rather easy qualitative data analysis.

Nowadays, a number of layouts, compact and portable, for in situ measurements are commercially available. These devices enable the sample spatial scanning to collect the information on the relative presence of various elements as a distribution map. However, due to the difficulty in handling x-ray radiation, these devices are characterized by low spatial resolution. Fine XRF studies are made possible only at large SR (synchrotron radiation) facilities, with concurrent higher experimental costs and longer time of analysis. In this respect, the development of x-ray optics as polycapillary lenses (polyCO), consisting of bundles of hollow glass channels delivering x-ray photons by multiple internal total reflections, has allowed to overcome these limitations. By this way, polyCO constitutes a very highly efficient tool for x-ray focusing, given their capability of delivering high flux beams emitted by conventional x-ray tubes, and enabling researchers to improve the spatial resolution of XRF measurements down to the micro scale size.

Even though XRF analysis is widely and routinely used for quality studies, on the contrary the quantitative evaluation of archeological samples still remains a strong hurdle in XRF Spectrometry, mainly due to significant matrix effects that influence the registered signal related to the chemical components under study.

In this respect the adoption of polycapillary optics on both source and detector represents a way to improve the signal to noise (S/N) ratio and hence to facilitate the determination of an almost quantitative analysis of the scanned sample.

The aim of the present work is to show the results obtained during the study of chemical elements that were present within two different archaeological artifacts (namely “Ms. 92”, “Buddhism Scroll 142 838” and “Scroll with Horses 142 846”) by means of XRF spectrometry, using the RXR (Rainbow X-ray) layout presently established at XLab Frascati, to make a semiquantitative analysis through the evaluation of fluorescence parameters, and to obtain indirect information on the provenience of the raw materials, the dating of the artifacts, and the manufacturing techniques.

Since the end of 90s, the National Laboratories of Frascati (last decade within the XLab Frascati laboratory) have been intensively involved in research projects dedicated to the design and characterization of customized layouts, such as ‘‘source-optics-detector”, for advanced applications in the fields of x-ray spectroscopy and x-ray microscopy.

The RXR station is the last facility that has been designed and developed for advanced x-ray micro fluorescence studies on 2- and 3-dimension stages (2D/3D µXRF), with an averaged spatial resolution of about 80 µm for the energy range 0.8 keV – 30 keV. This setup allows us to work with different experimental layouts including the one for total reflection XRF analysis (TXRF) to detect low concentration elements.

The RXR layout is based on a confocal geometry design that allows performing the elemental depth profiling for different types of samples and it provides information on 3D elemental distribution for a sample body through virtual stratigraphic sections. In addition, the “confocal” approach simplifies the measurements and permits to discriminate the depth dependent signals from high to low Z elements, avoiding also the negative superposition of images. The x-ray source is a conventional tube with Mo anode, maximum voltage 50 kV, and current 1 mA. It is coupled with a full lens, which produces a focal spot size of 50 µm in diameter. As a single polycapillary optics cannot be efficient in a wide energy range, two separated detector/half-lens couples are needed. In this way, it is possible to have one detector, combined with a semi-lens, efficient in the low-energy range (800 eV–5 keV) and another detector suitable for higher energies (4 keV–25 keV).

This layout allows users to perform fluorescence 2D/3D mapping for a wide energy range (starting from Na). The half lenses are combined with two SDD 30 mm² XGLab detectors 2 with a Be window for high energy, and a polymeric AP3.7 one – for low energy. The monocular microscope has a 0.75X to 3X microscope body and a 20X eyepiece. It is coupled with a CCD camera4 with 30-fps frame rate and 748 X 576 pixels resolution. The system has a three-axis xyz Bosch support structure, with a movement range of 300, 400, and 500 mm, respectively, with micrometric resolution, and the load capacity of about 100 kg. Over the Bosch support, an additional fine xyz positioning system permits carrying out 3D scanning with microscale accuracy, within the range of 30 mm for xy axis and 20 mm for z axis.

Fig. 1 Pictures of two artifacts studied by XRF Spectroscopy housed in the RXR facility: The Ancient Book (up left picture) and the Buddhist scroll (bottom right picture)

As previously anticipated, the quantitative analysis of XRF spectra is still an unsolved issue, mainly due to the “matrix effects” to which x-ray fluorescence is subject and that must be corrected in order to obtain accurate results. These effects are due to both absorption and enhancement effects, which are a consequence of the fact that both the target analytes and the matrix absorb and fluoresce in the x-ray region, thus affecting the magnitude of the signal.

Typically, the hand-held energy-dispersive x-ray analyzers dedicated to non-technical personnel (such as archeologists), are equipped with software using algorithms that automatically apply the matrix effect corrections. Although this is a very user-friendly approach, it typically does not allow the user to correctly evaluate the factors involved in the x-ray fluorescence process to obtain an adequate quantitative (or semi-quantitative) analysis. On the contrary, with the purpose of analyzing the artifacts through XRF spectrometry by means of our RXR apparatus to extract semi-quantitative evidences, we followed a methodology that has taken into account the following key factors as affecting the intensity of each XRF line registered during the scan.

• The cross section related to the photo-ionization process, pi, that is the probability that the electron is excited;
• The transition probability, f, related to the electrons moving from each atomic level above the vacant state;
• The probability of the radiative emission, that is the probability that the de-excitation transition determines the emission of a fluorescence photon.

As well-known, the emission of a fluorescence photon is a process subsequent to the absorption, that determines the emission of an electron belonging to one of the deep levels K, L, M, etc., and to the creation of a gap in a state of core. In particular, the value of the absorption coefficient is determined by the sum of the absorption by all the levels with binding energy lower than the energy of the incident photons.

For example, in the case in which the excitation energy is higher than the threshold K, we have for the absorption coefficient the following sum:

where µ_K and subsequent are the partial absorption coefficients for the different atomic levels K, L, M, while the second equality considers that the total absorption coefficient is constituted by the elastic (τ) and inelastic (coherent and inconsistent, σ_coh and σ_inc) cross sections.

The absorption ratio at the threshold r_K is defined as the fraction of the atomic absorption coefficient due to a particular absorption threshold, with respect to the photoelectric absorption coefficient due to the remaining thresholds. The r_K value can be obtained by measuring the absorption cross section before and after the threshold.

where τ_K +△ and τ_K-△ are the values of the photoelectric absorption coefficient measured before and after the absorption threshold. The values of the absorption ratio, constant as a function of the energy of the incident photons, are tabulated for each atomic species. For the region with an atomic number between 11 and 50 the trend for the K level can be expressed as a first approximation by the equation:

The probability that an incident photon excites electrons from a K level is given by p_K=(r_K-1)/r_K, an amount that must subsequently be multiplied by the mass absorption coefficient µ to obtain the overall probability. The second term to consider is the probability that, following the creation of a core holiday, a transition from a particular electronic level will be observed. This f_K-L2,3 value varies for each threshold and for each element.

The third factor is the probability of emission of a fluorescence photon w, which takes into consideration the two possible mechanisms of transition of the electrons from the levels above the empty one: the simultaneous emission of a fluorescence photon, or the emission of an electron secondary, due to the Auger effect, which is then emitted by a third atomic level placed at energies close to the energy of Fermi. The two mechanisms are competitive: if the fluorescence yield parameter ω is defined as the probability that a photon is emitted and the electron yield parameter η is defined in the same way, we can consider the identity

In particular, the parameter ω_K, which defines the emission probability of a photon K, is an increasing function of the atomic number. In the region of light elements, with Z <30, the probability of radiative emission is always less favored than non-radiative emission, and vice versa. Hence, the intensity of the fluorescence lines related to the i-th species can be written as:

where ρi is the concentration of the species i in the sample.

The proportionality of the equation takes into account many factors: geometric parameters, the angle of incidence and detection and matrix effects (this is a very important factor in case of thick samples as this aspect involves the phenomenon of re-absorption of secondary emissions by of the surrounding atoms).

Besides, the analysis of XRF data has to consider the parameters for the intensity correction due to effects of reducing the efficiency of low energy detector, which can be summarized as follows:

• the detector’s efficiency;
• the fluorescence efficiency which varies according to the emission threshold, the energy of the incident photons and the atomic number of the emitting element;
• the threshold absorption factor (determining the fraction of the photons absorbed at the threshold of interest;
• the relative line intensity, which determines the probability of emission of a particular fluorescence line (in particular K-L2,3);
• the mass absorption coefficient of each element as a function of the incident energy (Molybdenum x-ray tube with voltage 40kV, current 500 µA and polycapillary focusing lens).

Given all the considerations above and looking at our XRF experiments on the artifacts, we decided to exclude the matrix effects in our calculations because we have only conducted superficial XRF scan (the thickness of the scanned area was less than 100 µm of painting), in order to avoid any contribution from the bulk material. This simplified approach has led us to achieve sufficient information on the material composition that, in first approximation, can be considered as a semiquantitative analysis that resulted to be helpful to setup the further restoration procedure.

Before applying our methodology to the target samples, we tested our system on a Fluxana standard, in particular the FLX-S6M monitor glass. This standard is made with a very large list of elements (around 40 elements, from B₂O₃ to Yb₂O₅ to name a few), most of which are present in low concentration, that has made the qualitative and quantitative XRF analysis not easy both using our system and by means of the PyMCA supporting software.

During the experiment, the sample was placed on the focus of the main RXR lens, with beam focused less than 90 µm.

The acquisition time was set to 1 minute, that was the minimum time range to have a statistic of good quality, while the x-ray tube parameters were 40 kV as voltage and 500 µA as current. We applied the same experimental conditions during the characterization of both the archaeological samples.

To obtain the parameters described in the previous chapter it is necessary to know the excitation energy impinging on the sample, that is related both to the emission of the x-ray tube and to the polyCO efficiency with respect to varying energy, and the response of the detector (i.e. absorption of the window). The first parameter was obtained by placing the detector in front of the direct beam, attenuated by a pinhole in heavy material (lead), while the second parameter is supplied directly by the seller. The two parameters that we have taken into consideration are shown in Figure 2.

Fig. 2 The graphs representing the behavior both the detector efficiency (blue dotted line) and of the source emission (orange dotted line) vs. varying energy (keV).

Since the absorption coefficient µ is a function of energy, µ=µ(E), the fluorescence parameter for each emission line and for each element was weighted for the entire emission spectrum of the source/optical system. Consequently, we have calculated the final parameters for the lines K_α, K_β ed L_α, that are shown in Figure 3.

Fig. 3 Plot of the three emission lines Ka, Kb ed La with respect the varying Z atomic number, calculated following the proposed methodology

The evaluation of the semi-quantitative analysis, that does not consider the Matrix effect as approximation due to the “superficial” scan performed, has been compared with the open-source software PyMCA. In Tab. 1 the concentrations in percentage for the most abundant elements obtained with our analysis system (both without and with corrective XRF parameter), with PyMCA and the tabulated parameters are shown.

Table 1. Percentage of concentrations for the most abundant elements obtained with our analysis (both without and with corrective XRF parameter), with PyMCA and the tabulated parameters

The overestimation of Magnesium, Potassium and Calcium are mainly due to the following reasons: the K_α line of Mg is energetically very low and almost at the limit of the detector used, while the overestimation of K and Ca is due to the overlap of several L lines of elements that were present in Fluxana, and these lines are extremely difficult to separate and individually evaluate.

The first artifact studied with our XRF layout was an antique book “De bello italico adversus Gothos libri” (hereinafter Manuscript 92 – Ms. 92) attributed to Leonardo Bruni and dated back to XV century, belonging to the private collection of the “Casanatense Library”, located in Rome. We conducted XRF scan of different parts of the book, in particular the cover and some points within target pages, to primarily study the composition of the ink in order to evaluate the presence of the chemical elements and make hypotheses for the further phase of restoration of the damaged pages. The experimental layout of the station includes a focusing polycapillary lens with a 90 m FWHM spot size matching the Mo x-ray tube (Oxford Apogee 5000), a detector combined with dedicated semi-lenses (from X-Channel Techn.), an alignment system and an optical microscope with a CCD camera (Basler acA750-30gc). To align the sample in the vertical plane and to scan the surface we use a laser profilometer with a spatial resolution of 10 m (Micro-Epsilon NCDT 1401). The software adopted for control instrumentation and data acquisition is based on LabVIEW. The experimental conditions for each scan performed were set as 600 sec of acquisition time with a current of 800µA at a voltage of 45kV. In figure 4 we show the typology of samples analyzed by XRF and the related spectra collected.

Fig. 4 The picture shows the XRF spectrum collected by scanning the word “Laodomi” of a target page within the book, that is visible in the box on the up-left part of the spectrum

Generally, this chemical composition indicates iron-based metallic inks, presumably iron gall ink, that was used continuously from the Middle Ages until the 19th century. However, the presence of lead traces suggests a different ink composition, as it is not present in the rest of the text that was dated to the 15th century by the restorers. Moreover, the writings found in these papers, especially the one just described, showed differences in the layout and execution of the graphic signs and the ink is more marked and darker in some points with respects other ones.

This confirmed the hypotheses that the book has been managed by various readers and owners of the code. This assumption was also accompanied by the evidence of the compositional difference of the inks, which can be explained with a change in their preparation, that is related to different periods of time occurring subsequently to the period during which the first author started the writing of the book.

The objective of this study was the recognition of the chemical components present within the different inks revealed on Buddhist scrolls (hereinafter called emakimono 142 838 and “Scroll with Horses 142 846”) dated back to XVI century and coming from the Pigorini Museum of Rome thanks to the cooperation with the private collection of “Fondazione Ragusa”.

We carried out 2D scan of several points within different scrolls and we report, as example, the analysis of the green area shown in Figure 5:

Fig. 5 Image of one of the Buddhist scrolls examined by XRF Spectroscopy and a particular point (red circle) within a green area

The analyzed point, classified as “dark green”, showed a high signal attributed to the presence of copper. Together with visual examinations of the pictorial surface by microscopy analysis, it was possible to hypothesize that the author has used malachite, known also as rokusho. Copper carbonate, whose formula is Cu₂CO₃ (OH)₂, in naturally produced as a secondary mineral, nearby copper deposits and it is typically associated with azurite, cuprite and chrysocolla. Furthermore, it can be synthetically reproduced by combining the sulphate of copper with calcium carbonate, lime and ammonia salts.

This hypothesis has been confirmed by the comparison of the spectrum of the material, obtained using the non-invasive FT-IR ATR technique, and other sources extracted from the literature.

Fig. 6 XRF spectrum of the green ink analyzed during the characterization of the Buddhist scroll

Given the XRF results and following the methodology discussed in the previous sections, we evaluated the presence of the chemical species that are reported in Table 2.

Table 2. Calculation of the percentage of chemical elements found on the samples by means of the proposed approach

In this work we have presented the results obtained with XRF µSpectrometry through the RXR facility present at XLab Frascati, that is based on the combination of polyCOs with conventional x-ray sources, for the study of antique artifacts.

In particular, the objective of the work was the determination of chemical components of the inks present within several areas of two different sample typologies (an ancient manuscript and some Buddhist scrolls). As discussed in the paper, we aimed to the extraction of semi-quantitative results following a methodology that was primarily tested on an XRF standard, apart from the qualitative recognition of the elements constituting the inks. We reported the experimental results achieved on both the artifacts by mentioning two examples of points examined (the word “laudomi” of the manuscript and the green area of the Buddhist scroll), showing the XRF spectrum and the calculation of the element concentration (in %), that has led us to confirm the hypotheses coming from the bibliographic history of the artifacts, in order to fine-tune the further restoration procedure.