2018 atomic spectrometry update – a review of advances in X-ray fluorescence spectrometry and its special applications

Contents

• 1 Introduction

• 2 Chemical imaging using X-ray techniques
  • 2.1 Computed tomography and 3D XRF techniques
  • 2.2 Laboratory 2D XRF techniques

• 3 Synchrotron and large scale facilities

• 4 Grazing X-ray techniques including TXRF

• 5 Hand-held, mobile and on-line XRF techniques
  • 5.1 Hand-held and mobile XRF techniques
  • 5.2 On-line XRF techniques

• 6 Cultural heritage applications

• Conflicts of interest

• Abbreviations

• References

---

Abstract

This review describes advances in the XRF group of techniques published approximately between April 2017 and March 2018. The review is selective with the aim of providing a critical insight into developments in instrumentation, methodologies and data handling that represent a significant advance in XRF spectrometry. It is not the intention of the review to cover comprehensively the applications of XRF techniques except in those cases where the non-destructive and remote sensing nature of XRF analysis makes it particularly valuable and the method of choice. These applications concern samples which are irreplaceable and of great cultural value such as works of art and archaeological artefacts.

---

1 Introduction

Regular readers of this review will note that Margaret West has stepped down as co-ordinator. Margaret has contributed to this update as a writer for almost two decades and has been co-ordinator since 2007. Her enthusiasm and depth of knowledge of XRF techniques has helped galvanise the writing team which acknowledges the encouragement she provided in developing the skills of new members of the team. In addition, Christina Streli has stepped down as a writer. She has been a stalwart of the team for many years and we thank her for her valuable contributions.

Under Margaret’s guidance this ASU has been transformed from one that covered all aspects of XRF techniques including applications to one that focuses primarily on the major advances in instrumentation, methodologies and data processing. There is no longer an attempt to cover all aspects of applications. Only in those cases where XRF spectrometry is the method-of-choice and is unlikely to be considered in the application ASUs do we continue to highlight the unique advantages of remote sensing and non-destructive analysis. Such applications are primarily for the purpose of analysing samples of cultural value such as works of art and archaeological artefacts which are irreplaceable. For cultural heritage applications the use of portable and mobile XRF systems is well established and indispensable. Both macro-XRF scanning devices and μ-XRF imaging systems, combined with hyperspectral imaging, are used in the investigation of historical paintings. The challenge now is how to handle the impressive amount of data that these systems generate. For other applications, readers are referred to the ASUs on the analysis of environmental samples,¹ clinical and biological materials, foods and beverages² and metals, chemicals and materials³ as well as the ASU on elemental speciation.⁴

In addition to the changes in overall focus away from general applications, the long-standing separate sections on X-ray detectors, sources and data processing have now been removed and developments in those areas will now be covered in the section that they most closely align with. Furthermore, the section on laboratory instrumentation will now focus on 2D imaging, which is the most active area of development now that laboratory XRF spectrometers for bulk analysis are ubiquitous and in common use for a huge variety of applications.

Over the last few years there has been a marked shift in emphasis in the world of XRF research. This is particularly noticeable in the use of TXRF techniques for which commercial instrumentation is available and so emphasis is no longer on developing the technique but on applying it. This trend is highlighted by the marked reduction in the number of fundamental papers covered in this ASU. On the other hand, the current lack of commercial instrumentation for grazing techniques means that much of the development in this field remains fundamental and has seen substantial growth. The optimisation of grazing incidence angle techniques (GI-XRF) for nanomaterial characterisation has made layer thickness resolution of around 0.5 nm possible. Two beamlines, at SOLEIL and ELETTRA, are available for users to perform research on GI-XRF spectrometry combined with XRR. Direct sampling is another area of growth. The direct collection of particle-size-resolved aerosol samples on a quartz sample reflector allowed short sampling times of <1 h and LODs of a few pg m⁻³ for medium Z elements.

The use of XRF-CT continues to gain attention as a powerful element imaging tool both with respect to its developments and applications. While full 3D scanning was once considered prohibitively time-consuming, the development of fast scanning approaches using ED-detectors with high count rate capability has allowed XRF-CT to evolve towards a truly 3D element imaging technique. The use of SR-based sub-μm beams is becoming especially popular as this allows analysis of samples such as environmental particles or cryogenically frozen biological cells measuring only a few tens of μm in size. Because samples investigated using SR nanoprobes are of truly microscopic dimensions, the serious limitations associated with sample self-absorption effects when using XRF-CT are considerably mitigated and limited to elements having relatively low atomic numbers. In addition, new fast scanning approaches can be applied when the extreme beam intensity available at synchrotron sources is coupled with efficient XRF detection.

A list of abbreviations used in this review appears at the end. Although these abbreviations may not always be used in a strictly grammatically correct manner, the writing team feels that it is important to use terminology that would be easily recognised and understood by the XRF community.

2 Chemical imaging using X-ray techniques

2.1 Computed tomography and 3D XRF techniques

In an innovative spiral scanning approach for XRF-CT described by de Jonge et al.,⁵ the overheads intrinsic to these types of scanning measurements were reduced to virtually nil. This made possible fully fractionated tomography or higher dimensional methods such as volumetric tomography and XANES tomography. In addition, the spiral scanning method was combined with a Fourier ring correlation (FRC) analysis procedure to explore sources of resolution degradation. The extension to the FRC formalism enabled direct determination of the spatial resolution from the measured XRF sinogram data, thereby greatly enhancing the power of this type of tomography as a diagnostic tool.

In another interesting approach for accelerating data acquisition in XRF-CT experiments, Sasaya et al.⁶ investigated the feasibility of in vivo imaging of gold NPs for preclinical use. Full-field detection with a 2D pixelated detector (487 × 195 pixels of size: 172 × 172 μm) was used with a multi-pinhole collimator rather than the conventional single pinhole approach. The use of multi-pinhole collimation with a 2D detector and parallel beam volumetric illumination increased considerably the rate of acquiring multiple projections simultaneously and improved S/N of the projections. This preliminary study demonstrated that by using an incident X-ray energy of 25 keV and a photon flux of 5.0 × 10⁸ photons per mm² per s, the multi-pinhole XRF-CT approach achieved a data acquisition (real) time of 1200 s at a theoretical LOD of 0.1 mg mL⁻¹ Au and a spatial resolution of 0.4 mm. A similar approach to improving detection efficiency was taken by Zhang et al.⁷ The design concept and initial simulations were presented for a polychromatic full-field fan-beam XRF-CT device which was based on the use of pinhole collimators and linear-array photon-counting detectors. Monte Carlo simulations showed that regions containing 0.065% of Au and 0.04% of Gd by weight could be reconstructed. These reconstructions corresponded to a full scan time of 6 minutes. In comparison with an XRF-CT system with a pencil-beam source and a single-pixel detector, this full-field fan-beam solution with linear-array detectors resulted in significant scanning-time reduction and thus had potential for rapid in vivo element imaging.

In vivo element imaging optimised for the detection of metal NPs or clusters is an emerging research area with important medical applications. For example, the use of metal NPs as imaging contrast agents and radiosensitisers for tumour targeting and drug delivery in cancer therapy and diagnosis requires new methods for their 3D detection in vivo. Jung et al.⁸ demonstrated the feasibility of full-field XRF-CT-based K-shell-imaging of gadolinium oxide NPs and gold NPs in water phantoms using a pinhole-based XRF imaging system and polychromatic excitation. For a cylindrical water phantom of 5 cm diameter and height, the minimum concentrations for providing detectable contrast-to-noise ratios were 0.03 and 0.01% by weight for the gold and gadolinium oxide NPs, respectively. Jiang et al.⁹ carried out detailed Monte Carlo studies using the Geant4 simulation toolkit on a similar system using polychromatic excitation with a sheet-beam geometry to simulate the entire XRF-CT imaging process. Using PMMA phantoms loaded with gold NPs of various diameters and with various Au concentrations, the feasibility of quantitative XRF-CT on Au was demonstrated using a system that included polychromatic sheet-beam excitation and detection with a parallel collimator array in conjunction with a linear array of energy-resolving photon counting detectors. Particularly interesting was the quantitative evaluation of calculated Au concentrations by using (a) filtered back-projection and (b) maximum-likelihood-expectation reconstruction algorithms with and without self-absorption correction. A full-field fan beam XRF-CT setup corresponded¹⁰ to a large extent to the Monte Carlo simulation work of Zhang et al.⁷ and was based on the use of conventional X-ray tube excitation and detection by an energy-sensitive photon-counting detector array coupled with a single tungsten pinhole collimator. A phantom containing Gd solution could be scanned in 30 minutes using a polychromatic X-ray fan beam in a third-generation CT geometry, requiring only a single rotational scan for full 3D analysis. This was faster than typically achieved by other XRF-CT devices based on the use of a pencil beam in conjunction with a CT geometry and thus had potential for performing biomedical imaging applications.

Another direction of development of XRF-CT is represented by the addition of a number of complementary imaging modalities, based on various detected signals and optimised for various applications. George et al.¹¹ described an imaging platform which combined CT techniques based on the use of XRF, X-ray luminescence, X-ray Rayleigh scattering and X-ray transmission for monitoring and stimulating the therapeutic delivery of metal-containing NPs. In a demonstration of the feasibility of this approach for imaging metal-containing NPs used in X-ray-activated photodynamic therapy techniques, a gel sample containing three 600 μm diameter channels of Y₂O₃:Eu³⁺ NPs was scanned in both directions perpendicular to a 17.4 keV monochromatic beam with 200 μm step sizes over an 8 × 10 × 5 mm volume. An integrated approach combining X-ray μ-CT, XRF-CT and confocal XRF was demonstrated by Laforce et al.¹² Integrating μ-CT and element sensitive 3D XRF techniques in a single laboratory instrument proved to be a powerful tool for obtaining coupled 3D morphological and element distribution information from bio- and geo-materials. The spatial resolution was 10–20 μm for element imaging using confocal XRF and XRF-CT and 1 μm using μ-CT. Unlike typical microbeam scanning XRF instruments, this setup did not use an optical microscope for sample alignment. Instead, the first step of analysis was a fast μ-CT scan to explore the 3D morphology of the mm-sized sample and to decide which regions or volumes should be selected for subsequent element micro-imaging. The LODs using the XRF-CT and confocal XRF modules for element analysis in the atomic number range of 19–37 were 1–100 μg g⁻¹ for an acquisition time of 1000 s.

The (quantitative) reconstruction of XRF-CT datasets remains hampered mainly by self-absorption effects of the characteristic lines of the elements of interest within the (heterogeneous) sample matrix. Di et al.¹³ combined XRF and transmission CT datasets in a promising joint inversion approach for solving this problem. By reformulating the respective (forward) models of XRF and transmission CT, a link was established between these modalities so that they shared a common set of unknown variables. An iterative maximum likelihood algorithm, used to estimate the element distributions from the datasets obtained with two modalities (XRF and transmission CT), assumed that they both followed Poisson statistics. These data were then used to update the self-absorption terms in the forward model. The reconstruction method was tested on a borosilicate glass rod of known composition wrapped with gold and tungsten wires of 10 μm diameter and scanned by a 12.1 keV incident beam. Significant improvements in reconstruction quality were achieved by performing this joint inversion when strong self-absorption effects were present. In a similar approach for the quantitative reconstruction of trace-element distributions within low-Z samples investigated by XRF-CT, Huang et al.¹⁴ combined fluorescence and transmission CT data. The simple and practical iterative method used widely available transmission tomography reconstruction software for the fluorescence tomography data. The XRF sinograms were initially corrected for self-absorption effects using element-dependent mass-attenuation-coefficients calculated from sinogram asymmetry, achieved typically within 3–5 iterations. Subsequently conventional transmission CT reconstruction algorithms (such as filtered backprojection) could be applied to obtain absolute concentrations of trace elements. The procedure was tested in the trace element mapping of fish eye lenses of common bream (Abramis brama). Data obtained for air-dried samples were compared with those obtained by 2D μ-XRF scans of thin sections (280 μm thickness) using an X-ray beam of 20 μm FWHM and an incident X-ray energy of 16.2 keV. For the 6 elements studied (Br, Fe, Ni, Rb, Se, Zn), the concentrations of half of these converged to values only a few percent different from the expected results and all but one converged to values less than 20% different from the reference measurements.

A fundamental study by Viganò and Solé¹⁵ addressed general aspects of the reconstruction problems associated with emission and scattering tomography (including XRF-CT). They introduced a unified discrete representation that can be used to modify existing algorithms to reconstruct the data obtained from different types of experiments. The three induced-emission and scattering tomography modalities treated were based on detection using XRF, Compton scattering and XRD. The general nature of the approach allowed the underlying mathematics to be treated almost identically for the three modalities and to make the differences explicit only when required. The mathematical framework allowed modification and implementation of any kind of existing iterative algorithm to be made to reconstruct correctly the original sample structure. An especially powerful aspect of this proposed matrix representation was the ability to express the reconstruction algorithm as a minimisation problem.

Attention has been devoted to the fast and reliable handling of large multimodal tomographic datasets in order to be able to interpret the data as they are collected. Parsons et al.¹⁶ presented an accessible and flexible data processing framework, named Savu, that was able to handle both the variety and amount of data corresponding to multimodal and multidimensional scientific output such as those originating from XRF-CT experiments combined with X-ray transmission tomography. The software was demonstrated by the processing of XRF-CT and XRD-CT datasets generated by the microbeam scanning of a metal NP catalyst supported on graphite and loaded in a 400 μm outer diameter quartz capillary with a 10 μm wall thickness. Self-absorption correction of the XRF-CT data employed a simple version of McNear’s absorption correction.¹⁷ Ongoing work included the implementation of a HDF5 single-writer/multiple-reader Virtual dataset backend to achieve truly real-time data processing capability, application of PyMca for XRF fitting and the integration of various packages to treat coherent X-ray imaging datasets. A data-processing software (MMX-I) for multimodal X-ray imaging and tomography, reported by Bergamaschi et al.¹⁸ provided a new multi-platform open-source freeware for the processing and reconstruction of X-ray imaging data sets. The software could import raw data in HDF5 file format which is rapidly becoming a standard data format at various synchrotron facilities. The raw data could be reduced to 2D projections of the different modalities (e.g. absorption, dark field, differential phase contrast, XRF). These raw projections were further processed either by specific reconstruction methods (e.g. phase reconstruction for differential phase contrast) or by using tomographic reconstruction algorithms. The software package is now available for the user community.

In the context of using high intensity μ- and nanobeams of X-rays at SR facilities, Jones et al.¹⁹ presented extremely important guidelines on radiation dose limits for bioanalytical studies using scanning XRF imaging. As XRF-CT requires repeated scanning of the same regions to obtain the tomographic datasets, it is particularly important to take into account potential radiation damage effects in case of biological samples. Dose limits were established for three chemical-free specimen preparation techniques (lyophilisation, cryofixation and live). Below these limits, the μm-scale spatial distribution of specific analytes (Ca, Fe, K, Mn, Zn) was preserved. Element distributions within a cryofixed specimen were well maintained up to a dose exceeding 70 × 10⁶ Gy, whereas, in a lyophilised specimen some metal ion re-distribution was observed at doses exceeding 10 × 10⁶ Gy, concurrent with significant ultrastructural damage. Cell membrane rupture was the most plausible explanation for redistribution of metals in specimens hydrated at room temperature when doses exceeding 1.5 × 10⁶ Gy were used. From an analytical point of view, work by Lemelle et al.²⁰ which established sample preparation and data analysis requirements for quantitative XRF trace-element nanoimaging was seminal. The guidelines could be applied directly to the XRF-CT of metal traces in solid samples. Recommendations were given for the preparation of μm-sized samples and standards using adapted focused ion beam sample preparation. In addition to highlighting the need for appropriate sample preparation methods, the importance of highly detailed modelling of nano-XRF maps was emphasised for the accurate quantitative evaluation of nano-XRF imaging data-sets.

The use of XRF-CT combined with other modalities was nicely illustrated by sub-μm X-ray beam scanning applications in various scientific fields. For example, Sheppard et al.,²¹ in their impressive demonstration of in situ chemical imaging, employed XRF, XRD and STXM modalities to study Cu/ZnO/Al₂O₃@ZSM-5 (core@shell) catalysts used in the one-step conversion of synthesis gas to DME. Identical sample volumes were imaged stepwise with μm spatial resolution: first under oxidising and reducing atmospheres (imitating calcination and activation processes), and then under model reaction conditions for DME synthesis (H₂ : CO : CO₂ ratio of 16 : 8 : 1, up to 250 °C). The multimodal imaging methods provided insights into the active metal distribution and speciation within the catalyst and allowed differentiation between the catalyst core and zeolite shell to be made in a single imaging acquisition. The combination of data from several techniques (μ-XRF-CT, μ-XRD-CT, and STXM-CT) provided a more comprehensive characterisation and deeper understanding of the catalytic processes involved than could be achieved using a single acquisition mode or conventional (XRF, XAS, XRD or 2D) imaging. Metastable Cu₂O present during reduction was formed by interaction with the zeolite shell. A significant increase of mixed Cu oxidation states indicated the possible role of Cu₂O in methanol synthesis. Remarkably high resolution 2D and 3D element imaging results (50 nm voxel size for CT, 40 nm pixel size for XRF) for NP uptake studies by invertebrates were achieved by Cagno et al.²² in their study of the internal distribution of cobalt NPs in the nematode Caenorhabditis elegans. They employed synchrotron-based phase-contrast nano-CT combined with nano-XRF tomography. This study should pave the way for further nanotoxicology investigations based on nano-XRF-CT and phase contrast nano-CT.

Confocal XRF is another methodology for 3D XRF which has gained considerable attention recently. Based on the use of appropriate focusing and detector optics to detect elements present in a well-defined microscopic volume of the sample, the method can be used for depth resolved 1D/2D or full 3D element analysis of the material under investigation. It can also be extended to local XAS by employing energy-tunable SR sources. Several new (mobile) endstations, reported at various facilities, were optimised primarily for depth-resolved and/or 3D XRF/XAS micro-analysis. One such new endstation, established at the Beijing Synchrotron Radiation Facility, was dedicated²³ to chemical imaging studies on cultural heritage materials. A moderate confocal volume of 72 × 53 × 57 μm (FWHM) was measured at the Cu K-edge (8.979 keV). Another mobile setup, established²⁴ at the ESRF, used confocal μ-XRF and XAS in various scientific disciplines. The plug-and-play setup was based on polycapillary X-ray optics and enabled 3D confocal XRF and XAS of local micro-volumes of 8 × 8 × 11 μm (FWHM at 7 keV). Full analytical characterisation of the setup was presented together with two case studies: a geological study of inclusions in a deep Earth diamond; and a study in the field of cultural heritage, determining the composition of iron gall ink on a historical document. A laboratory setup at the National Laboratories of Frascati (Rome, Italy), named “Rainbow X-Ray” (RXR), was optimised for confocal XRF techniques. The RXR setup combined an X-ray tube (Mo, 50 kV, 1 mA) source with polycapillary focusing down to a 90 μm spot and two ED-detectors with polycapillary half-lens collimators. Depth resolution was 80 μm.

An interesting approach for improving the depth resolution of a conventional X-ray confocal microscopy system was proposed by Iida.²⁵ A thin wire was placed close to the sample surface, between the focus of the incident beam and the tip of the detection channel polycapillary. In addition, a slit that limited the horizontal acceptance (opening) angle of the detector was inserted behind the polycapillary. The resolution function of the combined system of the polycapillary and thin wire, determined as the product of the resolution functions of each part, gave improved resolution given that the focal size of the polycapillary was larger than the diameter of the thin wire. A detector polycapillary of 33 μm (FWHM at Fe-Kα) acceptance was used with a 10 μm diameter molybdenum wire. The FWHM of the depth profiles measured by the system equipped with the thin-wire were about 11 μm smaller than those obtained from confocal mode profiles (40–50 μm). This demonstrated that the polycapillary-wire mode could achieve considerably improved depth resolution. A new milestone for confocal XRF was represented by the work of Choudhury et al.²⁶ who used collimating channel array optics on the detection side to achieve superior spatial resolution for probing volumes down to 2.5 × 2.5 × 1.5 μm. The CCA optics made both confocal XRF mapping and XAS possible at a high spatial resolution of 2 μm. In comparison to polycapillaries, CCA optics do not suffer from energy-dependent spatial resolution effects and so can provide achromatic imaging for both low- and high-energy XRF. The CCA optics can therefore greatly enhance the current capabilities of confocal imaging in terms of spatial resolution, especially for light elements. One disadvantage of the CCA-based approach, however, is the particularly small working distance of 1–1.5 mm.

Alternative approaches to the use of confocal XRF have been reported. Tack et al.²⁷ combined a 1D-focused sheet beam with full-field ED pnCCD detection. An interesting comparison of the conventional confocal XRF approach and the full-field XRF type of sheet beam/pnCCD combination was provided by Rauwolf et al.²⁸ Although the spatial resolution using the latter combination was better than that provided by conventional confocal XRF setups, the system suffered from low detection efficiency arising from the detector optics used and so required extremely high brilliance SR sources for efficient use. Rauwolf et al.²⁹ used this confocal setup to measure abnormal levels of certain trace elements (e.g. Zn) in various cancer tissues. An in-house confocal setup comprised³⁰ a polycapillary full lens on the source (molybdenum target X-ray tube operated at 50 kV and 0.6 mA) and a half lens on the SDD. The system produced a 10 μm spot and a spatial resolution of 15–25 μm over the energy range 6–12 keV. The LODs for Fe, Mn and Ni were 25 mg L⁻¹ at the liquid surface and 70 mg L⁻¹ at a depth of 450 μm. The progress of dissolution of a steel sample into a saline solution was illustrated over a period of 10 days through use of maps. An innovative confocal XRF system for use at lower spatial resolution was based³¹ on a modified commercial 3D printer in which the extruder was replaced with an assembly composed of X-ray tube and a SDD. Use of simple collimator systems with internal diameters of 0.25–1.1 mm for both excitation and emerging fluorescence X-rays gave sub-mm spatial resolution in each direction.

Both improved approaches for absorption correction and more accurate quantitative data processing methods have been investigated for the evaluation of confocal XRF datasets. Liu et al.³² presented a beam-path (pixel)-based correction method that took into account the sample heterogeneity and the confocal XRF setup geometry. The former was addressed by assuming that the mass distribution of an element was the same as the XRF intensity distribution of the element across the pixels. A novel absorption correction procedure, based on measured absorption coefficients and element-specific mean-excitation-energies, was applied³³ to heterogeneous specimens with homogeneous matrix absorption. This procedure was particularly well-suited for biological samples with constant matrix composition and density and heterogeneous distribution of minor or trace elements. The linear absorption coefficients for the matrix, established by measuring spectra from depth scans of a floury endosperm sample, compared well with those obtained using FP calculations and assuming a cellulose matrix. The computation of source absorption correction was simplified by employing the long-standing approach of formulating an effective excitation energy rather than using the full excitation spectrum for each measured XRF line. A linear approximation of effective energy with depth was established for a sample depth of less than 1.5 mm. The authors described in detail how corrections were performed for each voxel and applied the correction procedure to 3D images of the top and bottom sections of millet seed to prepare 3D maps of the corrected concentration of trace elements in the seed. The impressive data set was collected with 15 s livetime per voxel, a 30 μm step size and approximately 30 steps in all three directions. Saving the entire X-ray spectrum for each voxel resulted in 27 000 spectra per map and a collection time of approximately 2 weeks! The simplifying assumptions in the correction algorithm allowed the data to be processed successfully and the maps provided an impressive ability to investigate the 3D micro-homogeneity of the trace elements in the studied millet seeds. The effectiveness of the correction was established by comparing 3D maps taken from opposite sides of the seed. A detailed investigation of any section through the seed was possible. A comprehensive quantitative reconstruction approach was presented by Szalóki et al.³⁴ for SR-based confocal-XRF-imaging. The theoretical approach used a generalised system of equations based on the FP method to calculate the 2D distribution of the trace, minor and major element concentrations within the sample cross-sections, corresponding to the measured confocal XRF dataset.

2.2 Laboratory 2D XRF techniques

Developments in laboratory-scale instruments and optics included an interesting study³⁵ in which ray-tracing methods were used in MATLAB to simulate X-ray beam paths through a single-bounce ellipsoidal capillary. The aim was to design and then fabricate an optimum X-ray optic for use in μ-XRF instruments. The optimal design of capillary was combined with a 50 W copper-target X-ray tube that had a 300 μm diameter spot and an SDD placed some 40 mm from the sample. The incident beam angle of 45° to the sample surface produced a 176 μm diameter circular spot. This setup was used to prepare an XRF map of a holly leaf using a dwell time of 20 s and a sampling step of 0.3 mm. The high intensity and long working distance of this setup and the ability to design optimised optics were valuable but the delivered spot size puts this in the realm of milli-XRF rather than μ-XRF spectrometry. A novel μ-EDXRF system with a much smaller spot size than in the above ellipsoidal setup was developed³⁶ for the in vivo XRF mapping of minor and trace elements in the leaves of living plants. The spectrometer incorporated a commercially available 50 kV, 1 mA rhodium target X-ray tube coupled with a commercially available polycapillary optic to deliver a spot size of 13.4 μm for the Rh Kα line and an effective beam diameter for the Fe K edge (7.111 keV) of 31.67 ± 0.15 μm. Detection was by means of a 1 mm thick SDD with an active area collimated down to 30 mm². Measurements (1000 s livetime) were made on a 13-element thin film prepared by drying a liquid RM to establish RSFs and LODs. The authors reported impressive LODs of 12–140 μg L⁻¹ and of around 15 fg absolute for Ca, Cu and Fe in a freestanding film. These concentration LODs were claimed to be similar to those obtained when using a fourth generation SR beamline. The key feature of this interesting spectrometer was its ability to hold and analyse live plants in order to establish their trace nutrient distribution on the microscale under different environmental stresses. High quality 2D maps were produced and the vitality of the plant before and after the XRF measurements was established by means of pulsed chlorophyll fluorescence analysis. Maximum photosynthetic efficiency in the mapped areas was clearly preserved when measurement settings were reduced to 1 s per pixel livetime and a 40 × 40 μm pixel size whereas, conversely, significant damage was observed when a smaller measurement area or longer dwell time was used. This approach offers the exciting prospect of measuring the distribution of trace and minor elements in healthy living plants in real time. On the subject of beam size, Gherase and Vargas³⁷ comprehensively described their method for determining the effective spot size delivered by a polycapillary lens for specific XRF peak energies. The data obtained for three wires of different metals using this scanning XRF method correlated well with data obtained from knife-edge measurements. The main benefit of this important work, however, was the ability to establish the effective beam spot size for specific characteristic XRF line energies. Such an approach will be particularly valuable for quantitative element mapping in confocal 3D geometries. An impressive full-field μ-EDXRF system described in a Japanese language paper³⁸ used a commercially available X-ray CCD with 1024 × 1024 pixels, each of 13 × 13 μm, and had a very respectable quantum efficiency of 70% at 6 keV and 40% at 8 keV. This detector was coupled directly to a linear polycapillary optic containing capillaries of 6.5 or 12 μm diameter. A molybdenum target X-ray tube operated at up to 40 kV and 20 mA was fitted with a collimator/shutter assembly to allow the source beam to be interrupted while the CCD array was read out and reset. The authors investigated the use of different timing and pixel binning/averaging methods and reported an impressive energy resolution of 140 eV and spatial resolution of 52 μm at 6.4 keV. Conventional calibration curve methods were used to good effect. The effectiveness and speed of the instrument, its operating regime and quantitative approach were demonstrated with quantitative maps of printed circuit board and steel samples. Wide-area μ-WDXRF spectrometry was used³⁹ to study the dissolution into HCl solution of Zn ions from solid zinc as it offered rapid measurements for a single element. Both these approaches have their merits but as both papers appeared in Japanese language publications only limited details are available. A portable μ-EDXRF system was claimed⁴⁰ to have benefits, such as ease of handling and relatively low cost, for the determination of elements in tomato plants and their fruit grown in a coastal environment. The instrument, originally designed for measurements on museum and art objects, was equipped with a molybdenum target X-ray tube capable of 50 kV and 0.7 mA and a SDD. The beam was collimated down to 200 μm diameter and the sample step size was typically 200 μm. Acquisition times of 800 s per step were used to map minor and trace elements but this resulted in a very long total acquisition time for each map. In addition, sample preparation such as washing and drying was required. Tomato fruits also had to be dehydrated for 3 days. The use of a mobile unit had no advantage whatsoever as the long mapping times necessitated use in a controlled, laboratory environment. Low-resolution XRF peak intensity maps were produced for nine elements using peak intensity divided by Compton peak intensity. Despite these being thin film samples, disappointingly there was no attempt to perform quantitative mapping so the results were of limited use for decision making.

The development of algorithms for processing spectra and performing matrix corrections in μ-XRF spectrometry, particularly confocal μ-XRF spectrometry, is challenging. Aida et al.⁴¹ used PCA to improve the quality of μ-XRF images for samples with low XRF P/B in trace element measurements. The authors used an in-house setup comprising a molybdenum target X-ray tube (50 kV, 0.5 mA) coupled to a commercially available polycapillary full lens to deliver a 10 μm spot at a working distance of 2 mm. Detection was by means of a 50 mm² SDD with an energy resolution of 130 eV at 5.9 keV and scans were taken using dwell times of 1 or 10 s and a step size of 10 μm over an area of 500 × 500 μm. The authors analysed the simple residues obtained from 5 μL drops containing 2.5 mg L⁻¹ of Cu or Fe. Each data set comprised 51 × 51 (2601) spectra, each of 2048 channels to which PCA was applied. The use of just two principal components improved the Fe Kα peak intensity by a factor >2 and the spatial resolution from 32.4 to 13.9 μm. Images were clearer and had much better contrast. However, these were very simple spectra from very simple model samples and wider application of the method has still to be demonstrated.

3 Synchrotron and large scale facilities

The outstanding properties of SR are high brilliance, linear polarisation in the plane of the SR storage ring and a tuneable continuum spectrum orders of magnitude more intense than any laboratory source. Continuing developments to improve the performance of storage rings and beamline setups include sophisticated X-ray optics and detectors as well as data handling. In addition the number of available facilities is increasing.

Understanding the fate and physical and chemical modifications of NPs in plants and their possible transfer into food chains requires specialised analytical techniques. In combination with spatially resolved XAS speciation, SR-XRF mapping offers multi-element detection with lateral resolution down to the tens of nm. A review by Castillo-Michel et al.⁴² primarily addressed the use of synchrotron-based μ- and nano-XRF mapping and XAS for studying the impact of engineered nanomaterials on plants which are essential components of ecosystems. The review focused on important methodological aspects such as sample preparation, data acquisition and data analysis of SR-XRF mapping and XAS. Das et al.⁴³ reviewed the probing of nanostructured materials using GI-XRF. Dispersion of metal NPs on flat surfaces has attracted considerable interest in many technological applications. Applications in medical physics, chemistry, biology and semiconductor quantum dot have driven rapidly the growth in the use of functionalised patterned nanostructures in the last two decades. The application to biomedical imaging received particular attention as these materials provide a direct interface at the subcellular scale. Reliable characterisation methods are therefore often needed to determine detailed structural properties such as average particle size, particle shape, internal structure, surface morphology, distribution of particles and their chemical composition. Grazing incidence XRF, a nondestructive probe, provided sensitive information about the physical and chemical nature of the NPs over the surface area (a few cm²) of a substrate. The potential capabilities of GI-XRF spectrometry to provide reliable and precise determination of the surface morphology of metal NPs dispersed on a flat surface were emphasised. The X-ray-standing-wave assisted fluorescence measurements were sensitive to the nature of dispersion of NPs on a substrate surface. The analyses of different types of nanostructured materials, e.g. the structure of a tungsten thin film and Au NPs deposited on a silicon surface, were presented. The average size of the Au NPs measured by GI-XRF spectrometry agreed well with data obtained using AFM.

Several papers described new beamlines. Newly available X-ray nanobeams in synchrotron radiation facilities open new research avenues in the nanosciences. However, a significant challenge is to concentrate efficiently, particularly for high-energy X-rays, a large photon flux into a very small focal spot. da Silva et al.⁴⁴ published an outstanding paper on the efficient concentration of high-energy X-rays for diffraction-limited imaging resolution. The authors demonstrated for the first time a sub-13 nm (FWHM) diffraction-limited X-ray focus size with 6 × 10⁹ photons per s formed using elliptically figured mirrors operating at 33.6 keV. This is the smallest and brightest focus spot achieved in this high-energy range and it will offer new opportunities in multidisciplinary fields for X-ray microscopy techniques, such as elemental imaging in which the focal spot size limits the resolution. It is routinely available for XRF studies at the ID16A nanoimaging beamline of the ESRF.

Chen et al.⁴⁵ described a new mobile endstation at the Beijing Synchrotron Radiation Facility for confocal depth-resolved fluorescence μ-XAS in the study of cultural heritage materials. An automatic program based on an algorithm aligned the two polycapillary half lenses efficiently. A study of sacrificial red glaze China (AD 1368–1644) was made on a general XAS beamline to demonstrate the capability of this depth-resolved system. As the endstation was mobile, the confocal system could be used to improve the function and flexibility of general XAS beamlines and to extend their capabilities to a wider user community.

A new setup for biological applications at the Canadian Light Source (CLS) facility consisted⁴⁶ of four beamlines devoted to studies ranging in scope from the atomic scale to cells, tissues and whole organisms. The Biological X-ray Absorption Spectroscopy (BioXAS) facility housed three beamlines devoted to XAS and multi-mode XRF imaging. The powerful features of these beamlines, which provided CLS users with an array of techniques for studying pressing biological questions, were described and future applications discussed. A multipurpose XRF endstation has been available for users since 2015 at the IAEA facility at Trieste (ELETTRA). Karydas et al.⁴⁷ provided a broad overview of the various analytical capabilities available, their intrinsic features and performance figures for the endstation. Crystal or multilayer monochromators were used to produce monochromatic X-rays (3.7–14 keV) from the output from the ELETTRA storage ring (2.0 or 2.4 GeV). Combinations of various advanced analytical probes (e.g. GI-XRF, XAFS and XRR measurements using various excitation and detection geometries) were possible to provide a comprehensive characterisation of different kinds of nanostructured and bulk materials. Menesguen et al.⁴⁸ described CASTOR, a new instrument for combined XRR-GI-XRF analysis at SOLEIL, which was dedicated to the characterisation of thin films with thicknesses in the nm range. Operation on the two branches of the metrology beamline made experiments possible over an impressive wide range of photon energies (45 eV to 40 keV). A heated sample holder allowed the sample temperature to be controlled up to 300 °C. Details of initial studies were given to illustrate the capabilities of the setup.

Special techniques have been developed to save analysis time and improve efficiency. A scan system⁴⁹ for the analysis of arbitrarily shaped samples at a SR facility had control and data acquisition systems which integrated motor control, detector triggering and data acquisition and storage so that motors did not need to be stopped during measurement. The 2D mapping of the arbitrary shapes and XRF data acquisition occurred synchronously. A standard gold mask was used for SR-XRF mapping to verify the validity of this method. Total scan time was reduced significantly. Photovoltaic (PV) devices require characterisation and optimisation over scales from cm to nm. Morishige et al.⁵⁰ used μ-SR-XRF, a valuable tool in the characterisation of PV materials and devices, to increase analytical throughput and sensitivity. Micro-XRF maps of element distributions in PV materials had high spatial resolution and excellent sensitivity and could be measured on both absorber materials and complete devices. An on-the-fly data collection system (flyscan) eliminated the time (300 ms) needed to move from one measurement position to another and enabled high-sensitivity (10¹⁴ atoms per cm²) maps to be collected over large areas (10 000 μm²) and with high-spatial resolution (<200 nm scale). Scanning time was reduced from 10 h to 84 minutes for a specific sample. In the characterisation of detrimental trace-metal-precipitate distributions in multicrystalline PV silicon wafers, flyscans increased the sample throughput of μ-XRF spectrometry by an order of magnitude. In addition, relatively large-area microscopy was possible. The transition metal distributions in a 50 μm diameter laser-fired contact of a silicon solar cell were mapped both before and after lasing.

Statistical and correlative analyses are increasingly important in the design and study of new materials such as semiconductors and metals. There are few non-destructive measurement techniques with the high spatial resolution necessary for correlating composition and/or structure with device properties. In the cases of polycrystalline and inhomogeneous materials, the added challenge is that nanoscale resolution is, in general, incompatible with the large sampling areas necessary for acceptable statistical representation of the specimen. This is of particular importance in the study of grain cores and grain boundaries in polycrystalline solar absorbers as their dissimilar behavior and variability throughout the samples makes it difficult to draw conclusions and ultimately optimise the material. West et al.⁵¹ presented a nanoscale in operando approach based on the multimodal utilisation of SR-nano-XRF and X-ray-beam-induced current. Data were collected for grain core and grain boundary areas and correlated pixel-by-pixel in fully operational Cu(In_1−xGa_x)Se²⁻ solar cells. Cells with low Ga concentrations had grain boundaries that overperformed compared with the grain cores whereas those with high Ga concentrations had boundaries that underperformed. These results demonstrated that nanoscale correlative X-ray microscopy could inform research on grain engineering with the aim of manufacturing high-efficiency solar cells at low-cost.

Details for only one new XRF detector system were published during the review period. Despite the constant technological improvements in the field of detector development, XRF with soft X-rays remains a challenge because of the low intrinsic fluorescence yield at energies <2 keV. A multi-element SDD system for fluorescence spectroscopy using soft X-rays, as described by Bufon et al.,⁵² will be installed at the TwinMic beamline of ELETTRA synchrotron (Trieste, Italy). In order to increase the detected count rate by up to an order of magnitude, the system was based on four trapezoidal-shaped monolithic SDD tiles (matrices) with six hexagonal elements, each equipped with an ASIC readout. These readouts had ultra-low noise and were made specifically for each application. The new architecture was claimed to be very versatile and adaptable to any XRF experimental setup. This detector system represents a remarkable improvement of the whole signal processing chain, leading to an energy resolution of 116 eV for an area of 924 mm² for Mg Kα (1.24 keV).

4 Grazing X-ray techniques including TXRF

The number of papers described in this year’s TXRF contribution is drastically reduced: a sign that the technique is maturing and most attention is now focused on applications. This section considers technical improvements to all components of the system including excitation beam and detectors in addition to advances in sample preparation. For an overall insight into recent developments, readers are referred to the special issue⁵³ on the 17th International Conference on Total Reflection X-ray Fluorescence Analysis and Related Methods.

As TXRF instrumentation is now fully mature for routine analysis, a working group was established to work on an ISO method for direct TXRF analysis of water samples. This analysis is limited mainly by the presence of salts and suspended particles, spectrum interferences and poor sensitivity for some trace and ultra-trace elements. Borgese et al.⁵⁴ assessed applicability, accuracy and precision of a method for trace element detection in waters. Calibrations were linear over the range 1 μg L⁻¹ to 10 mg L⁻¹. Statistical tests and data regressions were performed separately for each element of interest. Precision and accuracy were evaluated by analysis of a water RM prepared for an inter-laboratory comparison of element analyses of drinking water.

Although TXRF spectrometry is a fast and multi-element method and does not require complex sample pretreatment, its sensitivity is sometimes insufficient for environmental analysis so preconcentration is required. Panchuk et al.⁵⁵ suggested a very simple procedure based on a planar waveguide technique in which the sample was placed directly into the waveguide. The waveguide construction used just two glass reflectors so was simple and could be constructed in any laboratory. The S/N was improved considerably and typical LODs were 0.12 and 0.13 μg L⁻¹ for Cd and Hg, respectively. Buddhadev et al.⁵⁶ demonstrated the suitability of polychromatic X-rays for TXRF determination of trace elements in aqueous solutions and uranium oxide CRMs. A direct beam rather than a beam reflected from the multilayer was used for excitation of the analytes present in ng amounts. Uranium-based samples were dissolved and U was separated using solvent extraction. The advantage of this method over monochromatic excitation from the multilayer was not only that it was simpler and easier to use but that the LODs could be improved by a factor of 1.7.

Prost et al.⁵⁷ used quartz reflectors as sampling substrates in the quantitative TXRF analysis of directly collected aerosol samples on a quartz sample reflector. For the purpose of calibration, the spot patterns produced by a three-stage Dekati impactor were simulated on the surface of quartz reflectors using a nL deposition unit and multi-element standard solutions. A droplet containing 5 ng Y applied in the center of each reflector served as internal standard. Calibration samples and aerosol samples were prepared and measured in the same way. Linear calibration curves and good recovery rates of 10% were obtained. Aerosol samples were collected on quartz reflectors that had to be greased with petroleum jelly to prevent the bounce-off effect. The grease layer was removed by cold-plasma ashing and an internal standard added. No further chemical pretreatment was required for TXRF analysis. The LODs were 10 pg m⁻³ for medium-Z elements (Fe to Sr) in the largest size fraction (>10 μm) but typically only 30–100 pg m⁻³ for the samples collected on the two smaller stages (1–2.5 μm and 2.5–10 μm).

Grazing incidence and grazing emission XRF (GI-XRF and GE-XRF) spectrometries are techniques that enable non-destructive, quantitative analysis of element depth profiles with nm spatial resolution. Baumann et al.⁵⁸ combined a high-brilliance laser-produced-plasma (LPP) source with a scanning-free GE-XRF setup to achieve a large solid angle of detection. The detector, a pn CCD, was operated in a single-photon counting mode in order to utilise its energy dispersive properties. The GE-XRF profiles of the Ni-Lα and Ni-Lβ lines of a nickel–carbon multilayer sample, which displayed a lateral (bi)layer thickness gradient, were recorded at several positions. The prominent intensity minimum at grazing emission angles of 5–12°, as predicted by simulations of theoretical profiles, depended strongly on the bilayer thickness of the sample. The good agreement of results with values obtained by XRR, conventional XRF spectrometry and TEM demonstrated nm-resolved element depth profiling in the soft X-ray range and highlighted the potential for application to in situ process control in the semiconductor industry.

A newly developed code for GI-XRF spectrometry data evaluation (GIMPy) was presented by Brigidi and Pepponi.⁵⁹ The GI-XRF technique had improved sensitivity in comparison to conventional geometry and so allowed XRF spectrometry to be used for the analysis of thin films. The code applied a FP approach to quantitative XRF data. The electric field calculation in stratified media automatically delivered the total reflected intensity. Parameters modelled included the source, modulation of the primary beam, interactions with a layered sample, absorption of the emitted XRF intensities and the response function of semiconductor ED detectors. The last was obtained by comparing a simulation of the expected spectrum with that actually acquired. The FP part included signal enhancements by both cascade effects and secondary fluorescence. The code offered the possibility of taking into account the effects arising from deviations from ideal conditions: polychromatic excitation, beam divergence, beam size and shape, sample-inspected area and solid angle of detection. The effectiveness of the code was demonstrated on a set of semiconductor substrates (germanium, silicon and gallium arsenide) and shallow distributions of As in silicon.

The optical and electrical properties of transparent conducting oxide (TCO) thin films are strongly dependent on structural and chemical properties so non-destructive characterisation techniques are needed to probe the composition variations along deposited-film profiles. Rotella et al.⁶⁰ used a combination of GI-XRF spectrometry and XRR to get simultaneous information on structural properties (thickness and roughness) and chemical properties. The two techniques could be used with the same experimental set-up and the analysis was combined in a single software in order to refine the sample model. Whereas XRR was sensitive to the electron density profile, GI-XRF spectrometry was sensitive to the atomic density (i.e. the element depth profile). An XRR-GI-XRF analysis of indium-free TCO thin films (Ga-doped zinc oxide compound) was performed to correlate the optical properties of the films with the Ga dopant distribution. This combination of techniques had potential for providing insights into material composition and element depth profiling.

5 Hand-held, mobile and on-line XRF techniques

5.1 Hand-held and mobile XRF techniques

Portable XRF spectrometry is a powerful tool for making measurements non-destructively and within a short time frame. This technique is already well established but nevertheless challenges in improving detection limits remain. Pessanha et al.⁶¹ studied the performance of a portable EDXRF spectrometer using three different filters (Al, Cu and Al + Cu) placed between the X-ray tube and SRMs of different average-Z matrices. The LODs were improved for elements with emission energies of 5–15 keV, in particular for low-Z matrices. The use of filters did not improve the LODs in high-Z matrices but reduced interferences in the measurement of some elements. A new modelling method based on discrete wavelet transform (DWT) was introduced⁶² to determine As, Cr, Cu, Pb and Zn concentrations in soil samples. De-noising and baseline correction by the DWT method as a pre-processing procedure was effective for handling XRF spectrometry results. A comparison of the calibration curves and LODs for the raw spectral data and DWT-processed data demonstrated that better modelling results could be obtained using DWT. Accuracy was improved and the LODs were a factor of 1.5–5 lower for the processed data.

The presence of microplastics in the environment is of growing concern and might have negative effects on human and animal health. Samples of microplastics (n = 924) from two beaches in South West England were analysed⁶³ by field portable-XRF coupled with a ‘small-spot’ facility that collimated the X-ray beam to a width of 3 mm. The element content of plastics down to about 1 mm in diameter and 0.1 mm in thickness could be determined rapidly. Cadmium and Pb, detected in about 7% of the samples analysed, had the highest concentrations in red and yellow pellets or fragments and were associated with Se and Cr, respectively. Bromine, associated with Sb, was detected in over 10% of the samples analysed but was mainly found in neutrally-coloured pellets. Although the avian bioaccessibilities of Br, Cd and Pb in microplastics were low, accessible concentrations of Cd and Pb in brightly coloured fragments could exceed corresponding concentrations in the seabird diet by factors of 50 and 4, respectively. The same instrument was used⁶⁴ to measure the concentrations of various elements (including Br, Cd, Cl, Cr, Cu, Fe, Pb and Zn) in beached microplastics. In the laboratory, the XRF procedure provided results on average within 20% of concentrations determined by ICP-OES following acid digestion. Analysis of progressively smaller offcuts (to <1 mm) gave results comparable to those determined in original samples but the precision and LODs were poorer. When two operators worked together, up to 35 microplastic samples, each counted for 60 s, could be processed in an hour. The determination of various trace elements in both fresh and dry fucoid species of macroalgae demonstrated⁶⁵ the benefits of portable XRF spectrometry. When the low density mode with thickness correction was used, As could be quantified at dry weight concentrations of a few μg g⁻¹ and Br, Fe and Zn at concentrations of a few tens of μg g⁻¹. Measurement of Cu and Pb in fucoids was also possible in moderately to highly contaminated samples.

A field portable-XRF procedure was used⁶⁶ both in situ and in the laboratory to measure Sb concentrations in consumer products. Concentrations ranged from 60 to 60 000 μg g⁻¹ but Sb could be detected in only 18% of over 800 samples meaning that the concentration in 82% of the samples was <60 μg g⁻¹. The highest concentrations, encountered in white electronic casings, correlated with Br concentrations; a finding that would be consistent with the use of antimony oxides as synergistic flame retardants. Concentrations of Sb were highest (a few%) in plastic components of heat-generating electrical products but from a health perspective the presence of lower quantities of catalytic Sb in food packaging and small toys was of greatest concern. Co-association of Br and Sb in many products not requiring flame retardancy suggested that electronic casings are widely recycled. To investigate the extent to which kitchen utensils were contaminated with brominated flame retardants (BFR) and the potential for resultant human exposure, Kuang et al.⁶⁷ collected plastic kitchen utensils and screened them for Br content using a HH-XRF spectrometer. Whereas only three out of 27 utensils purchased after 2011 contained detectable concentrations of Br (≥3 μg g⁻¹), Br was detected in 31 out of the 69 utensils purchased before 2011. Eighteen utensils with Br contents >100 μg g⁻¹ and 12 new utensils were selected for GC-MS determination of BFRs. The predominant BFR in most utensils was BDE-209. Hennebert and Filella⁶⁸ used the HH-XRF technique to determine the chemical composition of plastics from electric and electronic equipment and for regulatory classification. The Br data (n = 4283), used as an indicator of BFR levels, could be valuable in the enforcement of EU regulation.

Portable XRF spectrometers are becoming more popular as a rapid diagnostic tool or to measure biomarkers in health related issues. In a portable EDXRF setup for urinary iodine diagnostics, the X-ray optical scheme was optimised⁶⁹ to provide the maximum sensitivity for revealing the iodine I-Lα and I-Lβ lines in a matrix of C, Cl, K, Na, P and S. The optimised parameters were: the material of the secondary target; the material and thickness of the filter; radiation incident and exit angles; aperture of collimation system; and the thickness of the jelly-like sample. The experimental LODs were 45 μg L⁻¹ (for I-Lα) and 75 μg L⁻¹ (for I-Lβ) at the theoretically calculated minimum value of 30 μg L⁻¹. This sensitivity was sufficient for urinary iodine diagnostics in the range from 50 to 200 μg L⁻¹ I. In a continuation of their research on the analysis of nails and nail clippings, Fleming and Ware⁷⁰ used a new portable XRF system with a higher tube current and larger radiation detector in order to achieve the increased sensitivity required for the determination of Cr. Initially, five measurements of 180 s (real time) duration were performed on six whole nail phantoms with Cr concentrations of 0, 2, 5, 10, 15 and 20 μg g⁻¹. Then these phantoms were converted to nail clippings using nail clippers and sorted into different mass groups of 20, 40, 60, 80 and 100 mg. The MDL for Cr determined from the whole nail intact phantoms was 0.88 ± 0.03 μg g⁻¹. For the clipping phantoms, the lowest MDL of 1.2 ± 0.1 μg g⁻¹ was achieved for the 40 mg clipping group. The use of higher mass collections did not lead to improved results. This MDL was comparable to Cr concentration levels seen in various studies involving human nail clippings. Disadvantages of the K-shell XRF technique such as the use of a radio-isotope source, the use of a liquid nitrogen cooled detector, the limitation in portability and the lengthy measurement times were tackled by the development of a new portable XRF device. Moreover, by using the high-energy KXRF spectrometer, Specht et al.⁷¹ was able to avoid the limitations of soft tissue thickness. The in vivo X-ray-tube-based KXRF bone Pb measurement system exploited recent advances in detector technology. The validity of the system was proven using Monte Carlo simulations. Use of a cadmium zinc telluride detector gave a LOD for Pb in bone of 6.9 μg g⁻¹, similar to that achieved using second-generation radioisotope-based KXRF bone Pb measurement systems. Desouza et al.⁷² used arsenic-doped skin-mimicking calibration phantoms to investigate the performance of two generations of HH-XRF spectrometers from the same manufacturer. The second generation analyser performed better than the first generation model, with an As Kα MDL of 0.462 ± 0.002 μg g⁻¹ in arsenic-doped resin calibration phantoms using 120 s measurement time. Radiation dosimetry studies showed that the newest system, like its predecessor, delivered an equivalent dose of 15–20 mSv to an approximately 1 cm² area of the skin. This performance was sufficient for studies of health effects in populations exposed to As.

The value of portable XRF spectrometers in providing rapid and quantitative sample characterisation for critical decision making on board geoscience research vessels was emphasised.⁷³ Working curves developed from portable XRF measurements using a suite of geological SRMs and well-characterised lavas permitted accurate quantitative measurements to be made on sample powders and rock surfaces for the determination of Ca, Cr, Cu, Fe, K, Mn, Sr, Ti, Rb, V, Zn and Zr. Although portable XRF spectrometry had been employed on previous cruises, a detailed, high-density chemostratigraphy of recovered core samples by portable XRF measurements of rock core surfaces was established for the first time during the International Ocean Discovery Program Expedition 352. Australian researchers⁷⁴ also proved that, despite the care required in validating data obtained using portable XRF systems, dynamic exploration campaigns in regolith-dominated terranes could achieve rapid turnaround times and low-cost measurement of a larger suite of elements than would normally be attempted. The analysis of regolith material by XRF spectrometry was challenging because of the large matrix effects that resulted from the variable Fe content (0–70%). For instance, the poor Pb data were associated with erroneous measured Bi concentrations. At high Fe and Pb concentrations, pile-up of the Fe Kα peak (6.405 keV) occurred at 12.8 keV, close to the Pb Lβ (12.614 keV) and Bi Lβ (13.023 keV) peaks.

Other interesting applications included a HH-XRF spectrometry method⁷⁵ for measuring in a cost- and time-efficient manner the release of Cu and Zn from antifouling paints. Calibration standards were prepared from a biocide-free antifouling paint mixed with increasing amounts of copper and zinc oxides. Subsequently, these paints were applied to a 6.3 μm thick Mylar film to obtain a wet film thickness of 100 μm, corresponding to 25–35 μm dry thickness. After drying, round discs (25 mm diameter) were punched out for XRF spectrometry analysis. In a field study in the Baltic Sea and Kattegat, the release of Cu increased with salinity whereas the release of Zn was independent of salinity. Park et al.⁷⁶ proposed a field method for measuring the size distribution (10 nm to 20 μm) of metal-containing particles, using a nano micro-orifice uniform-deposit impactor (nano-MOUDI) to collect particles according to size. Custom-made substrate holders were prepared to enable the direct measurement of nano-MOUDI substrates using the portable XRF system. The XRF results correlated linearly with those obtained using ICP-MS (R² = 0.84 and 0.91 for Cr and Fe, respectively). The mass of Fe detected by the XRF system was comparable to that detected by ICP-MS for loadings greater than 2.5 μg per substrate.

5.2 On-line XRF techniques

The popularity of on-line measurement is reflected in the increased number of publications. Furger et al.⁷⁷ determined the element composition of ambient aerosols with high temporal resolution using an on-line XRF spectrometer. During a 3 week field campaign in Switzerland ambient air was sampled through a PM₁₀ flow separator and subsequently collected on a Teflon filter tape. After each sampling interval the filter tape was moved into the XRF spectrometer for determination of 24 elements. The next sample could be collected simultaneously on the next section of filter tape. The high time-resolution (1 h) came at the cost of sensitivity as reflected in the MDLs which were poorer than those achievable using off-line methods. There was excellent correlation (R² ≥ 0.95) between the data for 10 elements (Ba, Ca, Cu, Fe, K, Mn, Pb, S, Ti and Zn) obtained by XRF spectrometry and those obtained by ICP-MS. However, the slopes of the regressions between the XRF data and the ICP-MS data varied from 0.97 to 1.8 (average 1.28) indicating generally higher concentrations measured by the XRF technique than by ICP-MS. Nevertheless, the on-line XRF system could be a powerful tool in air quality monitoring because of its multi-element characterisation at high-time resolution. A monitoring study in Bejing (China) used⁷⁸ a similar on-line XRF system to apportion contaminants in air to sources. The precision in interpreting source apportionment calculations was greatly improved by the detailed source profiles made possible by the high-time-resolution measurements of trace elements in PM_2.5. Norlin et al.⁷⁹ developed an XRF imaging method for on-line paperboard quality measurements in order to verify the coating thickness and structure of CaCO₃ on cellulose fibers. A laboratory scale setup used stepper motors, a SDD-based spectrometer and a collimated X-ray beam to give an image resolution of 0.5 mm. The XRF signals from the Ca atoms (3.7 keV) in the coating were combined with those from a Cu target (8.0 keV) placed behind the paper to measure simultaneously both the coating and the fibers. The material contents in the layers could then be determined from the absolute and relative intensities of these two signals. The authors indicated that the next step will be to use a tilted beam to improve the resolution and generate 3D information.

The average distance between an XRF analyser and the surface of coal samples changes with particle size and results in fluorescent intensity changes and subsequent inaccuracy of on-line XRF measurement. Zhang et al.⁸⁰ developed a method for correcting for this variable distance to improve the accuracy of on-line XRF measurements of coal particles. A theoretical study was conducted first, in which the relationship between the XRF intensity and the distance from the spectrometer to the sample surface was derived by considering five physical phenomena: (1) absorption of X-rays by the air, (2) irradiated surface area of the samples, (3) changes in exit angle, (4) changes in solid angle and (5) absorption of X-ray fluorescence by the air. The calculated theoretical curve agreed well with experimental data. When applied⁸¹ to real samples, their new correction method improved the accuracy of Fe determination significantly.

6 Cultural heritage applications

Research related to cultural heritage generally uses the XRF technique in combination with hyperspectral imaging and molecular analytical techniques. The value of the non-destructive nature of XRF techniques is highlighted in this section on application to the analysis of invaluable and irreplaceable cultural heritage samples.

A comprehensive and very interesting review of the application of scanning macro-XRF techniques using mobile instruments for the investigation of historical paintings was presented by Alfeld and de Viguerie.⁸² Both hyperspectral imaging in the visible and IR ranges and X-ray imaging were included. It could be concluded that the technical capabilities of the macro-XRF technique and hyperspectral imaging have reached a plateau and that, with the availability of commercial instruments, the focus of recent studies has been shifted from the development of methods to the applications themselves. An overview presented by Cotte et al.⁸³ of the current status of the instruments at the ID21 beamline (ESRF, France) included both X-ray and IR spectroscopies. Recent studies of cultural heritage highlighted the advantage of the ID21 platform as a multi micro-analytical platform specifically for 2D multi-chemical imaging. Trentelman⁸⁴ presented an overview of analytical imaging techniques for cultural heritage materials, from the macro-scale through to the micro- and nanoscales.

A novel mobile macro-XRF spectrometric scanning device based on a three-axis system provided⁸⁵ real-time element imaging of large paintings (110 × 70 × 20 cm) with a maximum scan speed of 1 cm s⁻¹, and a maximum pixel size of 500 μm in element images. In addition, when scanning was performed with samples in the focus position of the polycapillary, μ-XRF imaging could be performed with a lateral resolution of up to 25 μm. Micro-XRF and macro-XRF imaging spectroscopies were combined to measure the element composition and distribution of pigment materials in a painted wooden panel of uncertain origin and date. The potential of macro-XRF imaging for identifying pigment signatures and hidden paintings was demonstrated. A recently developed portable macro-XRF scanner (“CRONO”) was characterised⁸⁶ for its detection sensitivity and spectroscopic and spatial resolutions. The device, capable of scanning an area of 450 × 600 mm at a linear speed of 45 mm s⁻¹, was designed for in situ macro-XRF investigations of polychrome surfaces on heritage objects. The value of the system was illustrated by a study of a 15^th century panel painting representing the ‘Virgin with the child’. Scanning macro-XRF spectroscopy, able to cover a range of 80 × 60 cm, was also evaluated by Saverwyns et al.⁸⁷ for the non-invasive investigation into the authenticity of two paintings. Both paintings were scanned using a 500 μm X-ray beam, in steps of 500 μm and with a dwell time per step of 10 ms. The first painting, a still-life attributed to the 17th century Spanish painter Francisco de Zurbaran, was analysed both with point XRF analyses using a portable μ-XRF system and the macro-XRF scanner. The latter revealed a hidden painting and gave a clear answer on the question of authenticity. The second painting, attributed to the workshop or school of Pieter Paul Rubens, was investigated by macro-XRF spectroscopy and revealed a hidden stamp of the canvas manufacturer. Although macro-XRF scanning is slower than the more labour-intensive XRF point analysis, it can be considered more time-efficient. A methodological strategy was proposed by Sciutto et al.⁸⁸ in an attempt to overcome the limitations of macro-XRF imaging analysis (poor S/N) without compromising the investigated surface dimension and analysis time. In particular, high S/N maps were acquired on regions of interest selected after preliminary fast scans on the entire painting. Subsequently, the elaboration of the dataset was supported by the use of ad hoc analytical tools, such as correlation diagrams, averaged extracted spectra and line emission maps for each element. The proposed experimental workflow was proven on a masterpiece panel painting by Cimabue (1240–1302) depicting ‘Madonna Enthroned with the Child and Two Angels’.

Much research work was undertaken on the element imaging of paintings. A combination of large-scale and micro-scale element imaging (non-invasive macro-XRF spectrometry analysis, SEM-EDS analysis and μ-SR-XRF imaging) was employed⁸⁹ to inform the conservation strategy of van Eyck’s renowned Ghent Altarpiece. It was possible to visualise the original paint layers which had previously been hidden. Whereas, the distribution of the high-energy Hg-L and Pb-L emission lines revealed the exact location of hidden-paint losses, Fe–K maps demonstrated how and where these lacunae were filled-up using an iron-containing material. The results of this study helped in making the decision to remove all skilfully applied overpaints. A macro-XRF imaging study was conducted⁹⁰ on René Magritte’s ‘La condition humaine (1935)’ to obtain insights into the artist’s palette and technique. In addition, a third quarter of ‘La pose enchanteé’ (1927) was discovered. Element distribution maps were collected over an area measuring 54.2 × 73.2 cm, with a step size of 750 μm, a dwell time of 200 ms per step and a beam size of 0.8 mm. A new data analysis protocol developed by Galli et al.⁹¹ was applied to XRF data acquired from an unstudied work by Giotto: ‘God the Father with Angels’ (ca. 1330, San Diego Museum of Art). A portable XRF spectrometer, equipped with a x-y translator stage (with 100 mm total travel and 100 μm step size), was used to conduct in situ 2D element mapping. Combining the effectiveness of scanning-XRF spectrometry with the responsiveness of image spectroscopic analysis made it possible to identify Giotto’s palette for the flesh tones in ‘God the Father with Angels’. A multimodal chemical imaging approach was applied⁹² to analyse the production technology of an 1800 year-old painting (Fayum’s ‘Portrait of a woman’). Simultaneous recording of the data cubes from three hyperspectral imaging modalities (hyperspectral diffuse reflectance (400–2500 nm); luminescence (400–1000 nm); and macro-XRF spectrometry (2–25 keV)) enabled the spectra at each pixel in the image to be compared for the entire painting. This fusion of molecular and element data allowed a more thorough identification and mapping of the painting’s constituent organic and inorganic materials. Key information on the selection of raw materials, production sequence and the fashion aesthetics and chemical arts practiced in Egypt in the second century AD was revealed. Daniel et al.⁹³ confirmed the complementarity of the two point-by-point spectroscopic techniques HH-XRF hyperspectral imaging and HH-Raman spectroscopy for the identification of pigments in paintings. The main pigments used in Aragonese artworks, and especially in Goya’s paintings, were identified and mapped. The presence of pigments such as lead white, red and brown ochre and Prussian blue and vermillion confirmed that Goya and his contemporaries used pigments typical of those used in the eighteenth century. In addition, the use of other modern pigments during the 1980 restoration process was confirmed. The spatially-resolved spectroscopic imaging capabilities of the SRX beamline of NSLS-II were used⁹⁴ to study metal soap formation in a micro-sample removed from a 15^th century oil painting and to obtain information on element segregation and chemical compounds formed. Segregation of both Pb and Sn occurred in the soap-affected areas. Moreover, XANES analysis around the Pb-L₃ absorption edge differentiated Pb pigments from Pb soaps while μ-XANES analysis gave further information on the chemical heterogeneity in the paint film. Limitations of these techniques included possible damage to the sample by the synchrotron radiation and the lack of differentiation of lead azelate and palmitate.

The development of procedures for the rapid automated evaluation of full spectral XRF imaging data presents a challenge. A dataset generated from Hans Memling’s ‘Portrait of a man from the Lespinette family’ was used⁹⁵ to demonstrate the capabilities and the shortcomings of Simplex Volume Maximisation (SiVM); a novel approach for factorising complex data sets with non-negative constraints and low computing demands. The raw data were processed using the PyMCA and Datamuncher programmes and in parallel by SiVM. The results were not identical but similar to element distribution images and highlighted correlations between elements, and thereby paints, otherwise not obvious. A GEM-based detector with 2D readout can be used in full-field XRF imaging. Dabrowski et al.⁹⁶ described a system equipped with a standard 3-stage GEM detector (10 × 10 cm) and readout electronics based on dedicated full-custom ASICs and a DAQ system. A 2D spatial resolution of 100 μm and energy resolution of 20% FWHM for 5.9 keV could be obtained with a demonstrator system. Limitations of such a detector due to copper fluorescence radiation excited in the copper-clad drift electrode and GEM foils were discussed and the performance of the detector using chromium-clad electrodes reported. Using the detector with chromium-clad electrodes, the ratio between a measured Se fluorescence spectral line and the Cu fluorescence background increased by a factor of 7.

Original red ochre, original yellow ochre and transformed yellow ochre (nowadays showing a red colour) of wall paintings from Pompeian houses were analysed⁹⁷ using a HH-EDXRF spectrometer. The original red ochre and the transformed yellow ochre could be differentiated. The PCA of the multivariate data showed that the As content could be used to distinguish between both red-coloured ochres. The element composition and the conclusions drawn from the in situ analysis were confirmed by laboratory measurements using a benchtop EDXRF spectrometer. The wall painting in the “St. George” church in Staro Nagoričane (Republic of Macedonia) was examined by Robeva-Cukovska et al.⁹⁸ in order not only to characterise the materials and techniques used by the artists but also to reveal the presence of other materials due to later interventions and degradation processes. The use of complementary techniques, including XRF spectrometry, allowed the micro-stratigraphy, the pigments and binders used to be established. For the correct restoration of murals and icons in a Romanian church, pigments used in 1853 by a famous Romanian painter Gh. Tattarescu, were analysed⁹⁹ using a portable XRF spectrometer. The chemical element predominant in all spectra for the red areas was Hg, suggesting the pigment was cinnabar (mercury sulfide). In pink areas, the presence of both Hg and Pb suggested that cinnabar had been mixed with lead white. The yellow pigment was made from a mixture of chrome-yellow (PbCrO₄) and ochre (Fe based yellow ochre-Fe₂O₃·H₂O-hydrated iron oxide). The presence of zinc in a blue area (an iron-based pigment – very probably Prussian blue) indicated a late repainting with zinc white (ZnO) on the original lead white layer. Portable and laboratory XRF instruments were compared¹⁰⁰ for the discrimination of a group of yellow pigments containing Pb, Sb and Sn. Measurements with the portable system were performed with a spot size of 2 mm, whereas the laboratory μ-XRF spectrometer had a spot size of 25 μm obtained using polycapillary optics. The portable instrument had better spectral resolution for energy values over 10 keV whereas the laboratory μ-XRF spectrometer had a better resolution for energy values less than 10 keV. An acquisition time of 60 s was suitable for both instruments. Multivariate analysis of the data allowed artificial yellow pigments to be differentiated.

The colourful decoration of statues and buildings in antique times is commonly described by the term ‘antique polychromy’. Alfeld et al.¹⁰¹ discussed results obtained from the frieze of the Siphnian Treasury in the Sanctuary of Delphi (Greece). In the in situ investigation of antique polychromy, use of an in-house-built mobile instrument for macro-XRF imaging allowed the identification of significant traces invisible to the naked eye and not detectable by conventional XRF spot measurements or any other mobile, non-invasive method. A partial reconstruction of the polychromy could therefore be made. Furthermore, a novel approach for interpreting XRF data generated from complex-shaped samples was based on a 3D surface model derived from photogrammetry and fundamental parameter calculations. Gasanova et al.¹⁰² performed a non-destructive in situ analysis of polychromy on ancient Cypriot sculptures using a range of analytical techniques including portable μ-XRF spectrometry. Among the materials identified were Fe-containing (red, yellow, green) and Cu-containing (green and blue) pigments. The study demonstrated the complementarity of XRF and FORS. For example, traces of Pb-containing compounds in red iron pigments as identified by XRF analysis were not related to the components of polychromy as the FORS analysis excluded the presence of minimum (Pb₃O₄). Pelosi et al.¹⁰³ conducted an investigation on a group of 16^th to 18^th century terracotta sculptures (modelli) belonging to the extraordinary collection of Palazzo Venezia in Rome (Italy). A multi-technique approach including XRF spectrometry was chosen to study the morphology and composition of the surface-painted layers and made it possible to identify the pigments used. Langlois et al.¹⁰⁴ identified the composition of modern modelling materials used by Rodin on plaster sculptures and presented hypotheses on their degradation. Elemental and chemical analyses were performed on samples from 12 sculptures using multiple analytical techniques, including μ-SR-XRF spectrometry and μ-XANES. This thorough study of their composition and degradation was necessary to implement an appropriate restoration plan.

Smieska et al.¹⁰⁵ used SR-XRF spectrometry and SR-XRD mapping in the trace-element and composition analysis of azurite pigments in six illuminated manuscript leaves dating from the 13^th to 16^th centuries. The SR-XRF mapping revealed that the concentrations of several trace elements (As, Ba, Bi, Sb and Zr) correlated with that of azurite. This was confirmed by the SR-XRD maps. These elements could not always be detected by point XRF measurements. Variations of the trace element concentrations in azurite were greater between different manuscript leaves than within each individual leaf, suggesting the possibility that such impurities might reflect distinct mineralogical sources. By using SR-macro-XRF (beam size 50 or 100 μm diameter) and μ-SR-XRF (beam size ∼0.4 × 0.7 μm) maps together with μ-XANES data, Christiansen et al.¹⁰⁶ showed that carbon black inks on ancient Egyptian papyri from different time periods and geographical regions contained Cu. The composition of the Cu-containing carbon inks showed no significant differences that could be related to time periods or geographical locations. It was concluded that the same technology for ink production was probably used throughout Egypt for a period spanning at least 300 years. The Cu concentrations could be correlated with the three main components, viz. cuprite, azurite and malachite. Data from a portable macro-XRF spectrometry system, visible hyperspectral imaging and SR-macro-XRF spectrometry were used¹⁰⁷ in signal processing methods to reveal the provenance of a degraded manuscript recycled as binding material for a 16^th century printed edition of Hesiod’s ‘Works and Days’. Macro-XRF spectrometry measurements (1 mm beam size) were conducted with a step size of 1 × 1 mm (scanned area of 5 × 4 cm) or 100 × 100 μm (scanned area of 0.6 × 0.6 cm), whereas the SR-macro-XRF scanning measurements (100 × 125 μm beam size) were performed using a step size of 200 × 200 μm (scanned area 10 × 10 cm). The analytical techniques allowed hidden text to be revealed. Pigments and inks in the original manuscript could be identified as well as their sequence of use. Pessanha et al.¹⁰⁸ constructed a portable EDXRF spectrometer with orthogonal tri-axial geometry between the X-ray tube, the secondary target, the sample and the detector. Compared with other devices, the P/B was better and thereby also the LODs. This system allowed important elements (Bi and Pb) to be detected in an 18^th century paper document. These elements occur at concentrations below the LODs of a planar geometry setup. Improved P/Bs in a 19^th century bone sample allowed the quantitative determination of Br and Pb. Portable Raman and portable XRF spectroscopies combined with hyperspectral imaging were used by Mulholland et al.¹⁰⁹ to identify and map watercolour pigments used by the 18^th century botanical illustrator, Ferdinand Bauer, and to demystify the unusual colour code system found in his sketches. Eight pigments were identified unequivocally from 125 watercolour paintings analysed, suggesting that Bauer used a more traditional and more limited palette than previously considered and that his palette changed significantly in his later paintings.

Red and white porcelain sherds, excavated from the Mansion of Prince Qin (China), were selected¹¹⁰ to study their compositional characteristics and the painting techniques used. Analysis by EDXRF spectrometry revealed the homogeneous composition of the lime-alkali glazes whereas μ-SR-XRF mapping revealed that the writing and the painting used different processes. Machado et al.¹¹¹ used two non-destructive techniques, CT and μ-XRF spectrometry, to determine the structure and elemental composition of ceramic samples from two Brazilian archaeological sites. The chemical compositions determined by μ-XRF spectrometry revealed differences between the ceramic matrix and tempers/inclusions. The 3D models, generated from the μ-CT images, displayed the groups (ceramic matrix and tempers/inclusions) according to their density and so it was possible to visualise the distribution of the tempers inside the samples. The black decorations of Serra d’Alto figuline pottery (Matera, Italy) were characterised by Angeli et al.¹¹² using a portable μ-XRF system (102 samples, spot size 1 mm² acquisition time 90 s per spot). A black pigment based on Mn was used for the decoration of this pottery. A detailed composition of fine-paste ware (13^th to 14^th centuries) from five archaeological sites in Thailand and Indonesia could be provided¹¹³ by combining XANES and EDXRF spectrometry measurements. Sources of clay and production sites were classified according to Al₂O₃, SiO₂, α-Fe₂O₃ and γ-Fe₂O₃ content and trace element concentrations. The uniquely large contribution of α-Fe₂O₃ compared to that of γ-Fe₂O₃ in XANES spectra suggested that Nakhon Si Thammarat province of Thailand could be one of the fine-paste ware production areas in maritime southeast Asia. In situ non-destructive Raman spectroscopy and XRF spectrometry analyses of pollen from Lipari (Aeolian Island, Sicily) were performed¹¹⁴ with portable instruments on a selection of polychrome vases. Different pigments were identified: kaolin and gypsum, probably supplied locally, for white layers; Egyptian blue for blue hues; red ochre for brown-reddish hues; and cinnabar for pink and red-purple nuances. The identification of both Egyptian blue and cinnabar started an interesting discussion about the dating and circulation of the raw materials. The measurement parameters of a dedicated calibration procedure in the WDXRF spectrometry analysis of ancient ceramics, analysed as glass beads, were outlined by Georgakopoulou et al.¹¹⁵ Although the analytical performance of this method was good and relatively small sample amounts (1.5 g) could be analysed, a non-invasive approach is preferable for ceramic analyses.

The decorations on small artefacts from the Roman period were analysed¹¹⁶ with various μ-XRF techniques to show that oxides used in colourising enamels had much more varied chemical compositions than previously assumed. The scanning μ-XRF setup (2.5 × 2.5 mm, lateral resolution 0.05 mm) provided non-destructive analysis of colour motifs on enamel-decorated artefacts. The XRF unit used¹¹⁷ in the 2D and 3D μ-XRF analysis of artefacts from the Paleolithic period and Iron Age had confocal geometry. The LOD was 25 μg g⁻¹ with a 5% error. In the case of a prehistoric flint (late Pleistocene), it was possible to identify inclusions and to characterise their morphology by means of an intensity level diagram. Due to the high spatial resolution on the z-axis (80 μm), it was also possible to measure eight 2D fluorescence layers and reconstruct them using a 3D analysis. By selecting the Fe Kα-line, the authors identified the inclusion within the sample and successfully reconstructed its morphology by means of a 3D intensity diagram.

Energy dispersive-XRF spectrometry with a Monte Carlo simulation algorithm was used¹¹⁸ to determine the chemical composition of a selected group of 26 copper-based artefacts and fragments recovered at Perdigoes (Southern Portugal). All the analysed artefacts had a two-layered structure composed of the alloy substrate and a superficial layer formed by interaction of the alloy with soil. Complications in analysis resulting from this superficial layer were overcome through application of the X-ray Monte Carlo software. The artefacts were almost pure copper with <3% As and even lower concentrations of elements such as Bi, Pb and Sb. A collection of 33 anthropomorphic handle attachments of Roman situlae recovered from the archaeological site of Conimbriga (Central Portugal, 2^nd century BC – 5^th century AD), studied¹¹⁹ by μ-EDXRF spectrometry, SEM-EDS and metallographic techniques, was heterogeneous, consisting mainly of leaded copper and leaded bronzes with a wide range of Pb contents. The results suggested a local metallurgical tradition of using copper and copper alloy scraps and high additions of lead. Jin et al.¹²⁰ characterised a plated metallic fragment of the Han Dynasty from the Wushan County Museum by combining various analytical techniques, including WDXRF spectrometry (Rh tube, 40 kV, 95 mA). The bronze fragment had a ∼3 μm thick gilded layer and a ∼20 μm thick silvered layer. The high concentration of Hg and Hg-rich intermetallic compounds confirmed the use of a mercury gilding/silvering technology. An ordered stripe structure in the gilded surface corresponded to the formation of rod-like, intermetallic Au–Hg compounds. Specifically, a high concentration of Au was detected in the silvered layer. The protocol CHARMed PyMca was designed¹²¹ to improve the inter-laboratory reproducibility of EDXRF results for the wide range of copper alloys found in heritage materials. A reproducibility study was conducted using CHARMed PyMca and eight different EDXRF spectrometers of six different types. In comparison to a 2010 study,¹²² use of CHARMed PyMca gave a dramatic improvement in reproducibility and method sensitivity. For Pb, the inter-laboratory reproducibility improved from ±77 to ±18%, and for As, the LOD dropped from 0.24% to 0.13%.

The μ-EDXRF spectrometry analysis of Chalcolithic gold artefacts assigned to the Bell Beaker Culture in Portuguese Estremadura showed¹²³ the collection to be composed mainly of gold with 8.7–16.3% Ag and <0.04% Cu. The relative intensity of the Ag Kα and Ag Lα X-rays from these artefacts established the existence of a surface layer depleted in Ag. The penetration depth (about 25–30 μm) of the Ag Kα X-rays was sufficient for such gold alloys which have very smooth and homogeneous surfaces. Overall, the collection had a composition similar to that of Chalcolithic gold from Portuguese Estremadura but different from coeval gold from the southwestern Iberian peninsula. The composition and microstructural characteristics of fused gold objects were established¹²⁴ using a multidisciplinary approach which included portable μ-XRF spectrometry. The concentrations of Ag, Au and Cu varied according to the fusion temperature used. In the SR-XRF analysis of 11 gold samples from Bavaria, different excitation and detection conditions were applied¹²⁵ to confirm or disprove results obtained previously. The main impurities in the gold (>99.9% purity) were Ag (20–200 μg g⁻¹) and Cu (1–10 μg g⁻¹). These findings were in very good agreement with LA-ICP-MS measurements and provided basic information for discussions about provenance, processing techniques or authenticity.

Insights into the plating techniques used to produce counterfeit coins were provided¹²⁶ by μ-XRF analysis which has the capacity for conducting either selective spot analysis or element mapping of a larger area. It was possible to distinguish between authentic coins and forgeries very quickly and effectively to identify the metal-plating traces even if the plates were covered with a layer of green-coloured corrosion products. The Cu Kα/Cu Kβ ratio provided information on depth distribution of Cu and thereby the plating of a copper-based material could be recognised. The full characterisation of seven silver/copper coins could only be made¹²⁷ using both bulk and micro-area methods which included benchtop EDXRF and μ-XRF spectrometries.

A few interesting and remarkable applications are also brought to the reader’s attention. Elevated Hg levels (up to 1100 μg g⁻¹) were determined¹²⁸ in archaeological hair samples (n = 41) from an ancient burial in Xiongnu (Mongolia, 1^st century BC to 1^st century AD). The μ-SR-XRF spectrometry setup used polycapillary lenses in a confocal arrangement to establish Hg distribution in a cross section of hair shaft. The spatial resolution was 5 μm. Wouters et al.¹²⁹ demonstrated that μ-XRF element mapping was a viable option for the qualitative examination of thin sections of archaeological metal remains. This held true even in the case of thin sections covered with a borosilicate glass of 60 μm thickness since a wide array of elements (Ag, Au, Cu, Fe, Pb, potentially Sn, and Zn if not present in the coverslip) could be detected. This not only made element mapping faster, easier and more accessible than previously possible but also opened up possibilities for analysis of the vast archives of existing archaeological thin sections containing metallurgical components. A portable EDXRF spectrometer was used¹³⁰ in the non-invasive study of the element composition of eight musical instruments made by Antonio Stradivari between 1669 and 1734. There were substantial differences in the element compositions of these instruments and that of a modern instrument. The presence of Fe in association with Mn was ascribed to Umber earth pigments and that of Ca, Cl and K to wood sealer. Garcia-Florentino et al.¹³¹ quantified light elements (Z ≤ 20) in aqueous extracts and heavy elements (Z > 20) in acid extracts of materials from “built heritage” (mortars, black crusts, and calcium carbonate formations). The methodology involved EDXRF spectrometry measurements of thin films deposited on special sample retainers. Both external calibration and standard addition were used to analyse these real samples of “built heritage”. Although this methodology was not as easy to use as ICP-MS analysis, the authors promoted this EDXRF technology as a ‘green chemistry’ alternative. In a study of the colouration of late-antique and medieval glasses, Co, Cr, Cu, Fe, Mn, Ni, Ti, V and Zn were determined¹³² in glass fragments using TXRF spectrometry. Samples were analysed as slurries in Triton X114 and the method validated using the CRM BAM-S005 Type A (soda–lime–silica glass). In most cases the RSD was <10%. The majority of the existing colour variations were due to the deliberate addition of Co, Cu, Fe and Mn to the glass melt under appropriate melting conditions but in some cases colouration resulted from the presence of impurities in the raw materials.

Pessanha et al.¹³³ evaluated five methods of quantification in order to determine trace elements in exoskeletons of clams from the Bronze Age to the 16^th century AD (Tagus estuary, Portugal). Two EDXRF spectrometry setups were considered: a non-commercial benchtop system with tri-axial geometry and a commercially available μ-XRF system with a conventional geometry and vacuum capabilities. Comparison of results obtained from pressed pellet samples led to the conclusion that the most realistic results were obtained using calibration curves obtained with external standards and correction of the fluorescent intensities using the Compton scattering peak. The Fe content was lower for samples from the 12^th century than for samples from other centuries. To conclude, readers are referred to our companion ASU review on advances in the analysis of metals, chemicals and materials.³ The growing use of field-portable instrumentation for in situ analysis was recognised to have a substantial impact in the examination of cultural heritage artefacts, especially in relation to paintings, wall murals and other objects of unique historical value.

Conflicts of interest

There are no conflicts to declare.

Abbreviations

1D	one dimensional
2D	two dimensional
3D	three dimensional
AD	anno domini
AFM	atomic force microscopy
ASIC	application-specific integrated circuit
ASU	Atomic Spectrometry Update
BAM	Federal Institute for Materials Research and Testing (Germany)
BC	before Christ
BDE	brominated diphenyl ether
BFR	brominated flame retardant
CCA	collimating channel array
CCD	charge coupled detector
CLS	Canadian Light Source
CRM	certified reference material
CT	computed tomography
DAQ	data acquisition
DME	dimethyl ether
DWT	discrete wavelet transform
ED	energy dispersive
EDS	energy dispersive X-ray spectrometry
EDXRF	energy dispersive X-ray fluorescence
ESRF	European Synchrotron Radiation Facility
EU	European Union
FORS	fibre optics reflectance spectrometry
FP	fundamental parameter
FRC	Fourier ring correlation
FWHM	full width at half maximum
GC-MS	gas chromatography mass spectrometry
GEM	gas electron multiplier
GE-XRF	grazing emission X-ray fluorescence
GI-XRF	grazing incidence X-ray fluorescence
HH	hand-held
IAEA	International Atomic Energy Agency
ICP-OES	inductively coupled plasma optical emission spectrometry
ICP-MS	inductively coupled plasma mass spectrometry
IR	infrared
ISO	International Organization for Standardization
KXRF	K-line X-ray fluorescence
LA	laser ablation
LOD	limit of detection
LPP	laser-produced plasma
MDL	method detection limit
MOUDI	micro-orifice uniform-deposit impactor
NP	nanoparticle
NSLS	National Synchrotron Light Source
P/B	peak-to-background ratio
PCA	principal component analysis
PM	particulate matter
PMMA	polymethylmethacrylate
PV	photovoltaic
RM	reference material
RSD	relative standard deviation
RSF	relative sensitivity factor
RXR	Rainbow X-ray
SDD	silicon drift detector
SEM	scanning electron microscopy
SiVM	simplex volume maximisation
S/N	signal-to-noise ratio
SR	synchrotron radiation
SRM	standard reference material
SRX	submicron resolution X-ray spectroscopy
STXM	scanning transmission X-ray microspectroscopy
TCO	transparent conducting oxide
TEM	transmission electron microscopy
TXRF	total reflection X-ray fluorescence
WDXRF	wavelength dispersive X-ray fluorescence
XAFS	X-ray absorption fine structure spectrometry
XANES	X-ray absorption near edge structure
XAS	X-ray absorption spectroscopy
XRD	X-ray diffraction
XRF	X-ray fluorescence
XRR	X-ray reflectometry
Z	atomic number

References

1. O. T. Butler, W. R. L. Cairns, J. M. Cook, C. M. Davidson and R. Mertz-Kraus, J. Anal. At. Spectrom., 2018, 33(1), 8–56.
2. A. Taylor, N. Barlow, M. P. Day, S. Hill, M. Patriarca and M. White, J. Anal. At. Spectrom., 2017, 32(3), 432–476.
3. S. Carter, A. Fisher, B. Gibson, J. Marshall, R. Ben and I. Whiteside, J. Anal. At. Spectrom., 2017, 32(11), 2068–2117.
4. R. Clough, C. F. Harrington, S. J. Hill, Y. Madrid and J. F. Tyson, J. Anal. At. Spectrom., 2017, 32(7), 1239–1282.
5. M. D. de Jonge, A. M. Kingston, N. Afshar, J. Garrevoet, R. Kirkham, G. Ruben, G. R. Myers, S. J. Latham, D. L. Howard, D. J. Paterson, C. G. Ryan and G. Mccoll, Opt. Express, 2017, 25(19), 23424–23436.
6. T. Sasaya, N. Sunaguchi, S. J. Seo, K. Hyodo, T. Zeniya, J. K. Kim and T. Yuasa, Nucl. Instrum. Methods Phys. Res., Sect. A, 2018, 886, 71–76.
7. S. Y. Zhang, L. Li, R. Z. Li and Z. Q. Chen, Opt. Eng., 2017, 56(11), 7.
8. S. Jung, W. Sung and S. J. Ye, Int. J. Nanomed., 2017, 12, 5805–5817.
9. S. Jiang, P. He, L. Deng, M. Chen and B. Wei, Int. J. Biomed. Imaging, 2017, 2017, 7916260.
10. L. Li, S. Y. Zhang, R. Z. Li and Z. Q. Chen, Opt. Eng., 2017, 56(4), 5.
11. J. George, B. H. Yoon and L. J. Meng, J. Nucl. Med., 2017, 58, 2.
12. B. Laforce, B. Masschaele, M. N. Boone, D. Schaubroeck, M. Dierick, B. Vekemans, C. Walgraeve, C. Janssen, V. Cnudde, L. Van Hoorebeke and L. Vincze, Anal. Chem., 2017, 89(19), 10617–10624.
13. Z. W. Di, S. Chen, Y. P. Hong, C. Jacobsen, S. Leyffer and S. M. Wild, Opt. Express, 2017, 25(12), 13107–13124.
14. R. Huang, K. Limburg and M. Rohtla, AIP Adv., 2017, 7(5), 11.
15. N. R. Viganò and V. A. Solé, Meas. Sci. Technol., 2018, 29(3), 13.
16. Parsons.
17. D. H. McNear Jr, E. Peltier, J. Everhart, R. L. Chaney, S. Sutton, M. Newville, M. Rivers and D. L. Sparks, Environ. Sci. Technol., 2005, 39, 2210–2218.
18. A. Bergamaschi, K. Medjoubi, C. Messaoudi, S. Marco and A. Somogyi, J. Phys.: Conf. Ser., 2017, 849, 012060.
19. M. W. M. Jones, D. J. Hare, S. A. James, M. D. de Jonge and G. McColl, Anal. Chem., 2017, 89(22), 12168–12175.
20. L. Lemelle, A. Simionovici, T. Schoonjans, R. Tucoulou, E. Enrico, M. Salome, A. Hofmann and B. Cavalazzi, TrAC, Trends Anal. Chem., 2017, 91, 104–111.
21. T. L. Sheppard, S. W. T. Price, F. Benzi, S. Baier, M. Klumpp, R. Dittmeyer, W. Schwieger and J. D. Grunwaldt, J. Am. Chem. Soc., 2017, 139(23), 7855–7863.
22. S. Cagno, D. A. Brede, G. Nuyts, F. Vanmeert, A. Pacureanu, R. Tucoulou, P. Cloetens, G. Falkenberg, K. Janssens, B. Salbu and O. C. Lind, Anal. Chem., 2017, 89, 11435–11442.
23. G. Chen, S. Chu, T. Sun, X. Sun, L. Zheng, P. An, J. Zhu, S. Wu, Y. Du and J. Zhang, J. Synchrotron Radiat., 2017, 24, 1000–1005.
24. S. Bauters, P. Tack, J. H. Rudloff-Grund, D. Banerjee, A. Longo, B. Vekemans, W. Bras, F.E. Brenker, R. Van Silfhout and L. Vincze, Anal. Chem., 2018, 90, 2389–2394.
25. A. Iida, X-Ray Spectrom., 2017, 46(4), 225–228.
26. S. Choudhury, D. N. Agyeman-Budu, A. R. Woll, T. Swanston, T. L. Varney, D. M. L. Cooper, E. Hallin, G. N. George, I. J. Pickering and I. J. Coulthard, J. Anal. At. Spectrom., 2017, 32, 527–537.
27. P. Tack, B. Vekemans, B. Laforce, J. Rudloff-Grund, W. Y. Hernández, J. Garrevoet, G. Falkenberg, F. E. Brenker, P. Van Der Voort and L. Vincze, Anal. Chem., 2017, 89, 2123–2130.
28. M. Rauwolf, A. Turyanskaya, A. Roschger, J. Prost, R. Simon, O. Scharf, M. Radtke, T. Schoonjans, A. G. Buzanich, K. Klaushofer, P. Wobrauschek, J. G. Hofstaetter, P. Roschger and C. Streli, J. Synchrotron Radiat., 2017, 24, 307–311.
29. M. Rauwolf, B. Pemmer, A. Roschger, A. Turyanskaya, S. Smolek, A. Maderitsch, P. Hischenhuber, M. Foelser, R. Simon, S. Lang, S. E. Puchner, R. Windhager, K. Klaushofer, P. Wobrauschek, J. G. Hofstaetter, P. Roschger and C. Streli, X-Ray Spectrom., 2017, 46(1), 56–62.
30. R. Hosomi, J. Chin, T. Doi, K. Akioka and K. Tsuji, Bunseki Kagaku, 2017, 66(10), 713–718.
31. I. Szaloki, A. Gerenyi and G. Radocz, X-Ray Spectrom., 2017, 46(6), 497–506.
32. P. Liu, C. J. Ptacek, D. W. Blowes and Y. Z. Finfrock, J. Anal. At. Spectrom., 2017, 32(8), 1582–1589.
33. I. Mantouvalou, T. Lachmann, S. P. Singh, K. Vogel-Mikus and B. Kanngiesser, Anal. Chem., 2017, 89(10), 5453–5460.
34. I. Szalóki, A. Gerenyi, G. Radocz, A. Lovas, B. De Samber and L. Vincze, J. Anal. At. Spectrom., 2017, 32(2), 334–344.
35. J. Yang, Y. D. Li, X. Y. Wang, X. Y. Zhang and X. Y. Lin, Nucl. Instrum. Methods Phys. Res., Sect. B, 2017, 401, 25–28.
36. U. E. A. Fittschen, H. H. Kunz, R. Hohner, I. M. B. Tyssebotn and A. Fittschen, X-Ray Spectrom., 2017, 46(5), 374–381.
37. M. R. Gherase and A. F. Vargas, Nucl. Instrum. Methods Phys. Res., Sect. B, 2017, 395, 5–12.
38. M. Yamanashi, A. Yamauchi and K. Tsuji, Bunseki Kagaku, 2017, 66(12), 901–907.
39. S. Aida, Y. Takimoto, T. Sakumura, K. Matsushita, T. Shoji, N. Kawahara and K. Tsuji, Bunseki Kagaku, 2017, 66(12), 885–892.
40. J. Trebolazabala, M. Maguregui, H. Morillas, A. de Diego and J. M. Madariaga, Microchem. J., 2017, 131, 137–144.
41. S. Aida, T. Matsuno, T. Hasegawa and K. Tsuji, Nucl. Instrum. Methods Phys. Res., Sect. B, 2017, 402, 267–273.
42. H. A. Castillo-Michel, C. Larue, A. E. P. del Real, M. Cotte and G. Sarret, Plant Physiol. Biochem., 2017, 110, 13–32.
43. G. Das, A. Khooha, A. K. Singh and M. K. Tiwari, X-Ray Spectrom., 2017, 46(5), 448–453.
44. J. Cesar da Silva, A. Pacureanu, Y. Yang, S. Bohic, C. Morawe, R. Barrett and P. Cloetens, Optica, 2017, 4(5), 492.
45. G. Chen, S. Chu, T. Sun, X. Sun, L. Zheng, P. An, J. Zhu, S. Wu, Y. Du and J. Zhang, J. Synchrotron Radiat., 2017, 24(5), 1000–1005.
46. P. Grochulski, M. Fodje, S. Labiuk, T. W. Wysokinski, G. Belev, M. Korbas and S. M. Rosendahl, Nucl. Instrum. Methods Phys. Res., Sect. B, 2017, 411, 17–21.
47. A. G. Karydas, M. Czyzycki, J. J. Leani, A. Migliori, J. Osan, M. Bogovac, P. Wrobel, N. Vakula, R. Padilla-Alvarez, R. H. Menk, M. G. Gol, M. Antonelli, M. K. Tiwari, C. Caliri, K. Vogel-Mikus, I. Darby and R. B. Kaiser, J. Synchrotron Radiat., 2018, 25, 189–203.
48. Y. Menesguen, B. Boyer, H. Rotella, J. Lubeck, J. Weser, B. Beckhoff, D. Grotzsch, B. Kanngiesser, A. Novikova, E. Nolot and M. C. Lepy, X-Ray Spectrom., 2017, 46(5), 303–308.
49. X. Y. Lan, D. X. Liang and C. W. Mao, Nucl. Sci. Technol., 2017, 28(5), 4.
50. A. E. Morishige, H. S. Laine, E. E. Looney, M. A. Jensen, S. Vogt, J. B. Li, B. Lai, H. Savin and T. Buonassisi, IEEE J. Photovolt., 2017, 7(3), 763–771.
51. B. M. West, M. Stuckelberger, H. Guthrey, L. Chen, B. Lai, J. Maser, V. Rose, W. Shafarman, M. Al-Jassim and M. I. Bertoni, Nano Energy, 2017, 32, 488–493.
52. J. Bufon, A. Gianoncelli, M. Ahangarianabhari, M. Altissimo, P. Bellutti, G. Bertuccio, R. Borghes, S. Carrato, G. Cautero, A. Cicuttin, M. L. Crespo, S. Fabiani, M. Gandola, G. Giacomini, D. Giuressi, G. Kourousias, R. H. Menk, A. Picciotto, C. Piemonte, A. Rachevski, I. Rashevskaya, S. Schillani, A. Stolfa, A. Vacchi, G. Zampa, N. Zampa and N. Zorzi, X-Ray Spectrom., 2017, 46(5), 313–318.
53. L. Borgese, R. Dalipi, A. Riboldi, F. Bilo, A. Zacco, E. Bontempi and L. E. Depero, in 17th International Conference on Total Reflection X-ray Fluorescence Analysis and Related Methods, Brescia, Italy, 2017, p. 19.
54. L. Borgese and L. E. Depero, Spectrochim. Acta, Part B, 2018, 139, 83–84.
55. V. Panchuk, A. Goydenko, A. Grebenyuk, S. Irkaev, A. Legin, D. Kirsanov and V. Semenov, J. Anal. At. Spectrom., 2017, 32(6), 1224–1228.
56. K. Buddhadev, K. Sanyal and N. L. Misra, X-Ray Spectrom., 2017, 46(4), 277–282.
57. J. Prost, P. Wobrauschek and C. Streli, X-Ray Spectrom., 2017, 46(5), 454–460.
58. J. Baumann, C. Herzog, M. Spanier, D. Grotzsch, L. Luhl, K. Witte, A. Jonas, S. Gunther, F. Forste, R. Hartmann, M. Huth, D. Kaok, D. Steigenhofer, M. Kramer, T. Holz, R. Dietsch, L. Struder, B. Kanngiesser and I. Mantouvalou, Anal. Chem., 2017, 89(3), 1965–1971.
59. F. Brigidi and G. Pepponi, X-Ray Spectrom., 2017, 46(2), 116–122.
60. H. Rotella, B. Caby, Y. Menesguen, Y. Mazel, A. Valla, D. Ingerle, B. Detlefs, M. C. Lepy, A. Novikova, G. Rodriguez, C. Streli and E. Nolot, Spectrochim. Acta, Part B, 2017, 135, 22–28.
61. S. Pessanha, A. Samouco, R. Adao, M. L. Carvalho, J. P. Santos and P. Amaro, X-Ray Spectrom., 2017, 46(2), 102–106.
62. F. Li, A. X. Lu and J. H. Wang, Int. J. Environ. Res. Public Health, 2017, 14(10), 12.
63. A. Massos and A. Turner, Environ. Pollut., 2017, 227, 139–145.
64. A. Turner, Mar. Pollut. Bull., 2017, 124(1), 286–291.
65. A. Turner, H. Poon, A. Taylor and M. T. Brown, Sci. Total Environ., 2017, 593, 227–235.
66. A. Turner and M. Filella, Sci. Total Environ., 2017, 584, 982–989.
67. J. Kuang, M. A. E. Abdallah and S. Harrad, Sci. Total Environ., 2018, 610, 1138–1146.
68. P. Hennebert and M. Filella, Waste Manag., 2018, 71, 390–399.
69. I. F. Mikhailov, A. A. Baturin, A. I. Mikhailov, S. S. Borisova, M. V. Reshetnyak and D. I. Galata, J. X-Ray Sci. Technol., 2017, 25(3), 515–532.
70. D. E. B. Fleming and C. S. Ware, Appl. Radiat. Isot., 2017, 121, 91–95.
71. A. J. Specht, M. G. Weisskopf and L. H. Nie, Physiol. Meas., 2017, 38(3), 575–585.
72. E. D. Desouza, M. R. Gherase, D. E. B. Fleming, D. R. Chettle, J. M. O’Meara and F. E. McNeill, Appl. Radiat. Isot., 2017, 123, 82–93.
73. J. G. Ryan, J. W. Shervais, Y. Li, M. K. Reagan, H. Y. Li, D. Heaton, M. Godard, M. Kirchenbaur, S. A. Whattam, J. A. Pearce, T. Chapman, W. Nelson, J. Prytulak, K. Shimizu, K. Petronotis and I. E. S. Team, Chem. Geol., 2017, 451, 55–66.
74. M. F. Gazley, L. C. Bonnett, L. A. Fisher, W. Salama and J. H. Price, Aust. J. Earth Sci., 2017, 64(7), 903–917.
75. E. Ytreberg, M. Lagerstrom, A. Holmqvist, B. Eklund, H. Elwing, M. Dahlstrom and P. Dahl, Environ. Pollut., 2017, 225, 490–496.
76. J. H. Park, I. A. Mudunkotuwa, K. J. Crawford, T. R. Anthony, V. H. Grassian and T. M. Peters, Aerosol Sci. Technol., 2017, 51(1), 108–115.
77. M. Furger, M. C. Minguillon, V. Yadav, J. G. Slowik, C. Huglin, R. Frohlich, K. Petterson, U. Baltensperger and A. S. H. Prevot, Atmos. Meas. Tech., 2017, 10(6), 16.
78. Y. Y. Li, M. A. Chang, S. S. Ding, S. W. Wang, D. Ni and H. T. Hu, J. Environ. Manage., 2017, 196, 16–25.
79. B. Norlin, S. Reza, C. Frojdh and T. Nordin, J. Instrum., 2018, 13, 11.
80. Y. Zhang, W. B. Jia, R. Gardner, Q. Shan, X. L. Zhang, G. J. Hou and H. P. Chang, Radiat. Phys. Chem., 2018, 147, 118–121.
81. Y. Zhang, W. B. Jia, R. Gardner, Q. Shan, X. L. Zhang, G. J. Hou and H. P. Chang, Radiat. Phys. Chem., 2017, 141, 235–238.
82. M. Alfeld and L. de Viguerie, Spectrochim. Acta, Part B, 2017, 136, 81–105.
83. M. Cotte, E. Pouyet, M. Salome, C. Rivard, W. De Nolf, H. Castillo-Michel, T. Fabris, L. Monico, K. Janssens, T. Wang, P. Sciau, L. Verger, L. Cormier, O. Dargaud, E. Brun, D. Bugnazet, B. Fayard, B. Hesse, A. E. P. Del Real, G. Veronesi, J. Langlois, N. Balcar, Y. Vandenberghe, V. A. Sole, J. Kieffer, R. Barrett, C. Cohen, C. Cornu, R. Baker, E. Gagliardini, E. Papillon and J. Susini, J. Anal. At. Spectrom., 2017, 32(3), 477–493.
84. K. Trentelman, in Ann. Rev. Anal. Chem., ed. R. G. Cooks, J. E. Pemberton, Annual Reviews, Palo Alto, 2017, vol. 10, pp. 247–270.
85. F. P. Romano, C. Caliri, P. Nicotra, S. Di Martino, L. Pappalardo, F. Rizzo and H. C. Santos, J. Anal. At. Spectrom., 2017, 32(4), 773–781.
86. R. Alberti, T. Frizzi, L. Bombelli, M. Gironda, N. Aresi, F. Rosi, C. Miliani, G. Tranquilli, F. Talarico and L. Cartechini, X-Ray Spectrom., 2017, 46(5), 297–302.
87. S. Saverwyns, C. Currie and E. Lamas-Delgado, Microchem. J., 2018, 137, 139–147.
88. G. Sciutto, T. Frizzi, E. Catelli, N. Aresi, S. Prati, R. Alberti and R. Mazzeo, Microchem. J., 2018, 137, 277–284.
89. G. Van der Snickt, H. Dubois, J. Sanyova, S. Legrand, A. Coudray, C. Glaude, M. Postec, P. Van Espen and K. Janssens, Angew. Chem., Int. Ed., 2017, 56(17), 4797–4801.
90. A. Tavares da Silva, S. Legrand, G. Van der Snickt, R. Featherstone, K. Janssens and G. Bottinelli, Heritage Sci., 2017, 5, 37.
91. A. Galli, M. Caccia, R. Alberti, L. Bonizzoni, N. Aresi, T. Frizzi, L. Bombelli, M. Gironda and M. Martini, X-Ray Spectrom., 2017, 46(5), 435–441.
92. J. K. Delaney, K. A. Dooley, R. Radpour and I. Kakoulli, Sci. Rep., 2017, 7, 12.
93. F. Daniel, A. Mounier, J. Perez-Arantegui, C. Pardos, N. Prieto-Taboada, S. F. O. de Vallejuelo and K. Castro, Anal. Bioanal. Chem., 2017, 409(16), 4047–4056.
94. Y. Chen-Wiegart, J. Catalano, G. J. Williams, A. Murphy, Y. Yao, N. Zumbulyadis, S. A. Centeno, C. Dybowski and J. Thieme, Sci. Rep., 2017, 7, 9.
95. M. Alfeld, M. Wahabzada, C. Bauckhage, K. Kersting, G. van der Snickt, P. Noble, K. Janssens, G. Wellenreuther and G. Falkenberg, Microchem. J., 2017, 132, 179–184.
96. W. Dabrowski, T. Fiutowski, P. Fraczek, S. Koperny, M. Lankosz, A. Mendys, B. Mindur, K. Swientek, P. Wiacek and P. M. Wrobel, J. Instrum., 2016, 11, 9.
97. I. Marcaida, M. Maguregui, S. F. O. de Vallejuelo, H. Morillas, N. Prieto-Taboada, M. Veneranda, K. Castro and J. M. Madariaga, Anal. Bioanal. Chem., 2017, 409(15), 3853–3860.
98. L. Robeva-Cukovska, T. Sijakova-Ivanova and Z. Kokolanski, Maced. J. Chem. Chem. Eng., 2017, 36(1), 41–58.
99. D. Cristea-Stan, B. Constantinescu, C. Chiojdeanu and C. A. Simion, Rom. J. Phys., 2017, 62(1–2), 7.
100. G. Capobianco, C. Pelosi, G. Agresti, G. Bonifazi, U. Santamaria and S. Serranti, J. Cult. Herit., 2018, 29, 19–29.
101. M. Alfeld, M. Mulliez, P. Martinez, K. Cain, P. Jockey and P. Walter, Anal. Chem., 2017, 89(3), 1493–1500.
102. S. Gasanova, S. Pages-Camagna, M. Andrioti and S. Hermon, Archaeol. Anthropol. Sci., 2018, 10(1), 83–95.
103. C. Pelosi, D. Fodaro, L. Sforzini, C. Falcucci and P. Baraldi, Stud. Conserv., 2017, 62(5), 266–282.
104. J. Langlois, G. Mary, H. Bluzat, A. Cascio, N. Balcar, Y. Vandenberghe and M. Cotte, Stud. Conserv., 2017, 62(5), 247–265.
105. L. M. Smieska, R. Mullett, L. Ferri and A. R. Woll, Appl. Phys. A: Mater. Sci. Process., 2017, 123(7), 12.
106. T. Christiansen, M. Cotte, R. Loredo-Portales, P. E. Lindelof, K. Mortensen, K. Ryholt and S. Larsen, Sci. Rep., 2017, 7, 8.
107. E. Pouyet, S. Devine, T. Grafakos, R. Kieckhefer, J. Salvant, L. Smieska, A. Woll, A. Katsaggelos, O. Cossairt and M. Walton, Anal. Chim. Acta, 2017, 982, 20–30.
108. S. Pessanha, M. Alves, J. M. Sampaio, J. P. Santos, M. L. Carvalho and M. Guerra, J. Instrum., 2017, 12, 14.
109. R. Mulholland, D. Howell, A. Beeby, C. E. Nicholson and K. Domoney, Heritage Sci., 2017, 5, 19.
110. R. Wen, Y. Zhang, D. Wang and L. H. Wang, Anal. Methods, 2017, 9(30), 4380–4386.
111. A. S. Machado, D. F. Oliveira, H. S. Gama, R. Latini, A. V. B. Bellido, J. T. Assis, M. J. Anjos and R. T. Lopes, X-Ray Spectrom., 2017, 46(5), 427–434.
112. L. Angeli, S. Legnaioli, C. Fabbri, E. Grifoni, G. Lorenzetti, J. Guilaine, V. Palleschi and G. Radi, Microchem. J., 2018, 137, 174–180.
113. J. Jutimoosik, C. Sirisathitkul, W. Limmun, R. Yimnirun and W. Noonsuk, X-Ray Spectrom., 2017, 46(6), 492–496.
114. G. Barone, M. Di Bella, M. A. Mastelloni, P. Mazzoleni, S. Quartieri, S. Raneri and G. Sabatino, X-Ray Spectrom., 2018, 47(1), 46–57.
115. M. Georgakopoulou, A. Hein, N. S. Muller and E. Kiriatzi, X-Ray Spectrom., 2017, 46(3), 186–199.
116. M. Hlozek, T. Trojek, B. Komoroczy and R. Prokes, Radiat. Phys. Chem., 2017, 137, 243–247.
117. D. Hampai, A. Liedl, G. Cappuccio, E. Capitolo, M. Iannarelli, M. Massussi, S. Tucci, R. Sardella, A. Sciancalepore, C. Polese and S. B. Dabagov, Nucl. Instrum. Methods Phys. Res., Sect. B, 2017, 402, 274–277.
118. C. E. Bottaini, A. Brunetti, I. Montero-Ruiz, A. Valera, A. Candeias and J. Mirao, Appl. Spectrosc., 2018, 72(1), 17–27.
119. F. Lopes, R. J. C. Silva, M. F. Araujo and V. H. Correia, Mater. Manuf. Processes, 2017, 32(7–8), 827–835.
120. P. J. Jin, F. H. Ruan, X. G. Yang, H. X. Zou, J. Yi, Y. Zhang and Y. Zhao, Archaeometry, 2017, 59(2), 274–286.
121. A. Heginbotham and V. A. Sole, Archaeometry, 2017, 59(4), 714–730.
122. A. Heginbotham, A. Bezur, M. Bouchard, J. M. Davis, K. Eremin, J. H. Frantz, L. Glinsman, L.-A. Hayek, D. Hook, V. Kantarelou, A. G. Karydas, L. Lee, J. Mass, C. Matsen, B. Mccarthy, M. Mcgath, A. Shugar, J. Sirois, D. Smith and R. J. Speakman, in Metal 2010: Proceedings of the Interim Meeting of the ICOM–CC Metal Working Group, October 11–15, 2010, ed. P. Mardikian, C. Chemello, C. Watters, and P. Hull, Clemson University Press, Charleston, South Carolina, USA, Clemson, SC, 2011.
123. P. Valerio, A. M. M. Soares, M. F. Araujo and A. F. Carvalho, X-Ray Spectrom., 2017, 46(4), 252–258.
124. S. Scrivano, B. G. Tubio, I. Ortega-Feliu, F. J. Ager, A. Paul and M. A. Respaldiza, X-Ray Spectrom., 2017, 46(2), 123–130.
125. M. Radtke, U. Reinholz and R. Gebhard, Archaeometry, 2017, 59(5), 891–899.
126. M. Hlozek and T. Trojek, Radiat. Phys. Chem., 2017, 137, 234–237.
127. R. C. Fierascu, I. Fierascu, A. Ortan, F. Constantin, D. A. Mirea and M. Statescu, Nucl. Instrum. Methods Phys. Res., Sect. B, 2017, 401, 18–24.
128. V. V. Zvereva, V. A. Trunova, D. S. Sorokoletov and N. V. Polosmak, X-Ray Spectrom., 2017, 46(6), 563–568.
129. B. Wouters, C. Makarona, K. Nys and P. Claeys, Geoarchaeology, 2017, 32(2), 311–318.
130. T. Rovetta, C. Invernizzi, M. Licchelli, F. Cacciatori and M. Malagodi, X-Ray Spectrom., 2018, 47(2), 159–170.
131. C. Garcia-Florentino, M. Maguregui, E. Margui, I. Queralt, J. A. Carrero and J. M. Madariaga, Anal. Chem., 2017, 89(7), 4246–4254.
132. A. K. Detcheva, R. H. Velinova, A. K. Manoylova and E. H. Ivanova, Glass Technol.: Eur. J. Glass Sci. Technol., Part A, 2017, 58(5), 217–225.
133. S. Pessanha, C. Fonseca, J. P. Santos, M. L. Carvalho and A. A. Dias, X-Ray Spectrom., 2018, 47(2), 108–115.

---

Footnote

† Joint review coordinators.

---