2023 atomic spectrometry update – a review of advances in X-ray fluorescence spectrometry and its special applications

Christine Vanhoof a, Jeffrey R. Bacon b, Ursula E. A. Fittschen c and Laszlo Vincze d
aFlemish Institute for Technological Research (VITO), Boeretang 200, B-2400 Mol, Belgium. E-mail: Christine.vanhoof@vito.be
b59 Arnhall Drive, Westhill, Aberdeenshire, AB32 6TZ, UK. E-mail: bacon-j2@sky.com
cClausthal University of Technology, Institute of Inorganic and Analytical Chemistry, Arnold-Sommerfeld-Strasse 4, D-38678 Clausthal-Zellerfeld, Germany
dDepartment of Chemistry, Ghent University, Krijgslaan 281 S12, Ghent, B-9000, Belgium

Received 20th June 2023

First published on 17th July 2023


Abstract

The utilisation of SR sources has significantly enhanced the analytical capabilities of XRF spectrometry techniques. With the latest generation of facilities, SR-XRF spectrometry achieves remarkably high nm-scale resolution with excellent LODs at ppb levels. A noteworthy trend is the increasing use of SR-XRF spectrometry together with other X-ray spectroscopic and imaging techniques. This provides complementary information on elemental speciation as well as structural and morphological characteristics of samples. Sub-μm SR-XRF spectrometry has been extensively applied in diverse fields such as environmental and planetary studies, biomedical research, materials science and cultural heritage investigations. Methods for handling the huge datasets produced by macroXRF spectrometry have become essential for processing and classifying the element distributions collected from the analysis of paintings. Machine-learning-based correlations of element maps have been developed for the automatic identification of patterns as an alternative method of processing macroXRF spectrometry data from cultural heritage samples. The microanalytical capabilities of TXRF spectrometry have led to a steep increase in applications to biomedical problems with successful analyses of minute amounts of samples (ca. 20 mg) of, e.g., blood, placenta and heart tissue. The suspension-assisted preparation of theses samples and of mineralogical materials was improved in many studies by extending the common internal standard calibration with uni- and multivariate approaches. The development of a scan-free grazing-exit XRF spectrometer improved accuracy in the analysis of periodic surface structures. The degree of protonation of different thiol- or hydroxyl-bearing organic monolayers was successfully determined using grazing-incidence XRF and TXRF spectrometries in combination with other techniques.


1. Introduction

This review describes advances in the XRF spectrometry group of techniques published approximately between April 2022 and March 2023. 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 spectrometry techniques except in those cases where the non-destructive and remote-sensing nature of XRF spectrometry 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. For a wider appreciation of the applications of XRF spectrometry, this review should be read in conjunction with other related ASUs in the series, namely: environmental analysis;1 clinical and biological materials, foods and beverages;2 advances in atomic spectrometry and related techniques;3 elemental speciation;4 and metals, chemicals and functional materials.5

All the ASU reviews adhere to a number of conventions. An italicised phrase close to the beginning of each paragraph is intended to highlight the subject area of that individual paragraph. A list of abbreviations used in this review appears at the end. It is a convention of ASU that information given in the paper being reported on is presented in the past tense whereas the views of the ASU reviewers are presented in the present tense.

2. Chemical imaging using X-ray techniques

2.1. Computed tomography and 3D XRF spectrometry techniques

The XRF computed tomography (XRF-CT) and 3D XRF spectrometry techniques based on confocal excitation and detection strategies continue to be developed both in terms of data collection and (quantitative) evaluation. The XRF-CT technique is now regularly applied to the determination of 2D cross-sectional and full 3D elemental distributions at the nanoscale using SR facilities and at the microscale using laboratory instruments.

Luo et al.6 demonstrated an efficient approach to expand the useable tomographic angular range making it possible to analyse local areas of interest within samples with large width-to-thickness ratios. The authors performed a sequence of XRF-CT data collections from a range of areas (micro- and nanoscale) from the same specimen. Focussed ion beam (FIB) processing was applied between each of the XRF-CT scans. The precise sample manipulation using FIB allowed the direct exclusion of sample regions that represented obstacles to both the incident X-ray beam and the XRF spectrometry signals. By alleviating the geometric constraints imposed by the commonly used SiNx substrate and the limitations originating from the sample material itself, the quality of reconstructions of P and Ti distributions in HeLa cells exposed to Fe3O4/TiO2 nanocomposites was improved considerably.

Ge et al.7 developed a highly accurate quantitative XRF-CT method for analysing the 3D grain (boundary) distributions and compositions in a Ce0.8Gd0.2O2−δ–CoFe2O4 mixed ionic–electronic conductor system. The method provided statistically significant quantitative results for Fe, Ce, Co and Gd, illustrating the suitability of the approach for diverse investigations from the micro- to the nanoscale. A particularly interesting aspect of this work was that their XRF-CT reconstruction method did not rely on an idealised 2D cross-sectional geometry but took the entire 3D object and the associated escape path into account without compromising the absorption correction of fluorescence emitted from each element. The iterative correction of self-absorption relied on maximum-likelihood methods and required only XRF spectrometry data-sets to perform the correction. Absorption data were not needed. This innovative tomographic reconstruction method has the potential to advance significantly the field of XRF-CT reconstruction. The XRF tomography measurements were conducted at the Hard X-ray Nanoprobe (HXN) beamline of the National Synchrotron Light Source II (NSLS-II, Brookhaven, USA) by collecting for each sample a total of 61 2D XRF spectrometry projection images over a 180° angular range, each consisting of 160 × 160 pixels (80 nm).

The main motivation of a very-high-spatial-resolution experiment by Schulz et al.8 was to demonstrate the non-destructive capability of XRF-CT for identifying the location of implanted NPs within 3D nanostructures. The procedure was applied to samples of Si photonic band-gap crystals which had been functionalised by the targeted positioning of PbS quantum dots. The measurements were carried out at the ID16A beamline of the ESRF (Grenoble, France) using a 17 keV nanobeam of 23 × 37 nm2 (FWHM). The LOD for Pb was 0.4 ppm. This study was among the highest resolution XRF-CT experiments to date and demonstrated the high sensitivity of the technique for the non-destructive 3D characterisation of nanomaterials with photonic functionalities.

In preparation for the initial analysis of material returned from asteroid Ryugu by the Hayabusa2 spacecraft, De Pauw et al.9 carried out high-energy μXRF spectrometry and XRF-CT experiments on analogue materials originating from the Mighei-type Murchison meteorite. Optimisation for the detection of the lanthanides in mm-sized samples included the unique combination of high-energy excitation (90 keV) and sub-μm spatial-resolution (0.5 × 0.5 μm) at the ESRF ID15A beamline. High-energy synchrotron-radiation-based μXRF spectrometry used in conjunction with XRF-CT-based depth analysis represented an excellent approach to locate, to identify and to non-destructively map microscopic calcium–aluminum rich inclusions (<12 μm) in mm-sized extra-terrestrial samples. Measurement of the high-atomic-number (Z > 55) K-line XRF emission to a depth of ca. 1 mm was made, thereby enabling the detection of lanthanides from the entire sample cross section with LODs as low as 2 ppm for an acquisition time of 1 s per point. Following on from this initial study, Tack et al.10 used the same instrumentation to analyse primordial rock fragments up to 2.5 mm in size originating from Ryugu to provide 2D and 3D elemental distributions and quantitative composition information at the sub-μm level. To avoid exposure to air and potential contamination by terrestrial materials, the asteroid samples (rock fragments A0055 and C0076) were sealed in 20 μm polyimide foil caps for the μXRF spectrometry and XRF-CT analyses. The sub-μm excitation beam (90 keV, 1011 photons per s intensity) allowed the non-invasive investigation of these mm-sized samples to be carried out. By using an acquisition time of 600 s for each point analysed, REE compositions could be determined for specific mineral phases from the entire sample cross-section with LODs below 0.1 ppm.

An excellent example of an environmental application of XRF-CT was the work of Andresen et al.11 who studied the mechanism of sublethal Cd toxicity in soybean (Glycine max (L.) Merr.) by applying metabolomics and metalloproteomics established by HPLC-ICP-MS in conjunction with spatial distributions of Cd determined by microbeam XRF-CT. The plants were exposed to a range of Cd concentrations: background and non-toxic (0.5–50 nM); sublethally toxic (<550 nM); and lethally toxic (3 μM). The high-quality distribution maps acquired by XRF-CT demonstrated subcellular resolution at sub-μM Cd concentration levels. The distribution pattern of Cd changed with increasing concentrations. At low exposure levels, Cd was mainly present in the vacuoles and filled most of the cells. At high concentration levels, Cd-rich rings around the cells were observed, indicating the presence of Cd in the cell walls. The experiment was performed at the P06 beamline of PETRA III (Hamburg, Germany) using a 32 keV X-ray beam focused to a size of ca. 1.2 × 0.7 μm (FWHM) and with a photon flux of 4 × 109 photons per s.

All of the studies described above relied on the use of monochromatic and focused SR-beam excitation and scanning probes. However, there have also been significant advances in laboratory-based XRF-CT towards the use of full-field imaging. Li et al.12 demonstrated full-field in vivo imaging of gadolinium NPs using a benchtop cone-beam XRF-CT system equipped with a cadmium–zinc–telluride pixelated photon-counting-detector of 32 × 32 pixels, each of 1.6 mm size, and with a detector pinhole of 1 mm. The LOD for Gd of 2 mg mL−1 allowed areas with a local concentration level of 0.2% Gd to be successfully reconstructed based on a 45 min XRF-CT scan (150 kV, 3 mA, 0.2 mm copper beam filter) of a PMMA phantom.

A similar laboratory-based system, optimised for the quantitative XRF-CT imaging of gold NPs was characterised by Jayarathna et al.13 The benchtop setup used a spectroscopic pixelated CdTe detector system as a 2D array detector with 80 × 80 individual pixels of 250 μm size and a pinhole collimator of 2 mm for imaging. Each pixel could accumulate XRF spectrometry spectra in 800 channels in the energy range 3–200 keV at a frame rate of ca. 9 kHz. This provided an energy resolution of ca. 1 keV (FWHM) in the region of the gold K-lines (ca. 68 keV). Using a high-power tungsten-target X-ray source operated at 125 kV and 24 mA, the LODs for Au were 0.05 and 0.03 mg cm−3 with 5 and 10 s data acquisition times, respectively. This was possibly the highest sensitivity laboratory XRF-CT imaging result to date for cm-size objects using biologically relevant concentration levels of gold NPs at ca. 100 ppm.

Confocal detection methods for 3D XRF spectrometry analysis remained popular at both synchrotron and laboratory facilities with the quantitative aspects gaining more and more attention. Förste et al.14 presented a detailed quantification approach for the laboratory-based confocal-XRF spectrometry of homogeneous and layered samples. They applied an adapted fundamental parameter routine and carried out a detailed characterisation of the polycapillary lenses and associated spectrometers used. The quantification algorithm could handle full laboratory 3D XRF spectrometry datasets. Results for homogeneous bulk and layered RMs were typically within 30% of certified values for most detectable elements. Limburska and Trojek15 presented a universal confocal XRF spectrometry procedure for calculating accurately elemental depth profiles in homogeneous sample layers. Their approach was also based on the application of an adapted form of the fundamental parameter method able to calculate depth profiles for given layer thicknesses and compositions. The calculation model was tested by recording the experimental depth profiles of Ba, Ca, K and Zn in NIST SRM 1412 (multicomponent glass standard) using a confocal XRF spectrometer built in-house. The experimental depth profiles were reproduced by the model with excellent agreement, the accuracy seemingly being limited only by statistical uncertainties.

Accurate knowledge of transmission parameters is crucial for quantitative interpretation of confocal-XRF spectrometry measurements. Iro et al.16 studied the intensity and spatial beam-propagation properties of various detector polycapillary half lenses using a monochromatic confocal laboratory setup at the Atominstitut of TU Wien and a synchrotron setup installed on the BAMline at the BESSY II Synchrotron. The transmission efficiencies and spot-sizes established experimentally for three polycapillary lenses in the energy range 6–20 keV were in good agreement with theoretical values calculated using the Polycap software package and a newly presented analytical model for the transmission function.

In a high-profile planetary science application, Bazi et al.17 applied a fundamental parameter-based quantification scheme to sub-μm elemental-imaging-data obtained at beamline P06 of PETRA III (Hamburg, Germany) from samples collected from the Cb-type asteroid Ryugu. The potential for obtaining real-time quantitative elemental-information was considered a significant advantage during the initial analysis of this unique extra-terrestrial material. The derived composition of the bulk matrix of Ryugu grain C0033 was shown to be similar to that of Ivuna-type chondritic materials, a conclusion also reached when other bulk and/or destructive spectroscopic analysis methods were applied.

With the aim of better understanding and potentially mitigating NP pollution, confocal XRF spectrometry was18 one of several methods used to obtain the first evidence of NP uptake by the leaves and roots of beech (Fagus sylvatica L.) and pine (Pinus sylvestris L.) trees. Confocal (3D) analysis confirmed the presence of gold NPs in beech trichomes and leaf blades about 20–30 μm below the leaf surface. It was demonstrated that trees could take up gold NPs effectively and transport them through the plant system. The LODs for gold NPs were 0.005–0.01 μg kg−1.

In the field of cultural heritage, the use of both confocal XRF spectrometry and macroXRF spectrometry, together with SEM-EDX cross-sectional analysis, enabled Tapia et al.19 to study hidden layers in the painting Virgin and Child surrounded by saints and donor which is currently conserved in the Louvre Museum, Paris. The fact that macroXRF spectrometry did not offer depth resolution had previously hampered the collection of depth selective information on a painting's stratigraphy. The confocal μXRF spectrometry, performed on the LouX3D spectrometer installed at the C2RMF laboratory in Paris with a spot size of 50 μm, enabled elemental depth profiling and the characterisation of paint layers with thicknesses as little as 15–20 μm.

2.2. Laboratory 2D XRF spectrometry techniques

Bock et al.20 achieved sub-pixel spatial resolution by utilising the subtle misalignments which occur when the same area is scanned several times. The reconstruction of the original image was optimised for accuracy and time efficiency. Choosing one image as reference for computing the shifts for all others made the analysis time efficient (112 s for 48 XRF spectrometry images) but the reconstruction was poorer. If every image was compared to all other images, the computing time increased significantly to 2441 s for these 48 image scans. The compromise solution was the symmetric pairing of a limited number of image scans for the comparison. By limiting the number of comparisons per image to 4, the error in the reconstruction was halved and the analysis time reduced to only 228 s.

Homogeneous elemental RMs are a valuable tool in quantitative elemental microscopy. Rogoll et al.21 dissolved metal acetylacetonates in a UV-curing resin to prepare a set of RMs with trace amounts of Al, Ca, Cd, Co, Cr, Fe, La, Mg, Pb, Sr and Zn. The ability to stack the cured-resin layers made it possible to undertake depth-resolved analysis. Results for the μXRF spectrometry of 100 μm thick layers with element concentrations of 200 to 2000 mg kg−1 were in good agreement with those obtained by LA-ICP-MS.

The microscopic mineralogical characterisation of geological samples is usually undertaken by EPMA because routine procedures are widely available. In a comparison of EPMA, μLIBS and μXRF spectrometry for the mineralogical characterisation of a mica schist thin section, Fabre et al.22 highlighted the excellent acquisition speed of μLIBS and μXRF spectrometry. Although EPMA provided good spatial resolution (spot size ca. 1 μm), a major drawback was the relatively long data-acquisition-time of several hours for an area of a few mm2. In contrast, μLIBS (spot size ca. 15 μm) could be used to analyse an area of 9.6 cm2 in 4.5 h or μXRF spectrometry (spot size ca. 20 μm) could be used to analyse an area of 8.6 cm2 in 12 h. However, the poor detection capabilities for low-Z elements (e.g. Na) was a drawback of the use of μXRF spectrometry. In addition, element selection in μLIBS analysis was limited by the line-rich spectra and the need to minimise interferences. It was concluded that μLIBS and μXRF spectrometry were ideal tools for routine phase analysis of larger areas.

The use of standardless fundamental parameter calibration together with concentration values for certified silicate glasses allowed Liu et al.23 to determine accurately elements in several geological materials. A linear model was obtained from the correlation of the fundamental parameter prediction and the certified values. Applying this model to data obtained from mineral thin sections of albite, almandine, kaersutite, olivine, quartz and silimanite improved the accuracy of the determination of metal oxides (Al, Ca, Fe, Mg, Mn, Na and S) significantly compared to a fundamental-parameter-only calibration. The bias for most elements (but not Fe and Na) was reduced to 10%.

Nakae et al.24 improved the determination of the μXRF spectrometry focal spot size (ca. 11 μm) by thin wire (5–30 μm thick) scans and explored its limitations. To do so, they extended the commonly used correction procedure of subtracting the squared wire diameter from the squared determined spot size by additionally considering the overlap of the beam intensity profile with the wire at each point of the scan. They introduced correction factors to compensate for both the typical underestimation of the beam diameter and for the wire's self-absorption. According to their model calculation for a 10 μm diameter beam, the results from all the different correction approaches merged at the true beam diameter for thin wires (ca. 2 μm). The absorption correction provided correct results for wire diameters as large as 6 μm. It was concluded that the wire thickness should ideally be half of the beam diameter or less to obtain correct focal spot sizes for any absorber. The experimental results obtained from the 5 μm-wire scans had an error in the focal spot size of ca. 6% when a simple correction was applied but only 1% when the profile was taken into account. When the absorption correction was applied to 10 μm-wire scans, the error of 4% was still good but scans using 30 μm wires overestimated the focus significantly (17.09 μm).

Homogeneity is an important parameter monitored in the QC of fabricated MOX fuel pellets. Sanyal et al.25 analysed MOX fuel pellets, prepared by three fabrication routes, to provide size and concentration data for uranium agglomerates. The concentration of U (2–50%) could be determined with a high precision of <1%.

3. Synchrotron and large-scale facilities

This section of the review focuses on notable advancements in SR-XRF spectrometry, highlighting remarkable developments in both instrumental and computational aspects.

In their comprehensive review, Dinsley et al.26 illustrated the opportunities that SR techniques, including SR-XRF spectrometry imaging, offer to scientists working in the fields of environmental biology, ecology and evolutionary science. The elemental detection range and analytical characteristics of various spatially resolving elemental-analysis techniques (nano-SIMS, μPIXE, LA-ICP-MS, SEM-EDS/EPMA-WDS, STEM-EELS, STXM and SR-XRF spectrometry-imaging) were compared. In addition, aspects of sample-preparation methods appropriate for X-ray-based elemental-imaging and speciation were covered. Current uses of and trends in synchrotron techniques in ecology, evolution, and environmental biology studies were discussed. A survey of literature published in the last two decades showed that the number of publications in which synchrotron X-ray-based techniques were used increased from ca. 1000 to ca. 2500 articles per year.

Pushie et al.27 reviewed the use of synchrotron-based XRF microscopy techniques for the elemental imaging of biological tissues. A detailed overview of sample preparation and data acquisition methods for SR-XRF microscopy was mainly intended for researchers new to this technique. Consideration was given to improving data acquisition and quality and to the sample preparation of soft tissues. Sample pretreatment strategies were discussed together with factors that influence the behaviour of specific elements. This article serves as a valuable resource for those interested in utilising SR-XRF microscopy, providing both practical guidance on sample preparation and data acquisition and an in-depth discussion of the underlying factors that affect the quality and interpretation of XRF microscopy data.

Bonanni and Gianoncelli28 presented a comprehensive review of advancements in the soft X-ray regime for studying metabolic features in biological systems. The paper discussed the combined use of low-energy XRF spectrometry, STXM and XANES techniques and provided valuable information on synchrotron soft X-ray microspectroscopy beamlines suitable for metabolic research. Detailed information was given about the experimental capabilities offered by these beamlines, highlighting their relevance to life science applications and environmental research.

Harfouche et al.29 described the first general-purpose absorption and fluorescence spectroscopy beamline in the Middle East installed at the Synchrotron-light for Experimental Science and Applications in the Middle East (SESAME), Jordan. Optics previously installed and operated at the Rossendorf beamline (ROBL) of the European Synchrotron Radiation Facility (ESRF) were reused in the XAFS/XRF spectrometry beamline located at the bending-magnet D08 port. Use of this beamline, designed for the excitation-energy range 4.7–30 keV, made it possible to carry out XAFS experiments in the fluorescence mode for almost all elements from Ti and above at ppm concentrations.

Edwards et al.30 described a new synchrotron XRF-spectrometry-imaging instrument which incorporated a high-energy-resolution fluorescence detection (HERFD) XAS instrument on beamline 6-2 at the Stanford Synchrotron Radiation Lightsource (SSRL). A 57-pole 0.9 T wiggler source was coupled with a Si(111) monochromator and polycapillary optic. When a ca. 45 μm beam was used, a flux of ca. 1.3 × 1012 photons per s could be measured at 13.6 keV and a flux of ca. 8.7 × 1011 photons per s at 17.5 keV. A fast and continuous scanning system with a travel range of 250 × 200 mm facilitated the XRF spectrometry imaging by enabling multiple or large samples to be mounted. The Johann-type HERFD spectrometer consisted of seven spherically bent 100 mm-diameter crystals arranged on intersecting Rowland circles of 1 m diameter. A total solid angle of ca. 0.4% of 4π sr was covered. The available Bragg-angle-range of 64.5°–82.6° made it possible to study a wide range of emission lines. Elements in a sample were mapped rapidly by XRF spectrometry so that specific features could be selected for further analysis by HERFD-XAS. The higher spectral resolution of HERFD provided improved contrast and a better separation of interfering emission lines such as Rb Kα and U Lα1.

A dose-efficient multimodal X-ray-microscopy approach for the analysis of human tissues was demonstrated by Sala et al.31 at the hard X-ray nanoprobe beamline NanoMAX at MAX IV (Lund, Sweden). Data from fast and low-dose in-line holography scans were combined to produce quantitative electron-density maps prior to nanoscale XRF-spectrometry-data acquisition. The electron-density maps delivered morphological information with 200 nm spatial resolution and could be used to identify rapidly and reliably regions of interest (ROIs) for subsequent nanoXRF spectrometry scanning within each sample (human peripheral sural nerve biopsy of 2 μm thickness on a Si3N4 substrate). The absorbed dose during preliminary scans was nearly an order of magnitude lower than when an overview fluorescence image was used to cover the same area. Finding ROIs in this way represents a major advantage in the analysis of radiation-sensitive biological samples.

In a similar study, Quinn et al.32 used scanning XRF spectrometry with differential-phase-contrast detection for quantitative elemental imaging at the Nanoprobe Beamline of the Diamond Light Source (Oxfordshire, UK). Whereas fluorescence and absorption were sensitive only for elements with Z > 14, phase contrast imaging extended the sensitivity range to lower-Z elements by detecting shifts in the X-ray phase during transmission. The spatial resolution of 50–60 nm obtained using a detector of 512 × 512 pixels (55 μm size) complemented nanoXRF spectrometry measurements by providing sufficient detail for the identification of relevant structures in low-Z biological matrices.

An open source machine-learning software package (ROI-Finder) was developed33 for Bionanoprobe beamline 9-ID of the Advanced Photon Source (APS) (Lemont, Illinois, USA). In the analysis of samples containing biological cells treated with various chemicals, the algorithms distinguished the cells from the background and detected the different treatments. As demonstrated by the analysis of Escherichia coli (E. coli) bacteria, elemental images could be segmented automatically to extract morphological and element-content features. Application of Fuzzy k-means clustering revealed how the element contents of cells changed as a function of cell treatment.

An exciting development at Brookhaven National Laboratory (BNL, USA) for high-count-rate XRF spectrometry detection was34 the implementation of detector arrays consisting of a large number of small-area SDD pixels. The SDDs, which consisted of arrays of 32, 96 and 384-channels of 1 × 1 mm pixels, replaced the original diode-based pixel detector called Maia. A spectral resolution of 176 eV (FWHM) was achieved at an X-ray energy of 5.9 keV with a peaking time of 1 μs and a chip temperature of −13 °C. The deviation from linearity of <1% and count rates of up to 40 kcps per channel demonstrate potential for handling count rates of well over 10 Mcps using a 384-channel configuration.

4. Grazing X-ray techniques including TXRF spectrometry

The microanalytical capability of TXRF spectrometry and its ability to analyse undigested samples has led to a steep increase in its use in medical applications. Several papers reported on sample preparation strategies for and sources of error in the analysis of mammalian tissues. An excellent example of the potential of TXRF spectrometry was demonstrated by Hauser et al.35 in a study of mineral distributions in the human placenta. Minute amounts (10 mg) of powdered placenta were suspended in 1 mL HNO3 for analysis. The most crucial step in the preparation was the grinding of the sample. Grinding in a ball mill for 30 min twice rather than just once reduced the data spread by about an order of magnitude. The analysis of the suspended samples had RSDs of <10% and the results were similar to those obtained following microwave digestion of 250 mg samples. The LOD for Se was as low as 16 pg (500 s live time). The LODs for other elements were 1.3–1.6 times poorer for the analysis of the suspended samples than for digested samples. It was possible to distinguish between the maternal and fetal sides of the placenta and thereby reveal an accumulation of trace metals on the fetal side. Carvalho et al.36 studied both suspension and digestion for the preparation of specific tissues (colon, heart, liver, lung, muscle, intestine, skin, stomach, uterus, bladder, and aorta) with differing fat contents. Acid digestion was superior for tissues with high-fat contents (e.g. intestine and skin), in particular for the determination of low-Z elements which was hampered by matrix absorption in the analysis of suspended samples. The LOD for K was ca. 70 mg kg−1 in digested samples but ca. 90 mg kg−1 (both 2000 s live time) in suspended samples. The suspension in 1 mL HNO3 of as little as 20 mg of powdered low-fat materials such as heart was sufficient to yield results not significantly different to results obtained for digested samples. The optimised procedures were used with three analytical methods (ICP-OES, TXRF spectrometry and μEDXRF spectrometry) to determine elements in the different materials. There was no significant difference between the means for most elements. Even less material, 1 mm of a single hair, was required37 for the on-the-carrier digestion of human scalp hair and subsequent determination of Zn. In a search for chemical markers for early-stage detection of subclinical obesity, Szczerbowska-Boruchowska et al.38 studied earlobe samples from both obese and control rats. The study revealed that Br, Fe, K and especially Rb are robust discriminators for early-state obesity. The obese animals had significantly lower levels of Br (34%, p = 0.016), Fe (21%, p = 0.0040) and K (12%, p = 0.0088) than animals fed normal diets but 66% (p ≪ 0.05) higher concentrations of Rb. This research confirmed that obesity altered trace metal metabolism and identified Rb as a potent new marker of adiposity.

As TXRF spectrometry does not require sophisticated sample preparation, it is particularly suited to facilities with limited analytical support. For example, only simple sample-preparation-methods were needed39 to determine the concentration of essential micronutrients (Cu, Fe, Mn, and Zn) in coffee beans from Kenya in an assessment of the nutritional quality of the beans. No statistical difference was found40 between the results obtained in the WDXRF and TXRF spectrometric determinations of synthetic manganese dihydroquercetin and Se arabinogalactan, indicators for plant-derived drugs, but the RSD of the TXRF spectrometry results (2.4%) was lower than that (7%) of the WDXRF spectrometry results.

Mankovskii and Pejovic-Milic41 used La as an IS in an easy-to-use microanalytical procedure for quantifying Au and gold NPs in blood. To determine Au in whole blood and red blood-cells required a 1[thin space (1/6-em)]:[thin space (1/6-em)]5 dilution, whereas the analysis of plasma required only a 1[thin space (1/6-em)]:[thin space (1/6-em)]1 dilution to give accurate results. The determination of gold NPs in whole blood and plasma samples required digestion and a 1[thin space (1/6-em)]:[thin space (1/6-em)]1 dilution to obtain near 100% recovery but the analysis of red blood-cell samples required only a 1[thin space (1/6-em)]:[thin space (1/6-em)]5 dilution. Zambianchi et al.42 applied a Monte Carlo N-particle code to model the gold signal from dissolved ions, colloidal NPs and aggregated NPs and found that there was a drop in signal intensity for the aggregated NPs but not for the colloidal NPs. Parameters used in the simulation included those for the X-ray tube operating in the total reflection geometry and those for the certified gold NP solutions and IS (Ti) used. The line intensities in the modelled spectra of completely aggregated NPs were only 45% and 30% of the intensities of digested 50- and 100-nm-diameter NPs, respectively. It was suggested that as the TXRF spectrometry technique can quantify gold NPs held within just a few cancer cells, it has potential for detecting tumour cells circulating in blood.

The potential of TXRF spectrometry for determining elements in milk powder using suspension-assisted specimen preparation was evaluated43 by applying internal standardisation and external calibration. A set of milk powders was characterised by several laboratories and a subset taken for external calibration of the determination of Ca, Fe, K, and Zn. The accuracy for the low-Z elements Ca and K was improved, for example that for Ca improved from a bias of −1.6 to −0.05 mg g−1, bias being defined as the mean of the differences between the TXRF results and the reference value.

Freeze-drying of highly concentrated multi-elemental samples improved44 the accuracy and the LODs by a factor of 3 to 5 in comparison with the conventional drying of samples in ambient air onto siliconised reflectors. The specimens prepared with the new procedure were flatter and spread over a larger area than those heat-dried onto the hydrophobic carriers. This lower sample height reduced absorption in the analysis of a multi-element standard solution of 100 ppm Al, Ca and Fe and 10 ppm Ba, Cr, Mg, Pb and Sr using Ga as IS (a total elemental mass of 3600 ng on the carrier). However, the conventional sample preparation had the advantage of a lower sample spread and therefore optimal sample positioning in the detector field of view. As a result, the counts for elements other than Ca were higher than those achieved by the new procedure even though the elements were present at only a tenth of the concentration given above.

Dispersive SPME with β-cyclodextrin covalently linked to graphene oxide was developed by Kocot et al.45 for the enrichment of uranyl ions in water samples. Mineral, lake, river and artificial sea water samples were spiked with 5 and 10 μg L−1 UO22+ prior to analysis. Micro-amounts of the enriched phase were pipetted onto siliconised surfaces as small spots and measured by TXRF spectrometry. As there was no need for an elution step, the risk of analyte loss or sample contamination was reduced. The absorbance, as characterised by XPS, FTIR spectroscopy and Raman spectroscopy, was highly stable at high ionic strengths (up to 2 M). Absorption kinetics showed that the optimal pH for the extraction was 4.5 and that the absorption efficiency was constant over a wide range of ionic strength (90% recovery up to 2 M and 80% at 5 M). The absorption capacity of 87.7 mg g−1 was very promising for the preconcentration of uranyl ions. The UO22+ LOD of 0.014 μg L−1 (600 s live time) was extraordinarily low. Similar LODs were obtained46 for U by functionalising the surface of the quartz sample carriers with (3-amidoxy)triethoxysilane. The modified quartz carriers needed to be immersed for only 3 h in natural, ground, river and sea water samples for direct analysis by TXRF spectrometry.

The consumption of the noble metals Au, Ir, Os, Pd, Pt, Rh and Ru is constantly growing so it will become increasingly necessary to recover the metals from ores with relatively low concentrations. Maksimova et al.47 developed a procedure for the simultaneous determination of Au, Ir, Os, and Pt absorbed on a solid sorbent phase. No desorption was required for analysis by TXRF spectrometry. The two strongly basic anion-exchange resins used, one with one and the other with two pyridinium groups in the repeating unit, were the products of 4-vinylpyridine N-alkylation with either linear poly(p-vinylbenzylchloride) or poly(p-xylylene)dichloride. The maximum specimen size of ca. 14 μm was below the calculated critical thickness of 30 μm. Despite the introduction of errors in the spectra deconvolution, a bias from CRM values of only 15% for Pt and even less for Au and Ir was obtained using empirical calibration and multivariate regression algorithms. The LOQs were 0.31, 0.39, 0.43 and 0.43 mg kg−1 for Au, Ir, Os and Pt, respectively. Impressive LODs (e.g. 0.13 μg L−1 for Cr) in an aqueous solution were obtained by Akahane et al.48 with a portable low-power (5 W) TXRF spectrometer and a hydrophobic film-coated sample.

Accurate determination of elements in minerals, metal oxides and metal sulfides with minimal sample preparation remains an ongoing challenge. A review by Zhang et al.49 discussed different pretreatment strategies, calibration and data evaluation in the broader context of commercially important minerals and ores such as apatite, manganese ore, K-feldspar, granite and copper-nickel sulfide ore. A very good overview on sample preparation strategies for refractory materials was presented. In a study on copper-nickel sulfide ores, wet grinding improved50 homogeneity of the suspension and, thereby, the precision and accuracy of the determination of Cu, Fe, Ni and S. Ninety percent of the suspended particles were <6 μm in diameter. The R2 of >0.99 could be considered a good correlation for these high-density materials. The combination of internal standardisation with univariate or multivariate regression improved the accuracy compared to when only an IS was used, especially for the determination of Cu and Fe. For example, the RMSE for Fe decreased from 6.74 (IS only) to 1.95 wt% (IS and PLS).

The TXRF spectrometry technique can be considered a potent competitor to ICP-MS in time-resolved size-segregated elemental analysis of airborne particulate matter. It was shown51 to be more powerful than the conventionally used ICP-MS approach for samples collected on cascade impactors at high frequency. A major challenge, however, is traceable calibration based on RMs. This is because the aerosols are deposited in variable patterns on the impaction plates and any reference samples should ideally mimic these deposits in spatial distribution and elemental composition. Vigna et al.52 produced a flexible, reusable and low-cost perylene C shadow mask by photolithography. This mask was placed onto the acrylic carriers and Ti applied using an e-beam evaporator. Removal of the mask resulted in an arrangement of thin circular dots. The highly flexible patterning of the reflectors, made possible by the reusable micro stencils, facilitated the calibration for many kinds of impactors. The aspect ratios of the dots were repeated with an error of less than 4%. Although the sampling of aerosols directly onto carriers used subsequently for analysis offered the best LODs, most field campaigns collect aerosols on filters so the analysis of filters provides much wider scope for comparison between different sampling campaigns. Reference PTFE filters loaded with Pb at concentrations of 0.028 to 10.169 μg cm−2 were sandwiched53 between two 75 μm-thick polypropylene films for analysis under GI conditions. The linear range was 0.028 to 4.239 μg cm−2 and the RSDs <10%.

The optimal dimensions of a TXRF spectrometry specimen were evaluated by Tsuji et al.54 Mask vacuum evaporation was used to prepare a thin circular gold layer at specific positions on a flat glass substrate. By visualising the Au intensity over the glass substrate, it was concluded that a diameter of 5 mm could be detected. An optimal specimen height of <10 μm was found by adjusting the height of the gold layer by inserting polyimide thin films between the gold layer and the glass substrate.

Chemometry is being more widely used to assess information hidden within the TXRF spectrometry spectrum. Duarte et al.55 used variable importance in projection PLS-DA to obtain an accurate apportionment of provenance for various seafoods. Overall, the accuracy of classification was ca. 80%. A similar approach was successfully applied by Allegretta et al.56 to identifying the origin of bean seeds. No clear information could be extracted when only quantitative data were used but applying PCA to the quantitative data differentiated the two groups of samples based on where they were grown. An even better classification was achieved by using the raw TXRF spectrometry spectra in a fingerprinting approach when a PLS-DA classification model coupled with a GLSW filter was applied. By preprocessing the spectra with standard normal variate analysis, it was possible to identify the regions of the TXRF spectrometry spectrum which characterised the beans according to their origin. Elimination of the quantification step reduced errors due to the concentration estimation and speeded up the analysis. The microanalytical capabilities of TXRF spectrometry used with clustering and Pearson correlation facilitated57 a classification of individual kidney stones. This is important for understanding better the genesis of these agglomerates which can cause severe health problems.

A review by Revenko et al.58 discussed comprehensively the characterisation of thin films and surface coatings by XRF and TXRF spectrometries using mainly scattered radiation. The review included a discussion of a round-robin on the determination of the thickness of a multi-layer Au/Ni/Cu sample for which the SD of the reported thickness was 4.3–6.6%. Elemental concentrations derived from the measured thickness agreed well with the chemical analysis.

Marchi et al.59 used molecular dynamic simulations of reference-free GI-XRF spectrometry and XPS data in a study of self-assembled 7-mercapto-4-methylcoumarins on flat gold surfaces. It was possible to assess at a submolecular level the presence or absence of hydrogen from the S–H bond at the interface of gold and self-assembled monolayers. The MD with a specifically tailored force field simulated either thiol (S–H) or thiyl (S-) to determine the maximum dynamically stable densities. The GI-XRF spectrometry analysis provided an absolute quantification of the number of sulfur atoms in a dense self-assembled monolayer. The XPS spectra were fitted by using literature data on the binding energy of free and adsorbed thiols and thiyls. The results strongly supported chemisorption of the thiyls with S directly on the gold surface.

The combination of GI-XRF spectrometry and XRR with GI-XRD applied to Au/Cr bilayers showed60 that relatively large residual stress existed in the gold layer when the chromium layer was small. The authors concluded that mechanically stable Au/Cr bilayer systems can be produced if the thickness of the chromium binder layer was about 100 Å. It was clearly shown that the relatively greater thickness of the chromium-binder layer offered lower inter-mixing and in-plane residual stress for the top gold layer. Determining the actual thickness of gold and chromium layers will be extremely helpful for the development of an X-ray optical system with a gold layer as a reflecting medium. A combination of TXRF spectrometry and GI-XRD data together with data for surface pressure and molecular area made61 it possible to study the continuous ionisation of cardiolipin in a monolayer over a wide pH range. Cardiolipin is a unique phospholipid featuring two phosphate groups exclusively present in the plasma membrane of many bacteria and in mitochondrial and chloroplast inner membranes.

Lamellar Si3N4 nanostructured gratings with SiO2 surface layers were characterised62 by scan-free grazing-exit soft XRF spectrometry. A CMOS camera recorded simultaneously both the angle-dependent signals over a broad range of angles and an energy-dispersive spectrum at each angle. Experiments were carried out at the UE56-2 PGM-2 beamline at the BESSY II synchrotron facility. The GE-XRF spectrometry intensity map of the O Kα spectral line matched the simulation well, other than at ca. 1.2°, where the local maximum in the measurement was slightly less intense than the simulated values. Using a laser-produced plasma source, the intensity profile of the O Kα spectral line over the exit angle had the same trend but was several times noisier. A similar scan-free approach was applied by Skroblin et al.63 who compared GE-XRF spectrometry and GI-SAXS for the characterisation of periodic TiO2 nanostructures fabricated by a self-aligned double-patterning process. Both the angle-resolved fluorescence and the scattering intensities were calculated by modelling the X-ray standing-wave-field both close to and within the nanostructures. It was concluded that, in some cases, it was necessary to take into account the influences of roughness and imperfections to obtain a good match with the GE-XRF spectrometry measurements.

5. Hand-held and mobile XRF spectrometry techniques and planetary exploration

5.1. Hand-held and mobile XRF spectrometry techniques

Hand-held XRF spectrometry instrumentation has become widely used in a broad range of applications for problem solving. Potts and Sargent64 published a tutorial review for users of hand-held XRF spectrometers who did not necessarily have extensive training in XRF spectrometry methodology. A brief overview of the technique and its development was provided together with guidance and insights for promoting the reliable interpretation of hand-held XRF spectrometry data obtained in situ. In particular, an understanding of the depth of the XRF spectrometry excited volume, of the interpretation of data close to LODs and of factors that could affect the uncertainty budget was considered to be important for avoiding errors in the use of XRF spectrometry data. Lopez-Nunez65 reviewed studies on the characterisation of organic amendments and residues in which portable XRF spectrometry was used. Whereas elements such as Ca, Fe, K, Mn, Pb and Zn could be measured correctly, that was not the case for elements such as As, Cr and Ni. The reasons proposed for explaining this poor performance were: the inadequacy of comparing data obtained for an aqua-regia-soluble fraction with the total concentration as determined by XRF spectrometry; the poor understanding of the interfering effects of OM; the effect of sample moisture on the XRF spectrometry signals; and the need for a standardised measurement protocol. However, the speed and low cost of the procedure were nevertheless considered to be valuable advantages so the technique was seen to have considerable potential.

The utility of XRF spectrometry as a tool for monitoring compliance with regulatory limits for concentrations of halogenated flame retardants in waste polymers is well known. A critical review66 on the current knowledge about the presence of chlorinated-organophosphate and brominated flame retardants (Cl-OPFRs and BFRs) in the environment placed particular emphasis on the potential of XRF spectrometry for assessing the impact of increased recycling of consumer plastics. However, the review also identified the need for additional research to improve the verification of the presence of Cl-OPFRs.

Frahm et al.67 developed open-source calibration materials intended to aid the calibration of portable XRF instruments for archaeological ceramics. These standards consisted of well-characterised historical brick and geological specimens known as the BRICC (Bricks and Rocks for Instruments' Ceramic Calibration) sets. Each of the ten matched sets consisted of 12 brick and 8 geological specimens mounted in epoxy discs. The USGS CRM SBC-1 (Brush Creek Shale) was included with each set to assess accuracy and a high-purity silica blank was included to check for spectral interferences or other calibration issues. Information provided for the historical bricks and geological specimens included that on their origins and on the data used to derive the recommended values.

Portable XRF spectrometry was used68 both for determining macro and micronutrient compositions in fertilisers and for identifying possible contamination with trace elements. Empirical calibrations for 20 elements were constructed using 39 fertiliser samples. A further 17 fertiliser samples were used for validation. Samples were analysed as loose powders (<75 μm) and scanned for a maximum of 90 s. For the elements Ca, Fe, K, Mg, Mn, Mo, P, S and Zn the R2 values of the calibration models were ≥0.97 whereas those for the trace elements As, Cd, Co, Ni, Pb and Se were ≥0.80. The validation statistics were good but, as expected, dependent on concentration. Up to 1000 mg kg−1, the R2 values were in the range 0.78–0.99 (except for Fe for which R2 was 0.55) but at low concentrations (<20 mg kg−1) the R2 values were generally worse (especially for Cd) and ranged from 0.10–0.99. Nevertheless, the detailed data presented showed that portable XRF spectrometry could be used with high accuracy and precision to measure the major elements Ca and P, the micronutrients Cu and Mn and the trace elements As, Cr and Ni in fertilisers.

In the portable-XRF-spectrometry quantification of As, Cu, Fe, and Zn in organic materials (fungi, vegetation, and animal tissues), Zhou et al.69 used a derived mass-correction model to correct for the matrix effect observed in the analysis of intermediate-thickness samples for which not all the X-rays are absorbed but some pass through completely. The R2 values of the regression models were 0.878, 0.774, 0.802, and 0.901 for As, Cu, Fe and Zn, respectively. Infinitely-thick samples are defined as being thick enough for all the X-rays to be absorbed. The concentrations for such samples as determined by XRF spectrometry were linearly correlated with those determined by ICP-MS. The R2 values of the SLR model were 0.925, 0.960, 0.922 and 0.940 for As, Cu, Fe and Zn, respectively. Combining the mass-correction model and the SLR calibration model for calibration of the portable-XRF-spectrometry system resulted in relative errors of <22, <30, <34 and <18% for the determination As, Cu, Fe and Zn concentrations, respectively. The authors demonstrated that, for quantification of elements in intermediate-thickness samples, mass per unit area (surface density) should be considered rather than the sample thickness because sample density might not be uniform across organic matrices.

The elemental characterisation of air particulate matter samples by XRF spectrometry is widely used. In an interesting contribution, Chatoutsidou et al.70 presented methodologies for the optimisation and calibration of a hand-held XRF spectrometer and their subsequent application to the elemental quantification of unknown particulate-matter samples. Following investigation of the elemental sensitivities and LODs at various excitation conditions (voltage, filter), optimum operating conditions were established for five elemental ranges: Z = 11–12, 12 < Z < 17, 16 < Z < 23, 22 < Z < 31 and 30 < Z < 92. Subsequently, a number of RMs (both multi- and single-element) were used to obtain calibration curves for 24 elements (Al, As, Br, Ca, Cl, Co, Cr, Cu, Fe, K, Mg, Mn, Na, Ni, P, Pb, Rb, S, Se, Si, Sr, Ti, V and Zn). Application of weighted least-square regression analysis to fit the experimentally measured intensities with mass loadings gave high correlations (Pearson r > 0.98) and low statistical error for most of the elements. Elemental concentrations for 28 unknown particulate matter samples obtained using the hand-held instrument were in good agreement with those obtained using a benchtop XRF spectrometer. The difference between the two results was <40% for 14 of the 19 elements identified and, additionally, there was a strong correlation (r = 0.88–0.99) at concentrations higher than the LOQ.

The benefits and limits of portable XRF spectrometry for the characterisation and management of mining wastes was demonstrated by Barago et al.71 who studied samples from two decommissioned mining sites and classified the results obtained according to the data quality criteria established by the US-EPA. Air-dried samples were prepared either by sieving to <2 mm or by milling to a fine powder. The overall analytical data quality was influenced mainly by sample composition (due to interferences, matrix effects and low concentrations) rather than the heterogeneity of the sample. Milling the sample did not produce major variations in the overall data quality. The data for Pb, Sb, and Zn achieved the definitive US-EPA data quality criteria in at least one of the two datasets. Portable XRF spectrometry was shown to be useful for a preliminary quantification of elements in contaminated solid matrices as it did not require complex sample preparation and results were obtained relatively quickly. However, the study also showed that analytical results obtained using portable XRF spectrometry should not be accepted without careful evaluation and benefit from a well-defined QC protocol.

An innovative study was presented by Kim and Choi72 who developed an app for use with smart glasses to support workers performing soil contamination surveys in the field. A field worker wearing the smart glasses could move to soil sampling points while checking the satellite image, survey plan and real-time locations of other field workers. At each sampling point, the worker could use both hands to collect and pretreat soil samples and then measure the content of elements using a portable XRF spectrometry analyser. The results could be entered into the app using a wearable keyboard and shared in real-time with other field workers. A study at the Ilgwang mine in Korea revealed that the app supported field workers effectively and shortened the working time when compared to a study carried out previously at the same site without the use of smart glasses.

The feasibility of using a 109Cd-based portable XRF spectrometer to measure iron concentrations in skin from anaemic and beta-thalassaemic patients was assessed by Bangash et al.73 Their system used a SDD (resolution ≤136 eV) to measure the low-energy Fe Kα-lines. The system was calibrated using 3D-printed polylactic-acid phantoms filled with solutions of Fe at various concentrations (0–150 ppm). Normalisation of the Fe X-ray signals to a Ni X-ray signal improved the system's reproducibility. Use of the 3D phantoms and normalisation of the Fe signal gave a linear calibration (p < 0.001 and R2 > 0.999). The LOD of 1.35 ± 0.35 ppm achieved with a low radiation dose of 1.1 mSv to the skin surface and a real-time measurement of 1800 s was sufficient for this application.

5.2. Planetary exploration

The optical calibration of the Planetary Instrument for X-ray Lithochemistry (PIXL), a μXRF spectrometer mounted on the robotic arm of NASA's Perseverance rover, was described74 in an extended and valuable contribution. The system used a narrow 120 μm X-ray beam to scan target surfaces with high spatial resolution and thereby analyse the elemental chemistry of rocks and soils in detail. The PIXL subsystem (optical fiducial system, OFS) consisted of a micro context camera (MCC), two structured light illuminators and a floodlight illuminator and was used to correlate the X-ray measurements to the texture and structure of the sample derived visually. This revealed the distributions and variations of chemical elements within the rock. The preflight calibration of the OFS included optical calibration of the MCC, radiometric calibration of the floodlight system and geometric calibration of the structured illumination-beam together with an overall geometric calibration of the OFS and the X-ray beam. Results from the performance verification were presented. A software package, PIQUANT, developed by Heirwegh et al.75 for analysis of μXRF spectrometry data returned from PIXL was capable of supporting the quantitative elemental analyses of whole rock and geological materials. The PIQUANT software, which used an iterative fundamental parameters physics-based model to convert X-ray peak intensity into elemental concentration, had minimal reliance on calibration using standards. Correction for polycapillary optic transmission was incorporated to account for photon passage in the X-ray optics of the μXRF spectrometry system. The PIQUANT software architecture was discussed in detail and the performance was demonstrated using data recorded recently on a laboratory-based PIXL instrument which was used to emulate the geometry and functionality of the PIXL instrument on the rover.

6. Cultural heritage applications

The use of macroXRF spectrometry, always in combination with non-invasive imaging techniques such as μXRF spectrometry, IR reflectography and digital microscopy, has become a mature technique and therefore only novel applications are selected for discussion in this section. An interesting review worth reading was published by Silveira and Falcade76 who highlighted the principles, advantages and limitations of EDXRF spectrometry when applied to metallic cultural heritage samples.

Obtaining stratigraphical information about the distribution of different organic and inorganic components without damaging the sample is always a major challenge. Catelli et al.77 employed both a state-of-the-art analytical set-up, able to record VNIR, SWIR and XRF spectrometry data simultaneously, and innovative multivariate and multiblock high-throughput data processing for the analysis of multi-layered paintings. Elemental and molecular information could be obtained for both surface and subsurface layers across the investigated area. The chemometric strategy was highly efficient for the extraction and integration of the most useful information provided by the three different spectroscopies. In particular, multivariate exploratory analysis led to the identification and mapping of composition variability. Evaluation of the within- and between-block correlations revealed the relationship among pigments and binders. It was therefore possible to propose robust hypotheses for paint stratigraphy and execution techniques.

Gerodimos et al.78 presented two approaches (a spectroscopic approach and an exploratory data analysis approach) for the processing of the large number of spectra collected during macroXRF spectrometry analysis. The potentials of the applied methods were demonstrated by analysis of a notable 18th century Greek religious-panel-painting. The spectroscopic approach in which each one of the measured spectra was analysed separately led to the construction of single-element spatial distribution images. The statistical data analysis approach used a k-means algorithm to cluster all the spectra. Subsequently, dimensionality reduction algorithms (PCA, t-SNE) were used to reduce the thousands of channels of XRF spectrometry spectra to an easily perceived dataset of two-dimensional images. The two analytical approaches allowed the extraction of detailed information about the pigments used and paint layer stratigraphy and provided information on the state of preservation and potential restoration.

Cardinali et al.79 used an iterative and non-linear Fuzzy C-Means (FCM)-based analysis of the macroXRF spectrometry datacube to reveal details of the complex painting technique used by Edvard Munch for the famous painting The Scream. The dominating signal was eliminated after each FCM phase to classify better the remaining spectral contribution. The first stage of the FCM-based analysis led to classification of the element distributions into four main groups: three dominated by Cr, Hg and Zn and their corresponding variations in intensity and one delineating areas with low-Z elements, thin brushstrokes and/or contribution of the painting support. The second stage of the analysis revealed the combination and distribution of pigments used to create special shades. This was the case for ZnO and ultramarine, viridian and Thénard's blue paints in the sky section and for the Cd-pigments used for the face of the principal character and for the deep red-orange regions of the sky. The FCM technique was particularly effective for highlighting the distribution of the main pigments and, in some cases, their correlation with minor elements. This double-level approach therefore holds promise.

Machine learning, artificial intelligence (AI) and, more specifically, machine and deep learning have been applied to cultural heritage data sets to identify patterns automatically. Vermeulen et al.80 used the Julia programming language in an open-access, machine-learning approach. This gave faster data processing than when Python and R were used. A dictionary, constructed from the input data, took into account the whole spectrum and a number of experimental factors (compositional variations, paint thicknesses, attenuation of the X-ray signal by material layering, counting statistics and detector noise). Whereas identification of smalt, a pigment present in an 18th-century Mexican painting, by the elemental map approach required the creation of the various elemental maps and their intercomparison, the machine-learning approach created a single distribution map of the co-localised elements identified through an associated XRF spectrometry spectrum. As a result, no image manipulations were required to create the commonly used RGB composite images to highlight the co-presence of elements. The dictionary learning approach also identified pigment mixtures and extracted more information about previous conservation treatment than the classical elemental map approach. Jones et al.81 trained a convolutional neural network (CNN) using the fundamental parameter method to generate a synthetic dataset of XRF spectrometry spectra representative of pigments typically encountered in Renaissance paintings. The creation of the synthetic dataset and the training and fine-tuning of the CNN were described. Application of the classification model to a real XRF spectrometry dataset confirmed that the synthetic spectra, modelled as single layers of individual pigments, had characteristic element lines closely matching those found in real XRF spectrometry spectra. However, the method did not incorporate effects from the X-ray source so the synthetic spectra lacked the continuum and Rayleigh and Compton scatter peaks. When this initial network was based solely on synthetic spectra, the accuracy obtained for RMs was only 55% but application of transfer learning using a small quantity of real XRF spectrometry data improved the accuracy to 96%. Preliminary testing on spectra from the macroXRF spectrometry scanning of a historical painting was encouraging with a considerably reduced spectral signal and correct classification of four of the five spectra.

An open-source statistical tool for lithic sourcing, named SourceXplorer, was presented by McMillan et al.82 The application, built in the open-source R programming environment, facilitated holistic and step-wise multivariate investigations of the elemental relationships between archaeological lithics and potential geological sources. The characteristics of unknowns (e.g. artifacts) and natural sources (regardless of the material type or geographical context) could be compared qualitatively and quantitatively both with traditional low-dimensional scatterplots and with more sophisticated multivariate approaches (e.g. LDA and PCA). Users also had access to a guided interpretation of sourcing results via a series of post hoc tests based on variables produced by both LDA and PCA. The SourceXplorer's functionality was demonstrated by examining evidence for the procurement and use of lithic material from previously undocumented toolstone sources in southwestern British Columbia, Canada.

Further developments of a versatile macroXRF spectrometry scanner allowed83the simultaneous recording of photoluminescence (PL), reflectance image spectroscopy (RIS) and XRF spectrometry spectra on the same spot and delivered intrinsically aligned XRF spectrometry/PL/RIS maps. The five spectral datasets (XRF spectrometry; PL under 250 nm, 365 nm and 655 nm excitation; RIS at 400–1000 nm) were combined using open-source programs so that all the data could be correlated in one operation. A Fayum portrait from the collection of the Egyptian antiquities Department of the Louvre Museum was studied by scanning a 350 × 160 mm2 area in two passes with a resolution of 500 μm, a dwell time of 160 ms and a total scanning time of 10 h. This produced five datacubes of 700 × 320 pixels. The macroXRF spectrometry scanning highlighted the presence of lead white and, probably, green earth. Details such as the Greek madder lake or Egyptian blue signature on the portrait were only poorly revealed using UV photography alone but were clearly visible in the PL images. The study gave a better assessment of the undocumented 1950s restoration campaign, providing information about the pigments used by paint conservators at that time. The compact PL-RIS optical system did not limit operation of the macroXRF spectrometer and although the spatial resolution (100–500 μm) was only moderate this was considered to be a cost-effective alternative to hyperspectral imaging cameras.

In the paintings of the Old Masters, Sr is associated with gypsum ground panels and therefore not considered as important. However, an in-depth study of the distribution of Sr beneath the surface of Raphael's Entombment in the Borghese Gallery in Rome (Italy) by means of macroXRF spectrometry provided84 detailed information on a reapplication of the preparatory gypsum ground layers. The whole painting (174.5 × 178.5 cm2) was scanned using the mapping conditions: 1 mm collimator aperture; pixel size of 1 × 1 mm2 (medium setting); horizontal motor speed of 42 mm s−1 (maximum speed); acquisition time of 20 ms for each spectrum; and a tube setting of 50 kV and 200 μA. Twenty-one different maps (3.2 million pixels) were acquired in a total measurement time of about 18.5 h. Elevated Sr concentrations in the central area of the painting were attributed to substantial differences in the concentration of Sr within the gypsum ore and to two different and separate applications of gypsum over time. Furthermore, careful consideration of the shield effect and comparison of the macroXRF spectrometry map for Sr with the corresponding map for Pb and with radiographic images made it possible to establish when during the execution of the painting Raphael decided to eliminate a particular figure.

Angle-resolved XRF (AR-XRF) spectrometry can be used to sample at different angles of detection or irradiation. Orsilli et al.85 applied the technique to the analysis of gilded samples. As the intensity of a certain element depends on its position inside the sample, on the sample structure and on the geometry used during the analysis, a change in the geometry could be used to retrieve information on the sample structure and to measure the mass thickness (mass per unit area) of bilayer samples. To obtain the AR-XRF spectrometry profiles, the spectra were collected from −5° to +45° at 0.4° intervals with a dwell time of 2 s (a total of 126 spectra). For each sample, three angular scans were performed in different spots. An algorithm for the analysis of AR-XRF spectrometry data considered data pretreatment and exploited the fundamental parameter method to retrieve the fitting functions of the profiles. The thickness of the top-layer was calculated using both the layer's self-attenuation and the attenuation of the bulk signal. The two sets of results were in agreement, both having a RSD for the estimated thickness of <20%.

The elemental composition of gold jewels produced by the Castellani, an important family of goldsmiths in 19th century Europe, was characterised86 using a portable μXRF spectrometer specifically developed for jewellery analysis. The μXRF spectrometry system was equipped with polycapillary optics and focussed the primary X-ray beam down to 30 μm. The jewel under analysis was placed on a horizontal motorised stage with a 1 μm full-step resolution. The new development involved the addition of a second X-ray detector, positioned at a 45° take-off angle and fitted with a 20 μm zinc filter, to check for the presence of low amounts of Cd (±0.3%), a metal added to gold soldering only from the 19th century onwards. The zinc filter removed the overlapping AuLα and AuLβ peaks so the use of cadmium-containing solders in the analysed jewel could be assessed. The relative intensities of the Au X-ray lines were studied to check non-invasively for the presence of surface enrichments of gold metal.

7 Abbreviations

2Dtwo dimensional
3Dthree dimensional
AIartificial intelligence
APSAdvanced Photon Source
ARangle resolved
ASUAtomic Spectrometry Update
BFRbrominated flame retardant
BNLBrookhaven National Laboratory
BRICCBricks and Rocks for Instruments' Ceramic Calibration
CMOScomplementary metal oxide semiconductor
Cl-OPFRchlorinated-organophosphate flame retardant
CNNconvolutional neural network
CRMcertified reference material
CTcomputed tomography
DAdiscriminant analysis
EDSenergy dispersive spectrometry
EDXenergy dispersive X-ray
EDXRFenergy dispersive X-ray fluorescence
EELSelectron energy loss spectrometry
EPMAelectron probe microanalysis
ESRFEuropean Synchrotron Radiation Facility
FCMFuzzy C-Means
FIBfocussed ion beam
FTIRFourier transform infrared
FWHMfull width at half maximum
GEgrazing exit
GIgrazing incidence
GLSWgeneralised least square weighting
HERFDhigh energy resolution fluorescence detection
HPLChigh performance liquid chromatography
HXNHard X-ray Nanoprobe
ICPinductively coupled plasma
ISinternal standard
LAlaser ablation
LDAlinear discriminant analysis
LIBSlaser induced breakdown spectrometry
LODlimit of detection
LOQlimit of quantification
MCCmicro context camera
MDmolecular dynamics
MOXmixed oxide
MSmass spectrometry
NASANational Aeronautics and Space Administration
NISTNational Institute of Standards and Technology
NPnanoparticle
NSLSNational Synchrotron Light Source
OESoptical emission spectrometry
OFSoptical fiducial system
OMorganic matter
PCAprincipal component analysis
PIXEparticle-induced X-ray emission
PIXLPlanetary Instrument for X-ray Lithochemistry
PLphotoluminescence
PLSpartial least squares
PMMApolymethylmethacrylate
PTFEpoly(tetrafluoroethylene)
QCquality control
REErare earth element
RGBred green blue
RISreflectance image spectroscopy
RMreference material
RMSEroot mean square error
ROBLRossendorf beamline
ROIregion of interest
RSDrelative standard deviation
SAXSsmall angle X-ray scattering
SDstandard deviation
SDDsilicon drift detector
SEMscanning electron microscopy
SESAMESynchrotron-light for Experimental Science and Applications in the Middle East
SIMSsecondary ion mass spectrometry
SLRsimple linear regression
t-SNEt-distributed stochastic neighbour embedding
SPMEsolid phase microextraction
SRsynchrotron radiation
SRMstandard reference material
SSRLStanford Synchrotron Radiation Lightsource
STEMscanning transmission electron microscope
STXMscanning transmission X-ray microscopy
SWIRshortwave infrared reflectance
TXRFtotal reflection X-ray fluorescence
US-EPAUnited States Environmental Protection Agency
USGSUnited States Geological Survey
UVultraviolet
VNIRvisible and near infrared
WDSwavelength dispersive spectrometry
WDXRFwavelength dispersive X-ray fluorescence
XAFSX-ray absorption fine structure
XANESX-ray absorption near edge structure
XASX-ray absorption spectroscopy
XPSX-ray photoelectron spectroscopy
XRDX-ray diffraction
XRFX-ray fluorescence
XRRX-ray reflectometry
Z atomic number

Conflicts of interest

There are no conflicts of interest to declare.

References

  1. J. R. Bacon, O. T. Butler, W. R. L. Cairns, O. Cavoura, J. M. Cook, C. M. Davidson and R. Mertz-Kraus, J. Anal. At. Spectrom., 2023, 38(1), 10–56 RSC.
  2. M. Patriarca, N. Barlow, A. Cross, S. Hill, A. Robson, A. Taylor and J. Tyson, J. Anal. At. Spectrom., 2023, 38(3), 496–577 RSC.
  3. E. H. Evans, J. Pisonero, C. M. M. Smith and R. N. Taylor, J. Anal. At. Spectrom., 2022, 37(5), 942–965 RSC.
  4. R. Clough, C. F. Harrington, S. J. Hill, Y. Madrid and J. F. Tyson, J. Anal. At. Spectrom., 2022, 37(7), 1387–1430 RSC.
  5. S. Carter, R. Clough, A. Fisher, B. Gibson and B. Russell, J. Anal. At. Spectrom., 2022, 37(11), 2207–2281 RSC.
  6. Y. Q. Luo, T. Paunesku, O. Antipova, Y. Z. Liu, N. J. Zaluzec, Z. C. Di, G. Woloschak and S. Chen, Metallomics, 2022, 14(9) DOI:10.1093/mtomcs/mfac1025.
  7. M. Y. Ge, X. J. Huang, H. F. Yan, D. Gursoy, Y. Q. Meng, J. Y. Zhang, S. Ghose, W. K. S. Chiu, K. S. Brinkman and Y. S. Chu, Commun. Mater., 2022, 3(1), 37 CrossRef CAS.
  8. A. S. Schulz, C. A. M. Harteveld, G. J. Vancso, J. Huskens, P. Cloetens and W. L. Vos, ACS Nano, 2022, 16(3), 3674–3683 CrossRef CAS PubMed.
  9. E. De Pauw, B. J. Tkalcec, P. Tack, B. Vekemans, M. Di Michiel, F. E. Brenker and L. Vincze, Spectrochim. Acta, Part B, 2022, 188, 106346 CrossRef CAS.
  10. P. Tack, E. De Pauw, B. Tkalcec, M. Lindner, B. Bazi, B. Vekemans, F. Brenker, M. Di Michiel, M. Uesugi, H. Yurimoto, T. Nakamura, K. Amano, M. Matsumoto, Y. Fujioka, Y. Enokido, D. Nakashima, T. Noguchi, R. Okazaki, H. Yabuta, H. Naraoka, K. Sakamoto, S. Tachibana, T. Yada, M. Nishimura, T. Saiki, S. Tanaka, F. Terui, S. Nakazawa, S. L. Watanabe, Y. Tsuda and L. Vincze, Earth, Planets Space, 2022, 74(1), 146 CrossRef PubMed.
  11. E. Andresen, I. J. Flores-Sanchez, D. Bruckner, S. N. H. Bokhari, G. Falkenberg and H. Kupper, J. Hazard. Mater., 2023, 442, 130062 CrossRef CAS PubMed.
  12. L. Li, S. Y. Zhang, W. L. Zhang and H. B. Lu, Phys. Med. Biol., 2023, 68(3), 035020 CrossRef PubMed.
  13. S. Jayarathna, H. Moktan and S. H. Cho, IEEE Access, 2022, 10, 105074–105083 Search PubMed.
  14. F. Förste, L. Bauer, K. Heimler, B. Hansel, C. Vogt, B. Kanngiesser and I. Mantouvalou, J. Anal. At. Spectrom., 2022, 37(8), 1687–1695 RSC.
  15. K. Limburska and T. Trojek, Radiat. Prot. Dosim., 2022, 198(9–11), 661–665 CrossRef CAS PubMed.
  16. M. Iro, D. Ingerle, M. Radtke, A. G. Buzanich, P. Kregsamer and C. Streli, J. Synchrotron Radiat., 2022, 29, 1376–1384 CrossRef CAS PubMed.
  17. B. Bazi, P. Tack, M. Lindner, B. Vekemans, E. De Pauw, B. Tkalcec, F. E. Brenker, J. Garrevoet, G. Falkenberg, H. Yabuta, H. Yurimoto, T. Nakamura, K. Amano, M. Matsumoto, Y. Fujioka, Y. Enokido, D. Nakashima, M. Uesugi, H. Naraoka, T. Noguchi, R. Okazaki, K. Sakamoto, T. Yada, M. Nishimura, A. Nakato, A. Miyazaki, K. Yogata, M. Abe, T. Okada, T. Usui, M. Yoshikawa, T. Saiki, S. Tanaka, F. Terui, S. Nakazawa, S. Tachibana, S. Watanabe, Y. Tsuda and L. Vincze, Earth, Planets Space, 2022, 74(1), 161 CrossRef.
  18. P. Ballikaya, I. Brunner, C. Cocozza, D. Grolimund, R. Kaegi, M. E. Murazzi, M. Schaub, L. C. Schonbeck, B. Sinnet and P. Cherubini, Tree Physiol., 2023, 43(2), 262–276 CrossRef PubMed.
  19. J. Tapia, M. Eveno, T. Calligaro, L. Pichon, E. Laval, E. Ravaud and I. Reiche, Eur. Phys. J. Plus, 2023, 138(1), 46 CrossRef.
  20. F. Bock, A. Gruber, K. Leopold and H. Bruhn, J. Anal. At. Spectrom., 2023, 38, 1021–1031 RSC.
  21. A. Rogoll, K. Heimler, D. Friebel, E. Brendler and C. Vogt, J. Anal. At. Spectrom., 2022, 37(12), 2647–2661 RSC.
  22. C. Fabre, K. Trebus, A. Tarantola, J. Cauzid, V. Motto-Ros and P. Voudouris, Spectrochim. Acta, Part B, 2022, 194, 106470 CrossRef CAS.
  23. Y. Liu, C. Zhang, D. Zhang, T. Liu, H. Qiu, Q. L. Li and J. H. Li, At. Spectrosc., 2022, 43(5), 378–387 CAS.
  24. M. Nakae, T. Matsuyama, H. Ishi and K. Tsuji, X-Ray Spectrom., 2023 DOI:10.1002/xrs.3325.
  25. K. Sanyal, B. Kanrar, S. Dhara and R. V. Pai, J. Anal. At. Spectrom., 2022, 37(6), 1179–1185 RSC.
  26. J. M. Dinsley, H. S. Davies, M. A. Gomez-Gonzalez, C. H. Robinson and J. K. Pittman, Ecosphere, 2022, 13(12), e4312 CrossRef.
  27. M. J. Pushie, N. J. Sylvain, H. S. Hou, M. J. Hackett, M. E. Kelly and S. M. Webb, Metallomics, 2022, 14(6) DOI:10.1093/mtomcs/mfac1032.
  28. V. Bonanni and A. Gianoncelli, Int. J. Mol. Sci., 2023, 24(4), 3220 CrossRef CAS PubMed.
  29. M. Harfouche, M. Abdellatief, Y. Momani, A. Abbadi, M. Al Najdawi, M. Al Zoubi, B. Aljamal, S. Matalgah, L. U. Khan, A. Lausi and G. Paolucci, J. Synchrotron Radiat., 2022, 29, 1107–1113 CrossRef CAS PubMed.
  30. N. P. P. Edwards, J. R. R. Bargar, D. van Campen, A. van Veelen, D. Sokaras, U. Bergmann and S. M. M. Webb, Rev. Sci. Instrum., 2022, 93(8), 083101 CrossRef CAS PubMed.
  31. S. Sala, Y. H. Zhang, N. De la Rosa, T. Dreier, M. Kahnt, M. Langer, L. B. Dahlin, M. Bech, P. Villanueva-Perez and S. Kalbfleisch, J. Synchrotron Radiat., 2022, 29, 807–815 CrossRef CAS PubMed.
  32. P. D. Quinn, F. Cacho-Nerin, M. A. Gomez-Gonzalez, J. E. Parker, T. Poon and J. M. Walker, J. Synchrotron Radiat., 2023, 30, 200–207 CrossRef PubMed.
  33. M. A. Z. Chowdhury, K. Ok, Y. Luo, Z. Liu, S. Chen, T. V. O'Halloran, R. Kettimuthu and A. Tekawade, J. Synchrotron Radiat., 2022, 29, 1495–1503 CrossRef CAS PubMed.
  34. W. Chen, G. Giacomini, A. Kuczewski, J. Mead, A. K. Rumaiz and D. P. Siddons, J. Instrum., 2023, 18(1), P01016 CrossRef.
  35. S. Hauser, S. Andres and K. Leopold, Anal. Bioanal. Chem., 2022, 414(15), 4519–4529 CrossRef CAS PubMed.
  36. P. M. Carvalho, E. Margui, A. Kubala-Kukus, D. Banas, J. Machado, D. Casal, D. Pais, J. P. Santos and S. Pessanha, Spectrochim. Acta, Part B, 2022, 198, 106548 CrossRef CAS.
  37. T. Matsuyama, T. Hayakawa, F. Inoue and K. Tsuji, Anal. Sci., 2022, 38(5), 821–824 CrossRef CAS PubMed.
  38. M. Szczerbowska-Boruchowska, A. D. Surowka, B. Ostachowicz, K. Piana, A. Spaleniak, P. Wrobel, J. Dudala and A. Ziomber-Lisiak, Biochim. Biophys. Acta, 2023, 1867(2), 130279 CrossRef CAS PubMed.
  39. M. Nguli, M. J. Gatari, K. Shepherd, L. Njenga and J. Boman, X-Ray Spectrom., 2022, 51(3), 198–203 CrossRef CAS.
  40. E. V. Chuparina, A. S. Maltsev, E. V. Stolpovskaya and N. A. Neverova, Spectrochim. Acta, Part B, 2022, 197, 106542 CrossRef CAS.
  41. G. Mankovskii and A. Pejovic-Milic, Spectrochim. Acta, Part B, 2022, 194, 106466 CrossRef CAS.
  42. P. Zambianchi, G. Hermogenes and J. K. Zambianchi, Radiat. Phys. Chem., 2022, 193, 109937 CrossRef CAS.
  43. E. Margui, I. Queralt, D. Andrey and L. Perring, Food Chem., 2022, 383, 132590 CrossRef CAS PubMed.
  44. T. Matsuyama, Y. Tanaka, M. Nakae, T. Furusato and K. Tsuji, Analyst, 2022, 147(22), 5130–5137 RSC.
  45. K. Kocot, K. Pytlakowska, E. Talik, C. Krafft and R. Sitko, Talanta, 2022, 246, 123501 CrossRef CAS PubMed.
  46. K. Sanyal, S. Dhara, N. Gumber and R. V. Pai, Talanta, 2023, 254, 124129 CrossRef CAS PubMed.
  47. Y. A. Maksimova, A. S. Dubenskiy, A. V. Garmash, G. V. V. Pashkova, I. V. V. Shigapov, I. F. Seregina, L. A. Pavlova, P. Y. Sharanov and M. A. Bolshov, Spectrochim. Acta, Part B, 2022, 196, 106521 CrossRef CAS.
  48. Y. Akahane, S. Nakazawa and S. Kunimura, ISIJ Int., 2022, 62(5), 871–874 CrossRef CAS.
  49. Y. S. Zhang, Y. X. He, W. Q. Zhou, G. Q. Mo, H. Chen and T. Xu, Appl. Spectrosc. Rev., 2022, 428–441 Search PubMed.
  50. G. V. Pashkova, A. N. Zhilicheva, V. M. Chubarov, A. S. Maltsev, N. N. Ukhova, V. A. Pellinen, J. V. Sokolnikova, D. O. Kirsanov, V. V. Panchuk and A. E. Marfin, Spectrochim. Acta, Part B, 2022, 198, 106549 CrossRef CAS.
  51. S. Horender, K. Auderset, P. Quincey, S. Seeger, S. N. Skov, K. Dirscherl, T. O. M. Smith, K. Williams, C. C. Aegerter, D. M. Kalbermatter, F. Gaie-Levrel and K. Vasilatou, Atmos. Meas. Tech., 2021, 14, 1225–1238 CrossRef CAS.
  52. L. Vigna, M. Gottschalk, N. Cacocciola, A. Verna, S. L. Marasso, S. Seeger, C. F. Pirri and M. Cocuzza, Micro Nano Eng., 2022, 14, 100108 CrossRef CAS.
  53. P. Cirelli, F. Bilo, K. Tsuji, T. Matsuyama, G. Siviero, L. Pisani, A. Zacco, L. E. Depero, D. Eichert and L. Borgese, Spectrochim. Acta, Part B, 2022, 192, 106414 CrossRef CAS.
  54. K. Tsuji, N. Taniguchi, H. Yamaguchi and T. Matsuyama, X-Ray Spectrom., 2023 DOI:10.1002/xrs.3335.
  55. B. Duarte, R. Mamede, J. Carreiras, I. A. Duarte, I. Cacador, P. Reis-Santos, R. P. Vasconcelos, C. Gameiro, P. Re, S. E. Tanner and V. F. Fonseca, Foods, 2022, 11(17), 2699 CrossRef PubMed.
  56. I. Allegretta, G. Squeo, C. E. Gattullo, C. Porfido, A. Cicchetti, F. Caponio, S. Cesco, C. Nicoletto and R. Terzano, Food Chem., 2023, 401, 134124 CrossRef CAS PubMed.
  57. A. A. Shaltout, O. H. Abd-Elkader, P. Lassen and U. A. E. Fittschen, X-Ray Spectrom., 2023 DOI:10.1002/xrs.3331.
  58. A. G. Revenko, A. L. Tsvetyansky and A. N. Eritenko, Radiat. Phys. Chem., 2022, 197, 110157 CrossRef CAS.
  59. D. Marchi, E. Cara, F. F. Lupi, P. Honicke, Y. Kayser, B. Beckhof, M. Castellino, P. Klapetek, A. Zoccante, M. Laus and M. Cossi, Phys. Chem. Chem. Phys., 2022, 24(36), 22083–22090 RSC.
  60. M. A. Alam, M. K. Tiwari, A. Khooha, M. Nayak and C. Mukherjee, Surf. Interface Anal., 2022, 54(10), 1032–1040 CrossRef CAS.
  61. M. Sturm, O. Gutowski and G. Brezesinski, ChemPhysChem, 2022, 23(19), e202200218 CrossRef CAS PubMed.
  62. S. Staeck, A. Andrle, P. Honicke, J. Baumann, D. Grotzsch, J. Weser, G. Goetzke, A. Jonas, Y. Kayser, F. Forste, I. Mantouvalou, J. Viefhaus, V. Soltwisch, H. Stiel, B. Beckhoff and B. Kanngiesser, Nanomater, 2022, 12(21), 3766 CrossRef CAS PubMed.
  63. D. Skroblin, A. F. Herrero, T. Siefke, K. Nikolaev, A. Andrle, P. Honicke, Y. Kayser, M. Krumrey, C. Gollwitzer and V. Soltwisch, Nanoscale, 2022, 14(41), 15475–15483 RSC.
  64. P. J. Potts and M. Sargent, J. Anal. At. Spectrom., 2022, 37(10), 1928–1947 RSC.
  65. R. Lopez-Nunez, Appl. Sci., 2022, 12(14), 6944 CrossRef CAS.
  66. M. Alghamdi, M. A. E. Abdallah and S. Harrad, Emerging Contam., 2022, 8, 9–20 CrossRef CAS.
  67. E. Frahm, C. M. Carolus, A. Cameron, J. Berner, H. Brown, J. Cheng, J. Kalodner, J. L. L. Jr, A. Natale, S. Seibert, D. Sparks-Stokes and E. Wuellner, J. Archaeol. Sci.-Rep., 2022, 43, 103443 Search PubMed.
  68. G. E. Acquah, J. Hernandez-Allica, C. L. Thomas, S. J. Dunham, E. K. Towett, L. B. Drake, K. D. Shepherd, S. P. McGrath and S. M. Haefele, PLoS One, 2022, 17(1), e0262460 CrossRef CAS PubMed.
  69. S. B. Zhou, Q. M. Cheng, D. C. Weindorf, B. Y. Yang, Z. X. Yuan and J. Yang, J. Anal. At. Spectrom., 2022, 37(11), 2461–2469 RSC.
  70. S. E. Chatoutsidou, S. Papagiannis, D. F. Anagnostopoulos, K. Eleftheriadis, M. Lazaridis and A. G. Karydas, Spectrochim. Acta, Part B, 2022, 196, 106517 CrossRef CAS.
  71. N. Barago, E. Pavoni, F. Floreani, M. Crosera, G. Adami, D. Lenaz, F. L. Filon and S. Covelli, Appl. Sci., 2022, 12(23), 12189 CrossRef CAS.
  72. D. Kim and Y. Choi, Sustainability, 2022, 14(19), 12370 CrossRef CAS.
  73. S. U. K. Bangash, F. E. McNeill, M. J. Farquharson and D. R. Chettle, Biomed. Phys. Eng. Expr., 2022, 8(6), 065034 CrossRef PubMed.
  74. D. A. Klevang, C. C. Liebe, J. Henneke, J. L. Jorgensen, R. Sharrow, T. Setterfield, L. Wade, M. Sondheim, M. Foote, W. T. Elam, C. M. Heirwegh, J. Hurowitz and A. Allwood, Space Sci. Rev., 2023, 219(1), 11 CrossRef.
  75. C. M. Heirwegh, W. T. Elam, L. P. O'Neil, K. P. Sinclair and A. Das, Spectrochim. Acta, Part B, 2022, 196, 106520 CrossRef CAS.
  76. P. Silveira and T. Falcade, J. Cult. Herit., 2022, 57, 243–255 CrossRef.
  77. E. Catelli, Z. L. Li, G. Sciutto, P. Oliveri, S. Prati, M. Occhipinti, A. Tocchio, R. Alberti, T. Frizzi, C. Malegori and R. Mazzeo, Anal. Chim. Acta, 2023, 1239, 340710 CrossRef CAS PubMed.
  78. T. Gerodimos, A. Asvestas, G. P. Mastrotheodoros, G. Chantas, I. Liougos, A. Likas and D. F. Anagnostopoulos, J. Imaging, 2022, 8(5), 147 CrossRef PubMed.
  79. M. L. Cardinali, F. Albertin, L. Cartechini, I. C. A. Sandu, E. S. Tveit, A. Romani, C. Grazia, R. P. De Freitas, V. Bruni, D. Vitulano, F. Rosi and IEEE, Presented at 12th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS), Rome, Italy, 2022 Search PubMed.
  80. M. Vermeulen, A. McGeachy, B. J. Xu, H. Chopp, A. Katsaggelos, R. Meyers, M. Alfeld and M. Walton, J. Anal. At. Spectrom., 2022, 37(10), 2130–2143 RSC.
  81. C. Jones, N. S. Daly, C. Higgitt and M. R. D. Rodrigues, Heritage Sci., 2022, 10(1), 88 CrossRef CAS.
  82. R. McMillan, N. Waber, M. Ritchie and E. Frahm, J. Archaeol. Sci., 2022, 144, 105626 CrossRef.
  83. R. Moreau, L. Brunel-Duverger, L. Pichon, B. Moignard, D. Gourier and T. Calligaro, Eur. Phys. J. Plus, 2023, 138(1), 16 CrossRef.
  84. R. Alberti, T. Frizzi, M. Gironda, M. Occhipinti, T. Parsani, C. Seccaroni and A. Tati, J. Cult. Herit., 2022, 58, 130–136 CrossRef.
  85. J. Orsilli, A. Migliori, R. Padilla-Alvarez, M. Martini and A. Galli, J. Anal. At. Spectrom., 2023, 38(1), 174–185 RSC.
  86. R. Manca, S. Scrivano, C. Manfriani, F. J. A. Vazquez, I. Ortega-Feliu, M. Ferretti, M. A. Respaldiza and M. Benvenuti, Sci. Rep., 2022, 12(1), 7813 CrossRef CAS PubMed.

Footnote

Joint review coordinators.

This journal is © The Royal Society of Chemistry 2023