Atomic spectrometry update: 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 spectrometry

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

• 6 Cultural heritage applications

• Abbreviations

• References

---

Abstract

A highlight this year was the construction of elemental maps of microalgae cells at the organelle level using XRF-CT with a beam size of just 15 nm. These probably represent the highest ever spatial resolution achieved for XRF-CT images. An innovative approach for reconstruction algorithms used a deep convolutional neural network to correct for the self-absorption effects in the XRF-CT sinogram domain. Such corrections will enable the use of conventional tomographic reconstruction algorithms for XRF-CT analysis. A remarkable advantage of the proposed method was that it could correct for self-absorption effects without any prior knowledge of the scanning setups. A particularly innovative dual-energy X-ray-beam ptycho-fluorescence imaging method combined XRF spectrometry and X-ray ptychography scanning with simultaneous data-collection. To overcome the incompatibility of the different excitation and scanning conditions required for fast ptychography and high spatial resolution XRF spectrometry, two coaxial beams of different sizes were used on the sample simultaneously. In comparison to sequential scans, this combined approach has the potential to reduce the data collection time by a factor of 25. The very good performance of laboratory 2D-XRF spectrometry for the analysis of minute amounts of radioactive material has made it a potential alternative to TXRF spectrometry. Reconstruction algorithms are now capable of significantly enhancing imaging resolution in μXRF spectrometry. The forensic classification of incredibly small amounts of samples such as single fibres was accomplished using elemental profiles determined by TXRF spectrometry. Although characterisation of the low-Z matrix in milk and polymers by conventional TXRF spectrometry is challenging, it was successfully achieved by studying the Compton and Rayleigh scatter with both univariate and multivariate approaches. Portable XRF spectrometry is a well-established technique for a wide range of applications. Emphasis continues to be placed on improving calibration and reducing matrix and inter-element effects. The sample type remains a challenge for portable XRF analysis. Computer vision and statistical methods such as spectral angle mapper are being increasingly applied to maximise the information that can be extracted from the data. For example, an innovative approach exploited the elemental features from XRF spectra to improve the interpretation of molecular information provided by reflectance spectroscopy, and vice versa.

---

1 Introduction

This review describes advances in the XRF spectrometry group of techniques published approximately between April 2021 and March 2022. 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 ASU reviews 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 the ASU reviews 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 techniques

Three-dimensional (3D) chemical imaging using XRF techniques continues to evolve both for experimental strategies and for (quantitative) data evaluation and reconstruction. The wide range of research areas to which 3D-XRF spectrometry is applied includes material science, environmental and earth sciences, life sciences and biomedical imaging. This section highlights the contribution that X-ray-fluorescence computed-tomography (XRF-CT) and confocal XRF spectrometry can make to cutting-edge research in these subject areas.

2.1.1 X-ray fluorescence computed tomography (XRF-CT). Cross-sectional-mapping and fully 3D-elemental-imaging by XRF-CT can still be considered a novelty, used mainly at synchrotron micro- and nano-beam facilities for the analysis of μm- to mm-size samples. The 3rd and the newly developed 4th generation sources facilitate these applications at sub-μm spatial resolution and are typically used in combination with complementary imaging techniques, such as ptychography, absorption and phase-contrast CT, XRD-CT and 2D- and 3D-XAS.

In the field of cancer research, Chen et al.⁶ developed a multi-scale XRF-CT method for the examination of nanocomposite-treated biological samples to study interactions between NPs and cells. Flow cytometry was used to evaluate changes in cell cycles associated with non-targeted nanocomposite uptake by individual cells. Changes in cell populations were studied using the cervical-cancer cell-line HeLa (CCL-2 ATCC, Manassas, Virginia, USA). Single cell and cell population changes due to nanocomposite uptake were investigated by XFM techniques which included XRF-CT. Initially, a cluster of cells exposed to dopamine-coated Fe₃O₄ NPs with a TiO₂ shell was imaged with a 600 nm beam in 3D and then regions of interest were scanned at high resolution using XRF-CT with a beam size of 80 nm. This study was the first example of serial tomographic imaging at two different resolutions. The XFM and XRF-CT results indicated that non-targeted dopamine-coated Fe₃O₄/TiO₂ nanocomposites have a high affinity for the BIRC5 protein, a well-known cancer therapeutic target.

Associated with this study, Liu et al.⁷ used cryogenic XRF microscopy and XRF-CT on frozen hydrated neuroblastoma cells (the most common extracranial solid malignancy in children) which had been treated with nanocomposites. The study investigated the use of Fe₃O₄/TiO₂ NPs, either uncoated or coated with metaiodobenzylguanidine (MIBG) or 3,4-dihydroxyphenylacetic acid (DOPAC), as potential radiosensitisers for neuroblastoma. The in vitro cryogenic XRF-CT analysis revealed cytoplasmic as well as nuclear distributions of the nanoconstructs. Whereas the uptake of uncoated and MIBG-coated nanocomposites sensitised neuroblastoma cells to ionising radiation only modestly, cells exposed to DOPAC-coated nanocomposites had their sensitivity to radiation enhanced five-fold. The XRF imaging was carried out using several instruments under different conditions. At the sector 2-ID-D at APS (Illinois, USA), a beam spot size of 300 or 600 nm was used in combination with a cryo-jet for microscale resolution. At the Bionanoprobe (ANL, Illinois, USA) at sector 21 of the LS-CAT, samples were kept in vacuum at liquid nitrogen temperature and an 85 nm beam spot size was used to achieve nanoscale resolution.

The use of nanocomposites was also demonstrated by Saladino et al.,⁸ who developed (MoO₂, Rh, Ru)–SiO₂–Cy5.5 NP-based contrast agents that could be detected with various imaging modalities, including XRF-CT and (optical) confocal microscopy. The contrast agents could be detected without significant self-absorption by using the Mo, Rh and Ru K-lines. This, together with the detectability of these agents by confocal microscopy, provides a promising strategy for non-invasive diagnostics and multimodal bioimaging in cancer research.

Possibly the highest ever spatial resolution XRF-CT images, elemental maps at the organelle level using a beam size down to 15 nm, were obtained at the ESRF (Grenoble, France) ID-16A beamline in studies by Ashraf et al.⁹ The authors investigated the effects of the controlled exposure of Desmodesmus quadricauda, a freshwater green microalgae, to La by treating the algal cultures with nM concentrations of La under environmentally relevant conditions. At sublethal concentrations (128 nM), La accumulated in hotspots inside the cells in the cytoplasm. Lethal concentration levels (1387 nM) led to the release of ions (K, Zn) from the cells through leaky membranes and La filled most of the cells. At moderate concentrations, La had no clear positive effects on the cell growth or photosynthetic parameters but a dramatic decrease in cell counts occurred at higher concentrations.

Höller et al.¹⁰ studied the effects of overexpression of metal tolerance protein 8 (MTP8) on Mn tolerance during imbibition and on Fe and Mn storage in Arabidopsis thaliana seeds. The Fe, Mn and Zn concentrations in MTP8-overexpressing lines were determined by ICP-MS whereas the distributions of these metals in intact seeds were established by synchrotron XRF-CT at beamline P06 (PETRA III) at DESY (Deutsches Elektronen-Synchrotron, Hamburg, Germany). The extremely-high-quality elemental maps were recorded using a 400–500 nm beam at 12 keV excitation energy and a 384-element Maia detector in backscatter geometry. The MTP8 overexpression led to a strongly increased Mn tolerance of seeds during imbibition, excess Mn being loaded effectively into the vacuole. In mature seeds, however, MTP8 overexpression did not cause a consistent increase in Fe and Mn accumulations and it did not change the distribution patterns for these metals.

Kondo et al.¹¹ demonstrated the 3D trace-element imaging of a single hair by synchrotron XRF-CT at the BL05XU beamline of SPring-8 (Hyogo, Japan) with a beam of 2.5 μm (horizontal) × 1.3 μm (vertical) and an excitation energy of 20 keV. The cross-sectional distributions of Br, Cu, Fe and Zn were recorded at various positions-of-interest with a spatial resolution of 2 μm and at concentration levels of 4–20, 10–20, 5–20 and 200–300 ppm for Br, Cu, Fe and Zn, respectively. The method was considered attractive not only for research on dietary habits but also for forensic investigations.

Applications to geo- and planetary sciences included¹² the use of Sr as a proxy for the presence of Ca (and, thus, Ca-bearing components) throughout mm-sized samples of carbonaceous chondritic materials. As expected, the SR-XRF spectrometry and XRF-CT measurements indicated a clear, positive and preferential correlation between Ca and Sr, so Sr, which suffers much less than Ca from self-absorption effects due to its Kα energy of 14.1 keV, could be used as a good indicator for the presence of Ca in depths from which Ca-Kα lines cannot escape. The measurements were performed at the hard X-ray micro/nano-probe beamline P06 end station of PETRA-III at DESY (Hamburg, Germany) using an excitation energy of 19.1 keV, a beamsize of 1.2 (horizontal) × 0.7 (vertical) μm and an intensity of ca. 1.7 × 10¹¹ photons s⁻¹. The non-destructive XRF-CT technique was considered particularly suitable for the initial elemental analysis of unique and pristine material collected from primitive carbonaceous C-type asteroids and returned on missions such as Hayabusa 2 and OSIRIS-REx.

Relevant to the analysis of asteroid samples, Tack et al.¹³ demonstrated the use of non-resonant inelastic X-ray Raman scattering at beamline ID20 of the ESRF to obtain 3D distributions of low-Z elements (Al, Ca, Mg) at high concentrations. The measured 3D elemental-distribution-volumes agreed well with data obtained using either XRF-CT at ESRF ID15A or absorption CT experiments. The proposed approach used an X-ray beam (12.9 keV, 10¹³ photons s⁻¹) focused to a spot of 10 × 10 μm. Seventy two crystal analysers were used to measure the energy loss during inelastic X-ray scattering. This approach extended the information depth from the μm to the mm scale for Al, Ca and Mg, and was considered a valuable tool for the non-destructive investigation of low-Z elemental distributions within mm-sized extraterrestrial materials, especially when used in conjunction with XRF-CT data.

An impressive example of the use of XRF-CT was the work of Perotti et al.,¹⁴ who combined HR ptychographic X-ray nanotomography with multimodal chemical microtomography (XRF spectrometry and XRD analysis) to reveal the early history of forsteritic grains extracted from the matrix of the Murchison CM2.5 chondrite. The μXRF and μXRD tomography data were collected simultaneously at the microXAS beamline (X05LA) of the SLS (Villigen, Switzerland) using an X-ray beam of 17.2 keV to probe all relevant elements (including Fe and Ni) with a spatial resolution of 1 μm. The excellent resolution of 64 nm made it possible, when used together with 3D electron-density-maps obtained from ptychographic X-ray nanotomography, to identify trapped spherical inclusions containing Fe–Ni, but very little silica-rich glass, and void caps. The presence of the voids along with the overall composition and other features was consistent with chondrule-forming conditions and indicated that the grains experienced one or more heating events with peak temperatures close to the melting point of forsterite (ca. 2100 K) before cooling and contracting.

As part of the development of XRF-CT reconstruction algorithms, Gao et al.¹⁵ presented an innovative approach using a deep convolutional neural network to correct for the self-absorption effects in the sinogram domain. It was concluded that a well-trained neural network can correct fluorescence sinograms which exhibit considerable self-absorption effects. Such corrections enable the use of conventional tomographic reconstruction algorithms for XRF-CT analysis. A remarkable advantage of the proposed method was that it could correct for self-absorption effects without any prior knowledge of the scanning setups. The correction algorithm was verified experimentally using laboratory XRF-CT measurements (spatial resolution of ca. 25 μm) made on a test phantom composed of quartz capillaries filled with solutions of RbCl, FeNH₄(SO₄)₂ and K₂Cr₂O₇.

Gao et al.¹⁶ also presented a novel method, feasible in most laboratory setups, for the quantitative reconstruction of trace and low-Z (minor and major) elements determined by polychromatic excitation XRF-CT. By using model samples for experimental verification, uncertainties in the quantitative reconstruction of polychromatic XRF-CT data were determined to be 10–65% for Cr, Fe and Rb. This was deemed acceptable when the uncertainties in, e.g., the spectral distribution of the polychromatic X-ray tube spectrum were taken into account.

2.1.2 Confocal XRF spectrometry. A major improvement of sensitivities achievable in laboratory μXRF spectrometry and for confocal μXRF spectrometers was based¹⁷ on the use of polycapillary lenses specifically adapted to a high-brilliance liquid-metal jet source (LMJS). The LMJS had an effective minimal electron-source spot-size of 15 × 15 μm. The power was restricted to a maximum of 47 W to avoid damaging the source. The smallest X-ray spot diameter on the sample was 22 μm for Fe Kα radiation with a polycapillary lens optimised for spatial resolution. In comparison to a state-of-the-art conventional μXRF instrument, the LMJS source in conjunction with a lens optimised for transmission efficiency increased sensitivity by factors of 50, 20 and 24 for Cu, Sc and V, respectively. Considerable improvements in S/N and LODs were obtained¹⁸ for several Cu, Fe and Si microparticles, deposited on an acrylic plate to model surface contaminants, when confocal detection mode was used instead of conventional μXRF spectrometry. The 14.6, 21.9 and 43.5 fold improvements in S/N for Cu, Fe and Si resulted mainly from reduction of the scattering background which originated from the acrylic plate. The corresponding LOD values improved by 10.8–24.7 fold.

The same authors demonstrated¹⁹ that a new confocal line XRF (C-L-XRF) system could obtain elemental information over a larger area and at a higher intensity than conventional confocal point μXRF analysis. The spatial resolutions of the new setup in the horizontal and depth directions were ca. 2.9- and 2.6-fold larger than those of point focus confocal μXRF spectrometry. The XRF intensity was ca. 33 times higher than that of point focus confocal μXRF spectrometry. Considering the increased sensitivity and the large-area mapping capability, the C-L-XRF technique was expected to be especially important for depth-profiling wide-area samples such as layered structures.

A high-depth-resolution application, the study of mercury adsorption kinetics on sulfurised biochar, involved²⁰ a Ge microchannel array optic unit with etched channels as confocal collimator and a 2 μm primary beam at Beamline Sector 20ID of the Advanced Photon Source (APS, Illinois, USA). A confocal volume of 2 × 2 × 2 μm was achieved with a depth resolution of 2 μm. The authors investigated Hg diffusion (by confocal μXRF spectrometry) and transformation processes (by confocal μXANES) in untreated- and calcium-polysulfide-treated oak biochar. The main focus of the investigation was to evaluate the extraction efficiency of aqua-regia-based digestion procedures and the stability of Hg adsorbed by biochars. The same setup was used²¹ to obtain depth-resolved data for biochar-supported catalysts using a confocal volume of 8 μm³. Nitrate-based (from water-soluble salts) and hydroxide-based (from water-insoluble salts) catalysts of Co and Ni were prepared using wetness-impregnation and aqueous-dispersion methods. Whereas the metals associated with nitrate were dispersed mostly in the pores of the catalysts, those associated with hydroxide were distributed mainly on the surface. The high-resolution confocal μXRF spectrometry data were obtained using incident beam energies of 9200 and 9000 eV for Co and Ni, respectively.

Oguzturk et al.²² reported on the preparation of fumed silica-based anisometric supraparticles with catalytically active Pt-covered magnetite (Fe₃O₄/Pt) patches, suitable for self-propulsion. Use of confocal μXRF spectrometry made it possible to locate the catalytic Fe₃O₄/Pt NPs precisely within the mm-sized supraparticles with a resolution in the μm range and thereby provide detailed internal structural information on various supraparticles. The confocal μXRF studies were performed at TU Berlin using a laboratory setup with a Mo microfocus X-ray tube, two polycapillary lenses and an SDD. Operating conditions were 50 kV and 600 μA, 30 and 40 μm pixel size and 5 to 15 s measuring time per pixel.

In the field of biomedicine, Xu et al.²³ reported the first application of confocal μXRF spectrometry for evaluating hepatic iron deposition in liver fibrosis induced by thioacetamide. The results suggested that intrahepatic iron may be closely related to the pathogenesis of liver fibrosis induced by thioacetamide. The confocal setup used an X-ray tube with a Mo target operated at 20 kV and 0.5 mA. The sizes of the focal spots of the confocal and the primary polycapillary lenses were determined to be 33 and 32.4 μm, respectively, at 17.4 keV (Mo Kα). The samples were scanned using a point-by-point strategy with a dwell time of 300 s and step (pixel) size of 20 μm.

A novel EDXRF device for the simultaneous detection and treatment of cancer, currently at the prototype stage, was developed²⁴ within the framework of Project OXIRIS (orthovoltage X-ray-induced radiation system). The system had a rotatable X-ray source (120 kV, 1 mA) with interchangeable filters, emitting, as a net effect, a convergent X-ray radiation pointed at the focal spot. The sample, a 10 cm diameter phantom with a 1 cm diameter region infused with gold NPs acting as a biomarked artificial tumour, was positioned and scanned using an automatic sample holder. Four CdTe detectors (25 mm² active area) with conical collimators were employed. The focal spot was an approximate ellipse of 1.93 × 2.33 mm (FWHM). It was possible to detect Au K-fluorescence lines from a depth of 5 cm within tumour-like volumes down to a 0.5% w/w concentration of gold NPs used as biomarkers.

Improvements to confocal μXRF quantification schemes continued to be made. Cappuccio et al.²⁵ presented a thorough overview of the quantitative analysis capabilities of the fundamental parameter (FPM) approach, routinely available at the “Rainbow X-ray” facility (XLab, Frascati, Italy). A case study of decorative pigments covering two Japanese Buddhist scrolls provided quantitative results for 16 elements (Al to Sr) with concentrations of 0.04 to 73% (w/w). The estimated relative uncertainties of 1.8–75% depended on the concentration. Conclusions on the stoichiometry of the pigments used for the different colours were confirmed by literature data on the pigments available at the time the scrolls were manufactured.

Szaloki et al.²⁶ presented an optimised confocal μXRF analytical approach for determining the 2D distribution of elements in small (1–3 mm) biological objects at a 10–20 μm spatial resolution. A novel quantitative algebraic reconstruction model for confocal μXRF spectrometry datasets was based on an FPM algorithm optimised specifically for the monochromatic SR beam analysis of weakly absorbing biological samples. This FPM model was applied to the calculation of potassium concentration maps in the scanned cross-sections of cucumber hypocotyl samples.

Lin et al.²⁷ presented an excellent overview of their SR-based confocal μXRF spectrometry system. It included a detailed description of not only the quantitative FPM algorithm used but also the experimental setup established at beamline BL15U1 of the Shanghai synchrotron radiation facility. The setup and the associated quantification were characterised and benchmarked by analysis of several RMs which included NIST SRM 611 (trace elements in glass) and a set of thin metallic films (Au, Cr, Cu, Fe, Ni, Pb, Zn) sputter-deposited onto the surface of silicon chips with a typical thickness of <1 μm. The confocal volumes were 4.7 (horizontal) × 2.6 (vertical) × (22.8–11.4) (depth) μm for elemental lines from Cr-Kα (5.415 keV) to Pb-Lβ (12.614 keV). Detailed characterisation was given in terms of transmission efficiency and depth resolution of the polycapillary used in both focusing and confocal modes. The LODs ranged from 1.6 (Sr) to 5.6 mg kg⁻¹ (Fe) and the RSDs between nominal and calculated concentrations for the NIST SRM 611 were in the range 2–21% for Cu, Fe, Rb, Sr and Zn.

2.2 Laboratory 2D XRF techniques

Although laboratory μXRF spectrometry is more convenient to use than SEM and EPMA because it does not require a vacuum, the spatial resolution is still not sufficient and needs to be improved. High-resolution optics are available and widely applied at synchrotron facilities but they are not applied in laboratory setups because of the reduction in primary beam flux. The use of novel hardware and reconstruction algorithms in full-field μXRF spectrometry was reviewed in last year's ASU.²⁸ In this review period, Yang et al.²⁹ used oversampling and image restoration techniques in a comprehensive study to enhance the resolution of scanning μXRF spectrometry, which is more widely used than full-field μXRF spectrometry. The authors used two benchmark techniques (the mean squared error between measured and estimated values and the structural similarity index measure) to compare the image restoration procedures. Application to the Cr Kα false colour image of USAF 1951 (a microscopic optical resolution test chart of the US Air Force) provided individual patterns in the range from 0.250 to 912.3 line pairs per millimetre). The second method considered the structural information of an image, i.e. the interdependency of the pixels. A resolution of 11.3 lines mm⁻¹ was achieved when a 88 μm diameter beam and a step size of 5 μm were used without reconstruction. The image restoration algorithms all improved the resolution by 26.5% to 14.3 lines mm⁻¹. The Richardson–Lucy method was considered to give the best results because the images displayed the fewest artefacts.

An ongoing aim in developing new analytical procedures for characterising radioactive materials is to lower the sample requirement in order to limit contamination. Use of a new μXRF instrument, in which a Rh microfocusing tube and a polycapillary optic were combined, improved³⁰ LODs by an order of magnitude in comparison to a TXRF spectrometry approach. The instrument was evaluated³¹ by determining Pu and U fractions in fast breeder oxide and carbide fuels. The accuracy of ca. 0.2% was comparable to that obtained using electrochemical analysis but the sample requirement was reduced from a few mg to <1 μg. The LODs for Pu and U in quasi-matrix-free standards were reduced from the 2000–3000 ng mL⁻¹ obtained using cyclic voltammetry to ca. 60 ng mL⁻¹.

Determination of the relative amounts of Th and U in mixed oxide fuels was achieved³² by determining sensitivity from standards and by applying two thin-sample preparation techniques. In the first procedure, ca. 500 μg of the sample was scratched from the fuel pellet, suspended in 1% triton solution and pipetted onto a thin tape film. In the second procedure, a sticky tape was applied to the pellet and a few hundred ng of sample pulled off. The first procedure had a precision of 1% and a deviation from expected amounts of 3% and so was superior to the TXRF approach which had corresponding values of 2.6 and 5%. The second method was comparable to TXRF spectrometry with a precision of 2.96% and a deviation of 3.8%.

A comprehensive review on the application of XRF spectrometry in plant science covered³³ predominantly μXRF approaches which were considered useful for studies on plant physiology as they provided imaging and spatial mapping of elements in plant tissues. The mechanism of uptake, detoxification, and accumulation by the plant tissues could be studied. The feasibility of the in vivo imaging of elemental distributions in plant leaves will be important for these studies. The authors identified the increase in spatial resolution as a desirable improvement yet significant challenge. MicroXRF spectrometry was used³⁴ to study comprehensively seasonal variations in elemental composition and distribution in silver birch tree leaves. MicroXRF spectrometry performed on the roots of rice plants indicated³⁵ that, when compared with transplanted rice, directly seeded rice plants had less iron plaque covering the root surface in all growth stages but especially in the critical period of Cd accumulation, thereby weakening its role as an effective barrier to Cd uptake.

MicroXRF spectrometry is widely used in geological studies as it provides valuable information for sediments. MicroXRF spectrometry was³⁶ a powerful tool for interpreting the genesis of dolomitisation as it can be used to visualise the elemental distribution over a large area thereby to identify delicate genetic characteristics. Single point 10 s measurements gave a high level of accuracy (97–103% and a SD of <3%) for Ca, Fe and Mg. A 300 ms exposure time for each sampling point in a scan made it possible to study the enrichment and depletion of Mg and other elements. The elemental distributions in sulfidic, ferruginous and manganous zones revealed the effect of changing water redox conditions during the genesis of the dolomite nodule. MicroXRF spectrometry in combination with hyperspectral optical imaging was³⁷ particularly useful for studying paleoclimate on varved lake sediments. The C and Ti concentrations could be used as a temperature indicator whereas the Si concentration together with the mass accumulation rate could be used as an indicator for windy seasons.

3 Synchrotron and large scale facilities

Synchrotron-radiation-based XRF spectrometry represents the pinnacle of the XRF technique with spatial resolution now at nanoscopic levels and LODs at ppb levels approaching single atom sensitivity for the most efficiently detected elements for nanobeam applications. This is especially apparent in case of experiments performed at 3rd and the new 4th generation facilities. The use of SR-XRF spectrometry together with complementary X-ray spectroscopic/imaging techniques is becoming more common as both spatially resolved elemental information and speciation and structural/morphological imaging can be obtained. Scanning XRF spectrometry at the (sub-)μm scale continued to be used for elemental imaging in a wide range of research areas, including biomedical, environmental, material science and cultural heritage studies. The technique was used in conjunction mainly with micro- and nano-XAS and XRD methods at hard X-ray micro- and nano-probe facilities. This section focuses on notable advances in the field by highlighting a few outstanding instrumental and computational developments for SR-induced XRF spectrometry.

A new hard-X-ray nanoprobe beamline (I14) was installed³⁸ at the Diamond Light Source (Oxfordshire, UK) as a new facility for nanoscale microscopy. Emphasis was placed on multimodal analysis: elemental mapping by nanoXRF spectrometery; speciation mapping by XANES; structural phase mapping using nano-XRD analysis; and imaging by differential phase contrast and ptychography. The 185 m-long beamline operated over an energy range of 5–23 keV and provided a 50 nm beam size for routine experiments and a flexible scanning system for fast acquisition. The photon flux on the sample was ca. 5.4 × 10⁹ photons s⁻¹ at 12 keV, corresponding to a smallest focus of 50 nm. A flux of 5 × 10¹⁰ photons s⁻¹ with a horizontal spot size of 250 nm was possible when a secondary source aperture was opened. A four-element SDD (total area 200 mm²) in a backscatter geometry, located 17 mm from the sample and coupled with an Xspress3 pulse processor, gave an overall maximum count rate of 6 × 10⁶ cps.

In the context of using the soft regime of the X-ray spectrum for medical and biological imaging, Gianoncelli et al.³⁹ reported on the capabilities of the TwinMic soft X-ray microscopy end-station at the Elettra synchrotron (Trieste, Italy). The TwinMic beamline end-station was originally designed to provide a collection of imaging techniques in a single instrument with easy switching between the scanning transmission X-ray microscopy (STXM) and full-field modes. The STXM mode became dominant, mainly due to the development of the first low-energy X-ray fluorescence (LEXRF) system with imaging capabilities down to Z = 6(C). The LEXRF system had a spatial resolution of 0.4–2 μm with typical acquisition times of 1–10 s per pixel and a field of view of ca. 1000 μm. A notable development at TwinMic was the application of so-called compressive sensing for optimising the XRF data acquisition so that only the most significant data were collected. This was achieved by applying either: (i) a mask based on a previously acquired absorption image; (ii) an absorption threshold so that the XRF signal was acquired only in regions of specific interest; or (iii) a threshold on the XRF signal, thereby acquiring a longer XRF spectrum in the regions where a specific element of interest was present in higher concentrations.

A smart XRF-mapping-algorithm used⁴⁰ a Synchrotron-based Machine learning Approach for RasTer (SMART) mineral mapping. The Artificial Neural Network (ANN) based algorithm was developed initially for training and was subsequently applied as a mineral classifier for the characterisation at μm-scale resolution of mm-sized areas of rock thin sections. Initial training and testing were based on 192 coupled μXRF–μXRD data points (2 μm resolution) from a 0.25 mm² area of a 30 μm thick shale sample from the Eagle Ford rock formation (Texas, USA). The training involved the collection and evaluation of both μXRF spectrometry and μXRD data and ensured that the information about the mineral phases was correctly associated with the μXRF spectrometry data. Application of the ANN using only μXRF spectrometry data, accelerated and simplified the data analysis substantially. This was possible as the intensities of the fluorescent X-rays were sensitive to the concentrations of individual elements as well as to the densities of the corresponding mineral phase. In the testing phase, the minerals (aragonite, calcite, dolomite, pyrite, pyrolusite, pyrrhotite) were classified correctly with an accuracy >97%.

Software developed by Marini et al.⁴¹ was designed for automated evaluation of combined μXRF spectrometry and μXANES datasets for the characterisation of complex heterogeneous samples. The intuitive Python code MAP2XANES used basic Python modules and functions from Numpy, Scipy, Pandas, iPywidgets and Matplotlib libraries. The output of the data processing was a correlation between elemental distributions and spatial localisation of the chemical species. The software was demonstrated by the determination of Mn²⁺, Mn³⁺ and Mn⁴⁺ distributions in a mineral sample of hausmannite (Mn²⁺Mn₂³⁺O₄). The code could be used to: (1) visualise maps of elements and the scattering matrix; (2) simultaneously normalise a large set of μXANES data, including correction for self-absorption; (3) perform linear-combination-fit (LCF) analysis; and (4) overlap the LCF results with elemental distribution maps. The software was intended to be OpenSource, distributed under MIT licence and freely downloadable from the GitHub site of ALBA Synchrotron (Barcelona, Spain).

An innovative machine-learning-based quantification approach for synchrotron XRF spectrometry used⁴² ANNs trained using synthetic XRF spectra which had been generated by Monte Carlo simulations. The main advantage of this method when compared to conventional FPM was the possibility of quantifying the elemental concentrations directly from the XRF spectra thereby eliminating the need to use, e.g., spectral-fitting algorithms. All the main elements (Ag, Cu, Ni, Zn) in gold RMs (ERM 506, ERM 507, ERM 508) were determined simultaneously with RSDs of 0.1–12.6% relative to the certified concentrations.

A particularly innovative dual-energy X-ray-beam ptycho-fluorescence imaging method combined⁴³ XRF spectrometry and X-ray ptychography scanning with simultaneous data-collection. To overcome the incompatibility of the different excitation and scanning conditions required for fast ptychography and high-resolution XRF spectrometry, two coaxial beams of different sizes were used on the sample simultaneously. A pinhole beam was combined with a focused one using a Fresnel Zone Plate. The spatial resolution was 248 nm for the ptychography and 4 μm for the coupled μXRF imaging. The reduced acquisition time (ca. 1600 s) was ca. 25 times shorter than when a single beam was used for separate XRF and ptychography imaging of the same field-of-view on the sample.

In an attempt to improve the energy resolution of detection for low-energy SR-XRF spectrometry, Jagodzinski et al.⁴⁴ studied the properties of polycapillary lenses used as a detector collimator for a low-energy parallel-beam wavelength-dispersive spectrometer. The study concentrated on modelling the properties of the polycapillary optics and on simulations of X-ray transport in the spectrometer installed at the ID21 beamline of the ESRF (Grenoble, France). The X-ray transmission of the tested polycapillary optics was ca. 15% and the divergence of the collimated XRF-radiation changed from 8 to 3 mrad in the energy range 2–10 keV. The spectrometer had an energy resolution of 5–33 eV in the energy range of 1.4–6.5 keV.

4 Grazing X-ray techniques including TXRF spectrometry

Although TXRF spectrometry is usually considered to be the best XRF method for obtaining excellent low LODs, other geometries such as conventional XRF spectrometry and μXRF spectrometry can be optimised to have an ultra-low background and therefore very low LODs. Alam et al.⁴⁵ demonstrated that the absolute LODs of a few μg obtained when using a conventional EDXRF geometry could be improved to the levels obtained using TXRF spectrometry by preparing minute amounts (90 ng of a multielement RM) of sample on thin kapton foils (12–100 μm). Both techniques were applied at a synchrotron facility (BL-16 beamline of Indus-2) to make use of the polarisation of the synchrotron light and a geometry with an azimuth angle of 0° and a scattering angle 90°. Although the LODs obtained using EDXRF spectrometry were always higher than those obtained using TXRF geometry, very low LODs were nevertheless obtained. For example, the LOD for Cr of 0.08 ppm compared well with that of 0.04 ppm obtained by TXRF spectrometry. In a comparison of EDXRF and TXRF spectrometries for the analysis of air particulate matter generated at a welding site, the filter was analysed⁴⁶ directly by EDXRF spectrometry and after digestion in the TXRF analysis. There was no significant difference between the results obtained for Ca, Cu, Fe, K, Ni, Ti and Zn by the two procedures. Although the EDXRF procedure did not require digestion of the sample, it was unable to detect Al, P and S in the five samples collected at the welding site. The LODs for Al in similar samples were 3500 ng m⁻³ using EDXRF spectrometry and only 1400 ng m⁻³ using TXRF spectrometry. The LODs for the TXRF procedure were better for all elements than those obtained by EDXRF spectrometry; the smallest difference was 2.4 times for K and the largest difference was 6.0 times for Bi.

Demonstrating the accuracy of the elemental analysis of aerosols for air quality monitoring purposes is a major challenge. In particular, calibration of devices in the laboratory with quite simple model aerosols can contribute to uncertainties in field studies. Measurement uncertainties of 25% are generally regarded acceptable by regulatory bodies. The development of a facility producing multi-component model aerosols for instrument and procedure calibration would be a major step forward to improve significantly the accuracy of aerosol characterisation. Horender et al.⁴⁷ introduced a novel facility for the production of ambient-like model aerosols in the laboratory. The concentrations of the model aerosols could be adjusted from a few μg m⁻³ to ca. 500 μg m⁻³. The relative uncertainty (CI 95%) at 40 μg m⁻³ was 5.7%. The results were obtained by TXRF spectrometry analysis of model aerosols collected with a 13-stage low-pressure cascade impactor. The GIXRF and TXRF spectrometries were compared⁴⁸ for the determination of Al, C, Mg, N, Na and O in aerosols at the PGM beamline at the PTB laboratory, BESSYII, using an excitation energy of 1620 eV. Samples with low aerosol concentrations gave identical results in GIXRF and TXRF geometries whereas high-load samples gave lower elemental amounts when using the TXRF geometry. The results from the GIXRF analysis for the high-load samples could be used to correct for the low results obtained by TXRF analysis. The authors suggested the use of both geometries as this would provide both the benefit of low LODs obtained by TXRF spectrometry and the higher dynamic range provided by GIXRF spectrometry. Short sampling times for TXRF analysis could, however, potentially ensure ideally thin aerosol coverage on the sample carrier and reliable results for ambient aerosols. A 4 h sampling time, optimal for collecting ambient aerosols with a May-type impactor, yielded⁴⁹ an LOD of 0.1 ng m⁻³ for transition metals. Following analysis using TXRF spectrometry, samples with elevated concentrations of Br and Cu, which reflected pollution episodes, were selected for further study using SR-based XANES to determine the chemical species present in each sample. The possible sources were determined based on the elemental size distribution and the chemical species of Br or Cu being present in the samples. Determination of the Cu species using both TXRF spectrometry and XANES analysis made source assignment possible.

The binding of Fe ions by a monolayer film of gemini surfactant at the water air interface was studied⁵⁰ by angle-dependent TXRF spectrometry at the liquid-surface-scattering spectrometer (NSF's ChemMatCARS) at the APS. The angle scans showed that Fe was being accumulated in a thin surface layer on top of a Fe³⁺ aqueous solution. The use of TXRF spectrometry uniquely allowed access to the air liquid interface. Using a model developed in-house,⁵¹ surface ion concentrations were determined to be 0.037 to 0.069 ion Å⁻² at aqueous phase concentrations of 0.015 to 0.1 mM Fe.

In many forensic investigations, samples are only available in limited amounts so the micro analytical capabilities of TXRF spectrometry are especially useful. Komatsu et al.⁵² showed that the TXRF spectrometry detection of trace amounts of Ca, Cl, Cr, K, S and Zn in red silk single fibres made it possible to classify the fibres by the presence of metal-mordant dye. The presence of Cr and Zn as well as the Zn/Cr intensity ratios were the decisive parameters. The SR-XRF spectrometry measurement of Br and Co in single fibres allowed all the analysed fibres to be characterised. Single polyester fibers were also analysed⁵³ non-destructively for forensic purposes. Measurement of trace elements such as Co, Ge, Mn, Sb and Ti from additives and catalyst residues made it possible to establish individual elemental fingerprints for the fibres. Correlations between compositions and manufacturers were successfully established using PCA.

A method for preparing single apatite crystals on quartz reflectors with subsequent digestion used⁵⁴ stoichiometric information for P in the standardisation. Quantification using TXRF spectrometry provided sufficient LODs (1–20 mg kg⁻¹) for As, Br, Ce, Fe, K, La, Mn, Nd, Ni, Pb, Pr, Sr, Th, U, Y. The method was faster (200–1000 s for multielemental screening), more cost-efficient (Ar was not required) and more user-friendly than an LA-ICP-MS procedure.

The micro-analytical capabilities of TXRF spectrometry can play a vital role in clinical studies by determining elemental profiles and their anomalies for individual patients. Only 40–70 mg of a cleaned nail clipping was needed⁵⁵ to study profiles of As, Br, Ca, Cr, Cu, Fe, K, Mn, Ni, P, Ti, S, Se, V and Zn in patients with colon cancer. In comparison with the data from a control group, the nails of men with colon cancer had significantly lower concentrations of Ca (p < 0.05) and Zn (p < 0.05) and higher concentrations of As (p < 0.05), Cr (p < 0.05), Cu (p < 0.001), Fe (p < 0.001) and Se (p < 0.001).

Although determination of the low-Z matrix in TXRF spectrometry is usually omitted due to the lack of detectable fluorescence lines, both the coherent and incoherent scattering-cross-sections are a function of the energy and Mennickent et al.⁵⁶ successfully used the ratio of the prominent Compton and Raleigh scattering signals to determine the “effective Z′′ for different milk samples in a univariate calibration. When the model was applied to undiluted milk samples, it was possible to differentiate between raw and treated products and between samples with low and high fat contents. Even more information from the scattering region of the primary beam was extracted by Breuckmann et al.⁵⁷ when they applied a multivariate fitting approach. The C, H, N and O concentrations in polymers were determined using the scatter spectra acquired by WDXRF spectrometry and PLS regression. The correlations between measured and modelled results were very good for C, H and O (R² 0.912–0.986) and independent of the spectral resolution. The correlation for N was, however, rather poor (R² 0.183–0.558).

An article on the use of TXRF spectrometry in the field of biomedical science by Fernandez-Ruiz⁵⁸ included a short overview of its history, figures of merit and a tutorial on analytical procedures. This will be helpful for TXRF spectrometry practitioners, in particular for beginners in this field of study. A critical discussion on the sample preparation of a number of biomedical fluids was given by Margui et al.⁵⁹ The TXRF spectrometry approach allowed for rather simple sample treatment; human fluids were prepared after dilution and tissues by suspending a few mg of the powdered material.

Sample inhomogeneity can have a major influence on the quality of TXRF spectrometry results. Mankovskii and Pejovic-Milic⁶⁰ modelled its influence on the direct analysis of microtome thin slices. They used a profile-fitting based software (TOPAS) to simulate the TXRF instrument and a Python code to simulate the samples. The recovery was 98–105% for four distributions of Au and La: two were layered samples, one was homogeneous and the fourth was an inner square of gold on top of a La layer. This simulation toolkit could easily be adapted to other samples. The same authors directly quantified⁶¹ 10 nm gold NPs placed on either the upper or lower surfaces of thin (ca. 5 μm) slices of bovine liver spiked with a La IS. Determination of the gold NPs using TXRF spectrometry and TOPAS simulation had an excellent accuracy of 96–101% for all samples. The controlled drying of aqueous samples to give an ideal homogeneous deposit is an ongoing challenge. Tsuji et al.⁶² showed that glass substrates treated with an atmospheric-pressure He plasma had a small hydrophilic surface-area and could be used to produce a film-like residue from a small droplet of wine.

Whereas the ICP-AES and FAAS analysis of suspensions of herbal tea and plant leaf CRMs required⁶³ sample digestion, the use of TXRF spectrometry made direct analysis possible. The TXRF analysis did not require ultra-pure gases and could give results for 10–50 mg of sample in as little as 15–40 min. The ICP-AES analysis, on the other hand, took 2 h and consumed considerable amounts of ultra-pure Ar and the FAAS analysis took 5 h. In addition, both the latter techniques needed 100–500 mg of sample. Results for the three methods were not significantly different but the TXRF analysis provided better LODs overall (from 0.2 to 0.6 mg kg⁻¹) and better recoveries (97–113%) than the other methods for which the LODs were 0.1 to 1.6 mg kg⁻¹ and the recoveries 80–106%. Similar findings were made⁶⁴ for the biomonitoring of Cr, Cu, Ni, Pb and Zn in aquatic plants. The TXRF spectrometry method was faster than AAS analysis and less costly than ICP analysis, at the same time providing high quality results.

Matrix effects in TXRF spectrometry are usually neglected but can occur⁶⁵ at high concentrations. The efficient elimination of matrix and the preconcentration of Cd^II, Co^II, Cu^II, Ni^II, Pb^II and Zn^II from aqueous solutions with an enrichment factor of 133 has been achieved⁶⁶ using graphene oxide/carbon nanotube membranes. The LODs for a short measurement time of 600 s were 0.001–0.002 ng mL⁻¹ for all elements except Cd for which the LOD was 0.11 ng mL⁻¹. Low LODs were also obtained⁶⁷ by the same group by preconcentrating Hg^II from aqueous samples with a high ionic load (e.g. seawater). An LOD of 2.6 pg mL⁻¹ was achieved by using dispersive μSPE with thiosemicarbazide-functionalised carbon nanotubes. The Se concentrations (256.5 ± 22.5, 88.4 ± 1.42 and 61.3 ± 2.8 ng g⁻¹) determined directly in untreated minewater samples prepared using a 10 μL single-drop and 600 s measurement time agreed⁶⁸ within 1–3 SD of those obtained by ICP-MS analysis.

A review on low-angle X-ray-based analysis at the Jan Kochanowski University in Kielce illustrated⁶⁹ nicely the capabilities of sensitive surface and subsurface methods such as TXRF spectrometry, GEXRF spectrometry, GIXRF spectrometry, XRR and TRXPS. Recently, the Chambre d’Analyse Spectrométrique en Transmission ou en Réflexion (CASTOR), which combines low-angle-based techniques (XRR, TXRF spectrometry and GIXRF spectrometry), was commissioned⁷⁰ at Soleil. The GIXRF technique is often used in combination with XRR for the study of thin layers but can also be used to study regular nanostructures. A GIXRF-based methodology in the soft X-ray region was successfully applied⁷¹ to characterisation of the physical and compositional dimensions of a 2D regular nanostructure (pitch 100 nm, height 90 nm and line width 50 nm) of a silicon nitride grating covered by a silicon oxide layer. The sensitivities were improved by incorporating the N Kα and O Kα fluorescence signals originating from within the nanostructure. The incorporation of supporting experiments such as XRR for optical constant verification and machine learning techniques such as Bayesian optimisation decreased the computation requirements. The GIXRF reconstruction results agreed well with those obtained using SEM and AFM. The reconstruction also revealed unexpected effects on the nanostructure by, e.g., carbon contamination. The XRR technique was employed⁷² in combination with GIXRF techniques for the characterisation of a Ru/C/Ru waveguide. Whereas the bottom layers had densities close to that of the bulk density (within 95% for C and 97% for Ru), the ruthenium top layer had a slightly lower density (within 93% of the bulk density). The 10% thickness variation in the top cladding layer together with this marginal change in layer density affected the amplification of the wave field, in some cases by a reduction of more than three times.

Speciation of amorphous material in the laboratory is often challenging. A novel high-resolution table-top NEXAFS spectrometer with photon energies of 250–1000 eV, developed⁷³ for the absorption spectroscopy of thin samples, is also potentially applicable to low-angle analysis. The spectral resolution of E/ΔE = 1535 at 430 eV was close to that of typical synchrotron setups.

Standardisation of TXRF procedures is an ongoing effort. A TXRF spectrometry round-robin on pre-selected and well characterised samples reported by Unterumsberger et al.⁷⁴ was an important part of this effort. Sample preparation involved conventional μL deposition, nL dispensing and sub-monolayer thin-film sputtering. Excellent accuracy and precision data were obtained for the different kinds of well-characterised micro- and nano-scaled samples.

The absolute transition probabilities of L- and M- series of different Au and Pb species were studied intensively by Fernandez-Ruiz.⁷⁵ Intensity variations between the fluorescence lines of Au⁰ and chloride Au^II were observed. Whereas the Lβ line intensity decreased by 7.3%, that for Mαβ lines increased by 96%. Similar deviations were observed for metallic Pb⁰, nitrate Pb^II and sulfide Pb^II.

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

Portable XRF spectrometry has become well established for a range of applications as a result of the development of suitable calibration software by instrument manufacturers. In this section of the review, the focus is placed on the analytical XRF aspects and continued optimisation of the portable XRF technique. Details of dedicated applications are given in our companion ASU reviews.^1,2,5

Although EDXRF spectrometry is a well-established technique, new developments and software improvements continue to be made. In an EDXRF system using pyroelectric sources developed⁷⁶ for use in space exploration, a redesigned electronic readout for an SDD gave an energy resolution as low as 121.17 eV. The measurement repeatability was improved from 25.2 to 3.71% when the spectrum was renormalised using an indium thin wire reference. Possible drift effects in space due to aging were compensated for by applying a peak-search algorithm to find the best ROI for the fluorescence peak. An extensive test campaign was to be performed on a wide range of geological standards to determine the LODs of the current instrument. Lu et al.⁷⁷ based a new method for the correction of matrix effects on multiple linear regression analysis. The regression method was based on correction factors obtained for the major elements from analysis of 16 CRMs. Compared with calibration data obtained from simple linear regression analysis without taking major elements into account, the new method provided higher quality data, especially for Cd, Co, Mo, Pb, Sn, Ta, Tl and Zn. Almost all the data had a relative error of <20%. Hao et al.⁷⁸ validated a new approach to overcome the disturbance caused by pulse pile up in the measurement of traces of Ag in refined copper using portable XRF spectrometry. A 0.05 mm copper absorber sheet was used to suppress the pile up peaks resulting from the copper matrix, thereby enabling the X-ray characteristic peak signal of trace Ag to be enhanced by the elevated tube current. Moreover, the S/N was improved by the scattered-radiation normalisation method. A calibration for Ag from 0.01 to 0.10% (w/w) had an R² of 0.992, an RMSE error of 0.0029% (w/w) and an LOD of 0.0034% (w/w). The model had a prediction accuracy error of less than 0.0050% (w/w).

An interesting and comprehensive review on the use of the portable XRF spectrometry technique for the analysis of tropical soils by Silva et al.⁷⁹ highlighted the key differences between soils from tropical and temperate regions that need to be taken into account when using portable XRF spectrometry analysis. State-of-the-art achievements were presented and potential applications of portable XRF spectrometry to tropical soils research considered. In addition, a standard portable XRF methodology for tropical soil analysis was established. A nice review paper by Fleming⁸⁰ covered the use of portable XRF spectrometry for the measurement of As, Cr, Hg, Mn, Se, Zn and Pb in intact nails and nail clippings. Both early developments and recent studies were highlighted.

The portable XRF technique is widely used for the characterisation of soil samples but the influence of sample grain size and moisture content need to be taken into account if good quality data are to be produced. A comparative study by de Freitas et al.⁸¹ used three different EDXRF spectrometers (one benchtop, two portable) to study the influence of soil sample grain size. Soil samples were collected, ground and sieved into the particle size ranges: macerated (>2 mm), <2, <0.7, <0.3, <0.125 and <0.06 mm. Recoveries obtained for K and Zn when using the spectrometer made in-house were outside the acceptable range of 80–120%. The quality of data produced depended not only on the soil particle size but also on the sensitivity of the spectrometer and the element of interest. Whereas the optimum soil-particle size was <0.125 mm for the benchtop and one portable instrument, it was <0.06 mm for the other portable instrument. Nakano et al.⁸² investigated the effect of soil moisture on the on-site XRF analysis of hazardous metals in polluted soil. Wet soil samples (up to 30% water content, w/w) were prepared by exposing dry soils to a water mist generated by an ultrasonic humidifier for 0–95 min. The sensitivity of calibration graphs for six metals (As, Cd, Cr, Hg, Pb, Se) decreased with the increasing water content. The XRF intensities decreased about 15–20% when the moisture content was 30% (w/w). A new method for correction of this effect was based on Compton scattered X-rays and the soil water contents. As a result, the sensitivities of the calibration curves for any individual metal were identical at different water contents. Results for Cr in a geochemical RM were in good agreement with the expected value even at increased water content; for samples with the highest water content, the bias improved from 38 to 6%. Zhang et al.⁸³ developed a rapid field method for measuring metals in the residues from horizontal directional drilling (HDD) operations so that proper disposal decisions could be made in situ. Soil samples were spiked with four concentrations (up to 1000 mg kg⁻¹) of As, Cd, Cr, Cu, Ni, Pb and Zn. When the moisture content was reduced to <30% (w/w), the results obtained using portable XRF spectrometry were highly correlated with those obtained using the standard method. The slopes of the calibration graphs ranged from 0.83 ± 0.01 to 1.08 ± 0.01 (P < 0.001) for all metals. The procedure was used successfully for 59 HDD residues (moisture contents of 14–83%, w/w) analysed at sites in 26 US States. A commercially available filter press was used to remove excess water from samples with water contents of >30% (w/w) in order to achieve improved accuracy. The portable XRF procedure proved to be a reliable tool for the rapid detection of common metals in dried soils and HDD residues.

The performance of portable XRF spectrometry for the analysis of liquid matrices was evaluated by Kagiliery et al.⁸⁴ on 1440 samples. The study involved three elements (Cd, Cr, Pb), each at four different concentrations (125, 250, 500 and 1000 μg g⁻¹), three different film types (Kapton, Mylar, and Prolene) and five different liquid depths (4.29, 8.59, 17.18, 25.77 and 30.06 mm). Initially, there were significant differences between the expected values and those obtained using XRF spectrometry. As was to be expected, increased water thickness and, therefore, greater volume improved the XRF predictive accuracy. Of the three thin films evaluated, Kapton and Mylar film offered slightly better accuracy. An adapted calibration curve (linear for Cd and Cr, quadratic for Pb) was established using a water depth of 4.29 mm with all three films evaluated. As a result, there were no significant differences between the XRF predicted concentrations for different water depths and film types.

An experimental approach for in situ and ex situ analysis of plant leaves was used by Ribeiro et al.⁸⁵ to obtain the concentrations of both macro- (Ca, K, Mg, P, S) and micro-nutrients (Cu, Fe, Mn, Zn). The analysis of intact and fresh leaves in the field presented a considerable challenge as XRF spectrometry analysis is influenced by water content, natural heterogeneity and sample thickness. For on-site analysis, Pinus wood was used as a natural background stage and at least 1 cm thickness was required to avoid any influence from the soil surface. The plant thickness was evaluated by overlapping 1 to 10 leaves and gently compressing them between the X-ray source/detector and the wood background stage. For all macro- and micro-nutrients except Mg, the effect of thickness of the ground plant material (spinach and post oak leaves) was well represented by a polynomial inverse first model. The R² values ranged from 0.71 (Ca and Zn in spinach leaves) to 0.99 (K in post oak leaves). As was expected, the water content influenced the results so correction factors (using e.g. Compton scattering effect) needed to be applied. In the ex situ analysis of oven-dried and ground plant material, the thickness in the sample cup needed to be at least 1 cm to reach the spectroscopically infinite thickness of plant materials.

A handheld XRF spectrometer was evaluated by Horf et al.⁸⁶ for the challenging ambition of achieving a fast, simultaneous and on-site determination of several elements in organic liquid fertilisers. A set of 62 liquid pig and cattle manures as well as biogas digestates was collected and analysed for macro plant nutrients (Ca, K, Mg, P, S) and micro nutrients (Cu, Fe, Mn, Zn). The effects of four different sample preparation steps (untreated, dried, filtered, and dried filter residues) on XRF measurement accuracy were examined and the XRF results correlated with values obtained by acid digestion and ICP-AES measurement). The best correlations for most elements (but not for Ca, Mg and P) were obtained for the dried samples (R² from 0.64 to 0.86) and the dried filter residues (R² from 0.65 to 0.92). In contrast, liquid samples (untreated and filtered) had lower R² values of 0.02–0.68. The only exception was K which had R² values of 0.83 and 0.87, respectively. It was concluded that handheld XRF measurements for liquid manures and biogas digestates without special sample pre-treatment steps did not give reliable results other than for K. On the other hand, the analysis of dried and ground samples gave promising results. The performance of portable XRF spectrometry for the determination of Al, Ca, Cu, Fe, K, Mg, Mn, Na, P, S and Zn in biochar-based fertilisers was investigated by de Faria et al.⁸⁷ by comparing results with those obtained using either the standard modified dry-ashing method with ICP-AES analysis or WDXRF spectrometry. Adequate linear regression between methods was achieved for all elements and biochar samples. For the lighter elements (Al, Ca, Mg, Na, P, S), portable XRF spectrometry gave the lowest measured concentrations and the standard ICP-AES method the highest. In contrast, both XRF methods gave higher values for the heavier elements (Cu, Fe, Mn, Zn) than the standard ICP-AES method. Using portable XRF data, elements were characterised as being in three different categories. Zinc was in a definitive category (R² 0.85–1.00, RSD < 10%); Fe, Mn, P and S in a quantitative category (R² 0.70–1.00, RSD < 20%) and Al and Mg in a qualitative category (R² < 0.7, RSD > 20%). The portable XRF technique had encouraging potential for a rapid and chemical-free characterisation of biochar-based fertilisers.

6 Cultural heritage applications

The use of hand-held XRF spectrometry is indispensable in the study of cultural heritage materials and allows fast, non-invasive and non-destructive analysis of artworks, particularly in museums, conservation studios and during archaeological excavations. A review by Cesareo et al.⁸⁸ of the history and technical evolution of portable EDXRF devices for the study of works of art highlighted impressive developments in all the various instrumental components during the last 50 years. The sources used for portable XRF measurements evolved from radioactive sources to big X-ray tubes and then miniature air-cooled, low-power and dedicated tubes. The X-ray detectors changed from proportional gas counters to nitrogen-cooled Ge or Si semiconductor detectors to Peltier-cooled Si-drift detectors. Pulse height analysers switched over from big and heavy boxes to mini-electronic circuits incorporated in the detector box. Moreover, the measurement modality changed from single-point analysis to a complete scanning of the object to be studied. Gherardi⁸⁹ discussed the main advantages and limitations of the hand-held XRF technique in the study of cultural heritage materials, together with its potential for archaeological and conservation research.

The development of a prototype low-cost portable EDXRF spectrometer (FUXYA2020) for cultural heritage applications was presented by Ruschioni et al.⁹⁰ The geometry was optimised to meet the requirements for measuring both low Z (starting from Si) and medium–high Z (Co, Cu, Fe, Hg, Mn, Ni, Pb, Sb, Sr, Zn, Zr) constituents. The system consisted of: an X-ray tube with rhodium anode (4 W, 50 kV, 200 μA); two brass collimators with aluminium inserts (with 1 and 2 mm diameter holes); aluminium (254 μm) and copper, molybdenum, silver, tungsten (25.4 μm) filters; and an SDD (17 mm² active area, 500 μm thickness and 12.5 μm beryllium window). The case and the structural components were all 3D printed. The FUXYA2020's performance for qualitative analysis was tested by comparing data obtained for pigment layers with those obtained using a commercial (more expensive) XRF spectrometer. Elements typical for the analysed pigments (16) were detected within the XRF spectrometry limits. The Axil-QXAS software was used for data elaboration. The results for the quantitative analyses of 3 gold-based certified alloys agreed well with the certified values.

A laboratory-based XRF methodology presented by Laforce et al.⁹¹ for standardless quantitative analysis was based on a monochromatic XRF spectrometer (Mo anode X-ray tube, with an X-ray spot size of 220 μm on the sample) and Monte-Carlo-aided quantification. A laser triangulation system was used for adequate and repeatable positioning of the sample relative to the frame of reference of the spectrometer. The precision along the X-ray beam of 4 μm was much smaller than the depth of focus of the X-ray tube (∼250 μm). The upgraded μXRF instrument produced LOD values of <10 ppm for a wide range of elements (from K to Y) using a 100 μm thick MPI DING KL2-G (reference glass) as the RM. Archaeological flint artefacts from the Scheldt basin were analysed in an attempt to determine provenance using k-means clustering. The results contributed to a database created to assign lithic artefacts to specific geological outcrops.

Although XRF spectrometry and RS are commonly used for the characterisation of paintings and their synergetic application is useful for non-invasive scientific analysis of works of art, the value provided by comparing XRF spectrometry and RS data is not yet fully exploited. By applying computer vision and statistical methods such as spectral angle mapper (SAM), Galli et al.⁹² implemented an innovative approach that exploited the elemental features, obtained from XRF spectra, to improve the comprehension of the molecular aspects given by RS, and vice versa. The main issues behind the approach were elucidated and its capabilities were demonstrated in the case study of the painting Chariot Race by Giorgio De Chirico (1928–1929, oil on canvas, Pinacoteca di Brera, Milan, Italy). The results reflected the complexity of the painting. Even if only some of the spectra could be attributed to recognisable pigments, by combining the elemental distribution data and SAM maps a mixture of materials could be identified that matched the description given by the artist in his “Small Treatise on Pictorial Technique” (De Chirico in Abscondita, 2019). The importance of the macroXRF spectrometry scanning technique in the investigation of historical paintings was highlighted by Orsilli et al.,⁹³ who discussed its advantages and limitations and emphasised the challenges of handling the large data set and the inhomogeneity of the data itself. An impressive pioneering protocol called STEAM (Statistically Tailored Elemental Angle Mapper) was based on the SAM algorithm and combined macroXRF standard data with point XRF spectrometry data. Information from the scanned area, e.g. for pigments used either originally by the artist or in subsequent restoration, was used to construct a database which could be applied outside the scanned region. The protocol was independent of the analytical instrument and could be used to provide information for pigments at uncharacterised data points. The protocol, whose strength and capabilities was demonstrated on some reference pigment layers and a painting by Giotto, holds considerable promise for future research.

The performance of a modular macroXRF spectrometry scanner developed in-house by Lins et al.⁹⁴ was compared to that of a commercial instrument by analysing a Portuguese tile (azulejo) from the 17th century. Multiple X-ray detectors (3 SDDs), used to achieve shorter dwell times, were configured identically so that acquisition could be synchronised and the sum spectrum from all detectors viewed in live time. The commercial system had a better energy resolution and S/N ratio and so provided a clearer image (for the SbLα line) and overall better images (for the PbMα and SnLα lines) at the lower energy end of the spectrum. However, for the mid- to high-energy range (∼5 to 30 keV), the modular macroXRF scanner, equipped with an exchangeable 2 mm aluminium collimator, provided better results with higher counts and, therefore, better images for, for example, the PbLα and SbKα lines. Monte Carlo simulations were a powerful tool for determining the stratigraphy, composition and thickness of almost all pictorial layers of the analysed azulejo sample. Mazzinghi et al.⁹⁵ used a macroXRF spectrometry scanner for the analysis of the painting Entombment of Christ by the Flemish artist Rogier van der Weyden. The scanner was designed and developed specifically for cultural heritage applications by INFN-CHNet. The scanner was composed of an X-ray tube (Mo anode, max. 40 kV and 0.1 mA), 800 μm diameter collimator and an SDD (25 mm² active surface, 500 μm thickness). Operating conditions were 30 kV and 0.1 mA. The scanning velocity ranged from 1 to 2 mm s⁻¹ and the equivalent pixel size from 200 to 1000 μm. Traditional materials determined or hypothesised by the XRF spectrometry measurements included azurite, ultramarine, lead-white, vermilion, Cu-based green, Fe oxides/hydroxides (earth/ochres), lead–tin yellow, and gold (in thin lines in halos). Although the macroXRF spectrometry technique lacked specificity for a comprehensive characterisation, it nonetheless proved to be an invaluable tool for providing an overview or hypothesis of the painting materials and techniques used.

A multi-method approach is often used in cultural heritage applications to elucidate the objects under investigation such as drawings, manuscripts and paintings. Bicchieri et al.⁹⁶ reported the results of the non-destructive analysis of two drawings: Studi di gambe virili and Autoritratto. The use of three complementary techniques (macroXRF spectrometry, μRaman spectroscopy and AFM) with different penetration capabilities gave a complete chemical characterisation of the artworks. The macroXRF imaging was carried out using the LANDIS-X, a novel mobile X-ray scanner based on a real-time technology integrating macroXRF, μXRF and confocal-XRF spectrometries for performing both 2D and 3D elemental imaging. By using a spot size of a few hundreds of microns, the full area of 110 × 70 cm² was covered in less than 2 hours at 100 mm s⁻¹ with a pixel size of 1 mm. Macro-XRF measurements showed the presence of pentimenti – hidden marks or non-original compounds on the surface of the artworks – in the Studi di gambe virili and highlighted that the Autoritratto was drawn without a metal point preparatory sketch. The AFM topographies revealed that the Autoritratto urgently required chemical stabilisation. Colantonio et al.⁹⁷ included XRF spectrometry mapping data in their approach of integrating imaging data from multiple sources to investigate a page from the famous Persian illustrated manuscript “Humay meets the Princess Humayun” kept in the Louvre, Paris. Hypercolorimetric multispectral imaging, a portable 7-band multispectral system and its native digital-image-processing software, was used to integrate XRF imaging data with multispectral and colorimetric datasets. Study of the rich colour palette made it possible to identify original materials and restoration interventions and the eventual degradation of both pictorial layer and ancient paper. Digital treatments (e.g. multispectral similarity maps, cluster analysis, PCA) were used in order to achieve cross analysis of multisource data. In particular, the use of XRF elemental mapping in conjunction with multispectral images made it possible to map successfully the vergaut colour, a mixture of mineral arsenic-based orpiment pigment and organic dye indigo. The dark halos around figures and objects in a 17th century Western European easel painting were examined by Derks et al.⁹⁸ using IR photography and IR reflectography combined with complimentary macroXRF imaging. In contrast to the use of IR imagery for which the dark background prevented the halos from being visible, macroXRF imaging successfully visualised the hidden halos in the elemental distribution maps of Fe, K and Pb. More insight into the function of the dark halos and their optical effects on the adjacent colours was obtained. It became clear that resourceful artists probably adapted the halo technique for differently coloured backgrounds.

A comprehensive multi-material and multi-method approach based on the combination of SR-based X-ray micro-analytical techniques (XRD, XRF, XANES at SK-/AgL₃-/AsK-edges) and vibrational micro-spectroscopy methods was used by Monico et al.⁹⁹ to reveal the causes and mechanism of the darkening of “fake-gilded” decorations in tempera paintings which had originally consisted of an unusual mixture of As₂S₃ and metallic silver (Ag⁰). Measurements were performed at the scanning X-ray microscope endstation of beamline ID21 and at the hard X-ray nanoprobe beamline ID16b of the ESRF (Grenoble, France). The high specificity, sensitivity and lateral resolution of the analytical methods provided first-time evidence for the presence of black acanthite (α-Ag₂S), mimetite [Pb₅(AsO₄)₃Cl] and syngenite [K₂Ca(SO₄)₂·H₂O] as degradation products of the “fake-gilded” decorations in the Maestà by Cimabue (Church of Santa Maria dei Servi, Bologna, Italy). Study of the painting together with that of tempera paint mock-ups proved that: (i) Ag⁰ and moisture were key-factors for triggering the transformation of As₂S₃ to α-Ag₂S and As₂O₃; (ii) S²⁻-ions arising from the degradation of As₂S₃ were mainly responsible for the formation of α-Ag₂S; (iii) light exposure increased alteration of the paint components. Laboratory- (SEM-EDS and μRaman spectrometry) and SR-based (high resolution μXRF spectrometry and μXANES at the ID21 beamline of ESRF (Grenoble, France)) techniques were used by Marketou et al.¹⁰⁰ to investigate an unsuccessful attempt to produce a pellet of Egyptian blue (Kos, Greece), an ancient artificial pigment. The PCA of a large dataset of 171 μXANES spectra acquired on archaeological samples and on a series of reference copper compounds highlighted high variations in XANES features due to different speciation and orientation effects. The results indicated that, rather than the use of inadequate firing temperatures, the use of inappropriate starting materials with an unusually high Fe content led to production of the reddish cuprite (Cu₂O) instead of the desired Egyptian blue.

Although the number of publications dealing with new improvements in macroXRF scanners for cultural heritage applications has decreased since last year's review, the 3rd workshop on macroXRF ‘Macro X-ray Fluorescence Scanning in Conservation, Art and Archaeology’, which takes place in September 2022, will hopefully give a new boost to their use.

Abbreviations

2D	two dimensional
3D	three dimensional
AAS	atomic absorption spectrometry
AES	atomic emission spectrometry
AFM	atomic force microscopy
ANL	Argonne National Laboratory
ANN	artificial neural network
APS	advanced photon source
ASU	atomic spectrometry update
ATCC	American Type Culture Collection
BL	beamline
C-L-XRF	confocal line XRF
CHNet	cultural heritage network
CRM	certified reference material
CT	computed tomography
DESY	Deutsches Elektronen-Synchrotron
DOPAC	3,4-dihydroxyphenylacetic acid
EDS	energy dispersive X-ray spectrometry
EDXRF	energy dispersive X-ray fluorescence
EPMA	electron probe microanalysis
ERM	European Reference Materials
ESRF	European Synchrotron Radiation Facility
EU	European Union
FAAS	flame atomic absorption spectrometry
FPM	fundamental parameter method
FWHM	full width at half maximum height
GEXRF	grazing exit X-ray fluorescence
GIXRF	grazing incidence X-ray fluorescence
HDD	horizontal directional drilling
HR	high resolution
ICP	inductively coupled plasma
INFN	Istituto Nazionale di Fisica Nucleare
IR	infrared
IS	internal standard
LA	laser ablation
LCF	linear-combination-fit
LEXRF	low-energy X-ray fluorescence
LMJS	liquid-metal jet source
LOD	limit of detection
LS-CAT	life sciences collaborative access team
MIBG	metaiodobenzylguanidine
MIT	Massachusetts Institute of Technology
MS	mass spectrometry
MTP8	metal tolerance protein 8
NEXAFS	near edge X-ray absorption fine structure
NIST	National Institute of Standards and Technology
NP	nanoparticle
NSF	National Science Foundation
OXIRIS	orthovoltage X-ray-induced radiation system
PCA	principal component analysis
PGM	plane-reflection grating monochromator
PLS	partial least squares
PTB	Physikalisch-Technische Bundesanstalt
RM	reference material
RMSE	route mean square error
ROI	region of interest
RS	reflectance spectroscopy
RSD	relative standard deviation
SAM	spectral angle mapper
SD	standard deviation
SDD	silicon drift detector
SEM	scanning electron microscopy
SLS	Swiss Light Source
SMART	synchrotron-based machine learning approach for RasTer
S/N	signal-to-noise ratio
SR	synchrotron radiation
SRM	standard reference material
STEAM	statistically tailored elemental angle mapper
STXM	scanning transmission X-ray microscopy
TRXPS	total reflection X-ray photoelectric spectroscopy
TU	Technical University
TXRF	total reflection X-ray fluorescence
UK	United Kingdom
US	United States
USA	United States of America
USAF	United States Air Force
WDXRF	wavelength dispersive X-ray fluorescence
XANES	X-ray absorption near edge structure
XAS	X-ray absorption spectroscopy
XFM	X-ray fluorescence microscopy
XRD	X-ray diffraction
XRF	X-ray fluorescence
XRR	X-ray reflectometry
Z	atomic number

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., 2022, 37(1), 9–49.
2. M. Patriarca, N. Barlow, A. Cross, S. Hill, A. Robson, A. Taylor and J. Tyson, J. Anal. At. Spectrom., 2022, 37(3), 410–473.
3. E. H. Evans, J. Pisonero, C. M. M. Smith and R. N. Taylor, J. Anal. At. Spectrom., 2022, 37(5), 942–965.
4. R. Clough, C. F. Harrington, S. J. Hill, Y. Madrid and J. F. Tyson, J. Anal. At. Spectrom., 2021, 36(7), 1326–1373.
5. S. Carter, R. Clough, A. Fisher, B. Gibson, B. Russell and J. Waack, J. Anal. At. Spectrom., 2021, 36(11), 2241–2305.
6. S. Chen, R. O. Lastra, T. Paunesku, O. Antipova, L. Li, J. Deng, Y. Luo, M. B. Wanzer, J. Popovic, C. Jacobsen, S. Vogt and G. E. Woloschak, Cancers, 2021, 13(17) DOI:10.3390/cancers13174497.
7. W. Liu, S. Mirzoeva, Y. Yuan, J. J. Deng, S. Chen, B. Lai, S. Vogt, K. Shah, R. Shroff, R. Bleher, Q. L. Jin, N. Vo, R. Bazak, C. Ritner, S. Gutionov, S. Raha, J. Sedlmair, C. Hirschmugl, C. Jacobsen, T. Paunesku, J. Kalapurkal and G. E. Woloschak, Cancer Nanotechnol., 2021, 12(1) DOI:10.1186/s12645-12021-00081-z.
8. G. M. Saladino, C. Vogt, Y. Y. Li, K. Shaker, B. Brodin, M. Svenda, H. M. Hertz and M. S. Toprak, ACS Nano, 2021, 15(3), 5077–5085.
9. N. Ashraf, M. Vitova, P. Cloeten, A. Mijovilovich, S. N. H. Bokhari and H. Kupper, Aquat. Toxicol., 2021, 235 DOI:10.1016/j.aquatox.2021.105818.
10. S. Holler, H. Kupper, D. Bruckner, J. Garrevoet, K. Spiers, G. Falkenberg, E. Andresen and E. Peiter, Plant Biol., 2022, 24(1), 23–29.
11. R. Kondo, T. Yamato, A. Munoz-Noval, S. Honda, Y. Nishiwaki, K. Komaguchi and S. Hayakawa, J. Anal. At. Spectrom., 2021, 36(5), 1041–1046.
12. B. J. Tkalcec, P. Tack, E. De Pauw, B. Vekemans, T. Nakamura, J. Garrevoet, G. Falkenberg, L. Vincze and F. E. Brenker, Meteorit. Planet. Sci., 2022, 57(4), 817–829.
13. P. Tack, E. De Pauw, B. Tkalcec, A. Longo, C. J. Sahle, F. Brenker and L. Vincze, Anal. Chem., 2021, 93(44), 14651–14658.
14. G. Perotti, H. O. Sorensen, H. Haack, A. C. Andersen, D. F. Sanchez, E. van Kooten, E. H. R. Tsai, K. N. Dalby, M. Holler, D. Grolimund and T. Hassenkam, Astrophys. J., 2021, 922(2) DOI:10.3847/1538-4357/ac3826bc.
15. B. Gao, J. Aelterman, B. Laforce, L. Van Hoorebeke, L. Vincze and M. Boone, IEEE Trans. Nucl. Sci., 2021, 68(6), 1194–1206.
16. B. Gao, B. Laforce, L. Vincze, L. Van Hoorebeke and M. N. Boone, Anal. Chem., 2021, 93(4), 2082–2089.
17. L. J. Bauer, R. Gnewkow, F. Forste, D. Grotzsch, S. Bjeoumikhova, B. Kanngiesser and I. Mantouvalou, J. Anal. At. Spectrom., 2021, 36(11), 2519–2527.
18. H. Nakano, S. Komatani, T. Matsuyama and K. Tsuji, Anal. Sci., 2021, 37(10), 1447–1451.
19. H. Nakano, S. Sonoda, T. Matsuyama, S. Komatani and K. Tsuji, X-Ray Spectrom., 2021, 50(3), 224–230.
20. J. P. Tang, C. J. Ptacek, D. W. Blowes, Y. Y. Liu, Y. Feng, Y. Z. Finfrock and P. Liu, Chem. Eng. J., 2022, 428 DOI:10.1016/j.cej.2021.131362.
21. P. Roy, H. Jahromi, S. Adhikari, Y. Z. Finfrock, T. Rahman, Z. Ahmadi, M. Mahjouri-Samani, F. Feyzbar-Khalkhali-Nejad and T. S. Oh, Energy Convers. Manage., 2022, 252 DOI:10.1016/j.enconman.2021.115131.
22. H. E. Oguzturk, L. J. Bauer, I. Mantouvalou, B. Kanngiesser, O. D. Velev and M. Gradzielski, Part. Part. Syst. Charact., 2021, 38(6) DOI:10.1002/ppsc.202000328.
23. Q. Q. Xu, W. J. Xia, L. Z. Zhou, Z. W. Zou, Q. X. Li, L. J. Deng, S. Wu, T. Wang, J. D. Cui, Z. G. Liu, T. X. Sun, J. S. Ye and F. Z. Li, ACS Omega, 2022, 7(4), 3738–3745.
24. R. G. Figueroa, M. Valente, J. Guarda, J. Leiva, E. Quilaguilque, B. Casanelli and F. Leyton, X-Ray Spectrom., 2021, 51(3), 251–261.
25. G. Cappuccio, S. B. Dabagov, V. Guglielmotti, D. Hampai, M. Martini, C. Mazzuca, L. Micheli and M. Redi, Radiat. Phys. Chem., 2021, 188 DOI:10.1016/j.radphyschem.2021.109660.
26. I. Szaloki, A. Gerenyi, F. Fodor, G. Radocz, V. Czech and L. Vincze, Anal. Chem., 2021, 93(34), 11660–11668.
27. X. S. Lin, L. L. Zhang, J. H. Xu, Y. He, Y. Zheng, S. Yan, D. X. Liang and A. G. Li, J. Anal. At. Spectrom., 2021, 36(11), 2353–2361.
28. C. Vanhoof, J. R. Bacon, U. E. A. Fittschen and L. Vincze, J. Anal. At. Spectrom., 2021, 36(9), 1797–1812.
29. J. Yang, Z. J. Zhang and Q. M. Cheng, J. Anal. At. Spectrom., 2022, 37(4), 750–758.
30. K. Sanyal, B. Kanrar and S. Dhara, J. Anal. At. Spectrom., 2021, 36(4), 803–812.
31. K. Sanyal, B. Kanrar, S. S. Suresh and S. Dhara, J. Anal. At. Spectrom., 2022, 37(1), 130–138.
32. B. Kanrar, K. Sanyal and R. V. Pai, J. Anal. At. Spectrom., 2022, 37(4), 741–749.
33. V. K. Singh, N. Sharma and V. K. Singh, X-Ray Spectrom., 2021, 51(3), 304–327.
34. R. Kalliola, T. Saarinen and N. Tanski, Silva Fenn., 2021, 55(1) DOI:10.14214/sf.10444.
35. X. Deng, Y. X. Chen, Y. Yang, L. Peng, L. Si and Q. R. Zeng, Front. Environ. Sci. Eng., 2021, 15(6) DOI:10.1007/s11783-11021-11431-11785.
36. Y. K. Liu, W. Y. He, J. Y. Zhang, Z. W. Liu, F. Z. Chen, H. J. Wang, Y. T. Ye, Y. T. Lyu, Z. Y. Gao, Z. C. Yu, L. N. Bi and S. C. Zhang, Geofluids, 2022, 2022, 1–20.
37. P. D. Zander, M. Zarczynski, W. Tylmann, S. K. Rainford and M. Grosjean, Clim. Past, 2021, 17(5), 2055–2071.
38. P. D. Quinn, L. Alianelli, M. Gomez-Gonzalez, D. Mahoney, F. Cacho-Nerin, A. Peach and J. E. Parker, J. Synchrotron Radiat., 2021, 28, 1006–1013.
39. A. Gianoncelli, V. Bonanni, G. Gariani, F. Guzzi, L. Pascolo, R. Borghes, F. Bille and G. Kourousias, Appl. Sci., 2021, 11(16) DOI:10.3390/app11167216.
40. J. J. Kim, F. T. Ling, D. A. Plattenberger, A. F. Clarens, A. Lanzirotti, M. Newville and C. A. Peters, Comput. Geosci., 2021, 156 DOI:10.1016/j.cageo.2021.104898.
41. C. Marini, J. Roque-Rosell, M. Campeny, S. Toutounchiavval and L. Simonelli, J. Synchrotron Radiat., 2021, 28, 1245–1252.
42. A. Rakotondrajoa and M. Radtke, Mach. Learn. Sci. Tech., 2021, 2(2) DOI:10.1088/2632-2153/abc1089fb.
43. S. Cipiccia, F. Brun, V. Di Trapani, C. Rau and D. J. Batey, J. Synchrotron Radiat., 2021, 28, 1916–1920.
44. P. Jagodzinski, M. Pajek, D. Banas, A. Kubala-Kukus, J. Szlachetko, M. Cotte and M. Salome, Opt. Express, 2021, 29(17), 27193–27211.
45. M. A. Alam, M. K. Tiwari, A. Trivedi, A. Khooha and A. K. Singh, J. Anal. At. Spectrom., 2022, 37(3), 575–583.
46. M. Ghosh, S. Biswas and K. K. Swain, Spectrochim. Acta, Part B, 2022, 187 DOI:10.1016/j.sab.2021.106328.
47. 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(2), 1225–1238.
48. Y. Kayser, J. Osan, P. Honicke and B. Beckhoff, Anal. Chim. Acta, 2022, 1192 DOI:10.1016/j.aca.2021.339367.
49. O. Czompoly, E. Borcsok, V. Groma, S. Pollastri and J. Osan, Atmos. Pollut. Res., 2021, 12(11) DOI:10.1016/j.apr.2021.101214.
50. D. Sowah-Kuma, J. Rehman, A. Yeboah, W. Bu, C. Yan and M. F. Paige, J. Surfactants Deterg., 2021, 24(6), 897–907.
51. W. Bu, M. Mihaylov, D. Amoanu, B. Lin, M. Meron, I. Kuzmenko, L. Soderholm and M. L. Schlossman, J. Phys. Chem. B, 2014, 118(43), 12486–12500.
52. H. Komatsu, H. Takahara, W. Matsuda and Y. Nishiwaki, J. Forensic Sci., 2021, 66(5), 1658–1668.
53. H. Takahara, W. Matsuda, Y. Kusakabe, S. Ikeda, M. Kuraoka, H. Komatsu and Y. Nishiwaki, Anal. Sci., 2021, 37(8), 1123–1129.
54. A. S. Maltsev, A. V. Ivanov, G. V. Pashkova, A. E. Marfin and Y. A. Bishaev, Spectrochim. Acta, Part B, 2021, 184 DOI:10.1016/j.sab.2021.106281.
55. H. T. Phuong, N. A. Son, N. T. N. Ha, N. T. M. Sang, N. T. T. Linh, D. T. Binh, T. T. H. Loan, H. M. Dung, T. T. Anh and N. V. Dong, Spectrochim. Acta, Part B, 2021, 182 DOI:10.1016/j.sab.2021.106234.
56. D. Mennickent, R. D. Castillo, J. Araya and J. Y. Neira, X-Ray Spectrom., 2021, 51(2), 142–150.
57. M. Breuckmann, G. Wacker, S. Hanning, M. Otto and M. Kreyenschmidt, J. Anal. At. Spectrom., 2022, 37(4), 861–869.
58. R. Fernandez-Ruiz, X-Ray Spectrom., 2021, 51(3), 279–293.
59. E. Margui, J. Jablan, I. Queralt, F. Bilo and L. Borgese, X-Ray Spectrom., 2021, 51(3), 230–240.
60. G. Mankovskii and A. Pejovic-Milic, X-Ray Spectrom., 2021, 51(3), 262–270.
61. G. Mankovskii and A. Pejovic-Milic, X-Ray Spectrom., 2021, 51(3), 271–278.
62. K. Tsuji, T. Matsuyama, T. Fukuda, S. Shima, M. Toba, J. S. Oh and T. Shirafuji, J. Anal. At. Spectrom., 2021, 36(9), 1873–1878.
63. E. Margui, R. Dalipi, E. Sangiorgi, M. B. Stefan, K. Sladonja, V. Rogga and J. Jablan, X-Ray Spectrom., 2021, 51(3), 204–213.
64. N. Lara-Almazan, G. Zarazua-Ortega, P. Avila-Perez, C. Carreno-De Leon and C. E. Barrera-Diaz, X-Ray Spectrom., 2021, 50(5), 414–424.
65. S. Dhara, J. Anal. At. Spectrom., 2021, 36(2), 352–360.
66. M. Musielak, K. Kocot, B. Zawisza, E. Talik, E. Margui, I. Queralt, B. Walczak and R. Sitko, Spectrochim. Acta, Part B, 2021, 177 DOI:10.1016/j.sab.2021.106069.
67. M. Musielak, M. Serda, E. Talik, A. Gagor, J. Korzuch and R. Sitko, J. Anal. At. Spectrom., 2021, 36(7), 1533–1543.
68. B. Patarachao, D. D. Tyo, D. Chen and P. H. J. Mercier, Spectrochim. Acta, Part B, 2021, 177 DOI:10.1016/j.sab.2020.106053.
69. A. Kubala-Kukus, D. Banas, M. Pajek, J. Braziewicz, S. Gozdz, J. Szlachetko, J. Semaniak, L. Jablonski, P. Jagodzinski, M. Piwowarczyk, D. Sobota, I. Stabrawa, R. Stachura, K. Szary and J. Wudarczyk-Mocko, Acta Phys. Pol., A, 2021, 139(3), 247–256.
70. Y. Menesguen and M. C. Lepy, Phys. Status Solidi A, 2021, 219(9) DOI:10.1002/pssa.202100423.
71. A. Andrle, P. Honicke, G. Gwalt, P. I. Schneider, Y. Kayser, F. Siewert and V. Soltwisch, Nanomat, 2021, 11(7) DOI:10.3390/nano11071647.
72. R. D. Kiranjot and M. H. Modi, Surf. Interface Anal., 2021, 54(1), 52–58.
73. J. Holburg, M. Mueller, K. Mann, P. Wild, K. Eusterhues and J. Thieme, Anal. Chem., 2022, 98(8), 3510–3536.
74. R. Unterumsberger, B. Beckhoff, A. Gross, H. Stosnach, S. Nowak, Y. P. Stenzel, M. Kramer and A. von Bohlen, J. Anal. At. Spectrom., 2021, 36(9), 1933–1945.
75. R. Fernandez-Ruiz, Spectrochim. Acta, Part B, 2021, 180 DOI:10.1016/j.sab.2021.106207.
76. A. Carapelle, G. Lejeune, M. Morelle, O. Evrard, P. Leroux, J. Prinzie, Y. Cao, B. Van Bockel, F. Montfort and N. Martin, X-Ray Spectrom., 2022, 51(4), 388–393.
77. J. L. Lu, J. K. Guo, Q. Q. Wei, X. D. Tang, T. Lan, Y. R. Hou and X. Y. Zhao, Appl. Sci., 2022, 12(2) DOI:10.3390/app12020568.
78. J. Hao, F. S. Li, Q. Y. Wang, X. Y. Jiang, B. Y. Yang and J. Cao, Nucl. Instrum. Methods Phys. Res., Sect. A, 2021, 1013 DOI:10.1016/j.nima.2021.165672.
79. S. H. G. Silva, B. T. Ribeiro, M. B. B. Guerra, H. W. P. de Carvalho, G. Lopes, G. S. Carvalho, L. R. G. Guilherme, M. Resende, M. Mancini, N. Curi, R. B. A. Rafael, V. Cardelli, S. Cocco, G. Corti, S. Chakraborty, B. Li and D. C. Weindorf, Adv. Agron., 2021, 167, 1–62.
80. D. E. B. Fleming, X-Ray Spectrom., 2021, 51(3), 328–337.
81. M. G. de Freitas, F. R. dos Santos, P. S. Parreira and F. L. Melquiades, Spectrosc. Lett., 2021, 54(7), 560–570.
82. K. Nakano, S. Tobari, S. Shimizu, T. Ito and A. Itoh, X-Ray Spectrom., 2021, 51(1), 101–108.
83. H. Zhang, J. Antonangelo and C. Penn, Sci. Rep., 2021, 11, 1–12.
84. J. Kagiliery, S. Chakraborty, B. Li, M. Hull and D. C. Weindorf, EQA–Int. J. Envirol. Qual., 2021, 45, 27–41.
85. B. T. Ribeiro, D. C. Weindorf, C. S. Borges, L. R. G. Guimaraes and N. Curi, Spectrochim. Acta, Part B, 2021, 186 DOI:10.1016/j.sab.2021.106320.
86. M. Horf, R. Gebbers, S. Vogel, M. Ostermann, M. F. Piepel and H. W. Olfs, Sensors, 2021, 21(11) DOI:10.3390/s21113892.
87. A. J. G. de Faria, M. Rufini, A. D. Leite, B. T. Ribeiro, S. H. G. Silva, L. R. G. Guilherme and L. C. A. Melo, Environ. Technol. Innovation, 2021, 23 DOI:10.1016/j.eti.2021.101788.
88. R. Cesareo, G. E. Gigante, A. Castellano, S. Ridolfi and S. A. B. Lins, Braz. J. Phys., 2022, 52(2) DOI:10.1007/s13538-13021-01042-y.
89. F. Gherardi, Anal. Methods, 2021, 13(33), 3731–3734.
90. G. Ruschioni, F. Micheletti, L. Bonizzoni, J. Orsilli and A. Galli, Appl. Sci., 2022, 12(3) DOI:10.3390/app12031006.
91. B. Laforce, G. Fiers, H. Vandendriessche, P. Crombe, V. Cnudde and L. Vincze, Anal. Chem., 2021, 93(8), 3898–3904.
92. A. Galli, M. Caccia, S. Caglio, L. Bonizzoni, I. Castiglioni, M. Gironda, R. Alberti and M. Martini, Eur. Phys. J. Plus, 2021, 137(1) DOI:10.1140/epjp/s13360-13021-02183-13364.
93. J. Orsilli, A. Galli, L. Bonizzoni and M. Caccia, Appl. Sci., 2021, 11(4) DOI:10.3390/app11041446.
94. S. A. B. Lins, M. Manso, P. A. B. Lins, A. Brunetti, A. Sodo, G. E. Gigante, A. Fabbri, P. Branchini, L. Tortora and S. Ridolfi, Sensors, 2021, 21(5) DOI:10.3390/s21051913.
95. A. Mazzinghi, C. Ruberto, L. Castelli, C. Czelusniak, L. Giuntini, P. A. Mando and F. Taccetti, Appl. Sci., 2021, 11(13) DOI:10.3390/app11136151.
96. M. Bicchieri, P. Biocca, C. Caliri and F. P. Romano, X-Ray Spectrom., 2021, 50(4), 401–409.
97. C. Colantonio, L. Clivet, E. Laval, Y. Coquinot, C. Maury, M. Melis and C. Boust, Eur. Phys. J. Plus, 2021, 136(9) DOI:10.1140/epjp/s13360-13021-01909-13368.
98. K. Derks, G. Van der Snickt, S. Legrand, K. Van der Stighelen and K. Janssens, Heritage Sci., 2022, 10(1) DOI:10.1186/s40494-40021-00634-w.
99. L. Monico, S. Prati, G. Sciutto, E. Catelli, A. Romani, D. Q. Balbas, Z. L. Li, S. De Meyer, G. Nuyts, K. Janssens, M. Cotte, J. Garrevoet, G. Falkenberg, V. I. T. Suarez, R. Tucoulou and R. Mazzeo, J. Anal. At. Spectrom., 2022, 37(1), 114–129.
100. A. K. Marketou, F. Giannici, S. Handberg, W. de Nolf, M. Cotte and F. Caruso, Anal. Chem., 2021, 93(33), 11557–11567.

---

Footnote

† Joint review coordinators.

---